A Hybrid Deep Learning and Nature-Inspired Optimization Framework for Enhanced Cardiac Disease Detection Using Electrocardiogram Signals

Cardiovascular diseases (CVDs) remain the leading cause of mortality worldwide, necessitating early and accurate detection. Electrocardiogram (ECG) analysis is a primary diagnostic tool, but manual interpretation is subjective and time-consuming. Artificial Intelligence (AI) offers promise, yet its performance hinges on effective feature selection and model optimization. This paper proposes a hybrid framework that integrates deep learning-based feature extraction with a nature-inspired optimization algorithm for accurate and interpretable cardiac disease detection from ECG signals. The optimized feature set is fed into a multi-classifier system including Support Vector Machines (SVM), Random Forest (RF), and a Transformer-based neural network. Experimental validation on the MIT-BIH, PTB-XL, and Cleveland datasets demonstrates that the hybrid model significantly improves classification accuracy, F1-score, and computational efficiency compared to single-optimizer frameworks. Crucially, the system integrates interpretability mechanisms—SHAP values and attention maps—to highlight influential ECG features and signal regions, aiding clinical understanding. The results underscore the potential of combining deep learning with hybrid nature-inspired optimization for robust, efficient, and trustworthy automated cardiac diagnosis.