Prospective evaluation of arterial spin labeling and diffusion tensor imaging-magnetic resonance imaging for the non-invasive assessment of renal allograft dysfunction

Abstract

Background

The optimal management strategy for end-stage renal disease is renal transplantation, graft function must be monitored regularly postoperatively. This cross-sectional study aimed to explore the value of combining functional magnetic resonance imaging (MRI) parameters with laboratory parameters in assessing chronic allograft dysfunction (CAD), and to compare whether a combined approach was superior to single-parameter indicators.

Methods

A total of 86 subjects were enrolled in the study, of whom, 20 had stable renal function, and 66 had biopsy-confirmed CAD. Imaging was performed on a 1.5-T MRI system using T2-weighted imaging, arterial spin labeling (ASL), and diffusion tensor imaging (DTI). The serum creatinine, estimated glomerular filtration rate (eGFR), 24-hour urinary protein (24hUP), renal blood flow (RBF), and fractional anisotropy (FA) values of the subjects were measured. Correlation analysis was applied to assess MRI parameters’ association with eGFR, while receiver operating characteristic (ROC) curves were used to evaluate the diagnostic efficacy of fMRI parameters and clinical parameters for CAD.

Results

The subjects were categorized into CAD groups based on their eGFR levels. The control group had higher renal RBF [277.69±67.17 vs. 138.60 (99.54–193.51)] and FA values [cortex: 0.16 (0.14–0.16) vs. 0.13 (0.11–0.16); medulla: 0.32±0.06 vs. 0.24 (0.20–0.29)] than the CAD group (P<0.01). Cortical RBF decreased progressively across the CAD subgroups [group 1 (mild: 213.33±67.07) > group 2 (moderate: 151.14±53.21) > group 3 (severe: 92.89±35.62); all P<0.05]. Similarly, there was a gradual decrease in medullary FA across the CAD subgroups [group 1: 0.29±0.04; group 2: 0.24 (0.19–0.29); group 3: 0.20±0.06]. However, no statistically significant difference was found in medullary FA between groups 2 and 3 (P=0.102). The correlation analysis showed that cortical RBF and medullary FA were positively correlated with the eGFR in the CAD group (r=0.604, P<0.001; r=0.574, P<0.001). The combined RBF, medullary FA, 24hUP, and eGFR model (RBF-FA-24hUP-eGFR) had an area under the curve (AUC) of 0.95 [95% confidence interval (CI): 0.91–1.00], which was significantly better than the AUCs of the single indicators of 24hUP and medullary FA (AUC =0.78, 95% CI: 0.68–0.88; AUC =0.79, 95% CI: 0.69–0.89, P<0.05). Further, the combined RBF, medullary FA, and, 24hUP model (RBF-FA-24hUP) was significantly superior to single 24hUP in differentiating among the subgroups (all P<0.05). In the CAD subgroups, while the performance of RBF on its own was close to that of the RBF-FA-24hUP model, the AUC of the combined model showed an increasing trend compared with RBF. Notably, the RBF-FA-24hUP model (AUC =0.86, 95% CI: 0.76–0.97; P<0.001) also surpassed medullary FA alone (AUC =0.69, 95% CI: 0.54–0.85; P=0.023) in distinguishing between the subjects in group 2 and group 3 (P<0.05).

Conclusions

In this study, two multiparametric MRI models (RBF-FA-24hUP-eGFR and RBF-FA-24hUP) were developed and shown to be superior to non-invasive CAD assessment tools. These models outperformed conventional single-parameter methods in diagnosis and moderate-to-severe subgroup stratification. To a certain extent, these models could prevent unnecessary puncture biopsies, and reduce the occurrence of complications such as bleeding and infection. RBF in particular and FA showed utility as non-invasive biomarkers for CAD and risk stratification.

Introduction

Kidney transplantation is recommended for all chronic kidney disease (CKD) G4–G5 patients expected to progress to end-stage kidney disease (1,2). However, chronic allograft dysfunction (CAD) worsens the long-term prognosis of renal transplant patients. Recent research on the pathogenesis of CAD has revealed that tubular atrophy and interstitial fibrosis contribute to kidney function decline, which ultimately leads to irreversible scarring (3). Therefore, if the disease is not diagnosed and treated promptly, it will lead to transplanted kidney failure (4).

Serum creatinine and proteinuria are widely used to evaluate transplanted kidney function. Nevertheless, they exhibit limited sensitivity and can be influenced by other factors (5,6). The gold standard for diagnosing CAD is percutaneous puncture biopsy. However, it is limited by sampling limitations, invasiveness, sample bias, and the renal medulla’s high sensitivity to hypoxia, which prevents routine sampling (7-9). Therefore, to improve outcomes in renal transplant patients, a comprehensive non-invasive assessment urgently needs to be established.

Diffusion-weighted imaging (DWI) is a promising functional magnetic resonance imaging (MRI) technique for assessing transplanted kidney function (10). However, the direction of molecular diffusion varies in tissues. Conventional single-exponential DWI cannot reflect the directionality of water molecule movement or the actual diffusion process (11,12). Diffusion tensor imaging (DTI) compensates for the shortcomings of DWI in this regard. It is an imaging tool capable of characterizing the microstructural environment of renal tissues (13). DTI quantifies water molecule diffusion in at least six directions and reveals anisotropic information about water diffusion in the kidneys. DTI is more sensitive in detecting microstructures than global diffusion (14). Fractional anisotropy (FA) is a common quantitative measure (15,16). DTI continues to be extensively applied to the central nervous system of the brain; however, in recent years, it has also shown promising potential in renal research (17-20).

Conversely, arterial spin labeling (ASL) labels blood flow into the kidneys, enabling real-time renal perfusion monitoring. By magnetically tagging water protons in arterial blood as intrinsic tracers, ASL offers a non-invasive method for functional assessment (21). Multiple studies have shown its value in evaluating blood flow in kidneys (22-25). Notably, a recent study found that ASL shows promise in assessing the subclinical pathological status of patients with stable transplanted kidney function (5). Its high sensitivity enables the early clinical intervention and identification of the subclinical pathological status of patients. Both techniques can analyze CAD in transplanted kidneys from perfusion and diffusion perspectives. Moreover, they are suitable for repeated and longitudinal assessments. As a single technique can be affected by clinical and physiological factors, combining both could reduce interference and provide more accurate results.

This study explored the value of combining functional MRI and laboratory indicators in assessing CAD and evaluated their potential as non-invasive biomarkers. We present this article in accordance with the STROBE reporting checklist (available at https://qims.amegroups.com/article/view/10.21037/qims-2025-604/rc).

Methods

The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. The study was approved by the Institutional Ethics Committee of The First Affiliated Hospital of Soochow University (approval No. 2022-412), and informed consent was obtained from all the patients.

Study population

This was a prospective cross-sectional single-center study of renal transplant patients. Informed consent was obtained from all participants. From February 2022 to October 2024, a total of 86 patients were enrolled in the study, of whom, 20 had stable renal function and 66 had biopsy-confirmed CAD (Figure 1). The diagnostic workflow is described below.

Figure 1.

Workflow of patient selection. CAD, chronic allograft dysfunction; MRI, magnetic resonance imaging.

Patients were included in the CAD group if they had been diagnosed with CAD by biopsy. The histological features of CAD are interstitial fibrosis and tubular atrophy (3). Based on Kidney Disease Improving Global Outcomes CKD grading criteria, and the needle biopsy gold standard, the CAD patients were further subdivided into the following groups (26,27): group 1 [mild impairment; estimated glomerular filtration rate (eGFR) >60 mL/min/1.73 m², n=16], group 2 (moderate impairment; eGFR 30–60 mL/min/1.73 m², n=31), and group 3 (severe impairment; eGFR 30 mL/min/1.73 m², n=19).

Patients were included in the control group if they met the following inclusion criteria: had received a first-time renal transplant, had stable function (i.e., eGFR >60 mL/min/1.73 m², and ≤15% creatinine variation in three consecutive measurements within one year), and had no clinical/biochemical signs of CAD. No protocol biopsy was performed due to the absence of clinical indication.

Patients were excluded from the study if they met any of the following exclusion criteria: (I) refused to participate or showed physical intolerance; (II) had contraindications to MRI; (III) were pregnant or aged <18 years; (IV) had a severe hemorrhage in the transplanted kidney, giant cyst, or obvious atrophy; and/or (V) had poor-quality MRI images.

The clinical and laboratory data of the patients were obtained within 48 hours before and after the MRI scan (Table 1). The CKD Epidemiology Collaboration formula (CKD-EPI 2009) was used to determine the eGFR (28).

Table 1. Comparison of the clinical data between the control and CAD groups, and between the CAD subgroups.

Characteristics	Control group (N=20)	CAD group (total) (N=66)	CAD1 group (N=16)	CAD2 group (N=31)	CAD3 group (N=19)	P value*	P value**
Age (years)	38.50 [30.50–49.00]	42.44±9.52	41.19±10.52	42.45±9.27	43.47±9.47	0.294	0.784
Gender (male/female), n	11/9	49/17	13/3	22/9	14/5	0.101	1.000
BMI (kg/m2)	21.73±3.85	23.25±3.79	24.65±3.64	23.00±3.34	22.47±4.45	0.121	0.212
MAP (mmHg)	98.65±15.69	102.85±14.24	98.63±14.93	101 [94–114]	105.63±13.90	0.262	0.315
eGFR (mL/min/1.73 m2)	77.10±9.36	43.24±21.47	69.25 [64.18–74.88]†	42.12±8.05‡	21.5 [15.1–24.6]¶	<0.001	<0.001
Serum creatinine (μmol/L)	95.47±14.74	161.10 [124.10–231.45]	105.26±16.40†	163.21±31.58‡	290.7 [237.5–376.1] ¶	<0.001	<0.001
Hemoglobin level (g/L)	115.7±19.43	119.06±24.36	140.75±16.72†	120.55±20.18‡	98.37±18.91¶	0.519	<0.001
24h urine protein (g/24 h)	0.25±0.21	0.58 [0.30–1.98]	0.36 [0.14–1.00]	0.58 [0.18–2.64]	1.47±1.11¶	<0.001	0.031
Transplanted time (months)	1 [1–12]	51 [33–80]	37 [21–66]	54 [35–73]	67 [28–140]	<0.001	0.194
The time interval between biopsy and MRI examination	–	1 [0–2]	1 [0.25–2.75]	1 [1–2]	0 [0–1]	–	0.068
RBF (mL/100 g/min)	277.69±67.17	138.60 [99.54–193.51]	213.33±67.07†	151.14±53.21‡	92.89±35.62¶	<0.001	<0.001
Cortical FA	0.16 [0.14–0.16]	0.13 [0.11–0.16]	0.14±0.03	0.13±0.03	0.13 [0.12–0.18]	0.007	0.657
Medullary FA	0.32±0.06	0.24 [0.20–0.29]	0.29±0.04†	0.24 [0.19–0.29]	0.20±0.06¶	<0.001	<0.001

Quantitative data with a normal distribution are expressed as the mean ± standard deviation, and those with a skewed distribution are expressed as the median [Q1, Q3] unless otherwise specified. *, a comparison between CAD group and control group; **, a comparison between the three CAD groups; ^†, group 1 compared with group 2 (P<0.05); ^‡, group 2 compared with group 3 (P<0.05); ^¶, group 1 compared with group 3 (P<0.05). BMI, body mass index; CAD, chronic allograft disease; eGFR, estimated glomerular filtration rate; FA, fractional anisotropy; MAP, mean arterial pressure; MRI, magnetic resonance imaging; RBF, renal blood flow.

Scan protocol and image acquisition

The patients were positioned supine in feet-first orientation, with the scan centered on the transplanted kidney. A Philips 1.5T MRI scanner (Ingenia Ambition, Philips Healthcare, Best, the Netherlands) with a 28-channel phased array coil was used. The scanning range centered on the transplanted kidney, up to the lower edge of the costal arch, and down to the lower edge of the pubic symphysis. Before the scan, an abdominal compression bandage was applied to fix the area of the transplanted kidney, and a gravity sandbag was placed. Physical immobilization was adopted to reduce the respiratory motion artifacts. The patients were required to refrain from food and water for 4 hours before the scan. During the scanning process, the patients were instructed to engage in free breathing without respiratory triggering. The parameters for each sequence are listed in Table 2.

Table 2. MRI sequences and corresponding parameters.

Parameters	T2-weighted imaging	T2-weighted imaging	ASL	DTI
Orientation	Coronal	Transverse	Transverse	Coronal
TR/TE (msec)	1,086/80	1,300/257	3,963/15	2,300/67
Respiratory control	Free breath	Free breath	Free breath	Free breath
Voxel size (mm3)	1.45×1.45×5	0.90×0.91×1.80	3.75×3.75×8.00	2.50×2.53×5.00
FOV (mm2/mm3)	400×400	350×350	240×240×88	320×400
Number of slices	30	222	11	20
b-factor (s/mm2)	–	–	–	0 and 600
Total scan duration (min)	3	4	5	3
TSE factor	35	136	20	–
Diffusion directions	–	–	–	15

ASL, arterial spin labeling; DTI, diffusion tensor imaging; FOV, field of view; MRI, magnetic resonance imaging; TE, echo time; TR, repetition time; TSE, turbo spin echo.

Image processing

The post-processing of ASL was performed using software embedded in the scanning host. Renal blood flow (RBF) was quantified using the model proposed by Alsop et al. (29), and the calculation formula is expressed as follows:

$$\mathrm{RBF}=\frac{6000\cdot \lambda \cdot \left(S{I}_{\mathrm{control}}-S{I}_{\mathrm{label}}\right)\cdot e^{\frac{\mathrm{PLD}}{T_{1,\mathrm{blood}}}}}{2\cdot \alpha \cdot T_{1,\mathrm{blood}}\cdot S{I}_{\mathrm{PD}}\cdot \left(1-e^{\frac{\tau }{T_{1,\mathrm{blood}}}}\right)},\text{\hspace{0.17em}mL}/100\text{\hspace{0.17em}g}/\min$$ [1]

where λ represents the brain/blood partition coefficient in mL/g; SI_control and SI_label represent the time-averaged signal intensities in the control and label images, respectively; T_{1, blood} represents the longitudinal relaxation time of blood in seconds; α represents the labeling efficiency; SI_PD represents the signal intensity of a proton density weighted image; τ represents the label duration, and the post-labeling delay time is the time between the end of this pulse train and image acquisition.

The post-processing of DTI was performed on the Intellispace Portal workstation (Philips Healthcare, Best, the Netherlands). The FA map was calculated from the eigenvalues of the diffusion tensor using the following formula:

$$\mathrm{FA}=\sqrt{\frac{2}{3}}\cdot \frac{\sqrt{{\left({\lambda }_1-{\lambda }_{\mathrm{mean}}\right)}^2+{\left({\lambda }_2-{\lambda }_{\mathrm{mean}}\right)}^2+{\left({\lambda }_3-{\lambda }_{\mathrm{mean}}\right)}^2}}{\sqrt{{\lambda }_1^2+{\lambda }_2^2+{\lambda }_3^2}}$$ [2]

where λ_1, λ₂, and λ₃ are the three eigenvalues of the diffusion tensor, and ${\lambda }_{\mathrm{mean}}=\frac{{\lambda }_1+{\lambda }_2+{\lambda }_3}{3}$ is the mean diffusivity.

In outlining the regions of interest (ROI), areas of vascular aggregation, cysts, fluid accumulation, and areas of obvious artifacts were avoided. Two senior abdominal radiologists, who were blinded to the clinical data, independently evaluated all the images and manually drew the ROIs. For the DTI image processing, the researchers selected coronal T2-weighted imaging (T2WI), and divided the transplanted kidney into upper, middle, and lower parts by centering on the renal hilum (including the renal artery, vein, and collecting system). They manually drew three elliptical ROIs in the cortex and medulla respectively, positioned them at the upper, middle, and lower poles of the kidney, and copied them onto the corresponding FA maps. The area of each ROI was approximately 5–25 mm². The weighted average of the FA values was calculated (Figure 2 and Figure S1).

Figure 2.

Anatomical, ASL, and DTI images of the CAD subgroups and the control group. (A,E,I,M) Coronal T2-weighted images showing clear corticomedullary demarcation, which served as an anatomical reference. (B,F,J,N) DTI-FA maps showing corticomedullary structures with higher FA values in the medulla than the cortex. (C,G,K,O) Transverse T2-weighted images used for anatomical localization. (D,H,L,P) ASL maps. Panel (P) shows that the cortical RBF of patients with good function is clearly displayed, while the higher the CAD grade, the sparser the display of cortical RBF. (A-D) Images of a 27-year-old male from the CAD1 group with an eGFR of 62.4 mL/min/1.73 m²; (E-H) Images of a 50-year-old male from the CAD2 group with an eGFR of 39.8 mL/min/1.73 m². (I-L) Images of a 43-year-old female from the CAD3 group with an eGFR of 28.7 mL/min/1.73 m². (M-P) Images of a 30-year-old female from the control group with an eGFR of 101.2 mL/min/1.73 m². ASL, arterial spin labeling; CAD, chronic allograft dysfunction; DTI, diffusion tensor imaging; eGFR, estimated glomerular filtration rate; FA, fractional anisotropy; RBF, renal blood flow.

For the ASL image processing, the ASL was adjusted to a pseudo-color map to visualize the blood flow; black indicated low flow, and yellow indicated high flow. Using axial T2WI as the anatomical reference, images at the level of the maximal renal transverse diameter were selected. Cortical contours were manually delineated and subsequently copied onto the corresponding ASL images (Figure 2 and Figure S1).

Data analysis

SPSS 26.0 (IBM, Armonk, NY, USA) and GraphPad Prism 9.5 (GraphPad Software, LLC, San Diego, CA, USA) were used for the statistical analysis and data visualization, respectively. The Kolmogorov-Smirnov test was employed to assess data normality. The quantitative data conforming to a normal distribution are expressed as the mean ± standard deviation, the quantitative data with a skewed distribution are expressed as the median (Q1, Q3), and the categorical data are expressed as the ratio. The inter-observer and intra-observer reproducibility of the MRI measurements was evaluated using the intraclass correlation coefficient (ICC). A one-way analysis of variance was used for between-group comparisons of the normally distributed measures, and the Bonferroni method was used for two-by-two comparisons. The non-normally distributed data were analyzed using the Kruskal-Wallis H test, and post-hoc comparisons were conducted using the Mann-Whitney U test. For the categorical variables, the Chi-squared test or Fisher’s exact test was applied. Spearman and Pearson correlation coefficients were used to analyze the relationship between the MRI parameters and eGFR. A traditional receiver operating characteristic (ROC) curve analysis (non-machine learning approach) and binary logistic regression were used to evaluate the diagnostic efficacy of the parameters for the CAD group and its subgroups of patients with transplanted kidneys, and area under the curve (AUC) differences were compared using the DeLong test in SPSS 26.0 (IBM, Armonk, NY, USA). The power analysis (α=0.05) using PASS 15 (NCSS, LLC, Kaysville, UT, USA) demonstrated a power >0.99 for cortical RBF and medullary FA, and a power of 0.39 for cortical FA. A P value <0.05 indicated a statistically significant difference.

Pathologic analysis of transplanted kidneys

Ultrasound-guided percutaneous puncture biopsy was completed in the CAD patients within one week before or after the MRI examination. Electron microscopy techniques and section staining (including hematoxylin-eosin, periodic acid-Schiff, periodic acid-silver-methylamine, and Masson’s trichrome staining) were used to comprehensively analyze the samples at different histological levels. The pathologic analyses and diagnoses were conducted by a renal pathologist with 10 years of experience in diagnostic renal pathology, who was blinded to the MRI results (Figure 3).

Figure 3.

Histological examination of transplanted kidney results. (A,D,G) Results for a male from the CAD1 group 3 years post-transplantation. (B,E,H) Results for a male from the CAD2 group 2 years post-transplantation. (C,F,I) Results for a female from the CAD3 group 2 years post-transplantation. (A-C) PASM staining, 400×; (D-F) PAS staining, 400×; (G-I) Toluidine blue staining, G: ×6,000, H: ×2,000, I: ×2,500. CAD, chronic allograft dysfunction; PAS, periodic acid-Schiff; PASM, periodic acid-silver-methylamine.

Results

Comparison of the clinical and laboratory indices of the renal transplant patients in each group

A total of 86 patients were included in the study. The control group comprised 20 patients (11 males and 9 females), with an age range of 23–51 years and an average age of 38.50 (30.50–49.00) years. The CAD group comprised 66 patients (49 males and 17 females), with an age range of 24–67 years and an average age of 42.44±9.52 years. No statistically significant differences were found between the control and CAD groups in terms of age, gender composition, body mass index (BMI), mean arterial pressure (MAP), or hemoglobin levels (P>0.05). However, as expected, there were statistically significant differences between the control and CAD groups in terms of serum creatinine, the eGFR, and 24-hour urinary protein (24hUP) (P<0.05). The results of further pairwise comparisons showed a gradual decrease in the eGFR and hemoglobin levels, and a gradual increase in serum creatinine across the CAD1, CAD2, and CAD3 groups. Although 24hUP levels showed increased from CAD1 to CAD3, a statistically significant difference was only observed between the CAD1 and CAD3 groups (Table 1).

Comparison of the renal cortical RBF and FA of the renal transplant patients between the control and CAD groups, as well as among the CAD subgroups

There was good agreement between the inter- and intra-observer reproducibility of cortical RBF, cortical FA, and medullary FA, as assessed by the ICCs (Tables S1,S2).

Comparison of the renal cortical RBF of the renal transplant patients between the control and CAD groups, as well as among the CAD subgroups

As Figure 4A shows, the control group (277.69±67.17 mL/min/1.73 m²) had a higher value than the CAD group [138.60 (99.54–193.51) mL/min/1.73 m²] (P<0.001). Among the CAD subgroups, the CAD1 group (213.33±67.07 mL/min/1.73 m²) had a higher renal cortical RBF value than the CAD2 group (151.14±53.21 mL/min/1.73 m²) and the CAD3 group (92.89±35.62 mL/min/1.73 m²) (P<0.001). Post-hoc two-by-two comparisons showed that the differences between the groups were statistically significant.

Figure 4.

The comparison of medullary FA and cortical RBF in the CAD subgroups. (A) The medullary FA values decreased across the CAD1–CAD3 groups (no significant difference was found between CAD2 and CAD3). (B) Cortical RBF was significantly reduced in the CAD subgroups (all P<0.01). *, P<0.05; ***, P<0.001. CAD, chronic allograft dysfunction; FA, fractional anisotropy; ns, not significant; RBF, renal blood flow.

Comparison of the renal medullary FA and cortical FA of the renal transplant patients between the control and CAD groups, as well as among the CAD subgroups

As Figure 4B shows, the control group [0.16 (0.14–0.16)] had a higher cortical FA value than the CAD group [0.13 (0.11–0.16)] (P<0.01). However, no such statistically significant difference was found between the CAD subgroups. In terms of medullary FA, the difference between the control group (0.32±0.06) and the CAD group [0.24 (0.20–0.29)] was statistically significant. The medullary FA value was significantly higher in the CAD1 group (0.29±0.04) than both the CAD2 [0.24 (0.19–0.29)] and CAD3 (0.20±0.06) groups. The post-hoc analysis revealed that while the medullary FA value was decreased in the CAD3 group compared to the CAD2 group, this difference was not statistically significant (P=0.102).

Correlation analysis of each index in the CAD groups

The results of the correlation analysis showed that cortical RBF was positively correlated with the eGFR (r=0.604, P<0.001). Additionally, the medullary FA was positively correlated with the eGFR (r=0.574, P<0.001) (Figure 5). However, the cortical FA showed no significant correlation with the eGFR.

Figure 5.

Correlation analysis of MRI parameters with the eGFR and serum creatinine in CAD patients. (A) Transplanted renal cortical RBF was moderately positively correlated with the eGFR in the CAD groups (P<0.001). (B) Transplanted renal medullary FA was moderately positively correlated with the eGFR in the CAD groups (P<0.001). CAD, chronic allograft dysfunction; eGFR, estimated glomerular filtration rate; FA, fractional anisotropy; MRI, magnetic resonance imaging; RBF, renal blood flow.

Comparison of the efficacy of the combined model and individual indices in distinguishing between the patients in the control and CAD groups, as well as the patients in the CAD subgroups

The ROC curve analysis showed that RBF, medullary FA, 24hUP, and the eGFR could be used to effectively differentiate between patients in the control and CAD groups. The AUC values of each index are shown in Table 3. An AUC of 0.95 [95% confidence interval (CI): 0.91–1.00] was achieved by the combined RBF, medullary FA, 24hUP, and eGFR model (RBF-FA-24hUP-eGFR), which significantly outperformed the individual 24hUP (AUC 0.78, 95% CI: 0.68–0.88) and medullary FA (AUC 0.79, 95% CI: 0.69–0.89) (both P<0.05). Although the AUC of the RBF-FA-24hUP-eGFR model was higher than that of individual RBF (AUC 0.90, 95% CI: 0.83–0.97) or eGFR (AUC 0.91, 95% CI: 0.85–0.97), the difference was not statistically significant (P>0.05).

Table 3. The AUC values of the RBF-FA-24hUP-eGFR model, the RBF-FA-24hUP model, RBF, medullary FA, 24hUP, and eGFR in distinguishing between patients in the control and CAD groups, as well as among patients in the CAD subgroups.

Parameters	AUC (95% CI)	P
Control group vs. CAD group
RBF	0.90 (0.83, 0.97)	<0.001
Medullary FA	0.79 (0.69, 0.89)	<0.001
24hUP	0.78 (0.68, 0.88)	<0.001
eGFR	0.91 (0.85, 0.97)	<0.001
RBF-FA-24hUP-eGFR model	0.95 (0.91, 1.00)	<0.001
CAD1 vs. CAD2
RBF	0.75 (0.61, 0.90)	0.005
Medullary FA	0.74 (0.60, 0.88)	0.008
24hUP	0.65 (0.49, 0.80)	0.101
RBF-FA-24hUP model	0.81 (0.69, 0.94)	0.001
CAD1 vs. CAD3
RBF	0.96 (0.91, 1.00)	<0.001
Medullary FA	0.88 (0.75, 1.00)	<0.001
24hUP	0.78 (0.63, 0.94)	0.004
RBF-FA-24hUP model	0.99 (0.97, 1.00)	<0.001
CAD2 vs. CAD3
RBF	0.83 (0.71, 0.94)	<0.001
Medullary FA	0.69 (0.54, 0.85)	0.023
24hUP	0.58 (0.42, 0.74)	0.332
RBF-FA-24hUP model	0.86 (0.76, 0.97)	<0.001

AUC, area under the curve; CAD, chronic allograft dysfunction; CI, confidence interval; eGFR, estimated glomerular filtration rate; FA, fractional anisotropy; RBF, renal blood flow; RBF-FA-24hUP model, RBF, medullary FA, and 24hUP model; RBF-FA-24hUP-eGFR model, RBF, medullary FA, 24hUP, and eGFR model; 24hUP, 24-hour urinary protein.

In the subgroup analyses, the combined RBF, medullary FA, and 24hUP model (RBF-FA-24hUP) significantly outperformed the individual 24hUP in differentiating among all the subgroups (all P<0.05). In relation to the CAD1 vs. CAD2, the RBF-FA-24hUP model had higher AUC values 0.81 (95% CI: 0.69, 0.94) than individual RBF (AUC 0.75, 95% CI: 0.61–0.90) or medullary FA (AUC 0.74, 95% CI: 0.60–0.88). In relation to CAD 1 vs. 3, the RBF-FA-24hUP model had higher AUC values 0.99 (95% CI: 0.97–1.00) than individual RBF (AUC 0.96, 95% CI: 0.91–1.00) or medullary FA (AUC 0.88, 95% CI: 0.75–1.00). In relation to CAD 2 vs. 3, the RBF-FA-24hUP model (AUC 0.86, 95% CI: 0.76–0.97) was significantly superior to individual medullary FA (AUC 0.69, 95% CI: 0.54, 0.85) (P<0.05) (Figure 6 and Table 3).

Figure 6.

Efficacy of the RBF-FA-24hUP-eGFR model, RBF-FA-24hUP model, RBF, medullary FA, 24hUP, and eGFR in distinguishing between patients in the control and CAD groups, as well as among patients in the CAD subgroups. (A) ROC curves for the RBF-FA-24hUP-eGFR model, RBF-FA-24hUP model, RBF, medullary FA, 24hUP, and eGFR in distinguishing between patients in the control group and the CAD group. The combined RBF-FA-24hUP-eGFR model had an AUC of 0.95 (95% CI: 0.91, 1.00), outperforming the individual RBF, medullary FA, 24hUP, and the eGFR. (B-D) ROC curves for the RBF-FA-24hUP-eGFR model, RBF-FA-24hUP model, RBF, medullary FA, and 24hUP in distinguishing among the patients in the CAD subgroups. In the subgroup analyses, the combined RBF-FA-24hUP model significantly outperformed the individual RBF, medullary FA, 24hUP, and the eGFR. AUC, area under the curve; 24hUP, 24-hour urinary protein; CAD, chronic allograft dysfunction; CI, confidence interval; eGFR, estimated glomerular filtration rate; FA, fractional anisotropy; RBF, renal blood flow; RBF-FA-24hUP model, the combined RBF, medullary FA, and 24hUP model; RBF-FA-24hUP-eGFR model, the combined RBF, medullary FA, 24hUP, and eGFR model; ROC, receiver operating characteristic.

Discussion

The RBF-FA-24hUP-eGFR model (AUC =0.95) established in this study has the ability to differentiate between CAD and renal transplant patients with stable function. When patients present with a sudden elevation in serum creatinine or clinical proteinuria (>0.5–1.0 g/d), non-invasive MRI can be used to guide clinical decisions and reduce biopsy-related complications (30). Many previous studies on CAD have not included a histopathologic examination. In this prospective study using pathologic findings as the reference standard, functional MRI demonstrated higher diagnostic reliability than conventional clinical parameters in detecting CAD among renal transplant recipients. As contrast media may increase the risk of nephrogenic systemic fibrosis in CAD patients, this study employed the functional MRI technique. By avoiding the use of contrast agent, the safety and tolerability of CAD patients can be enhanced (14).

In recent years, multiple studies have shown that ASL has significant diagnostic value in kidney diseases through non-invasive renal perfusion measurements (5,31,32). Similarly, this study showed the significant advantages of ASL in assessing transplanted kidney function and identifying early CAD progression. Notably, compared to well-functioning renal cortical RBF, medullary RBF has a slower flow rate and significantly lower blood flow, measuring approximately one-tenth of that of cortical blood flow (5,33). It is challenging to accurately capture and quantify this low perfusion state using MRI, which contributes to the poor reproducibility of medullary RBF measurements. Thus, this study focused on the cortical RBF.

Similar to the findings of the present study, Artz et al. (34) found a good positive correlation between cortical perfusion and the eGFR in both autologous and transplanted kidneys. Additionally, consistent with previous research (22,31,35), the present study showed that the RBF value was significantly lower in patients with renal dysfunction than those with well-functioning transplanted kidneys. This may largely be due to the loss of peritubular capillaries, damage to renal endothelial cells, and endothelial arteritis during deterioration of renal function, all of which lead to a decrease in microcirculatory perfusion indexes (22,36,37). Our results showed that the RBF model had high diagnostic efficacy in differentiating between the CAD and renal transplant patients with stable function (AUC =0.90, 95% CI: 0.83–0.97). When the optimal RBF threshold value of 210.06 mL/100 g/min was used, the sensitivity and specificity of the model reached 0.83 and 0.85, respectively. This finding suggests that the model could serve as a routine clinical screening tool, thereby significantly reducing the need for invasive biopsy. Surveillance intervals can be dynamically adjusted according to measurements. CAD was recognized early and protective interventions were implemented, which ultimately improved the survival of transplanted kidneys. This study showed that ASL also had a good ability to discriminate between different CAD stages, and had a specificity of 100% in differentiating between mild and severe CAD. However, the ability of ASL to differentiate between mild and moderate CAD was limited, and while it had a sensitivity of 90%, its specificity was suboptimal. We hypothesize that the microcirculatory blood flow changes in the kidneys of patients with mild-to-moderate CAD are small. Overall, our findings support the use of ASL as a potential biomarker for severe CAD.

DTI, as a promising MRI technique for renal function assessment, has proven effective in the non-invasive evaluation of renal pathologies, including chronic glomerulonephritis (38), clear cell carcinoma (39), and diabetic nephropathy (40). FA was selected over apparent diffusion coefficient (ADC) as the focus of the study due to the unique anatomical structure of the renal medulla, which makes FA more sensitive than ADC in detecting pathological changes in transplanted kidneys (14). This study showed that the medullary FA values were consistently higher than the cortical FA values in both patients with stable renal function and those with CAD. These findings align with the observations of Kido et al., who reported elevated medullary FA values at 1.5T (0.42±0.05 vs. 0.14±0.03) and 3.0T (0.49±0.04 vs. 0.15±0.03) compared to cortical measurements (41). This is because the cortex has randomly arranged structures and the tubular structures in the medulla are arranged radially (16,42-44). The analysis showed a gradual decrease in medullary FA values across renal transplant patient groups: stable function, CAD1, CAD2, and CAD3, while cortical FA showed no significant differences among these groups. These findings are primarily attributed to the radial arrangement of the renal medullary tubules and microstructural alterations of the vasculature (13,16).

DTI demonstrated high diagnostic performance in discriminating between patients in the control and CAD groups. A further analysis showed that there was a large gap in the diagnostic efficacy of DTI across different CAD stages, and the overall effect was lower than that of ASL. Water molecule diffusion can be affected by intrinsic renal physiological factors (45). Additionally, FA values are affected by respiratory motion artifacts, renal vascular fluctuations, and MRI parameter settings (18,44).

Breath-triggered scanning is not currently used, as it significantly prolongs the scanning time and may increase patient discomfort, as well as the risk of motion artifacts. Before scanning, we immobilized the transplanted kidney region using abdominal compression bandages, and then placed gravity-based sandbags on the abdomen to minimize respiratory motion artifacts. The DTI protocol could be further optimized in the future (e.g., by adding respiratory gating to reduce respiratory artifacts and integrating cardiac gating to monitor vascular pulsations). Specifically, DTI showed high diagnostic accuracy in discriminating between mild and severe CAD, which is attributable to the significant microstructural differences in the renal tissues between these groups. DTI was able to effectively detect obvious changes in fibrosis and tubular atrophy. Conversely, the AUC was lower in distinguishing moderate to severe CAD. Presumably, CAD involves complex pathological changes such as renal medullary hypoxia, a loss of peritubular capillaries, and glomerulosclerosis. Consequently, the single FA index could not accurately reflect this dynamic change.

In addition, Notohamiprodjo et al. (46) showed that renal blood vessels also travel in a radial direction and have a close relationship with renal tubules; thus, microcirculatory perfusion also has a certain effect on the anisotropy of diffusion. Intravoxel incoherent motion diffusion (IVIM) established by biexponential modeling effectively addresses this limitation. IVIM that uses multiple b values can identify molecular diffusion and blood flow information thus reflecting more realistic physiopathological changes in tissues (16,47). Additionally, chronic hypoxia is an important factor affecting renal fibrosis, and blood oxygen level dependent (BOLD) MRI can effectively assess oxygenation levels (48). In a recent multiparametric MRI study, Bane et al. demonstrated that a multiparametric model combining IVIM, BOLD, DTI, and T1 mapping performed well in evaluating transplanted kidneys (49). In the future, we intend to add BOLD, IVIM, and other MRI techniques to our model to reduce confounding factors, and enable comprehensive multiparametric MRI evaluation. This will be an important area of our future research.

Limitations

This study showed that ASL and DTI have great potential in the assessment of CAD; however, this study still had a number of limitations. First, while this was a prospective study, its single-center design resulted in a limited sample size. The limited statistical validity of cortical FA may also be attributable to the limited sample size of our study. Therefore, future multicenter studies with larger cohorts are needed to expand the size of the stable cohort, validate differences between CAD subgroups, and confirm the external validity of our findings. Second, the current DTI sequence parameter settings used in this study were relatively basic, and did not achieve isotropic resolution or integrate techniques such as respiratory gating. In future studies, we intend to further optimize the DTI parameters (including isotropic voxels and respiratory triggering) to optimize the balance between the resolution and scan time. Third, no puncture biopsy was performed in the control group of this study, as this group of patients did not have clinical indications for puncture, and unnecessary puncture would have had a negative effect on the transplanted kidneys of these patients.

Conclusions

This study developed two multiparametric models (RBF-FA-24hUP-eGFR and RBF-FA-24hUP) that were shown to be superior non-invasive CAD assessment tools, outperforming conventional single-parameter methods in diagnosis and subgroup stratification. To a certain extent, the use of these models could prevent unnecessary puncture biopsies and reduce the occurrence of complications such as bleeding and infection. RBF in particular and FA showed utility as non-invasive biomarkers for CAD and risk stratification.

Supplementary

The article’s supplementary files as

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. The study was approved by the Institutional Ethics Committee of The First Affiliated Hospital of Soochow University (approval No. 2022-412), and informed consent was obtained from all the patients.

Data Sharing Statement

Available at https://qims.amegroups.com/article/view/10.21037/qims-2025-604/dss

qims-15-08-6882-dss.pdf ^{(66.6KB, pdf)}

DOI: 10.21037/qims-2025-604