Optimizing protocols for white matter tractography in animal models of genetic AD risk

Abstract

AbstractBackgroundAlzheimer's disease (AD) has a tremendous effect on the brain structural and functional integrity. As previous research has shown, AD impacts gray matter areas, but also neuronal tracts connecting them [1]. One of the ways this neuronal impact can be evaluated is through the combined use of diffusion imaging and tractography modeling. However, due to the complex nature of 3D modeling, it is important to evaluate not only the tractography models, but also the imaging protocols. We have evaluated several diffusion acquisition schemes in a mouse model expressing the APOE4 genotype, associated with genetic risk for AD. We have focused on fimbria, in order to recommend optimized diffusion imaging protocols in a larger cohort of mice.MethodsWe acquired diffusion images at 50 µm3 isotropic with 46 different directions and an 8‐fold compressed sensing acceleration factor for fixed brain specimens using high‐field MRI. We used six different b‐value variations on this acquisition scheme, including single, two, and three‐shells acquisitions. The diffusion signal was reconstructed using a Q ball Constant Solid Angle reconstruction[2], and tracts were built using a deterministic maximum direction getter. The smaller tracts were pruned, and models were evaluated using Linear fascicle evaluation (LiFE) [3] in regions including the fimbria (Fig.1).ResultWe ranked the quality of the tractography models according to the number of tracts and their average size in the fimbria, as well as the correspondence between the diffusion model and predicted tract results using Life error evaluation (Fig.2). Our results favored Acq 1 (single shell b = 4000s/mm2) which had the most consistent and longest tracks, and lowest errors, followed by Acq 3 (three shells, b = 2000, 3000, 4000), and 5 (double‐shell, b = 2000, 4000).ConclusionWhether it was for single, dual or triple shell, the acquisitions with the greatest model error were those with at least one lower b‐value. This would indicate that for tractography models, and despite the lower signal noise levels, the use of low b‐values lead to tract distortions. Other considerations, such as reducing bias should also be introduced so as to better estimate diffusion scalar parameters.