Machine Learning with Multiparametric MRI and Clinical Biomarkers for Noninvasive Renal Interstitial Fibrosis Staging

Renal interstitial fibrosis (RIF) is currently assessed by invasive biopsy. This prospective study developed and validated a noninvasive random forest model combining multiparametric MRI and clinical biomarkers for identifying severe RIF in 116 patients with biopsy-confirmed renal disease. Quantitative parameters were extracted from IVIM, ASL, phase-contrast MRI, T1 mapping, and BOLD sequences. Fibrosis was classified as mild (<25%) or severe (≥25%). In the held-out test set, the random forest model achieved an AUC of 0.89 (95% CI 0.82–0.96), sensitivity of 0.91, and specificity of 0.73, significantly outperforming clinical-only (AUC 0.63), MRI-only (AUC 0.63), and combined LASSO logistic regression (AUC 0.73) benchmarks. The model also demonstrated superior calibration (Brier score 0.154) and net clinical benefit on decision curve analysis. This integrated MRI–clinical model shows promise for noninvasive identification of severe RIF and warrants external prospective validation.