Towards Learned Optimal q-Space Sampling in Diffusion MRI

Fiber tractography is an important tool of computational neuroscience that enables reconstructing the spatial connectivity and organization white matter brain. takes advantage diffusion Magnetic Resonance Imaging (dMRI) which allows measuring apparent diffusivity cerebral water along different directions. Unfortunately, collecting such data comes at price reduced resolution substantially elevated acquisition times, limits clinical applicability dMRI. This problem has been thus far addressed using two principal strategies. Most efforts have extended towards improving quality signal estimation for any, yet fixed sampling scheme (defined through choice diffusion-encoding gradients). On other hand, optimization over also proven to be effective. Inspired by previous results, present work consolidates above strategies into a unified framework, in carried out with respect both model design concurrently. The proposed solution offers substantial improvements as well accuracy ensuing analysis means fiber tractography. While proving optimality learned models would probably need more extensive evaluation, we nevertheless claim schemes can immediate use, offering way improve dMRI without necessity deploying neural network used their estimation. We comprehensive comparative based on Human Connectome Project data. Code designs available https://github.com/tomer196/Learned_dMRI.