Classification of metabolic syndrome patients using implemented expert system

Abstract

This paper presents the development of an Expert System for the classification of metabolic syndrome (MetS). Two-layer feedforward Artificial Neural Network (ANN) with sigmoid transfer function is used for MetS classification. In accordance with international guidelines NHBL/AHA, classification is performed based on following input parameters: waist circumference, blood pressure, glucose level, HDL cholesterol and triglycerides. Samples for training of developed Expert System are obtained from 1083 patients at hospitals in Bosnia and Herzegovina. Testing of developed system is performed with 300 samples, also acquired from patients in hospitals in B&H by medical professionals. Out of 300 samples, 155 samples were of MetS while the rest was of healthy subjects. Developed Expert System correctly classified 283 MetS samples, therefore the sensitivity of 96% is achieved and specificity is 92,7%.