Usually you initialize nonlinear with linear first. Are you doing that?
Peace,
Matt.
On 12/7/17, 1:34 PM, "FSL - FMRIB's Software Library on behalf of DJP"
<[log in to unmask] on behalf of [log in to unmask]> wrote:
>Hi Everyone,
>
>I have some multiband epi data. I used topup to correct the distortion in
>the data and that has worked well. However, for about 5 (out of 30) Ss,
>the spatial normalization fails miserably. For spatial normalization, I
>am using non-linear transformation and I notice with linear those 5 Ss
>register quite well.
>
>Here is my question: Should I just use linear transformation for those 5
>Ss? If so, should I also use linear transformation for the remaining 25
>Ss so that my pipeline is consistent across subjects (upon visual
>inspection, their registration look good with linear or non-linear, but I
>understand that non-linear is more optimal).
>
>Many thanks for your time,
>DJP
|