Detection and Diagnosis of Cancer Regions in Oral MRI Images using NN and Random Forest Classification Methods

Abstract

            The formation of the abnormal cells in lips and tongue regions might create cancer cells, which could be curable if they are detected in earlier manner. The proposed method consists of pre-processing, enhancement, Gabor transform, feature extractions and classifications. The noises in the source oral image are detected and filtered using random valued filter, and the filtered image is enhanced using adaptive histogram equalization technique. The enhanced oral image is transformed into multi resolution image using Gabor transform and the features are extracted from this transformed image. In this work it is proposed to classify the features using Neural Network (NN) classification approach and to perform diagnosis of the cancer cells using random forest classifier. The classifiers are used to diagnose the segmented ca ncer regions. The proposed method obtains 98% of diagnosis rate for mild case and 99.3% for severe case. The performance evaluation of oral cancer segmentation using NN and Random Forest classification achieves 96.6% of sensitivity, 98.1% of specificity and 98.6% of oral cancer segmentation accuracy.