Deep-learning model predicts hepatocellular carcinoma recurrence effectively

A novel deep-learning model, Recurr-Net, significantly outperformed traditional methods in predicting hepatocellular carcinoma (HCC) recurrence post-curative surgery. Analyzed across 1,231 patients, Recurr-Net demonstrated an area under the receiver operating characteristic curve (AUROC) of 0.770 to 0.857 internally and 0.758 to 0.798 externally, while the historical microvascular invasion (MVI) showed much lower performance. This indicates Recurr-Net’s potential for superior pre-operative prognostication, crucial for risk stratification in HCC management.