summary: A new AI tool uses cognitive tests and MRI scans to predict Alzheimer’s disease progression with 82% accuracy, better than current methods. The tool could reduce the need for costly testing 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 accurately identified Alzheimer’s disease progression in 82% of cases.
- It uses non-invasive, low-cost data for predictions.
- Patients can be categorized into groups based on the rate at which the disease progresses.
sauce: University of Cambridge
Scientists at the University of Cambridge have developed an artificial intelligence tool that can predict whether people with early signs of dementia will remain stable or develop Alzheimer’s disease in four out of five cases.
The researchers say this new approach could reduce the need for invasive and costly diagnostic tests and improve outcomes at an earlier stage, when interventions such as lifestyle changes and new drugs may be most effective.
Dementia is a major global healthcare challenge, affecting more than 55 million people worldwide and costing an estimated $820 billion annually, with the number of patients expected to nearly triple over the next 50 years.
The leading cause of dementia is Alzheimer’s disease, accounting for 60-80% of cases. Early detection is important because treatment is likely to be most effective, but early diagnosis and prognosis of dementia may not be accurate without invasive or expensive tests such as a positron emission tomography (PET) scan or lumbar puncture, which are not available at all memory clinics.
As a result, up to a third of patients may be misdiagnosed, and others may receive a diagnosis too late to benefit from treatment.
A team led by scientists from the University of Cambridge’s School of Psychology has developed a machine learning model that can predict whether people with mild memory and thinking problems will develop Alzheimer’s disease, and if so, how quickly.
According to published research Clinical Medicineand has been shown to be more accurate than current clinical diagnostic tools.
To build their model, the researchers used non-invasive, low-cost patient data collected routinely from more than 400 patients in a US study cohort, including cognitive tests and structural MRI scans that show grey matter atrophy.
The researchers then tested their model using real patient data from an additional 600 participants from the US cohort and, importantly, longitudinal data from 900 people from memory disorder clinics in the UK and Singapore.
The algorithm was able to distinguish between people with stable mild cognitive impairment and those who progressed to Alzheimer’s disease within three years. Using only cognitive tests and MRI scans, it correctly identified those who went on to develop Alzheimer’s disease in 82% of cases and those who did not in 81% of cases.
The algorithm was approximately three times more accurate at predicting progression to Alzheimer’s disease than the current standard of care, i.e. standard clinical markers (such as gray matter atrophy and cognitive scores) and clinical diagnosis, indicating that the model can significantly reduce misdiagnosis.
This model allowed the researchers to classify Alzheimer’s patients into three groups, using data from each person’s first visit to the memory clinic: those whose symptoms remained stable (about 50% of participants), those whose Alzheimer’s disease progressed more slowly (about 35%), and those whose Alzheimer’s disease progressed more rapidly (the remaining 15%).
These predictions were validated by looking at six years of follow-up data, which is important because it helps identify people early on who may benefit from new treatments, as well as those who need closer monitoring as their condition may worsen quickly.
Importantly, the 50% of people who have symptoms such as memory loss but remain stable are better off being directed down a different clinical pathway as their symptoms may be caused by something other than dementia, such as anxiety or depression.
Lead author Professor Zoe Curzi, from the University of Cambridge’s School of Psychology, said: “Despite using only data from cognitive tests and MRI scans, we have developed a tool that is far more sensitive than current approaches in predicting whether mild symptoms will progress to Alzheimer’s disease, and if so, whether that progression will be fast or slow.”
“This has the potential to significantly improve patient health outcomes by showing us which patients need the most attention and removing anxiety for those who are predicted to remain in a stable condition. At a time when healthcare resources are under strain, this will also help eliminate the need for unnecessary invasive and costly diagnostic tests.”
The researchers tested the algorithm on data from the study cohort, but the algorithm was then validated using independent data including about 900 people who attended memory clinics in the UK and Singapore.
In the UK, patients were recruited through the NHS Memory Clinic Quantitative MRI Study (QMIN-MC), led by study co-author Dr Timothy Littman from Cambridge University Hospitals NHS Trust and Cambridgeshire and Peterborough NHS Foundation Trust (CPFT).
The researchers say this shows potential application to real patients and clinical settings.
Dr Ben Underwood, CPFT, Honorary Consultant Psychiatrist and Associate Professor of Psychiatry at the University of Cambridge, said: “Memory problems become common as we age, and in my clinical practice I see the uncertainty of whether this is the first sign of dementia causing great anxiety for patients and their families, and frustrating for doctors who want to give definitive answers.”
“The fact that information we already have may be able to reduce this uncertainty is encouraging and is likely to become even more important as new treatments emerge.”
“An AI model is only as good as the data used to train it,” said Professor Kurzi. “To ensure our model could be adopted in clinical practice, we trained and tested it not only on a research cohort, but also on routinely collected data from real memory clinic patients. This shows that our AI model can be generalized to real-world environments.”
The team now hopes to extend the model to other types of dementia, such as vascular and frontotemporal dementia, using different types of data, such as blood test markers.
“To tackle the growing health challenge posed by dementia, we need better tools to identify and intervene at the earliest possible stage,” Prof Kurzi added.
“Our vision is to scale AI tools to help clinicians put the right people on the right diagnosis and treatment pathway at the right time. Our tools will help match the right patients to clinical trials and accelerate the discovery of new drugs for disease-modifying treatments.”
About this AI and Alzheimer’s research news
author: Ben Underwood
sauce: University of Cambridge
contact: Ben Underwood – University of Cambridge
image: Image courtesy of Neuroscience News
Original Research: Open access.
“Robust and Interpretable AI-Derived Markers for Predicting Early Dementia in Real-World Clinical Practice” by Ben Underwood et al. Clinical Medicine
Abstract
Robust and interpretable AI-derived markers for early dementia prediction in real-world clinical practice
background
Early prediction of dementia would have a significant impact on clinical management and patient outcomes. However, sensitive tools for early patient stratification are still lacking, resulting in patients going undiagnosed or being misdiagnosed. Machine learning models for dementia prediction are rapidly expanding, but limited model interpretability and generalizability hinder clinical application.
Method
We build a robust and interpretable predictive prognostic model (PPM) and validate its clinical utility using routinely collected, non-invasive and low-cost (cognitive testing, structural MRI) real patient data. To increase scalability and generalizability to the clinic, we 1) train the PPM using clinically relevant predictors (cognitive testing, gray matter atrophy) that are common across study and clinical cohorts, and 2) test the PPM predictions using independent multi-center real data from memory clinics across countries (UK, Singapore).
Investigation result
PPM reliably predicts whether patients at early stages of disease (MCI) will remain stable or progress to Alzheimer’s disease (AD) (Accuracy: 81.66%, AUC: 0.84, Sensitivity: 82.38%, Specificity: 80.94%). PPM is generalized from the study to real memory clinic patient data and its predictions are validated against longitudinal clinical outcomes. PPM allows us to derive individual AI-derived multimodal markers (predictive prognostic indicators) that predict progression to AD more accurately than standard clinical markers (gray matter atrophy, cognitive scores, PPM-derived markers: Hazard Ratio = 3.42, p = 0.01) or clinical diagnosis (PPM-derived markers: Hazard Ratio = 2.84, p < 0.01), reducing misdiagnosis.
interpretation
Our findings were validated against longitudinal, multicenter patient data from across countries and provide evidence of a robust and explainable clinical AI-derived marker for early dementia prediction, with great potential for implementation in clinical practice.
Funding
The Wellcome Trust, the Royal Society, Alzheimer’s Research UK, the Alzheimer’s Drug Discovery Foundation Diagnostics Accelerator and the Alan Turing Institute.
