Comparative Study on Cancer Image Diagnosis using Soft Computing Techniques

Computer-aided diagnosis system (CAD) can be very helpful for radiologist in detection and diagnosing abnormalities earlier and faster than traditional screening programs. CAD as such employs several techniques to accomplish this task. In this paper, we propose to make a comparative study of two classification methods: One in which we utilize the texture features extracted from the images by directly feeding to the Neural Network based classifier stage to classify the images into benign or malign and in the other hybrid method, those texture features are made to undergo fuzzy discretization before feeding to the Neural Network classifier for the classification. The studies so far conducted using both the systems show that the hybrid system is far superior to the first method in its accuracy. Backward Propagation Network (BPN) algorithm is used in the training stage. General Terms Medical Image Mining, Data Mining, Cancer Diagnosis