Artificial intelligence-based model for the recurrence of hepatocellular carcinoma after liver transplantation.

Abstract

Artificial intelligence-based models might improve patient selection for liver transplantation in hepatocellular carcinoma. The objective of the current study was to develop artificial intelligence-based deep learning models and determine the risk of recurrence after living donor liver transplantation for hepatocellular carcinoma. The study was a single-center retrospective cohort study. Patients who underwent living donor liver transplantation for hepatocellular carcinoma were divided into training and validation cohorts (n = 192). The deep learning models were used to stratify patients in the training cohort into low- and high-risk groups, and 5-year recurrence-free survival was assessed in the validation cohort. The median follow-up period was 59.1 (33.9-72.4) months. The artificial intelligence model (pretransplant factors) had an area under the curve of 0.86 in the training cohort and 0.71 in the validation cohort. The largest tumor diameter and alpha-fetoprotein level had the greatest Shapley Additive exPlanations values for recurrence (>0.4). The 5-year recurrence-free survival rates in the low- and high-risk groups were 92.6% and 45% (P < .001). In the second artificial intelligence model (pretransplant factors + grade), the area under the curve for the validation cohort was 0.77, with 5-year recurrence-free survival rates of 96% and 30% in the low- and high-risk groups (P < .001). None of the low-risk patients outside the Milan and University of California San Francisco Criteria had recurrence during follow-up. The artificial intelligence-based hepatocellular carcinoma transplant recurrence models might improve patient selection for liver transplantation.