Cambridge scientists have created an AI tool that can predict, in four out of five cases, if people with early dementia will stay stable or develop Alzheimer’s.
“Our tool is more sensitive than current methods at predicting Alzheimer’s from mild symptoms,” said Zoe Kourtzi.
This new approach could reduce the need for expensive tests and improve early treatment with lifestyle changes or new medicines.
Dementia is a major global health issue, affecting over 55 million people and costing $820 billion annually. Cases are expected to nearly triple in the next 50 years.
The main cause of dementia is Alzheimer’s disease, making up 60-80% of cases. Early detection is crucial for effective treatment, but current methods like PET scans or lumbar punctures are expensive and not always available, leading to possible misdiagnosis or late diagnosis for many patients.
Scientists from the University of Cambridge have developed a machine learning model that predicts if and how fast a person with mild memory issues will develop Alzheimer’s. This new model is more accurate than current methods and uses non-invasive, low-cost data from cognitive tests and MRI scans. Their research was published in eClinical Medicine.
They tested the model with data from 600 US patients and 900 patients from UK and Singapore memory clinics.
The algorithm accurately predicted whether people with mild cognitive impairment would develop Alzheimer’s within three years. It identified future Alzheimer’s patients 82% of the time and those who wouldn’t develop it 81% of the time using only cognitive tests and MRI scans.
The algorithm was three times more accurate than current methods, reducing misdiagnosis significantly.
It also categorized patients into three groups: stable symptoms, slow progression, and rapid progression. These predictions were validated over six years, helping to identify who might benefit from early treatment and who needs close monitoring.
Importantly, the 50% of people with memory loss but stable symptoms could be directed to different treatments, as their issues might be due to anxiety or depression, not dementia.
Professor Zoe Kourtzi from Cambridge said, “Our tool, using only cognitive tests and MRI scans, is much more sensitive than current methods at predicting if and how quickly someone with mild symptoms will develop Alzheimer’s.”
This tool can improve patient care by identifying those who need close monitoring and reducing anxiety for those likely to remain stable, while also cutting down on unnecessary expensive tests.
The algorithm was tested on a research cohort and validated with data from nearly 900 patients in UK and Singapore memory clinics, showing its real-world applicability.
Dr. Ben Underwood, a psychiatrist at CPFT and University of Cambridge, said memory problems are common as we age and can cause worry about dementia. He finds it exciting that existing information might help reduce this uncertainty, which could become even more important with new treatments.
Professor Kourtzi emphasized that AI models need good data to work well. Their model, trained on real patient data, is promising for healthcare settings. They aim to expand it to other types of dementia and different data sources.
The goal is to improve early detection and treatment of dementia, match patients to clinical trials, and speed up new drug discoveries. This work involved collaboration with experts from the University of Birmingham and the National University of Singapore and was funded by various health and research organizations.
Journal reference:
Lee, LY & Vaghari, D et al. Robust and interpretable AI-guided marker for early dementia prediction in real-world clinical settings. eClinMed. DOI: 10.1016/j.eclinm.2024.102725.
>>> Read full article>>>
Copyright for syndicated content belongs to the linked Source : Tech Explorist – https://www.techexplorist.com/ai-outperforms-clinical-tests-predicting-alzheimers-disease/86117/#utm_source=rss&utm_medium=rss&utm_campaign=ai-outperforms-clinical-tests-predicting-alzheimers-disease