Advanced Automated Detection of Heart Diseases Using TQWT and Dual-Task Deep Learning Model

This study introduces a cutting-edge automated system for detecting and classifying 14 types of heart diseases, including arrhythmias and myocardial infarction, using ECG signals. The methodology leverages the Tunable-Q Wavelet Transform (TQWT) for precise feature extraction and Morphological Component Analysis (MCA) with SALSA for effective signal decomposition. A novel dual-task deep learning model, enhanced by a dynamic non-stationary redundancy factor, r(t,f)r(t, f)r(t,f), achieves exceptional accuracy (99.7%) while reducing computational cost by 25% and accelerating convergence by 30%. This approach offers a transformative solution for real-time heart disease diagnosis, advancing healthcare efficiency and patient outcomes.