Automatic Segmentation and Classification of Diffused Liver Diseases using Wavelet Based Texture Analysis and Neural Network

In this paper a computer aided diagnostic system for classifying diffused liver diseases from Computerized Tomography (CT) images using wavelet based texture analysis and neural network is presented. Liver is extracted from CT abdominal images using adaptive threshold and morphological processing. Orthogonal wavelet transform is applied on the liver to get horizontal, vertical and diagonal details. The statistical texture features like Mean, Standard deviation, Contrast, Entropy, Homogeneity and Angular second moment are extracted from these details and hence the eighteen features are used to train the Probabilistic neural network to classify the liver as fatty or cirrhosis. The proposed system is tested for 100 images. It produces an accuracy of 95%. The performance of the proposed system is also evaluated by calculating specificity, sensitivity, positive prediction value and negative prediction value. The performance measures of the above system are compared with the results evaluated by radiologists.