Preliminary evidence from a prospective DTI study suggests a posterior‐to‐anterior pattern of recovery in college athletes with sports‐related concussion

Abstract Objectives We compared the integrity of white matter (WM) microstructure to the course of recovery in athletes who sustained one sports‐related concussion (SRC), assessing individual longitudinal changes in WM fiber tracts following SRC using pre‐ and post‐injury measurements. Materials and Methods Baseline diffusion tensor imaging (DTI) scans and neuropsychological tests were collected on 53 varsity contact‐sport college athletes. Participants (n = 13) who subsequently sustained an SRC underwent DTI scans and neuropsychological testing at 2 days, 2 weeks, and 2 months following injury. Results Relying on tract‐based spatial statistics (TBSS) analyses, we found that radial diffusivity (RD) and mean diffusivity (MD) were significantly increased at 2 days post‐injury compared to the same‐subject baseline (corrected p < 0.02). These alterations were visible in anterior/posterior WM regions spanning both hemispheres, demonstrating a diffuse pattern of injury after concussion. Implicated WM fiber tracts at 2 days include the following: right superior/inferior longitudinal fasciculus; right/left inferior fronto‐occipital fasciculus; right corticospinal tract; right acoustic radiation; right/left anterior thalamic radiations; right/left uncinate fasciculus; and forceps major/minor. At 2 weeks post‐injury, persistently elevated RD and MD were observed solely in prefrontal portions of WM fiber tracts (using same‐subject contrasts). No significant differences were found for FA in any of the post‐injury comparisons to baseline. Plots of individual subject RD and MD in prefrontal WM demonstrated homogenous increases from baseline to just after SRC; thereafter, trajectories became more variable. Most subjects’ diffusivity values remained elevated at 2 months post‐injury relative to their own baseline. Over the 2‐month period after SRC, recovery of WM fiber tracts appeared to follow a posterior‐to‐anterior trend, paralleling the posterior–anterior pattern of WM maturation previously identified in the normal population. Conclusion These results suggest greater vulnerability of prefrontal regions to SRC, underline the importance of an individualized approach to concussion management, and show promise for using RD and MD for imaging‐based diagnosis of SRC.


| INTRODUC TI ON
Accruing scientific evidence regarding both the serious short-term and potential long-term consequences of concussive injury in college student athletes has received considerable media attention and raised awareness of concussion as an important public health problem. As such, efforts to identify objective neurobiological correlates of symptom-based diagnosis of and recovery from concussion have provided a deeper understanding of the physiological and structural consequences of concussive injury.
For example, diffusion tensor imaging (DTI) studies show evidence of structural alterations in college athletes who sustained a sports-related concussion (SRC) but did not experience a loss of consciousness (LOC) and did not score below 15 on the Glasgow Coma Scale (GCS) (Teasdale & Jennett, 1974). Specifically, in college athletes exhibiting prolonged symptoms (>1 month after SRC) without LOC, increased mean diffusivity (MD) was reported in parts of the left inferior/superior longitudinal and fronto-occipital fasciculi, the retrolenticular part of the internal capsule, and the posterior thalamic and acoustic radiations (Cubon, Putukian, Boyer, & Dettwiler, 2011). A comparison between measures taken 2 days and 2 weeks post-injury in varsity contact-sport college athletes also showed increased radial diffusivity (RD) in a cluster of right hemisphere voxels, spanning the posterior limb of the internal capsule, the retrolenticular part of the internal capsule, the inferior longitudinal fasciculus, the inferior fronto-occipital fasciculus (sagittal stratum), and the anterior thalamic radiation . These findings are supported by a recent study, conducted as part of the NCAA-DOD Care Consortium, which found that football players diagnosed with SRC displayed higher MD in frontal and subfrontal white matter (WM) fiber tracts compared to controls within 48 hr post-injury (Mustafi et al., 2017). Additionally, in the concussed group, a significant positive correlation was found between axial diffusivity (AD) and clinical measures including the Brief Symptom Inventory and the Sports Concussion Assessment Tool (SCAT). Fractional anisotropy (FA) measures were also found to be correlated with Standardized Assessment of Concussion (SAC) performance. The studies presented here highlight increased diffusivity, sometimes accompanied by a corresponding decrease in FA, following SRC. Similar trends are observed in DTI studies on mTBI. For instance, two additional studies (D'souza et al., 2015;Toth et al., 2013) report increased MD and decreased FA during the acute phase of mTBI as compared to age-and sex-matched healthy controls with FA decreases persisting at 1 month (Toth et al., 2013).
An additional TBSS study reports increased RD and decreased FA in the subacute phase of mTBI without post-concussion syndrome as compared to healthy controls with FA decreases persisting at 6 months (Messé et al., 2012). However, there are also literature reports of decreased MD following SRC. For instance, decreased MD, decreased RD, and increased FA were reported during both the acute and chronic phases of SRC (Henry et al., 2011), whereas recent TBSS analyses report widespread decreased MD, RD, and AD during the acute and subacute phases of SRC with decreased MD and RD persisting at 6 months post-concussion (Lancaster et al., 2016(Lancaster et al., , 2018. Although the directionality of diffusion metrics varies across studies and should be further addressed in future research, previous literature collectively suggests DTI values are abnormal in acute, subacute, and even chronic phases of SRC and mTBI as compared to control populations. Furthermore, structural differences in the deep WM have been detected in DTI studies assessing the effects of subconcussive blows to the head over the course of an entire season in college athletes playing high-risk sports (Koerte et al., 2012). For instance, increased RD and AD were found in the right precentral region, corona radiata, and both the anterior and posterior limb of the internal capsule in varsity college ice hockey players when comparing DTI structural metrics derived from pre-versus post-season DTI scans.
Additionally, lower cognitive function (CogState30) and decreased fractional anisotropy (FA) in temporo-occipital WM were found to be associated with a high-frequency heading rate (>885 -1,800 headings per year) in adult amateur soccer players (Lipton et al., 2013).
Alterations in the WM microstructure have also been observed in contact-sport college athletes with repeated subconcussive blows to the head, whereas no such changes have been identified in control participants (Bazarian et al., 2014;Lao et al., 2015).
Despite accruing evidence indicating structural and physiological differences after SRC, the course of recovery of WM fiber tracts continues to be poorly understood. This knowledge gap is especially salient when considering the ongoing myelination of the prefrontal portion of WM fiber tracts that continue into the mid-20s. Although the rate of increase in WM volume in the brain is known to slow after age 10, WM volume has been shown to continue to increase into early adulthood (Giedd, 2004;Iwasaki et al., 1997). A longitudinal MRI study of neurologically normal subjects spanning from children to young adults ranging from 5 to 25 years of age found the rate of WM volume increase to be linear in all four regions of the brain (occipital, temporal, parietal, and frontal) until the mid-20s (Giedd, 2004;Giedd et al., 1999). While multiple other studies have found myelination to continue into the early to mid-20s, still others have found myelination to continue for years beyond, until closer to age 30 (Bashat et al., 2005;Benes, Turtle, Khan, & Farol, 1994;Giedd et al., 1999;Giorgio et al., 2008;Lebel, Walker, Leemans, Phillips, & Beaulieu, 2008;Snook, Paulson, Roy, Phillips, & Beaulieu, 2005;Steen, Ogg, Reddick, & Kingsley, 1997). Utilizing histological autopsies, investigators found myelination to continue until age 29, while two separate DTI studies found myelination to occur between the ages of 8 to 27 years and 3 to K E Y W O R D S diffusion tensor imaging, mean diffusivity, mTBI, radial diffusivity, sports-related concussion 30 years, respectively (Benes et al., 1994;Lebel et al., 2008;Snook et al., 2005). Both studies found increases in FA in prefrontal WM tracts, thus indicating WM growth far into the late twenties (Lebel et al., 2008;Snook et al., 2005). Of particular interest is the finding that FA increases along WM fiber tracts in the adult age range were localized to more frontal regions (Benes et al., 1994). Additional studies have reinforced this finding, showing that prefrontal regions tend to develop quite slowly and are the last regions of the brain to become fully myelinated (Barnea-Goraly et al., 2005;Klingberg, Vaidya, Gabrieli, Moseley, & Hedehus, 1999;Schmithorst, Wilke, Dardzinski, & Holland, 2002;Yakovlev & Lecours, 1967). Taken together, these findings are consistent with the current understanding that the development of WM fiber tracts in the brain follows a posterior-to-anterior trend, leaving the prefrontal areas as some of the last brain regions to reach maturation. As a result, prefrontal areas may be more susceptible to concussive injury before full WM fiber tract maturation is reached.
Interestingly, a study involving high school varsity-level football players found that even without the diagnosis of SRC, post-season DTI measures demonstrated impaired WM tract diffusivity when compared to pre-season DTI measures (Davenport et al., 2014). Additionally, a statistically significant linear association was found between the measured total impact and change in WM tract diffusivity (Davenport et al., 2014). Based on evidence previously discussed that prefrontal WM fiber tracts and prefrontal portions of long WM fiber tracts are the last to mature in the normal population, it thus appears important to take the stage of brain development into consideration when evaluating concussive injury.
This study was designed to examine the temporal and spatial course of structural recovery after concussion in college-aged athletes. In order to elucidate individual differences, this study assessed each athlete's individual pattern of recovery in brain WM during the 2 months after injury using advanced DTI and a baseline, pre-injury scan. Since the unique use of a baseline scan allows for the examination of specific patterns of individual recovery and structural repair, the results of this study and others based on its findings might, in the future, provide an imaging correlate of neural recovery processes in the deep WM after SRC. Ultimately, this preliminary study might add critical information for an individualized approach to concussion management and prevention of re-injury. Y: scan included in analysis; N: scan excluded either due to excessive head motion or unavailability of the subject; DNR a : subject developed symptoms during return-to-play progression and, hence, did not return-to-play; DNR b : subject did not return-to-play, athletic season over; N/A: NP testing did not return to baseline.

respectively.
A total of 14 of the 53 subjects enrolled in this study subsequently sustained a sports-related concussion. These athletes were evaluated by athletic trainers and team physicians within 48 hr postinjury at University Health Services. The diagnosis of concussion was established using the criteria of the 4th International Consensus Conference on Concussion in Sport . Baseline and post-injury NP testing protocols were identical to those described in our previous study . All concussed athletes participated in NP testing within 24-48 hr after injury.
None of the athletes experienced a loss of consciousness (LOC), and further assessment by the GCS or clinical radiological examination was not warranted. Abnormal performance on NP tests was determined through comparison of post-injury NP test scores to the participant's baseline scores. Abnormality of ImPACT clinical composites was based on reliable change indices at the 0.8 confidence interval (Iverson, Lovell, & Collins, 2003). Scores on the traditional NP test performance were assessed using Princeton-specific normative data.
Return-to-activity decisions were made by the team physicians. After the athlete was symptom free and their clinical examination, including balance and NP test results, returned to baseline levels, he/she participated in a personalized return-toplay progression. Once an athlete was symptom free at rest, had successfully completed the physical activity program, and NP test results were back to baseline, he/she was cleared to return-toplay. Follow-up MRI scans and NP tests at 2 days, 2 weeks, and 2 months post-injury were performed on all 14 subjects (mean age 20.6 years, standard deviation [SD] 1.5, for subjects in Table 1) who sustained a concussion with the exception of subjects missing a time point due to subject unavailability or excessive head motion. One subject was completely excluded from analysis due to excessive motion in the baseline scan. For further details, see Table 1. Participants had no self-reported history of medical, genetic, or psychiatric disorder and presented without contraindications to MRI. History of concussion was obtained through self-report. It should be noted that it is difficult to evaluate number of previous concussions objectively in contact-sport athletes given the limitations and subjectivity of self-report. The study was approved by Princeton University's Institutional Review Board, and written consent was obtained from all athletes prior to their participation in the study.

| Image analysis
For each scan session, the diffusion dataset was concatenated with the 33 B0 volumes and then eddy current and motion corrected using the first B0 volume for reference (FSL software, https://www. fmrib.ox.ac.uk/fsl/, RRID:SCR 002823) (Smith et al., 2004). Images  fiber tracts in the brain, which is known to continue in the prefrontal cortex into early adulthood (Bashat et al., 2005;Benes et al., 1994;Giedd et al., 1999;Giorgio et al., 2008;Lebel et al., 2008;Snook et al., 2005;Steen et al., 1997). Additionally, superior axial slices reveal an anterior shift of significant voxels to adjacent regions from 2 days to 2 weeks post-injury (bottom rows in Figure 2) suggesting an ongoing or delayed process is occurring in prefrontal WM during recovery.

| 2 months post-injury to baseline
No significant differences (at p < 0.02) were found for RD or MD in the 2 months post-injury to baseline comparisons using paired, F I G U R E 1 Flow chart of data analysis between-session t tests on the WM skeleton. However, a trend of increased RD and MD in frontal cortex (at p < 0.05) was observed.
No significant differences or trends were found for FA in any of the post-injury comparisons to baseline.

| D ISCUSS I ON
Previous longitudinal studies assessing concussion typically evaluate pre-and post-season measurements. The current preliminary study, however, investigated pre-and post-injury measurements to monitor an athlete's progress toward recovery using DTI. This study revealed diffuse alterations throughout the brain in both posterior and anterior regions of deep WM immediately after only one diagnosed sports-related concussion and, more importantly, it uncovered a distinctive pattern of brain recovery that indicates a posterior-to-anterior progression. Diffusivity differences persisted in the prefrontal segment of WM fiber tracts at 2 weeks and, for some subjects, that these tracts may be more fragile or particularly vulnerable and less resilient to concussive injury. Remarkably, this preliminary finding regarding brain recovery following concussion appears to mirror the sequence of WM fiber tract maturation, suggesting that later maturing brain regions may be more susceptible to concussive injury.

While DTI offers a retrospective view of structural alterations
in the WM of the brain, it does not reveal the exact location of impact or the level of g-force with which the subject was hit. Studies have found that devices that collect those metrics may be useful for identifying concussive blows. For instance, finite element models (FEM) attempt to locate brain region-specific strain and stress responses to concussion by simulating the biomechanical event using recorded impact kinematics. The collection of accelerometer data for FEM analyses has progressed from laboratory-based impact reconstructions to actual helmeted and unhelmeted impacts from sub- be accounted for by the fact that sensor data interpretation rests on the assumption that the skull and brain move continuously as one, whereas the brain is actually surrounded by cerebrospinal fluid within the skull and is affixed to the brain stem. Hence, the brain can move within the skull. Given these anatomical facts and the findings described above, the addition of sensor data would, at this point in time, most likely not have broadened our DTI findings.
Examination of subjects' self-report of their respective impact locations revealed there was no tendency toward frontal impacts in our subject population. In fact, the site of impact was variable across subjects with most subjects (38%) reporting the side of the head Two previous DTI studies, assessing either athletes with prolonged symptoms or longitudinal post-injury differences, also reported structural alterations identified by increased diffusivity in nearly the same posterior anatomical region (sagittal stratum) but in contralateral hemispheres (Cubon et al., 2011;Murugavel et al., 2014). Specifically, RD decreased significantly in a comparison between 2 days and 2 weeks post-injury in the right sagittal stratum, indicating recovery in this particular brain region . Interestingly, in the current study, similar right hemisphere posterior anatomical regions also demonstrated increased RD and MD at 2 days post-injury when compared to their own baseline, but not at 2 weeks or 2 months post-injury. Collectively, these studies suggest that the posterior region recovers within 2 weeks after in- tern that brain development follows. Future studies will be needed to confirm this trend of recovery in college-aged athletes in addition to studies investigating how WM fiber tract injury and subsequent recovery patterns may be altered at different stages of brain development, namely childhood and adolescence. As such, additional research across other ages and stages of development must also be conducted in order to further understand how concussive injury may interact with WM fiber tract development.
A prior longitudinal study investigated concussive injury using a different imaging modality that assessed pre-and post-injury measurements of myelin water fraction. This study found that concussed athletes demonstrated myelin disruption at both 2 days and 2 weeks post-injury. However, these values normalized to pre-season values by 2 months (Wright et al., 2016). Other longitudinal studies typically assess concussion using pre-and post-season measurements with a particular focus on repetitive head impacts. For example, a DTI study comparing pre-and post-season measurements in a group of varsity ice hockey players (mean age 22 years) revealed increased diffusivity values in the right precentral region, right corona radiata, and the anterior and posterior limb of the internal capsule, over the course of one entire season (Koerte et al., 2012). Additionally, heading by amateur soccer players (mean age 30.9 years) was associated with poorer neurocognitive performance on memory tests and abnormal WM microstructure in temporal-occipital regions (Lipton et al., 2013). Taken together, these studies reveal effects of repetitive head impacts rather than monitoring the effects of and recovery from one clinically diagnosed concussion, as was the purpose of the current study. Bearing this in mind, it must be taken into consideration that, in the current study, subconcussive hits experienced during the time from the baseline scan to concussion and once the athlete returned-to-play may have contributed to diffusivity changes. However, these contributions would be highly variable and, on average, subjects demonstrated increased diffusivity at 2 days followed by a gradual decline at 2 weeks and 2 months post-injury (solid black line in Figure 4).
Remarkably, most subjects in the current study were not back to their baseline diffusion values by 2 months in prefrontal WM, an area underlying the lateral and dorsolateral frontal cortical regions.
These regions are known to be essential for complex behaviors such as executive control, decision-making, and impulse control. Recent work looking at subjects from 8 to 26 years of age using diffusion-weighted imaging (DWI) found a positive correlation between impulse control and age, noting that individuals with higher WM integrity of the frontostriatal tracts demonstrated better delayed gratification (Achterberg, Peper, Duijvenvoorde, Mandl, & Crone, 2016).
As WM fiber tracts matured into early adulthood, so too did future delay of gratification skills. A similar DWI study of subjects from 8 to 25 years of age also found that increases in frontal-striatal connections were correlated with improved impulse control, which was also found to be positively correlated with age (van den Bos, Rodriguez, Schweitzer, & McClure, 2015). Given that prefrontal WM tracts mature into the mid-20s and maturity has been found to be correlated with impulse control, our preliminary findings demonstrating prolonged recovery after injury thus raise concern regarding the consequences concussion may have for the development of executive function.
In light of these relationships, damage to WM fiber tracts in prefrontal regions may have potentially serious long-term effects.
Prominent neurofibrillary tangles (NFT) have been found in isolated perivascular foci within neurons and astrocytes at the base of the sulci limited to the frontal lobe in the early stages (mild pathology stages I and II) of chronic traumatic encephalopathy (CTE; Stein, Alvarez, & McKee, 2015). The stage of CTE has been positively correlated to the number of years an athlete has played American football, suggesting that prolonged duration to repeated blows to the head may lead to CTE (McKee et al., 2013). As of now, there is no method available to determine whether a young athlete will develop CTE and, as such, a link between concussive injuries experienced earlier in life and CTE cannot be inferred. However, given the regions of the brain associated with early stages of CTE, our findings that WM fiber tract recovery was delayed in the frontal regions of the brain are particularly salient. Our preliminary results of the current study as well as earlier studies Murugavel et al., 2014) suggest that neural recovery takes much longer than neurocognitive recovery, thus potentially indicating that athletes are returning-to-play before full neural recovery is complete. This is most apparent when evaluating individual subject diffusivities over time in prefrontal cortex, highlighting a unique advantage of indi- With these data from baseline, 2-day, 2-week, and 2-month scans, the current study was thus able to elucidate prefrontal WM as a brain region that exhibits pronounced increased diffusivity with prolonged variable recovery following concussion in athletes A recent longitudinal mTBI study that also utilized an individualized approach reported similar increases in RD and MD that additionally correlated with worse memory performance and worse somatic autonomy (Strauss et al., 2016). Interestingly, the authors suggest that increased RD may serve as a good early predictor of long-lasting dysfunction (Strauss et al., 2016). Even though post-injury FA abnormalities were also reported, all of the post-injury time points in the current study displayed insignificant FA results. These findings are consistent with results of our previous publications where diffusivity values were more sensitive to concussive injury, even across various SRC populations (Cubon et al., 2011;Murugavel et al., 2014).
Substantial evidence by both research previously discussed and the current preliminary study supports the validity and importance of an individualized longitudinal approach with baseline measurements to manage concussion (Strauss et al., 2016). Despite the low subject number in our current study, the DTI data show promise in providing a sensitive method to temporally monitor brain regions which are vulnerable to concussive injury and to assess each athlete's individual path to recovery. Future work should further evaluate water diffusivity in the prefrontal WM ROI as a potential diagnostic marker and as a tool to monitor the progression of recovery following SRC, observing whether it follows the same posterior-to-anterior trend noted in this study. Additional studies will need to include comparisons between gender, and other age ranges so as to observe the potentially varying effects of concussion at different stages of brain development. Given that this was a preliminary study, future studies with a larger subject number should also be conducted. One limitation of DTI studies in general is their inability to identify the exact underlying structural cause of any changes in diffusion measures; however, the power of DTI should not be lost by this statement, as DTI values are influenced by structural alterations such as axonal ordering, axonal density, degree of myelination, increased water content due to edema, axonal packing, or results of inflammatory processes (Beaulieu, 2002;Jones, Knösche, & Turner, 2013). Even though TBSS analyses provide enhanced alignment of deep WM fiber tracts across subjects, results are limited to larger bundles of these tracts and potential changes at the gray-white matter junction are not effectively captured. This is a second limitation of DTI in general, and methods to address this issue were not within the scope of the current study and should be investigated in future research.
While the current study's subject population did not experience persistent symptoms, future studies should investigate whether subjects with post-concussive syndrome also show prolonged changes in water diffusivity from baseline in the prefrontal WM ROI. To assess whether structural differences in prefrontal WM fiber tracts are predictive of neurocognitive outcome, correlations of diffusivity values with NP test results and clinical symptoms (symptom score, GAD anxiety score, and PHQ-9 depression screening) will be investigated and presented in a future publication.
In conclusion, despite its small sample size, the current preliminary study provides a longitudinal pre-/post-concussion DTI analysis to monitor recovery processes by tracking diffusion measures from baseline to 2 months after one SRC with no LOC. Within 2 days post-injury, increased water diffusivity (RD and MD) was widespread throughout the brain in all subjects and was followed by a posterior-to-anterior course of recovery reminiscent of brain development patterns. Elevated RD and MD persisted at 2 weeks post-injury in prefrontal WM, thereby highlighting fiber tracts that may, at this age level, be particularly vulnerable to concussive injury.
Moreover, RD and MD remained elevated at 2 months post-injury in most subjects despite normalization of NP test results, suggesting the recovery process in prefrontal WM was not complete by 2 months. Longitudinal, individualized analyses with baseline pre-injury scans, such as the current study, show promise in providing an imaging correlate of recovery processes after SRC with the potential of utilizing an individualized approach to concussion management and prevention of re-injury.

ACK N OWLED G M ENTS
The authors would like to thank Daniel Osherson PhD, for all his advice and support. The authors would also like to acknowledge study was not preregistered with or without an analysis plan in an independent, institutional registry. As of now, the investigators have no plans to make these data available to other researchers.

CO N FLI C T O F I NTE R E S T
No competing financial interests exist.