Computer Aided Diagnosis from CT Images Using Wavelet Features with Progressive Classification

Computer aided diagnostic system for classification of normal and infected tissue from computed tomography (CT) images of abdomen is presented. The region of interest (ROI) is segmented from CT images of normal tissues, cysts, stones and tumors using active contour models. The ROIs are fed to the next stage that is feature extraction stage. In the feature extraction stage, wavelet features are extracted from segmented ROIs. In this paper, five different features are used to form feature vectors. The classification is done by using the concept of progressive classification. The classification of input query image is done progressively by eliminating unlikely ROIs from the data set till only one ROI is left. The class of the left over ROI is assigned to the unknown ROI. The results show that CAD presented in this paper achieves a classification accuracy of 87.5% which is much better than the conventional texture extraction and classification systems.