100
K. H. Yang and H. Mao
high-speed game video recordings were first analysed to determine the relative
speed, direction, and impact sites on player’s heads. Data obtained were then used
to reconstruct the collisions with instrumented crash dummies to measure the three
linear and three angular accelerations. Finally, the kinematic responses were used to
drive a finite element (FE) model of the human brain to predict intracranial distortion
patterns. Viano et al. [97] reported that the strongest correlation of concussion was
with HIC and severity index (SI) measured directly from the Hybrid III dummy
and strain rate in the midbrain and fornix in FE simulations. Although this series
of experimental and computational studies represented the best efforts possible at
that time, many limitations did exist, for example, the validity and accuracy of the
3D reconstructions from the NFL game videos, the rightfulness of using helmeted
Hybrid III dummies to simulate the actual players’ head, and the fact that only one
FE model was used to represent all involved players. Kleiven [50] later extended
the 28 cases reported in the Viano study to include 58 players (25 concussed and
33 uninjured) and one motorcycle accident using the head model developed at
his institution. Despite this effort, the same limitations seen in the Viano study
continued to exist.
Recently, the kinematic data reconstructed at Biokinetics (e.g. [73]) were
found to include some inconsistencies, including the uncertainty of accelerometer
sensitivity due to a faulty accelerometer or amplifier (personal communication with
Professor Jeff Crandall at University of Virginia). Sanchez et al. [87] corrected these
inconsistencies using analytical techniques and found that the median peak angular
velocity for the previously reconstructed concussion cases were 16.6% lower than
that corrected in their study. The authors suggested that any conclusions drawn
previously using the kinematic data reported in the original reconstructions should
be re-examined.
To partially correct the indirect method (i.e. through video analyses and dummy
tests) used to acquire kinematic data described above, a variety of devices have been
invented to quantify the linear and angular accelerations during head impact [101].
In collegiate football, Rowson et al. [82] reported a positive correlation between the
magnitude of linear and angular accelerations using specially arranged accelerometer array imbedded in helmets albeit that the correlation was not statistically
significant. Their data were in contradiction to the statistically significant positive
correlation found in the study by Zhang et al. [110]. The reason for the differences
was that the same impact location was imposed throughout all experiments in the
Zhang’s study, in contrast to the varying locations of impact found in Rowson’s
real-world study.
Despite these efforts, real-world kinematic data collected from American football
players thus far have yet to provide sufficient specificity and sensitivity needed to
predict concussion with high degree of confidence [75]. While the low predictability
could be partially attributed to gender, age, position of the player, etc., inaccuracies
associated with using the helmet-mounted devices [39] could contribute even more
errors to the idea of using FE models driven by kinematic data, which are obtained
from head-/helmet-mounted sensors, to predict the risk of concussion.
K. H. Yang and H. Mao
high-speed game video recordings were first analysed to determine the relative
speed, direction, and impact sites on player’s heads. Data obtained were then used
to reconstruct the collisions with instrumented crash dummies to measure the three
linear and three angular accelerations. Finally, the kinematic responses were used to
drive a finite element (FE) model of the human brain to predict intracranial distortion
patterns. Viano et al. [97] reported that the strongest correlation of concussion was
with HIC and severity index (SI) measured directly from the Hybrid III dummy
and strain rate in the midbrain and fornix in FE simulations. Although this series
of experimental and computational studies represented the best efforts possible at
that time, many limitations did exist, for example, the validity and accuracy of the
3D reconstructions from the NFL game videos, the rightfulness of using helmeted
Hybrid III dummies to simulate the actual players’ head, and the fact that only one
FE model was used to represent all involved players. Kleiven [50] later extended
the 28 cases reported in the Viano study to include 58 players (25 concussed and
33 uninjured) and one motorcycle accident using the head model developed at
his institution. Despite this effort, the same limitations seen in the Viano study
continued to exist.
Recently, the kinematic data reconstructed at Biokinetics (e.g. [73]) were
found to include some inconsistencies, including the uncertainty of accelerometer
sensitivity due to a faulty accelerometer or amplifier (personal communication with
Professor Jeff Crandall at University of Virginia). Sanchez et al. [87] corrected these
inconsistencies using analytical techniques and found that the median peak angular
velocity for the previously reconstructed concussion cases were 16.6% lower than
that corrected in their study. The authors suggested that any conclusions drawn
previously using the kinematic data reported in the original reconstructions should
be re-examined.
To partially correct the indirect method (i.e. through video analyses and dummy
tests) used to acquire kinematic data described above, a variety of devices have been
invented to quantify the linear and angular accelerations during head impact [101].
In collegiate football, Rowson et al. [82] reported a positive correlation between the
magnitude of linear and angular accelerations using specially arranged accelerometer array imbedded in helmets albeit that the correlation was not statistically
significant. Their data were in contradiction to the statistically significant positive
correlation found in the study by Zhang et al. [110]. The reason for the differences
was that the same impact location was imposed throughout all experiments in the
Zhang’s study, in contrast to the varying locations of impact found in Rowson’s
real-world study.
Despite these efforts, real-world kinematic data collected from American football
players thus far have yet to provide sufficient specificity and sensitivity needed to
predict concussion with high degree of confidence [75]. While the low predictability
could be partially attributed to gender, age, position of the player, etc., inaccuracies
associated with using the helmet-mounted devices [39] could contribute even more
errors to the idea of using FE models driven by kinematic data, which are obtained
from head-/helmet-mounted sensors, to predict the risk of concussion.