Using AI to streamline dementia diagnosis and improve health equity for patients
At the University of Sheffield, AI is being used to reduce wait times and improve outcomes for people who may be at risk of developing dementia
As the world grapples with an ageing population and a rise in age-related cognitive disorders, early detection has become crucial. However, traditional diagnostic methods can be time-consuming, resource-intensive and often rely on subjective assessments.
To tackle the problems created by delays in the diagnostic processes, a team of researchers at the University of Sheffield, led by Dan Blackburn, a senior lecturer in neuroscience, and Heidi Christensen, professor in computer science, with support from Caitlin Illingworth, a neuroscience PhD student, are developing CognoSpeak, an online tool harnessing AI to analyse speech patterns and cognitive cues. In doing so, CognoSpeak can identify subtle markers of cognitive decline that might be missed during routine evaluations.
“Typically, if someone is worried about their memory, the first point of contact is their GP,” Blackburn explained on an episode of the THE Connect podcast. “GPs then carry out some assessments before deciding which specialist clinics to refer patients to. Currently, long waits mean people aren’t being diagnosed at early enough stages where interventions could make the biggest impact.”
At the University of Sheffield, collaboration has played a major role in the development of CognoSpeak, including a close connection between the campus and the local hospital. “About 10 years ago, Dan and I started working together,” Christensen said. “The initial inspiration for CognoSpeak came from the conversations Dan was having with patients. We wondered if there was a way of automating patient questioning while still picking up on the cues contained within their responses.”
CognoSpeak consists of a virtual agent that replicates the memory-probing questions and cognitive tests that have traditionally been carried out by GPs or consultants. This is coupled with software that analyses the audio being recorded. “Developing the tool has been a multidisciplinary effort over a number of years,” Christensen said.
Despite the potential benefits, the team behind CognoSpeak understands that the use of AI and automation, especially in a healthcare setting, can be accompanied by worries over a lack of human contact.
“We’ve done lots of work with patients with lived experience of dementia,” Blackburn noted. “Even among older generations who may still prefer to see a human doctor, there is an acceptance that pressures on health services mean this isn’t always possible. Overall, the feedback on CognoSpeak has been overwhelmingly positive.”
Another issue raised its head when the CognoSpeak team began engaging with members of the local community with different language backgrounds or cultural differences. “I travel to local community centres to start conversations about what dementia is, what normal healthy ageing is, and what we are hoping to do with CognoSpeak,” Illingworth said. “We found that sometimes the assessments we used were not suitable for individuals where English was not their first language. This wasn't necessarily something we considered before having conversations with local community members.
Although accessibility challenges – whether due to cultural biases or technology deserts – are certainly considerations for the CognoSpeak team, there remains much optimism regarding the platform.
“Looking to the future, I hope CognoSpeak can ultimately be put into the hands of GPs,” Christensen explained. “There are some challenges to overcome first, in terms of medical device approval, for example, but we are progressing all the time and working with other cohorts, such as stroke survivors, sufferers of motor neurone disease, people with Parkinson’s and others. So there are a lot of sister projects on the horizon.”
“I’m hopeful that CognoSpeak can help create more health equity throughout the dementia diagnostic process,” Illingworth added. “It can be a scary diagnosis to seek, so any way we can streamline and improve this could deliver major improvements for patients.”
This study is supported and funded by the National Institute for Health and Care Research (NIHR) via the Invention for Innovation Programme and the Sheffield Biomedical Research Centre. The views expressed are those of the authors and not necessarily those of the NIHR or the Department of Health and Social Care.
Built with Shorthand