Machine learning-based coronary artery disease diagnosis: A comprehensive review

Area of Science:

• Cardiology
• Medical Informatics
• Artificial Intelligence

Background:

• Coronary artery disease (CAD) is a leading cause of death and significant financial burden globally.
• Current diagnostic methods like angiography are invasive and carry risks.
• Machine learning (ML) presents a promising noninvasive, cost-effective approach for CAD detection.

Purpose of the Study:

• To conduct a comprehensive review of ML-based CAD diagnosis studies published between 1992 and 2019.
• To investigate the influence of dataset characteristics and ML techniques on diagnostic performance.
• To identify challenges and limitations in the current literature on ML for CAD.

Main Methods:

• Systematic literature review of studies on ML for CAD diagnosis.
• Analysis of factors including dataset size, geographical origin, features used, and stenosis data.
• Evaluation of various ML techniques, feature selection methods, and performance metrics.

Main Results:

• Significant heterogeneity exists across ML-based CAD studies regarding datasets and methodologies.
• Performance of ML models varies widely due to differences in data characteristics and applied techniques.
• Lack of standardization hinders the generalizability of reported ML achievements in CAD detection.

Conclusions:

• The variability in ML study designs impedes the reliable generalization of findings for CAD diagnosis.
• Further research is needed to standardize methodologies and validate ML models for clinical application.
• Addressing identified challenges is crucial for advancing ML-based CAD detection.