dcan-labs / infant-abcd-bids-pipeline Goto Github PK

Currently the pipeline requires intended-for fields to be added to fmap jsons to perform distortion correction on functional data. However, the pipeline does NOT require intended-for fields to perform DC on anat data (in PreFreeSurfer): it will find the fmaps and perform DC automatically without them. This should be documented

Hi everyone in DCAN-Lab!

Thank you for your contribution to simplifying the preprocessing progress of infant brain imaging data!
As a neophyte in the realm of brain imaging technology, I am immensely intrigued by the cerebral maturation of fetuses and neonates.After familiarizing myself with your infant-abcd-bids-pipeline, its convenience and advancement in preprocessing the images have made a profound impression to me.

Recently, a dataset consisting of images whose subjects' age were between 0-5 years has been collected by chance, prompting thoughts of utilizing the pipeline.Before preprocessing the entire dataset, a single-subject test dataset was utilized to evaluate it. However, it takes nearly 20 hours to run PreFreeSurfer on docker on Linux using 10 cores,and it appears that an error occurred during the PreFreeSurfer stage.Here are the screenshots depicting the errors:

Several endeavors have been undertaken to ascertain the root cause of the error, yet all of them have proven futile.:
1.reading PreFreeSurfer.err
2.rule out the possibility of the test dataset
3.add '--anat-only' or '--check-outputs-only' parameter

After encountering a bug, I proceeded to read the 'PreFreeSurfer.err' file that was generated in /subject/ses/logs/PreFreeSurfer, here is the '.err' report (in .txt format)

PreFreeSurfer_err.txt
PreFreeSurfer_out.txt

According to the .err report,I turned to the 'dcan-infant-pipeline' and tried to find where it goes wrong.At last,I found that something seems missing in dcan-infant-pipeline/PreFreeSurfer/scripts/T2wToT1wDistortionCorrectAndReg.sh. during my preprocessing progress.The wrong place was in line 234,likely saying that it unable to find TopupPreprocessingAll.sh or FieldMap workingdir

I am unsure of where to proceed, so I attempted a second trial to eliminate the dataset as a potential factor.

In order to retest the validity of the test-dataset,I have used bids-validator and it passed the validation.Due to its error report,I check the fmap and func files and their .json file in the dataset,it seems that nothing was wrong (Or maybe some particular parameters are needed in the .json file?I am not sure).Both modalities are present in the dataset.Then I attempted to apply the same pipeline to a different dataset which is compatible with fmriprep, it failed at the same stage once more.

To eliminate the possibility of func and fmap, I utilized the identical dataset, removed its func and fmap files, and appended the '--anat-only' parameter to solely execute anat. Despite generating additional files, this approach still proved unsuccessful. Presented below are the encountered errors:

Above are my bugs.I am not sure where it goes wrong,so some help of developers are necessary.
Here are some of my guess：

1.Some parameters are missing or wrong in func or fmap files or .json [such as intended for or any other things] (but it can pass bids validator,and tried add --anat-only parameter to exclude the possibility but failed,and tried another dataset)
I think this reason may be the root of the problem, but I am uncertain as to where to make revisions.

2.Due to this issue

I am not sure why it produces, but whenever I execute this pipeline, it consistently arises.

3.Some reasons I don't know

And my pipeline version is : v0.0.22 (latest version has been tried on Windows Docker Desktop,and failed at the same step[T2wToT1wDistortionCorrectAndReg.sh,line234] again)

One more question is that it takes too much time to run the pipeline. It appears that when running PreFreeSurfer on Linux Docker with 10 cores, it takes nearly 20 hours (and then report bugs).I am not sure why it causes that,and can I increase the speed?

Thank you for your time and attention!!Wish a solution can be made to run the pipeline fluently.I really appreciate your dedication to advancing this field and making it a better place.Your answer means a lot to me.

Kind Regards!
Zhao

Hello Developers,
In the latest version of the container there is an option to pass user generated aseg file to the pipeline. I was trying that option and it did not work correctly. On probing the code it seem that the aseg file is only copied as aseg_acpc.nii.gz without any transformation applied. The supplied aseg file is in the same space as the raw T1 and T2 images passed to the container. Since the T1 and T2 undergo may transformations, the aseg files should also go through the same transformations to be consistent with the processed T1 and T2 before being passed to the next stage. Your inputs to address this will be greatly appreciated.
Thank you,
Dhruman

The description of this flag does not clearly indicate that the input should be in bpm and not in Hz (as Hz would be the usual choice for a band-stop filter, this should be clarified)
--bandstop LOWER UPPER
parameters for motion regressor band-stop filter. It
is recommended for the boundaries to match the inter-
quartile range for participant group respiratory rate
(bpm), or to match bids physio data directly [3].
These parameters are highly recommended for data
acquired with a frequency of approx. 1 Hz or more
(TR<=1.0). Default: no filter

Currently, the BOLD images used for each run in the executive summary are slices 0.4, 0.5, and 0.6 from the run. If these slices have motion, then the "Task in T1: and "T1 in Task" images will show banding, which is an immediate QC fail even though the run may still be usable. The BOLD image used should therefore be an average of the BOLD images post-motion correction

User can pass in all arguments no matter what stages are asked to run. It may not be obvious that not all stages use all of the arguments, so they might expect different output even though the stage(s) that use the argument he passed in doesn't use it. Add a table that shows what args are passed to what stages.

The run commands live under the app folder, and it's very confusing. I didn't think to look there for quite awhile

https://github.com/DCAN-Labs/dcan-infant-pipeline