t2 and functional scan from same animal collected during different sessions

[geowk]

Hello,

Is there any recommendation on how to use the structural t2 scan of a rat with its functional scan that wasn't collected during the same session?

For example:

1. animal was scanned with a t2 scan.
   2.pulled out of the scanner
2. put back into the scanner
3. a 2nd localizer performed, but no t2 scan
4. 2 functional scans were collected.

We'd like to use the t2 for preprocessing with RABIES however the t2 and functional scans from the same animal are misaligned. Is there a way to account for this and produce quality normalized images to a template? Without accounting for it, the resulting functional images in standard template space do not look good.

Thanks.

-g

[CMonnin]

Hello,

have you tried renaming the t2 scan to match the functional scans and "tricking" BIDS into thinking they are from the same session? This should allow the automatic alignment to proceed.

[Gab-D-G]

Hi @geowk, having the scans acquired in the same session increase the likelyhood of registration success, because images are already mostly aligned. However, it is certainly possible to achieve successful registration without that initial alignment. Have you tried reviewing what may be causing the issues in the case of your image (incomplete brain coverage, signal drop-off, etc....)? Have you tried different registration parameters as listed here https://rabies.readthedocs.io/en/latest/nested_docs/registration_troubleshoot.html?

[Gab-D-G]

If you share the preprocessing QC folder here, or some representative samples from it, I can have a look.

[geowk]

Hello,

Apologies for the delayed response. Thank you both @CMonnin and @Gab-D-G for your responses suggestions.

1. We've tried tricking RABIES and making a copy of the t2 and splitting the runs into 2 sessions: 1a session= t2 and first run-01, 1b = session t2_Copy and 3 runs that are aligned together, but misaligned to the original t2 and 1st run due to being pulled out of the scanner. That still didn't work. It worked, but the image quality whether I used the unbiased template or fast template was poor after preprocessing. Tried the different multiotsu=false vs true, but image quality was still poor after processing.

2. Here is the RABIES command line that we recently ran. Here the everything was treated as 1 session.

singularity run -B ${template_dir}/:${template_dir}/
-B ${rabies_dir}/rabies_0.5.1/${temp_dir}:/tmp/
-B ${rabies_dir}/rabies_0.5.1//${i}/:/input_BIDS:ro
-B ${rabies_dir}/rabies_0.5.1/${preproc_dir}:/preprocess_outputs/ ${rabies_dir}/rabies.sif
-p MultiProc --local_threads 70
preprocess /input_BIDS/ /preprocess_outputs
--anat_template ${template_dir}/WHS_SD_rat_atlas_v4.01_brain.nii.gz
--brain_mask ${template_dir}/WHS_SD_v2_brainmask_bin.nii.gz
--WM_mask ${template_dir}/wm.nii.gz
--CSF_mask ${template_dir}/vascular.nii.gz
--vascular_mask ${template_dir}/vascular.nii.gz
--labels ${template_dir}/WHS_SD_rat_atlas_v4.01.nii.gz
--commonspace_reg masking=true,brain_extraction=true,template_registration=SyN,fast_commonspace=true
--anat_autobox
--bold_autobox
--commonspace_resampling 0.3125x0.3125x0.3125
--anat_inho_cor method=SyN,otsu_thresh=2,multiotsu=false
--bold_inho_cor method=Rigid,otsu_thresh=2,multiotsu=false

3. 1st aligned run overlayed on the t2.
   

2nd misaligned run overlayed on the t2.

Here is the t2 inho report

Here is the bold inho report 1st run

epi2anat t2 with1st bold run

Here is the bold inho report 2nd run (collected after being pulled out of the scanner)

bold 3rd run

bold 4th run

epi2anat t2 with 2nd, 3rd, and 4th runs

Here is what the images look like after RABIES preprocessing:

Bold 1st run:

2nd run

3rd run

4th run

Any comments, suggestions, recommendations, would greatly be appreciated.

Thanks,

-g

[Gab-D-G]

Hello @geowk, the processing indeed failed for all 3 runs which were from a different session (the first BOLD run looks good though). We can see that the mask is not great at the BOLD inhomogeneity correction step, and that in turn EPI2anat fails. On this page https://rabies.readthedocs.io/en/latest/nested_docs/registration_troubleshoot.html there various suggestion for either failures. The multiotsu will not help here, you do not have very pronounced inhomogeneity in your EPIs.

We can't predict in that case what is the most computationally efficient way to correct the issue, but I can suggest you try the most robust set of options to increase success likelyhood (although the --bold_robust_inho_cor option will increase computational time):

--bold_robust_inho_cor apply=true,masking=false,brain_extraction=false,template_registration=SyN \
--bold_inho_cor method=SyN,otsu_thresh=2,multiotsu=false \
--bold2anat_coreg masking=true,brain_extraction=true,registration=SyN \

Basically, --bold_robust_inho_cor will create an EPI average and give it a mask from your commonspace template (the mask needs to be good). then for the --bold_inho_cor step, this EPI average is used to derive a mask, which becomes easier because the EPI average is similar to your data. This may allow to get an improve EPI mask, and finally for EPI2Anat you add masking=true,brain_extraction=true so that you are using this EPI mask to improve registration.

Let me know how this goes!