Application of data mining techniques for detecting asymptomatic carotid artery stenosis

Asymptomatic carotid stenosis, one of the etiological factors for stroke, has several risk factors such as hypertension, cardiac morbidity, smoking, diabetes, and physical inactivity. Understanding and determining factors that predispose to asymptomatic carotid stenosis will help in the design of acute stroke trials and in prevention programs. The goal of this study is to explore rules and relationships that might be used to detect possible asymptomatic carotid stenosis by using data mining techniques. For this purpose, Genetic Algorithms (GAs), Logistic Regression (LR), and Chi-square tests have been applied to the patient dataset. Results of these tests have also been compared.

► LR can be used in computer-based decision support systems, but it does not provide a simple rule that helps the physician.
► GA produces simple rules that can be used in daily practice as they can be expressed in natural language unlike LR formulae.
► The two multivariate methods give better results than univariate approaches; LR yielding a bigger ROC AUC than GA.