Recent Advances in Heart Disease Prediction Using Optical Electrocardiograms (ECG) and a Hybrid Convolutional Block Attention Capsule Network: A Systematic Review

Abstract

Heart disease remains one of the leading causes of mortality worldwide, necessitating accurate and early diagnostic techniques. Optical electrocardiograms (ECG), often derived from photoplethysmography (PPG) and wearable sensors, have emerged as a promising non-invasive solution for continuous cardiac monitoring. This paper presents a systematic review of recent advances in heart disease prediction using optical ECG signals combined with hybrid deep learning architectures, particularly Convolutional Neural Networks (CNN), Block Attention Modules, and Capsule Networks. The integration of attention mechanisms enhances feature selection by focusing on relevant signal components, while capsule networks preserve spatial hierarchies and improve classification robustness. The review focuses on studies published in recent years, highlighting advancements in signal preprocessing, feature extraction, and hybrid model design. Comparative analysis indicates that hybrid attention-based capsule networks outperform traditional CNN and machine learning models in terms of accuracy, sensitivity, and generalization. The paper also discusses challenges such as noise sensitivity, data imbalance, interpretability, and computational complexity. Furthermore, the role of wearable devices and real-time monitoring systems is examined. The study concludes that hybrid AI-based frameworks leveraging optical ECG signals hold significant potential for improving heart disease prediction and enabling next-generation smart healthcare systems.