Research Data Reuse · Vision Science & Neuroscience

Reusing Archived Research Data in Vision Science and Neuroscience.

Infrastructure for making already-collected imaging datasets usable again, responsibly, by more than the team that first collected them.

Talk to Us ▶ Watch the Demo How It Works
UNIVERSITY OCT Lab NHS TRUST Imaging RESEARCH MRI Study CLINICAL Trial Data Undosa Tech GOVERNED REUSE

The thesis

A great deal of imaging and clinical data collected for biomedical research is used once and then becomes unreachable — sitting on lab servers and old drives, no longer findable, understandable, or approvable for reuse. The consent was given, the money was spent, the science was done, and the data is not lost, but it may as well be.

This is a governance and infrastructure problem more than a technical one, and it is the problem UndosaTech is being built to work on — starting with a single vertical, vision science, and adjacent neuroscience.

85%
of biomedical research investment is avoidably wasted — much of it because existing data is never made findable or reusable.
Chalmers & Glasziou, The Lancet (2009)
1.6M
existing retinal images trained RETFound, a foundation model that detects disease signals those images were never collected to reveal.
Zhou et al., Nature (2023)
20
institutions worldwide jointly trained one clinical model — without any patient data leaving any hospital.
Dayan et al., Nature Medicine (2021)
99%
of centralised model performance is achievable with federated training in multi-site medical imaging.
Sheller et al., Scientific Reports (2020)
Federated by Design

The data never leaves.

Analysis travels to the data and only aggregate, disclosure-controlled results return. For data custodians this changes the question an ethics committee is asked: instead of "may this dataset be copied to a third party?", it becomes "may an approved analysis run where the data already lives?" — with every run logged in a tamper-evident audit trail.

▶ Watch a training round
INSTITUTION A INSTITUTION B INSTITUTION C INSTITUTION D AGGREGATE MODEL model updates travel patient data stays on-site
Consent & ethics review before activation Statistical disclosure control on all outputs Tamper-evident audit trail UK GDPR · Article 89 research safeguards Security & governance detail →

What reuse delivers

Time compressed

A fresh prospective cohort takes years. A reused one is analysable in months.

New questions, old data

Foundation models can now extract signals from historical images that were not extractable when those images were collected.

Less waste

Every dataset reused is one fewer redundant cohort, one fewer patient asked to consent to a study that has already been done.

Cost avoided

Harmonising an existing cohort costs orders of magnitude less than collecting an equivalent one from scratch.

How It Works

From archived data to governed reuse

Five stages carry a dataset from a lab drive where it sits unused, to a resource a new research team can responsibly access. UndosaTech coordinates the pipeline and provides the tooling — structuring and analysis run inside the contributing institution, where the data stays.

01 Archived Data Sitting unused on a lab server 02 Governance Review Consent, ethics and provenance checked 03 Structure & Harmonise Mapped to a shared model, at the source 04 Governed Access Approved researchers apply and are vetted 05 Research Reuse New questions asked of old data
See What We Offer Get Involved