Abstract
There are no validated, objective diagnostic or prognostic biomarkers for sports-related concussion (SRC), which hinders evidence-based treatment for concussed athletes. While quantitative electrophysiology (EEG) and diffusion tensor imaging (DTI) are promising technologies for providing objective biomarkers for concussion, the degree to which they are related has not been systematically investigated in concussed athletes. This study examined whether diffusion metrics differentiated concussed athletes with prolonged recovery (n = 18) from non-conccused athletes (n = 13) and whether observed diffusion alterations related to EEG. Collegiate athletes (N = 31) completed EEG, neurocognitive, and magnetic resonance imaging. White matter diffusivity differed between the groups in multiple white matter tracts, including the corpus callosum, cingulum bundle, thalamic radiations, and inferior fronto-occipital, inferior longitudinal, and uncinate fasciculi, but not after correction for multiple comparisons. The enhanced Brain Function Index (eBFI), a measure that combines EEG and neurocognitive data, significantly correlated with altered diffusion in the concussed athletes. These preliminary findings suggest that the absolute deviation of diffusion metrics in concussed versus non-concussed athletes may have clinically utility. Results also suggested that the eBFI may be sensitive to early changes from sports-related concussion.
Introduction
Sports-related concussion (SRC), and associated mild traumatic brain injury (TBI) and its cognitive sequalae, represents a significant problem among collegiate athletes, particularly in sports where contact is associated with normal competitive play. 1 –3 Though urgently needed, no objective clinical tests or biomarkers currently exist for early and accurate identification and management of SRC. 4 –7 This significantly complicates clinical decisions regarding the appropriate time window for return to play (RTP), which is important given the significant risk for reinjury in recently concussed athletes, as well as the increased severity of the consequences of multiple SRCs within a short time window. 7 –9 The lack of validated, objective diagnostic and prognostic biomarkers for SRC is an obstacle to providing appropriate evidence-based diagnosis and clinical care for concussed collegiate athletes. However, quantitative electrophysiology (EEG) and advanced magnetic resonance imaging (MRI) are promising technologies that could potentially provide objective, valid biomarkers for SRC.
Identification of a biomarker of SRC and mild TBI more generally has proven to be quite difficult for several reasons. 6,10 Studies suggest that the underlying neuropathology of mild TBI is multi-faceted and involves complex neurochemical, structural, and functional changes in the brain. 4,11 –13 Moreover, the degree and spatial distribution of neural injury can vary widely, with microscopic alterations in widespread, heterogeneous brain regions. 6,11,12 Small signal changes within the brain are difficult to detect given the high degree of interindividual variability known to exist in brain structure and function in healthy young adults. 14 –16 This problem is particularly pronounced in the case of mild TBI, where brain changes are more subtle than those observed in cases of more severe TBI. 4,5
White matter diffusivity as a potential biomarker of sports-related concussion
Advanced diffusion MRI techniques, such as diffusion tensor imaging (DTI), have already provided promising results for non-invasive detection of mild TBI, including SRC. 4,6,10,17 Data from DTI studies suggest the presence of significant microstructural alterations after concussive and subconcussive TBIs in contact sport athletes. 18,19 DTI is also sensitive to alterations of white matter in both the early and chronic stages of SRC, and preliminary evidence indicates that this imaging technique has prognostic utility. 10,20 –22 Despite variability in DTI findings during the first few days post-SRC, several studies report that at least a subset of the diffusion metrics measured at 7 days post-concussion are altered in concussed student athletes. 23
There also is mounting evidence that DTI-based measures capture the increasingly severe effects of multiple concussive events. 10 Some studies support dose-response effects of subconcussive impacts on extent of white matter alterations, whereas other studies support greater alteration with more severe impacts as opposed to the number of impacts. 10,21,23 –25 Specific findings suggest that microstructural changes, as indexed by diffusion metrics like fractional anisotropy (FA) as well as mean diffusivity (MD), axial diffusivity (AD), and radial diffusivity (RD), are more indicative of pathology in contact-sport athletes with histories of SRC than in those with histories of exposure to repetitive subconcussive blows to the head, but without histories of SRC. 10,18,19,26
Electrophysiology as a potential biomarker of sports-related concussion
EEG, a functional brain imaging technique that records electrical signals generated by populations of neurons with high temporal resolution, provides unique insights into the time course of neuronal transmission. Clinically, EEG has the advantages of being ready availability at the point of care, quick to acquire, easy to administer with limited training, and cost-effective. EEG data acquired after mild TBI have revealed alterations in brain electrical activity relative to control participants, suggesting the utility of pursuing measures derived from scalp-recorded EEG as potential biomarkers of brain injury. 27
The EEG-based Brain Function Index (BFI), a derived biomarker recorded from frontal and fronto-temporal brain regions, is one biomarker thought to provide an index of functional brain impairment following a mild head injury. The BFI, which was validated in an independent prospective U.S. Food and Drug Administration (FDA) validation trial of the BrainScope One device (cleared as Ahead 300), 28,29 is derived from the EEG features associated with brain impairment, which best reflect current consensus on the physiology of concussion. These features include measures of connectivity (reflecting disruption in neuronal transmission between brain regions, e.g., phase synchrony and coherence), EEG signal complexity (reflecting disorganization in neural networks, e.g., fractal and scale-free dimensions), and shifts in the frequency spectra (including features reflecting changes in oxygen flow and glucose metabolism, e.g., alpha power). Brooks and colleagues 30 demonstrated the potential clinical utility of the BFI in a longitudinal evaluation of athletes post-SRC. An enhanced Brain Function Index (eBFI) has since been derived, which extends the EEG-only BFI to include multi-modal inputs. The eBFI includes selected neurocognitive and vestibular features and shows improved sensitivity to alterations after SRC and changes throughout recovery. 31
Current study aims
While both DTI and EEG techniques have shown promise generating potential biomarkers for SRC independently, they have not been systematically investigated in SRC. Recent data from military personnel with mild TBI suggest that the use of both techniques may be beneficial. 32,33 The objectives of the current study were to examine whether 1) DTI metrics differed between athletes with subacute concussion who had prolonged recovery versus non-contact athlete controls and 2) changes identified from DTI were related to EEG findings.
Methods
Participants
Thirty-one collegiate athletes between the ages of 17 and 24 years who sustained an SRC and had prolonged recovery (n = 18) and non-concussed athletes from non-contact sports (n = 13) were recruited from the athletic programs of University of South Carolina (SRC = 7, comparison = 7) and the University of Connecticut (SRC = 11, comparison = 6).
Sports-related concussion group
Athletes were included in the SRC group if they were diagnosed as having sustained a concussion by an athletic trainer or team physician on the basis of local institutional guidelines. Athletes in the SRC group also had a prolonged recovery, which was defined as RTP ≥14 days post-injury. All sites in the study followed a standard protocol conforming to NCAA policy guidelines for RTP. The RTP interval was calculated as the time period that included both the time during which the athlete reported symptoms consistent with concussion, and the time during which the athlete reported resolution of symptoms, and completed the gradual RTP protocol. Median (Mdn) RTP duration was 15 days as calculated in this manner. RTP in the SRC group occurred on an average of 22.4 (standard deviation [SD] = 8.67) days post-injury, and the total duration of symptoms for athletes in the concussed group occurred on average 18.35 days (Mdn = 14.5). The selection of this cut point was based on the larger group of participants from the overall study addressing the use of EEG in SRC, 34 where the median RTP was 14 days.
The RTP of the current sample of athletes with SRC ranged from 15 to 51 days (mean [M] = 22.44; SD = 8.67). All participants in the SRC group had a Glasgow Coma Scale score of 15. 35 One participant sustained a brief loss of consciousness, 2 reported post-traumatic amnesia, and 5 reported an altered mental state. The athletes in the SRC group were involved in the contact sports listed in Table 1. Two of the athletes in the SRC group had sustained a previous SRC in the year preceding the current injury (i.e., concussion history for the SRC group, M = 0.278; SD = 0.752; Mdn = 0).
Sample Demographic Characteristics
M, mean; SD, standard deviation; SCAT, Sport Concussion Assessment Tool; SRC, sports-related concussion.
Athlete comparison group
Athletes in the comparison group had no histories of concussion in the past year and were involved in non-contact sport(s) that are not typically associated with a high risk of sustaining a concussion or incurring blows to the head (see Table 1). One of the athletes in the comparison group had a lifetime history of one previous concussion without loss of consciousness. Recruitment efforts were made to ensure that the comparison group was demographically comparable to the participants in the SRC group.
Exclusion criteria for participants in both groups included: 1) scalp or skull abnormalities or whose clinical condition, such as TBI, would not allow placement of the electrodes; 2) taking any central nervous system active prescription medications, with the exception of medications prescribed for attention deficit/hyperactivity disorder; 3) history of neurological disease or neurosurgery (i.e., seizure disorder); or 4) history of a psychiatric disorder (e.g., psychosis, history of a substance or alcohol use disorder). 36
All study procedures were conducted in accordance with the Helsinki Declaration and approved by the institutional review boards of Baylor College of Medicine, University of Connecticut, and University of South Carolina. Informed consent (and assent, when appropriate) was obtained from each participant and from the parent or other legally authorized representative in the case of a minor.
Magnetic resonance imaging
All participants underwent an MRI that included diffusion-weighted imaging within a target of within 2 days post-injury (M = 2.57; SD = 0.95; Mdn = 2.92). To limit the variability that can be introduced in multi-site studies through the use of scanners from different vendors, 37 we limited this report to data from the two recruitment sites, which had comparable scanning platforms.
Diffusion magnetic resonance imaging
Diffusion MRI was collected on a Siemens 3T Prisma scanner at each site (Siemens Healthcare, Malvern, PA) using transverse multi-slice spin echo, single-shot, echo planar imaging sequences (9000 ms repetition time, 94 ms echo time, 59 slices, 0 mm gap, 350 mm field of view, and an isotropic 2.7 mm3 voxel size). Diffusion was measured along 64 directions (low b-value = 0; high b-value = 1300 sec/mm2).
Quality assurance of the magnetic resonance imaging data
All DTI data were evaluated for quality. To ensure consistent scanner performance throughout this longitudinal study, image quality was evaluated regularly according to American College of Radiology accreditation standards and performed using recommended measures of signal-to-noise ratio, field uniformity, gradient linearity, image distortions, and ghosting using the phantom that was provided by the vendor. Scans that had unacceptable image quality were deleted from further analysis and were not included in the current analyses. Given that DTI analyses are designed to detect the motion of water molecules, diffusion-weighted images can be particularly sensitive to head motion, which can result in misalignment between consecutive images as well as intensity changes. 38 Misalignment can be corrected by registering the diffusion-weighted images to each other; however, intensity alterations cannot be corrected. Therefore, the average volume-by-volume translation and rotation were used as covariates in the DTI analyses.
Quantitative tractography
Automated global deterministic tractography was performed on the diffusion MRI data. Diffusion-weighted images were initially preprocessed using the dtiInit pre-processing pipeline wrapper from Stanford open-source VISTASOFT package version 1.0 (
Common data elements for magnetic resonance imaging
All MRI data acquired as part of the research study were also reviewed by a single board-certified neuroradiologist (J.V.H.) for incidental findings and trauma-related abnormalities using the standardized Intra-agency Imaging Common Data Elements (CDEs) for Traumatic Brain Injury as well as additional elements which were included in the recently published CDEs for SRC, 40,41 including the presence of white matter hyperintensities (on fluid-attenuated inversion recovery [FLAIR] sequences) and the presence of cavum septum pallucidi. In coding the MRI findings, high-resolution T1-weighted, gradient recalled echo, FLAIR, and susceptibility-weighted imaging were utilized, as appropriate.
Electrophysiology and neurocognitive data
EEG and neurocognitive assessment data were collected within a target date of 2 days (M = 1.64; SD = 0.76; Mdn = 1.86), using prototypes of the BrainScope Ahead Concussion Assessment System. 29,31 The investigational system included an EEG acquisition module to collect brain electrical activity and a separate cognitive assessment module implemented on a tablet. Five to 10 min of “eyes closed counting” EEG, whereby participants were instructed to relax with eyes closed and perform a mental counting task (subtracting 7 sec), was obtained using hand-held BrainScope investigational devices. 28,29 The EEG was recorded from a limited frontal and frontotemporal electrode montage, including, specifically, the Fp1, Fp2, F7, F8, AFz, A1, and A2 locations of the expanded International 10-20 Electrode Placement System. The EEG data were acquired at a sampling rate of 1 kHz. All electrode impedances were <10 kW. Amplifiers had a band-pass filter from 0.3 to 250 Hz (3 dB points).
All participants in the SRC group were also assessed within 72 h post-injury, 5 days post-injury, on the RTP date, and 45 days after the RTP date (RTP +45). Comparison athletes were also assessed at the time intervals corresponding to a matched participant in the SRC group.
Quantitative electrophysiology feature extraction
Recording sites were remontaged to linked ears and down-sampled from 1 kHz to 100 Hz before any processing of the data. EEG recordings were processed through the BrainScope algorithms for artifact detection 34 to identify for removal any biological and non-biological contamination, including lateral and horizontal eye movement, electromyography muscle activity, high-frequency impulse artifacts, extreme low-amplitude EEG activity, and atypical electrical activity. The artifact-free EEG data were then submitted to fast Fourier transform to extract quantitative electrophysiology (QEEG) features of absolute and relative (%) power, mean frequency, inter- and intrahemispheric coherence, and asymmetry computed for the standard frequency bands. In addition to these traditional QEEG features, several additional features were included: fractal measures, information theory-based measures (entropy and wavelet entropy), and connectivity measures (phase lag, phase synchrony, and various across-region ratios of spectral power and coherence). All these EEG measures are described in detail elsewhere. 34 All features were transformed for Gaussianity and regressed against age.
Enhanced multi-model Brain Function Index
The eBFI was based on a binary classifier trained to separate the participants with SRC from a normative sample. 31 The feature set offered to the classifier builder was composed of EEG features, a reduced set of neurocognitive assessment features (throughputs), and a clinical sign/symptom feature reflecting vestibular/balance symptoms/issues (numeric feature scored on a 7-point Likert scale). EEG and neurocognitive features were all transformed to age-regressed z-transformed scores (z scores). The index was based on the outputs of the binary classifier (discriminant scores) and mapped to a percentile scale (percentile ranking from 1 to 100; i.e., percentage of normative population that performed more poorly) for statistical interpretability. Features that contributed most to the classifier/index included EEG coherence, phase synchrony, and asymmetry EEG features; neurocognitive features; and a clinical sign and symptoms feature reflecting vestibular/balance issues. Greater details on the methodology for classifier development and derivation of such a multi-modal index are described elsewhere. 31
Neurocognitive data acquisition and clinical signs/symptoms
Select measures from a neurocognitive assessment were included in the calculation of the eBFI. The neurocognitive tests were collected on a Microsoft Surface tablet (Microsoft Corp., Redmond, WA) after EEG data acquisition and included the simple reaction time (repeated twice), procedural reaction time, go/no-go reaction time, and code substitution (including the delay) subtests of the Automated Neuropsychological Assessment Metric battery. 42 Results included age- and sex-normed speed, accuracy, and throughput (a multi-variate combination of speed and accuracy). Signs and symptoms were collected using the Sports Concussion Assesment Tool 3 (SCAT 3) symptom inventory questionnaire, which uses a 0- to 6-point Likert scale. 7,43
Statistical analysis
Demographic characteristics for the SRC and comparison athlete groups were compared using univariate analyses of variance for continuous variables and chi-square for categorical variables. Multiple one-way analyses of covariance (ANCOVAs) were performed to examine group differences on DTI metrics (i.e., FA, MD, RD, or AD) of the 100 nodes along each of the white matter tracts with age, sex, and head motion (i.e., the average volume-by-volume translation and rotation) as covariates. Group differences were determined on the basis of differences in “clusters” of five or more contiguous segments within a specific tract in order to isolate the spatial location(s) of differences in white matter diffusivity between the groups. The false discovery rate (FDR) was used to correct for multiple comparisons for group comparisons on DTI metrics. 44 Pearson correlation coefficients were used within the SRC group to assess whether the DTI metrics of clusters that differed between the groups were associated with the initial post-injury multi-modal eBFI.
Results
Demographic and injury characteristics
Demographic characteristics for each group are shown in Table 1. Groups did not differ on age, sex, handedness, and race/ethnicity. With the exception of concentration scores (SRC < comparison), groups did not differ on symptoms at the time of the EEG.
Common Data Elements imaging findings
One participant in the SRC group demonstrated possible acute trauma-related pathology (i.e., focal hemorrhage in the right anterior temporal pole). Four participants in the SRC group had incidental findings of 1) minor prominence of the cerebellar folia, 2) slightly pronounced extra-axial and perivascular spaces, 3) developmental venous anomaly in the right frontal area, or 4) arachnoid cyst with unusual configuration of the frontal horns of the lateral ventricles. One athlete in the comparison group had a small right mid-cranial arachnoid cyst. White matter hyperintensities were noted in 2 additional participants in the SRC group, but not in the comparison group. Cavum septum pallucidi (usually 1–2 mm) were also noted in 6 participants in each group.
Group differences in diffusion tensor imaging metrics and correlation with extended Brain Function Index
Results are summarized in Table 2 and Figure 1. None of the group differences in DTI metrics survived FDR correction, although regions of interest (ROIs) with nominally significant effects were used in Pearson correlation analyses to examine the effects of DTI metrics and eBFI results.
(
Summary Data for Tract Clusters that Differed Significantly in Diffusivity between the SRC Group and the Comparison Athlete Group
Note: None of the differences survived correction for multiple correction (i.e., FDR).
SRC, sports-related concussion; IFOF, inferior frontal-occipital fasciculus; UF, uncinate fasciculus; ILF, inferior longitudinal fasciculus; SD, standard deviation; DTI, diffusion tensor imaging; eBFI, enhanced Brain Function Index; FDR, false discovery rate.
Mean diffusivity
One-way ANCOVAs revealed several regions of interest within the tracts comprising the genu of the corpus callosum (i.e., forceps minor), left and right inferior fronto-occipital fasciculi, and left uncinate fasciculus. The SRC group had lower MD values within these regions as compared to the comparison athlete group. MD measured in two of these clusters correlated significantly with the eBFI data, including the right (r = 0.39, p = 0.043) and left (r = 0.49, p = 0.006) inferior frontal-occipital fasciculus, where higher diffusivity was related to higher eBFI score. There was a marginally significant correlation between MD and eBFI in the corpus callosum (r = 0.33, p = 0.083).
Axial diffusivity
For AD, there were significant ROIs in the following tracts: genu and splenium (i.e., forceps major) of the corpus callosum; right and left inferior frontal-occipital; inferior longitudinal; and uncinate fasciculi. In each region, higher AD was observed in the SRC group as compared to the comparison group, except that the opposite pattern was observed for AD values in the right inferior frontal-occipital fasciculus. Of these regions, significant correlations were observed between the eBFI and callosum forceps minor (r = 0.34, p = 0.033), and left inferior frontal-occipital (r = 0.39, p = 0.035) and inferior longitudinal (r = 0.51, p = 0.005) fasciculi, where higher eBFI was associated with higher AD. There was a marginally significant correlation between AD and eBFI in the left uncinate fasciculus (r = 0.33, p = 0.083).
Radial diffusivity
Clusters were detected in the splenium of the corpus callosum, bilateral inferior frontal-occipital fasciculus, left uncinate fasciculus, right cingulum bundle, and right thalamic radiation, whereby RD values were lower in the SRC group as compared to the comparison group. Of these regions, significant correlations were observed between the eBFI and splenium of the corpus callosum (r = 0.44, p = 0.020) and the right inferior frontal-occipital fasciculus (r = 0.43, p = 0.024). Higher eBFI was associated with higher RD in the SRC group.
Fractional anisotropy
Finally, for FA, clusters were detected in the left inferior frontal-occipital (SRC > comparison) and inferior longitudinal (SRC < comparison) fasciculi. FA values were higher in the left inferior frontal-occipital fasciculus, but lower in the left inferior longitudinal fasciculus, in the SRC group as compared to the comparison athlete group. A significant correlation was observed between the eBFI and left inferior longitudinal fasciculus (r = 0.51, p = 0.005), with higher eBFI associated with higher FA in the group of athletes with SRC.
Further exploratory analyses
Given that there remains considerable lack of consistency in the direction of change in DTI metrics in the subacute post-injury interval in the literature, we also created a standard score for each region, which was calculated using the mean and standard deviation of the comparison group to create a normalized MD, RD, AD, and FA. These standardized scores were then correlated with the eBFI score. Although there were no significant correlations between MD and eBFI using this method, there were marginally significant differences in the genu of the corpus callosum (forceps minor; r = 0.31, p = 0.102), right (r = 0.34, p = 0.077) and left (r = 0.43, p = 0.067) inferior frontal-occipital fasciculus, and left uncinate fasciculus (r = 0.34, p = 0.106). Additionally, AD in the callosal forceps minor (r = −0.36, p = 0.052), corpus callosum major (r = −0.36, p = 0.062), and left inferior longitudinal fasciculus (r = −0.37, p = 0.051) were marginally related to eBFI. For RD, the left uncinate fasciculus (r = −0.45, p = 0.015) and right inferior frontal-occipital fasciculus (r = −0.51, p = 0.005) were significantly related to eBFI. Finally, for FA, a significant correlation between the left inferior longitudinal fasciculus (r = −0.38, p = 0.043) and eBFI was demonstrated. In these correlations, a greater deviation of the DTI metric from the norm was associated with lower eBFI.
Discussion
We compared DTI metrics and a composite, quantitative EEG measure in collegiate athletes with prolonged recovery (exceeding 14 days) from SRC to a group of non-contact, comparison athletes with no histories of previous concussion. We found significant between-group differences in DTI metrics in multiple tracts, including the corpus callosum, inferior fronto-occipital fasciculus, inferior longitudinal fasciculus, uncinate fasciculus, cingulum bundle, and thalamic radiations, that did not survive correction for multiple comparison. Moreover, the extent of observed DTI-related white matter alterations correlated with a novel composite measure that incorporates components of EEG measurement, cognitive testing, and vestibular clinical signs, the eBFI (i.e., enhanced multi-modal Brain Function Index).
These results were further supported by the results of an exploratory analysis that disclosed significant correlations between a standardized DTI metric and eBFI. Though preliminary, our findings suggest that the absolute deviation of DTI metrics from non-contact comparison athletes may be clinically useful. The correlation with an EEG-based index suggests the potential for such measures to be used at the early point of care as a surrogate for neuroimaging in the initial assessment of concussion.
The current findings are important and clinically relevant because the extant DTI literature is inconsistent regarding the directionality of aberrant metrics such as FA and MD post-SRC. DTI is the most commonly utilized diffusion MRI technique in imaging studies of SRC. 10,17 Microstructural white matter changes, as indexed by diffusion metrics, have been reported in studies of collegiate athletes who sustained an SRC at post-injury intervals ranging from <24 h to 6 months, and a subset of studies have reported longitudinal imaging data. 10,45,46 As reviewed elsewhere, 10 studies have reported altered diffusivity in several brain regions post-SRC, including subregions of the corpus callosum, corona radiata, inferior longitudinal fasciculus, anterior and posterior limbs of the internal capsule, and frontotemporal white matter, especially of the right hemisphere. Despite variability in DTI findings during the first few days post-SRC, most studies have found that at least a subset of the diffusion metrics measured at 7 days are altered in adolescents with SRC as compared to controls. 45,46
Contextual caveats require consideration in the interpretation of the current results within the published literature to date. First, the direction in which the metrics are altered is inconsistent across studies; some investigators have reported increased FA, whereas others have found decreased FA, and the directionality of other DTI metrics is also variable. 45 For instance, another study of collegiate athletes with SRC showed increased MD relative to a non-contact-sport comparison athlete group. 47 Differences in results could reflect our current focus on athletes whose recovery was delayed (i.e., RTP ≥14 days post-injury). Additionally, differences in methodology, including DTI processing and post-injury interval, could contribute to differences between results.
Second, characteristics of the control group may impact detection of significant between-group differences in DTI metrics. 48 Emerging evidence indicates that exposure to a season of contact sport associated with repetitive, subconcussive TBIs is sufficient to alter DTI metrics even in players without a concussion. 23 Although some studies have compared patients with mild TBI to patients with extracranial or orthopedic injuries, recent evidence suggested that DTI in both populations differs from healthy controls with no histories of concussion. 48
The current study is the first to combine DTI and EEG to examine SRC. The relationships of these modalities have been previously examined in blast-related concussion in soldiers and military personnel. Sponheim and colleagues 33 reported a significant correlation between changes in mean FA of four major white matter tracts related to frontal interhemispheric communication and changes in phase synchrony of the EEG between frontal and frontotemporal regions. Wang and colleagues 32 examined the relation of these modalities in chronic phases of mild TBI in military personnel, also observing a positive relation between FA and their measure of EEG phase synchrony (i.e., weighted phase lag).
Our investigation extends existing reports in several important ways. First, despite evolving literature in DTI-related findings and separate literature on EEG-derived findings in athletes with SRC, this represents one of the first reports to examine the relation between these quantitative modalities. The eBFI constitutes a novel and efficient quantitative approach for assessment of SRC, through the use of an empirically derived composite measure that incorporates salient EEG, cognitive, and clinical features; this composite is related to DTI-derived alterations in white matter, particularly in the frontal and temporal areas, which are highly vulnerable to traumatic forces, such as rotational acceleration of the brain, which occur in concussion.
Whereas the EEG features at the core of the eBFI are recorded from only frontal and frontotemporal locations, the eBFI showed significant correlations with tracts that go to more posterior regions, suggesting the sensitivity of this EEG-based index to reflect integrity of white matter tracts, even when distant from the recording electrodes. Finally, our approach to imaging analysis enables more precise investigation of specific foci along major tracts that may be affected in SRC. In addition to traditional analysis of common DTI metrics (i.e., FA, MD, AD, and RD), we incorporated an approach to examine variability in the direction of the differences by utilizing absolute deviation from expected values.
This study was limited to examination of diffusion MRI, EEG, and clinical data in the subacute post-injury interval of SRC in athletes with a relatively prolonged recovery (i.e., RTP ≥14 days). Our decision to focus on a comparison of athletes who had prolonged clinical recovery from SRC and a non-concussed comparison group of athletes involved in non-contact sports provided a sharp contrast for this initial study concerning the correlation of DTI with eBFI. We reasoned that a comparison between athletes with SRC and athletes involved in non-contact sports who were largely free of previous TBI (i.e., non-concussed) would enable greater clarity for us to study the effects of recent SRC.
It is possible that the inclusion of a comparison group comprised of athletes involved in contact sports may have better isolated the effects of subacute SRC from the cumulative effects of previous concussion and repetitive TBIs (or subconcussive blows to the head). However, our primary goal was to study and validate the diagnostic and prognostic features of the eBFI for subacute SRC in athletes by correlating the eBFI with DTI metrics, which are more established biomarkers of white matter injury after SRC. Further investigation may be needed to determine the consistency of these findings in longer post-injury intervals.
Additional limitations of this study include the lack of detailed history concerning early exposure to TBIs, rates of attention-deficit/hyperactivity disorder, and the lack of long-term follow-up. Finally, although only data from the time of injury was discussed in this article, it could have been beneficial to examine symptoms at the time that the RTP decision was made. Whereas the NCAA guidelines for graded RTP were followed at each site, the protocol is intentionally designed to test whether symptoms not present at rest are elicited with varying degrees of activity. The response to this exercise can be variable among athletes, and therefore we elected to define and record clearance for RTP (which would indicate symptom resolution at all levels of activity) rather than the duration from injury to the initial report of symptom resolution.
Conclusions and Future Directions
Our findings suggest that the enhanced multi-modal Brain Function Index (i.e., eBFI), a novel composite quantitative measure that incorporates EEG features, limited cognitive testing, and clinical signs after concussion, relates to white matter changes observable by DTI in contact sports athletes with subacute SRC. These initial findings warrant further investigation in a larger sample.
Footnotes
Acknowledgments
The views, opinions, and/or findings contained in this report are those of the author(s) and should not be construed as an official Department of the Navy position, policy, or decision unless so designated by other documentation. We acknowledge the technical and intellectual contributions of Drs. Leslie Prichep and Arnaud Jacquin. Finally, we thank the athletes for their participation in this study.
Funding Information
The clinical study was funded, in part, by a contract to BrainScope Company Inc. from the U.S. Navy (Naval Health Research Center), contract #W911QY-14-C-0098.
Author Disclosure Statement
No competing financial interests exist.