Identification of Diffusion‐Relaxation Compartments Based on Correlation to Renal Pathological Scores: An Explorative Study on Lupus Nephritis

ABSTRACT Background Renal histologic findings are beneficial in the diagnosis and treatment of lupus nephritis (LN), but require an invasive biopsy. Diffusion‐relaxation correlated spectrum imaging (DR‐CSI) has previously been applied to assess renal fibrosis, yet current approaches for compartment identification lack quantified standards. Purpose: This exploratory study is aimed at establishing a spectrum‐pathology correlation analysis for identifying possible DR‐CSI compartments, and explore the diagnostic value of the compartment fractions on common pathologic features. Study Type Cross‐sectional. Population/Subjects 66 LN patients (58 females) and 20 healthy volunteers (17 females). Field Strength/Sequence 3 T scanner/Axial renal DR‐CSI scan with 34 contrasts (TE 55~170 ms, b‐values 0~1200 s/mm 2 ). Assessment Pathologic classification, activity index, and chronicity index were obtained through biopsy. The estimated glomerular filtration rate (eGFR) was assessed, representing renal function. Four compartments, short‐T2 (S), restricted‐diffusion (R), long‐T2 (L), and pseudo‐diffusion (PD), were identified. Compartment fractions F X were calculated (X: compartment label). Statistical Tests Spearman correlation for spectrum‐pathology relationship. Student's t ‐test or Mann–Whitney U ‐test for group comparison. ROC analysis for diagnostic performance. Significance criteria: p < 0.05. Results LN patients had significantly higher F R (12.3% ± 5.6% vs. 7.6% ± 2.3%), lower F PD (36.3% ± 9.1% vs. 29.6% ± 13.0%), lower F L (30.1% ± 12.1% vs. 24.8% ± 7.6%), and higher ADC (1.95 ± 0.19 vs. 1.78 ± 0.22 μm 2 /ms) than healthy volunteers. Patients with impaired renal function had significantly higher F R (11.1% ± 5.3% vs. 15.4% ± 5.2%), higher F L (27.8% ± 10.8% vs. 36.0% ± 13.5%), lower F PD (32.3% ± 12.3% vs. 23.3% ± 12.7%), and lower ADC (1.82 ± 0.19 vs. 1.69 ± 0.23 μm 2 /ms). Proliferative LN patients had significantly increased F L (31.8% ± 12.1% vs. 24.8% ± 11.2%) than non‐proliferative ones. For identifying patients with impaired function, F R showed numerically highest AUC (0.736), sensitivity (0.842), and balanced accuracy (0.730) at optimal point for Youden index. For identifying proliferative patients, F L showed numerically highest AUC (0.679), sensitivity (0.800), and negative predictive value (0.474) among single univariate models. Data Conclusion The compartment fractions can identify proliferative LN and impaired function, highlighting the potential of a histology‐driven compartment identification strategy. Evidence Level 2. Technical Efficacy Stage 2.