E.G. Evdakova1

1 St. Petersburg State Electrotechnical University "LETI" (Saint Petersburg, Russia)
1 kat355@mail.ru

Timely detection of dangerous arrhythmias is of great importance in clinical practice. The rapid implementation of resuscitation measures is critical for providing necessary assistance to patients in life-threatening situations. In cardiac monitoring, attention should be paid to recognizing precursors of serious disturbances. Effective analysis of short fragments of the electrocardiographic signal is important for solving these tasks.

The study aims to conduct a comparative analysis of various arrhythmia recognition algorithms. Special attention is given to the impact of the sample partitioning method on classification quality. Two-second fragments of the electrocardiographic signal are considered. A transition to the spectral domain is performed to identify dangerous arrhythmias.

The task of binary classification of dangerous and non-dangerous ventricular arrhythmias was addressed. A database was used, developed based on an existing publicly available database from PhysioNet. The original database contained thirty-minute recordings, while the created one included cut two-second fragments containing six classes of arrhythmias ranked by their degree of threat to the patient's life. Data partitioning for training, in which it is not considered whether the fragments from patient records that were included in the training set are also in the test set, does not lead to model overfitting. It also does not degrade its generalization ability. Only spectral descriptions of signal fragments (amplitude spectrum) normalized by total power were used in training. In the studied database, it is possible to select a test set without ensuring that records from one patient can be only in train or test set. It is preferable to use the smoothed signal spectrum for model training, as it has a greater generalization ability. Furthermore, data synthesis methods can increase the volume of the training set and enhance classification efficiency.

Evdakova E.G. The influence of the method for forming the training and test sets on the accuracy assessment of binary classification of ventricular arrhythmias. Biomedicine Radioengineering. 2025. V. 28. № 2. P. 33–37. DOI: https:// doi.org/10.18127/j15604136-202502-05 (In Russian)

1. Kolk M.Z.H., Deb B., Ruipérez-Campillo S., Bhatia N.K., Clopton P., Wilde A.A.M., Narayan S.M., Knops R.E., Tjong F.V.Y. Machine learning of electrophysiological signals for the prediction of ventricular arrhythmias: systematic review and examination of heterogeneity between studies. EBioMedicine. 2023. V. 89. P. 104462. https://doi.org/10.1016/j.ebiom.2023.104462.
2. Fira M., Costin H.-N., Goras L. Prediction and Detection of Ventricular Fibrillation Using Complex Features and AI-Based Classification. Applied Sciences. 2024. V. 14(7). P. 3050. https://doi.org/10.3390/app14073050.
3. Amann A., Tratnig R., Unterkofler K. Reliability of Old and New Ventricular Fibrillation Detection Algorithms for Automated External Defibrillators. Biomedical Engineering Online. 2005. 4. № 60. P. 1–23. http://dx.doi.org/10.1186/1475-925X-4-60.
4. MIT-BIH Malignant Ventricular Ectopy Database. Massachusetts Institute of Technology https://www.physionet.org/content/vfdb/1.0.0/
5. Nemirko A., Manilo L., Tatarinova A., Alekseev B., Evdakova E. ECG Fragment Database for the Exploration of Dangerous Arrhythmia (version 1.0.0). PhysioNet. https://doi.org/10.13026/kpfg-xs25.
6. Priori S.G., Blomstrom-Lundqvist C., Mazzanti A. et al. 2015 ESC Guidelines for the management of patients with ventricular arrhythmias and the prevention of sudden cardiac death. European Heart Journal. 2015. V. 36(41). P. 2793–2867.
7. Glikson M., Nielsen J.C., Kronborg M.B. et al. 2021 ESC Guidelines on cardiac pacing and cardiac resynchronization therapy. European Heart Journal. 2021. P. 1–94.
8. Bayers de Luna A.B., Coumel P., Leclercq J.F. Ambulatory sudden cardiac death: mechanism of production of fatal arrhythmia on the basis of data from 157 cases. American Heart Journal. 1989. V. 117(1). P. 151–159.
9. Bigger J.T.Jr. Identification of patients at high risk for sudden cardiac death. American Journal of Cardiology. 1984. V. 54(9). P. 3D–8D.
10. Manilo L.A., Nemirko A.P., Evdakova E.G. Ventricular Arrhythmia Classification Using Classical and Neural Network Approaches, accepted for publication in the international journal of the Russian Academy of Sciences “Pattern Recognition and Image Analysis. Advances in Mathematical Theory and Applications”. 2024. V. 34. № 4 (IMTA-2024).
11. Manilo L.A., Nemirko A.P., Evdakova E.G., Tatarinova A. Comparing Binary Classification Results for Dangerous Ventricular Arrhythmias Using 2 and 4 Seconds Short ECG Signal Fragments. In: 2024 International Conference on Information Processes and Systems Development and Quality Assurance (IPS), Saint Petersburg, Russian Federation. 2024. P. 36–39. DOI: 10.1109/IPS62349.2024.10499488.