Differentiation of low- and high- grade hepatocellular carcinomas with texture features and a machine learning model in arterial phase of contrast-enhanced MR

Texture has been a recognized feature for biological aggressiveness of hepatocellular carcinomas (HCCs). However, texture feature alone may not be optimal to characterize malignancy of HCC. Computer-aided techniques combined with multi-feature fusion may be a method of choice for the preoperative assessment of the aggressiveness of HCC. To this end, a computer-aided method in the combination with machine learning technique based on texture analysis for malignancy differentiation of HCCs was desmonrated and high classification performance(AUC>0.9) of the classifier was achieved to differentiate low- and high- grade of HCCs.