A Hybrid Deep Learning Framework for Early Cardiovascular Disease Prediction: SDCN-CHDMO-ResVGG Approach

Abstract

Being a major cause of morbidity and mortality, cardiovascular diseases (CVDs) require precise predictive models for early detection. This work suggests a deep learning-based method to improve diagnostic performance by combining feature selection, classification, and optimization. Chaotic Dwarf Mongoose Optimization (CH-DMO) selects optimal features, while Self-Attention Assisted Dense Capsule Network (SDCN) ensures robust feature extraction. Harris Whale Residual VGG Network (ResVGG) improves classification accuracy using residual learning, with M-Square normalization enhancing feature homogeneity. Experimental results on benchmark CVD datasets show superior classification accuracy, lower computational complexity, and improved generalization over existing methods. The findings highlight the model’s potential to assist healthcare professionals in early CVD detection, improving clinical decision-making and patient outcomes.