• Split into 2 datasets:
  • BIDS compliant dataset created from DICOMs: 346 subjects
  • Dataset of participants with missing DICOMs: 12 subjects


• BIDS dataset created from DICOMs validated with bids-validator:

  
  $ bids-validator "/media/localadmin/HagmannHDD/Seb/PsychoDev/NiftiBIDSCompliant" --ignoreSubjectConsistency
  bids-validator@1.7.2
  
  [...]
  
          Summary:                   Available Tasks:        Available Modalities:
          9212 Files, 47.25GB        rest                    MTR                  
          346 - Subjects                                     T1w                  
          6 - Sessions                                       T2map                
                                                             dwi                  
                                                             bold                  
                                                             T1map                
                                                             fieldmap              
  									

• Creation of pre-configured empty Datalad dataset


• Copy the content of the initial BIDS dataset with a few problems such as the use of the acquisition date as a session label.


• Edit the CHANGES

  
  0.1 2021-06-05
      - Initial commit of BIDS dataset after reconversion of DICOMs to Niftis using dcm2niix
  										



• Save the dataset with version tag 0.1

• Add and run a script in code/ folder to fix BIDS validation errors


• Validate dataset with the bids-validator


• Edit the CHANGES

  
  0.2 2021-07-07
      - Create bidsify_dataset.py in code/ folder
      - Run code/bidsify_dataset.py that:
        * Add CHUV prefix to BIDS participant_id labels
        * Change session labels from acquisition date to scan1, scan2, ...
        * Add sub-<label>_sessions.tsv/.json files that stores the participant age in month and the acquisition time of each session
        * Generate participants.tsv/.json files that stores the participant sex and the group
    										



• Save the dataset with version tag 0.2


• Publish the dataset to our instutional data storage accessible via ssh

Until now

• Dataset can be installed/cloned with:

  
  	$ datalad clone ssh://<user>@stockage-horus.chuv.ch/:/archive/PRTNR/CHUV/RADMED/phagmann/biopsycho/NiftiBIDSCompliant
  	  										

  Note: one needs to be inside CHUV intranet or to use CHUV VPN and have access to stockage-horus to be able to clone the dataset

In the near future

• Dataset could be installed/cloned with:

  
  	$ datalad clone git@github.com:NCCR-SYNAPSY/ds-biopsycho
  	  										

  Note: everybody would be able to clone the dataset from GitHub but one still needs to be inside CHUV intranet or to use CHUV VPN and have access to stockage-horus to be able to retrieve the content of the files

BIDS-compliant solution

• Add an extra layer of directories and file names in the form of ses-<label> to encode multiple sessions (visits)
• Use a subject label prefix identifying the site (CHUV here)

├── sub-CHUVA001
│   ├── ses-scan1
│   │   ├── anat
│   │   │   ├── sub-CHUVA001_ses-scan1_run-1_T1w.json
│   │   │   ├── sub-CHUVA001_ses-scan1_run-1_T1w.nii.gz
│   │   │   ├── ...					
						

Reference: Longitudinal and multi-site studies section of the specifications

BIDS-compliant solution

• Use BIDS _run-<label>_ entity to differenciate scans with same acquisition parameters
• Use BIDS _acq-<label>_ entity to differenciate similar scans with different acquisition parameters

├── sub-CHUVA022
│   ├── ses-scan1
│   │   ├── anat
│   │   │   ├── sub-CHUVA022_ses-scan1_run-1_T1w.json
│   │   │   ├── sub-CHUVA022_ses-scan1_run-1_T1w.nii.gz
│   │   │   ├── sub-CHUVA022_ses-scan1_run-2_T1w.json
│   │   │   ├── sub-CHUVA022_ses-scan1_run-2_T1w.nii.gz
│   │   │   ├── ...			
│   │   ├── dwi
│   │   │   ├── sub-CHUVA022_ses-scan1_acq-dsiNdir129_run-1_dwi.bval
│   │   │   ├── sub-CHUVA022_ses-scan1_acq-dsiNdir129_run-1_dwi.bvec
│   │   │   ├── sub-CHUVA022_ses-scan1_acq-dsiNdir129_run-1_dwi.json
│   │   │   ├── sub-CHUVA022_ses-scan1_acq-dsiNdir129_run-1_dwi.nii.gz
│   │   │   ├── sub-CHUVA022_ses-scan1_acq-hardiNdir147Bval3000_run-1_dwi.bval
│   │   │   ├── sub-CHUVA022_ses-scan1_acq-hardiNdir147Bval3000_run-1_dwi.bvec
│   │   │   ├── sub-CHUVA022_ses-scan1_acq-hardiNdir147Bval3000_run-1_dwi.json
│   │   │   ├── sub-CHUVA022_ses-scan1_acq-hardiNdir147Bval3000_run-1_dwi.nii.gz		
							

Reference: Entities section of the specifications

• At least participant_id, age, sex and group MUST BE REPORTED


• If the study IS longitudinal:
  • participant_id, sex and group MUST be reported in the participants.tsv file
  • age MUST be reported in the session.tsv file


• If the study IS NOT longitudinal:
  • participant_id, age, sex, age, and group MUST be reported in the participants.tsv file


• Age MUST BE REPORTED in months to account for cases with multiple acquisition sessions in a given year


• All columns (except participant_id) MUST BE DESCRIBED in the participants.json / session.json

• Never keep the acquisition date in the session label: this might enable the identification of the scanned individual even if MRI was anonymized or de-identified


• Keep the README of the dataset the most informative: it is the first information a user will have access to.


• De-identify MRI with a defacing tool such as pydeface if the intent is to release the dataset publicly.
  
  Please check this paper (Athena E. Theyers et al, 2021) for a comparison of multiple defacing tools.


• Install bids-validator locally if the dataset is big; the online validator is very slow. See the bids-validator Github repo for more details.