Type 2 diabetes mellitus is a prevalent chronic disease worldwide and diabetic nephropathy (DN) has become the main leading cause of end-stage renal disease. The onset of DN is always ambiguous, and renal function deteriorates progressively in clinical phase. Diffusion Tensor Imaging (DTI) as a noninvasive technique can provide valuable information, which is based on the Brownian motion of water. The reduction of fractional anisotropy (FA) and apparent diffusion coefficient (ADC) values of cortex and medulla in diabetics with impaired renal function may be related to glomerulosclerosis, interstitial fibrosis, and tubular damage of DN. However, the diffusion of water molecules in biological tissue like in kidney, is restricted and does not follow a Gaussian distribution. Diffusion Kurtosis Imaging (DKI) can reflect the degree of restriction of hydrogen diffusion movement¹, and might detect the diffusion changes of kidney diseases, such as diabetes, earlier and more sensitive than DTI. Our study aimed to compare the diffusion changes of DKI and DTI in kidneys of DN and healthy controls.Four patients with diabetes (mean age 45.0±3.6 years, 2 females and 2 males) and 11 healthy controls (mean age 24.1±6.7 years, 5 females and 6 males) were recruited. Subjects with history of renal diseases, hypertension and other vascular diseases, and abnormal findings in kidney on MRI were excluded from the study. All four patients with chronic kidney disease due to diabetic nephropathy shows increasing 24 hours proteinuria. The FA and ADC were measured for the cortex and medulla on DTI images. Coronal-oblique DKI was obtained with following parameters: field of view, 230× 230 mm², TE/TR, 49/596ms, slices, 9; slice thickness, 4 mm with no intersection gap, and b values of 0, 300, 500 and 700 s/mm² on 3 gradient directions. DTI was acquired with an oblique-coronal fat-saturated imaging sequence with the following parameters: diffusion directions, 6; b values, 0 and 300 s/mm²; TR/TE, 405/43ms; averages, 2; slices, 9; slice thickness, 4 mm with no intersection gap; field of view, 230× 230 mm²; matrix, 128 × 128; sense, 4. The mean kurtosis (MK) and mean diffusion (MD) values were measured for the cortex and medulla for DKI model. For DKI and DTI, respiratory-triggered acquisition was used to weaken the impact of respiratory motion. Three sections nearest to the renal hilum were selected for region of interest (ROI) analysis. For each selected section, three ellipsoid ROIs of approximately 10-15 pixels were placed in the medulla, and an ROI of 80-120 pixels was manually delineated to cover the renal cortex. The left and right kidney ROIs were averaged for each subject for the cortex and medulla after excluding significant left and right differences. The parameters of cortex and medulla were compared between two groups by using Student t-test, performed with SPSS 17.0 software (SPSS Inc., Chicago, IL, USA) and results with P values less than 0.05 were considered statistical significant. Table 1 summerizes mean values of DKI and DTI parameters between healthy control group and patient group with diabetes nephropathy. Mean ADC, MD values in the cortex and mean FA, MD, MK values in medulla were lower in patients with DN than healthy controls. This result was similar to the previous study². There were no significant differences of mean FA and MK values in the cortex and medullar ADC values between two groups. The reduction of MD and MK values in the cortex and medulla was significant. This might indicate that the parameters from DKI was more sensitive and there were diffusion changes in both cortex and medulla of patients with DN. Figure 1 illustrates image examples of MD, MK, ADC and FA of kidneys in a patient with DN and a healthy volunteer.DKI was able to reflect diffusion information with higher sensitivity compared to DTI. DKI can be used for detecting renal changes in diabetes nephropathy of patients with diabetics, and monitoring renal function noninvasively.