Brains are inherently continuous structures. Yet, in the field of diffusion MRI, we have to rely on voxelized representations of the brain. As a result, the diffusion MRI community often uses the acquired data to model the (micro-)structure of the brain in independent voxels. This means valuable spatial correlations might be lost. In our previous work we have looked into implicit neural representations to make use of these correlations, which produced some nice results. There are however many more ways to approach this.
Are you interested in the brain? Do you want to create innovative ways of using MRI data to make continuous representations? Send us an email!
Maxime Chamberland
Tom Hendriks