Classification of Liver Tumor using Modified SFTAbased Multi Class Support Vector Machine

Abstract

Liver tumor is one of the most common diseases and it increase the mortality rate. So detection at its early stage is very important and this is done through different non-invasive imaging techniques such as US scan, MRI, CT, PET etc. In this work CT images are used for the detection and classification of liver tumor. The CT imaging is compatible with MRI in respect of image quality and economy. For the classification of liver tumor, the CT image is segmented first for obtaining the boundaries of liver and liver tumor in it. Segmentation of liver and its tumor is done using adaptive threshold method. A modified SFTA algorithm is used to extract the features from tumor and these features are used for tumor classification. In this work SVM is used as classifiers. The performance analysis is taken between SFTA based multi SVM classifier and modified SFTA based multi SVM classifier. The results show that modified SFTA based SVM classifier provides accurate results of 94.531% over SFTA based multi SVM classifier.