A noninvasive method for coronary artery diseases diagnosis using a clinically-interpretable fuzzy rule-based system

Area of Science:

• Cardiology
• Computer Science
• Medical Informatics

Background:

• Coronary artery disease (CAD) is a leading cause of death and disability globally.
• Current diagnostic methods like invasive coronary angiography are expensive and carry risks.
• There is a need for noninvasive, cost-effective CAD detection tools.

Purpose of the Study:

• To design a computer-based, noninvasive system for diagnosing coronary artery disease (CAD).
• To develop a system that provides clinically interpretable rules for CAD diagnosis.

Main Methods:

• Utilized the Cleveland CAD dataset for analysis.
• Employed a neuro-fuzzy classifier (NFC) with a scaled conjugate gradient algorithm.
• Applied feature selection methods (MLR, sequential FS) to reduce attributes and enhance classifier performance.

Main Results:

• The MLR + NFC model achieved 84% accuracy, 79% sensitivity, and 89% specificity.
• Key predictive features identified include age, ST/heart rate slope, exercise-induced angina, fluoroscopy, and thallium-201 stress scintigraphy.
• The system demonstrated substantial agreement with the gold standard diagnosis.

Conclusions:

• The developed computer-based system is a promising tool for screening coronary artery disease (CAD) patients.
• The noninvasive approach offers a potentially safer and more accessible method for CAD assessment.
• The system's interpretability aids clinical decision-making in CAD diagnosis.