Dear Zak,
>
> I needed to create warp files to apply to data to create non-linear transforms quickly. My impression is that this is the appropriate sequence of transforms in order to go from DTI space to the most accurate standard space:
>
> DTI --(flirt)--> T1 --(flirt)--> MNI --(fnirt)--> MNI non-linear
>
> However, the fnirt command allows for the inclusion of a linear transform step, --aff, so I did the following:
>
> DTI --(flirt)--> T1 --(fnirt with --aff T12MNI_mat)--> MNI non-linear
>
> And I used fnirt's --jout command to create a warp for quicker use with the same subjects in the future. My question: given that I used the --aff command in fnirt in the same step used to create the --jout, does the --jout warp now include the affine step or not?
>
> I ask because I was getting wildly incorrect transforms when using applywarp (with the above described--jout warp) AND the --premat T12MNI_mat. However, when I ONLY use the --jout warp on my T1 data it looks fine.
yes, the question is a little confusing. The --jout flag will give you a Jacobian field, i.e. a scalar field describing the local stretches/compressions. You should not use that as input for applywarp, which expects a warpfield. For that purpose you should use the output from the --cout flag, which contains the spline coefficients for a warp field. The --cout field also contains the linear transform you supplied with the --aff flag, so if you also use that transform it will be applied twice. And that would not be good.
Jesper
>
> Please advise.
>
> Thanks,
> Zak
|