Rethinking selection criteria for liver transplantation in hepatocellular carcinoma: from morphology to artificial intelligence

INTRODUCTION: Liver transplantation (LT) is a uniquely curative treatment for hepatocellular carcinoma (HCC), as it removes both the tumor and the underlying cirrhotic liver. However, organ scarcity makes candidate selection a central clinical and ethical issue. Over the past three decades, selection has evolved from simple morphologic thresholds toward increasingly multidimensional models of oncologic risk. AREAS COVERED: This review examines the evolution of selection criteria for LT in HCC, from the Milan criteria and subsequent morphology-based expansions to models incorporating tumor biology, serum biomarkers, treatment response, waiting-list dynamics, and artificial intelligence. The literature discussed includes landmark morphologic criteria, AFP-based models, PIVKA-II and AFP-L3 integration, downstaging strategies, dynamic scores such as TRAIN and NYCA, and machine-learning models including MoRAL-AI, TRIUMPH, and TRAIN-AI. EXPERT OPINION: The field is moving from categorical eligibility rules toward individualized, continuous risk estimation. Future selection should not aim merely to expand transplant access, but to identify patients most likely to achieve durable oncologic benefit. Artificial intelligence, radiomics, liquid biopsy, and multi-omic profiling may refine risk prediction, but their clinical adoption will require prospective validation, transparency, calibration, transportability, and integration into allocation systems.