=Paper=
{{Paper
|id=Vol-2711/paper23
|storemode=property
|title=Evaluating the Effectiveness of Electrocardiological Study Using Cardiological Decision Support Systems
|pdfUrl=https://ceur-ws.org/Vol-2711/paper23.pdf
|volume=Vol-2711
|authors=Anna Filatova,Inna Skarga-Bandurova,Eugene Brezhniev,Mohamad Fahs
|dblpUrl=https://dblp.org/rec/conf/icst2/FilatovaSBF20
}}
==Evaluating the Effectiveness of Electrocardiological Study Using Cardiological Decision Support Systems==
https://ceur-ws.org/Vol-2711/paper23.pdf
Evaluating the Effectiveness of Electrocardiological
Study Using Cardiological Decision Support Systems
1
Anna Filatova [0000-0003-1982-2322], 2Inna Skarga-Bandurova [0000-0003-3458-8730],
3
Eugene Brezhniev [0000-0003-2073-9024], 4Mohamad Fahs [0000-0001-7776-3311]
1,4National Technical University “Kharkiv Polytechnic Institute”,
Kyrpychova street, 2, Kharkiv, 61002, Ukraine
filatova@gmail.com
2Oxford Brookes University, Wheatley Campus, Oxford, OX33 1HX, UK
iskarga-bandurova@brookes.ac.uk
3National Aerospace University “Kharkov Aviation Institute”,
Chkalov street, 17, Kharkiv, 61070, Ukraine
e.brezhnev@csn.khai.edu, fahes_93mohamad@hotmail.com
Abstract. This work is devoted to evaluating the effectiveness of the electro-
cardiological study process without using and using cardiological decision sup-
port systems. To assess the effectiveness, analytical expressions of the probabil-
istic-time characteristics of the developed structural model of the electrocardio-
logical study process are used. An analysis of the time characteristics of the
model is performed when different initial conditions are set for three different
types of electrocardiological studies: the study is conducted for the first time,
the study is repeated as a result of screening, the study is repeated after treat-
ment. The work shows that the use of cardiological decision support systems
based on the developed methods for analyzing biomedical signals with locally
concentrated features reduced the average time required for the electrocardio-
logical study of each of the considered types.
Keywords: Electrocardiological study; Cardiological decision support system;
Probabilistic-time characteristic; Biomedical signals with locally concentrated
features
1 Introduction
Currently, a sharp increase in the amount of information processed in solving tradi-
tional medical problems has led to the introduction of various medical information
systems (MIS) into modern medicine, from simple electronic medical records to com-
plex decision support systems (DSS) [1-3]. The electrocardiological (ECG) study
process is based on the analysis of biomedical signals (BMS) with locally concentrat-
ed features (LCF) associated with the cyclic work of the heart and cardiovascular
system [4]. Various computerized cardiological systems, including cardiological DSS,
are used to automate the collection and processing of such information. The authors
developed the structural model of the ECG study process in the form of a probabilis-
Copyright © 2020 for this paper by its authors. Use permitted under Creative Commons
License Attribution 4.0 International (CC BY 4.0). ICST-2020
tic-time graph, which made it possible to obtain analytical expressions for the analysis
of this process given initial conditions (the presence or absence of previous examina-
tions and treatment), as well as determine the criteria for the effectiveness of ECG
studies [5].
2 Literature review
A review of the literature showed that most often attention is paid to the analysis of
individual stages of the ECG study process, among which the following:
• detection of indications for examination;
• recording and digitization of BMS with LCF;
• preprocessing of BMS with LCF;
• morphological analysis of BMS with LCF;
• detection of diagnostic indicators;
• diagnostics and issuing a diagnostic report.
The quality and effectiveness of the ECG study depend on the quality of the re-
cording BMS with LCF.
As a result of preprocessing of BMS with LCF, there are most often performed re-
moving artifacts from the signal by methods based on the use of various types of fil-
ters [6, 7] and of wavelet transform [8-10], compensation of the isoline drift by meth-
ods based on interpolation of the ECG isoelectric line [11, 12].
One of the difficult and critical stages is the stage of the morphological analysis of
BMS with LCF for which various methods are used:
─ analysis of BMS with LCF in the time domain using modern classification methods
such as cluster analysis and pattern recognition [13, 14], probabilistic classification
[15], neural networks [16], fuzzy clustering [17, 18];
─ analysis of BMS with LCF in the time-frequency domain, for example, local (win-
dow) Fourier transform (spectral-time mapping) and wavelet transform [19, 20], as
well as in the phase plane [4];
─ morphological filtration of BMS with LCF using the multichannel matched mor-
phological filter proposed by the authors [21].
Diagnostic features are formed in the form of parameters of the found structural el-
ements based on the morphological analysis of BMS with LCF [22, 23]. Thus, errors
at the stage of the morphological analysis of BMS with LCF can lead to the incorrect
diagnostic solutions.
Therefore, the quality of the ECG study directly depends on the quality of the mor-
phological analysis of BMS with LCF.
Different MISs are used with varying degrees of effectiveness at each of the listed
stages. However, a systematic analysis of the ECG study process without using and
using cardiological DSS is not found in literary sources.
3 Formal problem statement
The aim of the work is to analyze the effectiveness of the ECG study process without
using and using cardiological DSS based on the morphological analysis of BMS with
LCF.
To achieve this goal, the following tasks are solved:
─ to determine the average examination time under various initial conditions by the
developed structural model;
─ to evaluate the effectiveness of ECG studies without using and using cardiological
DSS by the developed criterion.
4 A structural model of a patient's ECG study
Let us consider the structural model developed in [5] for the process of ECG study,
shown in Fig. 1.
f19(z)
f29(z)
f39(z)
f49(z)
S5 f59(z) S9
f45(z)
f57(z) f69(z) 1
S4 f79(z)
f22(z) f46(z)
f21(z) f34(z) S6
f67(z)
S0 f01(z) S1 f12(z) S2 f23(z) S3 f37(z) S7 f78(z) S8
f32(z)
f31(z) f77(z) 1
Fig. 1. The structural model M S of the ECG study
To describe the passage of the ECG study process from the initial state to the final
state the arc function f ( pij , t ij ) = f ij ( z ) of a probabilistic-time graph is defined such
that when finding the products of the arc functions, the probabilities pij are multi-
plied and the times tij are summed:
t
f ij ( z ) = pij z ij , (1)
where z – a parameter of the arc function, the degree of which characterizes the time
of transition from one state to another ( z 1 ).
The following states are identified in the structural model M S : S 0 – the beginning
of the study; S1 – indications were defined; S 2 – morphological analysis of BMS
with LCF was performed; S 3 – pathological changes were identified; S 4 – compari-
son with previous ECG studies was performed; S 5 – dynamics evaluation was com-
pleted; S 6 – evaluation of treatment effectiveness was completed; S 7 – the diagnostic
decision was made; S 8 – recommendations were issued (the end of the ECG study);
S 9 – a set of states that do not lead to the goal (the state of uncertainty); f ij ( z ) ,
i, j = 0;9 – arc function by (1).
In [5], it is indicated that the structural model M S is no state associated with the
direct recording of the ECG signal because the duration of the ECG signal recording
is strictly regulated by the protocol of the type of the ECG study and can vary from
several minutes to several hours and days. That is, this time cannot be optimized, and
the duration of the recording process does not affect the effectiveness of the ECG
study.
The generating function of the graph shown in Fig. 1 has the following form:
F ( z ) = F08 ( z ) + F09 ( z ) , (2)
where
F08 ( z ) =
=
( (
p01 p12 p 23 p78 p34 z t34 p 45 p57 z t45 +t57 + p 46 p67 z t46 +t67 + p37 z t37 z t01 +t12 +t23 +t78 ) )
(1 − p z ) 77
t77
(1 − p (p (1 + p p z )z + p p z )z − p p z − p z );
12 21 23 32
t23 +t32 t21
23 31
t23 +t31 t12
23 32
t23 +t32
22
t22
F ( z ) = p ((1 − p )z + p ((1 − p − p − p )z +
09 01 12
t19
12 21 22 23
t29
+ p ((1 − p − p − p − p )z + p ((1 − p − p )z +
23 31 32 34 37
t39
34 45 46
t49
+ p ((1 − p )z + p (1 − p − p )z
45 57 )z + t59
57 77 78
t57 +t79 t45
+ p ((1 − p )z + p (1 − p − p )z
46 67 )z )z +
t69
67 77 78
t67 +t79 t46 t34
+ p (1 − p − p )z
37 )z )z )z (1 − p z )
77 78
t37 +t79 t23 t12 t01
77
t77
(1 − p (p (1 + p p z )z + p p z )z − p p z − p z ).
12 21 23 32
t23 +t32 t21
23 31
t23 +t31 t12
23 32
t23 +t32
22
t22
Using the generating function (2), it is possible to determine the probability and the
average time of an ECG study by following expressions:
PECG = F ( z ) z =1 ;
dF ( z )
TECG = .
dz z =1
Since the analytical expressions for the probability PECG and the average time TECG
of an ECG study are too cumbersome, in [5] the authors developed a program in the
Matlab language for getting these analytical expressions as well as an analytical ex-
+
pression of the probability PECG of a successful ECG study which has the following
form taking into account restrictions:
+ p01 p12 p23 p78 ( p34 ( p45 p57 + p46 p67 ) + p37 )
PECG = ;
( )
(1 − p77 ) 1 − p12 ( p21 (1 + p23 p32 ) + p23 p31 ) − p23 p32 − p22
p34 p37 = 0;
(3)
p34 + p37 ( 0;1 ;
p p = 0;
45 46
p45 + p46 ( 0;1 .
Also in [5], there were proposed the following criteria for the effectiveness of the
ECG study process by the average time taken to complete the study and the probabil-
ity of its successful completion:
TECG → min;
p p = 0;
34 37
p34 + p37 ( 0;1 ;
p p = 0;
45 46
p45 + p46 ( 0;1 ;
PECG
+
→ max;
p34 p37 = 0;
p34 + p37 ( 0;1 ;
p p = 0;
45 46
p45 + p46 ( 0;1 .
Let us use the obtained analytical expressions that describe the probabilistic-time
characteristics of the ECG study process under given initial conditions (the presence
or absence of previous examinations and treatment), as well as use the proposed crite-
ria of the effectiveness for analysis and optimization of the entire process and its indi-
vidual stages.
5 Experiments and results
To analyze the probabilistic-time characteristics of the ECG study process, it is neces-
sary to set the initial conditions.
According to the structural model M S of the ECG study (Fig. 1), there are three al-
ternative ways of transition from the initial state S 0 to the final state S 8 which corre-
spond to three different types of ECG studies [5]:
• the study is conducted for the first time;
• the study is repeated as a result of screening;
• the study is repeated after treatment.
In this case, let us consider a simplified version of the model when the ECG study
process does not go into a state of uncertainty S 9 , that is
p19 = p29 = p39 = p49 = p59 = p69 = p79 = 0 and t19 = t29 = t39 = t49 = t59 = t69 = t79 = 0 .
Since all transitions from the current state S i form a complete group of events, the
following expressions can be written:
p01 = p12 = p57 = p67 = 1 ; (4)
p23 + p21 + p22 = 1 ; (5)
p34 + p37 + p31 + p32 = 1 ; (6)
p45 + p46 = 1 ; (7)
p78 + p77 = 1 . (8)
−
Let us denote PECG – probability of transition to a state of uncertainty S 9 , then,
taking into account the simplified model ( p19 = p29 = p39 = p49 = p59 = p69 = p79 = 0 )
− +
PECG = 0 , which means PECG = 1 for any admissible probability values in the expres-
sion (3), that is, the examination will surely end successfully. However, the time taken
to complete the examination will depend not only on the time of each stage but also
on the corresponding probabilities.
In this case, the analytical expression for the average time Tsimp of the ECG study,
calculated according to the simplified model, is too cumbersome, but it is easy to
obtain using the program in the Matlab language, if, taking into account (4) and (8),
the following substitution is made at the end of the program, given in [5]:
Tsimpl = subs(T,[p01, p12, p57, p67, p77, t19, t29, t39,
t49, t59, t69, t79],[1, 1, 1, 1, 1-p78,
0, 0, 0, 0, 0, 0, 0]);
Let us analyze the average time of conducting an ECG study using a simplified
model M S separately for each of the cases under different initial conditions. Moreo-
ver, in each of the cases we will consider the average execution time of each stage for
three options:
─ using cardiological DSS with the module of morphological analysis of BMS with
LCF (DSS1) developed by the authors [21];
─ using cardiological DSS in which the morphological analysis of BMS with LCF is
performed in a semi-automatic mode (DSS2);
─ without using any MIS (without MIS).
In all experiments, we take p21 = p31 = 0 . Then according to (5) p22 = 1 − p23 .
5.1 Analysis of the time characteristics of the model for the case when the
ECG study is conducted for the first time
If an ECG study is conducted for the first time, then p34 = 0 , and then a simplified
structural model of an ECG study has the form shown in Fig. 2.
f22(z)
f21(z)
S0 f01(z) S1 f12(z) S2 f23(z) S3 f37(z) S7 f78(z) S8
f32(z)
f31(z) f77(z) 1
Fig. 2. Simplified structural model of an ECG study: the study is conducted for the first time
The analytical expression for the average time T1 of the ECG study, which is con-
ducted for the first time, was obtained from the Matlab, performing the following
substitution at the end of the program given in [5]:
T1 = simplify(subs(Tsimpl, p34, 0));
Let us consider the dependence T1 ( p23 ) under the following initial conditions:
p37 = p78 = 1 . The initial conditions for the average execution time of each stage are
given in Table 1. It should be noted here that only the time t12 differs when conduct-
ing an ECG study with DSS1 and DSS2 (Table 1).
Table 1. The average time of each stage (in minutes): the study is conducted for the first time
The average time t 01 t12 t 23 t37 t 78 t 22 t32
DSS1 7 2 2 2 2
DSS2 7 5 2 2 2 1.1t12 1.2t37
without MIS 12 7 3 7 7
Table 2 shows the minimum and maximum values of the average time T1 of the
ECG study without using and using cardiological DSS in experiments at different
values p32 .
Table 2. Values T1 (in the format mm:ss) in experiments at various values p32
Values DSS1 DSS2 without MIS
p32 min T1 max T1 min T1 max T1 min T1 max T1
0 15:00 17:12 18:00 23:30 36:00 43:42
0.1 15:26 17:53 18:26 24:33 37:06 45:40
0.3 16:42 19:51 19:42 27:34 40:17 51:17
0.5 19:00 23:23 22:00 33:00 46:00 61:23
Figure 3 shows the curves of the change in the average time T1 ( p23 ) of the ECG
study without using and using cardiological DSS at various initial values p32 .
a) b)
c) d)
Fig. 3. Graphs of changes in the average time T1 ( p23 ) of the ECG study
without using and using cardiological DSS:
a) p32 = 0 ; b) p32 = 0.1 ; c) ; p32 = 0.3 d) p32 = 0.5
From an analysis of the data presented (Table 2 and Fig. 3), we can conclude that
in each of the experiments the following trend is observed:
─ max T1DSS1 min T1DSS2 (for p32 = 0 and p32 = 0.1 );
─ max T1DSS1 min T1DSS2 (for p32 = 0.3 );
─ max T1DSS1 min T1without MIS and max T1DSS2 min T1without MIS (for all values p32 ).
5.2 Analysis of the time characteristics of the model for the case when the
ECG study is repeated as a result of screening
If the ECG study is repeated as a result of screening, then p37 = 0 and p46 = 0 , and
then the simplified structural model of the ECG study has the form shown in Fig. 4.
f22(z)
f21(z)
S0 f01(z) S1 f12(z) S2 f23(z) S3 f34(z) S4 f45(z) S5 f57(z) S7 f78(z) S8
f32(z)
f31(z) f77(z) 1
Fig. 4. Simplified structural model of an ECG study: the study is repeated
as a result of screening
The analytical expression for the average time T2 of the ECG study that is repeated
as a result of screening was obtained from the Matlab, performing the following sub-
stitution at the end of the program given in [5]:
T2 = simplify(subs(Tsimpl,[p37, p45, p46],[0, 1, 0]));
Let us consider the dependence T2 ( p23 ) under the following initial conditions:
p34 = p45 = p57 = p78 = 1 . The initial conditions for the average execution time of
each stage are given in Table 3. As in the first case, only the time t12 differs when
conducting an ECG study with DSS1 and DSS2 (Table 3).
Table 3. The average time of each stage (in minutes):
the study is repeated as a result of screening
The average time t 01 t12 t 23 t34 t 45 t57 t 78 t 22 t32
DSS1 7 2 2 2 1 1 2
DSS2 7 5 2 2 1 1 2 1.1t12 1.1t34
without MIS 12 7 3 10 3 2 7
Table 4 shows the minimum and maximum values of the average time T2 of the
ECG study without using and using DSS in experiments at different values p32 .
Figure 5 shows the curves of the change in the average time T2 ( p23 ) of the ECG
study without using and using cardiological DSS at various initial values p32 .
Table 4. Values T2 (in the format mm:ss) in experiments at various values p32
Values DSS1 DSS2 without MIS
p32 min T2 max T2 min T2 max T2 min T2 max T2
0 17:00 19:12 20:00 25:30 44:00 51:42
0.1 17:26 19:53 20:26 26:33 45:26 54:00
0.3 18:42 21:51 21:42 29:34 49:34 60:34
0.5 21:00 25:23 24:00 35:00 57:00 72:24
a) b)
c) d)
Fig. 5. Graphs of changes in the average time T2 ( p23 ) of the ECG study
without using and using cardiological DSS:
a) p32 = 0 ; b) ; p32 = 0.1 c) p32 = 0.3 ; d) p32 = 0.5
From an analysis of the data presented (Table 4 and Fig. 5), we can conclude that
in each of the experiments, the trend described for the first case remains:
─ max T2DSS1 min T2DSS2 (for p32 = 0 and p32 = 0.1 );
─ max T2DSS1 min T2DSS2 (for p32 = 0.3 );
─ max T2DSS1 min T2without MIS and max T2DSS2 min T2without MIS (for all values p32 ).
5.3 Analysis of the time characteristics of the model for the case when the
ECG study is repeated after treatment
If the ECG study is repeated after treatment, then p37 = 0 and p45 = 0 , and then the
simplified structural model of the ECG study has the form shown in Fig. 6.
f22(z)
f21(z)
S0 f01(z) S1 f12(z) S2 f23(z) S3 f34(z) S4 f46(z) S5 f67(z) S7 f78(z) S8
f32(z)
f31(z) f77(z) 1
Fig. 6. Simplified structural model of an ECG study: the study is repeated after treatment
The analytical expression for the average time T3 of the ECG study that is repeated
as a result of screening was obtained from the Matlab, performing the following sub-
stitution at the end of the program given in [5]:
T3 = simplify(subs(Tsimpl,[p37, p45, p46],[0, 0, 1]));
Let us consider the dependence T3 ( p23 ) under the following initial conditions:
p34 = p46 = p67 = p78 = 1 . The initial conditions for the average execution time of
each stage are given in Table 5. As in the first two cases, only the time t12 differs
when conducting an ECG study with DSS1 and DSS2 (Table 5).
Table 5. The average time of each stage (in minutes): the study is repeated after treatment
The average time t 01 t12 t 23 t34 t 46 t 67 t 78 t 22 t32
DSS1 7 2 2 2 1,5 1 2
DSS2 7 5 2 2 1,5 1 2 1.1t12 1.1t34
without MIS 12 7 3 10 3,5 2 7
Table 6 shows the minimum and maximum values of the average time T3 of the
ECG study without using and using cardiological DSS in experiments at different
values p32 .
Table 6. Values T3 (in the format mm:ss) in experiments at various values p32
Values DSS1 DSS2 without MIS
p32 min T3 max T3 min T3 max T3 min T3 max T3
0 17:30 19:42 20:30 26:00 44:30 52:12
0.1 17:56 20:23 20:56 27:03 45:56 54:30
0.3 19:12 22:21 22:12 30:04 50:04 61:04
0.5 21:30 25:54 24:30 35:30 57:30 72:54
Figure 7 shows the curves of the change in the average time T3 ( p23 ) of the ECG
study without using and using cardiological DSS at various initial values p32 .
a) b)
c) d)
Fig. 7. Graphs of changes in the average time T3 ( p23 ) of the ECG study
without using and using cardiological DSS:
a) p32 = 0 ; b) p32 = 0.1 ; c) p32 = 0.3 ; d) p32 = 0.5
From an analysis of the data presented (Table 6 and Fig. 7), we can conclude that
in each of the experiments, the trend described for the first two cases remains:
─ max T3DSS1 min T3DSS2 (for p32 = 0 and p32 = 0.1 );
─ max T3DSS1 min T3DSS2 (for p32 = 0.3 );
─ max T3DSS1 min T3without MIS and max T3DSS2 min T3without MIS (for all values p32 ).
6 Conclusion
In this work, using the simplified structural model of an ECG study, analytical ex-
pressions were obtained to calculate the average execution time of this process for
three different types of studies: the study is conducted for the first time, the study is
repeated as a result of screening, the study is repeated after treatment.
Using the obtained analytical expressions, an analysis of the time characteristics of
the ECG study was performed without using and using cardiological DSS separately
for each of the considered study types. The above data show that the use of any cardi-
ological DSS significantly reduces the time for the ECG study of each of the consid-
ered types, even if the worst option of the ECG study using any cardiological DSS
was being compared with the best option of the ECG study without using any MIS.
Moreover, if cardiological DSS is used with an improved module for the morphologi-
cal analysis of BMS with LCF (DSS1) then even the best option for conducting the
ECG study using DSS2 in almost all cases is inferior in time to the worst option for
conducting the ECG study using DSS1.
Further studies are aimed at experimental verification of the effectiveness of ECG
studies both in time and in the probability of successful completion of the considered
process under various initial conditions using the proposed full structural model.
References
1. Timchik SV, Zlepko SM, Kostishyn SV (2016) Klasyfikacija medychnyh informacijnyh
system i tehnologij za integral'nym sukupnym kryterijem [Classification of medical infor-
mation systems and technologies by the integrated cumulative criterion]. Information pro-
cessing systems, 3 (140), pp 194–198 (in Ukrainian)
2. Doan DCh, Kroschilin AW, Kroschilina CW (2015) Obsor podchodow k probleme prin-
jatija reschenij w medizinckich invormazionnych cictemach w uclowijach neopredelen-
nocti [Overview of approaches to the problem of decision-making in medical information
systems in the face of uncertainty], Basic research, 12 (1), pp 26–30 (in Russian)
3. Gucew AW, Sarubina T W (2017) Podderzhka prinjatija wratschebnych reschenij w
medizinckich invormazionnych cictemach medizinckoj organisazii [Support for the adop-
tion of medical decisions in the medical information systems of a medical organization].
Doctor and information technology, 2, pp 60–72 (in Russian)
4. Fainzylberh LS (2015) Obobschennyiy metod obrabotki tsiklicheskih signalov slozhnoy
formyi v mnogomernom prostranstve parametrov [A generalized method for processing
cyclic signals of complex shape in a multidimensional parameter space]. International Sci-
entific and Technical Journal “Problems of control and informatics”, 2, pp 58–71 (in Rus-
sian)
5. Filatova AE, Povoroznyuk AI, Fahs M. (2020) Application of probabilistic-time graphs for
evaluating the effectiveness of the electrocardiological study process. Scientific Journal
Applied Aspects of Information Technology, 3 (1), pp 405–417.
doi:10.15276/aait.01.2020.3
6. Leski JM, Henzel N (2004) ECG baseline wander and powerline interference reduction us-
ing nonlinear filter bank. Signal Process, 35 (4), pp 781–793
7. Iravanian S, Tung L (2002) A novel algorithm for cardiac biosignal filtering based on fil-
tered residue method. IEEE Transactions on Biomedical Engineering, 49 (11), pp 1310–
1317. doi:10.1109/TBME.2002.804589
8. Ho CY-F, Ling BW-K, Wong TP-L, Chan AY-P, Tam PK-S (2003) Fuzzy multiwavelet
denoising on ECG signal. Electronics Letters, 39 (16), pp 1163–1164.
doi:10.1049/el:20030757
9. Ercelebi E (2004) Electrocardiogram signals de-noising using lifting-based discrete wave-
let transform. Computers in Biology and Medicine, 34 (6), pp 479–493.
doi:10.1016/S0010-4825(03)00090-8
10. Poornachandra S, Kumaravel N (2005) Hyper-trim shrinkage for denoising of ECG signal.
Digital Signal Process, 15 (3), pp 317–327. doi:10.1016/j.dsp.2004.12.005
11. Blinow PA, Micheew AA (2009) Analis metodow kompencazii drejva isolinii jelektrokar-
diocignala [Analysis of compensation methods for the drift of the contour of an electrocar-
diogram signal]. Herald of RSRTU, Ryazan, Russian Federation, RSRTU, 4 (30), pp 94–
97 (in Russian)
12. Blinow PA (2011) Algoritmy uctranenija drejva isolinii jelektrokardiocignala na ocnowe
preobrasowanija ego cpektra [Algorithms for eliminating the drift of the isoline of an elec-
trocardiogram based on the conversion of its spectrum]: the abstract of the dissertation of
the candidate of technical sciences (05.11.17). Ryazan, Russian Federation, 16 p. (in Rus-
sian)
13. Abawajy JH, Kelarev AV, Chowdhury M (2013) Multistage approach for clustering and
classification of ECG data. Computer Methods and Programs in Biomedicine, 112 (3), pp
720–730. doi:10.1016/j.cmpb.2013.08.002
14. Pasollia E, Melgani F (2015) Genetic algorithm-based method for mitigating label noise
issue in ECG signal classification. Biomedical Signal Processing and Control, 19, pp 130–
136. doi:10.1016/j.bspc.2014.10.013
15. Ruvinskaya VM, Shevchuk I, Michaluk N (2019) Models based on conformal predictors
for diagnostic systems in medicine. Applied Aspects of Information Technology, 2 (2), pp
127–137. doi:10.15276/aait.02.2019.4
16. Serbest K, Bozkurt MR, Eldoğan O (2015) Classification of cardiac arrhythmias with arti-
ficial neural networks according to gender differences. Journal of Engineering Science and
Technology, 10(9), pp 1144–1149
17. Ceylana R, Özbaya Y, Karlikb B (2009) A novel approach for classification of ECG ar-
rhythmias: Type-2 fuzzy clustering neural network. Expert Systems with Applications, 36
(3), pp 6721–6726. doi:10.1016/j.eswa.2008.08.028
18. Doğan B, Korürek M (2012) A new ECG beat clustering method based on kernelized
fuzzy c-means and hybrid ant colony optimization for continuous domains. Applied Soft
Computing, 12 (11), pp 3442–3451. doi:10.1016/j.asoc.2012.07.007
19. Dubrowin WI, Twerdochleb JuW, Chartschenko WW (2014) Awtomatisirowannaja cicte-
ma analisa i interpretazii JeKG [Automated ECG analysis and interpretation system]. Ra-
dio electronics, computer science, management, 1, pp 150–157 (in Russian)
20. Kolomejzewa AW, Mischugowa GW, Mul AP, Rjabych GJu (2010) Primenenie wejwlet-
preobrasowanija i metoda Proni dlja identivikazii biogennych cignalow [Application of
wavelet transform and Proni method for identification of biogenic signals]. Herald of
DSTU, Rostov-on-Don, Russian Federation, DSTU, 10, 4(47), pp 455–465 (in Russian)
21. Povoroznyuk AI, Filatova AE, Zakovorotniy AYu, Shehna Kh (2019) Development of
method of matched morphological filtering of biomedical signals and images. Automatic
Control and Computer Sciences, 53 (3), pp 253–262. doi:10.3103/S014641161903009X
22. Selcan Kaplan Berkaya, Alper Kursat Uysal, Efnan Sora Gunal, Semih Ergin, Serkan
Gunal, M. Bilginer Gulmezoglu (2018) A survey on ECG analysis. Biomedical Signal
Processing and Control, 43, pp 216–235. doi:10.1016/j.bspc.2018.03.003
23. Lyon A, Minchole´ A, Martı´nez JP, Laguna P, Rodriguez B (2018) Computational tech-
niques for ECG analysis and interpretation in light of their contribution to medical advanc-
es. J. R. Soc. Interface 15: 20170821. doi:10.1098/rsif.2017.0821