Skip to content

weird inhomgeneity correction bold errors #475

Description

@mapepe18

Hello rabies team,

I've been testing several things with my dataset to experiment with many different preprocessing and confound correction conditions. So far, everything seems fine, but now I was testing one last preprocess step that I did not try until now (so basically you can assume that I tested the same preprocessing pipeline with different conditions on the same dataset, easily up to 8 or 9 different pipelines and everything seemed ok).

The current preprocessing I employed just added the despiking option. This is the command:

docker run --rm
-v C:\Users\UPVUSER\Desktop\trabajo_Manuel\atlas_normal:/atlas_normal 
-v C:\Users\UPVUSER\Desktop\trabajo_Manuel\bids2\rsm:/rsm 
-v D:\Nueva_carpeta\fmri_bad_not_use\prepro\rsm_prep_levi6:/rsm_prep_levi6 
ghcr.io/cobralab/rabies 
-p MultiProc --local_threads 6 --scale_min_memory 1.5 
preprocess /rsm /rsm_prep_levi6 --resampling_space both --bold_nativespace 
--apply_despiking --anat_inho_cor method=Rigid,otsu_thresh=2,multiotsu=false 
--anat_template /atlas_normal/anat_template0.nii.gz 
--brain_mask /atlas_normal/brain.nii.gz 
--WM_mask /atlas_normal/wm.nii.gz 
--CSF_mask /atlas_normal/csf.nii.gz 
--labels /atlas_normal/atlas_rsm_levi.nii.gz 
--vascular_mask /atlas_normal/csf.nii.gz 
--apply_STC --tpattern alt+z 
--bold2anat_coreg masking=true,brain_extraction=true,registration=SyN  ##Originally SyN, optionally use Rigid

Everything goes right until the inhomogeneity correction of the bolds to genereate the registration templates. Basically the results are weird, like part of the images are erases randomly, also the raw image also appears to have the background erased from the beginning.

I'm attaching three different examples (each example I attach the bad result I obtained now vs a result obtained with the same preprocessing minus the deskiping option)
this is how it looks:

Example 1:
With despiking:

Image

No despiking:

Image

Example 2:

with despiking:

Image

No despiking:

Image

Example 3:

With despiking

Image

No despinking:

Image

As you can see the results are weird, there are some cases (the majority) that look like example 3, but even that one is weird, as you compare with the no despinking one (the raw image has a big portion of the background erased).

I wanted to know I you have come across a situtation like this, I've seen some times there are errors more related to bad cropping but nothing like this, I could not find anything in the docs or issues referring to problems with despiking like this. Maybe you know what could be happening.

At this point it is seems clear to me that it is something related to the despiking, because that is the only difference in the preprocessing. Additionally I repeated the preprocessing with despiking 3 times and in all of them this happens. And it did not happen in any of the other preprocessing variations I tried that did not employ despiking but I would like to share it to see if anyones knows anout results like this.

By the way the inhomogenity correction in the anatomic aqucisition works fine, not problems there, just in the bolds.

As always thanks in advance

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions