2013 8th International Workshop on Systems, Signal Processing and their Applications (WoSSPA)

DETECTION OF SOME HEART DISEASES USING FRACTAL DIMENSION AND

CHAOS THEORY
Ibticeme Sedjelmaci* - F.Bereksi.Reguig*
E-mail : ibticeme_sedjelmaci@yahoo.fr
E-mail : fethi.bereksi@mail.univ-tlemcen.dz
* Laboratoire de gnie biomdical, Dpartement dlectronique Biomdicale
Facult de Technologie ;
Universit ABOUBEKR BELKAD
Tel: 213 43 28 56 89, Fax: 213 43 28 56 86

may uncover abnormalities in the time series data, which

are not apparent using conventional linear statistic
methods.

ABSTRACT

2. HEART RATE VARIABILITY

This study evaluates the changes in heart rate variability

for 13 signals ECG signals taken from the MIT-BIH
arrhythmia database to detect some major heart disease
(APC, PVC, RBB, LBB) with fractal dimension. Fractal

The non-stationary signal HRV (Heart Rate Variability)

is the variation between two consecutive heartbeats, this
variation can shows a current disease, indicates about
heart failure. HRV is used to extract information on the
control of autonomic nervous system on heart rate
responsible for regulating cardiac activity. It is well
know that HRV may in fact represent a complex
phenomenon and reflect the nonlinear fluctuations in a
fractal or chaotic manner.
It is well known that HRV may in fact represent a
complex phenomenon and reflect the nonlinear
fluctuations in a fractal [4, 5] or chaotic manner [6, 7].

dimension is one of the best known parts of fractal

analysis. A huge number of dimensions have been
defined in various fields. We choose the regularization
dimension [1] for detection and prediction of some
hearts failure. Nonlinear analysis based on chaos theory
and fractal analysis techniques may quantify
abnormalities. This article emphasizes changes in time
series applied on patients with heart disease.

Key words: Electrocardiogram signals ECG - fractal

dimension --- fractal analysis --- chaos theory- MIT-BIH
Data base.

3. FROM FRACTALS TO CHAOS THEORY

Fractal analysis has been developed to study the
irregular complex objects [8]. Many natural phenomena
in several domain like physics, biology and medicine,
could have a fractal behavior. It has also been applied
with a certain success in biomedical domain [9].

1. INTRODUCTION
A normal ECG contains waves, intervals, segments, and
one complex. An interval in an electrocardiogram
includes one segment and one or more waves and a
segment is the region between two waves. RR interval
measure the interval between two consecutive R waves
and ST segment starts from the end of the QRS and
terminates at the onset of the T wave. They are important
in diagnosis.
The variability of the RR intervals, known as Heart
Rate Variability (HRV), is extracted from the
electrocardiogram and used to extract information on the
control of autonomic system on heart rate.
The traditional linear measures as time and frequency
cannot provide all significant information in the signal.
A number of new methods have been recently developed
to quantify complex heart rate dynamics [2, 3]. They

978-1-4673-5540-7/13/$31.00 2013 IEEE

4. CHAOS IN HEALTHY HEART?

Chaotic systems generate signals which can undergo
detection and characterization phenomena and in
synthesize signals for a variety of signal processing
applications [10]. A variety of methods of estimation
[11, 12] and detection of chaotic signals are developed in
many applications. In some cases, the chaotic signal is a
form of noise or other unwanted interference signal. We
are often interested in detecting of some major heart
disease from HRV and ST segment.
When there is a loss of fractal complexity or
dynamic, the regular physiological process exits are poor

89

in information and the system loses its capacity to adjust

to the new factors that continuously change and can
become highly predictable [13].
At first, conventional medical methods assigned
disease and aging to factors that disturbed an ordered
and automatic system, but the discover is that young and
healthy heart have chaotic behavior more than the old
and abnormal one [7], so the heart does not become
regular until it loses its flexibility and ability to adapt.
However, the researches want to clarify if whether
cardiac chaos represents normal or abnormal behavior.
The question is: How heart can have such variety of
beat frequencies and why this variety is suddenly
reduced in the case of a heart disease?

fa = f X a
 

Condition (1) insures that X a tend to the Dirac

distribution and f a to f in the sense of distributions as a
goes to 0. Since f a S , the length of its graph a on K
is finite and given by :

La = 1 + f a' (t ) 2 dt


(3)

log(La )
a 0 log(a )

dim R () = 1 + lim


(4)

dim R ( ) is called the regularization dimension of the

graph .

5. FRACTAL DIMENSION

7. MIT-BIH DATABASE

Physiological signals show different kinds of shapes

with varied complexity. Their structure gives a particular
part in the set of irregular signals. To describe them the
measure of the amplitude and the frequency is typical.
However these measures cannot allow a comparison in
groups of patients where the absolute frequency of
rhythms can be different from a person to another and
can depend on other factors as sex and age.
Fractal dimension provides description of
singularities of a signal [14]: most of the time, the
relevant information in a signal doesn't contain in its
amplitude but in the variations of its regularity.
The analyzed method determined the Regularization
dimension (RD) of RR intervals and ST segment.

In 1975, the laboratories of the Beth Israel Hospital in

Boston and MIT achieved a database named MIT / BIH.
Distributed in 1980, it contains 48 records retrieved a
half an hour, obtained from 47 subjects in the laboratory
BIH arrhythmia. The series "100" have 23 records
randomly selected from a set of 4000 records ambulatory
24-hour joined in a population of hospitalized patients
(60%) and outpatient (40%). Series "200" represent the
remaining records 25 taking into consideration
arrhythmias rarely observed.
These recordings are sampled at a frequency fs = 360
Hz, with a resolution of 11 bit over a range of 10mV.
Each record is annotated independently by two
cardiologists or more, about 110,000 annotations are
included in the database.

6. REGULARIZATION DIMENSION
There are many ways to describe a dimension which are
not necessary equivalents. The regularization dimension
is a new way of evaluating the regularity of a graph of a
function. Two main ways of measuring the regularity of
a non-differentiable function exist [1]:
The first one is based on the investigation of the
Hlder properties. These can be considered from a
global point of view.
The second way of evaluating the regularity f is to
measure the dimension of its graph:
Let be the graph of a bounded function f : R R
and x(t) a kernel function of Schwartz class S such that:
X =1

(2)

8. RESULTS AND DISCUSSION

We analyzed the electrocardiograms from a group of
healthy subjects and those with some different heart
failure.
After the detection of the RR intervals and ST
segments, the regularization dimension (RD) is
calculated. The obtained results are illustrated in figures
(figure1 to figure15), and margins of each graph are
resumed in the corresponding tables (tables1 and table2).
Each point represents a single RR interval RD or ST
segment RD, and each case (normal and various
pathologies) is defined by a color and a different
geometric shape.

(1)

1 t
X be the dilated version of X at
a a
scale a. let f a be the convolution of f with Xa :

8.1.

Let X a ( t ) =

Detection

Detection of abnormal intervals from normal ones and

each pathology from others.

90

8.1.1
RR intervals
Signal
Age RD margins of RR intervals

RD_RR (N)

69

RDN=1,3-1,4; RDAPC= ; 1,25-1,32

RDPVC=1,204

212

32

RDN= 1,7-2; RDRBBB= 1,29-1,4

233

57

RDN= 1,2-1,6; RDPVC=1-1,3

124

77

RDRBBB= 1,25-1,4; RDPVC= 1,1-1,2;

RDJP= 1,15-1,26; RDFVN= 1,2-1,3

1.5
1.4

RD

100

1.3

232

76

RDRBBB= 1,55-1,9; RDAPC=1,3-1,6

214

53

RDLBBB= 1,1-1,35;
1,13

1.2
1.1
1
0

RDAPC= 1,05-

RD_RR (N-APC)

RD_RR (N-PVC)

10000
Peak R

15000

20000

RD_RR (RBBB)

RD_RR (JP)

RD_RR (PVC)

RD_RR (FVN)

1.4

1.45

1.35
1.3

1.4

1.25

RD

1.35

RD

5000

Figure 3. RD of RR interval for signal 233, N:

Normal beat, PVC: Premature Ventricular Contraction.

Table 1: Signals with their values of RD margins for RR

intervals
RD_RR (N)

RD_RR (PVC)

1.6

1.2

1.3

1.15

1.25

1.1
1.05

1.2

1
85000

1.15
430000

450000

470000

490000

510000

Peak R

530000

550000

125000

135000

Figure 4. Signal 124, RBBB: Right Bundle Branch

Block, PVC: Premature Ventricular Contraction, JP:
Nodal (Junctional) Premature beat, FVN: Fusion of
Ventricular and Normal beat.

RD_RR (RBBB)

1.9

1.9

1.8

1.8

RD_RR (APC)

RD_RR (RBBB)

1.7

RD

1.7

RD

105000
115000
Peak R

570000

Figure 1. RD of RR intervals for signal 100, APC:

Atrial Premature Contraction, PVC: Premature
Ventricular Contraction.
RD_RR (N)

95000

1.6
1.5

1.6
1.5

1.4

1.4

1.3

1.3

1.2
9000

19000

29000
Peak R

39000

49000

Figure 2. RD of RR intervals for signal 212, N:

Normal beat, RBBB: Right Bundle Branch Block.

5000

10000

15000
Peak R

20000

25000

30000

Figure 5. RD of RR interval for signal 232, APC:

Atrial Premature Contraction, RBBB: Right Bundle
Branch Block.

91

RD_RR (PVC)

RD_RR (LBBB)

1.35

RD_ST(N)

1.3

1.5

1.5

1.25

RD_ST(RBBB)

1.6

RD

RD

1.4

1.2

1.15
1.1

1.2

0.5

1.1

1.05

1
0

10000

20000
Peak R

30000

1.3

1
0

40000

212

32

RDN= 1,2-1,6; RDRBBB= 1,08-1,23

231

72

RDN= 1,2-1,3; RDRBBB= 1,07-1,21

119

51

RDN= 1,10-1,22; RDPVC=1,05-1,11

221

57

RDN= 1,27-1,6; RDPVC=1,15-1,25

RD_ST(RBBB)

1.35

50000

60000

RD_ST(N)

1.3

RD

1.25
1.2

1.15
1.1
1.05
1
0

10000

20000

30000
ST_on

40000

50000

60000

Figure 8. RD of ST interval for signal 231, N: Normal

beat, RBBB: Right Bundle Branch Block.

8.1.2
ST segments
In electrocardiogram, the ST segment detection is
important, for some pathology this segment is modified.
After detection of ST segments [15], we calculate RD
dimensions of each one (RD_ST) and listed in Table 2.
In all calculations, the ST segment includes the T wave.
So RD_ST means the dimension of interval starting at
the end of complex QRS and terminating at the end of T
wave.
RD margins of ST segments

30000
40000
ST_on

A significant difference was found between healthy and

unhealthy segments (Fig.7 and Fig.8).

The decrease in values of abnormal intervals compared

with normal or a second kind of pathology intervals is
clear. The normal interval may display complex
nonlinear dynamics.
The variations of RR interval exhibited strongly and
consistently chaotic behavior in all healthy intervals, but
were frequently interrupted by periods of non-chaotic
fluctuations in patients with heart diseases. Chaotic
dynamics in different heart troubles data is not very
discernible, exhibited low degree variability and
suggested a weaker form of chaos. These findings
suggest that cardiac chaos is prevalent in healthy heart,
and a decrease in such chaos may be indicative of
diseases.

Age

20000

Figure 7. RD of ST interval for signal 212, N: Normal

beat, RBBB: Right Bundle Branch Block.

Figure 6. RD of RR interval for signal 214, LBBB:

Left Bundle Branch Block, PVC: Premature Ventricular
Contraction.

Signal

10000

RD_ST(N)

RD_ST(PVC)

RD

1.21
1.19
1.17
1.15
1.13
1.11
1.09
1.07
1.05
0

10000

20000
ST_on

30000

40000

Figure 9. RD of ST interval for signal 119, N: Normal

beat, PVC: Premature Ventricular Contraction.

Table 2: Signals with their values of RD margins for ST

segments

92

RD_ST(N)

RD_ST(PVC)

RD_RR (N)

1.6
1.55
1.5
1.45
1.4
1.35
1.3
1.25
1.2
1.15
1.1

RD_RR (N-PVC)

1.35
1.3

RD

RD

1.25
1.2
1.15
1.1

10000

20000
ST_on

30000

20000

Peak R

40000

60000

Figure 10. RD of ST interval for signal 221, N:

Normal beat, PVC: Premature Ventricular Contraction.

Figure 12. RD of RR interval for signal 116, N:

Normal beat, PVC: Premature Ventricular Contraction.

The increased parameter variability in the abnormal ST

segments group correspond to the decrease in detection
for chaos and fractal characteristics (Fig. 7, 8, 9, 10) and
confirm that the degree of cardiac chaos was decreased
in the heart diseases.

The RD of normal RR intervals (Fig 11, 12, 13) which

come immediately after dimension of abnormal RR
intervals can have low values but higher ones than PVC
intervals. The cardiac diseases can affect the normals
beats.

8.2.

Prediction:

RD_RR (N)

1.6

RD_RR (N)

1.32
1.3
1.28
1.26
1.24
1.22
1.2
1.18
1.16
1.14

1.5

RD

1.4
1.3
1.2
1.1
1
0

5000

Peak R

10000

15000

Figure 13. RD of RR interval for signal 228, N:

Normal beat, PVC: Premature Ventricular Contraction.

RD_RR (N-PVC)

RD_RR (N)

RD_RR (N-PVC)

1.4
1.35
1.3

RD

RD

In another point of view, and after calculating

dimensions for some RR intervals between two waves
normal and abnormal (RD_RR(N-PVC)), we compare its
with the ones between two normal waves (RD_RR(NN)). We detect the difference in dimension for the
intervals before the pathology, so we can know the
existence of pathology just before. The knowing about
this existence can help us for prediction.
We choose the signals with Normal beat and
Premature Ventricular Contraction:
119 (51 years old), 116 (68 years old),
228 (80 years old), 105 (73 years old),
215 (81 years old).

RD_RR (N-PVC)

1.25
1.2
1.15
1.1

10000

20000
30000
Peak R

40000

50000

Figure 11. RD of RR interval for signal 119, N:

Normal beat, PVC: Premature Ventricular Contraction.

100000

200000
Peak R

300000

400000

Figure 14. RD of RR interval for signal 105, N:

Normal beat, PVC: Premature Ventricular Contraction.

93

RD_RR(N)

1.7

RD_RR(N-PVC)

[3]
P. Smrka, R. Bittner, P. Vysok, K. Hna, Fractal
and Multifractal Properties of Heartbeat Interval Series in
Extremal States of the Human Organism, Measurement
Science Review, Vol. 3, pp. 13-15, 2003.

RD_RR(PVC-PVC)

1.6
1.5

[4]
Diosdado. A. Muoz, Coyt. G. Glvez, Uribe. B. M
Prez Oscillations in the evaluation of fractal dimension of
RR intervals time series, Engineering in Medicine and
Biology Society, 2010 Annual International Conference of the
IEEE, pp. 4570 4573, 2010.

RD

1.4
1.3
1.2
1.1

[5]
HV. Huikuri, TH. Mkikallio, CK. Peng, AL.
Goldberger, U. Hintze, M. Mller, Fractal correlation
properties of R-R interval dynamics and mortality in patients
with depressed left ventricular function after an acute
myocardial infarction, Circulation, pp. 47-53, 2000.

1
0.9
7000

14000

21000
Peak R

28000

35000

Figure 15. RD of RR interval for signal 215, N:

Normal beat, PVC: Premature Ventricular Contraction.

[6]
G. Krstacic, D. Gamberger, A. Krstacic, T. muc, D.
Milicic, The Chaos Theory and Non-linear Dynamics in Heart
Rate Variability in Patients with Heart Failure Computers in
Cardiology, pp. 957 959, 2008.

The heartbeat series of patients with PVC were

characterized by the distinct presence of low-amplitude,
low-frequency and decrease in variability. Most of the
time, such decrease in regularity may contribute to a
decrease in chaotic contents. Indeed, Heart failures may
all reflect a decrease in complexity of the heartbeat
nonlinear dynamics.

[7]
C. D. Wagner, P. B. Persson, Chaos in the
cardiovascular system : an update, Cardiovascular Research,
Vol. 40, pp. 257-264, 1998.
[8]
R. Lopes, P. Dubois, I. Bhouri, H. Akkari-Bettaieb,
S. Maouche, N. Betrouni, La gomtrie fractale pour lanalyse
de signaux mdicaux: tat de lart, IRBM31, pp189---208,
2010.

9. CONCLUSION

[9]
Nahina Islam, Nafiz Imtiaz Bin Hamid, Adnan
Mahmud, Sk. M. Rahman, Arafat H. Khan, Detection of
Some Major Heart Diseases Using Fractal Analysis,
International Journal of Biometrics and Bioinformatics (IJBB),
Vol. 4, pp. 63-70, 2010.

The main goal of study was to investigate the clinical

and prognostic significance of non-linear methods as
fractal analysis and chaos theory and to detect dynamic
changes in signals with healthy and unhealthy segments.
The nonlinear methods based on chaos theory and
fractal analysis may have significant clinical advantage
of ECG.
The regularization dimension RD is very sensitive to
changes of irregular structure of signals for both RR
interval and ST segment.
The normal heartbeat may display complex nonlinear
dynamics and that such cardiac chaos may be a useful
physiological marker for the diagnosis and management
of certain heart trouble. The analysis applied on RR
intervals and ST segments of patients with heart failure
shows decrease of dimension so that signals lose fractal
characteristics and regularity of the signal.

[10]
Joon Suh Lee and Kun Soo Chang, Applications of
chaos and fractals in process systems engineering, Journal of
Process Control, Vol. 6, pp. 7187, 1996.
[11]
P. Gonalvs, P. Abry, G. Rilling, and P. Flandrin,
Fractal dimension estimation: empirical mode decomposition
versus wavelets, IEEE International Conference on Acoustics
Speech and Signal Processing, Vol. 3, pp. 1153-1156, 2007.
[12]
Michael J. Vrhel, Chuihee Lee, and Michael Unser,
Fractal dimension estimation using the fast continuous
wavelet
transform,
Biomedical
Engineering
and
Instrumentation Program, Vol. 2569, 1995.
[13]
C. S. Poon and C. K. Merrill, Decrease of cardiac
chaos in congestive heart failure, Nature, Vol. 389, pp. 492495, 1997.

10. REFERENCES

[14]

P.Y. Muller, N. Contento, H. Rix, Fractal

Dimension on ECG, Engineering in Medicine and Biology
Society, Proceedings of the Annual International Conference of
the IEEE, Vol. 14, pp. 977-978, 1992.

[1]
F. Roueff, J. L. Vhel, A Regularization Approach
to Fractional Dimension Estimation, Proceeding of Fractals,
1998.

N. Dib and F. Bereksi Reguig, Algorithm for

[15]
Automatic Detection of ECG Waves, Journal of Mechanics in
Medicine and Biology, Vol. 11, pp.1529, 2011.

[2]
P. van Leeuwen and H. Bettermann, The status of
nonlinear dynamics in the analysis of heart rate variability,
Herzschr Elektrophys, Vol. 11, pp. 127-130, 1999.

94