Spindle Autoencoder-CNN hybrid model for cardiac arrhythmia classification

Cardiac arrhythmias, characterized by irregular heart function, disrupt normal blood circulation and are commonly detected using electrocardiograms (ECGs). ECG is widely preferred due to its cost-effectiveness, ease of application, and high reliability in diagnosing heart rhythm disorders. To enhance diagnostic efficiency, automatic arrhythmia detection systems have been increasingly developed. In this study, we propose a novel deep learning-based classification framework that integrates a Modified Spindle Autoencoder (MSCAE) with a Convolutional Neural Network (CNN). Unlike traditional autoencoders, the Spindle Autoencoder utilizes deeper and symmetric hidden layers to extract complex and meaningful representations from ECG signals. These learned features are then analyzed by the CNN to capture spatial relationships through its convolutional layers. The proposed model was trained and evaluated on the MIT-BIH Arrhythmia Database, which contains both normal and abnormal heartbeat recordings. The ECG signals were preprocessed, segmented into individual beats, and subjected to feature extraction to improve classification performance. The integrated MSCAE-CNN architecture achieved an accuracy of 98.78 % under various experimental conditions. Performance metrics demonstrate that the proposed method outperforms existing approaches in classifying arrhythmias. These results underline the clinical potential of the model in providing rapid and accurate ECG-based arrhythmia detection for medical decision-making.