Hybrid Spiking Neural Network-Attention Model for Multi-Class Heart Disease Detection

Heart disease is one of the leading global health issues, and early detection and diagnosis are crucial. This paper presents a hybrid Spiking Neural Network (SNN) - Attention Model for Classifying Cardiac Disease using Electrocardiogram (ECG) Images. The model successfully captures the intricate ECG patterns and their more generalized correlations by merging the powerful contextual learning of the Attention mechanism with the important temporal processing capability of SNNs. In this study, the model was tested on ECG images. The data sets have four classes: history of Myocardial Infarction (MI), Abnormal heartbeat, Myocardial Infarction (MI) and normal heart function. The Attention mechanism module improves classification by observing complex interrelationships between classes, whereas the hybrid approach uses SNNs to process the ECG waveforms using spike-based efficiency. Our model outperforms the previous models by far and compares favorably with the contemporary attention-based CNN and SNN architectures. It outperformed the best current state-of-the-art models in sensitivity and specificity and achieved a better accuracy of $97.78 \%$