vitjВрач и информационные технологииMedical Doctor and Information Technologies1811-01932413-5208Pirogov National Medical and Surgical Center10.25881/18110193_2025_4_44vitj-273Research ArticleОРИГИНАЛЬНЫЕ ИССЛЕДОВАНИЯORIGINAL RESEARCHСверточная нейронная сеть для обнаружения нарушений сердечного ритмаConvolutional neural network for detecting cardiac arrhythmiasНовиковаО. А.NovikovaO. A.

к.т.н.

г. Москва

PhD

Moscow

ol-novikova@bk.ruЕрмоловА. Е.ErmolovA. E.

г. Москва

Moscow

a.e.ermolov@mail.ruПервый Московский государственный медицинский университет имени И.М. СеченоваРоссияI.M. Sechenov First Moscow State Medical UniversityRussian FederationМИРЭА – Российский технологический университетРоссияMIREA – Russian Technological UniversityRussian Federation202523122025044455Copyright © Новикова О.А., Ермолов А.Е., 20252025Новикова О.А., Ермолов А.Е.Novikova O.A., Ermolov A.E.This work is licensed under a Creative Commons Attribution 4.0 License.https://www.vit-j.ru/jour/article/view/273

Целью исследования является разработка и оценка архитектуры нейронной сети для автоматизированного анализа электрокардиографических (ЭКГ) сигналов, позволяющей выявлять нарушения сердечного ритма.

Материалы и методы

Материалы и методы. Для обучения использованы открытые базы данных ЭКГ (34570 записей). Работа проводилась на основе данных шести стандартных отведений (I, II, III, aVR, aVL, aVF). Предобработка сигналов включала удаление полиномиального тренда, вейвлет-фильтрацию, медианную фильтрацию, сглаживание и нормализацию. Для классификации использовалась сверточная нейронная сеть, которая обрабатывает сигналы каждого из шести отведений по отдельности. Итоговое решение формируется методом взвешенного голосования с эмпирически определенными коэффициентами, отражающими вклад каждого канала на основе предварительного анализа метрик. В качестве референсного теста использовалась клиническая аннотация, выполненная врачами-кардиологами.

Результаты

Результаты.  Предложенная  модель  демонстрирует высокие показатели качества классификации: accuracy – 0,97 (95% ДИ 0,96-0,98), precision – 0,98 (95% ДИ 0,97-0,99), recall – 0,98 (95% ДИ 0,97-0,99), specificity – 0,94 (95% ДИ 0,92-0,96), F1-score – 0,98 (95% ДИ 0,97-0,99), ROC-AUC – 0,99 (95% ДИ 0,98-1,00), PR-AUC – 0,96 (95% ДИ 0,94-0,97).

Заключение

Заключение. Полученные результаты подтверждают эффективность предложенного метода в задачах диагностики сердечно-сосудистых заболеваний. Предложенная методика может быть адаптирована для диагностики более широкого спектра сердечных заболеваний, что делает ее актуальной для внедрения в практическую кардиологию.

The aim of the study is to develop and evaluate the architecture of a neural network for automated analysis of electrocardiographic (ECG) signals to detect cardiac arrhythmias.Materials and methods. Open ECG databases (34,570 records) were used for training. The study was conducted using data from six standard ECG leads (I, II, III, aVR, aVL, aVF). Signal preprocessing included polynomial trend removal, wavelet filtering, median filtering, smoothing, and normalization. A convolutional neural network, which processes signals from each of the six leads separately, was used for classification. The final decision is formed using a weighted voting method with empirically determined coefficients reflecting the contribution of each channel based on a preliminary analysis of metrics. A clinical annotation performed by cardiologists was used as a reference standard.Results. The proposed model demonstrates high classification performance: accuracy – 0.97 (95% CI 0.96-0.98), precision – 0.98 (95% CI 0.97-0.99), recall – 0.98 (95% CI 0.97-0.99), specificity – 0.94 (95% CI 0.92-0.96), F1-score – 0.98 (95% CI 0.97-0.99), ROC-AUC – 0.99 (95% CI 0.98-1.00), PR-AUC – 0.96 (95% CI 0.94-0.97).Conclusion. The obtained results confirm the effectiveness of the proposed method for diagnosing cardiovascular diseases. The proposed method can be adapted for diagnosing a wider range of cardiac diseases, making it relevant for implementation in practical cardiology.

ЭКГклассификациясверточная нейронная сетьобработка сигналаECGclassificationconvolutional neural networksignal processingReferencesZhang C, Li J, Guo P, et al. A configurable hardware-efficient ECG classification inference engine based on CNN for mobile healthcare applications. Microelectronics Journal. 2023; 141: 105969. doi: 10.1016/j.mejo.2023.105969.Zhang C, Li J, Guo P, et al. A configurable hardware-efficient ECG classification inference engine based on CNN for mobile healthcare applications. Microelectronics Journal. 2023; 141: 105969. doi: 10.1016/j.mejo.2023.105969.Шутов Д.В., Дроздов Д.В., Газашвили Т.М. и др. Методические и нормативные аспекты создания набора данных ЭКГ для тестирования автоматизированных алгоритмов и систем искусственного интеллекта // Врач и информационные технологии. – 2021. – №4. – С. 4-15. doi: 1025881/18110193_2021_4_4.Shutov DV, Drozdov DV, Gazashvili TM, et al. Methodology and regulatory implications of ECG database formation for testing automated algorithms and AI systems. Medical doctor and information technology. 2021; 4: 4-15. (In Russ.) doi: 1025881/18110193_2021_4_4.Zhang Y, Liu S, He Z, et al. A CNN Model for Cardiac Arrhythmias Classification Based on Individual ECG Signals. Cardiovasc Eng Technol. 2022; 13(4): 548-447. doi: 10.1007/s13239-021-00599-8.Zhang Y, Liu S, He Z, et al. A CNN Model for Cardiac Arrhythmias Classification Based on Individual ECG Signals. Cardiovasc Eng Technol. 2022; 13(4): 548-447. doi: 10.1007/s13239-021-00599-8.Cheng J, Zou Q, Zhao Y. ECG signal classification based on deep CNN and BiLSTM. BMC Med Inform Decis Mak. 2021; 21: 365. doi: 10.1186/s12911-021-01736-y.Cheng J, Zou Q, Zhao Y. ECG signal classification based on deep CNN and BiLSTM. BMC Med Inform Decis Mak. 2021; 21: 365. doi: 10.1186/s12911-021-01736-y.Maweu BM, Dakshit S, Shamsuddin R, et al. CEFEs: A CNN Explainable Framework for ECG Signals. Artif Intell Med. 2021; 115: 102059. doi: 10.1016/j.artmed.2021.102059.Maweu BM, Dakshit S, Shamsuddin R, et al. CEFEs: A CNN Explainable Framework for ECG Signals. Artif Intell Med. 2021; 115: 102059. doi: 10.1016/j.artmed.2021.102059.Андриков Д.А., Курбанов С.В. Алгоритмическое обеспечение спектральной обработки кардиограмм // Вестник Российского университета дружбы народов. Серия: Инженерные исследования. – 2024. – Т.25. – №2. – C. 111-120. doi: 10.22363/2312-8143-2024-25-2-111-120.Andrikov DA, Kurbanov SV. Algorithmic support for spectral processing of cardiograms. RUDN Journal of Engineering Research. 2024; 25(2): 111-120. (In Russ.) doi: 10.22363/2312-8143-2024-25-2-111-120.Павлов А.Н., Руннова А.Е., Храмов А.Е. Частотно-временной анализ нестационарных процессов: концепции вейвлетов и эмпирических мод // Известия вузов. ПНД. – 2011. – Т.19. – №2. – С.141-157. doi: 10.18500/0869-6632-2011-19-2-141-157.Pavlov AN, Filatova AE, Hramov AE. Time-frequency analysis of nonstationary processes: concepts of wavelets and empirical modes. Izvestiya vuzov. PND. 2011; 19(2): 141-157. (In Russ.) doi: 10.18500/0869-6632-2011-19-2-141-157.Kaggle. Shaoxing and Ningbo Hospital ECG Database. Available at: https://www.kaggle.com/datasets/bjoernjostein/shaoxing-and-ningbo-first-hospital-database. Accessed January 2, 2025.Kaggle. Shaoxing and Ningbo Hospital ECG Database. Available at: https://www.kaggle.com/datasets/bjoernjostein/shaoxing-and-ningbo-first-hospital-database. Accessed January 2, 2025.Kaggle. PTB-XL ECG dataset. Available at: https://www.kaggle.com/datasets/khyeh0719/ptb-xl-dataset.Accessed January 2, 2025.Kaggle. PTB-XL ECG dataset. Available at: https://www.kaggle.com/datasets/khyeh0719/ptb-xl-dataset.Accessed January 2, 2025.Kaggle. Georgia 12-Lead ECG Challenge Database. Available at: https://www.kaggle.com/datasets/bjoernjostein/georgia-12lead-ecg-challenge-database. Accessed January 2, 2025.Kaggle. Georgia 12-Lead ECG Challenge Database. Available at: https://www.kaggle.com/datasets/bjoernjostein/georgia-12lead-ecg-challenge-database. Accessed January 2, 2025.Kaggle. China Physiological Signal Challenge in 2018. Available at: https://www.kaggle.com/datasets/bjoernjostein/china-physiological-signal-challenge-in-2018. Accessed January 2, 2025.Kaggle. China Physiological Signal Challenge in 2018. Available at: https://www.kaggle.com/datasets/bjoernjostein/china-physiological-signal-challenge-in-2018. Accessed January 2, 2025.Kaggle. China 12-Lead ECG Challenge Database. Available at: https://www.kaggle.com/datasets/bjoernjostein/china-12lead-ecg-challenge-database. Accessed January 2, 2025.Kaggle. China 12-Lead ECG Challenge Database. Available at: https://www.kaggle.com/datasets/bjoernjostein/china-12lead-ecg-challenge-database. Accessed January 2, 2025.Safdar M, Nowak R, Palka P. Pre-Processing techniques and artificial intelligence algorithms for electrocardiogram (ECG) signals analysis: A comprehensive review. Computers in Biology and Medicine. 2024; 170: 107908. doi: 10.1016/j.compbiomed.2023.107908.Safdar M, Nowak R, Palka P. Pre-Processing techniques and artificial intelligence algorithms for electrocardiogram (ECG) signals analysis: A comprehensive review. Computers in Biology and Medicine. 2024; 170: 107908. doi: 10.1016/j.compbiomed.2023.107908.Lin M., Hong Y., Hong S., et al. Discrete Wavelet Transform based ECG classification using gcForest: A deep ensemble method. Technol Health Care. 2024; 32(S1): 95-105. doi: 10.3233/thc-248008.Lin M., Hong Y., Hong S., et al. Discrete Wavelet Transform based ECG classification using gcForest: A deep ensemble method. Technol Health Care. 2024; 32(S1): 95-105. doi: 10.3233/thc-248008.Pal A, Rai H, Agarwal S, et al. Advanced Noise-Resistant Electrocardiography Classification Using Hybrid Wavelet-Median Denoising and a Convolutional Neural Network. Sensors. 2024; 24(21): 7033. doi: 10.3390/s24217033.Pal A, Rai H, Agarwal S, et al. Advanced Noise-Resistant Electrocardiography Classification Using Hybrid Wavelet-Median Denoising and a Convolutional Neural Network. Sensors. 2024; 24(21): 7033. doi: 10.3390/s24217033.Mokhtari R, Belhaouari S, Kassoul K, et al. ECG Heartbeat classification using Progressive Moving Average Transform. Research Square. 2024; 10: 21203. doi: 10.1038/s41598-025-88119-9.Mokhtari R, Belhaouari S, Kassoul K, et al. ECG Heartbeat classification using Progressive Moving Average Transform. Research Square. 2024; 10: 21203. doi: 10.1038/s41598-025-88119-9.Yousuf A, Hafiz R, Riaz S, et al. Inferior Myocardial Infarction Detection from lead II of ECG: A Gramian Angular Field-based 2D-CNN Approach. Electrical Engineering and Systems Science. 2024; 2302: 13011. doi: 10.1109/lsens.2024.3450176.Yousuf A, Hafiz R, Riaz S, et al. Inferior Myocardial Infarction Detection from lead II of ECG: A Gramian Angular Field-based 2D-CNN Approach. Electrical Engineering and Systems Science. 2024; 2302: 13011. doi: 10.1109/lsens.2024.3450176.Narotamo H, Dias M, Santos R, et al. Deep learning for ECG classification: A comparative study of 1D and 2D representations and multimodal fusion approaches. Biomedical Signal Processing and Control. 2024; 93: 106141. doi: 10.2139/ssrn.4555989.Narotamo H, Dias M, Santos R, et al. Deep learning for ECG classification: A comparative study of 1D and 2D representations and multimodal fusion approaches. Biomedical Signal Processing and Control. 2024; 93: 106141. doi: 10.2139/ssrn.4555989.Guo K, Wang S, Ma L, et al. ECG signals’ detection using a Dropout Deep Convolutional Neural Network. Reasearch Square. 2022; 10: 21203. doi: 10.21203/rs.3.rs-1591115/v1.Guo K, Wang S, Ma L, et al. ECG signals’ detection using a Dropout Deep Convolutional Neural Network. Reasearch Square. 2022; 10: 21203. doi: 10.21203/rs.3.rs-1591115/v1.Liu C, Wan Z, Ouyang C, et al. Zero-Shot ECG Classification with Multimodal Learning and Test-time Clinical Knowledge Enhancement. Electrical Engineering and Systems Science. 2024; 2403: 06659. doi: 10.48550/arXiv.2403.06659.Liu C, Wan Z, Ouyang C, et al. Zero-Shot ECG Classification with Multimodal Learning and Test-time Clinical Knowledge Enhancement. Electrical Engineering and Systems Science. 2024; 2403: 06659. doi: 10.48550/arXiv.2403.06659.Ebrahimi Z, Loni M, Daneshtalab M, et al. A review on deep learning methods for ECG arrhythmia classification. Expert Systems with Applications: X. 2020; 7: 100033. doi: 10.1016/j.eswax.2020.100033.Ebrahimi Z, Loni M, Daneshtalab M, et al. A review on deep learning methods for ECG arrhythmia classification. Expert Systems with Applications: X. 2020; 7: 100033. doi: 10.1016/j.eswax.2020.100033.Xie C, Wang Z, Yang C, et al. Machine Learning for Detecting Atrial Fibrillation from ECGs: Systematic Review and Meta-Analysis. Rev Cardiovasc Med. 2024; 25(1): 8. doi: 10.31083/j.rcm2501008.Xie C, Wang Z, Yang C, et al. Machine Learning for Detecting Atrial Fibrillation from ECGs: Systematic Review and Meta-Analysis. Rev Cardiovasc Med. 2024; 25(1): 8. doi: 10.31083/j.rcm2501008.Fiorina L, Chemaly P, Cellier J, et al. Artificial intelligence–based electrocardiogram analysis improves atrial arrhythmia detection from a smartwatch electrocardiogram. Eur Heart J Digit Health. 2024; 5(5): 535-541. doi: 10.1093/ehjdh/ztae047.Fiorina L, Chemaly P, Cellier J, et al. Artificial intelligence–based electrocardiogram analysis improves atrial arrhythmia detection from a smartwatch electrocardiogram. Eur Heart J Digit Health. 2024; 5(5): 535-541. doi: 10.1093/ehjdh/ztae047.Hygrell T, Viberg F, Dahlberg E, et al. An artificial intelligence–based model for prediction of atrial fibrillation from single-lead sinus rhythm electrocardiograms facilitating screening. EP Europace. 25(4): 1332-1338. doi: 10.1093/europace/euad036.Hygrell T, Viberg F, Dahlberg E, et al. An artificial intelligence–based model for prediction of atrial fibrillation from single-lead sinus rhythm electrocardiograms facilitating screening. EP Europace. 25(4): 1332-1338. doi: 10.1093/europace/euad036.Barker J, Li X, Kotb A, et al. Artificial intelligence for ventricular arrhythmia capability using ambulatory electrocardiograms. European Heart Journal – Digital Health. 5(3): 384-388. doi: 10.1093/ehjdh/ztae004.Barker J, Li X, Kotb A, et al. Artificial intelligence for ventricular arrhythmia capability using ambulatory electrocardiograms. European Heart Journal – Digital Health. 5(3): 384-388. doi: 10.1093/ehjdh/ztae004.Gruwez H, Barthels M, Haemers P, et al. Detecting Paroxysmal Atrial Fibrillation From an Electrocardiogram in Sinus Rhythm: External Validation of the AI Approach. JACC: Clinical Electrophysiology. 9(8): 1771-1782. doi: 10.1016/j.jacep.2023.04.008.Gruwez H, Barthels M, Haemers P, et al. Detecting Paroxysmal Atrial Fibrillation From an Electrocardiogram in Sinus Rhythm: External Validation of the AI Approach. JACC: Clinical Electrophysiology. 9(8): 1771-1782. doi: 10.1016/j.jacep.2023.04.008.Gadaleta M, Harrington, Barnhill E, et al. Prediction of atrial fibrillation from at-home single-lead ECG signals without arrhythmias. npj Digit. Med. 6: 229. doi: 10.1038/s41746-023-00966-w.Gadaleta M, Harrington, Barnhill E, et al. Prediction of atrial fibrillation from at-home single-lead ECG signals without arrhythmias. npj Digit. Med. 6: 229. doi: 10.1038/s41746-023-00966-w.

The authors declare that there are no conflicts of interest present.