Classifying Atrial Fibrillation Using Deep Residual Networks (ResNet) on ECG Signals

Atrial Fibrillation (AF) is a complex most prevalent arrhythmia with many complications that affect the affected individual’s overall health as his/her propensity to stroke and heart failure. This study seeks to investigate the effectiveness of Deep Residual Networks (ResNet) in the categorisation of AF from ECG signals given that this will help improve the AF diagnostic accuracy and speed. The ResNet model is based on deep architecture and uses residual connections to learn various patterns from ECG data. The process includes data preprocessing: filtering, normalization and seg-mentation; model training and testing. The results show a high accuracy of the method, the sensitivity, and specificity, which is higher than in many traditional and some modern techniques. The use of ResNet-based approach suggested for automated AF detections seems to be promising with its clinical application and the consequent effects on patients technology. Improvements for future work may be made to fine-tune the model and the integration of the proposed solution with other diagnostic aids for easier diagnosis of arrhythmias.