Optimum Subset Approach for Automatically Finding Effective Risk Markers in Coronary Artery Diseases

Assessing the lipid-related blood indexes is the most common method for identifying individuals at high risk of coronary artery diseases (CAD). However, the current approach requires testing a series of blood lipid parameters to predict CAD risk, which has two drawbacks: 1) examining almost all lipid-related indexes increases the cost of medical examinations; 2) clinicians face difficulty in identifying which blood lipid parameters are linked to CAD. To address these issues, a new optimum subset approach is proposed to automatically find the effective risk markers for CAD from the serial blood lipid parameters. Based on the identified risk markers, the prediction accuracy for CAD was 84.34%, with the specificity of 81.23% and sensitivity of 82.17%. By effectively identifying relevant risk markers, this method improves testing efficiency and allows doctors to focus on more effective clinical indexes.