Home

Fulltext - ETH E-Collection

1. orig SA proc contro proc sh 0 9 0 8 eee nn 0 8 Megs ei eee 0 8 eae ee acral 0 7 0 7 0 6 5 0 6 0 5 0 5 0 44 04 0 3 0 3 0 24 0 5 iS N 5 0 0 000712121 12 12 12 12121212102 00000000 orgetting 10 forgetting 1 forgetting 0 1 Figure 47 shows the four ARV processed curves see text during a fast pulse amplitude transition from hand closing to hand opening for three differeent forgetting weights p 10 1 and 0 1 The transition time is 0 9 s The original SA curve containing voluntary SEMG and the SA has a much higher ARV than the processed SA and both control curves It is mainly generated by the SA A comparison between the original and the processed control shows that the forgetting weight p 1 does not affect the SEMG as much as a forgetting weight p 0 1 and the SA during the transition is reduced significantly A fast convergence of the SA extraction algorithm can be obtained with a forgetting weight p smaller than 1 as shown in Figures 46 and 47 Thus a forgetting weight p is a good compromise between fast adaptation and not influencing the voluntary SEMG activity Changing pulse widths or pulse amplitudes for example during transitions from grasp to release are more challenging for the algorithm Here the shape of the SA is changing under some conditions A fast adaptation to such changes is required In such transitions are shown with changing pulse wi
2. Abbreviations Medical Abbreviations ADL AP ASIA CNS EMG FES MD OT PT SA SCI SEMG Technical Abbreviations ANN ARV BOSFET field CMRR DMA FFT FIR FSR EPROM GAL GUI PC RAM RMS SPI S N Activities of daily living Action potential American Spinal Injury Association Central nervous system Electromyographic Functional electrical stimulation Medical doctor Occupational therapist Physical therapist Stimulation artifact Spinal cord injured Surface EMG Artificial neural network Average rectified mean value Bidirectional Metal Oxide Semiconductor Effect transistor MOSFET Common mode rejection ratio Direct memory access Fast fourier transform Finite impulse response Force sensitive resistor Electrically programmable read only memory Gate address logic Graphical user interface Personal computer Random access memory Root mean square Synchronous serial port Signal to noise ratio 12 1 Introduction 1 1 Background The development of new rehabilitation technology like neuroprostheses has been more the field of activity of universities and small spin off companies than of globally operating companies The main challenges in this field lie more in the breadth of the different medical aspects that are not clearly described or sometime even in a contradictory manner than in the depth of finding an exact solution to a well described phenomenon as in other engineering disciplines R
3. Harding G W 1991 A method for eliminating the stimulus artifact from digital recordings of the direct cortical response Comp Biomed Res 2 pp 183 95 Haugland M 1997 A flexible method for fabrication of nerve cuff electrodes In Proc 18th Int Conf IEEE EMBS Amsterdam Netherlands pp 359 60 Haugland M Lickel A Haase J and Sinkjaer T 1999 Control of FES thumb force using slip information obtained from the cutaneous electroneurogram in quadriplegic man JEEE Trans Rehab Eng 2 pp 215 27 Hermie J H Freriks B Merletti R Stegeman D Blok J Rau G Disselhorst Klug C and Haegg G 1999 SENIAM European recommondations for surface electromyography Enschede Netherlands Roessingh Research and Development 1999 Hill T L Eisenberg E Chen Y and Podolsky R J 1975 Some self consistent two state sliding filament models of muscle contraction Biophys J pp 335 72 Hines A E Crago P E Chapman G J and Billian C 1996 Stimulus artifact removal in EMG from muscles adjacent to stimulated muscles J Neurosci Meth 1 pp 55 62 Huxley A F 1957 Muscle structure and theories of contraction Prog Biophys Chem pp 257 318 Ijezerman M J T S S in t Groen F A C G Klatte M A P Snoeck G J Vorsteveld J H C Nathan R H and Hermens H J 1996 The NESS Handmaster orthosis restoration of hand function in C5 and stroke patients by means of electr
4. The transistor output is used as a reference for the closed loop voltage regulated current controller to regulate the depolarizing stimulation pulse see scheme B in Figure 16 The charge balancing stimulation pulse is generated as follows Immediately after the depolarization pulse a second open loop voltage to current converter allows an inverse current flow with an exponential curve shape that balances the charge in the stimulated muscles see scheme C in Figure 14 The curve has its shape from a discharging capacitor that is charged during the depolarization pulse As a result the entire stimulation current pulse form is composed of a depolarizing rectangular current pulse with constant amplitude followed by a hyperpolarizing exponentially decreasing current pulse that balances the charge see Figure 15 A high voltage source of approximately 180 V DC generated by a PICO DC DC converter powers the constant current stimulation pulse generator Because of the rather high electrode skin impedance several kQ such a high DC voltage is needed for the generation of 50 to 100 mA current pulses that are used for supramaximal stimulation of leg muscles Four pairs of electrodes are sequentially connected to the pulse generation circuit using 8 BOSFET switches Only one pair of electrodes is connected to the pulse generation circuit at a time all others are floating In case of an emergency the stimulation can be disabled at any time by turning
5. Vodovnik L Crochetiere W J and Reswick J B 1967 Control of a skeletal joint by electrical stimulation of antagonists Med Biol Eng 2 pp 97 109 Vodovnik L Long C 2nd Reswick J B Lippay A and Starbuck D 1965 Myo electric control of paralyzed muscles IEEE Trans Biomed Eng 3 pp 169 72 Walker D D and Kimura J 1978 A fast recovery electrode amplifier for electrophysiology Electroencephal Clin Neurophys 6 pp 789 92 143 Wichmann T 2000 A digital averaging method for removal of stimulus artifacts in neurophysiologic experiments J Neurosci Meth 1 pp 57 62 Winters J M and Woo S L Y 1990 Multiple muscle systems biomechanics and movement organization Springer Verlag New York Zach G A Koch H G Wolfensberger M and Michel D 1995 Demographie und Statistik der Querschnittlahmung In Ouerschnittl hmung ganzheitliche Rehabilitation Hg G A Z ch Verlag Dr Felix W st AG K ssnacht pp 16 9 Zahalak G I 1992 An overview of muscle modeling In Neural prostheses replacing motor function after disease or disability Hg R B Stein et al Oxford University Press New York pp 17 57 Zajac F E 1989 Muscle and tendon properties models scaling and application to biomechanics and motor control CRC Crit Rev Biomed Eng pp 359 410
6. al 2000 The last reference gives also detailed information about what is achievable in the prosthetic field 2 1 1 Generation of Action Potentials The basic principle to contract a muscle using FES is to artificially generate action potentials in efferent muscle nerve fibers called motor neurons Efferent nerve fibers see are the descending axons from the brain to the muscles whereas the afferent nerves are the ascending axons that provide the sensory information to the CNS Cell body soma Initial segment of Node of ranvier Schwann cell Terminal buttons Figure 1 The motor neuron Reprinted from Karu 1992 An action potential AP can be generated by depolarizing the nerve In normal steady state conditions there is a difference of the electrical potential between the inside and the outside of the nerve membrane of 70 90 mV The potential inside the nerve membrane is electrically negative with respect to the outside potential This difference is due to a high concentration of cellular anions and a poor concentration of sodium Na ions inside the nerve membrane In the resting state the concentration of potassium K ions inside the nerve is higher than outside However the cell membrane in osmotic equilibrium keeps more cellular anions inside the nerve than cations Furthermore the membrane is more permeable to potassium than to sodium and other cations In equilibrium the resting potential inside the membrane is at a
7. 2 Principle and Function of Neuroprostheses for Grasping 22 2 1 2 Propagation of Action Potentials The local influx of sodium ions and the polarization of the nerve affects the neighboring sodium channels to operate in the same manner with a slight time delay This results in a propagation of the AP The AP propagates along the nerve with a propagation velocity of about 30 120 m s depending on the nerve type and reaches the axon terminal where the neurotransmitter acetylcholine is released This neurotransmitter diffuses into the synaptic gap and is absorbed by the so called motor end plate receptors They are located on the muscle membrane The received acetylcholine causes a depolarization of the muscle membrane and initiates a contractile muscle twitch 2 1 3 Excitability of Nerve Fibers The excitability of nerve fibers depends on the following factors e the distance of the excitatory electrode cathode to the nerve e the diameter of the nerve fiber e the applied electrical charge The closer a stimulation electrode is positioned to a nerve the lower the stimulation intensity can be chosen to excite it Additional tissue between the electrode and the nerve reduces the voltage gradient between them The excitability of nerve fibers changes also with the change of the fiber diameter The larger the nerve fiber is in diameter the easier the nerve can be electrically excited This effect is reported in literature as reverse recruitment or
8. Bie Aba digiti min Opp poll F Opp digiti min Fl poll b sup h FI digiti min Lumbricales I II Palmar interossei Lumbricales Ill IV Dorsal interossei see dorsum Fl poll b deep h Ant inter branch Ant inter bran Adductor pollicis 1993 Florence P Kendall Author grants permission to reproduce for personal use but not for sale Figure 60 The motor points of the upper extremity muscles Motor points are the locations where the peripheral nerves enter the muscle At these positions the muscles can be stimulated with the best selectivity Reprinted from Kendall et al 1993 Theoretically more proximal muscles that would allow the subjects to have a better reaching control could also be stimulated It was shown by the Cleveland group that the reaching task could be improved by constantly stimulating the M triceps brachii during the whole grasp sequence Crago et al 1998 C4 to C6 complete SCI subjects have no voluntary control of the triceps muscle Elbow extension can only be achieved with help of gravity A pre activation of this muscle allows the subject to control elbow extension and flexion with the voluntarily controlled biceps muscle The stimulation of more 7 Neuroprosthesis for Grasping 115 muscles is principally possible but it complicates the system on the stimulation side as well as on the control side and was not tried yet by our group 7 3 Control Strategies for FES G
9. The digital SEMG control scheme was used with a C6 tetraplegic subject that had voluntary control over his wrist extensors The neuroprosthesis was commanded by activating the measured M flexor carpi radialis with specific Morse code like patterns see Section 7 3 3 Potentially every muscle with voluntary activity can be used with this strategy but other muscles were not yet tested The digital SEMG control strategy has the advantage compared to the analog SEMG control strategy that it is less sensitive to voluntary arm movements of the arm that carries the recording EMG electrodes for example the contralateral arm can be used to support the grasp task The push button and the sliding potentiometer control strategies were used with C5 and C6 SCI subjects Both strategies allow also bilateral grasp Contralateral arm movements do not affect the control of the neuroprosthesis Our general findings are that after a four week learning phase the SEMG control strategies are faster and more intuitive for the subjects to use However the analog SEMG control strategy using proportional control is very tiring for the subject It is impossible to hold an object for a prolonged time because of muscle fatigue of the control muscles The modified analog control strategy that integrates the control variable redressed this problem but has to be tested more intensively with SCI subjects The digital SEMG control strategy works satisfactory if only a few stim
10. The parameter single or double pulses enables double pulse twitches for the specified channels with an inter pulse interval of 5 ms An inter pulse interval of 5 ms is reported to be optimal to recruit additional nerve fibers compared to single pulses Karu et al 1995 Double pulse twitches are also reported to reduce habituation when afferent nerves are stimulated to generate reflexes such as the flexion reflex 4 3 4 Compensation of the Stimulation Recruitment Curves One of the simplest muscle models the Hammerstein model describes an electrically stimulated muscle using a static non linearity followed by a second order linear system The static non linearity of the Hammerstein model corresponds to the recruitment curve of the stimulated muscle The recruitment curve can be obtained by stimulating a muscle with a slowly increasing pulse width or pulse amplitude and by measuring the muscle contraction force Figure 24 For each stimulation channel the normalized control variable or the normalized stimulation pattern is mapped to a look up table that compensates the static non linear recruitment characteristics of the muscle 4 PC Based Rapid Prototyping FES System 61 In both the event triggered stimulation pattern generator software and the continuously controlled stimulation software the normalized stimulation pattern or the normalized control variable is mapped to the recruitment curve compensation look up tables to gen
11. afterwards summed subtracted and or scaled depending on the set mode The output of the controller is a control variable ranging from 1 to 1 Continuous Controller Mode Control input Preprocessing Figure 22 The continuous controller generates the control variable for the two control algorithm sliding potentiometer control and analog EMG control The control variable from the continuous controller is mapped to the recruitment curve compensation module using a look up table Figure 23 for the thumb and one for the fingers With this method different stimulation sensor output mapping can be done for the thumb and for the fingers For example the thumb can only be stimulated during hand closing and not during hand opening Figure 23 The control variable that is generated in the continuous controller is mapped differently to the finger and thumb stimulation channels For this task two look up tables are used 4 3 3 Stimulation Parameter Setup Module For each stimulation channel the stimulation parameters stimulation amplitude stimulation pulse width stimulation frequency and single or double pulses are set Table 2 shows the parameters the range of the parameters and the resolutions The parameter stimulation pulse width defines the maximal stimulated pulse duration in Us when the normalized stimulation pattern is 1 hand opening or 1 hand closing 4 PC Based Rapid Prototyping FES System 60
12. iFa Institut f r Automatik Diss ETH No 14481 Surface Functional Electrical Stimulation FES Neuroprostheses for Grasping Thierry Keller Eidgen ssische Technische Hochschule Z rich Swiss Federal Institute of Technology Zurich Diss ETH No 14481 Surface Functional Electrical Stimulation FES Neuroprostheses for Grasping A dissertation submitted to the SWISS FEDERAL INSTITUTE OF TECHNOLOGY ZURICH for the degree of Doctor of Technical Sciences presented by Thierry Keller Dipl El Ing ETH born 9 Mai 1968 citizen of Oberthal BE accepted on the recommendation of Prof Dr M Morari examiner Prof Dr V Dietz co examiner 2001 Foreword Doing research and writing a thesis has a lot to do with curiosity persistence and motivation Curious I was since my early childhood when I was questioning almost everything and my parents never got tired to explain or to guess the right answer They appertain my innermost thank for all the freedom they gave me in making my own decisions and in supporting me in every sense Persistence is a necessity for engineers who want to get things moved and I never lacked of The third attribute motivation is where friends colleagues and supervisors can really help that a work becomes successful And I had not a few I would like to thank Jan Schultheiss who brought me into the challenging field of rehabilitation engineering in offering me to work with him as a research e
13. 17 64 De Luca C J 1997 The use of the surface electromyography in biomechanics J Appl Biomech pp 135 63 Del Pozo F and Delgado J M R 1978 Hybrid stimulator for chronic experiments IEEE Trans Biomed Eng 1 pp 92 4 Donaldson N D Perkins T A and Worley A C 1997 Lumbar root stimulation for restoring leg function stimulator and measurement of muscle actions Artif Organs 3 pp 247 9 Dorgan S J and O Malley M J 1997 A nonlinear mathematical model of electrically stimulated skeletal muscle JEEE Trans Rehab Eng 2 pp 179 94 Duchene J and Goubel F 1993 Surface electromyogram during voluntary contraction processing tools and relation to physiological events Crit Rev Biomed Eng 4 pp 313 97 Epstein C M 1995 A simple artifact rejection preamplifier for clinical neurophysiology Am J EEG Technol 1 pp 64 71 Fisekovic N and Popovic D B 2001 New controller for functional electrical stimulation systems Med Eng Phys 6 pp 391 9 Frankel H L Hancock D O Hyslop G Melzak J Michaelis L S Ungar G H Vernon J D and Walsh J J 1969 The value of postural reduction in the initial management of closed injuries of the spine with paraplegia and tetraplegia Paraplegia 3 pp 179 92 Freeman J A 1971 An electronic stimulus artifact suppressor Electroencephal Clin Neurophys 2 pp 170 2 Gareiss V 2001 Sprachsteuerung von Neuroprothesen Semes
14. 1999 data analysis tool pack 5 Portable FES System The chapter presents briefly our first generation portable FES system the ETHZ ParaCare portable FES system and then describes more in detail the Compex Motion system that was developed in collaboration with our industrial partner Compex SA The following topics are covered in this chapter e The ETHZ ParaCare portable FES system e The concept of the Compex Motion stimulator e A description of the Compex Motion hardware that was developed by our industrial partner and is currently sold as a therapeutic stimulator The here described hardware functions are those of the new FES system and differ in some aspects from those of the sold therapeutic device e The newly developed Compex Motion assembler software e The Compex Motion programming software This part describes in detail how the Compex Motion stimulator can be programmed It lists all available functions and parameters and explains how they are used to cover a wide variety of applications The main achievements were e to develop a concept that brings together the requirements of a very flexibly programmable stimulator with the demands of a commercially uniform product e to deal with the constraints of an already existing hardware that could only be modified in its function by changing the software firmware e to fit all volitional and required functions into the limited hardware capabilities of the stimulator T
15. Brindley G S Polkey C E Rushton D N and Cardazo L 1986 Sacral anterior root stimulators for bladder control in paraplegia the first 50 cases J Neurol Neurosurg Psychiat pp 1104 14 Buckett J R Peckham P H Thrope G B Braswell S D and Keith M W 1988 A flexible portable system for neuromuscular stimulation in the paralyzed upper extremity IEEE Trans Biomed Eng 11 pp 897 904 Cameron T Loeb G E Peck R A Schulman J H Strojnik P and Troyk P R 1997 Micromodular implants to provide electrical stimulation of paralyzed muscles and limbs JEEE Trans Biomed Eng 9 pp 781 90 136 137 Cameron T Loeb G E Richmond F J R Peck R A Schuman J H Strojnik P and Troyk P 1993 Micromodular electronic devices to activate paralyzed muscles and limbs In Proc 14th Int Conf IEEE EMBS pp 1242 3 Compex SA 1996 Biofeedback version 2M4456 In Manual Compex SA Ecublens pp 2 10 Compex SA 1998 Compex electrodes for use with Compex muscle stimulator In Manual Hg Compex SA Ecublens p 2 Crago P E Memberg W D Usey M K Keith M W Kirsch R F Chapman G J Katorgi M A and Perreault E J 1998 An elbow extension neuroprosthesis for individuals with tetraplegia IEEE Trans Rehabil Eng 1 pp 1 6 Curt A 1996 Klinische Neurophysiologie in der Paraplegiologie In Querschnitt l hmung Hg V Dietz W Kohlhammer GmbH Stuttgart pp
16. ER an a ere are E unsre ge on OB fT cpenantntte tea tenennccececenenee 75 ears 7 o 0 64 0 6 0 5 0 5 0 4 0 4 0 3 0 2 0 1 0 12 12 12 12 12 118 6 4 1 00 000000 0 0 000 0 0 00 0 1 4 6 8 41 12 12 12 12 12 0 0000 00 0 0 1 4 6 8 11 12 12 12 12 12 12 12 12 12 12 11 8 6 4 10 00 00 00 0 0 15 17 0 9 eet NR EEE 0 Gata a Ta me Berne a 0 84 Ps stants lenken Me 0 75 0 71 0 64 0 64 0 54 0 54 0 44 0 44 0 34 0 34 0 24 0 24 0 1 ATIRAR RTE Tea ann anna Srarmanmenannnetne u 01 0 0 12 12 12 12 12 12 12 114 0000000 0 0 00 0 0 0 0 0 0 7 12 12 12 12 12 12 12 0 00 0 0 0 0 0 0 6 12 12 12 12 12 12 12 12 12 12 12 12 12 12 10 2 00 000 0 0 0 Figure 49 shows the four ARV processed curves see text for different pulse width and amplitude transitions from hand opening to hand closing left panels and hand closing to hand opening right panels for a forgetting weight p 1 The transition times are 2 7 s in panels I and 3 left and right and 0 9 s in panels 2 and 4 left and right The panels I and 2 left and right show pulse width transitions and the panels 3 and 4 left and right show pulse amplitude transitions The original SA curves containing voluntary SEMG and the SAs have a much higher ARV than the processed SAs and the control curves During the transitions the processed SAs are significantly lower when the pulse amplitudes are changing than when the pulse widths are changing The ARV curves in Figure 49 increase during th
17. The M extensor carpi radialis and brevis and the M extensor carpi ulnaris articulate the wrist joint The depolarization electrode is placed such that both muscle groups are balanced in the ulnar radial articulation of the wrist joint and therefore only causes wrist extension Extensor digitorum Extensor d Gath minimi Figure 55 The stimulation of the M extensor digitorum and digiti minimi produces hand opening as shown in the Figure Reprinted from Kendall et al 1993 Actually for hand opening the contraction of the M extensor digitorum and digiti minimi and the articulation of the metacarpophalangeal and interphalangeal joints finger joints are required and not the articulation of the wrist joint The M extensor 7 Neuroprosthesis for Graspin 11 digitorum extends the metacarpophalangeal and interphalangeal joints of the second through fifth digits and the M extensor digiti minimi additionally only the fifth digit little finger as shown in If wrist extension during hand opening is too strong such that the subject cannot place the fingers nicely around the object prior to the grasp task hand opening should only be stimulated during the release task 7 2 2 Electrode Positions for Finger Flexion Finger flexion is generated by placing a stimulation electrode pair on the anterior forearm either distal of the medial epicondyle of the humerus depolarization electrode and proximal of the wrist joint charge balancing elect
18. and up to eight independent telemetry channels for system functions Overall timing constraints power limitations a limited number of lead wires and limited implant capsule size do not allow to use all the above features at once The first fully implantable sensor that can be used with the new implant is a wrist joint angle transducer that uses three Hall effect sensors It is capable of detecting the orientation of the wrist position for two degrees of freedom A permanent magnet is implanted in an articulated bone and the transducer in the opposing bone Currently the system is in its trial phase Other fully implantable sensors that have been tried are EMG sensors 2 Principle and Function of Neuroprostheses for Grasping 39 However problems with the quality and resolution of the measured EMG signals were reported and the Hall sensors could potentially be susceptible to electromagnetic noise This new generation of implantable stimulators has a very high potential to improve future implanted FES applications NEC FESMate FES System In early 1980s the Sendai FES group lead by Y Handa developed microcomputer controlled neuroprostheses for grasping and walking By the end of the 1980s a portable PC programmable FES system consisting of a NEC PC 98LT personal computer and an 8 bit microcontrolled stimulator with 16 D A stimulation and 3 A D recording channels was developed Handa et al 1989 The system was used to restore the grasp function
19. sampling frequency is set appropriately Sensors like force sensitive resistors FSR push buttons sliding potentiometers or analog joysticks can be sampled with a sampling frequency of less than 100 Hz Normally the stimulation frequency ranging from 20 to 50 Hz is chosen to sample those types of sensors Raw EMG signals due to their larger bandwidth have to be sampled with a much higher sampling frequency Therefore a sampling frequency of 8 kHz is used see Chapter 6 For all sensors a gain of 1 on the data acquisition card is chosen Weak sensor signals e g EMG signals are amplified with external amplifiers The gain and an offset compensation can be adjusted by the software Usually the hardware gains are adjusted such that the signals use the whole range of the data acquisition card 5 to 5 V Additionally a software gain after sampling the signal normalizes the sensor signal in the range of 1 1 and an offset compensation sets the neutral position of the sensors 4 PC Based Rapid Prototyping FES System 57 4 3 2 Sensor Signal Processing The normalized and offset compensated sensor signals are processed according to the chosen control strategy Of the five implemented control strategies push button control sliding potentiometer control voice control EMG event triggered control and analog EMG control only the EMG control strategies need a more sophisticated signal pre processing This will be discussed in Chapters 5 and
20. shades in curves 2 and 4 that can be eliminated by blanking the signal during the stimulus As mentioned before in all shown SA curves only the first 12 5 ms post stimulus are plotted concatenated The rest of the curve is SA free and can be used to measure the voluntary SEMG activity without performing a SA removal signal processing 8 6 EMG Signals Features Signal Acquisition Stimulation Artifact Removal 9 PW Transition 25us 50us 75us 100us 125yus 150us 17515 200us 225us 250us raw SA without oie HAL YYYYYYY YY Yo processed SA forgetting 10 processed SA forgetting 1 processed SA forgetting 0 1 raw SA with voluntary EMG impeded dy 44 fh LP processed SA forgetting 10 processed SA forgetting 1 processed SA forgetting 0 1 Figure 46 Three different forgetting weights p were tested Curves 2 and 6 show that if the adaptation algorithm converges too slowly during pulse width changes the residual SA exceeds the voluntary contraction The weight p must be sufficiently small A good compromise is a forgetting weight of 1 Each panel of the Figures 47 49 and 50 shows the concatenated smoothened ARVs of the following SEMG signal segments 1 the first 12 5 ms post stimulus without SA removal algorithm original SA 2 the first 12 5 ms post stimulus with SA removal algorithm processed SA 3 the last 12 5 ms pre stimulus without SA removal algorithm original control 4 the last 12 5 ms pre s
21. the wrist retainer splint has 105 7 Neuroprosthesis for Graspin 106 two main functions 1 to stabilize the wrist of subjects with no voluntary wrist control and 2 to cover and to protect the stimulation electrodes C6 or lower injured subjects that have voluntary wrist control do not need a wrist retainer splint for stabilization They can stabilize their wrist voluntarily The electrodes and the cable connectors are covered using a gaze to prevent electrode loss or torn cable connectors C5 or subjects with higher lesion that have no voluntary wrist control need a wrist splint retainer Different splints either made of leather or textile fabric and synthetic fiber are custom tailored by the hospital orthopedic workshop Two examples of wrist retainers are shown in rs The wrist is retained in a 30 extended position with an inwrought plastic or a metal bracer The bracer is positioned on the bottom side of the splint and fixes the palm with the ulna The purpose of the wrist splint retainer is to prevent wrist flexion during the stimulation of the finger flexors which are located in the forearm under the wrist flexors It stabilizes the hand in a physiological position for hand grasp Although the finger flexor electrodes are carefully placed some wrist flexion caused by the stimulation cannot be prevented The method to stimulate the wrist extensors during finger flexion doesn t work in every case due to muscle denervation or lack of enoug
22. trial our ETHZ ParaCare team participated with 6 subjects 36 systems were manufactured The system was built very compact with some excellent engineering solutions Despite its excellent engineering the system has some disadvantages e The stimulator control box is exposed to collision and sudden impacts for example tetraplegic subjects open doors with their forearms e Frequently the electrodes loose contact with the steel mesh resulting in a loss of function e The LVDT sensor is too delicate and had to be replaced or readjusted frequently Results of the multicenter trials showed improvements in performing activities of daily living ADL with and without the device after 6 months of use Power grasp and handling of big objects was improved but in many subjects with C6 C7 SCI the benefit of using the neuroprosthesis in ADL over time was not anymore significantly higher The Bionic Glove contributed mostly to improve hand grasp as a therapeutic aid Popovic et al 1999 A further development and commercialization of the Bionic Glove was stopped in 1999 by bankruptcy of the startup company Other Surface FES Systems for Grasping Many other FES devices for grasping were developed for research purposes but were seldom used for ADL In addition to systems similar to those mentioned above some EMG controlled systems were reported Keller et al 1998 Saxena et al 1995 Thorsen 1998 The systems from Saxena and Thorsen enhanced the teno
23. 1993 In the example of Figure 37 five seconds stationary SEMG activation of the M carpi radialis was recorded and analyzed The used SEMG recording electrode has a built in preamplifier with a gain of 1400 It provides an output voltage between 0 and 5 V with an offset of 2 5 V if no SEMG activity is measured The amplified SEMG signal that was measured had a mean of 2 5 V and a standard deviation of 0 273 V The median voltage was 2 49 V and the data had a kurtosis of 0 546 and a skew of 0 237 This can be shown by plotting the histogram of the SEMG signal 150 bars and the calculated Gaussian distribution using the mean and standard deviation of the recorded SEMG data 6 1 2 SEMG Stationarity For our applied SEMG processing algorithms the randomness plays a less significant role than the SEMG stationarity SMEG stationarity means that a single time window of the recorded SEMG is sufficient to describe the condition of the muscle activation Simplified this can be assumed when mean ergodicity i e the time invariant 6 EMG Signals Features Signal Acquisition Stimulation Artifact Removal 85 characteristic of the mean and more general of the autocorrelation function can be shown A signal is mean ergodic if T Rei 0 or Re gt 0 as T gt amp 0 In our case R T is the autocorrelation function of the SEMG Different studies also showed Inbar et al 1984 Paiss et al 1987 Popivanov et al 1986 that a SEMG sequence is stationar
24. 2001 The programming language a high level C like language allows one to program 7 Neuroprosthesis for Grasping 119 the recognition sequence the communication with other devices through ports and the choice of the speech recognition type Sensory Inc 2000 The internally used voice recognition algorithm itself can not be modified We have chosen to use the speaker dependent continuous listening mode Gareiss 2001 In this mode the speaker has to wake up the system by a speaker defined keyword a call Then the voice recognition system is listening and tries to recognize the following phrase by a classification process The following commands are used to control grasp with the neuroprosthesis grasp release abort stronger and weaker The three commands grasp release and abort control the stimulation sequences shown in The two commands stronger and weaker increase or decrease the stimulation amplitude using 8 different levels In order to achieve both the sequencing control grasp and release and the analog control of the stimulation amplitude the output voltage of the voice recognition system between 0 and 5 V is divided in four ranges see Figure 62 Within one range the offset voltage from the lower to the higher range boundary defines the stimulation amplitude level For example if the output voltage is 3 125 V grasp is initiated with amplitude level 4 2 5 V 4 0 1625 V The voice recognition control system s output voltage of
25. A trial consisted of eight sequences During the transitions either the pulse widths from 0 250 us or the pulse amplitudes from 0 12 mA for finger extensor and flexor muscles and 0 8 mA for the thenar muscle were linearly increased or decreased The exact timing parameters for stimulation patterns are listed in The control frequency of the pulses was 10 Hz i e the pulse widths or the pulse amplitudes were changed every 100 ms The stimulation frequency was 20 Hz sequence transition overlap hand close transition overlap hand open 10 mA 8 mA 12 mA F flexors thenar M F extensors 1 start 1 s PW Os 5s 2 5 4 s PW 0 6s 2s 5 4 s PW 0 6s 2s 3 2 7 s PW 0 4 s 2s 2 7 s PW 0 4 s 2s 4 1 8 s PW 0 28 2s 1 8 s PW 0 28 2s 5 0 9 s PW 0 1s 2s 0 9 s PW 0 1s 2s 6 2 7 s AMP 0 3 s 2s 2 7 s AMP 0 4 s 2s 7 0 9 s AMP 0 1s 2s 0 9 s AMP 0 1s 2s 8 1 s PW 0 2 s 5s end Table 7 The stimulation protocol of one trial consisted of eight concatenated stimulation sequences During the transitions either the pulse width was changed between 0 and 250 us marked with PW or the pulse width was constantly 250 us and the pulse amplitude was changed marked with AMP Two such trials were conducted one without and one with voluntary muscle contraction A Compex Motion constant current stimulator and Compex 5050MED self adhesive electrodes CompexSA 1998 were used to stimulate
26. Biofeedback sensors are attached to the tripolar electrode pads with three snap buttons per pad see Figure 53 Both the EMG and the stimulation electrodes can be placed on the patient s body during the whole day without causing any skin allergies or irritation Figure 53 Tripolar electrode pads are stuck on the skin above the selected recording muscle for example the M pars clavicularis of the contralateral deltoid muscle 7 Neuroprosthesis for Grasping 108 Experienced family members or health care personnel need not more than 5 min to completely mount the electrodes and to do the cabling 7 2 Electrode Placement This subsection describes the placing procedure of the self adhesive surface stimulation electrodes at optimal positions for SCI subjects The optimal position is defined by the following factors e the targeted function e the grade of muscle denervation of the targeted and of adjacent muscles e the displacement of the electrode due to motion and wear e the sensation of the subject The position of the electrodes that allows the best control or the strongest grip is not always the optimal position For some subjects it took several weeks to find the optimal electrode positions In many cases the subject s muscles are more or less atrophied what makes it even more difficult to find the muscles although they are not denervated A commonly practiced procedure is to place the anodic electrode distal to the targeted muscl
27. D D 1987 An externally powered multichannel implantable stimulator for versatile control of paralyzed muscle EEE Transactions on Biomedical Engineering 7 pp 499 508 Smith B Tang Z Johnson M W Pourmehdi S and Gazdik M M 1998 An externally powered multichannel implantable stimulator telemeter for control of paralyzed muscle IEEE Trans Biomed Eng 4 pp 463 75 Solomonow M 1984 External control of the neuromuscular system JEEE Trans Biomed Eng pp 752 63 Solomonow M Baratta R Miwa T Shoji H and D Ambrosia R 1985 A technique for recording the EMG of electrically stimulated skeletal muscle Orthopedics 4 pp 492 5 Takahashi K Hoshimiya N Matsuki H and Handa Y 1999 Externally powered implantable FES system Jap J Med Elec Biol Eng 1 pp 43 51 Takahashi K Kikuchi M Takeuchi S Hoshimiya N Mastuki H and Handa Y 1995 Externally powered implantable FES system In Proc 6th Int Symp on Micro Machine and Human Science MHS 95 Cat No 95TH8079 pp 121 6 Thorsen R 1998 Restoration of hand function in tetraplegics PhD thesis Technical University of Denmark Copenhagen Thorsen R 1999 An artefact suppressing fast recovery myoelectric amplifier IEEE Trans Biomed Eng 6 pp 764 6 Van B A Goudswaard P and Schomaker L R B 1984 Amplitude and bandwidth of the frontalis surface EMG effects of electrode parameters Psychophysiology pp 699
28. E g the stimulation intensities for the thumb muscles channel 3 and 4 are not varied as much as for the finger extensor and flexors channel I and 2 7 3 3 Digital SEMG Control The digital SEMG control strategy controls the grasp task using the pre processed SA free EMG signal from a voluntary controlled muscle like a Morse code The Compex Motion stimulator is capable to record and pre process the EMG signal in real time using one or two Compex EMG Biofeedback sensors The artifact free rectified and low pass filtered EMG activity as described in Chapter 5 that is higher than a predefined threshold level is interpreted as active and EMG activity levels below the threshold are detected as inactive Short and long active phases can be distinguished thus different codes can be generated The advantage of using EMG activity instead of 7 Neuroprosthesis for Grasping 122 a push button is the more direct commanding of the neuroprosthesis without any limb articulation This allows the subjects a more natural activation of the grasp and release action The portable system uses two different user interaction primitives to distinguish between grasp and release The hand grasp EMG activation pattern is chosen such that voluntary motion of the arm does not result in an activation of the grasp sequence Therefore two short active phases within 1 5s are chosen The release EMG activation pattern is chosen to be one long active phase sinc
29. M ext carpi l b pronation C7 M triceps active stretching of lat upper arm active functional M pamaris 1 elbow and wrist C6 C7 hand tenodesis M ext carpi uln active finger grasp M ext poll 1 extension M ext digit com C8 intrinsic thumb active wrist and hypoesthesia active hand muscles finger flexion med forearm M flex carpi M flex digit com Thl Mm lumbricales active fist and hypoesthesia active hand incl Mm interossei precision grip med forearm _ precision grip Table 1 Typical motor sensory impairments for segmental lesions of the cervical spinal cord and the resulting impaired hand function Adapted from Curt 1996 2 Principle and Function of Neuroprostheses for Grasping 37 Clinically the pathological hand function is classified in passive functional hand active functional hand and active hand In subjects with passive and active functional hand the intrinsic finger muscles and the long finger extensors and flexors cannot be voluntarily activated Hand opening and closing is performed indirectly using the so called tenodesis grasp In the tenodesis grasp a palmar grasp making the fist is performed by dorsally extending the wrist The finger flexor tendons are shortened by an active voluntarily performed wrist extension movement With a special treatment by the occupational therapist the finger flexor tendons and muscles become shortened such that a functional fing
30. Pappas and Sabine Mangold for their contribution in the FES project To work in such a research family was and further is a pleasure Great support from the medical side I also got from Armin Curt He always believed in the potential of FES opened us the way to clinical applications motivated patients and therapists to try our prototypes and educated and provided me with the medical knowledge necessary for developing assistive and therapeutic devices for SCI subjects On the technical side and for managing the project I was encouraged supported and advised by my supervisor Prof Manfred Morari head of the Automatic Control Laboratory ETHZ I was specially impressed by his way of leadership giving his people enough freedom to evolve own ideas and to put them into practice but decidedly intervening if projects were in danger to slip off the track He always evaluated our actions from a more general perspective contributed an additional point of view and brought in additional dimensions Also his humanity and concernment in critical situations were very formative and exemplary to me Besides our group I would like to thank Prof Dejan Popovic who helped our group with his enormous knowledge about the field of FES and from whom I learned to separate the wheat from the chaff He was always in our favor and provided us with information suggestions and advice The last and most intimate thanks belong as Milos used to say to my real boss my
31. SEMG signal The switching methods potentially cause additional transients and adaptive filters may have a slow convergence in the case the SA is changing as it is the case in FES applications Software artifact subtraction methods Blogg et al 1990 Kiss et al 1989 McGill et al 1982 Wichmann 2000 subtract a more or less pure SA from the mixed signal The presented methods differ in the way how the pure SA is obtained Sub motor threshold stimulation off nerve recording double pulse stimulation within the refractory period of the nerve fiber or ensemble averaging of the SA contaminated mixed signal are some of the presented methods For the control of neuroprostheses the proposed SA subtraction algorithms cannot be used because the produced SA changes with the action e g grasp or release over time During stimulation an a priori extracted SA cannot be adjusted to the measured SA in real time since the changes of the SA can be non linear and depend on many unpredictable factors e g electrode tissue impedance changes As a result residual or newly generated SAs can exceed the voluntary SEMG activity by orders of magnitudes if one of the above presented methods is used in combination with changing SAs 6 4 Moving Ensemble Averaging Stimulation Artifact Removal Algorithm The objective was to develop an algorithm that removes the SA generated by electrical stimulation from recorded SEMG of voluntarily activated muscles in real time e
32. a 9 5 case The digital and analog circuits of the stimulation device are built on separate boards This gives the flexibility to potentially combine different analog pulse generation boards with the same digital microcontroller board Either current or voltage regulated stimulators with different shapes of the stimulation pulses can be used The digital board provides control signals for two analog circuit boards with four stimulation channels each The stimulation device is battery powered and all the connections to the PC are galvanically separated by HCPL 2630 optocouplers 4 1 2 Digital Circuit Board The digital circuit board sequentially receives the stimulation data from a 200 MHz Pentium PC via the digital ports of the LabPC multifunction I O board The stimulation data consists of 1 the stimulation amplitudes and pulse widths for up to eight channels 2 the on off state of a so called doublet pulse feature For the stimulation channels 1 2 5 and 6 the doublet pulse feature generates two consecutive stimulation pulses doublets with 5 ms inter pulse delay instead of one single pulse Doublets reduce habituation and potentate the stimulus response Karu et al 1995 3 the actual stimulation frequency 4 the on off state of each channel and 5 the on off state of the high voltage DC DC converter From pulse to pulse all the parameters can be changed for each channel except the stimulation frequency and the on off state of the DC D
33. a deeper layer as it is the case for the finger extensors and flexors They can be stimulated with higher amplitudes but all muscles above the target muscle are activated as well unless they are denervated Sometimes a slight misalignment from the motor point of the target muscle reduces significantly the activation of the muscles closer to the skin surface Remaining activation of unwanted muscles has to be compensated either by co activating antagonist muscles or by using retainer splints or other orthotic devices 7 Neuroprosthesis for Graspin 109 Typical positions of the electrode pairs for finger extension finger flexion and thumb opposition flexion are shown in Figures 54 56 and 58 They are all used with SCI subjects with levels of lesion ranging from C4 to C6 7 2 1 Electrode Positions for Finger Extension If a hand opening prior to hand closing is preferred a pair of electrodes is placed on the posterior surface of the forearm as shown in Stimulation through these electrodes causes muscle contraction of the following muscles M extensor carpi radialis and brevis M extensor carpi ulnaris and M extensor digitorum and digiti minimi Electrode placement Finger extension Depolarization electrode Figure 54 The electrode placement of the finger extensor stimulation electrodes is not very critical The depolarization electrode can be located over the M externsor digitorum and is not very sensitive to displacement
34. analog input B This feature is used to control hand grasp with an analog sliding potentiometer By pushing the sliding potentiometer towards one end the grasp forces of the finger extensors are increased and by pushing the slider towards the other end finger and thumb flexors are contracted It is possible to use the analog input signals to trigger the start stop or to command the continuation of the stimulation sequences It is also possible to jump in the stimulation sequence depending on the measured sensor signal This feature will be discussed later in this chapter in Section 5 2 2 Stimulation Outputs The device generates rectangular current regulated stimulation pulses Four different pulse shapes are available of which one can choose one that is applied to all the channels Monophasic constant current pulses Figure 2da This mode is preferably used for research studies where it is important to generate a specific number of APs at a specific location The lack of charge compensation makes this mode unsuitable for long time stimulation because it can produce skin irritation Bipolar biphasic constant current pulses Figure 29p This pulse mode is preferably used for the stimulation of big muscles like the femoral muscle group where excitation of other unwanted muscles is unlikely The stimulation pulses are symmetric with respect to the anode and cathode Therefore APs are excited at the anodal electrode as well as at the cath
35. based The user interaction trigger criterion is checkgd in between these two icons If is detected the pattern player jumps to the service Figure 32 Different human interaction primitives can be used for the interactive control of the pulse width pattern generation in the time line a the user interaction primitive b the user branch primitives and c the user interrupt primitives Fast TTL trigger primitive This special primitive is used to trigger very accurately the stimulation pulses with a 5 V digital TTL logic signal on the falling edge negative slope from 5 V to 0 V External devices can trigger the stimulator with an accuracy of better than 500 us The other human interaction primitives have only a resolution of 100 ms 5 Portable FES System 74 Normally the stimulator continues to stimulate all the channels with the pulse frequency amplitude and pulse width that is set by the last instruction in the time line until the next primitive is read For example when the stimulator is waiting for an user interaction the stimulator remains stimulating with the pulse amplitude and pulse width set prior to the user interaction primitive Only the fast TTL trigger primitive always stops the stimulation after all four channels are synchronized and waits for a negative slope TTL trigger detected at the input C Internally all parameters for the next stimulation pulses that occur in the time lines are armed and the stimulator waits
36. bottom 5 Portable FES System 70 5 4 1 Stimulation Modes and Frequency The default stimulation frequency and the stimulation modes are set globally for all stimulation channels The following stimulation modes can be selected e monophasic amp biphasic see Figure 29 a and b e monopolar amp bipolar see Figure 29 b and c e non alternate amp alternate see b and d e time based pulse based In time based mode the specifications for how long a pulse width primitive is executed has a resolution of 100 ms As already mentioned in the Section 5 1 Basic_Concep the stimulation patterns are built with stimulation pulse width primitives with a specified execution time duration for each primitive Independent of the stimulation frequency pulse width ramps and constants that determine the stimulation patterns are specified in the unit seconds and a resolution of 100 ms In pulse based mode the duration of the stimulation pattern primitives are specified in number of pulses Only the pulse based mode allows one to change the stimulation pulse widths and amplitudes from pulse to pulse e master amp slave An unlimited number of stimulators can be synchronized to work in parallel One of the stimulators is set as the master device and all other stimulators are slaves The master stimulator dictates the control frequency and the slave stimulators follow this control frequency In single device mode the stimulator must be set to master
37. combined with a fast and intuitive graphical drag and drop programming technique makes the neuroprosthesis for grasping applicable in an early rehabilitation phase where changes of the subject s neurological condition are always present and demand an adjustment of the neuroprosthesis parameters the number of the stimulated muscles and the stimulationelectrode positions 1 4 Structure of the Thesis The thesis is structured into nine chapters The second chapter provides basic information about the physiological processes involved in muscle contraction using FES and describes the influence of the stimulation parameters pulse width pulse amplitude pulse shape and stimulation frequency It presents the concepts of the different currently used stimulation electrodes and their influence on muscle selectivity which is an important factor in neuroprostheses for upper extremities A subchapter briefly describes the tetraplegic subject focussing on the functional deficits in the upper extremities since he she is not only the customer of the system but also a part of it At the end of the chapter an overview of the currently existing and used neuroprostheses for grasping divided in implanted and surface FES systems is given The third chapter describes the concept of the ETHZ ParaCare FES systems and of the Compex Motion stimulator It points out the requirements for a modern and flexible stimulator needed for neuroprosthetic applications T
38. graphically programmed Two different control schemes are implemented An event triggered stimulation pattern generator software or a continuously controlled stimulation software can be chosen to control hand grasp The event triggered stimulation pattern generator software is used for the following control strategies 1 push button control 2 EMG event triggered control and 3 voice commanded control The continuously controlled stimulation software is selected for 4 analog slider control and 5 analog EMG control All control strategies use two normalized control variables one for the fingers and one for the thumb They range from 0 to 1 for finger and thumb extension and from 0 to 1 for finger flexion and thumb adduction flexion Both neuroprostheses software versions for hand grasp consist of the following modules a sensor signal acquisition module b sensor signal processing output is the control variable with a range of 1 1 and stimulation pattern generation module c stimulation parameter setup module d graphical interface for compensating the stimulation recruitment curves e data acquisition and data storage module 4 3 1 Sensor Signal Acquisition Module The rapid prototyping FES system for hand grasp allows the implementation of a variety of different sensor types The sensors are used to detect the subjects intention to grasp or to release an object with a desired grasp force Depending on the type of sensor the
39. if it affects all four stimulation channels although it is placed only in one channel dark green or if it has to be places in the time lines of all active channels light green The blue colored and the dark green colored primitives can be placed in the time line of any channel The light green colored primitives must be place in the time lines of all active channels A channel is considered active if the end primitive is not yet reached The seven different primitive categories are described in more detail Pulse width primitives The pulse width primitives blue icons play the most important role They define the stimulation intensities and the duration of the stimulation patterns Per channel two different pulse width ramp up and pulse width ramp down primitives four constant pulse width primitives a delay primitive holds the value of the last stimulated pulse width constant and a no stimulation primitive pulse width 0 us are available For all these primitives the duration of the pulse train can be 5 Portable FES System 72 entered in the text box below the primitive icon The shapes of the pulse width ramps and values of the constant pulse widths are defined in the pulse width sub window Pulse amplitude primitives The change amplitude primitive blue icon allows one to change the stimulation pulse amplitude at any time in the stimulation sequence The new amplitude is not set immediately to the new value when the primitive is
40. in a fast loop until a negative slope TTL trigger signal occurs General purpose primitives They have no influence on the pattern duration and provide the channel independent functions dark green icons playing a sound displaying a text or turning off the stimulator If the end primitive blue icon is reached in the time line all activities of this channel are terminated Randomization primitives The stimulation pulse frequency amplitudes and pulse widths can be randomly varied by a predefined percentage These primitives can be used for scientific applications There is some evidence that varying the stimulation pulse frequency can reduce fatigue significantly Graupe et al 2000 All available primitives are shown and specified in Table 5 Pulse Width Primitives Primitive name Description Param 1 Param 2 constant pulse width Generates a pulse train constant pulse width Width us Duration s 4 different values are available per 0 16000 0 1 25 5 channel pulse width ramp up Profile for changing the pulse width 2 Width us Duration s different profiles are available per channel 0 16000 0 1 3 1 profiles are described with 16 values pulse width ramp down Profile for changing the pulse width 2 Width us Duration s different profiles are available per channel 0 16000 0 1 3 1 profiles are described with 16 values no stimulation Pulse width is set equal to 0 Width
41. inputs are available e without signal pre processing Inputs A B C or A B e with EMG signal pre processing EMGA EMGB or EMGA EMGB An EMG Electromyographic signal is a measured small voltage lt 1 mV on a muscle surface It is produced by muscle APs and represents the muscular activity The chosen real time EMG pre processing algorithm calculates a rectified low pass filtered muscle activity level from a pre amplified SEMG and removes the stimulation artifact by artifact blanking more details follow in Chapter 6 e with EMG signal pre processing and integrating the result IEMGA IEMGB or IEMGA IEMGB Therefore the pre processed EMG signals EMGA and EMGB are integrated and a small constant value is subtracted from the result The subtraction of the constant value is performed to linearly decrease the integrated EMG activity if no additional EMG activity is measured Like this the IEMG activity level can be kept constant with a small amount of additional muscle activation It decreases with no muscle activation and increases with a muscle activation that is bigger than the subtracted value Figure 33 User interaction sub window Here the seven user interactions A to F are set At the bottom of the window the user interactions used for the user branches and the user interrupt are selected 5 Portable FES System 77 Setting the Trigger Criteria A general trigger module detects the trigger criterion in the interaction inp
42. is described by a random process This claim often comes from qualitative observations of APs recorded from muscles The efferent nerves conduct many thousand APs refer Section 2 1 4 that are fired by the CNS asynchronously Indeed the electroneurographic ENG recordings from such nerves look quite random Thus the practitioners very often characterize the SEMG the muscular response to a nerve AP by Gaussian assumptions with mean and standard deviation However the histogram analysis presented in shows that the SEMG is not exactly Gaussian distributed white noise 6 EMG Signals Features Signal Acquisition Stimulation Artifact Removal 84 3 5 SEMG V 2 5 2 0 1 8 1 6 1 4 1 2 gt 1 0 0 8 0 6 0 4 0 2 0 0 em 1 4 1 8 2 1 2 4 27 3 0 3 4 3 7 Amplified SEMG G 1400 V Histogram SEMG une Gaussian Distribution Figure 37 The histogram of a recorded SEMG signal 4 order band pass filter 100 4000 Hz sampling frequency 10 kHz is not exactly Gaussian distributed Nevertheless many authors make Gaussian assumptions in the characterization of SEMG recordings It is important to mention that the used SEMG electrode has a built in preamplifier with a gain of 1400 and an offset of 2 5 V This fact makes the application of commonly used statistical tests and classification algorithms rather difficult In most studies the randomness of the SEMG is not considered Duchene et al
43. lengths as shown in The skeletal muscle has its optimal operating point at a sarcomere length of about 2 2 2 um For example the optimal operating point of the finger flexors is at 30 wrist extension Skelettmuskel l I I I j I l l l I I ich der max Spannung isometrische Spannung des Maximums l N I N t 0 cP GaP a a a a A 12 14 16 18 20 22 a 30 32 34 36 Va 2 20 Aktin Myosin Sarkomer Bere Sarkomerl nge km nach Gordon u Mitarb Figure 8 Maximal isometric contraction force of skeletal and heart muscles Reprinted from Silbernagel etal 1991 2 1 5 Influence of the Stimulation Frequency A single stimulation pulse generates an AP that propagates along the nerve and results in a muscle twitch A single muscle twitch lasts only about 100 200 ms By applying trains of pulses 3 10 Hz the stimulated muscles experience tremor Increasing the stimulation frequency puts the single muscle twitches closer together they start overlapping to the extreme that the muscle is not able to relax anymore This continuous muscle contraction is called tetanization For stimulation frequencies above 25 Hz the tremor becomes very small and tetanic contraction is obtained Increasing the stimulation frequency smoothens the tetanic contraction further but has the disadvantage of increasing muscle fatigue During physiological contractions the muscle fibers are activated randomly through th
44. levels are switched and by a step size of 1 25 V the stimulator switches between grasp release and abort Menu Parameter Value Interaction input Input C Interaction type Push button Trigger criteria level 1 3 75 V peak valley 1 valley time 1 O I s longer shorter 1 longer all other values 0 Table 9 A voltage between 2 5 and 3 75 V indicates that the verbal grasp command has been detected Menu Parameter Value Interaction input Input C Interaction type Push button Trigger criteria level 1 3 75 V peak valley 1 peak time 1 O I s longer shorter 1 longer all other values 0 Table 10 A voltage above 3 75 V indicates that the verbal release command has been detected Menu Parameter Value Interaction input Input C Interaction type Push button Trigger criteria level 1 2 5 V peak valley 1 valley time 1 0 1 s longer shorter 1 longer all other values 0 Table 11 A voltage below 2 5 V indicates that the verbal abort command has been detected 7 Neuroprosthesis for Graspin 121 if Set Lookup table Pulse Amplitude Look up Tables Channel 3 Act 2 Amp Load Look up 3 Sensor Save Look up 4 Analog pulse amplitude control E chi Ch2 Ch3 Ch4 Figure 63 The look up tables of the analog stimulation amplitude control submenu can be programmed to set the stimulation amplitudes for each stimulation intensity level appropriate to each muscle
45. muscle functions but no finger functions became able to grasp hold and release objects used in activities of daily living and improved their level of independence They were the ideal candidates for using the system chronically as a grasp aid Incomplete SCI subjects could mainly profit from the system as a training device There is strong evidence that FES training improves their grasp capacity and plays a significant role in the reorganization of the remaining intact pathways and the plasticity of the central nervous system This last result could not be proven but will be the focus of a future multicenter study with the developed portable neuroprosthesis for grasping Kurzfassung Die Arbeit beschreibt die Entwicklung von Systemen und Methoden welche mittels funktioneller Elektrostimulation FES die Greiffunktion von querschnittgel hmten Personen verbessern Solche Greifhilfen werden auch als Greifneuroprothesen bezeichnet Die hier vorgestellten Greifneuroprothesen wurden speziell f r die Erstrehabilitation von tetraplegischen Patienten entwickelt und am schweizerischen Forschungs und Behandlungszentrum ParaCare der Universit tsklinik Balgrist in Zusammenarbeit mit dem Institut f r Automatik der ETH Z rich angewendet Die Greifneuroprothesen basieren auf dem Prinzip der Oberfl chenelektrostimulation und unterscheiden sich von kommerziell erh ltlichen Systemen durch eine h here Flexibilit t In einer ersten Phase wurden Prototypen f
46. nerve degenerates over time and leaves the muscle denervated In principle it is also possible to stimulate denervated muscle by directly stimulating the motor endplate In such a case the needed currents to generate muscle contraction are much higher than for innervated muscles see Figure 4 In practical applications with surface stimulation electrodes a long pulse duration of 200 300 ms must be applied With such a long pulse duration a relatively moderate current about 20 mA can be chosen But because of the long pulse duration only 2 to 3 pulses s can be generated per stimulation channel A tetanic muscle contraction can only be achieved on big muscles with multiple stimulation channels using the carrousel method alternating change of the stimulated location on a muscle using multiple electrodes Alternatively with shorter pulse duration and high currents of several hundred mA tetanic contraction can be achieved In this case special care has to be taken that the current densities on the skin do not become too high One has to use big electrodes with good contact to the skin For chronic treatment FES of denervated muscles can not be recommended because in contrast to spastic muscles a trained denervated muscles immediately atrophy if the intensive muscle training is interrupted Nevertheless the Vienna FES group chronically trains in a special FES program denervated lower limb muscles of SCI subjects Kern et al 1999 They could show
47. off all BOSFETs with a hardware switch A buzzer can be switched on in parallel to the stimulation electrodes to provide an acoustic feedback of the stimulation pulses 4 1 6 Asynchronous Communication between Stimulation Device and PC The PC software of the rapid prototyping FES system is programmed in LabVIEW and runs on Microsoft Windows95 98 operating system This operating systems has no real time capabilities No exact timing for the transfer of the stimulation data and for the execution of the stimulation pulses can be programmed for this operating system Therefore an asynchronous communication interface between the PC and the stimulation device is the only solution to this problem The stimulator generates an accurate timing of the stimulation pulses using the timers of the HC11 microcontroller A 8 Kbit dual ported RAM buffers the data between the stimulator and the PC For the data transfer from the PC to the FES stimulator three digital 8 bit ports port A B and C of the LabPC board are used They are set up to run in a special mode in which the ports A and B are used to transfer the data and the storage address and port C provides a hardware handshake with the dual ported RAM This hardware handshake prevents loss of data and acknowledges to LabVIEW that the written data is stored in the dual ported RAM All electrical connections between the PC and the stimulator are galvanically separated by optocouplers In addition to the hardw
48. order to open the hand The shape of the ramp function can be set individually for each subject Tests with SCI subjects have shown that a hand opening duration of about 1 second is a good compromise between a fast hand opening and the subjects comfort After the hand opening phase the finger extensors remain constantly stimulated for 2to 5s depending on the subject s grasp skills During that time the subject can place the fingers around the targeted object Then the stimulation of the finger extensors is decreased using the same stimulation pulse ramp function in reversed direction and the stimulation of the finger extensors is stopped marker in Figure 61 In order to close the hand the stimulation pulse width of the finger flexors is increased With a short time delay of 0 8 s also the thumb flexors are stimulated and the hand grasps the object The delay of the thumb stimulation is very essential It prevents the thumb from being grasped by the fingers Push button control Hand closure remains stimulated until the push button is pressed again If so the release sequence is initiated The fingers are opened and remain open for 2s to allow the subject to release the grasped object If the push button is pressed during any transition in the prehension task the time between the first push button touch and completed hand closure the system interprets this as an interrupt and changes from the grasp to the release sequence Like this the
49. portable and ready for daily use e two analog input channels the analog input channels can measure two independent voltage signals between O and 5 V with a maximal sampling frequency of 4 kHz e programmable chip card credit card like chip cards are used to save all user dependent information of the stimulation parameters This is the most outstanding feature that this device can provide compared to most other existing electrical stimulators This feature makes the stimulator completely user and application independent By inserting a specific chip card into the stimulator all its parameters and stimulation patterns can be adjusted for the desired application In its product palette Compex SA has chip cards for training atrophied muscles of the lower and upper extremities iontophorese stimulation for drug delivery denervated muscle stimulation and programs for pain treatment The following features are missing and because of that the Compex 2 stimulator cannot be used for scientific and neuroprosthetic applications e arbitrary programmable stimulation patterns e complete independence of the four stimulation channels e implementation of different sensor systems and man machine interfaces e pulse to pulse control of the stimulation amplitude and pulse width e analog control of the stimulation amplitude and pulse width e precise triggering of the stimulation pulses with known timing e stimulation of more than four muscle groups Th
50. repetitive angular conformation change of the myosin heads combined with a docking undocking process to the actin takes place in the following order 1 docking the myosin head to actin filament 2 changing the angle from 90 to 50 C 3 undocking and 4 returning to 90 makes the myosin slide inside the actin filaments what causes muscle contraction The underlying chemical processes are described in Silbernagel et al 1991 I gt 10 100 um 100 1000 um 1 Myofibrille 2 Muskelfaser Zelle 3 Faserb ndel A Structure of fasciated muscle fibers Aktin Tropomyosin Troponin Aktinfilament Sarkom r lt ae H Zo ne lt Aktin filament Z Scheibe Myosin Z Scheibe filament A Band 1 6 um Myosinfilament Myosink pfchen ae Pa a Myosinmolek l L l Band ae B Structure of a sarcomere 10 nm Beweglich keit pe Myosinschaft schweres Meromyosin C Myosin molecule Figure 7 Sarcomere anatomy Reprinted from Silbernagel et al 1991 2 Principle and Function of Neuroprostheses for Grasping 26 Each cycle causes a contraction of 1 of the muscle length A muscle can contract about 50 of its initial length It can perform the full contraction in 100 200 ms It was pointed out before that the muscle dynamics are non linear One of the reasons is the non linear maximal isometric contraction force for different sarcomere
51. shown in configuration A the wrist flexors M flexor carpi radialis and M palmaris longus are also activated and cause unwanted wrist flexion In subjects with partial denervation of the wrist actuators finger flexion without much wrist flexion can be achieved The remaining wrist flexion can be compensated either by voluntary co activation of the wrist extensors during finger flexion possible for C6 or lower lesioned subjects by stimulating the wrist extensor muscles during finger flexion to stabilize the wrist joint in a preferred 30 7 Neuroprosthesis for Grasping 111 extended angle or by using a wrist retainer splint Co contraction of the wrist muscles can be used to stiffen the wrist joint In a 30 wrist extension angle the finger flexor muscles have their optimal force length relation and therefore provide the strongest grasp Figure 57 The M flexor digitorum superficialis flexes the proximal interphalangeal joints of the second through fifth digits The M flexor digitorum profundus flexes the distal interphalangeal joints of the second through fifth digits Reprinted from Kendall et al 1993 If the subjects are not oversensitive to the electrical stimulation and if configuration A produces too much wrist flexion then configuration B can be applied It provides finger flexion without much wrist flexion when applying enough current The electrical field that is generated with such an electrode configuration can activate
52. sliding resistor as man machine interface see Chapter 7 3 4 The GUI shown in Figure 36 allows one to graphically edit the values of the four look up tables The look up tables can also be loaded from a text file The x axis represents the sensor signal voltage range and the y axis represents the stimulation pulse amplitude range between 0 mA and the actual set pulse amplitude This can be either the default amplitude see main GUI in Figure 30 or a in the time line programmed amplitude or the amplitude the user has adjusted by the control buttons on the stimulator The control frequency the frequency the mapping is performed without EMG pre processing is 100 Hz and with EMG pre processing is 10 Hz 5 Portable FES System 81 5 4 6 Chip Card Download Upload It was mentioned before that all subject dependent stimulator settings are stored on credit card like chip card The LabVIEW programmed PC software allows all the stimulator settings to be uploaded and downloaded from and to the chip card via a RS 232 serial port Thus the stimulator setting parameters on the chip card can be uploaded into the LabVIEW programming software and from there be stored on a hard disk From a library delivered with the LabVIEW programming software preset parameter settings can be written to the chip card The Compex Motion stimulator software firmware is programmed completely user independently By inserting a programmed chip card into the stimulator s card
53. subject can faster retry the grasp process if he she fails to grasp the object The stimulation sequences that are programmed with the PC software on the chip card of the portable system are shown in Figure 6 In case the subject fails to grasp the object he she can press the push button for longer than 1 s to interrupt the grasp The pattern player immediately for all channels jumps to the interrupt sequence see 61 right bottom and opens the hand within 0 5 s Then the pattern player jumps back to the initial position in the time line and waits for the push button to be pressed again for a next grasp trial 7 Neuroprosthesis for Graspin 11 Channel 1 Finger flex User interaction Channel 2 Finger ext T Jump to first see ea Channel 3 Thumb flex Channel 2 Time line of finger flexors Figure 61 The time lines and the resulting stimulation sequence modulation of the pulse widths of 3 stimulation channels in the case that the finger flexors and the thumb flexors are stimulated using the push button the voice controlled or the digital surface EMG control strategy To initiate and to release the grasp task the trigger criterion parameters for the User Interaction A primitive and in Figure 61 are set the following refer to Section 5 4 4 for the trigger criteria features 7 Neuroprosthesis for Grasping 118 Menu Parameter Value Interaction input Push button
54. techniques that can be easily implemented in actual electrical stimulators using a microcontroller for the real time processing of SEMG signals They blank or sample and hold the SEMG signal during the SA while loosing all signal information during that time For low stimulation frequencies and few stimulation channels these techniques can be applied to control a neuroprosthesis But for higher stimulation frequencies or many stimulation channels the blanking time especially with current regulated stimulators becomes too long and the SEMG signal looses its stationarity characteristics SA filtering methods Del Pozo et al 1978 Epstein 1995 Grieve et al 2000 Knaflitz et al 1988 Millard et al 1992 Parsa et al 1998 Solomonow et al 1985 reduce the SA using linear non linear or and adaptive filtering gain switching slew rate limiting or constant current voltage switching techniques They try to preserve more of the SA contaminated SEMG signal by reducing the SA spike low pass filters slew rate limiters Epstein 1995 Solomonow et al 1985 by reducing the slowly decaying SA tail gain or current voltage switching methods Del Pozo et al 1978 Knaflitz et al 1988 Millard et al 1992 or by estimating the SA and filtering it adaptive filter methods Grieve et al 2000 Parsa et al 1998 However because the SEMG signal and the SA overlap in time and frequency domains all applied filters influence the quality of the
55. that denervated and atrophied muscles recover after an intensive training of 9 months and more Such recovered atrophied muscles can be functionally stimulated tetanized with pulses with 10 30 ms pulse width Kern et al 2001 Subjects participating in this program report benefits for their cardiovascular system resulting in an improved quality of life 2 2 Stimulation Electrodes 2 2 1 Cuff Electrodes Cuff electrodes are wrapped around the nerve bundle where they stimulate the nerve at the closest possible position The cuffs are made of stainless steel or other conductive bio compatible material Two or more cuffs are Tripolar Spiral Cuff Electrode molded in a silastic tube that provides a stable distance and diameter of the cuffs to the nerves and therefore provides well defined and stable electrical properties Basic models exist for bipolar or tripolar electrode configurations newer models can have 12 or more channels Because the nerve bundles consist of many motorneurons that lead to different muscle groups cuff electrodes have a poor muscle selectivity Brindley et al Brindley et al 1986 developed tripolar cuff electrodes for sacral anterior root stimulation that have been successfully implanted in several hundred 2 Principle and Function of Neuroprostheses for Grasping 30 subjects for bladder contraction and voiding The Brindley electrodes are assembled like a book with 4 sheets and wrapped around three roots per ram
56. the dummy stimulator the stimulator generates the received stimulation pulse widths and amplitudes from pulse to pulse in real time One software allows one to trigger arbitrary stimulation sequences and to easily implement rule based controllers The other software controls on line and in real time the stimulation intensities of all channels using EMG sensors sliding resistors or other analog sensors The stimulation and control parameters are adjusted in GUIs New control concepts sensor systems and sensor data pre processing algorithms can be tested by adding new data processing modules in the LabVIEW software The main application of the stationary rapid prototyping FES system is to test new ideas for stimulating and controlling muscle groups to rehabilitate lost muscle functions in SCI subjects The portable FES system interprets and controls arbitrary stimulation sequences that are pre programmed on a credit card like memory card using a GUI software installed on a PC The GUI software uses a drag and drop technique to program the stimulation sequences This is done by sequentially placing icons called primitives on a time line that describes the chronological sequence of the tasks that will be carried out by a stimulation channel There are four such time lines one for each stimulation channel A total of 56 primitives are available in order to take the advantage of all the flexibility of the system The drag and drop technique makes it
57. us Duration s 0 0 1 25 5 delay Keeps the actual pulse width at the previous Time s no level for the given time interval 0 1 25 5 Pulse Amplitude Primitives Primitive name Description Param 1 Param 2 Tue change amplitude Changes the amplitude from the previous to Ampl mA Duration s A a new value in a specified time interval 0 120 0 819 2 change is linear Pulse Frequency Primitives Primitive name Description Param 1 Param 2 T change frequency Changes the stimulation frequency Freq Hz no 4 different values are available and they 0 250 apply to all stimulation channels Sequence Control Primitives z P Primitive name Description Param 1 Param 2 I jump back to marker Jumps back n times in the sequence to the No of jumps no A marker primitive where n 1 255 or infinite 1 255 or EA n 0 infinite er jump to first Jumps back to the beginning of the no no sequence synchronize Synchronizes independent stimulation no no sequences in all 4 stimulation channels 5 Portable FES System Human Interaction Primitives Primitive name Description Param 1 Param 2 user interaction This primitive waits for a specific user action to trigger a stimulation sequence Any sensor and triggering criteria can be used no no user branch Two trigger criteria set with the user interaction primitive are used to ge
58. width Activates a stochastic variation of the pulse width in the specified channel s The pulse width varies randomly about the nominal value within a specified range 0 100 following a uniform probability distribution function Deviation no random amplitude Activates a stochastic variation of the pulse amplitude in the specified channel s The actual amplitude varies randomly about the nominal value within a specified range 0 100 following a uniform probability distribution function Deviation no Table 5 The available set of primitives that can be used to build individual stimulation sequences for each stimulation channel 5 4 4 Settings for Human Interaction Primitives The human interaction primitives are configured in the user interaction submenu shown in Figure 33 The setup of the user interaction primitives plays the most important role The user branch and user interrupt primitives operate using the settings of the user interaction primitives The setup is performed in a two step procedure First the 5 Portable FES System 76 necessary sensor signal channel s the analog inputs A and B or the Compex push button input C and a signal processing algorithm are chosen as interaction input In a second step the trigger criterion is defined Each user interaction primitive is defined by one trigger criterion applied to one interaction input The following interaction
59. wife Claudia She was the person that supported me in all three attributes curiosity persistence and motivation through the thesis In evening long discussions she prickled my curiosity trained my persistence and enlivened my motivation She gave me the strength the relief and the necessary spunk Abstract The main objective described in this thesis was to develop systems and methods that use functional electrical stimulation FES to improve the grasp function in spinal cord injured SCI subjects Such systems are called neuroprostheses for grasping The transcutaneous surface neuroprostheses were developed for neurologically not stable SCI subjects and mainly applied during their first rehabilitation at ParaCare University Hospital Balgrist Zurich in a collaboration with the Automatic Control Laboratory of the Swiss Federal Institute of Technology Zurich For these subjects highly flexible systems are required that are commercially not available The existing implantable technology cannot be applied that early In a first phase successfully functioning prototypes of neuroprostheses for grasping were developed Experiments with SCI subjects demonstrated that the neuroprostheses for grasping can significantly improve the quality of life of SCI subjects Since the first phase established the feasibility of using neuroprostheses for grasping to effectively improve the SCI subjects hand function the project was focussed on resolving a nu
60. with percutaneous needle electrodes The stimulation patterns for the different stimulation channels were derived from standardized EMG data from able bodied subjects The stimulation patterns could have trapezoidal trajectories The system generated monopolar constant voltage stimulation pulses ranging from 0 to 15 V The grasp tasks were commanded with a head tilt switch or respiration controlled pneumatic commands or with a wrist watch push button In a second generation the number of stimulation channels were increased to 30 Since 1994 in a collaboration with NEC Inc the Sendai group developed a fully implantable 16 channel stimulator 200 of these stimulators were manufactured Takahashi et al 1999 Takahashi et al 1995 Thus far the NEC FESMate stimulators have been almost exclusively implanted for research purposes Annually the Japanese Ministry of Education Science and Culture finances the implantation of 10 FES systems at the Sendai FES Clinic The portable controller box for the implanted system and the percutaneous stimulators which are used for therapeutic electrical stimulation TES are programmed with the same PC system The trapezoidal stimulation pattern for the different muscle groups are determined heuristically using standardized EMG patterns For the implanted system the sets of stimulation patterns are downloaded via RS 232 serial port to the controller box After receiving a trigger signal from the control sensor mostl
61. 6 All other control strategies require as pre processing only linear gain and offset changes of the sensor signals Because of the modular concept of the software further control strategies can easily be added Trigger Unit en Rolaasing rule EI OR Grasping ge 05 AND WOT C3 Figure 19 In the event triggered stimulation pattern generator software the trigger unit combined with the rule based controller activates the stimulation patterns The event triggered stimulation pattern generator software shown in differs from the continuously controlled stimulation software in the trigger unit Figure 20 and the stimulation pattern generation module Figure 21 In the trigger unit up to 7 different trigger conditions can be defined The recording channel the processing algorithm and a trigger criterion can be chosen for each condition In a sub window one of the trigger criteria positive or negative slope above or below threshold peak or valley or double peak or double valley can be defined In addition the trigger conditions C1 to C7 can be operated with logical statements Logical AND OR and NOT operations are used to build the logical statements The first of the two statements defines the grasp rule and the second statement defines the release rule 4 PC Based Rapid Prototyping FES System 58 Trigger input Preprocessing Trigger criteria detect Trigger Unit Releasing rule C1 OR C2 Grasping rule C5 AND NOT C3 Fi
62. C converter that are the same for all stimulated channels The digital board converts all data from the PC to accurately timed signals for the analog circuit board using a Motorola MC68HC11 microcontroller The core of the digital circuit board consists of a MC68HC11 evaluation board EVB including the microcontroller eight Kbytes battery backup static RAM and an eight Kbytes EPROM that stores the monitor program and the final software code For testing purposes modified assembler code can also be downloaded to the static RAM via a RS232 serial communication port instead of reprogramming the EPROM The monitor program can either execute the software code in the EPROM or in the static RAM Further the main power supply the latches for stimulation pulse multiplexing an address logic GAL and an eight Kbit dual ported RAM that synchronizes the asynchronous communication between PC and stimulator are also implemented on the digital circuit board shows the stimulation parameter specifications and the timing constraints Parameter Range Resolution Processing Time Dim Multiplexing 2x4 channels on off off on 200 us on 2ms on off 80 us no processing time max 10 ms DC DC converter 2x on off 200 V Double pulse interval on off 5 ms fix Channel enable 8 channels onoff OOOO 0 O no processing time Table 2 Stimulation parameter specification and processing time 4 PC Based Rapid Prototyping FES System 50 4 1 3 Stimu
63. FES field like Liberson Liberson et al 1961 who is regarded as the inventor of FES and Long et al Long et al 1963 who proposed the first FES based system for grasping used surface electrodes for the stimulation of lower and upper extremities Later especially for upper limb FES applications the demand for a higher muscle selectivity motivated people to start using percutaneous intramuscular needle electrodes They can be inserted very precisely to the targeted muscles even if the muscles are in deeper layers Unfortunately the danger of electrode breakage and infection is always present with percutaneous electrodes so FES groups in different centers started to develop fully implantable systems The electrodes developed with such systems have the potential to be very selective while the risk of infection is reduced Also the stimulation pulses do not have to pass through the skin where they can stimulate the skin receptors that provoke pain and unwanted reflexes Generally stronger forces can be produced with implanted electrodes Because the electrodes can be placed closer to the motor nerve a better muscle selectivity can be obtained as well Deep muscles that are difficult to reach with surface electrodes can easier be stimulated with implanted electrodes Compared to surface electrodes implanted ones do not need daily donning and doffing The wires from the stimulator to the electrode can also be routed inside the body This provides
64. G amplifiers and the electrical stimulators SA blanking SA filtering and SA subtraction methods The SA blanking methods can be used in real time if the stimulation frequency is low and only a few channels are used because the SEMG signal is lost during blanking The other methods were proposed for an offline SA removal e A developed and tested algorithm that subtracts an extracted SA from the SA contaminated SEMG signal The SA is extracted from the SA contaminated SEMG signal using an ensemble averaging algorithm To allow an adaptation of 82 6 EMG Signals Features Signal Acquisition Stimulation Artifact Removal 83 the extracted SA to changes a moving window with exponential forgetting is used The algorithm can be applied in real time For the control of our neuroprostheses we also used electromyographic EMG signals recorded from muscles the subjects can contract voluntarily EMG activity is a random like weak electrical voltage signal produced by muscle APs during muscle contraction The SEMG recording electrodes perform due to their big size compared to a single muscle fiber and distance to the muscle membrane a temporal spatial summation of many single muscle APs The asynchronous firing of the nerve APs that propagate through hundredths of nerve fibers to a single muscle during voluntary muscle activation are converted into muscle APs that can be recorded as a random looking SEMG signal An increase of muscle activation resu
65. MG controlled neuroprosthesis J Neurorehab pp 251 5 Keller T and Popovic M R 1999 Rapid stationary FES prototyping system for walking and grasping neuroprostheses In Proc 4th Ann Conf IFESS Sendai Japan pp 159 62 Keller T Popovic M R Pappas I P I and M ller P Y 2001 Functional electrical stimulator Compex Motion In Proc 7th Vienna International Workshop of Functional Electrical Stimulation Vienna Austria pp 22 5 Kendall F P McCreary E K and Provance P G 1993 Muscles testing and function Williams amp Wilkins Baltimore Kern H Hofer C Modlin M Forstner C Raschka Hogler D Mayr W and Stohr H 2001 FES training of denervated muscles in human In Proc 7th Vienna Int Workshop on FES Vienna pp 2 5 Kern H Hofer C Strohhofer M Mayr W Richter W and Stohr H 1999 Standing up with denervated muscles in humans using functional electrical stimulation Artif Organs 5 pp 447 52 Kiss I and Shizgal P 1989 Improved artifact rejection and isolation of compound action potentials by means of digital subtraction J Neurosci Meth 3 pp 219 29 Knaflitz M and Merletti R 1988 Suppression of stimulation artifacts from myoelectric evoked potential recordings IEEE Trans Biomed Eng 9 pp 758 63 140 Kornfield M J Cerra J and Simons D G 1985 Stimulus artifact reduction in nerve conduction Arch Phys Med Rehabil 4 pp 232 5 Lanmu
66. MG recording M pars clavicularis Figure 43 Three pairs of stimulation electrodes were placed on the finger extensors channel 1 the finger flexors channel 2 and the thenar channel 3 muscles on the forearm Two Compex EMG Biofeedback sensors recorded the voluntary SEMG activity of the M extensor carpi radialis and the M pars clavicularis of the contralateral deltoid muscle that are shown in the left graph Pre study tests showed no significant difference of the SEMG signal quality from unprepared skin compared to shaved and rubbed skin For the study we decided to leave the skin unprepared because the EMG control strategy should also work under those more challenging conditions since they are more realistic like those in real applications 6 EMG Signals Features Signal Acquisition Stimulation Artifact Removal 96 raw raw contro SA SA free 7 W SEMG hu extracted i control lt N i i I I I i I l 1 extracted i sa I processed control U AN I 1 l i y processed SEMG with removed SA processed SA raw SMG Hemp IODINE without SA A B A B Tee j 100 ms Figure 44 shows the processing steps that were performed to obtain the processed SEMG signal with removed SA from the raw SEMG with the SA and qualitatively compares curve with a raw SEMG without SA The SEMG curves a
67. Trigger criterion level 1 3 5 V peak valley 1 valley time 1 0 2 s longer shorter 1 longer all other values 0 Table 8 With this simple setup of the trigger module of the portable system the push button activation is detected By setting the time 1 longer than 0 2 s trigger artifacts like voltage spikes are suppressed In the User Interrupt trigger criterion User Interaction B the parameter time 1 is set to 1 s All other parameters are the same as for User Interaction A The consequence of such a setup is that if the push button is pressed shortly between 0 2 and 1 s grasp or release is initiated and if the push button is pressed longer than 1 s the stimulation is interrupted For some subjects the finger extensors are not stimulated The reason is that some subjects prefer to place the fingers around the object and to hold the object with the passive stiffness of their fingers and then press the push button to close the fingers with a strong FES produced grasp force We have experienced with those subjects that the loss of voluntary hand opening is not that limiting for daily activities The stimulation patterns for those subjects are modified such that only the finger and thumb flexors are activated see Figure 61 curves for finger and thumb flexion 7 3 2 Voice Control A voice recognition systems for the control of a neuroprosthesis was used first by the Haifa group in 1990 Nathan et al 1990 However this control
68. a typical grasp release and abort sequence is shown in release grasp zZ stronger 2 gt Q gt o Figure 62 The output signal of the voice recognition system generated by a DAC is divided in four ranges of which the upper three are used for abort grasp and release commands Within each range the stimulation amplitude is regulated using the analog input and the amplitude look up tables As mentioned before the stimulation sequences in the portable FES system Compex Motion are programmed equally as for the push button control strategy Figure 61 The parameters for the user interaction primitives are set differently see Tables 9 10 11 The parameters of Table 1 are used for the user interrupt primitive In addition to the stimulation sequence triggering the analog output voltage of the voice recognition system is mapped to predefined stimulation amplitudes using the look up tables of the analog control capabilities of Compex Motion see Section 5 4 5 and Figure 63 With the commands stronger and weaker the stimulation intensities for each muscle are adjusted according to the look up tables see Figure 63 The look up tables are divided into four ranges Each of it has a range of 1 25 V In each range the mapping of eight 7 Neuroprosthesis for Grasping 120 levels is done equally such that the eight amplitude levels set the same stimulation amplitudes By a step size of 0 1625 V amplitude
69. ackwards the finger and thumb flexors are increasingly stimulated Instead of a linear sliding resistor a one dimensional touch pad from Interelectronic Inc based on the force sensitive resistor FSR technology can be used Figure 64B Figure 64 Three different types of sliding resistors A Conventional sliding resistor B FSR based sliding resistor and C Standard potentiometer that is unsuitable as a control interface for tetraplegic subjects 7 Neuroprosthesis for Grasping 124 The resistor of the 20x105 mm FSR pad changes linearly to the touched x position It is rather insensitive to the applied force unlike the normal FSR that are commonly used as foot switches The pad can be used as a voltage divider same as a normal sliding resistor By connecting a capacitor in parallel to the output the divided voltage remains constant even when the pad is not touched anymore Like this the function of the pad is exactly the same than of a potentiometer based on a sliding resistor The pad has the advantage of being easier to mount on an arm support of an electrical wheelchair than a common sliding resistor In the time lines of all four stimulation channels in the Compex Motion programming software the stimulation pulse widths are set constantly to 250 us and the constant pulse width primitives are executed in an endless loop The output voltage of the sliding potentiometer is recorded with the analog input A or B It controls the pulse amp
70. al time 3 Concept of the ETH ParaCare FES Systems 44 2 Scalable computational power The decision to take a standard PC and the Windows 95 98 OS as hardware platform was mainly driven by the demand for high computational power that can occur in real time processing of physiological data Of course one always has to have in mind when developing a control strategy that at a later stage it has to be packed into a low power portable system with limited calculation power For testing concepts however a straight forward programming without much time consuming optimization of the algorithm requires a fast CPU Due to the fact that a standard PC is used and also due to the upgrade compatibility of improved LabVIEW versions the stationary FES prototyping system can always be improved in performance without changing the software concept New more demanding algorithms can be added and implemented modularly 3 Flexibility of using different software programs and setups The stationary rapid prototyping system can be used as a neuroprosthesis for walking in combination with a treadmill or as a neuroprosthesis for grasping with different control software programs The type of neuroprosthesis that is implemented only depends on the used LabVIEW program and can be changed within minutes This high flexibility makes the system very versatile It can be used for training the subjects muscles for determining the gait or grasp patterns of different neuropros
71. are handshake a software handshake is implemented as well Each time the microcontroller reads all stimulation parameters from the dual ported RAM a read data byte is written by the microcontroller in the dual ported RAM On the other hand LabVIEW overwrites the read data byte with written whenever new data is written by the PC The stimulator stops the stimulation if the microcontroller detects a read instead of written byte and waits until new data is written from the PC to the stimulator and the written byte is set 4 PC Based Rapid Prototyping FES System 53 4 1 7 Multi Function Board The National Instruments low cost multifunction card LabPC National Instruments Corp 1994 was chosen for the data acquisition and communication with the stimulator Its main features are e A 12 bit A D converter with eight multiplexed channels an overall sampling frequency of 83 3 kHz and an input voltage range of 5 to 5 V e Two double buffered 12 bit D A converters e Three 8 bit digital I Os e Windows 95 98 NI DAQ drivers for LabVIEW with the capability of hardware buffered acquisition and buffered analog output The system is able to acquire eight sensor signals and to store them to the hard drive with 8 kHz sampling frequency in parallel to the real time data processing and the stimulation with 25 Hz loop frequency 4 2 Assembler Software of the Stimulator The software for the HC11 microcontroller is written in assembler programm
72. ated muscles require a 100 1000 times longer stimulation pulse width than innervated muscles Data from wrist extensor surface stimulation stimulation frequency 35 Hz Adapted from Baker et al 1993 Between the motor threshold and the near maximum motor response the increase of the stimulus intensity excites more and more nerve fibers that generate more and more muscle force The stimulus force or stimulus torque relationship is depicted in so called recruitment curves They show the static non linearity of the force output of an electrically excited nerve muscular system shows the recruitment curve of a quadriceps muscle for isometric contraction The system dynamics also behave non linearly The non linear static and dynamic properties of the nerve muscular system have been described Dorgan et al 1997 Hill et al 1975 Huxley 1957 Riener et al 1996 and reviewed Winters et al 1990 Zahalak 1992 Zajac 1989 by many authors using several different models 2 Principle and Function of Neuroprostheses for Grasping 24 120 100 Torque ft lbs S 8 20 40 60 80 1 00 120 140 Amplitude mA Figure 5 Curves that show the static stimulus intensity force relation are called recruitment curves The data was obtained from a quadriceps femoris with surface electrodes pulse duration 100 us and a frequency of 35 Hz Reprinted from Baker et al 1993 2 1 4 Muscle Contraction The detailed mechanisms that cause musc
73. bout 80 mV with respect to the outside of the nerve cell An AP can be characterized as a short depolarization of a nerve fiber with a duration of approximately 400 us During that time the inside nerve potential changes from 80 to 40 mV as a result of fast opening of sodium selective channels and the inflow of sodium ions driven by the large sodium concentration outside the nerve and the voltage gradient Shortly after the sodium inflow potassium channels open potassium ions flow out and a repolarization takes place In a third phase the osmotic pressure reestablishes the ionic concentrations of the resting condition see Figure 2 2 Principle and Function of Neuroprostheses for Graspin 20 Aen Ina kurzzeitig erh ht Overshoot 3 Repolarisation gx erh ht A Depolarisation und Repolarisation Potential Leitf higkeit g Ruhepotential Hyperpolarisation Zeit Figure 2 Electrical nerve properties during an action potential Reprinted from Silbernagel et al 1991 An AP can be triggered by an electrical stimulus that is applied to the excitable tissue of a nerve fiber with a pair of stimulation electrodes A short current pulse depolarizes the muscle nerve fibers close to the cathode The current pulse induces a flow of positive ions from the anode to the cathode and a flow of negative ions in opposite direction The positive charge is absorbed at the cathode Close to the cathode the posit
74. cause of the nature of all Microsoft Windows operating systems all software packages programmed for this OS are not executed in hard timing real time LabVIEW supports most standard PC interfaces like RS232 serial ports and many non standard interfaces like National Instruments data acquisition boards Using a standard 200 MHz PC soft timing closed loop real time applications can be programmed for control frequencies up to 40 Hz This fulfills the timing requirements for human controlled neuroprostheses for walking or grasping Three different versions of the controller software one for walking and two for hand grasp were developed for the rapid prototyping FES system The two controller software versions for neuroprostheses for grasping are subject of this section The proposed rapid prototyping controller software for hand grasp controls two different tasks e Task 1 take an object and hold it grasp e Task 2 release an object release 4 PC Based Rapid Prototyping FES System 56 Depending on the controller software version either pre adjusted stimulation patterns that perform either a palmar or a key grasp are executed In some control strategies also the grip force can be varied at any time For both tasks up to 4 stimulation channels can be used The grip force is controlled by changing the duration of the stimulation pulses The adjustment of the stimulation pulse widths for the different muscles that generate a specific grip force are
75. ce our objective is not to manufacture stimulators and since assembling the devices is time consuming we decided to find an industrial partner to manufacture the stimulators for us Because our portable systems are not CE labeled they can only be used for basic research with ethics committee approval but not for clinical use The ethics committee approval gave us in the past enough room to study the feasibility of our neuroprostheses Nevertheless there is a need for a commercial CE approved system that can be given to our SCI subjects for clinical treatment and for home use 3 Concept of the ETH ParaCare FES Systems 45 In 1999 we established a collaboration with Compex SA one of the leading electrical stimulator manufacturers world wide The objective of the collaboration was to enhance their state of the art stimulator Compex 2 a therapeutic electric stimulator The hardware can potentially provide all necessary features that are needed for neuroprostheses and other FES research applications but the stimulator s software is not designed for such applications The main features of the Compex 2 devices are e compact size and portability its size is 120x55x180 mm it weights 420 g The stimulator can be carried around like a MP3 player e low battery consumption the stimulator can work for 8 to 12 h if all four stimulation channels are used e rechargeable battery the rechargeable battery can be fully charged in 2 h It makes the system
76. cle e g on the quadriceps muscles more force can be generated with symmetric biphasic pulses 2 1 7 Current or Voltage Regulated Stimulation For FES two different types of neurostimulators are used voltage regulated and current regulated stimulators The excitation of an AP with a voltage regulated stimulator depends very much on the impedance of the underlying tissue Muscle contraction is much weaker if the voltage loss between the electrode and the tissue is higher A change of the impedance occurring from dried out skin under the surface electrodes affects strongly the muscle response Current regulated stimulators with their very high output impedance are more stable and independent of large impedance changes They provide a better control of the muscle contraction than voltage regulated stimulators because they compensate electrode tissue impedance changes On the other hand if the electrode tissue impedance becomes partially very high a high current density can cause skin irritation or even skin burns under the part of the electrode with the lower impedance In this case and in the case of an electrode failure the current regulator increases the 2 Principle and Function of Neuroprostheses for Grasping 29 stimulation voltage up to its limit Stimulation voltages of several hundred volts can be produced 2 1 8 Stimulation of Denervated Muscles If a motor neuron looses the connection to the spinal cord due to a peripheral lesion the
77. control signals for the stimulator are transmitted via radio frequency using a magnetic coil Since the inductive coupling is very sensitive to misalignment between the stimulator s receiver coil and the transmission coil precise positioning of the coil above the stimulator implant in shoulder region is required The shoulder position transducer is taped on the skin overlying the sternum and is controlled with the contralateral shoulder Johnson et al 1990 It monitors two axes of shoulder motion protraction retraction and elevation depression The control strategy can be varied for different shoulder motion capabilities of the individual subjects In the normal case the protraction retraction axis is used as control signal for hand opening and closing The shoulder elevation depression axis is used for logical commands such as holding a stimulation level or to establish a zero level for the protraction retraction axis An additional switch allows toggling between palmar and lateral grasp Keith et al 1988 Often in combination with the implantation of the Freehand system an arthrodesis of the interphalangeal joint of the thumb is done This surgery simplifies thumb positioning and allows the subject to have more force at the thumb tip with fewer muscles stimulated In subjects where the synchronization of the finger flexors is poor the index finger is not nicely flexed with the other fingers an additional surgical intervention can be done The l
78. cy 5 Portable FES System 68 5 3 Compex Motion Controller Program Firmware The stimulator controller program is programmed using the timer controlled multitasking features of the HC11 Due to limited memory space and computational power it is implemented in assembler code Most of the basic routines like the serial port communication routine the stimulation pulse generation routines and the display driver routines were taken from the standard Compex 2 stimulator software Only minor changes had to be introduced Other subroutines such as the display routines the stimulation pattern player primitive parser the data acquisition routines and the data processing routines were newly written Apart from the display routines and the serial communication routines all subroutines are timer controlled and run in real time Different timers see Table 4 are used to control the control frequency the watchdog and time counter the stimulation frequency and the data acquisition frequency The timer concept allows the stimulator to process multiple tasks in real time using a single processor Special care has to be taken that all tasks require less processing time than given by the timer interrupt Routine Timer frequency Called every milisecond 1 kHz 1 ms battery scan 1 ms watchdog 10 ms control frequency 100 ms slow A D scan 10 ms pulse sequence interpreter timed mode 100 ms look up table routine timed mode 10 m
79. d 7 ms the third 14 ms and the fourth 17 ms after the first pulse Stimulation Pulse Bosfet Switch 4 Bosfet Switch set Eyman Bose oui 3 i osfet Switch 2 Bosfet Switch ON _ Bosfet Switch 1 Voltage DAC set Voltage DAC Voltage DAC ON Stimulation Pulses 500 1000 1500 2000 2500 3000 3500 0 Time ps Time ms Figure 15 The left plot shows a stimulus pulse and the BOSFET switch timings The right plot shows the timings of all four stimulation channels when single pulse is activated The charge balanced stimulation pulses of the four stimulation channels are demultiplexed using eight two per channel high voltage analog BOSFET switches All BOSFETSs are controlled with an 8 bit latch 74HC374 The controlling signal for the BOSFETs are switched on 250 us prior to the stimulation pulse width The off on delay of the BOSFET is specified with 220 us The BOSFETs for single pulses are enabled for 2 ms and for double pulses for 7 ms 4 1 4 Power Supply For safety reasons the stimulation device is battery powered A 8 4 V Ni Cd battery pack with a capacity of 3 2 Ah allows 10 hours of stimulation Two voltage regulators one for the digital and the analog low voltage circuits and one for the high voltage DC 4 PC Based Rapid Prototyping FES System 51 DC converter transform the battery voltage into stable 5 V voltage sources The battery voltage is monitored and a sav
80. d selling the devices in some other countries The NESS Handmaster is a forearm orthosis that stimulates the finger flexors and extensors and the thenar muscle group Ijezerman et al 1996 It can be easily mounted by the subject using a clasp system that traps the forearm between two plastic shells and fixes the electrodes attached to the inner side of the shells to the dedicated stimulation sites The orthosis can be mounted without voluntary finger activity The stimulator box equipped with the power button a push button and an adjustable potentiometer is connected to the orthosis with a fixed cable Tests carried out with the Handmaster showed that the system is very convenient for stroke subjects whereas SCI subjects with partial muscle denervation can only benefit with some restriction There is not enough freedom for placing the electrodes inside the shells which makes it difficult to adjust the electrodes for all potential users The neuroprosthesis is too short to stimulate the finger flexors at a proximal position of the forearm where the wrist flexors are less activated Another limiting factor is the stiff construction of the orthosis that limits the range of motion With an attached Handmaster supination of the hand is not possible Although the splint causes some limitations in the range of motion and restricts the electrode placement on the forearm the design of the splint is very good as it can be put on and off by most users withou
81. de leading producer and patent owner Axelgaard et al 2001 of self adhesive electrodes is Axelgaard Mfg Co Inc 2 2 6 Other Electrodes for Transcutaneous Stimulation Other electrodes for surface stimulation are metal plates covered with fabric tissue or carbon electrodes Both types are not self adhesive and require water or special electrode gel to equally distribute the electrical current over the electrode surface The wetted fabric tissue equally distributes the current over the entire metal plate in order to prevent skin burns With this type of electrode one has to be careful that the electrode does not dry out In the best case if completely dry such a dried out electrode isolates the metal plate from the skin But while drying out unequally distributed electrical fields under the electrodes may cause severe skin burns Carbon electrodes are more safe The carbon rubber has a rather high electrical resistance that prevents high voltage drops in different regions of the electrode and provides a better current distribution Additionally the rubber always keeps the skin underneath the electrode a little bit humid by the sweat Carbon rubber electrodes and tissue covered metal electrodes are fixed to the skin with elastic straps or built in a garment or brace as it is the case for example with the Handmaster neuroprosthesis 2 2 7 Discussion Implanted Electrodes versus Surface Electrodes In the beginning of the 1960s pioneers in the
82. der Blair et al 1933 Natural voluntary contraction with a relatively weak force e g a well controlled precision grasp mainly involves fatigue resistant type I muscle fibers that are innervated by motorneurons with a small diameter The large fast nerve fibers that are easier to excite by FES innervate type II muscle fibers Those muscles have a fast high twitching force but are fast fatiguing For normal grasp tasks type I muscle fibers are used and on demand for faster reaction or a higher force type II fibers are recruited In FES with lower stimulation intensity first large nerve fibers connected to type II muscles are recruited and only with an increased stimulus also small nerve fibers connected to type I muscle fibers start being recruited The question how a nerve has to be stimulated to obtain the best response can be answered as follows The stimulus signal has the task to provoke a depolarization of the nerve in order to excite an AP The osmotic pressure balances the voltage inside the nerve fiber to about 80 mV therefore the artificial depolarization of the nerve has to be faster than the re polarization maintained by the membrane charge pumps This demands a sharp slope of the stimulation current In order to provide a fast depolarization good stimulators have stimulation current slopes greater than 5 10 A s In practice the excitability of the nerve is measured by recording the motor thresholds the weakest stimulus that provo
83. desis grasp Both systems stimulated the finger flexors if the artifact blanked rectified and binary integrated EMG signal of the wrist extensor muscles was above a threshold value The system acted like a tenodesis grasp amplifier similar to the Bionic Glove The system proposed in Keller et al 1998 provided a continuously controlled grasp force using real time measured EMG activity from voluntary controllable muscles and will be discussed later in this thesis 3 Concept of the ETH ParaCare FES Systems The ETH ParaCare FES systems are basically two different systems one is a stationary rapid prototyping system and the other is a portable system named Compex Motion The stationary system can be modified very quickly and is used to explore new concepts Its control software is programmed in LabVIEW a graphical programming language specially designed for system control data acquisition and signal processing The portable system has been developed to provide the promising concepts to our patients on a daily basis In general commercially available FES systems lack flexibility in the way how the stimulation patterns are programmed and how the systems are controlled The few commercially available Neurostimulators like Automove Stiwell Compex 2 or others have been developed for therapeutic applications e g for the training of atrophied muscles or for building up muscles The stimulation patterns of most of these systems are often very
84. dths and constant amplitudes Figure 48 A and B or with changing the amplitudes and constant pulse widths Figure 48 C Changes of the pulse amplitudes at constant pulse widths of 250 us do not show such strong changes of the SA Even a SA generated with a 4 mA stimulus shows a similar SA shape as one with a higher amplitude although such a stimulus is already below the motor threshold 6 E MG Signals Features Signal Acquisition Stimulation Artifact Removal A PW transition in 1 8 s from hand closing to hand opening from 100 us finger and thumb flexion to 100 us finger extension in 600 ms 100us 75us 50us flex 25us flex 75us 100us raw SA flex flex 25us ext 50us ext ext ext PW change proc SA PW change B fast PW transition in 0 9 s from hand opening to hand closing from 150 us finger extension to 150 us finger and thumb flexion in 500 ms 150us 100us 50s flex 100us 150us raw SA ext ext 50us ext flex flex PW change proc SA PW change C fast AMP transition in 0 9 s from hand opening to hand closing from 6 mA finger extension to 6 mA finger and thumb flexion in 500 ms 6 mA 4 mA 2 mA flex 4 mA 6 mA raw SA ext ext 2mA ext flex flex AMP change proc SA AMP change 10 Figure 48 A and B For weak stimulations PW less than 100 us the SA changes during transitions The curves show 200 ms flexion B extension 200 ms B 100 ms overlapping and 200 ms extension B f
85. e capacitance is about 0 03 uF regardless of the preparation The underlying tissues and fluids can be modeled as a pure resistive volume conductor In general the resistance is about 100 to 500 Q The electrode skin impedance is responsible for the slowly decaying SA tail Stimulators There are two types of stimulators current or voltage regulated Voltage regulated stimulators have a low output impedance whereas current regulated stimulators have a high output impedance in the order of MQs The low output impedance of voltage regulated stimulators can reduce the SA by discharging the electrode skin tissue capacitance after a stimulus Current regulated stimulators do not shortcut such a capacitance Additionally a stray capacitance in the range of 100 200 pF between the outputs and ground have to be considered These are responsible for the fast decaying SA tail 6 EMG Signals Features Signal Acquisition Stimulation Artifact Removal 92 6 3 2 Methods to Remove Stimulation Artifacts in SEMG Signals Many different methods that remove the SA in SEMG or similar neurophysiological signals were proposed in the last 30 years They can be divided in three main categories SA blanking SA filtering and SA subtraction methods Hardware Babb et al 1978 Freeman 1971 Minzly et al 1993 Roby et al 1975 and software Handa et al 1990 Hines et al 1996 Keller et al 1998 artifact blanking or sample and hold blanking methods are simple
86. e group and to scan the whole region from the center of the target muscle in proximal direction with a small pen like cathodic electrode The applied stimulation pulses should be biphasic asymmetric single pulse twitches with a rather high amplitude The interpulse interval should be in the order of 1 second Observing the muscle body the hand and finger responses to the single stimulation pulses one can easily find the motor points of the targeted muscles Because of the relatively high amplitude even partially atrophied muscles respond adequately Completely atrophied muscles do not respond to the stimulation twitches Once the motor points are found larger self adhesive electrodes can be placed on the motor points By stimulating the regions where the muscle should be located for some days one can train and re strengthen atrophied muscles For finger extensors and flexors we use either 4 cm x 4 cm square electrodes or round electrodes with different diameters They are customized to the appropriate size Larger sized electrodes are used to compensate for the displacement of the skin with respect to the motor point occurring from arm movements These electrodes are placed with their center close to the motor point but displaced a bit in opposite direction of the observed skin movement The use of too large electrodes can also stimulate adjacent muscles and cause unwanted contractions Critical is also the fact that the targeted muscles are mostly in
87. e purpose of our collaboration with Compex SA was to enhance the stimulator capabilities with the above listed features that are necessary in scientific and 3 Concept of the ETH ParaCare FES Systems 46 neuroprosthetic applications A similar concept like the one already applied with the stationary system was developed The new device runs under the new brand name Compex Motion It can be programmed with a PC software using GUIs All parameters of the Compex Motion that are stored on the chip card can be graphically modified Logical groups of parameters like stimulation patterns recruitment curves and look up tables can be stored and loaded from libraries Also the whole chip card content can be stored and loaded Additionally implemented features that were introduced by us merged the portability of the Compex 2 device with the features and flexibility of the stationary rapid FES prototyping system The modularity of the stationary FES system s PC software program was generalized such that the complexity of the stationary system fits into the small Compex 2 device and that the GUIs of the Compex Motion PC software represent a logically structured and intuitive image of the chip card content The GUIs of the Compex Motion PC programming software see Figure 30 Section 5 4 allow the users to change all the parameters that are stored on the chip card Different libraries for general scientific FES grasp and walking applications are delivered with t
88. e software power down procedure is activated whenever the rechargeable battery shows a voltage drop below 7 V 4 1 5 Analog Circuit Board The design of the analog circuit was partially taken with permission from a stimulator developed by the Edmonton group Gauthier 1994 Keller et al 1995 Slight modifications were done in the choice of the resistor values and in the design of the current controller The implemented analog board of the stimulator generates charge balanced current pulses for four stimulation channels A 5 V DC DC converter and four HCPL 2630 optocouplers isolate the high voltage circuit from the digital board The optocouplers galvanically isolate the three SPI signals and the pulse width signal The SPI controlled 6 bit D A converter adjusts the input voltage of the voltage regulated constant current controller that defines the pulse amplitude Using a bipolar transistor the pulse amplitude is multiplied with the pulse width signal that comes from an output compare pin of the HC11 Pulse generation circuit A atte Fa ts td Fa J K 21 al a I LT Depolarization circuit zu B m uF i PE 1 Muscle Model I nr Figure 16 A The analog circuit for the pulse generation B The circuit that is involved in the generation of the depolarizing stimulation pulse C The circuit for the generation of the exponentially shaped charge compensating pulse 4 PC Based Rapid Prototyping FES System 52
89. e sometimes a SA is not canceled perfectly during stimulation and can potentially mimic an EMG activation pattern shows the applied trigger criterion for the grasp and for the release command Menu Parameter Value Interaction input Input A Interaction type EMG Trigger criteria level 1 1 5 V peak valley 1 peak time 1 0 5 s longer shorter 1 shorter time 2 0 5 s longer shorter 2 shorter level 3 1 5 V peak valley 3 peak time 3 0 5 s longer shorter 3 shorter Table 12 The activation pattern Morse code that commands grasp in the digital EMG control mode Each of the two EMG activity peak has to be shorter than 0 5 s The release trigger criteria is set to detect one peak lasting longer than 2 seconds With our subjects we either have selected the extensor carpi radialis muscle if voluntary controllable e g from C6 or lower injured subjects or a shoulder muscle as control muscle In high lesioned SCI subjects we have chosen the contralateral deltoid muscle Those subjects have barely control over their contralateral arm and thus less unwanted activation of the control muscle caused by arm movements Menu Parameter Value Interaction input Input A Interaction type EMG Trigger criteria level 1 1 5 V peak valley 1 peak time 1 2s longer shorter 1 longer all other values 0 Table 13 If the processed EMG is higher than the 1 5 V for more than 2 s the release pattern of the neuropro
90. e transitions from grasp and release This increase is the smallest for changing pulse amplitudes If the SA free SEMG parts 2 6 EMG Signals Features Signal Acquisition Stimulation Artifact Removal 103 12 5 to 50 ms post stimuli are added to calculate the ARV this increase during pulse amplitude transitions is much smaller All four ARV curves calculated from the SEMG recorded from the ventral deltoid muscle are almost congruent This indicates that after the software blanking of 2 5 ms the SA is almost completely eliminated In this case the SA subtraction algorithm is not needed However it does also not falsify the recorded SEMG activity proc SA 15 SEHE orig control orig SA e e s proc control proc SA 14 saasa origcontrol orig SA sessssesssee proc control Dt E A 0 8 07 0 7 0 6 0 6 05 0 5 250 250 250 250 250 250 250 200 50 0 0 0 0 0 0 0 0 250 250 250 250 250 250 250200 50 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 150 250 250 250 250 250 250 250 0 0 0 0 0 0 0 0 0 150 250 250 250 250 250 250 250 12 12 12 12 12 12 12 11 4 0 0 0 0 0 0 0 0 12 12 12 12 12 12 12 114 0000000 0 0 0 0 0 0 0 0 0 0 6 12 12 12 12 12 12 12 0 0 0 0 0 0 0 0 0 6 12 12 12 12 12 12 12 Ext Carpi Rad M Deltoideus Figure 50 The four ARV curves of SEMG recorded above the ventral deltoid muscle right panels only slightly differ from each other durin
91. eaker can be flexibly programmed to play different melodies and alerts and a dot matrix LED display see Figure 27 provides the most relevant stimulation information to the user Figure 26 The Compex 2 stimulator was initially designed for therapeutic and for medical applications The modification of its original software increases its flexibility and allows the implementation of different neuroprostheses The new device with the enhanced capabilities is called Compex Motion The control elements on the stimulator are a power on off button and two and labeled toggle buttons per stimulation channel that are used to adjust the stimulation 5 Portable FES System 6 amplitudes Additional functions of the control elements are displayed on the stimulator screen The internal NiMH battery provides power for at least 8 h of stimulation The battery can be recharged in two hours Amplitude Channel 1 4 bargraph Functional specification of the Pulse duration on off button u bargraph Stimulation time i Amplitude value mA Battery charge status Figure 27 The 64x256 dot matrix graphical display of the Compex Motion stimulator displays the actual amplitudes and pulse widths of the four stimulation channels 5 2 1 Inputs The two input channels A and B and the Compex push button input C part of the multipurpose I O connector can be used to control the stimulator with almost any type of sensor shows the connector int
92. easy to compose rapidly precisely timed stimulation sequences including customized pulse width ramps loops branches pauses user interaction rules and displayed texts The primitives used for user interactions define how a subject can interact with the stimulator and can be customized to individual needs For example the user can initiate or terminate a stimulation sequence via a predetermined analog or digital sensor signal 132 9 Conclusions 133 curve profile detected at the input ports A and or B of the stimulator Two different sensor signal curve profiles can be used to select between two different stimulation sequences Sensors such as EMG sensors force sensitive resistors gyroscopes foot switches and push buttons were successfully applied with the user interaction primitives Continuous regulation of the stimulation intensity can be achieved in real time using an analog input signal i e the pulse amplitude depends on the voltage level of the input signal This dependence can be arbitrarily defined by look up tables that can be imported as an ASCII file and can be edited both graphically and numerically Each stimulation channel has its own look up table Sensors such as EMG sensors sliding resistors one dimensional touch pads and potentiometers were successfully used with this feature The portable system uses the same hardware as the already commercially available Compex 2 stimulator from Compex SA which is a therapeu
93. ehabilitation engineering requires the interdisciplinary collaboration of medical doctors therapists and engineers and all of them have to be aware of the uncertainties and limitations of their techniques In medical fields like diagnostic and intensive care technology plays a dominant role One has realized how the huge investments and efforts into diagnostic systems like MRI scanners X ray and ultrasound equipment and in patient monitoring systems improved diagnosis prognosis and patient surveillance As a result of these efforts it was possible to tremendously shorten hospitalization time and more people could be cured In contrast to the acute medicine rehabilitation therapy is dominated by manual treatment methods mainly provided by therapists Here similar improvements can potentially be expected from rehabilitation engineering that has its operating field on the other side of the chain after the surgical intervention the intensive care and the acute medicine The main task of rehabilitation engineering is to help faster improve or and cure after the acute medical intervention by developing new technologies concepts methodologies and assistive tools In cases where no complete cure can be expected rehabilitation engineering can help to further assist and improve the conventional therapy Rehabilitation engineering deals with the human subject as an integral system that is influenced by an unknown number of factors and combinations It i
94. ency Its small size and very low power consumption makes it ideal for portable applications 6 2 4 Filtering For the signal conditioning of the SEMG a band pass filter is used According to SENIAM recommendations they were proposed as a result of a Concerted Action on SEMG in the Biomed 2 Program of the European Union the high pass cut off frequency should be set between 10 and 20 Hz Hermie et al 1999 The high pass filter is necessary to remove movement artifacts generated by electrode dislocations and other low frequency variations like skin impedance and static charge changes Standard amplifiers use second or fourth order high pass filters In special cases for example for a fast artifact recovery the high pass cut off frequency is set up to 100 Hz although most of the SEMG s power is in lower frequencies The low pass filter reduces high frequency noise created by the electrode and the amplifier Additionally it removes high frequency components of the SMEG that violate the Nyquist sampling theorem The low pass cut off frequency in standard applications is set to 300 500 Hz Thus the main part of the SEMG spectrum is amplified In standard applications second to sixth order filters are used Due to their higher high pass filter cut off frequency fast artifact recovery SEMG amplifiers also have a higher low pass cut off frequency to compensate for the loss of low frequency signal portions 6 2 5 Signal Processing In all our appl
95. equence 5 Portable FES System 7 Time line empty box ZSe 25 i Generated stimulation sequence Primitives Pulse width us Time s Figure 31 A stimulation sequence a train of stimulation pulses is composed by placing function primitives into a time line using a drag and drop technique In this example the stimulation pulse train is composed using three primitives The first primitive increases from pulse to pulse the width for 3 s the second stimulates with a constant pulse width for 9 4 s and the third decreases the pulse width from pulse to pulse in 2 6 s The above curve depicts the stimulation pulse width profile of the generated Figure 29 29 stimulation sequence The single pulses have a shape as shown in The pulse width ramp primitives the first and third primitive can be defined by a 16 point curve of any shape They are defined in the pulse width PW sub window Additionally with insert and delete buttons a given stimulation pattern can be lengthened or shortened A copy paste function of the whole time line is available through a pull down menu accessible with the right mouse button 5 4 3 Stimulation Primitives There are seven different categories of primitives see Table 5 to build the stimulation sequences Three different background colors of the primitive icons visualize to the programmer if the primitive affects only the channel in which it is placed blue or
96. er und Ger tesoftware f r ein kommerziell erh ltliches portables Elektrostimulationsger t welche es erm glicht das Ger t f r unterschiedlichste FES Anwendungen und als portable Greifneuroprothese zu verwenden Klinische Versuche haben gezeigt dass die Greiffunktion bei allen Versuchspersonen mit der Neuroprothese verbessert werden konnte Insbesondere konnten Personen mit ausreichender proximaler Armmotorik aber ohne willk rliche Fingerfunktion nur mit der Greifhilfe Objekte greifen halten und wieder loslassen was ihre Selbst ndigkeit erheblich erh hte Solche Personen stellen die ideale Benutzergruppe f r einen t glichen Gebrauch der Greifneuroprothese dar Personen mit einer inkompletten Querschnittl hmung konnten ihre Greiff higkeit mittels FES Training erheblich verbessern Es gibt deutliche Hinweise darauf dass funktionelles Training mit der Greifneuroprothese eine wichtige Rolle in der Reorganisation der noch intakten Nervenbahnen und der Plastizit t des zentralen Nervensystems spielt Dieses Ergebnis konnte noch nicht bewiesen werden steht aber im Zentrum einer geplanten Multicenter Studie mit der entwickelten Greifneuroprothese Table of Contents FO WOLG E De near pees E Ses Eee 3 PRS UAC a re Eee eee eee 5 RUE ZIAS SUING oie seid ca canaiicedanasenadeisvawedasagewodeungeetsduanubaviautanauigarbaelsagacasigambdeteapabautes 7 Table of COMMENTS ass iki 9 Abbreviallonsschhnn ul 12 2 3 The Tetraplegic Subject
97. er flexion can be achieved by wrist extension The functional hand is passive when a dorsal extension of the wrist is performed by supination of the hand using the M biceps brachii In case of an active functional hand the subject can perform the tenodesis grasp by contraction of the M carpi radialis longus and brevis which directly produce dorsal wrist extension In subjects with an active hand additionally to the wrist actuators the long finger extensors and flexors are innervated Only the intrinsic muscles cannot be voluntarily controlled Subjects with active hand can perform hand opening and closing but have no pin grip or other precision grip In general they are not candidates for current neuroprostheses for grasping 2 4 Currently Available Neuroprostheses for Grasping The upper limb shoulder arm hand fingers has more than 30 degrees of freedom that are impossible to control or command with state of the art technology A good review that shows the complexity of the upper limb and describes which actuator is involved in which upper limb movement is provided in Kendall et al 1993 Popovic et al 2000 Almost all existing upper extremity neuroprostheses are aimed to improve the grasp function The reduced set of commands of such neuroprostheses mainly hand opening and hand closing can relatively easily be operated by SCI subjects In addition to hand opening and closing a few laboratory systems provide the possibility to contro
98. erate the stimulation pulse widths for each stimulated muscle see Figure 24 With this method the open loop stimulation patterns can be smoothened and linearized 4 3 5 Data Acquisition and Data Storage Routines Time discrete stimulation amplitudes and pulse widths of all stimulated muscles and up to eight A D converted sensor signals from the multifunction card can be recorded and stored to the hard drive The stimulation amplitudes and pulse widths are stored with the same sampling frequency as the sensor signals The data are stored in a bitstream integer format with 16 bit resolution although the acquired data has only a resolution of 12 bits Each recording file consists of a header with information about the recorded data and of the 16 bit data itself For all channels the header includes a measurement identification string the channel number the channel name the sampling frequency the starting time of the measurement the physical unit of the data and a predefined amplification and offset of the data The last 3 items allow the storage of the data in their physical unit and magnitude On a 200 MHz Pentium PC with 64 Mbytes RAM the system is able to store all the data with a maximum sampling frequency of 8 kHz per channel This is sufficient to store surface EMG recordings De Luca 1997 Duchene et al 1993 as well as to record kinematic data The stored data format is compatible with the commercially available Soleasy Alea Solution GmbH
99. erface of the stimulator The input channels A and B can be used as analog inputs for voltages with a range of 0 to 5 V or as digital 5 V TTL compatible inputs In addition they have a 5 V DC power output that can be used to supply active sensors like EMG amplifiers or gait phase detection sensors The input C is a pulled up 5 V digital input that can be used for the Compex push button or also for a force sensitive resistor FSR Besides as input C the multipurpose I O connector is used to charge the stimulator s batteries to interconnect multiple Compex Motion stimulators in order to synchronize them and to program the chip card with a PC via the asynchronous serial communication port COM port of the HC11 The voltage requirements of the inputs are listed in Insertion slot for Input A the chip card multipurpose V O connector Input C Stimulation outputs 5 5 Portable FES System 66 Figure 28 The back of the stimulator has four stimulation outputs a multipurpose I O connector two analog and TTL inputs A and B and the chip card slot The analog inputs can be configured to control the stimulation amplitudes in real time They can control the actual stimulation amplitudes via a programmable look up table for each stimulation channel Hence it is possible to increase the amplitude of channel 1 to keep the amplitude channel 2 constant and to decrease the amplitude of channel 3 with a voltage increase of the sensor attached to the
100. ery flexibly If spontaneous recovery happens so that the benefit of using FES vanishes the FES treatment can be reversed completely The exact electrode positions of implanted electrodes are very difficult to evaluate during the surgery The resulting function depends much on the pressure on the electrode the position and the tissue itself All these factors are completely different during the implantation when a dislocation of the electrode can be done and after the surgery For every change a new surgical intervention is needed In contrary the location of surface electrodes can be adjusted at any time until the optimal places are found Because of that more than 50 of the subjects need a second or even a third follow up surgery There are several reasons for it Electrodes but also the implanted transmission coils need to be dislocated as a result of hyper sensation in several subjects In some cases broken leads must be replaced Compared to the surface stimulation technology the implantation technology is very expensive The devices the hospitalization time and the second rehabilitation process are very cost intensive With surface electrodes significantly cheaper neuroprostheses can be offered The implantation procedure is a major surgery with the risk of loosing some functions Not every subject is willing to undergo such an intrusion and risk Those subjects first should have the possibility to test if FES really improves their dai
101. es to electrode arrays which are used to trace temporal spatial changes of propagating AP s The most frequently used electrode configuration is the bipolar configuration This configuration rejects common mode noise and provides a good S N ratio The bipolar electrodes are directly connected to the bipolar inputs of an instrumentation amplifier and the ground electrode a conductive strip is fixed around the limb see Figure 38 The interelectrode distance in the bipolar configuration has a filtering effect It was shown in the early 1970s by Lindstr m Lindstrom et al 1977 that when a signal s t is propagating with a velocity v from one electrode to the other interelectrode distance d the recorded signal is s t 3 or s t The Fourier transform from the above function is F s a t s f 2i sola S v 6 EMG Signals Features Signal Acquisition Stimulation Artifact Removal 86 It has zeros for multiple frequencies f V d In practical applications for SEMG amplitude analysis an interelectrode distance of 10 20 mm is recommended It is important that d remains constant when comparing different SEMG measurements c Electrode la b Electrode la T Electrode 1b Electrode Electrode 2a a a N GND OND m u Electrode nn GND Electrode Electrode 1b Electrode 2b Figure 38 a Monopolar b bipolar and C multipolar electrode amplifier configuration Multipolar electrodes are of
102. ethod provides every 250 ms see Section one value that represents the SEMG activity If the data pieces are smaller than 250 ms the calculated ARV must be smoothened either by taking the mean value of several ARVs or by low pass filtering the ARVs Xi Both the rectifying and low pass filtering and the ARV processing methods are implemented in the stationary FES system The Compex Motion FES system uses the second method 6 3 Stimulation Artifact Removing Techniques The recording and processing of SEMG is more complex when the recorded muscles or muscle groups close the recorded one are stimulated In this case the stimulation pulses and muscle and tissue responses to the stimuli are measured together with the SEMG All those stimuli evoked effects we define as stimulation artifacts SAs since for controlling neuroprostheses we are only interested in SEMG from voluntary contracted muscles and not in any evoked responses If the stimulation site is close to the recorded muscle site the SA exceeds the SEMG signal by orders of magnitude This immediately drives the EMG amplifiers into saturation If the EMG amplifiers are AC coupled which is almost always the case for drift compensation the filter and the electrode tissue capacitors are fully charged with each stimulus After the SA the electrode tissue and the filter capacitors discharge slowly with an inverse exponential decay Due to the high gain the amplifiers remain saturated for
103. executed in the time line Instead the current amplitude is linearly increased or decreased to the new amplitude value during a chosen time duration see 2nd parameter in Table 5 which is set in the frequency and amplitude sub window Pulse frequency primitives The stimulation frequency can also be programmed to change at any time in the stimulation sequence The new frequency is set immediately after being executed in the time line As the stimulation frequency is the same for all channels it is changed whenever a change frequency primitive dark green icon appears in the time line of any of the four stimulation channels Sequence control primitives The stimulation sequence or a part of it can be repeated by inserting marker and jump back to marker primitives blue icons in the time line A jump back to marker primitive and its corresponding marker primitive defines a sub sequences that is repeated for the indicated number of times Up to four different jump back to marker and marker pairs can be cascaded to build single or nested loops The marker and jump back to marker primitives only apply to the channel in which time line they are placed The synchronize primitive is used to synchronize the stimulation sequences of the different channels The execution of the primitives in a channel is halted when a synchronize primitive is reached in the time line If all other active channels have also reached the synchronize primitive all channels con
104. f denervation in the upper extremities motor and sensory performance and impairment Also the general physical and psychological condition were taken into account The PT and OT examined the patients muscle status according to the Frankel classification Frankel et al 1969 and more recently according to the modified ASIA American Spinal Injury Association score Maynard et al 1997 Especially in an early phase of the rehabilitation it was important to assess the potential of voluntary arm movement For a successful 127 8 Results with the Neuroprosthesis for Grasping 128 application of a neuroprosthesis for grasping the proximal arm muscles should be voluntarily controllable to the extent that the hand can be voluntarily moved to the mouth The subjects were informed that the neuroprosthesis is a device that potentially helps them to artificially regain more hand function but that it cannot reverse their SCI or help in reconnecting central nerve tracts We signed with all subjects an agreement that informed them about all potential risks and guaranteed them to stop the FES treatment at any time All our tests were approved by the local ethics committee With half of the patients we started very early even before mobilization with a training phase The muscles were trained once a day to accommodate to the stimulation and to become fatigue resistant This training period was also used to find the optimal electrode positions A train
105. ful to find the optimal stimulation electrode locations These optimal locations have to be modified slightly to take into account muscle skin motion for different arm positions and unwanted activation of adjacent muscles 3 7 Neuroprosthesis for Graspin 114 J ANTERIOR VIEW RIGHT N POSTERIOR VIEW RIGHT Long thoracic N Serratus ant Accessory N cran XI Subclavian N Sternocleidomastoid Dor Mid amp low trap Subclavius scap N Upper trapezius Lat and Med pectoral Ns Lev scap Suprascapular N Pectoralis major _ pe N Rhom min Supraspinatus Pectoralis minor Infraspinat Rhom maj a p Axillary N Musculocutaneus N Coracobrachialis Biceps s h L subscap N Biceps I h Brachialis Subscap Radial N Teres maj T Triceps long h Triceps lat h Thoracodor N Triceps med h Latiss dor Brachialis Median N Brachioradialis Pron teres _ Ai 4 Ext carpi r l Fi carpi rad Al Anconeus Ulnar N Palmaris long Ext carpi r b Fl carpi ulnaris Fl dig super Supinator Bl Ext digitorum N en Ph dig prot te i Yellow Cords peripheral nerves and anterior Fi poll 1 divisions from which they arise f f R gt Pron quad W I Green Cord peripheral nerves and posterior Nt I divisions from which they arise Wh 4 leg Palmaris brevis Abd poll b
106. g transitions from release to grasp This indicates that the SA blanking during the stimuli of 2 5 ms completely removed the SA from the ventral deltoid muscle SEMG No further processing is needed However on the other hand the SA subtraction algorithm does also not affect the voluntary SEMG activity 6 4 5 Discussion and Conclusions A novel SA removal method for real time applications is presented The algorithm subtracts a moving ensemble averaged SA from the SA contaminated SEMG of a voluntarily activated muscle Because of the random nature of voluntary EMG activity the recorded voluntary SEMG activity is canceled by ensemble averaging over a period lasting from pulse to pulse Thus the ensemble averaged SEMG recording represents only the SA Subtracting this SA from the raw SEMG for each period results in a SA free SEMG signal that can serve as a measure of voluntary muscle activation Previously extracted SAs are weighted with an exponential forgetting weight to allow an adaptation of the ensemble averaged SA to changes of the stimuli The algorithm is capable of eliminating SA tails in presence of voluntary SEMG activity even if the SA shapes are changing due to changing stimuli The stimulation spikes cannot be eliminated We suggest to blank the signal during that saturated period see shaded periods in Figure 45 In our case the blanking period is fixed to 2 5 ms 6 EMG Signals Features Signal Acquisition Stimulation Artifact Remova
107. gure 20 The trigger unit allows a flexible definition of the processing steps of the recorded sensor signals in order to generate a hand open or a hand close trigger If the grasp or the release rule is detected i e if the logical states of all trigger conditions stated in the rule are fulfilled then the stimulation patterns either for grasp or release are executed releasing Load Pathan Save Pattern Edit Pattern Figure 21 The stimulation patterns for the grasp and the release task can be imported and exported in ASCII format and can be graphically edited The stimulation patterns Figure 21 are normalized to 1 1 and free of a physical unit 1 stands for a completely opened hand and 1 for a completely closed hand The upper graph in Figure 21 shows the stimulation pattern for grasp and the graph below shows the stimulation pattern for release These stimulation patterns are mapped in the recruitment curve compensation module for details see Section 4 3 4 to the four stimulation channels and control the pulse widths of the stimulator The continuously controlled stimulation software Figure 18 has the same structure as the event triggered stimulation pattern generator software Instead of the trigger unit a continuous controller Figure 29 maps the sensor signals either EMG or sliding potentiometer signals to the control variable The sensor signals are pre processed and 4 PC Based Rapid Prototyping FES System 5
108. h selectivity between the finger and the wrist extensors Ww d The splint is worn over the self adhesive stimulation electrodes The electrode cables are fed through small punched wholes Special care has to be taken to prevent pressure sores resulting from non padded electrode cables under the splint Outside the splint the electrode cables are connected to the multi core cable The multi core cable is guided inside the subject s clothing along the arm to the stimulator on the backrest of the wheelchair Generally our tetraplegic subjects use the neuroprosthesis for grasping when they are sitting in the wheelchair Thus the portable stimulator is either fixed to the backrest of the electrical wheelchair or is placed in a meshed bag behind the back rest When using a push button or a sliding resistor to control the neuroprosthesis these b Figure 51 Splints made of A leather B textile fabric and synthetic fiber 7 Neuroprosthesis for Graspin 107 sensors are fixed with a Velcro fabric on the armrest of the wheelchair together with the power ON OFF button The sensors and their cabling are permanently fixed on the wheelchair and can be plugged into the portable stimulator Figure 52 The control and power button are placed easily reachable at the armrest of the wheelchair The SEMG sensors if used as control tool contact the skin using tripolar self adhesive pads that are placed over the measured muscles The Compex EMG
109. h frequency pulses mostly fast nerve fibers big diameter are blocked hindering contraction of fast muscles The authors suggested this method to be applicable against the reversed recruitment order of artificially stimulated muscles Although the phenomenon is not completely understood some literature provides support to the notation that high frequency stimulation maintains the endplates in a temporary refractory state preventing the muscle from contracting Solomonow 1984 2 1 6 Waveform of Stimulation Pulses The waveform of the current pulse plays a significant role in the daily application of FES APs should be generated by sharp edged negative current pulses with a pulse duration of at least 40 us see intensity pulse duration curve in Figure 4 Shorter pulses or pulse slopes less than 10 A s need a largely increased current to generate an AP Stimulation pulses longer than 300 us applied with surface electrodes only recruit a few more nerve fibers but generate more pain since they stimulate preferably afferent nerves that are sensitive to a longer pulse duration If such sharp edged negative monophasic current pulses are applied over a long period of time the unidirectional ion current flow has the potential for ion accumulation and skin irritation Therefore biphasic stimulation pulses that remove the induced electric charge are used for FES applications In practical applications either symmetric or asymmetric biphasic stimulation pulse
110. h of the function primitives needs a user specified time to be executed by the stimulator and therefore controls the duration of the stimulation patterns There are four time lines one for each of the four stimulation channels The time line and the function primitives are discussed in more detail later in this chapter The lower left area in the main GUI window shows buttons that open sub widows for setting up additional parameters that are needed by the function primitives In the lower right area the stimulation mode the default stimulation frequency and the serial port number can be selected At the lower right bottom of the main GUI window there are buttons for reading and writing all settings to from the chip card and a save button that stores all settings to the hard drive In a sub menu after storage the same or a different setting can be loaded into the software Finally a Quit button can be used to store and to quit the software ew Conpex FES ee User Intonation Pulse Width Fi Ampldude Preg nr CoO Analog Amp Kontroll User Inleracions Read Chip Card HEVE Tew Program Chip Cand Liat Figure 30 Main window of the Compex Motion graphical user interface software It shows four horizontal time lines associated with each stimulation channel center and right pulse amplitude and pulse width safety limits left pulse type settings center bottom memory chip card functions right bottom and setup functions left
111. hannels 2 and 6 starts All 8 stimulation channels are activated in this way If doublet pulses are enabled the BOSFETs are turned on for 7 ms instead of 2 ms allowing a second pulse of equal duration 5 ms after the first pulse being generated for the specified channels 4 PC Based Rapid Prototyping FES System 54 initialization gt battery voltage check power button and turn off stimulator software power down check software handshake check if new data from PC turn on bosfet ch 1 and 5 set DA converter for pulse amplitude ch 1 and5 set DC DC converter set buzzer ch 1 and 5 generate pulses for 8 stimulation channels turn off bosfet ch 1 and5 i 3 a turn on bosfet set stimulation ch 2 and 6 frequency generate pulse width turn off bosfet ch 4 and 8 Figure 17 The simplified flow chart visualizes the sequencing of the software routines written in assembler The last routine in the stimulation loop adjusts the stimulation frequency All the routines sequentially executed need all together 20 ms The stimulation frequency routine is a simple waiting loop that meets the desired frequency between 20 and 50 Hz After finishing the waiting routine the program counter jumps back to the battery and power button check routine 4 3 LabVIEW Software The rapid pro
112. hase the feasibility of applying neuroprostheses for grasping using surface stimulation electrodes to improve the grasp function in SCI subjects in an early phase of rehabilitation was demonstrated This thesis is focussed on resolving a number of scientific engineering and clinical questions which stand in the way of commercially available attach and go devices that require minimal training adaptation and maintenance Specifically the most important research goals of this thesis project were 1 The development of a hardware platform FES system that can facilitate fast testing of concepts and methods based on FES for the restoration of the grasp function in SCI subjects This platform was used to develop control strategies that are suitable for controlling the neuroprostheses Two successfully functioning prototypes of neuroprostheses for grasping were developed A stationary rapid prototyping and a portable FES system Experiments with SCI subjects conducted in our laboratory demonstrated that the neuroprostheses for grasping could significantly improve the grasp function in SCI subjects 2 The development of control strategies that command the neuroprostheses for grasping to perform different types of grasps One of the main focus was on exploring control strategies that use electromyographic EMG signals from voluntarily activated muscles during FES to command the grasp task 3 The development of new programming and controller fir
113. he fourth chapter presents with the stationary rapid prototyping FES system This system is mainly used to develop new control strategies stimulation patterns and general concepts of which the successful concepts are implemented in the portable system The chapter describes the stimulation hardware and the LabVIEW programmed FES controllers and the recording software that runs on a PC The fifth chapter briefly describes the ETHZ ParaCare portable FES system which was used as a demonstrator of a flexible neuroprosthesis for grasping and was used with 1 Introduction 17 some of the subjects in activities of daily living ADL before the Compex Motion system was completed Six such portable FES systems were built They were very useful for testing and showing our concepts and convincing our industrial partner Compex SA to start a collaboration with the goal to enhance their stimulator Compex 2 to be able to perform the functions needed for neuroprosthetic applications The main part of the chapter describes the Compex Motion concept the hardware the firmware and the stimulator programming software The sixth chapter first addresses the main characteristics of recorded surface electromyographic SEMG signals from voluntarily activated muscles the measurement hardware the filtering techniques and our processing algorithms of the SEMG The next part of the chapter describes the proposed techniques used to eliminate stimulation artifacts SAs w
114. he nerve and depolarize the nerve rapidly Once the inactivation gate h is closed the sodium influx is stopped The re polarization process of the nerve starts taking place It is driven by the osmotic pressure that causes an efflux of cations through non specific leakage channels After 600 us the resting condition of about 80 mV is established Through this process the gate m is closed and h is opened and ready for another AP l membrane transmembrane potential 500 1000 time us Figure 3 A simplified model uses an activation gate m and an inactivation gate h to describe the behavior of voltage driven sodium channels that play a major role in the process of generating APs The Figure shows A the stimulus B the conditions of the two gates h and m and C the transmembrane potential as a function of time during a depolarizing stimulus that generates an AP Adapted from Grill et al 1995 Hyperpolarization of the nerve generated by reversing the stimulation current can also produce an AP known as anodic break excitation During long time hyperpolarization 500 us or longer with positive current pulses the fast activation gate m closes a bit more and the inactivation gate h opens fully in resting position it is only 75 open At the end of the stimulation pulse the activation gate m opens a bit and because the inactivation gate is fully open an influx of sodium ions can initiate an AP if the activation gate opens sufficiently
115. he numerical values for the time duration trigger parameters can only be set with an accuracy of 100 ms 5 4 5 Analog Control In addition to the trigger control capabilities user interactions that interactively controls the stimulation patterns programmed in the time lines a continuous analog control of the stimulator pulse amplitude is implemented too This analog amplitude control can be used independently of the primitives in the time line For each channel one of the sensor inputs A or B can be chosen to serve as a control variable Using a look up table consisting of 64 values the measured analog voltage between 0 V and 5 V is mapped and scaled to the stimulation pulse amplitude 5 Portable FES System 8 Set Lookup table Edit Amp2 Look up Edit Amp4 Look up Figure 36 A sensor signal that is connected to the analog input A can also control the stimulation amplitude continuously Four look up tables map the selected analog input signal 0 5 V x axis to the stimulation amplitudes of the four channels y axis 0 actual amplitude The look up tables in the figure are set such that the stimulation amplitude of channel I is increased with sensor signal voltages above 2 8 V from 0 mA to the preset stimulation amplitude and channels 2 3 and 4 are increased with sensor signal voltages below 2 2 V The analog control capability for example is used to control hand opening and closing with a neuroprosthesis for grasping using a
116. he previously described rapid FES prototyping system is very useful for the development of customized neuroprostheses After the evaluation of the optimal electrode positions the grasp and release sequences and the control strategy are determined individually to fit the subject s needs and skills After the individual setup the SCI subject can use the rapid FES prototyping system for the conditioning of the muscles and for the training of the grasp skills However the stationary system is not built for everyday use outside the hospital Therefore a portable FES system had to be 62 5 Portable FES System 6 developed A modified version of the stationary stimulator was designed The asynchronous communication interface was dropped and replaced by signal conditioning low pass filters for the eight internal A D converters of the HC11 The assembler software was extended with routines for stimulation pattern generation data acquisition and real time sensor signal processing of push buttons analog sliding potentiometers and EMG sensors The system was packed in a box of size 122x55x250 mm Six such devices were built Figure 25 The first prototype of the portable system consisted of four stimulation channels and up to six 8 bit analog input channels that were used to measure signals from different sensors The HC11 was able to process in real time two EMG channels and controls the stimulation sequences The system was powered with a Li ion bat
117. he software allowing the user a faster and easier programming of the stimulator On the stimulator itself only a few parameters like the stimulation amplitudes can be adjusted Almost all FES applications using surface technology that we could foresee can be done with the Compex Motion stimulator Only when high computational power is required e g for closed loop muscle force control the stationary rapid prototyping system is required Also new control algorithms and concepts can be implemented faster in the LabVIEW control software of the stationary system than in the Compex Motion 4 PC Based Rapid Prototyping FES System In this chapter the stationary rapid prototyping FES system is described The following topics are presented and discussed e The composition of the system e The hardware design of the electrical stimulator e The assembler software that generates the stimulation pulses with a high precision and safety standard e The FES controller and the recording software that runs on a PC and is programmed in LabVIEW The main achievements were e to overcome the synchronization problems of the two basically independent systems stimulator and PC and to provide from pulse to pulse controllable and exactly timed stimulation pulses e to create a flexible and easy to use modular programming platform for the development of new control strategies and stimulation patterns for neuroprostheses for grasping Several microprocessor
118. hich are always present in the SEMG signal of a muscle close to stimulated muscles and must be removed The last part of the chapter describes a novel technique that is capable of eliminating in real time the major part of the slowly decaying SA tail which is the main disturbance of the SEMG signal recorded close to a stimulation site In control strategies using SEMG signals the presented SA removal algorithm can be successfully applied The seventh chapter describes the developed neuroprostheses for grasping First the components the cabling and the fixation of the systems on the subjects electrical wheelchairs are presented The second subchapter explains the electrode positions specifies the expected muscle contractions and the resulting limb functions Another subchapter describes in detail the control strategies the stimulator setups and the stimulation patterns used with the Compex Motion portable FES systems At the end of the chapter advantages and limiting factors of EMG control strategies are discussed and compared with the other control strategies In the eight chapter the main results obtained with the neuroprostheses for grasping in our clinical trials are summarized The main emphasis here is put on explaining our rehabilitation procedure Unfortunately the functional outcome could not be assessed consistently because the systems were always in a development phase and the functional outcome depended much on the actual developmen
119. ic and d charge balanced alternating symmetric bipolar constant current stimulation pulses recorded from a Compex Motion device The stimulation pulses of the four channels are generated consecutively with one DC DC converter A demultiplexer distributes the generated pulses to the four stimulation outputs This method reduces the size and the costs of the pulse generation hardware but has the disadvantage that the stimulation frequency is the same for all stimulation channels However by choosing the smallest common multiple as basic stimulation frequency different stimulation frequencies for different channels can be generated by introducing appropriate pauses between the pulses For example by choosing 60 Hz as stimulation frequency channel 1 can stimulate with 30 Hz having a pause of two interpulse intervals between two pulses and channel 2 can stiulate with 20 Hz having a pause of three interpulse intervals while channel 3 stimulates with 60 Hz lists the specifications of the Compex Motion device Digital VO C Rx T SV TTL fF Table 3 Stimulation parameter specifications The column Range specifies the maximal possible range settings for each parameter if possible The maximum pulse duration parameter and the stimulation frequency have a common constraint Because of the multiplexing of the four stimulation channels the sum of all active pulse widths plus 5 ms processing time determines the maximum possible stimulation frequen
120. ical stimulation J Rehab Sci 3 pp 86 9 139 Inbar G F and Noujaim A E 1984 On surface EMG spectral characterization and its application to diagnostic classification IEEE Trans Biomed Eng 9 pp 597 604 Johnson M W and Peckham P H 1990 Evaluation of shoulder movement as a command control source IEEE Trans Biomed Eng 9 pp 876 85 Karu Z Z 1992 Optimization of force and fatigue properties of electrically stimulated human skeletal muscle Master thesis Karu Z Z Durfee W K and Barzilai A M 1995 Reducing muscle fatigue in FES applications by stimulating with N let pulse trains IEEE Trans Biomed Eng 8 pp 809 17 Keith M W Kilgore K L Peckham P H Wuolle K S Creasey G and Lemay M 1996 Tendon transfers and functional electrical stimulation for restoration of hand function in spinal cord injury J Hand Surg Am 1 pp 89 99 Keith M W Peckham P H Thrope G B Buckett J R Stroh K C and Menger V 1988 Functional neuromuscular stimulation neuroprostheses for the tetraplegic hand Clin Orthop Rel Res pp 25 33 Keller T and Treichler M 1995 Das Stimulationsger t Commstim In Bilineare adaptive Regelung f r funktionelle elektrische Stimulation Diploma Thesis Automatic Control Laboratory ETH Zurich Zurich pp 47 53 Keller T Curt A Popovic M R Dietz V and Signer A 1998 Grasping in high lesioned tetraplegic subjects using the E
121. ically sealed iridium electrodes on both ends The glass capsule houses an ASIC and a radio frequency coil that together constitute the stimulator 256 such BIONs can be individually controlled with an external device using an eight bit programmable unique address for each BION aeae Cameron et al 1993 The control signal and the supply voltage of the BIONs are transmitted using an inductive coil The BIONs can be implanted into muscles or near GI i D cannery muscle nerves using mamal invasive surgery tube techniques The BIONs are injected with an Bar insertion tool made from a 12 gauge ne angiocatheter that can be used to test stimulate the location before permanently injecting the A electrodes Cameron et al 1997 Circuit board Preferred applications are therapeutic electrical stimulation of deep muscles e g electrical stimulation to treat shoulder subluxation At the moment neuroprosthetic applications are not considered because the required transmitter lol power is at the moment too high for most portable applications Next generation devices might have a more efficient power and transmission scheme to be able to build portable systems 2 2 5 Self Adhesive Electrodes for Transcutaneous Stimulation Self adhesive electrodes for transcutaneous stimulation use a gel to contact a conductive member with the subject s skin The electrode is built in a multi layer configuration as shown in Figure 1 consist
122. ications we are interested in the activity level of the voluntarily activated muscle of which we record the SEMG This activity level can be obtained from a SA free SEMG signal by detecting its envelope The envelope can be detected by rectifying and low pass filtering the recorded EMG signal Another measure of the SEMG activity is the averaged rectified mean value ARV of a stationary piece of the SEMG data In the first method the low pass filtering can produce a significant time delay since a cut off frequency of about 1 5 Hz must be chosen to fulfill the stationarity requirements for SEMG signals To overcome this problem for example a phase shift compensated 2 order low pass filter can be used The phase shift compensation is performed by filtering the data with a low pass filter reversing the data and filtering it again with the 6 EMG Signals Features Signal Acquisition Stimulation Artifact Removal 89 same filter and reversing again the data With the first filtering the data is shifted to right on the time scale delayed and with the second filtering the data shifted back because of the reversing of the data This method can also be performed in real time if more than one sample is available per time step over sampling technique The ARV is processed by applying the following equation 1 N 2 N is the number of samples and has to be chosen such that the piece of the SEMG signal is stationary In other words this processing m
123. igaments of the finger flexors that are attached to the PIP and DIP joints can be sutured together Or in case of severe deficits caused by muscle denervation muscle tendon transfers combined with FES are applied by the Cleveland group Muscle tendon transfer is a surgical intervention that connects the tendon coming from a voluntary controllable muscle or a stimulated muscle with intact motorneuron innervated muscle to a tendon that controls a desired function which could not be controlled because the original muscle was denervated or not active Theoretically all innervated muscles are candidates for muscle tendon transfers However the primary actuators are left intact The preferred muscles for tendon transfers to improve hand grasp are the palmaris longus the flexor carpis ulnaris the extensor carpi ulnaris and the brachioradialis Keith et al 1996 The currently used stimulator for the Freehand system has only a one way communication direction from the external controller box to the implant In 1998 the Cleveland group introduced a new generation of implantable stimulators Smith et al 1998 The new generation is now able to provide back telemetry for implanted control sensors The device is a modular system that can be configured prior to fabrication It can have up to 32 independent biphasic stimulation channels up to eight telemetry channels either for EMG or other sensors with independent sampling frequencies and pulse powering
124. ill et al McGill et al 1982 the SA is divided in three segments 1 an artifact spike coincident with the stimulation pulse 2 a fast decaying SA tail segment and 3 a slowly decaying SA tail segment A similar segmentation was also proposed by Harding Harding 1991 as shown in 41 Our recordings of SAs can be divided in three or four segments see Figure 40 as well 6 EMG Signals Features Signal Acquisition Stimulation Artifact Removal 91 v A P P f B P C e E Figure 41 A The SA was segmented in four parts B stimulus C fast exponential decay from amplifier saturation D recovery from overshooting of the fast decay and E slowly decaying SA tail from the electrode tissue coupling Reprinted from Harding 1991 The fast and slowly decaying SA tail segments are attributed to the capacitive components of the electrical stimulator the EMG amplifier and the electrode skin impedance Electrode Skin Interface The interface between the stimulation electrode and the skin in the case of surface stimulation consists of different inhomogeneously conductive layers see also the multi layer electrode construction in Section 2 2 5 Figure 10 that influence the impedance in a non linear manner The keratinous and the epidermis layer for example form a highly resistive layer that is very dependent from the skin preparation It can be below 10 KQ for prepared abraded skin or several MQs for unprepared skin Th
125. imulation comfort and applicability for neuroprostheses for grasping e the tetraplegic subject focussing on the functional deficits in the upper extremities e the currently existing and used implantable and surface neuroprostheses for grasping Neuroprostheses based on FES are systems that artificially generate muscle contractions obeying commands of the user A neuroprosthesis combines artificial parts electrical stimulator wires and electrodes and natural parts nerves and muscles of the human body with the function to overcome a neuronal lesion in the central nervous system CNS The lesion can be in the spinal cord Spinal Cord Injury or in the brain e g stroke It is obvious that the natural parts of the body significantly influence the function of the neuroprosthesis This first chapter gives a brief overview of the involved parts and provides some information about the most important currently available neuroprostheses 2 1 Muscle and Nerve Properties and Activation Mechanisms during Functional Electrical Stimulation FES In this subsection the basic nerve and muscle properties will be addressed More information about nerve properties can be found in Guyton et al 1996 Muscle function and properties are described in detail in Karu 1992 Silbernagel et al 1991 More clinically relevant aspects of FES are discussed in Baker et al 1993 Popovic et 18 2 Principle and Function of Neuroprostheses for Grasping 19
126. in a broad sense bridge interrupted or damaged neural connections between the brain and upper extremity muscles using a technique called Functional Electrical Stimulation FES Unlike the name suggests state of the art neuroprostheses are not even approximately able to replace the neural structure and function of nerves nerve bundles the damaged part of the spinal cord in case of Spinal Cord Injury SCI or a lesioned part of the brain e g in stroke subjects As a matter of fact neuroprostheses for grasping require intact motor neurons that connect the spinal cord with the upper extremity muscles A motor neuron is a peripheral efferent nerve a pathway from the central nervous system to the muscle Neuroprostheses bridge central lesions either of the spinal cord or the brain by detecting interpreting and commanding a desired motor action which results in a limb movement For the detection of the desired actions man machine interfaces and sensor systems are used The interpretation is performed by a microcontroller unit using a control scheme and the motor action is commanded using FES In some publications FES is also referred to as Functional Neuromuscular Stimulation FNS or Neuromuscular Electrical Stimulation NMES All three expressions refer to the same method the artificial generation of action potentials APs in peripheral efferent nerves with the goal to produce muscle contraction Therefore short electrical current pulses depolari
127. in combination with the rapid prototyping FES system A LabVIEW interface for the setup of all filter and gain parameters is implemented in the stationary rapid FES prototyping controller software It allows a fast setup of the EMG amplifier Table d shows all parameters and ranges that can be specified for each channel In combination with the stationary rapid prototyping FES system either the flexible and Table 6 EMG amplifier parameters and their ranges The Compex EMG Biofeedback sensor has two active electrodes and the ground electrode arranged in a symmetric triangle configuration with an interelectrode distance of 25 mm The amplifier of the Compex EMG Biofeedback sensor has a fixed 4 order butterworth band pass filter with a bandwidth of 100 4000Hz and a fixed amplification gain of 1400 The high pass cut off frequency of 100 Hz is chosen relatively high to allow a faster recovery of the filter after a stimulation artifact SA more details about SAs see Section 63 The EMG Biofeedback sensor is powered with 5 V and has a rail to rail output between 0 and 5 V In resting condition no SEMG activity an output voltage of 2 49 V can be measured 6 EMG Signals Features Signal Acquisition Stimulation Artifact Removal 88 Figure 39 a The Madaus stationary SEMG amplifier can be flexibly adjusted to different gains and filter frequencies left b The Compex EMG EMG Biofeedback sensor right has a fixed gain and filter frequ
128. infinite impulse response IIR t eh x t N is the number of samples filter can be calculated y nh x n t p yhak 1 biep pti is the weight that controls the forgetting and t 1 is the time of the previous stimulation pulse The curve Y t bi 2 r rvfe is the ensemble averaged SEMG curve that consists mainly of the pure SA By subtracting the moving ensemble averaged SA curve Y t from the SEMG curve X t an SA free signal is obtained The simplicity of the algorithm and the recursive formulation allows the implementation of the algorithm in a real time microcontroller system The optimal forgetting weight p and the performance of the SA cancellation were experimentally evaluated as discussed below 6 4 2 Validation Experiment The SA subtraction algorithm was tested in an experiment with three subjects There were no differences observed in the SAs of the different subjects For testing the algorithm the finger extensors the finger flexors and thumb thenar muscles were stimulated with different stimulation patterns similar to those used in the stimulation of a grasp and release task The hand opening and closing was alternately stimulated as shown in Figure 42 6 EMG Signals Features Signal Acquisition Stimulation Artifact Removal 94 Stimulation Patterns and Parameters The stimulation patterns for hand opening and closing overlapped for a short time period to keep the SA alive
129. ing language A simplified program flow chart in Figure 1 shows the most important procedures and the program sequences The initialization procedure consists of the latching of the power retaining circuit the configuration of the microcontroller in extended mode the initialization and start of the continuous A D conversion for the battery check the setup of the asynchronous and the synchronous serial ports and the initialization of all program variables If no data is sent from the PC to the stimulator a short program loop only checks the battery voltage and the state of the stimulator s power on off button The power down procedure turns off the BOSFETs the high voltage DC DC converter and the buzzer This function also turns off the port that holds the power retaining circuit and waits in an endless loop until the residual power of the stimulator burns out 0 1 s If there are new data at the dual ported RAM the high voltage DC DC converter and the buzzer are set according to their values in the dual ported RAM The pulse generation routine turns on the BOSFETs la 1b 5a and 5b Then the new amplitude values for channel 1 and 5 are sent via the SPI serial port to the D A converters on the analog boards The timers for the output compare function are configured to switch the output compare pins of the HC11 for the actual pulse widths of channel 1 and 5 After 2 ms the BOSFETs la 1b 5a and 5b are turned off and the same procedure for the c
130. ing of two layers of hydrogel The skin interface layer includes an electrically conductive gel with a relatively low peel strength for removably contacting the subject s skin It has a wet feeling and can be removed relatively easily from the skin The conductive gel is made from copolymers derived from polymerization of acrylic acid and N vinylpyrrolidone A second hydrogel layer connects the substrate a low resistive material with the skin hydrogel layer This second conductive gel layer has a relatively high peel strength that provides very good adhesion to the substrate 2 Principle and Function of Neuroprostheses for Graspin 3 Substrate ern SUBSTRATE I Substrate Adheres and bonds to any common substrate media such os HYDROGEL LAYER 2 Hydrogel fabric silver carbon film etc _SCRIM LAYER f Fa Polyester serim Controlled scrim placement prevents slippage and delamination SKIN HYDROGEL LAYER 1 Multiple re odhesions to the skin ie wi Ye Figure 10 Recent self adhesive stimulation electrodes are manufactured in a multi layer configuration Reprinted from Axelgaard 2001 As material for the substrate conductive fabric carbon film or other conductive materials are used A wiring cable connects the electric stimulator to the self adhesive electrode substrate Between the two hydrogel layers a scrim layer can be introduced This scrim layer can be used to prevent slippage of the two hydrogel layers or it is u
131. ing session consisted of e 2 min weak stimulation either with a low stimulation intensity or a low stimulation frequency to warm up the muscle e 14 min strong stimulation to strengthen and train the muscle with a stimulation frequency of 25 Hz e 4 min of low frequency stimulation 2 Hz to maintain an increased blood flow through the muscles without generating further fatiguing stimulation After two to four weeks or after the first mobilization the functional training was started The subjects learned to place the hand to be able to grasp an object to control the neuroprosthesis with one of the presented control strategies and to grasp the object in a useful manner During the functional training the subjects were supervised by the OTs and the FES researchers The grasp patterns and the control strategies were modified during that time to improve the grasp performance In this learning phase dummy objects were used during the therapy sessions Tools of ADL like tooth brushes forks knives pens were modified e g with a thicker handle to make hand grasp easier During this functional training period the subjects only used the system during the occupational therapy sessions Figure 66 In functional training sessions the SCI subjects learns how to use the neuroprosthesis for grasping In the clinical environment the push button control strategy is preferred In a third period the neuroprosthesis was provided during the day The sti
132. is needed for the development of neuroprostheses and of other FES applications Therefore a rapid stationary FES prototyping system Keller et al 1999 and a portable system Keller et al 2001 have been developed The system requirements were 42 3 Concept of the ETH ParaCare FES Systems 43 e safe stimulation e sufficient stimulation power to stimulate muscles in SCI subjects e full control over the stimulation amplitude frequency and pulse width e real time control of the pulse width and amplitude that allows one to implement closed loop control strategies e fast and easy programming of stimulation patterns e fast adaptation of the system to different subjects e user independent system e ability to handle different man machine interfaces and sensor systems e low power consumption e small size and portability The rapid stationary prototyping system consists of 1 a four channel current regulated FES stimulator 2 an eight channel multifunction data acquisition board and 3 a standard Pentium PC that executes the controller software written in LabVIEW and can be readily modified using graphical user interfaces GUIs The system is mounted on a wheeled rack and can be transported to various hospital facilities The main purpose of the stationary system is to serve as a platform for the development and testing of new concepts for neuroprostheses All above mentioned requirements except the last two are fulfilled with this sy
133. ive potential outside the nerve membrane with respect to the potential inside the cell is decreased In other words the nerve fiber inside the membrane has a less negative potential As a consequence voltage gated sodium ion membrane channels are triggered to open The function of the voltage gated sodium ion membrane channels can be described using a model with two sodium channel gates a fast sodium channel activation gate m and slow sodium inactivation gate h Grill et al 1995 see Figure 3 In reality the opening and closing of the sodium membrane channels are not driven by gates but by conformation changes in the membrane spanning domains of the channel proteins that 2 Principle and Function of Neuroprostheses for Grasping 21 form the sodium channel Guy et al 1986 In resting state the activation gate is almost completely closed and the inactivation gate is about 75 open Changes of the electrical potential inside the nerve with respect to the outside influence the membrane spanning of the sodium channel protein A depolarizing pulse negative current pulse opens the activation gate m and at the same time starts closing the inactivation gate h as shown in Figure 3 The time constant of the activation gate Tm is 100 times faster than the time constant of the inactivation gate 7 Therefore during the time gate m is open and gate h is not fully closed sodium ions driven by the concentration gradient flow through the channel inside t
134. kes a muscle response for different stimulus pulse duration and intensities The result is plotted in a so called intensity duration curve Such a curve shows the shortest pulse duration for a given stimulus intensity that provokes a motor response shows intensity duration curves of innervated and denervated muscles Typically for innervated muscles a pulse width longer than 300 us does not produce much more muscle contraction for a given stimulation intensity e g 2 Principle and Function of Neuroprostheses for Grasping 23 40 mA and a pulse width shorter than 50 us needs a very high stimulus intensity to reach motor threshold The motor threshold is the level of the weakest muscle response to a stimulus Similarly for denervated muscles a non linear intensity duration behavior can be observed but the required pulse duration has to be more than 100 times longer 140 innervated muscle A near maximum denervated muscle 120 4 innervated Motor response motor threshold muscle x 100 motor T threshold E 80 60 gt oO AQ Sn 2 eat ARE eee as un yks leas cass cei 2S dae cage Saeeacncd ee 20 0 T K T T T 1 0 01 0 1 0 3 1 10 100 1000 pulse width ms Figure 4 Curves of equal motor response for different pulse widths and pulse amplitudes For intact motorneurons stimulation pulses longer than 300 us do not increase the motor response if the stimulation amplitudes are higher than 40 mA Denerv
135. l 104 We allow a fast adaptation of the extracted SA to changes of the stimuli by choosing a small forgetting weight p 1 It can be shown that for pulse amplitude modulated stimulation patterns with a constant pulse width of 250 us the SA removal performance is very good The SA from fast changing pulses whose pulse widths are changed and shorter than 150 ms cannot be eliminated completely because the SA significantly changes from pulse to pulse If there is no possibility to use changing pulse amplitudes instead of changing pulse widths to modulate the stimulation patterns we recommend to blank both the SA spike and slowly decaying SA tail In this case only a low stimulation frequency below 25 Hz can be used to retrieve a SA free stationary SEMG measure in a reasonable time since the number of samples that can be used to calculate the ARV values become small and need more smoothening We suggest further to slightly overlap the stimulation pulses during transitions for example from hand opening to closing and always to stimulate sub threshold if no muscle contraction is desired to sustain the SA as long as the SEMG activity is needed for the control of the neuroprosthesis 7 Neuroprosthesis for Grasping This chapter describes the developed neuroprostheses for grasping The components the cabling and the fixation of the system on the subject s electrical wheelchair The preferred electrode positions for stimulating hand opening and ha
136. l parts of the elbow movement i e elbow extension by stimulating the M triceps brachii Currently available neuroprostheses used either implanted or surface stimulation electrodes associated with its advantages and disadvantages 2 4 1 Implanted FES Systems Freehand System Cleveland In the late 1970s P H Peckham and his collaborators started with the restoration of hand functions in tetraplegic subjects using FES The first systems used percutaneous intramuscular electrodes to stimulate the peripheral nerves of the finger extensors and flexors and of the thumb extensors flexors abductors and adductors Both the stimulator and the controller were external devices controlled by a shoulder position transducer Buckett et al 1988 The main deficiencies of these systems were about 10 breakage of the electrode leads within the first year and infections at the skin portal In some cases also skin irritation or burns were reported To overcome the electrode problems in 1987 an externally powered multichannel implantable FES stimulator for hand grasp was introduced Smith et al 1987 This first generation implantable technology has been made commercially available as Freehand System 2 Principle and Function of Neuroprostheses for Grasping 38 from NeuroControl Inc It consists of an implantable eight channel stimulator with epimysial electrodes an external controller box and a shoulder position transducer Electrical power and the
137. lass 4 to grasp a fork or a spoon and eat with it 5 to grasp an apple and eat it 6 to grasp a pencil and write with it 7 to brush the teeth and 8 to shave using an electrical or a manual razor Figure 70 The subject with complete C5 tetraplegia performs an EMG controlled the grasp of a Tetra Pack containing milk Currently subjects B C and D use the neuroprosthesis for grasping in daily living activities B two years D seven months and C six months Subject A was released from our hospital in early 1997 before we were able to provide him with a portable FES system 9 Conclusions Two FES systems a stationary rapid prototyping and a portable FES system were developed and tested Both systems can be used in various FES applications as a therapeutic and muscle training device general research electrical stimulator and as a neuroprosthesis In comparison to most other existing FES systems which have very restricted and limited possibilities to change the stimulation patterns and training sequences to adjust the stimulation intensities for different channels and to externally control the systems both the stationary rapid prototyping and the portable FES systems offer a high flexibility to adapt and control the stimulation patterns for almost any kind of transcutaneous FES applications The rapid prototyping FES system offers this flexibility with two different modularly structured LabVIEW software programs which control
138. lation Amplitude and Stimulation Pulse Control Signals The stimulation pulses for all four stimulation channels are generated from a single voltage regulated constant current controller The controller has two inputs one input controls the pulse amplitude and the other one controls the pulse width The pulse amplitudes are generated with a 6 bit D A converter placed on the analog boards The D A converter has a serial input port that is compatible with the synchronous serial port SPI of the HC11 Because of the rather slow serial communication between the microcontroller and the D A converter a new pulse amplitude has to be sent 120 us prior to the start of the stimulation pulse This time is necessary for the D A converter to receive and to process the SPI signals The accurate stimulation pulse widths are generated with the hardware implemented output compare timer function of the HC11 The output compare feature allows one to pre program a timer controlled toggling 0 to 5V of a hardware output pin of the microcontroller The 16 bit timer of the HC11 runs at 2 MHz and is pre programmed for each pulse to switch on and off an output compare pin for the duration of the stimulation pulse A resolution of 500 ns can be achieved using this method Therefore the pulse widths can be adjusted in a range of 0 to 500 us with an accuracy of 500 ns The right plot in one 14 shows the timings of the four stimulation channels The second pulse is generate
139. le averaged SA does not affect the EMG activity because of the randomness of SEMG A moving window with exponential forgetting for the SA averaging adapts the extracted SA to changes of the artifact caused by changes in the stimulation pulses With a relatively small averaging window a fast adaptation can be obtained which is needed for rapidly changing stimulation patterns for example in transitions from grasp to release It could be shown that changes of the stimulation amplitudes during stimulation produced significantly less changes in the SA than changing the pulse duration Therefore it is recommended to change the stimulation amplitude during transitions from grasp to release The SA of a 2mA stimulation pulse produces almost the same SA than a stimulation pulse that generates full muscle contraction The SA that is recorded far away of a stimulation site e g the contralateral deltoid muscle is mainly produced by the direct stimulus and not followed by a slowly decaying SA tail These very short SAs of maximal 3 ms for four stimulation channels can be eliminated using SA blanking techniques The SA subtraction method is only required if the long lasting SA tails of 10 20 ms disturb the voluntary SEMG activity The SA subtraction method is implemented in the stationary FES system The portable system uses a software SA blanking algorithm With a more powerful microcontroller than the HC11 the ensemble averaging SA subtraction technique could a
140. le contraction are described in Karu 1992 or Silbernagel et al 1991 Only a brief summery of the most important processes is given here A muscle consists of about 100 2000 motor units and the number of muscle fibers in a motor unit amounts to 5 1000 see and Figure 7 Each muscle fiber again consists of a bundle of myofibrils that are incased by the sarcolemma oc SARCOMERE MYOFIBRIL Figure 6 Morphology of skeletal muscle Reprinted from Baker et al 1993 2 Principle and Function of Neuroprostheses for Grasping 25 The motor nerve axon terminal is connected to the motor unit at the neuromuscular junction called motor end plate It is located at the muscle fiber s midpoint causing a symmetric muscle fiber contraction in both directions Whenever a nerve AP reaches the motor endplate acetylcholine is released over a small gap called synaptic cleft and generates a muscle AP This process lasts about 6 ms The muscle AP propagates along the muscle fiber in both directions and is transported through the transverse tubulesarcoplamic reticulum system T tubules to the myofibrils see Figure 7 In a myofibril there are about 1500 actin and 3000 myosin filaments large polymerized proteins which form the contractile muscle structure The actin filaments are held in Z discs around the myosin filaments as shown in The myosin heads arranged with a joint like connection around the myosin filament adhere to the actin filament A
141. lectrodes the connection of all the cables and the chosen man machine interface and the putting on of the garment or brace The time that was needed to install the neuroprosthesis was too long and has to be reduced A solution to this problem will be the development of a garment with already included electrodes This garment has to be designed such that the electrode positions can be flexibly chosen to the subjects needs The electrode cabling should only consist of plugging one cable to one connector for all channels The garment itself has to be fitted individually to the subject or at least three to four different sizes will be needed With such a garment the donning and doffing time can be reduced to to 2 minutes which will be acceptable for the subjects Our future plan is to develop the above mentioned garment and to combine it with the Compex Motion FES system to have a uniform neuroprosthesis for grasping that allows to make the absolutely necessary individual adaptations to fit the subjects needs in terms of control strategy stimulation pattern electrode positions and man machine interface We made the experience that this high flexibility is needed to be able to provide a system that can be used for a wide range of disabilities with a high performance of the neuroprosthesis The uniformity of the system on the hardware side will be demanded by the manufacturer to be able to produce an affordable system It will also be required f
142. lexion C The amplitude modulated SAs remain constant PW 250 us The forgetting weight is p 1 6 EMG Signals Features Signal Acquisition Stimulation Artifact Removal 10 The SA subtraction algorithm has difficulties with pulse width transitions when the pulse widths become shorter than 120 us and are rapidly changed Then the SAs are changing significantly from pulse to pulse A and B in and cannot be completely eliminated by the algorithm dla peee orig control orig SA ssesssseeen proc control proc SA 1200 few eee orig control orig SA serere proc control proc SA 0 94 ET A DOA TESO g 0 84 OB ne HE ete eae 071 va 7 0 7 0 6 0 6 0 54 0 5 A 0 44 ur 04 0 34 0 3 0 24 63 0 1 eR o1 0 o 250250250250250230180130 80 30 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 30 80 130 180 230 250 250 250 250 250 000 O O 0 O 0 0 30 80 130180230250 250250250250 250 250 250 250 250 230 180 130 80 30 0 0 0 0 0 0 0 0 0 1 a 0 94 Er ao OG EANES A NETTE ae u gt 75 RA ee en a 0 84 S pereectentontenteatensenameaeeneessesn 0 74 Se 0 74 RS BES 0 6 u 0 6 Nat 0 55 0 54 0 44 0 4 0 34 0 34 0 24 0 24 0 17 0 14 mr 250 250 250 250 250 250 250200 50 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0O 150 250 250 250 250 250 250 250 0 0 0 0 0 0 0 0 0 150 250 250 250 250 250 250 250 250 250 250 250 250 250 250 200 50 0 0 0 0 0 0 0 0 1 i 1 58
143. limited Either the stimulators stimulate repetitively ON OFF intervals with an adjustable interval time or they can perform ramp up ramp down of the stimulation intensity for each interval Additionally it can be defined whether all channels stimulate simultaneously or whether they alternate The stimulation patterns cannot be freely chosen nor controlled in real time This is by no means sufficient to control limb movements Similar problems can be observed in existing surface FES neuroprostheses for grasping such as the Bionic Glove and the Handmaster which have entirely fixed stimulation patterns If a muscle or two cannot be stimulated in a subject due to denervation or if some additional muscles have to be stimulated to provide a better grasp function the fixed design of these neuroprostheses does not allow such changes Also using different man machine interfaces and or control strategies that might be more appropriate for the subject can t be done All these limitations clearly show that it is necessary to improve the existing technology To do so almost every research center that is involved in improving and adapting the FES technology to fit their subjects needs has to develop their own stimulators After working with the Bionic Glove as one of the multicenter trial sites and with a stimulator called Commstim that was also developed by Prochazka s group decision was made to build an own FES system that provides all the flexibility that
144. litudes of the selected channels For each channel see in real time the recorded potentiometer voltage is mapped to the stimulation amplitude using a look up table The X axis in ems eee the measured sensor signal voltage in the range of O to 5 V and the y axis indicates the stimulation amplitude The stimulation amplitude ranges from 0 mA to the set default amplitude actual amplitude depending on the potentiometer position Suet Lisk p iabk Figure 65 The sliding potentiometer that is connected to the analog input A controls the grasp task The channel I stimulates the finger extensors for hand opening and the channels 2 and 3 control the finger and thumb flexors for hand closing 7 Neuroprosthesis for Grasping 125 7 3 5 Analog SEMG Control In the analog SEMG control strategy the pre processed EMG signals from two voluntary controllable muscles for example the ventral and dorsal branches of the deltoid muscle are used to control the grasp force Keller et al 1998 The gains of two processed SEMG signals are carefully adjusted that equal muscle activation results in equally measured EMG activity Then both signals are subtracted from each other to eliminate co contraction activities of both branches The resulting control variable controls the grasp in the same way as the potentiometer output voltage in the sliding potentiometer control strategy In the case of more ventral deltoid activity hand opening is stimulated When
145. ller H Bijak M Mayr W Rafolt D Sauermann S and Thoma H 1997 Useful applications and limits of battery powered implants in functional electrical stimulations Artif Organs 3 pp 210 2 Lanmuller H Mayr W Stohr H Thoma H and Unger E 1990 20 channel implant for functional electrostimulation Biomed Tech Berlin Suppl 3 pp 122 3 Liberson W T Holmquest H J Scott D and Dow M 1961 Functional electrotherapy stimulation of the peroneal nerve synchronized with the swing phase of the gait of hemiplegic patients Arch Phys Med Rehab pp 101 5 Lickel A 1998 Restoration of lateral hand grasp in a tetraplegic patient applying natural sensory feedback PhD thesis Aalborg University Aalborg Lindstrom L H and Magnusson R I 1977 Interpretation of myoelectric power spectra a model and its applications Proc IEEE 5 pp 653 62 Long C and Masciarelli V 1963 An electrophysiological splint for the hand Arch Phys Med Rehabil pp 449 503 MadausMedizin Elektronik 1993 Gebrauchanweisung VP Verst rker In Reference Manual Madaus Freiburg pp 1 26 Maynard F M Jr Bracken M B Creasey G Ditunno J F Jr Donovan W H Ducker T B Garber S L Marino R J Stover S L Tator C H Waters R L Wilberger J E and Young W 1997 International standards for neurological and functional classification of spinal cord injury American Spinal Injury Ass
146. lso be implemented in the portable system 9 Conclusions 135 In our rehabilitation center 10 subjects used one of our neuroprostheses for grasping All of them could benefit in one or more ways from the FES technology Although the applied neuroprostheses for grasping were in development and used different technologies a lot of experience and some conclusions could be gained e In all subjects the neuroprostheses were able to improve the grasp function and force e Complete injured C5 and C6 subjects could benefit from the neuroprosthesis for grasping as permanent grasp aid whereas incomplete subjects mainly profited from the neuroprosthesis as a training tool e Using the neuroprosthesis for functional training incomplete subjects often developed their grasp skills to a level where the additionally produced grasp force provided by the neuroprosthesis was not further needed in ADL e In clinical applications the push button control strategy was the preferred one although the SEMG strategies are more intuitive and permit a faster control The main reason was the higher reliability during the development phase of the neuroprostheses Wrong triggering or a false activation of the grasp as it sometimes happened with the EMG control strategies were never observed with the push button or sliding resistor strategy e The donning of our prototype neuroprostheses took about 5 to 10 minutes This included the proper placement of the stimulation e
147. lts in an increase of the muscle contraction force In the recorded SEMG signal such an increase can be observed as a higher signal amplitude that is related to the higher number of generated muscle APs and an increased firing rate The SEMG signal can be modulated by the intensity of muscular activity SEMG activity can be voluntarily mechanically e g stretch reflex electrically or magnetically evoked The latter two methods use magnetic or electric stimulators Magnetic stimulators generate strong magnetic pulses of about one Tesla or more that can also provoke APs like the electric stimulation Many different types of SEMG responses are used in the medical field to examine and also to investigate the electrical pathways of the CNS and the peripheral nervous system This chapter concentrates on voluntarily generated SEMG that can be recorded and processed to be used as a control variable for neuroprostheses 6 1 Characteristics of SEMG The main characteristics of voluntary generated SEMG signals are the randomness and the stationarity Other SEMG characteristics like different propagation velocities of APs in smaller or larger muscle fibers which can be observed in the SEMG signal as a right shift of the median frequency when larger Type II muscle fibers are activated are less important for controlling commanding neuroprostheses 6 1 1 SEMG Randomness In many studies the firing activity of motorneurons during voluntary muscle activation
148. lug breaks that could easily be fixed these two command interfaces were the most reliable ones and therefore preferred by the clinicians therapists and the patients Although the push button and the sliding resistor strategies seem to be optimal command interfaces for clinical applications they have some disadvantages that makes it worth to explore alternative control interfaces and strategies One of the main disadvantage is the need of some cognitive feedback The place of the push button and place and position of the sliding resistor have to be remembered and visual feedback and one hand is needed to command these interfaces All that makes the commanding rather slow The voice control strategy can be commanded without the need of one hand Bilateral handling of the target objects can be performed Although user dependent voice recognition was implemented the system requires a keyword prior to each verbal 9 Conclusions 134 command to prevent unintended execution of the grasp or release task Each command consists of at least two words which delay the grasp by at least one second In addition the rather poor recognition rate of 80 90 due to wrong intonation or environmental noise doubles or triples the time needed to recognize the user s intention Gareiss 2001 The recognition rate can be enhanced by the subject with training to better than 95 when the environmental noise is low Wrong executions the detection of a command that was no
149. ly activities This can be done relatively easily with surface electrodes 2 3 The Tetraplegic Subject Subjects with a SCI at cervical level region around the neck suffer from tetraplegia According to demographic statistics in Switzerland every year 28 residents per million become SCI subjects Zach et al 1995 Compared to other countries this incidence rate is average About 30 of all SCI subjects suffer from tetraplegia 2 3 became SCI by an accident trauma 2 3 1 Clinical Classifications The clinical classifications of tetraplegia and also paraplegia distinguish between sensory and or motor impairment complete or incomplete SCI the severeness of the injury and the level of spinal lesion The level of spinal lesion describes above which neurophysiological level the spinal cord is intact for motor and sensory functions 2 Principle and Function of Neuroprostheses for Graspin 35 CERVICAL NERVES THORACIC NERVES eosousundle soon v r l LUMBOSACRAL ENLARGEMENT LUMBAR 3 NERVES 1 SACRALNERVES _ 3 4 5 Figure 12 left The subdivision of the spinal cord lateral view in the cervical thoracic lumbar and sacral parts right The regions of the sensory skin nerves The indicated sensory points are examined with pin prick and light touch to assess the sensory impairment Reprinted from Popovic et al 2000 The spinal lesion levels are divided into four regions cervical C2 C8 th
150. mber of scientific engineering and clinical questions which stand in the way of commercially available attach and go devices that require minimal training adaptation and maintenance In collaboration with one of the world s leading manufacturers of electrical stimulators the Swiss company Compex SA the goal of developing a flexible FES device with commercial strength could be achieved Specifically the most important research goals of this thesis project were 1 The development of a hardware platform FES system that facilitates fast testing of concepts and methods based on FES for the restoration or improvement of the grasp function in SCI subjects 2 The development of control strategies that allow the user of the neuroprosthesis to perform different types of grasps One of the goals was to explore control strategies that use electromyographic signals EMG from voluntarily activated muscles during FES to command and or control the grasp function 3 The development of new firmware and software programs for an existing commercially available electrical stimulator that enabled us to use the stimulator for FES applications and to build portable neuroprostheses for grasping First clinical and in field tests with the portable neuroprosthesis for grasping proved its applicability in an early rehabilitation phase In all SCI subjects a better grasp performance could be obtained with the system SCI subjects that had proximal arm
151. mode 5 4 2 Stimulation Sequence The GUI software uses a drag and drop technique to program the desired stimulation sequences This is done by sequentially placing icons called primitives on a time line that describes the chronological sequence of the tasks that will be carried out by a stimulation channel The term time line stands for an array of boxes as shown in 31 that can be filled with the primitive icons There are four such time lines see 30 one for each stimulation channel The programmer can stack the primitives in any desired chronological order For example a simple stimulation sequence as shown in can be composed using three pulse width primitives The first primitive a ramp up primitive defines of a train of pulses lasting for 3 s whose widths are increased from pulse to pulse in a ramp like manner The second primitive a constant pulse width primitive defines a train of pulses with equal widths lasting for 9 4 s and the third primitive a ramp down primitive defines a train of pulses with decreasing widths lasting for 2 6 s The desired duration of each pulse width primitive can be numerically entered in a text box below the primitive icon see time line in Figure 31 The primitive sequence in each time line can have a maximal length of 254 primitives By pressing the left mouse button over a primitive icon or an empty box the primitive icon can be selected from a pop up menu in order to create or edit a stimulation s
152. more dorsal activity is measured the neuroprosthesis stimulates the finger and thumb flexors for hand closing The amplitude of the pre processed voluntary EMG Control EMG activity determines the grasp force This very direct way to control the neuroprosthesis has the disadvantage that the fast fatiguing voluntary EMG activity hinders the subject to hold an object for a prolonged amount of time To overcome this problem the analog SEMG strategy that was published in Keller et al 1998 can be slightly modified The control variable can be integrated and scaled before it is mapped to the stimulation pulse amplitude A scaling factor adjusts the sensitivity of the new control variable Of course int is also limited the imaginary 0 to 5 V To eliminate drift a dead band a in which the original control variable is not integrated is introduced The following recursive equation transforms the old EMG signal to din Gintn Pintn 1 On 3 if n gt a and Pine Pintn 1 if Onl lt a With the modified control strategy the subject can start hand opening by activating the ventral deltoid muscle By keeping the ventral deltoid muscle active the grasp force is increased No EMG activity keeps the stimulation constant Stimulation is reduced by activating the dorsal part of the deltoid muscle If the dorsal part of the deltoid muscle remaines activated the hand is closed with increased force until no voluntary deltoid muscle activity i
153. more freedom to the user Especially when using many surface electrodes that are all wired to the stimulator the user looks like a Christmas tree On the other hand there are also some disadvantages of using implanted electrodes for FES Implanted electrodes have to be supplied with stimulation current in some manner This can either be achieved by a huge transmission coil as it is the case for the BIONs see Section 2 2 4 or by cables that have to be routed from the stimulator through the whole 2 Principle and Function of Neuroprostheses for Grasping 34 limb to the muscles The cabling makes the surgery very complicated and increases the risk of loosing some of the functions that were present before the surgery Implanted electrodes can only be implanted 1 5 to 2 years after injury when spontaneous recovery in SCI subjects can be excluded For SCI subjects this means that they have to undergo a second rehabilitation period and some of the learned tricks cannot be used anymore or have to be modified after the implantation Surface electrodes can be applied during early recovery and rehabilitation phases SCI subjects can learn to use FES combined with other assistive tools from the beginning on Although the body functions of most subjects are exposed to big changes due to partial recovery surface FES can be started immediately after mobilization and for muscle training and conditioning even before Electrode positions can be chosen and modified v
154. mulator was fixed to backrest of the electric wheel chair and the electrodes were put on in the morning and taken off in the evening So the subjects could use the system during the day and explore the capabilities of the system in ADL The self adhesive electrodes 8 Results with the Neuroprosthesis for Graspin 12 could be attached to the skin for the whole day They did not dry out To keep them in position we either used Micropore tape or an elastic sock In all our subjects we never experienced any skin burns and only once skin irritation due to stimulation When taking off the electrodes one could observe that the skin underneath the electrodes was a little bit red colored but this disappeared within minutes Figure 67 The neuroprosthesis for grasping allows a complete C5 SCI subject to perform ADL tasks he cannot do without the neuroprosthesis In this early stage of the neuroprosthesis development we also had to live with some inconveniences of the systems that should also be mentioned here There were problems with the wiring In the beginning we often had broken wires We tried to use thin wires in a multi core cable But at the cable ends where the single cables connected the self adhesive electrodes using 2 mm Multicontact connectors the single cables often broke A solution to that problem will be a multi core cable with a single connector that will be plugged to a special electrode glove splint The two leads cables tha
155. mware software programs for an existing commercially available stimulator Compex 2 The new software concept elevated the stimulator with the brand name Compex 1 Introduction 16 Motion to a new dimension of applications in basic and applied research in terms of flexibility versatility and applicability The new concept and features enabled us to build new portable neuroprostheses for grasping based on the commercially available stimulator 4 The performance of first clinical trials and in field tests with the portable neuroprosthesis for grasping The main contribution of the thesis are new concepts the implementation and testing of flexibly programmable neuroprostheses as a therapeutic tool or a permanent aid for improving the grasp function in SCI subjects using surface FES technology Therefore a stationary rapid prototyping and a flexibly programmable portable FES system five different control strategies and a new algorithm for the real time removal of stimulation artifacts in recorded EMG signals from voluntarily contracted muscles were developed All concepts were successfully tested in a clinical environment Resulting from this work a new generation of versatile portable stimulators Compex Motion are on the way toward commercialization which will bring to neuroprostheses unprecedented flexibility of the stimulation patterns and sequences the man machine interfaces and the control strategies This high flexibility
156. nd were measured above the M ext carpi rad The small artifact in curve was produced by the stimulator although amplitude and pulse width were set to zero Curve is obtained by concatenating the processed SA and the SA free SEMG signal The processed SA is calculated by subtracting curve from curve The performance of the algorithm is tested by processing a control raw SEMG signal that has no SA with the same algorithm Therefore curve is subtracted from curve and gives as result curve that qualitatively resembles to a raw SEMG signal see curve The curves and are the results Y t from the ensemble averaging algorithm 6 EMG Signals Features Signal Acquisition Stimulation Artifact Removal 97 6 4 3 Signal Processing The validation of the algorithm was performed off line although the algorithm had been made to operate in real time In a first step the SEMG signals recorded from the two muscles during the trials were cut into pieces of 500 samples corresponds to the time between two stimulation pulses 50 ms starting at 0 6 ms pre stimulus From each piece curve in Figure 44 shows two such pieces the two parts A and B were used for the signal processing A from 0 6 ms pre stimulus to 11 9 ms post stimulus containing the SA curve Figure 44 B from 34 8 ms to 49 3 ms post stimulus which was SA free curve Figure 44 Both signals A and B were processed as shown in performing the f
157. nd closing The five applied control strategies push button analog sliding resistor voice digital EMG and analog EMG control All the stimulator setups parameters and stimulation patterns used with the Compex Motion FES system The following conclusions about the presented neuroprostheses for grasping can be made The neuroprostheses for grasping could be adapted to SCI subjects with different disabilities and an appropriate control strategy could be found for all of them The flexibility of the system was necessary to individually treat or provide an optimized grasp function Some of the chosen components for example the used multi core cables were not robust enough for the clinical environment and had to be replaced with more robust ones The mounting procedure of the self adhesive stimulation electrodes took about 5 10 minutes This is for daily use too long and has to be improved 7 1 Components and Fixation The neuroprostheses for grasping consist of 1 the stationary or portable stimulator see Chapters 4 and 5 2 up to four pairs of the Compex self adhesive surface stimulation electrodes 5052MID CompexSA 1998 3 a wrist retainer splint and 4 man machine interfaces like push buttons sliding resistors or the Compex EMG Biofeedback sensors 2M4456 CompexSA 1996 All the above mentioned components of the neuroprostheses for grasping were already described in detail except the wrist retainer splint Basically
158. ne and waits until the trigger criterion is fulfilled before it continues to process the time 5 Portable FES System 73 and Figure 33 line Up to seven different user interaction primitives can be defined see Figure 324 Figure 33 User branch primitive The pattern player similar to the user interaction primitive waits at the user branch primitive position until one of two specified user interactions is fulfilled see and Figure 33 If the first user interaction trigger criterion is fulfilled the pattern player jumps in the time line to the user branch jump position and continues to play the stimulation pattern If the second user interaction trigger criterion is fulfilled the pattern player continues to play the primitives that are placed right after the user branch primitive in the time line User interrupt primitive see Figure 32 and Figure 33 The user interrupt on and off primitives define a range in the time line within which the pattern generator can be interrupted and forced to jump to a defined service routine in the case that a predefined user interaction trigger criterion is fulfilled This set of primitives can serve as a emergency routine in case some unwanted event happens Channel 1 Time based The pattern player waits here until the user trigger criterion is to ist no The pattern player waits here until either the continue or the jump interaction trigger criterion is Channel 1 Time
159. nerate branching If criterion 1 is fulfilled the program proceeds with the next primitive in the time line If criterion 2 is fulfilled the program jumps to a marker in the time line and proceeds with the next primitive after the marker no no user interrupt ON OFF One trigger criterion set with the user interaction primitive is used to generate an interrupt If this criterion is fulfilled at any time in the time line between the ON and OFF primitives the program jumps to a predefined marker and proceeds with the next primitive after the marker no no TTL trigger Stops stimulation and waits until a negative slope TTL signal is detected at input C no no General Primitives Primitive name Description Param 1 Param 2 end Terminates stimulation in the channel time line specified no no turn off Turns off the stimulator no no Text text Displays two text lines with 8 characters in each text line Disp time s 1 59 no E sound Generates a melody 2 different short melodies are available no no Randomization Primitives Primitive name Description Param 1 Param 2 random frequency Activates a stochastic variation of the frequency The frequency varies randomly about the nominal value 0 100 following a uniform probability distribution function Deviation no Ei random pulse
160. ng 10 pp 761 70 Parsa V Parker P and Scott R 1998 Convergence characteristics of two algorithms in non linear stimulus artefact cancellation for electrically evoked potential enhancement Med Biol Eng Comp 2 pp 202 14 Peckham P H 2001 Restoration of function current challenges and future opportunities through electrical stimulation J Rehabil Res Dev 1 pp 13 6 Peckham P H and Creasey G H 1992 Neural prostheses clinical applications of functional electrical stimulation in spinal cord injury Paraplegia 2 pp 96 101 Peckham P H Thrope G Woloszko J Habasevich R Scherer M and Kantor C 1996 Technology transfer of neuroprosthetic devices J Rehabil Res Dev 2 pp 173 83 Popivanov D and Todorov A 1986 Statistical procedures for interference EMG power spectra estimation Med Biol Eng Comp 4 pp 344 50 Popovic D and Sinkjaer T 2000 Control of movement for the physically disabled Springer Verlag London Popovic D Stojanovic A Pjanovic A Radosavljevic S Popovic M Jovic S and Vulovic D 1999 Clinical evaluation of the bionic glove Arch Phys Med Rehabil pp 299 304 Popovic M R Keller T Pappas I P Dietz V and Morari M 2001 Surface stimulation technology for grasping and walking neuroprosthesis JEEE Eng Med Biol Mag 1 pp 82 93 Prochazka A Gauthier M Wieler M and Kanwell Z 1997 The bionic glove an electrical stimulat
161. ngineer In this first phase I was essentially supported by our former technician and my friend Hannes Wichser who helped me with the design and construction of the first two versions of electrical stimulators He motivated me to continue the project after Jan left the group My intimate thanks belong to one of my best friends Milos R Popovic I learned him to know as a very professional group leader with both excellent technical and human expertise We often spent hours with very vivid discussions about the accomplished work the next goals general concepts and strategic decisions I would not be where I m now without his advice help and support Of course the whole project could not be done without the support and collaboration of the ETH ParaCare team the engineers researchers clinicians and therapists Such an intense and close collaboration between the clinical and research staff was only possible due to the integrative commitment of my co supervisor Prof Dr Volker Dietz head of the spinal cord injury rehabilitation center ParaCare University Hospital Balgrist It was always very impressive for me to see and feel how strong he is committed to research therapy and medicine and how he acknowledges both the technical and medical efforts Special thanks go to Gery Colombo who besides his own research and developments was responsible as ParaCare lab leader for the excellent functioning infrastructure and equipment and to my teammates Ion
162. nterface for programmable chip cards built in the stimulator 5 Portable FES System 64 The memory card interface of the Compex Motion stimulator can read and write special Compex chip cards with a capacity of 2 Kbytes The concept is to store all subject specific parameters like stimulation parameters stimulation patterns and sensor settings on the chip card A task specific chip card can be inserted in any Compex Motion stimulator turning the stimulator into the desired neuroprosthesis rehabilitation or research device A graphical stimulator programming software for PC s was developed which programs via the PC s serial port the chip card inserted in the stimulator Depending on the kind of application different PC software versions can be created e g one for programming neuroprostheses for walking or one for programming neuroprostheses for grasping The presented programming software version has a general concept allowing the user to adjust all functions and features of Compex Motion and to create stimulation sequences for neuroprostheses for walking and grasping 5 2 Compex Motion Hardware The Compex Motion stimulation device see Figure 26 has four stimulation channels two input channels that can be used either for measuring analog signals or digital TTL signals and one digital input channel that is used for the existing external Compex push button The stimulator is packaged in a small metal housing and weights 420 g A buzzer loudsp
163. ociation Spinal Cord 5 pp 266 74 McGill K C Cummins K L Dorfman L J and Berlizot B B 1982 On the nature and elimination of stimulus artifact in nerve signals evoked and recorded using surface electrodes IEEE Trans Biomed Eng 2 pp 129 37 McLean L Scott R N and Parker P A 1996 Stimulus artifact reduction in evoked potential measurements Arch Phys Med Rehabil 12 pp 1286 92 Millard R E McAnally K I and Clark G M 1992 A gated differential amplifier for recording physiological responses to electrical stimulation J Neurosci Meth 1 pp 81 4 Minzly J Mizrahi J Hakim N and Liberson A 1993 Stimulus artefact suppressor for EMG recording during FES by a constant current stimulator Med Biol Eng Comp 1 pp 72 5 Nathan R H 1979 Functional electrical stimulation of the upper limb charting the forearm surface joint moments Med Biol Eng Comp 6 pp 729 36 Nathan R H 1992 The isometric action of the muscle drives of the wrist joint ASME J Biomech Eng p 162 141 Nathan R H and Ohry A 1990 Upper limb functions regained in quadriplegia a hybrid computerized neuromuscular stimulation system Arch Phys Med Rehabil 6 pp 415 21 National Instruments Corp 1994 LabPC User manual National Instruments Corp Austin TX Paiss O and Inbar G F 1987 Autoregressive modeling of surface EMG and its spectrum with application to fatigue IEEE Trans Biomed E
164. odal electrode Monopolar biphasic constant current pulses Figure 2de The stimulation pulses are asymmetric but charge balanced The charge balancing pulse is chosen such that it does not excite an AP It is implemented such that the excitation pulse is four times higher in amplitude and four times shorter in pulse width than the charge balancing pulse This type of pulses is used to stimulate smaller muscle groups in case a high muscle selectivity is needed Alternating bipolar symmetric constant current pulses Figure 29d This type of pulses can be selected instead of bipolar biphasic constant current pulses If the stimulation pulses are high there might be some residual charge left in the tissue using bipolar biphasic constant current pulses because the pulses are always applied in the same chronological order and can have a slightly asymmetric shape produced by the non linear behavior of the tissue By alternating the pulses these possibly unbalanced charge capacities in the tissue can be prevented Therefore this pulse form is recommended for the simulation of big muscle groups with high stimulation intensities 5 Portable FES System 67 d ie 0 0 0 5 1 0 1 5 2 0 2 5 3 0 3 5 50 0 50 5 510 515 52 0 52 5 53 0 53 5 time ms Figure 29 Pulse shapes of Compex Motion a charge unbalanced monophasic b charge balanced symmetric bipolar biphasic c charge balanced asymmetric bipolar biphas
165. of Alberta Prochazka et al 1997 whose commercialization failed in a first step the Belgrade FES system Fisekovic et al 2001 and the ETHZ ParaCare neuroprostheses for grasping from ETH Zurich and University Hospital Balgrist Zurich Keller et al 1998 Keller et al 1999 Popovic et al 2001 presented in this thesis of which the Compex Motion system is in the commercialization process In addition to the above mentioned neuroprostheses many other FES systems were proposed that improve the grasp function Haugland et al 1999 Lickel 1998 Rakos et al 1999 Saxena et al 1995 Thorsen 1998 but they have not been commercialized or widely applied clinically 1 3 Motivation Aim and Contribution The motivation to develop the ETHZ ParaCare neuroprostheses and Compex Motion was driven by three main factors 1 the lack of a flexible commercially available neuroprosthesis for surface FES 2 the observed need of a very flexible FES system especially if used in an early phase of rehabilitation and 3 the expressed interest and need for such a system by the medical and therapeutic staff of the rehabilitation center ParaCare which agreed with the needs expressed by the FES community at FES related conferences Several development steps were required to reach the goal of a flexibly programmable portable neuroprosthesis that is able to improve the grasp function in SCI subjects and that is accepted by clinicians and users In a first p
166. ollowing steps 1 cutting the SA part A from 0 6 ms pre stimulus to 11 9 ms post stimulus or the control part B from 34 8 ms to 49 3 ms post stimulus provided curves and in 2 processing the previously cut 125 samples with the moving ensemble average algorithm provided curves and in Figure 44 3 subtracting the extracted average SA obtained in step 2 from the raw signals from step provided curves and in 4 The result from step 3 was concatenated with the SA free curve in and the residual SAs during the stimuli were blanked see shaded residual SAs in Figure 45 The result of the SA subtraction method was further processed to obtain a measure of the voluntary SEMG activity from curve Therefore the average rectified mean value ARV was calculated for each part and a 2 order phase shift compensated low pass filter with a cut off frequency of 0 8 Hz smoothened the result The filter frequency was set such that it fulfilled the stationarity requirements For the validation the curves and of were ARV processed without adding SA free SEMG curve in Figure 44 to the processed SA This worst case condition would occur when the stimulation frequency would be 80 Hz what is considered to be the limit for the SA subtraction method 6 4 4 Results The recorded SA from the wrist extensor muscle consists of a saturated SA spike during the stimuli a fast and a slowly decaying SA tail whereas the SA recorded fr
167. om the contralateral deltoid muscle is only present during the stimulus and can be relatively easily eliminated using a simple software blanking algorithm The following results mainly refer to the SA removal from SEMG signals recorded from the M extensor carpi radialis Figures 45 46 and 48 and show the raw and processed signals obtained by the SA subtraction algorithm and the Figures 47 49 and 50 show the smoothened ARVs of the concatenated 12 5 ms SA pieces During the constant stimulation phases constant stimulation amplitude and pulse width the SAs are almost completely eliminated from 6 EMG Signals Features Signal Acquisition Stimulation Artifact Removal 9 the recorded SEMG signals for both electrode locations see curves and without voluntary EMG M ext carpi rad raw SA processed l SA with voluntary EMG M ext carpi rad Ser raw SA processed SA without voluntary EMG si ff processed SA A An A y er with voluntary EMG er raw SA processed SA Figure 45 The curves 1 3 5 and 7 show the first 12 5 ms after stimulus of four concatenated SAs recorded from the M extensor carpi radialis and the M pars clavicularis of the contralateral deltoid muscle The curves 4 and 8 show the voluntary SEMG signal without the extracted SA There are some spikes left in the processed wrist extensor SEMG see gray
168. onths push button accepted K M 1973 C5 C6 complete right 2 years sl resistor p button accepted L M 1970 C6 C7 complete right 2 years push button voice accepted Subjects C and D accepted the device but probably didn t use it much in their ADL This is our estimate since they very rarely asked for new electrodes Table 14 Experimental results with the neuroprosthesis for grasping The neuroprostheses provided in all subjects an increased palmar and lateral grasp force Without the neuroprosthesis the subjects could not articulate the fingers into flexion or if they had a tenodesis grasp they could move their fingers passively With the neuroprostheses we measured grasp forces between the fingers and the thumb of ION or even more 8 Results with the Neuroprosthesis for Graspin 131 Figure 69 The subject with a complete C5 tetraplegia performs an EMG controlled hand grasp of a Phone receiver 1 finger extension is controlled by the EMG signal from the ventral branch of the contralateral deltoid muscle 2 and 3 palmar grasp is controlled by the EMG signal from the dorsal branch of the deltoid muscle Some of the tasks the subjects listed in were able to perform with the neuroprosthesis were 1 to grasp to lift and to place a variety of objects up to 2 kg 2 to lift a telephone receiver to dial a number to maintain a conversation and to hang up 3 to pour a liquid from a bottle into a glass and to drink it from the g
169. or a comparison of our system to the already existing Handmaster and Freehand systems in terms of a multicenter trial although all three systems are used for different applications Bibliography Alea Solution GmbH 1999 Soleasy 4 0 Professional Version User Manual Alea Solution GmbH Zurich Axelgaard J 2001 Patented multi layer hydrogel AmGel Technologies Internet http www amgel com tech htm Axelgaard J and Heard S 2001 Medical electrode method of manufacture US Patent No 6 198 955 B1 USA Babb T L Mariani E Strain G M Lieb J P Soper H V and Crandall P H 1978 A sample and hold amplifier system for stimulus artifact suppression Electroencephal Clin Neurophys 4 pp 528 31 Baker L L McNeal D R Benton L A Bowman B R and Waters R L 1993 Neuromuscular electrical stimulation a practical guide Rehabilitation Engineering Program Los Amigos Research and Education Institute Rancho Los Amigos Medical Center USA Baratta R Ichie M Hwang S K and Solomonov M 1989 Orderly stimulation of skeletal muscle units with tripolar cuff electrode IEEE Trans Biomed Eng 8 pp 836 43 Blair E and Erlanger J 1933 A comparison of the characteristics of axons through their individual electrical responses Amer J Physiology pp 524 64 Blogg T and Reid W D 1990 A digital technique for stimulus artifact reduction Electroencephal Clin Neurophys 6 pp 557 61
170. or garment that provides controlled grasp and hand opening in quadriplegia Arch Phys Med Rehabil pp 1 7 Rakos M Freudenschuss B Girsch W Hofer C Kaus J Meiners T Paternostro T and Mayr W 1999 Electromyogram controlled functional electrical stimulation for treatment of the paralyzed upper extremity Artif Organs 5 pp 466 9 Riener R Quintern J and Schmidt G 1996 Biomechanical model of the human knee evaluated by neuromuscular stimulation J Biomech 9 pp 1157 67 Roby R J and Lettich E 1975 A simplified circuit for stimulus artifact suppression Electroencephal Clin Neurophys 1 pp 85 7 Rushton D N Donaldson N d N Harper V J Perkins T A Taylor P N and Tromans A M 1996 Lumbar anterior root stimulator for lower limb control in 142 paraplegia In Neuroprosthetics from basic research to clinical applications Hg A Pedotti et al Springer Verlag Berlin pp 611 21 Saxena S Nikolic S and Popovic D 1995 An EMG controlled grasping system for tetraplegics J Rehabil Res Dev 1 pp 17 24 Sensory Inc 2000 Voice Extreme development kit manual Sensory Inc Sunnyvale CA Sensory Inc Neural Networks and their use 2001 Sensory Inc Sunnyvale CA Sensory Inc 2001 RSC 300 364 Data Book Sensory Inc Sunnyvale CA Silbernagel S and Despopoulos A 1991 Taschenatlas der Physiologie Thieme Smith B Peckham P H Keith M W and Roscoe
171. or microcontroller FES stimulators were developed to improve upper limb functions in spinal cord injured SCI and stroke subjects They can be devided into two main categories external devices for percutaneous intramuscular or for transcutaneous surface stimulation Buckett et al 1988 Handa et al 1989 Ijezerman et al 1996 Nathan et al 1990 Prochazka et al 1997 Saxena et al 1995 and internal devices for implanted systems Lanmuller et al 1997 Lanmuller et al 1990 Smith et al 1987 Smith et al 1998 Takahashi et al 1999 Takahashi et al 1995 Most of these systems were built for one specific application and did not have an open architecture They generally operated with pre programmed stimulation patterns for each stimulated muscle group that were stored on an EPROM A fixed set of sensors combined with a control algorithm triggered the pre programmed timed stimulation sequences Some systems allowed changes of the stimulation intensity either during the initialization phase or during stimulation on line A separate PC software often allowed the modification of trigger levels and stimulation sequences and allowed the user to download the settings to the stimulation units 47 4 PC Based Rapid Prototyping FES System 48 In addition to these features the proposed rapid prototyping FES system provides more functions on a modular basis that allow a variety of different sensors and control strategies to be combined with the
172. oracic T1 T12 lumbar L1 L5 and sacral S1 S5 see Figure 12 Diagnostic and functional tests determine the motor sensory impairments and the classification in spinal lesion levels More detailed information about the classification of motor sensory impairments can be found in Popovic et al 2000 shows the distribution of the tetraplegic subjects according to their level of lesion 29 31 16 12 6 Tetraplegic Subjects in 4 C1 C2 C3 C4 C5 C6 C7 Lesion Level Figure 13 Distribution of SCI subjects with cervical lesion according to the lesion level n 95 Adapted from Z ch et al 1995 In addition para tetraplegia can be divided in three clinical syndrome groups 1 the anterior cord syndrome leads in case of a cervical lesion to motor impairment of shoulder hand and fingers spastic lower extremities and muscle denervation of muscles innervated from nerves originating around the lesion 2 the central cord syndrome 2 Principle and Function of Neuroprostheses for Grasping 36 leads in case of a cervical lesion mainly to loss of temperature and pain perception upper more than lower limb weakness but can also lead to spastic lower extremities if the anterior horn cells are affected and 3 the rare posterior cord syndrome leads in case of a cervical lesion to the loss of position perception and to spastic lower extremities The po
173. ousands of nerve fibers The APs are fired asynchronously with a firing rate between 0 3 5 Hz depending on the desired force and the fatigue level of the muscle This asynchronous firing results in tetanic contraction although the stimulation frequency for a single motor unit is low whereas in artificially stimulated nerves the AP s are generated all at the same time The spatio temporal distribution of AP s as it is produced during natural contractions can not be generated artificially with FES For tetanic contractions the muscle has to be stimulated with a rather high stimulation frequency around 20Hz or higher The high stimulation frequency reduces the recovery time of the muscle fibers and produces a faster fatiguing of the muscles 2 Principle and Function of Neuroprostheses for Grasping 27 Besides increased muscle fatigue there is also another limit to increasing the stimulation frequency for muscle activation With stimulation pulses at frequencies above 600 Hz muscle contraction can be hindered Studies from Baratta et al Baratta et al 1989 showed that by applying a high frequency stimulation train of 600 Hz or higher using a nerve cuff electrode distal to an electrode with normal 20 50 Hz stimulation pulses one can selectively prevent the muscle from contracting With a higher amplitude of the high frequency pulses muscles innervated with slower nerves smaller diameter are prevented from contraction With a low amplitude of the hig
174. r Nr 6 is long the signal has to cross the second trigger level after time 2 and if parameter Nr 6 is short the signal has to cross the second trigger level before time 2 6 shorter longer 2 shorter the time between the two peaks valleys has to be shorter than time 2 longer the time between the two peaks valleys has to be longer than time 2 7 level 2 Sets the trigger level threshold of the second peak above threshold or valley below threshold 8 peak valley 2 Sets the above or below threshold condition for level 2 9 time 3 Sets the time that the signal has to be above or below threshold of level 2 10 short long 3 Defines if the sensor signal has to be shorter or longer above or below level 2 than time 3 The easiest trigger criterion see example 3 in Figure 39 the detection of a short peak can be defined by setting level J and time 1 Level 2 and time 3 are neglected if time 2 is set to zero Figure 35 shows three different examples of sensor signal patterns and the trigger parameters to detect them Figure 35 The three examples show different patterns of interaction input signal curves The pattern in example 1 and 2 can be distinguished by setting the trigger parameter as visualized The third sensor signal pattern can be detected by a simple trigger criterion shown in example 3 5 Portable FES System 79 1 In the first example a sensor signal pattern of
175. r ein station res und ein portables System entwickelt Versuche mit diesen Prototypen im Rahmen einer Machbarkeitsstudie zeigten dass sich mit diesen Greifneuroprothesen die Greiffunktion der Probanden signifikant verbessern liess In einer zweiten Phase richtete sich das Forschungsinteresse auf die L sung von wissenschaftlichen technischen und klinischen Fragen welche im Hinblick auf eine m gliche Kommerzialisierung eines FES Systems gel st werden mussten In Zusammenarbeit mit einem der weltweit f hrenden Hersteller von Neurostimulatoren der Schweizer Firma Compex SA wurde ein vielseitig anwendbares FES System als Basis f r Greifneuroprothesen entwickelt welches alle Anforderungen der Neuroprothesen erf llt Die wichtigsten Forschungsziele der Arbeit waren 1 Die Entwicklung einer Hardware Plattform FES System welche ein schnelles Testen von Konzepten und Methoden zur Wiedererlangung oder Verbesserung der Greiffunktion bei Tetraplegikern erlaubt 2 Die Entwicklung von unterschiedlichen Steuerungsarten welche es den Benutzern der Neuroprothese erm glichen unterschiedliche Greifarten auszuf hren Das Schwergewicht lag dabei auf der Entwicklung von elektromyographischen EMG Steuerungsarten mit welchen Querschnittgelahmte mit einer hohen Tetraplegie mittels willk rlicher Aktivierung ausgew hlter Schulter oder Unterarmmuskeln die Greiffunktion steuern und nachregeln K nnen 8 3 Die Entwicklung einer neuen Programmi
176. rasping Tetraplegic subjects have a very limited voluntary control over their upper extremities Functional electrical stimulation can help to improve their grasp capabilities Due to the limited voluntary movements of upper limbs the control of such neuroprostheses becomes an important issue The main objective in the development of grasp control strategies is to provide to a SCI subject an intuitive and easy to use control tool that does not restrict the subject s range of motion Additionally the control tool has to be aesthetically and socially accepted This means that bulky control tools or unnatural body movements during the control task are not desired by the subjects From the technical point of view the control tools have to provide a reliable control output Further the control tools have to be practical in sense of attaching them to the wheelchairs or to the subjects For the ETHZ ParaCare neuroprostheses five different control strategies were developed implemented in the rapid prototyping system as well as in the portable system and finally tested with tetraplegic SCI subjects Depending on the level of lesion and the subject s preference different control sensors and the control strategies can be chosen Also the subject s future place of living all subjects were in their first rehabilitation is taken into consideration when choosing a control strategy in order to provide them the most suitable one We have developed the follo
177. reader and turning on the stimulator all programmed stimulation parameters are set 6 EMG Signals Features Signal Acquisition Stimulation Artifact Removal Of all implemented control strategies that are described in the next chapter the EMG control strategies require the biggest recording an signal processing This chapter provides the necessary information about surface electromyography SEMG and stimulation artifacts SAs that are always present if SEMG is recorded during FES The chapter is structured in four main parts that describe e The most important characteristics of SEMG of voluntarily activated muscles in our applications They are SEMG randomness Because of the randomness and the zero mean of recorded SEMG signals from voluntarily activated muscle recordings performed at different times can be added or subtracted resulting in a random like signal with zero mean SEMG stationarity a minimal duration of SEMG from voluntarily activated muscles of about 250 ms is required that the activity level can be estimated or measured e The recording filtering and signal processing techniques The influence of the recording electrodes and SEMG amplifiers are discussed and the used amplifiers and signal processing techniques are presented e Proposed methods that reduce stimulation artifacts in SEMG signals Discussed are The characteristics of the SA the influence of the electrode skin impedance and the influence of the SEM
178. rejected by four subjects for the following reasons Subject G was emotionally unstable and refused to collaborate with our group Subject H already had a good tenodesis grasp and did not benefit much from the neuroprosthesis As for subjects F and E they improved their grasp function during FES training to the point that they did not need the neuroprosthesis any longer Our tests showed that the best candidates for the proposed neuroprosthesis for grasping were subjects with C4 C5 or C5 complete SCI lesions or equivalent Subjects with lower lesion levels such as C6 complete SCI subjects or incomplete SCI subjects could also benefit from the neuroprosthesis for grasping However the success rate with these subjects was much lower since they often have a partially functional grasp subject sex born __ disability arm __ after injury control strategy outcome A M 1962 C5 complete right 8 months proportional EMG _ accepted B M 1979 C4 incomplete right 3 months sliding resistor accepted C F 1959 C5 complete right 5 years push button accepted D M 1966 C4 C5 complete right 2 months push button accepted E F 1928 C6 incomplete right 2 months push button rejected F M 1977 C6 incomplete left 7 months discrete EMG rejected G M 1983 C6 incomplete right 4 months push button rejected H M 1935 C3 incomplete right 2 months push button rejected I M 1967 C6 incomplete right 2 m
179. rode 2 2 3 Epimysial Electrodes Epimysial electrodes consist of a platinum disc coated with silastic material or Teflon on the back side The electrode cables are made of strong leads configured in a double helix and cast in silastic tubes The one side coated platinum discs are sutured in an invasive surgery to the outer epimysium membrane of the muscle body They produce an asymmetric electrical field that can be directed to stimulate a specified muscle close to the motor point without stimulating other muscles In practical applications the electrode is aligned close but not at the motor point of the targeted muscle The sub optimal electrode placement reduces the sensitivity of the nerves to small electrode displacements generated by limb movements Additionally it prevents an all or none stimulation with a poor control of the muscle when changing the stimulation intensity Smith et al 1987 This type of electrode was developed by the Cleveland group and is used in the commercially available Freehand system The epimysial electrode is selective and stable over time but has the disadvantage of being difficult to implant Epimysial Elecrode 2 Principle and Function of Neuroprostheses for Graspin 3 2 24 Transponder Electrodes BIONs BION is a microminiature electrical stimulator that can be injected into muscles or near nerves It consists of a cylindrical glass capsule of 2 mm diameter and 13 mm length with hermet
180. rode see Figure 56 A or more ulnar and proximal to the wrist depolarization electrode and distal of the medial epicondyle of the humerus charge balancing electrode if the deeper muscles should be reached see B Electrode placement Finger flexion A Finger flexor motor points B Deep muscle stimulation a Depolarization electrodes Figure 56 Two different electrode configurations can be used to generate finger flexion The electrode positions shown in A stimulate the M flexor digitorum superficialis and profundus at their motor points but activate also the wrist flexor muscles if not denervated The configuration shown in B mainly generates an electrical field in the deeper layers activates M flexor digitorum superficialis and profundus and not the wrist flexors but requires a higher amplitude than configuration A In configuration A the motor points see Figure 60 of the finger flexors M flexor digitorum superficialis and profundus see are stimulated The electrode distal of the medial epicondyle of the humerus flexes with low stimulation amplitudes digits four and five and only with increased amplitude also digits three and two By placing a second small electrode medial about 12 15 cm proximal from the wrist joint the parts of M flexor digitorum profundus and superficialis that actuate the second and third digits can be activated with lower amplitude When the finger flexors are stimulated at the positions
181. s second 1000 ms battery check 1000 ms minute 60 sec hours 60 min stimulation frequency 1 100 Hz 8 1000 ms pulse sequence interpreter pulse mode 8 1000 ms look up table routine pulse mode 8 1000 ms fast A D scan 8 kHz 250 us EMG processing 100 ms Table 4 Processing times and timer rates of the real time routines Since the Compex Motion controller software contains parts that are property of Compex SA the detailed program structure program flowcharts or the source code cannot be published in this thesis However the program concept and methods are presented in adequate detail 5 4 Compex Motion Programming Software The main window of the Compex Motion programming software is shown in The Compex Motion programming software allows one to program all stimulation parameters stimulation patterns and user interactions using GUIs On the left side of the GUI s main window in Figure 30 the channel dependent default amplitudes maximum pulse amplitudes and maximum pulse widths and the muscle names can be 5 Portable FES System 69 entered Right of these controls the stimulation patterns are built using graphical icons that can be placed in a consecutive order in so called time lines A time line shown in consists of empty boxes that can be filled with the graphical icons The icons called function primitives can be dragged and dropped from an icon pool that pops up when the left mouse button is pressed in the region of a box Eac
182. s are used Asymmetric stimulation pulses are produced such that an AP is only generated under one of the two stimulation electrodes whereas symmetric biphasic pulses generate APs under both stimulation electrodes Figure J shows the most commonly used pulse forms 2 Principle and Function of Neuroprostheses for Grasping 28 Figure 9 Commonly used pulse forms The depolarization pulse is for all pulse forms rectangular with a pulse duration up to 300 us and the hyperpolarization pulse is a not existent monophasic pulse b the same as depolarization pulse symmetric biphasic pulse c longer than the depolarization pulse with a sub threshold amplitude asymmetric rectangular biphasic pulse or d shorter than the depolarization pulse with a sub threshold pulse duration asymmetric exponentially decreasing biphasic pulse All biphasic pulse forms have in common that the charge is balanced The main difference between symmetric and asymmetric pulses is that the hyperpolarization pulse positive pulse of the asymmetric pulse has either a small amplitude Figure ora small pulse duration Figure 4a such that the stimulation intensity under the charge balancing electrode also called indifferent electrode is below motor threshold Asymmetric biphasic pulses are preferably used for the activation of small muscles e g for finger flexion extension if a greater muscle selectivity is needed If the electrode pair can be placed on the same mus
183. s faced with the fact that exact knowledge about the human system and the ongoing processes during rehabilitation are not completely understood and therefore make an engineering approach rather difficult Simple fault tolerant and robust approaches have to be chosen for practical clinically applicable solutions In the last few years the availability of portable battery powered high computational power in small size and miniaturized sensor systems made the application of modern technology for assistive devices for rehabilitation much more feasible 13 1 Introduction 14 One of the cutting edge fields in rehabilitation engineering is the field of neuroprosthetics as it combines almost perfectly the artificial world of electronics with the human circuits called pathways with the aim to improve lost or pathological functions The field of neuroprosthetics which has been existing for more than 40 years experienced a renaissance in the last decade The main concepts that were invented by the founders of the field Liberson Vodovnik or in case of neuroprostheses for grasping by Long and Masciarelli Liberson et al 1961 Long et al 1963 Vodovnik et al 1967 Vodovnik et al 1965 were brought to a clinical applicable commercial level by people like Peckham Nathan and Prochazka Ijezerman et al 1996 Peckham 2001 Peckham et al 1992 Peckham et al 1996 Prochazka et al 1997 Neuroprostheses for grasping are artificial systems that
184. s measured or the control variable hits its limit This control scheme is a little bit more complicated to learn but it allows the subject to control the grasp over a prolonged time Both control schemes are implemented in the portable system They can be selected in the analog control window see Figure 64 in the analog pulse amplitude control menu The control variable is processed be choosing the EMGA EMGB input and the control variable ipt is processed by choosing IEMGA IEMGB see Section 5 4 4 Two active Compex EMG Biofeedback sensors are connected to the analog inputs A and B 7 Neuroprosthesis for Grasping 126 and record the voluntary SEMG control signals The look up tables of are set the same as for the analog sliding potentiometer control 7 4 Advantages and Limiting Factors of EMG Control Strategies Compared to Push Button and Potentiometer Control Strategies The analog SEMG control strategy which uses the voluntary EMG activity of the contralateral deltoid muscle is preferably used for SCI subjects with a complete lesion at the level C4 and C5 We made use of this control strategy in neuroprostheses for unilateral grasp with the arm that had a better reaching function The contralateral arm that controlled the neuroprosthesis was only used to stabilize the upper body Additional arm movements with the contralateral arm during the grasp task would have generated unwanted control signals and therefore were not desired
185. sed to strengthen the multi layer substrate A new type of self adhesive electrodes the Ultrastim electrode uses a scrim layer to redistribute the stimulation current that it receives from a metal connector pin on a garment see Figure 11 Therefore the Ultrastim electrode can be positioned more freely on the garment The second hydrogel layer delivers the stimulation current obtained from the metal pin to a scrim layer made from a good conductive carbon film The scrim layer homogeneously redistributes the stimulation current and provides it via a first self adhesive hydrogel layer to the skin UltraStim Electrode _ Construction p h Candecfive hydrogel layer fo miare alevirade ip garmani z Conductive fim with patented Gwrent Comoling technalogy ON pa h Conductive w bylage loyer for skin contact Figure 11 The specially designed electrode Ultrastim from Axelgaard Mfg Co Inc can be placed individually on the garment It is connected to the stimulator with a connector pad Reprinted from Axelgaard 2001 The main goal of the multilayer construction of the self adhesive electrode is to provide a balanced most equally distributed stimulation current density over the whole electrode to prevent the skin from burns Additionally the electrode substrate skin impedance is 2 2 Principle and Function of Neuroprostheses for Grasping 33 kept as low as possible to excite the least afferent nerves possible The worldwi
186. several milliseconds before they recover see Figure 40 6 EMG Signals Features Signal Acquisition Stimulation Artifact Removal 90 vl 10 ms a b ee ee Figure 40 Typical SAs generated on the finger extensor muscles when the SEMG is recorded a on the same muscle b on the finger flexor muscles and c on the contralateral ventral deltoid muscle Special fast recovery SEMG amplifier circuits that help to reduce the amplifiers recovery time from saturation after a SA were proposed by Walker and Thorsen Thorsen 1999 Walker et al 1978 The following precautions can help to reduce the SA significantly Kornfield et al 1985 McGill et al 1982 McLean et al 1996 e optimized placement of the recording electrodes McGill proposes in McGill et al 1982 to place the electrodes on an equipotential line perpendicular to the stimulation electrodes although SEMG recordings can be easier interpreted if the recording electrodes are aligned with the muscle fiber direction see Section 6 2 1 e careful preparation of the skin shave and scrub the skin and apply gel to reduce the electrode skin impedance e shielded short stimulation and recording cables that minimize motion artifacts and electromagnetic interspersing e new not previously used stimulation and recording electrodes 6 3 1 Characteristics of Stimulation Artifacts in Measured SEMG In the 1982 published a paper on the nature of SAs by McG
187. stem The stationary system has three main advantages compared to other existing systems 1 The LabVIEW programming environment The easy to use and very powerful programming language LabVIEW in combination with National Instruments multifunction data acquisition boards is used to perform fast and reliable data recording and data processing tasks Although LabVIEW is only available for non real time operating systems OS of SUN MacIntosh and PC it is possible to program quasi real time closed loop applications using relatively low loop frequencies of less than 30 Hz Such loop frequencies are sufficient for controlling FES applications because the response time to a stimulus typically has a time delay of 60 to 80 ms The National Instrument data acquisition board can profit from the direct memory access DMA featured by the Windows OS and supported by LabVIEW Data acquisition does not require any CPU time Thus multichannel EMG data acquisition and processing can be done during closed loop stimulation using packet streaming technologies LabVIEW provides large libraries with all necessary subroutines needed to program such an application Additionally visualization tools like buttons bars sliders and graphs with zoom and other features can be used to build easy to use and multifunctional GUIs Such user interfaces are very important since they allow users during trials with subjects to react fast and to immediately adjust parameters in re
188. sterior syndrome occurs rather seldom in traumatic SCI subjects with cervical lesion A good clinical description of the syndrome groups can be found in Curt 1996 2 32 Hand Function For SCI subjects standing and walking are often considered the most important motor functions they would like to recover whereas the hand functions reaching and grasp are even more important for their independence in daily living Depending on the level of the spinal cord lesion tetraplegic subjects have a loss of finger hand arm and or shoulder functions see Table D This not only comprises motor function loss but also sensory and tactile function loss or impairment For some functions tetraplegic subjects can compensate the impairment with special skills and compensatory movements level of characteristic characteristic sensory hand function lesion innervated muscle movement impairment C3 C4 M trapezius shoulder elevation entire arm has no shoulder M sternocl no sensation control no hand M levator scap function C5 M teres min control of shoulder shoulder no hand function M inf sup spin girdle active normal lat or passive M deltoideus elbow flexion and proximal arm functional hand M biceps elevation impaired M brachialis M supinator C6 M pectoralis active proximal lat upper arm passive or active M teres maj arm adduction radicular C6 functional hand M pronator ter active distal arm
189. sthesis is started With a strong EMG activity of more than 3 V lasting for at least 1 5 s the grasp task can be interrupted like in the push button control mode The scale with the unit Volts of 7 Neuroprosthesis for Grasping 123 the processed EMG signal does not represent the real measured voltage of the EMG activity The EMG Biofeedback sensor has a fixed gain of 1400 and the pre processing software is equipped with an adjustable software gain that amplifies the processed EMG to activity levels with hard limits between 0 V no activity and 5 V full activity For each subject the software gain is chosen such that a normal activation results in a processed EMG activity in the order of 1 5 V and a strong activation of the control muscle is higher than 3 V 7 3 4 Sliding Potentiometer Control This strategy allows a continuous control of the force during grasp and release using a sliding potentiometer The portable system can be programmed such that the stimulator s analog input A or B measures the sliding potentiometer position and controls the pulse amplitudes of the four stimulation channels Depending on the position of the slider hand opening or closing is commanded In the neutral position in the middle of the sliding potentiometer no muscle is stimulated Pushing the sliding potentiometer forwards increases the stimulation pulse amplitude of the finger extensors and generates hand opening By pulling the sliding potentiometer b
190. stimulator using the same stimulation hardware and software All parameters can be set up with a graphical user interface GUI A fully operational data acquisition system is integrated with the capacity to record up to eight measured sensor signals with an overall sampling frequency of 83 3 kHz Additionally all relevant stimulation data are stored on a hard drive in real time The rapid prototyping FES system see consists of three main parts e A four channel PC controlled constant current FES stimulator The stimulator can be upgraded with a second analog output board with four additional channels to become an eight channel device e An eight channel multifunction data acquisition board LabPC from National Instruments e Two different LabVIEW programmed software versions of the grasp controller with GUIs are implemented on a standard Pentium PC 1 an event triggered pattern generator and 2 a continuous analog controller Figure 14 The rapid prototyping FES system consists of a wheeled rack a standard PC a four channel expandable to eight channels current regulated stimulator and different sets of sensors Because the system is assembled on a wheeled rack it can be easily transported to different rehabilitation facilities in the hospital 4 PC Based Rapid Prototyping FES System 49 4 1 Hardware 4 1 1 Electrical Stimulation Device The electrical stimulation device is built on Europe format printed boards that are housed in
191. strategy is not implemented in any of the commercially available systems although for a handicapped voice control represents an elegant way to control a neuroprosthesis and is widely used in environment control units The main reason that it is not commercialized might be the rather poor recognition performance and the demanding computational power of many voice recognition systems However in the last few years the voice recognition technology for portable devices made big improvements since high computational power in portable form is now available We have decided to buy the off the shelve product Voice Extreme from Sensory Inc Voice Extreme is a low power programmable artificial neural network ANN based voice recognition system Sensory Inc 2001 The ANN is hardware implemented in the RSC364 microcontroller which also includes AD DA converters for the microphone and the loudspeaker and memory for storing the sound files Sensory Inc 2001 Additional hardware in the size of the Compex Motion stimulator that can be placed under the stimulator was built It consists of the Voice Extreme rapid prototyping module RPM 364 a voltage regulator that converts the 5 V supply voltage of Compex Motion to 3 V used by the RPM 364 a push button that switches between learning and recognition mode some LEDs that display the status a DA converter that generates the output signal for controlling the stimulator an audio amplifier and a speaker Gareiss
192. t come with the Compex Motion device are stable but for neuroprostheses for grasping they are too thick Another problem occurred in our first generation portable system There one could easily break the battery connector when the battery was not plugged in with care In the rough clinical environment careful handling with electronic devices could not necessarily be assumed Such failures are natural in prototypes but they were sometimes annoying for the subjects Nevertheless as shown in Table 14 six subjects accepted the neuroprosthesis The term accepted is used to indicate that the neuroprosthesis was able to generate the desired function and that the subject adopted the prosthesis and used it to perform daily living functions When it is stated that the system was rejected this means that the FES system could not generate the desired function due to physiological reasons or the subject refused to use the neuroprosthesis despite the fact that it performed successfully or the subject recovered to the point that he she could generate the desired function without using the neuroprosthesis 8 Results with the Neuroprosthesis for Grasping 130 Figure 68 A C5 complete SCI subject 1 grasps 2 holds and 3 writes with a pencil using the sliding resistor control strategy The required grasp force for writing is very low Therefore the subject can write for 20 or more minutes without much muscle fatigue The neuroprosthesis for grasping was
193. t external aid The Bionic Glove The development of Prochatzka s Bionic Glove started at the University of Alberta Canada in 1989 Prochazka et al 1997 Similar to the Handmaster easy donning and doffing of the system were major concerns for developing the system As a result the neuroprosthesis stimulates the same muscle groups as the Handmaster but has a 2 Principle and Function of Neuroprostheses for Grasping 41 different control strategy and therefore targets a different segment of users A linear variable differential transducer LVDT measures the wrist angle and controlles hand opening and closing to augment tenodesis grasp The finger extensors are stimulated during wrist flexion and the finger flexors during wrist extension A dead zone between wrist extension and flexion can be adjusted individually This very intuitive way to control hand grasp can only be applied to people with functional C6 C7 SCI and to stroke subjects with active wrist function Three self adhesive stimulation electrodes are placed over the motor points of the target muscles and one balancing anodic electrode is placed proximal to the wrist crease On their back the electrodes have a contact stud A glove made of neoprene with inlaid stainless steel meshes at the electrode positions establishes the contact between the electrodes and the stimulator located on the glove The stimulator including rechargeable batteries weights only 200 gr For a multicenter
194. t spoken by the user never occurred Therefore the voice recognition technology of the used Voice Extreme platform by Sensory Inc is a safe but not very reliable control strategy The EMG control strategies are characterized by being a fast and intuitive man machine interface for commanding neuroprostheses for grasping It is in the nature of the humans to control finger and hand movements by muscular activation although in case of SCI subjects different muscles are used than normally After a retraining phase one or more muscle groups that can be voluntary activated act either as command or continuous control interface between the brain and the stimulated muscles Thus an automatism can be achieved that is quite convenient for the subjects Technically challenging is the elimination of the SAs in the SEMG caused by FES SAs are present in the whole body during stimulation and generate a positive feedback if SEMG is used in closed loop Close to the stimulation site the SA is about ten times higher than SEMG from maximally contracted muscles EMG amplifiers become immediately saturated In this thesis a novel SA removal algorithm is presented that is able to reduce the SA in real time even from changing SAs An ensemble averaged SA is subtracted from the recorded SA contaminated SEMG The stationarity and randomness of voluntary SEMG keeps the EMG activity almost completely unchanged while the SA is eliminated Residual voluntary SEMG in the ensemb
195. t state It will be the task of a multicenter trial performed with the Compex Motion system in the near future to assess the functional outcome in activities of daily living ADL and to show the applicability of the Compex Motion neuroprosthesis for grasping in early rehabilitation for restoring and or improving the hand function in SCI and stroke subjects Our results provide strong promise of a successful application The ninth chapter the conclusion gives a brief summery of the developed FES systems their performance and limitations and provides an outlook for future enhancements 1 5 Acknowledgements The work presented in the thesis was supported by grants from the Swiss National Science Foundation Switzerland SPP Biotechnology Project No 5002 044895 and the Federal Commission for Technology and Innovation Switzerland Project No 4891 1 2 Principle and Function of Neuroprostheses for Grasping This introductory chapter provides basic information about e the physiological processes involved in the generation of action potentials how they propagate through the nerve fibers and how muscle contraction is caused e how muscles are artificially activated by electrical stimulation e the influence of the different stimulation parameters pulse width pulse amplitude pulse shape and stimulation frequency e the concepts of the different currently used stimulation electrodes and their influence on muscle selectivity st
196. ten built in multi bipolar configuration They are mainly used to trace temporal spatial changes of propagating APs The electrode shape circular or rectangular small displacements of the electrode with respect to the muscle belly or changes in the interelectrode distance considering a minimum of 10 mm have no significant influence to the mean rectified SEMG signal Van et al 1984 Only the electrode size in longitudinal direction with respect to the main fiber direction of the recorded muscle has an integrative influence on the SEMG recording With increasing electrode size the electrode sensitivity is increasing for low frequencies and decreasing for high frequencies Therefore in case of circular electrodes the electrode size should not exceed 10 mm Placing bipolar electrodes orthogonal to the main fiber direction results in a different raw signal recording that is more difficult to interpret compared to the longitudinal configuration 6 2 2 Amplifiers The raw SEMG signal of a voluntary activated muscle has an amplitude of 500 to 1000 uV The output impedance of the recording electrodes is rather high variable and strongly influenced by the skin impedance The preparation of the skin determines the skin impedance It ranges from 10 Q for prepared skin shaved and roughened up to 10 Q for unprepared dry skin Therefore the main requirement for the SEMG amplifier is a very high input impedance in the range of 10 Q The band
197. ter thesis Automatic Control Laboratory ETH Zurich Zurich Gauthier M 1994 Schematics Commstim Fax Edmonton University of Alberta Graupe D Suliga P Prudian C and Kohn K H 2000 Stochastically modulated stimulation to slow down muscle fatigue at stimulated sites in paraplegics using functional electrical stimulation for leg extension Neurol Res 7 pp 703 4 138 Grieve R Parker P A Hudgins B and Englehart K 2000 Nonlinear adaptive filtering of stimulus artifact EEE Trans Biomed Eng 3 pp 389 95 Grill W M and Mortimer T 1995 Stimulus waveforms for selective neural stimulation IEEE Eng Med Biol pp 375 85 Guy H R and Seetharamulu P 1986 Molecular model of the action potential sodium channel Proc Nat Acad Sci pp 508 12 Guyton A C and J E Hall J E 1996 Textbook of medical physiology W B Saunders Company Philadelphia Handa T Takahashi H Saito C Handa Y Ichie M Kameyama J and Hoshimiya N 1990 Development of an FES system controlled by EMG signals In Proc 11th Ann Int Conf IEEE EMBS pp 2273 No 5 Handa Y Hoshimiya N Iguchi Y and Oda T 1989 Development of percutaneous intramuscular electrode for multichannel FES system JEEE Trans Biomed Eng 7 pp 705 10 Handa Y Ohkubo K and Hoshimiya N 1989 A portable multi channel FES system for restoration of motor function of the paralyzed extremities Automedica pp 221 31
198. tery that provided energy for amp h of stimulation At the same time a collaboration with the therapeutic and sports electrical stimulation device manufacturer Compex SA was established Its state of the art stimulator Compex 2 could satisfy our hardware needs However the software of the device can t fulfill the requirements of a FES system The Compex 2 is a four channel stimulation device with the capability to record up to two analog sensor signals It is small 30x80x148 mm and has a 64x165 dot matrix graphical display Originally Compex SA provides three different stimulation software versions for all kind of therapeutic and training applications In a joint project Compex SA and our team decided to reprogram the Compex 2 device and to enhance it with a new software that provides all features necessary for all foreseeable applications involving surface stimulation technology This enhanced version of the Compex 2 stimulator was given the new brand name Compex Motion 5 1 Basic Concept of the Compex Motion Stimulator One of the main requirements to an electrical stimulation system that it can be used as a neuroprosthesis or as a general research tool is flexibility Flexibility can only be warranted by allowing many parameters to be flexibly changed Two features of the Compex Motion hardware allowed to create a concept that a single device can be used for a wide range of applications 1 a PC compatible serial port and 2 a memory card i
199. the deeper muscle layers M flexor digitorum superficialis and profundus without activation of the wrist flexors located in the upper layer In addition to the finger flexors in configuration B the M flexor pollicis longus is also activated through the median nerve and causes thumb flexion at the interphalangeal joint Subjects with a sensory impairment above T1 level have a reduced sensation to electrical stimulation at electrode positions A but only subjects with a sensory impairment above level C6 feel reduced sensation for electrical stimulation produced at the location shown in configuration B 7 2 3 Electrode Positions for Thumb Flexion Opposition The thumb position and the way how the thumb is stimulates determined the type of grasp Our neuroprostheses performes either a palmar or a lateral grasp The two types of grasp can be achieved with three different configurations by stimulating A the thenar muscles B the median nerve or C the thumb flexors 7 Neuroprosthesis for Graspin 112 Electrode placement Thumb opposition flexion A thenar muscle B median nerve C thumb flexors Depolarization electrode Figure 58 Three different electrode positions can be chosen for thumb flexion opposition depending on the subjects sensation to the stimuli The median nerve stimulation is the most painful if the subject have full sensation but provides the best thumb flexion for a pinch grasp For the palmar grasp the thenar m
200. the finger extensors channel 1 during hand opening and the finger flexors channel 2 and the thenar muscle channel 3 during hand closing F extensors F flexors Thenar M transition over s 2 jap transition hand closed transition re transition hand opened Figure 42 The stimulation sequence was repeated with different transition times To change the stimulation intensity either the pulse width or the pulse amplitude were changed 6 EMG Signals Features Signal Acquisition Stimulation Artifact Removal 95 SEMG Recording SEMG signals were recorded from the wrist extensor muscle of the stimulated hand and from the contralateral deltoid muscle Two trials were conducted one without voluntary muscle activation and one with voluntary muscle activation In the second trial the subjects had to raise the hand of the contralateral side and had to extend the wrist of the stimulated hand to produce voluntary contractions during the entire trial Two Compex 2M4456 EMG Biofeedback sensors Compex SA 1996 gain 1400 bandwidth 100 4000 Hz were placed on the skin one between the finger flexor stimulation electrodes over the M extensor carpi radialis and one on the M pars clavicularis of the contralateral deltoid muscle see Figure 43 The sampling frequency was 10 kHz EMG recording M extensor carpi radialis Electrical Stimulation Finger extensors Finger flexors and Thumb thenar muscle E
201. theses and for recording and processing sensor signals that are used to control the system All parameters within one software can be stored in setup files and can be recalled Refinements of the parameters can be logged during different sessions conducted with subjects The stationary rapid prototyping FES system is very useful for the development of new concepts and for training the subjects in the initial phase of their rehabilitation when their mobility is still limited For later application of the FES technology where the patients are expected to take the system home we have developed a portable system Therefore we built six portable systems that use the same stimulation technology as the stationary system Additionally the six systems are enhanced with the capability to read and process the sensor information of push buttons sliding resistors and EMG signals from voluntary activated muscles during stimulation The processing and controlling software is implemented in assembler and has to be modified for each subject according to the parameters found with the stationary system Once programmed the systems can provide exactly the same function as obtained with the stationary system Some parameters can be changed in a setup menu the stimulator has a small display to visualize the parameters while major modifications of the control strategy must be programmed in assembler The portable systems worked reliably in ADL with six SCI subjects Sin
202. tic device with a fixed set of stimulation program libraries Its newly developed firmware and the LabVIEW programming software adds flexibility to the system that it can be used as research stimulator and neuroprosthesis It will be commercialized as Compex Motion with an excellent chance be become successful since its flexibility to serve as general device for many different applications in therapy clinical research and rehabilitation is novel and superior to all existing portable transcutaneous stimulators The presented thesis describes the hardware firmware and programming software of the stationary and portable FES systems and demonstrates the application of the FES systems as neuroprostheses for grasping It discusses the proper electrode placements the advantage and disadvantage of surface FES the installation of the neuroprostheses and proposes five different control strategies that command hand opening and closing All five control strategies were developed to be used with the stationary and portable neuroprosthesis The following strategies were proposed push button sliding resistor voice digital EMG and analog EMG control For clinical applications the push button and sliding resistor control strategies are the most robust ones They are easy to install to learn and to apply by the therapists and patients Additionally they are very fail safe what is an important feature in a clinical environment Except of a few wire and p
203. timulus with SA removal algorithm processed control Each ARV is calculated from 10 ms the first 2 5 ms are blanked raw or with the SA subtraction algorithm processed SEMG segments The consecutive ARVs are smoothened using a 2 order phase shift compensated butterworth low pass filter with a filter frequency of 0 8 Hz With the smoothening the non stationary ARVs from the 10 ms small SEMG parts are made stationary This algorithm represents the signal 9 6 EMG Signals Features Signal Acquisition Stimulation Artifact Removal 100 processing algorithm as it is used for the SEMG control strategies and can be performed in real time In the case the SAs are completely removed the generated ARV curves are a measure of the voluntary EMG activity The results show a very good match of the three ARV curves processed SA original control and processed control during the stimulation phases with constant pulse widths They are almost congruent see Figure 49 left panels when the pulse widths and amplitudes are constant Differences between the original control and the processed control would indicate that the algorithm influences the voluntary SEMG activity instead of only removing the SA This is the case for a very small forgetting weight p 0 1 see Figure 47 circle in right panel 1 orig control orig SA proc contro proc sh 1 orig control orig SA proc controi proc S 1 orig control
204. tinue with the execution of the next primitives Thus the synchronize primitive has to be entered into all active channels light green icon As already mentioned a channel is considered to be active as long as the end primitive is not executed The synchronize primitive number displayed in the text box below the icon indicates which synchronize primitives belong to each other in the different channels Human interaction primitives The human interaction primitives allow the user to control the stimulation pattern execution using external signals from sensors or man machine interfaces The primitives must be entered into all channels and are automatically synchronized light green icon As a consequence a human interaction is only detected when the primitives of all channels prior to the user interaction and user branch are executed Only the user interrupt primitive see below makes an exception from this rule Three different primitive types can be used to interrupt the stimulator from executing the stimulation pattern and or to do branching operations in the time line depending on a trigger criterion User interaction primitive The trigger criteria that are applied to the recorded signals from the two input channels A and B or the Compex push button input C are defined for each of the user interaction primitives see Section 5 4 4 If the user interaction primitive is placed in the time line the stimulator stops at that position in the time li
205. tivities of daily living they could not perform without it e SCI subjects with a complete C5 lesion were the best candidates for using the proposed neuroprostheses for grasping on daily basis as a grasp aid Subjects with lower lesion levels such as C6 or incomplete SCI subjects could also benefit from a better grasp function Some of them developed a partially functional grasp such that the neuroprosthesis was not anymore needed for simple grasp tasks e The applied neuroprostheses for grasping were all in development and used different technologies with different constraints Thus the systems could not be systematically assessed and compared The ETHZ ParaCare and Compex Motion neuroprostheses for grasping were tested with 10 SCI subjects see Table 14 Subjects A B D E F G and H were in house patients during their first rehabilitation Subject C was an ambulant patient subject I came for a second rehabilitation and L was rehabilitated in an other SCI unit He was the first subject that received our neuroprosthesis for grasping in an other center The recruitment of the subjects and also the testing of our neuroprostheses was performed in collaboration with occupational therapists OT physical therapists PT and medical doctors MD of our center A specialized FES group consisting of OT and PT was formed to support all FES activities If a subject was recognized as a potential candidate the MD examined the nerve conductivity grade o
206. totyping FES system provides to rehabilitation engineers medical doctors and to physical and occupational therapists a versatile and quickly adjustable tool for the adaptation of neuroprostheses to SCI and stroke subjects The complexity of such neuroprostheses demands a well structured design of the programming software Although the users ask for total flexibility in the adjustment of all parameters and 4 PC Based Rapid Prototyping FES System 55 curves the program should still have an A ce This can be achieved by using graphical user interfaces see and Figure 19 A fu Alena FES Spec Continuous Controller Figure 18 Graphical user interfaces allow fast and user friendly modifications of stimulation parameters stimulation patterns and settings of the man machine interface The figure shows the continous controller software of the rapid prototyping FES system for grasping For the generation of the graphical user interfaces the LabVIEW virtual instrument programming environment was chosen This programming language allows a very fast and easy implementation of complex instrumentation software It is suitable for quasi real time application on Windows PCs The expression quasi real time is used here for soft timing real time which means that on average over multiple loop cycles all data are processed in real time whereas in hard timing real time all processing tasks belonging to one loop cycle are processed within the cycle Be
207. two peaks a short and a long peak is detected In order to detect such a pattern the parameters are set such that the first peak has to be shorter above level 1 1 than time 1 3 The second peak has to come later than time 2 5 and has to be longer above level 2 7 than time 3 9 If these trigger parameters are fulfilled the criterion is fulfilled If this is the case the stimulator that has been waiting continues with the stimulation sequence 2 In the second example the user interaction primitive waits for sensor signal pattern with a long peak and a short valley The parameters are set such that the peak has to be longer above level 1 1 than time 1 3 The valley level 2 7 has to occur later than time 2 5 and has to be shorter below level 2 7 than time 3 9 Such a trigger criterion can be nicely distinguished from the setting in example 1 Therefore the trigger criterion in example 1 and the trigger criterion in example 2 can be used for the user branch function of the stimulator It can distinguish between the two signal patterns to control for example either a palmar or a lateral grasp 3 In the third example as already discussed a simple sensor signal pattern of only one peak is detected by the following parameter settings The peak has to be longer above level 1 1 than time I 3 By setting time 2 to 0 00 s the level 2 and the time 3 entries are ignored Note that 100 ms is the control frequency Therefore t
208. ulation channels are used The strategy could only be tested in a SCI subject with a first generation portable FES system that used a SA blanking technique With more than 2 stimulation channels this FES system produced positive feedback because of the residual SA tail that disturbed the control signal This constraint should be solved with the better SA removal algorithm described in Chapter 6 The push button and the sliding potentiometer control strategies are very robust to environmental disturbances and easier to handle for the therapists or the health care personnel Normally the push button or the sliding resistor is mounted on the wheelchair whereas the EMG sensor has to be placed every day very carefully on the patient Thus the simpler control strategies are preferably used in the hospital Nevertheless the SEMG control strategies offers a more natural way to control the neuroprostheses With the new Compex Motion stimulator these control strategies are available and can be used in activities of daily living 8 Results with the Neuroprosthesis for Grasping This chapter provides results we obtained from clinically testing the ETHZ ParaCare and the Compex Motion neuroprostheses for grasping with 10 SCI subjects in our rehabilitation center They can be summarized as follows e The neuroprostheses provided in all subjects an increased palmar and or lateral grasp force and function e With the neuroprostheses the subjects could perform ac
209. under laboratory conditions More recently the group combined the slippage control scheme with the fully implanted Freehand system and a percutaneous neural activity recording system which is still sensitive to infection lead breakage and environmental noise An implantable neural signal amplifier with telemetry is under development 2 4 2 Surface FES Systems As an alternative to implanted systems various surface FES systems for hand grasp were developed that have comparable capabilities The Handmaster One of the pioneers in developing and building neuroprostheses for grasping using surface electrodes is the Beer Sheva group in Israel Toward the end of the 1970s Nathan charted the forearm surface for FES applicability Nathan 1979 He could show that out of fifteen muscles that articulate hand and fingers thirteen can be stimulated using surface electrodes Nathan 1992 The Beer Sheva group developed a special high resolution electrode array using carbonized rubber A voice controlled command interpreter combined with a 24 channel stimulator and two electrode arrays on upper arm and forearm activated arm and finger muscles for hand opening and closing and arm movements Nathan et al 1990 The system was far too complicated to be used outside laboratory conditions Reducing the system to 3 stimulation channels and a simple push button control strategy NESS Inc chose a rather slow but steady market approach starting in 1994 in Israel an
210. us Although three stimulation channels can be used per ramus the muscle selectivity is rather poor Attempts to stimulate the lumbar anterior roots for walking Donaldson et al 1997 Rushton et al 1996 using Brindley electrodes failed due to poor muscle selectivity For the same reason cuff electrodes are not used for FES of upper extremities at this time With the design of new multichannel electrodes people try to overcome this problem by only stimulating parts of the nerve bundle 2 2 2 Percutaneous Intramuscular Electrodes Recent percutaneous stimulation electrodes are made of multistrand fine filament stainless steel wires and are coated with teflon The wires are manufactured in a spiral configuration and housed in a silastic tube to reduce electrode breakage Handa et al 1989 Percutaneous intramuscular electrodes are implanted using a long hypodermic needle The procedure can be done minimal invasive Thus percutaneous electrodes can be implanted more easily than epimysial electrodes see below that need an almost uncovered target muscle for implantation Percutaneous electrodes are commonly used with external systems if a good muscle selectivity is required Especially the Sendai group and in the 1980s the Cleveland group used such percutaneous electrodes for FES of upper extremities A disadvantage of percutaneous electrodes is the susceptibility to infection migration problems and electrode breakage Percutaneous Wire Elect
211. uscles of the thumb mainly the M opponens pollicis using configuration A of are stimulated to bring the thumb into opposition of the fingers The palmar grasp can be used for lifting heavier objects like glasses tetra packs or books The lateral grasp is performed by stimulating either the M flexor pollicis brevis and the M opponens pollicis through the median nerve see configuration B in Figure 58 or the M adductor pollicis a deep muscle by placing the depolarizing electrode on the dorsal side of the hand between thumb and index finger configuration C The neutral electrode in C is placed on the palmar side proximal to the wrist With this configuration thumb adduction can be stimulated even with surface electrodes The lateral grasp is used to hold smaller and lighter objects like pens eating tools floppy disks etc The stimulated muscles of all three configurations can be seen in 7 Neuroprosthesis for Graspin 11 Flexor pollicis brevis Opponens pollicis Figure 59 The M opponens pollicis opposes the carpometacarpal joint of the thumb that the thumb can oppose the fingers for a palmar grasp The M adductor pollicis adducts the carpometacarpal joint such that the thumb moves toward the plane of the palm The M flexor pollicis brevis flexes the metacarpophalangeal and carpometacarpal joints toward the little finger Reprinted from Kendall et al 1993 The use of a motor point location chart shown in is very help
212. ut signal The module is able to distinguish different signal patterns in the interaction input signal Two trigger levels the signal slopes positive or negative at the trigger level cross point the time duration between the two trigger level cross points and the above or below threshold time duration are used as criterion to classify the different trigger patterns Figure 34 shows the GUI that is used to set the ten trigger parameters One can see graphically the two trigger levels and the three time duration With these trigger parameters sensor signal patterns consisting of maximum two peaks or two valleys can be distinguished Be Set Imggen E 55 Lt 4 4 a 3 2 2 i 1 a a A 4 Figure 34 All parameters of a trigger criterion can be entered either numerically or graphically Additionally the trigger criterion can be stored or loaded The following ten parameters have to be defined in order to detect a signal with two peaks or two valleys 1 level 1 Sets the trigger level threshold of the first peak above threshold or valley below threshold 2 peak valley 1 Sets the above or below threshold condition for level 1 3 time 1 Sets the time that the signal has to be above or below threshold of level 1 4 shorter longer 1 Defines if the sensor signal has to be shorter or longer above or below level 1 than time 1 5 Portable FES System 78 5 time 2 Sets the time between the two peaks or two valleys If paramete
213. ven if 6 EMG Signals Features Signal Acquisition Stimulation Artifact Removal 93 the SA is changing over time The recorded SEMG signal is processed in pieces of data with the duration of one interpulse interval Real time in this context means one result from pulse to pulse 6 4 1 Algorithm From pulse to pulse the algorithm subtracts an extracted pure SA from the SA contaminated SEMG signal The SA is extracted from the SA contaminated SEMG signal using an ensemble averaging algorithm Ensemble averaging means that SEMG curves between two stimulation pulses are summed and divided by the number of curves Because of the ensemble averaging the random like SEMG is canceled out and a pure artifact signal is obtained This artifact can be subtracted from the last SEMG curve and if the SA did not change an SA free SEMG signal can be obtained Changes of the SA cannot be eluded in FES applications because the SA changes with changes of the stimulation intensity An adaptation of the extracted SA to changes of the stimulation intensity is necessary Therefore a moving window with exponential forgetting was implemented in the ensemble averaging algorithm Earlier SEMG curves are less weighted than later curves Like this the ensemble averaged SA exponentially forgets its past and adapts to changes of the SA The SEMG curve between two pulses at time can be written as X t ct t x 2 For each sample x n t the following recursive first order
214. width requirements for SEMG are not very high The SEMG signal frequency ranges from 0 to 5000 Hz but has 95 of the power density below 400 Hz Normally SEMG is recorded with differential amplifiers A high common mode rejection ratio CMRR is desired in order to reduce artifacts from electromagnetic interference CMRRs of more than 100 dB are normal 6 EMG Signals Features Signal Acquisition Stimulation Artifact Removal 87 Nowadays available low noise OP Amps and instrumentation amplifiers satisfy all the requirements needed for SEMG recordings 6 2 3 Specifications of the Used SEMG Amplifiers Two different SEMG amplifiers are used to control the neuroprostheses with SEMG activity from voluntary contracted muscles e A Madaus eight channel stationary EMG amplifier from TPM GmbH Madaus Medizin Elektronik 1993 e Compex EMG Biofeedback sensors the SEMG amplifier is integrated in the sensor from Compex SA Compex SA 1996 adjustable Madaus SEMG amplifiers 9h or the Compex EMG Biofeedback sensors are used With the portable Compex Motion stimulator only the Compex EMG Biofeedback sensors with build in SEMG amplifier see Figure 39b are applied The Madaus EMG amplifier is a software controlled eight channel stationary EMG amplifier with variable gain and adjustable low pass and high pass filters It can be remotely controlled via a standard PC serial communication port Due to its size standard 19 3 HE it is only used
215. wing control strategies for the event triggered stimulation pattern generator software of the rapid prototyping system e push button control e voice control e digital SEMG control For the continuously controlled stimulation software of the rapid prototyping system the following control strategies were developed e analog sliding potentiometer control e analog SEMG control All five control strategies are also used with the portable Compex Motion FES system In the following the five implemented control strategies will be described how they are used with the portable system The control task for all the strategies is to command hand opening and hand closing Hand closing either performed a palmar grasp allowing the subject to hold bigger objects like cans glasses telephone receivers etc or a lateral grasp allowing the subject to hold smaller objects like floppy disks keys eating utilities pencils etc The analog sliding potentiometer the analog SEMG and the voice control strategy also allow a control of the grasp force and not only the start of the grasp and release sequences 7 Neuroprosthesis for Grasping 116 7 3 1 Push Button Control The push button control strategy can be implemented very easily in the portable system By pressing a push button the grasp task is initiated and the following grasp pattern is executed First the stimulation pulse width of the finger extensor muscles is increased following a ramp function in
216. y a push button the controller box generates digitally coded stimulation pulses that are transmitted via radio frequency signal to the implanted stimulator using a transmission coil The transmission coil consists of a signal and a power coil The signal coil sends a 24 bit signal for each of the 16 stimulation channels and receives an 8 bit status signal per channel from the implant Each implant requires a controller box Other Implanted Systems for Grasping The center for sensory motor interaction in Aalborg SMI developed neuroprostheses for the restoration of lateral hand grasp using natural sensory feedback In the first version they used the percutaneous intramuscular FES system from the Cleveland group After experiencing some problems with electrodes in the first subject 6 out of 11 electrodes broke they decided to use NEC San ei Instruments Ldt percutaneous electrodes from the Sendai group for the second subject Additionally self made nerve recording cuff electrodes Haugland 1997 were implanted around the palmar digital nerve that recorded the neural activity from the cutaneous mechanoreceptors of the index finger Lickel Haugland et al 1999 Lickel 1998 developed a control strategy that used the slippage information obtained from the recorded index finger nerve signals 2 Principle and Function of Neuroprostheses for Grasping 40 to control the lateral grasp force The group could show that the proposed system worked at least
217. y when the sequence lasts for at least 250 ms 6 2 SEMG Recording Techniques SEMG recordings are strongly influenced by the type of electrodes the quality of the amplifier and the used recording filters 6 2 1 Electrodes Various SEMG recording electrodes are available They differ in the contact material in the number and configuration of the electrode poles and in the electrode size and shape Two types of electrodes are commonly used Electrodes with direct metal contact to the skin or floating electrodes using a electrolytic paste as an interface between the skin and the electrode The electrode paste reduces the electrode skin impedance Rarely self adhesive recording electrodes consisting of conductive rubber are used Their main disadvantage is the relatively high impedance of the conductive rubber On the other hand they can easily be mounted The most frequently used recording electrodes are made from Ag AgCl They are stable over time and have very low noise Electrodes made from AgCl Ag Au etc are also used Popular electrode configurations are monopolar bipolar and multipolar configurations The monopolar configuration is used when the muscle body is very small or very close to another muscle The reference electrode in the monopolar configuration often is a conductive strip that is fixed around the measured limb This configuration provides a rather low signal to noise ratio but has a good muscle selectivity The same appli
218. ze the motor neuron and generate nerve APs The nerve APs propagate along the motor neuron branch and are transmitted to the motor units that articulate the limb by muscular contraction The pulses are provided to specifically selected muscles or muscle groups through electrodes that are placed close to the motor neurons of the selected muscles The electrodes must be placed in number and location such that they produce useful synergic muscle contraction patterns for the intended limb function 1 2 State of the Art of Neuroprostheses Only a few neuroprostheses that restore or improve the grasp function in SCI and stroke subjects are commercially available and or clinically used They can be divided into two main categories implanted systems and non invasive systems Both categories have their own advantages and disadvantages Presently two systems are commercially available in Europe and USA 1 the Freehand system from Neurocontrol Inc Keith et al 1988 Smith et al 1987 an implantable FES system and 2 the Handmaster from Ness Inc Ijezerman et al 1996 a system using transcutaneous surface stimulation electrodes A third system the FESMate system from NEC Inc Takahashi et al 1999 1 Introduction 15 also an implantable system is used for both upper and lower extremities It is applied in Japan Korea and Taiwan mainly in research Other clinically used neuroprostheses for grasping are the Bionic Glove from the University

Fulltext - ETH E-Collection

Contents

Download Pdf Manuals

Related Search

Related Contents