Skip to content

Latest commit

 

History

History
executable file
·
71 lines (57 loc) · 6.09 KB

README.md

File metadata and controls

executable file
·
71 lines (57 loc) · 6.09 KB
Raw

BIDS Preprocessing Pipeline

This repo contains scripts for minimal containerized preprocessing for Mackey Lab child data in BIDS. These scripts using containers, which you can learn more about here, and work on our longitudinal data. They operate in two chunks.

First script

The first chunk is run by the new_subj_first script. This does the below:

  • Convert to nifti format with Heudiconv, put into CBPD_bids directory (see here).
  • Fix TOPUP fieldmaps for diffusion (see here).
  • Assign IntendedFor field to TOPUP fieldmaps (see here).

Next, if a subject has fallen asleep or we need to discard some data, the BOLD niftis must be edited to reflect this after running the first script (i.e. if some number # of TRs are removed, for example, for sleeping), and subjects or runs that are bad or incomplete (total number of volumes < 130) will be documented in the CBPD Scanning Notes.

Note: Adding to the .bidsignore does not affect running of any downstream tools such as MRIQC, which will run on these subjects anyways. This simply tells the bids-validator to ignore these files when checking whether the folder is valid.

Check the MRI protocol notes (here) or CBPD Scanning Data (here) for whether a participant has fallen asleep, and at what time. If we didn't note when they fell asleep, mark the whole run to discard for sleeping. A script for discarding extra TRs is here.

Second script

The second chunk is run by new_subj_second script. This does the below:

  • Run MRIQC on that subject.
  • Re-run MRIQC group command to add the new subject to the group MRIQC file.
  • Pull only a subset of the columns of MRIQC IQMs into the MRIQC_Filtered_Columns.csv, used to update the CBPD_Scanning_Data Google sheet.
  • Run cross-sectional Freesurfer on that subject’s chosen T1, including hippocampal subfields. The T1 used is documented on the CBPD_Scanning_Data Google sheet, and a copy is saved for posterity in this repo in fmriprep/CBPD_Scanning_Data_MRI_Quality.csv. Each timepoint has a separate Freesurfer SUBJECTS_DIR (see here).
  • Run cross-sectional Freesurfer on any experimental T1 sequences (no longer used). Each experimental sequence has a separate Freesurfer SUBJECTS_DIR.
  • When Freesurfer is done, run fMRIprep specifically on this session, which will then run with precomputed session-specific Freesurfer inputs. (see here).

Then, someone should put their eyes on fMRIprep .html files for the subject, and copy MRIQC outputs (including # of resting-state vols kept) into CBPD Scanning Data file.

Note: Here we treat longitudinal timepoints as separate subjects for Freesurfer and fMRIprep, given that we might expect significant anatomical change between timepoints. Longitudinal Freesurfer pipelines will be run separately, as they are not being used for functional preprocessing (per APM 03/2020).

Requirements

Most scripts pull Singularity containers from their location in /cbica/projects/cbpd_main_data/tools, but you do need a few utilities accessible from your path. I recommend installing them all into a Conda environment (maybe your base environment).

  • Check that Singularity is accessible from your path (and if not, add it): 
     which singularity
  • Freesurfer : Use the 6.0.0-make-fix version if you're going to be brain-editing. You will need to add this to your .bashrc. Add the two lines below to your .bashrc, and check with echo $FREESURFER_HOME once you've logged in. 
     module unload freesurfer/5.3.0
 module load freesurfer/6.0.0
    More information on the make-fix version is here.
  • Python packages in your Conda environment: python-dateutil, dcm2niix, pandas
    For CfN: You need a current version of dcm2niix, do not use the one in /data/picsl/mackey_group/BPD/envs/bpd_py! You may need to remove that from your path. 
       conda create -n <env-name> python=3.7
   conda activate <env-name>
   conda config --append channels conda-forge
   conda install python-dateutil dcm2niix pandas
 	conda install -c conda-forge jq
   pip install pybids
    If you're using these scripts and you didn't install these packages into your base environment, change lines 8-10 of new_subject_first.sh and new_subject_second.sh to reflect your username and activating your Conda environment.

Utilities

  • Copying over dicoms from Rico (computer at SC3T), detecting which have changed in the last four days
  • Copying over dicoms from Flywheel; to do this you need the Flywheel Command-line Interface (CLI) in your $PATH and a Flywheel Upenn Account. Run the line below while logged in to your user account on CUBIC (this is already set up for the cbpdmaindata project user): 
     echo "export PATH=\$PATH:/cbica/projects/cbpd_main_data/tools/flywheel/linux_amd64" >> ~/.bashrc
  • Backing up dicoms to the hard drive (plug in drive, put in IDs, and run the loop)
  • Removing extra TRs for sleeping participants
  • Fixing error about Conflicting study identifiers found (see here for re-registered participants)
  • Fixing protocol names of wrongly-named dicom headers
  • Pulling a subset of MRIQC image quality metrics

Using the data

Please read the README and CHANGES at the top-level of the BIDS dataset!