A Medical Decision Support System Based on Ensemble of Complex-Valued Radial Basis Function Networks

Abstract

The use of machine learning techniques for medical diagnosis has become increasingly common in recent years because, most importantly, the computer-aided diagnostic systems developed for supporting the experts have provided effective results. The authors aim in this chapter to improve the performance of classification in computer-aided medical diagnosis. Within the scope of the study, experiments have been performed on three different datasets, which include heart disease, hepatitis, and BUPA liver disorders datasets. First, all features obtained from these datasets were converted into complex-valued number format using phase encoding method. After complex-valued feature set was obtained, these features were then classified by an ensemble of complex-valued radial basis function (ECVRBF) method. In order to test the performance and the effectiveness of the medical diagnostic system, ROC analysis, classification accuracy, specificity, sensitivity, kappa statistic value, and f-measure were used. Experimental results show that the developed system gives better results compared to other methods described in the literature. The proposed method can then serve as a useful decision support system for medical diagnosis.