Concepts

Identifiers and privacy

What in a dataset can identify a person, what BIDS Manager removes on its own, what it leaves, and what to check before a dataset leaves your hands.

A BIDS dataset is meant to be shared. Raw data is not: DICOM headers carry the participant's name, birth date and patient ID, recording files carry what was typed into the acquisition software, and a head image carries a face. BIDS Manager removes what it can find on its own. You are still responsible for the dataset; the checks below are the ones that matter.

What BIDS Manager removes

As the last step of converting each subject, from every sidecar the conversion writes:

Removed Kept, and why
The participant's name, ID, birth date, age, sex, weight and size
The accession number of the visit
The names of the referring, performing and requesting physicians and of the operator
The department's address InstitutionName and InstitutionAddress: the standard recommends them, and they describe the site, not the person
The study and series identifiers (UIDs): they identify no one, and they are what traces an image back to its series

MR spectroscopy needed its own pass: a NIfTI-MRS file carries its metadata inside the image file, where no sidecar step reaches, and the same identifiers are removed there, together with OriginalFile, the path of the source file on the converting computer.

What it leaves, and you should check

Where What What to do
participants.tsv In 1.5.0, the conversion copies given_name, family_name and patient_id from the DICOM headers into it Delete the three columns (Sharing a dataset). Listed in Known issues.
Anatomical images A face can be rendered from any head MRI, and from some PET Deface them (Defacing and skull stripping). The quality check reports A face is visible.
EEG and MEG recording files Their headers can hold a name or a birth date, as typed at the recording BIDS Manager does not change what a recording file holds. Check their headers, and anonymise at the source if needed.
Free text A task description, a note, a README that names a person or a date Read them.
Labels and file names A subject label made from initials or a hospital number, a session named after a date Rename them in the table before converting, or with Rename entity... after. The Converter's Overview flags task and other labels that differ for nearly every recording; it does not check subject or session labels, so read those yourself.
Dates Acquisition dates and times in _scans.tsv and sidecars Some data-sharing rules require shifting or removing them.
sourcedata/ The raw files, headers and all, when Keep in sourcedata/ is set to copy or archive, and the images as they were before defacing if you chose to keep them there Do not share sourcedata/ from a dataset of people.
.bidsmgr/ The project's history, including the originals of the files an Editor operation rewrote, deleted or defaced, and the raw folder's path Never share it. It is listed in .bidsignore.

The DICOM header viewer

The DICOM header viewer marks the identifying values and hides them until you turn on Show identifying values; a copied or exported header leaves them out unless you turn on Include identifying values.

Before sharing

Sharing a dataset goes through the steps on the course's sample: the Dashboard's Ready to share list, participants.tsv, defacing, the README and the citation file, and a last validation.