Machine Learning–Based Prediction of Structural MRI Measures from Military Environmental Exposures in Veterans
Jasmin M. Francisco
Mentor: Dr. Peter Bayley, PhD, War Related Illness and Injury Study Center (WRIISC), U.S. Department of Veterans Affairs; Department of Psychiatry and Behavioral Sciences, Stanford University School of Medicine; Dr. Ansgar Furst, PhD, War Related Illness and Injury Study Center (WRIISC), U.S. Department of Veterans Affairs; Department of Psychiatry and Behavioral Sciences and Department of Neurology and Neurological Sciences, Stanford University School of Medicine.
Date/Time: August 21st, 2025 at 1:45 PM.
Abstract: Combat Veterans often experience numerous environmental exposures that lead to a wide range of symptoms and medical conditions post-deployment. To address these unique health concerns, the U.S. Department of Veterans Affairs (VA) established the War Related Illness and Injury Study Center (WRIISC) in 2001 as a national program for Veterans facing issues related to military environmental exposures (MEE). Since then, WRIISC research has included a strong focus on MEE; however, their impact on brain structure remains unclear. Magnetic resonance imaging (MRI), when combined with detailed exposure histories, offers a promising approach for identifying objective neural markers of exposure-related changes. To address this uncertainty, we applied a machine learning–based predictive modeling approach to evaluate whether these exposures can reliably predict structural brain variation, providing a data-driven pathway to uncover subtle neural effects that may not be detectable through traditional analyses.
Within this framework, we analyzed a cohort of 286 Veterans evaluated at WRIISC. The dataset included MRI data processed with FreeSurfer by the WRIISC Neuroimaging team and the self-reported WRIISC Military Environmental Exposures Questionnaire. The analysis tested whether MEE could predict structural brain variation. For each MRI-derived metric, the single most correlated exposure was selected as the predictor, and linear regression models with 10-fold cross-validation were applied. Across more than 7,000 MRI–exposure prediction tests, linear models showed weak explained variance (mean R² ≈ 0.10). The highest-performing model predicted cortical thickness in the right medial orbitofrontal cortex based on chemical nerve agent exposure, but the predictive power was weak (mean R² = 0.072), and dropped further after adjusting for age, sex, and traumatic brain injury (mean R² = 0.017). When both exposures and covariates were included, age consistently emerged as the strongest predictor of cortical thickness, explaining up to ~18–20% of the variance (mean R² = 0.196). Although prediction errors were low across all models in this analysis (on average ≈ 5.5% RMSE and 4.2% MAE relative to mean MRI values), the weak R² values showed that exposures alone and exposures adjusted for covariates explained little variance. When exposures and covariates were included together, age consistently emerged as the prominent predictor, confirming that environmental exposures did not meaningfully predict MRI outcomes.
These findings highlight the difficulty of predicting brain structure from individual exposures, suggesting that the contribution of single exposures to brain variation is small compared with the broader set of factors influencing Veterans’ health. A key limitation is the absence of diagnostic information on Gulf War Illness (GWI), a chronic multisymptom condition frequently observed in Veterans evaluated at WRIISC. Prior studies demonstrate that GWI influences brain structure, so its absence from the dataset limits the interpretability of our findings. In addition, MEEs were assessed through self-report several years after deployment, raising the possibility of recall bias. Future research should integrate broader covariates known to influence brain structure, such as post-traumatic stress disorder (PTSD) and educational level, incorporate longitudinal MRI data, and clarify how MEEs may contribute to post-deployment health conditions, including GWI.
The views expressed are those of the presenter and do not necessarily represent the views or policy of the VA or the U.S. Government. The presenter has no financial conflicts of interest to disclose