Related Experiment Video
Updated: Aug 6, 2026
Mouse Model of Metabolic Dysfunction-Associated Steatotic Liver Disease with Fibrosis
Published on: July 18, 2025
Machine Learning Prediction of Liver Fibrosis in Patients With Metabolic Dysfunction-Associated Steatotic Liver
Salvatore Petta1, Grazia Pennisi1, Ciro Celsa1
1Section of Gastroenterology and Hepatology, Department of Health Promotion, Mother and Child Care, Internal Medicine, and Medical Specialties (PROMISE), University of Palermo, Palermo, Italy.
Background & Aims:
Accurate staging of liver fibrosis is crucial for risk stratification in patients with metabolic dysfunction-associated steatotic liver disease. We aimed to develop and validate artificial intelligence-based models capable of distinguishing fibrosis stages.
Methods:
We developed and validated machine learning models to predict fibrosis stages in more than 3600 biopsy-confirmed patients with metabolic dysfunction-associated steatotic liver disease using 22 clinical features, liver stiffness measurement, and controlled attenuation parameter. Models included Feature Tokenizer Transformer, TabNet variants with distance-aware losses, ordinal Multilayer Perceptron with the COnditional RAnk Logits framework. Three centers were prespecified for geographic external validation. In the remaining centers, we used a stratified 75/25 split into a training pool and an internal random test set, tuned hyperparameters by stratified 10-fold cross-validation on the training pool, and trained 1 model per imputed dataset (M = 5) with an internal 80/20 train-validation split for early stopping and operating-point selection. Performance was pooled across imputations using Rubin's rules and reported for the internal random test set and the held-out centers.
Results:
Feature Tokenizer Transformer and TabNet achieved the highest performance in binary tasks: area under the receiver operating characteristic curve = 0.860 and 0.855 (F ≥3 vs 0-F2) and area under the receiver operating characteristic curve = 0.800 and 0.788 (F ≥2 vs 0-F1), significantly outperforming the Fibrosis-4 index. For F ≥3, Feature Tokenizer Transformer had significantly higher area under the receiver operating characteristic curve than liver stiffness measurement (P = .014) but only marginally higher than AGILE3+, and, together with TabNet, the highest accuracy and the lowest gray zone (8.2% and 8.4%, respectively). For multiclass staging, ordinal models performed best with Multilayer Perceptron with the Consistent Rank Logits achieving quadratic weighted kappa = 0.616.
Conclusions:
Deep learning models with ordinal-aware architectures can accurately predict liver fibrosis stages using routinely available clinical data, offering a scalable alternative to biopsy without requiring specialized biomarkers.
