To perform a longitudinal analysis of the effect of traumatic brain injury in high-contact sports on microstructural measures of white matter integrity along fiber bundles.

Players and Image acquisition

A total of 210 brains scans were acquired from 46 high-contact (football) players and 24 low-contact (volleyball) players over the course of 3-4 years, in accordance with IRB and HIPAA. Players were scanned at the start and end of each season and within 24-96 hours of a concussion on a 3T scanner (GE, Milwaukee, WI, USA) using an 8-channel head coil. We acquired whole-brain structural T1-weighted images (FSPGR BRAVO, 0.93x0.93x1mm, 5 minutes), multi-echo T2* gradient echo (SWAN, TR 40-45, 5-8 TEs from 5-35, 1x1x1mm, ARC 2x2) and DTI: b-values= 2500|800 s/mm2, number of directions =60|30, b0s= 6|3, TE=81.6 ms, TR=8500 ms, matrix=128x128, FOV 24cm, slice thickness=2mm. In addition to imaging, high-impact players were evaluated using the standardized sport concussion assessment tool II (SCAT II) at baseline and after concussion.

DTI processing

Using FSL 5.0 we corrected for eddy currents using Eddy and distortion using Fugue, with the field map from the complex multi-echo gradient echo acquisition. In addition to fractional anisotropy (FA) and mean diffusivity (MD), we employed the neurite orientation dispersion and density imaging (NODDI) model [1] which produces maps of orientation dispersion index (ODI), intracellular volume fraction (FICVF) and isotropic CSF fraction (FISO) (Figure 1).

Automated Fiber Quantification

We used the software AFQ [2], which automatically produces 20 fiber bundles and quantifies diffusion metrics along the tracts. Briefly, AFQ first performs whole brain tractography, then segments tracts using waypoint regions of interests. It then refines and cleans the bundles based on a probabilistic fiber tract atlas and finally computes diffusion parameters at 100 nodes (locations) along the tracts.

Statistical analysis

We performed three distinct experiments:

1) Baseline comparison of DTI changes between the groups using the Mann Whitney test along 100 nodes of each of 20 fiber bundles for FA, MD, FICVF, FISO, and ODI.

2) Longitudinal analysis of DTI changes using multiple linear regressions with sport, time, their interaction and age as predictors of diffusion indices. A tract was considered significant if the p-value of the regression model was less than the family wise p-value, and the p-value of the interaction (sport * time) was less than alpha (0.05).

3) We correlated the diffusion metrics in the tracts (averaged across football players) to baseline SCAT2 scores.

Multiple comparison correction was performed within track as well as across all 20 tracks for each contrast (FA, MD, FICVF, FISO, ODI) using a non-parametric permutation method [3]. Tracts with less than 5 contiguous significant nodes were not considered significant. Statistical analyses were performed in MATLAB and Stata.

1) At baseline, increased MD in high-contact sports was present along the superior longitudinal fasciculus (SLF) (Table1, Figure 2). We also observed changes in both corticospinal tracts (FICVF and FISO) and the parahippocampal cingulum hippocampus in the left temporal lobe (ODI).

2) For the longitudinal analysis, FA increased, MD decreased, and FICVF increased along the callosum forceps minor (CFM) in the high-contact sport compared to the low-contact sport (Figure 3). We also observed an increase in FISO in the right corticospinal tract (CST), corresponding to our baseline analysis.

3) FICVF correlated with SCAT II scores at baseline in the Left inferior fronto-occipital fasciculus (IFOF) (r=0.372, p=0.032) and in the left Uncinate (r=0.361, p=0.0388) (Figure 4).

This large longitudinal analysis employed not only standard diffusion indices (FA & MD) [4,5], but also more sophisticated diffusion metrics using the NODDI model, to demonstrated longitudinal changes along fiber bundles, specifically the CFM, CST and SLF. Increases in FA and FICV along the CFM may be due to axonal swelling, altered myelination, be secondary to changes undetected in the periphery of the brain, or may be secondary to differences unrelated to brain injury (sports-specific training, cognitive differences) [6]. An analysis of immediate post-concussive data in comparison to the current findings should elucidate to what extends the current findings are due to injury. We have also shown that diffusion metrics (intracellular volume fraction) correlate with cognition as assessed by the SCAT scores, likely supporting cognitive/language functional components of the testing. Future work will include a pre and post concussion analysis and will look at SCAT scores longitudinally and their correlation with brain changes.

This study presents novel evidence suggesting that more sophisticated diffusion models may be useful clinical tools to examine white matter integrity and cognitive deficit in traumatic brain injury.