Machine Learning Coronary Artery Disease Prediction Based on Imaging and Non-Imaging Data.

Area of Science:

• Cardiology
• Medical Imaging
• Machine Learning

Background:

• Obstructive atherosclerotic disease prediction is crucial for clinical decision-making.
• Coronary Artery Disease (CAD) poses significant health risks.
• Accurate risk stratification is essential for patient management.

Purpose of the Study:

• To develop a machine learning model for predicting coronary artery disease (CAD) risk.
• To identify patients at high and low risk of CAD.
• To integrate imaging and non-imaging data for enhanced predictive accuracy.

Main Methods:

• A gradient boosting classifier was employed for prediction.
• Methodology included data preprocessing, Easy Ensemble for class imbalance, recursive feature elimination, and model evaluation.
• Hyper-parameter tuning was performed using randomized search optimization with 3-fold cross-validation.

Main Results:

• The study included 187 participants with suspected CAD from EVINCI and ARTreat clinical studies.
• The predictive model achieved an overall accuracy of 0.81.
• Both imaging (geometrical, blood flow) and non-imaging data were utilized for training.

Conclusions:

• The developed machine learning model effectively predicts CAD risk.
• Combining imaging data with traditional CAD risk factors offers an integrated predictive approach.
• This integrated model holds promise for improved patient risk stratification.