Info
Background:
Renography evaluates kidney function by tracking contrast uptake, transit, and clearance over time. This is clinically important in obstructive kidney injury, where impaired urine drainage and hydronephrosis can progress to altered acid–base homeostasis and loss of renal function. Dynamic contrast-enhanced MRI and nuclear scintigraphy employ probe kinetics to detect differences in the function of the left and right kidneys, but they do not provide pH-sensitive molecular information which could clarify if there is baseline renal disease as well. Chemical exchange saturation transfer (CEST) MRI can generate renogram-like dynamic signal curves and kidney pH maps from the same acquisition, enabling additional characterization of kidney tissue.
Hypothesis/Objective:
To develop the Uptake–Plateau–Decay (UPD) framework for analyzing dynamic CEST MRI data and test whether voxel-wise kinetic parameter maps can act as biomarkers for detecting kidney injury following obstruction in a mouse model.
Methods:
For model development, healthy C57BL/6J mice received I45DC-diGlu at a dose of 250 mg/kg (n=5), 500 mg/kg (n=5), or 750 mg/kg (n=3) and were imaged at 11.7T using dual-offset CEST MRI at 7.7 and 4.5 ppm. Signal-enhancement (SE%) curves were extracted from renal cortex, outer medulla, and inner medulla. Standard log-normal (LN) and our new UPD model were fit in parallel and compared using corrected Akaike information criterion (AICc), with model performance assessed using R², RMSE, residual distributions, and autocorrelation analysis. For disease application, an iopamidol-enhanced dynamic CEST MRI dataset from control mice (n=3) and unilateral ureteral obstruction (UUO) mice imaged on day 1, 2, or 3 after obstruction (n=3/time point) was re-analyzed using the same framework. Contralateral and obstructed kidneys were compared using pH mapping, renogram-like SE% curves, voxel-wise LN versus UPD model selection, kinetic biomarker distributions, and Wilcoxon rank-sum testing.
Results:
I45DC-diGlu produced dose-dependent renal CEST enhancement and stable ratiometric pH mapping (Figure. I-II) . Voxel-wise pH maps reproduced the expected corticomedullary gradient: cortex 6.86 ± 0.087, outer medulla 6.80 ± 0.073, and inner medulla 6.61 ± 0.079 (Figure. III). Dose escalation produced marked kinetic changes, with SEmax increasing from ~3–6% at 250 mg/kg to ~18–30% at 750 mg/kg, AUC increasing from ~2–3 to ~32–61 %·min, and absolute t1/2 increasing from ~1.6–1.9 min to ~3.2–3.3 min. AICc-based model selection revealed a dose-dependent transition from LN-described unimodal kinetics at lower doses to UPD-preferred rise–plateau–washout behavior at 750 mg/kg (Figure. IV). Residual diagnostics supported model stability, with near-zero-centered residuals, lower RMSE for the preferred model, high R2 values for well-described curves, and limited higher-lag autocorrelation.
In the UUO application, dynamic iopamidol CEST MRI detected disease-associated alterations in renal pH and probe handling (Figure. V). Cohort-level analysis showed significantly prolonged TTP in obstructed kidneys compared with contralateral kidneys (p<0.001) and elevated renal pH in UUO kidneys compared with contralateral kidneys (6.79 ± 0.11 vs 6.67 ± 0.08). Voxel-wise UPD analysis identified a focused set of obstruction-associated kinetic biomarkers. Our measurements detected a delayed TTP, altered AUC, prolonged absolute t1/2, altered washout rate (kout), and reduced plateau enhancement in obstructed kidneys, consistent with delayed renal transit, impaired clearance, and plateau-like kinetic behavior(Figure. S1). Parametric maps of TTP, AUC, absolute t1/2, kout, and plateau enhancement revealed spatially heterogeneous renal probe-handling abnormalities that were not fully captured by ROI-averaged curves(Figure. S2). Secondary biomarkers, including SEmax and kup, provided supportive evidence of altered peak enhancement and early uptake dynamics. Wilcoxon rank-sum testing of voxel-wise distributions demonstrated significant contralateral-versus-UUO differences across key primary kinetic biomarkers. Voxel-wise ΔAICc analysis further showed greater UPD preference in obstructed kidneys across days 1–3, indicating spatially heterogeneous uptake–plateau–decay behavior associated with obstruction (Figure. VD).
Conclusion:
UPD-based dynamic CEST MRI renography converts renal CEST time courses into quantitative biomarkers of probe uptake, retention/plateau behavior, washout, and pH homeostasis. By extending prior ROI-level dynamic CEST analysis to voxel-wise kinetic phenotype mapping, this approach provides a quantitative molecular imaging framework for assessing spatially heterogeneous renal functional impairment.
Image/Figure:
Click to View
Image/Figure Caption:
Figure. UPD-based dynamic CEST MRI renography enables quantitative assessment of renal probe kinetics and obstructive kidney injury.
(I) Schematic of the Uptake–Plateau–Decay (UPD) model, which describes dynamic CEST signal evolution as uptake, transient plateau, and washout phases. Parameters include B baseline signal, A uptake amplitude, kup uptake rate, kout washout rate, t0 bolus arrival, t1–t2 plateau bounds, P plateau level, Dp plateau drift, and derived washout half-life t1/2 = ln(2)/kout. These parameters enable extraction of SEmax, TTP, AUC, t1/2, uptake/washout rates, and plateau parameters. (II–IV) Model development using I45DC-diGlu in healthy kidneys. I45DC-diGlu generated dose-dependent renal pH maps at 250, 500, and 750 mg/kg (II), resolved corticomedullary pH heterogeneity with higher pH in cortex than outer and inner medulla at 750 mg/kg (III), and produced plateau-containing high-dose renal CEST kinetics that were captured by UPD fitting at 750 mg/kg (IV). (V) Disease application using iopamidol in unilateral ureteral obstruction (UUO). Iopamidol-enhanced dynamic CEST MRI demonstrated altered pH distributions, kidney-specific renogram curves, and significant voxel-wise shifts toward UPD model preference in obstructed kidneys, indicating spatially heterogeneous impairment of renal probe handling.
Author
Zinia Mohanta
Johns Hopkins University School of Medicine