Binary Classification of Single-Lead Electrocardiogram Using Machine and Deep Learning Methods

Abstract

This study presents a comparative analysis of machine and deep learning methods for binary classification of single-lead electrocardiogram (ECG) signals. Five cardiac data classification models were evaluated: decision tree and random forest classification, a long short-term memory (LSTM) network with pre-extraction of random forest and decision tree data, and the proposed hybrid model with parallel feature extraction on a stream of electrocardiogram signals. Experiments were conducted on the publicly available PhysioNet Computing in Cardiology Challenge 2017 dataset labeled as "Normal"or "Atrial Fibrillation."Simulation results showed that the hybrid model with parallel feature extraction is inferior in classification accuracy to the model with pre-extraction of morphological features using the random forest method. These results indicate that machine learning methods, in particular random forest, offer a robust and efficient solution for automatic ECG classification, balancing high performance with low computational cost. This makes them suitable for integration into wearable devices and large-scale screening applications.