Transforming dementia diagnosis

Eye-tracking and machine learning for early, accessible cognitive assessment.

Why early diagnosis matters

Dementia affects one in eleven people over the age of 65 in the UK, placing immense pressure on individuals, families and health systems. Yet diagnosis remains slow, reactive, resource-intensive, and often insensitive to subtle cognitive changes. Most people only enter the diagnostic pathway once symptoms become noticeable – memory lapses, confusion, or difficulty managing daily tasks. By that point, cognitive decline may already be significant, and the underlying neurodegeneration is largely irreversible.

Current assessments typically involve multiple GP visits, pen-and-paper tests, specialist clinical appointments, and sometimes imaging or biomarker analysis such as brain scans or blood tests. These processes are not only intensive for the NHS – they rely on symptoms being noticed, meaning early-stage decline is often missed altogether.

Early detection offers the best opportunity for intervention before substantial cognitive decline occurs – this is where UWE Bristol is leading bold innovation.

Can eye movements reveal cognitive decline earlier?

Evidence from psychology and neuroscience shows that subtle changes in eye-movement patterns can signal cognitive impairment long before clinical symptoms emerge. When tasks require more mental effort, subtle eye movement patterns can reveal struggle - hesitation, slower visual processing, difficulty tracking information across a screen.

However, traditional eye-tracking equipment is expensive (£1,000s–£100,000s), highly technical, and requires precision calibration. Tasks used in clinical settings are often unfamiliar or anxiety-inducing for older adults. As a result, eye-tracking has yet to be widely adopted in mainstream healthcare.

Dr Wenhao Zhang, Associate Professor of Computer Vision and Machine Learning at UWE Bristol, is changing that.

Eye-tracking using any camera

Dr Zhang and the research team are developing a low-cost, accessible eye-tracking system that works through the kinds of cameras people already own such as laptops, tablets or phones. There is no specialist hardware involved, no calibration rituals, and no need for a clinical environment. Instead, the system uses advanced machine-learning models that can detect and follow the movement of the eyes, analyse tiny changes in how they behave over time, and identify patterns associated with attention, memory, and other cognitive processes relevant to early decline. Crucially, it can also build a personalised baseline for each user, making it possible to recognise unusual deviations long before symptoms become obvious.

This shift, from occasional clinic-based testing to continuous, home-based monitoring, opens the door to earlier, more effective intervention at scale.

Making assessment intuitive: A gamified cognitive test

To make assessment feel natural and engaging, especially for older adults, the team created a gamified cognitive test as part of a BRACE-funded research project. The result is a fruit-picking game that combines mobile-game inspiration with well-established cognitive paradigm. In the game, apples and pears sweep across the screen, and players tap the correct basket corresponding to the colour each fruit changes to, to harvest the fruit. The game subtly increases in difficulty through changes in speed, visual obstruction and fruit-colour combinations, making it an enjoyable way to probe attention, working memory and reaction time.

As people play, the system records how their eyes move, how quickly they respond, how often they make errors, and how their performance shifts under greater cognitive demand. These behavioural and visual data, analysed using machine learning to automatically uncover additional predictive signatures, may reveal early decline long before traditional tests would detect it.

The BRACE-funded phase has already produced a substantial dataset (soon to be made open access) alongside peer-reviewed outputs, additional major manuscripts in development, and increasing engagement with patients, carers, the research and professional community through patient and public involvement sessions, sector events, and conferences, for example, Alzheimer’s Research UK conferences. It highlights UWE Bristol’s dedication to meaningful innovation that blends creativity, technology and societal impact.

Towards naturalistic, everyday testing

The research is now evolving beyond gamified tasks to explore how cognitive changes appear during everyday digital activities such as reading news online, browsing on a smartphone or completing common onscreen interactions. Eye-movement analysis during reading has long been recognised as an informative method in cognitive science, and the team is adapting this approach for real-world use in home environments.

This work forms the focus of a new PhD project, which will test the system with patient populations to understand how well these naturalistic tasks can support early diagnosis in practice.

Building a strong interdisciplinary foundation

This pioneering research builds on a series of funded, collaborative projects:

UWE Bristol-funded interdisciplinary project: Early pilot studies brought together machine vision, psychology and biomedical science.
EPSRC FARSCOPE CDT PhD: Advanced the single-camera method, resulting in a peer-reviewed publication.
BRACE-funded project (2023–2025): Enabled the creation and validation of the gamified cognitive test, extensive data collection and increased visibility through publications and sector events.
Current PhD project: Extends the work into naturalistic assessment and clinical testing.

Across these stages, the research demonstrates the sustained, cross-college collaboration that defines UWE Bristol’s RISE strategy – combining technological innovation with real-world healthcare application.

Cross-sector innovation

Although this case study focuses on dementia, Dr Zhang’s broader work in the Centre for Machine Vision has generated impact across multiple sectors.

Collaborative machine-vision research has produced new, award-winning technologies now commercialised in agriculture, demonstrating how expertise in machine vision and learning can translate across disciplines, including health and industrial applications.

This showcases the RISE commitment to enterprise that fuels progress, with applications spanning healthcare, farming and digital technology.

Towards scalable, preventative healthcare

This research has the potential to transform how we detect and monitor dementia. A system that people can use at home, with devices they already have, dramatically reduces pressure on clinical services and opens up the possibility of regular, even daily, assessment. Over time, this creates a richer and more accurate picture of an individual’s cognitive health, potentially enabling decline to be spotted much earlier than current systems allow. Machine-learning personalisation strengthens this further, tailoring the assessment to each person’s baseline rather than relying on one-size-fits-all thresholds.

By releasing its datasets openly, the team is also accelerating progress across the global research community.

Together, these advances position UWE Bristol at the forefront of creativity-led healthcare innovation – championing solutions that are accessible, intuitive and built for real-world impact.