Hi Patrick,
It is likely that this is due to partial-volume effects - e.g. transformations moving the edge of the mask into a voxel, which results in voxels with fractional values. You can use fslmaths to threshold and rebinarise the output mask, a threshold of 0.5 would be a good starting point if you are looking to preserve total mask volume ( fslstats can be used to check this ).
Hope this helps,
Kind Regards
Matthew
> Hi experts
>
> So I'm currently working on a problem and I want to run my process by you and ask a question.
>
> We have two sets of T1 scans (month 1 and 6) for each subject which we used for MRS voxel placement. We have an in house program which produces a square binary mask showing the MRS voxel placement and we want to numerically compare the voxel placement at month 1 and 6 to see if we are extracting data from the same brain regions.
>
> My process so far has been to register the 6 month scan to 1 month scan using flirt (the registrations are fine). At that point i used the output matrix of that registration to transform the binary mask at 6 month to the 1 month scan using:
>
> flirt -in 6_month_mask -ref 1_month_mask -out transform_mask -init matrix -applyxfm
>
> What I've found is that the number of voxels in the 6 month binary mask increases by approximately 23% after applying the above transform (averaged over 30 subjects).
>
> So I have a two part question:
> 1. Does anyone know why the mask would change this much?
> 2. Is there anyway to determine the dimensions of the mask (its a simple square box) before and after transform to ensure that the proportions are still correct.
|