Non‑invasive nomogram predicting moderate‑severe renal interstitial fibrosis and tubular atrophy integrating biochemical markers and ASL‑MRI histogram features in chronic kidney disease
Year: 2026
Type: preprint
Abstract: Objective This study aimed to construct a non‑invasive predictive model for renal interstitial fibrosis and tubular atrophy (IFTA) in patients with chronic kidney disease (CKD) by integrating clinical biochemical indicators and histogram features derived from arterial spin‑labeling magnetic resonance imaging (ASL‑MRI), and to evaluate its diagnostic performance and clinical translational value. Methods A prospective cohort of 98 CKD patients undergoing renal biopsy and 3.0T three‑dimensional pseudo‑continuous arterial spin‑labeling (3D‑pCASL) examination between November 2024 and May 2026 were enrolled. Patients were divided into mild IFTA group (n = 61) and moderate‑severe IFTA group (n = 37) according to pathological IFTA grading. Clinical biochemical markers including serum creatinine (Scr) and estimated glomerular filtration rate (eGFR) were collected. Mean renal cortical perfusion and 17 histogram parameters from ASL‑MRI were extracted. Univariate analysis and least absolute shrinkage and selection operator (Lasso) regression were used to screen predictive factors, followed by collinearity assessment before building a multivariate logistic regression combined model. Two comparator models (biochemical‑only model and ASL mean perfusion‑only model) were built simultaneously. Receiver operating characteristic (ROC) curve, 1000‑time bootstrap internal validation, decision curve analysis (DCA), and eGFR‑stratified subgroup analysis were performed for comprehensive model evaluation. Results Univariate analysis revealed inter‑group differences in Scr, eGFR, blood urea nitrogen, serum uric acid, 24‑hour urinary protein (24hUP), and 13 ASL histogram parameters. Lasso regression retained five core predictors: Scr, eGFR, 24hUP, Median and TotalEnergy. Variance inflation factor (VIF) < 5 for all variables indicated no severe multicollinearity. Multivariate logistic regression showed that only 24hUP was an independent risk factor for moderate‑severe IFTA. The AUC was 0.737 for the ASL mean perfusion‑only model, 0.904 for the biochemical‑only model, and 0.946 for the combined multi‑modal model. Delong test showed the combined model outperformed the ASL mean perfusion‑only model significantly (P < 0.001), whereas the AUC improvement compared with biochemical‑only model was not statistically significant (P = 0.092). After 1000‑bootstrap internal validation, the corrected AUC reached 0.921; Hosmer‑Lemeshow test P = 0.845 indicated good calibration. DCA demonstrated favorable net clinical benefit across risk thresholds. Subgroup analysis showed stable performance with AUC = 0.919 in patients with eGFR ≥ 60 mL/min/1.73 m² and AUC = 0.985 in patients with eGFR < 60 mL/min/1.73 m². A nomogram was plotted for individualized risk estimation of moderate‑severe IFTA. Conclusion ASL‑MRI histogram analysis quantifies heterogeneous renal cortical perfusion and compensates limitations of single mean perfusion metrics for focal parenchymal injury detection. The multi‑modal combined model integrating biochemical markers and ASL histogram features achieves good discriminative power, favourable calibration and stable subgroup performance, showing promising potential for risk stratification and longitudinal follow‑up in CKD patients, although the incremental diagnostic value added by ASL‑MRI histogram features did not reach statistical significance compared with biochemical markers alone.
Source: Research Square
Institution: Ningbo Medical Center Lihuili Hospital
Language: English
Subfield: Nephrology
Field: Medicine
Domain: Health Sciences
Open Access status: green
OpenAlex ID: W7225783018