Fuzzy based system for coronary artery disease prediction using subtractive clustering and risk factors data

Abstract

Over the past three decades, coronary artery disease (CAD) has been considered one of the most common fatal diseases worldwide. Consequently, early diagnosis and prediction are essential, as they can significantly reduce patient mortality and treatment costs. This study aims to design an automatic expert system using fuzzy logic theory to predict CAD. Thus, aiding physicians to identify diseases at an early stage and assess their severity. This system generates fuzzy rules automatically from training dataset through a subtractive clustering method and employs the Sugeno Fuzzy Inference Engine to produce an output indicating the patient's condition. Feature selection is performed using filter methods such as variance analysis, Mutual Information, and Pearson's Correlation Coefficient to identify the most relevant factors affecting heart disease. The implementation is conducted on publicly available UCI heart disease datasets, and the system's performance is evaluated based on accuracy, specificity, and sensitivity metrics. The findings indicate a classification accuracy of 99.61 %, achieving a sensitivity rate of 100 % and a specificity rate of 99.20 %. These findings highlight the system's potential as an effective diagnostic and early prevention tool, ultimately improving clinical outcomes in CAD treatment. • Based on risk factors a supervised fuzzy CAD system is introduced. • Clinical data is segmented using the subtractive clustering method. • The fuzzy rules and membership functions are generated automatically from the dataset. • Features selection is performed using statistical algorithms. • The proposed system with a minimal number of rules is developed and evaluated.