Docstoc

Empirical Mode Decomposition Analysis of Heart Rate Variability

Document Sample
Empirical Mode Decomposition Analysis of Heart Rate Variability Powered By Docstoc
					                                                         (IJCSIS) International Journal of Computer Science and Information Security,
                                                         Vol. 8, No. 6, September 2010




                  Empirical Mode Decomposition Analysis of Heart Rate Variability
                                            C.Santhi.M.E., Assistant Professor,
             Electronics and Communication Engineering, Government College of Technology, Coimbatore-641 013
                                                        N.Kumaravel Ph.D
                                             Professor, Head of the Department,
                         Electronics and Communication Engineering, Anna University,Chennai-600 025.

Abstract                                                                power in the LF and HF bands (LF/HF) provides the
           The analysis of heart rate variability (HRV)                 measure of cardiac sympathovagal balance. Empirical Mode
demands specific capabilities not provided by either                    Decomposition (EMD) retains the intrinsic nonlinear
parametric or nonparametric spectral estimation methods.                nonstationary property of the signal. Any intrinsic timescale
Empirical mode decomposition (EMD) has the possibility of               derived from the signal is based on the local characteristics
dealing with nonstationary and nonlinear embedded                       timescale of the signal [2-4]. EMD carries out layer upon
phenomena, for a proper assessment of dynamic and                       layer sifting and obtains ordered array components from
transient changes in amplitude and time scales of HRV                   smallest scale (highest frequency) to largest scale (lowest
signal. In this work EMD and a non-linear curve fitting                 frequency) [4]. Empirical mode decomposition has the
technique are used to study half an hour HRV signal and its             possibility of dealing with nonstationary and nonlinear
intrinsic mode function obtained from 20 healthy young                  embedded phenomena, and owing to its suitability for a
control subjects, 20 healthy old control subjects and 20                proper assessment of the dynamic and transient changes in
subjects with long term ST. The intrinsic oscillations are              amplitude and in frequency of the HRV components [2& 3].
measured by means of its meanperiod and variance.                                 Application of EMD to half an hour HRV data
Significant meanperiod reduction is observed in the intrinsic           yields nine intrinsic mode functions (IMFs). The first scale
time scales of healthy old control subjects and subjects with           represents the highest frequency or the shortest period
long term ST, which is used to classify the three groups of             component of the signal. The second scale represents the
HRV signal with high sensitivity and specificity. The                   lower frequency or the longer period component of the
estimated slope using the non-linear curve fitting technique            signal. Similarly the last IMF represents the lowest time
represents the flexibility of the cardiovascular system. The            scale present in the HRV signal. The first two scales contain
main advantage of this method is it does not make any prior             more than 85% of total signal power. The meanperiod and
assumption about the HRV signal being analyzed and no                   variance of IMFs are computed as time domain measures.
artificial information is introduced into the filtering method.         The variance of IMF decreases exponentially with respect to
                                                                        increasing timescales (meanperiods). Using nonlinear curve
Index Terms- Empirical Mode Decomposition, Heart Rate                   fitting technique the IMFs variations are estimated. The
Variability, Intrinsic Mode Functions, RR intervals,                    estimated parameter represents the flexibility of the
nonlinear curve fitting.                                                cardiovascular system.The methodology is applied to HRV
                                                                        signal obtained from 20 healthy young control subjects, 20
1. Introduction
                                                                        healthy old control subjects and 20 subjects with long term
Over the last 20 years there has been widespread interest in            ST. The intrinsic time scale of IMF 2 classifies the three
the study of variations in the beat-to-beat interval of heart           groups HRV signal with high sensitivity and specificity.
known as heart rate variability (HRV) or RR interval
                                                                        2. Empirical Mode Decomposition (EMD)
variations. HRV has been used as a measure of mortality
primarily with patients who have undergone cardiac                                EMD is a procedure oriented adaptive method for
surgery. Clinical depression strongly associated with                   decomposing non-linear non-stationary signals. The
mortality with such patients may be seen through a decrease             components resulting from EMD are called Intrinsic Mode
in HRV [1]. HRV is a non invasive measure of autonomic                  Functions (IMFs) [2]. The IMFs are amplitude frequency
nervous system balance. Heart rate is influenced by both                modulated intrinsic signals. The IMF’s represents the
sympathetic and parasympathetic (vagal) activities of ANS.              oscillatory modes imbedded in the signal. It should satisfies
The influence of both branches of the autonomic nervous                 the following two conditions. 1. In the whole data set the
system (ANS) is known as sympathovagal balance reflected                number of extrema’s and the number of zero crossings must
in the RR interval changes. A low frequency (LF)                        be either equal or differ by at most one. 2. At any point the
component provides a measure of sympathetic effects on the              mean value of the envelope defined by the local minima and
heart and generally occurs in a band between 0.04 Hz and                the envelope defined by the local maxima is zero.
0.15 Hz. A measurement of the influence of the vagus nerve
in modulating the sinoatrial node can be made in the high
frequency band (HF) loosely defined between 0.15 and 0.4
Hz known as respiratory sinus arrhythmia (RSA), and is a
measure of cardiac parasympathetic activity. The ratio of




                                                                  255                               http://sites.google.com/site/ijcsis/
                                                                                                    ISSN 1947-5500
                                                                                                                                                                                                                                             (IJCSIS) International Journal of Computer Science and Information Security,
                                                                                                                                                                                                                                             Vol. 8, No. 6, September 2010

                  1.8
                                                  Heart rate variability
                                                                                                                           0.2
                                                                                                                             0
                                                                                                                          -0.2
                                                                                                                                 0                200         400
                                                                                                                                                                          Intrinsic mode functions



                                                                                                                                                                             600            800           1000           1200      1400
                                                                                                                                                                                                                                                            Step 5: Check h(t) for the conditions of an Intrinsic Mode
                                                                                                                                                                                                                                                                      Functions. [2]
                  1.7                                                                                                      0.2
                                                                                                                             0
                                                                                                                          -0.2
                                                                                                                                 0                200         400            600            800           1000           1200      1400
                  1.6                                                                                                      0.2
                                                                                                                             0
                                                                                                                          -0.2
                                                                                                                                 0                200         400            600            800           1000           1200      1400
                                                                                                                           0.2



                                                                                                                                                                                                                                                               If h(t) is an IMF compute residue r(t)=x(t)-h(t) and again
                  1.5                                                                                                        0
                                                                                                                          -0.2
                                                                                                                                 0                200         400            600            800           1000           1200      1400
                                                                                                                           0.2
                  1.4                                                                                                        0
  rr intervals




                                                                                                                          -0.2



                                                                                                                                                                                                                                                            the process is repeated to extract the next IMF. If h(t) is not
                                                                                                                                 0                200         400            600            800           1000           1200      1400
                                                                                                                           0.1
                  1.3                                                                                                        0
                                                                                                                          -0.1
                                                                                                                                 0                200         400            600            800           1000           1200      1400
                                                                                                                           0.1
                                                                                                                             0



                                                                                                                                                                                                                                                            an IMF x(t) is replaced with h(t) and the procedure is
                  1.2                                                                                                     -0.1
                                                                                                                                 0                200         400            600            800           1000           1200      1400
                                                                                                                          0.05
                                                                                                                             0
                  1.1                                                                                                    -0.05
                                                                                                                                 0                200         400            600            800           1000           1200      1400
                                                                                                                          0.01



                                                                                                                                                                                                                                                            repeated from step 1. Fig.6 shows all IMFs of the signal
                                                                                                                             0
                    1                                                                                                    -0.01
                                                                                                                                 0                200         400            600            800           1000           1200      1400
                                                                                                                          1.4
                                                                                                                          1.3
                  0.9



                                                                                                                                                                                                                                                            x(t).
                        0    200     400           600         800                1000          1200          1400               0                200         400            600            800           1000           1200      1400
                                                     beat number




Fig.1.RR interval signal                                                                                                Fig.2. Intrinsic Mode Functions                                                                                                           The process ends when the range of residue is below a
                                                                                                                          0.3
                                                                                                                                                                            Detrended HRV signal
                                                                                                                                                                                                                                                            predetermined level or the residue has a monotonic trend. In
                                                                                                                                                                                                                                                            order to guarantee that the IMF components retains enough
                                                 Reconstructed signal
                  1.8


                                                                                                                          0.2
                  1.7



                  1.6                                                                                                     0.1
                                                                                                                                                                                                                                                            physical sense in both amplitude and frequency
                                                                                                                                                                                                                                                            modulations, the sifting process is stopped by limiting the
                  1.5
                                                                                                                            0


                  1.4
  rr intervals




                                                                                                                         -0.1

                  1.3
                                                                                                                         -0.2                                                                                                                               size of standard deviation(SD) which is computed from two
                                                                                                                                                                                                                                                            consecutive sifting results.
                  1.2

                                                                                                                         -0.3
                  1.1


                                                                                                                         -0.4
                    1
                                                                                                                                                                                                                                                                                T
                                                                                                                                                                                                                                                                       SD = ∑ [ h1( k −1) (t ) − h1k (t ) / h 21( k −1) (t )] (1)
                  0.9
                        0    200    400           600         800             1000              1200          1400
                                                                                                                         -0.5
                                                                                                                                0                 200         400             600            800            1000          1200      1400
                                                                                                                                                                                                                                                                                                                2
                                                    beat number




Fig.3.Reconstructed signal                                                                                                       Fig.4.Detrended signal                                                                                                                        t =0

                                                                                                                                                                                                                                                            where k represents number of siftings.
          Figs 1-4 explain the efficiency of EMD for RR
interval signal. The ECG data has been collected from the                                                                                                                                                                                                             The process of finding an intrinsic mode function
biomedical website [7] http://www.physionet.org. The RR                                                                                                                                                                                                     requires number of iterations and the process to find all the
intervals are derived from half an hour ECG signal by                                                                                                                                                                                                       IMFs requires further more iterations. As a result of this
identifying the QRS complexes. The signal is manually                                                                                                                                                                                                       iterative procedure finally yields many IMFs and a residue.
edited and only noise free ectopic free segments are used for                                                                                                                                                                                               By summing up all the IMF functions and the residue, the
the analysis. A real time RR interval signal and its EMD                                                                                                                                                                                                    original signal is reconstructed, given by the mathematical
decomposed IMFs are shown in Fig.1&2. Application of                                                                                                                                                                                                        formulae
EMD to real time RR interval signal identifies eight to nine                                                                                                                                                                                                                           n
IMFs. The IMFs are zero mean amplitude frequency                                                                                                                                                                                                                          X (t ) = ∑ hi (t ) + r (n)                                (2)
modulated signal. The decomposition is adaptive and                                                                                                                                                                                                                                   i =1
lossless. The original RR interval signal is reconstructed
using decomposed IMFs (Fig.3). The nonstationary trend is                                                                                                                                                                                                   Where each hi represents an intrinsic mode function and r(n)
removed when the residue or monotonic trend (last IMF) is                                                                                                                                                                                                   either a mean trend or a constant.
omitted while reconstructing the signal (Fig.4).                                                                                                                                                                                                                      For each IMF the meanperiod and variance are
          From the RR intervals the HRV signal or ∆RR                                                                                                                                                                                                       computed. The meanperiod is the ratio of distance between
signal (Ri+1-Ri) is obtained by computing successive                                                                                                                                                                                                        the first and last zero-crossings to number of zero-crossings
difference between consecutive RR intervals. The obtained                                                                                                                                                                                                   of IMF.
HRV signal and its IMFs are shown in Fig.5 and Fig.6.
Matlab 7.1 tools are used for the analysis.                                                                                                                                                                                                                           The obtained RR interval signal using ECG
                                                                                                                                                                                                                                                            represents the response of the cardiovascular system to ANS
3. Methodology
                                                                                                                                                                                                                                                            activities not the ANS activities themselves. The
SIFTING ALGORITHM:                                                                                                                                                                                                                                          characteristics of cardiovascular system determine how the
                                                                                                                                                                                                                                                            system responds to ANS activity and can alter significantly
                                                                                                                                                                          Intinsic mode functions of HRV signal
                                           Heart rate variability signal                                                               0.5
                 0.6                                                                                                                     0
                                                                                                                                      -0.5
                                                                                                                                             0          200         400             600            800            1000      1200      1400
                 0.5                                                                                                                   0.2
                                                                                                                                         0



                                                                                                                                                                                                                                                            the characteristics of the HRV signal. The response
                                                                                                                                      -0.2
                                                                                                                                             0          200         400             600            800            1000      1200      1400
                 0.4                                                                                                                   0.1
                                                                                                                                         0
                                                                                                                                      -0.1
                                                                                                                                             0          200         400             600            800            1000      1200      1400
                 0.3                                                                                                                  0.05



                                                                                                                                                                                                                                                            characteristics are often nonlinear in nature. The IMFs
                                                                                                                                         0
                                                                                                                                     -0.05
                                                                                                                                             0          200         400             600            800            1000      1200      1400
                 0.2                                                                                                                  0.05
                                                                                                                                         0
                                                                                                                                     -0.05
                                                                                                                                             0          200         400             600            800            1000      1200      1400



                                                                                                                                                                                                                                                            capture the all the variations present in the HRV signal.
                 0.1                                                                                                                  0.02
                                                                                                                                         0
                                                                                                                                     -0.02
                                                                                                                                             0          200         400             600            800            1000      1200      1400
                   0
                                                                                                                                      0.02
                                                                                                                                         0
                                                                                                                                     -0.02



                                                                                                                                                                                                                                                            Plotting the variance of all IMFs against its meanperiods
                 -0.1                                                                                                                        0          200         400             600            800            1000      1200      1400
                                                                                                                                                 -3
                                                                                                                                             x 10
                                                                                                                                         2
                                                                                                                                         0
                                                                                                                                        -2
                 -0.2                                                                                                                        0          200         400             600            800            1000      1200      1400
                                                                                                                                                 -3
                                                                                                                                             x 10
                                                                                                                                         5



                                                                                                                                                                                                                                                            gives a nonlinear function. The variance of IMF decreases
                                                                                                                                         0
                                                                                                                                        -5
                 -0.3                                                                                                                        0          200         400             600            800            1000      1200      1400
                                                                                                                                      0.01
                                                                                                                                         0
                 -0.4                                                                                                                -0.01
                        0   200    400         600             800         1000          1200          1400                                  0          200         400             600            800            1000      1200      1400




                                                                                                                                                                                                                                                            with increasing meanperiod and this behavior is
                    Fig.5.HRV signal                                                                                 Fig.6. Intrinsic Mode Functions
                                                                                                                                                                                                                                                            approximated using a geometric function
Step 1: All the minima and maxima of the HRV Signal x (t),                                                                                                                                                                                                            Y= aXb                                       (3)
         are located.                                                                                                                                                                                                                                       where Y represents vector of IMF’s variance,. X represents
Step 2: Spline Interpolate the minima and maxima points to                                                                                                                                                                                                  vector of meanperiods of IMFs, a is constant and b is the
         obtain lower and upper envelopes of the signal.                                                                                                                                                                                                    exponential decrease of the function. The IMFs meanperiod
                                                                                                                                                                                                                                                            and variance of healthy young control subjects, healthy old
Step 3: Compute mean envelope                                                                                                                                                                                                                               control subjects and long term ST subjects vary
                                   m (t)=(maxima’s+minima’s)/2.                                                                                                                                                                                             significantly. The variations in the IMF are quantified by
Step 4: Subtract local mean from the original Signal to                                                                                                                                                                                                     the parameter b. The parameter b represents the flexibility
        obtain local details h(t)=x(t)- m(t).                                                                                                                                                                                                               of cardiovascular system to ANS activities. The parameter b
                                                                                                                                                                                                                                                            is estimated using nonlinear curve fitting technique
                                                                                                                                                                                                                                                            explained below.


                                                                                                                                                                                                                                                      256                                    http://sites.google.com/site/ijcsis/
                                                                                                                                                                                                                                                                                             ISSN 1947-5500
                                                                                                    (IJCSIS) International Journal of Computer Science and Information Security,
                                                                                                    Vol. 8, No. 6, September 2010

                                                                                                                   0.15Hz; 3. Very low frequency band from 0.01Hz to
Taking logarithm of equation (3), gives                                                                            0.04Hz..
  ln Y = ln a + b ln X                                                                             (4)                                                      Relative powers of IMF 1 and IMF 2

           *                    *                                          *
putting Y =ln Y , X =ln X , A = ln a then the above                                                                                 0.9
                                                                                                                                    0.8
                                                                                                                                    0.7

nonlinear equation becomes Y* = A* + bX* which is a linear
                                                                                                                                    0.6
                                                                                                                                    0.5
                                                                                                                     Relative power
                                                                                                                                    0.4                                                                                           Relative powers of IMF 1
                                                                                                                                    0.3                                                                                           Relative powers of IMF 2
equation in X. The corresponding normal equations are                                                                               0.2
                                                                                                                                    0.1
                                                                                                                                      0




∑Y        = NA* + b∑ X *
                                                                                                                                                  1         3        5   7     9     11   13       15        17    19
      *
                                                                                                   (5)
                                                                                                                                                                Healthy young control records



                                                                                                                   Fig 8: Relative powers of IMF 1 and IMF 2
∑X    *
          Y =A
           *     *
                           ∑X            *
                                              + b∑ X                                   *2
                                                                                                   (6)                      0.01



                                                                                                                           0.005
                                                                                                                                                            IMF1
                                                                                                                                                                                               4

                                                                                                                                                                                               3

                                                                                                                                                                                               2
                                                                                                                                                                                                   x 10
                                                                                                                                                                                                        -3
                                                                                                                                                                                                                          IMF2




                                                                                                                                                                                               1
                                              Nonlinear response of IMFs
                 0.9                                                                                                          0                                                                0
                                                                                                                                   0        0.1       0.2          0.3   0.4       0.5             0         0.1    0.2          0.3   0.4    0.5
                                                                                                                                       -3                                                               -4
                 0.8                                                                                                               x 10                     IMF3                                   x 10                   IMF4
                                                                                                                              2                                                                8

                 0.7                                                                                                         1.5                                                               6

                                                                                                                              1                                                                4
                 0.6
                                                                                                                             0.5                                                               2

                                                                                                                              0                                                                0
                 0.5                                                                                                               0        0.1       0.2          0.3   0.4       0.5             0         0.1    0.2          0.3   0.4    0.5
                                                                                                                                       -4                   IMF5                                        -4                IMF6
                                                                                                                                   x 10                                                            x 10
                 0.4                                                                                                          8                                                                8

                                                                                                                              6                                                                6
                 0.3
                                                                                                                              4                                                                4

                 0.2                                                                                                          2                                                                2

                                                                                                                              0                                                                0
                                                                                                                                   0        0.1       0.2          0.3   0.4       0.5             0         0.1    0.2          0.3   0.4    0.5
                 0.1



                  0
                       0   20       40   60         80        100          120   140   160   180




                Fig.7. Curve fitting                                                                               Fig 9: Welch periodogram of IMFs
N represents number of IMFs. Solving the normal equations                                                          IMFs                     Peak                                                             Peak power in
using least mean square method the variables ‘a’ and ‘b’ are
estimated [5]. The simulated response function using the                                                                                    frequency                                     in                 ms2
estimated parameter is shown in Fig.7.                                                                                                      Hz
4. Results and Discussion
                                                                                                                   IMF1                     0.2891                                                           0.01
EMD and curve fitting techniques are applied to half an
hour HRV signal of 20 healthy young control subjects, 20                                                           IMF2                     0.13                                                             0.003
healthy old control subjects and 20 subjects with long term                                                        IMF3                     0.068                                                            0.002
ST. Empirical mode decomposition adaptively decomposes
the half an hour HRV signal into number of Intrinsic mode                                                          IMF4                     0.03                                                             0.00069
functions (Fig.6). The analysis is done with ∆RR intervals.                                                        IMF5                     0.021                                                            0.0007
∆RR (Ri+1-Ri) represents the difference between successive
beat intervals. The IMFs are measured by their absolute                                                            IMF6                     0.01                                                             0.00062
variance, relative variance and meanperiods. The
                                                                                                                   Table-1 Spectral values of IMFs
meanperiod is the ratio of distance between the first and last
zero-crossings to the number of zero-crossings of IMF. First                                                       The meanperiod of IMF2 of healthy young controls subjects
3 IMFs contains more than 92% of total variance. The IMF1                                                          are significantly higher compared to healthy old controls
represents the highest frequency or the shortest period                                                            subjects and subjects with long term ST. Considering
component of the signal. The IMF2 contains the lower                                                               meanperiod of IMF2 (2.9724 secs) as threshold value, we
frequency or the longer period component of the signal.                                                            classified the healthy young control subjects and subjects
Since the 1st and 2nd IMF contains more than 85% of total                                                          with long term ST with 95% sensitivity and 90% specificity.
power they are very significant.                                                                                   The classification is shown in Fig. (10). A threshold value
Relative powers are computed with respect to total power                                                           of 2.8 secs classifies the healthy old controls subjects and
considering all IMFs except the residue with zero                                                                  subjects with long term ST with 90% sensitivity and 70%
meanperiod. In healthy young subjects an increase in                                                               specificity shown in Fig .(11).
relative power of IMF1 decreases the relative power of
IMF2 (Fig.8). IMF 1 and IMF 2 are in phase opposition                                                                              IMF 2 meanperiod of healthy young and subjects with long term
                                                                                                                                                                 st
representing different components of the HRV signal. The
original signal is interpolated to 2 Hz for a meaningful                                                               4
                                                                                                                     3.5

frequency measure. The Welch periodogram (with window                                                                  3
                                                                                                                     2.5
width 1024 and window overlap of 512 samples) of IMFs of                                                               2
                                                                                                                                                                                                                                                    Healthy young
                                                                                                                                                                                                                                                    Sub. With long term st
a healthy young control subject are shown in Fig.(9). Table-                                                         1.5
                                                                                                                       1

1 gives the peak frequency(Hz) and absolute spectralpower                                                            0.5


(ms2-miliseconds square) of IMFs The figure shows the
                                                                                                                       0
                                                                                                                            1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20

frequency spectrum of the IMFs falls in the recognized                                                             Fig.10. Meanperiod comparison of healthy young subjects
spectral bands of HRV signal: 1.High frequency band from                                                           and subjects with longterm ST.
0.15Hz to 0.5Hz; 2. Low frequency band from 0.04Hz to




                                                                                                             257                                                                                       http://sites.google.com/site/ijcsis/
                                                                                                                                                                                                       ISSN 1947-5500
                                                                                                                                                                                         (IJCSIS) International Journal of Computer Science and Information Security,
                                                                                                                                                                                         Vol. 8, No. 6, September 2010

                                   IMF 2 meanperiod of healthy old and sub.with long term st                                                                                                               rate variability analysis”, Med.Bio.Eng.Comput., 2001, 39,
 4.5
                                                                                                                                                                                                           471-479.
  4                                                                                                                                                                                                        [3]E.P.Souza Neto, M.A.Custaud, J.C.Cejka,P.Abry,
 3.5
  3                                                                                                                                                                                                        J.Frutoso, C.Gharib, P.Flandrin, “Assessment of
 2.5
  2
                                                                                                                                          Healthy old controls
                                                                                                                                          sub.with long term st
                                                                                                                                                                                                           Cardiovascular Autonomic Control by the Empirical Mode
 1.5                                                                                                                                                                                                       Decomposition”, Methods Inf Med 2004;43:60-5.
  1
 0.5                                                                                                                                                                                                       [4] N.E.Huang, Z.Shen, S.R.Long, M.C.Wu, H.H.Shih,
  0
       1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20                                                                                                                                                  etal.1998. “The empirical mode decomposition and the
                                                                                                                                                                                                           Hilbert spectrum for nonlinear and nonstationary time series
Fig.11. Meanperiod comparison of healthy old subjects and                                                                                                                                                  analysis” Proc.R.Soc.A, vol 454, pp.903-995.
subjects with longterm ST.                                                                                                                                                                                 [5] B.V.Ramana, “Higher Engineering Mathematics”, Tata
                                                                                                                                                                                                           McGraw-Hill Publishing Company Limited,New Delhi.
The parameter b of IMFs of the three groups HRV signal                                                                                                                                                     [6] HRV Analysis Software 1.1, developed by The
are estimated, average plots are shown in Fig. (12). The                                                                                                                                                   Biomedical Signal Analysis Group, Department of Applied
estimated parameter b of healthy young control subjects,                                                                                                                                                   Physics,       University      of       Kuopio,     Finland.
healthy old control subjects and long term ST subjects are -                                                                                                                                               http://venda.uku.fi/research/biosignal
1.49, -1.43 and -1.39 (average values only). The more                                                                                                                                                      [7] www.physionet.org.
negative value represents the flexibility of the system. The                                                                                                                                               [8] Jan W.Kantelhard, Stephan A, Armin Bunde, 2002,
healthy young control subject’s cardiovascular system is                                                                                                                                                   Multifractal Detrended      Fluctuation     Analysis      of
more flexible than healthy old subjects and longterm ST                                                                                                                                                    Nonstationary Time Series, Physica A 316, 87-114.
subjects. The flexibility of the system decreases in healthy
old control subjects and longterm ST subjects. The absolute
powers of healthy young control subjects are significantly
higher compared to healthy old subjects and long term ST
subjects as shown in Fig.(13) (average values only). The
higher values of absolute power represent more fluctuating
power in the signal. The results show the HRV of healthy
young control subjects contains higher power, longer time
scales and more adaptive to ANS activities compared to
healthy old control subjects and subjects with long term ST.
                                                Meanperiod vs relative powers of IMFs
                         0.7
                                                                                                                                       Absol ut e powe r s I M F1& I M F2
                         0.6
                                                                         * healthy young controls
                                                                         o healthy old controls
                         0.5
                                                                         + longterm st subjects                              0.004
                                                                                                                             0.003
                   ers




                         0.4
                                                                                                                Absol ut e
        elative pow




                                                                                                                             0.002
                                                                                                                power s                                                     Absolute power of IMF1
                         0.3
                                                                                                                             0.001
       R




                                                                                                                                                                            Absolute power of IMF2
                         0.2                                                                                                    0
                                                                                                                                          1           2           3
                         0.1

                                                                                                                                     1. Y oung 2. Ol d 3. Long t er m ST
                          0
                               0     50   100            150          200               250         300   350
                                                           Meanperiod




                          Fig.12 Correlation graphs Fig.13. Absolute powers

5. Conclusion
          In order to cope up nonlinearity and nonstationarity
issue of HRV signal EMD and nonlinear curve fitting
techniques are used in this work. The IMFs of HRV signal
are negatively correlated. The frequency spectrum of first
two IMFs falls in the recognized HF and LF spectral bands
of HRV signal. The meanperiod of IMF2 classifies half an
hour HRV signal of healthy young control subjects, healthy
old control subjects and subjects with long term ST with
high sensitivity and specificity. The nonlinear curve fitting
technique estimates the flexibility of cardiovascular system.
The method is simple, adaptive and no artificial information
is introduced in the analysis.
6. References
 [1] R. M. Carney, J. A. Blumenthal, P. K. Stein, L.
Watkins, D. Catellier, L. F. Berkman, S. M. Czajkowski, C.
O'Connor, P. H. Stone, K. E.Freedland, “Depression, Heart
Rate Variability, and Acute Myocardial Infarction,”
Circulation, vol. 104, no. 17, pp. 2024 – 2028, 2001.
[2] J.C.Echeverria, J.A.Crowe, M.S.Woolfson, B.R.Hayes-
Gill, “Application of empirical mode decomposition to heart


                                                                                                                                                                                                     258                            http://sites.google.com/site/ijcsis/
                                                                                                                                                                                                                                    ISSN 1947-5500

				
DOCUMENT INFO
Description: IJCSIS is an open access publishing venue for research in general computer science and information security. Target Audience: IT academics, university IT faculties; industry IT departments; government departments; the mobile industry and computing industry. Coverage includes: security infrastructures, network security: Internet security, content protection, cryptography, steganography and formal methods in information security; computer science, computer applications, multimedia systems, software, information systems, intelligent systems, web services, data mining, wireless communication, networking and technologies, innovation technology and management. The average paper acceptance rate for IJCSIS issues is kept at 25-30% with an aim to provide selective research work of quality in the areas of computer science and engineering. Thanks for your contributions in September 2010 issue and we are grateful to the experienced team of reviewers for providing valuable comments.