A groundbreaking new artificial intelligence model has been developed by researchers at the USC Leonard Davis School of Gerontology and King’s College London. This innovative tool is capable of measuring how fast a patient’s brain ages non-invasively through magnetic resonance imaging (MRI) scans, marking a significant advancement in understanding, preventing, and treating cognitive decline and dementia.
Andrei Irimia, an associate professor at USC Leonard Davis School of Gerontology who also serves as a visiting associate professor of psychological medicine at King’s College London, emphasizes the importance of this novel measurement. “This could change how we track brain health both in research labs and clinics,” he said. Knowing one’s brain aging rate can be incredibly powerful for early intervention.
Irimia is the senior author of a study that outlines the new model’s predictive capabilities; it was published on February 24, 2025, in Proceedings of the National Academy of Sciences.
Understanding Biological Age Versus Chronological Age
Biological age differs from chronological age based on how well an individual’s body functions and cellular aging. Two people sharing the same birthdate can have markedly different biological ages due to their lifestyle choices or health conditions. Common methods for measuring biological age, such as blood tests assessing epigenetic aging and DNA methylation, are not suitable for evaluating brain health because of barriers preventing blood cells from crossing into the brain.
The earlier research by Irimia highlighted MRI scans’ potential to non-invasively assess the brain’s biological age. However, previous models had limitations in accurately tracking changes over time since they analyzed single MRI scans rather than longitudinal data.
A More Precise Method of Measuring Brain Aging
To address these challenges, researchers developed a three-dimensional convolutional neural network (3D-CNN). This model was trained and validated on more than 3,000 MRI scans from cognitively normal adults. Unlike cross-sectional approaches that estimate brain age at one point in time, the longitudinal method compares baseline and follow-up scans for each individual to accurately pinpoint neuroanatomic changes tied to accelerated or decelerated aging.
The 3D-CNN also generates interpretable “saliency maps,” highlighting specific brain regions crucial for determining the pace of aging. When applied to a group of 104 cognitively healthy adults and 140 Alzheimer’s disease patients, the new model’s calculations closely correlated with changes in cognitive function tests conducted at both time points.
“The alignment between these measures and cognitive test results suggests that this framework may serve as an early biomarker for neurocognitive decline,” said Paul Bogdan, associate professor of electrical and computer engineering. “It also demonstrates its applicability to individuals regardless of their cognitive status.” The model could potentially characterize both healthy aging patterns and disease trajectories more effectively in the future.
Looking Ahead
The study noted that the new model can distinguish different rates of brain aging across various regions, which may provide insights into how different pathologies develop based on genetics, environment, and lifestyle factors. Additionally, it demonstrated differing paces of brain aging between men and women, offering potential explanations for gender differences in neurodegenerative disorders like Alzheimer’s.
Irimia expressed excitement about the model’s potential to identify people with faster-than-normal brain aging before they exhibit any symptoms of cognitive impairment. This early detection could significantly improve treatment outcomes since new drugs targeting Alzheimer’s have proven less effective when started after significant disease progression has occurred.
