a computer screen with a lot of text on it

World leading science for early dementia prediction

Spun out from the University of Cambridge, we've invented smart ways to use biologically-informed AI for clinical impact

We need new solutions for dementia

We currently lack precise ways to work out each person's subtype and stage of dementia - especially early on when interventions can make the biggest difference.

Good drugs won't work if given to people who can't benefit from them
It's hard to tell patients apart - we need to know the type and stage of dementia.
It's hard to tell patients apart - we need to know the type and stage of dementia.

This means that when we study a group of patients, we're often blind to the important differences in disease between people.

Studying a mix of people with different dementia types and/or stages makes it much harder to spot the key signals that help us make progress.

Mixed groups of subtypes/stages means its hard to find new biomarkers and test new treatments
Mixed groups of subtypes/stages means its hard to find new biomarkers and test new treatments
black blue and yellow textile

Predict years ahead

Diagram showing dementia progression with vs. without AI‑based patient stratification
Diagram showing dementia progression with vs. without AI‑based patient stratification
Prodromic transforms understanding of patients:

  • Patients who were apparently ‘the same’ show very different trajectories

  • Prodromic distinguishes them at baseline with threefold higher sensitivity than standard approaches

  • Subgroups diverge statistically after just 18 months

The Lancet: eClinicalMedicine, 2024

Diagram showing dementia progression with vs. without AI‑based patient stratification
Diagram showing dementia progression with vs. without AI‑based patient stratification

Transforming the development of new treatments

Graphs showing trial results with vs. without stratification using Prodromic Alzheimer's prediction
Graphs showing trial results with vs. without stratification using Prodromic Alzheimer's prediction
Prodromic uncovers a treatment effect in a large pharmaceutical trial

  • Therapy believed not to help patients based on standard analysis

  • Stratifying the patients using only baseline measures revealed that the treatment had worked for a specific group.

Nature Communications, 2025
JPAD, 2026

Alluvial plots showing dementia progression is slower for patients receiving treatment
Alluvial plots showing dementia progression is slower for patients receiving treatment
Less progression in the treatment group

  • "Slowly progressive" patients less likely to become "Rapidly progressive" in the treatment group relative to the placebo.

Using Prodromic increases trial efficiency

  • Modelling trial results with vs. without Prodromic demonstrates the benefits of stratification.

  • Standard approach - 1,524 patients needed

  • Using Prodromic - only 164 patients needed

Statistical power analysis shows much higher ROC performance using Prodromic
Statistical power analysis shows much higher ROC performance using Prodromic
Graphs showing trial results with vs. without stratification using Prodromic Alzheimer's prediction
Graphs showing trial results with vs. without stratification using Prodromic Alzheimer's prediction
Alluvial plots showing dementia progression is slower for patients receiving treatment
Alluvial plots showing dementia progression is slower for patients receiving treatment
Statistical power analysis shows much higher ROC performance using Prodromic
Statistical power analysis shows much higher ROC performance using Prodromic

Accelerating new treatments

Prodromic's solutions can identify the right patients for clinical trials to deliver new treatments faster and at lower cost.

Based on re-analysis a previous trial, using Prodromic's technology means:

  • Fewer patients needed

  • Higher statistical power

  • Faster trial readout

  • Lower costs

Cupped hands holding medicine pills
Cupped hands holding medicine pills

Research papers

Prodromic is built on a decade of innovation driven from Cambridge. The work underpinning our technology has been published in a series of papers in leading, international peer-reviewed journals:

Giorgio et al (2020) Modelling prognostic trajectories of cognitive decline due to Alzheimer's disease. Neuroimage: Clinical 26, 102199

Giorgio et al (2022) A robust and interpretable machine learning approach using multimodal biological data to predict future pathological tau accumulation. Nature Communications 13, 1881

Lee et al (2024) Robust and interpretable AI-guided marker for early dementia prediction in real-world clinical settings. The Lancet: eClinicalMedicine 74, 102725

Vaghari et al (2025) AI-guided patient stratification improves outcomes and efficiency in the AMARANTH Alzheimer's Disease clinical trial. Nature Communications.

Welchman & Kourtzi (2026) Solving the 'Goldilocks problem' in dementia clinical trials with multimodal AI. Journal of Prevention of Alzheimer’s Disease.

an abstract photo of a curved building with a blue sky in the background

Coverage of our science

BBC News: AI may diagnose dementia in a day

The Times: AI could diagnose dementia before symptoms show

Sky News: New AI tool could be a game-changer in battle against Alzheimer's

Alan Turing Institute: AI could detect dementia after single brain scan

University of Cambridge: AI outperforms tests at predicting progress of Alzheimer's Disease

hello@prodromic.ai

© Prodromic Ltd. 2026. All rights reserved.

Privacy Policy