PCA-SVM based CAD System for Focal Liver Lesions using B-Mode Ultrasound Images
The contribution made by texture of regions inside and outside of the lesions in classification of focal liver lesions (FLLs) is investigated in the present work. In order to design an efficient computer-aided diagnostic (CAD) system for FLLs, a representative database consisting of images with (1) typical and atypical cases of cyst, hemangioma (HEM) and metastatic carcinoma (MET) lesions, (2) small as well as large hepatocellular carcinoma (HCC) lesions and (3) normal (NOR) liver tissue is used. Texture features are computed from regions inside and outside of the lesions. Feature set consisting of 208 texture features, (i.e. 104 texture features and 104 texture ratio features) is subjected to principal component analysis (PCA) for finding the optimal number of principal components to train a support vector machine (SVM) classifier for the classification task. The proposed PCA-SVM based CAD system yielded classification accuracy of 87.2% with the individual class accuracy of 85%, 96%, 90%, 87.5% and 82.2% for NOR, Cyst, HEM, HCC and MET cases respectively. The accuracy for typical, atypical, small HCC and large HCC cases is 87.5%, 86.8%, 88.8%, and 87% respectively. The promising results indicate usefulness of the CAD system for assisting radiologists in diagnosis of FLLs.
Defence Science Journal, 2013, 63(5), pp.478-486, DOI:http://dx.doi.org/10.14429/dsj.63.3951
Where otherwise noted, the Articles on this site are licensed under Creative Commons License: CC Attribution-Noncommercial-No Derivative Works 2.5 India