Artificial Intelligence in the Evaluation of Abnormal Liver Tests and MASLD: Emerging Applications in Risk Stratification and Clinical Decision Support

Abstract

Background: Abnormal liver tests and metabolic dysfunction-associated steatotic liver disease (MASLD) are common, yet the central clinical task is identifying advanced fibrosis, cirrhosis, or an alternative liver disease that requires timely investigation or referral. Methods: We conducted a narrative review of PubMed, Ovid MEDLINE, Embase, Scopus, and Web of Science from database inception through April 2026, supplemented by searches of Google Scholar, reference lists, and relevant society guidelines. Results: Artificial intelligence (AI) can integrate clinical, laboratory, longitudinal, imaging, and elastography data to support risk stratification, identify missing investigations, and assist referral workflows. Representative studies reported promising discrimination for selected outcomes, but populations, reference standards, and validation methods were heterogeneous; calibration and external validation were often absent or incompletely reported. Clinical use should begin with standard pattern recognition, exclusion of competing etiologies and urgent red flags, and age-aware interpretation of the fibrosis-4 index before second-line testing or referral. Conclusions: AI is an emerging adjunct, not a replacement for clinical judgment. Evidence that it improves outcomes, reduces missed advanced fibrosis, or increases referral efficiency remains limited. Responsible adoption requires transparent models, clinician verification, external validation, calibration, workflow evaluation, and post-deployment monitoring.