Introduction

Estimations by the Centers for Disease and Prevention show that 300,000 sport-related mild traumatic brain injuries, also known as concussions, were occurring annually in the United States,35 and concussions have become a growing concern around the world in recent years.16 Repetitive concussions can cause long-term neurodegenerative processes among athletes.34 In most grading systems, the identification of concussion is based on symptoms. According to the “Evidence-Based Cantu Grading System for Concussion,” the severity of the concussion can be categorized as one of the following: (1) grade 1 or mild (no loss of consciousness); (2) grade 2 or moderate (loss of consciousness lasting less than 1 min); grade 3 or severe (loss of consciousness lasting more than 1 min or posttraumatic amnesia lasting longer than 24 h).4 Contributing factors can be categorized into (1) the intensity of the impact; (2) the number of impacts; and (3) type of injury: direct impact or acceleration-deceleration. Recent surveys showed that concussions could occur in (1) various sports, (2) at different ages, and (3) among both male and female athletes. For instance, Patel et al.25 reported the occurrence of 189 concussions in the NBA from 1999 to 2018. Also, using instrumented helmets, Wilcox et al.39 recorded 37,411 impacts during three seasons of collegiate ice hockey.

The concussion prevention programs designed to date can be categorized into three main groups: (1) improving helmet design and the manufacturing process; (2) devising new rules to limit the number and severity of head impacts; and (3) educating coaches, athletes, and judges of the importance of concussion prevention.11 Although these programs have been effective in reducing the effects of concussion, they each have limitations. For example, previous research showed that certain protective equipment might prevent superficial head injury, but these items were suboptimal for concussion prevention in sport.33 Also, through an in-lab experimental study with four American football helmets, Cournoyer et al.8 showed that linear acceleration sensed over four (out of six) tested locations of the helmet increased up to 14 g after repetitive impacts. Therefore, secondary measures, such as specific physical training programs, also must be taken into consideration.

Neck strengthening can be considered the primary concussion prevention approach introduced so far.6 For example, neck strengthening programs have been effective in increasing the neck strength in athletes and reducing neck pain in pilots.15,31 Therefore, we hypothesized that regular, customized, self-directed neck strengthening training could reduce the risk of long-term complications of concussion, specifically in youth and female athletes.5,14 For instance, Hislop et al.14 showed that for school-aged male rugby players, those who completed the training program suffered 72% fewer overall match injuries, 72% fewer contact-related injuries, and 50% fewer days lost due to contact injuries. However, to understand the effectiveness of a neck strengthening program, injury metrics are required to quantify the risk of concussion based on measurable metrics before and after training. To this end, several injury metrics have been introduced and reported in the literature, including head injury criterion, peak translational acceleration, peak rotational velocity, Gadd severity index, and brain injury criterion.30 Although previous research successfully showed a correlation between head injury metrics and the severity of the concussion, no attempt has been made to evaluate the effect of neck strengthening using such metrics.

Commonly, muscle strength has been measured using a hand-held dynamometer or myometer. As a cost-effective alternative tool for high schools, Collins et al.6 proposed the application of a hand-held tension scale. It has also been demonstrated that the risk of concussion was directly associated with smaller neck circumference and weaker overall neck strength. Hall et al.12 introduced a custom-designed frame, equipped with a digital force gauge as a standardized tool for isometric neck strength measurement. However, previous works reported inconsistent reliability for these measurement technologies due to the procedural differences like considering the mean value versus the maximum value of the contraction force or different contraction times.12

Moreover, Mihalik et al.18 showed that, because neck strength is a protective factor during head impacts, the static cervical muscle strength measurements might not be able to capture the head accelerations during impacts. They suggested that future studies should measure head impact biomechanics dynamically. Other research has shown that, in contrast to in-lab static measurements, in-field measurement of the head kinematics during impacts was correlated with the risk of concussion. For example, Tierney et al.36 showed the positive effect of isometric neck strength and girth on reducing the head acceleration, correlation coefficients of − 0.48 and − 0.47, respectively, caused by impacts during soccer among male and female young adults. Similarly, Bretzin et al.2 showed that neck girth and strength had negative correlations with linear acceleration (correlation coefficients = − 0.60) and rotational velocity (correlation coefficients = − 0.65) of the head during head impacts of soccer players. However, experimental assessment of the neck response to concussive impacts in a laboratory setup is not ethical, and only three studies have used custom-designed frames to measure head kinematics during sub-concussive head impacts.10,32,37 These studies measured head kinematics using stationary and expensive motion-capture systems only available in clinical motion analysis laboratories. Moreover, the custom-designed frames only simulated direct head impacts and not “body checks.”

The objective of this study was to develop and validate an instrumented device to objectively evaluate the effect of neck strength on concussions caused by successive “body checks” that may occur during American football and ice-hockey, and similar impacts during soccer and other contact sports. For this purpose, we used a novel Concussion Active Prevention Testing Device (CAP Corp, Canada) composed of a custom-designed frame to measure neck strength during a controlled whiplash test (provisional patent7). We equipped this device with inertial measurement units (IMUs) to measure head kinematics during a sub-concussive chest impact in a precise and repeatable manner. Notably, the proposed testing device used IMUs to measure the effect of neck muscle strength on the kinematic response of the head during impacts applied to the chest (not directly to the head). The repeatability and sensitivity of the proposed testing device were evaluated through an experimental study by measuring five standard head injury metrics (HIMs): (1) repeatability of the testing device for HIM measurement was validated via the coefficient of multiple correlations (CMC) and intra-class correlation coefficient (ICC) among repeated trials; and (2) sensitivity of the measured HIMs to neck strength was validated by comparing HIMs of male and female participants. We used the HIMs previously established in the literature to assess the sensitivity and repeatability of the proposed testing device and experimental procedure in recording the head kinematics in response to controlled impact. As such, this study did not aim to evaluate the performance of any HIM.

The remainder of this paper has been organized as follows. “Materials and methods” section provides the details of the proposed testing device, the HIMs used, experimental study, calibration procedure, and repeatability/sensitivity evaluation. Repeatability of the testing device, and its ability to identify sex differences in HIMs, are presented in the “Results” section. Finally, the “Discussion” section highlights the capabilities of the proposed testing device and its comparison with the literature, and limitation/future works.

Materials and Methods

Participants Demographic and Recruitment

The repeatability and sensitivity of the proposed testing device were evaluated through an experimental study with fourteen able-bodied participants (7 males: 35 ± 10 years old, 85 ± 14 kg, 179 ± 7 cm; 7 females: 27 ± 5 years old, 70 ± 8 kg, 167 ± 6 cm). The Research Ethics Board Committee of the University of Alberta approved the study protocol, and written consent was obtained from all participants. Participants were recruited through poster or other means of advertisement.

Concussion Active Prevention Testing Device (CAPTD)

The CAPTD was designed to simulate a controlled sub-concussive chest impact (Fig. 1a). For this purpose, a rope connected to a series of weights at one end was connected to the chest of the participant through a custom-designed harness. When the study coordinator released the lever, an impulsive impact caused by the sudden release of weights was transferred to the participant’s chest via the rope and harness. The intensity of the impact can be increased by adding weights. Also, the delay between releasing the lever and applying the force to the participant’s chest, known as the level of impact anticipation, can be controlled by the slack of the rope. Finally, the direction of the force can be tuned by changing the height of the horizontal adjustable rod.

Figure 1
figure 1

(a) The custom-designed testing device; (b) The local coordinate system defined for the head kinematics measurement using two IMUs attached over the head and under the chin.

Full size image

The head kinematics were measured via two commercially available IMUs (MTws, Xsens Technologies, The Netherlands) attached to the participant’s head (over the head and under the chin) with a head harness. Each IMU included a tri-axial accelerometer (range: ± 16 g) and a tri-axial gyroscope (range: ± 2000 deg s−1) and measured the 3D linear acceleration and 3D angular velocity in a coordinate system shown in Fig. 1b. Additionally, another IMU was attached to the rope to identify the impact instant. All IMUs recorded data with a sampling frequency of 100 Hz (100 samples/second) and resolution of 16-bits/sample synchronously and transferred data wirelessly to a computer. To ensure that the sampling frequency of the IMUs was high enough to capture the head kinematics and avoid aliasing, frequency domain analysis (plotting the power spectral density versus frequency for the raw accelerometer and gyroscope readouts) was performed. The power spectral density graphs showed that almost all of the frequency contents of the signals with the heaviest weight (W3 = 52 kg) were contained in the range of 0 to 20 Hz. Therefore, in line with the Nyquist–Shannon sampling theorem, signals were low-pass filtered using a zero-phase 4th-order digital Butterworth filter with the cut-off frequency of 40 Hz to remove the high-frequency noise and avoid aliasing.

Head Injury Metrics

Five HIMs were calculated using the head kinematics obtained by the head and chin IMUs to evaluate the effectiveness of a customized training program. The following HIMs were selected as they were shown to be predictive metrics of the risk of concussion among young and adult athletes of both genders.18,27,30,38

(1) Range: Difference between maximum and minimum values of a measured translational acceleration or angular velocity of the head center of gravity (CG).

(2) Peak: Maximum of the absolute value of the measurements above.

(3) Gadd severity index (GSI): GSI was introduced by the National Operating Committee on Standards for Athletic Equipment as in Eq. (1):

$${\text{GSI}} = \mathop \smallint \limits^{T} A_{CG,r} \left( t \right)^{2.5} {\text{d}}t,$$
(1)

where ACG,r(t) was the resultant acceleration of the head CG in m s−2 and T was the duration of the ACG,r in seconds.

(4) Head injury criterion (HIC): HIC was introduced by the National Highway Traffic Safety Administration as in Eq. (2)38:

$${\text{HIC}} = \left\{ {\left( {t_{2} - t_{1} } \right)\left[ {\mathop \smallint \limits_{{t_{1} }}^{{t_{2} }} A_{\text{CG,r}} \left( t \right){\text{d}}t/\left( {t_{2} - t_{1} } \right)} \right]^{2.5} } \right\}_{\hbox{max} }$$
(2)

where t1 and t2 were determined to maximize the HIC. In practice, to calculate HIC, we integrated ACG,r(t) in a window of t2 − t1 = 50 ms symmetrically around the peak of ACG,r(t). Both GSI and HIC were developed to assess the risk of injury in football helmet and car crash tests.

Since the head and chin IMUs measured acceleration in their local coordinate system, and not the head CG acceleration, a transformation was required to convert accelerations measured by IMUs to head CG acceleration as in Eq. (3):

$$\overrightarrow {{A_{\text{CG}} }} = \overrightarrow {{A_{\text{IMU}} }} - \vec{\omega } \times \left( {\vec{\omega } \times \vec{r}} \right) - \dot{\vec{\omega }} \times \vec{r}$$
(3)

where \(\overrightarrow {{A_{\text{CG}} }}\) was the head CG acceleration in m s−2, \(\overrightarrow {{A_{\text{IMU}} }}\) was the measured acceleration by the head or chin IMU,\(\vec{\omega }\) and \(\dot{\vec{\omega }}\) were the angular velocity (in rad s−1) and acceleration of the head, respectively, and \(\vec{r}\) was the position vector between head CG and the head or chin IMU. We placed the head and chin IMUs in a way that the head CG was along their y-axis and thus \(\vec{r}\) was along their y-axis. Also, the head CG is located approximately 40% below the vertex.41 Thus, Eq. (3) was simplified as Eq. (4):

$$\begin{aligned} A_{{{\text{CG,}}x}} &= A_{{{\text{IMU,}}x}} - r_{y} \left( {\omega_{x} \omega_{y} - \dot{\omega }_{z} } \right) \hfill \\ A_{{{\text{CG}},y}} &= A_{{{\text{IMU}},y}} - r_{y} \left( {\omega_{x}^{2} + \omega_{z}^{2} } \right) \hfill \\ A_{{{\text{CG}},z}} &= A_{{{\text{IMU,}}z}} - r_{y} \left( {\omega_{y} \omega_{z} + \dot{\omega }_{x} } \right) \hfill \\ \end{aligned}$$
(4)

where subscripts x, y, and z indicate each vector’s components in the frame shown in Fig. 1. Thus, ry was approximated by 40% of skull height (as shown in Fig. 1b). Also, \(\dot{\vec{\omega }}\) calculated by a four-point central difference numerical differentiation.

In addition to ACG,r(t), we calculated the range, peak, GSI, and HIC metrics for head resultant angular velocity (ωr) in the same fashion, since previous research showed that ωr was an accurate and predictive metric for evaluating the risk of injury.13,29 As the whiplash test was conducted in the sagittal plane of the body, we also assessed the head CG acceleration in anterior (ACG,x) and upward (ACG,y) directions and head angular velocity in the sagittal plane (ωs) in addtion to ACG,r and ωr.

(5) Head range of motion (ROM): The range of rotational motion of the head in the sagittal plane (rotation around z-axis according to Fig. 1b) calculated using the orientation provided by the IMU.23 However, as the duration of the test was short (less than 15 s), strap-down integration of the head angular velocity could alternatively be used for orientation estimation.

Experimental Study

To assess the repeatability and sensitivity of the testing device, the protocol of the experiment was designed as follows. After kneeling in the appropriate position according to Fig. 1a and selecting the appropriate weight (will be described later), the rope was used to connect the rod carrying weights to the chest of the participant via the custom-made harness. Then, the participant was asked to lean on the backrest, close their eyes, and tilt his/her head 20° backward and wait for a few seconds until the study coordinator released the lever.

To have the same experimental condition among all participants, the height of the rope was adjusted to make the rope horizontal and insert horizontal force at the very beginning of the impact. To lower the level of impact anticipation, the lever was released randomly after a few seconds. To choose the weights, several preliminary experiments were performed with multiple participants in which head kinematics was measured for various weights. The effect of increasing weight was then evaluated on repeatability and sensitivity of the HIMs. Additionally, to ensure the safety of the participants, an upper limit was considered for the selected weights; see “Limitations and future works” section for more details. According to the results of the preliminary study, the main experiments were performed with three different weights (W1 = 23 kg, W2 = 43 kg, W3 = 52 kg) to assess the effect of impact severity on the head kinematics. For each participant and each weight, three trials were performed to assess the repeatability of the testing device and experimental conditions. The impact instant was detected by applying a threshold (3.3 g, 2.3 g, and 3.2 g for W1, W2, and W3, respectively) to the resultant acceleration of the rope IMU. After impact detection, the IMU readouts were cut for 2 s, [impact-0.5 s, impact + 1.5 s] to capture the head kinematics before, during, and after impact.

IMU Local Frame Calibration

It was not practical to visually align the IMU local frame with the reference coordinate system shown in Fig. 1b. Therefore, to align the IMU local frame with the reference frame, we used a calibration procedure introduced by our research group.21,22 To this end, before the main trials, we asked participants to kneel upright for 10 s and then perform ten consecutive lumbar flexion/extensions. Acceleration readouts during quiet sitting were then used to align the vertical axis of the IMU local frame with y-axis of the reference frame while the planar angular velocity of the IMU during lumbar flexion/extension was utilized to align the frontal and lateral axes of the IMU local frame with the x-axis and z-axis of the reference frame, respectively.21

Frequency Domain Analysis

The power spectrum of the IMU readouts for all trials with the highest weight showed that the sampling frequency of 100 Hz was sufficient for capturing head kinematics during the whiplash test.

Repeatability of the Testing Device

To assess the intra-participant repeatability (repeatability between the three trials for each participant and each weight), two assessments were performed: (1) the CMC1 was calculated for ACG,x, ACG,y, ACG,r, ωs, and ωr time-series among the three trials for each (participant, weight) pair; and (2) the ICC17 was calculated for the HIMs mentioned in “Concussion Active Prevention Testing Device (CAPTD)” section of three trials for each (participant, weight) pair. For the ICC, the degree of absolute agreement for three independent measurements under the fixed levels of the column factor (two-way mixed model, interaction absent) was calculated. The closer the value of CMC or ICC to 1, the more repeatable the device.

Sex Differences in Head Injury Metrics

The HIMs introduced in “Concussion Active Prevention Testing Device (CAPTD)” section were used as indicators of the risk of concussion. According to the literature, the cervical muscle of females was generally weaker than males.9,40 Therefore, we hypothesized that our device must show sex differences in the measured HIMs. To show the sensitivity and effectiveness of the proposed device in capturing head kinematics for each (sensor position, weight) pair, the HIMs were compared between male and female participants. To this end, we used the Jarque–Bera test to verify the normality of the distribution of HIMs (significance level = 5%). Then, we evaluated the equality of the variance for two data sets with normal distributions using the Bartlett test. Finally, based on the tested normality and equality of variance, two-sample t test or Wilcoxon rank-sum test (significance level = 5%) was applied to detect significant differences between HIMs of male and female participants. Inter-participant repeatability of the measured HIMs among male and female participants was compared via a two-sample F-test.

Results

Figure 2a shows the 3D acceleration and angular velocity of the head IMU before and after IMU local frame calibration. Frequency domain analysis of impacts with the heaviest weight, W3, showed that the dominant frequency contents of the acceleration and angular velocity signals lay at frequencies lower than 20 Hz, and therefore, the sampling frequency the IMUs, 100 Hz, was enough to capture head kinematics during impacts, Fig. 3. Also, it should be noted that a higher sampling frequency will be required for capturing the head kinematics during real impacts in contact sports compared to what was used in this study for controlled sub-concussive impacts.

Figure 2
figure 2

(a) 3D acceleration and angular velocity of the head IMU before and after IMU sensor frame calibration; (b) Representative waveforms of ACG,x, ACG,r, ωs, and ωr for one participant, obtained using the raw data of the head IMU for three trials with W3 = 52 kg.

Full size image
Figure 3
figure 3

Power spectral density graph of acceleration and angular velocity of the head IMU in the anterior direction and sagittal plane, respectively, for a representative trial with W3 = 52 kg of one participant (the frequency is plotted on logarithmic scale).

Full size image

Figure 2a shows that after calibration, the cross-talk between IMU axes was reduced such that the gravitational acceleration was sensed only through the vertical axis of the accelerometer during the motionless period in the beginning and that the angular velocity was sensed in the sagittal plane during the impact. Figure 2b shows representative waveforms of ACG,x, ACG,r, ωs, and ωr for three trials and that the head moved in the posterior direction (negative ACG,x and ωs) immediately after impact due to neck and head inertia, and then moved in the anterior direction (positive ACG,x and ωs).

According to Table 1, the lowest CMC for ωs was 0.80 ± 0.07, while very good correlations were obtained for ωr (minimum CMC of 0.88 ± 0.10 for chin IMU). For ACG,x, very good correlations were obtained with the head IMU (minimum CMC of 0.90 ± 0.10), while moderate and good correlations were observed among trials using the chin IMU. For ACG,r, the highest correlation for both head and chin IMUs were obtained for W3, 0.80 ± 0.18 and 0.83 ± 0.12, respectively. Finally, the intra-participant repeatability of the head and chin IMUs were similar. However, for the ACG,x, the Wilcoxon signed-rank test showed significantly higher (p < 0.05) repeatability for head IMU compared to chin IMU for all testing conditions.

Table 1 Mean values ± standard deviation (among all participants) of multiple correlation coefficient (CMC) obtained for three trials of each participant.
Full size table

The color map in Table 2 shows that by increasing the weight, higher ICCs were obtained for HIMs for both head and chin IMUs in general. For the range of head CG acceleration and angular velocity, excellent ICC values (minimum ICC of 0.90) were obtained with both W2 and W3, while for W1, excellent ICC values (minimum ICC of 0.91) were calculated only for ωs, ωr, ACG,y. For the peak of ACG,x, ACG,r, ωs, and ωr signals, excellent ICC values were obtained with both W2 and W3, while similar to the range, excellent ICC values were calculated only for ωs, ωr, ACG,y. For GSI, good and excellent ICC values were obtained among the three trials for both ACG,r and ωr, while for the HIC, the three trials showed moderate ICC values (minimum ICC of 0.73) for ACG,r signal. Finally, the ROM of the head in the sagittal plane was measured with good and excellent repeatability for all testing conditions.

Table 2 Intra-class Correlation Coefficient (ICC) and lower/upper band ICCs [LB, ICC, UP] obtained for head injury metrics including range, peak, GSI, HIC, and ROM between three trials of each participant.
Full size table

According to Table 3 and Fig. 4, significantly lower (p < 0.05) range and peak values were recorded for ACG,x, ωs, and ωr signals for male participants using the head IMU for all testing conditions. Similarly, both GSI and HIC showed significantly lower values for ACG,r and ωr in the head IMU signals, except for W3, where males only tended to have lower HIC values compared to females. Also, Fig. 4k shows that male participants had significantly lower head ROM, measured by both IMUs, in comparison to female participants. Finally, the HIMs obtained from ωr signals of female participants had significantly higher inter-participant variability compared to their male counterparts.

Table 3 Comparison of head injury metrics between male and female participants for each (sensor, weight) pair.
Full size table
Figure 4
figure 4

The head injury metrics (HIM: range, peak, GSI, HIC, and ROM) obtained by the head IMU shown as boxplot for male (M) and female (F) participants and for the weights of 23 kg (W1), 43 kg (W2), and 52 kg (W3). Similar to Table 2, significantly stronger muscle strength in male participants is indicated by † and significantly higher inter-participant repeatability in HIMs in male participants was indicated by ‡.

Full size image

Discussion

Given that head inertia remains constant in adult male athletes, muscular strength and activation time of the cervical muscles would be the only contributing factor in reducing the motion of the head, and thus the risk of concussion. Therefore, there has been an immediate need for concussion prevention programs such as neck strengthening, specifically in youth and female sports, to avoid the occurrence of concussion at an epidemic level. However, as shown by Naish et al.,20 not all neck training programs were successful in increasing neck strength, and the effectiveness of each training program must be evaluated separately. Moreover, Wilcox et al.39 found that concussion prevention strategies need to be sport- and gender-specific, with considerations for team and session type.

Therefore, the present study proposed, for the first time, a testing device to objectively measure the effect of neck strength on concussions caused by successive “body checks.” For this purpose, the proposed testing device simulated impacts applied to the chest (not directly to the head) and simultaneously measured the resulted head motion to characterize the effect of neck muscle strength on the risk of concussion. The primary outcome of this research was the introduction of a testing device with the following advantages over similar counterparts: (1) having high sensitivity and repeatability in measuring various HIMs during sub-concussive impacts; (2) being affordable and accessible for sport facilities; (3) having a simple experimental procedure without the need for an experienced operator.

Repeatability of the Testing Device

Manual measurement of neck strength may result in poor reliability due to procedural differences like considering the mean value versus the maximum value of the contraction force or different contraction times. To address this issue, the developed testing device simulated the same testing condition by automatic measurement of head kinematics during impacts using IMUs. The following parameters can be controlled in a repeatable manner.

  1. 1.

    Impact severity: by using similar weights, reducing the slack of the rope, and asking participants to lean on the backrest, the same impact severity can be simulated.

  2. 2.

    Impact direction: by having a constant height for the adjustable rod, the direction of the force can be kept constant.

  3. 3.

    IMU local frame: by applying the proposed IMU local frame calibration, the effect of IMU attachment and head initial inclination can be minimized.

  4. 4.

    Level of anticipation (cervical muscle onset time): by asking the participants to open/close their eyes or announcing the lever release, the level of anticipation can be controlled.

To assess the repeatability of the proposed testing device, CMC and ICC values were calculated for acceleration and angular velocity waveforms and HIMs, respectively, for the three trials associated with each (participant, weight) pair. Good and very good repeatability was obtained for ωs and ωr signals among the three trials (minimum CMC of 0.80 ± 0.07). While very good repeatability was obtained with head IMU for ACG,x for all weights, ACG,y and ACG,r signals showed moderate and good correlations. Therefore, ACG,x, ωs and ωr time-series can be measured with high repeatability using the proposed testing device. Regarding HIMs, good and excellent repeatability was obtained for all HIMs and testing conditions, except for HIC obtained from ACG,r.

Sex Differences in Head Injury Metrics

Previous research works reported lower neck strength and higher concussion rates in female athletes in comparison to their male counterparts.9,19,40 Therefore, we assessed the ability of our testing device in differentiating between various levels of neck strength by comparing the HIMs between male and female participants. According to Table 3, head IMU was more successful than chin IMU in differentiating between male and female participants; the head IMU detected 38 significant differences out of 45 conditions (15 injury metrics × 3 weights) while the chin IMU only detected 21 significant differences. A possible explanation is that the estimated position vector between each IMU and the head CG had smaller errors for the head IMU. Therefore, we recommend placing the IMU on the top of the head for HIMs measurement with this testing device.

The measurement of the head IMU in Table 3 and Fig. 4 showed that several head kinematic parameters and HIMs were significantly greater in female participants than in male participants. In particular, female participants had significantly greater values in the following HIMs.

  1. 1.

    Range and peak values for the head angular velocity in the sagittal plane (ωs), according to Figs. 4a and 4b;

  2. 2.

    Range and peak values for the head accelerations (ACG,x and ACG,r), according to Figs. 4c to 4f;

  3. 3.

    GSI and HIC, according to Figs. 4g to 4j;

  4. 4.

    Head ROM in the sagittal plane, according to Fig. 4k.

For example, for W3, the [25th, 50th, 75th] percentiles of the peak ωs among all male and female participants were [3.1, 4.4, 4.9] and [4.8, 6.6, 7.2] rad s−1, respectively. Also, the [25th, 50th, 75th] percentiles of the peak ACG,x among all male and female participants were [9.1, 10.3, 12.8] and [13.4, 15.0, 16.9] m s−2, respectively. Notably, since the non-parametric statistical analysis was employed for comparison, the [25th, 50th (median), 75th] percentiles of the data among male and female participants were reported here.

As such, HIMs during the controlled trunk impact conditions, in general, showed a significantly greater risk of concussion in female participants. In addition, higher inter-participant variability in the measured HIMs was observed for female participants. For instance, for W2, the interquartile range of the peak ωs for male and female participants were 1.6 and 2.6 rad s−1, respectively. This indicates a larger variation of the neck muscle activity strategies among female participants compared to their male counterparts that may recommend individual-specific training programs for female athletes. Further analysis is needed for such a recommendation.

Testing the Device Safety

The average linear acceleration associated with concussions, based on Hybrid III crash test dummies simulated using a video game, as reported by Pellman et al.,26 was 98 ± 27 g. More recently, Rowson et al.28 reported an average concussive linear acceleration of 105 ± 27 g collected from football players. Additionally, the latter study showed that in the most conservative case,26 the probability of injury associated with peak head accelerations of 50 g was less than 15%. During our experiments, the average ACG,r values for male and female participants were recorded as 1.9 ± 0.3 g and 2.3 ± 0.5 g, respectively, which are nearly 50 times lower than the average concussive linear acceleration reported in the literature. Also, the peak ACG,r measured with our device was 3.1 g and 3.6 g which was nearly 25 times lower than the average concussive linear acceleration. Therefore, it can be concluded that the simulated sub-concussive impacts with the proposed device testing were safe and the risk of concussion during this experiment was nearly 1–2% based on the most conservative injury risk curve.26 Nonetheless, all the injury risk curves present stochastic models and their precision cannot be guaranteed for all individuals.

Also, the measured peak ACG,x and ACG,r during our tests are comparable to those measured during various daily activities, such as standing, sitting, walking, running, and nodding the head, reported by Bussone et al.3 (near or less than 10 and 20 m s−2 for ACG,x and ACG,r, respectively). However, there are activities such as fast running, jumping jacks, and vertical leaping that showed higher peak linear acceleration. At the same time, by comparing the measured peak ωs during our tests with the resultant angular velocity recorded in the same study, 10 out of 13 activities had a smaller peak of the resultant angular velocity (less than 2 rad s−1), and the remaining 3 had similar values to our studies. Therefore, this device might not be recommended for very frequent testing or for those with history of back or neck pain, brain injury, or other neuromuscular impairments. Additionally, there might be a need for weight (W) adjustment based on body weight, body height, and physical condition, particularly for youth.

To further reduce the risk of concussion or even discomfort, the use of smaller testing weights should be recommended. Table 3 shows that conducting the experiments with W1 = 23 kg and W3 = 52 kg were both sensitive to neck strength differences between our male and female participants at the same level, i.e., in both cases 13 (out of 15) HIMs in Table 3 showed a significant difference between the male and female groups. At the same time, there was no significant difference between the CMC values of waveforms obtained by W1 and W3. Thus, W1 could be considered the testing weight to obtain responsive and repeatable results while reducing the risk of concussion or discomfort during the experiment. Nevertheless, the smallest testing weight that obtains responsive and repeatable results for each individual should be further investigated.

Limitations and Future Works

A number of factors would limit the generalization of the proposed testing device.

  1. 1.

    The repeatability and sensitivity of the proposed testing device were evaluated for fourteen participants only and must be further validated with more participants.

  2. 2.

    The ability of the testing device to differentiate between various levels of neck strength must be assessed by measuring head kinematics before and after a training program.

  3. 3.

    In the present work, weights were kept constant for all participants. However, by selecting weights as a percentage of the body weight, a more reliable comparison could be conducted between youth/adult or male/female.

Also, the application of the proposed testing device could be generalized in the future by modifying the frame structure to simulate impacts from other directions. Furthermore, other wearable sensors such as EMG could be added to the setup to measure other predictive measures of concussion, such as muscle activation response.