2026 American Academy of Neurology Abstract Website

Hailey Brigger¹, Ian Johnson¹, Annabelle Shanks¹, Steph Maynez¹, Alison Champagne¹, Cyprien Rivier¹, Megan Johnson², Matt De Both², Danielle Metz², Darian Chambers², Calla Price², Yareli Barraza², Bethine Moore², Ignazio Piras², Arjun Agrawal², Jua Iglesias Gonzalez³, Annabel Sorby-Adams⁴, W. Kimberly⁴, Adam De Havenon¹, Matthew Huentelman², Kevin Sheth¹
¹Department of Neurology, Center for Brain & Mind Health, Yale New Haven Hospital and Yale School of Medicine, ²Early Detection and Prevention, The Translational Genomics Research Institute (TGen), ³Computer Science & Artificial Intelligence Laboratory, Massachusetts Institute of Technology, ⁴Department of Neurology, Massachusetts General Hospital and Harvard Medical School

Objective:

To develop a portable MRI brain-age model for community brain health screening and examine its associations with vascular risk factors.

Background:

Brain-age modeling estimates biological brain age from MRI-derived volumes, providing a quantitative marker of neurovascular health. Portable MRI (pMRI) enables imaging outside of traditional hospital settings, offering opportunities for large-scale, accessible brain health screening. By integrating pMRI with data-driven modeling, we aimed to establish a scalable approach to monitor brain health across community sites.

Design/Methods:

T2-weighted pMRI scans were analyzed from 1,311 participants (mean age 50.8 ± 18.8 years, 70.6% female) across 14 community sites in Arizona and Southern California. Sixteen regional brain volumes were segmented with WMH-Synthseg, a machine-learning tool optimized for pMRI. These volumes were entered into a LASSO regression model (5-fold cross-validation) to predict chronological age. Performance was evaluated using mean absolute error. Multivariable associations between the corrected brain-age gap and vascular risk factors were evaluated. Additional analyses examined predictors of extreme deviation.

Results:

The model achieved a mean absolute error of 8.7 years (95% CI, 8.3-9.0) and R²=0.68 (95% CI, 0.65-0.70). In multivariable analyses, BMI, high blood pressure, and high cholesterol were associated with greater brain-age deviation, while sex, diabetes, and race were not. Participants with a brain-age gap ≥10 years older were more likely to have high cholesterol (OR=2.2[1.50-3.10], p<0.001), whereas those with a brain-age gap ≥10 years younger were more often female (OR=1.6[1.10-2.30], p=0.01) and did not have high blood pressure (OR=0.64[0.46-0.90], p<0.01). No single demographic strongly predicted deviation magnitude (R²=0.007).

Conclusions:

Portable MRI brain-age modeling offers accessible estimates of biological brain-age using low-field data acquired in community settings. The model identifies expected neuroanatomical aging patterns and vascular correlates of accelerated brain aging. These results support the feasibility of population-level brain-health screening with portable MRI and highlight the possibility to track modifiable risk factors in real-world environments.