Hepatic tumor detection in ultrasound images

Hepatic tumors are tumors that grows on or in the liver. They are classified into benign and malignant tumors. Hepatocellular carcinoma is the most frequent malignant tumor in the liver. Ultrasound is the first line investigation carried out by the physician for any abnormalities in the liver. The only golden standard for detection of liver tumor is needle biopsy, but it is invasive and causes secondary infection and bleeding at that site. In this work we present a non invasive method for detection of hepatic tumors based on ultrasound images and classification is done to differentiate the tumors in the liver. The proposed method consist of three stages namely segmentation, feature extraction and classification. In the first stage the ultrasound image containing the tumor is segmented using Fuzzy C means clustering algorithm. In the second stage gray level co-occurrence matrix features are extracted from the segmented image and Haralick texture features are extracted. In the third stage consist of training the extracted features using SVM and classification is done for normal and abnormal image. The Fuzzy C means clustering combined with SVM outperforms the other classifiers with a sensitivity of 98%.