White and gray matter diffusion-derived differences in type 2 diabetes mellitus: a free-water-corrected diffusion MRI...

Type 2 diabetes mellitus (T2DM) is frequently accompanied by complex brain microstructural abnormalities and cognitive decline. Conventional diffusion tensor imaging (DTI) metrics can reflect both tissue diffusion properties and extracellular free-water effects, complicating the interpretation of T2DM-related brain changes. We aimed to characterize free-water-corrected diffusion alterations in white matter (WM) and cortical gray matter (GM) and examine their associations with cognitive performance and fasting blood glucose.

We enrolled 48 patients with T2DM and 48 healthy controls (HCs). Diffusion measures included conventional fractional anisotropy (FA), mean diffusivity (MD), axial diffusivity (AD), and radial diffusivity (RD), together with free-water-corrected FA (FAt), free-water-corrected MD (MDt), free-water-corrected AD (ADt), and free-water-corrected RD (RDt), as well as the free-water fraction (FW), estimated using a free-water elimination (FWE) DTI model. Tract-based spatial statistics (TBSS) was used to evaluate all nine diffusion metrics in WM, whereas gray matter-based spatial statistics (GBSS) was used to evaluate six metrics in the cortical GM skeleton (AD, MD, RD, ADt, MDt, and RDt). Voxel-wise association analyses were performed using FMRIB Software Library (FSL) randomise based on a general linear model with nonparametric permutation testing to examine the relationships of imaging metrics with fasting blood glucose and Mini-Mental State Examination (MMSE) scores, while adjusting for age, sex, and total intracranial volume in the primary models; sensitivity models additionally included diagnostic group as a covariate.

Compared with HCs, patients with T2DM showed widespread diffusion abnormalities across the WM and GM skeletons, including lower FAt and higher MDt, ADt, and RDt. No skeleton-wide between-group difference in the estimated FW was detected. In the primary analyses, higher fasting blood glucose was associated with higher FW in selected WM regions and greater cortical GM diffusivity, whereas lower MMSE scores were associated with widespread WM and GM diffusion abnormalities.

T2DM-related diffusion abnormalities persisted after free-water correction, whereas no skeleton-wide between-group difference in the estimated FW was detected. Integrating FWE with complementary spatial analyses of WM and cortical GM may help characterize diffusion-derived brain phenotypes and their clinical correlates in T2DM.