Introduction

The Centers for Disease Control and Prevention has concluded that mild traumatic brain injury (mTBI), also known as concussion, is at epidemic levels with an estimated 1.6–3.8 million cases occurring annually in the United States.7 Interest in understanding the pathophysiology of the injury, as well as potential short- and long-term effects, has never been greater. Recent studies suggest a link exists between concussion history and developing mild cognitive impairment, clinical depression, and early onset of Alzheimer’s disease,21,22 creating a public and scientific debate about effective prevention strategies for concussions. A 2007 epidemiological study conducted by the National Collegiate Athletic Association (NCAA) on American Football revealed that concussion accounted for 6.8% of all game injuries and 5.5% of all practice injuries, making it the third most common game injury and fourth most common practice injury.14 Given the high number of injuries per athlete exposure and frequency of plays resulting in blows to the head, researchers have been using the football field as a living laboratory to study brain injury since the 1960s.28,40 To better understand the etiology of concussion and the relationship between head impacts and injury, however, a large number of measured impacts from multiple athletes, both concussed and non-concussed, are required.

In order to quantify concussion biomechanics in football, the National Football League (NFL) commissioned a series of studies to quantify the characteristics of impacts associated with concussion diagnosis.33,37–39 Using video reconstruction techniques, Pellman et al. were able to determine the impact velocity for 31 of 182 cases of head impact (both with and without diagnosis of concussion) captured on film. By matching the obtained impact velocity and observed direction, these 31 cases were re-enacted in a laboratory using Hybrid III (HIII) anthropomorphic test devices (ATDs). Head linear and rotational accelerations were measured from accelerometers in the ATD’s head, permitting calculation of injury metrics such as Gadd Severity Index (GSI) and Head Injury Criterion (HIC).19,45 Due to practical limitations (e.g., complexity and cost of reconstruction), the dataset created by the NFL was biased toward injury events when compared with actual rates of injury per head impact sustained in play. Nonetheless, risk curves were generated from these data which have become commonly cited thresholds for concussion injury. In addition to the developed injury risk curves, several follow-up studies were also conducted to characterize the relationship between head impact measures and clinical outcome, including neuropsychological testing, symptomatology, and repeat injury.34–36

To assess the accuracy of acceleration measures derived from the NFL studies, Newman et al. quantified the error associated with reconstruction variables, including effects of noise in the data acquisition system, processing inaccuracy, and determining the impact velocity from multiple camera angles. Of these variables, deriving input parameters from video was identified as the primary source of error, with the estimated difference between actual and reconstructed head acceleration potentially reaching levels of up to 17% for peak linear acceleration and 25% for peak rotational acceleration.31 In an additional study to quantify the effect of head, neck, and torso coupling, Beusenberg et al. found that if a player’s neck coupling with the head (e.g., stiffness, strength, etc.) deviates from that of the HIII head and neck, peak linear acceleration can be altered by more than 15% and peak rotational acceleration will be drastically different.3 Due to the complexity of these reconstructions and the beneficial insight they provide, these errors were deemed to be within an acceptable level, and output measures from the reconstructions were considered reasonable for use in estimating the actual head kinematics for this group of impacts.31

Subsequently, the NFL commissioned Biokinetics (Ottawa, Canada) to develop a linear impactor test device to replicate head impacts to a HIII dummy head and neck at the impact velocities derived from the laboratory reconstructions.39 The linear impactor system has been proposed as a standardized helmet-testing device but has not yet been adopted by governing bodies or standards organizations.25,29 The benefit of a laboratory system for impacting helmets is to allow parametric evaluation of test parameters (e.g., helmet models, impact direction and energy, etc.) on biomechanical response (e.g., head acceleration, GSI, etc.) by pairing consistent test conditions with precision measurement devices and established mathematical techniques in a controlled environment. Similar to the full ATD reconstructions, though, the linear impactor test does not fully replicate on-field conditions as it standardizes the biomechanical response, which can affect biomechanical response to impact.25 It has also proven to be challenging to achieve measurement repeatability across multiple test laboratories because of potential differences in impactor surfaces and HIII head and neck properties.

Head Impact Telemetry (HIT) System technology (Simbex, Lebanon, NH; Sideline Response System, Riddell, Chicago, IL) records the frequency, location, and magnitude of impacts sustained by football players during live play,4,11,16,20,27,44 and was specifically developed to enable research to better understand the relationship between measured parameters of head kinematics and concussion. This technology incorporates an array of non-orthogonal accelerometers (Analog Devices, Inc., Cambridge, MA), data acquisition, and RF telemetry hardware into self-contained inserts placed in commercially available helmets. While physically connected to the helmet, the instrumented insert acts as an effective spring to maintain contact with the head during impact and to decouple head acceleration from helmet acceleration.26 Because traditional mathematical approaches to calculate linear and rotational acceleration at the head center of gravity, such as those employed by the HIII, require precise mounting of accelerometers that is not practical for field implementation, the HIT System uses proprietary, simulated-annealing optimization algorithms to estimate linear and rotational acceleration.8,10 While this approach has been shown to be mathematically robust, on-field variables that could potentially affect the measured and computed output variables (e.g., linear and rotational acceleration, impact location, impact duration, etc.) include relative motion between the head and the sensors and separation of the sensors off the head during the impact. The first on-field use of the system occurred during the 2003 football season,17 and, since then, the HIT System has been validated in both laboratory and on-field settings.6,10,17,18,26 To date, usage has spread to all levels of play (youth, high school, collegiate, and professional) which has generated a head impact database that now includes more than 1.5 million recorded head impacts.42

Given that data obtained from both the HIT System and the NFL studies are now commonly cited in the literature, it is important to understand the similarities and differences in output from these two approaches. To address this issue, the NFL’s Mild Traumatic Brain Injury Committee requested a laboratory comparison of the output from the HIT System with the Hybrid III ATD under conditions that mimicked impacts occurring in the NFL that are associated with and without diagnosed concussion. The purpose of this study was to correlate measures of head impact biomechanics recorded by both systems when tested on the linear impactor with impact velocities and impact locations similar to those identified in the NFL laboratory reconstructions.

Methods

Experimental Setup

The head of a 50th percentile HIII ATD (Humanetics, Plymouth MI) mounted to a standard HIII neck was used to replicate the response of a football player’s head. Per manufacturer’s specification, the cable in the HIII neck was tensioned to 1.1 Nm (10 in·lb.). Two nylon stockings were stretched over the HIII head before fitting the helmet to reduce friction and to provide a more realistic interface between the helmet and the headform skin. The HIII head-and-neck system was connected to a table fitted with linear bearings, permitting the head-and-neck assembly to translate in the direction of impact (Fig. 1). The sliding table allows for both more realistic kinematics and improved equipment durability while not affecting the head response, since the head acceleration impulse occurs before significant neck bending.39

Figure 1
figure 1

A pneumatic linear impactor (Biokinetics, Inc) was employed to replicate on-field reconstruction of head impacts. The ram mass, impactor surface, and target velocities were selected to best simulate on-field head impacts occurring in the National Football League

Full size image

Impacts were delivered to the HIII using a pneumatically activated linear impactor (Biokinetics Ottawa, Canada; Fig. 1).29,39 The linear impactor uses a pressure storage vessel with attached high-speed solenoid valve to accelerate a ram by in-rush of high-pressure air into a pneumatic cylinder. The ram detaches from the cylinder to insure that the ram is not being driven at the moment of impact. An impacting face is attached to the ram via a metal plate. The impacting face is made of ultra-high molecular weight polyethylene, and has the same curvature as that of a football helmet and a surface diameter of 127 mm (5 in). A 35-mm (1–3/8 in) layer of vinyl nitrile foam (VN600, DerTex, Saco, Maine), the same material used in many models of football helmets, was placed behind the impactor surface to mimic helmet-to-helmet contact. Impact velocity was measured by passing the impactor ram through a light gate just before contact with the helmet. The light gate also acted as a trigger for HIII data collection.

A new medium Revolution helmet (Riddell, Elyria, OH) with standard inflatable liner and 1.5-in. jaw pads was instrumented with a HIT System sensor insert. A G2EG-R style facemask was used for all impacts. The facemasks and attachment hardware were visually inspected between each impact and replaced with new equipment at first sign of deformation. The helmet was initially fit to the instrumented headform according to the manufacturer’s fitting specifications and then re-adjusted between each trial to maintain consistent test conditions.

A high-speed camera was used to record the first impact trial at each reconstruction condition. Video was recorded at 1000 Hz and was triggered via the impactor light gate thereby synchronizing video with HIII data. High-speed video was used to confirm consistent helmet placement and to qualitatively review the quality of impact reconstruction.

Primary Impact Sites

Four primary impact sites, designated as A, B, C, and D (Fig. 2), were identified by Pellman et al.37 as points of contact that most frequently result in concussion for NFL athletes. In that study, the average impact velocity for impacts associated with and without concussion was determined to be 9.3 ± 1.9 m/s and 7.0 ± 2.6 m/s, respectively. For this study, each site was impacted at four target speeds: 4.4, 7.4, 9.3, and 11.2 m/s. The highest three velocities correspond to the average speed, ±1 standard deviation, of video-reconstructed impacts sustained before diagnosis of concussion in NFL players.38 4.4 m/s represents the average velocity, −1 standard deviation, of all sub-concussive impacts evaluated by the NFL. This lower impact velocity was selected to insure inclusion of conditions that may not result in injury, but occur most often in the field.20,24,30 Between three and five trials were conducted for each site and speed combination (Table 1).

Figure 2
figure 2

Four primary impact sites were (A, B, C, and D) were identified as points of contact that most frequently result in mTBI for NFL athletes. Each site was impacted at four target speeds: 4.4, 7.4, 9.3, and 11.2 m/s

Full size image
Table 1 Kinematic parameters were recorded from instrumented helmets and a Hybrid III ATD following 54 linear impactor tests to correlate output measures from on-field and laboratory measurement systems
Full size table

Secondary Impact Sites

After initial testing and analyses were conducted, two additional impact sites, A′ and A′′ (Fig. 2), were added to the test protocol at the request of the then co-chairman of the NFL mTBI Committee to characterize HIT System measurement capability during conditions where extensive helmet movement was anticipated to occur. Sites A′ and A′′ were identified by Craig (2007) who, when reevaluating video from the initial NFL studies, suggested a sub-category of site A impacts existed that resulted in chinstrap loading.9 The two additional sites were variations of the original site A, located directly in line with the mid-sagittal plane of the headform, and tilted away from the impactor at −10° and −20° angles, respectively. Based on the recommendations made by Craig,9 and previous experience which predicted that facemask deformation and HIII neck damage were likely to occur, the 11.2 m/s speed was not used for A′ and A′′ tests. Three trials were conducted at each test condition except for A′′ at 4.4 m/s which had four trials.

Laboratory Measurements

The HIII head was outfitted with nine accelerometers (Endevco Corp., San Juan Capistrano, CA) in a 3–2–2–2 configuration.32 This arrangement allows for the resolution of three-dimensional (3D) angular accelerations using linear accelerometers, and correction for centripetal accelerations.15 The accelerometer signals were filtered initially with a CFC1000 hardware filter in the signal-conditioning device. The data were collected using a custom Labview program, post-processed using an additional CFC180 filter,31 and used for calculation of linear and rotational head accelerations. Direction of the impact recorded by the HIII was determined by converting the 3D peak linear acceleration vector into a spherical coordinate space.

Instrumented Helmet Measurements

Accelerometer data were collected by the instrumented helmet’s on-board microprocessor at 1000 Hz over a 40-ms window for each event. Data collection occurred when any of the six accelerometers exceeded a pre-filtered threshold of 10 g (8-ms pre-trigger and 32-ms post-trigger). An on-board 400 Hz low-pass filter was applied to the data after a 0.5-Hz AC hardware filter removed any DC offset. Data were then transmitted wirelessly to a laptop computer where each event was processed using a simulated-annealing optimization algorithm that solved for linear acceleration magnitude at the head center of gravity.10 Impact location was estimated by taking the reflection of the 3D direction vector of peak linear acceleration which is optimized from the peak output of each of the six accelerometers. The directional vector is then converted to a spherical head coordinate system (azimuth and elevation) that has been previously reported by Greenwald et al.20 Rotational acceleration at the time of peak linear acceleration was calculated as described by Rowson et al. using the modeled equations of motion for a force acting on the head, the anterior–posterior and medial–lateral components of the peak linear acceleration vector, and the directly measured relationship between linear and rotational acceleration or on-field head impacts.43

Data Analysis

Linear acceleration time series data from both the HIT System and HIII were processed in real time to obtain the resultant linear acceleration for each impact. Because a simultaneous trigger was not feasible, the linear acceleration resultants were synchronized post-processing by minimizing the RMS error between resultants.1,41 HIII data were then truncated to a 40-ms time window to allow direct time series comparison with the instrumented helmet. Resultant linear acceleration data from both systems were utilized to calculate two impact metrics used for relating time-weighted head acceleration to risk of injury, GSI and HIC15. GSI was calculated for each impact event19:

$$ {\text{GSI}} = \int\limits_{{}}^{T} {a\left( t \right)^{2.5} dt} $$
(1)

where a(t) is the linear resultant acceleration of the head CG, and T is the impact duration. HIC15 was evaluated over an incremental time window38,45 of maximum duration t 2 − t 1 = 15 ms:

$$ {\text{HIC}}_{15} = \left\{ {\left( {t_{2} - t_{1} } \right)\left[ {\int\limits_{{t_{1} }}^{{t_{2} }} {a\left( t \right)dt/\left( {t_{2} - t_{1} } \right)} } \right]^{2.5} } \right\}_{\max } $$
(2)

where t 1 and t 2 are the initial and final points of the time window and a(t) is the linear resultant acceleration of the CG.

To establish a correlation between severity measures from the HIT System and HIII, linear regression analysis (Microsoft Excel 2010) was performed on all acceleration and severity measures—peak linear acceleration, peak rotational acceleration, GSI, and HIC15.1 All regressions were performed both on the entire dataset and by impact site using Eq. (3):

$$ y = mx + y_{0} $$
(3)

where x is the HIII measure, y is the HIT System measure, and the linear slope, m, is the relationship between the measurements. For all conditions, y 0 was constrained to be zero since both systems have a baseline output of zero when not impacted. The coefficient of determination (r 2) was also calculated for each regression as a measure of goodness of fit. Finally, the average absolute location difference between the estimated instrumented helmet location and the direction vector (in spherical coordinates) of the peak HIII linear acceleration was calculated for each impact. The overall average difference between the two measures along with the difference by impact site was calculated.

Results

Primary Impact Sites

Fifty-four impacts were measured by the instrumented helmet and HIII over four target velocities and impact sites previously defined by the NFL as potentially injurious (Locations A–D). A high correlation for peak linear acceleration was found between the two systems, r 2 = 0.903, with a slope of 1.009 for the linear fit trendline (Fig. 3) indicating a 0.9% overprediction of the HIT System compared to the HIII for all locations and velocities. Analysis by impact site again showed high correlation between the two systems (r 2 = 0.822–0.987) with no linear slope deviating more than 8.4% of nominal for any impact location (Table 2).

Figure 3
figure 3

The HIT System overestimated Hybrid III peak linear acceleration by 0.9% and underestimated peak rotational acceleration by 6.1% for all primary impact sites. Acceleration measures for all impacts were correlated; however, linear acceleration was higher (r 2 = 0.903) than rotational (r 2 = 0.528) primarily due lower HIT System rotational acceleration estimates at impact site A

Full size image
Table 2 Results from linear regression analysis between HIT System and HIII impact measures
Full size table

Peak rotational acceleration was correlated for B, C, and D sites (r 2 = 0.710–0.981) but not for A location (r 2 = 0.415). For all impact sites, the slope of the fit trendline was 0.939 (r 2 = 0.528); however, the coefficient of correlation (r 2 = 0.800) improved when only considering sites B, C, and D (Fig. 3).

The coefficient of correlation between the two measurement systems for GSI and HIC15 were r 2 = 0.846 and 0.787, respectively with no individual impact site correlation lower than r 2 = 0.609. Slopes of the fit trendlines for all impact events were 0.948 and 1.061 for GSI and HIC15, respectively (Table 2). Impact location recorded by the HIT System had an average absolute location difference of 31.2 ± 46.3° for all impacts when compared to the acceleration direction of the HIII (Table 2). Most of this error can be attributed to six impacts to the A site, which were symmetrically opposite of the expected. When considering location measurement without the A impact site, the average difference for remaining sites was only 13.2 ± 6.3°, well within the overall 0.127-m diameter of the impactor surface (10° absolute error equates approximately to 0.0127 m). The average impact location as estimated by both systems can be seen in Fig. 4 for all impact locations.

Figure 4
figure 4

The average absolute difference between impact location estimated by the instrumented helmet and HIII for all impacts was 31.2 ± 46.3° for all primary impact sites (A, B, C, and D). Arrows shown on the headforms above, identify the average impact location by both systems for each impact sites. A * is placed to the left of the arrow indicating impact location estimated by the instrumented helmet. Six of the twelve impacts to site A were identified by the instrumented helmet as being symmetrically opposite of the actual location, and so two arrows are provided at that location for the instrumented helmet to differentiate those events

Full size image

Secondary Impact Sites

For the 19 additional frontal impacts at sites A′ and A′′, the HIT System had a moderate overall peak linear acceleration correlation with the HIII (r 2 = 0.556) and greatly overpredicted the HIII as indicated by a factor of 2.673. While the HIT System overpredicted the HIII at both sites, linear acceleration was highly correlated for site A′ while no correlation was identified for site A′′ (Table 2). Rotational acceleration, GSI, and HIC15 all followed similar trends of overall moderate correlation (r 2 = 0.438, 0.609, and 0.649), high correlation for site A′, no correlation for site A′′, and the HIT System overpredicting the HIII for each measure (m = 1.627, 5.346, and 3.835). Impact location recorded by the HIT System for A′ and A′′ locations had an average absolute location difference of 42.1 ± 13.4°, which was consistent with the results at the A, B, C, and D locations (Table 3). The source of this discrepancy can be attributed to location estimates from the HIII that did not correspond to the point of impact for either impact site, but rather went directly through the nose.

Table 3 Results from linear regression analysis between HIT system and HIII impact measures
Full size table

Discussion

Researchers have long sought to determine the relationship between head kinematics following impact and the pathophysiology of traumatic brain injury. One approach to understanding this relationship for mTBI has been to treat the football playing field, where athletes may sustain more than 2000 impacts to the head during a season,11–13 as a living laboratory. Specifically, researchers have monitored the impacts sustained by football players and have attempted to correlate their impact exposure with signs and symptoms of injury. Over the last decade, two primary approaches have been implemented for quantifying head kinematics following impacts experienced during play: laboratory reconstruction of impacts recorded on video and on-field measurement using instrumented helmets. The aim of this study was to correlate HIT System output with Hybrid III ATD output in a laboratory test using a linear impactor system under realistic input conditions.

Measures of peak linear acceleration, GSI, and HIC15 obtained from the instrumented helmet and HIII were found to be correlated (r 2 = 0.903, 0.846, and 0.787), and the overall system-to-system relationship was less than 6.1% for each measure (0.9%, 5.2%, and 6.1%, respectively). Higher deviations for both GSI and HIC15 were not unexpected given the reliance of these measures on the entire 40 ms linear acceleration resultant which leads to more potential points of deviation than maximum value measures. While little difference in correlation was found between impact sites for peak linear acceleration, site A resulted in the lowest correlation for both GSI and HIC15. A review of high-speed video data for these impacts provided insight to the discrepancy between the A site and other impact sites. For the A site, the impactor ram drove the helmet downward and counter clockwise toward the HIII before head acceleration began (approximately 5 ms of helmet contact before head acceleration). Once the impactor surface settled into the opening of the facemask, the helmet and headform moved simultaneously. This helmet movement prior to head acceleration caused a discrepancy in the linear acceleration resultant time series between the two systems from which GSI and HIC15 are calculated.

In these tests, the instrumented helmet underestimated peak rotational acceleration compared to the HIII by 6.1%. The overall correlation between the two systems, however, was only moderate and lower than the other impact measures. This can be attributed primarily to measurements at the A site which had low correlation with the HIII (r 2 = 0.415) when compared with the other primary sites B, C, and D, (r 2 = 0.800). The processing algorithm used in conjunction with the instrumented helmet in this study calculates rotational acceleration at the time of peak linear acceleration from the equations of motion for force applied on the head.43 The unknown parameters of head mass, head moment of inertia, and the point of rotation about the neck are determined by the relationship of peak linear and rotational acceleration of on-field impacts obtained from a similar device paired with a full six degree of freedom processing algorithm.8,41 Compared to the full six degree of freedom helmet, this approach reduces cost of the instrumented helmet and the computation time required for processing head acceleration measures, thus allowing researchers to conduct studies on a larger sample of players. This approach, however, also limits rotational acceleration calculations to two degrees of freedom since acceleration about the axis extending through the neck cannot be calculated. Similar to findings from previous studies evaluating the accuracy of rotational acceleration measures from the HIII head and neck,3,31 results from these tests indicate that the HIT System estimation of rotational acceleration is an accurate approach when analyzing and reporting results based on impact distributions; however, care should be taken when reporting rotational acceleration from either system for single impact events.

The average difference between the direction of head acceleration estimated by the instrumented helmet and HIII was 31.2° for all primary test impacts. Based on HIII dimensions, a 10° offset correlates to approximately 0.0127 m of head circumference, and, given the impactor size (diameter = 0.127 m), these results indicate the instrumented helmet measures are well within practical reason for on-field studies. The instrumented helmet identified the direct opposite acceleration for six of the twelve site A impacts causing a higher mean difference at this site compared to all others. It is important to note that the HIT System algorithm used in this study optimizes for impact location based on the temporal relationship of peak acceleration values obtained from each of the six embedded accelerometers. While the resultant head acceleration corresponds with the head CG, the estimation of location best approximates impact location on the helmet. In most cases the impact location and direction of acceleration is the same, but, there are occasions where it is not. A good example of this can be seen in the average estimated impact locations of both the HIII and HIT System at sites A′ and A′′ (Fig. 5). While the HIT System correctly identifies the impact location as being low on the facemask, the direction at peak acceleration measured by the HIII is directly through the nose. Incorrect identification of acceleration direction for the six site A impacts is related to the facemask and impactor surface interaction previously described. Initial contact caused the helmet to move downward before head acceleration-mimicking motion typically resulting from an impact to the back of the head. It is unclear if this phenomenon is limited to the conditions present in these specific laboratory tests or if there is the potential for misidentifying impact location in the field; however, on-field performance of the HIT System has previously been verified through video review without description of inaccurate location measurements and, anecdotally, have not been recorded under similar test conditions at other independent test laboratories.

Figure 5
figure 5

The HIT System’s processing algorithm to estimate impact location best approximates the impact site on the helmet. The average location difference between the HIT System and HIII for sites A′ and A′′ was 42.1 ± 13.4°; however, the HIT System correctly identified the impact location as being low on the facemask, while the estimate of impact location from HIII was directly through the nose

Full size image

Nineteen additional frontal impacts were measured at two sites (A′ and A′′) with increased neck angle relative to the impactor, causing primary loading of the chinstrap and large deflection of the helmet off the head. In these tests, the aim was to challenge the HIT System’s ability to retain contact with the head during the impact. For these 19 additional impacts, the instrumented helmet had high correlation with the HIII for all A′ measures (r 2 ≥ 0.883) and low correlation for all A′′ measures (r 2 ≤ 0.308). All four severity metrics were greatly overpredicted by the instrumented helmet. As with the A location, high-speed video illustrated excessive helmet departure from the head 6 ms before initial head acceleration (Fig. 6). Maximum lift of the helmet off the head was estimated to be in the range of 1–3 in. depending on impact velocity. For the highest impact speeds, the entire gap between the face and facemask was closed as the bottom facemask bar was driven into the HIII chin. From these observations, it can be assumed that the instrumented helmet was potentially no longer in contact with the head at the time of initial acceleration and most likely not in contact at the time of peak acceleration. The large overestimation of linear acceleration would therefore be attributed to a high component of helmet acceleration artifact measured by the instrumented helmet.26 To account for helmet movement that will inevitably occur during play, the instrumented helmet is designed to isolate head acceleration from helmet acceleration via the use of spring-like actuators.26 Accelerometers are embedded in the actuators so that head contact is maintained even during helmet movement. These actuators can account for the majority of helmet movement that occurs during a typical impact, however, in the event that the helmet is dislodged from the head before the acceleration pulse, as was the case in these laboratory tests, the system most likely will overestimate head acceleration. This represents a practical limitation of the system that must be considered by adopters of the technology.

Figure 6
figure 6

High speed video of a 9 m/s impact at the A′′ site. Top Row: (1) Initial contact occurs, (2) 1 ms after contact—facemask compresses, (3) 6 ms after contact—facemask is bent, helmet has lifted off the head, and the head begins to accelerate. Bottom Row: (4) 10 ms after contact—impactor surface pushes the facemask into the chin, (5) 16 ms after contact—the helmet experiences maximum deflection off the head

Full size image

The results from this study predict that impacts to the A′ and A′′ tests measured by the instrumented helmet will result in much higher accelerations (e.g., 2–5 times the actual value) for a given impact velocity when compared to other impact locations. It is important to note, however, that, independent of the large differences in acceleration magnitudes recorded between the systems at A′ and A′′, the impact location was closely identified. While impacts will inevitably occur on the field at these sites, we do not believe that the A′ and A′′ sites and, in some instances, the A site, as currently tested using the linear impactor represent realistic cases of helmet impact and helmet motion that occurs on the football field. Using data collected with the HIT System, Mihalik et al. have reported statistically lower average peak linear acceleration for impacts to the frontal region of the head than to the top.27 Similarly, Greenwald et al. reported frontal impacts are statistically indistinguishable in terms of impact severity from side of the head impacts and are much lower than either impacts to the top or back of the head following analysis of on-field head impact data from 13 high school and collegiate football teams.20 In the most extensive data collection reported to date, Crisco et al. recorded head impacts with the HIT System from three collegiate football teams over three seasons, resulting in 286,636 recorded impacts.12 Front of the head impacts (i.e., front 90° quadrant of the head and below an elevation level of 65° that generally relates to mid-forehead)20 accounted for 41.7% of these events. To compare with the laboratory tests reported here, data from the cited study were separated into three general levels of elevation to estimate forehead, mid facemask (similar to A and B sites), and low facemask (similar to A′ and A′′ sites) impacts (Table 4). Impacts to the lower portion of the facemask do occur on the field (13.3% of all Front impacts, 5.5% of all impacts); however, a comparison of median and top 95th percentile levels of linear and rotational acceleration by impact elevation does not support the A′ and A′′ laboratory test results which suggest that the HIT System will consistently measure low facemask impacts by a factor of 2–3 times higher than those of mid facemask and helmet shell impacts. While the A′ and A′′ impact sites do provide insight into potential limitations of using instrumented helmets, the discrepancy in on-field results with those predicted by linear impactor tests also demonstrates the difficulty of creating a biofidelic reconstruction of live events in a laboratory setting.

Table 4 On-field data support that impacts to the lower portion of the facemask do occur; however, a comparison of median and top 95th percentile levels of linear and rotational accelerations by elevation does not support the A′ and A′′ laboratory test results, which suggests that the HIT System will consistently measure low facemask impacts by a factor of 2–3 times higher than those of mid facemask and helmet shell impacts
Full size table

We chose to use regression analysis as the primary means for correlating output between the HIT System and HIII. One of the primary benefits of the HIT System is its ability to record large amounts of in vivo impact data that are inclusive of all on-field scenarios that can be experienced by different playing positions and skill levels.5,11,12,27,42,44 It can reasonably be assumed that each of the 286,636 impacts reported in the previously described study by Crisco et al. represents a specific combination of input conditions that could influence the player’s head kinematics. Because of this variability in conditions, researchers have primarily chosen to present data from the HIT System in the form of impact distributions inclusive of a variety of impact conditions (e.g., impact location, impact severity, playing position, etc.).

By utilizing regression analysis, an expected relationship between the overall distributions of output from the two systems can be obtained as well as an indicator for the strength of that relationship. This relationship, though, should only be assumed for conditions included in the test distribution. For example, considering the A, B, C, and D impact sites and associated test velocities were previously found to be representative of concussion in the NFL, it would be expected from these test results that HIII measures obtained by Pellman et al. would closely match those recorded by the HIT System for concussive injuries when compared as a group. This close relationship has been confirmed in several studies,4,23,42 including one analysis that found no statistical difference in peak linear acceleration, peak rotational acceleration, HIC15, or GSI between 55 on-field impacts associated with concussion and those obtained from the 25 reconstructions of concussive impacts conducted by the NFL.2 The relationships provided by regression analysis do not, however, provide an estimate of how the two systems may differ for individual impacts. One approach to obtaining this estimate would be to calculate the percent difference for each impact. For all trials conducted at the primary impact sites, this method would result in linear acceleration differences between 0.1–38.9% for individual impacts simulated in this study. While there is benefit to this approach, it has only limited utility considering that the association can only be applied for the exact input conditions, which are rarely known for on-field impacts. Because the exact differences for single impact measures between the HIT System and HIII remains unclear because of the impracticality of evaluating all possible on-field scenarios in the laboratory, researchers should use caution when analyzing results from isolated events. In general, though, the relationship found through regression suggests distributions of measurements obtained on-the field with the HIT System will be nearly identical to those obtained in the laboratory as long as the laboratory tests reflect field conditions. This relationship appears to be strong even for distributions of relatively few data points, as shown by the 54 impacts evaluated in this study.

Athletic environments, American football in particular, have long been viewed as a potentially rich test-bed for understanding concussions. Over the last decade, two methods have been employed to obtain head acceleration measures occurring from actual impact events. Laboratory reconstructions utilize precision measurement equipment to recreate on-field impacts but are limited by cost, technical expertise, and the necessary assumptions of the human surrogate models. The HIT System is an on-field measurement system that allows for large-scale data collection by actively measuring head acceleration of athletes during play. Given that the current knowledge base has been developed using both techniques, the current study was required to establish the relationship between these two measurement methodologies. From these laboratory tests, a relationship has been established between the two systems, particularly when considering peak metrics such as linear and rotational acceleration for distributions of impacts; however, it is important for users of HIT System technology to understand the practical limitation of ensuring proper fit of a player’s helmet. This limitation could result in error for single impacts that is similar to those previously reported for laboratory reconstructions. Results from this study indicate that measurements from the two methods of study are correlated and provide a link that can be used to better interpret findings from future study using either technology.