Introduction

Chronic glomerulonephritis is the most common cause of end-stage renal disease in China [1]. The condition develops over a period of several years, up to 20 years, and is followed by progression to end-stage renal disease. Structural abnormalities in chronic glomerulonephritis include kidney hypertrophy, thickening of the glomerular basement membrane, nodular and diffuse glomerulosclerosis, tubular damage, and interstitial fibrosis [1–3]. At present, kidney biopsy is the only method to evaluate histopathological changes in chronic glomerulonephritis. However, biopsy is an invasive procedure and may cause complications, and thus non-invasive imaging techniques are needed.

Diffusion-weighted imaging (DWI) of the kidney is a feasible approach in the characterization of renal diseases and is used in an increasing number of studies [4–10]. DWI visualizes water motion at the molecular level and provides useful information on parenchyma microstructure and function, with the major advantages of not using ionizing radiation or potentially nephrotoxic contrast agents [11, 12]. The apparent diffusion coefficient (ADC) is a quantitative parameter calculated from MR-DWI images, which is altered by various physiological and pathological conditions of the renal system. Previous studies have shown that ADC values are decreased in chronic kidney disease, depending upon the degree of renal function impairment [13–15]. Togao et al. [16] demonstrated a relationship between decreased ADC and renal fibrosis.

The predominant function of the kidneys is to transport water, and the anisotropic diffusion properties in the kidneys are the result of structures such as vessels, tubules, and collecting ducts that are oriented in a radial fashion [17]. Diffusion is a three-dimensional process, and molecular mobility in organized tissue is not necessarily the same in all directions. DWI does not allow analysis of diffusion in multiple directions, but tissue anisotropy can be analysed using diffusion tensor imaging (DTI), which provides additional information on diffusion direction and degree of directed diffusion (fractional anisotropy, FA) by analysing water diffusion in different directions. DTI was first used by Ries et al. in the kidneys of healthy volunteers [18]. Renal DTI has recently been used for the detection of allograft dysfunction in humans and ischemia-reperfusion injury in a rat model, in which FA was significantly reduced [19, 20].

The purpose of this study was to investigate whether DTI can be used to assess renal function and to determine the correlation between diffusion anisotropy and the degree of renal pathology in disorders such as glomerulosclerosis and tubulointerstitial fibrosis in patients with chronic glomerulonephritis.

Materials and methods

The study population

A total of 75 patients (39 men and 36 women) admitted to the nephrology department in our hospital were retrospectively recruited to participate in this study. Patients had chronic kidney disease (CKD) stages 1 to 5, with 15 cases per stage. Kidney biopsies were performed in all patients with CKD stage 1–3; biopsy was performed in only one patient with stage 4 CKD. Before renal biopsy, all subjects underwent functional MRI and laboratory tests, and MRI scans were performed within seven days of hospital admission. The estimated glomerular filtration rate (eGFR) was calculated using the Modification of Diet in Renal Disease (MDRD) formula for Chinese subjects, which is eGFR (mL/min per 1.73 m2) = 186 × (Scr)-1.154 × (age)-0.203 × 1.233 × (0.742 if female), as recommended by the Chinese Society of Nephrology. The diagnosis of CKD was based on a chronic decline in eGFR (<60 ml/min per 1.73 m2) and was classified into stages as follows: stage 1, eGFR >90; stage 2, 60–89; stage 3, 30–59; stage 4, 15–29; and stage 5, <15. Chronic glomerulonephritis was diagnosed based on clinical history, laboratory findings, and renal biopsy results. Twenty volunteers with no history of renal disease, hypertension, diabetes, or vascular disease were included in the study. The study was approved by the local ethics committee, and informed consent was obtained from the participants prior to MRI examination.

MR protocol

All examinations were performed using a 1.5 T whole-body scanner (MAGNETOM Avanto, Siemens Healthcare, Erlanger, Germany) with a maximum gradient strength of 45 mT/m, a minimum rise time of 225 μs, and 32 receiver channels. For DTI, fat-saturated coronal EPI sequences were performed in six directions using the Siemens multidirectional diffusion-weighted protocol with b-values of 0–600 s/mm2.

The remaining parameters for respiratory-triggered imaging were as follows: 10 slices; slice thickness, 6 mm with no intersection gap; TR, 1400 ms; TE, 76 ms; bandwidth, 1370 Hz/pixel; field of view in read direction, 400 mm; field of view in phase direction, 100 %; voxel size, 2.1 × 2.1 × 6 mm3; partial Fourier factor, 6/8; number of excitations (NEX), 4; phase encoding direction, left to right; k-spaced parallel imaging technique, GRAPPA with an acceleration factor 2; and acquisition time, 2 minutes 4 seconds to 3 minutes. The kidneys were imaged consecutively with 10 slices in an oblique coronal orientation, slightly tilted, parallel to the long axis of the kidney.

MR data analysis

Images were analysed on an external workstation using Syngo software (Siemens Healthcare) by two readers who were blinded to histopathological and laboratory findings. For DTI analysis, the commercially available Neuro 3D Task Card software (Siemens Healthcare) was used. The diffusion tensor was determined by the orientation and magnitude of diffusion. The degree of diffusion anisotropy was calculated and depicted in parameter maps of FA; apparent diffusion coefficient maps were calculated based on a mono-exponential fitting model. On ADC maps of the same kidney, two authors measured the parenchyma and three regions of interest (ROIs) from the upper, middle, and lower pole of the cortex and medulla. ROIs with a size of 10–50 pixels each were placed into the cortex and medulla of the kidney and were copied to FA maps (Fig. 1). Mean ADC and FA values were determined separately in all patients. The ADC and FA values in parenchyma were also measured.

Fig. 1
figure 1

The region of interest from parenchyma (A) and three regions of interest from the upper, middle, and lower pole of the cortex and medulla (B) are shown on the FA maps

Full size image

Histology

Formalin-fixed kidney tissue was embedded in paraffin. Coronal sections (2 μm) were prepared and were subjected to Masson staining to visualize the accumulation of tubulointerstitial fibrosis and tubular damage. Periodic acid-silver methenamine (PASM) staining was used to determine glomerulosclerosis and mesangial expansion within renal glomeruli. For quantification of tubulointerstitial fibrosis and tubular damage, 10 representative images of each patient were evaluated. The percentage of Masson-positive area per total tissue area was used as a measure of tubulointerstitial fibrosis. For the assessment of glomerulosclerosis, 10 representative glomeruli in each patient were evaluated, and the mesangial index was expressed as the percentage of PASM-positive area per total glomerular tuft area. Renal histology was analysed by one author.

Statistical analysis

Statistical analysis was performed using SPSS 13.0. Values are given as mean and standard deviations. P values of 0.05 were considered statistically significant. The consistency of measurements between the two authors was assessed using the Kappa test. Mean values for physiological parameters, histology, ADC, and FA within the groups were calculated. Parameters were compared among groups using one-way analysis of variance (ANOVA), followed by post hoc multiple comparison with the Fisher LSD test. Within the groups, FA and ADC values in the renal cortex and medulla layers were compared using the paired t-test. The correlation between histopathological parameters and FA and ADC values in the anatomical layers was determined by linear regression analysis (r is the regression coefficient).

Results

Trend of renal ADC and FA values among stages

For each patient, the image quality of renal DTI was satisfactory for further evaluation, with no motion artefacts, distortion artefacts, or morphological abnormalities. With regard to the ADC value in renal parenchyma, cortex, and medulla, the consistency of the two authors was high (r = 0.88, 0.90, and 0.91, respectively), similar to that for the FA value (r = 0.84, 0.87, 0.91, respectively). The ADC and FA values were obtained in five different disease stages (Table 1, Fig. 2). The ADC value in the renal parenchyma ranged from 1.86 to 2.33 × 10-3 mm2/s. The FA value in the renal parenchyma ranged from 0.18 to 0.27. The ADC and FA values tended to decrease with increased disease stage (Fig. 3), and significant differences were observed among the five different stages. The ADC and FA values were not statistically different between the control group and stage 1 patients (P = 0.52 and 0.54, respectively). Significant differences between the control group and stage 2–5 patients were observed (P = 0.00), similar to the differences between stage 1 and stage 2–5, stage 2 and stage 3–5, stage 3 and stage 4–5, and stage 4 and stage 5 patients. Significant differences were also found in Scr and eGFR (P = 0.00).

Table 1 Comparison between control group and groups with different stages of chronic glomerulonephritis
Full size table
Fig. 2
figure 2

ADC map (A–E) and FA map (F–J) in different disease stages, with A–E and F–J each representing stages 1–5. On ADC maps, low signal area of renal medulla decreased with increased stages. On FA maps, high signal area of renal medulla decreased with increased stages. The difference between cortex and medulla were significantly reduced with increasing disease stage. In stage 1, low signal area of renal medulla on ADC maps and high signal area of renal medulla on FA maps were obvious; in stages 4 and 5, renal cortex and medulla were not differentiated

Full size image
Fig. 3
figure 3

ADC values (A) and FA values (B) in different disease stages. ADC and FA were significantly reduced with increasing disease stage

Full size image

Corticomedullary gradient of ADC and FA values among stages

Due to a lower difference between the cortex and medulla and lower significance of renal biopsy, patients with stage 4 and 5 CKD were not analysed. The cortex and medulla layers of the kidney were identified on MRI in stage 1–3 patients. In general, the ADC was significantly lower in the medulla than in the cortex in the different stages (P < 0.05). Inversely, FA was significantly higher. ADC and FA were significantly reduced with increasing stage (P < 0.05).

In stage 1–3 patients, ADC values in the cortex and medulla were found to be statistically different, similar to the FA values in the cortex and medulla (Table 2).

Table 2 Comparison of ADC and FA value in cortex and medulla among groups stage 1–3
Full size table

Histopathology of chronic glomerulonephritis

From stage 1 to stage 3, the percentage of the PASM-positive area per total glomerular tuft area was 8.21 ± 2.15, 18.27 ± 2.37, and 28.4 0 ± 3.14, respectively, and significant differences among the groups were observed (F = 228.13, P < 0.05). The percentage of the Masson-positive area per total tissue area was 8.58 ± 1.60, 17.73 ± 1.83, and 27.41 ± 2 .77, respectively (F = 228.13, P < 0.05). The comparisons among the groups showed statistical differences, indicating that glomerulosclerosis and tubulointerstitial fibrosis significantly increased with higher disease stage (Fig. 4).

Fig. 4
figure 4

Histology of chronic glomerulonephritis. The degree of magnification is 200. A–D and E–H each represent stages 1–4. (A–D) PASM-stained sections of the renal cortex show progressive glomerular changes, such as glomerulosclerosis (black area within the glomerular tuft). (E–H) Masson-stained section reveals progressive tubulointerstitial fibrosis, i.e., increased Masson-positive (blue) area

Full size image

Correlation between ADC/FA values and renal histopathology

Significant negative correlations between the percentage of glomerulosclerosis and FA in the renal cortex and medulla were observed (r = -0.74, -0.70, respectively, P < 0.05). Correlations between glomerulosclerosis and ADC in the renal cortex and medulla were also observed (r = -0.65, -0.66, respectively, P < 0.05). The degree of tubulointerstitial fibrosis was negatively correlated with FA in the medulla and renal cortex (r = -0.76, -0.70, respectively, P < 0.05). Correlations were also found between tubulointerstitial fibrosis and ADC in the medulla and renal cortex (r = -0.68, -0.67, respectively, P < 0.05).

Discussion

To date, there are only a few studies investigating the value of DTI in the assessment of renal disease. Measurements of DTI parameters within the kidneys, such as ADC and FA, have shown high repeatability and reproducibility, with low inter- and intra-observer variability, in healthy volunteers [21, 22].

In the present study, we demonstrated that the ADC and FA values reflected the level of renal function and the degree of histopathological changes. There were significant differences in renal ADC and FA values at different disease stages. A possible reason for this is that renal function was assessed using the MDRD formula for Chinese subjects [23], which is more sensitive and more accurate than the original MDRD formula. In the renal parenchyma, the ADC value decreased with higher disease stage, indicating a correlation between chronic kidney disease and ADC value, which is in accordance with previous studies [14–16]. The pathology of chronic glomerulonephritis, such as glomerulosclerosis and tubulointerstitial fibrosis, leads to a reduction in blood flow and blood perfusion, resulting in a decrease in the amount of global diffusivity. The FA value in the parenchyma decreased with higher disease stage. A possible explanation for this is the fact that glomerulosclerosis and tubulointerstitial fibrosis destroy renal anisotropic architecture.

Magnetic resonance DTI demonstrated higher ADC values in the cortex than the medulla, and significantly higher FA values in the medulla than in the cortex. These findings were consistent with renal anatomy in previous studies [11, 20]. It has been shown that cortical and medullary FA is reduced in transplanted kidneys compared with normal kidneys and that medullary FA correlates with eGFR [19]. In a rat model, Cheung et al. observed a reduction in medullary FA five hours after ischemia-reperfusion injury, which was associated with cellular oedema and necrosis at histology. Hueper et al. [24] found a significant negative correlation between the reduction in FA and pathologies in diabetic nephropathy, and concluded that these pathologies impaired diffusion along the radially oriented tubules, which explained the reduction in FA. Lu et al. [25] found that medullary FA was reduced in diabetic patients, with and without impaired renal function, when compared with controls, indicating the value of DTI for the early identification of diabetic nephropathy. To date, the histopathological correlation with chronic glomerulonephritis has not been demonstrated.

In our study, renal biopsies were performed in patients with stage 1 to 3 CKD, as early intervention can reverse chronic nephritis and delay pathological changes, whereas advanced pathological changes are irreversible in patients with stage 4 and 5 CKD. In stages 1 to 3, we were able to demonstrate that histopathological changes of chronic glomerulonephritis such as glomerulosclerosis and interstitial fibrosis were negatively correlated with ADC and diffusion anisotropy (FA) in MR DTI, and therefore three stages of chronic glomerulonephritis were evaluated in the present study. In our study, quantification of diffusion anisotropy revealed that the FA in various anatomical layers was significantly reduced in patients with chronic glomerulonephritis compared to lower disease stages, indicating restriction of directed diffusion.

The mechanisms of renal pathologies that act to decrease diffusion anisotropy, however, remain unclear. One theory is that tubular dilatation in chronic glomerulonephritis removes some of the directionality of diffusivity along the tubules, thereby contributing to FA restriction. Furthermore, tubular dilatation and glomerulosclerosis are associated with cellular damage and renal dysfunction, which may influence diffusion direction. Cellular debris might, for example, congest the tubules and thus impair directed diffusion. Interestingly, FA reflects the pathological changes in the kidney with potentially greater sensitivity than ADC, as different pathologies were correlated with FA in specific anatomical layers. Quantifying diffusion parameters separately for each anatomical layer, as performed in this study, may help to differentiate renal pathologies. Similar to what was demonstrated in the study by Gaudiano et al. [26], FA value appears to be the more useful parameter for assessing renal damage.

The shortcomings of the current study are as follows. Firstly, this study did not include different pathological types of chronic nephritis, which influence disease outcome. Secondly, we did not evaluate the sensitivity of ADC and FA values using statistical methods. The results from the present study will be further used to test renal function and pathology in a study using statistical methods.

In conclusion, MR DTI identified renal pathologies associated with chronic glomerulonephritis by showing reduced ADC and FA values, representing different stages of disease. Diffusion anisotropy was negatively correlated with the degree of glomerulosclerosis and interstitial fibrosis. Therefore, MR DTI may be valuable for non-invasive monitoring of chronic glomerulonephritis.