The diagnostic efficacy of diffusion tensor imaging in children with chronic kidney disease: correlation with histopathology and serum biomarkers

Background Children with chronic kidney disease (CKD) usually present with disease impact on growth besides car‑ diovascular problems that not only impact the patient’s health during childhood but also affect their adult life. We aimed to identify the diagnostic role of diffusion tensor imaging (DTI) in CKD in pediatric using its metrics: apparent diffusion coefficient (ADC) and fraction anisotropy (FA). Results This prospective study was performed on thirty‑five CKD patients (16 girls, 19 boys; mean age 12.3 ± 2.6 years) and 19 sex‑ and age‑matched controls. Both groups underwent renal DTI and renal function tests. Based on renal biopsy, patients with CKD were further categorized into sclerotic CKD ( n = 25) and non‑ sclerotic CKD ( n = 10). Mean FA renal medulla/cortex in CKD (0.18 ± 0.18 and 0.20 ± 0.17) was lower significantly ( p = 0.001) than volunteers’ (0.31 ± 0.19, 0.27 ± 0.18). The cutoff FA of renal medulla/cortex used for CKD diag‑ nosis was 0.22 and 0.23 with AUC of 0.828, 0.838 and accuracy of 80.8%, 82.8%. Mean of renal medulla/cortex ADC in CKD (2.13 ± 0.23 and 1.93 ± 0.22 × 10 −3 mm 2 /s) was higher significantly ( p = 0.001) than that of volunteers’ (1.67 ± 0.15 and 1.64 ± 0.133 × 10 −3 mm 2 /s. ADC cutoff value of renal medulla/cortex used for CKD diagnosis was 1.86 and 1.74 × 10 −3 mm 2 /s with AUC of 0.827, 0.82, 0.827, and 0.911, and accuracy of 80.6%, 79.6%, 82.8%, and 84.2%. Renal medulla/cortex FA in sclerotic CKD was significantly different ( p = 0.001) from non‑sclerotic CKD (0.25 ± 0.07 and 0.26 ± 0.08). Cortical and medullary FA in CKD patients correlated with e‑GFR ( r = 0.363, r = 0.317) and serum creati‑ nine ( r = − 0.467, r = − 0.383). Conclusions Renal cortical/medullary FA can assist in diagnosing pediatric CKD, predict sclerotic CKD, and correlate with some serum biomarkers.

besides cardiovascular problems, that not only impact the patient's health during childhood but also affect their adult life [1,2].Renal fibrosis refers to the deposition of the pathological matrix in the interstitial space, glomerular capillaries' walls, and around arterioles secondary to metabolic, toxic, immunological, and mechanical insults.Scarring of renal parenchyma can negatively impact renal function by progressive loss up to end-stage renal failure that entails permanent dialysis or renal transplantation.End-stage kidney disease manifests histologically as fibrotic lesions affecting different compartments; tubulointerstitial fibrosis, glomerulosclerosis, and vascular sclerosis.The histological pattern can predict and contribute to the functional demise of the kidney [3,4].
Serum biomarkers as blood urea nitrogen level, serum creatinine, and estimated glomerular filtration rate (e-GFR) are helpful parameters for assessing renal function and grading the disease severity.Still, they cannot reflect morphological changes in the kidney.Ultrasonography, CT, and MR can provide anatomic images without functional information.Contrast media used in MRI and CT studies can induce systemic nephrogenic fibrosis and nephrotoxicity, respectively.Hence, this limits their use in patients with chronic kidney disease.Scintigraphy is a well-established imaging tool that merely assesses renal function by measuring glomerular filtration rate; however, it results in radiation exposure and has a lower spatial resolution.Evaluation of renal parenchymal microstructure is crucial to diagnose and monitor kidney diseases.It is now established by renal biopsy, which is an invasive procedure with hazardous side effects and sampling bias [5,6].
Diffusion-weighted imaging (DWI) is a noninvasive imaging tool that depends on the diffusivity of water molecules within a tissue.It can give an idea about the structural changes of the tissue occurring in different pathologies in addition to its ability to quantitatively assess the magnitude of water diffusion [7][8][9][10].Diffusion tensor imaging (DTI) is a powerful technique based on DWI, as DWI allows assessment of the water random motion and provides information about how the tissue cellular structures are aligned.It enables the analysis of anisotropic properties of the tissues by determining the principal directions of preferential water diffusion, which can be expressed quantitatively by DTI metrics: apparent diffusion coefficient (ADC) and fractional anisotropy (FA) [9][10][11].DTI was used in the assessment of renal parenchyma in the normal population [12] and various renal disorders: renal allograft [13,14], renal fibrosis [15,16], glomerulonephritis [17], diffuse kidney parenchymal disease [18][19][20], lupus nephritis [21], and diabetesinduced nephropathy in humans [22] and rats [23].This work aims to identify the diagnostic value of DTI in pediatric CKD and detect the correlation between the DTI parameters, clinical and pathological stages of CKD disease.

Patients
Parents of enrolled children signed informed consent, and institutional review board approval was obtained.The study included 38 children prepared for renal biopsy due to chronic kidney disease with high serum creatinine levels and low GFR.All patients were diagnosed with chronic kidney disease based on Kidney Disease Improving Global Outcomes (KDIGO) guidelines [1].Patients with chronic liver, blood disease, or neoplastic disease were excluded.Finally, the study group of 35 patients underwent conventional MR imaging and renal DTI.A control group of 19 children of matched age and sex were included in the study.The control group comprised children without a history of renal diseases who were referred to perform abdominal MRI to assess other pathologies rather than a renal disease and had normal renal function tests.

Image analysis
Diffusion tensor imaging data were processed by a workstation with DTI functional tool software (Extended Philips MR Workspace 2.6.3.2).Two expert radiologists with ten and twenty years of experience in MR imaging performed the analysis.Both observers were not informed by patients' laboratory and clinical data.Axial T1 and T2 were assessed to exclude any gross abnormalities.FA and ADC color maps were extracted.T2 images with distinctive cortico-medullary differentiation were applied directly to the opposing coregistered FA and ADC maps to measure the cortical and medullary FA and ADC accurately.Three regions of interest (ROI single-pixel seed) (Fig. 3) were placed in different regions in the cortex and medulla of both kidneys, approximately in the anterior, middle, and posterior regions away from the hilar vessels.The ADC and FA of different cortical and medullary regions were calculated.The mean values of both kidneys representing the renal FA and ADC for each subject were used for the statistical analysis (Fig. 1).

Renal biopsy
Under sterilized conditions and local anesthesia, an ultrasound-guided biopsy was performed.Patients lied in prone or lateral decubitus position, and a 16-gauge core needle was used.Cores were sent for light and electron microscopy evaluation.Pathologists were asked to assess the histopathological pattern of CKD and categorize the study group into sclerotic (Fig. 2) and nonsclerotic renal disease (Fig. 3) based on the presence of glomerulosclerosis.

Serum and urinary biomarkers
Serum creatinine was calculated by using COBAS INTE-GRA 400 plus device.The Bedside Schwartz equation was used for estimating e-GFR from serum creatinine [24].

Statistical analysis
Data were collected, revised, and sent for analysis using a statistical package for social science (SPSS, version 22.0.Armonk, NY: IBM Corp).Quantitative data were represented by the mean and standard deviation of the ADC, FA, e-GFR, and serum creatinine.Testing of the statistically significant difference was judged at (0.05) level.To compare between two groups, one sample student t test was used.The Chi-square test was used for comparison between more two groups.The optimal cutoff point utilizing the area under curve analysis (AUC), diagnostic capability, sensitivity, specificity, and accuracy of the cortical and medullary FA and ADC was performed using the receiver operating characteristic (ROC) analysis.It was used to differentiate between diseased and control groups and between the sclerotic and non-sclerotic that was highlighted with Masson trichrome stain (B).Renal cortex FA was 0.17 and ADC was 1.88 × 10 −3 mm 2 /s, FA for medulla was 0.15 and ADC was 1.92 × 10 −3 mm 2 /s groups.Spearman's rank-order correlation was used to determine the linear relationship's direction and strength between renal parenchyma FA and ADC and laboratory parameters (e-GFR and serum creatinine).A Kappa agreement (GraphPad Software Inc. CA) was estimated by cross-tabulation for categorical variables with Kappa as follows (slight agreement from 0.01 to 0.20, fair agreement from 0.21 to 0.40, moderate agreement from 0.41 to 0.60, good agreement from 0.61 to 0.80, and perfect agreement from 0.81 to 0.99).

Result
The study group comprised 35 patients (19 boys, 16 girls with a mean age of 12.23 ± 2.79 years), while the control group was composed of 19 children (10 boys and 9 girls; mean 11.0 ± 2.50 years).Table 1 shows CKD patients' age, sex, clinical, urinary biomarkers, serum, and DTI parameters versus the control group.All scans show a normal morphological appearance of both kidneys with no signal abnormalities.The etiology of CKD varied from obstructive uropathy, hypertension, hypoplastic or dysplastic kidneys, reflux nephropathy, focal segmental glomerulosclerosis, polycystic kidney disease, and juvenile-onset lupus erythematosus.
Figure 5 demonstrates the correlation between FA with serum creatinine and e-GFR.
The Kappa coefficient indicated good to excellent interobserver agreement as regards the FA and ADC reading in renal cortex and medulla (Table 3).

Discussion
In this study, the main finding was a significant difference between the cortical and medullary FA and ADC in CKD patients and the control group.Moreover, we found a significant difference between FA in patients with sclerotic CKD versus non-sclerotic CKD.A correlation was found between DTI and laboratory parameters, including serum creatinine and e-GFR.
The FA value represents the ability of diffusion tensor imaging to detect anisotropic diffusion.It is mainly related to the preferential direction of medullary molecular diffusion due to the radial orientation of the renal vessels and tubules, in addition to the contribution to microcirculation [19,20].In this study, the mean FA value of the renal cortex/medulla in patients with CKD is statistically reduced from that of a control group.This reduction in diffusion anisotropy is possibly caused by structural parenchymal changes in CKD, such as cellular infiltration, interstitial fibrosis, and tubular atrophy in various CKD.Medullary FA was a sensitive and accurate marker for the detection of CKD, which is coping with many authors who considered medullary FA as a marker for renal structural integrity alteration and a useful potential index for kidney disease diagnosis in children [25][26][27].Additionally, Li et al. concluded that medullary Fig. 4 A ROC of FA.The cutoff values of FA of the cortex/medulla used to diagnose CKD were 0.23, 0.22, AUC is 0.828, 0.848, and accuracy is 82.9%, 80.7% respectively.B ROC of ADC.The cutoff values of ADC of the cortex/medulla used to diagnose CKD were 1.75, 1.85 × 10 −3 mm 2 /s, AUC is 0.910, 0.81, and accuracy is 84.1%, 79.5% respectively FA is a noninvasive tool of assessing pediatric renal allograft health and correlates to several renal allograft histopathology scores [28].Apparent diffusion coefficient or mean diffusivity quantifies cellular and membrane density where an increase in mean diffusivity indicates disease processes such as edema or necrosis.Whereas the FA is related to the water molecule transport in the collecting tubules, the ADC is mainly influenced by perfusion [20].Disrupted microcirculation and molecular diffusion by structural parenchymal changes influence the cortical and medullary perfusion and subsequently affect the ADC [18].In this study, the ADC value of the renal cortex/medulla in patients with CKD is statistically higher than that of volunteers.There is a discrepancy regarding the relationship between ADC and CKD in the current literature.In agreement with our results, Ye et al. found that cortical and medullary ADC values showed a trend toward an increase in the diabetic group with early-stage kidney disease as compared to the control group [29].Gaudiano et al. found no significant difference for the cortical and medullary ADC values between the control and the CKD groups [19].Contrary to our results, some studies reported that the ADC value was lower in the disease group than in control subjects [26][27][28][29][30]. Mao et al.  added that the ADC value is significantly lower in the medulla than the cortex for both patients and controls (p = 0.01) [31].This discrepancy may be secondary to different parameters used for DTI, various hydration conditions of the patients, various disease duration, in addition to variable grades of active inflammation, arteriolosclerosis, and glomerulosclerosis.
Glomerulosclerosis and tubulointerstitial fibrosis are consequences of CKD progression secondary to an imbalance between excessive synthesis and reduced breakdown of the extracellular matrix replacing the renal cells [4].The correlation of DTI parameters with CKD pathological subtypes was not studied before, it is needed to understand the types of structural changes and their effect on FA and ADC.In this study, FA of renal cortex/medulla was significantly different in patients with sclerotic CKD compared to non-sclerotic kidney disease.Deposition of the hyaline matrix in the renal cortex/medulla in sclerotic renal disease results in diffuse and nodular glomerulosclerosis, interstitial fibrosis, and tubular damage that will hinder the preferential diffusion of water molecules across the renal tubules in addition to blood stasis [23,32].This will result in reduced FA value in both the renal cortex/medulla as demonstrated by our results.ADC value also reduced in both the renal cortex/ medulla despite being statistically insignificant likely due to a small number of the study population.
The renal medullary and cortical FA in patients with CKD are significantly correlated with serum creatinine and e-GFR in this study.Spearman's correlation was tested and resulted in a negative and positive correlation between the FA of the cortex/medulla and serum creatinine and e-GFR respectively.One study reported that there is a reducing trend in medullary FA values with higher stages of CKD [33].However, others could not find this correlation in their study and attributed it to the marked heterogeneity of the study population [19].Similar to our results, Saini et al. reported a negative correlation between FA of the renal medulla with serum creatinine (r = − 0.785) and a strong positive correlation between medullary FA with e-GFR in renal parenchymal disease (r = 0.598).In the current study, there was no significant correlation between renal cortical/ medullary ADC with serum creatinine or GFR [34].In a recent study, ADC extracted from intravoxel incoherent motion (IVIM) showed a positive correlation with e-GFR (r = 0.72) and a negative correlation with renal interstitial infiltrations (r = − 0.44) [35].Moreover, Ren et al. reported a positive correlation between cortical ADC extracted from IVIM and e-GFR of the transplanted kidneys [36].
The first limitation of this study is the relatively small number of patients and control groups, further multicentric studies including a larger number of pediatric CKD are recommended.Additionally, we could not assess the utility of DTI in differentiation between histopathological types of CKD.The second limitation was that DTI parameters were calculated through a uni-seed region of interest of the renal cortex/medulla.Multiparametric MR techniques such as intravoxel incoherent motion diffusion-weighted MRI (IVIM-DWI) with advanced post-processing methods such as histogram analysis in addition to artificial intelligence may improve the results since overall renal microstructural changes can be detected through more parameters.

Conclusions
We concluded that renal cortical and medullary FA can help in the diagnosis of CKD and could predict sclerotic CKD.FA cortex/medulla correlate with some serum biomarkers.

Fig. 1
Fig. 1 Region of interest localization.DTI color map of kidney shows the localization of ROI within the cortex/medulla of both kidneys in a child with CRPD

Fig. 5
Fig. 5 Scatter diagrams show correlation between FA of renal cortex/medulla, serum creatinine, and GFR

Table 1
Demographic, serum, and DTI parameters of CKD patients and controls

Table 2
Mean FA & ADC cortex and medulla according to the presence of sclerosis

Table 3
Inter-observer agreement for DTI parameters