In this work we analyze the incidence of voxels with physically impossible model parameters, reconstructed from diffusion-weighted data that is acquired using different sampling schemes. Our results show that for cumulants up to order $$$4$$$ constrained least squares can be used to compute a reliable reconstruction of the cumulant expansion of the signal from realistic acquisitions, with spherical sampling producing fewer unsatisfied model constraints compared to space-filling sampling. Voxels where reconstruction is likely to fail are shown to be consistently localized near the white matter-gray matter interface and in deep brain structures.
This abstract and the presentation materials are available to members only; a login is required.