Scientific background, objectives, and related publications for the BRAID project.
The Brain fRAilty Integrated through Data federation (BRAID) project evaluates and enhances the DARE-UK TREvolution infrastructure to support federated neuroimaging and multimodal analytics. Our aim is to advance the long-term vision of a scalable, privacy-preserving national framework for cross-TRE research on brain health and frailty.
Brain frailty refers to the risk of poor clinical outcome — including mortality — resulting from pre-existing but not yet clinically manifest brain damage from ageing and subclinical disease. Unlike chronological ageing, brain frailty reflects cumulative structural and functional deterioration that increases vulnerability to neurodegeneration, hypoxia, or neurotoxicity.
Neuroimaging studies have identified robust indicators of brain frailty — cortical thinning, white-matter hyperintensities, microbleeds, and disrupted functional connectivity — that correlate with physical frailty, accelerated cognitive decline, poor mental health, and adverse outcomes including dementia, stroke, and treatment-related neurotoxicity.
Testing compute capacity, workflow management, access procedures, interoperability, provenance tracking, and containerised pipelines within the TREvolution technology stack.
Evaluating whether analytical outputs — spatial maps, summary statistics, numerical vectors, trained models — pose any risk of re-identification before leaving the TRE.
Adapting multi-modal MRI processing and brain age regression models into 5Safes-TES (Task Execution Service) compatible workflows deployable across distributed TRE nodes.
BRAID employs public and staged, synthetic-data-driven approach using the DPUK TRE as the primary federated node, with Nottingham University Hospitals as a secondary node. All research uses either fully anonymised publicly available or synthetic data derived from real neuroimaging cohorts, ensuring privacy whilst enabling rigorous technical validation.
Data used to develop the project will be obtained from ADNI and from High-fidelity synthetic datasets are generated from neuroimaging and clinical data (QMIN-MC, OBHC, MEMENTO, Whitehall II, CamCAN) using differentially private vine copula methods developed through DPUK. Datasets are evaluated for utility, quality, and privacy protection.
Multi-modal anatomical MRI processing (T1-weighted, T2-weighted, susceptibility-weighted) using FSL and UK Biobank pipelines. Outputs include brain age regression, cortical thickness, white matter hyperintensity segmentation, and subcortical volume analysis.
All datasets are explicitly tagged and documented according to FAIR principles. Metadata specifications ensure reproducibility of federated neuroimaging pipelines.
Systematic evaluation of analytical outputs — spatial maps, summary statistics, numerical vectors, and trained models — using semi-automated tools (SACRO, SACRO-ML) to assess re-identification risk and ensure safe egress before results leave the TRE.
Note: Public Involvement & Engagement (PIE) activities are embedded throughout all stages of the project, ensuring public perspectives shape technical design, governance decisions, and the responsible development of federated methods. See the PIE Activities page for details.
BRAID is part of the DARE UK Research Exemplars programme. Visit the official DARE UK website to learn how BRAID contributes to the broader UK TRE ecosystem and the TREvolution programme.
View BRAID on DARE UK ↗Key publications from the BRAID team underpinning the project's methodology. This list will be updated as the project progresses.
Publications directly arising from the BRAID project will be added here as they become available. All code and workflows will be openly shared via GitHub under an MIT licence.