Mapping default mode connectivity alterations following a single season of subconcussive impact exposure in youth football

Abstract Repetitive head impact (RHI) exposure in collision sports may contribute to adverse neurological outcomes in former players. In contrast to a concussion, or mild traumatic brain injury, “subconcussive” RHIs represent a more frequent and asymptomatic form of exposure. The neural network‐level signatures characterizing subconcussive RHIs in youth collision‐sport cohorts such as American Football are not known. Here, we used resting‐state functional MRI to examine default mode network (DMN) functional connectivity (FC) following a single football season in youth players (n = 50, ages 8–14) without concussion. Football players demonstrated reduced FC across widespread DMN regions compared with non‐collision sport controls at postseason but not preseason. In a subsample from the original cohort (n = 17), players revealed a negative change in FC between preseason and postseason and a positive and compensatory change in FC during the offseason across the majority of DMN regions. Lastly, significant FC changes, including between preseason and postseason and between in‐ and off‐season, were specific to players at the upper end of the head impact frequency distribution. These findings represent initial evidence of network‐level FC abnormalities following repetitive, non‐concussive RHIs in youth football. Furthermore, the number of subconcussive RHIs proved to be a key factor influencing DMN FC.

To address this issue, we performed a prospective and longitudinal cohort study to examine changes in rsfMRI FC of the DMN in youth tackle football players (n = 50; ages 8-14 years). Employing seed-based correlation and a combination of voxel-wise and regionof-interest (ROI) approaches, the primary aim was to test the hypothesis that football-related exposure to subconcussive RHIs over the course of a single season leads to aberrant network-level connectivity of the DMN relative to healthy non-collision sport controls. Second, we performed an exploratory analysis contrasting in-and off-season changes in DMN FC in football player subsample (n = 17) with sevenmonth follow-up data availability (i.e., beginning of the subsequent season). Lastly, we tested the hypothesis that a higher number of experienced head impacts is associated with more adverse FC outcomes in football players. excluded. An additional set of 28 players were removed due to enrollment exclusion criteria, which included self/parent-reported history of neurological and/or psychiatric illness, concussion within the last year, and MRI contraindication (i.e., motion and susceptibility artifacts). Of the remaining 84 pairs of the preseason and postseason scans, 34 included sessions by the same player across multiple seasons. We opted to only use the first enrollment season to limit the inclusion of multiple sessions by the same player. A total of 50 male football players were ultimately retained for study inclusion (average age = 11.5 [SD = 1.2] years) ( Figure 1). Of note, of the 50 total players, 17 had availability of multiple MRI acquisition sessions across backto-back seasons. These player scans were retained for an exploratory analysis and the basis for this analysis is described in Section 2.3. Two football players included in the study had self/parent-reported history of a single concussion more than a year prior to the season and none were diagnosed with a concussion during the season.
Youth male control athletes with no prior tackle football experience were also recruited for this study. Controls were recruited from local non-collision sports teams during a scheduled player/parent meeting. Control participants were represented across six different sports (seven basketball, seven baseballs, two soccer, two tennis, one swimming, one karate). Of the 33 control athletes who elected to enroll in the study, 20 (average age = 11.4 [SD = 1.2] years) were retained for study inclusion based on the aforementioned exclusion criteria. Demographic variables for football and control participants are presented in Table S1.
This study was conducted in compliance with the Health Insurance Portability and Accountability Act. All participants and their parental guardians provided signed assent and consent, respectively, and were in full understanding of experimental objectives. All experimental procedures were approved and monitored by the Wake Forest School of Medicine Institutional Review Board Committee and conducted in ethical accord with the Declaration of Helsinki.

| Head impact telemetry system data acquisition
Football participants were fitted with helmet-based accelerometers using the Head Impact Telemetry System (HIT System: Simbex, Lebanon, NH) Crisco, Chu, & Greenwald, 2004) for real-time acquisition of head impact kinematics during all practice and game sessions. Only those players who were enrolled in the study were equipped with the HIT System. Only players with a head circumference that could adequately fit a medium or large size Riddell Speed® or Riddell Revolution® helmet were permitted to participate, as the head impact sensor system was not compatible with smaller helmet sizes. The HIT System measures the location and magnitude of head impacts using an array of six springmounted single-axis accelerometers in contact with the head surface.
Data acquisition was initiated upon detection of peak resultant linear acceleration of the head above 10 g. An important consideration is that indirect head impacts (e.g., direct impacts to the torso) can initiate data acquisition due to the associated change in peak resultant acceleration of the head above this threshold. All games and practices were video recorded and any accelerations due to activities other than head impacts (e.g., dropped helmet) were excluded from the analysis. For the current study, we calculated the total number of head impacts, head impact severity based on percentiles calculated from the distribution of linear (measured in g) and rotational acceleration (measured in rad/s 2 ), and risk-weighted exposure (RWE CP ). RWE CP represents a measure of cumulative head impact exposure encompassing the number and severity of experienced head impacts, in which the peak resultant linear and rotational accelerations of each impact are non-linearly weighted using the combined probability risk function and summed for each player over the course of the season Urban et al., 2013).

| MRI acquisition and processing
All 50 football players included in this study were scanned at two separate time points, including up to 1 month preceding (i.e., preseason) and following (i.e., postseason) the season. The average time between preseason and postseason acquisition sessions was 4.7 (SD = 0.8) months. This sample of 50 players is henceforth referred to as Cohort 1. Players were asked to enroll across multiple seasons if they remained on a team selected to participate in the study. Of the F I G U R E 1 Flowchart of study enrollment. The left panel shows the number of teams and respective players who were recruited for the study between 2015 and 2017. The right panel describes the process of recruitment to final enrollment for Cohorts 1 and 2 50 players who were included in the study, a subsample (n = 17) was scanned during the subsequent season. This provided the opportunity to examine, on an exploratory basis, DMN FC dynamics between the postseason acquisition scan of the first enrollment season and the preseason acquisition scan of the subsequent season (i.e., off-season).
The average time between these MRI acquisition sessions was 7.2 (SD = 0.6) months. This subsample of 17 players, representing four different teams, is henceforth referred to as Cohort 2. Control participants were scanned at two-time points an average of 4.0 (SD = 0.9) months apart. The timing of initial and follow-up acquisition scans did not always coincide with the preseason and postseason of their respective sport.
MRI sequences were acquired on a 3 Tesla Siemens Skyra MRI scanner using a 32-channel head/neck radiofrequency coil (Siemens Medical, Erlangen, Germany). Participants laid supine and were instructed to remain still with their eyes closed and free of external thought for the experimental duration. The rsfMRI data were acquired using a gradient EPI sequence with the following parameters: repetition time (TR) = 2,000 ms, repetitions = 190, echo time (TE) = 25 ms, flip angle = 90 , field of view (FOV) = 61 × 64, 3.5 mm 3 isotropic resolution. Structural T1-weighted images were acquired for spatial normalization using a 3D-MPRAGE sequence with the following parameters: TR = 2,300 ms, TE = 2.98 ms, flip angle = 90 , FOV = 256 × 256 mm, 1 mm 3 isotropic resolution.

| Functional MRI data preprocessing
Processing of rsfMRI data was carried out using afni_proc.py in Analysis of Functional NeuroImages (AFNI: Version 13.0.3). Preprocessing for each scan included removal of three base volumes to account for scanner magnetization equilibrium, de-spiking of extreme time-series outliers, acquisition-dependent slice-timing correction, and rigid-body volume registration. Anatomical and functional images were co-registered, skull stripped, and non-linearly warped to the MNI avg152 template with a resampled isotropic resolution of 2 mm. Functional volumes were spatially smoothed to improve the signal-to-noise ratio using a 4 mm Gaussian FWHM kernel. Nonspecific or spurious sources of variance from the BOLD time series, including six head motion derivatives describing rigid-body transformations, as well as mean global, CSF, and local white matter signals, were regressed. The regression step also included censoring of consecutive functional volumes >0.5 mm in relative motion and time points with >10% of total voxels identified as signal outliers. Across pre-and postseason scans the average number of censored volumes was 11.62 (SD = 12.21, 3.1% of total TRs) and 11.35 (SD = 15.12, 3.03% of total TRs) for football and control participants, respectively. Lastly, the BOLD time series data was bandpass filtered between 0.008 and 0.1 Hz.

| Functional connectivity statistical analysis
A seed-based correlation approach was used to examine FC using an a priori selected spherical seed region within the posterior cingulate cortex (PCC [MNI coordinates: 0, −53, 25; radius = 6 mm]), a central node of the DMN (Fox et al., 2005;Greicius, Krasnow, Reiss, & Menon, 2003). For each residual time series (preseason, postseason), Pearson correlation coefficients were computed between the seed time series and that of all other voxels and converted to Z-scores using the Fisher r-to-z transformation. Preseason Z-score maps for each participant were subtracted from their postseason counterpart to compute in-season delta FC maps.
Group-level analyses were anatomically constrained to regions comprising the DMN using a masking procedure derived from control participants. To this end, voxels that demonstrated a positive time series correlation with the seed location (p < .05 uncorrected) were obtained across control participants, separately for both preseason and postseason scans. Positive correlation maps were then concatenated across scans and averaged across control participants to generate a single uncorrected DMN mask. Positive cerebellar correlations were discarded due to variable and incomplete spatial coverage across participants. To refine the mask, a one-sample t test in AFNI with the Clustsim option was performed on control datasets at preseason. The Clustsim option provides the cluster-extent volume required to limit false-positive correlations with the seed location.
This option advised a cluster-extent volume ≥ 400 mm 3 , equivalent to p < .005 (p < .05 FWER-corrected). Positive correlation maps were then re-obtained using the advised cluster-extent threshold, concatenated across scans, and averaged across participants. As shown in Figure 2a, the resultant mask is in agreement with previous literature, comprising established DMN regions that exhibit positive time-series correlations with the PCC under task-negative conditions (Fox et al., 2005;Greicius et al., 2003).
Group-level DMN FC differences between Cohort 1 football players and controls were examined using two separate approaches.
In the first analysis, voxel-wise independent-sample t-tests in AFNI (3dttest++) were used to compare FC maps, separately for preseason, postseason, and in-season delta. Age and BMI were entered as covariates for preseason and postseason tests and time between acquisition sessions was entered as an additional covariate when comparing delta maps. Group-level effects were considered significant at a minimum cluster volume of 400 mm 3 (p < .05 FWER-corrected), in accordance with the aforementioned Clustsim output.
A drawback of the voxel-wise approach is that test statistics are computed across thousands of voxels and significant group-level effects depend on a rigorous cluster-extent threshold in accounting for multiple comparisons. Limiting statistical tests to a condensed set of defined regions is advantageous in reducing the severity of the alpha level adjustment for multiple test statistics (Poldrack, 2007). To this end, a region-of-interest (ROI) approach was used in a second analysis to examine group differences in mean Z-score values within subregions of the DMN. In this approach, clustered voxels from the Delta comparisons included time between acquisition scans as an additional covariate. Significant multivariate group main effects were further decomposed using univariate tests across all ROIs and considered significant at p < .05 FDR-corrected.

| Longitudinal connectivity changes in football players
In Cohort 2 football players, we tested whether in-season delta FC, representing a 4-month period of RHI exposure, differed from that of off-season delta FC, representing a 7-month withdrawal from football-related activity. A first step included the previous ROI analysis to identify group-level differences between Cohort 2 football players and controls. The purpose of this first step was to determine whether Cohort 2 football players were representative of the group as a whole, as well as identify a subset of ROIs that best-characterized football and control participants according to in-season delta FC. Second, regions identified as displaying significant in-season delta FC group effects in the ROI analysis were decomposed longitudinally in Cohort 2 football participants. To this end, postseason Z-scores for each identified ROI and each participant were subtracted from their subsequent preseason counterpart scan to compute off-season delta Z-scores. Separately for each region, a two-way linear mixed model was used to examine differences between in-and off-season delta Zscores, with time (in-season, off-season) and participants as the within-and between-subjects random factors, respectively. Age and BMI were included in the models as covariates. The main effects of time across ROIs were considered significant at p < .05 FDRcorrected.

| Effect of head impact frequency on connectivity outcomes in football players
In a final exploratory analysis, we sought to examine whether a high and the low number of cumulative head impacts in football players F I G U R E 2 Control default mode network and ROI clusters. (a) Three-dimensional rendering of the default mode network (red) comprising voxels that exhibit a significant positive time-series correlation with that of the seed region in the posterior cingulate cortex for control participants (p < .05 FWER-corrected). (b) Axial, sagittal, and coronal color-coded depiction of architectural hubs comprising the FWER-corrected default mode network used in the ROI analysis. PFC/l-DFC (yellow), PCC/precuneus (orange), l-TPC (red), r-TPC (light blue), r-DFC (lime green), r-MFG (navy), l-IFG (white), l-LTC (dark blue), r-LTC (cyan), l-PHF I and II (salmon), and r-PHF (pink) yielded differential FC outcomes. In Cohort 1, football participants were partitioned into low and high impact groups according to the lower and upper third percentiles, respectively, of the head impact frequency distribution across participants. This split resulted in 18 low impact players with head impact values ranging from 22 to 151 and 18 high impact players with head impact values ranging from 349 to 2016. Due to the small sample size in Cohort 2 and to preclude further player exclusion, low and high impact groups were partitioned according to the 50th percentile of the head impact frequency distribution (413 impacts). This split resulted in eight players in the low impact group with head impact values ranging from 43 to 341 and nine players in the high impact group with head impact values ranging from 413 to 2016. Players who experienced 413 impacts (i.e., 50th percentile) were included in the high impact group. Table 1 provides descriptive statistics of head impact kinematics based on each respective partition. Specifically, for Cohort 1, we examined whether the number of head impacts differentially influenced the change between preseason and postseason FC in a subset of ROIs that distinguished football and control participants according to significant postseason group effects. For Cohort 2, we examined whether the number of head impacts differentially influenced in-and off-season delta FC in a subset of ROIs that distinguished football and control participants according to significant in-season delta group effects. ROI Z-scores were decomposed based on the number of impacts using a two-way linear mixed model, with time and participants as the within-and between-subjects factors, respectively. Age and BMI were included in the model as covariates.
Separately for each cohort and impact allocation, the main effects of time across ROIs were considered significant at p < .05 FDRcorrected.
T A B L E 1 Descriptive statistics of head impact kinematics

| Head impact frequency differentially influences connectivity outcomes
A linear mixed model was used to examine whether a high and low total number of cumulative head impacts over the course of the season differentially influenced FC outcomes. In the first analysis F I G U R E 3 Cohort 1 group-level voxel-wise differences. Axial slice representation demonstrating significant postseason group-level voxel-wise differences in functional connectivity of the default mode network between control and Cohort 1 football participants. Group averaged Z-score correlation maps in control and football participants are displayed in the left and center panels, respectively, whereas significant group-level t-statistic maps between control and football participants are displayed in the right panel. Red and blue voxels in the average Z-score correlation maps demonstrate a significant positive and negative correlation with the posterior cingulate cortex seed, respectively, whereas the blue voxels in the t-statistic maps represent significantly reduced functional connectivity in football players compared with controls. All maps are thresholded at p < .05 FWER-corrected  Nakamura, Hillary, & Biswal, 2009). This pattern of hyperconnectivity has been hypothesized to reflect a compensatory mechanism to increase neural resource utilization and re-establish networklevel communication following severe disruption (i.e., so-called "Hyperconnectivity Hypothesis" [Hillary & Grafman, 2017;Hillary et al., 2015]) and may underlie more complex pathophysiological and clinical features associated with TBI compared with subconcussive RHIs.
The underlying changes in DMN FC following subconcussive RHI exposure in collision-sport athletes with respect to a control group has not been well characterized. A study in collegiate female rugby F I G U R E 6 Head impact frequency and default mode connectivity. Grouped bar plots demonstrating the mean (± SEM) Z-score values for (a) preseason and postseason in Cohort 1 football participants and (b) in-and off-season delta in Cohort 2 football participants. For Cohort 1, the x-axis denotes a set of 10 ROIs that characterized football and control participants according to group-level effects (football vs. control) in the postseason ROI analysis, as well as the global default mode network (DMN). For Cohort 2, the x-axis denotes a set of seven ROIs that characterized football and control participants according to group-level effects (football vs. control) in the in-season delta FC ROI analysis, as well as the global DMN. The left and right subplots depict Z-scores obtained from players in the low and high impact frequency partitions for each football cohort, respectively. Asterisks represent a significant time main effect (p < .05 FDR-corrected) players demonstrated a DMN hyper-connectivity pattern with the PCC at both the start and 2-3 months following the season compared with non-collision sport controls (Manning et al., 2020). The respective hypo-versus hyper-connectivity pattern between our work and their study could be explained by several factors, including age, gender, and inter-sport differences in head impact dynamics. One study using a whole-brain correlation approach in high-school football players reported both an increase and decrease in the number of significant positive connections with the PCC/precuneus at several time points during the season compared with a baseline control value, including an increase in postseason connections (Abbas, Shenk, Poole, Breedlove, et al., 2015). While this study provides a perspective on the total number of regional correlations between the PCC and anatomical parcellations within and beyond the DMN, our work characterized group differences with respect to the relative strength of functional connections within a specific set of network hubs that consistently show a positive correlation with the PCC under task-negative conditions (Fox et al., 2005;Greicius et al., 2003).
While it may be unlikely that a single season of youth football RHI exposure leads to adverse long-term clinical outcomes observed for former career football players (McKee et al., 2009;Mez et al., 2017;Roberts et al., 2019;, it is also important to recognize the potential implications. Koerte et al. (2017) demonstrated that soccer players exposed to RHIs from ball heading did not improve in executive control performance during the season akin to their non-contact sport control counterparts. This was posited to reflect suppressed developmental benefit in athletes exposed to RHIs.
This finding is consistent with another study that demonstrated impaired oculomotor executive control function following mTBI in a college-aged cohort (Webb, Humphreys, & Heath, 2018). Functional abnormalities within the PCC and DMN are characteristic of a broad spectrum of neurological and psychiatric conditions (Buckner et al., 2008;Leech & Sharp, 2014;Zhang & Raichle, 2010). Abnormal DMN function has been shown to be related to cognitive impairment following neurotrauma. For example, Mayer et al. (2011) demonstrated that DMN FC predicted cognitive complaints during the subacute phase following a concussion. Moreover, attentional and information processing task performance has been shown to correlate with DMN FC following TBI Sharp et al., 2011). It remains unclear, however, whether subconcussive head impacts, independent of concussive head injury, represent a catalyst for long-term clinical impairment. Future, large-scale, multimodal, and longitudinal studies will be critical to determining the neuronal mechanisms underlying abnormal FC changes in youth tackle football players, as well as the potential long-term neurological outcomes.

| Longitudinal DMN connectivity changes in football players
A subsample of football players with off-season follow-up data (Cohort 2) provided the opportunity to examine whether in-season delta FC, representing a period of RHI exposure, differed from that of off-season delta FC, representing a 7-month withdrawal from football-related activity. Regions identified as displaying significant group-level in-season delta FC effects in the ROI analysis were decomposed longitudinally in Cohort 2 football participants (i.e., inseason vs. off-season). The results showed that in-season delta FC yielded a negative mean across all ROIs in football players, whereas an opposite compensatory effect was observed for off-season delta FC ( Figure 5).
The preseason (Abbas, Shenk, Poole, Breedlove, et al., 2015;. The positive change in off-season delta FC is in slight contrast to the work by , which reported a reduced number of whole-brain connections with the PCC at initial and 6-month postseason time points relative to preseason. This was interpreted to reflect a long-term repair process underlying mechanical stress of cumulative RHI exposure. This is consistent with prior work showing persistent long-term reduced DMN FC in collegiate football players (Zhu et al., 2015) and adults (Mayer et al., 2011) following a concussion. Our findings do not support a long-term outcome, as regions displaying a negative inseason change in FC encountered a positive rebound effect during the 7-month off-season interval. This discrepancy could be attributed to methodological differences.  examined the total number of regional connections between the PCC and whole-brain anatomical parcellations, whereas our work examined the relative FC strength of withinnetwork DMN architecture. An alternative view is that additional accrual of RHIs with increasing levels of play (Broglio et al., 2009;Broglio et al., 2011;Broglio, Surma, & Ashton-Miller, 2012;Daniel et al., 2012;Kelley et al., 2017;Urban et al., 2013;Urban et al., 2019) leads to more consequential long-term recovery outcomes. It is possible that reduced accumulation of head impacts over a shorter participation duration in youth players better supports off-season compensation of dysfunctional connectivity patterns. In support of this interpretation, a subsequent study in highschool players by  reported a deviation in DMN connections from control values even at preseason.  Koerte, Ertl-Wagner, Reiser, Zafonte, & Shenton, 2012;Manning et al., 2020;Mayinger et al., 2018). Diffusion MRI studies incorporating the HIT system have pointed to an association between the number and severity of subconcussive head impact exposure and abnormal brain microstructure in youth, high-school, and collegiate football players (Bahrami et al., 2016;Bazarian et al., 2014;Chun et al., 2015;Davenport et al., 2016;McAllister et al., 2014). Aside from acute changes, age at first exposure (≤ 12 years) to RHIs in professional football players may be associated with long-term damage of commissural white matter and more severe neurobehavioral impairment (Alosco et al., 2018;Stamm, Koerte, et al., 2015).  The results from this analysis provide initial evidence that the number of experienced head impacts at the youth football level may be a key contributing factor to abnormal DMN FC outcomes. It is possible, however, that other factors contributed to the FC disparity between low and high impact groups. As shown in Table 1, high impact players in Cohort 1 were also characterized by a significant increase in median peak resultant linear acceleration and RWE CP . A key aim for future work should include combined consideration for both the number and severity of head impacts on FC outcomes. Positional play is also associated with differences in head impact exposure, with fundamental tackle positions (i.e., linemen, linebackers) engaging in more frequent collisions compared with skilled ball-handling positions (Broglio et al., 2009;Crisco et al., 2010;Crisco et al., 2011).
Youth football players in the current study were characterized by vari- The results from this analysis emphasize the importance of limiting head impacts in youth football and support implementation of targeted rule interventions and policy changes to reduce the number of sustained head impacts in players (Kerr et al., 2015;Ocwieja et al., 2012). These results may also serve to inform parental decisionmaking regarding youth football participation. Many parents are in support of age-related restrictions on tackling (Chrisman et al., 2019) and have concerns about later-life neurological detriments stemming from their children sustaining football-related head injuries (Kroshus, Bowen, Opel, Chrisman, & Rivara, 2020). Certainly, this perspective is understandable given widespread media attention surrounding studies with overwhelming evidence of CTE in former professional football players . It is important to emphasize, however, that FC changes and associated short-term and long-term outcomes related to subconcussive RHIs, especially in youth collision sport athletes, are not yet fully understood. Future longitudinal studies will play an important role in shaping our perspectives regarding the social benefits and safety risks associated with youth football participation.

| Limitations
There are several important limitations of the current study. First, the current seed-based methodology does not account for possible widespread network-level responses to RHIs across other intrinsic RSNs previously shown to be affected by sport-related neurotrauma (Bharath et al., 2015;Mayer, Bellgowan, & Hanlon, 2015). Future studies implementing a more comprehensive exploratory approach (e.g., independent component analysis [e.g., Calhoun & Adalı, 2012]) are important to better characterize intrinsic functional architecture changes underlying RHI exposure in youth sport cohorts. Second, while the initial and follow-up scans for players coincided with the preseason and postseason for football players, the initial and followup scans did not always coincide with the preseason and postseason of the respective sport for controls. Although the average inter-scan MRI acquisition interval was consistent for both football and control groups (i.e., 4 months), the DMN connectivity profile of control participants could have been influenced by the absence of sport-related involvement between initial and follow-up scans. Third, the HIT System used for acquisition of head impact kinematic data is associated with some individual impact detection and acceleration measurement error. However, this device limitation has been shown to be comparable with other head impact devices and well within an acceptable range of error . A fourth limitation is the arbitrary grouping of high-and low-impact groups while examining the effects of head impact frequency on DMN FC outcomes. For Cohort 1, the upper and lower 30th percentile of the head impact frequency distribution was used to separate players, while players falling between those cutoff points were removed. This procedure was used to provide adequate separation of players corresponding to the head impact frequency distribution, as well as an adequate sample size within each impact partition. For Cohort 2, the 50th percentile was used in order to limit the exclusion of players from an already small pool of participants. An important step for future work may include a sensitivity analysis to determine the number of head impacts required to induce network-level changes in FC. Lastly, it is important to emphasize the exploratory nature of the analyses for Cohort 2 due to its relatively small sample size.

| CONCLUSION
This study represents initial evidence in a high-risk youth collisionsport cohort that a single season of subconcussive head impact exposure in the absence of concussion causes reduced network-level FC of widespread DMN regions compared with non-collision sport controls. In the longitudinal analysis, in-season delta FC was characterized by a negative directional change between preseason and postseason, whereas an opposite, compensatory effect was observed for offseason delta FC between postseason and follow-up. Lastly, the number of experienced head impacts in youth football players proved to be a key contributing factor to FC alterations. These findings extend evidence from other neuroimaging modalities and advance our understanding of the underlying pathophysiology characterizing subconcussive head impact exposure in youth football players.