APPLICATION OF FUZZY-MLP MODEL TO ULTRASONIC LIVER IMAGE CLASSIFICATION

Abstract

In this paper, we propose the application of fuzzy-MLP in theclassification of ultrasonic liver images. The four sets of ultrasonic liverimages used in the experiment are: normal, liver cysts, alcoholic cirrhosisand carcinoma.To deal with the sample images efficiently, we extract textural features fromthe Pathology Bearing Regions (PBRs) of the ultrasound liver images. Theselected features for the classification are entropy, energy and maximumprobability-based texture features extracted using gray level co-occurrencematrix second-order statistics. The fuzzy-MLP model is constructed for theselected features classify various categories of ultrasonic liver images.The efficacy of Fuzzy-MLP model and conventional artificial neural network(ANN) has been compared on the basis of the same feature vector. A testwith 82 training data and 110 test data for all the four classes shows 92.73%classification accuracy for the proposed fuzzy-MLP model. It is comparedwith the 81.82% counterpart provided by conventional ANN method.