Computerized features for LI-RADS based computer-aided diagnosis of liver lesions

Abstract

Liver Imaging Reporting Data System (LI-RADS) aims to standardize liver lesion imaging findings and diagnostic reports, and it is used as an accurate noninvasive diagnosis and staging method of hepatocellular carcinoma (HCC) nowadays. In this study, we proposed several computerized features for LI-RADS based computer-aided diagnosis of liver lesions. We used several popular machining learning approaches for computerized LI-RADS classification (benign and malignant classification) with our proposed features. The performance of each method was evaluated by using ROC curve and the best AUC score was 0.965 reached by the gradient boosting classifier.