The judgment of the three major features of LI-RADS by radiologists is subjective and time-consuming. We proposed an explainable and quantitative algorithm based on DCE MRI to recognize the three major features and then get LI-RADS grades together with tumor diameter. The AUC is 0.96, 0.92, 0.70 in the validation set and 0.98, 0.90, 0.76 in the testing set for arterial phase hyper-enhancement (APHE), washout, and capsule. The overall accuracy of LI-RADS grades is 0.68 and 0.71 for the validation and testing set. The developed automatic LI-RADS grading system can provide explainable results for HCC diagnosis with high efficiency.
This abstract and the presentation materials are available to members only; a login is required.