Machine Learning-Based Prediction of Coronary Artery Disease Using Clinical and Behavioral Data: A Comparative Study

Area of Science:

• Cardiology and Artificial Intelligence
• Biomedical Data Science
• Predictive Analytics in Healthcare

Background:

• Coronary artery disease (CAD) is a major global health concern requiring accurate early diagnosis.
• Machine learning (ML) offers advanced tools for integrating complex patient data to improve diagnostic capabilities.
• Multidimensional patient data integration is key for enhancing risk stratification and clinical management of CAD.

Purpose of the Study:

• To develop and compare supervised ML algorithms for early CAD diagnosis.
• To evaluate the efficacy of demographic, anthropometric, biochemical, and psychosocial parameters in CAD prediction.
• To identify the most effective ML model for distinguishing CAD patients from controls.

Main Methods:

• Retrospective analysis of 300 adult patients (165 with CAD, 135 controls).
• Utilized a dataset including 21 biochemical markers, body composition, and eating behavior scores.
• Trained and evaluated six ML algorithms (k-NN, SVM, ANN, LR, NB, DT) using 10-fold cross-validation.

Main Results:

• The k-NN model achieved the highest accuracy (98.33%) and AUC (0.99), followed by SVM and ANN.
• CAD patients showed elevated glucose, triglycerides, LDL cholesterol, and abdominal obesity, with lower vitamin B12.
• Psychosocial factors like emotional eating had limited impact on model performance.

Conclusions:

• Supervised ML models, especially k-NN, SVM, and ANN, demonstrate high accuracy in CAD detection.
• Integrating diverse clinical and behavioral data enhances CAD prediction beyond traditional biomarkers.
• This approach shows promise for improving early CAD diagnosis and patient risk stratification.