Automated Surface-Based Segmentation of Deep Gray Matter Regions Based on Diffusion Tensor Images Reveals Unique Age Trajectories Over the Healthy Lifespan

Abstract

Many studies have demonstrated unique trajectories of deep gray matter (GM) volumes over development and aging, suggesting but not measuring microstructural alterations over the lifespan. Only a few studies have measured diffusion tensor imaging (DTI) parameters in deep GM or reported these values across a wide age range in a large cohort. To enable efficient DTI studies of deep GM in large cohorts without the need of T1-weighted images, an automated segmentation technique is proposed here that works solely on parametric maps calculated from DTI. The algorithm segments the globus pallidus, striatum, thalamus, hippocampus and amygdala per hemisphere by deforming 3D models of these structures to their boundaries visible on the contrast provided by diffusion tensor maps and images alone. This new DTI-only method is compared against standard T1-weighted image segmentation for (i) 1.25 mm isotropic diffusion data from the Human Connectome Project (HCP) test-retest cohort (n=44) and (ii) 1.5 mm isotropic test-retest diffusion data from a local normative study (n=24). Dice coefficients of voxel overlap between methods in the HCP test-retest cohort were high (>0.7) for 7 of 10 structures, but were low for the left globus pallidus (0.54) and left/right amygdala (0.67, 0.69). The proposed DTI-only segmentation qualitatively appeared more accurate and yielded smaller volumes than T1w for 8/10 structures in both cohorts, with the exception of the globus pallidus which showed larger volumes in the HCP test-retest data but lower volumes in the local normative study data. The DTI-only segmentation method was then applied to two local single site development/aging lifespan cohorts (cohort 1: n=365 5-90 years, cohort 2: n= 164 5-74 years) to assess age changes in volume, fractional anisotropy (FA) and mean diffusivity (MD). In both cohorts, MD trajectories were quadratic for all five structures, decreasing slightly and then increasing after [~]30-35 years. In cohort #1, FA trajectories remained flat from 5 to [~]25 years and then started to decrease for the globus pallidus and hippocampus and over 5 to 90 years, FA decreased linearly for amygdala, increased linearly for striatum, and remained constant for the thalamus. In the second cohort, using an alternate acquisition protocol, the FA trajectories of all 5 structures across all ages were similar, except for the globus pallidus and thalamus which both increased in value from 5 [~] 20 years and likely reflect differences in acquisition details. Notably, the development and aging trajectories for DTI were distinct from those of the deep GM volumes. The proposed automated deep GM segmentation method on DTI-only will facilitate the analysis of deep GM DTI (currently ignored in nearly all studies despite the data there within the field-of-view) and will be advantageous particularly for studies that do not have a T1-weighted scan, as in many clinical populations.