AI takes lead over clinical tests in Alzheimer’s prediction
Share
Cambridge University researchers have developed an artificial intelligence tool that could transform how we diagnose and manage Alzheimer’s disease.
This innovative algorithm outperforms current clinical methods, accurately predicting in 82% of cases whether individuals with mild cognitive impairment will develop Alzheimer’s within three years.
Key features of the AI tool:
- Uses readily available data from cognitive tests and MRI scans
- More accurate than invasive and expensive diagnostic tests like PET scans
- Can categorise patients into stable, slow-progressing, or fast-progressing groups
- Validated using real-world data from memory clinics in the UK and Singapore
The tool’s potential benefits include:
- Earlier interventions when treatments may be most effective
- Reduced need for invasive diagnostic procedures
- Better allocation of healthcare resources
- Decreased anxiety for patients unlikely to progress to Alzheimer’s
Professor Zoe Kourtzi, the study’s senior author, emphasises that this AI model could significantly improve patient care by identifying those needing the closest attention while reassuring others likely to remain stable.
The researchers trained their model on data from over 400 individuals in a US research cohort and tested it on 1,500 participants from the US, UK, and Singapore. This diverse dataset ensures the tool’s applicability in real-world clinical settings.
Looking ahead, the team plans to expand its model to other forms of dementia and incorporate additional data types, such as blood test markers. Their ultimate goal is to create a scalable tool that can help clinicians make more informed decisions about patient care and accelerate drug discovery for disease-modifying treatments.
As the global dementia crisis grows, with cases expected to triple in the next 50 years, innovative solutions like this AI tool offer hope for better management and potential prevention of Alzheimer’s disease.
Upcoming Events
6th Annual Aged Care Week
June 18, 2025
Subscribe
We send emails,
but we do not spam
Join our mailing list to be on the front lines of healthcare , get exclusive content, and promos.