[MINC-users] mincreeample transformation problem

[Vladimir S. FONOV]

Hello,

I think your attachment didn't go through - can you post it online 
somewhere.

Also, can you post output of mincheader on both T1w and EPI files that 
cause problems?

On 13-04-02 12:10 PM, Colin Shaun Hawco, M wrote:
> Hello all,
>
> I am having a problem with mincresample for a single participant in my study. I am attempting to normalize EPI data into MNI space. My basic pipeline is something like this:
>
> calculate transform, high-res T1 to MNi space (mincANTS)
> calculate transform, EPI to T1 (to account for any movement between scans, mritoself)
> concat transform
>
> I then apply the transform, using mincresample. Because I am analyzing my data in SPM, I am transforming it into the same space/voxel size as the SPM-EPI template (this has several advantages for me). So my command is, paraphrased:
>
> mincresample -like EPI_templace.mnc -transform epi2mni.xfm epi.mnc epi_normalized.mnc
>
> This works very well on all my data sets, save one.
>
> In one case, i was unable to do the T1 scan due to time constraints (technical problems delayed the scan). However, I was able to get a T1 scan from a different day from this participant. But my pipline fails, and when I apply the transform, the EPI image is shifted, such that the actual data is being partially moved outside the range of the file. I attached an image to show what I mean.
>
> Upon further investigation, it seems the transforms themselves are not to blame. the problem is when I run mincresample and use the EPI_template.mnc. So, for example, if I do this, I get what looks like a good reg, but the data has the wrong voxel size and dimensions:
>
> mincresample -like epi_file.mnc -transform epi2mni.xfm epi_file.mnc epi_normalized.mnc
>
> In this case, when I do -like the file I am transforming, the data is not shifted out of range, and appears to overlay on the MNI152 brain. But as soon as I do -like EPI_template.mnc, things go wrong, and the data is majorly shifted in the range of the file, and the registration with the MNI brain is garbage.
>
> This was a complex and challenging study, with only 16 usable data sets (counting this one), so i would really like to recover this data. Does anyone have any idea why I could be having this problem? I supect there is some confusion over transforming the origin, but I am not sophisticated with minctools enough to figure it out. Also, I'd like to emphasize, the pipeline works on all 15 other participants.
>
> I really appreciate any help.


-- 
Best regards,

  Vladimir S. FONOV ~ vladimir.fonov <at> gmail.com