EXPANDING THE POSSIBILITIES OF ELECTROCARDIOGRAM SIGNALS PROCESSING AND ANALYSIS

Abstract

Modern functional diagnostics of cardiovascular diseases have a large number of different instrumental research methods at their disposal, but one of the most common is electrocardiography (ECG). Since its discovery, this method of studying the bioelectrical activity of the heart has been a leading method in the diagnosis of rhythm and conduction disorders, coronary heart disease, and other diseases of the cardiovascular system [2]. For traditional methods of ECG analysis, such as visual inspection, the doctor identifies characteristic visible features on the ECG. An unusually long PR interval, for example, indicates a conduction defect in the atria, or a prolonged QT interval can lead to an abnormal heart rhythm. Unfortunately, for many medical problems, such significant signs cannot be identified so easily. In addition, standard characteristics such as PR interval, QRS width, or ST level are ambiguous in many cases. As a result, sophisticated feature extraction methods are used. Such methods try to find new features that allow the diagnosis of CVD based on the ECG. These features can be obtained from the ECG in the time domain, ECG in the time-frequency domain, or ECG in the frequency domain. These features can then be processed using various approaches such as visual inspection or machine learning algorithms [3].

With the growth of computing performance and the development of digital signal processing, it is important to solve the problems of classification/clustering of ECG signals, as well as forecasting problems using this approach. Achieving high results in solving these problems is impossible without providing an in-depth analysis of such signals, which can be achieved by using wavelet processing with the subsequent use of neural networks. Unlike the Fourier transform, the wavelet transform provides a two-dimensional representation of the signal, with the scale and time offset considered independently, making it possible to analyze signals in two independent spaces simultaneously - scale and time. The results of ECG wavelet analysis contain not only information about the distribution of cardiac signal energy by frequency components but also information about the time coordinates at which certain frequency components are detected or at which rapid changes in the frequency components of the cardiac signal occur.

This paper reviews existing studies using the wavelet transform (WT) to detect electrocardiogram (ECG) complexes, analyze ECG signals to identify features of the studied signal complexes and analyze decomposed signals to obtain in-depth diagnostic data. The reviewed studies show the high efficiency of using this method for signal analysis and open the way for further research in the direction of identifying additional information from the ECG signal, which would serve as the basis for early diagnosis of CVD.