Machine Learning Prediction of Liver Fibrosis in Patients with Metabolic Dysfunction Associated Steatotic Liver Disease (MASLD).

BACKGROUND: Accurate staging of liver fibrosis is crucial for risk stratification in patients with metabolic dysfunction-associated steatotic liver disease (MASLD). We aimed to develop and validate AI-based models capable of distinguishing fibrosis stages. METHODS: We developed and validated machine learning models to predict fibrosis stages in more than 3,600 biopsy-confirmed MASLD patients using 22 clinical features, liver stiffness measurement (LSM), and controlled attenuation parameter. Models included Feature Tokenizer Transformer (FTT), TabNet variants with distance-aware losses, ordinal MLP with the CORAL framework. Three centers were pre-specified 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 one 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: FTT and TabNet achieved the highest performance in binary tasks: area under the receiver-operating characteristic curve (AUROC)=0.860 and 0.855(F≥3vsF0-F2) and AUROC=0.800 and 0.788(F≥2vsF0-F1), significantly outperforming FIB-4. For F≥3, FTT had significantly higher AUROC than LSM(p=0.014) but only marginally higher than AGILE3+, and, together with TabNet, the highest accuracy and the lowest grey zone (8.2% and 8.4%, respectively). For multiclass staging, ordinal models performed best with MLP-CORAL 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.