Artificial Intelligence Techniques for Heart Disease Prediction Using Optical Electrocardiograms (ECG) and a Hybrid Convolutional Block Attention Capsule Network: Trends and Challenges

Abstract

Cardiovascular diseases (CVDs) remain the leading cause of mortality worldwide, necessitating early and accurate diagnostic techniques. Electrocardiogram (ECG) signals, particularly optical ECG representations derived from wearable and imaging technologies, provide rich physiological information for detecting cardiac abnormalities. Recent advancements in artificial intelligence (AI), especially deep learning, have significantly improved heart disease prediction by enabling automatic feature extraction and classification. This review explores state-of-the-art AI techniques focusing on hybrid architectures that integrate Convolutional Neural Networks (CNN), attention mechanisms, and Capsule Networks for enhanced ECG-based diagnosis. CNN models effectively capture spatial patterns, while attention mechanisms highlight critical features, and Capsule Networks preserve hierarchical relationships between ECG signal components. Studies demonstrate that hybrid models achieve superior accuracy (often exceeding 97%) compared to traditional machine learning approaches. This paper presents a comprehensive analysis of recent literature, comparative evaluation of models, and discussion of challenges such as data imbalance, interpretability, and computational complexity. Furthermore, trends including multimodal learning, wearable integration, and explainable AI are highlighted. The study concludes that hybrid attention-based capsule architectures represent a promising direction for reliable, real-time heart disease prediction systems.