Share
Healthy Minds: AI for better brain & mental health
Reported by Adisetu Joy Malih, CSaP Policy Intern (Jan-April 2023)
CSaP's Policy Leaders Fellowship met earlier this month to explore the opportunities and challenges of Artificial Intelligence systems in healthcare. Leading the discussion was Professor Zoe Kourtzi, Professor of Experimental Psychology in the Department of Psychology at the University of Cambridge.
Professor Kourtzi began her presentation by noting that dementia has become a global challenge. According to recent research, every two seconds someone is diagnosed with dementia. Around 50% of individuals who experience symptoms of dementia, like memory loss, remain undiagnosed, and about 35% of dementia cases are misdiagnosed. After stressing the urgency of predicting dementia early and developing tailored interventions to improve patient outcomes, Professor Kourtzi introduced her team’s work on early diagnosis of dementia using AI with machine learning tools.
An early prediction has significant implications for the treatment and management of dementia
Professor Kourtzi explained that part of their work involved developing machine learning tools to detect dementia at a very early stage using data which includes brain scans, biological data, memory tests and information on demography. Explaining how these machine learning tools generate scores that group patients into stable, slowly progressing and rapidly progressing, Professor Kourtzi emphasised that for patients diagnosed with mild cognitive impairment, the machine learning tools were 80% accurate in identifying those that later developed dementia.
Professor Kourtzi said that it takes several years from when structural changes begin in the brain to when symptoms start to show. She also argued that current diagnostic systems involve several scans and cognitive tests that are expensive, time-consuming and, most importantly, cause patients' anxiety while they wait for results. Arguing that AI provides a window of opportunity for early intervention via predicting dementia early, Professor Kourtzi said that lifestyle changes can effectively help slow 40% of dementia.
Professor Kourtzi also noted that in addition to early diagnosis, her team is looking at how AI can help clinicians provide tailored interventions. Within this context, she approached cardiovascular diseases that impair cognition and said that the machine learning tools had identified individuals diagnosed with mild cognitive impairment who went down the dementia pathway due to co-morbidities such as cardiovascular disease.
Emphasising that dementia is a global challenge that requires a global solution, Professor Kourtzi highlighted that AI has the potential as a clinical decision support system for the diagnosis, treatment, and management of dementia. She further argued that AI could guide the patient at the right time to the right pathway and discussed that AI can make clinical trials faster, more impactful and adaptable.
Concerns about AI systems in healthcare
At the end of the seminar, a question and answer session was held with audience participation and the discussion broadened on different themes such as return on investment, ethical concerns and distrust of AI. Within this scope, Professor Kourtzi talked about how machine learning tools are transferable, interpretable, and generalisable along with her team's work on return on investment. She also noted that more work is needed to gain public trust in AI, and informed the audience that her team engage with the public to better understand their worries.
During the session, participants agreed that regulating AI systems in healthcare is currently a grey area that needs much work clarifying questions like what should be done to move from research to adaptation and who is responsible for regulation.
Professor Kourtzi concluded the lunch discussion by emphasising that AI offers opportunities for early intervention in individuals with dementia but tackling dementia will require a global approach.
Image by Hal Gatewood - Unsplash