Resume: A new AI tool predicts Alzheimer’s progression with 82% accuracy using cognitive tests and MRI scans, outperforming current methods. The tool could reduce the need for expensive tests and improve early intervention.
Alzheimer’s disease is the leading cause of dementia, affecting more than 55 million people worldwide.
Key Facts:
- The AI tool correctly identified Alzheimer’s progression in 82% of cases.
- It uses non-invasive, low-cost data for predictions.
- Patients can be divided into groups based on the speed at which the disease progresses.
Source: University of Cambridge
Cambridge scientists have developed an artificial intelligence tool that can predict in four out of five cases whether people with early signs of dementia will remain stable or develop Alzheimer’s disease.
According to the team, this new approach could reduce the need for invasive and costly diagnostic tests while improving treatment outcomes early on, when interventions such as lifestyle changes or new medications still have a chance to work optimally.
Dementia is a major global health challenge, affecting more than 55 million people worldwide, with an estimated annual cost of $820 billion. The number of cases is expected to nearly triple in the next 50 years.
The leading cause of dementia is Alzheimer’s disease, which accounts for 60-80% of cases. Early detection is crucial, as this is when treatments are likely to be most effective, but early diagnosis and prognosis of dementia may not be accurate without the use of invasive or expensive tests such as positron emission tomography (PET) scans or lumbar punctures, which are not available at all memory clinics.
This can result in a third of patients being misdiagnosed, or in others being diagnosed too late, making treatment ineffective.
A team led by scientists from the University of Cambridge’s Department of Psychology has developed a machine learning model that can predict whether and how quickly someone with mild memory and thinking problems will develop Alzheimer’s disease.
In research published in eClinical Medicinethey demonstrate that it is more accurate than current clinical diagnostic tools.
To build their model, the researchers used routinely collected, noninvasive, and inexpensive patient data — cognitive tests and structural MRI scans showing gray matter atrophy — from more than 400 individuals who were part of a US research cohort.
They then tested the model using real patient data from an additional 600 participants from the US cohort and – importantly – longitudinal data from 900 people from memory clinics in the UK and Singapore.
The algorithm was able to distinguish between people with stable mild cognitive impairment and those who developed Alzheimer’s within a three-year period. It was able to correctly identify people who developed Alzheimer’s 82% of the time and those who did not 81% of the time, based only on cognitive tests and an MRI scan.
The algorithm was about three times more accurate in predicting progression to Alzheimer’s than the current standard of care; that is, standard clinical markers (such as gray matter atrophy or cognitive scores) or clinical diagnosis. This shows that the model could significantly reduce misdiagnosis.
Using data from each person’s first visit to the memory clinic, the model allowed the researchers to divide people with Alzheimer’s into three groups: those whose symptoms remained stable (about 50 percent of participants), those whose Alzheimer’s would slowly worsen (about 35 percent), and those whose disease would worsen more quickly (the remaining 15 percent).
These predictions were validated when we looked at six years of follow-up data. This is important because it can help identify those people early so they can benefit from new treatments, while also identifying those people who need to be monitored closely because their condition is likely to deteriorate rapidly.
Importantly, the 50% of people who have symptoms such as memory loss but remain stable may be better referred to another clinical treatment pathway, as their symptoms may be due to causes other than dementia, such as anxiety or depression.
Lead author Professor Zoe Kourtzi from the Department of Psychology at the University of Cambridge said: “We have developed a tool that, despite only using data from cognitive tests and MRI scans, is much more sensitive than current methods for predicting whether someone will progress from mild symptoms to Alzheimer’s – and if so, whether this progression will be rapid or slow.
“This has the potential to significantly improve patient wellbeing by showing us which people need the most immediate care, while removing 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.”
Although the researchers tested the algorithm on data from a research cohort, it was validated using independent data from nearly 900 individuals attending memory clinics in the United Kingdom and Singapore.
In the UK, patients were recruited through the Quantitative MRI in NHS Memory Clinics Study (QMIN-MC) led by study co-author Dr Timothy Rittman from Cambridge University Hospitals NHS Trust and Cambridgeshire and Peterborough NHS Foundation Trusts (CPFT).
According to the researchers, this shows that the method can be applied in a real clinical setting with patients.
Dr Ben Underwood, Honorary Consultant Psychiatrist at CPFT and Assistant Professor in the Department of Psychiatry at the University of Cambridge, said: “Memory problems are common as we get older. In clinic I see how uncertainty about whether these could be the first signs of dementia can cause a lot of worry for people and their families, and how frustrating it can be for clinicians who would rather give definitive answers.
“The fact that we can reduce this uncertainty with the information we already have is exciting and will likely become even more important as new treatments emerge.”
Professor Kourtzi said: “AI models are only as good as the data they are trained on. To ensure ours has the potential to be applied in a healthcare setting, we trained and tested it on routinely collected data, not just from research cohorts, but from patients in real memory clinics. This shows that it is generalisable to a real-world setting.”
The team now hopes to extend their model to other forms of dementia, such as vascular dementia and frontotemporal dementia, using other types of data, such as markers from blood tests.
Professor Kourtzi added: “If we are to tackle the growing health challenge of dementia, we need better tools to identify and intervene in dementia at the earliest possible stage.
“Our vision is to scale our AI tool to help clinicians assign the right person to the right diagnostic and treatment pathway at the right time. Our tool can help match the right patients to clinical trials, accelerating the discovery of new drugs for disease-modifying treatments.”
About this AI and Alzheimer’s research news
Author: Ben Underwood
Source: University of Cambridge
Contact: Ben Underwood – University of Cambridge
Image: The image is attributed to Neuroscience News
Original research: Open access.
“Robust and interpretable AI-driven marker for early prediction of dementia in real-world clinical settings” by Ben Underwood et al. eClinical Medicine
Abstract
Robust and interpretable AI-driven marker for early dementia prediction in real-world clinical settings
Background
Early prediction of dementia has major implications for clinical management and patient outcomes. Yet, we still lack sensitive tools to stratify patients early, resulting in patients being misdiagnosed or undiagnosed. Despite the rapid expansion of machine learning models for dementia prediction, limited model interpretability and generalizability hamper translation to the clinic.
Methods
We build a robust and interpretable predictive prognostic model (PPM) and validate its clinical utility using real-world, routinely collected, non-invasive, and inexpensive (cognitive tests, structural MRI) patient data. To improve scalability and generalizability to the clinic, we: 1) train the PPM with clinically relevant predictors (cognitive tests, gray matter atrophy) that are common in research and clinical cohorts, 2) test PPM predictions with independent multicenter real-world data from memory clinics in different countries (UK, Singapore).
Findings
PPM robustly predicts (accuracy: 81.66%, AUC: 0.84, sensitivity: 82.38%, specificity: 80.94%) whether patients in early disease stages (MCI) will remain stable or progress to Alzheimer’s disease (AD). PPM generalizes from real-world patient data research in memory clinics and its predictions are validated against longitudinal clinical outcomes. PPM allows us to derive an individualized AI-driven multimodal marker (i.e., predictive prognostic index) that predicts progression to AD more accurately than standard clinical markers (gray matter atrophy, cognitive scores; PPM-derived marker: hazard ratio = 3.42, p = 0.01) or clinical diagnosis (PPM-derived marker: hazard ratio = 2.84, p
Interpretation
Our results provide evidence for a robust and explainable clinical AI-driven marker for the prediction of early dementia that is validated using longitudinal, multicenter patient data from different countries and has great potential for application in clinical practice.
Financing
Wellcome Trust, Royal Society, Alzheimer’s Research UK, Alzheimer’s Drug Discovery Foundation Diagnostics Accelerator, Alan Turing Institute.