Feature extraction and classification of ultrasound liver images using haralick texture-primitive features: Application of SVM classifier

This paper describes the feasibility of selecting features from the gray level co-occurrence matrix (GLCM) with 12 haralick features based on texture to classify ultrasonic diseased liver into fatty, cyst and cirrhosis. The objective in this work is the selection of the most discriminating parameters for liver disease classification. The diagnosis scheme includes three modules: preprocessing, feature analysis and classification modules. The images were preprocessed by using Anisotropic Diffusion speckle reduction method. Then the features, derived from the gray level co-occurrence matrix with twelve haralick features are extracted from both entire image and pathology bearing region (PBR) in the image. The analysis of the obtained results suggested that diseases like cyst, fatty, cirrhosis can be diagnosed with only five features namely contrast, auto correlation, Angular Second Momentum, cluster shade and cluster prominence out of 12 features belonging to haralick features. The result show that the Support Vector Machine classifiers with five haralick features show the classification accuracy rate is comparatively better when it is compared with the feature extraction by other methods for the same datasets which is our earlier work. The dataset used in each phase of the work are authenticated datasets provided by doctors. The results at each phase have been evaluated with doctors in the relevant field.