An estimated 10% to 20% of Veterans from the wars in Iraq and Afghanistan have suffered traumatic brain injury (TBI). Many Veterans from prior conflicts have experienced TBI as well. Some studies have reported a link between TBI and increased risk of cognitive impairment and dementia even after years of active life post-injury, but few have studied Veterans. Many people with TBI do not develop dementia; however, we have no tools to predict which individuals are at highest risk and would benefit from careful follow-up and recruitment into clinical trials to prevent post-TBI dementia. Practical biomarkers for identifying patients at highest risk for dementia after TBI are desperately needed to inform individual patient management, dementia prevention strategies, and clinical trials. Furthermore, understanding the underlying etiology of post-TBI dementia (clinical dementia subtypes) could further inform clinical care and prevention for Veterans with TBI. Measures of structural brain changes rapidly measured on neuroimaging modalities, particularly structural brain magnetic resonance imaging (MRI), are well-established predictors of cognitive decline and risk for dementia in the general population. Large population-based samples of TBI-exposed Veterans are needed in order to leverage advances in neuroimaging and biomarker discovery via artificial intelligence approaches to develop robust and generalizable models to predict post-TBI dementia and to discover neuroimaging biomarkers to characterize dementia subtypes among TBI-exposed individuals. The Veterans Health Administration (VHA) has recently made the nationwide VHA imaging data available to researchers; we have the unprecedented opportunity to merge this nationwide structural MRI with our existing nationwide VHA cohort of 1.6 million TBI-exposed and unexposed Veterans to create the largest military-relevant TBI MRI dataset that has (to our knowledge) ever been created and directly address the major knowledge gaps described. We have assembled a team of experts in dementia, TBI, neuroimaging, and prognostic modeling with track records of successful completion of high-impact research. The propose a 3-year project will cost-efficiently harness the newly available wealth of nationwide clinical neuroimaging data and merge with our existing cohort of 1.6 million TBI-exposed and unexposed Veterans with up to 12 years of follow-up in order to (1) create a large, nationwide, high-quality cohort of ~200,000 TBI-exposed and unexposed Veterans with MRI imaging data; (2) predict which TBI-exposed Veterans will go on to develop dementia; and (3) identify prevalence of specific subtypes of dementia among TBI-exposed versus unexposed Veterans. We expect that we will (1) produce the largest military-relevant MRI dataset with expertly curated TBI exposure and dementia outcome and up to 12 years of follow-up (with the option of continued follow-up via VHA electronic medical records) and (2) develop a method for predicting 5+-year risk of post-TBI dementia using routinely collected clinical MRI. This work may (1) directly inform clinical care of Veterans and identify a high-risk subset that may be ideal for further studies of underlying mechanisms of post-TBI dementia and clinical trials for prevention and (2) facilitate discovery of the nationwide epidemiology of neuroimaging biomarker-supported dementia subtypes in TBI-exposed versus unexposed Veterans receiving care within VHA. This work may directly inform public health planning within the Department of Defense and VHA and generate testable hypotheses regarding underlying etiology of post-TBI dementia.