Прогнозування кардіоміопатії у пацієнтів з постійною шлуночковою електрокардіостимуляцією за допомогою методів машинного навчання
Pacing-induced cardiomyopathy is a notable issue in patients needing permanent ventricular pacing. Identifying risk groups early and swiftly preventing the ailment can reduce patient harm. However, current prognostic methods require clarity. We employed machine learning to develop predictive models...
Gespeichert in:
| Datum: | 2024 |
|---|---|
| Hauptverfasser: | , , , , |
| Format: | Artikel |
| Sprache: | Englisch |
| Veröffentlicht: |
The National Technical University of Ukraine "Igor Sikorsky Kyiv Polytechnic Institute"
2024
|
| Schlagworte: | |
| Online Zugang: | https://journal.iasa.kpi.ua/article/view/285956 |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| Назва журналу: | System research and information technologies |
| Завантажити файл: |  |
Institution
System research and information technologies| _version_ | 1867334437675466752 |
|---|---|
| author | Perepeka, Eugene Lazoryshynets, Vasyl Babenko, Vitalii Davydovych, Illia Nastenko, Ievgen |
| author_facet | Perepeka, Eugene Lazoryshynets, Vasyl Babenko, Vitalii Davydovych, Illia Nastenko, Ievgen |
| author_institution_txt_mv | [
{
"author": "Eugene Perepeka",
"institution": "Amosov National Institute of Cardiovascular Surgery, Kyiv"
},
{
"author": "Vasyl Lazoryshynets",
"institution": "Amosov National Institute of Cardiovascular Surgery, Kyiv"
},
{
"author": "Vitalii Babenko",
"institution": "National Technical University of Ukraine \"Igor Sikorsky Kyiv Polytechnic Institute\", Kyiv"
},
{
"author": "Illia Davydovych",
"institution": "National Technical University of Ukraine \"Igor Sikorsky Kyiv Polytechnic Institute\", Kyiv"
},
{
"author": "Ievgen Nastenko",
"institution": "National Technical University of Ukraine \"Igor Sikorsky Kyiv Polytechnic Institute\", Kyiv"
}
] |
| author_sort | Perepeka, Eugene |
| baseUrl_str | http://journal.iasa.kpi.ua/oai |
| collection | OJS |
| datestamp_date | 2024-05-23T07:09:36Z |
| description | Pacing-induced cardiomyopathy is a notable issue in patients needing permanent ventricular pacing. Identifying risk groups early and swiftly preventing the ailment can reduce patient harm. However, current prognostic methods require clarity. We employed machine learning to develop predictive models using medical data. Three algorithms — decision tree, group method of data handling, and logistic regression — formed models that forecast pacing-induced cardiomyopathy. These models displayed high accuracy in predicting development, signifying soundness. Factors like age, paced QRS width, pacing mode, and ventricular index during implantation significantly influenced predictions. Machine learning can enhance pacing-induced cardiomyopathy prediction in ventricular pacing patients, aiding medical practice and preventive strategies. |
| doi_str_mv | 10.20535/SRIT.2308-8893.2024.1.03 |
| first_indexed | 2025-07-17T10:28:19Z |
| format | Article |
| fulltext |
E.O. Perepeka, V.V. Lazoryshynets, V.O. Babenko, I.V. Davydovych, I.A. Nastenko, 2024
Системні дослідження та інформаційні технології, 2024, № 1 33
UDC 004.852 + 616.12-07
DOI: 10.20535/SRIT.2308-8893.2024.1.03
CARDIOMYOPATHY PREDICTION IN PATIENTS
WITH PERMANENT VENTRICULAR PACING USING
MACHINE LEARNING METHODS
E.O. PEREPEKA, V.V. LAZORYSHYNETS, V.O. BABENKO,
I.V. DAVYDOVYCH, I.A. NASTENKO
Abstract. Pacing-induced cardiomyopathy is a notable issue in patients needing
permanent ventricular pacing. Identifying risk groups early and swiftly preventing
the ailment can reduce patient harm. However, current prognostic methods require
clarity. We employed machine learning to develop predictive models using medical
data. Three algorithms — decision tree, group method of data handling, and logistic
regression — formed models that forecast pacing-induced cardiomyopathy. These
models displayed high accuracy in predicting development, signifying soundness.
Factors like age, paced QRS width, pacing mode, and ventricular index during im-
plantation significantly influenced predictions. Machine learning can enhance pac-
ing-induced cardiomyopathy prediction in ventricular pacing patients, aiding medi-
cal practice and preventive strategies.
Keywords: permanent ventricular pacing, risk factors, artificial intelligence, fore-
casting, machine learning.
INTRODUCTION
Right ventricular myocardial pacing remains dominating method in providing
medical care to patients with various potentially fatal bradyarrhythmias, even
though at the beginning of the 21st century, a relation between this form of car-
diac pacing and the left ventricular contractility impairment [1], as well as dete-
rioration of clinical outcomes in the distant period [2; 3].
According to data from various sources, the incidence of pacing-induced
cardiomyopathy (PICM) in patients with conventional right ventricular pacing
and with preserved initial left ventricle ejection fraction (LVEF) ranges from 7.5
to 26% [4–10].
The risk of heart failure hospitalizations (HFH) and overall mortality are
significantly higher among patients with PICM, as was shown in a large retro-
spective study by Sung Woo Cho et al. [10]. Though in patients with initially re-
duced systolic function of the left ventricle and high burden of ventricular pacing,
the factors of deterioration of the clinical outcomes are well established [3], in
patients with preserved LVEF, they have not yet been fully studied. Along with
the wide availability and significant global experience of using this method of
cardiac pacing in clinical practice, there is a growing number of publications fo-
cusing on the adverse effects of right ventricular myocardial pacing (and investi-
gating risk factors that led to them), one of which is the development of the so-
called pacing-induced cardiomyopathy, which is characterized by a decrease in
the left ventricle contractility and negative remodeling of the heart chambers,
E.O. Perepeka, V.V. Lazoryshynets, V.O. Babenko, I.V. Davydovych, I.A. Nastenko
ISSN 1681–6048 System Research & Information Technologies, 2024, № 1 34
The identification of risk factors and prediction of PICM development in patients
with an implanted pacemaker is an objective of significant importance for modern
medicine, considering the appearance of modern physiological methods of cardiac
pacing (such as conduction system pacing) which allow preventing or minimizing
the negative consequences of right ventricular myocardial pacing [11–14]. It is
important to note that machine learning and artificial intelligence are becoming
more prevalent in healthcare, particularly in cardiology. These technologies have
successfully predicted disease cases and identified pathologies [15]. However,
studies that apply machine learning to indicate PICM were not found after analyz-
ing various literature sources.
The primary focus of research in the intersection of cardiology and machine
learning is centered around the prediction and diagnosis of diseases, including
ischemic heart disease (IHD) [16; 17], HF [18], atrial arrhythmias [19; 20], and
others, using data from patients’ medical records, imaging, and biosignals. In the
context of PICM, the scientific community focuses on studying risk factors and
developing preventive measures [21; 22]. Thus, the use of machine learning can
contribute to identifying patients at considerable risk of PICM, which will allow
the introduction of prompt and effective therapeutic interventions or other inva-
sive strategies. This study focuses on figuring out the possibilities of using the
machine learning methodology to predict the development of PICM in patients
with permanent ventricular pacing.
Specific tasks due to the urgency of the problem are determined by the fol-
lowing aspects:
1. Development of PICM prediction models based on various machine
learning algorithms using the available medical dataset.
2. Comprehensive evaluation of constructed models using classification
metrics including (but not limited to) accuracy, sensitivity, and specificity.
3. A detailed study of the importance of individual factors included in the
model in the context of their influence on predicting PICM.
The objective of the study is the construction of detailed prognostic models
for the development of PICM and the identification of critical factors that con-
tribute to the occurrence of this complication.
MATERIALS AND METHODS
In this research, we used anonymized data from patient examinations performed
at the State Institution “M. Amosov National Institute of Cardiovascular Surgery”
of the National Academy of Medical Sciences of Ukraine within the framework
of the cooperation agreement with the National Technical University of Ukraine
“Igor Sikorsky Kyiv Polytechnic Institute”.
Before initiating the study, the M. Amosov National Institute of
Cardiovascular Surgery performed a bioethical evaluation of the research
protocol. We analyzed data on thirty-four patients, of which nine (26.5%) were
diagnosed with PICM, which was determined with an LVEF of less than 45%.
Left ventricle ejection fraction was within normal limits in the remaining twenty-
five patients (73.5%).
The study included only those patients who met the following criteria: avail-
ability of echocardiographic data at the time of pacemaker (PM) implantation;
Cardiomyopathy prediction in patients with permanent ventricular pacing using …
Системні дослідження та інформаційні технології, 2024, № 1 35
total percentage of ventricular pacing at the time of examination is not less than
90%; preserved LVEF at the time of implantation (≥ 50%); age restrictions pa-
tients (18–80 years at the time of implantation and control examination, respec-
tively); this was to be a primary PM implantation without previous endocardial
lead extractions or power source replacements.
For the study, the M. Amosov National Institute of Cardiovascular Surgery
systematically collected and documented data, which included gender, age, the
period from PM implantation to follow-up, as well as the main and concomitant
diagnoses of the patients.
In addition, the data from echocardiographic and electrocardiographic
studies were collected, as well as cardiac pacing parameters at two stages: at the
time of hospitalization and during the control examination.
It is important to note that the used database has seventeen attributes, de-
scribed in detail in Table 1.
T a b l e 1 . Attributes of the selected database
Attribute1 Data type Symbolic
notation
PICM Binary y
Gender Binary x1
Age Continuous integer x2
Time from pacemaker implantation to follow-up Continuous integer x3
LVEF at the time of implantation Continuous integer x4
LA diameter at the time of PM implantation Continuous integer x5
Width of native QRS complex Continuous integer x6
Width of paced QRS complex Continuous integer x7
Presence of atrial arrhythmias (including AF) Binary x8
Right ventricle pacing site Binary x9
Structural heart diseases Binary x10
Diabetes mellitus Binary x11
Hypertension Binary x12
Ischemic heart disease Binary x13
Pacemaker type (single-chamber/dual-chamber) Binary x14
Rate-adaptive pacing mode Binary x15
Left ventricle EDI at the time of PM implantation Continuous х16
The purpose of applying machine learning technologies was to find key in-
put (independent) variables x that correlate with the presence of cardiomyopathy,
represented as an output (dependent) variable y . Machine learning aims to iden-
tify patterns and relationships between variables through data processing. Ma-
chine learning algorithms are designed to explore dependencies in data and show
trends that may not be clear. A substantial number of scientific developments con-
firmed this hypothesis, where authors considered tasks from various subject areas,
including medicine [23–25].
1 Accepted abbreviations: PICM — stimulation-induced cardiomyopathy; AP — artificial pace-
maker; LVEF — left ventricular ejection fraction; LA — left atrium; AF — atrial fibrillation; EDI
— end-diastolic index.
E.O. Perepeka, V.V. Lazoryshynets, V.O. Babenko, I.V. Davydovych, I.A. Nastenko
ISSN 1681–6048 System Research & Information Technologies, 2024, № 1 36
As can be seen from Table 1, the output variable y is dichotomous, which
writes down the need to solve the classification problem.
Taking this into account, we decided to use three simple classification algo-
rithms: decision tree [26], group method of data handling (GMDH) [27], and lo-
gistic regression [28].
Decision trees are one of the most convenient algorithms because of their
visual interpretation and ability to oversee numerical and categorical data. They
work by partitioning the space of input variables into regions corresponding to
different classes of the output variable. However, they can be prone to overfitting,
especially with complex data.
GMDH is an algorithm that creates a model based on a pairwise comparison
of objects. Its main advantage is the high interpretability of the results, which
supplies the possibility of a clear understanding of the classification mechanisms.
However, due to high computational complexity, GMDH may only be effective
for a small volume of data.
Logistic regression is a statistical algorithm commonly used to predict the
probability of an event occurring by applying a logistic function. This method
works well on two-class problems but can run into issues with non-linear relation-
ships or many categorical variables.
RESULTS
Before building PICM prediction models, we divided the patient sample into a
train (80%, or twenty-seven patients) and a test (20%, or seven patients) using a
stratification method, which preserves the class ratio between subjects in each
sample. We measured the performance of each algorithm by its accuracy (propor-
tion of correctly classified patients), sensitivity (proportion of correctly classified
patients with pathology), and specificity (proportion of correctly classified
healthy patients) [29]. The performance of the selected classification algorithms is
presented in Table 2.
T a b l e 2 . Evaluation of the constructed PICM prediction models by classification
metrics
Train (80%) Test (20%)
Classifier
Accuracy Sensitivity Specificity Accuracy Sensitivity Specificity
Decision tree 1.000 1.000 1.000 1.000 1.000 1.000
GMDH 1.000 1.000 1.000 1.000 1.000 1.000
Logistic regression 0.852 0.857 0.850 1.000 1.000 1.000
Using the scikit-learn (the Python library), we implemented the classifica-
tion model (Figure) based on the decision tree method. The tree has a depth of
five and consists of nine leaves.
According to the data presented in Table 2, this model shows 100% accuracy on
the test sample, indicating its reliability and the absence of overfitting phenomena.
Six input variables: 152763 ,,,, xxxxx , and 16x , were used in the model and
are illustrated in the tree (Fig. 1). We expressed each variable’s impact through
the tree’s weights: 303.03 x , 193.06 x , 146.07 x , 127.02 x , 118.015 x ,
113.016 x . The weighting coefficients were figured out using the Gini index [26]. It
Cardiomyopathy prediction in patients with permanent ventricular pacing using …
Системні дослідження та інформаційні технології, 2024, № 1 37
considers the number of times each variable was used to split the data and how
effective this split was. The final weighting of each variable is a normalized value
based on the overall reduction in the Gini index caused by each variable. A vari-
able with a higher importance value is considered more “important” to the model.
Applying the GMDH, we generated a model for which the formula below
is given:
16121221413167
2
16 006.0006.0643.00009.0001.0 xxxxxxxxxy
736.0519.0018.00003.0031.0 1211124
2
7159 xxxxxxx
We conducted the training process by using GMDH Shell DS software. As
shown in Table 2, the results show that the GMDH model is completely accurate
x16 ≤ 67.674
gini = 0.5
samples = 270
value = [135.0, 135.0]
class = 2
x6 ≤ 152.454
gini = 0.366
samples = 100
value = [60.75, 19.286]
class = 1
x3 ≤ 33.641
gini = 0.476
samples = 170
value = [74.25, 115.714]
class = 2
4. gini = 0.0
samples = 80
value = [54.0, 0.0]
class = 1
x2 ≤ 68.34
gini = 0.254
samples = 60
value = [13.5, 77.143]
class = 2
5. x2 ≤ 60.71
gini = 0.384
samples = 20
value = [6.75, 19.286]
class = 2
7. x3 ≤ 51.819
gini = 0.475
samples = 110
value = [60.75, 38.571]
class = 1
8. gini = 0.0
samples = 10
value = [6.75, 0.0]
class = 1
9. gini = 0.0
samples = 10
value = [0.0, 19.286]
class = 2
10. gini = 0.0
samples = 30
value = [0.0, 57.857]
class = 2
11. x15 ≤ 1.15
gini = 0.484
samples = 30
value=[13.5, 19.286]
class = 2
12. gini = 0.0
samples = 60
value = [40.5, 0.0]
class = 1
13. x6 ≤ 128.428
gini = 0.451
samples = 50
value = [20.25, 38.571]
class = 2
14. gini = 0.0
samples = 10
value = [0.0, 19.286]
class = 2
15. gini = 0.0
samples = 20
value = [13.5, 0.0]
class = 1 16. x7 ≤ 160.001
gini = 0.5
samples = 40
value=[20.25, 19.286]
class = 1
17. gini = 0.0
samples = 10
value = [0.0, 19.286]
class = 2
18. gini = 0.0
samples = 30
value = [20.25, 0.0]
class = 1
19. gini = 0.0
samples = 10
value = [0.0, 19.286]
class = 2
Decsion tree model
E.O. Perepeka, V.V. Lazoryshynets, V.O. Babenko, I.V. Davydovych, I.A. Nastenko
ISSN 1681–6048 System Research & Information Technologies, 2024, № 1 38
in testing, indicating no model overfitting. This model includes ten independent
variables, namely: 121471316 ,,,, xxxxx , 92411 ,,, xxxx , and 15x . The weight of
each of these variables is found based on the change in the model’s predicted val-
ues when replacing the variable’s actual values with its average value. As a result,
the following variable weights were obtained: %7.8316 x , %9.5313 x ,
%2.537 x , %7.3914 x , %4.3312 x , %8.3011 x , %6.174 x , %4.62 x ,
%6.19 x , %2.015 x .
In the third step, we applied a logistic regression model. The general form of
the logistic model is defined by formula:
ye
p
1
1
,
where p is the calculated probability of occurrence of a given event (PICM in
this context); e is the basis of natural logarithms (2.713); y is the linear regression
equation. The following logistic regression model was obtained:
179.8119.0758.1044.0065.0 161572 xxxxy .
We conducted the training procedure using the scikit-learn package of the
Python programming language. The complexity of this model is four. Interest-
ingly, this model includes variables also used in the earlier models: 1572 ,, xxx ,
and 16x .
DISCUSSION
While analyzing constructed models for predicting pacing-induced cardiomyopa-
thy based on the data in Table 2, it was found that only the logistic regression
model failed to present an ideal result for the entire sample (with a classification
accuracy of 85.2% in the train, despite 100% accuracy in the test). The observed
phenomenon can be explained by the intrinsic simplicity of the logistic model in
contrast to the other comparable models utilized in the research.
The developed decision tree model was structurally simple and included
only six independent variables. While the results of the classification estimation
are excellent, this model may be prone to misprediction of new data due to the
limited initial sample size. The GMDH model wins here by incorporating ten in-
dependent variables for prediction. Additionally, the algorithm for constructing
such a model allows non-linear combinations of variables, which sensitively in-
creases their predictive power.
The identified combinations of factors influencing the PICM development
align with the latest global publications. The three prediction models include the
following independent variables: x2 (patient’s age), x7 (width of paced QRS com-
plex), x15 (rate-adaptive pacing mode), and x16 (left ventricular EDI at the time of
PM implantation).
Among them, variable x7 has a significant impact, especially in the decision
tree (0.147) and GMDH (53.2%), with one of the highest weighting coefficients.
There are also independent variables that were not included in any of the models,
such as x1 (patient gender), x5 (LA diameter at the time of PM implantation), x8
(presence of atrial arrhythmias), and x10 (structural heart diseases).
Cardiomyopathy prediction in patients with permanent ventricular pacing using …
Системні дослідження та інформаційні технології, 2024, № 1 39
The modeling results obtained during the study open the possibility of pre-
dicting undesirable clinical consequences of right ventricular pacing based on
combinations of the most informative factors. That makes it possible to prevent
the influence of these factors or intervene at the stage of medical care provided,
choosing more physiological cardiac pacing methods.
CONCLUSION
The study successfully developed models for predicting pacing-induced cardio-
myopathy (PICM) based on various machine learning algorithms using an avail-
able medical dataset of thirty-four patients.
Methods used — including decision tree, group method of data handling
(GMDH), and logistic regression — allowed robust predictive models to be cre-
ated. On the test sample, all of them showed 100% prediction accuracy.
Obtained results demonstrated the high efficiency of the used machine learn-
ing algorithms in terms of the accuracy of the PICM prediction, the absence of
overfitting, and the ability of the models to classify adequately normal and patho-
logical states of patients.
A detailed study of values included in the models allows an understanding of
their role in developing PICM.
The most significant data included in the models were patient age, paced
QRS complex width, rate-adaptive pacing mode, and left ventricular end-diastolic
index (EDI) at the time of pacemaker implantation.
The developed models can serve as a basis for further improving diagnostic
and treatment technologies for PICM prevention strategies.
REFERENCES
1. J.C. Nielsen, L. Kristensen, H.R. Andersen, P.T. Mortensen, O.L. Pedersen, and
A.K. Pedersen, “A randomized comparison of atrial and dual-chamber pacing in 177
consecutive patients with sick sinus syndrome,” Journal of the American College of
Cardiology, 42(4), pp. 614–23, 2003. doi: 10.1016/S0735-1097(03)00757-5.
2. M.O. Sweeney et al., “Adverse Effect of Ventricular Pacing on Heart Failure and
Atrial Fibrillation Among Patients With Normal Baseline QRS Duration in a Clinical
Trial of Pacemaker Therapy for Sinus Node Dysfunction,” Circulation, 107(23), pp.
2932–2937, 2003. doi: 10.1161/01.CIR.0000072769.17295.B1.
3. The DAVID Trial Investigators*, “Dual-Chamber Pacing or Ventricular Backup Pac-
ing in Patients With an Implantable Defibrillator,” JAMA, 288(24), pp. 3115–3123,
2002. doi: 10.1001/jama.288.24.3115.
4. S. Khurshid et al., “Incidence and predictors of right ventricular pacing-induced cardio-
myopathy,” Hear Rhythm, 11(9), pp. 1619–1625, 2014. doi: 10.1016/j.hrthm.2014.05.040.
5. E.L. Kiehl et al., “Incidence and predictors of right ventricular pacing-induced car-
diomyopathy in patients with complete atrioventricular block and preserved left ven-
tricular systolic function,” Hear Rhythm, 13(12), pp. 2272–2278, 2016. doi:
10.1016/j.hrthm.2016.09.027.
6. A. Abdin et al., “Incidence and predictors of pacemaker induced cardiomyopathy:
A single-center experience,” Journal of Electrocardiology, 57, pp. 31–34, 2019. doi:
10.1016/j.jelectrocard.2019.08.016.
7. Mohamed Abdelmohsen Sayed, Haitham Abd El Fatah Badran, Said Khaled, and
Emad Effat Fakhry, “Predictors of right ventricular pacing-induced left ventricular
dysfunction in pacemaker recipients with preserved ejection fraction,”
Herzschrittmachertherapie + Elektrophysiologie, 33(3), pp. 312–318, 2022. doi:
10.1007/s00399-022-00880-w.
E.O. Perepeka, V.V. Lazoryshynets, V.O. Babenko, I.V. Davydovych, I.A. Nastenko
ISSN 1681–6048 System Research & Information Technologies, 2024, № 1 40
8. H.T. Perla, S. Chandra Srinath Patloori, A. Manickavasagam, D. Chase, and
J. Roshan, “Do the predictors of right ventricular pacing-induced cardiomyopa-
thy add up?” Indian Heart Journal, 73(5), pp. 582–587, 2021. doi:
10.1016/j.ihj.2021.07.011.
9. J.H. Kim, K.-W. Kang, J.Y. Chin, T.-S. Kim, J.-H. Park, and Y.J. Choi, “Major de-
terminant of the occurrence of pacing-induced cardiomyopathy in complete atrioven-
tricular block: a multicentre, retrospective analysis over a 15-year period in South
Korea,” BMJ Open, 8(2), pp. e019048, 2018. doi: 10.1136/bmjopen-2017-019048.
10. S.W. Cho et al., “Clinical features, predictors, and long-term prognosis of pacing-
induced cardiomyopathy,” European Journal of Heart Failure, 21(5), pp. 643–651,
2019. doi: 10.1002/ejhf.1427.
11. M. Abdelrahman et al., “Clinical Outcomes of His Bundle Pacing Compared to
Right Ventricular Pacing,” Journal of the American College of Cardiology, 71(20),
pp. 2319–2330, 2018. doi: 10.1016/j.jacc.2018.02.048.
12. P.S. Sharma et al., “Clinical outcomes of left bundle branch area pacing compared to
right ventricular pacing: Results from the Geisinger-Rush Conduction System Pacing
Registry,” Hear Rhythm, 19(1), pp. 3–11, 2022. doi: 10.1016/j.hrthm.2021.08.033.
13. P. Vijayaraman et al., “Outcomes of His-bundle pacing upgrade after long-term right
ventricular pacing and/or pacing-induced cardiomyopathy: Insights into disease pro-
gression,” Hear Rhythm, 16(10), pp. 1554–1561, 2019. doi: 10.1016/j.hrthm.2019.03.026.
14. L.M. Rademakers, S. Bouwmeester, T.P. Mast, L. Dekker, P. Houthuizen, and F.A.
Bracke, “Feasibility, safety and outcomes of upgrading to left bundle branch pacing
in patients with right ventricular pacing induced cardiomyopathy,” Pacing and Clini-
cal Electrophysiology, 45(6), pp. 726–732, 2022. doi: 10.1111/pace.14515.
15. K. Basu, R. Sinha, A. Ong, and T. Basu, “Artificial intelligence: How is it changing
medical sciences and its future?” Indian Journal of Dermatology, 65(5), pp. 365–370,
2020. doi: 10.4103/ijd.IJD_421_20.
16. Ch Anwar Ul Hassan et al., “Effectively Predicting the Presence of Coronary Heart
Disease Using Machine Learning Classifiers,” Sensors, 22(19), 7227, 2022. doi:
10.3390/s22197227.
17. O. Petrunina, D. Shevaga, V. Babenko, V. Pavlov, S. Rysin, and I. Nastenko, “Com-
parative Analysis of Classification Algorithms in the Analysis of Medical Images
From Speckle Tracking Echocardiography Video Data,” Innovative Biosystems and
Bioengineering, 5(3), pp. 153–166, 2021. doi: 10.20535/ibb.2021.5.3.234990.
18. J. Wang, “Heart Failure Prediction with Machine Learning: A Comparative Study,”
Journal of Physics: Conference Series, 2031(1), 12068, 2021. doi: 10.1088/1742-
6596/2031/1/012068.
19. M. Varshneya, X. Mei, and E.A. Sobie, “Prediction of arrhythmia susceptibility
through mathematical modeling and machine learning,” Proceedings of the National
Academy of Sciences, 118(37), 12068, 2021. doi: 10.1073/pnas.2104019118.
20. E.R. Pujadas et al., “Atrial fibrillation prediction by combining ECG markers and CMR
radiomics,” Scientific Reports, 12(1), 18876, 2022. doi: 10.1038/s41598-022-21663-w.
21. P.Z. Fruelund et al., “Risk of Pacing-Induced Cardiomyopathy in Patients with High-
Degree Atrioventricular Block—Impact of Right Ventricular Lead Position Con-
firmed by Computed Tomography,” Journal of Clinical Medicine, 11(23), 7228,
2022. doi: 10.3390/jcm11237228.
22. J. Mizner, P. Jurak, H. Linkova, R. Smisek, and K. Curila, “Ventricular Dyssyn-
chrony and Pacing-induced Cardiomyopathy in Patients with Pacemakers, the Utility
of Ultra-high-frequency ECG and Other Dyssynchrony Assessment Tools,” Ar-
rhythmia & Electrophysiology Review, 11, e.17, 2022. doi: 10.15420/aer.2022.01.
23. I.H. Sarker, “Machine Learning: Algorithms, Real-World Applications and Research
Directions,” SN Computer Science, 2(3), 160, 2021. doi: 10.1007/s42979-021-00592-x.
24. A. Rajkomar, J. Dean, and I. Kohane, “Machine Learning in Medicine,” The New Eng-
land Journal of Medicine, 380(14), pp. 1347–1358, 2019. doi: 10.1056/nejmra1814259.
25. V. Babenko et al., “Classification of Pathologies on Medical Images Using the Algo-
rithm of Random Forest of Optimal-Complexity Trees,” Cybernetics and Systems
Analysis, 59(2), pp. 346–358, 2023. doi: 10.1007/s10559-023-00569-z.
Cardiomyopathy prediction in patients with permanent ventricular pacing using …
Системні дослідження та інформаційні технології, 2024, № 1 41
26. B. Charbuty, A. Abdulazeez, “Classification Based on Decision Tree Algorithm for
Machine Learning,” Journal of Applied Science and Technology Trends, 2(1),
pp. 20–28, 2021. doi: 10.38094/jastt20165.
27. V. Vaishnav, J. Vajpai, “Assessment of impact of relaxation in lockdown and
forecast of preparation for combating COVID-19 pandemic in India using Group
Method of Data Handling,” Chaos, Solitons & Fractals, 140, 110191, 2020. doi:
10.1016/j.chaos.2020.110191.
28. C. El Morr, M. Jammal, H. Ali-Hassan, and W. El-Hallak, “Logistic Regression,” in
book “Machine Learning for Practical Decision Making,” pp. 231–249, 2022. doi:
10.1007/978-3-031-16990-8_7.
29. Ž.Ð. Vujovic, “Classification Model Evaluation Metrics,” International Journal
of Advanced Science and Computer Applications, 12(6), 2021. doi:
10.14569/IJACSA.2021.0120670.
Received 15.08.2023
INFORMATION ON THE ARTICLE
Eugene O. Perepeka, ORCID: 0000-0001-9755-8825, Amosov National Institute of Car-
diovascular Surgery, Ukraine, e-mail: eugeneperepeka@gmail.com
Vasyl V. Lazoryshynets, ORCID: 0000-0002-1748-561X, Amosov National Institute of
Cardiovascular Surgery, Ukraine, e-mail: lazorch@ukr.net
Vitalii O. Babenko, ORCID: 0000-0002-8433-3878, National Technical University of
Ukraine “Igor Sikorsky Kyiv Polytechnic Institute”, Ukraine, e-mail: vba-
benko2191@gmail.com
Illia V. Davydovych, ORCID: 0000-0001-9987-8267, National Technical University of
Ukraine “Igor Sikorsky Kyiv Polytechnic Institute”, Ukraine, e-mail:
bkmzbkmz6@gmail.com
Ievgen A. Nastenko, ORCID: 0000-0002-1076-9337, National Technical University
of Ukraine “Igor Sikorsky Kyiv Polytechnic Institute”, Ukraine, e-mail: nas-
tenko.e@gmail.com
ПРОГНОЗУВАННЯ КАРДІОМІОПАТІЇ У ПАЦІЄНТІВ З ПОСТІЙНОЮ
ШЛУНОЧКОВОЮ ЕЛЕКТРОКАРДІОСТИМУЛЯЦІЄЮ ЗА ДОПОМОГОЮ
МЕТОДІВ МАШИННОГО НАВЧАННЯ / Є.О. Перепека, В.В. Лазоришинець,
В.О. Бабенко, І.В. Давидович, Є.А. Настенко
Анотація. Кардіоміопатія, спричинена кардіостимуляцією, є важливою про-
блемою для пацієнтів, які потребують постійної шлуночкової кардіостимуля-
ції. Раннє виявлення груп ризику та швидка профілактика недуги можуть зме-
ншити шкоду для пацієнтів. Однак сучасні методи прогнозування потребують
доопрацювання. Застосовано машинне навчання для розроблення прогностич-
них моделей на основі медичних даних. Три алгоритми — дерево рішень, гру-
па оброблення даних та логістична регресія — сформували моделі, які прогно-
зують кардіоміопатію, спричинену кардіостимуляцією. Ці моделі показали
високу точність у прогнозуванні розвитку, що свідчить про їх надійність.
Ключові фактори, такі як вік, ширина QRS, режим кардіостимуляції та шлуно-
чковий індекс під час імплантації, суттєво впливали на прогнози. Машинне
навчання може покращити прогнозування кардіоміопатії, спричиненої кардіо-
стимуляцією, у пацієнтів, які перебувають на шлуночковій електрокардіости-
муляції, допомагаючи медичній практиці та профілактичним стратегіям.
Ключові слова: постійне ритмоведення шлуночків, фактори ризику, штучний
інтелект, прогнозування, машинне навчання.
|
| id | journaliasakpiua-article-285956 |
| institution | System research and information technologies |
| keywords_txt_mv | keywords |
| language | English |
| last_indexed | 2025-07-17T10:28:19Z |
| publishDate | 2024 |
| publisher | The National Technical University of Ukraine "Igor Sikorsky Kyiv Polytechnic Institute" |
| record_format | ojs |
| resource_txt_mv | journaliasakpiua/2c/475fbfc17414db9c78847b2080cd202c.pdf |
| spelling | journaliasakpiua-article-2859562024-05-23T07:09:36Z Cardiomyopathy prediction in patients with permanent ventricular pacing using machine learning methods Прогнозування кардіоміопатії у пацієнтів з постійною шлуночковою електрокардіостимуляцією за допомогою методів машинного навчання Perepeka, Eugene Lazoryshynets, Vasyl Babenko, Vitalii Davydovych, Illia Nastenko, Ievgen permanent ventricular pacing risk factors artificial intelligence forecasting machine learning постійне ритмоведення шлуночків фактори ризику штучний інтелект прогнозування машинне навчання Pacing-induced cardiomyopathy is a notable issue in patients needing permanent ventricular pacing. Identifying risk groups early and swiftly preventing the ailment can reduce patient harm. However, current prognostic methods require clarity. We employed machine learning to develop predictive models using medical data. Three algorithms — decision tree, group method of data handling, and logistic regression — formed models that forecast pacing-induced cardiomyopathy. These models displayed high accuracy in predicting development, signifying soundness. Factors like age, paced QRS width, pacing mode, and ventricular index during implantation significantly influenced predictions. Machine learning can enhance pacing-induced cardiomyopathy prediction in ventricular pacing patients, aiding medical practice and preventive strategies. Кардіоміопатія, спричинена кардіостимуляцією, є важливою проблемою для пацієнтів, які потребують постійної шлуночкової кардіостимуляції. Раннє виявлення груп ризику та швидка профілактика недуги можуть зменшити шкоду для пацієнтів. Однак сучасні методи прогнозування потребують доопрацювання. Застосовано машинне навчання для розроблення прогностичних моделей на основі медичних даних. Три алгоритми — дерево рішень, група оброблення даних та логістична регресія — сформували моделі, які прогнозують кардіоміопатію, спричинену кардіостимуляцією. Ці моделі показали високу точність у прогнозуванні розвитку, що свідчить про їх надійність. Ключові фактори, такі як вік, ширина QRS, режим кардіостимуляції та шлуночковий індекс під час імплантації, суттєво впливали на прогнози. Машинне навчання може покращити прогнозування кардіоміопатії, спричиненої кардіостимуляцією, у пацієнтів, які перебувають на шлуночковій електрокардіостимуляції, допомагаючи медичній практиці та профілактичним стратегіям. The National Technical University of Ukraine "Igor Sikorsky Kyiv Polytechnic Institute" 2024-03-29 Article Article application/pdf https://journal.iasa.kpi.ua/article/view/285956 10.20535/SRIT.2308-8893.2024.1.03 System research and information technologies; No. 1 (2024); 33-41 Системные исследования и информационные технологии; № 1 (2024); 33-41 Системні дослідження та інформаційні технології; № 1 (2024); 33-41 2308-8893 1681-6048 en https://journal.iasa.kpi.ua/article/view/285956/296304 |
| spellingShingle | постійне ритмоведення шлуночків фактори ризику штучний інтелект прогнозування машинне навчання Perepeka, Eugene Lazoryshynets, Vasyl Babenko, Vitalii Davydovych, Illia Nastenko, Ievgen Прогнозування кардіоміопатії у пацієнтів з постійною шлуночковою електрокардіостимуляцією за допомогою методів машинного навчання |
| title | Прогнозування кардіоміопатії у пацієнтів з постійною шлуночковою електрокардіостимуляцією за допомогою методів машинного навчання |
| title_alt | Cardiomyopathy prediction in patients with permanent ventricular pacing using machine learning methods |
| title_full | Прогнозування кардіоміопатії у пацієнтів з постійною шлуночковою електрокардіостимуляцією за допомогою методів машинного навчання |
| title_fullStr | Прогнозування кардіоміопатії у пацієнтів з постійною шлуночковою електрокардіостимуляцією за допомогою методів машинного навчання |
| title_full_unstemmed | Прогнозування кардіоміопатії у пацієнтів з постійною шлуночковою електрокардіостимуляцією за допомогою методів машинного навчання |
| title_short | Прогнозування кардіоміопатії у пацієнтів з постійною шлуночковою електрокардіостимуляцією за допомогою методів машинного навчання |
| title_sort | прогнозування кардіоміопатії у пацієнтів з постійною шлуночковою електрокардіостимуляцією за допомогою методів машинного навчання |
| topic | постійне ритмоведення шлуночків фактори ризику штучний інтелект прогнозування машинне навчання |
| topic_facet | permanent ventricular pacing risk factors artificial intelligence forecasting machine learning постійне ритмоведення шлуночків фактори ризику штучний інтелект прогнозування машинне навчання |
| url | https://journal.iasa.kpi.ua/article/view/285956 |
| work_keys_str_mv | AT perepekaeugene cardiomyopathypredictioninpatientswithpermanentventricularpacingusingmachinelearningmethods AT lazoryshynetsvasyl cardiomyopathypredictioninpatientswithpermanentventricularpacingusingmachinelearningmethods AT babenkovitalii cardiomyopathypredictioninpatientswithpermanentventricularpacingusingmachinelearningmethods AT davydovychillia cardiomyopathypredictioninpatientswithpermanentventricularpacingusingmachinelearningmethods AT nastenkoievgen cardiomyopathypredictioninpatientswithpermanentventricularpacingusingmachinelearningmethods AT perepekaeugene prognozuvannâkardíomíopatííupacíêntívzpostíjnoûšlunočkovoûelektrokardíostimulâcíêûzadopomogoûmetodívmašinnogonavčannâ AT lazoryshynetsvasyl prognozuvannâkardíomíopatííupacíêntívzpostíjnoûšlunočkovoûelektrokardíostimulâcíêûzadopomogoûmetodívmašinnogonavčannâ AT babenkovitalii prognozuvannâkardíomíopatííupacíêntívzpostíjnoûšlunočkovoûelektrokardíostimulâcíêûzadopomogoûmetodívmašinnogonavčannâ AT davydovychillia prognozuvannâkardíomíopatííupacíêntívzpostíjnoûšlunočkovoûelektrokardíostimulâcíêûzadopomogoûmetodívmašinnogonavčannâ AT nastenkoievgen prognozuvannâkardíomíopatííupacíêntívzpostíjnoûšlunočkovoûelektrokardíostimulâcíêûzadopomogoûmetodívmašinnogonavčannâ |