Research Project
10266848
Project Summary/Abstract Traumatic brain injury (TBI) is a major public health issue globally, and while neuroimaging has been useful in understanding disruption in brain structure and function after injury, there are a number of factors that attenuate its prognostic ability. For example, there is tremendous heterogeneity in outcome after injury which is only partially explained by injury severity. Cost frequently limits sample size in neuroimaging studies, yet given the myriad factors that have been shown to influence patient outcome (age, injury severity, socioeconomic status), small samples and mass univariate testing often result in many studies being grossly under-powered. One solution is to combine data and create novel data sharing platforms, and the Enhancing Neuroimaging Genetics through Meta-Analysis (ENIGMA) consortium has supported this kind of collaboration for over a decade across a range of clinical disorders. The goal of this proposal is to develop tools and data processing procedures for use in the ENIGMA Brain Injury working group. In the R61 phase, we aim to develop and test a workflow for harmonized processing of behavioral data (Aim 1) as well as structural and functional (resting-state) MRI data (Aim 2). For Aim 1 of the R61, the goal is to offer a decision tree of procedures that is data-dependent, allowing investigators to establish common cognitive endpoints across cohorts that collect a range of neuropsychological and clinical measures. This proposal will create sharable procedures, flexible tools, and generalizable guidelines for best practices for extracting common cognitive endpoints from distinct behavioral test batteries (R61 Aim 1). In Aim 2 of the R61, we develop an image processing pipeline called Harmonization and Aggregation for Functional and structural imaging data PIPEline; HAF-PIPE) that allows for aggregation of non-equivalent imaging data. A primary goal is to decentralize ComBat, an open-source data harmonization tool, so that it can be used in a virtual sharing environment. Following satisfaction of the R61 Go/No-Go criteria, which is the curation of the dataset including 13 cohorts, extraction of common cognitive endpoints, and creation of HAF- PIPE, we will move to the R33 phase. In the R33 phase, we will leverage the large, harmonized dataset and apply a machine learning technique (CorEx - Correlation Explanation) to identify patient clusters within each patient population studied. HAF-PIPE and the procedures and guidelines from the R61 phase will then be extended to additional patient populations and made available to other ENIGMA working groups. The harmonized data, along with the tools and procedures for creating them, will be accessible to researchers following proposal submission and approval as a curated dataset. With success, this proposal holds the promise of significantly advancing data curation, harmonization, and sharing in the clinical neurosciences. We anticipate that our proposal will significantly advance our understanding of factors that impact outcome after injury and will yield a tool that will be useful across the neuroimaging community.
