Using AI and speech technology to transform the detection of neurological conditions

CognoSpeak is a new tool for detecting the early signs of dementia. Using a virtual AI agent to guide patients through conversations at home, the way they speak can be analysed for thousands of hidden clues about brain health. 

Timely detection of neurological conditions such as dementia remains a challenge in healthcare. Across the UK, memory and cognitive assessment services are under increasing pressure, with growing patient demand and limited access to specialist testing leading to long waiting times. These delays can slow access to treatment and support at a critical stage for both individuals and their families.

New technologies that enable earlier and more accessible assessments could help ease this pressure, allowing clinics to see more patients and ensure they get the right care sooner. 

At the University of Sheffield, Professor Heidi Christensen, Dr Dan Blackburn, Dr Leslie Ing, and Professor Christopher McDermott are pioneering the use of artificial intelligence and speech technology to improve the early detection of neurological conditions. Through the development of two digital tools, CognoSpeak and CognoMND, the team is helping to transform how cognitive changes are identified and assessed. 

This project is supported by a grant from LifeArc and supported by the NIHR Sheffield Biomedical Research Centre and NIHR HealthTech Research Centre in Long term Conditions.

Detecting cognitive change through conversation

Developed by an interdisciplinary team led by Professor Heidi Christensen from the University’s School of Computer Science and Dr Dan Blackburn from the School of Medicine and Population Health, CognoSpeak uses a virtual agent displayed on a screen to guide patients through a series of conversations and cognitive tasks.

The system asks the patient simple, memory-probing questions, similar to those used in outpatient consultations - and guides them through tasks like describing pictures and verbal fluency exercises. Behind the scenes, it then analyses speech and language patterns for subtle indicators of neurological conditions, including dementia and Alzheimer's disease. 

CognoSpeak is non-invasive and can be completed online from the comfort of a patient’s home, making the process more convenient and less stressful for patients. 

“What we wanted to do was develop a tool that people could use at home. We focused on speech because it is something that can be easily recorded on personal devices, making remote use possible”, explains Dr Leslie Ing, Clinical Research Fellow at the Sheffield Institute for Translational Neuroscience (SITraN).

“From our clinical experience, we noticed that people with thinking difficulties often speak in subtly different ways. Using the technology we have developed in-house, we are now able to extract thousands of features from the speech signal”, adds Dr Ing.

CognoSpeak has been developed over the past eight years and has already been tested with more than 350 participants, including follow-up data from over 40 individuals. Early findings show an accuracy of 87% in detecting mild cognitive impairment and Alzheimer’s disease, as well as over 90% sensitivity (meaning more than 90% of MCI and dementia cases are correctly identified) in real-world clinical and home settings. The tool has also been adapted for use with stroke survivors and people living with Motor Neuron Disease (MND), demonstrating its ability to support the assessment of different causes of cognitive impairment. 

This work builds on Sheffield’s internationally recognised strengths in artificial intelligence, speech technology and neuroscience research, bringing together expertise from healthcare, computer science and clinical practice to tackle one of the biggest healthcare challenges facing an ageing population.

“An important aim for us is the translational aspect - turning research into a practical tool that can help make a difference to patients’ lives. By bringing together expertise in computer science, speech analysis, and neurology, we are able to make this work in an integrated way”, adds Dr Ing.

Addressing gaps in MND care

The next phase of the research project focuses on MND. 

Around half of individuals with MND experience changes in thinking, behaviour or personality. Despite this, cognitive symptoms can sometimes receive less attention during clinical care, where immediate priorities often focus on physical challenges such as mobility, swallowing and breathing.

At the same time, there is a shortage of specialist staff able to carry out recommended cognitive assessments.

To address this gap, CognoMND is being developed to identify cognitive and behavioural changes associated with MND using the same AI-driven speech analysis approach.

Led by Professor Christopher McDermott, Dr Dan Blackburn, and Dr Leslie Ing, the tool aims to detect subtle signs of cognitive change that may not be easily identified during routine appointments.

“For healthcare providers, this kind of tool helps focus limited time and resources on the people who need care most urgently. In MND, many of the most immediate concerns relate to physical symptoms, and as a result, cognitive changes can sometimes receive less attention, even though they can have a significant impact on patients and families. This means that people experiencing thinking difficulties may not always be identified early enough to receive the support they need”, explains Dr Ing. 

“A tool like CognoMND could help identify those at risk earlier, allowing them to be prioritised for appropriate assessment and care”, adds Dr Ing.

For some patients, cognitive testing helped identify or make sense of changes, or validated concerns. Even when results showed no impairment, the process helped reassure and reduce uncertainty for both patients and their families. 

“It’s a relief to have a bit of clarity, even just knowing that it’s not something else going on,” one patient shared.

For caregivers, testing could also provide an objective lens to support communication and reduce conflict, especially in the presence of differing insight.

As one caregiver put it: “Sometimes you need more than just the opinion of the person or the caregiver, a more scientific way of saying, ‘yeah, that tracks.’”

The team is now working towards larger clinical trials to ensure the CognoMND system is accurate, inclusive and effective across diverse patient populations.

Shaped by lived experience 

A defining feature of this research is its strong commitment to patient involvement. From the outset, a lived experience advisory group has worked in partnership with researchers to guide the development of both tools.

This group has helped to shape everything from the design of the interface to how tasks are presented. Their feedback has also guided how the digital agent engages with users and how the platform accommodates different preferences.  

The CognoMND team have worked alongside people with MND and care-partners as collaborators, placing lived experience at the heart of the design process, helping build the technology around the needs of those it is intended to support.

“The feedback from the lived experienced groups helped ensure the platform was accessible and practical for a diverse range of users. They even contributed to the design of the user interface and overall usability, including the clarity of instructions, button size and placement, and how tasks were presented and completed”, explains Dr Ing.

“I think our set up is unique: the clinicians who actually see patients are in constant conversation with the computer scientists building the technology, so what we learn in the clinic directly shapes what gets built, and the tool keeps improving as a result. The back-and-forth between clinic and lab means the tool gets built around the people it's actually for - people living with MND”, adds Dr Ing.

A more accessible future for diagnosis 

As demand for neurological care continues to rise, tools like CognoSpeak and CognoMND offer a new pathway for early detection - supporting clinicians while improving access to assessment for patients.

"As a clinician, I see the devastating impact conditions like dementia and MND can have on patients and families, and not enough people are getting tested early enough."

Dr Leslie Ing

Clinical Research Fellow, Sheffield Institute for Translational Neuroscience (SITraN)

“We are interested in how digital technology can help us innovate the way healthcare is delivered, while ensuring that the tools we develop remain clinically informed, transparent and meaningful for patients. Developing the technology in-house allows us to better understand how it is designed, tested and implemented, which is essential if it is going to be used safely and effectively in real healthcare settings”, explains Dr Ing.

Looking ahead, the aim is to evaluate the system through clinical trials involving a diverse group of people living with MND. This will help ensure the tool is not only accurate, but also accessible and usable for the people it is intended to support.

Using AI and speech technology to transform the detection of neurological conditions | Research | The University of Sheffield 

The NIHR funds, enables and delivers world-leading health and social care research that improves people's health and wellbeing and promotes economic growth.

NIHR Biomedical Research Centres (BRCs) are collaborations between NHS organisations and universities. They bring together academics and clinicians to translate scientific discoveries into potential new treatments, diagnostics and technologies.