In a groundbreaking study published in Nature Mental HealthResearchers at Queen Mary University of London have developed a new method to predict dementia with more than 80% accuracy up to nine years before clinical diagnosis. The method, which outperforms traditional memory tests and measures of brain shrinkage, relies on detecting changes in the brain’s default mode network (DMN) using functional magnetic resonance imaging (fMRI).
Dementia is an umbrella term used to describe a variety of diseases characterized by a progressive decline in cognitive function severe enough to interfere with daily living and independent functioning. It affects memory, thinking, orientation, comprehension, calculation, learning ability, language, and judgment.
Alzheimer’s disease is the most common cause of dementia, accounting for 60–70% of cases. Other types of dementia include vascular dementia, dementia with Lewy bodies, and frontotemporal dementia.
Dementia is a progressive disease, meaning that symptoms worsen over time, often leading to significant impairment in daily activities and quality of life. There is currently no cure for dementia, and treatments focus primarily on symptom management and support for patients and their caregivers.
Early diagnosis is important because it paves the way for interventions that can slow disease progression, improve quality of life, and give people and their families more time to plan for the future. Traditional diagnostic methods, such as memory tests and brain scans to detect atrophy, often detect the disease only after significant neuronal damage has occurred. These methods are not sensitive enough to detect the very early changes in brain function that precede clinical symptoms.
“Predicting who will develop dementia in the future will be essential to developing treatments that can prevent the irreversible loss of brain cells that causes the symptoms of dementia,” said Charles Marshall, who led the research team from the Centre for Preventive Neurology at Queen Mary’s Wolfson Institute of Population Health. “Although we are increasingly able to detect the brain proteins that can cause Alzheimer’s disease, many people live for decades with these proteins in their brain without developing symptoms of dementia.”
“We hope that the measure of brain function we have developed will allow us to be much more precise about whether a person will actually develop dementia, and when, so that we can determine whether they might benefit from future treatments.”
The study used a nested case-control approach, using data from the UK Biobank, a large-scale biomedical database. The researchers focused on a subset of participants who had undergone functional magnetic resonance imaging (fMRI) scans and were diagnosed with or later developed dementia. The sample included 148 dementia cases and 1,030 matched controls, ensuring a robust comparison group by matching age, sex, ethnicity, handedness, and geographic location of the MRI scan center.
Participants underwent resting-state fMRI (rs-fMRI) scans, which measure brain activity by detecting changes in blood flow. The researchers specifically targeted the default mode network (DMN), a network of brain regions that are active at rest and involved in higher-level cognitive functions such as social cognition and self-referential thinking.
Using a technique called dynamic causal modeling (DCM), they analyzed the rs-fMRI data to estimate the actual connectivity between different regions of the DMN. This method goes beyond simple correlations to model the causal influence of one brain region on another, providing a detailed picture of neural connectivity.
The researchers then used these connectivity estimates to train a machine learning model. This model aimed to distinguish between individuals who were likely to develop dementia and those who would not. The training process involved a rigorous cross-validation technique to ensure the model’s reliability and avoid overfitting. In addition, a prognostic model was developed to predict the time to dementia diagnosis, using similar data and validation techniques.
The predictive model achieved an area under the curve (AUC) of 0.824, indicating excellent performance in distinguishing future dementia cases from controls. This level of accuracy is significantly higher than traditional diagnostic methods, which often struggle to detect dementia at an early stage.
The model identified 15 key connectivity parameters within the DMN that differed significantly between future dementia cases and controls. Among these, the most notable changes included increased inhibition from the ventromedial prefrontal cortex (vmPFC) to the left parahippocampal formation (lPHF) and from the left intraparietal cortex (lIPC) to the lPHF, as well as attenuated inhibition from the right parahippocampal formation (rPHF) to the dorsomedial prefrontal cortex (dmPFC).
In addition to its diagnostic capabilities, the study also developed a prognostic model to predict the time to diagnosis of dementia. This model showed a strong correlation (Spearman’s ρ = 0.53) between predicted and actual times to diagnosis, indicating its potential to provide valuable timelines on disease progression. The predictive power of these connectivity patterns suggests that changes in the DMN may serve as early biomarkers for dementia, providing a window into the disease process years before clinical symptoms appear.
Additionally, the study explored the relationship between DMN connectivity changes and various dementia risk factors. They found a significant association between social isolation and DMN dysconnectivity, suggesting that social isolation may exacerbate neural changes associated with dementia. This finding highlights the importance of considering environmental and lifestyle factors in dementia risk and opens up potential avenues for intervention.
“By using these analysis techniques with large datasets, we can identify people at high risk for dementia and also discover what environmental risk factors pushed these people into a high-risk zone,” said co-author Samuel Ereira. “There is huge potential to apply these methods to different brain networks and populations, to help us better understand the interactions between environment, neurobiology and disease, both in dementia and potentially in other neurodegenerative diseases. fMRI is a non-invasive medical imaging tool, and it takes about 6 minutes to collect the necessary data on an MRI scanner, so it could be integrated into existing diagnostic pathways, especially where MRI is already in use.”
Despite the promising results, some caveats should be noted. One limitation of the study is the use of data from the UK Biobank, which may not be fully representative of the general population. Participants in this cohort tend to be healthier and less socioeconomically deprived. Future research should validate these findings in more diverse and representative samples.
“One in three people with dementia never receive a formal diagnosis, so there is an urgent need to improve the way people with the condition are diagnosed. This will be even more important as dementia becomes a treatable condition,” Julia Dudley, head of strategic research programmes at Alzheimer’s Research UK, told the Science Media Centre.
“This study provides exciting insights into the early signs that a person may be at higher risk of developing dementia. While this technique needs to be validated in further studies, if validated it could be a promising addition to the toolkit of methods for detecting the diseases that cause dementia as early as possible. Early and accurate diagnosis is essential to unlocking personalised care and support and, soon, accessing the first-of-their-kind treatments that are on the horizon.”
Eugene Duff, a researcher at the Institute of Dementia Research at Imperial College London, added: “This work shows how detailed analysis of brain activity measured by MRI can predict a future diagnosis of dementia. Early diagnosis of dementia is useful for many reasons, particularly as better pharmaceutical treatments become available.”
“Measures of brain activity may complement cognitive, blood and other markers in identifying people at risk for dementia. The brain modeling approach they use has the advantage of potentially clarifying which brain processes are affected in the early stages of the disease. However, the cohort of patients diagnosed was relatively small (103 cases). Further validation and direct comparisons of predictive markers are needed.”
The study, “Early Detection of Dementia with Efficient Default Mode Network Connectivity,” was authored by Sam Ereira, Sheena Waters, Adeel Razi, and Charles R. Marshall.