Linear discriminant analysis and principal component analysis to predict coronary artery disease

Area of Science:

• Biomedical Sciences
• Data Mining
• Cardiology

Background:

• Coronary artery disease is a leading cause of mortality globally.
• Myocardial ischemia poses a significant public health challenge.
• Advanced data analysis is crucial for understanding cardiovascular diseases.

Purpose of the Study:

• To apply data mining techniques for myocardial ischemia detection.
• To evaluate the efficacy of linear discriminant analysis and principal component analysis.
• To support clinical decision-making in diagnosing coronary artery disease.

Main Methods:

• Analysis of 22 features from 10,265 patient records.
• Implementation of linear discriminant analysis using Knime and R.
• Application of principal component analysis for feature reduction prior to classification.

Main Results:

• Classification accuracies of 84.5% (Knime) and 86.0% (R) were achieved.
• High specificity (over 97%) was consistently obtained.
• Sensitivity ranged from 62% to 66% for patient classification.

Conclusions:

• Data mining offers a practical approach to aid clinicians in diagnosing myocardial ischemia.
• Principal component analysis enhances classification performance by reducing feature dimensionality.
• These techniques provide valuable tools for managing coronary artery disease.