Artificial intelligence in hepatology: A comprehensive scoping review of clinical applications, challenges, and future directions

Abstract

Background and aims

Artificial intelligence (AI) is increasingly integrated into hepatology, yet the existing evidence is fragmented. This scoping review systematically mapped clinical AI applications across hepatology, summarized validated outcomes, and highlighted implementation challenges and research priorities.

Methods

Following the Arksey–O’Malley and Levac frameworks and reported in accordance with PRISMA-ScR, a systematic search of PubMed, Embase, Scopus, Web of Science, and IEEE Xplore (January 2018–July 2025), complemented by grey-literature screening, identified relevant studies. Two reviewers independently screened and extracted data. From 3214 records, 75 studies met inclusion criteria. (PROSPERO ID: CRD420251159117).

Results

Most studies were retrospective and single-center. Imaging models achieved area-under-curve values of 0.80–0.95 for fibrosis staging, lesion detection, and volumetry, often comparable with expert radiologists. Digital-pathology algorithms enabled objective quantification of fibrosis and steatosis. Machine-learning tools improved prediction of disease progression, readmission, and mortality compared with conventional scores, while AI-based transplantation models enhanced donor–recipient matching and graft-survival forecasting. Natural-language-processing systems facilitated early complication detection from electronic health records. Common barriers included small datasets, limited external validation, and model interpretability concerns.

Conclusions

AI demonstrates strong potential for diagnostic, prognostic, and workflow enhancement in hepatology. Its responsible translation requires multimodal data integration, explainable modeling, prospective and multicenter validation, and equity-focused deployment strategies.