<![CDATA[Cardiovascular diseases remain the leading cause of mortality on a global scale, with myocardial infarction (MI) representing a critical and life-threatening condition. Electrocardiography (ECG) is a widely utilized method for the detection of myocardial infarction (MI), and artificial intelligence (AI) has demonstrated a promising performance in the automated ECG-based diagnosis. However, most existing studies emphasizepredictive accuracy while failing to provide substantial evidence that model decision logic aligns with clinical reasoning, thereby limiting clinical adoption. This present study evaluates the interpretability of three rule-based machine learning classifiers—Decision Tree, RIPPER, and Rough Set—for MI detection from ECG signals, including a comparison between models with and without feature selection. Interpretability of the system is assessed through rule complexity analysis and a standardized qualitative clinical validation protocol involving three cardiologists, based on contemporary AHA/ESC ECG diagnostic guidelines. The findings indicate that the Rough Set classifier attains the optimal overall performance, with 80% of its generated rules demonstrating clinically aligned, thereby outperforming the other models regarding interpretability. The findings demonstrate the benefit of guideline-based clinical validation for advancing trustworthy ECG-based MI diagnostic systems.]]>
Copyrights © 2025
