Hepatic lesions classification by ensemble of SVMs using statistical features based on co-occurrence matrix

Liver diseases are among the leading causes of death worldwide. The most useful approach for controlling the growth of diseases to reach at severe condition is to treat these diseases at the early stages. Early treatment requires early diagnosis, which needs an accurate and reliable diagnostic procedure. The aim of this study is to develop a computer-aided diagnostic system to achieve aforementioned objective. Computed tomography is one of the most common and robust imaging techniques for the detection of liver lesions. Although in recent years the quality of CT images has been significantly improved, however in some cases image interpretation by human beings is often limited. We developed an automated system to detect and classify liver anomalies using CT images. Region of interest from CT images is segmented using active contours (snakes) algorithm and segmented image is used to extract statistical features based on co-occurrence matrix. To facilitate the classification of hepatic lesions, support vector machine is used. The results show that it is possible to automatically identify patients with liver lesions like hemangioma, hepatoma or cirrhosis based on texture features and that the machine performance is satisfactory and can assist human experts.