AI In Diagnosing Metabolic Dysfunction- Associated Steatotic Liver Disease (Masld): A Narrative Review

Abstract

Formerly known as non-alcoholic fatty liver disease (NAFLD), metabolic dysfunction-associated steatohepatitis (MASH), fibrosis, cirrhosis, and hepatocellular carcinoma (HCC), MASLD is one of the most common diseases in the world, affecting 28% of adults and being the primary cause of chronic liver disease. Although the liver biopsy is the gold standard for diagnosing MASLD, the hunt for trustworthy non-invasive diagnostic techniques has been spurred by its invasive nature, sample inaccuracy, expenses, and consequences. For improving MASLD diagnosis, staging, and prognosis, artificial intelligence (AI) which includes machine learning, deep learning, radiomics, and digital pathology can be regarded as breakthrough technology. The current narrative review summarizes research on the use of AI-assisted techniques in comparison to histological, imaging, or clinical reference standards that was done between January 2018 and July 2026 and was found in PubMed, Embase, Scopus, Web of Science, ScienceDirect, and the Cochrane Library. With AUROCs ranging from 0.85 to 0.99 for steatosis diagnosis and 0.82 to 0.97 for fibrosis staging, AI-based imaging models utilizing ultrasound, computed tomography, magnetic resonance imaging, and elastography demonstrated exceptional diagnostic effectiveness. When it came to detecting clinically significant fibrosis and MASH, machine-learning algorithms that incorporated regular laboratory markers, metabolomics, proteomics, and data on the gut microbiome frequently outperformed conventional non-invasive algorithms like FIB-4 and the NAFLD Fibrosis Score. Diagnostic accuracy comparable to that of hepatopathologist experts was achieved by digitizing liver biopsies and used deep learning to analyse these images. HCC risk classification and surveillance were enhanced by using AI to create prediction models. The clinical use of these technologies is hampered by a number of factors, including inconsistent methodology, inadequate external validation, biases in data selection, and the interpretability of the results produced, despite the encouraging advancements made thus far. Because AI offers precise and minimally intrusive diagnoses, it generally has enormous potential for the treatment of MASLD.