MRI TECHNIQUES

Blood-oxygen-level-dependent MRI

Blood-oxygen-level-dependent (BOLD) MRI assesses tissue oxygenation. In a prospective study of 112 CKD patients, elevated cortical _R_₂* (indicating low oxygenation) independently predicted a three-fold higher risk of progression to end-stage outcomes over ∼3 years [30] . An outer-cortex _R_₂* above the 90th percentile likewise conferred a ∼3× risk of renal replacement therapy or ≥30% creatinine increase in hypertensive CKD, even after adjustment for baseline eGFR, proteinuria, and other factors [30]. Advanced BOLD methods confirm true tissue hypoxia: in stage 3–4 CKD, uncorrected _R_₂* values no longer correlated with GFR, whereas iron oxide-corrected BOLD imaging clearly showed cortical and medullary hypoxia. Furthermore, acute changes in cortical oxygenation can be captured. In albuminuric diabetic patients, a single 50 mg dose of the SGLT2 inhibitor dapagliflozin lowered cortical _R_₂* by ∼9% within 6 h [31], reflecting improved oxygenation that would be undetectable by standard lab measures. Although BOLD MRI requires additional scanner time and technical expertise, it is a non-contrast technique that can be incorporated into multiparametric MRI protocols and provides valuable physiologic information for both risk assessment and treatment monitoring. While its thresholds are relative and vary with magnetic field strength, higher cortical _R_₂* values consistently correspond to greater disease risk. Across CKD and hypertensive populations, higher cortical _R_₂ values on BOLD MRI* are consistently associated with worse renal outcomes, including about a three-fold higher risk of disease progression [32, 33]. Absolute _R_₂* thresholds vary substantially due to differences in field strength, hematocrit, and correction methods, making relative comparisons and longitudinal changes more reliable than fixed cutoffs. These features support BOLD MRI as a sensitive functional biomarker rather than a static diagnostic measure.

ASL MRI

ASL MRI provides quantitative, contrast-free maps of RBF. ASL studies consistently show reduced perfusion in CKD. Cortical RBF may be nearly halved by stage 3 CKD compared to healthy kidneys, and flow correlates with eGFR [4, 5]. For example, a threshold of ∼143 ml/min/100 g distinguished CKD patients from controls with an AUROC ∼0.98 [34]. ASL measurements decline progressively with advancing CKD stage and have a rank correlation ∼0.8 with eGFR [35]. Even in patients with preserved GFR, ASL can uncover hidden deficits: in biopsy-proven early CKD (eGFR ≥ 90), cortical ASL-RBF and IVIM-derived perfusion fraction f were significantly lower than normal; perfusion fraction achieved an AUC ∼0.92 for detecting early injury [26, 36]. In diabetes, ASL-based radiomics can detect subclinical disease: one radiomic signature differentiated normoalbuminuric ss from healthy controls with AUC 0.865, and a combined imaging–clinical model predicted progression to overt nephropathy with ∼74% accuracy (AUC 0.734) [37]. Because ASL is contrast–free, quantitative, and now standardized in research protocols, it is increasingly feasible to incorporate as a cost–effective biomarker within a single multiparametric MRI exam to phenotype ischemic patterns and monitor responses without gadolinium. ASL MRI consistently demonstrates a significant decrease in cortical kidney blood flow as CKD advances, with perfusion values frequently falling to almost half by stage 3 illness and exhibiting a strong correlation with eGFR. Perfusion values of about 143 ml/min/100 g consistently distinguish patients with CKD from healthy controls, and the method has a high diagnostic accuracy [34–36]. Rather than actual biological diversity, differences between research are primarily due to technical issues.

Intravoxel incoherent motion–diffusion-weighted imaging

Diffusion MRI, including intravoxel incoherent motion (IVIM) and diffusion tensor imaging (DTI), assesses renal microstructure and microvascular flow. In a biopsy-phenotyped CKD cohort, a lower cortical IVIM perfusion fraction f—and, to a lesser extent, a reduced true diffusion coefficient D—were independently associated with progression to end-stage kidney disease over five years. Adding f and D to a model with baseline eGFR and fibrosis improved the prognostic AUC from 0.886 to 0.955, and each 1% drop in f corresponded to ∼21% higher ESKD risk [38]. Diffusion metrics also flag subtle diabetic kidney injury. In type 2 diabetes mellitus (T2DM), medullary diffusion parameters showed high diagnostic power for incipient nephropathy: the IVIM-derived true diffusion (D) achieved AUC ∼0.95 and DTI-based fractional anisotropy (FA) ∼0.91 for distinguishing early (microalbuminuric) DKD, and combining D with FA yielded near-perfect discrimination (AUC 0.99) [39]. While absolute thresholds vary by scanner and protocol, lower f and D and lower medullary FA signify worse microvascular/tubular integrity. Because diffusion sequences are fast, contrast–free, and easily appended to MRI, they provide a cost–efficient route to risk–stratify patients and identify fast progressors in targeted clinical pathways and trials. Lower IVIM perfusion fraction and diffusion measurements are consistently associated with worse outcomes and disease progression in biopsy-characterized CKD and diabetic patients. Although the precise cutoff levels vary based on the scanner and imaging methodology, the observed effects are clinically significant [38, 39]. Diffusion MRI is a helpful predictive supplement rather than a stand-alone diagnostic technique because the majority of the variability is technical rather than biological.

Fibrosis imaging (T₁ mapping)

Mapping of native (non-contrast) _T_₁ relaxation times provides a surrogate of tissue fibrosis. In a cohort of 119 CKD patients (G1–G4) followed for up to 3 years, those with the highest cortical _T_₁ values had a three-fold higher risk of requiring renal replacement or having a ≥30% rise in creatinine; adding _T_₁ to a clinical model improved the event prediction AUC from 0.83 to 0.88 [40]. Similarly, _T_₁ mapping can detect fibrotic changes in DKD. A recent study using radiomics of cortical _T_₁ maps achieved an AUC of ∼0.93 for any fibrosis and ∼0.90 for >25% fibrosis, outperforming models based on Δ_T_₁ or eGFR alone. The same _T_₁-based radiomic signature detected biopsy-proven fibrosis with 82% sensitivity and 91% specificity (AUC ∼0.93) in a training cohort, and yielded AUC 0.94 with 100% sensitivity and 86% specificity on validation [41]. These findings support native _T_₁ mapping as a noninvasive imaging biomarker of renal fibrosis and prognosis. Although absolute cutoffs differ by field strength and sequence, a higher–than–expected cortical _T_₁ and/or a rising _T_₁ trend over time flag significant scarring. _T_₁ mapping is quick, reproducible, and contrast–free, making it practical to integrate into multiparametric exams for risk stratification and longitudinal monitoring. Across different stages of CKD, renal fibrosis and worse clinical outcomes, such as a higher risk of creatinine advancement or the necessity for renal replacement treatment, are consistently associated with higher cortical native _T_₁ levels. Direct comparison between studies is limited since the exact _T_₁ cutoff values vary depending on the scanner and imaging sequence. Therefore, changes over time and relative increases in cortical _T_₁ are more therapeutically valuable than constant thresholds.

Proton-density fat-fraction mapping

Proton-density fat-fraction (PDFF) is a Dixon-based quantitative MRI metric that reports the fraction of fat protons relative to total (fat + water) protons and resolves cortical versus medullary steatosis with a coefficient of variation <5% [42, 43]. In T2DM, higher cortical PDFF parallels increases in BOLD R2*, consistent with hypoxia, linking lipid load to oxygenation changes [40]. Whole-kidney PDFF maps are acquired in ∼3–4 min, detect fat fractions from ∼0.4% to >50%, and are more accurate than CT for parenchymal fat quantification; PDFF can be incorporated alongside intravoxel incoherent motion–diffusion-weighted imaging (IVIM/DWI) and native T1/T2 mapping within a single multiparametric session [15, 44]. Clinically, PDFF is most useful in metabolic phenotypes (T2DM, obesity, hypertension) for early phenotyping—when albuminuria may precede a measurable change in GFR—and for longitudinal therapy monitoring within the same visit [15, 43, 44]. Among advanced implementations, iterative decomposition of water and fat with echo asymmetry and least-squares estimation—iron quantification (IDEAL-IQ) demonstrates significantly higher renal parenchymal PDFF in T2DM vs. controls (_P _< 0.001), with PDFF correlating negatively with eGFR (r = –0.437) and positively with creatinine (r = 0.421); repeatability is excellent (r = 0.81; mean difference ∼0.004%), and diagnostic performance is strong (AUC 0.857 for distinguishing T2DM from controls), supporting clinical applicability as an adjunct biomarker [45]. Reporting should include PDFF (%) separately for cortex and medulla and acquisition details (field strength/sequence); while universally accepted disease thresholds are not yet established and may be scanner-dependent, center-specific reference ranges, longitudinal trends, and concordance with complementary MRI biomarkers (e.g. BOLD R2*) are recommended for interpretation [43–45]. Renal PDFF is consistently associated with reduced kidney function and hypoxic symptoms in all metabolic types of kidney disease. Although measurements are repeatable, precise disease-specific cutoff values have not yet been established and still rely on the scanner in use. Therefore, rather than rigid diagnostic classification, PDFF appears to be more helpful for early illness characterization and long-term follow-up.

Multiparametric MRI and other advances

Many of the above MRI techniques can be combined into a single multiparametric scanning session. Modern 3 T protocols (≈45 min) now integrate phase-contrast flow mapping, ASL perfusion, BOLD _R_₂* mapping, diffusion-weighted imaging, and quantitative _T_₁ mapping [32]. Using such a multiparametric approach, a study of 38 diabetic CKD patients and 20 controls found that reductions in RBF (by phase-contrast MRI) and cortical perfusion (by ASL) most clearly distinguished DKD from healthy kidneys [32]. Furthermore, in a 151-patient CKD cohort (including 29 DKD) followed for ∼3.8 years, a panel of four MRI biomarkers (perfusion, oxygenation, diffusion, and _T_₁) was evaluated: although all markers correlated with disease, only cortical BOLD _R_₂* remained an independent predictor of subsequent eGFR decline when all four were considered together [32]. These MRI sequences require no exogenous contrast and have shown robust repeatability across visits (within-subject coefficient of variation ∼4% for _T_₁ and diffusion parameters, ∼11% for perfusion) [46]. Accordingly, quantitative MRI is increasingly feasible for routine use to stratify “fast” vs. “slow” progressors and may ultimately serve as a noninvasive surrogate for biopsy in assessing renal fibrosis [32]. Additionally, when evaluation of renal vasculature is needed, non-contrast MR angiography techniques (e.g. time‐SLIP) can reliably detect renal artery stenoses with ∼90% accuracy vs. CT angiography [47]. While mpMRI requires scanner time and specialty expertise, bundling multiple contrast–free biomarkers into one visit can streamline care pathways and reduce repeated testing, supporting cost–effectiveness in specialized centers. By integrating measurements of perfusion, oxygenation, diffusion, and fibrosis into a single scan, multiparametric MRI adds prognostic value across several cohorts. Even after controlling for other variables, oxygenation-based metrics such cortical BOLD _R_₂* frequently continue to be independently predictive, despite the fact that many of these markers are linked to the severity of the disease. Current repeatability results indicate mpMRI as a potential method in specialized clinical contexts, despite technical complexity and restricted availability still being obstacles.