Processing details

When running BIDSonym, the following processing steps are executed:

1. running BIDS-validator:

Before anything else happens, the BIDS-validator is run in order to evaluate if the provided bids_dataset is valid with regard to the BIDS specification. If that’s the case, BIDSonym will continue with the subsequent steps, if not, it will stop and provide the user with a message of why the bids_dataset was evaluated as not valid. Users should then conduct the changes necessary to make it valid, as otherwise BIDSonym won’t run. This step and its implementation is based on and borrowed from the respective fmriprep functionality, thus all credits go to their amazing developer team. The same accounts for the copyright. Below you see an example of the BIDS-validator output, providing information about the bids_dataset and a warning telling you that the README is missing.

2. brain extraction applied to non-defaced images:

In order to evaluate the success of the de-facing, especially with regard to “cut out too much”, a brain extraction/skull stripping procedure will be applied to the images before the de-facing. The respective brain mask will be overlaid on the de-faced images within the visual QA report to allow an easy assessment of potentially too stringent outcomes. A respective example is shown below where the upper panel displays a non-defaced image and the lower one a defaced image. The reddish overlay represents a brain mask that was extracted before the image was defaced.

3. moving of non-de-identified data:

The non-de-identified data, that is structural images and sidecar JSON files, will be moved from the bids_dataset directory to bids_dataset/sourcedata/bidsonym. In case the de-identification was not successful (too much or too little information deleted), the non-de-identified version can easily be copied back from to the bids_dataset directory without the necessity to run the corresponding DICOM to Nifti in BIDS conversion again.

4. evaluation of metadata:

The metadata found in both, the header of the images and sidecar JSON files will gathered and saved in a tabular data file (.tsv) of the form metadata field : value to the bids_dataset/sourcedata/bidsonym/ directory. Additionally, a third column problematic will indicate if the value of a certain metadata field is potentially problematic or sensitive. By default all values will considered not problematic (no). Only if the user specifies a list of strings which might be included in problematic or sensitive information (e.g., ‘name’) BIDSonym will search for these strings and mark respective values of metadata fields as problematic (yes)

5. defacing of images:

Subsequently, the chosen defacing algorithm will be applied to the structural images, aiming to remove features that could potentially allow or aid the identification of participants’ identity (e.g., their face). Depending on the algorithm chosen, more or less features are removed and the sufficiency needs to be evaluated by the user (supported through the visual QA). If the -deface_t2w flag is set, structural T2 weighted images will de-faced by using the respective de-faced T1 weighted image of the same subject as a mask. All de-faced images will be written to the bids_dataset directory.

6. de-indetification of metadata fields:

If indicated by the user via a list of strings, the values of certain metadata fields will be replaced by the string deleted_by_bidsonym. As the image files, metadata files will be moved to bids_dataset/sourcedata/bidsonym/ and de-identified files will be written to bids_dataset.

7. renaming of original files:

All files, including imaging and metadata, that are stored under bids_dataset/sourcedata/bidsonym/ will be renamed to indicate their respective status. Thus the descriptor _desc-nondeid is added to the respective file names.