MRI-Based Brain Age and Dementia Risk — A 2026 Clinical Study
Summary ID02-HA-RS-00084
Research Summary
Original Publication
Comparison of Brain Age With Standard MRI Assessment for Dementia Risk Stratification in Subjective and Mild Cognitive Impairment
Journal
Neurology, .
- Category
- 02Healthy Aging
- Evidence
- Retrospective longitudinal cohort study
- Summary Type
- Latest
What Was Studied?
Can an MRI scan reveal information about brain aging that is not apparent during a standard radiological assessment?
Researchers at Amsterdam University Medical Center investigated whether artificial intelligence could extract additional information from brain MRI scans to help predict future dementia.
The study included 799 patients who had attended a specialist memory clinic:
- 412 had subjective cognitive decline. They reported memory or other cognitive difficulties, but standard testing did not confirm measurable impairment.
- 387 had mild cognitive impairment. Their cognitive difficulties were objectively measurable, but they did not have dementia.
Participants were followed for up to 10 years, with a median follow-up of 3.2 years.
The researchers used two independently developed computer models to estimate each participant’s brain age from MRI images.
They then calculated the brain-predicted age difference, or brain-PAD: the difference between the computer-estimated brain age and the person’s chronological age.
For example, a 60-year-old person with an estimated brain age of 65 would have a brain-PAD of +5 years.
The researchers examined whether this difference predicted future dementia and whether it added useful information beyond conventional MRI assessments, cognitive testing and, where available, amyloid biomarkers.
Key Findings
1. Higher Brain Age Was Associated With Greater Dementia Risk
During follow-up, 235 participants (29.4%) developed dementia.
After adjusting for chronological age, sex and baseline cognitive performance, each additional year of brain-PAD was associated with a 6% higher relative hazard of dementia diagnosis.
Hazard ratio: 1.06 (95% CI 1.04–1.08).
After further adjustment for standard MRI assessments, the association remained statistically significant but was smaller.
Hazard ratio: 1.04 (95% CI 1.02–1.06).
These findings indicate an association between older-appearing brain structure and subsequent dementia diagnosis. They do not establish that accelerated brain aging causes dementia.
2. Brain Age Was Most Informative in Patients Without Measurable Cognitive Impairment
The strongest findings occurred among patients with subjective cognitive decline.
In this group, each additional year of brain-PAD was associated with a 9% higher relative hazard of dementia after adjustment for clinical information and conventional MRI findings.
Hazard ratio: 1.09 (95% CI 1.03–1.15).
This suggests that automated MRI analysis may provide additional prognostic information in patients who report cognitive difficulties but do not yet show measurable impairment on standard cognitive testing.
However, only 39 participants in this group developed dementia. This relatively small number limits the precision of the subgroup findings and requires cautious interpretation.
3. Additional Predictive Value Was Modest
In patients with subjective cognitive decline, adding brain-PAD to a model containing clinical information and conventional MRI findings increased its C-index from 0.77 to 0.79.
The C-index measures how well a prediction model distinguishes between individuals with different outcomes.
Although the additional brain age information was statistically significant, the increase in discrimination was small.
When amyloid status was included, the C-index remained unchanged at 0.84.
This distinction is important: a biomarker may be independently associated with disease risk without substantially improving individual clinical predictions.
4. Results Were Less Consistent in Mild Cognitive Impairment
Among patients who already had measurable cognitive impairment, the additional value of brain age was less clear.
Using the BrainageR algorithm, brain-PAD did not significantly improve prediction beyond conventional MRI findings.
Hazard ratio: 1.01 (95% CI 0.99–1.04).
A second algorithm, cNeuro, showed a statistically significant association after similar adjustments.
This difference demonstrates that brain age estimates and their predictive performance can depend on the computational method used.
5. A Younger-Appearing Brain Was Associated With Lower Observed Risk
Researchers identified a brain-PAD threshold of −2.6 years within the subjective cognitive decline group.
At this threshold, the negative predictive value ranged from 0.99 at two years to 0.91 at ten years.
These findings suggest that lower brain-PAD values may help identify patients less likely to develop dementia during follow-up.
However, this threshold was developed and evaluated within the same cohort, without independent external validation.
Its apparently high negative predictive value may partly reflect the relatively low frequency of dementia in this subgroup.
The positive predictive value was also relatively low, ranging from 0.05 to 0.31.
Therefore, these results should not be interpreted as evidence that brain age can reliably exclude future dementia in individual patients.
6. MRI Equipment and Algorithms Affect Brain Age Estimates
The researchers used MRI scans acquired on eight different scanners.
They found scanner-related differences in estimated brain age, although the overall association with dementia remained relatively stable.
The mean absolute error in estimating chronological age was approximately 6.6 years for BrainageR and 5.2 years for cNeuro in this patient cohort.
These values describe the difference between algorithm-estimated age and actual chronological age. They do not establish the accuracy of measuring an individual’s biological brain age.
Consequently, a brain age estimate should not be interpreted as a precise measurement of biological brain aging.
Standardisation and calibration are essential before universal clinical thresholds can be considered.
Why It Matters for Longevity
Brain health is a central component of healthy aging.
Maintaining memory, cognitive function and independence is often more meaningful than extending lifespan alone.
MRI-based brain age estimation is attracting interest because it attempts to translate complex structural brain information into a single, understandable measure.
However, biological brain age is not equivalent to chronological age, and neither is a direct measure of an individual’s future cognitive health.
Importantly, a single brain age measurement does not establish how quickly a person’s brain is aging. It also cannot independently diagnose neurodegeneration.
An older-appearing brain does not mean that dementia is inevitable. Equally, a younger-appearing brain does not guarantee protection.
For longevity medicine, the potential value of brain age lies in whether it can eventually contribute reliable information beyond established clinical assessment.
This study represents progress towards that goal, particularly in the early evaluation of cognitive complaints.
Further prospective studies are needed to establish whether brain age testing meaningfully improves clinical decisions and outcomes.
Clinical Perspective
This study provides evidence that AI-derived brain age contains information associated with future dementia risk.
Its most interesting finding concerns patients who report cognitive difficulties despite normal results on conventional cognitive testing.
In this group, brain age estimation may capture MRI-derived information that is not fully reflected in conventional visual ratings.
Nevertheless, several limitations restrict immediate clinical application.
First, this was a retrospective observational study conducted in a specialist memory clinic. Its participants were not representative of healthy adults undergoing preventive screening.
Second, the additional predictive value was modest. A statistically significant association did not consistently translate into better discrimination between patients who did and did not develop dementia.
Third, the two algorithms produced different results in patients with mild cognitive impairment.
Fourth, most dementia diagnoses were Alzheimer’s disease. Of the 235 participants who developed dementia, 178 were diagnosed with Alzheimer’s disease. The findings may therefore not apply equally to other forms of dementia.
Fifth, the subgroup with subjective cognitive decline included only 39 patients who developed dementia. This limits confidence in the precision and generalisability of the subgroup estimates.
Finally, the study did not demonstrate that providing brain age estimates to clinicians changes medical decisions, improves patient outcomes or prevents dementia.
The findings therefore support further clinical validation rather than routine use of brain age as a standalone diagnostic or screening test.
Key Takeaway
AI-estimated brain age from MRI was associated with future dementia risk, particularly in patients reporting cognitive difficulties without measurable impairment. Its additional predictive value was modest, and further validation is required before routine clinical use.
Reviewed and Summarized by
Source Publication
- Title:
- Comparison of Brain Age With Standard MRI Assessment for Dementia Risk Stratification in Subjective and Mild Cognitive Impairment
- Authors:
- De Vries S, Farkas K, Rhodius-Meester HFM, et al.
- Journal:
- Neurology, August 2026.
