Publication

Abstract : Liver cancer is one of the common cancer among all other cancers and also it leads to death. Some traditional algorithms are used in the liver segmentation process. But still it has some limitations, such as not effective liver tissue segmentation in the presence of large dataset, less accurate result and difficult to apply tumor segmentation for large intensities of tumor region, high computational cost, required to improve the performance of tiny parts of liver and the misclassification of the tumors in near liver boundaries. To overcome these issues, an efficient method is developed. Initially, the image is localized and these images are preprocessed by median filter and active contour segmentation technique. Then the segmented liver image is given to the novel Edge Prediction based Segmentation (EPBS) algorithm in order to segment the tumor from the liver. The novel Window Analysis based Pattern formation technique is used to forming the Patterns over the segmented image. From this obtained pattern, the features are extracted and offers the necessary information that are present in the segmented image. Finally, the normal and abnormal classes are classified by novel Supervised Machine Learning Classifier. The performance of the proposed technique is analyzed by using 3D IR CAD dataset with varies parameters such as, Volumetric Overlap Error (VOE), accuracy, precision, recall, and the coefficients are Jaccard, dice, and kappa. The proposed method of EPbs-VDBC performance is higher than other classification techniques of Support Vector Machine (SVM), probabilistic neural network (PNN) and Relevance Vector Machine (RVM).