Diagnostics of the Arterial Hypertension by Means of the
Discriminant Analysis
Analysis of the Heart Rate Variability Signals Features Combinations
Vladimir Kublanov
1
, Anton Dolganov
1
and Yan Kazakov
2
1
Research Medical and Biological Engineering Centre of High Technologies, Ural Federal University,
Mira 19, 620002, Yekaterinburg, Russian Federation
2
Ural State Medical University, Repina 3, 620028, Yekaterinburg, Russian Federation
Keywords: Heart Rate Variability, Arterial Hypertension, Classification, Discriminant Analysis.
Abstract: The paper presents investigation of the diagnostic possibilities of the arterial hypertension using linear and
quadratic combinations of the heart rate variability signals features. For this study, two groups were
considered: healthy volunteers and patients suffering from the arterial hypertension of the II-III degree. For
the study, features of statistical, geometric, spectral, nonlinear and multifractal methods were investigated.
Results of the analysis have shown that among studied combinations four feature sets (heart rate, features of
the VLF frequency band and LF/HF ratio) have the highest classification accuracy – 93%.
1 INTRODUCTION
According to the World Health Organization, arterial
hypertension is among the most common diseases in
the world nowadays. In (Feng et al., 2014) it had been
shown that nearly 40% of people aged 45 years had a
hypertensive disorders. Among the individuals with
hypertension around 40% were unaware of their
condition. In the 2000 there were 972 million people
suffering from the arterial hypertension. According to
current predictions this number will increase to 1,56
billion.
One of the main problems concerning arterial
hypertension is late detection for apparently healthy
people. Therefore, the task of the detection of the
arterial hypertension symptoms is among urgent ones.
The heart rate variability (HRV) signals (R-R
intervals) can be used in this task. Among the
advantages of this signals are safeness, prevalence,
repeatability, ease of the record and relative
cheapness (Kamath et al., 2012).
Studies of the hypertension patients generally use
just couple of statistical and/or spectral features for
the analysis (Scheffler et al., 2013). Most of the
studies devoted to this topic imply analysis of the
long-term Holter monitor (Melillo et al., 2012).
However, for the clinical diagnostics in many cases it
is more appropriate to use short-term signals, about 5
minutes length.
Nowadays, great variety of methods are applied
for the HRVy signals processing: statistical, spectral,
non-linear and multifractal. For the most part, during
one study features of single method are used. As an
example in (Melillo et al., 2014) authors have studied
different features of the non-linear methods for
detecting stress state – Poincare plot, approximate
entropy, correlation dimension and recurrence plot .
On the other hand, there were studies that compares
informativeness of features set of different methods.
In (Ebrahimi et al., 2013) authors have studied 4 sets
of features – time domain features, non-linear
features, discrete wavelet transform features and
empirical mode decomposition features. Results of
that study have shown possibilities of different sets
for automatic sleep staging. However, in that work, as
well as in other known works, the possibilities of each
methods were studied separately and combinations of
features from different methods were not tested.
Application of different methods combination
may increase classification accuracy as incorporation
of various methods data increase informative capacity
of knowledge about the studied object (Wiener,
1961). Because of that, the goal of this work is to
develop methodology of the arterial hypertension
diagnostic by the short-term time series (TS) of HRV
Kublanov V., Dolganov A. and Kazakov Y.
Diagnostics of the Arterial Hypertension by Means of the Discriminant Analysis - Analysis of the Heart Rate Variability Signals Features Combinations.
DOI: 10.5220/0006107902910298
Copyright
c
2017 by SCITEPRESS Science and Technology Publications, Lda. All rights reserved
signals using combined estimates, and to study
effectiveness of this methodology.
2 MATERIALS AND METHODS
As the combined estimates in this study, we have used
linear and quadratic combination of two and more
features. Features were obtained by the statistical,
geometric, spectral and multifractal methods. For
evaluation of the diagnostic (classification)
robustness the discriminate analysis have been
applied. All quantifications were performed by the in-
house software developed in MATLAB version
2014b (The MathWorks Inc., Natick, MA).
2.1 Recorded Data
The clinical part of the study was performed in the
Sverdlovsk Clinical Hospital of Mental Diseases for
Military Veterans (Yekaterinburg, Russian
Federation). For the HRV signals registration the
electroencephalograph-analyzer “Encephalan-131-
03” (“Medicom-MTD”, Taganrog, Russian
Federation) was used. The rotating table Lojer
(Vammalan Konepaja DY, Finland) performed the
spatial position change of the patient during passive
orthostatic load – the lift of the head end of the table
was up to 70
o
from the horizontal position.
Participants of this study: 30 relatively healthy
volunteers and 41 patients suffering from the arterial
hypertension of II and III degree. The signals of HRV
were recorded in two functional states: functional
peace (state F) and passive orthostatic load (state O).
Length of the signal in mentioned state was about 300
seconds.
2.2 Classification
As the classification method, we adopted
discriminant analysis (DA) (Jain, 2010). For this
study, we have trialed linear and quadratic DA.
Linear DA aims to find such linear combination of the
features that can be used for adequate separation
between two classes. In turn, quadratic DA aims to
find quadratic combination of the features for
separation. In case of the current study, two classes
are healthy volunteers and patients with the arterial
hypertension.
Evaluation of the classifiers efficiency was
computed with standard measures for binary
classification performance:
Total classification accuracy (ACC)



;
(1)
Sensitivity (SEN)


;
(2)
Specificity (SPE)


;
(3)
where: P, the total number of patients with arterial
hypertension; N, the total number of healthy
volunteers; TP – True Positive, the number of
correctly classified patients with arterial
hypertension; TN – True Negative, the number of
correctly labelled healthy volunteers; FP – False
Positive, the number of people incorrectly classified
as patients with arterial hypertension; FN – False
Negative, the number of people incorrectly classified
as healthy volunteers (Sokolova and Lapalme, 2009).
For the performance measures evaluation
estimation we adopted 5-fold cross-validation
scheme (Bock et al., 2010). This technique imply
developing five classifiers according to following
steps:
division of the original dataset randomly into 5
subsamples (i.e. 8 patients for a group with
arterial hypertension and 6 volunteers for
healthy group);
successive exception of one subsample (testing
subset);
development of a classifier with the remaining
4 subsamples (training subset);
testing of classifier with the excluded
subsample;
computation of the binary classification
measures;
averaging of the performance measures over 5
classifiers.
Division of the original dataset into 5 subsamples
allowed obtaining person-independent testing.
2.3 Properties of Short-Term HRV
Measures
In this work, we investigated diagnostic possibilities
of the arterial hypertension by the combination of the
different methods of the short-term HRV signals
analysis estimates. Prior to the processing the original
time series were cleaned from the artifacts. By the
artifacts in this study, we considered values of the R-
R intervals that differed from the mean by more than
three values of standard deviation. NN is the
abbreviation for the “normal to normal” time series,
i.e. without artifacts. For spectral and multifractal
analysis NN time series were interpolated using cubic
spline interpolation with the 10 Hz sampling
frequency. Interpolation was performed in MATLAB
software by the interp1 function with method
spline’.
2.3.1 Statistical Features
Statistical methods are used for the direct quantitative
evaluation of the HRV time series. Main quantitative
features are:
M, the mean value of the R-R intervals;
HR, the Heart Rate, in inverse ration to the M;
SDNN, the standard deviation of the R-R
intervals;
CV, the coefficient of variation, defined as ratio
of standard deviation SDNN to the mean M,
expressed in percent;
RMSSD is the square root of the mean of the
squares of the differences between successive
elements in NN;
NN5O, the number of pairs of successive
elements in NN that differ by more than 50 ms
(Malik, 1996).
2.3.2 Geometric Features
Geometric methods analyze distribution of the R-R
intervals as a random numbers. The common features
of these methods are:
М
0
, the mode, the most frequent value in the R-
R interval. In case of the normal distribution is
close to the mean M;
VR, the variation range, is the difference
between the lowest R-R interval and the highest
R-R interval in the time series. VR shows
variability of the R-R interval values and
reflects activity of the parasympathetic
department of the autonomic nervous system
(ANS);
АМ
0
, the amplitude of the mode, is a number of
the R-R intervals that correspond to the mode
value. AM
0
shows the stabilizing effect of the
heart rate management, mainly caused by the
sympathetic activity (Malik, 1996).
The following indexes are derived from common
geometric features:
SI, the Stress Index that reflects centralization
degree of the heart rate and mostly characterize
the activity of the sympathetic department of
the ANS

АМ
М
∙
;
(4)
IAB, the Index of the Autonomic Balance,
depends on the relation between activities of
the sympathetic and parasympathetic
department of the ANS:

АМ

; (5)
ARI, the Autonomic Rhythm Index, which
shows parasympathetic shifts of the autonomic
balance: smaller values of the ARI correspond
to the shift of the autonomic balance to the
parasympathetic activity:

М
∙
;
(6)
IARP, the Index of Adequate Regulation
Processes, that reflects accordance of the
autonomic function changes of the sinus node
as a reaction of the sympathetic regulatory
effects on the heart

АМ
М
.
(7)
2.3.3 Spectral Features
Spectral analysis is used to quantify periodic
processes in the heart rate by the means of the Fourier
transform (Fr). The main spectral components of the
HRV signal are High Frequency – HF (0.4 – 0.15 Hz),
Low Frequency – LF (0.15 – 0.04 Hz), Very Low
Frequency – VLF (0.04 – 0.003 Hz), and Ultra Low
Frequency – ULF (lower than 0.003 Hz)
(Malik, 1996). For 300 seconds short-term time series
ULF spectral component is not analyzed.
The studied quantitative features of spectral
analyzes are
Spectral power of the HF, LF, VLF
components
Total power of the spectrum – TP;
Normalized values of the spectral components
by the total power - HF
n
, LF
n
and VLF
n
;
The LF/HF ratio, also known as the autonomic
balance exponent;
IC, the Index of centralization
C


;
(8)
IAS, the Index of the Subcortical nervous
centers Activation



. (9)
2.3.4 Wavelet Transform
For nonstationary time series one can also use the
wavelet transform (wt), that can simultaneously study
time-frequency patterns. The general equation for
continuous wavelet transform is as follows:
,


,
(10)
where: – the scale, b – the shift , ψ – the wavelet
basis , st – analyzed signal (Addison, 2005).
In MATLAB continuous wavelet transform is
implemented by the cwt function. Moreover, the
connection between the scale and the analyzed
frequency is in accordance with the following:

,
(11)
where:
– the central frequency of the wavelet basis,
called by the centfrq function, f
s
– sampling frequency
of the analyzed signal, f – the analyzed frequency
(Mallat, 2009).
It is possible to acquire same spectral features by
means of the wavelet transform:
Spectral power of the HF, LF, VLF
components
Normalized values of the spectral components
by the total power - HF
n
, LF
n
and VLF
n
;
The LF/HF ratio.
Additionally, standard deviations SDHF(wt),
SDLF(wt), SDVLF(wt) of the HF
wt
(t), LF
wt
(t) and
VLF
wt
(t) TS were tested as features. HF
wt
(t), LF
wt
(t)
and VLF
wt
(t) are TS of the HF, LF and VLF spectral
components respectively, acquired by means of the
wavelet transform.
Moreover, one can study informational
characteristics of the wavelet transform by analyzing
the 




 function. As the features of





 is possible to use number of the
dysfunctions N
d
, maximal value of the dysfunction
(LF/HF)
max
, and intensity of the dysfunction
(LF/HF)
int
. By the dysfunction, we consider values of
function that suppress decision threshold .
According to our previous studies =10 (Kublanov,
2008). For wavelet transform computation in this
work, we used wavelet Coiflet of the fifth order.
2.3.5 Nonlinear Feature
As the nonlinear feature in this study we have used
the Hurst exponent calculated by the aggregated
variance method. The variance can be written as
followed



|

|

,
(12)
where H is the Hurst exponent (Rubin et al. 2013).
H can be defined as the slope exponent in the
following equation:
log

∆
log||,
(13)
where

∆
– is the standard deviation of the ΔХ
increments, corresponding to the time period s, с – the
constant.
Note, that H > 0,5 correspond to the process with
trend, so-called persistent process, contrary H < 0,5
correspond to anti-persistent processes that have a
tendency for trend change, H = 0,5 is the random
process (Mandelbrot, 2003).
2.3.6 Multifractal Features
As the nonlinear method we adopted the multifractal
detrended fluctuation analysis (MFDFA) (Stanley et
al., 1999). Algorithm and application features of the
MFDFA method to estimation of short-term TS are
described in details in (Ihlen, 2012).
Figure 1: The characteristic features of the multifractal
analysis.
Fig. 1 represents the main features of the
multifractal spectrum estimated by the MFDFA
method. Here, H
0
is the height of the spectrum,
represents the most probable fluctuations in the
investigated time scale boundary of the signal; H
2
is
the generalized Hurst exponent (also known as
correlation degree);
min
represents behavior of the
smallest fluctuations in the spectrum;
max
represents
behavior of the greatest fluctuations in the spectrum;
W =
max
-
min
, is the width of multifractal spectrum
that shows the variability of fluctuations in the
spectrum. Multifractal characteristics are quantitative
measures of the self-similarity and may characterize
functional changes in the regulatory processes of the
organism.
In addition, we also tested so-called
1/2
width
measure of the spectrum, which is defined as
W
1/2
=|H
2
-H
0
| (Makowiec et al., 2011)
In this study, we investigated time scale
boundaries that correspond to the LF and VLF
frequency bands: (6-25) sec and (25-300) sec
respectively. In our earlier works and by other authors
it was noted that multifractal analysis of the HF
component is not informative because of the noising,
(Makowiec et al., 2012).
3 RESULTS AND DISCUSSION
In this study, we wanted to test all possible
combinations of the features. However, the number of
k-combinations of 53 features is quite high for k
equaled to 2, 3 and 4 - 1378, 23426 and 292825,
respectively. In order to decrease computation time
and remove redundant results (formed by the features
that are already combination of previous features) we
have decided to use for further test only those
combinations that are formed by non-correlated
features.
The correlation coefficient was calculated in
MATLAB software by the corrcoef function. The
threshold for the correlation coefficient was set to be
lower than 0,25. Using this threshold number of
combination was reduced to 629 for two features
combinations, to 1985 for three features combinations
and to 1995 for four features combinations.
3.1 Single Feature Test
For the F state none of the tested features showed
ACC higher than 75 %. The best results of the
classification efficiency estimation for the state O are
presented in table 1. Another 8 combinations have
ACC higher than 76%.
Table 1: Efficiency of the classification for the linear DA
for state O of the single features, %.
feature SEN SPE ACC
H
2
VLF
90 73 83
VLFn(Fr)
83 76 80
VLFn(wt)
85 73 80
According to the data shown in table 6 one can
conclude that application of the single feature
for the arterial hypertension classification is
not sufficient. Therefore, application of two
and more features combinations is justified.
3.2 Two Features Combination Test
Three combinations of features combinations (M and
VLFn(wt),VLFn(Fr) and VLF(wt), VLFn(Fr) and
SDVLF(wt)) for the signals recorded in state F
obtained by the linear DA. achieved the highest
results (ACC 77%). Another 14 combinations that are
formed by the statistical and spectral features, have
ACC not less than 75 %.
Table 2 presents the highest results of the
classification efficiency for two features
combinations of the signals recorded in state O
obtained by the linear DA. Another 42 combinations
that are formed by the statistical, spectral and
multifractal features have ACC not less than 80 %.
Table 2: Efficiency of the classification for the linear DA
for state O of the two feature combinations, %.
features SEN SPE ACC
HFn(Fr), W
1/2
LF 95 73 86
VLFn(Fr), LF/HF(Fr) 93 76 86
VLFn(Fr), LF/HF(wt) 93 76 86
VLFn(Fr), H 95 73 86
LF/HF(Fr), VLFn(wt) 93 76 86
HFn(wt), W
1/2
LF 95 73 86
VLFn(wt), LF/HF(wt) 93 76 86
VLFn(wt), H 93 76 86
Two combinations of features (VLFn(Fr) and
LF/HF(Fr), VLFn(wt) and H) for the signals recorded
in state F obtained by the quadratic DA achieved the
highest results (ACC 76%).
Table 3 presents the highest results of the
classification efficiency for two features
combinations of the signals recorded in state O
obtained by the quadratic DA. Another 42
combinations that are formed by the statistical,
spectral and multifractal features have ACC not less
than 80 %.
According to the data presented in tables 2-3 one
can conclude:
features of the signals recorded in state O allows
to reach higher classification results than those
recorded in state F;
for two features combinations the highest results
are obtained by combinations of the spectral and
multifractal features;
application of two features combination improves
classification efficiency compared to application
of single feature, however it is not possible to
achieve simultaneously high specificity and
sensitivity.
Table 3: Efficiency of the classification for the linear DA
for state O of the two feature combinations, %.
features SEN SPE ACC
VLFn(Fr), H 95 73 86
LF/HF(Fr), VLFn(wt) 88 83 86
HR, H2 VLF 93 73 84
VLFn(Fr), SDVLF(wt) 88 79 84
VLFn(Fr), LF/HF(wt) 83 86 84
VLFn(Fr), (LF/HF)
max
85 83 84
SDVLF(wt), VLFn(wt) 90 76 84
VLFn(wt), LF/HF(wt) 85 83 84
VLFn(wt), W
1/2
LF 88 80 84
VLFn(wt), H 93 73 84
3.3 Three Features Combination Test
The highest result of the classification efficiency for
the signals recorded in state F obtained by the linear
DA was achieved by the combination of M0,
(LF/HF)max, H2 VLF (ACC 78%). Another 17
combinations that are formed by the statistical,
spectral and multifractal features have ACC not less
than 77 %.
Table 4 presents the highest results of the
classification efficiency for three features
combinations of the signals recorded in state O
obtained by the linear DA. Another 83 combinations
that are formed by the statistical, spectral and
multifractal features have ACC not less than 85 %,
while having SPE and SEN not less than 75 %.
Table 4: Efficiency of the classification for the linear DA
for state O of the three feature combinations, %.
features SEN SPE ACC
HR, VLFn(Fr), H
2
LF 95
87 91
VLFn(Fr), LF/HF(Fr), W
1/2
LF 93 87 90
VLFn(Fr), LF/HF(wt), W
1/2
LF 93 87 90
VLFn(Fr), (LF/HF)int, W
1/2
LF 93 87 90
The highest result of the classification efficiency
for the signals recorded in state F obtained by the
quadratic DA was achieved by three combinations
(VLFn(Fr), (LF/HF)int, H; VLFn(wt), (LF/HF)int, H;
VLFn(wt), H0 LF, H) with ACC 77%.
Table 5 presents the highest results of the
classification efficiency for three features
combinations of the signals recorded in state O
obtained by the quadratic DA. Another 65
combinations that are formed by the statistical,
spectral and multifractal features have ACC not less
than 85 %, while having SPE and SEN not less than
75 %.
Table 5: Efficiency of the classification for the quadratic
DA for state O of the three feature combinations, %.
features SEN SPE ACC
LF/HF(Fr), SDVLF(wt),
VLFn(wt)
90
89 90
SDVLF(wt), VLFn(wt),
LF/HF(wt)
90 89 90
HR, VLFn(Fr), SDVLF(wt) 85 93 89
VLFn(Fr), LF/HF(Fr),
VLF(wt)
88 89 89
VLFn(Fr), LF/HF(Fr),
SDVLF(wt)
88 89 89
VLFn(Fr), LF/HF(Fr), W
1/2
LF
90 87 89
VLFn(Fr), VLF(wt),
LF/HF(wt)
88 89 89
VLFn(Fr), (LF/HF)
max
, W
VLF
93 83 89
LF/HF(Fr), VLFn(wt), W
1/2
LF
90 87 89
SDVLF(wt), VLFn(wt),
(LF/HF)
int
93 83 89
VLFn(wt), (LF/HF)
max
, W
VLF
90 87 89
According to the data presented in tables 4-5 one
can conclude:
features of the signals recorded in state O allows
to reach higher classification results than those
recorded in state F, same result as for two feature
combinations;
for three features combinations the highest results
are obtained by combinations of the spectral
features as well as combination of spectral,
statistic and multifractal features;
application of three features combination
improves classification efficiency compared to
application of two features combination, it is
possible to achieve high accuracy (more than 90
%), while maintaining high level of specificity(up
to 89 %) and sensitivity (up to 95 %).
3.4 Four Features Combination Test
Five feature combinations of the signals recorded in
state F obtained by the linear DA (IAB, IAS,
(LF/HF)int, W1/2 VLF; VLF(Fr), IAS, LF/HF(wt),
max
VLF; LF/HF(Fr), IAS, VLF(wt),
max
VLF; IAS,
VLF(wt), LF/HF(wt),
max
VLF; HF(wt), (LF/HF)int,
H2 VLF, H) achieved the highest results of the
classification efficiency (ACC 79%). Another 27
combinations that are formed by the statistical,
spectral and multifractal features have ACC not less
than 77 %.
Table 6 presents the highest results of the
classification efficiency for four features
combinations of the signals recorded in state O
obtained by the linear DA. Another 93 combinations
that are formed by the statistical, spectral and
multifractal features have ACC not less than 85 %,
while having SPE and SEN not less than 75 %.
Table 6: Efficiency of the classification for the linear DA
for state O of the four feature combinations, %.
features SEN SPE ACC
HR, VLFn(Fr), VLF(wt),
(LF/HF)
max
90 93 91
HR, SDVLF(wt),
VLFn(wt), (LF/HF)
max
90 93 91
HR, VLFn(Fr),
SDVLF(wt), (LF/HF)
max
87 93 90
HR, VLF(wt), VLFn(wt),
(LF/HF)
max
90 90 90
Three feature combinations for the signals
recorded in state F obtained by the quadratic DA
(ARI, IAS, Nd, W1/2, VLF; ARI, IAS, (LF/HF)int,
W1/2 VLF; VLFn(Fr), (LF/HF)max,
min
LF, W1/2
VLF) achieved the highest classification efficacy
(ACC > 75%).
Table 7 presents the highest results of the
classification efficiency for four features
combinations of the signals recorded in state O
obtained by the quadratic DA. Another 77
combinations that are formed by the statistical,
spectral and multifractal features have ACC not less
than 85 %, while having SPE and SEN not less than
75 %.
Table 7: Efficiency of the classification for the quadratic
DA for state O of the four feature combinations, %.
features SEN SPE ACC
HR, SDVLF(wt), VLFn(wt),
(LF/HF)
max
93
93 93
HR, VLFn(Fr), SDVLF(wt),
(LF/HF)
max
93 90 91
M, VLFn(Fr), VLF(wt),
(LF/HF)
max
93 86 90
HR, VLFn(Fr), VLF(wt),
(LF/HF)
max
90 90 90
HR, VLF(wt), VLFn(wt),
(LF/HF)
max
90 90 90
VLF(Fr), VLFn(Fr), (LF/HF)
int
,
W VLF
93 87 90
According to the data presented in tables 6-7 one
can conclude:
features of the signals recorded in state O allows
to reach higher classification results than those
recorded in state F, same result as for two and
three feature combinations;
for four features combinations the highest results
are obtained by combinations statistical features
HR and M and spectral features;
application of four features combination further
improves classification efficiency compared to
application of three features combination; it is
possible to achieve relatively high accuracy,
specificity, sensitivity; the specific combinations
of HR and spectral features achieves ACC, SPE
and SEN, all higher than 90 %.
4 CONCLUSIONS
The paper described results of the diagnostic
possibilities test of statistic, geometric, spectral, non-
linear and multifractal features for discrimination of
the arterial hypertension.
Obtained results have shown that classification
efficiency increases as number of features in
combination increases. For four features
combination, formed by HR, VLF estimates and
LF/HF ratio, accuracy, sensitivity and specificity
suppress 90%. Linear and quadratic DA have shown
about the same results of the classifier efficiency.
Results of the current study have higher
classification efficiency compared to our previous
works. There we analyzed comparable sample of
subjects, using single feature in two-dimensional
space: “state F – state O” (Kublanov et al., 2016).
Furthermore, current results suppress classification
efficacy of some other authors. In particular results of
the Artificial Neural Network and Logistic
Regression Analysis models for patients with
hypertension (Tang et al., 2013).
In our opinion this results confirms scientists’
interpretation of the arterial hypertension
development mechanisms. The activation of the
sympathetic nervous system takes important part in
the initialization of the arterial hypertension,
maintenance of the increased vascular tone as well as
increased cardiac output. Role of the vascular system
regulation central mechanisms disorders, including
lost balance of suprasegmental autonomic regulation
and development of the anxiety and depression
disorders (Parati and Esler, 2012, 2013 ESH/ESC
guidelines for the management of arterial
hypertension, 2013).
Results of our study shows application
possibilities of the combined estimates of the short-
term time series heart rate variability signals for the
arterial hypertension diagnostics. In future works, our
research group will continue study this problem on
larger sample of subjects in order to improve
robustness of the classification as well as compare
discriminate analysis performance versus other
methods on the same sample of subjects.
ACKNOWLEDGEMENTS
The work was supported by Act 211 Government of
the Russian Federation, contract 02.A03.21.0006.
REFERENCES
2013 ESH/ESC guidelines for the management of arterial
hypertension: The Task Force for the management of
arterial hypertension of the European Society of
Hypertension (ESH) and of the European Society of
Cardiology (ESC), 2013. European Heart Journal, 34
(28), 2159–2219.
Bock, R., Meier, J., Nyúl, L.G., Hornegger, J., and
Michelson, G., 2010. Glaucoma risk index: Automated
glaucoma detection from color fundus images. Medical
Image Analysis, 14 (3), 471–481.
Ebrahimi, F., Setarehdan, S.-K., Ayala-Moyeda, J., and
Nazeran, H., 2013. Automatic sleep staging using
empirical mode decomposition, discrete wavelet
transform, time-domain, and nonlinear dynamics
features of heart rate variability signals. Computer
Methods and Programs in Biomedicine, 112 (1), 47–57.
Feng, X.L., Pang, M., and Beard, J., 2014. Health system
strengthening and hypertension awareness, treatment
and control: Data from the China health and retirement
longitudinal study. Bulletin of the World Health
Organization, 92 (1), 29–41.
Ihlen, E.A.F., 2012. Introduction to multifractal detrended
fluctuation analysis in Matlab. Frontiers in Physiology,
3 JUN.
Jain, A.K., 2010. Data clustering: 50 years beyond K-
means. Pattern Recognition Letters, 31 (8), 651–666.
Kamath, M.V., Watanabe, M., and Upton, A., 2012. Heart
Rate Variability (HRV) Signal Analysis: Clinical
Applications. New York: CRC Press.
Kublanov, V., Dolganov, A., and Borisov, V., 2016.
Application of the Discriminant Analysis for
Diagnostics of the Arterial Hypertension - Analysis of
Short-Term Heart Rate Variability Signals. Presented at
the 4th International Congress on Neurotechnology,
Electronics and Informatics, 45–52.
Kublanov, V.S., 2008. A hardware-software system for
diagnosis and correction of autonomic dysfunctions.
Biomedical Engineering, 42 (4), 206–212.
Makowiec, D., Rynkiewicz, A., Gałaska, R., Wdowczyk-
Szulc, J., and Żarczyńska-Buchowiecka, M., 2011.
Reading multifractal spectra: Aging by multifractal
analysis of heart rate. EPL (Europhysics Letters), 94
(6), 68005.
Makowiec, D., Rynkiewicz, A., Wdowczyk-Szulc, J., and
Zarczynska-Buchowiecka, M., 2012. On reading
multifractal spectra. multifractal age for healthy aging
humans by analysis of cardiac interbeat time intervals.
Acta Physica Polonica B, Proceedings Supplement, 5
(1), 159–170.
Malik, M., 1996. Heart rate variability: Standards of
measurement, physiological interpretation, and clinical
use. Circulation, 93 (5), 1043–1065.
Mallat, S., 2009. A Wavelet Tour of Signal Processing.
Mandelbrot, B.B., 2003. Multifractal Power Law
Distributions: Negative and Critical Dimensions and
Other ‘Anomalies,’ Explained by a Simple Example.
Journal of Statistical Physics, 110 (3–6), 739–774.
Melillo, P., Izzo, R., De, L., and Pecchia, L., 2012. Heart
rate variability and target organ damage in hypertensive
patients.
BMC Cardiovascular Disorders, 12.
Melillo, P., Pacifici, E., Orrico, A., Iadanza, E., and
Pecchia, L., 2014. Heart rate variability for automatic
assessment of congestive heart failure severity. IFMBE
Proceedings, 41, 1342–1345.
Parati, G. and Esler, M., 2012. The human sympathetic
nervous system: Its relevance in hypertension and heart
failure. European Heart Journal, 33 (9), 1058–1066.
Rubin, D., Fekete, T., and Mujica-Parodi, L.R., 2013.
Optimizing Complexity Measures for fMRI Data:
Algorithm, Artifact, and Sensitivity. PLoS ONE, 8 (5).
Scheffler, P., Muccio, S., Egiziano, G., Doonan, R.J., Yu,
A., Carli, F., and Daskalopoulou, S.S., 2013. Heart rate
variability exhibits complication-dependent changes
postsurgery. Angiology, 64 (8), 597–603.
Sokolova, M. and Lapalme, G., 2009. A systematic analysis
of performance measures for classification tasks.
Information Processing and Management, 45 (4), 427–
437.
Stanley, H.E., Amaral, L.A.N., Goldberger, A.L., Havlin,
S., Ivanov, P.C., and Peng, C.-K., 1999. Statistical
physics and physiology: Monofractal and multifractal
approaches. Physica A: Statistical Mechanics and its
Applications, 270 (1–2), 309–324.
Tang, Z.-H., Liu, J., Zeng, F., Li, Z., Yu, X., and Zhou, L.,
2013. Comparison of Prediction Model for
Cardiovascular Autonomic Dysfunction Using
Artificial Neural Network and Logistic Regression
Analysis. PLoS ONE, 8 (8).
Wiener, N., 1961. Cybernetics Or Control and
Communication in the Animal and the Machine. MIT
Press.