Automated characterization of the gray matter white matter distribution demonstrates age-related decline

Neuroimaging techniques offer valuable insights into the structural characteristics of the brain. For example, a salient feature of the cerebrum is the distinct transition of voxel intensity at the interface of gray matter (GM) and white matter (WM). Leveraging the inherent difference in tissue composition - lower GM and higher WM signal on T1-weighted (T1W) magnetic resonance imaging (MRI) - we introduce a novel metric, distance between peaks of WM and GM intensities, which unlike GM/WM contrast does not rely on segmentation and demonstrate its efficacy in capturing age-related effects. A single 3D T1W whole brain MRI (MP-RAGE, 1 mm isotropic voxels) image was acquired at 3 Tesla from each of 178 healthy participants (18-91 years (54.78 ± 21.37), 103 Female) between 2019 and 2023. Before peak differentiation calculation, non-brain tissue was removed, and voxel intensities were corrected for RF coil profile. We define peak differentiation as the difference between means of two Gaussians fitted to the T1W voxel intensities, scaled by their common standard deviation. Age dependence of peak differentiation is examined using a linear model with biological sex included as a covariate. Similar analysis was performed by limiting to voxels within 5 mm of the Freesurfer-defined cerebral gray matter-white matter interface (GWI) in six lobes: orbitofrontal, frontal, occipital, temporal, parietal, and cingulate. Reduced whole-brain peak differentiation was associated with older age ([Formula: see text] = - 0.012, 95% CI [-0.0132, -0.0105], p < 0.001), indicating a decline in WM-GM distribution sharpness with advancing age. Each GWI subregional peak differentiation exhibited the same pattern of decline with age, but differed in magnitude region (p < 0.001). The cingulate region exhibited the sharpest decline with age ([Formula: see text] = -0.014, 95% CI [ - 0.0160, - 0.0126], p < 0.001). We demonstrate feasibility of an automated whole brain peak differentiation measurement to characterize brain aging in a group of healthy individuals, with minimal reliance on segmentation or other processing, with implications for understanding normal trajectories and identifying pathological deviations from expected aging processes.