Volumetric analysis of traumatic brain injury

Proposal details

Title: Volumetric analysis of cortical and subcortical structures
Research Area(s): Brain Imaging
Background: Traumatic brain injury (TBI) is a major cause of morbidity and mortality in the U.S. resulting in an estimated 3.2 million individuals living with long-term disability (Zaloshnja 2008). Injury results in a variety of neurocognitive deficits and methods of predicting outcome are limited. TBI has many forms including acute vs. chronic types as well as focal vs. diffuse axonal injury. TBI occurs as a result of acceleration-deceleration as well as rotational forces, which can sheer axons and blood vessels (Smith 2003). Cerebral atrophy is a known effect of TBI due to a variety of suggested but poorly understood mechanisms. Recent studies have evaluated changes in cortical and subcortical structural volumes in TBI patients and their impact on long-term outcome (Ariza 2006, Bigler 2001, Bigler 2006, Ding 2008, Gale 2005, MacKenzie 2002, Sidaros 2009, Tomaiulo 2004, Tomaiulo 2005, Warner 2010, Xu 2010, De la Plata 2007). These studies utilized a variety of methods and magnetic resonance image (MRI) sequences to calculate cortical volumes. A study of declarative memory deficits failed to correlate findings in TBI patients with hippocampal volume and instead showed left parietal region thinning as the best predictor of memory impairment (Palacios 2012). In another longitudinal follow-up, subtle changes in the corpus callosum, sagittal stratum, superior longitudinal fasciculus and optic radiations correlated with distinct neuropsychological measures (Farbota 2012a). The use of quantitative cerebral blood flow has been used to predict outcome within the first 12 hours after severe TBI (Kaloostian 2012). This study used xenon CT for quantitative CBF analysis and showed that a pattern of hyperemic flow volumes could correlate to poor outcomes independently of Glasgow Outcome Score at 6-months. A longitudinal study of subjects using tensor-based morphometry showed volume loss in cortical and white matter regions, which correlated well with neuropsychological evaluation (Farbota 2012b). The heterogeneity of patients, MRI sequences, sample size, follow-up and neuropsychiatric testing make correlation of brain imaging to long-term outcome an incomplete area. The use of automated techniques for evaluating cortical and subcortical structure volumes has greatly advanced the ability to analyze a large number of patient images as well as detect subtle changes. One recent study evaluated changes in cortical and subcortical brain volumes by MRI in acute and chronic TBI subjects using 3D Slicer (http://www.slicer.org), a freely available imaging tool (Irimia 2011). However, one limitation from this study included the use of semi-automated methods necessitating user inputs and thereby introducing bias. Evaluation of working memory in a subset of pediatric children with moderate-to-severe TBI showed left parietal lobe thinning and diminished white matter integrity of the frontal lobe and cingulum, as assessed by diffusion tensor imaging, as correlating with decreased performance (Wilde 2011). Furthermore, this study used the freely available software Freesurfer (http://surfer.nmr.mgh.harvard.edu/) for quantitating areas of interest. Previous studies have compared automated from user-defined quantitation of cortical and subcortical structure volume as well as cortical thinning in patients with TBI (Bigler 2010, Merkley 2008, Desikan 2006). These studies have shown strong concordance of automated vs. user-defined methods thus supporting the use of such automated techniques. The significant advantage of automated methods can involve the assessment of large datasets of images with greater speed and accuracy as well as an unbiased approached towards clinical questions.
Aims: The purpose of this study will be to utilize the BRAINnet data in order to analyze cortical and subcortical brain volumes with data contained in the BRAINnet database as well as neuropsychiatric scores as well as disability and quality of life outcomes. Automated analysis of brain volumes will be performed. Various secondary analyses can include assessing brain volumes by subgroups, including younger vs. older patients, males vs. females, therapeutic strategies, etc. This data will serve to form a normative database for which future studies utilizing imaging from TBI patients can be performed.
Method: The BRAINnet data set contains a multi-center aggregated set of patient demographic and survey variables. Analysis will involve the use of Freesurfer (http://surfer.nmr.mgh.harvard.edu/), a freely available and validated software for automated analysis of brain volumes. Correlation of volumes to clinical outcomes, prognosis, and follow-up surveys will be performed. Eventual number of images analyzed can be in the hundreds due to the automated analysis proposed. Method for statistical analysis of such variables to Freesurfer analyzed brain images has been previously described (Li 2012) as well as a general framework for such analysis of volumes to outcome (De la Plata 2007, Warner 2010). Strengths of the BRAINnet dataset will include a multi-center approach, long-term follow-up, and organization of a large variety of patient data. Nonetheless several limitations of this study may be possible. One such limitation will include the heterogeneity in the types of MRI sequences from institutions. The Freesurfer software suggests the use of thin-cut T1 MRI sequences, which may or may not be available on all subjects or in all institutions. This may introduce a variable in the automated analysis that may be accounted for in a multivariate analysis, but will possibly alter volume quantitation. Another limitation can include variability in follow-up survey and MRI data may limit the ability to track volume changes in these subjects after their hospital courses. This will involve the use of multivariate linear regression and Cox Proportional Hazards modeling. ANOVA and Chi-squared analysis will be utilized to assess subjects categorized by Glasgow Coma and Glasgow Outcome Scores. Comparison of intrasubject and intersubject structural volumes will be made. In addition, correlation of structural volumes with neuropsychological scales will be made. Additional analyses will be planned during the data analysis.