Implementation of a Capsule Network Algorithm for ECG Signal Classification in Arrhythmia Detection

Cardiovascular diseases remain one of the primary contributors to global mortality rates.Insufficient early diagnostic tools and the limited availability of cardiology experts pose challenges to accurate arrhythmia identification, which may elevate the likelihood of severe complications and fatal outcomes.This study applies the Capsule Network (CapsNet) model to categorize ECG signal patterns associated with arrhythmic conditions.CapsNet demonstrates a strong capability in capturing intricate spatial relationships within sequential physiological data, including cardiac waveforms.By maintaining spatial hierarchies and dependencies between features, CapsNet effectively identifies subtle morphological variations in cardiac waveforms.The data in this study were obtained through electrocardiogram (ECG) recordings of subjects grouped into three activity categories: sitting, walking, and running.The acquired ECG datasets were subsequently utilized to train and evaluate the classification model for detecting potential arrhythmic events from observed cardiac activity patterns.The CapsNet algorithm achieved an accuracy rate of up to 91% in the classification process.The findings suggest that the CapsNet framework could serve as an effective approach for the early identification of arrhythmias.