Researchers at Queen Mary University of London have built a test that predicts dementia with 82% accuracy, and it can do so up to nine years before a person receives an official diagnosis.
That is not a typo.
Nine years is an enormous window. It is the difference between watching a disease take hold and actually having the time to slow it down.
The findings, published in the journal Nature Mental Health, are being called a first of its kind.
The test uses a standard brain scan and a machine learning model to detect changes in a part of the brain that most doctors would not even be looking at yet.
The core finding is straightforward: your brain reveals warning signs of dementia long before memory loss or confusion ever shows up.
And now, for the first time, we have a reliable way to read those signs.
How the Study Was Conducted
The research team, led by Professor Charles Marshall of Queen Mary’s Centre for Preventive Neurology, analyzed over 1,100 brain scans sourced from UK Biobank, a massive biomedical database containing health information from half a million people across the United Kingdom.
The scans used were functional MRI scans, commonly called fMRI scans.
Functional MRI is different from a regular brain scan.
A standard MRI shows you the structure of the brain, what it looks like physically.
An fMRI shows you the brain in action, capturing how different regions communicate with each other in real time.
What makes this study particularly clever is what the researchers were looking at during those scans.
They were not asking participants to solve puzzles or take memory tests.
They simply asked people to lie still and do nothing.
When the brain is at rest, a specific network of interconnected regions becomes active.
This network is called the default mode network, or DMN.
It is the part of your brain responsible for daydreaming, self-reflection, and mental wandering.
Think of it as the brain’s background program, always quietly running.
The researchers trained a machine learning model to estimate the strength and pattern of connections between ten different regions within the default mode network.
Each participant was then assigned a probability score, indicating how closely their brain’s resting connectivity matched the pattern of someone who would go on to develop dementia.
Those scores were then checked against the long-term medical records held by UK Biobank.
The results were striking.
Findings From the Study
The model correctly predicted who would go on to develop dementia with more than 80% accuracy, in some cases flagging people who would not receive a clinical diagnosis for another nine years.
That alone would be remarkable.
But the researchers went further.
They compared their method against the two most widely used tools in early dementia detection: memory tests and measurements of brain shrinkage.
The fMRI-based model outperformed both.
Memory tests, which have long been a cornerstone of cognitive screening, can miss early-stage dementia entirely because memory problems only become visible after significant brain damage has already occurred.
Brain shrinkage, another common marker, suffers from a similar problem.
By the time the brain visibly shrinks on a scan, the disease has been quietly progressing for years.
The default mode network, it turns out, is the first neural network affected by Alzheimer’s disease.
It shows subtle but measurable disruptions years before the brain physically deteriorates or memory begins to fade.
The fMRI scan required to collect this data takes approximately six minutes.
Lead author Samuel Ereira, an Academic Foundation Programme Doctor at the Centre for Preventive Neurology, noted that fMRI is already used in many clinical settings and could be integrated into existing diagnostic pathways without major disruption.
What Most People Get Wrong About Dementia Diagnosis
Here is where things get uncomfortable.
Most people, including many in the medical community, still think of dementia as something you diagnose when the symptoms become obvious.
The patient forgets names.
They get lost on familiar roads.
They repeat the same question three times in a single conversation.
That is when the doctor orders the tests.
That is when the conversation begins.
But by that point, the disease has often been silently destroying brain cells for a decade or longer.
The reason so many drug trials for Alzheimer’s have failed is not necessarily because the drugs did not work.
It is because they were tested on people who were already too far along for any intervention to make a meaningful difference.
A 2025 analysis published in Alzheimer’s and Dementia confirms that early identification of the disease is now considered essential for treatment to succeed.
Two new drugs, lecanemab and donanemab, have recently received regulatory approval in the United States and are showing genuine promise at slowing the progression of early-stage Alzheimer’s.
The keyword is early.
These drugs work best, and arguably only, when administered before the damage becomes too severe.
Which is exactly why a test that can spot the disease nine years in advance is not just a scientific curiosity.
It is a clinical breakthrough that changes the entire logic of treatment.
How This Discovery Applies to Real Life
Consider what nine years actually means in practice.
A 58-year-old flagged by this test would have until age 67 before symptoms would likely appear under the old diagnostic timeline.
That is nine years to make meaningful lifestyle changes.
Nine years to participate in clinical trials while the disease is still in its earliest stages.
Nine years to take advantage of emerging treatments that have a chance of actually working.
Nine years to plan, legally, financially, emotionally, for what lies ahead.
According to the World Health Organization, dementia currently affects around 57 million people worldwide, with nearly 10 million new cases diagnosed every year.
By 2050, that number is projected to reach 139 million.
The economic cost already surpasses 1.3 trillion dollars annually and is expected to grow dramatically.
Most of that burden falls on unpaid family caregivers.
The emotional weight is harder to quantify, but it is just as real.
A test that can identify people at high risk years before symptoms appear has the potential to shift that entire equation.
The Default Mode Network: Why This Part of the Brain Matters
The default mode network is not the flashiest part of neuroscience.
It does not get the same attention as memory regions like the hippocampus, or the emotional centers tied to the amygdala.
But researchers have known for years that the DMN is among the first casualties of Alzheimer’s disease.
The network connects regions across the brain that work together when you are not actively focused on a task.
When you zone out on a long drive.
When you replay a conversation from yesterday.
When you imagine how a future event might unfold.
All of that quiet inner life runs through the default mode network.
And when it starts to falter, the disruptions are subtle.
Too subtle to notice consciously.
Too subtle to show up on a memory test.
But apparently not too subtle for a well-trained machine learning model looking at a six-minute brain scan.
What Comes Next
The research team is clear that this test is not ready to walk into your local clinic tomorrow.
The next step is validation in larger and more diverse populations.
The UK Biobank, while extraordinarily valuable, skews toward healthier, more educated participants, a well-known limitation in large-scale biomedical research.
The model also needs to be tested across different ethnic groups, age ranges, and people with other health conditions that may affect brain connectivity.
There are also meaningful questions about what happens when someone receives a positive result.
What do you tell a 55-year-old that their brain patterns suggest they may develop dementia within the next decade?
How do healthcare systems support that person?
What interventions are available, and how do you ensure people can access them equitably?
These are not reasons to slow the science.
They are reasons to invest urgently in the infrastructure, the ethics, and the policy that would allow this test to do the most good.
The Lancet Commission on Dementia Prevention has identified 14 modifiable risk factors, including high blood pressure, physical inactivity, and social isolation, that collectively account for nearly 45% of dementia cases worldwide.
A screening tool that provides a nine-year warning window gives people real time to act on those risk factors.
That is not nothing.
That is, potentially, everything.
A Final Thought
We have spent decades treating dementia as a disease you manage after it arrives.
This research suggests we may finally be entering an era where we can prepare before it does.
The brain, it turns out, has been leaving clues all along.
We just needed the tools, and the patience, to learn how to read them.
If this kind of research matters to you, share it with someone who needs to hear it.
And if you are curious about what the early warning signs of cognitive decline actually look like, it is worth starting that conversation with your doctor before you feel like you need to.
References and Further Reading
- First-of-its-kind test can predict dementia up to nine years before diagnosis — Queen Mary University of London
- Early detection of dementia with default-mode network effective connectivity — Nature Mental Health
- 2025 Alzheimer’s Disease Facts and Figures — Alzheimer’s Association
- Dementia Fact Sheet — World Health Organization
