Image misalignment and failure in spatial inhomogeneity correction

[ParsaNilchian]

Hi all,

we are encountering a few issues when trying to preprocess our 7T mouse fMRI data collected on a Bruker scanner.

1. When attempting the inhomogeneity correction the preprocessing crashes (crash files attached).
2. When I turn off the inhomogeneity correction with the rabies call below, the pipeline runs but the data looks misaligned and the SNR seems really high.

I have sent you all QC reports, rabies log (without inhomogeneity correction), images, and crash files in the preprocess_QC_report folder separately. I have no experience running this pipeline before so I do not know if something might be off with our data or if I am calling the pipeline incorrectly. I am really grateful for your help with this issue. Thanks!

#RABIES call for final version that ran:
singularity run
-B ${data_input}:/input_BIDS:ro
-B ${preproc_output}:/preprocess_outputs
-B ${TMPDIR}:${TMPDIR}
${container} -p MultiProc preprocess /input_BIDS /preprocess_outputs
--apply_STC
--TR ${TR}s
--anat_inho_cor method=disable,otsu_thresh=2,multiotsu=false
--bold_inho_cor method=disable,otsu_thresh=2,multiotsu=false
--commonspace_reg masking=true,brain_extraction=false,template_registration=SyN,fast_commonspace=false

[Gab-D-G]

Hi @ParsaNilchian, I was not able to download the .zip folder that you shared by email, and without accessing any log file or example QC images I cannot be of any help. Can you attach the log.txt file for the run where you get a crash, and some example registration QC images for the case where you get registration failures? You don't need to upload the entire QC folder, only a few representative examples of you issue.

[ParsaNilchian]

Thanks for your response. The only log file that I currently have is the one from the last run without the spatial inhomogeneity correction (attached below). It does not allow me to attach any other files due to the formats. I tried sharing the complete folder a different way now - let me know if it worked this time.

rabies_preprocess.log

[Gab-D-G]

Hi Parsa,

I had a look at the QC files you shared, and since you are not running any anatomical inhomogeneity correction, every stage of the registration pipeline fails. The EPI inhomogeneity correction and registration requires that anatomical preprocessing was succesful.

Can you provide me the log and QC files that you obtain when applying inhomogeneity correction?

Also, I can make 2 observations regarding your EPI image from the QC report below:

1. It looks like you image orientation and axes are wrong (cortical slice upside-down, in the wrong axis, etc...). You should check the metadata of your images to make sure that image orientation follows proper conventions, otherwise your images won't align properly with the mouse template in RABIES. You can check this examples here https://rabies.readthedocs.io/en/latest/preproc_QC.html as a reference for proper RAS file orientation.
2. Distortions appear to be very heavy in your images, and there is little brain contrast. It will be difficult to carry registration on those images, and I would suspect that it will be difficult to observe proper BOLD signal at the analysis stage. I strongly suggest you revise your MRI protocols, this will simplify your fMRI experiments greatly in the future. The example from the RABIES documentation https://rabies.readthedocs.io/en/latest/preproc_QC.html where also acquired on a 7T, using a cryoprobe, so scanner strength should not be the limiting factor. You may want to consult these references for some guidance regarding scanning protocols: https://www.biorxiv.org/content/10.1101/2025.05.19.653309v1.abstract, https://www.nature.com/articles/s41593-023-01286-8 (this is in rats though)

[ParsaNilchian]

Thanks for your responses and sorry for the delay.

1. Reorientation: From what I understood our images (anat and func) may have been in LPI so I used fsl to reorient them as shown below. I noticed however that in fsleyes even our anatomical images look slightly "blurry" which I am not sure might be related to the contrast issue you raised:

Before (last submission):
fslhd sub-Mouse1_ses-1_run-02_T2w.nii.gz | grep qform
qform_name Scanner Anat
qform_code 1
qform_xorient Right-to-Left
qform_yorient Posterior-to-Anterior
qform_zorient Superior-to-Inferior

Reoriented (new submission):
fslhd sub-Mouse1_ses-1_run-02_T2w.nii.gz | grep qform
qform_name Scanner Anat
qform_code 1
qform_xorient Left-to-Right
qform_yorient Posterior-to-Anterior
qform_zorient Inferior-to-Superior

2. Distortions: I have raised this issue with our team and forwarded the papers. Thank you for your suggestions - I am not directly involved with the data collection but we wanted to see if we can get the pipeline potentially working with the data we currently have and then revise/improve for future data collection.

3. New QC and log files with anat inhomogeneity correction after reorientation:

My singularity call:

singularity run
-B ${data_input}:/input_BIDS:ro
-B ${preproc_output}:/preprocess_outputs
-B ${TMPDIR}:${TMPDIR}
${container} -p MultiProc preprocess /input_BIDS /preprocess_outputs
--apply_STC
--TR ${TR}
--commonspace_reg masking=true,brain_extraction=false,template_registration=SyN,fast_commonspace=false

Log file:
rabies_preprocess.log

QC folder: uploaded to same box folder that you have access to.

Thanks again!

[gdevenyi]

Before (last submission):
fslhd sub-Mouse1_ses-1_run-02_T2w.nii.gz | grep qform
qform_name Scanner Anat
qform_code 1
qform_xorient Right-to-Left
qform_yorient Posterior-to-Anterior
qform_zorient Superior-to-Inferior

Reoriented (new submission):
fslhd sub-Mouse1_ses-1_run-02_T2w.nii.gz | grep qform
qform_name Scanner Anat
qform_code 1
qform_xorient Left-to-Right
qform_yorient Posterior-to-Anterior
qform_zorient Inferior-to-Superior

To add a clarification here, the ultimate arbiter of correct orientation is viewing the data in ITK-snap. The anatomical directions labelled (Left, Right, Superior, Inferior) in ITK snap should correspond to the correct directions in the image. The NIFTI metadata can say whatever it says, but if the image doesn't show correctly there, the orientation is incorrect.

[ParsaNilchian]

Thank you for the clarification - that makes sense. I did look at the images, and if I am not mistaking, it looked correct this time. Will wait to hear back on QC report and log file.