(© Mikki Orso - stock.adobe.com)

New AI model detects accelerated brain aging 10 times more accurately than previous methods In a nutshell

Scientists developed a new method to measure how fast brains age using MRI scans, achieving 93% accuracy in predicting brain age changes in healthy adults.

People with faster-than-normal brain aging showed greater cognitive decline over time and were more likely to develop future cognitive impairment, suggesting the method could help identify Alzheimer’s risk before symptoms appear.

The AI model created “saliency maps” revealing which brain regions age differently between men and women, and between people in their 50s versus 70s, offering new insights into how aging affects the brain.

LOS ANGELES — A new study has uncovered a revolutionary way to measure how quickly our brains are aging—not just how old they appear to be. Using advanced deep learning techniques and longitudinal MRI scans, researchers have created a method to calculate the “pace of aging” in the brain, which could dramatically improve our ability to identify individuals at risk for cognitive decline and neurodegenerative diseases like Alzheimer’s.

The research, published in the Proceedings of the National Academy of Sciences (PNAS), introduces a novel three-dimensional convolutional neural network (3D-CNN) that can estimate the pace of brain aging with remarkable accuracy—far surpassing existing methods.

“This is a novel measurement that could change the way we track brain health both in the research lab and in the clinic,” said Andrei Irimia, associate professor at the USC Leonard Davis School of Gerontology, in a statement. “Knowing how fast one’s brain is aging can be powerful.”

Why Chronological Age Doesn’t Tell the Full Story

While chronological age—the number of candles on your birthday cake—is the same for everyone born on the same date, biological age can vary significantly between individuals. Some people’s bodies and brains age faster or slower than the calendar would suggest. Previously, scientists could estimate a person’s “brain age” from a single MRI scan, but this only provided a static snapshot of accumulated aging effects since birth.

Pace of brain aging could provide the most accurate calculation of biological age yet. (© Aryazu – stock.adobe.com)

The new approach goes further by measuring the rate at which aging is occurring—the pace of brain aging—which offers more meaningful insights into an individual’s current brain health trajectory.

“The pace of brain aging conveys the rate of aging-related alteration in neurobiological system integrity,” the researchers explain in their paper. “For example, faster pace reflects faster adverse cognitive changes contributing to morbidity and mortality.”

Blood Tests Fall Short

Some common measures of biological age use blood samples to measure epigenetic aging and DNA methylation. However, this approach is problematic for brain assessment, as Irimia explained: “The barrier between the brain and the bloodstream prevents blood cells from crossing into the brain, such that a blood sample from one’s arm does not directly reflect methylation and other aging-related processes in the brain.”

The team, led by researchers from the University of Southern California, trained their longitudinal model on MRI scans from 2,055 cognitively normal adults, validated it with an additional 1,304 participants, and then applied it to an independent cohort of 104 cognitively normal adults and 140 patients with Alzheimer’s disease.

What makes this approach unique is that it examines the differences between two MRI scans of the same person taken at different time points. By calculating how much the brain has changed between scans and dividing by the time elapsed, the system determines the pace at which that person’s brain is aging.

The results were impressive. The new model achieved a mean absolute error of just 0.16 years (7% mean error) when predicting brain age changes in cognitively normal individuals. This dramatically outperformed the best cross-sectional model, which had an error of 1.85 years (83% error)—more than ten times worse.

How Brain Aging Differs Among People

Beyond just measuring the pace of aging, the team combined their longitudinal model with an interpretable CNN saliency approach to map which specific brain regions were aging more rapidly in different populations. This revealed fascinating differences in aging patterns between sexes, across different decades of life, and between cognitively normal individuals and those with cognitive impairment.

“The 3D-CNN also generates interpretable ‘saliency maps,’ which indicate the specific brain regions that are most important for determining the pace of aging,” said Paul Bogdan, associate professor at the USC Viterbi School of Engineering.

Scientists found notable differences in pace of aging between men and women, as well as older and younger adults.(© Tyron Molteni – stock.adobe.com)

In healthy females, brain aging was most apparent in the right precentral gyrus, postcentral gyri, precunei, superior parietal lobules, and paracentral lobules. Males, on the other hand, showed more aging-related changes in the left transverse frontopolar gyrus, right supramarginal gyrus, subcentral gyri, and parts of the cingulate gyri.

The researchers also found that individuals in their 50s showed different patterns of brain aging compared to those in their 70s. Younger individuals experienced more aging in the left lateral temporal lobe and right medial occipital lobe, while those in their 70s showed accelerated aging in the right central and postcentral gyri and left cingulate gyrus.

Pace Of Aging Sheds Light On Alzheimer’s Disease

Perhaps most importantly, the team discovered that their pace of aging estimates correlated significantly with changes in cognitive test scores across multiple domains. People whose brains were aging faster than expected typically showed greater declines in memory and executive function over time.

“Rates of brain aging are correlated significantly with changes in cognitive function,” Irimia said. “So, if you have a high rate of brain aging, you’re more likely to have a high rate of degradation in cognitive function, including memory, executive speed, executive function, and processing speed. It’s not only an anatomic measure; the changes we see in the anatomy are associated with changes we see in the cognition of these individuals.”

This correlation was especially strong for the ADAS13 (Alzheimer’s Disease Assessment Scale), a test commonly used to track cognitive decline. The researchers found that individuals with faster-than-expected brain aging showed significant increases in ADAS errors between baseline and follow-up scans, indicating more rapid cognitive deterioration.

“The alignment of these measures with cognitive test results indicates that the framework may serve as an early biomarker of neurocognitive decline,” Bogdan said. “Moreover, it demonstrates its applicability in both cognitively normal individuals and those with cognitive impairment.”

Study authors hope that the pace of aging could help people identify the risk for neurodegenerative diseases and disorders, like Alzheimer’s, sooner. (Photo by Unsplash+ in collaboration with Getty Images)

Remarkably, the researchers found that their pace of aging measurements could help predict which cognitively normal individuals would later develop cognitive impairment. Among the participants studied, 31 were diagnosed with cognitive impairment after their follow-up MRI, while 211 remained cognitively normal. The pace of aging was significantly higher in those who later developed impairment, and it correlated with how quickly they progressed to a diagnosis.

If physicians could identify individuals whose brains are aging more rapidly before they show symptoms, preventive treatments or lifestyle interventions might be initiated earlier, potentially slowing or even preventing cognitive decline.

A Model For The Future Of Brain Health?

Irimia said he is excited about the potential for the new model to identify people with faster-than-normal brain aging before they show any symptoms of cognitive impairment. While new drugs targeting Alzheimer’s have been introduced, their efficacy has been less than hoped for, potentially because patients might not be starting the drug until there is already substantial pathology present in the brain.

“One thing that my lab is very interested in is estimating risk for Alzheimer’s; we’d like to one day be able to say, ‘Right now, it looks like this person has a 30% risk for Alzheimer’s.’ We’re not there yet, but we’re working on it,” Irimia said. “I think this kind of measure will be very helpful to produce variables that are prognostic and can help to forecast Alzheimer’s risk. That would be really powerful, especially as we start developing potential drugs for prevention.”

Since the study also demonstrated that the pace of brain aging in certain regions differed between the sexes, it might help identify why men and women face different risks for neurodegenerative disorders, including Alzheimer’s.

The researchers acknowledge that their approach is currently limited by their sample size and suggest that future studies with larger and more diverse samples could improve the model’s performance across different populations. Still, their innovative approach represents an important step toward a future where we might not just live longer, but maintain healthier brains throughout our extended lifespans.

By focusing on how quickly individuals’ brains are changing, rather than just their current state, this research opens new avenues for personalized medicine and preventive interventions. After all, when it comes to brain health, it’s not just about how old your brain is—it’s about how quickly it’s getting there.

Paper Summary Understanding the Methodology

The researchers developed a sophisticated approach using deep learning to measure brain aging rates. They created a longitudinal model (LM) that analyzes the differences between two brain MRI scans taken from the same person at different times. The model uses a three-dimensional convolutional neural network (3D-CNN) that was trained on the difference between MRI volumes acquired from the same person at baseline and follow-up timepoints. Rather than estimating brain age at each timepoint separately (as traditional methods do), this model directly calculates the change in brain age across the follow-up interval. By dividing this change by the chronological time elapsed between scans, the system determines the pace at which a person’s brain is aging. The researchers trained their model on 2,055 cognitively normal adults from the UK Biobank and Alzheimer’s Disease Neuroimaging Initiative (ADNI), validated it on 1,304 cognitively normal adults, and tested it on an independent cohort of 104 cognitively normal adults from the National Alzheimer’s Coordinating Center (NACC) and 140 patients with Alzheimer’s disease from ADNI.

Key Findings

The study produced several significant findings. First, the longitudinal model estimated the pace of brain aging with remarkable accuracy, achieving a mean absolute error of just 0.16 years in cognitively normal individuals—far better than existing cross-sectional models, which had errors of 1.85 years or more. Second, faster-than-expected brain aging correlated significantly with cognitive decline across multiple domains, particularly with changes in the Alzheimer’s Disease Assessment Scale (ADAS) scores. Third, the pace of brain aging was significantly higher in cognitively normal individuals who later developed cognitive impairment, suggesting its potential as a predictive biomarker. The team also mapped anatomical variations in regional brain aging rates, revealing differences according to sex, decade of life, and cognitive status. For example, women showed more aging-related changes in areas like the precentral gyrus and parietal lobules, while men showed more changes in the frontopolar and supramarginal gyri.

Limitations

The researchers acknowledge several limitations to their approach. Because they estimate the pace of aging as the change in brain age divided by the time elapsed between scans, their measurement represents the average pace over the follow-up interval rather than the instantaneous rate of brain aging. The accuracy of their estimates may be limited for very short follow-up intervals (where anatomical changes are subtle) or very long intervals (where the average pace may not reflect recent changes). The model also performed better for cognitively normal individuals than for those with Alzheimer’s disease, likely because it was trained exclusively on cognitively normal adults. Additionally, the model is most effective for estimating the pace of aging in individuals whose MRI follow-up intervals are similar to those in the training set (primarily between 1.5 and 3.0 years).

Discussion and Implications

This research has major implications for understanding and monitoring brain health during aging. The ability to measure the pace of brain aging opens new possibilities for early detection of accelerated aging that might indicate neurodegenerative disease risk. The strong correlation between the pace of aging and cognitive decline suggests that this measurement captures meaningful neurobiological changes that affect cognitive function. The finding that currently cognitively normal individuals with faster brain aging are more likely to develop future cognitive impairment could lead to earlier interventions before symptoms appear. The anatomical specificity provided by the saliency mapping technique could help researchers better understand sex differences in brain aging and develop more targeted interventions.

Funding and Disclosures

The research was supported by several funding sources, including the National Institutes of Health (NIH) under Grants R01 NS 100973, RF1 AG 082201, and R01 AG 079957, the Department of Defense under contract W81XWH-18-1-0413, the National Science Foundation under CAREER Award CPS/CNS-1453860, grants MCB-1936775 and CNS-1932620, the U.S. Army Research Office under grant W911NF-23-1-0111, DARPA under a Young Faculty Award and under Director Award N66001-17-1-4044, an Intel Faculty Award, Northrop Grumman, the Hanson-Thorell Research Scholarship Fund, the Undergraduate Research Associate Program, the Center for Undergraduate Research in Viterbi Engineering (CURVE) at USC, and anonymous donors. The authors declared no competing interests.

Publication Information

This study, titled “Deep learning to quantify the pace of brain aging in relation to neurocognitive changes,” was published in the Proceedings of the National Academy of Sciences (PNAS) on February 24, 2025. The research was led by senior author Andrei Irimia and first author Chenzhong Yin, along with other collaborators including Phoebe E. Imms, Nahian F. Chowdhury, Nikhil N. Chaudhari, Heng Ping, Haoqing Wang, and Paul Bogdan from the University of Southern California and King’s College London, along with the Alzheimer’s Disease Neuroimaging Initiative.