Researchers at Cambridge University have developed an artificial intelligence (AI) model that is effective at predicting if patients who show early dementia symptoms will remain stable or develop Alzheimer's.
In findings published in eClinical Medicine, the machine-learning tool outperformed current standard diagnostic tools used for prediction.
Alzheimer's disease is the main cause of dementia, accounting for 60% to 80% of cases. However, while early diagnosis is critical for more effective intervention and better outcomes, the effective tools available, such as positron emission tomography (PET) scans or lumbar puncture, are often still costly, invasive, and are not accessible for all patients.
To build the model, the research team, led by scientists at the University of Cambridge used data from cognitive tests and structural magnetic resonance imaging (MRI) scans showing gray matter atrophy from over 400 people in the U.S. who were part of a research cohort. Cognitive tests and gray matter atrophy are standard clinical markers for predicting the progression to Alzheimer's; the data are collected routinely, noninvasively, and are low-cost.
The team then tested the algorithm using real-world patient data from an additional 600 patients from the same cohort, as well as using longitudinal data from 900 patients at memory clinics in the U.K. and Singapore. In a statement, senior author Professor Zoe Kourtzi noted that this was important for validating the model, as it needed to apply to real-world settings.
"AI models are only as good as the data they are trained on. To make sure ours has the potential to be adopted in a healthcare setting, we trained and tested it on routinely-collected data not just from research cohorts, but from patients in actual memory clinics. This shows it will be generalizable to a real-world setting," Kourtzi said.
The results showed that the model was approximately three times more accurate than the standard clinical markers in differentiating patients with mild cognitive impairment that was stable and patients who progressed to Alzheimer's within a three-year period, correctly identifying those who developed Alzheimer's with an accuracy rate of 82%, and correctly identifying those who didn't with an accuracy rate of 81% solely from cognitive tests and an MRI scan.
Furthermore, the model was effective at stratifying the patients who developed Alzheimer's into three groups using data from each person's first visit to the memory clinic: Patients whose symptoms would remain stable (around 50% of the cohort), patients who would progress to Alzheimer's slowly (around 35%), and patients who would progress to Alzheimer's more rapidly (approximately 15%). Follow-up data for the patients over six years was examined to validate the predictions.
"This has the potential to significantly improve patient well-being, showing us which people need closest care, while removing the anxiety for those patients we predict will remain stable. At a time of intense pressure on healthcare resources, this will also help remove the need for unnecessary invasive and costly diagnostic tests," Kourtzi noted.
In future work, the Cambridge team hope to apply the algorithm to other forms of dementia, such as vascular dementia and frontotemporal dementia, and will incorporate different types of data, such as markers from blood tests.
