An ECG Processing and Analysis Technique Based on Neural Network

The article is devoted to the study of processing and analysis of the main indicators of an electrocardiogram (ECG) based on neural networks. An algorithm for decision support for an ECG analysis using a neural network for training vector quantization is proposed. For the study, the following functions were selected such as the duration of the QRS complex, the RR interval, the amplitude of the R wave and the change in the slope of the ST segment and the heart rate. These five functions are the five inputs to the neural network learning vector quantization. At the same time, methods of preliminary processing and analysis of the extraction of ECG functions based on the obtained ECG database of a medical institution in MATLAB are given. A generalized algorithm for the generated LVQ network and the architecture of the LVQ neural network created using MATLAB are proposed. A method for training a neural network to classify an ECG signal based on the training data obtained in the process of extracting signs - 5 inputs (RR interval, R wave amplitude, QRS complex duration, ST segment slope, heart rate) into one of five classes (bradycardia, tachycardia, premature ventricular contraction (PVC), myocardial infarction or in the absence of disease class) is proposed.