Running MRIQC group w/o bids_dir

[Steven]

What are the BIDS validation issues?

[mdemi]

Working on the fixes, but approx 250 subjects have .json and/or .nii.gz files that are empty. Which causes an MRIQC decoding error

[Steven]

One workaround to try is making a fake BIDS directory containing just symlinks/shortcuts with subject data you want to analyze, and then running MRIQC on that. That way it will not be taking up much extra space on your system (presumably a HPC if you have 10k subjects). Just make sure to bind the original and symlinked drives in Singularity.

[mdemi]

Thanks for the recommendation, Steven! When you say “just make sure to bind the original and symlinked drives in singularity”, what do you mean by that? Sorry, semi-singularity newb.

[Steven]

singularity run -e -B ${original_bids_dir},${fake_bids_dir}

Basically, you will input your fake directory to be analyzed, but Singularity needs to be able to find the original data (i.e. the target of the shortcuts), so both drives need to be available.

[effigies]

Edit: I’m a bad reader and should feel bad. Sorry for the noise.

[mdemi]

Created the symbolic links using:
for folder in $(find ${input}/${sub}/ses-${ses} -mindepth 1 -maxdepth 1 -type d -not -name "dwi" ) ; do ln -s "$folder" "${sub}/ses-${ses}/"; done; done

Got an error similar to when using the --participant-labels, e.g., mriqc: error: Querying BIDS dataset at </bids_dir> got an empty result. Please, check out your currently set filters. Retrying atm by removing the --session_id

Perhaps something wrong with my singularity command?

singularity run --cleanenv \
-B ${in_dir}:${bids_dir} \
-B ${bids_dir}:/bids_dir \
-B ${mriqc_in}:/output_dir \
-B ${work_dir}/work_dir/mriqc_v23_1_0_grp:/wd \
${sif_img} \
/bids_dir /output_dir group \
--session-id ses-2YearFollowUpYArm1 \
-vv \
-w /wd

[Steven]

Hi @mdemi,

Can you try running singularity shell into your container (using the same -e and -B arguments) and do ls /bids_dir and see if your BIDS folder shows up as expected?

Best,
Steven

[mdemi]

Thanks – tried it and I get a list of subject directories as expected…

singularity shell --cleanenv \
-B ${in_dir}:${bids_dir} \
-B ${bids_dir}:/bids_dir \
-B ${mriqc_in}:/output_dir \
-B ${work_dir}/work_dir/mriqc_v23_1_0_grp:/wd \
${sif_img}

However, ls /bids_dir/sub-EXAMPLE/ses-EXAMPLE/anat shows no files. Not sure if that is expected.

[Steven]

Which of your names arguments are the original and fake bids directories?

[mdemi]

@Steven – The original of bids subject directories: in_dir. The fake sym linked of subject direcotirs: bids_dir

[Steven]

Ah so here’s the thing. Your symlinks may be pointing to something with full path of your original bids directory, BUT in the container, the path is renamed! Try not renaming your directories when you bind in your singularity command.

[mdemi]

Thanks, @Steven ! So, do you recommend something like the below?

singularity run --cleanenv \
-B ${in_dir}:${bids_dir} \
-B ${mriqc_in}:/output_dir \
-B ${work_dir}/work_dir/mriqc_v23_1_0_grp:/wd \
${sif_img} \
/home/mdemiden/bids_dir /output_dir group \
-vv \
-w /wd

[Steven]

Remove the colon statements in your -B arguments, and replace /output_dir and /wd with their paths

[Steven]

singularity run --cleanenv \
-B ${real_bids_dir},${fake_bids_dir},${mriqc_dir},${work_dir} \
${sif_img} \
${fake_bids_dir} ${mriqc_dir} group \
-vv \
-w ${work_dir}

[Steven]

And if all of your important drives are all subdirectories of a more parent drive, you can just only bind the parent drive that contains all of them.

[Yunhong_Wang]

Hi @Steven

My question is that can I get both participant and group level reports while running mriqc?
For example, I specify both participant and group argument like this:

singularity run \
        --cleanenv \
        -B ${BIDS_DIR}:/data:ro \
        -B ${OUTPUT_DIR}:/output \
        -B ${WORK_DIR}:/work \
    ${MRIQC} /data /output \
    participant group \
    --participant-label "$subjectID" \
    -w /work \
    -vv \
    --n_cpus 4 \
    --mem 16000 \
    --write-graph \
    --fd_thres 0.2 \
    --float32 \

Best,
Yunhong

[Steven]

Hi @Yunhong_Wang,

No, you cannot do participant and group at the same time.

Best,
Steven

[Yunhong_Wang]

Hi @Steven

I run the mriqc participant level first and then run group level. However, the group report does not present the interactive figure for me. It’s weird.

The group_bold.html is here:

And the whole output of group analysis:

Best,
Yunhong