Implementation of Neuro Fuzzy System for Diagnosis of Multiple Sclerosis

Abstract

-Medical diagnosis is often done by expertise and experience of phisician, but sometimes may lead to misdiagnosis. Multiple sclerosis (MS) is a disease of the central nervous system. In this disease, body produces antibodies that attack and damage the Myelin, in which the myelin sheath (the insulation for nerve fibers) is in trouble and the damage to myelin in the central nervous system cause to disconnect between brain and other organs. The major problem is the lack of diagnosis. To improve diagnosis, Adaptive Neuro-Fuzzy Inference System (ANFIS) is used. ANFIS main idea is that using the way the nervous system of biological for data processing in order to learn and create the knowledge. This system uses neural network for learning, classification capabilities and modifying. There are several ways to train neural network. In this study, we use hybrid approach to train. This hybrid approach uses Back Propagation(BP) and Least Square Error(LSE). ANFIS has the ability to combine the linguistic power of fuzzy system with numeric power of neural network. For optimizing the input/output, the K-fold cross validation has been used. Implementation has been done in MATLAB. The Data set consist of 600 patients that each one has 6 columns, 5 of them is input and 1 of them is output that shows diagnosis.