R-PEAKS IDENTIFICATION METHOD IN ELECTROCARDIOGRAM SIGNALS

Abstract

The rapid development of artificial intelligence, particularly deep learning models, has significantly expanded the spheres of its usage. One of these areas is medicine. One of the important areas of using artificial intelligence in the medical field is the processing of electrocardiograms (ECG). The ECG remains an important tool for the accurate diagnosis of cardiac pathologies in clinical practice. However, accurate processing and analysis of ECG signals, especially the identification of R-waves, remains a challenge. Inaccurate detection of R-peaks can significantly affect further analysis and diagnosis of the disease. Therefore, this study presents an improved method for detecting R-peaks in ECG signals.

The authors propose a method that includes the integration of knowledge into the ECG signal, processing of this information using a convolutional neural network, and post-processing of the CNN model results to detect R-waves. A feature of the proposed method is the integration of knowledge about the reference cardiac cycle into the input ECG signal. The idea of using such integration of knowledge in the ECG signal is that the neural network will better detect R peaks in the ECG signal with integrated knowledge and will be more resistant to changes in signal artifacts.

The accuracy of the proposed method was verified by using ECG signals from 4 databases, namely: MIT-BIH, QT, CPSC-2020, and UoG. The method was evaluated using an error of ±25 ms using statistical metrics accuracy, precision, recall, and F1-score. The obtained statistical indicators are compared with those of existing methods. The proposed method showed the best result on 3 out of 4 test data sets compared to its analogues. Thus, the proposed method is a reliable and efficient solution for detecting R-peaks in ECG signals.