Diffusion tensor imaging (DTI) is emerging as a promising technique for structural and functional evaluation of the kidneys^1,2. Diffusion sequences employing echo planar imaging (EPI) readout are prone to geometric distortions due to static field inhomogeneities arising from different susceptibilities. Geometric distortion correction has its main application in brain MRI, but recent studies have shown its applicability to body organs^3,4. There is little experience with this technique for renal imaging, which is subject to distortions due to surrounding organs and adjacent bowel gas. The purpose of this work is to evaluate efficacy of distortion correction using EPI-based diffusion images acquired with opposite phase encoding (PE) polarities, in healthy controls and patients with Type 1 diabetes (T1DM).Ten T1DM patients (6M/4F, mean age [range] 44[19-62]y, mean eGFR [range] 84 [55-91], and six healthy volunteers (3M/3F, mean age [range] 44[28-63]y) were scanned using a 3T system (Skyra, Siemens) including a respiratory-navigated DTI sequence with the following parameters: TR/TE=1800/81 ms, FOV=400x400 mm², matrix=340x340, slice thickness 3 mm, 7 slices, b-values 0 and 500 s/mm² in 12 directions and 4 averages with a feet-to-head [FH] PE and EPI readout in oblique/coronal orientation. Additionally, one b0 acquisition with identical parameters and reversed head-to-feet [HF] PE was acquired. Breath-hold T2-weighted HASTE images (T2WI) were obtained covering the full extent of the kidneys with: TR/TE=1000/95 ms, FOV=420x420 mm², matrix=640x640, slice thickness 5 mm in identical orientation. Geometric distortion correction of DTI was carried out according to the method described in⁵ using CMTK⁶. An experienced abdominal radiologist selected a single representative slice from T2WI and matching slice from DTI datasets for subsequent evaluation. The same radiologist segmented bilateral kidneys in each subject on the selected T2WI slice, to obtain an area as the reference standard for each kidney. After a two week period, the same radiologist and an inexperienced resident radiologist independently segmented the matching uncorrected and corrected DTI-b0 images in random order. Segmentation and parametric maps were produced using non-commercial software developed using Matlab R2014. DTI parametric maps were fitted for both uncorrected and corrected series including the apparent diffusion coefficient (ADC), fractional anisotropy (FA), and the three tensor eigenvalues (λ₁,λ₂ and λ₃). Segmented area was compared with Pearson correlation and Wilcoxon signed ranks test. Intraclass correlation (ICC) between readers was assessed.Geometric distortion correction was successfully performed in all cases (Figure 1). Table 1 presents kidney area segmented on T2WI, uncorrected DTI-b0, and corrected DTI-b0 images. Excellent agreement was found between both readers with an ICC of 0.97 and 0.98 for kidney segmentation in uncorrected DTI-b0 and corrected DTI-b0 respectively. Figure 2 presents a scatter/boxplot for direct comparison of T2WI segmented area with areas obtained by the experienced reader for uncorrected and corrected DTI-b0 images. Pearson correlation between segmented area on T2WI and corrected DTI-b0 was higher (r=0.904; p<0.001), than the correlation between the area segmented in T2WI and uncorrected DTI-b0 (r=0.840; p<0.001). Median [range] of DTI parametric maps for all cases are presented in Table 2 using the experienced reader segmentation. Figure 3 shows ADC and FA parametric maps for one case before and after distortion correction.Our results show there is a significant improvement in the agreement between the area segmented on T2WI (the anatomic reference standard) and the area segmented on corresponding slices in the DTI series after distortion correction. This includes T1DM patients with a spectrum of renal function. This has implications for more reliable segmentation of renal cortex and medulla in DTI, with better correlation to higher resolution anatomic images, of importance when renal pathology is present. In our small cohort, the parameters derived from DTI did not seem to be affected. This observation strengthens its applicability as parameter quantification would not be affected by the processing routine. Nevertheless, it is important to consider that statistical comparison was carried out accounting only for the median value of each parameter in the single-slice kidney segmentation. In this study the PE trajectory for the complete DTI acquisition was chosen as FH, based on⁷. Therefore, geometric distortion resulted in artefactual elongation of the kidney (compression in the reversed HF b0 image) in the PE direction, as clearly depicted in Figure 1. The choice of PE in this study required that for most cases, the stomach needed to be masked out of the images before correction, as it caused significant artifact affecting the left kidney. Future work to evaluate how DTI parameters might be affected with smaller ROIs to segment renal cortex from medulla using DTI parametric maps with reference to anatomic imaging is planned.