Heart disease (HD) is a prevalent and deadly disease that mainly affects middle-aged and elderly individuals and has a higher risk in men as compared to women. In Bangladesh, non-communicable disorders (NCDs) represent 67% of all mortality, especially cardiovascular disease, which alone is responsible for around 30% of deaths. Globally, HDs claim the lives of one-third of the population.
HD is diagnosed through a clinical assessment, investigation of risk factors, and medical imaging. Yet they cannot reliably forecast whether HD will occur, especially during the pre-symptomatic stages in which disease signs are absent. As healthcare data become increasingly accessible and abundant, many researchers have started using machine learning (ML) methods in this problem domain to achieve better diagnostic accuracy.
This study is motivated to help bridge the research gap between model interpretability and prediction accuracy for HD detection. Prior work has largely focused on the enhanced predictive accuracy of ML models for HD, often trading off interpretability and feasibility in real-world clinical settings.
【MORE】