Fast consensus optimisation in diffusion MRI modeling

This work explores how to obtain parameter maps from a diffusion MRI model under 30 seconds and how to leverage multiple plausible solutions into a consensus solution using a framework similar to MR fingerprinting. The method is tested with a two compartments model and a mean kurtosis model and compared with a standard optimisation solver. Visual improvements are shown on parameter maps, showing less implausible values than the standard method, but without the use of constraints or assumptions. Access to a distribution of plausible values allows to compute most likely value from this distribution in the case of parameter degeneracies.