Interpretation of Clinical Data Based on C4.5 Algorithm for the Diagnosis of Coronary Heart Disease

Area of Science:

• Medical Informatics
• Machine Learning in Healthcare
• Cardiology Data Analysis

Background:

• Clinical data interpretation for coronary heart disease (CHD) diagnosis often uses opaque data mining algorithms.
• Existing black-box models struggle to elucidate relationships between examination attributes and CHD incidence.

Purpose of the Study:

• To develop and evaluate a novel system for interpreting clinical examination results for CHD diagnosis.
• To leverage the C4.5 decision tree algorithm for transparent data interpretation and improved diagnostic accuracy.

Main Methods:

• A system integrating synthetic minority oversampling technique (SMOTE), feature selection, and the C4.5 classification algorithm was developed.
• K-fold cross-validation was employed for rigorous system performance evaluation.
• Key performance metrics included sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and area under the curve (AUC).

Main Results:

• The proposed system achieved a sensitivity of 74.7%, specificity of 93.7%, PPV of 74.2%, NPV of 93.7%, and AUC of 84.2%.
• The decision tree output provided interpretable insights into attribute relationships relevant to CHD.

Conclusions:

• The C4.5 algorithm enables data interpretation into a decision tree format, enhancing clinician understanding.
• The developed system demonstrates improved performance in diagnosing coronary heart disease, offering category-specific insights.