Home

USER'S GUIDE - Biosignal Analysis and Medical Imaging Group

1. Poincare plot SD1 SD2 Recurrence plot Mean line length Lmean Max line length Lmax Recurrence rate REC Determinism DET Shannon Entropy ShanEn Other Approximate entropy ApEn Sample entropy SampEn Detrended fluctuations DFA 1 Detrended fluctuations DFA 0 2 Correlation dimension D2 i y a 06 08 1 12 14 16 18 log n beats Results are calculated from the non detrendedselected RR series 24 Jan 2005 11 27 18 Mika Tarvainen Kubios HRV Analysis version 2 0 beta 1 Department of Applied Physics Department of Applied Physics University of Kuopio University of Kuopio Finland Figure 5 3 Sample run 1 results for the standing period of the orthostatic test Kublos HAV Andiveis Biosignal Analysis and Medical Imaging Group N Department of Physics University of Kuopio FINLAND version 2 0 beta 5 2 Sample run 2 Time varying analysis 56 5 2 Sample run 2 Time varying analysis In the second sample run we show how to make time varying analysis for the whole or thostatic HRV measurement First of all we need to enable the time varying analysis from the preferences To do this select Edit preferences from the File menu and check the time varying analysis from Analysis options category of the preferences window If you do not want to view any of the time domain frequency domain or nonlinear analysis results and nor do you wish to include any of them in the re
2. current analysis have not been saved user is prompted to do so EA Dd Cc 0 All the above actions are also available on the user menus The File menu includes Open Save Results Save Results As Edit Preferences Close and Quit commands The Open Save Results Edit Preferences and Close commands work exactly as the corresponding toolbar buttons The difference between the Save and Save As commands is that when the results have already been saved the Save command automatically overwrites these results whereas the Save As command asks the user for a new file name The Quit command of the File menu is for exiting from the software The View menu includes Markers menu and Report Sheet command The latter works as the corresponding toolbar button The Markers menu on the other hand is for displaying possible stimuli or event markers presented in the experimental procedure and stored in the data file If no markers are found from the data file the Markers menu will be disabled Finally the Help menu includes the About HRV Analysis Software command which opens the same about dialog as the corresponding toolbar button 4 3 Saving the results The analysis results can be saved by selecting Save Results or Save Results As from the File menu or by pressing the save button on the toolbar This will open a file save dialog in which the saving type can be selected There are three different types in which the results can be saved That is the r
3. Brown noise integral of white noise l lt a lt 1 5 Different kinds of noise a 1 1 f noise 05 lt a lt l Large values are likely to be followed by large value and vice versa a 0 5 white noise 0 lt a lt 0 5 Large value is likely to be followed by small value and vice versa Typically in DFA the correlations are divided into short term and long term fluctuations In the software the short term fluctuations are characterized by the slope a obtained from the log n log F n graph within range 4 lt n lt 16 Correspondingly the slope az obtained from the range 16 lt n lt 64 characterizes long term fluctuations see Fig 3 2 dy Ku bios HAV Analysis Biosignal Analysis and Medical Imaging Group i Department of Physics 2 0 bet eee University of Kuopio FINLAND 3 3 Nonlinear methods 27 0 8 1 e 41 2 D S 1 41 1 61 1 8 0 6 0 8 1 1 2 1 4 1 6 1 8 log n Figure 3 2 Detrended fluctuation analysis A double log plot of the index F n as a func tion of segment length n a and ag are the short term and long term fluctuation slopes respectively 3 3 5 Correlation dimension Another method for measuring the complexity or strangeness of the time series is the corre lation dimension which was proposed in 13 The correlation dimension is expected to give information on the minimum number of dynamic variables needed to model the underlying system and it can be obtained as follows Similarly as
4. Department of Physics eee pena University of Kuopio FINLAND 4 2 The user interface 36 RR Interval Series Options Artifact correction Apply Samples for analysis Number of samples Sample 1 Range s 1 351 Length s 350 Remove trend components Method Smoothn priors z Lambda 500 Figure 4 3 The RR interval series options segment of the user interface HO MARKERS R Move Remove Add Eo D0 02 01 00 02 02 00 02 03 00 02 04 00 02 05 00 02 06 00 02 07 00 02 08 00 02 09 00 02 10 Time h min s Range 00 00 00 00 01 40 00 03 20 00 05 00 00 06 40 00 10 00 00 11 40 Time h min s Range 73 HE SSSA gt Figure 4 4 The data browser segment of the user interface two ways If only RR data is available the artifact corrections described in Section 4 2 1 are displayed in the RR axis If the ECG is measured these corrections can be made by editing the misdetected R peak as follows Each detected R peak is marked in the ECG axis with a mark Each mark can be moved or removed by right clicking it with the mouse see Fig 4 4 In addition new R peak markers can be added by either right clicking some other marker and selecting Add or by pressing the uppermost button on the right hand side of the ECG axis Moved or added R peak markers are by default snapped to closed ECG maximum but manual positioning can also be achieved by pressing the middle button on the right
5. Full installation Selected components Install Kubios HAY Analysis Install MATLAB Component Runtime uncheck only if MATLAB Component Rui O Start Menu folder Kubios HRY Analysis gt Install Cancel Kibios HAY Analysis Biosignal Analysis and Medical Imaging Group vercion OO beta Department of Physics University of Kuopio FINLAND 1 2 Installation 9 8 Wait while the installer copies the necessary files on your system This can take a few minutes 18 Setup Kubios HRV Analysis Installing Please wait while Setup installs Kubios HAY Analysis on pour computer Extracting files C Program FilestKubios HRY Analysis MCRtemp MCAlnstaller msi Cancel 9 After the installer has finished copying the necessary files the MATLAB Compo nent Runtime installer starts Choose Next to continue the installation NOTE If you have selected not to install the MATLAB Component Runtime earlier in the Select Components page you can skip to section 15 i MATLAB Component Runtime MATLAB Component Runtime Cancel vercion OO beta Department of Physics Ku bios HAV Analysis Biosignal Analysis and Medical Imaging Group University of Kuopio FINLAND 1 2 Installation 10 10 The MATLAB Component Runtime setup wizard starts Choose Next to continue i MATLAB Component Runtime Welcome to the MATLAB Component Runtime Setup Wizard The installer will guide you through the steps required
6. General analysis al al Neuroscan_data cnt 20 11 03 15 24 59 HRV Analysis General Results Page 3 4 RR Interval Time Series Results for single samples sample 2 2 00 06 30 00 12 10 Detrending method Smoothn priors A 500 Pars en pey Pt 00 01 40 00 03 20 00 05 00 00 06 40 00 08 20 00 10 00 00 11 40 Selected Detrended RR Series T Pear Y rt 00 00 00 00 00 50 00 01 40 00 02 30 00 03 20 00 04 10 00 05 00 Time h min s Time Domain Results Distributions Variable Units Mean RR s STD RR SDNN s Mean HR 1 min STD HR 1 min RMSSD ms NN50 count pNN50 RR triangular index 0 8 0 9 1 a li 50 55 60 65 70 TINN ms RR s HR beats min Frequency Domain Results FFT spectrum Welch s periodogram 128 s window with 50 overlap AR Spectrum AR model order 16 not factorized 0 04 0 04 0 03 02 0 01 0 0 i 0 2 0 3 0 4 y E 0 2 0 3 0 4 Frequency Hz Frequency Hz Frequency Peak Power Power Frequency Peak Power Power Band Hz ms Band Hz ms VLF 0 0 04 Hz 0 0352 254 15 1 VLF 0 0 04 Hz 0 0391 197 7 LF 0 04 0 15 Hz 0 0859 1167 69 3 LF 0 04 0 15 Hz 0 0781 1218 72 5 HF 0 15 0 4 Hz 0 1523 263 15 6 HF 0 15 0 4 Hz 0 1523 266 15 8 Total 1683 Total 1680 LF HF 4 441 LF HF 4 583 Nonlinear Results Poincare Plot Detrended fluctuations DFA Variable
7. Installation 11 12 Confirm your selections by choosing Next i MATLAB Component Runtime Confirm Installation The installer is ready to install MATLAB Component Runtime on your computer Click Next to start the installation 13 The installation begins The process takes some time due to the quantity of files that are installed i MATLAB Component Runtime Installing MATLAB Component Runtime MATLAB Component Runtime is being installed Please wait Cancel Biosignal Analysis and Medical Imaging Group Department of Physics University of Kuopio FINLAND Kubios HRV Analysis version 2 0 beta 1 2 Installation 12 14 When the installation completes click Close to close the MATLAB Component Run time installer i MATLAB Component Runtime Installation Complete MATLAB Component Runtime has been successfully installed Click Close to exit 15 Finally to exit the setup wizard click the Finish button 18 Setup Kubios HRV Analysis Completing the Kubios HRV Analysis Setup Wizard Setup has finished installing Kubios HRY Analysis on pour computer The application may be launched by selecting the installed icons Click Finish to exit Setup Now all the components required for the Kubios HRV Analysis software are installed You can launch the Kubios HRV Analysis by selecting it from the created Start Menu folder or by clicking the Desktop icon if created Please note that
8. PSD s Hz o Teny Bard E cH qe LF 000 15 Hy 00703 7164 DzasZ 510 HF 0 150 HH Tot 1560 o4 02 1 log n heat Result oe ciodakd tom be rorrde rerdedselecied AR selez 12 1 4 16 12 Kublos HRV Aredysis version 20 bets 1 Deparmenl or Applied Physics Uriuersi y of Kuglo Findand Figure 4 11 The first report sheet including all the time domain frequency domain and nonlinear analysis results calculated by the software Kubios HRV Analysis version 2 0 beta Biosignal Analysis and Medical Imaging Group Department of Physics University of Kuopio FINLAND 4 3 Saving the results 45 Report Page 2 SEE e File Edit Page BSASSA HRV Analysis Time Varying Results O Pane wiw bra dngle cample 00 12 16 Derrerdirg me hod SmooPn prior A S00 RRinterval Time Boris a Beleo d Cetended RR DHe t oz a2 Time Varing awite Window wid h GOs ard wirdow shi 5 Mean RR BTD ard quiten HA scade Mean ARG HR eskimiry pNNSOCH E i 2 Peak tequency HI Frequency HI LEAF ralo Bard power 1 174k Z0DS 05 15 12 Mika Taren Kublos HRV Aredysis version ZO bets 1 Deparmenl ot Applied Physics Deparment or Applied Physics Urduers ly of Kuoglo Urduersi y of Kucglo Findand Figure 4 12 The second report sheet including all the time varying analysis results calcu lated by the software Kubios HAV Analysis Biosignal Analysis and Medical Imaging Group Department of Ph
9. R12 or higher The MAT file includes a single structured array variable named Res The Res variable includes the numeric results as well as the RR interval data and all the analysis options This saving option is aimed for MATLAB users and makes the further analysis or processing of the HRV data in MATLAB much easier The Res structure includes four fields which are shortly described as follows f_name File name of the analyzed data file f_path Full path for the analyzed data file CNT Basic information of the data file the field name refers to Neu roscan CNT file for historical reasons HRV Used analysis options RR interval data and all analysis results The HRV field is clearly the most essential one of these fields The HRV field includes six fields the contents of which are shortly described as follows Param The analysis options used in the calculation of the results Data The RR interval data Statistics Time domain analysis results Frequency Frequency domain analysis results NonLinear Nonlinear analysis results TimeVar Time varying analysis results vercion oO beta Department of Physics Kubios HAV Analysis Biosignal Analysis and Medical Imaging Group University of Kuopio FINLAND 4 4 Setting up the preferences 47 Preferences User information User Details Analysis options Name Mika Tarvainen Advanced settings Department Department of Applied Physics Organization University of Kuop
10. The software is a considerable extension to the previous version version 1 1 of the software described in 32 The software has been programmed using MATLAB 7 0 4 Release 14 Service Pack 2 The MathWorks Inc and was compiled to a deployable standalone application with the MATLAB Compiler 4 2 The MATLAB Component Runtime MCR is required for running Kubios HRV Analysis Kubios HRV analysis is an advanced tool for studying the variability of heart beat intervals Due to its wide variety of different analysis options and the easy to use interface the software is suitable for researchers and clinicians with varying premises The software is mainly designed for the analysis of normal human HRV but should also be usable e g for animal researchers The developed software includes all the commonly used time and frequency domain variables of HRV The frequency domain variables being calculated for both nonparametric Fourier transform based and parametric autoregressive modeling based spectrum esti mates In addition the software calculates several nonlinear variables such as the Poincar plot recurrence plot analysis detrended fluctuation analysis approximate and sample en tropies and correlation dimension Furthermore the software also includes time varying analysis options For example the time varying changes in all the time and frequency domain variables are supported The variation of the frequency domain variables is obtai
11. spectrum estimate on the other hand does not depend on the sample length but can be decreased by averaging several shorter samples which leads to decreased frequency resolution How to select the AR model order for the AR spectrum estimate The AR method is a parametric method which can be used also for spectrum estimation In this method the RR interval series is modeled with an autore gressive model of specific order The roots of the AR polynomial which are actually complex conjugate pairs correspond to the spectral peaks in the AR spectrum Thus the order of the AR model has to be at least twice the number of expected spectral peaks in the spectrum In practice the order is however always higher than this minimum and the few extra roots do not disturb the spectrum estimate Even though an exaggerated model order can induce spuri ous peaks into the AR spectrum estimate and distort the results The AR model order naturally depends on the interpolation rate of the RR interval series but in many cases the default order of 16 is reasonable Why do I get a warning Negative component power in AR spectrum esti mate When factorization is used see the Analysis options for the AR spectrum es timate the AR spectrum is divided into separate components VLF LF and HF components The power of each component is estimated using a method presented in 19 This method works for well separated AR roots but for roots close to
12. 10 00 00 11 40 Color map Jet he Time varying spectrum estimation method Kalman smoother ba Adaptation coeff 0 05 57 Figure 5 4 Sample run 2 time varying results for the orthostatic test recording a The time varying trend of the pNN50 parameter b the time varying spectrum estimate using the spectrogram method and c the time varying spectrum estimate using the Kalman smoother method Kubios HRV Analysis version 2 0 beta Biosignal Analysis and Medical Imaging Group Department of Physics University of Kuopio FINLAND 5 2 Sample run 2 Time varying analysis 58 Neuroscan_data cnt 20 11 03 15 24 59 HRV Analysis Time Varying Results Page 2 2 Results for a single sample 00 12 18 RR Interval Time Series Detrending method Smoothn priors A 500 00 01 40 00 03 20 00 05 00 00 06 40 00 08 20 00 10 00 00 11 40 Selected Detrended RR Series T 0 2 J 00 00 00 00 01 40 00 03 20 00 05 00 00 06 40 00 08 20 00 10 00 00 11 40 Time h min s Time Varying Results Window width 60 s and window shift 5 s Mean RR STD and equivalent HR scal 50 60 70 RMSSD bold line pNN50 medium line Mean RR s HR beats min 60 4 40 4 20 PNN50 T E a Qa Q c4 Time varying spectrum spectrogram Frequency Hz VLF bold line LF medium line and HF thin line peak frequency Peak frequ
13. 1219 7255 934 5564 STD RR ms E 45 8613 gt 42 0675 3 Mean HR 1 min 49 3307 64 5556 STD HR 1 min 2 4044 3 5681 RMSSD ms H 58 8540 H 27 9642 NN50 count 119 20 pNN50 3 42 8058 3 5 6818 E SDANN ms SDNN index ms 3 3 Geometric parameters RR tri index A 0 084577 A 0 088074 A TINN ms H 235 0000 230 0000 A Frequency Domain Results FFT spectrum AR spectrum FFT spectrum AR spectrum Peak frequencies z VLF Hz 0 039063 0 039063 0 035156 0 039063 LF Hz 0 058594 0 070313 0 089844 0 078125 HF Hz 0 304688 0 285156 0 152344 0 152344 Absolute powers A VLF ms 2 154 6365 286 2022 526 1911 196 2865 LF ms 2 3 556 0111 764 4204 1556 0408 1228 9957 HF ms 2 3 818 5592 909 6497 384 9678 265 8349 Relative powers VLF 10 1122 14 6001 21 3275 11 6069 LF 5 36 3594 38 9956 63 0691 72 6736 HF 53 5283 46 4043 15 6034 15 7195 Normalized powers LF n u 40 4498 45 6624 80 1666 82 2164 HF n u 59 5502 54 3376 19 8334 17 7836 LF HF ratio 0 6793 0 8403 4 0420 4 6232 Nonlinear Results 4 Poincare plot SD1 ms 41 982675 20 192471 SD2 ms E 80 323877 96 721774 Recurrence plot analysis RPA Mean line length beats 8 0360 7 14 4666 Max line length beats 43 H 343 Recurrence rate REC 5 24 7243 39 0836 Determinism DET 3 96
14. 3 3 3 Dample entropy s s aces Pa aoa a G ea eee a ae ae ee 3 3 4 Detrended fluctuation analysis 2 3 3 5 Correlation dimension gt s a aaa s eoe aa e 3 3 6 Recurrence plot analysis o s soa aa cee a a ee EA 3 4 Time varying methods 3 5 Summary of HRV parameters s s ss s ia dcia s dakota d 4 Software description 4L lnputdatadormats 4 sa saaa a a dbi id 4 2 Thewuser interface s s oie an ee d eee ea eee es 4 2 1 RR interval series options a s a damad deuaa aaa 4 22 Data DEOWSE sasi iae ee ee ee SE eh a N a AD 3 jiAMAIYSISOPhIONS a daa wi BEB sek eee ee RE oe Ai ADA Results view e sore ee 64 bb dae eGR G bee ee ee aaaa anes 4 2 5 Menus and toolbar buttons sacc s s sdp eumeus a a a 0004 da Saving the results era a ARE ee ee ee ee ee BY 4 3 1 ASCI Ef o o ee 13 13 13 14 14 16 17 18 19 20 22 22 23 23 24 25 25 26 27 28 30 30 4 3 2 Report sheet 4 3 3 MATLAB MAT file 4 4 Setting up the preferences 5 Sample runs 5 1 Sample run 1 General analysis 5 2 Sample run 2 Time varying analysis A Frequently asked questions B Troubleshooting References 59 62 63 Chapter 1 Overview The Kubios heart rate variability HRV analysis software is developed by the Biosignal Analysis and Medical Imaging Group at the Department of Physics University of Kuopio Kuopio Finland
15. Il Nonlinear Time varying 0 5 0 4 0 3 0 2 Ta gt En E a 3 gt 2 iva 0 1 0 00 00 00 00 03 20 00 05 00 00 06 40 00 08 20 00 10 00 Time h min s Color mal e Options a Select variable Time varying computations Time varying spectrum estimation method Frequency domain Window width s 60 Spectrogram Jens E Time varying spectrum Grid interval s 5 Adaptation coeff 0 1 Figure 4 10 The results view segment of the user interface time varying results view se lected hand side of the axis When the time varying spectrum is selected for view a color bar indicating the power values is also shown on the right The color map of the spectrum can be changed with the Color map button The adjustable options for the time varying analysis include the window width and grid interval for the moving window which is used to calculate the results In addition the time varying spectrum can be estimated using either the spectrogram method or the Kalman smoother method The latter one is a parametric approach where the time varying AR parameters are solved with the Kalman smoother algorithm The adaptation speed of the algorithm can be adjusted manually by changing the Adaptation coeff value For bigger values of this coefficient the algorithm adapts faster to local changes in the signal with the expense of increased variance The default value for the adaptation coefficient is 0 1 4 2 5 Menus an
16. L H Carney and J P Saul Time and frequency domain methods for heart rate variability analysis a methodological comparison Psychophys tol 32 492 504 1995 F Lombardi T H Makikallio R J Myerburg and H Huikuri Sudden cardiac death role of heart rate variability to identify patients at risk Cardiovasc Res 50 210 217 2001 A Malliani M Pagani F Lombardi and S Cerutti Cardiovascular neural regulation explored in the frequency domain 84 2 482 492 August 1991 J Malmivuo and R Plonsey Bioelectromagnetism Principles and Applications of Bioelectric and Biomagnetic Fields Oxford University Press Web Edition 1995 S L Marple Digital Spectral Analysis with Applications Prentice Hall 1987 Department of Physics ee pena University of Kuopio FINLAND dy Kubios HAV Analysis Biosignal Analysis and Medical Imaging Group References 65 28 29 30 31 32 33 34 35 36 37 38 39 40 pS Al 42 J Mateo and P Laguna Improved heart rate variability signal analysis from the beat occurrence times according to the IPFM model IEEE Trans Biomed Eng 47 8 985 996 August 2000 J Mateo and P Laguna Analysis of heart rate variability in the presence of ectopic beats using the heart timing signal IEEE Trans Biomed Eng 50 3 334 343 March 2003 M Merri D C Farden J G Mottley and E L Titlebaum Sampling frequency of the electrocardiogram for spectra
17. Users Desktop dy Ku bios HAV Analysis Biosignal Analysis and Medical Imaging Group i Department of Physics 2 0 bet ee University of Kuopio FINLAND 1 4 Software home page 14 Delete the possible Start menu entries from C Documents and Settings All Users Start Menu Programs Remove the Kubios registry entry from HKEY_LOCAL_MACHINE SOFTWARE Kubios HRV PLEASE NOTE THAT MODIFYING THE WINDOWS REGISTRY CAN CAUSE SERIOUS PROBLEMS THAT MAY REQUIRE YOU TO REINSTALL YOUR OP ERATING SYSTEM USE THE INFORMATION PROVIDED AT YOUR OWN RISK Manual removal of the Kubios HRV Analysis entry from the Windows Add or Re move Programs list requires modifying registry A thorough instructions on how to manually remove programs from the Add or Remove Programs list is available on the Microsoft support web site at http support microsoft com kbid 314481 MATLAB Component Runtime The MATLAB Component Runtime can be completely uninstalled manually by deleting the following files folders and registry and system path entries Delete the install folder by default C Program Files Mathworks MATLAB Component Runtime and all the subfolders and files in it Remove the MATLAB Component Runtime entry lt MCR install dir gt v72 runtime win32 from the system path Manual removal of the MATLAB Component Runtime entry from the Win dows Add or Remove Programs list requires modifying registry A thor ough instruc
18. are saved in HRV2 folder located in the user specific Application Data folder The preference files are found from the folder C Documents and Settings lt username gt Application Data HRV2 where lt username gt is the name of your user profile The folder will include three files hrv_pref dat user_pref dat and HRVprefs mat The hrv_pref dat file includes all the preferences for the analysis options user_pref dat includes the user information prefer ences and HRVprefs mat all the preferences related to the usability of the software These files are created when the software is started for the first time and they will be updated whenever the preference values are edited The original settings of the preferences can thus be restored by deleting these files 1Note that the Application Data folder is hidden by default and is not visible in the Windows File Explorer if the Show hidden files and folders is not selected from the Folder Options section of the File Explorer dy Kubios HAV Analysis Biosignal Analysis and Medical Imaging Group i Department of Physics 2 0 bet ee University of Kuopio FINLAND Chapter 5 Sample runs In this chapter we present as an example two sample runs of the software Both of the sample runs are made for the sample data file distributed with this software The sample data is measured from a healthy young male during an orthostatic test The change in the posture is known to be reflected i
19. case of AR spectrum on the other hand if factorization is enabled distinct spectral components emerge for each frequency band with a proper selection of the model order and the absolute power values are obtained directly as the powers of these components If factorization is disabled the AR spectrum powers are calculated as for the FFT spectrum The band powers in relative and normalized units are obtained from the absolute values as described in Table 3 1 3 3 Nonlinear methods Considering the complex control systems of the heart it is reasonable to assume that non linear mechanisms are involved in the genesis of HRV The nonlinear properties of HRV have been analyzed using measures such as Poincar plot 5 6 approximate and sample entropy 40 12 detrended fluctuation analysis 36 37 correlation dimension 15 17 and Version 0 beta Department of Physics dy Rubies HAV Analysis Biosignal Analysis and Medical Imaging Group University of Kuopio FINLAND 3 3 Nonlinear methods 24 1000 950 f 900 F 850 f RR ms 800 f 750 F 700 F 650 650 700 750 800 850 900 950 1000 RR ms Figure 3 1 Poincar plot analysis with the ellipse fitting procedure SD1 and SD2 are the standard deviations in the directions x and x2 where xa is the line of identity for which RR RRj41 recurrence plots 47 46 49 During the last years the number of studies utilizing such methods have increased su
20. estimates for the negative natural logarithm of the conditional probability that a data of length N having repeated itself within a tolerance r for m points will also repeat itself for m 1 points SampEn was designed to reduce the bias of ApEn and has a closer agreement with the theory for data with known probabilistic content 20 3 3 4 Detrended fluctuation analysis Detrended fluctuation analysis DFA measures the correlation within the signal The corre lation is extracted for different time scales as follows 36 First the RR interval time series is integrated k RR RR k 1 N 3 15 j l where RR is the average RR interval Next the integrated series is divided into segments of equal length n Within each segment a least squares line is fitted into the data Let y k denote these regression lines Next the integrated series y k is detrended by subtracting the local trend within each segment and the root mean square fluctuation of this integrated and detrended time series is calculated by 3 16 This computation is repeated over different segment lengths to yield the index F n as a function of segment length n Typically F n increases with segment length A linear rela tionship on a double log graph indicates presence of fractal scaling and the fluctuations can be characterized by scaling exponent a the slope of the regression line relating log F n to logn Different values of a indicate the following a 1 5
21. in the calculation of approximate and sample entropies form length m vectors Uj uj RR RRj41 RRj4 m 1 j 1 2 N m 4 1 3 17 and calculate the number of vectors uz for which d uj ug lt r that is 7 nbr of ug d uj uk lt r Wont Vk 3 18 where the distance function d u us is now defined as d uj ur gt uy 1 ug 1 3 19 l 1 Next an average of the term C7 r is taken 1 N m gt 1 Version 20 beta Department of Physics Kibios HRW Analysis Biosignal Analysis and Medical Imaging Group University of Kuopio FINLAND 3 3 Nonlinear methods 28 Ot or 0 5 1t e O D gt 1 51 2 2 ol 25t 3 1 4 1 2 1 0 8 0 6 0 4 log r Figure 3 3 Approximation of the correlation dimension D from the log r log C r plot which is the so called correlation integral The correlation dimension Da is defined as the limit value D m lim lim bg ON 3 21 r 0 N logr In practice this limit value is approximated by the slope of the regression curve logr logC r 17 The slope is calculated from the linear part of the log log plot see Fig 3 3 The slope of the regression curves tend to saturate on the finite value of D when m is increased In the software a value of m 10 was selected for the embedding 3 3 6 Recurrence plot analysis Yet another approach included in the software for analyzing the complexity of the time series is the so c
22. occurrence times That is the n th RR interval is obtained as the difference between the R wave occurrence times RR tn tn 1 In some context normal to normal NN may also be used when referring to these intervals indicating strictly intervals between successive QRS complexes resulting from SA node depolarization 44 In practice the NN and RR intervals appear to be the same and thus the term RR is preferred here The time series constructed from all available RR intervals is clearly not equidistantly sampled but has to be presented as a function of time i e as values tn RRn This fact has to be taken into account before frequency domain analysis In general three different approaches have been used to get around this issue 44 The simplest approach that have been adopted in e g 2 is to assume equidistant sampling and calculate the spectrum directly from the RR interval tachogram RR intervals as a function of beat number see the left panel of Fig 2 3 This assumption can however cause distortion into the spectrum Rubies HRY Andiysis Biosignal Analysis and Medical Imaging Group f Department of Physics version 2 0 beta University of Kuopio FINLAND 2 3 Preprocessing of HRV time series 19 V EG G t Deriveo PU R WAVE PROCESSES ta t INTERVAL TACHOGRAM INTERVAL FUNCTION WITH 2 POSSIBLE WAYS OF INTERPOLATION Figure 2 3 Derivation of two HRV signals from ECG redrawn from 41 the interva
23. of Kuopio FINLAND Chapter 4 Software description 4 1 Input data formats The Kubios HRV Analysis supports the following binary data formats Neuroscan continuous CNT files Compumedics Limited Biopac AcqKnowledge files Biopac Systems Inc and European data format EDF files When any of these data format files are read to the software Kubios HRV automatically tries to determine the ECG channel from the channel labels Tf the ECG channel cannot be determined or more than one channels are identified as ECG channels the software prompts the user to select the appropriate channel Due to internal design restrictions of Kubios HRV the channel labels should only contain alphabets numbers and underscores If the channel labels contain other characters such as spaces or plus signs etc these characters are changed to underscores Furthermore the channel label should start with an alphabet If this is not the case Ch is added to the beginning of the channel label In addition to the three binary formats support for ASCII text files is also provided The input ASCII file can include either RR interval values or ECG data in one or two column format That is The RR interval values can be given as Type 1 Type 2 0 759 0 759 0 759 0 690 1 449 0 690 0 702 2 151 0 702 0 712 2 863 0 712 0 773 3 636 0 773 The RR interval values above are given in seconds but millisecond values can also be given Correspondingly the ASCII ECG data
24. regularized least squares solution arg min z HO A Da H6 2 3 where A is the regularization parameter and Dg indicates the discrete approximation of the d th derivative operator This is clearly a modification of the ordinary least squares solution to the direction in which the side norm D H0 gets smaller In this way prior information about the predicted trend H0 can be implemented to the estimation The solution of 2 3 can be written in the form 6 HTH X HTDI DH HTz 2 4 and the estimate for the trend which is to be removed as Strend HOy 2 5 The selection of the observation matrix H can be implemented according to some known properties of the data z For example a generic set of Gaussian shaped functions or sigmoids can be used Here however the trivial choice of identity matrix H I RN is used In ap Kubios HAV Analysis Biosignal Analysis and Medical Imaging Group Department of Physics 2 0 bet eee University of Kuopio FINLAND 2 3 Preprocessing of HRV time series 21 Magnitude Magnitude 0 0 1 0 2 0 3 0 4 0 5 Relative frequency b Figure 2 4 a Time varying frequency response of L N 1 50 and A 10 Only the first half of the frequency response is presented since the other half is identical b Frequency responses obtained from the middle row of cf bold lines for A 1 2 4 10 20 100 and 500 The corresponding cut off frequ
25. should be as 32 4 1 Input data formats 33 ASCII File Import ASCII Import Options Number of header lines 4 Column separator Tab Space Data specifications ECG Data column Preview of Data File ECG Measurement 4 10 2005 University of Kuopio Department of Applied Physics FINLAND s EKG mV 000 DOZ DOO0OO0FPFNNNOONMNM 0000000000000 00o0o0oO0X Figure 4 1 The interface for importing customized ASCII data files into the software Type 1 Type 2 0 173 0 0 173 0 119 0 002 0 119 0 025 0 004 0 025 0 091 0 006 0 091 0 218 0 008 0 218 where the first column on the second format type is the time scale in seconds for the ECG data The sampling rate of this example file is thus 500 Hz If ECG data is given according to the first type user is requested to enter the sampling rate manually In addition to above ASCII text files a custom ASCII file option is also provided Using this option you can import ASCIT files including header lines and or several data columns Once you have selected an input file an interface for importing the file into Kubios is opened This interface is shown in Fig 4 1 Through this interface you can specify the following Kibios HAV Analysis Biosignal Analysis and Medical Imaging Group P Department of Physics version 2 0 beta University of Kuopio FINLAND 4 2 The user interface 34 required details corresponding to your data file e Number of he
26. sinus arrhythmia RSA which is considered to range from 0 15 to 0 4 Hz 3 In addition to the physiological influence of breathing on HRV this high frequency HF component is generally believed to be of parasympathetic origin Another widely studied component of HRV is the low frequency LF component usually ranging from 0 04 to 0 15 Hz including the component referred to as the 10 second rhythm or the Mayer wave 3 The rhythms within the LF band have been thought to be of both sympathetic and parasympathetic origin 3 even though some researchers have suggested them to be mainly of sympathetic origin 25 The fluctuations below 0 04 Hz on the other hand have not been studied as much as the higher frequencies These frequencies are commonly divided into very low frequency VLF 0 003 0 04 Hz and ultra low frequency ULF 0 0 003 Hz bands but in case of short term recordings the ULF band is generally omitted 44 These lowest frequency rhythms are characteristic for HRV signals and have been related to e g humoral factors such as the thermoregulatory processes and renin angiotensin system 3 Even though HRV has been studied extensively during the last decades within which numerous research articles have been published the practical use of HRV have reached general consensus only in two clinical applications 44 That is it can be used as a predictor of risk after myocardial infarction 24 18 and as an early warning sign of diabetic neu
27. the starting of Kubios HRV dy Kibios HAV Analysis Biosignal Analysis and Medical Imaging Group f Department of Physics 2 0 bet ee University of Kuopio FINLAND 1 3 Uninstalling the Kubios HRV Analysis 13 Analysis also starts the MATLAB Component Runtime and may take some time especially with older computers 1 3 Uninstalling the Kubios HRV Analysis The Kubios HRV Analysis can be uninstalled either automatically using the uninstaller or manually if the uninstaller fails for some reason Both methods for uninstallation are described in the following 1 3 1 Automated uninstall The preferred and the most straightforward way of uninstalling the Kubios HRV Anal ysis is to use the automated uninstaller The uninstaller can be launched by selecting Uninstall Kubios HRV Analysis from the software s Start menu folder the default Start menu folder is Kubios HRV Analysis The software can also be uninstalled from the Add or Remove Programs under the Windows Control Panel These are the two rec ommended ways of uninstalling the Kubios HRV Analysis and either one of them should always be used for uninstallation If however for some reason the uninstaller fails a man ual uninstallation may be necessary Additionally note that the Kubios HRV Analysis uninstaller does not uninstall the MATLAB Component Runtime Furthermore the unin staller does not remove your preferences settings These have to be deleted manually from C
28. 0 02 08 00 02 09 Range Sample 1 Range s 10 350 Length s 340 Remove trend components Method Smoothn priors ha Lambda 500 Analysis Options Frequency bands VLF Hz 0 0 04 LF Hz 0 04 015 HF Hz 0 15 04 00 00 00 00 01 40 00 03 20 00 05 00 Sample Analysis Type single samples y 4 TENES Time Domain Frequency Domain Common limit H 04 1 00 06 40 Time h min s 1 TE Range 737 L_ 1 00 10 00 2 Nonlinear FFT spectrum estimation results Interpolation of RR series Interpolation rate Hz 4 Spectrum estimation FFT spectrum Window width s Window overlap AR spectrum AR model order Use factorization Apply Changes Y Automatic 0 1 Peak He 02 0 3 0 4 Frequency H2 Power PSD s7 H2 Frequency Band Time Varying E AR spectrum estimation results Sample 1 02 03 0 4 Frequency Hz Power Power 0 0391 0 0469 0 2813 8 0 40 49 0 09 14 6 39 0 46 4 0 8 Figure 5 1 Sample run 1 the lowest frequency trend components from the RR series These trend components affect on the time and frequency domain variables and thus by removing the trend from the data enables these variables to better describe the LF and HF variability which we are interested of We select to remove the trend with the smoothness priors based method Once
29. 5712 99 3427 Shannon entropy H 2 8420 A 3 4933 Detrended fluctuation analysis DFA alpha 1 1 0698 1 3655 alpha 2 0 9310 0 8909 Others H Approximate entropy ApEn 1 0406 1 0194 3 Sample entropy SampEn 1 7331 1 1493 Correlation dimension D2 4 0176 A 3 2946 A Biosignal Analysis and Medical Imaging Group Department of Physics University of Kuopio FINLAND Kubios HRV Analysis version 2 0 beta 5 1 Sample run 1 General analysis Neuroscan_data cnt 20 11 03 15 24 59 Page 1 4 HRV Analysis General Results RR Interval Time Series Results for single samples sample 1 2 00 00 10 00 05 50 T T Detrending method Smoothn priors A 500 Cerri tena 00 06 40 00 01 40 Selected Detrended RR Series T fig 0 2 00 00 00 00 03 20 00 05 00 00 08 20 00 10 00 00 11 40 ny i UN ath ol i 1 i i L i 00 00 50 00 01 40 00 02 30 00 03 20 00 04 10 00 05 00 Time h min s Time Domain Results Distributions Variable Mean RR STD RR SDNN Mean HR STD HR RMSSD NN50 pNN50 Units 6 s 1 min 1 min ms count RR triangular index TINN ms Frequency Domain Results FFT spectrum Welch s periodogram 128 s window with 50 overlap JO a ae m 0 03 e 0 02 1 2 RR s 50 HR beats min 55 AR Spectrum 0 04 AR model order 16 not f
30. Documents and Settings lt username gt Application Data HRV2 If you want to remove the Kubios HRV Analysis permanently from your computer you should also uninstall the MATLAB Component Runtime The MATLAB Component Runtime can be uninstalled from the Add or Remove Programs under the Windows Con trol Panel Note however that if you are uninstalling Kubios HRV Analysis just to update it with a newer version you do not necessarily need to uninstall the MCR detailed instructions of such situations will be given at the Kubios HRV Analysis web site 1 3 2 Manual uninstall Kubios HRV Analysis The manual uninstallation should be conducted only if the automated uninstallation fails The Kubios HRV Analysis can be completely uninstalled manually by deleting the following files folders and registry entries e Delete the install folder by default C Program Files Kubios HRV Analysis and all the subfolders and files in it e Delete the HRV2 folder if exists from C Documents and Settings lt username gt Application Data NOTE The Application Data folder is a hidden folder and you may need to en able the Show hidden folders option in the Folder Options found under Control Panel e Delete the Kubios HRV Analysis quick launch icons if exist from C Documents and Settings lt username gt Application Data Microsoft Internet Explorer Quick Launch and desktop icons if exist from C Documents and Settings All
31. HRV frequency bands and Update analysis results options which have already been described in Sections 4 2 1 4 2 3 and 4 2 4 The Advanced settings category shown in Fig 4 15 includes QRS detection Spec trum estimation and Time varying spectrum estimation options In the QRS detection options you can set up the prior guess for the average RR interval By default this prior guess is estimated automatically This may not however always work in which case the prior guess for the RR interval value should be fixed to the supposed value The spectrum estima tion options include one additional option compared to those described in Section 4 2 3 i e points in frequency domain option The point in frequency domain is given as points Hz and corresponds by default to the window width of the FFT spectrum If spectrum in terpolation is desired the points in frequency domain can be increased The time varying spectrum estimation options were described in the end of Section 4 2 4 The Report settings category shown in Fig 4 16 includes two options First of all the contents of the report sheet can be selected by checking the General and or Time varying results options If either one of these is unchecked only one report sheet will be printed Secondly the paper size of the report sheet can be changed between A4 210x297 mm and Letter 8 5x11 inch size The default paper size is A4 All modifications for the preferences are saved by pressing either the A
32. Kubios HRV Analysis version 2 0 beta USER S GUIDE June 13 2006 Mika P Tarvainen Ph D Mika Tarvainen uku fi and Juha Pekka Niskanen Biosignal Analysis and Medical Imaging Group BSAMIG http bsamig uku fi Department of Physics University of Kuopio Kuopio FINLAND et Tilt PRORQ Sc A MATLAB Copyright 1984 2005 The MathWorks Inc MATLAB is a registered trademark of The MathWorks Inc Contents 1 Overview LL System requirements s isoo 24 4 bee eee 2 eee eee eG aw eR x 1 2 Installation 4 545 424 4430 2 24 4h eee ed 1 3 Uninstalling the Kubios HRV Analysis 2 2000 1 3 1 Automated uninstall 2 2 2 0 0 0000020022 eee 1 3 2 Manual uninstall o e p i sorea 5 ew ee ew a a a N a S 1 4 Software home Page oos e aeta a y BE a ed a ee eee eo 1 5 Structure of this guide 2 ee ee 2 Heart rate variability 2 1 Heart beat period and QRS detection 00 0 2 2 Derivation of HRV time Series e e e 2 3 Preprocessing of HRV time series o 00000 ee eee 2 3 1 Smoothness priors based detrending approach 3 Analysis methods 3 1 Time domaim methods 044 4 4444 h4 44440 0b4454 4444 3 2 Frequency domain methods 0 00 eee eee eee 3 3 Nonlinear Methods e 225 4 dba eee nee EO eA YEE de Seek Pomeare plot caos ear Dosis pura eee oS SS Se BEX 3 3 2 Approximate entropy ie et a tewas egetei p ee
33. V signals can be divided into technical and physiological artifacts The technical artifacts can include missing or additional QRS complex detections and errors in R wave occurrence times These artifacts may be due to measurement artifacts or the computational algorithm The physiological artifacts on the other hand include ectopic beats and arrhythmic events In order to avoid the interference of such artifacts the ECG recording and the corresponding event series should always be manually checked for artifacts and only artifact free sections should be included in the analysis 44 Alternatively if the ap Kubios HAV Analysis Biosignal Analysis and Medical Imaging Group f Department of Physics 2 0 bet eee University of Kuopio FINLAND 2 3 Preprocessing of HRV time series 20 amount of artifact free data is insufficient proper interpolation methods can be used to reduce these artifacts see e g 21 22 29 Another common feature that can alter the analysis significantly are the slow linear or more complex trends within the analyzed time series Such slow nonstationarities are characteristic for HRV signals and should be considered before the analysis The origins of nonstationarities in HRV are discussed e g in 3 Two kinds of methods have been used to get around the nonstationarity problem In 48 it was suggested that HRV data should be systematically tested for nonstationarities and that only stationary segments should be analy
34. a Department of Physics Kibios HAV Analysis Biosignal Analysis and Medical Imaging Group University of Kuopio FINLAND 1 2 Installation 7 4 Next select the components you want to install only the checked components will be installed The Kubios HRV Analysis will always be installed but you may choose not to install the MATLAB Component Runtime by unchecking the box next to it NOTE MATLAB Component Runtime is required to run Kubios HRV Analysis so uncheck it only if you have MATLAB Component Runtime version 7 2 already installed 16 Setup Kubios HRV Analysis Select Components Which components should be installed Select the components you want to install clear the components you do not want to install Click Next when you are ready to continue tall Kubios HAY Analysi 2 9 ME Install MATLAB Component Runtime uncheck only if MATLAB Component Runtime has been installed earlier 228 2 MB Current selection requires at least 238 4 MB of disk space 5 Then select the Start Menu folder in which the program s shortcuts are created The default Start Menu folder is Kubios HRV Analysis You can select a different Start Menu folder by clicking Browse You can also choose not to create a Start Menu entry i Setup Kubios HRV Analysis Select Start Menu Folder Where should Setup place the program s shortcuts Setup will create the program s shortcuts in the following Start Menu folder To conti
35. actorized m 0 03 e 0 02 a 0 0 2 0 3 Frequency Hz Peak Hz Power Power ms Frequency Band Frequency Band Peak Hz Power ms 0 0391 0 0469 0 2813 150 809 923 1882 0 876 8 0 43 0 49 0 VLF 0 0 04 Hz LF 0 04 0 15 Hz HF 0 15 0 4 Hz Total LF HF VLF 0 0 04 Hz LF 0 04 0 15 Hz HF 0 15 0 4 Hz Total LF HF 0 0391 0 0703 0 2852 286 764 910 1960 0 840 Nonlinear Results Poincare Plot Detrended fluctuations DFA Variable Poincare plot SD1 SD2 Recurrence plot Mean line length Lmean Max line length Lmax Recurrence rate REC Determinism DET Shannon Entropy ShanEn Other Approximate entropy ApEn Sample entropy SampEn Detrended fluctuations DFA 01 Detrended fluctuations DFA a2 Correlation dimension D2 06 08 1 12 1 4 109 n beats 16 1 8 Results are calculated from the non detrendedselected RR series 24 Jan 2005 11 27 18 Mika Tarvainen Department of Applied Physics University of Kuopio Figure 5 2 Sample run 1 results for the Kubios HRV Analysis version 2 0 beta Kubios HRV Analysis version 2 0 beta 1 Department of Applied Physics University of Kuopio Finland lying period of the orthostatic test Biosignal Analysis and Medical Imaging Group Department of Physics University of Kuopio FINLAND 5 1 Sample run 1
36. ader lines e Column separator tab space comma or semicolon e Data type ECG or RR e Data column the ordinal number of data column e Data units uV V or mV for ECG ms or s for RR e Time index column the ordinal number of time indexes e Time units units of time indexes in ms or s e ECG sampling rate in Hz if no time index column defined for ECG Once you have specified the above values for your file press OK to proceed to analysis 4 2 The user interface The developed HRV analysis software is operated with a graphical user interface which consists of only one window This user interface window is shown in Fig 4 2 The user interface is divided into four segments 1 the RR interval series options segment on the top left corner 2 the data browser segment on the top right corner 3 the analysis options segment on the bottom left corner and 4 the results view segment on the bottom right corner Each of these segments are described in Sections 4 2 1 4 2 2 4 2 3 and 4 2 4 respectively 4 2 1 RR interval series options The RR interval series options shown in Fig 4 3 include three functions Artifact correction Samples for analysis and Remove trend components The artifact correction options can be used to correct artifacts from a corrupted RR interval series The user can select between very low low medium strong and very strong correction levels In addition a custom level in seconds can be set The corrections to b
37. alled recurrence plot RP analysis In this approach vectors uj RRj RRj47 RRja m vyr J 1 2 N m 1 r 3 22 where m is the embedding dimension and 7 the embedding lag The vectors uj then represent the RR interval time series as a trajectory in m dimensional space A recurrence plot is a symmetrical N m 1 7 x N m 1 7 matrix of zeros and ones The element in the j th row and k th column of the RP matrix ie RP j k is 1 if the point uj on the trajectory is close to point uz That is 1 d uj uk lt r RP j k l 0 otherwise Gay where d u ux is the Euclidean distance given in 3 19 and r is a fixed threshold The structure of the RP matrix usually shows short line segments of ones parallel to the main diagonal The lengths of these diagonal lines describe the duration of which the two points are close to each other An example RP for HRV time series is presented in Fig 3 4 dy Kubios HAV Analysis Biosignal Analysis and Medical Imaging Group i Department of Physics 2 0 bet TROT 0 Beta University of Kuopio FINLAND 3 3 Nonlinear methods 29 ips LAP Ed 4h A r 4 PA oy Yes Time min Figure 3 4 Recurrence plot matrix for HRV time series black 1 and white 0 Methods for quantifying recurrence plots were proposed in 47 The methods included in this software are introduced below In the software the following selections were made The embedding dimen
38. and Medical Imaging Group University of Kuopio FINLAND 4 2 The user interface 35 Kubios HRY Analysis DER File View Help ra f File Info HO MARKERS File mame Neuroscan_data cnt Rec date 20 11 03 Rec time 15 24 59 Channel label EKG Sampling rate 500 Hz Data length 12 min 17 s RR Interval Series Options Artifact correction Apply aa H Unc 0 02 00 02 01 00 02 02 00 02 03 00 02 04 00 02 05 00 02 06 00 02 07 00 02 08 00 02 09 none Time h min s Range Samples for analysis Number of samples 1 Sample 1 Range s 1 351 Length s 350 Remove trend components Method Smoothn priors ha Lambda 500 fi 1 L 1 00 01 40 00 03 20 00 05 00 00 06 40 00 08 20 00 10 00 00 11 40 Time h min s A Analysis Options Range 737 Frequency bands EAS wun El oo Time Domain Frequency Domain Nonlinear Time varying LF Hz 0 04 015 VIEW RESULTS 2 Z a IS o N HF Hz 0415 04 Interpolation of RR series Interpolation rate Hz Spectrum estimation FFT spectrum Window width s SO DI o2 08 04 7 i o2 o3 04 AR spectrum Frequency Hz Frequency Hz AR model order Frequency Peak Power Power Power Frequency Power Power Use factorization d He 98 Band 36 0 0391 10 3 11 9 _ a _ _ _ __ 0 0586 355 24 0 Apply Changes 0 3047 54 2 64 1 Y Automatic Apply 07 0 4 PSD s7 H2 Figure 4 2 The gra
39. bstantially The downside of these methods is still however the difficulty of physiological interpretation of the results 3 3 1 Poincar plot One commonly used nonlinear method that is simple to interpret is the so called Poincar plot It is a graphical representation of the correlation between successive RR intervals i e plot of RRj as a function of RR as described in Fig 3 1 The shape of the plot is the essential feature A common approach to parameterize the shape is to fit an ellipse to the plot as shown in Fig 3 1 The ellipse is oriented according to the line of identity RR RR 1 5 The standard deviation of the points perpendicular to the line of identity denoted by SD1 describes short term variability which is mainly caused by RSA It can be shown that SD1 is related to the time domain measure SDSD according to 5 1 SD1 3SDSD 3 5 The standard deviation along the line of identity denoted by SD2 on the other hand de scribes long term variability and has been shown to be related to time domain measures SDNN and SDSD by 5 SD2 2SDNN 5SDSD 3 6 The standard Poincar plot can be considered to be of the first order The second order plot would be a three dimensional plot of values RR RRj 1 RRj 2 In addition the lag can be bigger than 1 e g the plot RR RR 2 Version U beta Department of Physics Kibios HAV Analysis Biosignal Analysis and Medical Imaging Group University of Kuo
40. d on the feedback obtained from the users of the previous version of this software Finally in Appendix B workarounds for some commonly encountered technical problems are given Version 0 beta Department of Physics K bios HAV Analysis Biosignal Analysis and Medical Imaging Group University of Kuopio FINLAND Chapter 2 Heart rate variability Heart rate variability HRV describes the variations between consecutive heartbeats The rhythm of the heart is controlled by the sinoatrial SA node which is modulated by both the sympathetic and parasympathetic branches of the autonomic nervous system Sympa thetic activity tends to increase heart rate HR and its response is slow few seconds 3 Parasympathetic activity on the other hand tends to decrease heart rate HR and mediates faster 0 2 0 6 seconds 3 In addition to central control there are some feedback mechanisms that can provide quick reflexes One such mechanism is the arterial baroreflex This reflex is based on baroreceptors which are located on the walls of some large vessels and can sense the stretching of vessel walls caused by pressure increase Both sympathetic and parasympathetic activity are influenced by baroreceptor stimulation trough a specific baroreflex arc Fig 2 1 The continuous modulation of the sympathetic and parasympathetic innervations results in variations in heart rate The most conspicuous periodic component of HRV is the so called respiratory
41. d toolbar buttons The user menus and toolbar buttons are located on the upper left hand corner of the user interface There are all together three user menus and seven toolbar buttons The toolbar button icons and their actions are given below Version 0 beta Department of Physics Kibios HAV Analysis Biosignal Analysis and Medical Imaging Group University of Kuopio FINLAND 4 3 Saving the results 42 Open new data file button is for opening a new data file for anal ysis If the results of the current analysis have not been saved user is prompted to do so Save results button is for saving the analysis results The results can be saved in ASCII PDF and MATLAB MAT file format see Section 4 3 for details Report sheet button opens one or several report sheet windows which include all the analysis results see Section 4 3 2 for details Edit preferences button opens a preferences window in which you can e g change the default values for analysis options see Section 4 4 for details About HRV analysis software button opens the about dialog of the software which includes the version number and contact informa tion Also the Kubios HRV Analysis End User License Agreement can be viewed in the about dialog Open Kubios HRV User s Guide button opens the Kubios HRV User s Guide this document PDF file using the default PDF viewer of the system Close file button closes the current data file If the results of the
42. e made on the RR series are displayed on the RR interval axis To make the corrections press the Apply button A piecewise cubic spline interpolation method is used in the corrections You can reverse the correction by pressing the Undo button or by selecting none as the correction level It should be noted that artifact correction generates missing or corrupted values into the RR series by interpolation and can thus cause distortion into the analysis results If ECG is measured the corrections should always be done by editing the R peak marks in the ECG data as described in Section 4 2 2 In the Samples for analysis options the part s of the RR interval series to be analyzed can be selected by editing the Number of samples Range End and Length values If more than one sample is selected the analysis can be done either for the single samples separately or by merging the samples into one before analysis This selection is visible under the RR series axis when multiple samples are selected The range of the samples can also be changed by moving resizing the patch over the RR series as described in Section 4 2 2 Sometimes the RR interval time series includes a disturbing low frequency baseline trend component Detrending options can be used to remove this kind of trend components Detrending options include removal of the first second or third order linear trend or the Version 00 beta Department of Physics bios HAV Analysis Biosignal Analysis
43. each other power estimates can yield even negative values which are obviously highly erroneous and a warning message is displayed Because of this and some other drawbacks the AR spectrum factorization should be used judi ciously Instead by disabling the factorization more robust and in that sense more reliable results are obtained Version 0 beta Department of Physics Kibios HRW Analysis Biosignal Analysis and Medical Imaging Group University of Kuopio FINLAND 61 Why do the power values of Kubios HRV Analysis 2 0 spectrum estimates differ from those of the version 1 1 The differences in power values are due to two changes First of all in version 2 0 the mean of the RR interval data is removed before spectrum estimation This decreases significantly the VLF power value of the FFT based spectrum estimate Secondly the scaling of the spectrum estimates is changed as follows In version 1 1 the spectrum estimates were scaled such that the total power from 0 f Hz was equal to the variance of the RR data where f is the sampling frequency of the RR data i e the interpolation rate In version 2 0 on the other hand the total power from 0 f 2 Hz is equal to the variance How to select the window width in time varying analysis Here the selection of the window width is a compromise between the accuracy of the calculated variable and its time resolution For time domain variables the longer the window the more RR inte
44. ead the following important information before continuing Please read the following License Agreement You must accept the terms and conditions of this Agreement before continuing with the installation END USER LICENSE AGREEMENT This End User License Agreement the Agreement set forth the terms and conditions according to which Kuopio University grants to the end user the Licensee the right to use the Kubios HRV Analysis version 2 0 beta software Please read these terms and conditions carefully before installing copying downloading accessing or otherwise using the software By installing copying downloading accessing or otherwise using the Kubios HRV Of O I do not accept the Agreement 3 Next select the destination folder in which the software should be installed To se lect the default destination C Program Files Kubios HRV Analysis click the Next button If you want to select a different folder click the Browse button If the selected folder exists setup will ask if you a confirmation before installing in this folder al To Setup Kubios HRV Analysis Select Destination Folder Where should Kubios HAY Analysis be installed O Setup will install Kubios HRW Analysis into the following folder To continue click Next If you would like to select a different folder click Browse C Program Files Kubios HAY Analysis Browse Atleast 10 3 MB of free disk space is required vercion oO bet
45. encies are 0 213 0 145 0 101 0 063 0 045 0 021 and 0 010 times the sampling frequency this case the regularization part of 2 3 can be understood to draw the solution towards the null space of the regularization matrix Dg The null space of the second order difference matrix contains all first order curves and thus Da is a good choice for estimating the aperiodic trend of RR series With these specific choices the detrended nearly stationary RR series can be written as Betas z HO I I DE Da 7 z 2 6 In order to demonstrate the properties of the proposed detrending method its frequency response is considered Equation 2 5 can be written as Zsta z where L I I D D3 7 corresponds to a time varying finite impulse response highpass filter The frequency response of for each discrete time point obtained as a Fourier transform of its rows is presented in Fig 2 4 a It can be seen that the filter is mostly constant but the beginning and end of the signal are handled differently The filtering effect is attenuated for the first and last elements of z and thus the distortion of end points of data is avoided The effect of the smoothing parameter A on the frequency response of the filter is presented in Fig 2 4 b The cutoff frequency of the filter decreases when A is increased Besides the A parameter the frequency response naturally depends on the sampling rate of signal z Version 0 beta De
46. ency Hz Band power LF HF ratio 0 6 0 4 0 2 E I i 1 I 1 1 I 00 00 00 00 01 40 00 03 20 00 05 00 00 06 40 00 08 20 00 10 00 00 11 40 17 Mar 2005 10 00 51 Mika Tarvainen Kubios HRV Analysis version 2 0 beta 1 Department of Applied Physics Department of Applied Physics University of Kuopio University of Kuopio Finland Figure 5 5 Sample run 2 time varying results for the orthostatic test measurement Ku bios HRV Analysis Biosignal Analysis and Medical Imaging Group Department of Physics open 20 beta University of Kuopio FINLAND Appendix A Frequently asked questions Based on the feedback and user experiences obtained from the users of the previous version of the HRV analysis software 32 we have collected here a bunch of frequently asked ques tions An answer to each question is given below the question Some of the questions are concerned with the optimal value of some analysis parameter Often these parameter values are however more or less case specific for example the length of the selected HRV data may change the preferred settings for FFT spectrum calculation Thus some of the answers might be more or less vague but hopefully still helpful What is the best way to treat artifacts in RR interval series When doing HRV analysis one should always make sure that the measured HRV series does not include any artifacts such as ectopic beats or missed extra QRS detections Thu
47. ent of Physics 2 0 bet eee University of Kuopio FINLAND 4 2 The user interface Report Page 1 File Edit BSN ALA PIO ECG Sj gnal Weurcs can daaa 201100 poner 0007 00 0007 15 mr 0007 345 Time Amins 174 2005 51624 Mia Tauren Kublos HRY Aralysis version ZO be la 1 Deparment or Applied Physics Deparimenior Applied Physics Universi ty of Kuopio Unduersi ty of Kuopio Firiard Figure 4 5 The printout of the ECG signal generated by the software version 2 0 beta 38 Department of Physics dy bios HRY Andiysis Biosignal Analysis and Medical Imaging Group University of Kuopio FINLAND 4 2 The user interface 39 Analysis Options Frequency bands LF Hz D 0 04 LF Hz 015 HF Hz 015 F 04 Interpolation of RR series Interpolation rate Hz 4 Spectrum estimation FFT spectrum Window width s Window overlap 96 AR spectrum AR model order Use factorization VIEW RESULTS Time Domain _ Frequency Domain Nonlinear ll Time Varying Time Domain Results variable i Value Distributions Statistical Measures Mean RR 1219 3 STD RR SDNN 51 9 Mean HR i 49 37 STD HR i 271 RMSSD 66 7 NNSO 126 pNNSO 44 2 Geometric Measures HRY triangular index 0 080 TINN ms 345 0 50 55 Calculated from the non detrended selected RR series HR beats min Figure 4 7 The results view segment of the user interface time domain results view se
48. eriods is discussed and the derivation of HRV time series is described The rest of the chapter is focused on the preprocessing of HRV data and gives a detailed description of the smoothness priors based detrending approach In Chapter 3 the analysis methods included in the software are described The de scriptions of the methods are divided into time domain frequency domain nonlinear and time varying categories and a summary of the methods is given at the end of the chapter For most of the methods exact formulas for the different variables are given and possible parameter selections are pointed out In Chapter 4 the description of the features and usage of the software is given First the input data formats supported by the software are described and then the user interface through which the software is operated is described Then different options for saving the analysis results are described and finally instructions on how to set up the preference values for the analysis options are given In Chapter 5 two sample runs of the software are presented The first sample run describes how to analyze the lying and standing periods of the orthostatic test measurement distributed along this software separately as stationary segments The second sample run on the other hand describes the time varying analysis procedure of the same measurement In Appendix A some frequently asked questions with answers are given The questions are selected base
49. esults can be written in an ASCII text file for further inspection the report sheets generated from the results can be saved in a PDF file or the results can be saved in a MATLAB MAT file 4 3 1 ASCII file When the ASCII text file is selected for the saving type the numeric results of the analysis will be written in an ASCII text file The resulting text file includes the following information in the enumerated order Version 0 beta Department of Physics K bios HAV Analysis Biosignal Analysis and Medical Imaging Group University of Kuopio FINLAND 4 3 Saving the results 43 Software user and data file informations Used analysis parameters Samples selected for analysis Time domain results Frequency domain results Nonlinear results Time varying results Ce lee St A RR interval data and spectrum estimates The columns of the file are separated with semicolons so that the results could easily be imported to e g spreadsheet programs such as the Microsoft Excel for further inspection 4 3 2 Report sheet The software generates two printable report sheets which present all the analysis results The first report sheet shown in Fig 4 11 includes all the time domain frequency domain and nonlinear analysis results and the second report sheet shown in Fig 4 12 includes all the time varying analysis results The RR interval data and the sample selected for analysis are presented on the two axes on top of both sheets a
50. hand side of the ECG axis The changes made in R peak markers will be automatically updated to RR interval series The selected sample s yellow patches in the RR axis can be modified with mouse as follows Each sample can be moved by grabbing it from the middle with the left mouse button and resized by grabbing it from the left or right edge When more than one samples are selected a sample can be removed by right clicking it with the mouse Version iO beta Department of Physics bios HAV Analysis Biosignal Analysis and Medical Imaging Group University of Kuopio FINLAND 4 2 The user interface 37 In addition the data browser segment includes buttons for displaying a printout of the ECG recording on the left hand side of the ECG axis moving the ECG axis view to the beginning of a selected sample on the left hand side of the ECG axis scrolling the markers of the recording session below the ECG axis and changing the RR series display mode on the right hand side of the RR axis An example of the ECG printout is shown in Fig 4 5 The ECG signal is displayed in a similar window as the report sheet and has thus e g the same kind of exporting functions see Section 4 3 2 for details 4 2 3 Analysis options The analysis options segment shown in Fig 4 6 includes three subcategories Frequency bands Interpolation of RR series and Spectrum estimation All of these options are con cerned with frequency domain analysis The ver
51. he MATLAB Component Runtime about 230 MB of hard disk space 1 2 Installation Kubios HRV Analysis was developed using MATLAB 7 0 4 R14SP2 and compiled to a deployable standalone Windows application with the MATLAB Compiler 4 2 Therefore the MATLAB Component Runtime MCR needs to be installed for Kubios HRV Analysis 1 Make sure that you have administrator rights and run the Kubios HRV Analysis in staller file This will launch the setup wizard To proceed with the installation click the Next button or click the Cancel button to exit installation 18 Setup Kubios HRV Analysis Welcome to the Kubios HRY Analysis Setup Wizard This will install Kubios HAY Analysis version 2 0 beta 1 on your computer It is recommended that you close all other applications before continuing Click Next to continue or Cancel to exit Setup Cancel vercion OO beta Department of Physics Ku bios HAV Analysis Biosignal Analysis and Medical Imaging Group University of Kuopio FINLAND 1 2 Installation 6 2 The next setup window displays the License Agreement for the software Read the License Agreement carefully before proceeding If you accept all the terms and con ditions of the Agreement you are allowed to continue the installation IF YOU DO NOT AGREE WITH ALL OF THE TERMS AND CONDITIONS OF THE LICENSE AGREEMENT YOU CANNOT CONTINUE THE INSTALLATION 18 Setup Kubios HRV Analysis License Agreement Please r
52. he data also affects ApEn When N is increased the ApEn approaches its asymptotic value The tolerance r has a strong effect on ApEn and it should be selected as a fraction of the standard deviation of the data SDNN This selection enables the comparison of different data types A common selection for r is r 0 2SDNN which is also used in this software 3 3 3 Sample entropy Sample entropy SampEn is similar to ApEn but there are two important differences in its calculation 40 20 For ApEn in the calculation of the number of vectors uz for which d uj ux lt r also the vector u itself is included This ensures that C r is always larger than 0 and the logarithm can be applied but at the same time it makes ApEn to be biased In sample entropy the self comparison of uj is eliminated by calculating C7 r as nbr of ug d uj up lt r Gre N m VkA j 3 12 Version 0 beta Department of Physics Kibios HRW Analysis Biosignal Analysis and Medical Imaging Group University of Kuopio FINLAND 3 3 Nonlinear methods 26 Now the value of C r will be between 0 and 1 Next the values of C7 r are averaged to yield 1 N m 1 C r esl 2 Ent 3 13 and the sample entropy is obtained as SampEn m r N In C r C Ar 3 14 The values selected for the embedding dimension m and for the tolerance parameter r in the software are the same as those for the approximate entropy calculation Both ApEn and SampEn are
53. he presented time domain frequency domain nonlinear and time varying measures of HRV calculated by the software are summarized in Table 3 1 For each measure preferred units and a short description is given In addition a reference to the equation in which the specific measure is defined is given when possible and related references are given for some of the measures Version 0 beta Department of Physics Kibios HAV Analysis Biosignal Analysis and Medical Imaging Group University of Kuopio FINLAND 3 5 Time Domain Frequency Domain Nonlinear Time Varying Summary of HRV parameters 31 Table 3 1 Summary of the HRV measures calculated by the software Measure RR STD RR SDNN HR STD HR RMSSD NN50 pNN50 HRV index TINN triangular Peak frequency Absolute power Relative power Normalized power LF HF SD1 SD2 ApEn SampEn Da DFA Q1 Q2 RPA Lmean Lmax REC DET ShanEn Units ms ms 1 min ds 1 ms beats beats Time domain measures RR SDNN HR STD HR RMSSD NN50 and pNN50 Frequency domain measures Peak frequencies absolute powers relative powers normalized powers and LF HF ratio Note Time varying spectrum is estimated using the spectrogram or Kalman smoother method Nonlinear measures ApEn and SampEn Kubios HRV Analysis version 2 0 beta Description References The mean of RR intervals Standard deviation of RR intervals E
54. he time series is measured by the variable Imax DET Elisa M Se Re Finally the Shannon information entropy of the line length distribution is defined as 3 27 Imax ShanEn 5 ni Inn 3 28 l lmin where n is the number of length lines divided by the total number of lines that is Ni lanos Ny U lmin 3 29 n 3 4 Time varying methods The time varying methods of the software include the trends of the time domain measures RR SDNN HR SD of HR RMSSD NN50 and pNN50 For frequency domain measures the trends are obtained for VLF LF and HF peak frequencies VLF LF and HF band powers and LF HF ratio In addition trends are calculated for the nonlinear measures ApEn and SampEn The trends for the time domain and nonlinear measures are obtained by using a moving window the length and shift of which can be changed The frequency domain measures trends are instead obtained from a time varying spec trum estimate The time varying spectrum is estimated either by using the moving win dow FFT which is also known as the spectrogram method or with the Kalman smoother algorithm The Kalman smoother algorithm is an iterative algorithm for estimating the parameters of a time varying model In the software a time varying AR model is used to model the HRV signal The adaptation of the Kalman smoother algorithm affecting on the resolution of the spectrum can also be altered 3 5 Summary of HRV parameters T
55. hich you are interested in e g watch out the lower limit of the LF band 59 60 How to select the interpolation rate of the RR series The interpolation rate is related to the cubic spline interpolation that is used for converting the RR interval series to equidistantly sampled series The default value for the interpolation rate is 4 Hz which works well for normal human HRV data It should be noted that the interpolation rate should be at least twice as high as the highest expected frequency in the RR interval series When changing the interpolation rate it should be remembered that it affects on the smoothness priors based detrending method i e when decreasing the interpolation rate also the value of the smoothness priors method should be decreased How to select the window width and overlap for the FFT spectrum estimate The FFT spectrum in the software is calculated using the Welch s periodogram method where one or more overlapped segments are extracted from the data Then FFT spectrum is calculated for each segment and as a result the average of the segment spectra is calculated The selection of the window width and overlap in this method is simply a trade off between the frequency resolution and variance of the spectrum estimate The frequency resolution of FFT spectrum is roughly the reciprocal of the sample length i e the frequency resolution of the FFT spectrum of a 100 second sample is 0 1 Hz The variance of the FFT
56. ic indices in heart rate variability of normal subjects and heart transplanted patients Cardiovascular Research 31 441 446 1996 P S Hamilton and W J Tompkins Quantitative investigation of QRS detection rules using the MIT BIH arrhythmia database IEEE Trans Biomed Eng 33 12 1157 1165 December 1986 B Henry N Lovell and F Camacho Nonlinear dynamics time series analysis In M Akay editor Nonlinear Biomedical Signal Processing Dynamic Analysis and Mod eling volume II chapter 1 pages 1 39 IEEE Press New York 2001 H V Huikuri T H Makikallio P Raatikainen J Perki m ki A Castellanos and R J Myerburg Prediction of sudden cardiac death appraisal of the studies and methods assessing the risk of sudden arrhythmic death 108 1 110 115 July 2003 S J Johnsen and N Andersen On power estimation in maximum entropy spectral analysis Geophysics 43 681 690 June 1978 D E Lake J S Richman M P Griffin and J R Moorman Sample entropy analysis of neonatal heart rate variability ajp 283 R789 R797 September 2002 N Lippman K M Stein and B B Lerman Nonlinear predictive interpolation a new method for the correction of ectopic beats for heart rate variability analysis J Electrocardiol 26 514 S19 1993 N Lippman K M Stein and B B Lerman Comparison of methods for removal of ectopy in measurement of heart rate variability Am J Physiol 267 1 H411 H418 July 1994 D A Litvack T F Oberlander
57. ing window settings used for time domain and nonlinear variables are also utilized for the spectrogram The Kalman smoother method on the other hand is based on time varying AR modeling and does not utilize the same kind of moving window as the spectrogram Thus the window width value does not apply to the frequency domain variables if the Kalman smoother is used for spectrum estimation The Grid interval is however utilized for the Kalman smoother method as well The differences between the spectrogram and Kalman smoother methods have been discussed in 42 where they were applied to nonstationary EEG signals In brief it can be said that the Kalman smoother is computationally more complex but yields better resolution than the spectrogram The spectrogram is also more robust and requires only the moving window settings to be defined The Kalman smoother methods on the other hand requires the fixing of both the adaptation coefficient and the AR mode order Spectrogram and Kalman smoother spectrum estimates for the orthostatic measurement are presented in Figs 5 4 b and c The adaptation coefficient of the Kalman smoother method was set to 0 1 and the AR model order to 16 The results of the time varying analysis can be saved as in the first sample run If the general analysis was disabled from the preferences window only time varying results will be included in the PDF and ASCII text files In the PDF file the time varying results are presen
58. io These details will be displayed in the report sheet and ASCII results file Report settings Figure 4 13 Set up preferences window of the software User information category The variable names of the different fields are more or less self descriptive and are not docu mented here 4 4 Setting up the preferences All the analysis options that can be adjusted in the user interface have some default values These preference values will be used every time the program is started Any changes made on these values in the user interface only apply for the current session The preference values are designed to be more or less suitable for short term HRV recordings and may sometimes need to be redefined This can be done by selecting Edit Preferences from the File menu or by pressing the corresponding toolbar button This will open the preferences window in which the preference values can be redefined The preferences are divided into four categories User information Analysis options Advanced settings and Report settings In the User information category shown in Fig 4 13 you can set up your personal contact information Name Department and Organization This information will only be included in the bottom left corner of the report sheet and in the beginning of the ASCII text file including the analysis results That is the user information is meant just for indicating the person who has carried out the analysis The Analysis opti
59. it only generates a wrapper executable that starts the MATLAB Component Runtime MCR and runs the heavily crypted Matlab M files of the compiled application on top of the MCR Although this has many advantages the main disadvantage is that the starting of the MATLAB Component Runtime takes roughly the same time as starting MATLAB This can be anything from 10 to 50 seconds depending on the speed of the computer The report sheets print in black and white although I m printing to a color printer Due to a bug in the MATLAB Compiler 4 printing directly to a printer does not work in a MATLAB created standalone applications Therefore the printing system had to be developed in a different way for the Ku bios HRV Analysis Currently the printing system uses GNU Ghostscript http www gnu org software ghostscript ghostscript html to print exported postscript files The generic Windows printing device included in the rather old version of GNU Ghostscript that is provided with the MATLAB Component Runtime seems to have some problems printing in color This will probably change when the printing bug is resolved hopefully in the future versions of MATLAB Compiler For now the workaround is to export the report sheets to a PDF file and print the resulting PDF file using e g Adobe Acrobat Reader 62 References 1 10 11 V X Afonso ECG QRS detection In W J Tompkins editor Biomedical Digital Signal Processi
60. ith mouse and drag it to the desired direction Close button is for closing the report sheet P The File menu includes Export to Export All to PDF Print Current Page Print All Pages dy Kubios HAV Analysis Biosignal Analysis and Medical Imaging Group i Department of Physics eee ae University of Kuopio FINLAND 4 3 Saving the results Report Page 1 File Edit 44 RSMALAA IP HRY Analysis General Results RR interval Time Bories Bolsotd Cobended RA Banoe 0 1 Tims Comain Re wite Vartable UNE va Vean AR ms 12157 STO AR DNN Mean HR STO HR RISD LLES pH RR harguiar index TINN ms Frequenoy Comain Re wite Weurce can_dais cnl 211 00 152455 Page 12 wiw bra dngis cample Deterdirg me hod moon priors X 00 00 10 00 00 14 40 Dishbulons 11 12 13 D 55 RR 5 HR Q eakiminy FFT eotum Qivelch s penodogram 255 5 window wih 50 cuedap AR Bpsotum AR model omer 16 mol tsciodzed D PSD sH1 og Lf 0040 15 kn 0055 HF Den Hg oDer Monlinear Re wit v ale UNE vE Polnoars plot S01 ns S02 ms Reourrenoe plot Me lire lergh Lmeay Meat Nex line lerg h Lmag Qeab Reocumence rr REC Delemirism D ET Shernon Enkopy Erener Ober Approdmale entropy Ap Er Sample entropy amp Ey De kerded tucholore DFA a1 De kerded tucholore DFA a2 Comelalon dimersion 02 174k 20S 05 17 57 Mia Taren Deparmeni ot Applied Physics Urduers ty of Kuopio oo
61. j 3 3 j 1 Another measure calculated from successive RR interval differences is the NN50 which is the number of successive intervals differing more than 50 ms or the corresponding relative amount pNN50 x 100 3 4 3 2 Frequency domain methods 23 In addition to the above statistical measures there are some geometric measures that are calculated from the RR interval histogram The HRV triangular index is obtained as the integral of the histogram i e total number of RR intervals divided by the height of the histogram which depends on the selected bin width In order to obtain comparable results a bin width of 1 128 seconds is recommended 44 Another geometric measure is the TINN which is the baseline width of the RR histogram evaluated through triangular interpolation see 44 for details 3 2 Frequency domain methods In the frequency domain methods a power spectrum density PSD estimate is calculated for the RR interval series The regular PSD estimators implicitly assume equidistant sampling and thus the RR interval series is converted to equidistantly sampled series by interpolation methods prior to PSD estimation In the software a cubic spline interpolation method is used In HRV analysis the PSD estimation is generally carried out using either FFT based methods or parametric AR modeling based methods For details on these methods see e g 27 The advantage of FFT based methods is the simplicity of i
62. l tachogram left panel and the interval function right panel 28 This distortion becomes substantial when the variability is large in comparison with the mean level Furthermore the spectrum can not be considered to be a function of frequency but rather of cycles per beat 9 Another common approach adopted in this software is to use interpolation methods for converting the non equidistantly sampled RR interval time series also called the interval function to equidistantly sampled 44 see the right panel of Fig 2 3 One choice for the interpolation method is the cubic spline interpolation 28 After interpolation regular spectrum estimation methods can be applied The third general approach called the spectrum of counts considers a series of impulses delta functions positioned at beat occurrence times 10 This approach relies on the generally accepted integral pulse frequency modulator IPFM which aims to model the neural modulation of the SA node 41 According to this model the modulating signal is integrated until a reference level is achieved after which an impulse is emitted and the integrator is set to zero The spectrum of the series of events can be calculated e g by first lowpass filtering the event series and then calculating the spectrum of the resulting signal 9 2 3 Preprocessing of HRV time series Any artifact in the RR interval time series may interfere the analysis of these signals The artifacts within HR
63. l analysis of the heart rate variability IEEE Trans Biomed Eng 37 1 99 106 January 1990 I P Mitov A method for assessment and processing of biomedical signals containing trend and periodic components Med Eng Phys 20 9 660 668 November December 1998 J P Niskanen M P Tarvainen P O Ranta aho and P A Karjalainen Software for advanced HRV analysis Comput Meth Programs Biomed In Press M Pagani Heart rate variability and autonomic diabetic neuropathy Diabetes Nutri tion amp Metabolism 13 6 341 346 2000 O Pahlm and L S rnmo Software QRS detection in ambulatory monitoring a review Med Biol Eng Comput 22 289 297 July 1984 J Pan and W J Tompkins A real time QRS detection algorithm IEEE Trans Biomed Eng 32 3 230 236 March 1985 C K Peng S Havlin H E Stanley and A L Goldberger Quantification of scaling exponents and crossover phenomena in nonstationary heartbeat time series Chaos 5 82 87 1995 T Penzel J W Kantelhardt L Grote J H Peter and A Bunde Comparison of detrended fluctuation analysis and spectral analysis for heart rate variability in sleep and sleep apnea IEEE Trans Biomed Eng 50 10 1143 1151 October 2003 G D Pinna R Maestri A Di Cesare R Colombo and G Minuco The accuracy of power spectrum analysis of heart rate variability from annotated RR lists generated by Holter systems Physiol Meas 15 163 179 1994 S W Porges and R E Bohrer The analysis of
64. lected bottom left corner of the user interface When unchecked one or more changes to options can be made without updating breaks and when finished with changes the Apply button can be pressed to update the results The time domain results view shown in Fig 4 7 displays the time domain variables in a table and the RR interval and heart rate histograms in the two axes Most of the results are calculated from the detrended RR series if detrending is applied but there are two obvious exceptions i e mean RR interval and mean HR which are marker with the symbol The frequency domain results view shown in Fig 4 8 displays the results for both FFT and AR spectrum estimation methods Both methods are applied to the detrended RR series The spectra of the two methods are presented in the two axes FFT spectrum on the left and AR spectrum on the right The frequency axes of the spectra are fixed to range from O Hz to the upper limit of HF band plus 0 1 Hz Thus for the default frequency band settings the frequency axis range is 0 0 5 Hz The power axes of the spectra on the other hand can be adjusted with the options on the upper left corner of the frequency domain results view The power axes can be selected to have either common or separate upper Version 0 beta Department of Physics Kibios HAV Analysis Biosignal Analysis and Medical Imaging Group University of Kuopio FINLAND 4 2 The user interface 40 VIEW RESULTS Frequency Domai
65. mplementation while the AR spectrum yields improved resolution especially for short samples Another property of AR spectrum that has made it popular in HRV analysis is that it can be factorized into separate spectral components The disadvantages of the AR spectrum are the complexity of model order selection and the contingency of negative components in the spectral factorization Nevertheless it may be advantageous to calculate the spectrum with both methods to have comparable results In this software the HRV spectrum is calculated with FFT based Welch s periodogram method and with the AR method Spectrum factorization in AR method is optional In the Welch s periodogram method the HRV sample is divided into overlapping segments The spectrum is then obtained by averaging the spectra of these segments This method decreases the variance of the FFT spectrum The generalized frequency bands in case of short term HRV recordings are the very low frequency VLF 0 0 04 Hz low frequency LF 0 04 0 15 Hz and high frequency HF 0 15 0 4 Hz The frequency domain measures extracted from the PSD estimate for each frequency band include absolute and relative powers of VLF LF and HF bands LF and HF band powers in normalized units the LF HF power ratio and peak frequencies for each band see Table 3 1 In the case of FFT spectrum absolute power values for each frequency band are obtained by simply integrating the spectrum over the band limits In the
66. n Nonlinear Time Varying lini Y FFT spectrum estimation results Common Y limit 04 p iis 0 04 Time Domain AR spectrum estimation results gt D 0 03 0 02 PSD s Hz Sos B PSD s2 Hz gt i 0 01 Ls D 0 1 0 2 0 3 0 4 Frequency Hz Frequency Hz Frequency Power Power Frequency Peak Power Power Band mms 96 Band Ha ms 96 VLF 148 10 3 VLF 0 0039 355 11 9 LF 512 35 5 LF 0 0430 717 24 0 HF 782 54 2 HF 0 2734 1912 64 1 LEJHF 07 LF HF 0 4 Figure 4 8 The results view segment of the user interface frequency domain results view selected Honlinear Analysis Results Variable Poincare plot SD1 SD2 Recurrence plot Mean line length Lmean Maximum line length Lmax Recurrence rate REC Determinism DET Units ms ms beats beats 96 98 Shannon Entropy ShanEn Other Approximate Entropy ApEn Sample Entropy SampEn Detrended fluctuations DFA 0 1 Detrended fluctuations DFA o2 Correlelation dimension D2 14 1 2 13 1 4 06 08 1 12 14 16 18 RR 5 log y N beats Calculated from the non detrended selected RR series Figure 4 9 The results view segment of the user interface nonlinear results view selected Y limits If common Y limit is selected it can also be entered manually into the edit box below the selection button The selected power axis options apply also for the report sheet The results for b
67. n the low frequency and high frequency HRV in an opposite way That is when subject stands up after lying for few minutes a strong decrease in the HF power and a more gradual increase in LF power are observed In addition a strong increase in heart rate in observed immediately after standing up which aims to compensate the sudden decrease in blood pressure In the first sample run this data file is analyzed by considering the lying and standing periods separately whereas in the second sample run we perform a time varying analysis for the whole measurement 5 1 Sample run 1 General analysis In the first sample run we show how to make the general analysis i e time domain frequency domain and nonlinear analysis for the lying and standing periods of the or thostatic measurement separately This task can be easily accomplished in a single session First start the software and open the data file into the user interface At this point you can edit any of the analysis options to fit your demands If you are about to analyze several data files with the same options you better make these changes straight to the preferences Here since we wish to do only general analysis we want to disable the time varying analysis from the software To do this select Edit preferences from the File menu and uncheck the time varying analysis from Analysis options category of the preferences window Then press the OK button Now the Time varying button in the results
68. nd the analysis results below them When Save Results have been selected these report sheets can be saved in a single PDF file by selecting Report figure as the saving type in the save dialog In this case the report sheets will not be displayed but just saved in the selected PDF file If you wish to view the report sheets and or to export them into some other file format choose Report sheet from the View menu or just press the corresponding toolbar button This will open the report sheet windows for view The report sheet windows include 8 toolbar buttons and File and Page menus on the upper left hand corners of the windows The toolbar button icons and their actions are given below E Export figure button opens an export dialog where the report sheet can be exported into one of the various file formats listed in Table 4 1 Print button opens a print dialog where the report sheet can sent to the selected printer Export all pages to PDF file button is for exporting both report sheets into the selected PDF file Zoom in button if for zooming in magnifying the report sheet Zoom out button is for zooming out the report sheet OD ie Reset to original size button can be used to restore the original zoom level This also resets the size of the corresponding report sheet window to its original size Move visible area button is for moving the visible area of the zoomed report sheet in the report window just grab the sheet w
69. ned through time varying spectrum estimation for which the commonly used spectrogram and the statistically sophisticated Kalman smoother methods are available In addition to the usability of the program user interface attention has also been paid in the presentation and saving of the analysis results On the one hand the software generates two report sheets which include all the analysis results in a compact form These report sheets can be either printed or exported to various file formats including portable document format PDF On the other hand the results can also be saved in an ASCII text file from which they can be imported to a spreadsheet program such as Microsoft Excel for further inspection 1Kuopio University has only limited rights to the software These limited rights are governed by a certain license agreement between Kuopio University and The MathWorks Inc 1 1 System requirements 5 1 1 System requirements The system requirements given below should be considered as recommended system re quirements The software may work also with lower system specifications but will probably function slower or with reduced usability e Microsoft Windows XP 2000 operating system e 1 GHz or higher Intel compatible 32 bit x86 processor e 256 MB of RAM 512 MB or higher recommended e Desktop resolution of 1024x768 or higher e The MATLAB Component Runtime version 7 2 e The Kubios HRV Analysis will require about 10 MB and t
70. ng chapter 12 pages 237 264 Prentice Hall New Jersey 1993 G Baselli S Cerutti S Civardi F Lombardi A Malliani M Merri M Pagani and G Rizzo Heart rate variability signal processing a quantitative approach as an aid to diagnosis in cardiovascular pathologies Int J Bio Med Comput 20 51 70 1987 G G Berntson J T Bigger Jr D L Eckberg P Grossman P G Kaufmann M Malik H N Nagaraja S W Porges J P Saul P H Stone and M W Van Der Molen Heart rate variability Origins methods and interpretive caveats Psychophysiol 34 623 648 1997 H J Braune and U Geisen rfer Measurement of heart rate variations influencing fac tors normal values and diagnostic impact on diabetic autonomic neuropathy Diabetes Res Clin Practice 29 179 187 1995 M Brennan M Palaniswami and P Kamen Do existing measures of Poincar plot geometry reflect nonlinear features of heart rate variability JEEE Trans Biomed Eng 48 11 1342 1347 November 2001 S Carrasco M J Cait n R Gonz lez and O Y nez Correlation among Poincar plot indexes and time and frequency domain measures of heart rate variability 25 6 240 248 November December 2001 H Dabire D Mestivier J Jarnet M E Safar and N Phong Chau Quantification of sympathetic and parasympathetic tones by nonlinear indexes in normotensive rats amj 44 H1290 H1297 1998 I Daskalov and I Christov Improvement of resolution in measurement of elect
71. nue click Next If you would like to select a different folder click Browse ry Kubios HRV Analysis C Don t create a Start Menu folder Kibios HAV Analysis Biosignal Analysis and Medical Imaging Group Department of Physics See pena University of Kuopio FINLAND 1 2 Installation 8 6 In the next setup window you can select if the installer creates Desktop and or Quick Launch icons for software Select the desired additional icons and click Next to con tinue 18 Setup Kubios HRV Analysis Select Additional Tasks Which additional tasks should be performed Select the additional tasks you would like Setup to perform while installing Kubios HRY Analysis Click Next to continue Additional icons O Create a desktop icon 7 The setup is now completed and the installer is ready to install the Kubios HRV Analysis and MATLAB Component Runtime if selected on your computer Check the selected installation options summarized in the setup window and click the Install button to start the installation If you want to change any of the installation settings click the Back button 18 Setup Kubios HRV Analysis Ready to Install Setup is now ready to begin installing Kubios HAY Analysis on your computer Click Install to begin the installation or click Back if you want to review or change any settings Destination folder A C Program FilestKubios HRW Analysis E Setup type
72. ons category shown in Fig 4 14 includes some basic analysis op tions The default input data type can be set to one of the file formats mentioned in Section 4 1 and the selected data type will be used as default every time a new data file is opened Te analysis to be performed options include general analysis i e time domain frequency domain and nonlinear analysis and time varying analysis Only the selected analysis will be performed For example if time varying analysis is not necessary it can be checked off to speed up the functioning of the software In addition the analysis options category includes dy Kubios HRV Analysis Biosignal Analysis and Medical Imaging Group Department of Physics 2 0 bet eee University of Kuopio FINLAND 4 4 Setting up the preferences 48 Preferences A A ES 3 User information Default Input Data Type Data type Neuro Scan CNT files Analysis to be Performed V General analysis M Time varying analysis RR Interval Samples Number of samples Sample analysis type RR Interval Detrending Detrending method smoothing Parameter HRY Frequency Bands Very low frequency YLF Low frequency LF Advanced settings Report settings High frequency HF Update Analysis Results Update mode Atomatic Del Figure 4 14 Set up preferences window of the software Analysis options category RR interval samples RR interval detrending
73. oth spectra are displayed in tables below the corresponding spectrum axes The nonlinear results view shown in Fig 4 9 displays all the calculated nonlinear vari ables in one table All the variables are calculated from the original non detrended RR series The Poincar plot and the DFA results are also presented graphically in the two axes In the Poincar plot left hand axis the successive RR intervals are plotted as blue circles and the SD1 and SD2 variables obtained from the ellipse fitting technique are presented for details see Section 3 3 1 In the DFA plot right hand axis the detrended fluctuations F n are presented as a function of n in a log log scale and the slopes for the short term and long term fluctuations a and ag respectively are indicated for details see Section 3 3 4 The time varying results view shown in Fig 4 10 displays the time varying trend of the selected variable The variable is selected using the two buttons on the lower left hand corner of the view Selectable variables are divided into time domain frequency domain and nonlinear categories The trend of the selected variable will appear immediately in the axis The line style of the trend can be changed by pressing the button on the right Biosignal Analysis and Medical Imaging Group Department of Physics University of Kuopio FINLAND Kubios HRV Analysis version 2 0 beta 4 2 The user interface 41 VIEW RESULTS Time Domain l Frequency Domain
74. partment of Physics Kibios HAV Analysis Biosignal Analysis and Medical Imaging Group University of Kuopio FINLAND Chapter 3 Analysis methods In this chapter the analysis methods used in the software are introduced The presented methods are mainly based on the guidelines given in 44 The presentation of the methods is divided into four categories i e time domain frequency domain nonlinear and time varying methods The methods summarized in Table 3 1 3 1 Time domain methods The time domain methods are the simplest to perform since they are applied straight to the series of successive RR interval values The most evident such measure is the mean value of RR intervals RR or correspondingly the mean HR HR In addition several variables that measure the variability within the RR series exist The standard deviation of RR intervals SDNN is defined as SDNN 3 1 where RR denotes the value of 7 th RR interval and N is the total number of successive intervals The SDNN reflects the overall both short term and long term variation within the RR interval series whereas the standard deviation of successive RR interval differences SDSD given by SDSD y E ARR5 E ARR 3 2 can be used as a measure of the short term variability For stationary RR series E ARR E RRj 41 E RR 0 and SDSD equals the root mean square of successive differences RMSSD given by N 1 RMSSD F7 Y RRy41 RR
75. periodic processes in psychophysiological research In J T Cacioppo and L G Tassinary editors Principles of Psychophysiology Physical Social and Inferential Elements pages 708 753 Cambridge University Press 1990 J A Richman and J R Moorman Physiological time series analysis using approximate entropy and sample entropy Am J Physiol 278 H2039 H2049 2000 O Rompelman Rhythms and analysis techniques In J Strackee and N Wester hof editors The Physics of Heart and Circulation pages 101 120 Institute of Physics Publishing Bristol 1993 M P Tarvainen J K Hiltunen P O Ranta aho and P A Karjalainen Estimation of nonstationary EEG with Kalman smoother approach an application to event related synchronization ERS IEEE Trans Biomed Eng 51 3 516 524 March 2004 Department of Physics ee ae University of Kuopio FINLAND dy Kubios HRW Analysis Biosignal Analysis and Medical Imaging Group References 66 43 M P Tarvainen P O Ranta aho and P A Karjalainen An advanced detrending method with application to HRV analysis IEEE Trans Biomed Eng 49 2 172 175 February 2001 44 Task force of the European society of cardiology and the North American society of pac ing and electrophysiology Heart rate variability standards of measurement physio logical interpretation and clinical use 93 5 1043 1065 March 1996 45 N V Thakor J G Webster and W J Tompkins Optimal QRS detector Med Biol Eng Compu
76. phical user interface of the developed HRV analysis software trend can be removed using the so called smoothness priors method presented in 43 In the smoothness priors method the smoothness of the removed trend can be adjusted by editing the Lambda value The bigger the value of Lambda the smoother is the removed trend The trend to be removed from the RRI series is shown over the selected part of the RR series as a red line 4 2 2 Data browser The data browser segment shown in Fig 4 4 displays the measured ECG signal and the extracted RR interval series It should be noted that if only RR interval data is given as input the ECG axis will not be displayed and the RR series axis will be bigger in size The ECG and RR interval data can be scrolled with the two sliders The position of the ECG axis is displayed as a green patch in the RR axis This patch can also be moved with the left mouse button The range of both axes can be changed by editing the Range values and also the Y limits of the axes can be manually changed by editing the edit boxes on the left hand side of the axes The ECG and RR interval axes can also be scrolled together by locking the axes by pressing the button connecting the Range values In addition to the visualization of the ECG and RR interval data the main function of this segment is to enable correction of corrupted RR interval values This can be done in Kibios HAV Analysis Biosignal Analysis and Medical Imaging Group
77. pio FINLAND 3 3 Nonlinear methods 25 3 3 2 Approximate entropy Approximate entropy ApEn measures the complexity or irregularity of the signal 12 40 Large values of ApEn indicate high irregularity and smaller values of ApEn more regular signal The ApEn is computed as follows First a set of length m vectors uj is formed Uj RR RRj41 RRj4m 1 j 1 2 N m41 3 7 where m is called the embedding dimension and N is the number of measured RR intervals The distance between these vectors is defined as the maximum absolute difference between the corresponding elements i e d uj up max RRjin RRk4n n 0 m 1 3 8 Next for each u the relative number of vectors uz for which d uj uk lt r is calculated This index is denoted with C r and can be written in the form nbr of Luz d uj uk lt r k N m 1 ij 3a Cr r Due to the normalization the value of C r is always smaller or equal to 1 Note that the value is however at least 1 N m 1 since u is also included in the count Then take the natural logarithm of each C7 r and average over j to yield 1 N m 1 Finally the approximate entropy is obtained as ApEn m r N r 97 1 r 3 11 Thus the value of the estimate ApEn depends on three parameters the length m of the vectors uj the tolerance value r and the data length N In this software the value of m is selected to be m 2 The length N of t
78. port sheets or results file you should at the same time uncheck the general analysis from the preferences Then press the OK button and the time varying analysis will be enabled in the user interface For the time varying analysis we wish to include the whole measurement period and thus we set the Range of the RR interval sample to 0 737 seconds This can be done using the Range edit boxes or by resizing the sample patch with the mouse For the same reasons as in the first sample run we use again the smoothness priors detrending method with the Lambda value of 500 The time varying analysis results can then be viewed in the results view segment press the Time varying button The variable in view can be selected from the two pop up buttons on the lower left hand corner of the results view For example in Fig 5 4 a the time domain variable pNN50 has been selected for view All the time domain variables as well as the two nonlinear variables are calculated using a moving window the width and time shift of which can be changed by editing the Window width and Grid interval values Here we have used a 60 second window and a 10 second grid interval The frequency domain variables on the other hand are obtained from the time varying spectrum estimate for which there are two different methods available These are the spectrogram and the Kalman smoother methods The spectrogram method is simply a moving window Fourier transformation method The same mov
79. pply or the OK button The Apply button applies the modifications to the current session while the OK saves the preferences but they will be applied only in the next session A session is considered to be ended when the program is restarted or Close file is selected If on the other hand a new file is opened without first closing the previous file preferences will not be applied but the local settings changes made in the user interface are applied for the new file as well Note also that by pressing the Apply button in the preferences window your local settings will be replaced with the updated preferences Version U beta Department of Physics dy Ku bios HAV Analysis Biosignal Analysis and Medical Imaging Group University of Kuopio FINLAND 4 4 Setting up the preferences 49 Preferences Advanced settings User information QRS Detection Options Analysis options Prior guess for RR interval automatic E vanced settings Average interval s Spectrum Estimation opine Interpolation of RR series 4 Hz Report settings Points in frequency domain 256 points Hz FFT spectrum using Welch s periodogram method Window width 256 s Window overlap so AR spectrum l AR model order l Use spectral factorization No zl Time Varying Spectrum Estimation Options Window width s Grid interval 10 s Spectrum estimation method Spectrogram Kalman smooth adapt coef 01 Apply OK Cancel Figure 4 15 Se
80. q 3 1 The mean heart rate Standard deviation of intantaneous heart rate values Square root of the mean squared differences between successive RR intervals Eq 3 3 Number of successive RR interval pairs that differ more than 50 ms NN5O divided by the total number of RR intervals Eq 3 4 The integral of the RR interval histogram divided by the height of the histogram 44 Baseline width of the RR interval histogram 44 VLF LF and HF band peak frequencies Absolute powers of VLF LF and HF bands Relative powers of VLF LF and HF bands VLF VLF ms total power ms x 100 LF LF ms total power ms x 100 HF HF ms total power ms x 100 Powers of LF and HF bands in normalized units LF n u LF ms total power ms VLF ms HF n u HF ms total power ms VLF ms Ratio between LF and HF band powers The standard deviation of the Poincar plot perpendicular to SD1 and along SD2 the line of identity 5 6 Approximate entropy Eq 3 11 40 12 Sample entropy Eq 3 14 40 Correlation dimension Eq 3 21 15 17 Detrended fluctuation analysis 36 37 Short term fluctuation slope Long term fluctuation slope Recurrence plot analysis 47 7 49 Mean line length Eq 3 26 Maximum line length Recurrence rate Eq 3 24 Determinism Eq 3 27 Shannon entropy Eq 3 28 Biosignal Analysis and Medical Imaging Group Department of Physics University
81. required to be 1 2 ms and thus the sampling frequency of the ECG should be at least 500 1000 Hz 44 If the Version 0 beta Department of Physics Ku bios HAV Analysis Biosignal Analysis and Medical Imaging Group University of Kuopio FINLAND 2 2 Derivation of HRV time series 18 branches Purkinje Ventricular muscle Gah Sle bie adler ala la 3 Time ms 0 100 200 300 400 500 600 700 Figure 2 2 Electrophysiology of the heart redrawn from 26 The different waveforms for each of the specialized cells found in the heart are shown The latency shown approximates that normally found in the healthy heart sampling frequency of the ECG is less than 500 Hz the errors in R wave occurrence times can cause critical distortion to HRV analysis results especially to spectrum estimates 30 The distortion of the spectrum is even bigger if the overall variability in heart rate is small 38 The estimation accuracy can however be improved by interpolating the QRS complex e g by using a cubic spline interpolation 8 It should be however noted that when the SA node impulses are of interest there is an unavoidable estimation error of approximately 3 ms due to fluctuations in the AV nodal conduction time 41 2 2 Derivation of HRV time series After the QRS complex occurrence times have been estimated the HRV time series can be derived The inter beat intervals or RR intervals are obtained as differences between suc cessive R wave
82. rocar diogram RR intervals by interpolation Med Eng Phys 19 4 375 379 June 1997 R W DeBoer J M Karemaker and J Strackee Comparing spectra of a series of point events particularly for heart rate variability data IEEE Trans Biomed Eng 31 4 384 387 April 1984 R W DeBoer J M Karemaker and J Strackee Spectrum of a series of point events generated by the integral pulse frequency modulation model Med Biol Eng Comput 23 138 142 March 1985 G M Friesen T C Jannett M A Jadallah S L Yates S R Quint and H T Nagle A comparison of the noise sensitivity of nine QRS detection algorithms EEE Trans Biomed Eng 37 1 85 98 January 1990 63 References 64 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 Y Fusheng H Bo and T Qingyu Approximate entropy and its application in biosignal analysis In M Akay editor Nonlinear Biomedical Signal Processing Dynamic Analysis and Modeling volume II chapter 3 pages 72 91 IEEE Press New York 2001 P Grassberger and I Procaccia Characterization of strange attractors Phys Rev Lett 50 346 349 1983 P Grossman Breathing rhythms of the heart in a world of no steady state a comment on Weber Molenaar and van der Molen Psychophysiol 29 1 66 72 January 1992 S Guzzetti M G Signorini C Cogliati S Mezzetti A Porta S Cerutti and A Malliani Non linear dynamics and chaot
83. ropathy 4 33 In addition HRV has been found to correlate with e g age mental and physical stress and attention see e g the review in 3 The term HRV refers in general to changes in heart beat interval which is a reciprocal of the heart rate This is also the case here The starting point for HRV analysis is the ECG recording from which the HRV time series can be extracted In the formulation of the HRV 16 2 1 Heart beat period and QRS detection 17 Vasomotor sympathetic Sympa thetic Sympa thetic Baroreceptors Figure 2 1 The four baroreflex pathways redrawn from 41 Variation in venous volume AV left ventricular contractility VC sympathetic and parasympathetic vagal control of heart rate HR stroke volume V cardiac output CO total peripheral resistance TPR and arterial blood pressure BPa time series a fundamental issue is the determination of heart beat period 2 1 Heart beat period and QRS detection The aim in HRV analysis is to examine the sinus rhythm modulated by the autonomic ner vous system Therefore one should technically detect the occurrence times of the SA node action potentials This is however practically impossible and thus the fiducial points for the heart beat is usually determined from the ECG recording The nearest observable activ ity in the ECG compared to SA node firing is the P wave resulting from atrial depolarization see Fig 2 2 and thu
84. rval samples are used in the calculation of the variable value In the frequency domain a longer window corresponds to better frequency resolution of the time varying spectrum On the other hand using shorter window the time variation of the changes in RR interval series characteristics is better observed Version 0 beta Department of Physics K bios HAV Analysis Biosignal Analysis and Medical Imaging Group University of Kuopio FINLAND Appendix B Troubleshooting Kubios HRV Analysis fails to start and gives the following error message This application has failed to start because MCLMCRRT72 DLL was not found Re installing the application may fix this problem This error message is produced if the Kubios HRV Analysis cannot find the MATLAB Component Runtime MCR Verify that you have the MATLAB Component Runtime v7 2 installed on your system The MATLAB Component Runtime is included in the Kubios HRV Analysis installer If you have the MCR installed and Kubios HRV Analysis still fails to start make sure that the system path contains the entry for MCR i e the lt MCR install dir gt v72 runtime win32 directory is found in the system path Note that in order to modify the system path you need to have ad ministrator privileges The Kubios HRV Analysis seems to take ages to start The MATLAB Compiler 4 has changed dramatically from the earlier versions It is now more of a deployment tool than a compiler because
85. s if ECG measurement is available it is recommended that the ECG measurement corresponding to the analyzed RR interval series is checked for artifacts In the software this can be easily done by scrolling the ECG signal Any misdetections can be corrected straight and possible ectopic beats can be excluded from the analysis by suitable selections for the analyzed sample range When several RR interval samples have been selected what does the merge samples do When the Merge samples option is selected for the Sample analysis type the RR interval samples selected for analysis are simply merged into one sample by concatenating the samples m How to select the value of A in the smoothness priors based detrending A is the regularization parameter in the smoothness priors based detrending approach see Section 2 3 1 for details The value of this parameter changes the smoothness of the estimated trend i e a bigger value corresponds to a smoother trend As discussed in Section 2 3 1 the smoothness priors detrending method can be compared to a time varying highpass filter and A adjusts the cut off frequency of the filter A practical way to observe what this cut off frequency is for a specific A value is as follows Start for example from a big value e g A 10000 and decrease it to some desired value and observe from the FFT spectrum which frequencies were eliminated Just remember to make sure that you are not eliminating frequencies of w
86. s the heart beat period is generally defined as the time difference between two successive P waves The signal to noise ratio of the P wave is however clearly lower than that of the strong QRS complex which results primarily from ventricular depo larization Therefore the heart beat period is commonly evaluated as the time difference between the easily detectable QRS complexes A typical QRS detector consists of a preprocessing part followed by a decision rule Several different QRS detectors have been proposed within last decades 45 34 35 16 11 For an easy to read review of these methods see 1 The preprocessing of the ECG usually includes at least bandpass filtering to reduce power line noise baseline wander muscle noise and other interference components The passband can be set to approximately 5 30 Hz which covers most of the frequency content of QRS complex 34 In addition preprocessing can include differentiation and or squaring of the samples After preprocessing the decision rules are applied to determine whether or not a QRS complex has occurred The decision rule usually includes an amplitude threshold which is adjusted adaptively as the detection progresses In addition the average heart beat period is often used in the decision The fiducial point is generally selected to be the R wave and the corresponding time instants are given as the output of the detector The accuracy of the R wave occurrence time estimates is often
87. s 53 series axis becomes highlighted Note that you can force a common Y limit for the spec tra of both samples by setting a common Y limit value manually in the frequency domain results view For example we have here fixed the Y limit value to 0 04 s Hz Once we are done with the analysis we wish to save the analysis results in all possible formats This can be done by selecting Save Results from the File menu or just by pressing the save button on the toolbar Then select Save all txt mat pdf as the save type and enter a file name You do not need to give any extension to the file name The numeric results of the analysis will be saved in the txt text file and in the mat MATLAB file and the report sheets in the pdf file The generated PDF file will now include two pages one for the results of the first RR interval sample the lying period and one for the second sample standing period These report sheet pages are shown in Figs 5 2 and 5 3 In the text file the results for the two samples are presented side by side as can be seen from the partial results text file given below Performed Analysis General analysis Yes Time varying analysis No RR Interval Samples Selected for Analysis A Sample 1 Sample 2 Sample limits s H 10 350 390 720 Sample Analysis Type Single samples RESULTS FOR SINGLE SAMPLES GENERAL RESULTS A SAMPLE 1 A SAMPLE 2 H Time Domain Results z Statistical parameters Mean RR ms gt
88. should always be at least twice the number of spectral peaks in the data The second option is whether or not to use spectral factorization in the AR spectrum estimation In the fac torization the Ar spectrum is divided into separate components and the power estimates of each component are used for the band powers The factorization however has some serious problems which can distort the results significantly The main problems are the selection of the model order in such a way that only one AR component will result in each frequency band and secondly negative power values can result for closely spaced AR components Thus the selection of not to use factorization in AR spectrum is surely more robust and in that sense recommended 4 2 4 Results view The results for the selected RR interval sample are displayed in the results view segment The results are divided into time domain frequency domain nonlinear and time varying results The results of each section are displayed by pressing the corresponding button on the top of the results view segment The results are by default updated automatically whenever any one of the the sample or analysis options that effect on the results is changed The updating of the results can be time consuming for longer samples and in that case it might be useful to disable the automatic update by unchecking the Automatic check box in the dy Kubios HAV Analysis Biosignal Analysis and Medical Imaging Group f Departm
89. sion and lag were selected to be m 10 and 7 1 respectively The threshold distance r was selected to be ym SD where SD is the standard deviation of the RR time series The selection are similar to those made in 7 The first quantitative measure of RP is the recurrence rate REC which is simply the ratio of ones and zeros in the RP matrix The number of elements in the RP matrix for T 1 is equal to N m 1 and the recurrence rate is simply given as 1 N m 1 j k 1 The recurrence rate can also be calculated separately for each diagonal parallel to the line of identity main diagonal The trend of REC as a function of the time distance between these diagonals and the line of identity describes the fading of the recurrences for points further away The rest of the RP measures consider the lengths of the diagonal lines A threshold lmin 2 is used for excluding the diagonal lines formed by tangential motion of the trajectory The maximum line length is denoted lmax and its inverse the divergence 1 DIV 3 25 has been shown to correlate with the largest positive Lyapunov exponent 46 The average diagonal line length on the other hand is obtained as l E LN mean a 3 26 l l lmin dy Kibios HRW Analysis Biosignal Analysis and Medical Imaging Group i Department of Physics eee ae University of Kuopio FINLAND 3 4 Time varying methods 30 where N is the number of length lines The determinism of t
90. t 21 343 350 May 1983 46 L L Trulla A Giuliani J P Zbilut and C L Webber Jr Recurrence quantification analysis of the logistic equation with transients Phys Lett A 223 4 255 260 1996 47 C L Webber Jr and J P Zbilut Dynamical assessment of physiological systems and states using recurrence plot strategies J Appl Physiol 76 965 973 1994 48 E J M Weber C M Molenaar and M W van der Molen A nonstationarity test for the spectral analysis of physiological time series with an application to respiratory sinus arrhythmia Psychophysiol 29 1 55 65 January 1992 49 J P Zbilut N Thomasson and C L Webber Recurrence quantification analysis as a tool for the nonlinear exploration of nonstationary cardiac signals Med Eng Phys 24 53 60 2002 veson lO Beta Department of Physics K bios HAV Analysis Biosignal Analysis and Medical Imaging Group University of Kuopio FINLAND
91. t up preferences window of the software Advanced settings category Preferences eS EOS User information Include in Report Analysis options Y v General results Advanced settings Y Time varying results Paper Size a 210 x 297 mm ASCII File Settings Field delimiter Semicolon hd Decimal symbol Dot Report settings Figure 4 16 Set up preferences window of the software Report settings category Rubies HRY Andiysis Biosignal Analysis and Medical Imaging Group veson lO beta Department of Physics University of Kuopio FINLAND 4 4 Setting up the preferences 50 In addition to the actual analysis options there are various other editable options which have mainly influence on the usability of the software Such options are the Range and Y limit values of the ECG and RR interval axes the RR series display mode button Y limit options for the power spectra and time varying results display mode button color map selection and displayed variable selections The values of these options are preserved in memory and thus any changes made to them will be applied in the future sessions Also the preference directories path from where the data file is searched for and in which the results are saved are preserved in memory The last nine opened data files will also appear in the File menu of the user interface and can be reopened from there All the preferences and preserved options used by the software
92. ted in one page shown in Fig 5 5 Version 0 beta Department of Physics Kibios HRW Analysis Biosignal Analysis and Medical Imaging Group University of Kuopio FINLAND 5 2 Sample run 2 Time varying analysis Time Domain Frequency Dornain Nonlinear VIEW RESULTS Time Varying PNN50 amp w 5 0 00 00 00 00 01 40 00 03 20 00 05 00 00 06 40 Time h min s Options Select variable Time varying computations Time domain s Window width s 60 PNNSO Grid interval s 5 Time Domain Frequency Domain Nonlinear VIEW RESULTS 00 03 20 00 10 00 00 11 40 Time varying spectrum estimation method Spectrogram ag Adaptation coeff 0 1 0 5 0 4 0 3 0 2 T gt 2 E a 3 F 2 irs 0 1 0 00 00 00 00 01 40 00 03 20 00 05 00 00 06 40 Time h min s Options Time varying computations Window width s 60 Grid interval s i 5 Select variable Frequency domain E Time varying spectrum Time Domain Frequency Domain l Nonlinear VIEW RESULTS 00 08 20 00 10 00 00 11 40 Color map Jet z Time varying spectrum estimation method Spectrogram be Adaptation coeff 01 Time Varying Frequency Hz 00 00 00 00 01 40 00 03 20 00 05 00 00 06 40 Time h min s Options Time varying computations Window width s 50 Gridinterval s 5 Select variable Frequency domain ha Time varying spectrum kd c 00 08 20 00
93. the detrending method is selected red lines appear over the RR interval data indicating the removed trend components The smoothness of the removed trend in the smoothness priors method can be adjusted by changing the Lambda value The smoothness priors detrending method can be compared to a high pass filter in which the cutoff frequency is determined from the lambda value bigger lambda corresponds to lower cutoff Since we are now interested in LF and HF frequencies we wish to make sure that the detrending does not remove those frequencies This can be easily done by changing the Lambda value and looking at the FFT spectrum Here we set the Lambda value to 500 The time domain frequency domain and nonlinear analysis results for the selected sam ples can then be viewed in the results view segment Just make sure that the results have been updated check that the Automatic is checked in Apply changes and if not press the Apply button Press the Time domain Frequency domain or Nonlinear button to view the corresponding results At first the results are shown for the first sample To take a look at the results of the second sample press the gt button on the top right corner of the results view segment the text on the left changes to Sample 2 and the second sample in the RR Biosignal Analysis and Medical Imaging Group Department of Physics University of Kuopio FINLAND Kubios HRV Analysis version 2 0 beta 5 1 Sample run 1 General analysi
94. tions on how to manually remove programs from the Add or Remove Programs list is available on the Microsoft support web site at http support microsoft com kbid 314481 PLEASE NOTE THAT MODIFYING THE WINDOWS REGISTRY CAN CAUSE SERIOUS PROBLEMS THAT MAY REQUIRE YOU TO REINSTALL YOUR OPERATING SYSTEM USE THE IN FORMATION PROVIDED AT YOUR OWN RISK 1 4 Software home page The Kubios HRV Analysis version 2 0 beta home page on the web can be found at http kubios uku fi where you can find current information on the software and download possible updates and related material 1 5 Structure of this guide The aim of this guidebook is to help the user to get started with the Kubios HRV analysis It should not however be thought of as being an easy to follow step by step manual but more like a reference material from which you can probably find answers to your problems related to HRV analysis or usability of the software The structure of this guide is as follows dy Kubios HAV Analysis Biosignal Analysis and Medical Imaging Group Department of Physics eon AO a University of Kuopio FINLAND 1 5 Structure of this guide 15 After the overview chapter from where you will find useful information about the system requirements and installation an introduction to heart rate variability is given in Chapter 2 This chapter starts with a short discussion on the control systems of heart rate after which the extraction of heart beat p
95. to install MATLAB Component Runtime on your computer NOTE THIS INSTALLATION SHOULD TAKE ABOUT 5 MINUTES TO COMPLETE WARNING This computer program is protected by copyright law and international treaties Unauthorized duplication or distribution of this program or any portion of it may result in severe civil or criminal penalties and will be prosecuted to the maximum extent possible under the law 11 Next select the destination folder in which the MATLAB Component Runtime should be installed To select the default destination C Program Files MathWorks MATLAB Component Runtime click the Next button If you want to select a different folder click the Browse button You can view available and required disk space on your system by clicking Disk Cost You can also choose whether you want to install the MATLAB Component Runtime for just yourself or others Finally choose Next to continue i MATLAB Component Runtime Select Installation Folder The installer will install MATLAB Component Runtime to the following folder To install in this folder click Next To install to a different folder enter it below or click Browse Folder C Program Files Mathworks MATLAB Component Runtime Install MATLAB Component Runtime for yourself or for anyone who uses this computer O Just me Kibios HAV Analysis Biosignal Analysis and Medical Imaging Group Department of Physics See pena University of Kuopio FINLAND 1 2
96. view segment is disabled and only general analysis will be performed The next thing to do is to select the RR interval samples to be analyzed First change the number of samples value to 2 This will make two samples shown as yellow patches in the RR interval axes Then change the sample ranges to cover the periods or interest as shown in Fig 5 1 The easies way to change the sampled ranges is to edit them with the mouse as described in Section 4 2 2 but the ranges can also be changed by editing the Range values in RR interval series options segment Then check that the Sample analysis type option under the RR axis is set for Single samples Then analysis results are calculated for both samples separately If on the other hand Merge samples is selected then the two samples are first merged into one sample and the analysis results are calculated for this merged sample Since we are now only interested in the changes in LF and HF bands we wish to remove 51 5 1 Kubios HRV Analysis File View Help E f File Info File name Rec date Rec time 15 24 59 Channel label EKG Sampling rate 500 Hz Data length 12 min 17 s RR Interval Series Options Artifact correction Apply Level none Undo Samples for analysis Number of samples Neuroscan_data cnt 20 11 03 HO MARKERS Sample run 1 General analysis 52 DoK 00 02 01 00 02 02 00 02 03 00 02 04 00 02 05 Time h min s 00 02 06 00 02 07 0
97. y low frequency VLF low frequency LF and high frequency HF bands of HRV frequency domain analysis can be adjusted by editing the VLF LF and HF values The default values for the bands are VLF 0 0 04 Hz LF 0 04 0 15 Hz and HF 0 15 0 4 Hz according to 44 The default values for the bands can be restored by pressing the Defaults button The RR interval time series is an irregularly time sampled series as discussed in Section 2 2 and thus spectrum estimation methods can not be applied directly In this software this problem is solved by using interpolation methods for converting the RR series into equidis tantly sampled form As the interpolation method a piecewise cubic spline interpolation is used The sampling rate of the interpolation can be adjusted by editing the Interpolation rate value By default a 4 Hz interpolation is used The spectrum for the selected RR interval sample is calculated both with Welch s peri odogram method FFT spectrum and with an autoregressive modeling based method AR spectrum In the Welch s periodogram method the used window width and window overlap can be adjusted by editing the corresponding value The default value for window width is 256 seconds and the default overlap is 50 corresponding to 128 seconds In the AR spectrum there are also two options that can be selected First the order of the used AR model can be selected The default value for the model order is 16 but the model order
98. ysics eee ae University of Kuopio FINLAND 4 3 Saving the results 46 Table 4 1 Supported file formats to export the report sheet File Format Extension Enhanced Metafile emf Encapsulated Postscript eps Encapsulated Postscript color eps Encapsulated Postscript level 2 eps Encapsulated Postscript level 2 color eps Adobe Illustrator file ai JPEG image file Jpg TIFF image file tif TIFF no compression image file tif Portable Network Graphics file png Portable Document Format pdf Close and Close All commands The Export to Export All to PDF Print All Pages and Close commands are also given as toolbar buttons described above The last command Close All can be used for closing all report sheets simultaneously The Edit menu contains only one option Copy to Clipboard which copies the contents of the corresponding report sheet window to the Windows clipboard This can be used to quickly copying the report sheet as an image into another program The Page menu includes commands for changing for the previous or the next report sheet page Prev page and Next page commands respectively and for changing the sheet by its page number However the Page menu is not shown if only one report sheet window is open 4 3 3 MATLAB MAT file In addition to saving the numeric results into an ASCII text file or saving the report sheets in a PDF file the analysis results can also be saved in a MATLAB MAT file compatible with MATLAB
99. zed Representativeness of these segments in comparison with the whole HRV signal was however questioned in 14 Other methods try to remove the slow nonstationary trends from the HRV signal before analysis The detrending is usually based on first order 23 31 or higher order polynomial 39 31 models In addition this software includes an advanced detrending procedure originally presented in 43 This approach is based on smoothness priors regularization 2 3 1 Smoothness priors based detrending approach Let z RN denote the RR interval time series which can be considered to consist of two components Z Zstat Ztrend 2 1 where Zstat is the nearly stationary RR interval series of interest Ztrena is the low frequency aperiodic trend component and N is the number of RR intervals Suppose that the trend component can be modeled with a linear observation model as trend H0 e 2 2 where H R is the observation matrix 0 R are the regression parameters and e is the observation error The task is then to estimate the parameters by some fitting procedure so that Ztrena H 6 can be used as the estimate of the trend The properties of the estimate depend strongly on the properties of the basis vectors columns of the matrix H in the fitting A widely used method for the solution of the estimate is the least squares method However a more general approach for the estimation of is used here That is the so called

USER'S GUIDE - Biosignal Analysis and Medical Imaging Group

Contents

Download Pdf Manuals

Related Search

Related Contents