There is a continuing interest in the understanding and prevention of head injuries in sport. Although early research focused on severe head injuries, in the last decade there has been greater interest in understanding sports-specific loadings and injury biomechanics related to concussion (25,31,35). There has also been an increase in the availability of data describing the incidence and prevalence of concussion in sports, including football, the topic of this article (16,18). From the medical point of view, effort has been made toward a better definition and severity grading of these events (19). From the biomechanical perspective, the challenge is to define both injury mechanisms and biomechanical indicators, or injury criteria, for the assessment of injury risk.
Head impact dynamics in contact sports have been analyzed using in-game measurements (7,25,34), video analysis (21), including reconstructions using anthropomorphic test devices (10,27,31,38), and numerical methods (2,35,40). These studies concern mostly the analysis of helmeted impacts, as in American Football. The analysis of the biomechanical indicators associated with these impacts is prerequisite for the development of methods to mitigate the effects of head impacts in these sports. For example, it has been observed that head impacts may be more severe than those performed in laboratories to assess padded headgear and protective helmets (20).
Finally, there has been some debate regarding risk compensation or behavioral adaptation in sport related to the use of padded clothing and protective equipment (12,22). Without biomechanically based methods for measuring impact severity, it is difficult to investigate these behavioral factors in sport. For these reasons, it was decided to continue a program of research into concussion that was commenced in the 1990s. In the first phase of the program, a field study had been undertaken, in which videos of the impact dynamics from a population of a hundred players with medically verified concussion were analyzed (21). To refine this analysis, a protocol of numerical simulation reproducing these real-life impacts was developed using MADYMO. This protocol and its validation were presented previously (8). In this article, a study of the biomechanics of concussive impacts experienced by unhelmeted football players is presented.
Background to the case reconstructions.
The numerical reconstructions build on available data obtained previously from a two-dimensional video analysis of concussive events that occurred during football games. A set of a hundred concussive head impacts in rugby and Australian football had been analyzed and reported (21). Both the causes and the symptoms associated with these events characterized them as sports concussions, as defined by McCrory et al. (19). Informed written consent by each player and ethics approval for the study were obtained and described in the same article (21). For impacts between two players or one player and the ground, this protocol had enabled an evaluation of the impacted player's head kinematics, pre- and postimpact, allowing an evaluation of associated head impact dynamics. Also available from this previous study, were the anthropometry (mass and height) of each player involved in an impact as well as medical assessment of the injury (definition and duration of the symptoms, concussion grade). For this study, the concussion grade had been defined as follows: grade 1-no loss of consciousness (LOC); grade 2-LOC <1 min; and grade 3-LOC >1 min.
Great care had been taken in the estimation of the players' kinematics from the video of each concussive event; in particular, the possible calculation errors induced by a two-dimensional analysis had been evaluated. On the basis of this analysis, a first selection of videos had been performed, restricting the analysis to impacts where the closing movements occurred within an estimated 20° of the plane of the camera. With this first selection and based on comparisons with calibrated reconstruction in a laboratory environment, the authors had evaluated the error in reconstructing each player's initial kinematics to be below 10%. However, it appeared that many impacts generated head movements that occurred out of the plane, yielding possible inaccuracies in the dynamic calculations that were inherent to the two-dimensional analysis. To take into account precisely these out-of-the-plane components, to refine the initial calculations, and to estimate head accelerations and related injury indicators, a three-dimensional numerical analysis was chosen for the present study. The dynamics of such contact events strongly depend on the players' effective masses involved in the impact; therefore, it was decided to model the whole players. The rigid-body MADYMO (TNO Automotive Safety Solutions) human facet models were chosen for these simulations. These models are used widely in automotive crashworthiness assessment, and their behavior has been validated against several sled test (frontal, lateral, and rear-end) as well as blunt test (head, shoulder, thorax, and abdomen) impact configurations (36). They also offered an accurate description of the body shape geometry and their relative simplicity compared with an FE model allowed for an easy personalization to the players' morphometry.
Twenty-seven cases of impact between two players were chosen, where the impact occurred close to the plane of the camera. These cases were then reconstructed and simulated using the following protocol: first, the models were positioned using HYPERMESH v6.0 (Altair Engineering, Inc.) to reproduce the relative position of each player just before the impact(Fig. 1). The masses and the inertias of each model's body segments were calculated based on the known anthropometry using GEBOD (6) scaling equations and were lumped into the model's segments. The closing velocities previously assessed from the videos were used as input to the simulations.
The initial MADYMO human model was designed to be used as a seated, unaware occupant in an automotive impact environment. It only models passive behaviors of the body segments and joints that are not representative of the muscular tension that is expected to be found in a standing/running player, especially if the player braces himself in anticipation of an impact. However, it was difficult to both assess the awareness of the injured players on some of the videos and transfer appropriately this information into the model. Furthermore, a preliminary sensitivity analysis performed in the coronal plane (8) had confirmed that, compared with other variables, neck stiffness had a relatively low influence on the head's peak dynamics for direct impacts (30). Therefore, it was decided to model a generic state of muscular activation in each player's model by adding joint restraint torques into the model so that it could just maintain a standing upright position in a presimulation under a 1g gravity. Finally, the initial position of the model, the initial velocity of each body segment, and the stiffness of each joint were fine-tuned to obtain a satisfactory match between the kinematic behavior of the impacting bodies compared with the real event on video. The restraint torques used for the shoulder, the elbow, the hip, the knee, and the ankle were chosen within the range of associated maximal isometric joint torques reported in the literature.
Biomechanical output data.
All simulations were carried out using MADYMO v6.2.2. Simulations were performed on a 200-ms time frame with a time step of 10−3 ms, allowing for the description of both the impact and the kinematics shortly thereafter. The dependant variables chosen as output were the head impact power (HIP) (26), the head impact criterion (HIC15), the peak linear acceleration at the head's center of gravity (CG), and the peak angular acceleration of the head. The linear peak velocity change (PVC) and peak angular velocity change (PAVC) were also calculated.
To reproduce the conditions of these concussive impacts with drop tests and to assess the effectiveness of headgear, it is necessary to obtain an estimate of the energy that is transmitted to the head during an average impact. The impact being inelastic, there is no conservation of the kinetic energy of the two-segment system. If the head is assumed to be a rigid body, the change in total energy is equal to the work of the nonconservative forces acting on the head:
where PEi and PEf are the initial and the final potential energy of the head, respectively; KEi and KEf are the initial and the final kinetic energy of the head, respectively; and ΔU is the work performed by the external forces and moments on the head during the impact.
On the basis of this equation and assuming the work of the neck forces and the change in potential energy to be negligible compared with the change in kinetic energy and work of the contact forces, McIntosh et al. (21) had defined the head impact energy (HIE), equal to the difference in the head's kinetic energy between just before and just after the impact, as an estimate of the energy that was transferred to the head due to the action of these forces. For means of comparison with this previous analysis, this HIE was calculated in the present study. Practically, an estimate of the normal to the contact was defined as the direction of the change in the head's velocity vector. The final velocity vector was assessed just after the impact and, along with the initial head velocity vector, was projected on this normal to calculate the KEi and KEf and the impulse.
The HIE, impulse, and PVC results were compared with the results obtained through the previous video analysis. Because the variables' distributions were not Gaussian, a Wilcoxon matched pairs nonparametric test was used to test for difference between the two sets of results. The relationship between each biomechanical variable and concussion grading was assessed using binomial logistic regression analysis. The outcome variable, that is, concussion severities 1, 2, and 3, was assessed in pairs or was grouped; for example, severity 1 versus severity 2 and severities 1 and 2 combined versus severity 3. To quantify whether the relationship between the severity grading and a variable was statistically significant, the −2 log likelihood ratio significance test was used. Significance was measured at the level α < 0.05 for all statistical analyses. The statistical software SPSS version 15.0.1 (SPSS Inc., Chicago, IL) was used for these analyzes.
Figure 1 presents an example of the MADYMO reconstruction and simulation of one of the impacts (case 16) between two Australian Rules Football players, which resulted in a grade 2 concussion. Table 1 presents the full results of the peak values reached by each biomechanical variable stratified according to each concussion grade, for each reconstruction. There were nine simulations for each concussion grade. HIC values for concussion ranged from 87 to 994. The greatest HIC was reached for one of the most severe impacts, where peak values of 100 J in HIE and of 43 kg·m·s−1 impulse were reached. The overall mean values for HIC, peak linear, and angular acceleration were 359, 103g, and 8022 rad·s−2, respectively. Although the descriptive statistics for the results stratified to concussion severity demonstrated high SD (68% and 69% of the mean value for HIC and HIP, respectively), a common trend between injury severity and some of the biomechanical parameters was observed. The mean for the HIE was 56 J.
Figure 2 presents the peak rotational acceleration function of peak linear acceleration of the head's CG and PAVC, respectively. Concussions were associated with combined peak angular and linear accelerations of 3470 rad·s−2 and 586 m·s−2 (60g) and peak angular accelerations and PAVC of 3470 rad·s−2 and 15.6 rad·s−1, respectively.
Results obtained for the HIE, impulse, and PVC were compared with the equivalent results from the previous video analysis (21), first with the same 27 cases then with the full set of the 100 video cases. The associated mean values are presented in Table 2. Mean simulation results for HIE, impulse, and PVC were slightly higher than their equivalent calculation from the video analysis (Wilcoxon's P < 0.05).
Table 3 presents the results of the significance tests for the logistic regressions between the severity grades and the biomechanical parameters. There was a significant relationship between severity grading and some of the parameters, depending on the grades or combinations of grades considered. For example, HIP, HIC, peak linear acceleration, PVC, HIE, and impulse provided a significant correlation with change from grade 1 to grade 3. Overall, the HIP provided the strongest correlation with the change in severity, whereas no significant relationship was observed between the peak angular acceleration and the concussion severity using this method. No significant relationship was observed between the biomechanical parameters and the outcome variable concussion severity grade 1 and grade 2.
In this study, 27 real-life concussive head impacts in football (rugby and Australian Rules Football) were reconstructed numerically. These three-dimensional reconstructions allow refining the knowledge of the dynamics associated with concussive head impacts from a previous video analysis (21). In particular, parameters such as head acceleration, change in velocity, and impact energy were assessed, which allow defining equivalent drop-test conditions for headgear evaluation in these sports.
General comments and limitations.
There are many degrees of freedom in numerical reconstructions of impacts. Assumptions regarding the model's geometry and mechanical properties influence the results. Errors may also come from the evaluation of boundary conditions and from their transfer from the video to the model. As an example, the reconstruction of the initial position of the players just before the impact was performed by assessing the videos frame by frame. For many videos, these frames were blurry, which may have resulted in an imprecise assessment of both the location and the orientation of the head impact. Therefore, before reconstructing the real-life impacts, a sensitivity analysis of the influence of various independent parameters on the dynamics of the head impact had been performed (8). The influence of the possible error made in the estimation of parameters such as neck stiffness, contact stiffness, and coefficient of friction had been evaluated between a low and a high level that were representative of the range of physiological values found in the literature. The influence of an error made in evaluating the impact velocity had been assessed between a low and a high level that represented a possible error of ±10% around a mean value, representative of the closing velocities assessed in the video analysis. This possible error margin was based on the previous video analysis (21). Finally, the influence of possible errors in the assessment of the location and orientation of the contact had been assessed.
The sensitivity analysis showed that errors in the evaluation of the location of the head contact point and of the impact velocity had potentially the largest influence on the biomechanical parameters chosen as possible injury predictors. The contact stiffness had the second highest effect on these predictors. To limit these effects in the reconstruction of the real-life events, refined head contact properties were implemented for the MADYMO head model that were evaluated against a range of experimental boundary conditions from the literature (8). To limit the effects of possible error in the assessment of the impact velocity due to the two-dimensional analysis, 27 cases out of the 100 from the initial database were selected and reconstructed. These videos were chosen based on their clear description of the event; allowing both the determination of accurate boundary conditions and the assessment of the reliability of the simulations. In particular, videos were chosen where the closing movement between the players occurred in or close to the plane of the cameras to minimize errors made in the estimation of the velocity. Finally, the peak head angular acceleration was strongly influenced by the selected location of the contact point. This factor is inherent to any similar reconstruction process, whether it involves computer modeling or physical testing.
Notwithstanding these factors, difficulties also reside in the assessment of both the presence and the severity of all injury types. Reasons are numerous and relate to variability both in the injured human and in the impact situations as well as the range of experimental data acquired. The validity of global mechanical parameters and injury criteria to assess injury risk also remains a point of argument and controversy (14). For example, the HIC is based on the experimental assessment of the presence or absence of fractures, and it is a significant extrapolation to use it as a general predictor of mild traumatic brain injury or concussion, as it is now characterized in sport. However, previous studies have found that both impact and translational acceleration (11,13,29) and rotational acceleration (1,9) could be associated with brain injury. Intracranial pressure and shear stress in the brain were, respectively, described as injury mechanisms associated with these loadings. Moreover, correlations between these global criteria and the risk of concussion have also been found (27,33,40). They are also simple to calculate, generally highly reliable in impact testing, and may be used effectively for means of comparisons, for example, for headgear evaluation.
Discussion of the results compared with the literature.
As seen from Table 2 and although they show similar trends, the average HIE, impulse, and PVC obtained from the 27 numerical reconstructions are slightly higher than their counterparts obtained from the video analysis. The video analysis was two-dimensional and only allowed assessing projections of the velocity vectors in the plane of the camera. The current numerical reconstructions allowed for a full three-dimensional assessment of this specific variable, which would explain the higher values obtained in this case.
Other numerical studies on head injury biomechanics have been performed previously (3), and a larger number were aimed at describing the mechanisms of injury and therefore used more precise FE models (15,24,33,40). Only a few studies reported on real-life accident reconstruction, either using dummy human surrogates (27,31) or numerical models (2,28,40). Finally, thanks to improvements in measuring techniques, several field measurements have been reported in the literature (4,7,25,34). Comparable results from these studies are presented together with our results in Table 4.
From this comparison, it appears that the levels reached by the biomechanical parameters associated with concussion in the present study are comparable to results from similar studies including helmets. In boxing, Walilko et al. (37) used an instrumented ATD to measure the average impact characteristics of punches from seven boxers in the flyweight to super heavyweight classes. Although the impulse levels were comparable (25 ± 8 N·s), the average energy transmitted to the head in these reconstructions was significantly lower (17.2 ± 8.9 J) than that in our study. Because the instrumented dummy was standing still in their study, this would confirm that high-impact energy levels found in rugby of Australian football are due to the combination of both impacting and impacted player's velocities. Similarly, Pincemaille et al. (32) found high angular acceleration values (up to 16,000 rad·s−2) while measuring nonconcussive blows with boxers, but all other biomechanical indicators associated with concussion in our study appear to reach significantly higher levels than with the average boxing punches measured by these authors.
An interesting point concerns the comparison between the American football (players wearing a helmet) and the football codes (rugby and Australian football) from the present study. In their reconstructions of 25 concussive events in American football (31), the authors evaluated PVC and "energy of impact" of the head to be 7.2 m·s−1 and 118 J, respectively. The calculation of this energy does not represent the change in kinetic energy of the head because it appears to be calculated as
However, this value is interesting for means of comparison. When using this formula on our results, it appears that both this "energy of impact" and the peak change in head velocity for concussion were found to be significantly higher (P = 0.002 and 0.004, respectively) in the study of Pellman et al. (31) study than that in ours (118 ± 59 J compared with 83 ± 40 J and 7.2 ± 1.8 m·s−1 compared with 5.8 ± 1.4 m·s−1, respectively). Pellman et al. (31) assumed similar average head masses (helmet mass was not taken into account) for the calculation of the impact energy. In this context, a player is exposed to a defined energy of impact, and the outcomes of that exposure are the dynamic responses of the head, for example, accelerations, and the injury. The use of a helmet interposes an energy attenuating device between the exposure and the outcomes, which will modulate the head's dynamics and thus injury likelihood. Similar injuries and biomechanical head injury indicators, as indicated by similar levels of HIC and accelerations, were observed in the two groups (helmeted and unhelmeted), however, with more severe impacts (as shown by the higher impulse and energy of impact in helmeted cases. This is an intriguing observation with regards to evidence for risk compensation or behavioral adaptation to the presence of a helmet. Are we seeing helmeted players increasing the severity of the impacts they are exposed to while maintaining the same risk as an unhelmeted player, or are these more representative of fundamental differences between the football codes?
Consequences for injury risk assessment.
Although the logistic regression analysis is not intended to provide a quantitative evaluation of the concussion severity, it is remarkable to observe that a significant correlation exists between the medical grading of the concussion and some of the biomechanical parameters. In that regard (Table 3), the HIP appears to allow the best discrimination between the grades. This conclusion reflects previous results of a similar evaluation performed by Newman et al. (27) and suggests that such a global criterion, taking into account both translational and rotational components, may allow for an improved assessment of concussion severity. No significant discrimination was found between grades 1 and 2: this could be explained by the fact that there might be only a slight difference in terms of medical assessment between the two first grades compared with the third grade, at which durations of both LOC and posttraumatic amnesia may be much greater. Such grading is directed toward medical management and is not necessarily an even discrete scale. In a recent consensus at the Prague International Conference on Concussion in Sports, it was suggested (19) that it should be replaced by a two-level assessment that differentiates between simple and complex concussion. Our biomechanical results are in agreement with this proposal.
Limitations related to both the reconstruction process and the injury assessment have been presented above. It is acknowledged that factors such as concussion history or impact history during the game may also affect this evaluation (34). However, although no control (no-injury) cases were included in this study, our results are in close agreement with previously (40) suggested values for a tolerable reversible brain injury. Figure 2 shows combined concussion thresholds in peak angular acceleration and PVC, which are also very close to the 5% strain curve obtained by Margulies et al. (17) when assessing the risk of diffuse axonal injury. Because grade 1 concussions were associated with mean HIC values of 230, HIP values of 8830 W, and peak linear accelerations of 86g, these could be added to the pool of existing tolerance values proposed for this specific injury. Due to the inherent risk of error in the evaluation of the peak angular acceleration, more cases are needed to confirm the levels found in this study.
Consequences for headgear testing.
It is still unclear if headgear is effective in protecting players from concussion in football. In American football (39), linear acceleration was shown to be significantly attenuated by the presence of the helmet, but not rotational acceleration. Although not detailed in our results, a calculation of the rotational part of the change in the head's kinetic energy yielded an average value of 15.8 ± 13 J. This rotational component is not currently implemented into headgear testing but may represent a significant contributor to the risk of concussion. Cantu et al. (5) reported that helmets reduced the number of more severe TBIs whereas this appeared not to be the case for headgear and concussion in rugby (23), a point that needs further research and that may be related with behavioral aspects. On the basis of our results, the common trend between the HIE and the severity grading also suggests that attenuating the impact energy (even if only its linear component) would reduce the severity of the concussion.
Current headgear testing methods include headform drop tests performed on rigid anvils. The following paragraph discusses how the present results may be relevant to refine the tests. The calculation of the HIE is an estimate of the energy that is transferred to the head during an impact. However, it is acknowledged that neither the head nor the impacting segment can be assumed to be rigid bodies during the impact. Irrespective of the coefficient of restitution, it is not possible to determine accurately the part of the impact energy that effectively contributes to the deformation of brain tissue and injury. In their comparison between lateral head impact reconstructions with dummies and anvil drop tests with headforms, Gilchrist et al. (10) proposed that the "effective impact energy" transferred to the head was the energy transferred up to the point where the head and the impacting segment had the same normal velocity. This energy includes the energy that is partly stored, partly dissipated in the deformation of the tissues, and would be the same as for a plastic impact. Because the HIE calculation also includes the part of the energy that is transferred to the head during the remainder of the contact, it constitutes an upper boundary of the impact energy that may be responsible for the brain injury. In our numerical simulations, it was possible to assess the head's velocity at the moment where both the head and the impacting segment shared the same normal velocity. By using this vector for the calculation of the final kinetic energy of the head, an average impulse of 25 kg·m·s−1 was calculated for the 27 cases, with average impulses of 22, 24, and 28 kg·m·s−1 for concussion grades 1, 2, and 3, respectively. The resulting average effective impact energy for concussion was 47 J, with mean values of 39, 42, and 60 J for grades 1, 2, and 3, respectively. These last values include energy losses through the deformation of the human scalp that are not modeled by rigid headforms (10); however, they constitute a reasonable estimate of the energies to be used for equivalent drop-test conditions. Current headform compliance drop tests for Rugby Union headgear, as defined in the International Rugby Board's performance specifications for players' equipment, result in a 0.3-m high, 14-J impact. The present study confirms that these test conditions may not be representative of the severity of an average concussive event.
In conclusion, 27 cases of medically verified concussion from rugby union and Australian football were reconstructed using numerical simulation. The simulations were able to refine and to add to data obtained from the previously performed video analysis. Grade 1 concussions were associated with mean HIC15 of 230, HIP of 8830 W, and peak linear acceleration of the head's CG of 86g. These results confirm suggested tolerance levels for concussion from other studies. Grade 1 concussions occurred for impacts involving mean equivalent drop-test energies of 39 J and impulses of 22 kg·m·s−1. These values confirm previous findings and should contribute to improve the experimental design of protective headgear testing. Head impacts associated with impulses as high as 40 kg·m·s−1 and impact energies superior to 100 J were measured for more severe impacts, and a common trend existed between the severity grading and the energy transferred to the head. This point, together with the comparison with a similar study in American football, suggests that the energy attenuation provided by a helmet may effectively contribute to reduce the risk and severity of concussion. Further refinement of these values is required through ongoing studies incorporating nonconcussive head impacts as well as expanding the present data set of concussive cases. It is also expected that these results will prove beneficial as an input to more refined models to study not only the risk but also the injury mechanism associated with these impacts.
The results of the present study do not constitute endorsement by ACSM.
1. Abel J, Gennarelli T, Segawa H. Incidence and severity of cerebral concussion
in the rhesus monkey following sagittal plane angular acceleration. In: Proceedings of the 22nd Stapp Car Crash Conference, SAE paper 780886
. Warrendale (PA): SAE; 1978. p. 33-53.
2. Baumgartner D, Willinger R, Shewchenko N, Beusenberg M. Tolerance
limits for mild traumatic brain injury
derived from numerical head
impact replication. In: Proceedings of the IRCOBI Conference
; 2001 Oct 10-12: Isle of Man (UK). Bron (France): IRCOBI; 2001. p. 353-5.
3. Beusenberg M, Shewchenko N, Newman J, DeLange R, Cappon H. Head
, neck, and body coupling in reconstructions of helmeted head
impacts. In: Proceedings of the IRCOBI Conference
; 2001 Oct 10-12: Isle of Man (UK). Bron (France): IRCOBI; 2001. p. 295-310.
4. Brolinson PG, Manoogian S, McNeely D, Goforth M, Greenwald R, Duma S. Analysis of linear head
accelerations from collegiate football impacts. Curr Sports Med Rep
5. Cantu RC, Mueller FO. Brain injury
-related fatalities in American football, 1945-1999. Neurosurgery
6. Cheng H, Obergefell L, Rizer A. The development of the GEBOD program. In: Proceedings of the 1996 fifteenth Southern Biomedical Engineering Conference SBEC-96
; 1996 Mar 29-31: Dayton (OH). New York (NY): IEEE; 1996. p. 251-4.
7. Duma SM, Manoogian SJ, Bussone WR, et al. Analysis of real-time head
accelerations in collegiate football players. Clin J Sport Med
8. Fréchède B, McIntosh A. Use of MADYMO's human facet model
to evaluate the risk of head injury
in impact. In: Proceedings of the 21st ESV conference, paper No. 07-0300
; 2007 Jun 18-21: Lyon (France). Washington DC: NHTSA; 2007. p. 1-14.
9. Gennarelli T, Thibault L, Ommaya A. Pathophysiologic responses to rotational and translational accelerations of the head
. In: Proceedings of the 16th Stapp Car Crash Conference, SAE paper 720970
. Warrendale (PA): SAE; 1972. p. 296-308.
10. Gilchrist A, Mills NJ. Protection of the side of the head
. Accid Anal Prev
11. Gurdjian ES. Recent advances in the study of the mechanism of impact of the head
-a summary. Clin Neurosurg
12. Hagel B, Meeuwisse W. Risk compensation: a side effect of sport injury
prevention? Clin J Sport Med
13. Hodgson V, Thomas L, Gurdjian E, Fernando O, Greenberg S, Chason J. Advances in understanding of experimental concussion
mechanisms. In: Proceedings of the 13th Stapp Car Crash Conference, SAE paper 690796
. Warrendale (PA): SAE; 1969. p. 18-37.
14. King AI, Yang KH, Zhang L, Hardy W. Is head injury
caused by linear or angular acceleration? In: Proceedings of the IRCOBI Conference
; 2003 Sep 25-26: Lisbon (Portugal). Bron (France): IRCOBI; 2003. p. 1-12.
15. Kleiven S, Von Holst H. Consequences of head
size following trauma to the human head
. J Biomech
16. Koh JO, Cassidy JD, Watkinson EJ. Incidence of concussion
in contact sports: a systematic review of the evidence. Brain Inj
17. Margulies SS, Thibault LE. A proposed tolerance
criterion for diffuse axonal injury
in man. J Biomech
18. McCrory P. Neurological injuries in rugby and Australian Rules Football. In: B. Jordan, P. Tsaris, and R. Warren (editors.) Sports Neurology
. Philadelphia (PA): Lippincott Williams & Wilkins; 1998. p. 441-9.
19. McCrory P, Johnston K, Meeuwisse W, et al. Summary and agreement statement of the 2nd International Conference on Concussion
in Sport, Prague 2004. Br J Sports Med
20. McIntosh AS, McCrory P. Impact energy attenuation performance of football headgear. Br J Sports Med
21. McIntosh AS, McCrory P, Comerford J. The dynamics of concussive head
impacts in rugby and Australian Rules Football. Med Sci Sports Exerc
22. McIntosh AS. Risk compensation, motivation, injuries and biomechanics in competitive sports. Br J Sports Med
23. McIntosh AS, McCrory P. Preventing head
and neck injury
. Br J Sports Med
24. Miller RT, Margulies SS, Leoni M, et al. Finite element modeling approaches for predicting injury
in an experimental model
of severe diffuse axonal injury
. In: Proceedings of the 42nd Stapp Car Crash Conference, SAE paper 983154
. Warrendale (PA): SAE; 1998. p. 155-66.
25. Naunheim RS, Standeven J, Richter C, Lewis LM. Comparison of impact data in hockey, football, and soccer. J Trauma
26. Newman J, Barr C, Beusenberg M, et al. A new biomechanical head injury
assessment function-the maximum power index. Stapp Car Crash J
27. Newman J, Barr C, Beusenberg M, et al. A new biomechanical assessment of mild traumatic brain injury
. Part 2: results and conclusions. In: Proceedings of the IRCOBI Conference
; 2000 Sep 20-22: Montpellier (France). Bron (France): IRCOBI; 2000. p. 223-33.
28. O'Riordain K, Thomas PM, Phillips JP, Gilchrist MD. Reconstruction of real world head injury
accidents resulting from falls using multibody dynamics. Clin Biomech (Bristol, Avon)
29. Ono K, Kikuchi A, Nakamura M, Kobayashi H, Nakamura N. Human head tolerance
to sagittal impact-reliable estimation deduced from experimental head injury
using subhuman primates and human cadavers skulls. In: Proceedings of the 24th Stapp Car Crash Conference, SAE paper 801303
. Warrendale (PA): SAE; 1980. p. 379-89.
30. Ono K, Kaneoka K, Hattori S, et al. Cervical vertebral motions and biomechanical responses to direct loading of human head
. In: Proceedings of the IRCOBI Conference
; 2002 Sep 18-20: Munich (Germany). Bron (France): IRCOBI; 2002. p. 235-48.
31. Pellman EJ, Viano DC, Tucker AM, Casson IR, Waeckerle JF. Concussion
in professional football: reconstruction of game impacts and injuries. Neurosurgery
32. Pincemaille Y, Troseille X, Mack P. Some new data related to human tolerance
obtained from volunteer boxers. In: Proceedings of the 33rd Stapp Car Crash Conference, SAE paper 892435
. Warrendale (PA): SAE; 1989. p. 177-90.
33. Ruan JS, Khalil TB, King AI. Finite element modeling of direct head
impact. In: Proceedings of the 37th Stapp Car Crash Conference, SAE paper 933114
. Warrendale (PA): SAE; 1993. p. 69-81.
34. Schnebel B, Gwin JT, Anderson S, Gatlin R. In vivo study of head
impacts in football: a comparison of National Collegiate Athletic Association Division I versus high school impacts. Neurosurgery
35. Shewchenko N, Withnall C, Keown M, Gittens R, Dvorak J. Heading in football. Part 2: biomechanics of ball heading and head
response. Br J Sports Med
. 2005;39(suppl I):i26-i32.
36. TNO. MADYMO human models manual version 6.2.2 [Software manual]. TNO MADYMO, Delft (The Netherlands); May 2005. p. 219.
37. Walilko TJ, Viano DC, Bir CA. Biomechanics of the head
for Olympic boxer punches to the face. Br J Sports Med
38. Withnall C, Shewchenko N, Gittens R, Dvorak J. Biomechanical investigation of head
impacts in football. Br J Sports Med
. 2005;39(suppl I):i49-i57.
39. Zhang L, Ramesh S, Yang KH, King AI. Effectiveness of the football helmet assessed by finite element modeling and impact testing. In: Proceedings of the IRCOBI Conference
; 2003 Sep 25-26: Lisbon (Portugal). Bron (France): IRCOBI; 2003. p. 27-38.
40. Zhang L, Yang KH, King AI. A proposed injury
threshold for mild traumatic brain injury
. J Biomech Eng