Dear Calvin,
>
> Thanks for the response. Thresholding does not seem to be the issue though. I am using a T1 image to warp to standard space. I believe the issue is that my patient T1 scans have lesions, while the reference image does not and so the warp is correcting for the differences in intensities in those voxels by essentially shrinking them.
That sounds spot on.
>
> When I apply the transformation matrix I get from that process to my lesion masks, I'm guessing the same thing is happening.
>
> Is there a way to mask out the lesions on the T1 images itself by reducing the intensity contrast so FNIRT does not try to remove them?
Yes, you need to specify an —inmask, which should be a binary mask with the value one for every voxel that should be included in the registration, and zero for those that should not.
The easiest is to load your patient T1 scan into fslview and manually draw a mask that covers the lesion/lesions. Save it to (for example) my_mask and then “invert it” (i.e. turn ones to zero and vice versa) with
fslmaths my_mask -sub 1 -mul -1 my_inverted_mask
Draw the mask with a little bit of margin (1-2 voxels).
Finally add
—inmask=my_inverted_mask
to the fnirt command line.
Good luck Jesper
>
> Thanks,
> Calvin
|