A novel repetitive head impact exposure measurement tool differentiates player position in National Football League

American-style football participation poses a high risk of repetitive head impact (RHI) exposure leading to acute and chronic brain injury. The complex nature of symptom expression, human predisposition, and neurological consequences of RHI limits our understanding of what constitutes as an injurious impact affecting the integrity of brain tissue. Video footage of professional football games was reviewed and documentation made of all head contact. Frequency of impact, tissue strain magnitude, and time interval between impacts was used to quantify RHI exposure, specific to player field position. Differences in exposure characteristics were found between eight different positions; where three unique profiles can be observed. Exposure profiles provide interpretation of the relationship between the traumatic event(s) and how tissue injury is manifested and expressed. This study illustrates and captures an objective measurement of RHI on the field, a critical component in guiding public policy and guidelines for managing exposure.

Characteristics of RHI exposure. Frequency of head impacts. A total of 3439 (2941 confirmed; 498 suspected) head impacts were documented for eight player positions during 32 regular season games (Table 1). Frequencies are presented per position based on total frequency, average per game, and estimated frequency for a 16-regular game season in professional ASF (excludes practices). The Kruskal-Wallis H test showed strong evidence of a statistically significant difference in impact frequency between the mean ranks of player position, χ 2 (7) = 187.21, p = 0.000: QB mean rank = 43.50; RB mean rank = 150.09; WR mean rank = 56.53; TE mean rank = 170.64; OL mean rank = 200.89; DL mean rank = 213.44; LB mean rank = 134.02; DB mean rank = 58.89. Dunn's pairwise showed specifically that differences were between QB, WR, DB, and the remaining five positions (Table 1). Additionally, significant differences were found between LB and OL (p = 0.008); LB and DL (p = 0.000); and RB and DL (p = 0.017). Independent of position, the average number of impacts a player received during game play over a 16-game season was estimated at approximately 184 impacts. Impact frequencies are estimated for an average professional football career (ave. yrs.) and for the greatest number of games played (># games) documented in NFL player history (Table 1; Pro-Football-Reference.com, 2014; 2018). Suspected and multiple (pile-up) impacts are presented for the eight positions (Table 1). Suspected impacts accounted for approx. 15% of the total impacts recorded; RB with the greatest number recorded, followed by LB and DL. RB entered a pile-up twice as many times as the closest second (DL).This results in a conservative estimate of overall frequencies.
With player position collapsed, helmet represent 49% of the total documented impacts. Shoulder impacts were the second most common event type with 23% of the total. The remaining two event categories accounted for the fewest documented impacts (ground = 13%, hip/thigh = 2%). 'Other' events described 13% of impacts (back = 3.4%, chest = 1.7%, stomach = 1%, arm = 2.5%, hand = 2.5%, leg = 1.2%, knee = 0.6%, field goal post/ wall = 0.1%). Most often LB, TE, OL and DL experience helmet impacts, and these were predominantly to the front location (Fig. 1). Impacts with the ground are more frequent among the QB, WR and DB positions, accompanied by a noticeable shift to include the side and rear head locations. An even distribution between shoulder and ground event types can be observed for the RB. This position also experienced recurrent head impacts to their opponents' torso represented in the 'other' category ( Fig. 1). Side location was greatest for the RB player position.
Magnitude of brain tissue maximum principal strain (MPS). The maximum principal strain (MPS) resulting from exemplar impacts for eight ASF field positions ranged from 4.6-50.6%. The mean MPS resulting from exemplar reconstructions of each event type individually and collapsed for each ASF position is presented in Table 2. MPS resulting from all exemplar impacts was further categorized into five levels of magnitude: very low = <8%, low = 8-16.9%, moderate = 17-25.9%, high = 26-34.9%, very high = 35%+.

Time interval between head impacts.
Time between impacts is presented as averages per game, both as an overall time and distributed by MPS magnitude level (  www.nature.com/scientificreports www.nature.com/scientificreports/ longer time intervals (Fig. 3). BSE/T documented for LB and RB, (profile 2) consist of all levels of MPS magnitude of mid-range impact frequency and intervals. Finally, OL and DL (profile 3) positions experience the highest frequency of head impacts within short time intervals, predominately of low strain magnitude (Fig. 3). The TE position is unique in that the cumulative BSE/T was not significantly different to either field positions in profile 2 or profile 3, resulting in an overlap between the two profiles.

Discussion
The complex nature of head injury and one's history, predisposition, symptom expression, and RHI exposure results in an uncertainty on how to manage trauma to the brain. Focusing on one head impact metric to predict injury has proven challenging and typically results in low injury prediction sensitivity, particularly when using impact sensors to measure biomechanical metrics 43,48 . This research has employed a novel method of objectively quantifying RHI that captures a spectrum of severity in ASF. Difference in the characteristics of RHI between player field positions in National Football League was described.
This study yielded lower impact frequencies over a season in comparison to previous reports using impact sensors 36,38 , which include impacts sustained during practices. However, over the course of a 12-game season an average of 128 impacts per player was reported in collegiate football, comparable to the present findings 49 . Average impact frequency per game ranged from 2.3-22.1 depending on position, which are lower than previously reported excluding individual frequencies for RB, TE, and OL [36][37][38] . The higher frequencies documented for TE and RB can perhaps be attributed to their multiple roles on the field, and may also describe the lower frequencies reported for WR in comparison to Crisco and associates 37,38 . Their studies do not include frequency data for the TE position, and therefore may have been considered either a receiver or a lineman, thereby increasing the count for receivers. Most notable are the lower count estimations for the QB and DB. This discrepancy may be explained in a number of ways: as missed impacts from being limited to those within camera view, the tendency of impact sensors to over report hit counts, or differences in play between collegiate, high school, and professional level football. The OL and DL positions sustained the highest number of impacts throughout a game; predominantly from collisions to the top/front of the head (Fig. 1) which has been associated with detectable changes in the dorsolateral prefrontal cortex after sub-concussive head impacts 11 .
Based on the 32 game analysis, the average NFL football athlete may experience upwards of 1000 impacts from game play throughout their career, with closer to 6000 head impacts estimated for the greatest number of games played on record (Table 1). This estimation merely accounts for professional level game play and not the history of the athlete, where length of career and age of exposure influence risk 15,[20][21][22] , and may, in addition to genetic and environmental factors, partly explain why not all former ASF athletes show signs of long-term clinical dysfunction 17,20,50 .
Reports identifying risk associated with RHI are limited in describing tissue deformation associated with common impacts experienced during a game 11,13,30,51 . Impact magnitudes obtained from the use of helmet impact sensors consistently report averages in the vicinity of 20 to 30 g and 1500 to 2000 rad/s 2 for linear and rotational  www.nature.com/scientificreports www.nature.com/scientificreports/ head accelerations, respectively 36,38,39,41,49,52 . These averages are found across all players, providing insufficient information on magnitude differences 36,[39][40][41]52 . Further, peak head accelerations alone provide a less stable measure of impact severity than the brain tissue response 44,45 . Given that serum biomarker levels of axonal injury associate with higher hours of contact 8 , a better understanding of what magnitudes of tissue strain associate with game play provides a useful analysis for risk management strategies. Few differences in strain magnitudes were found between positions based on exemplar impacts in the current study, indicating that all positions may be exposed to a similar range in event severities. How the frequency of events distributed by strain magnitude provided more distinction between positions. QB, WR, and DB experienced the lowest impact frequencies but were among the positions most likely to sustain impacts of higher MPS magnitudes of above 26%, largely from head to ground contact (Fig. 1), and are among those reported as most vulnerable to sustaining 'extreme' hits 39 and concussions 53 . Close to 95% of impacts received by OL, DL and TE were estimated at below 17% MPS and under 50% probability of concussion risk ( Table 2; Fig. 2) 46,47 . Although these strain magnitudes are considered relatively low, repetitive low magnitude impact frequencies may exhibit structural and/or functional changes that are associated with long-term mental health concerns reported by current and retired football athletes 14,18,[54][55][56][57] . Further, as little as 5-15% strain levels have been associated with functional impairment of signal transmission in vitro [58][59][60][61][62] .
Few impacts were recorded below 8% MPS across all positions. Pile-up situations presumably would consist of this level of impact although the nature of this scenario is such that impacts could not be confirmed and counted (Table 1). In vitro strains of 5% exhibit the minimum level of injury required to observe minor undulations and induce a calcium influx 61 . Observed undulations are caused by the immediate breakage and buckling of microtubules that are a consequence of mechanical failure 63,64 . Ahmadzadeh and researchers 65 have also modeled macroscopic strains of 5% causing microscopic changes in the form of protein unfolding and microtubule rupture, a consistent neuropathology of tauopathy and brain diseases 20,66,67 . All impacts documented in this investigation were at or above these strain levels.
Time interval between impacts showed an inverse relationship with impact frequency (Fig. 3), which was an expected result. Impacts were documented on average every 6-30 minutes during game play in professional ASF. Oliver and researchers 8 have demonstrated that elevations in biomarkers indicative of neuronal damage consistently increase at various time points throughout a season in ASF starter athletes (20-40 + plays/game), a population predominantly represented (starters) in the current study. Moreover, these measurements indicate that neurobiological recovery continues even with cessation of activity and RHI 6,8,14 . The relationship between interval and frequency becomes less predictable when examined individually at each magnitude level (Table 3). Although limited in sample size, the shortest time intervals between impacts of high strain magnitudes (>26% MPS) were reported for QB and TE. The importance of interval between high magnitude impacts has been described as an inconsistency between clinical symptom recuperation and neurobiological recovery 6,68 . Although time intervals are much longer in these reports, the very short intervals recorded here could influence the cumulative effect of sub-concussive impacts. Neurobiological damage may be present in the absence of clinical symptoms, and without a sufficient recovery there is a risk of cumulative injury 6,8 . The density to which sub-concussive impacts are experienced in succession may also collectively lead to concussion 35 .
The present study identified three profiles based on the combined characteristics of brain trauma exposure, BSE/T. Profile 1 identified positions that were exposed to low impact frequency of proportionately higher strain magnitudes above 17% MPS, experienced within 25-30 min time intervals. Player positions that are susceptible to a more even distribution of strain magnitudes at moderate frequencies within on average 13 min time intervals was identified as profile 2. Profile 3 was identified as positions that sustained the highest impact frequencies of www.nature.com/scientificreports www.nature.com/scientificreports/ magnitudes predominately below 17% MPS on average every 6-7 min time intervals. Positions that are recognized as more vulnerable to high magnitude events, such as DB, WR and QB, report the highest incidence of concussions 23 . Although lowest in frequency count, over 90% of the impacts documented for the QB were above moderate strain magnitudes, putting them above the 50% probability of concussion, essentially each time they are hit 46,47 . Sustaining a concussion does not inevitably lead to long-term mental disorders, nor does experiencing a greater number of concussive events consistently result in more severe cognitive decline among athletes 20 . This type of exposure, very few hits of high strains, may speak to the exhibition of immediate signs of acute injury, however coupled with sufficient recovery time could be less perilous for long-term damage 69 .
Long-term exposure to RHI associates with the inability to recover from injury, mental health disorders and, in most severe cases, degenerative brain disease years after exposure 14,20,70,71 . Consequently, positions identified in profile 3 are among the highest for confirmed cases of CTE 25 . Profile 2, perhaps the most dangerous, seems to have a less predictable brain trauma exposure, as they are exposed to relatively high impact frequency and are vulnerable to high impact magnitudes.
The TE position showed an overlap between the profile 2 and profile 3, possibly owing to their multiple roles on the field, and thereby successfully identifying TE as both a lineman and a short distance receiver. The RB position too, is unique in that they represent those players responsible for both running and passing plays, and is reflected in their exposure profile. With a low number of players on the field in a TE and/or RB position at any given time, they are considered among the riskiest positions for concussion and continuing neurological damage 23,25 . In light of the current distribution among position of various forms of diagnosed head injury, one can appreciate how an exposure profile may assist in the understanding of how tissue injury is manifested and expressed.
Few studies have examined RHI as a cumulative exposure metric based on multiple characteristics. These metrics are derived from self-reported history and/or head impact sensors, where typically higher exposure loads correlate with objective clinical measures 4,5,42,72 . The current study enhances our understanding of risks associated with RHI by providing tissue deformation magnitudes for a spectrum of common head impacts, giving contextual meaning to the cumulative strain experienced on the field. In practice, the RHI exposure profiles identified here can be used to estimate individualized exposures based on their position played, percentage of play, and number of games/seasons played. Exposure indices derived from this research can be used in comparative analyses involving current and former athletes, particularly useful in those with higher levels of play, presenting with neurological and cognitive health-related outcomes. As advancements in measurement techniques are becoming more refined and sensitive to all levels of head impacts, an objective method of measuring and quantifying RHI exposure is required to interpret clinical outcomes, guide legislation and intervention strategies and provide league officials, athletic directors, coaches and athletes information to make decisions concerning RHI in ASF.
Clinical outcomes, neither from individual events nor multiple events were documented in the current study, however, the range in impact frequency, strain magnitude and time between impacts demonstrates that RHI can be described using various combinations of exposure characteristics. There is agreement among scientists, neurologists and clinicians of the necessity to understand the relationship between the biomechanics of RHI, with neuroimaging, fluid biomarkers, genotype, symptom expression and altered behavior and cognition. One can appreciate that the population studied here may be at risk for a full spectrum of injury, regardless of field position, based on the intrinsic nature of the sport. The results, however, provide insight of the possible relationships between the characteristics of brain trauma exposure and specific outcomes. This measurement tool was effective at identifying the subtleties between player field positions and supports the notion that RHI can be described using various combinations of exposure characteristics. The brain's many physiological responses under impact loading reflect the complexity of injury mechanisms 8,14,34,73,74 . Without an understanding of the event(s), there is little reference for change. Future applications using this method will be instrumental in connecting cumulative impact events to physiological response and injury outcomes. Connecting the characteristics of exposure to specific neuronal damage and neurobiological responses is required to estimate risk, establish dose-response relationships, and management of RHI. Establishing indices of exposure based on the interaction of impact frequency, strain magnitude and time interval will provide a measurement tool that, when applied universally, grants an opportunity to make informed decisions regarding risk mitigation strategies and public health policy for mental health protection and injury prevention for a myriad of sporting environments.

Materials and Methods
Game video analysis. Films of 32 professional ASF games from the 2009 to 2015 seasons were reviewed.
One regular season game played by each of the 32 league teams was chosen at random for analysis. One starter player in each of eight positions was tracked and all head impacts were documented. When a tracked player was taken off the field, the substitution player was tracked for the same position. Each position was tracked throughout the entire game. Initial analysis involved Reviewer 1 clipping video of all head impacts that occurred and documenting the team name, position played, player name and number, type of event, time of impact, head location of impact, and estimated velocity level of impact based on recognized averages of jogging, running and sprinting speeds 75 . The impact confirmation process then consisted of every impact clip viewed by Reviewer 2, then Reviewer 3 to designate impacts as either confirmed or suspected (or multiple suspected), and discarding those that did not meet either definition. The final Reviewer 4 ensured that all documented confirmed impacts were consistent with the inclusion criteria. Velocity level was later confirmed using video analysis software (see Physical reconstruction of head impact events).
Head impact frequency counts. Impacts were designated as confirmed, suspected, or multiple. Impacts were logged as confirmed if: contact with the head was clearly visible; the event type could be determined; the head Scientific RepoRtS | (2020) 10:1200 | https://doi.org/10.1038/s41598-019-54874-9 www.nature.com/scientificreports www.nature.com/scientificreports/ location could be determined. Suspected impacts were treated as events in which the player being tracked appeared to have experienced a head impact; however, due to any of the following reasons an impact could not be confirmed: poor video quality, poor camera angle of view, point of contact was obscured by another player. Multiple impacts were documented for circumstances in which players were tackling or being tackled and entered a pile-up situation, suggesting that multiple impacts were suspected; however, the specifics of those events could neither be distinguished nor confirmed.
Time interval between head impacts. Clock time of impact was recorded and used to calculate the time intervals. Total game time and time of play was used to calculate the percentage of stop time for all games independently. Interval between every impact was multiplied by the stop time specific to each game. Intervals were calculated in minutes. A two-minute interval was added between the last impact in the first quarter and the first impact in the second quarter, similarly between the third and fourth quarter. A 12-minute interval was added between impacts occurring on either side of half time. Two games went to overtime where a 5-minute interval was added between the end of the fourth and beginning of the overtime. Time intervals between each impact within level of magnitude were calculated. Impacts were further categorized into four time intervals for the calculation of BSE/T and are as follows; very low = <15 min, low = 15-30 min, moderate = 31-90 min, high = 91 min+.
Physical reconstruction of head impact events. Four primary event types were reconstructed in laboratory; helmet, shoulder, hip/thigh and ground. All other events were documented as 'other' . To appropriately reconstruct an event, the following five parameters were determined and categorized: velocity, location, orientation, effective mass, and compliance. These parameters create unique head kinematic motions that subsequently influence the magnitude of neural tissue loading and physical responses 46,47,[76][77][78][79][80][81][82][83][84][85] . Inbound velocity, and head location and orientation were obtained during video analysis; effective mass and compliance were approximated using known measurements and available literature [86][87][88] .
Velocity, location, orientation. Inbound velocity was calculated by establishing the distance separating the player's head from the contact surface (i.e. opponent shoulder) three to five frames prior to the impact using Kinovea software (version 0.8.20). Collision velocity (helmet, shoulder, hip/thigh) was calculated using field lines and video recording speed as described in a number of earlier studies 89,90 . Fall (ground) velocities were estimated by calculating the resultant velocity in a two-step process. Horizontal velocities were calculated using a marker system and perspective grid of known dimensions. The coordinates of the marker (representing camera movement) and the coordinates of the player's head (representing camera + player movement) were documented for three frames prior to contact and velocity was calculated based on displacement measurements. The calculated velocity of the marker (camera) was subtracted from the player (camera + player) velocity for each frame and an average velocity was calculated. Vertical distances were established by measuring the distance from the player's head to the ground two to three frames prior to contact using the known helmet width or length measurement as a reference 91  Head locations were categorized as front, front boss, side, rear boss, rear and crown and were labelled using the reference system illustrated in Fig. 4. In the transverse plane, the head was divided into 8 sectors of 45°, representing 5 head locations, with eyes forward being 0°. All impacts occurring at the top of the helmet were categorized as crown (Fig. 4a-1). Side and boss locations were treated as one location independent of occurring on either the right or left side of the head. In the sagittal plane, the distance between the base and top of the helmet was divided into five equal levels of elevation (minus crown location) (Fig. 4a-2). Similar methods used and outlined by Rousseau 89 deemed the level of precision was consistent with previous investigations. The orientation of impact was estimated within a tolerance of 15° as smaller increments have little effect on headform dynamic response 92,93 . Location and direction were verified using a high speed imaging PCI-512 Fastcam running at 2 kHz and Photron Motion Tools computer software (Photron, San Diego CA) (Fig. 4b).
Compliance, mass. Compliance and mass were achieved using appropriate reconstruction equipment and selected based on the type of impact event. A pneumatic linear impactor (collisions) and a drop rig (falls) were used to replicate the four event types. The linear impactor consisted of a stationary steel frame secured to a cement floor supporting a 1.28 ± 0.01 m long cylindrical, free-moving impactor arm (13.1 ± 0.1 kg) (Fig. 5). A Hybrid III 50 th -percentile adult male head form and neutral unbiased neck form 94 (mass 6.65 kg ± 0.01 kg) was attached to a 12.78 ± 0.01 kg sliding table to allow for post impact movement. The inbound velocity matched the values obtained from the categorization of impacts. Shoulder and hip/thigh events utilized a nylon disc (diameter 13.2 mm) covered with a 142 mm thick layer of vinyl nitrile 602 foam attached to the impacting end of the arm to simulate human shoulder and hip compliance 88,89 (Fig. 5a,b). Additionally, a pro football shoulder pad was attached to the impacting end for shoulder collisions (Fig. 5a). The striking mass was 15.5 kg and 15.2 kg for shoulder and hip/thigh events respectively, representing a realistic effective mass of player-to-player collisions 87,88 . Helmet events employed an additional Hybrid III head and neck form attached to the impacting end of the cylindrical arm via an L-frame and angled wedge to represent the striking player (Fig. 6c). Four different angled wedges (15°, 30°, 45° and 60°) were used to obtain the desired neck angle of the striking player. A heavier free-moving cylindrical impacting arm (16.1 ± 0.1 kg) and a stiff Hybrid III neck form were used for these collisions to represent the higher mass and minimal neck bending of a striking player 95 for a total striking mass of 24.0 kg. A monorail drop tower with turf anvil was be used to simulate the ground events (Fig. 5d) www.nature.com/scientificreports www.nature.com/scientificreports/ assembly was dropped unrestrained from a predetermined height onto a turf surface to duplicate the contact surface of a football field.
An exemplar event was chosen for reconstruction for every permutation in impact condition (event type X velocity level X head location) experienced by each of the eight positions. This resulted in 249 head impact event exemplars. Table 4 presents the exemplar event type classification for each field position. All event reconstructions consisted of three impact trials for a total of 747 simulated impacts. Impacts that met the following criteria were chosen as exemplar events in this study: the event was captured on video, multiple lines were visible in a plane of view of the camera to determine plane perspective using distortion correction algorithms, and the head location could clearly be determined. If a condition consisted of two or three impacts than an average of the calculated velocities was used for the reconstruction. The midrange value of a velocity level (i.e. low collision = 3.25 m/s) was used for reconstructions in which the exemplar represented a category consisting of four or more impacts. Finally, for exemplars representing only one impact condition, the calculated velocity for that event was used for the physical reconstruction.
Collection system. Nine single-axis Endevco 7264C-2KTZ-2-300 accelerometers (Endevco, San Juan Capistrano CA) mounted in a 3-2-2-2 array in the headform were sampled at 20 kHz and filtered using the SAE J211 class 1000 protocol 96,97 . A SLICE free motion headform was outfitted with three single-axis accelerometers and 3 angular rate sensors fixed near the headform center of gravity. The accelerometer signals were passed through a TDAS Pro Lab system (DTS, Calabasas CA) before being processed by TDAS software. An electronic time gate measured the inbound velocity just prior to head impact using National Instrument's VI Logger software.
finite element brain modelling. The University College Dublin Brain Trauma Model (UCDTBM) was used to calculate the brain deformations resulting from the exemplar head impacts. The UCDBTM model geometry was developed from CT and MRI imaging of the head of a male cadaver, and the version used in this research was composed of the scalp, skull, pia, falx, tentorium, CSF, grey and white matter, cerebellum, and the brain stem. Overall, the brain was composed of approximately 26,000 hexahedral elements 98,99 . Model material characteristics were established from earlier research [100][101][102][103] , presented in Table 5, modelled using a linearly viscoelastic model combined with large deformation theory. The behaviour of the tissue is characterized as viscoelastic in shear with a deviatoric stress rate dependent on the shear relaxation modulus, and defined as elastic in compression. The interaction between the CSF and the brain represents the brain/skull interface and is modeled with solid elements with a high bulk modulus and low shear modulus to allow for sliding behavior similar to fluids. This interface uses a friction coefficient of 0.2 and the contact interaction specified no separation 104 . The shear modulus characteristics of the brain tissue, modeled as viscoelastic is given by Eq. (1): www.nature.com/scientificreports www.nature.com/scientificreports/ where G ∞ , is the long term shear modulus, G 0 , is the short term shear modulus, and β is the decay factor 99 .
Validation of the model was conducted by comparing simulation responses to cadaveric testing that measured intracranial pressure 105 and relative brain skull motion 106 . Comparisons to real world reconstructions of traumatic brain injury incidents were also conducted as a further validation 107 .
Magnitude of impact. Impact magnitude was measured and characterized using MPS. This metric has been identified as being among the best predictors of severity as it provides a more complete analysis of the brain's mechanical response under loading that is difficult to capture using peak head acceleration alone 44,46,47 . MPS was calculated from exemplar head impact events using the three-dimensional head motion dynamic response curves obtained from the Hybrid III physical model reconstructions and applied to the centre of gravity of the finite element brain model 47 . Five MPS magnitude categories were established to capture a spectrum of severity that occurs in the sporting environment. These categories were based on biomechanical event reconstructive research and anatomical tissue analysis that associate with, physiological brain responses to axonal stretch, reported   www.nature.com/scientificreports www.nature.com/scientificreports/ concussion, sub-concussion and clinical outcomes. Moderate magnitudes represent a category based on a 50% risk of concussion injury measured in the cerebrum white or grey matter 46,47,59 . Strains below this range were considered very low 60,61,65,108,109 and low magnitudes 8,110 ; high magnitudes correspond to those reported as average estimates to sustaining a concussive level injury 89,111 . A very high category was included to designate impacts that put athletes at substantially higher risk and represent a 50% risk of loss of consciousness and persistent symptoms 112,113 . MPS magnitudes for each condition were categorized as follows: very low = <8%, low = 8-16.9%, moderate = 17-25.9%, high = 26-34.9%, very high = 35%+. Strain magnitude values resulting from exemplar head impact reconstructions were assigned to each head impact within their represented impact condition category, thereby assigning an MPS magnitude to each head impact within the four primary event types.
cumulative RHi exposure. An exposure index, BSE/T, utilizing the three characteristics was created to quantify cumulative RHI for each field position throughout each game. This index was calculated by multiplying the frequency of impacts within each brain tissue strain magnitude category by its severity # (very low = 1; low = 2; moderate = 3; high = 4; very high = 5), then multiplied by the time interval in which those impacts were experienced (very low = 4; low = 3; moderate = 2; high = 1). This resulted in five values (one for each magnitude category) which were added together and divided by the total impact frequency experienced throughout the game to account for the number of impacts that created the RHI exposure.
Statistical differences between American-style football player positions. Nonparametric rank based Kruskal-Wallis H tests were conducted to examine statistically significant effects of player field position on impact frequency, strain magnitude, time interval between impacts, and BSE/T. Impact frequency, strain magnitude, time interval and BSE/T variables were considered as dependent variables and eight player positions as independent variables. Frequencies, magnitudes, and intervals of confirmed head impacts resulting from the four primary event types were included in the analysis. Interval was included for games where >1 impact was documented within the same game: (QB, n = 24), (RB, n = 32), (WR, n = 26), (TE, n = 32), (OL, n = 32), (DL, n = 32), (LB, n = 31), (DB, n = 28). Post hoc Dunn adjusted by the Bonferroni correction for multiple tests were performed when significance was found. Additionally, a contingency table analysis was performed to examine the frequency distribution differences in strain magnitude categories amongst player field position. A Bonferroni correction was applied for multiple comparisons. Statistical level was set at α = 0.05 for all analysis and performed using IBM SPSS Statistics 24.0.

Data availability
Raw and processed physical and computational data, and game video clips are currently being stored at the University of Ottawa, Canada. It is available upon request.  Table 5. Brain tissue material properties and characteristics used for the UCDBTM components.