Hi - theoretically bedpostx is designed to suppress fibre orientations that are not supported by the data. so even if you ask for 3 fibres you may actually only have 1 or 2 with a significant volume fraction. That said, in practice you may be able to converge faster to sampling the posterior distribution if you use the simpler model to start with.
Cheers
Saad
> On 30 Mar 2018, at 08:06, Ryan Hyon <[log in to unmask]> wrote:
>
> Hi all,
>
> I am working with a 32-direction dataset. I'm wondering if the default (3 fibers modeled) is too ambitious for a dataset with only 32 directions.
>
> I've checked to make sure my V1 eigenvectors look fine from dtifit's output, and I've successfully run bedpostx with the default -n 3 option. However, I'm getting low waytotals from using probtrackx. I'm wondering if modeling only 1 or 2 fibers could improve my results. Or would it have no effect at all?
>
> Thank you!
|