Quantitative and qualitative analysis of head and body impacts in American 7v7 non-tackle football

In-game video recording

The study was designed after consulting with 7v7 coaches, players and parents about typical contact in the sport and their questions about player safety and appropriate protective equipment. Games were played by 12U, 14U and HS teams. The games took place between July 2018 and March 2019 at two sites: site 1, an outdoor American football field, and site 2, an indoor multipurpose playing field. Both of these sites had artificial turf. Qualitative impact data were collected at both sites but only data from site 1 were used for quantitative head impact reconstruction.

For site 1, the site was scanned with a 3-D colour laser scanner (LS120, Faro, USA). The laser scans were registered together using registration spheres and field markings in postprocessing to generate a 3-D point cloud of the stadium. This point cloud information was used for subsequent image calibration for head impact reconstruction. Fourteen stationary action cameras (Hero6, GoPro, USA) were mounted on tripods around the stadium such that any point on the field of play was within 30 m of at least two cameras. The cameras used 41° field of view (FOV) lenses and recorded at 2.7 K/120 fps with a shutter speed of 1/1920 s and a white balance of 5000 K. These parameters were selected following an internal validation study.22 For site 2, fifteen cameras were magnet mounted on I-beams around the field of play. The site was indoors so the camera frame rate was reduced to 4K/60 fps and used an automatic shutter speed to optimise the image exposure. At both sites an additional camera with a 120° FOV GoPro Hero6 lens with 4K resolution and 60 (site 1) or 30 (site 2) frame rate was placed to provide an overall view of the field. Camera times were synchronised via an external clock manually. There were two levels of synchronisation used. All cameras were first roughly synchronised to time of day using a clock. This synchronised the cameras to within approximately 1 s. After an impact was identified for tracking, the video was analysed frame by frame to align videos from multiple cameras. This synchronisation procedure was done by aligning a discrete event in the video, such as a hand or foot touching the ground.

Qualitative video analysis

The video analysis methods were based on previous work to evaluate the nature and frequency of head impact in contact sports.23 24 The overall camera was reviewed to identify plays with contact, and the game number, play number, play type, time of day and the approximate location on the field where the play of interest occurred were documented. The location on the field referenced the yardage line and the lateral position on the field. The play type was documented as a short pass (0–10 yards), medium pass (10–20 yards) or a long pass (>20 yards). The distance of the pass was identified as the distance the football travelled in the air from the quarterback to reaching the receiver.

Based on the field location of the contact, the play of interest was identified in three 41° FOV camera views. A 3 s video clip was extracted to assess the type of impact and the location of impact on the head (figure 1). If multiple contacts occurred, they were separated into first contact, second contact and third contact. Type of impact was classified as: body to body (B2B), body to ground (B2G), head to body (H2B), head to ground (H2G) and head to head (H2H). A worked example is provided in online supplementary material A.

Request permission

Quantitative video analysis (head impact reconstruction)

All plays involving head contact recorded at site 1 were targeted for video-based reconstruction of head kinematics at impact. Additionally, a set of plays in which offence and defensive players fell to the ground but no head contact occurred (B2G contact) were extracted for comparison. Model-based image matching was used to reconstruct head translational velocity (V) and rotational velocity (ω) over time, similar to previously published methods.25 26 In a laboratory validation of our model-based image matching implementation, we determined the mean absolute errors in the estimated change in resultant translational velocity and rotational velocity (ΔV_R and Δω_R, respectively) during simulated H2G and H2H impacts to be ±0.24 m/s (±10.7%) and ±3.4 rad/s (±21.8%), respectively.22

A worked example of the model-based image matching process is provided in online supplementary material A. For each impact, three separate camera views were identified and reviewed. The two primary (or best) views of the impact were used for reconstruction. Three-second video clips from the two primary views were extracted and uploaded into head tracking software (PFTrack, The Pixel Farm, UK) along with colour laser scan data of the stadium. Common points were selected in the video clips and laser scan data to align the camera views. Subsequently, the National Operating Committee on Standards for Athletic Equipment (NOCSAE) and Hybrid III headforms were fit to the head in both camera views in each video frame from approximately 150 ms preimpact to 150 ms postimpact or for 18 frames in the cases of no head impact (figure 1). Global head position, orientation and translational and rotational velocities were calculated from the positional tracking data. Transformation matrices were applied to transform the data to the local head coordinate system (HCS, figure 1). Rotational speed about the head x-axis and z-axis was reported as an absolute value, since biomechanically there would be no difference from a left/right z-axis rotation (ie, the motion that results in ‘No’) or a left/right x-axis rotation (ie, ear to shoulder motion). For the y-axis, flexion (−ve) and extension (+ve) were reported with their respective signs in place, because y-axis rotation can result in significantly different cervical spine kinematics. The positional and rotational data were verified by exporting them into CAD software (Polyworks, InnovMetric, Canada) while overlaying the laser scan data.

Several kinematical measures were calculated from the velocity data. The maximum preimpact translational velocity in the field coordinate system and in the XY plane (ie, field plane) were calculated as the maximal velocity in the 100 ms prior to impact or the maximal in cases with no head impact. The maximal rotational speed was reported as the maximal speed in the entire tracking sequence. The change in translational and rotational velocities due to impact (ΔV and Δω, respectively) were calculated by analysing the velocity curves and video to assess difference in speed between the time of contact and the end of contact. Contact was defined through frame-by-frame review of the video. ΔV and Δω were calculated along each axis, and the sum of the squares was taken to represent ΔV_R and Δω_R. The angle of the head velocity vector relative to the XY plane in the HCS during the preimpact motion of the head was also calculated (α).

Quantitative video analysis

Contact type

B2G and B2B were the most common contact types across all ages (figure 2). When the first contact was B2G, it typically involved a player falling to the ground after diving for a ball or tripping. When B2B was the first contact, the play was typically two players attempting to catch a pass resulting in B2B contact with one or both players falling to the ground subsequently. Second contacts were mostly B2G followed by H2G. Thirty-nine cases involved a third contact, which consisted of 28 H2G, 9 B2G and 2 H2B. No H2H contacts were identified. The small sample size for 12U and 14U precludes statistical analysis by age group. Qualitatively, the results were similar across age groups.

Request permission

Figure 2

Distribution of contact types by first, second and third contacts of the play for high school (HS; left) and under 12–under 14 (12U–14U) combined (right). Head contacts typically occurred on the second or third contact of the impact play, and the majority were head to ground (H2G). No head-to-head (H2H) contacts were detected across the three age groups. B2B, body-to-body contact; B2G, body-to-ground contact; H2B, head-to-body contact; H2G, head-to-ground contact; H2H, head-to-head contact.

Head contact

Head contact occurred on 4.8% of plays, resulting in a head contact incidence of 3.5 per 1000 athlete-plays (table 1). The majority of head contacts were H2G (75.0%), followed by H2B (25.0%). Head contacts typically occurred on the second or third contact of the impact play. For H2G, the majority of head contacts involved the rear upper head region (occiput) or the side upper head region (temporal/parietal) striking the ground and always occurred as the second or third contact. For H2B, impact locations were more distributed across the head but still favoured the side and rear locations (figure 3). Qualitatively, contact location on the head was similar in 12U and 14U compared with HS.

Request permission

Figure 3

Distribution of head impact locations for head-to-body (H2B) and head-to-ground (H2G) impacts. Values are summed across all recorded games. Trends were similar across age groups.

Head impact reconstruction

Eighteen plays were extracted for video reconstruction (H2G: 9, B2G: 4, running: 5). Online supplementary material B provides data for each examined impact. Across the 13 plays involving H2G and B2G impacts, the maximal preimpact velocity of the player’s head in the ground (XY) plane was 5.0±1.8 m/s, with a downward vertical component of 3.7±1.0 m/s. Evaluating H2G in isolation produced similar results, with a maximum preimpact velocity in the ground plane of 5.0±2.0 m/s and a downward vertical component of 3.7±1.1 m/s (table 2). The velocity vector of the head relative to the HCS XY plane (α) was calculated to be −47.6±26.8°. Based on velocity of players running, the typical game speed was 7.4 m/s±1.2 m/s in the ground plane. The head ΔV_R for H2G cases was 3.0±1.1 m/s with a median of 3.0 m/s and a range of 1.5–4.9 m/s over a Δt of approximately 0.06 s (table 3). The head Δω_R was 23.4 rad/s±10.8 rad/s with a median of 21.8 rad/s and a range of 10.1–43.3 rad/s.

Head contact incidence

Overall the frequency of head contact reported herein is lower than the rates for youth tackle football, even accounting for the longer duration of tackle football games.20 27–29 Differences in impact-recording methodology and definition of AE preclude direct comparison, particularly since sensor-based impact counts tend to overestimate impacts compared with video analysis.24 30 31 However, conservative estimates would suggest approximately two head impacts per player per 10 game-minutes20 29 32 compared with 0.1 head impacts per player per 10 game-minutes for non-tackle in the present study. Head contact incidence in 7v7 non-tackle appears to be less than non-helmeted contact sports, such as rugby, Australian Rules Football (ARF) and international football. Conservative estimates for participants in these sports suggest 0.2–0.7 head impacts per player per 10 game-minutes.33–39

Mechanism of head contact

The mechanism of head impact was different from other contact sports. In international football, one of the most common sources of injurious head impacts was found to be another player’s head,40 which has led to rule changes and the development of soft protective headgear.41 42 Likewise in rugby, H2H and head-to-shoulder contact during tackle events were the most common sources of head injury.13 15 43 44 Similarly, reducing helmet-to-helmet contact in American tackle football has been a key target of injury prevention efforts until recently.23 45 No H2H impacts were identified in this study.

The most common source of head contact was impact with the ground, typically in a multicontact sequence of B2G-H2G or B2B-B2G-H2G. The player first made contact with their body to the ground, and as their body decelerated on impact with the ground their head followed to make contact. Attempting to catch a pass was the most common scenario producing H2G, often with the defender and receiver making contact when challenging for the ball, causing a fall to the ground. H2B was another notable source of head impacts. Unlike H2G, several H2B contacts occurred on the first contact of the play. In most cases, as with H2G, the defender and the receiver were challenging in the air for a pass and collided.

Head contact location

The distribution of head impacts in non-tackle football favoured the side (44%) and rear (44%), with very few to the front (1%), face (10%) or top (1%). In contrast, tackle football players experience the highest percentage of impacts to the front of the head (33%–52%), followed by smaller and relatively even distribution across the rear (18%–40%), side (13%–19%) and top (10%–15%).27 46–48 Rugby and ARF players have been found to experience a relatively heavier distribution to the front of the head and a lower distribution to the rear of the head than 7v7.33 49 After excluding headers, international football players experience more front and top of head hits and fewer rear of head impacts than 7v7.50 The discrepancy in head impact distribution is likely due to the unique nature of each sport. Whereas the other football disciplines involve tackling, which is associated with high percentage of head impacts, 7v7 does not involve tackling, and instead, head impact results from colliding with another player’s body or falling to the ground when attempting to catch a pass. The unique distribution of head impacts in non-tackle football highlights the need for sport-specific head protection and likewise sport-specific headgear certification standards. For example, these data indicate a certification standard designed for American tackle football would not be applicable to non-tackle 7v7 football.

Head impact speed and energy

Head velocities during nine H2G impacts and four B2G impacts were estimated with model-based image matching.25 26 Our internal validation study determined similar errors to previous work.26 51 52 For H2G impacts, maximum preimpact translational and rotational head velocities averaged 5.9±2.2 m/s and 21.5±9.2 rad/s, respectively. Kent et al.53 reported higher V_R (mean: 8.3 m/s) and lower ω_R(mean: 13.5 rad/s) for H2G impacts associated with concussion in professional American Football (National Football League, NFL). The direction of the velocity vector (mean: -41.2°) was similar to the H2G impacts in the present study. Similarly, Pellman and colleagues51 reported higher head impact closing velocities for NFL head impacts resulting in head injury as well as non-injurious impacts. In international football, head contusions resulting from H2H impacts were associated with closing speeds of 1.3–2.5 m/s, while head injuries from elbow-to-head game impacts occurred at speeds ranging from 1.0 to 5.3 m/s (3.0±1.7 m/s).52 Examining ΔV_R, the 7v7 H2G impacts were associated with a ΔV_R similar to the one non-injurious NFL H2G impact (2.9 m/s) and notably less than injurious head impacts.51 53 In contrast, Δω_R was similar between the NFL injurious head impacts and the head impacts recorded in the current study. This raises the question of whether, by itself, Δω_R is a biomechanical predictor of injury as there were no head injuries noted on these plays.

Limitations

The study is limited by the relatively small sample size, particularly for 12U and 14U players, which precluded statistical comparisons across age groups. This study was not designed to evaluate injury epidemiology, and the majority of head contacts evaluated in this study were not associated with injury. At site 1 there were no injuries that required medical attention to our knowledge. At site 2, we observed two definitive injuries (clavicle fracture and a lower limb injury) and two instances where a player struck their head (one H2G, one H2B) and appeared shaken but returned to play on the next set. The lack of H2H contacts and the relatively low-peak impact severities (ΔV_R) would suggest that non-tackle football incurs a lower risk of head injury compared with other contact sports, confirming prior epidemiological studies.3 6 8 However, future work should incorporate systematic medical reporting to identify the risk of head injury and the associated injurious head biomechanics.