Home

Kubios HRV USER'S GUIDE

1. Kubios HRV version 2 2 http kubios uef fi USER S GUIDE June 5 2014 Mika P Tarvainen Ph D kubios softwareQuef fi Biosignal Analysis and Medical Imaging Group BSAMIG http bsamig uef fi Department of Applied Physics http www uef fi sovfys University of Eastern Finland http www uef fi Kuopio FINLAND UNIVERSITY OF EASTERN FINLAND MATLAB Copyright 1984 2012 The MathWorks Inc MATLAB is a registered trademark of The MathWorks Inc Contents 1 Overview Ll System requirements iieii a a e de ed aaa A LLL Windows x e s od ook a Ba ee As eh eee A od ee S L12 UR bok ee ee A RA DR SE eM eR ee a a De ee 4 MES MacOS 2662548 fed due Cae Ge A Be SP hae eR EA eg es 1 2 Installation 3 44 4 4 et0e bobo dee eke ee ee bb eed SG KEP See eh d KA A IN e RR AS MACS e o ae aa A ee a ds de de de ba Se eS E 1 3 Wnimstallation 22 6 dg vrs We had dd o e a a ba ee A o A O A ST R32 o AAN A LS MacOS ee p fas 8 ae Sa Oe ra a eee eH oe ie ai a Software home pace c soe id ems rg ee we ae a AR ws Lb Structure of thiseuldes i vale RSA e o REE SEHD EEN DROS 2 Heart rate variability 2 1 Heart beat period and QRS detection o e e 2 2 Derivation of HRV time series 2 cco ccro cc nm e a a ee ee a 2 3 Preprocessing of HRV time series o et 2 3 1 Smoothness priors based detrending approach o e 3 Analysis methods 3 1 Time domain
2. 4 4 Setting up the preferences 5 Sample run 5 1 Samplerun References Chapter 1 Overview Kubios HRV 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 can also be used e g for animal research Kubios HRV includes all the commonly used time and frequency domain variables of HRV The frequency domain variables are calculated for both nonparametric Fourier transform based and para metric autoregressive modeling based spectrum estimates In addition several nonlinear HRV variables are calculated such as Poincar plot recurrence plot analysis detrended fluctuation analysis approximate and sample entropies and correlation dimension The Kubios HRV heart rate variability HRV analysis software is developed by the Biosignal Analysis and Medical Imaging Group BSAMIG at the Department of Applied Physics University of Eastern Finland Kuopio Finland The first version of the software was published already at the end of 2002 and is described in 33 The current version is the third published version of the software The versions published so far including the current version and a short description of their main features are listed below e Version 1 1 released
3. and was compiled to a deployable standalone application with the MATLAB Compiler 4 17 The MATLAB Compiler Runtime MCR version 7 17 is required for running Kubios HRV and is included in the Kubios HRV installers lUniversity of Eastern Finland has only limited rights to the software These limited rights are governed by a certain license agreement between University of Eastern Finland and The MathWorks Inc 1 1 System requirements 5 1 1 System requirements The system requirements given below should be considered as recommended system requirements The software may work also with lower system specifications but will probably function slower or with reduced usability 1 1 1 Windows e Operating system Microsoft Windows 7 or later e Memory 512 MB of RAM 1024 MB or higher recommended e Disk space about 500 MB e Desktop resolution of 1024x768 or higher e The MATLAB Compiler Runtime 7 17 32 bit version installation 1 1 2 Linux e Operating system a Linux distribution with kernel 2 4 x or 2 6 x and glibc glibc6 2 3 4 and above e Memory 512 MB of RAM 1024 MB or higher recommended e Desktop resolution of 1024x768 or higher e Disk space about 460 MB e The MATLAB Compiler Runtime 7 17 included in the Kubios HRV installer 1 13 Mac OS e Operating system Mac OS X e Memory 512 MB of RAM 1024 MB or higher recommended e Desktop resolution of 1024x768 or higher e Disk space about 460 MB e The MATLAB Compiler Ru
4. 41 a Da Correlation dimension Eq 3 21 17 19 DFA Detrended fluctuation analysis 37 38 El 01 Short term fluctuation slope E a2 Long term fluctuation slope s RPA Recurrence plot analysis 47 9 49 Lmean beats Mean line length Eq 3 26 Lmax beats Maximum line length REC Recurrence rate Eq 3 24 DET Determinism Eq 3 27 ShanEn Shannon entropy Eq 3 28 Kubios HRV 3iosignal Analysis and Medical Imaging Group Department of Applied Physics University of Eastern Finland Kuopio FINLAND Chapter 4 Software description 4 1 Input data formats Kubios HRV supports the following data formats 1 Biopac AcqKnowledge files Biopac Systems Inc acq 2 European data format EDF files edf 3 General data format GDF files gdf 4 ECG ASCII data files txt dat 5 Polar HRM files Polar Electro Ltd hrm 6 Suunto SDF STE files Suunto Ltd sdf ste 7 Garmin FIT files Garmin Ltd fit 8 RR interval ASCII files txt dat 9 Custom ASCII data files txt dat 10 Kubios HRV Matlab MAT files mat The three first ones Biopac ACQ EDF and GDF are binary data formats 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 If the ECG channel cannot be determined or more than one channels are identified as ECG channels the software prompts the user to select
5. HR stroke volume V cardiac output CO total peripheral resistance TPR and arterial blood pressure BP Sinus node Common bundle Bundle branches gt Purkinje Ventricular muscle 1 i 1 T T T T T T T 1 0 100 200 300 400 500 600 700 Time ms Figure 2 2 Electrophysiology of the heart redrawn from 27 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 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 35 36 18 13 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 35 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 fiducia
6. The printout of the ECG signal generated by the software analysis 27 The very 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 Kubios HRV version 2 2 3iosignal Analysis and Medical Imaging Group Department of Applied Physics University of Eastern Finland Kuopio FINLAND 4 2 The user interface 28 Analysis Options Frequency bands VLF Hz 0 0 04 LF Hz 0 04 0 15 HF Hz 0 15 04 Interpolation of RR series Interpolation rate Hz 4 FFT spectrum Window width s 256 Window overlap 50 AR spectrum AR model order 16 Use factorization No Figure 4 7 The analysis options segment of the user interface 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 equidistantly 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 us
7. r plot which is the so called correlation integral The correlation dimension D2 is defined as the limit value D2 m lim lim DEEN 21 r gt 0N gt logr ean In practice this limit value is approximated by the slope of the regression curve logr logC r 19 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 Dz when m is increased In the software a default value of m 10 was selected for the embedding 3 3 7 Recurrence plot analysis Yet another approach included in the software for analyzing the complexity of the time series is the so called recurrence plot RP analysis In this approach vectors uj RR RRj47 RRyz m ajr s G 1 2 N m 1 7 3 22 where m is the embedding dimension and 7 the embedding lag The vectors u then represent the RR interval time series as a trajectory in m dimensional space A recurrence plot is a symmetrical N m 1 r 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 i e RP j k is 1 if the point u on the trajectory is close to point uz That is 1 d uj uk lt r a 0 otherwise 3 23 where d u uz 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 di
8. 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 depo larization 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 t RR 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 29 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 11 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 in
9. 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 right corner of the frequency domain results view The power axes can be selected to have either common or separate upper Kubios HRV Biosignal Analysis and Medical Imaging Group Department of Applied Physics version 2 2 University of Eastern Finland Kuopio FINLAND 4 2 The user interface 29 VIEW RESULTS Time Domain _ Frequency Domain Noninear Time Domain Results Distributions Variable Value Units Mean RR 1222 0 ms STD RR SDNN 47 151 ms Mean HR 49 240 1 min STD HR 2 4467 1 min RMSSD 60 567 ms INNSO 107 pNNSO 43 673 HRV triangular index 10 250 s HR beats min TINN 230 00 ms Set fixed axes limits RR s HR 1 min Calculated from the non detrended selected RR series Frequency Domain Results Variable VLF LF HF LF HF FFT Results Peak Hz 0 039063 0 042969 0 30469 Power ms2 162 82 61826 872 70 Y Show EDR Power 9 8454 37 385 52 770 EDR 0 28 Hz Power n u 41 467 58 533 AR Results Peak Hz 0 039063 0 066406 0 28125 Power ms2 304 37 812 99 975 00 Power 14 547 38 855 46 598 Power n u 45 470 54 530 Spectrum Y limits Common Fix to PSD 87Hz o 8 PSD 87Hz Figure 4 9 The results view segment of the user interface freque
10. 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 different correction levels define thresholds very low 0 45 sec low 0 35 sec medium 0 25 sec strong 0 15 sec very strong 0 05 sec for detecting RR intervals differing abnormally from the local mean RR interval For example the correction level medium will identify all RR intervals which are bigger smaller than 0 25 seconds compared to the local average Furthermore the above correction thresholds are for 60 beats min heart rate for higher heart rates the thresholds are smaller because the variability is expected to decrease when HR increases The corrections to be made on the RR series are displayed on the RR interval axis To make
11. Analysis and Medical Imaging Group Department of Applied Physics version 2 2 University of Eastern Finland Kuopio FINLAND 4 3 Saving the results 32 Page File Edit x GH AALA O gdf_ecg_data gdf 20110325 10112500 z I HRV Analysis Results Page 1 1 RR Interval Time Series Results for a single sample 00 00 22 g e 13 i 1 1 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 CAE a i So 19 0 1 i o2 ben 00 00 00 00 00 50 00 01 40 00 02 30 00 03 20 00 04 10 Time h min s Time Domain Results Distributions Variable Units Value Mean RR ms 1222 0 STD RR SDNN ms 47 2 Mean HR 1 min 49 24 STD HR min 2 45 RMSSD ms 80 6 NNEO count 107 pNN50 5 43 7 RR triangular index 10 250 TINN ms 20 0 Frequency Domain Results FFT spectrum Welch s pe odogram 256 s window with 50 overlap AR Spectrum AR model order 16 not factorzed Tr EDR 0 28 Hz y j kE m 0 02 m 0 02 1 z l amp i a 0 01 a 0 01 E a l i N 0 A o 0 1 0 2 0 3 0 4 0 5 E i E Frequency Hz Frequency Pesk Power Power Power Hz ms nu Band Hz ms nu VLF 0 0 04 Hz 0 0391 163 98 VLF 0 0 04 Hz 0 0391 204 14 5 LF 0 04 0 15 Hz 0 0420 618 37 4 415 LF 0 04 0 15 Hz 0 0664 813 38 9 45 5 HF 0 15 0 4 Hz 0 3047 873 52 8 58 5 HF 0 15 0 4 Hz 0 2813 975 46 6 54 5 Total 1654 Total 2092 LF HF 0 708 LF HF 0
12. Hz 0 0 04 LF Hz 0 04 015 HF Hz 0 15 04 l 1 00 00 05 00 00 10 Undo Find marker 800 lt lt Time h min s RR s 00 11 40 737 00 01 40 Sample Analysis Type Single samples X AO ___ A ee 00 03 20 00 10 00 Range s vew resurs Time Domain Frequencydomain noninear Frequency Domain Results Variable VLF LF FFT Results Peak Hz Power ms2 Power Power n u Spectrum Y limits Common X HF LF HF Fixto 0 04 Y Show EDR EDR 0 28 Hz 0 039063 0 042969 134 34 650 19 10 888 38 403 43 096 0 28516 858 52 50 709 56 904 PSD s Hz AR Results Peak Hz Power ms2 Power Power n u 0 039063 0 070313 293 39 766 37 14 746 38 517 45 179 0 28516 929 93 46 737 54 821 PSD s Hz Figure 5 1 Sample run 1 the Automatic is checked in Apply changes and if not press the Apply button Press the Time domain Frequency domain or Nonlinear buttons to view the corresponding results Note that the results are shown only for one of the samples at a time To take a look at the results of the other sample press the lt or gt button on the top right corner of the results view segment the text on the left changes to indicate which sample s results are shown this sample will also be highlighted in the RR series axis Note that you can force a common Y limit for the spectra of
13. Medical Imaging Group Department of Applied Physics version 2 2 University of Eastern Finland Kuopio FINLAND Chapter 5 Sample run In this chapter we present a sample run of the software The sample run is made for the GDF data file gdf_ecg_data gdf 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 in 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 is observed immediately after standing up which aims to compensate the sudden decrease in blood pressure In the sample run this data file is analyzed by considering the lying and standing periods separately 5 1 Sample run In this sample run we show how to make the time domain frequency domain and nonlinear analysis for the lying and standing periods of the orthostatic 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 The next thing to do is to select the RR interval sa
14. P O Ranta aho and P A Karjalainen Software for advanced HRV analysis Comput Meth Programs Biomed 76 1 73 81 2004 M Pagani Heart rate variability and autonomic diabetic neuropathy Diabetes Nutrition 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 Kubios HRV Biosignal Analysis and Medical Imaging Group Department of Applied Physics version 2 2 University of Eastern Finland Kuopio FINLAND References 44 37 C K Peng S Havlin H E Stanley and A L Goldberger Quantification of scaling exponents and 38 39 40 41 42 43 44 45 46 47 48 49 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 fluctu ation 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 periodic processes in psycho
15. 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 26 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 fre quency 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 re search 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 25 20 and as an early warning sign of diabetic neuropathy 5 34 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 hea
16. in September 2002 This was the first version of the Kubios HRV software which was distributed only for Windows operating systems It included the basic HRV time domain and frequency domain analysis features This version supported only ASCII RR interval data files as input e Version 2 0 released in October 2008 The second version of the software was released for both Windows and Linux operating systems The most significant new features in this version included Suunto SDF STE and Polar HRM data file support RR interval artifact correction options several new nonlinear analysis features and improved user interface e Version 2 1 released in July 2012 The current version is released for Windows and Linux operating systems The main new features compared to previous versions include ECG data support with integrated QRS detection algorithm saving the analysis results in SPSS friendly ASCII file and several usability related improvements e Version 2 2 released in May 2014 The current version is released for Windows Linux and Mac operating systems New features compared to previous version include several improvements for input data support support for Garmin FIT files and EDF annotations was added and Biopac file support updated multiscale entropy MSE computation and several minor functionality and usability modifications The latest version of Kubios HRV has been developed using MATLAB Release 2012a The Math Works Inc
17. 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 The estimated cutoff frequency of the detreding method is also shown next to the Lambda value 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 in such a way that the cutoff frequency is below 0 04 Hz The effect of detrending can also be verified by inspecting how it changes the FFT spectrum Here we set the Lambda value to 500 The time domain frequency domain and nonlinear analysis results for the selected samples can then be viewed in the results view segment Just make sure that the results have been updated check that 37 References 38 By kubios HRV QAMatlablWRV v2 matlab datagdf ecg data gdf file View Help a ECG mv Hz Data length h min s RR Interval Series Options Artifact correction Apply Level none Samples for analysis 2 Sample 1 lt lt OS Start h min s 00 00 29 Length h min s 00 05 00 Remove trend components Method Smoothn priors x Lambda 500 f 0 035Hz Analysis Opti Frequency bands VLF
18. 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 6 1 SD1 z5DSD 3 5 The standard deviation along the line of identity denoted by SD2 on the other hand describes long term variability and has been shown to be related to time domain measures SDNN and SDSD by 6 SD2 2SDNN SDSD 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 3 3 2 Approximate entropy Approximate entropy ApEn measures the complexity or irregularity of the signal 14 41 Large values of ApEn indicate high irregularity and smaller values of ApEn more regular signal The ApEn is computed Kubios HRV 3iosignal Analysis and Medical Imaging Group Department of Applied Physics version 2 2 University of Eastern Finland Kuopio FINLAND 3 3 Nonlinear methods 15 1000 950 F 900 F 850 F ms j 1 800 F RR 750 F 700 F 650 F 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 xz is the line of identity for which RR RRj 1 as follows First a set of length m
19. vectors uz is formed uj RR RRj 1 ia 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 ux max RRjin RRk4n n 0 m 1 3 8 Next for each u the relative number of vectors uz for which d uj ug lt r is calculated This index is denoted with C7 r and can be written in the form nbrof uz d u uz lt r ES N m 1 Vk 3 9 Due to the normalization the value of C7 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 ONG ag 2 In C r 3 10 Finally the approximate entropy is obtained as ApEn m r N P r 97 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 default value of m is set to be m 2 The length N of the 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 s
20. 824 Nonlinear Results Variable Units Value Poincare plot SD1 ms 429 SD2 ms 50 2 Recurrence plot po Mean line length Lmean beats 72 2 Max line Lmax beats 38 z Recurrence rate REC 19 61 Determinism DET 95 19 Shannon Entropy ShanEn 2 701 Other Approximate entropy ApEn 1 011 Sample entropy SampEn 1 848 Detrended fluctuations DFA a1 0 801 5 Detrended fluctuations DFA 2 0 219 06 08 1 12 14 16 18 Correlation dimension D2 3 734 log N beats Results are calculated from the non detrended selected RR series 30 Mar 2011 15 05 11 Kubios HRV version 2 1 Test User Department of Applied Physics University of Eastern Finland Kuopio Finland Figure 4 11 The first report sheet including all the time domain frequency domain and nonlinear analysis results calculated by the software The File menu includes Export All to PDF Print Current Page Print All Pages Close and Close All commands The 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 simul taneously The Edit menu Windows version only contains Copy to Clipboard option 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 o
21. Epos Eeces ESRI Paade reset esse ir HH i ened Hite oc aoe Moase Seese Sease Sesaat ice HH 00 00 31 00 00 32 00 00 33 00 00 34 i 00 00 35 EE 00 00 36 j t HEHH HIEL HHHH Hil hE PE EEE HETEN HHH Poar AE 00 00 39 1 00 00 44 t 00 00 45 00 00 46 00 00 47 00 00 48 00 00 49 00 00 50 00 00 51 j f H i a HE MEE i di E HERET SS A E a o a o Vy HE W W yii YY 7 E iy 00 00 52 00 00 53 00 00 54 00 00 55 00 00 56 00 00 58 00 00 59t HEHEHE HHHH HEHEHEH HHEH H E H 00 01 00 00 01 01 00 04 03 00 01 04 q 00 01 06 09 01 07 j i i i li HE HEERE H e inne DEREN EE E FEH H vy Vy y 00 01 08 00 01 09 00 01 10 goat A HE E eenen ESEE ESGSEE i a HERET PEH ponina 00 01116 00 01 17 00 01 18 00 01 19 oratra 00 01 22 i i l t i 14 FEH F al SHE AE F na 00 01 23 00 01 24 00 01 25 00 01 26 00 01 27 00 91 28 l Seay gibt esas cases overeat peste nn ASE EDEN sent eee SEE 3 00 01 30 00 01 31 i 00 01 32 00 01 38 00 01 34 t 00 01 35 00 01 36 00 01 37 f i i i i H H PEH HIE HEE HH H Hh Ea ee OE W HE H FETES Ht 00 01 38 00 01 39 00 01 40 00 01 41 00 01 42 00 01 43 00 01 44 00 01 45 Time h min s 25 Oct 2011 14 15 56 Kubios HRV version 2 1 Test User Department of Applied Physics University of Eastern Finland Kuopio Finland Figure 4 6
22. HRV data and gives a detailed description of the smoothness priors based detrending approach Kubios HRV Biosignal Analysis and Medical Imaging Group Department of Applied Physics version 2 2 University of Eastern Finland Kuopio FINLAND 1 5 Structure of this guide 7 In Chapter 3 the analysis methods included in the software are described The descriptions of the methods are divided into time domain frequency domain and nonlinear 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 So if you want to learn how to use all the functionalities of the software this is the chapter to read 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 Kubios HRV 3iosignal Analysis and Medical Imaging Group Department of Appli
23. RV 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 abso lute 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 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
24. S Mezzetti A Porta S Cerutti and A Malliani Non linear dynamics and chaotic 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 Modeling 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 Circulation 108 1 110 115 July 2003 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 26 514 519 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 L H Carney and J P Saul Time a
25. a of these devices it should however be noted that the RR intervals must be measured stored in beat to beat If only averaged data e g HR values at every 5 seconds are stored one can not perform HRV analyses In addition Polar Suunto and Garmin file formats a support for plain RR interval ASCII text files is provided The input ASCII file can include RR interval values in one or two column format That is The RR interval values can be given as Type 1 0 759 0 690 0 702 0 712 0 773 Type 2 0 759 0 759 1 449 0 690 2 151 0 702 2 863 0 712 3 636 0 773 So in the second type of input the first column includes the time indexes of R wave detections zero time for the first detection and second column the RR interval values The RR interval values above are given in seconds but millisecond values can also be given In addition to above file formats a custom ASCII file option is also provided Using this option you can import ASCII 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 required details corresponding to your data file e Number of header lines e Column separator tab space comma or semicolon Kubios HRV version 2 2 Biosignal Analysis and Medical Imaging Group Department of Applied Physics University of Eastern Finl
26. agonal 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 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 dimension and lag were selected to be m 10 default value and 7 1 fixed respectively The threshold distance r was selected to be ym SD default value where SD is the standard deviation of the RR time series The selection are similar to those made in 9 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 7 1 is equal to N m gt l and the recurrence rate is simply given as 1 N m gt 1 j k 1 Kubios HRV 3iosignal Analysis and Medical Imaging Group Department of Applied Physics version 2 2 University of Eastern Finland Kuopio FINLAND 3 4 Summary of HRV parameters 19 Time min Figure 3 4 Recurrence plot matrix for HRV time series black 1 and white 0 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 meas
27. alysis and Medical Imaging Group Department of Applied Physics version 2 2 University of Eastern Finland Kuopio FINLAND 4 2 The user interface 26 ECG mv N w t lE tle 00 00 30 00 00 35 00 00 40 00 00 45 Find marker 800 w lt lt Time h min s Range s 15 00 00 00 00 01 40 00 03 20 00 05 00 00 08 40 00 08 20 00 10 00 00 11 40 Time h min s Range s 737 gt Lock Figure 4 5 The data browser segment of the user interface 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 Lock button on the right side of the RR axis slider 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 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 misde
28. am RR RR RR RR RR 1 2 3 4 5 RR interval series with two possible ways of interpolation RR ee a RR RR RR o ee A get 7 t t t t ts Figure 2 3 Derivation of two HRV signals from ECG the interval tachogram middle panel and inter polated RR interval series bottom panel 2 3 1 Smoothness priors based detrending approach Let z R denote the RR interval time series which can be considered to consist of two components Z 2stat 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 Ztrena H0 e 2 2 where H RY 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 2trena H 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 regularized least squares solution 6 arg min z HO Da H0 11 2 3 where A is the regularization parameter and Dg indicates the discrete approxima
29. and Kuopio FINLAND 4 2 The user interface 23 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 s or date time format 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 Finally the software supports also MATLAB MAT files saved by the software itself When you are using Kubios HRV you can save the analysis results also into a Matlab MAT file as described in Section 4 3 3 These result files include all the analysis results and analysis parameters exactly as they where when you saved the results In addition these files include the raw data ECG or RR data Therefore you are able to return to already analyzed data simply by opening the saved MAT file again in Kubios HRV The software will open with the settings that you have used when saving the results e g including the selected analysis samples So this is the simplest way to return to previously analyzed data and perhaps change some analysis parameter and re compute the results In addition the MAT file are also useful for anyone working with Matlab 4 2 The user interface The developed HRV analysis software is operated with a graphical user interface This user interface window is shown in Fig 4
30. 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 in the nat 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 Fig 5 4 Note that you can also save the results in the SPSS friendly formatted text file if you select this option from the software preferences If this option is selected the text file looks like the one shown in Fig 5 5 That is all the parameter values are written in one row The column headers indicate the corresponding parameter and the sample _s1 indicates here the first sample i e from the lying period After the results for the first sample you will find the same results for the seco
31. ctrum AR spectrum Peak frequencies VLF Hz 039063 039063 0 035156 0 039063 LF Hz 0 042969 0 078313 0 089844 0 078125 HF Hz 285156 285156 156258 152344 Absolute powers VLF ms 2 184 3352 293 3944 388 8022 181 0716 LF ms 2 658 1852 766 3733 1102 6417 1132 1685 HF ms 2 858 5197 929 9297 347 5892 242 0337 Figure 5 4 Sample run 1 results saved in an ASCII file Sample meanrr sdnn meanhr sdhr rmssd nn50 pnn50 sdann sdnni hrvtri tinn vlfpeakfft lfpeakfft hfpeakfft vlfpowfft lfp 1 1221 653491 47 148614 49 252546 2 460168 60 538309 108 000000 44 081633 10 250000 225 000000 0 039063 0 042969 2 926 250796 38 813623 65 071712 3 423940 25 735385 13 000000 4 024768 9 818182 190 000000 0 035156 0 089844 0 15 Figure 5 5 Sample run 1 results saved in an SPSS friendly ASCII file Kubios HRV Biosignal Analysis and Medical Imaging Group Department of Applied Physics version 2 2 University of Eastern Finland Kuopio FINLAND References 10 11 12 13 JEN A 15 V X Afonso ECG QRS detection In W J Tompkins editor Biomedical Digital Signal Processing 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 cardiovasc
32. d has a closer agreement with the theory for data with known probabilistic content 21 3 3 4 Multiscale entropy MSE Multiscale entropy MSE is an extension of SampEn in the sense that it incorporates two procedures 8 1 A course graining process is applied to the RR interval time series Multiple course grained time series are constructed for the time series by averaging the data points within non overlapping windows of increasing length 7 where 7 represents the scale factor and is selected to range between T 1 2 20 The length of each course grained time series is N T where N is the number of RR intervals in the data For scale 7 1 the course grained time series is simply the original beat to beat RR interval time series 2 SampEn is calculated for each course grained time series SampEn as a function of the scale factor produces the MSE MSE for scale factor 7 1 returns standard SampEn computed from the original data points 3 3 5 Detrended fluctuation analysis Detrended fluctuation analysis DFA measures the correlation within the signal The correlation is extracted for different time scales as follows 37 First the RR interval time series is integrated k gt RR RR k 1 N 3 15 j 1 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 yn k denote these regression lines Nex
33. d to e g spreadsheet programs such as the Microsoft Excel for further inspection Note that there is also an alternative for the ASCII file format The analysis results can be also saved in an SPSS friendly format where all the parameter values are saved in one row The column headers are named according to the parameter names but in SPSS compatible strings If you have analyzed more than one sample the results for different samples are presented in separate rows one row per analyzed sample This functionality was implemented in order the user s to be able to easily construct SPSS compatible text files for their patient data That is if you are analyzing data of several patients just save the SPSS fiendly text files for each patient and then finally copy paste all the results in one text file with the column headers one row for each patient This text file can then be easily imported to SPSS NOTE you need to select this alternative format from the Preferences under Report settings see Section 4 4 because by default it s not selected 4 3 2 Report sheet The software generates a printable report sheet which present all the analysis results The report sheet shown in Fig 4 11 includes all the time domain frequency domain and nonlinear analysis results The RR interval data and the sample selected for analysis are presented on the two axes on top of both sheet and the analysis results below them If multiple analysis samples
34. draw the solution towards the null space of the regularization matrix D 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 Estat 2 HO I a I ADD ye 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 sta Lz where L I I AD7D3 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 Kubios HRV 3iosignal Analysis and Medical Imaging Group Department of Applied Physics version 2 2 University of Eastern Finland Kuopio FINLAND Chapter 3 Analysis methods In this c
35. e 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 results can be Kubios HRV 3iosignal Analysis and Medical Imaging Group Department of Applied Physics version 2 2 University of Eastern Finland Kuopio FINLAND 4 3 Saving the results 31 written in an ASCIT 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 Software user and data file informations Used analysis parameters Samples selected for analysis Time domain results Frequency domain results Nonlinear results RR interval data and spectrum estimates MPA A The columns of the file are separated with semicolons so that the results could easily be importe
36. ed The spectrum for the selected RR interval sample is calculated both with Welch s periodogram 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 cor responding 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 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 factorization 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 r
37. ed Physics version 2 2 University of Eastern Finland 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 Sympathetic 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 HRJ 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 sinus arrhyth mia 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
38. election for r is r 0 25DNN which is also the default value in this software 3 3 3 Sample entropy Sample entropy SampEn is similar to ApEn but there are two important differences in its calculation 41 21 For ApEn in the calculation of the number of vectors uz for which d u ux lt r also the Kubios HRV Biosignal Analysis and Medical Imaging Group Department of Applied Physics version 2 2 University of Eastern Finland Kuopio FINLAND 3 3 Nonlinear methods 16 vector uj itself is included This ensures that C7 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 Ci r as br of d u up lt Cry nbrof uz d u uz lt r TEF Gua N m Now the value of C7 r will be between 0 and 1 Next the values of C r are averaged to yield 1 N m 1 o ae Ea 2 cr r 3 13 and the sample entropy is obtained as SampEn m r N In C r C r 3 14 The default values set 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 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 an
39. ern Finland Kuopio FINLAND 4 4 Setting up the preferences 36 Report settings User information Ce Advanced settings Advanced settings ll X Figure 4 16 Set up preferences window of the software Report settings category there All the preferences and preserved options used by Kubios HRV are saved in user specific folders Windows XP C Documents and Settings lt username gt Application Data KubiosHRV Windows Vista Windows 7 C Users lt username gt AppData Roaming KubiosHRV Linux nome lt username gt kubioshrv 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 preferences 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 be restored by deleting these files 1Note that the Application Data folder in Windows XP and AppData folder in Windows Vista and 7 are hidden by default and are 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 Kubios HRV Biosignal Analysis and
40. esults are divided into time domain frequency domain and nonlinear 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 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 8 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 x symbol The frequency domain results view shown in Fig 4 9 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 0 Hz to the upper limit of HF band plus 0
41. ether three user menus and seven toolbar buttons The toolbar button icons and their actions are given below Kubios HRV Biosignal Analysis and Medical Imaging Group Department of Applied Physics version 2 2 University of Eastern Finland Kuopio FINLAND 4 3 Saving the results 30 VIEW RESULTS Time Domain Frequency Domain Montinear Nonlinear Results Poincare plot Detrended fluctuations DFA i Value Units Po 42 919 ms Poincare plot SD2 50 245 ms Recurrence plot Lmean 7 2218 beats Recurrence plot Lmax 38 beats Recurrence plot REC 19 612 Recurrence plot DET 95 189 Recurrence plot ShanEn 2 7015 DFA alpha 1 0 80066 DFA alpha 2 0 21881 ApEn 1 0107 SampEn 1 8481 Correlation dimension D2 3 7339 Figure 4 10 The results view segment of the user interface nonlinear results view selected 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 Print results button is for printing the current results without open ing report sheet windows 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 wind
42. fied by adding or removing samples and by changing the start time or length of the sample If more than Kubios HRV 3iosignal Analysis and Medical Imaging Group Department of Applied Physics version 2 2 University of Eastern Finland Kuopio FINLAND 4 2 The user interface 24 lg al ae Se EE Kubios HRV Q Matab HRV 2 matiab datagdtecs cataad I a a File View Help x 69S 8 200 0 ECG mv Hz Data length h min s RR Interval Series Options Artifact correction Apply 00 00 00 mm A 00 00 03 00 00 04 00 00 05 00 00 06 00 To ETT 00 TO 00 00 10 Level none Undo Find marker 800 y lt lt gt gt Time h min s Ra 10 Samples for analysis 4 Sample 1 Start h min s 00 00 21 Length h min s 00 05 00 Remove trend components Method Smoothn priors i Lambda 0 035 Hz 00 00 00 00 01 40 00 03 20 00 05 00 00 08 40 00 08 20 00 10 00 00 11 40 a Range s 737 A f jar Frequency bands vircia o oos vewnesurs Time Domain Frequency Doman noninear LF Hz 0 04 015 HF Hz 0 15 04 Time Domain Results Distributions Variable Value Units Mean RR 1221 7 ms STD RR SDNN 46 586 ms Mean HR 49 252 1 min STD HR 2 4368 1 min RMSSD 60 068 ms INNSO 106 pNNSO 43 443 45 0 HRV triangular index 9 8000 HR beats min TINN 225 00 ms Set fixed axes limits RR s HR 1 min 55 Calculated
43. from the non detrended selected RR series Figure 4 2 The graphical user interface of the developed HRV analysis software RR Interval Series Options Artifact correction Apply Level none i Undo Samples for analysis 1 Sample 1 Start h min s 00 00 21 Length h min s 00 05 00 Remove trend components Method Smoothn priors x Lambda 500 f 0 035 Hz Figure 4 3 The RR interval series options segment of the user interface one sample is selected the analysis can be done either for the single samples separately or by merging the samples into one long sample before analysis This selection is visible under the RR series axis when multiple samples are selected The starting point and length of the samples can also be changed by moving resizing the patch over the RR series as described in Section 4 2 2 This section also describes how to add remove samples to from RR series axes 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 trend can be removed using a method called smoothness priors which was presented in 43 In the smoothness priors method the smoothness of the removed trend can be adjusted by editing the Lambda value The smoothness priors method is basically a time varying high pass filter a
44. h includes the MATLAB Compiler Runtime MCR version 7 17 and the Kubios HRV application bundle First install the MCR on your computer under the MCRinstaller folder run InstallForMacOSX After you have installed the MCR move the Kubios HRV application bundle into Applications on your computer Kubios HRVis then ready to be launched Note that the zip file includes also a Sample Data folder which includes few sample files by which you can use to make test runs with Kubios HRVor to examine the structure of certain Kubios compatible file formats The zip file includes also a run_kubioshrv sh script which you need to run if you are experiencing problems in opening Kubios HRV This script will add the necessary entries of MCR into system path see the readme file for its usage 1 3 Uninstallation 1 3 1 Windows The preferred and the most straightforward way of uninstalling Kubios HRV is to use the auto mated uninstaller The uninstaller can be launched by selecting Uninstall Kubios HRV from the software s Start menu folder the default Start menu folder is Kubios HRV The software can also be uninstalled from the Add or Remove Programs under the Windows Control Panel The uninstaller does not remove your preferences settings These have to be deleted manually from C Users lt username gt AppData Roaming KubiosHRV 1 3 2 Linux Open terminal and change directory to the Kubios HRV install directory Run the command sh uninstall_
45. hapter 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 three categories i e time domain frequency domain and nonlinear 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 N SDNN O RR RR 3 1 1 where RR denotes the value of jth 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 E ARR 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 71 Y RR 41 RR 3 3 j l Another measure calculated from successive RR interval differences is the NN50 which is the number of successive intervals differing more than 50
46. have been selected a report sheet is generated for each sample When Save Results have been selected the report sheet s 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 sheet s will not be displayed but just saved in the selected PDF file If you wish to view the report sheet s and or to export it 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 window includes 7 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 Print button opens a print dialog from which the report sheet can sent to the selected printer Export all pages to PDF file button is for exporting all report sheets into a single PDF file Zoom in button if for zooming in magnifying the report sheet Zoom out button is for zooming out the report sheet PPP D 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 with mouse and drag it to the desired direction Close button is for closing the report sheet B Kubios HRV 3iosignal
47. ided 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 default values Correspondingly the slope az obtained by default from the range 16 lt n lt 64 characterizes long term fluctuations see Fig 3 2 3 3 6 Correlation dimension Another method for measuring the complexity or strangeness of the time series is the correlation dimension which was proposed in 15 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 in the calculation of approximate and sample entropies form length m vectors uj us RR RRj41 RRj m 1 j 1 2 N m 1 3 17 and calculate the number of vectors uz for which d uj ug lt r that is nbr of Luz uz ux lt r Ci r Nomai Vk 3 18 where the distance function d uj uz is now defined as dluz ur Y 40 0 3 19 l 1 Next an average of the term C r is taken 1 N m 1 a 231 2 CP r 3 20 Kubios HRV Biosignal Analysis and Medical Imaging Group Department of Applied Physics version 2 2 University of Eastern Finland Kuopio FINLAND 3 3 Nonlinear methods 18 1 5 D log C r ol log r Figure 3 3 Approximation of the correlation dimension D2 from the log r log C
48. ing the ECG axis view to the beginning of a selected sample on the right 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 6 When clicking on the button for displaying a printout of the ECG recording a popup window will appear in which you can select the range for the ECG to be printed e g the whole recording or the range of the analysed sample In addition you can adjust print speed in mm sec of the ECG in this popup window Once you have defined the range for ECG printout and clicked the OK button 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 7 includes three subcategories Frequency bands Interpo lation of RR series and Spectrum estimation All of these options are concerned with frequency domain Kubios HRV 3iosignal Analysis and Medical Imaging Group Department of Applied Physics version 2 2 University of Eastern Finland Kuopio FINLAND 4 2 The user interface gdf_ecg_data gdf 20110325 10112500 ECG Signal Page 1 4 z 25 mm sec gt 1 i i ie i H H Hei Bt
49. kubios sh and follow instructions given on screen NOTE You may need root privi leges for running the uninstaller if you have installed Kubios HRV as root 1 33 Mac OS Move the installed applications the MCR and Kubios HRV application to trash 1 4 Software home page The Kubios HRV home page on the web can be found at http kubios uef fi where you can find current information on the software and download possible updates and related material Tf you have any trouble or questions regarding the software please check first if your particular problem or question has an answer in the FAQ troubleshooting section at the software homepage 1 5 Structure of this guide The aim of this guide is to help the user to get started with Kubios HRV 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 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 periods is discussed and the derivation of HRV time series is described The rest of the chapter is focused on the preprocessing of
50. l data Statistics Time domain analysis results Frequency Frequency domain analysis results NonLinear Nonlinear analysis results The variable names of the different fields are more or less self descriptive and are not documented 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 12 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 options category shown in Fig 4 13 includes some basic analysis option
51. l 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 required to be 1 2 ms and thus the sampling frequency of the ECG should be at least 500 1000 Hz 44 If the sampling frequency of the Kubios HRV Biosignal Analysis and Medical Imaging Group Department of Applied Physics version 2 2 University of Eastern Finland Kuopio FINLAND 2 2 Derivation of HRV time series 10 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 31 The distortion of the spectrum is even bigger if the overall variability in heart rate is small 39 The estimation accuracy can however be improved by interpolating the QRS complex e g by using a cubic spline interpolation 10 or some model based approach 4 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 42 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 successive R wave occurrence times That is the n th RR interval is obtained as the difference between the R wave occurrence times RR
52. methods e x saa eao oe Sa da e o eee e e eel es 3 2 Frequenecy domain methods s ce scini e000 442 ea eee ed aaa A 3 3 Nonlinear methods iras heh eee we eee es a a ee ee Re Ocul Pomicare plot a srep ieuane aema eee ee eek eb h A a ee Se 3 3 2 Approximat entropy o s sa beet bv a eb eee e bbb ede a be dee Jiao SAMPE ENIO e lt a suda apar SOAR eS hae eee SRE Rae 3 3 4 Multiscale entropy MSE 2 2 2 00 0 eee ee eee 3 3 5 Detrended fluctuation analysis e o 3 3 6 Correlation dimension Li BI aa a ee 3 3 7 Recurrence plot analysis ce se e wata ae kale a e 3 4 Summary of HRV parameters a ca secre kagad an ms dd A Ra 4 Software description 41 Tnp tdata formats s ss se a 62 4 rea bbe dab awe ee bbe a Pe ol e gae goi a 42 Thenserinterface r 44 4 ea aed ae ee ha ladrar BAS BEE Ea 42 1 RR interval series options e daaa 4 4 428 Se be eee ee ae Aes 4 2 2 Data browser oca demas ERD Ye ae we eae oe aed aoe 423 Analysis Options e sa w fous deg A ee as bea Rm eK a a he GE we EA dod 4 24 Results View 22 4544 04 00 beet ada ea be eee Gee hea a ae a 42 5 Menus and toolbar buttons lt lt ss ra sc 46 laa a a ee A ew eS 4 3 Saving the results sas o sor a A ae a a ee ee 43 1 ASCE 24 4 gee eS e PEE eee a ee REE RS ODS wee aS ered ASO Reportssheet s se rre 00848 EAS REM See a ew ewe 43 3 Matlab MAT file s s uk dic een e Bh wd eee ge Baa cha J AAAA AA OOA O en ata A
53. mples to be analyzed First add a new sample to the RR interval axes because we want to analyze both the lying and standing periods To do this you can simply right click the RR axes press Yes to the Add sample popup window and OK to verify the sample properties Now you will have 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 samples 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 Start and Length 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 the low est 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 the detrending method is selected red lines appear over the RR
54. ms or the corresponding relative amount NN50 NN50 pNN50 Vol 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 x 100 3 4 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 13 3 3 Nonlinear methods 14 interval series is converted to equidistantly sampled series by interpolation methods prior to PSD estima tion 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 28 The advantage of FFT based methods is the simplicity of implementation while the AR spectrum yields improved resolution especially for short samples Another property of AR spectrum that has made it popular in H
55. narity test for the spectral anal ysis of physiological time series with an application to respiratory sinus arrhythmia Psychophysiol 29 1 55 65 January 1992 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 Kubios HRV Biosignal Analysis and Medical Imaging Group Department of Applied Physics version 2 2 University of Eastern Finland Kuopio FINLAND
56. ncy domain 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 both spectra are displayed in tables below the corresponding spectrum axes If ECG is measured an estimate of the respiration frequency is also computed This estimate i e electrocardiogram derived respiration EDR is shown as a vertical line in both spectrum estimates The EDR value is also shown on the right below the spectrum Y limit options The nonlinear results view shown in Fig 4 10 displays all the calculated nonlinear variables 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 2 respectively are indicated for details see Section 3 3 5 4 2 5 Menus and toolbar buttons The user menus and toolbar buttons are located on the upper left hand corner of the user interface There are all tog
57. nd frequency domain methods for heart rate variability analysis a methodological comparison Psychophysiol 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 Circulation 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 Prentice Hall International 1987 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 electrocar diogram for spectral 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
58. nd its cut off frequency can be adjusted with the Lambda parameter the bigger the value of Lambda the smoother is the removed trend The estimated cut off frequency for the given Lambda value is presented next to the Lambda value edit box In addition the trend to be Kubios HRV Biosignal Analysis and Medical Imaging Group Department of Applied Physics version 2 2 University of Eastern Finland Kuopio FINLAND 4 2 The user interface 25 00 01 40 00 10 00 Value Units 790 44 ms 37 626 ms 77 136 1 min 32 046 1 min 38 817 ms 2 3715 6 4364 875 ms Value 790 83 ms 24 594 ms 76 219 1 min 3 6300 1 min 18 269 ms 1 9763 6 1498 165 00 ms Calculated from the non detrended selected RR series 1 1 00 01 40 00 10 00 Range s 3600 80 70 2 HR beats min mam J b Figure 4 4 Artifact correction a the artifact corrected series is visualized on top of the raw RR interval series b Corrected RR interval series 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 5 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 Kubios HRV Biosignal An
59. nd sample _s2 Kubios HRV version 2 2 dy Biosignal Analysis and Medical Imaging Group Department of Applied Physics University of Eastern Finland Kuopio FINLAND References File Edit Page S MIRAR IP HRV Analysis Results RR Interval Time Series 00 00 25 gdf_ecg_data gdf 20110325 10112500 Page 1 2 Results for single samples sample 1 2 00 02 30 Time h min s Time Domain Results Variable Units Mean RR ms STD RR SDNN ms Mean HR min STD HR t min RMSSD ms NN5O count pNNSO RR triangular index TINN ms Frequency Domain Results 0 04 FFT spectrum Welch s periodogram 150 s window with 50 overlap Distributions 0 03 PSD s4Hz 0 04 50 55 HR beats min AR Spectrum AR model order 16 not factorized o 8 PSD 5 Hz o 8 LF 0 04 0 15 Hz HF 0 15 0 4 Hz Total LF HF Nonlinear Results Variable Poincare plot SD1 SD2 Recurrence plot Mean line length Lmean Max line length Lmax Recurrence rate REC Determinism DET Shannon Entropy ShanEn LF 0 04 0 15 Hz HF 0 15 0 4 Hz Total LF HF 0 0391 0 0703 0 2852 0 824 Detrended fluctuations DFA we 068 08 1 12 14 16 18 fog N beats Results are calculated from the non detrended selected RR series 31 Mar 2011 10 02 25 Test User Kubios HRV version 2 1 Department of Applied Physics University of Eastern Finland Kuopi
60. ntime 7 17 installation 1 2 Installation 1 2 1 Windows Make sure that you have administrative privileges you will need them to install Kubios HRV In order to install Kubios HRV on a Windows computer you need to first install the MATLAB Compiler Runtime MCR version 7 17 on your computer You can find the MCR installer from Kubios HRV home page http kubios uef fi After you have installed the MCR run the Kubios HRV installer file Follow the instructions given in the setup wizard to complete installation You can launch the Kubios HRV by selecting it from the created Start Menu folder or by clicking the Desktop icon if created Please note that the starting of Kubios HRV also starts the MATLAB Compiler Runtime and may take some time especially with older computers 1 2 2 Linux Run the Kubios HRV installer by typing sh KubiosHRV linux 2 2 x86 run in the terminal and follow the instructions given on screen The Kubios HRV installer also includes the MATLAB Compiler Runtime NOTE If you run the installer as root Kubios HRV will be installed on your computer system wide for all users If you want to install Kubios HRV only for yourself run the installer as local user Kubios HRV 3iosignal Analysis and Medical Imaging Group Department of Applied Physics version 2 2 University of Eastern Finland Kuopio FINLAND 1 3 Uninstallation 6 1 23 Mac OS Download the zip file from Kubios HRV home page http kubios uef fi whic
61. o Finland Figure 5 2 Sample run 1 results for the lying period of the orthostatic test Kubios HRV version 2 2 Biosignal Analysis and Medical Imaging Group Department of Applied Physics University of Eastern Finland Kuopio FINLAND 39 References File Edit Page 40 G N AALA O RR Interval Time Series Variable Mean RR STD RR SDNN Mean HR STD HR RMSSD NN50 pNN5O RR triangular index TINN 0 04 0 03 0 02 PSD s Hz Frequency LF 0 04 0 15 Hz HF 0 15 0 4 Hz Total LF HF Nonlinear Results 31 Mar 2011 10 02 25 Time Domain Results Units ms ms 1 min min ms count ms Frequency Domain Results FFT spectrum Welch s periodogram 150 s window with 50 overlap AR Spectrum AR model order 16 not factorized gdf_ecg_data gdf 20110325 10112500 HRV Analysis Results Page 2 2 Results for single samples sample 2 2 00 11 52 Distributions 0 8 0 85 09 095 1 1 05 1 1 5 6 6 70 RR s HR beats min 0 04 PSD 64 Hz o o 8 8 Frequency Band 0 0281 LF 0 04 0 15 Hz 0 0781 HF 0 15 0 4 Hz 0 1823 Total LF HF Poincare Plot 0 05 0 0 05 0 1 0 15 0 15 0 1 005 0 0 05 01 06 08 1 12 14 16 18 RR s log n beats Results are calculated from the non detrended selected RR series Kubios HRV version 2 1 Department of Applied Physics University of Eastern Finland Kuopio Finland Figure 5 3 Sam
62. ow 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 infor mation Also the Kubios HRV End User License Agreement can be viewed in the about dialog Y 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 current analysis have not been saved user is prompted to do so bAt amp P All the above actions are also available on the user menus The File menu includes Open Save Results Save Results As Print Results 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 fil
63. physiological 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 Westerhof editors The Physics of Heart and Circulation pages 101 120 Institute of Physics Publishing Bristol 1993 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 2002 Task force of the European society of cardiology and the North American society of pacing and electrophysiology Heart rate variability standards of measurement physiological interpretation and clinical use Circulation 93 5 1043 1065 March 1996 N V Thakor J G Webster and W J Tompkins Optimal QRS detector Med Biol Eng Comput 21 343 350 May 1983 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 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 E J M Weber C M Molenaar and M W van der Molen A nonstatio
64. ple run 1 results for the standing period of the orthostatic test Kubios HRV version 2 2 Biosignal Analysis and Medical Imaging Group Department of Applied Physics University of Eastern Finland Kuopio FINLAND References 41 version 2 1 released May 2811 Analyzed by Test User File name Q Matlab HRV_v2_matlab datagdf_ecg_data gdf Measurement date 20110325 10112508 File type GDF Channel label ECG Data length e8 13 12 h min s Measurement rate 508 Hz Parameters Number of samples 2 Detrending method Smoothn priors lambda 5ee Frequency bands VLF 0 04 Hz LF 0 04 0 15 Hz HF 0 15 0 4 Hz Interpolation rate 4 Hz Points in frequency domain 256 points Hz FFT spectrum options Window width 158 s Window overlap 50 X AR spectrum options AR model order 16 Use factorization No RR Interval Samples Selected for Analysis A Sample 1 Sample 2 Sample limits s 5 29 329 412 712 Sample Analysis Type Single samples RESULTS FOR SINGLE SAMPLES SAMPLE 1 Time Domain Results Statistical parameters Mean RR ms 1221 6535 STD RR ms 47 1436 Mean HR 1 min 49 2525 STD HR 1 min 2 4602 RMSSD ms 60 5383 NN58 count 108 PNNSO 44 80816 SDANN ms SDNN index ms Geometric parameters RR tri index 18 258008 TINN ms 225 0000 Us us us us us us us us us Us ws us us us ws ue us ue 208 2000 Frequency Domain Results FFT spectrum AR spectrum FFT spe
65. r the next report sheet page and for changing the sheet by its page number However the Page menu is not shown if only one report sheet window is open Kubios HRV Biosignal Analysis and Medical Imaging Group Department of Applied Physics version 2 2 University of Eastern Finland Kuopio FINLAND 4 4 Setting up the preferences 33 4 3 3 Matlab MAT file In addition to saving the numeric results into an ASCIT text file or saving the report sheet s in a PDF file the analysis results can also be saved in a MATLAB MAT file compatible with MATLAB 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 Neuroscan 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 interva
66. roup Department of Applied Physics version 2 2 University of Eastern Finland Kuopio FINLAND 4 4 Setting up the preferences 35 Advanced settings User information Analysis options User information Analysis options Advanced settings Figure 4 15 Set up preferences window of the software Advanced settings category IT 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 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 e g the Range and Y limit values of the data axis and various visualization options The values of these options are preserved in memory and 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 Kubios HRV Biosignal Analysis and Medical Imaging Group Department of Applied Physics version 2 2 University of East
67. rt 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 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 nervous 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 activity in the ECG compared to SA node firing is the P wave resulting from atrial depolarization see Fig 2 2 and thus 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 depolarization Therefore the heart beat period is commonly evaluated as the time difference between 2 1 Heart beat period and QRS detection 9 Vasomotor sympathetic AV Sympa thetic Sympa R thetic S a BPa Sympa thetic Figure 2 1 The four baroreflex pathways redrawn from 42 Variation in venous volume AV left ventricular contractility VC sympathetic and parasympathetic vagal control of heart rate
68. s 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 In addition the analysis options category includes RR interval samples RR interval detrending 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 Figa 4 14 and 4 15 is divided into two pages The first page includes QRS detection and 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 e g for some animal data in which case the prior guess for the RR interval value should be fixed to the supposed value The spectrum estimation options include one additional option compared to those described in Section 4 2 3 i e points in frequency domain option The points in frequency domain is given as points Hz and corresponds by default to the window width of the FFT spectrum If spectrum interpolation is desired the points in frequency domain can be increased The second page of the Advanced settings category includes options for nonlinear analysis The em bedding dimension m and the tolerance value r used in for the computation of Approximate entropy ApEn and Sample entropy SampEn can be modified Note that
69. t 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 1 N F n 4 57 2 lk yn k 3 16 k 1 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 relationship 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 Kubios HRV Biosignal Analysis and Medical Imaging Group Department of Applied Physics version 2 2 University of Eastern Finland Kuopio FINLAND 3 3 Nonlinear methods 17 1 2f log F n 1 4 1 6t 1 8 l i 06 08 1 12 14 16 1 8 logn Figure 3 2 Detrended fluctuation analysis A double log plot of the index F n as a function of segment length n a and ag are the short term and long term fluctuation slopes respectively a 1 5 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 div
70. tected 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 5 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 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 You can also add a new sample to a specific location in the RR series by right clicking the RR axis The new sample will start from the clicked time instant and the length of the new sample is by default same as the previous sample After right clicking the RR axis a small popup window opens in which the sample start time and length can be accepted modified When more than one samples are selected a sample can be removed by right clicking it with the mouse In addition the data browser segment includes buttons for displaying a printout of the ECG recording on the right hand side of the ECG axis mov
71. ted by the software version 2 2 Measure Units Description References RR ms The mean of RR intervals STD RR SDNN ms Standard deviation of RR intervals Eq 3 1 s HR 1 min The mean heart rate 3 STD HR 1 min Standard deviation of intantaneous heart rate values E RMSSD ms Square root of the mean squared differences between successive RR intervals Eq A 3 3 NN50 Number of successive RR interval pairs that differ more than 50 ms E pNN50 NN50 divided by the total number of RR intervals Eq 3 4 E HRV triangular The integral of the RR interval histogram divided by the height of the histogram index 44 TINN ms Baseline width of the RR interval histogram 44 a Peak frequency Hz VLF LF and HF band peak frequencies 3 Absolute power ms Absolute powers of VLF LF and HF bands E Relative power Relative powers of VLF LF and HF bands 4 VLF VLF ms total power ms x 100 a LF LF ms total power ms x 100 2 HF HF ms total power ms x 100 2 Normalized power n u Powers of LF and HF bands in normalized units 7 LF n u LF ms total power ms VLF ms E HF n u HF ms total power ms VLF ms LF HF Ratio between LF and HF band powers SD1 SD2 ms The standard deviation of the Poincar plot perpendicular to SD1 and along SD2 the line of identity 6 7 ApEn Approximate entropy Eq 3 11 41 14 SampEn Sample entropy Eq 3 14
72. terpolation 29 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 12 This approach relies on the generally accepted integral pulse frequency modulator IPFM which aims to model the neural modulation of the SA node 42 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 11 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 HRV 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 Alternativel
73. 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 above binary formats Kubios HRV supports also ASCII ECG data which must be given in the form 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 these ECG file support options Kubios HRV supports three RR interval file formats as described below Kubios HRV supports the following RR interval file formats First of all data of three commonly used heart rate monitor manufacturers are supported These are SUUNTO SDF STE POLAR HRM 21 4 1 Input data formats 2 3310 2 6430 2 8990 3 0360 3 0330 2 8720 2 5310 2 0450 1 4750 0 9050 Figure 4 1 The interface for importing customized ASCII data files into the software 22 files and GARMIN FIT files When analyzing dat
74. 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 cause distortion into the analysis results Note that 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 An example of artifact correction can be seen in Fig 4 4 In this case the analyzed RR interval sample includes two clear artifacts In order to remove these artifacts a medium level correction was selected The effect of correction can be verified from the user interface i e the corrected series is displayed on top of the raw RR interval series as can be seen from Fig 4 4 a The percentage of corrected beats is also displayed on the right side of the RR series axis The correction level can be accepted and the correction performed by pressing the Apply button The corrected series is shown in Fig 4 4 b As can be seen from Figs 4 4 a and b the correction of just few artifacts has a very significant effect on the time domain analysis results Thus even single artifacts should always be taken care of prior to HRV analysis In the Samples for analysis options the part s of the RR interval series to be analyzed can be modi
75. 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 nonlinear mechanisms are involved in the genesis of HRV The nonlinear properties of HRV have been analyzed using measures such as Poincar plot 6 7 approximate and sample entropy 41 14 detrended fluctuation analysis 37 38 correlation dimension 17 19 and recurrence plots 47 46 49 During the last years the number of studies utilizing such methods have increased substantially 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 6 The standard deviation of the points perpendicular to
76. the tolerance value is adjusted in relation to the standard deviation of the RR interval data Next limits of the short term N1 and long term fluctuations used in the Detrended fluctuation analysis DFA can be modified Finally the Kubios HRV 3iosignal Analysis and Medical Imaging Group Department of Applied Physics version 2 2 University of Eastern Finland Kuopio FINLAND 4 4 Setting up the preferences 34 Figure 4 13 Set up preferences window of the software Analysis options category embedding dimension used both in the computation of the Correlation dimension D2 and in the Recur rence plot analysis RPA and the threshold level used in RPA can be modified For more information on the meaning of these different options see Section 3 3 The Report settings category shown in Fig 4 16 includes the following options The paper size of the report sheet can be changed between A4 210x297 mm and Letter 8 5x 11 inch size The default paper size is A4 In addition the field delimiter and decimal point used when saving the results in an ASCII file can be selected Here you can also select to save the results in an SPSS friendly format described in Section 4 3 1 The Custom Print Command option allows the use of an external program to print the report sheets in PostScript format All modifications for the preferences are saved by pressing the OK button Note that the OK button Kubios HRV Biosignal Analysis and Medical Imaging G
77. tion 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 Hz 2 4 and the estimate for the trend which is to be removed as Serena HO 2 5 Kubios HRV 3iosignal Analysis and Medical Imaging Group Department of Applied Physics version 2 2 University of Eastern Finland Kuopio FINLAND 2 3 Preprocessing of HRV time series 12 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 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 frequencies are 0 213 0 145 0 101 0 063 0 045 0 021 and 0 010 times the sampling frequency 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 RY NY is used In this case the regularization part of 2 3 can be understood to
78. ular 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 Na garaja 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 T Bragge M P Tarvainen P O Ranta aho and P A Karjalainen High resolution QRS fiducial point corrections in sparsely sampled ECG recordings Physiol Meas 26 5 743 751 2005 H J Braune and U Geisen rfer Measurement of heart rate variations influencing factors 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 IEEE Trans Biomed Eng 48 11 1342 1347 November 2001 S Carrasco M J Cait n R Gonz lez and O Yanez Correlation among Poincar plot indexes and time and frequency domain measures of heart rate variability J Med Eng Technol 25 6 240 248 November December 2001 M Costa A L Goldberger and C K Peng Multiscale entropy analysis of biological signals Physical Rev E 71 021906 2005 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 Daskalo
79. ures consider the lengths of the diagonal lines A threshold lin 2 is used for excluding the diagonal lines formed by tangential motion of the trajectory The maximum line length is denoted Imax and its inverse the divergence 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 Pan LN Imean Bas N 3 26 l l lmin where N is the number of length lines The determinism of the time series is measured by the variable DET Uli M 3 27 Eici RP Finally the Shannon information entropy of the line length distribution is defined as lmax ShanEn y n Inn 3 28 l lmin where n is the number of length lines divided by the total number of lines that is Ni lax Ny U Lmin 3 29 n 3 4 Summary of HRV parameters The presented time domain frequency domain and nonlinear 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 Kubios HRV Biosignal Analysis and Medical Imaging Group Department of Applied Physics version 2 2 University of Eastern Finland Kuopio FINLAND 3 4 Summary of HRV parameters 20 Table 3 1 Summary of the HRV measures calcula
80. v and I Christov Improvement of resolution in measurement of electrocardiogram 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 IEEE Trans Biomed Eng 37 1 85 98 January 1990 Y Fusheng H Bo and T Qingyu Approximate entropy and its application in biosignal analy sis 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 42 References 43 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 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
81. y if the amount of artifact free data is insufficient proper interpolation methods can be used to reduce these artifacts see e g 22 23 30 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 analyzed Representativeness of these segments in comparison with the whole HRV signal was however questioned in 16 Other methods try to remove the slow nonstationary trends from the HRV signal before analysis The detrending is usually based on first order 24 32 or higher order polynomial 40 32 models In addition this software includes an advanced detrending procedure originally presented in 43 This approach is based on smoothness priors regularization Kubios HRV 3iosignal Analysis and Medical Imaging Group Department of Applied Physics version 2 2 University of Eastern Finland Kuopio FINLAND 2 3 Preprocessing of HRV time series 11 Derived RR intervals RR RR RR RR RR lt gt a gt a gt a gt a gt ty t t t t ts RR interval tachogr

Kubios HRV USER'S GUIDE

Contents

Download Pdf Manuals

Related Search

Related Contents