Alterations of Renal Function in Patients with Diabetic Kidney Disease: A BOLD and DTI Study

Objectives Our study aims to determine the patterns of renal oxygenation changes and microstructural changes by BOLD and DTI with deteriorating kidney function in patients with diabetic kidney disease (DKD). Methods Seventy-two patients with type 2 diabetes mellitus (DM) and twenty healthy controls (HCs) underwent laboratory examinations, and renal BOLD and DTI images were obtained on a 3T-MRI machine. R2∗, fractional anisotropy (FA), and average diffusion coefficient (ADC) values were evaluated. DM patients were divided into three subgroups (Group-DI/DII/DIII, based on urinary albumin-creatinine ratio (UACR)) and a nondiabetic kidney disease group (Group-NDKD). D-value and MCR of R2∗ and FA were proposed to evaluate the differentiation between medulla and cortex of the individual kidney among HCs and three subgroups for reducing individual differences. Comparisons were made between NDKD and kidney function-matched DKD patients. Correlations between MRI parameters and renal clinical indices were analyzed. Results Compared with Group-HC/DI, medullary R2∗ and FA values were significantly different in Group-DII/III. The D-value of R2∗ and FA in Group-III were significantly smaller than that in Group-HC. However, only MCR of R2∗ in Group-III was significantly smaller than that in HCs. Medullary R2∗ and FA were negatively associated with serum creatinine (SCr) and cystatin C (Cys C) and positively associated with eGFR. Conclusions With renal function declining, BOLD and DTI could capture alterations including the first rising and then falling medullary R2∗, continuously declining medullary FA, and apparent cortex-medullary differentiation in DKD patients. The MRI parameters showed renal changes accompanied by varying degrees of albuminuria, sharing common involvement in DKD and NDKD patients, but it was hard to distinguish between them. BOLD seemed more sensitive than DTI in identifying renal cortex-medullary differentiation.


Introduction
Diabetes usually results in complications, affecting the nerves, eyes, kidneys, and cardiovascular system, which leads to decreased quality of life and a greater risk of death than similarly aged healthy people [1]. DKD has been reported to have an incidence of 20% ∼ 40% in diabetic patients in European and American countries [2,3]. In 2011, the prevalence of DKD in hospitalized patients in China began to exceed that of glomerular nephritis-related chronic kidney disease and the gap has been increasing [4,5].
e American Diabetes Association (ADA) and Kidney Disease: Improving Global Outcomes (KDIGO) group suggest that the detection of the urinary albumin/creatinine ratio (UACR) and eGFR should be appropriately frequent to better diagnose and treat DKD [6,7]. e clinical symptoms of early DKD are not obvious. Patients may have dry mouth, anorexia, stomach upset, fatigue, weakness, mild eyelid edema, or even numbness, but usually do not have symptoms such as moderate edema, nausea, and vomiting. When urinary microalbumin (uAlb) or SCr is used as a basic means to predict and screen DKD, once it is abnormal, kidney function already declines significantly [8,9]. Biomarkers, including indicators of glomerular damage and tubular damage, inflammatory factors, proteomics, and even microRNAs, require in-depth clinical validation [10].
Moreover, the renal lesions of DKD cannot be visually displayed by routine ultrasound, CT, and MRI examination.
It is essential to assess kidney damage in DKD patients via a noninvasive and intuitive method. Sufficient evidence indicates that chronic hypoxia and interstitial fibrosis in the kidneys are involved in the occurrence and development of DKD [11][12][13][14].
e early pathological manifestations are increased glomerular volume, increased mesangial matrix, and thickened basement membrane, and the late stage is mostly sclerotic changes. Blood oxygenation level-dependent magnetic resonance imaging (BOLD-MRI) is the only noninvasive technology for evaluating the renal oxygen content in renal living tissue instead of invasively inserting microelectrodes into the kidneys directly [15]. In BOLD, deoxygenated hemoglobin is paramagnetic, and its content will affect the uniformity of the magnetic field in the surrounding local tissues, which is transformed into the T2 * signal. Since R2 * � 1/T2 * , anything that could reduce the concentration of deoxygenated hemoglobin decreases the R2 * value. An increase in the R2 * value corresponds to a decrease in oxygenated hemoglobin, a decrease in partial pressure of oxygen, and tissue hypoxia. e hyperperfusion of the renal medulla and its oxygen partial pressure gradient allow BOLD-MRI to detect slight fluctuations in deoxyhemoglobin in the renal medulla, making it easier to evaluate renal oxygenation. Diffusion tensor imaging (DTI) is the only noninvasive technology for evaluating water molecular diffusion in the renal living tissue, which can detect the altered structures in the medulla by abnormal diffusion of water molecules in patients with early DKD [16,17]. Diffusion refers to the random and irregular movement of molecules, which is an important physiological activity of the human body. Renal DTI mainly describes the range of diffusion movement of water molecules in unit time by ADC and describes the proportion of anisotropic components of water molecules in the whole diffusion tensor by FA to reflect the renal microstructure. e purpose of our research was to explore the relationships between BOLD and DTI alterations and the renal function decline in DKD patients and to explore whether there is a difference between the renal damage exhibited by NDKD and DKD in DM.

Participants.
is study was approved by the Ethics Committee. All participants signed an informed consent form before participating in the experiment. We recruited 80 patients from the Department of Nephrology and Endocrinology in our hospital and 20 healthy volunteers matched the sex and age criteria from December 2020 to June 2022. e patient's inclusion criteria were as follows: (1) age 18 to 80 years; (2) type II DKD, in accordance with the 2020 American Diabetes Association diagnosis, eGFR >30 ml/ min/1.73 m 2 ; (3) type II simple DM, no renal complications or biochemical abnormalities; (4) nondiabetic kidney disease (NDKD) patients (type II diabetes with CKD, e.g., hypertensive nephropathy, nephrotic syndrome, and tubulointerstitial disease), eGFR >30 ml/min/1.73 m 2 ; (5) no serious complications of diabetes, such as diabetic ketosis, hyperglycemic hyperosmolarity syndrome, and hypoglycemia; no diuretic use within one month and no renal replacement therapy; and (6) no MRI contraindications. Eight cases were completely excluded because of multiple or large kidney cysts (the presence of a cyst larger than 4 cm in diameter in one kidney or the presence of multiple cysts with the sum of their diameters larger than 4 cm, n � 6) and eGFR <30 ml/min/1.73 m 2 (eGFR too low, no control, n � 2). Finally, 62 T2DM and DKD patients, and 10 NDKD patients were included in this study. 62 patients were divided into three subgroup according to UACR: DI group (UACR ≤ 30 mg/g); DII group (30 < UACR ≤ 300 mg/g); DIII group (UACR> 300 mg/g), and NDKD group (30 ml/ min/1.73 m 2 < eGFR < 90 ml/min/1.73 m 2 ).
Healthy volunteers were included accordant with the following specific criteria: no history of urinary system disease, no diabetes, and no use of any drugs that damage the kidneys; blood pressure control within 140/90 mmHg; no abnormalities in routine blood tests and liver and kidney function tests; good cooperation with MRI examinations; and no MRI contraindications.

Study Design.
We used a 3.0-Tmachine (Signa 750w, GE Healthcare, USA) with a 16-channel body coil. Before the scans, all the subjects were abstained from consuming food and water for 4 hours and underwent breathing training. Routine T2 and T1 sequences were scanned after localization to exclude significant renal lesions.

Scanning Parameters.
Coronal renal BOLD scanning: we used a multiple fast gradient recalled echo (MFGRE) sequence with 8 echoes with the following parameters: TR � 100 ms, TE � 3.3-26.2 ms, FOV � 32 cm × 32 cm, NEX � 1, slice thickness � 5 mm, spacing � 0.4 ms, matrix size � 256 × 256, bandwidth � 62.50 kHz per pixel, and flip angle � 20°. Images were acquired during a 14 second breath hold with an acquisition time of approximately 27 s, and the central line was placed at the renal hilum, with five layers in front and behind.

Image Analysis.
We measured the renal cortical R2 * (CR2 * ), medullary R2 * (MR2 * ), renal cortical FA (C.FA), renal medullary FA (M.FA), cortical average diffusion coefficient (C.ADC), and medullary average diffusion coefficient (M.ADC). All data were processed by two well-trained doctors independently with three years and six years of experience in imaging diagnosis, who were blinded to the population group before the measurement. We selected layers close to the renal hilum and manually drew regions of interest (ROIs) on BOLD and DTI images at the upper, middle, and lower poles of the cortex and the medulla of each kidney with ellipses of approximately 40-60 mm 2 . We tried to avoid renal blood vessels, edges, collecting systems, and cysts. Examples of pseudocolor images of the R2 * map, FA map, and ADC map are displayed in Figure 1.

Statistical Analysis.
Statistical analysis was performed using IBM SPSS Statistics 26.0, and the statistical significance was defined as p < 0.05. e parameters are expressed as the mean ± standard deviation (SD). One-way ANOVAs with the Bonferroni correction were used to test differences in R2 * , ADC, FA values among groups as well as D values (Δ, the measured medulla value-the corresponding cortical value) and MC ratios (MCR, the measured medulla value divided by the corresponding cortical value) of R2 * and FA across groups. A Spearman correlation was also computed between the MRI parameters and clinical indices of renal function.

General Clinical Characteristics of All Participants.
e general clinical information and laboratory results are shown in (Table 1), including sex, age, hemoglobin (HGB), body mass index (BMI), eGFR, blood urea nitrogen (BUN), hemoglobin A1c (HbA1c), uAlb, UACR, and Cys C. ere was no statistically significant difference in sex, age, or BMI among the HCs and the three diabetic subgroups. HbA1c in the three diabetes subgroups showed no difference. e mean eGFR of the DI group (115.25 ± 16.25 mL/min/ 1.73 m 2 ) was slightly higher than that of the HC group (109.49 ± 9.70 mL/min/1.73 m 2 ) (p > 0.05). HGB, SCr, BUN, uAlb, and Cys C were all significantly different among the three diabetes subgroups. e differences were mainly between the DI group and the DII/DIII group based on posthoc multiple comparisons (p < 0.05). Table 2 shows that there was no significant difference in sex, age, and clinical data between the DKD group (30 ml/min/ 1.73 m 2 < eGFR < 90 ml/min/1.73 m 2 ) and NDKD group.
Since renal puncture was an invasive method, most patients refused it when they were hospitalized. Among all the recruited subjects, 21 were definitely diagnosed with diabetes kidney disease and 10 more were diagnosed with NDKD by renal pathology. e specific pathological results and staging are shown in Table 3. We performed paired-samples t test between the left and right kidneys ( Figure 2). ere was no difference in cortical and medullary R2 * between the left and right kidneys (p > 0.05). However, FA and ADC values of right kidneys were significantly lower in both cortex (t � −6.054, p < 0.001; t � −0.530, p < 0.001) and medulla (t � −4.162, p < 0.001; t � −3.364, p < 0.01) than those of the left kidneys. Considering that the left kidney was most likely affected by the heart and blood vessels, we decided to individually utilize the DTI data of the right kidney.

Comparison of
R2 * , FA, and ADC values of the cortex were not significantly different among the HCs and the three diabetes subgroups (p > 0.05) ( Table 4).
Moreover, medullary R2 * values in DI group were significantly higher than those in the HC group (p < 0.001).
Medullary FA values in the HC group were higher than those in DI group (p � 0.005). Medullary ADC values were not significantly different among HCs, the DI, DII, and DIII group. We found no difference in R2 * , FA, and ADC values between DKD and NDKD (p > 0.05) (Figure 3).
To reduce individual differences, we calculated the D values (Δ) and MC ratios (r) between the renal medulla and the corresponding cortex ( Figure 4). Relative to the stable differences between the cortex and medulla of the kidneys in the HC group, the changes in the DIII group implied an decreasing functional and structural differentiation of the cortex and medulla as renal function deteriorated.

Correlative Analysis.
Spearman correlation analyzed the relationships between the cortical and medullary R2 * , FA values, and the eGFR, BUN, HbA1c, uAlb, and Cys C values among the DM patients ( Figure 5). e eGFR had a moderate and positive correlation with the MR2 * value and a positive but weak correlation with the M.FA (p < 0.001). In contrast, SCr was moderately and negatively correlated with MR2 * and negatively but weakly correlated with the M.FA (p < 0.001). Cys C had a weak and negative correlation with the MR2 * and M.FA (p < 0.01, p < 0.05). In addition, there was a correlation between SCr and the CR2 * value, but the correlation was weak.

Discussion
e present study verified the (patho) physiological differences between the kidney cortex and medulla. Based on the trends in R2 * and FA values, we can get a general idea of the Computational Intelligence and Neuroscience  changes in renal oxygenation levels, water molecular diffusion, and renal microstructure during the progression of diabetes to DKD. Our results revealed that the R2 * and FA values in the renal medulla were higher than those in the cortex in all the groups, and the same result was also seen in previous studies [21][22][23][24]. is demonstrated that the renal medulla was in a state of "physiological hypoxia" and provided evidence of a regular arrangement of radiating medullary loops, collecting ducts, and tiny vessels. R2 * and FA values in DKD patients showed significant differences from those in the HCs. e decline in FA was expected because changes in renal pathology, including glomerulosclerosis, interstitial fibrosis, tubular injury, and inflammatory cell infiltration, altered the anisotropy of wellaligned tubules [25,26]. Notably, in our study, medullary R2 * had a clear tendency to rise at the very early stage of   Computational Intelligence and Neuroscience DKD and rapidly decline once renal function was significantly impaired, which was same with the previous studies [27,28]. e oxygenation level of the human kidney depends on the careful and coordinated physiological structure, and its coordination is mainly reflected in the balance between the oxygen supply and oxygen consumption of the kidneys. From a hemodynamic point of view, blood flow and the oxygen-carrying capacity of arterial blood are mainly responsible for the activities of supplying oxygen to the kidneys. Processes such as renal tubular reabsorption also utilize the oxygen transported to the kidneys. e blood flow of the renal cortex is significantly higher than the blood flow of the renal medulla, and the renal medulla is in a "physiological hypoxia state". e difference in blood flow between the cortex and medulla helps maintain the osmotic pressure gradient, which is beneficial for the body to concentrate urine. e progression of kidney disease often changes in function earlier instead of that in structure. e level of renal oxygenation is a very potential and particularly suitable indicator for the early evaluation of DKD. At present, chronic hypoxia has been found in animal models of DKD, and its possible mechanism is related to factors such as hypoxia-inducible factors, cytokines and inflammatory mediators, and renal tubular epithelial cell trans differentiation. In animal models, improving chronic hypoxia can indeed delay the progression of chronic kidney disease to a certain extent [11].
BOLD-MRI, using deoxyhemoglobin as an endogenous contrast agent, relies on its paramagnetic properties to obtain images that are sensitive to local tissue oxygen concentration. e increase of deoxyhemoglobin content in the blood causes the inhomogeneity of the magnetic field in the surrounding local tissue, which leads to the rapid dephasing of protons in the tissue and thus the signal decreases on T2 * WI images. According to formulas (1) and (2), e apparent relaxation rate R2 * can be obtained by calculating the monoexponential curve fitting of the signal intensity of a series of T2 * WI images at different echo times to the TE time.
e ratio of oxyhemoglobin to deoxyhemoglobin is related to the partial pressure of oxygen in the blood. e partial pressure of oxygen in the capillaries being in equilibrium with the surrounding tissue, the signal  Legends: e results are expressed as the mean ± SD. R2 * , rate of spin dephasing (Hz); FA, fractional anisotropy; ADC, average diffusion coefficient (×10 −3 mm 2 /s). p: ANOVA was used to compare the values among all above groups, and a p value < 0.05 was considered significant. p # : Post-hoc paired comparisons between the HC group and the DII group; p & : Post-hoc paired comparisons between the HC group and the DIII group; p * : Post-hoc paired comparisons between the DI and the DIII group. 6 Computational Intelligence and Neuroscience changes detected by BOLD-MRI can be interpreted as changes in the partial pressure of oxygen in the tissue. e decrease of R2 * value indicates that the concentration of oxygenated hemoglobin increases in the local tissue, the proportion of deoxygenated hemoglobin decreases, and the local partial pressure of PaO 2 increases. BOLD-MRI technology was first applied to the study of brain function, and its basis mainly depends on the close connection between hemodynamics and brain neural activity. Prasad et al. were the first scholars to apply BOLD-MRI to assess the level of renal oxygenation in humans. ey also found that the medullary R2 * value of patients with kidney disease showed a significant decreasing trend after using diuretics, suggesting that diuretics can improve renal hypoxia [29]. In this study, for the subjects recruited, we excluded those who used diuretics and those taking antihypertensive drugs containing diuretics in the past month to prevent interference errors to R2 * values.
R2 * is proportional to the content of deoxyhemoglobin, so an increase in R2 * represents a decrease in PO 2 and tissue hypoxia [30]. DM patients might experience renal medulla hypoxia, which is related to the highly increased oxygen consumption and less blood supply in the medulla, such as active reabsorption of excess sodium, glomerular hyperfiltration, and increased Na+/K+-ATPase activity in a hyperglycemic state [28,31]. Significant elevation of MR2 * was shown in DM patients without kidney disfunction in our study. Considering that these patients had higher HGB values and eGFR levels compared with healthy controls and DKD patients, the elevated eGFR and HGB seemed to verify that the kidney tissue was in a state of hypoxia compensation. At this stage, some diabetic patients may experience increased blood pressure, and it is recommended that blood pressure should not be controlled too low. e removal of the high filtration state will cause insufficient renal blood perfusion and accelerate renal damage [32]. Besides, R2 * was also affected by other factors, such as hemodynamic and structural changes, red blood cell volume and the vascular volume fraction, vessel geometry, and applied pulse sequence parameters, including factors influencing the oxygen dissociation curve [33]. In the later stages of DKD, hypoxia indeed takes place in the medulla despite the decreased R2 * in diabetic patients with a moderately decreased eGFR. It was inferred that the eGFR decreased and the tubular reabsorption decreased, along with the metabolic products of renal tubular epithelial cells, interstitial inflammatory cells, and endothelial cells that stimulate interstitial inflammation and fibrosis in the kidney, so that oxygen consumption was decreased [34]. e phenomenon of "hypoxic remission" of the medulla appeared. Additionally, there was no significant  difference on MR2 * between the DKD and NDKD, implying that diabetic patients with different etiologies of CKD were consistent in their changes in renal oxygenation levels.
DWI imaging can noninvasively evaluate the movement of water molecules in vivo. In recent years, the study of renal DWI under physiological or pathological conditions has gradually become a hot topic. Diffusion tensor imaging (DTI-MRI) is a technology developed based on the DWI technology. It was an MRI technique proposed in 1994 to analyze white matter fiber tracts. It reflects the tensor information of the diffusional motion of water molecules in vivo by applying diffusion-sensitive gradients in at least 6 noncollinear directions. So, the description of water molecular motion is more precise. DTI employs parameter values to reflect tissue diffusion characteristics. An FA value, ranging from 0 to 1, measures the proportion of the anisotropic component of the diffusion of water molecules in the overall diffusion tensor.
e DTI postprocessing program of the workstation can correct the EPI distortion. e formula (3) required to generate the FA map is as follows. λ 1 , λ 2 , and λ 3 are the eigenvalues of the diffusion tensor.
e larger the value, the more restrictive the tissue is to the direction of diffusion of water molecules. Compared with DWI, DTI-MRI can not only observe the restricted degree of the diffusion movement of water molecules but also show the differences in the directionality of the diffusion movement of water molecules and provide information about the changes in the ultrastructure of the tissue, helping evaluate the subtleties in the structure of the tissue. is makes it possible to apply DTI imaging to the evaluation of kidney-related physiological  In general, cortical blood flow was abundant, and the radially arranged tubule structure in the medulla limited the diffusion of water molecules, so the ADC value was higher and the FA value was lower in the cortex [35]. In our study, the cortical and medullary ADCs did not differ significantly between HCs and the three diabetic subgroups. But diabetic patients had slightly lower cortical ADCs than compared with HCs, and DKDs with moderately decreased eGFR had higher medullary ADCs than the other groups. e reason relied on the possibility that diabetic patients had higher blood viscosity than HCs, and the disruption of the initially regularly arranged microstructure in the renal medulla could no longer well restrict the free diffusion of water molecules. Medullary FA decreased gradually with the decline in eGFR in diabetic patients since the kidneys in early-stage DKDs were likely to undergo microscopic changes such as tubular ectasia, tubular epithelial cell edema and necrosis, and interstitial inflammatory cell infiltration and proliferation, which destroyed the microstructures and inhibited the movement of water molecules [36,37]. e ADC and FA results in Feng's DTI study were similar to our study. It seemed that FA had more potential than ADC in determining the diffusion of water molecules and the microstructural changes in the human kidneys [33]. However, in some DWI studies, the advantages of apparent diffusion coefficient (ADC) values were reflected. Wang et al. found that the cortical ADC value in the DKD murine model was higher than in the blank model [38]. Emre et al. revealed that ADC values of renal parenchyma could be used for early diagnosis and clinical staging of patients with kidney disease [39]. Xu et al. found a negative correlation between renal parenchymal ADC values and the degree of interstitial fibrosis [40].
Medullary R2 * and FA had positive correlations with eGFR and negative correlations with SCr and Cys C, which is consistent with Lu L's study [41]. Serum Cys C is filtered through the glomerulus, entirely reabsorbed, catabolized, and metabolized by the proximal tubular epithelium. In addition, Cys C is not disturbed by inflammatory reactions, tumors, hemolysis, bilirubin, triglycerides, etc., and is also independent of sex, age, muscle mass, and diet. Cys C is a more sensitive indicator than SCr for the diagnosis of renal impairment. Mojiminyi et al. [42] suggested that Cys C is a more sensitive and practical indicator for detecting DKD. e advantages of BOLD and DTI allow them to refrain from radiation and there is no need for an injection of a contrast agent. e renal cortex and medulla worked together and somewhat differently and maintained a clear functional and structural differentiation. D-value and MCR compared the difference between the cortex and medulla in an individual kidney rather than focusing only on the changes in medullary R2 * and FA values. Compared with HCs and simple DM patients, statistically reduced corticalmedullary differentiation of the kidney was shown in DKD patients with microalbuminuria and massive albuminuria. With the improvement of technical methods and the continuous optimization of postprocessing analysis, DTI will surely show more special applications in scientific research and clinical practice. Its prospects are very broad, especially combined with IVIM technology and BOLD-fMRI in the field of nephrology.

Limitations of the Study.
e sample size of this study is relatively small. e number of cases in the subgroups requires amplification. In addition, this study only investigated BOLD and DTI parameters in assessing the changes in renal oxygenation levels and microstructure. It did not fully elucidate the mechanisms of renal hypoxia or provide insight into the specific changes in the structure in DKD combined with the pathological basis. Alterations in microcirculatory perfusion could also affect the result. At present, there is no unified technical standard for BOLD-MRI to detect renal oxygenation at home or abroad. e normal value of the relaxation rate is related to different field strengths. In DTI, diffusion gradients could cause eddy currents, which have an effect on the final parameters. Combining with arterial spin labeling (ASL) and IVIM, BOLD and DTI may be useful for a more comprehensive assessment of renal functional status.

Conclusion
In conclusion, our study shows that BOLD and DTI are effective and safe ways to estimate renal function through the oxygenation level, water molecular movement, and microstructural alterations in patients with DKD and it proves that BOLD performs better than DTI. With renal function declines, alterations can be found including the firstly rising and then falling medullary R2 * , continuously declining medullary FA and apparent cortex-medullary differentiation in DKD patients.