Dear Jeff,
The process is as follows:
1 - choose one image to serve as the original reference (the "most average" image, although it shouldn't matter too much which)
2 - register each image to this chosen reference using first FLIRT and then FNIRT
3 - average the results of all of the resampled images
4 - repeat the process but use the result of the average image from step 3 as the reference image
You can repeat several times if it seems to get better, but we find that one repeat is usually enough for a good result.
If you want your resulting average template in a different orientation (e.g. in MNI space) then you can get this by doing a registration of your result to this other space and then combining that transformation with the transformation of each subject to the average in order to get a sharp template in the new space, effectively resampling each image once (with the combined transformations) directly to this new space. However, depending on how blurry the average is, you may be just as well off applying the transformation directly to the average (which is effectively doing two transformations).
I hope this helps.
All the best,
Mark
On 31 Jul 2012, at 02:01, Jeff Thompson wrote:
> Hello,
>
> In the FAQ on the FSL website there is a description of creating a study specific template image with FLIRT. Could someone that has used FNIRT before describe this process using FNIRT? Particularly creating a template image for beagle brains.
>
> Secondly should this template creation be done before skull stripping or after? ( I have manually skull stripped dog brains).
>
> I would like to use the resulting template in the freesurfer pipeline for cortical analysis.
>
> Any help would be much appreciated,
>
> -Jeff Thompson
>
|