Multimodal Artificial Intelligence in Chronic Liver Disease – Current Applications, Clinical Insights and Future Directions: A Literature Review

Introduction: Chronic liver disease is a complex condition that may not be fully assessed using a single data source. Multimodal artificial intelligence can combine different types of data and may improve disease evaluation. This review examined current applications, performance, and validation of multimodal artificial intelligence models in chronic liver disease.Methods: A literature search was conducted in PubMed using the “in the last 5 years” publication-date filter, corresponding to January 26, 2021, through January 26, 2026. Relevant studies were identified using keywords related to chronic liver disease, artificial intelligence and multimodal data integration. After screening titles and abstracts, full-text articles were assessed for eligibility. Studies were included if they applied artificial intelligence or machine learning methods to integrate multimodal data in chronic liver disease. In total, 11 studies met the inclusion criteria and were included in the final analysis.Results: Reported area under the receiver operating characteristic curve (AUC) values varied across studies and validation settings. Three main types of approaches were identified: (i) models combining imaging data with clinical or laboratory data (seven studies); (ii) models integrating multiple imaging modalities, sequences, or imaging-derived signals (three studies); and (iii) a model integrating genetic and clinical data (one study). Among the seven studies combining imaging with clinical or laboratory data, four reported higher AUCs for integrated models than for models using fewer data sources in internal validation or test analyses. Separately, only three studies tested at least one outcome using data from an independent institution or an independent data source.Discussion: The findings suggest that multimodal integration may provide useful predictive information by combining different aspects of chronic liver disease. However, differences in study populations, clinical tasks, and validation designs limited comparisons across studies. Limited external validation and difficulties in interpreting complex models remain important barriers to clinical translation.Conclusion: Multimodal artificial intelligence may support the diagnosis and risk prediction of chronic liver disease by integrating diverse data sources. However, external validation in larger and more diverse populations, direct comparisons with models using fewer data sources, and improved model interpretability are needed before broader clinical implementation.