Type 2 diabetes mellitus (T2DM) is a prevalent disease and have considerably higher risk of developing cognitive impairment. Some Diffusion Tensor Imaging (DTI) studies have focused on the white-matter (WM) alterations in brain. Widespread WM integrity changes were observed in several important regions and linked with neuropsychological function.^1,2 Diffusional kurtosis imaging (DKI) is an extension of DTI by taking non-Gaussian diffusion behavior into consideration, allowing more comprehensive characterization of diffusion in tissues. From a DKI scanning, both standard DTI metrics, as well as metrics reflecting the diffusional kurtosis can be computed. This study aims at investigating brain microstructural changes in WM of T2DM patients using DKI, and making a comparison with the diffusion tensor metrics.

Subjects: With approval of the Institutional Review Board, 30 T2DM patients (DM group; based on diagnostic criteria of American Diabetes Association; 60.6±6.0 years old; 17 females) and 28 healthy controls (HC group; 58.5±6.2 years old; 18 females) were recruited. A battery of neuropsychological tests (Montreal Cognitive Assessment, Mini-Mental State Examination, Trail Making Tests, Auditory Verbal Learning Test, Hachinski test, and Activity of Daily Living test) were performed at first. Plasma fasting/postprandial glucose and Glycated hemoglobin A1c (HbA1c) were also recorded. Imaging: On a 3 Tesla MRI scanner (Discovery MR750, GE Health Care, Waukesha, Wisconsin, USA) with a 32-channel head coil, axial DKI images were obtained using a single-echo echo planar imaging sequence (TR/TE = 5000/98ms, FOV = 24×24cm², matrix size = 128×128, slice thickness = 4 mm, slice spacing = 0, slice number = 46, NEX = 1, b-values = 0, 1250, and 2500 s/mm², diffusion encoding directions = 25 for each nonzero b-values). Data processing: The FMRIB Software Library (FSL) with tract-based spatial statistics (TBSS) ^3,4 was utilized to analyze the whole-brain diffusion metrics, including fractional anisotropy (FA), mean kurtosis (MK), axial kurtosis (K_∥), and radial kurtosis (K_⊥), and compare group differences using a standard atlas. Additionally, based on the skeleton created by FSL, regional FA and MK values were also evaluated on the JHU WM tractography atlas, by which the entire WM was parceled into 48 ROIs. MK changes were correlated with neuropsychological scores, glycosylated hemoglobin A1c level, and disease duration. Statistical analyses were carried out using SPSS software (SPSS Inc., Chicago, IL).

In the whole-brain TBSS analysis, the T2DM patients exhibited abnormalities in 29.6% and 35.4% of WM regions as measured by FA and MK, when compared to healthy controls (Fig.1). A reduction in MK of the DM group was contributed more by the decreased K_⊥ than the decreased K_∥ (abnormalities in 26.0% and 10.5% of WM regions respectively). In the atlas-based analysis, MK detected more ROIs (28/48) with microstructural changes in WM than FA did (13/48), such as pontine crossing tract (PCT), superior longitudinal fasciculus, and corpus callosum (Fig.2). Fig.3 shows the box plots of diffusion tensor and kurtosis metrics in PCT as an example between the DM and HC groups. No detectable changes in FA and MD are seen. In contrast, there are changes in MK and K_⊥ detected within this crossing tract. Among T2DM patients, MK values negatively correlated with disease duration in the genu of corpus callosum (R = -0.512), and positively correlated with neuropsychological scores in the cingulum (hippocampus) (R = 0.466) (with age, gender as the covariates, Fig.4).

Our results suggest that the MK is more sensitive than FA when detecting WM abnormalities in diabetic brain, especially when detecting changes within crossing fibers. The reduced DKI parameters indicated decreased structural complexities ⁵ in the WM in T2DM patients compare to healthy controls. These alterations in brain may be caused by the compromised fiber tracts and neuron loss in some regions, such as the corona radiata, internal capsule, cingulum (hippocampus) and corpus callosum. These findings complement conventional DTI and provide more information to characterize and pinpoint brain microstructural changes in WM of T2DM patients, and potentially provide valuable information to study diabetic encephalopathy, including cognitive impairment. It suggests that DKI is potentially a valuable technique for investigating the diabetic brain.