Increasing Sensitivity, Specificity and PPV for Liver Tumor Segmentation and Classification Using Enhanced GLCM
Sreeraj R1* and Raju G21Research Scholor, Baharathiyar University, Coimbatore, India.
2Department of IT, Kannur University, Kerala, India.
Corresponding Author E-mail: sreerajr@gmail.com
Abstract: The paper presents an automatic segmentation and classification of liver tumor segmentation in CT images. Computed Tomography (CT) is a standout amongst the most generous medical imaging modalities. CT images are extensively used for liver tumor diagnosis. The precise identification of liver tumor classification and segmentation is based on the accuracy. The decrease in sensitivity, specificity and positive predictive value (PPV) directly affects the accuracy of classification and segmentation. This paper mainly focuses on improving sensitivity, specificity and PPV using an enhanced gray level co-occurrence matrix (GLCM) method. The proposed method uses LiverCT image as a source which has been preprocessed and segmented using Adaptive Threshold Segmentation method. Using the region of interest selection, 13 texture features has been extracted using GLCM method and classification is achieved using SVM classifier. We have tested our proposed method with 100 images and a comparative analysis has been made with three classifiers such as support vector machine(SVM), KNN and Bayesian in which SVM performs better. The result also shows a promising increase in specificity and sensitivity while using SVM. The proposed method achieves 99.4% sensitivity, 99.6% specificity, 97.03% PPV and hence overall accuracy is 99.5% which shows a commendable identification rate of liver tumor.
Keywords: Computed Tomography (CT); Liver Segmentation; Specificity; Sensitivity; PPV; NPV; GLCM Back to TOC