BIDSonym: a BIDS App for the pseudo-anonymization of neuroimaging datasets
Why this work is in the frame
A frame that forgets how it found something cannot be audited. These are the routes that admitted this work.
Bibliographic record
Abstract
sharing is important and beneficial To that end, Ethic Review Boards and data sharing platforms typically require that uploaded datasets are provided in anonymized or pseudoanonymized form, limiting participant reidentification. However, the (pseudo-)anonymization process is deceptively complex; attempts at ensuring data privacy must take into consideration all dataset components, including imaging modalities, as well as national legal and ethical frameworks. Several algorithms have been developed to (pseudo-)anonymize imaging datasets but they offer limited solutions. Some are attached to specific software and some are limited to specific computing environments; most miss an in-depth assessment and treatment of the metadata attached to the dataset or lack the capacity to automatize (pseudo-)anonymization across large datasets. BIDSonym was created to address these points in one simple, flexible, and general tool that offers users an array of automated (pseudo-)anonymization options to augment participant privacy in neuroimaging datasets. There are two components of neuroimaging datasets that arguably pose the largest risk to maintaining participant privacy: the structural images and accompanying metadata (e.g., metadata text files or information embedded in image file headers). Structural images contain visible identifiable participant information via facial features like the eyes, nose, and mouth, and privacy is usually addressed through a process called "defacing," within which all or a subset of these features are removed from the final structural data files. The metadata text files may additionally contain identifiable participant data through the recording of acquisition time and location, and personal details such as date of birth, height, and weight. Here, privacy is maintained by removing or blurring this information from the final dataset. BIDSonym addresses both vulnerabilities in neuroimaging datasets, obviating the need for multiple steps within a data sharing pipeline to ensure participant privacy.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
Full frame distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.002 | 0.002 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.000 | 0.001 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.000 | 0.001 |
| Open science | 0.004 | 0.001 |
| Research integrity | 0.000 | 0.000 |
| Insufficient payload (model declined to judge) | 0.000 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from it