Personalized Profiles of Pathology in Pediatric Traumatic Brain Injury

点击翻译按钮获取中文摘要

Project summary/abstract Children and adolescents have the highest rate of traumatic brain injury (TBI) in the general population, but current tools for examining structural and functional deficits from MR data have several key limitations. First, there is no gold standard procedure for considering lesions in imaging analysis, and second, existing tools have been built on adult populations. We propose to develop a workflow that addresses both of these issues and supports the extension of novel tools for advanced, multimodal analysis, and to use that workflow to identify factors associated with outcome. Personalized Profiles of Pathology (P3) will include registration to age-appropriate templates and lesion segmentation, allowing for voxelwise lesion symptom mapping, morphometric analyses, shape analysis, and network diffusion modeling as part of the package, with options for longitudinal analysis as well. This workflow will include and extend novel pipelines. Voxelwise lesion symptom mapping examines the correspondence between lesion location and specific symptoms, but has been underpowered in existing applications. Network diffusion modeling uses diffusion MRI data from healthy individuals to model the spread of pathology. While this is currently used on chronically injured patients to estimate the epicenter of injury, in this proposal, longitudinal data will be used to validate predictions of pathology spread. Tract-wise statistical analysis similarly uses healthy data to predict the degree of disconnection based on lesion location. Symmetric multivariate linear reduction reduces high dimensional imaging, cognitive, and clinical data to components, revealing patterns of disruption. By including all of these individual approaches across multiple sites, P3 will allow for multi-modal examination of the impact of TBI on pediatric patients with greater statistical power. In Aim 1, we will develop and test P3 on cohorts from eight sites. With input from clinical experts in neurology, rehabilitation, neuropsychology, radiology, and brain development, and technical expertise from mathematics, computer science, and neuroimaging analytics, we will ensure that P3 is statistically and computationally valid and clinically relevant. In Aim 2, we will extract common neuropsychological endpoints from disparate scales across cohorts and use these measures with brain metrics generated by P3 to determine factors associated with outcome and identify subgroups within the patient population. In Aim 3, we will distribute P3 to a network of beta-testing sites to run locally, allowing for further improvement and validation, and disseminate P3 to the research community. Through meta-analysis or harmonization paired with mega-analysis, we will combine effects across sites and examine consistency in effect size, location, and direction. Education will occur through tutorials at national and international conferences, site visits, and through written documentation. P3 will be made available online, with continuing support from the development team. The ultimate goal of P3 is to better understand heterogeneity in post-injury outcome to inform future treatment development.