AI prediction of individual brain health risk from real-world clinical datasets
The computational partner for dementia prevention programmes.
Aim
Many emerging dementia therapies are likely to be most effective before substantial neurodegeneration has occurred. Yet patients are often identified only after symptoms become established and diagnosis is made. Eyzan's aim is to identify individuals at increased risk years before routine clinical diagnosis, using scalable predictive methods developed for real-world healthcare data. This supports earlier intervention, clinical trial recruitment and future treatment pathways.
What we do
AI designed for real-world healthcare data.
Organisations running dementia prevention programmes, clinical trials, and longitudinal cohort studies hold datasets with far more prognostic information than conventional analysis extracts — or are planning studies where the design choices made now will determine whether AI prediction is feasible at scale. Eyzan works at both stages.
We convert existing datasets into personalised, confidence-calibrated predictions of individual future clinical risk, and advise on how planned studies should be designed to maximise AI prediction performance from the outset. Eyzan Ltd is a medical consultancy, analytics and software development company specialising in predictive AI for preventive medicine and clinical research.
Our core expertise is not simply AI — it is developing methods that work with real-world clinical data: heterogeneous NHS imaging acquired across different scanners and protocols, linked electronic health records, blood biomarkers, genetic risk scores.
Complementary to
What brain health organisations already do
Clinical trial infrastructure, biomarker diagnostics, data platforms, and biobanking are established capabilities in the field. Eyzan adds the AI neuroimaging computation that makes individual-level risk stratification possible — enabling more precise trial recruitment, earlier intervention, and personalised prevention programmes.
| Partners provide | Eyzan provides |
|---|---|
Clinical trial infrastructure Site management, participant recruitment, regulatory compliance |
AI risk stratification Individual patient prediction of future risk, progression, and clinical trajectories |
Biomarker diagnostics Neuroimaging and blood-based biomarkers including plasma pTau217 and APOE genotyping |
Biomarker diagnostics Confidence-calibrated multimodal AI prediction and computational modelling |
Data infrastructure Longitudinal cohort data, biobanking, linked health records |
Study design advisory Advising on study and data-collection design to maximise scalable individual patient prediction from real-world healthcare data. |
Track record
Working with AI and clinical data since 2009
2009
AI development with NHS patient neuroimaging datasets began
80%
Accuracy in highest-confidence subgroup (~35% of scans)
5–10 yrs
Advance prediction of NHS dementia diagnosis before clinical recognition
MRC Programme Grant funding supported development and validation in a regional NHS Safe Haven. Methods were subsequently replicated at national scale in the Scottish National Safe Haven. The key insight from both programmes is that accurate AI prediction of future dementia is a solved problem in curated research datasets. The unsolved problem—and Eyzan's focus—is making it work reliably with real-world clinical data. Our methods were developed and validated using routine clinical imaging collected during everyday healthcare rather than research-grade datasets.
Age is the strongest predictor of dementia risk. Our models were therefore developed and validated using age-matched cohorts, ensuring that predictions reflect information contained within the brain scan independent of age. The 80% accuracy reported for our MRI models therefore represents age-independent predictive performance. Scan-derived predictions can then be combined with age and other clinical variables to provide an overall personalised risk estimate for an individual patient.
Tailored Predictive Models
Predictive models are not universally transferable between datasets. Differences in populations, biomarkers, imaging protocols and outcome definitions mean that models must be developed and validated for the data and objectives of each organisation. Eyzan develops and validates predictive models using existing healthcare, biomarker and research datasets, supporting organisations from feasibility assessment through model development, validation and deployment.
From Feasibility to Deployment
Eyzan uses a stage-gated development process, allowing organisations to evaluate feasibility before committing to full deployment. Each stage is undertaken under a separate agreement with defined objectives, deliverables and costs.
Stage 1 – Feasibility: Assessment of whether prediction is feasible using existing data resources.
Stage 2 – Model Development and Validation: Development and validation of predictive models using anonymised labelled data, with reporting of standard performance metrics. Predictions can then be generated for anonymised datasets with labels withheld from Eyzan, enabling independent verification of performance by the organisation.
Stage 3 – Real-World Validation: Evaluation on newly acquired data not available during model development, demonstrating that predictive performance is maintained outside the original development dataset.
Stage 4 – Deployment: Operational deployment through software licensing or bespoke prediction services.
Contact
To discuss whether Eyzan's methods could add value to your datasets or programmes:
Enquiries