A Hybrid Data Mining Model to Predict Coronary Artery Disease Cases Using Non-Invasive Clinical Data

Area of Science:

• Cardiology
• Artificial Intelligence
• Medical Informatics

Background:

• Coronary artery disease (CAD) diagnosis relies on invasive angiography, which is costly and time-consuming.
• There is a need for alternative, noninvasive methods for accurate CAD diagnosis and severity assessment.

Purpose of the Study:

• To develop and evaluate a novel hybrid machine learning method for CAD diagnosis using noninvasive clinical data.
• To identify key risk factors associated with CAD.
• To compare the performance of various supervised learning algorithms for CAD prediction.

Main Methods:

• A hybrid approach combining Correlation-based Feature Subset (CFS) selection with Particle Swarm Optimization (PSO) for risk factor identification.
• K-means clustering for data segmentation.
• Supervised learning algorithms including Multi-Layer Perceptron (MLP), Multinomial Logistic Regression (MLR), Fuzzy Unordered Rule Induction Algorithm (FURIA), and C4.5 for CAD case modeling.
• Validation on clinical data from Indira Gandhi Medical College, Shimla, and the benchmarked Cleveland heart disease dataset.

Main Results:

• The hybrid model achieved the highest prediction accuracy of 88.4% using Multinomial Logistic Regression (MLR) on the clinical dataset.
• MLR outperformed other techniques on the Cleveland heart disease dataset as well.
• The proposed hybridized model demonstrated an accuracy improvement of 8.3% to 11.4% for classification algorithms on the Cleveland data.

Conclusions:

• The developed hybrid machine learning method is a promising tool for the noninvasive diagnosis of coronary artery disease.
• The approach offers improved prediction accuracy compared to existing methods.
• This technique facilitates efficient identification of patients with CAD using readily available clinical data.