AI in Liver Disease and Hepatology

Liver disease covers several distinct problems: fatty change, inflammation, fibrosis, cirrhosis and cancer. AI tools for hepatology therefore perform different tasks. Image models may analyze ultrasound or other scans for steatosis or fibrosis, while pathology systems score features on a liver biopsy. Risk models can also estimate outcomes for patients awaiting liver transplant. A result from one task should not be treated as evidence for another; a model that classifies biopsy slides does not automatically diagnose a patient from a routine ultrasound. A current example is AIM-NASH, an AI-based histology measurement tool qualified by the U.S. FDA as a drug-development tool for MASH clinical trials. It scores biopsy components including steatosis, inflammation and fibrosis. Pathologists remain responsible for final interpretation and review the whole slide and AI output before accepting or rejecting scores. This is a specific clinical-trial use, not a general-purpose diagnostic clearance. Research has also examined ultrasound-based fibrosis and fatty-liver assessment, comparing models with biopsy or other reference tests. Transplant allocation is a separate area: the GEMA-AI study trained and internally validated a liver-waitlist model using UK registry data and tested it externally with Australian patients. It estimated which candidates might face death or removal from the list within a defined period, and modeled different prioritization from existing scores. That study did not make GEMA-AI a universal allocation rule. Transplant decisions involve policy, equity, organ availability and multidisciplinary review. Across all these applications, model outputs need validation in the intended population and should be interpreted with clinical context.