New AI Model Maps Aging Across Brain Regions

Artificial intelligence can now generate detailed maps highlighting differences in how distinct parts of the brain age, according to a study published in the journal Proceedings of the National Academy of Sciences. Developed by University of Southern California researchers, the deep learning model analyzes magnetic resonance imaging (MRI) scans from nearly 15,000 individuals to map local brain age and its correlation with cognitive function across the lifespan.

Deep Learning Model Maps Local Brain Age Across 15,000 Scans

Traditional neuroimaging biomarkers typically reduce a person’s structural brain changes to a single numerical age estimate. This broad approach often conceals crucial regional differences occurring beneath the skull. To address this limitation, researchers led by Andrei Irimia, an associate professor at the USC Leonard Davis School of Gerontology, trained a deep-learning neural network using MRI scans from 14,748 cognitively normal adults aged 19 to 100.

The training data spanned six large public datasets, including the UK Biobank, the Human Connectome Project, and the Alzheimer’s Disease Neuroimaging Initiative. By establishing this baseline, the AI model measures local brain age at the voxel level—the three-dimensional units comprising an MRI scan—rather than assigning a single blanket age to an individual.

Frontal and Temporal Lobes Show Advanced Aging in Healthy Adults

When analyzing cognitively healthy participants, Irimia’s team found consistent structural patterns across the lifespan. The frontal and temporal lobes—areas responsible for decision-making, memory, and higher cognitive functions—frequently appeared biologically older than the parietal and occipital regions, which handle spatial awareness and sensory processing.

Furthermore, the data revealed asymmetry between hemispheres. The brain’s right hemisphere tended to show slightly more advanced aging than the left hemisphere. According to the study findings, this lateralized pattern persisted regardless of whether the individual participants were right-handed or left-handed.

Accelerated Aging Patterns in Alzheimer’s Disease and Cognitive Impairment

Beyond healthy cohorts, the research team tested the deep learning model using MRI scans from more than 1,900 additional participants enrolled in the Alzheimer’s Disease Neuroimaging Initiative. This test group included cognitively normal adults, individuals with mild cognitive impairment, and patients diagnosed with Alzheimer’s disease.

As cognitive impairment progressed, regional discrepancies intensified. Participants with mild cognitive impairment or Alzheimer’s disease demonstrated significantly older local brain ages in structures historically targeted early by Alzheimer’s pathology. These vulnerable structures include the hippocampus, the amygdala, and several deep brain regions integral to memory processing.

Additionally, older local brain age correlated directly with poorer performance on standard cognitive assessments. The strongest statistical relationships appeared in patients with Alzheimer’s disease, indicating that regional structural metrics gain predictive value as neurodegeneration advances.

Future Validation and Clinical Translation Pathways

While the map generation method marks a notable technical step forward for neuroscience, Irimia emphasized that the approach remains strictly a research tool. Because the model relied primarily on research-quality MRI data, it requires extensive validation using diverse clinical datasets before doctors can integrate it into routine patient care.

Moreover, the underlying study relied heavily on cross-sectional data. Consequently, future longitudinal studies are necessary to confirm whether tracking local brain aging can accurately forecast which healthy individuals will eventually transition to mild cognitive impairment or Alzheimer’s disease.

Frequently Asked Questions

What is local brain age?

Local brain age refers to an AI-driven measurement that estimates the apparent biological age of specific, localized regions of the brain based on MRI scans, rather than assigning a single age to the entire organ.

How was the AI model trained?

Researchers trained the deep-learning neural network using structural MRI scans from 14,748 cognitively normal adults aged 19 to 100 drawn from major public repositories like the UK Biobank and the Human Connectome Project.

Can doctors use this tool to diagnose Alzheimer’s disease today?

Not yet. According to the study authors, the model is currently a research tool that requires additional validation using diverse clinical datasets before deployment in standard medical settings.

---