The policy question
Sharing brain imaging data supports checking results and reusing existing research. Protecting participants requires understanding what the released information can reveal and under what conditions. A useful policy assessment therefore asks both how a person could become identifiable and what protection would cost in scientific usefulness. 1, 2
What the evidence says
Brain imaging outputs differ. Full-head images may retain recognisable facial information; a regional summary or connectivity matrix retains different features. The working consensus manuscript examines these representations separately, across modalities and processing levels. It focuses on imaging properties under stated assumptions rather than issuing a complete legal determination for every use. 2
Distinctiveness is not the same as recovering an identity. Matching a person's brain-connectivity patterns across recordings is a demonstrated scientific capability. Whether this enables identification outside that research setting depends on additional information and opportunities for linkage. 2, 3
Accompanying tables deserve attention. The metaprivBIDS paper found that age and demographic combinations were important potential identification routes in selected OpenNeuro metadata. Clinical scores contributed less to matching in these analyses while remaining sensitive if connected to a person. The study assessed potential vulnerabilities; it did not attempt to identify participants. 1
What these studies do not establish
They do not provide a universal probability of someone being identified from a brain scan. The metadata datasets were selected to cover different use cases, not to estimate the prevalence of privacy problems in all research data. The consensus manuscript uses expert judgements, with assumptions about the data and participants; its ordinal ratings are not percentages. 1, 2
The manuscript's imaging assessment also does not automatically extend to people with visually noticeable skull or brain abnormalities, or to combinations with identifying metadata. A policy built around one modality-wide label would miss these conditions. 2
Implications for research policy
The following are implications drawn for this website from the two sources, rather than final recommendations endorsed by the consensus group.
| Policy priority | Practical implication |
|---|---|
| Assess complete releases | Include images, recordings, derivatives and accompanying metadata in disclosure review |
| Support proportionate access | Consider public and controlled access in relation to the actual information and intended use |
| Preserve scientific value | Test whether proposed coarsening, noise or image transformations affect the analyses the data should support |
| Resource the work | Provide time and expertise for curation, privacy assessment, documentation and maintenance |
| Make decisions reviewable | Record evidence, assumptions and reasons for revisiting an assessment |
The metadata paper illustrates why blanket removal of clinically valuable variables may target the wrong part of the problem. The contributors' knowledge of the study population and variables remains important alongside quantitative tools. 1
Keep the evidence current
New matching methods, accessible reference information and changing acquisition practices can alter an assessment. The website will accept studies and corrections through GitHub review, with a dated explanation when the guidance changes. New evidence should be evaluated for its study conditions and relevance, rather than automatically treated as proof that every dataset has the same risk. 2
Sources
- Published study: Kibsgaard et al. (2026), Assessing metadata privacy in neuroimaging. OpenNeuro analysis notebooks.
- Ongoing consensus work: Pernet et al., An expert consensus on identification likelihood in NeuroImaging data. Working manuscript consulted 6 November 2026, unpublished and subject to revision. Associated analysis.
- Finn et al. (2015), Functional connectome fingerprinting: identifying individuals using patterns of brain connectivity.
Published findings, the working consensus manuscript and editorial implications are identified separately. These audience summaries await expert review; final consensus ratings are not reproduced here.