Introduction:: The potential risk for long-term cumulative neurodegeneration from mild traumatic brain injuries (mTBI) resulting from repetitive brain trauma while soccer heading is still poorly understood. Studies in soccer have found conflicting results on correlation between the head kinematics and mTBI symptoms, and the head loading parameters governing these injuries have remained elusive. Such findings urge us to validate the fidelity of head kinematics measurement in soccer heading.
This study aims to adapt and validate a user-friendly wearable sensor system from an existing, commercially available device, for high fidelity head kinematics time history measurements in soccer. Previous studies have used validated instrumented mouthguards to obtain head kinematics; however, there is a need for alternative wearable sensors when mouthguards are not a feasible option due to comfort or player preference. Therefore, we select a user-friendly headband and optimize its instrumentation and filtering for head kinematics data collection.
We performed a two-phase experimental study to optimize the headband instrumentation and evaluate the kinematics data fidelity. First, in-laboratory soccer ball impacts on an anthropomorphic test dummy (ATD) were performed over a range of impact locations and intensities. The kinematics data was compared against a reference sensor rigidly fixed to the ATD. The data from the lab tests was used to develop the filtering scheme and optimize the sensor number and locations in the headband. The finalized headband was then evaluated on-field against a video photogrammetry reference sensor under typical soccer heading scenarios (throw-ins, goal-kicks, and corner-kicks).
Materials and Methods:: We designed a headband mounted sensor setup to be worn just below the hairline of the forehead for head kinematics measurements during soccer heading. Five Blue Trident (BT) inertial measurement units (IMU) were embedded in a headband and were positioned at the back of the head and in direct contact with the head surface.
The laboratory tests were performed on a 50th percentile male Hybrid-III head and neck, rigidly bolted to a flat elevated surface. A total of 65 impacts were performed at five different locations on the Hybrid-III head (Figure 1A) by aiming a 12 psi soccer ball from the JUGS Soccer Ball Machine at 3 meters distance. Kinematics data from a rigidly attached DTS IMU sensor at the Hybrid-III center of gravity served as the reference data. The frequency distribution of the data at any time, obtained from a continuous wavelet transform, showed the noise patterns and helped create the filtering scheme. The filtered kinematics data was used to optimize the number and locations of the BT sensors in the headband.
Head kinematics data was then obtained from on-field soccer heading using the finalized headband on human subjects and were evaluated against Optotrak (OT) high speed video photogrammetry. Three headers were performed for each type (throw-ins, goal-kicks, and corner-kicks) by aiming a soccer ball towards the head from 13-18 meters distance at varying speeds. This allowed the fidelity of the time-history data to be evaluated under typical on-field conditions. Figure 2 provides the flowchart of the study design.
Results, Conclusions, and Discussions:: In this study, we conducted experimental tests to optimize the headband instrumentation for head kinematics measurements of soccer headers. We found that in laboratory tests, as the relative angle between the sensor and the impact location (∠ϴ) (Figure 1B) increased, the noise in the data decreased. The filtered angular velocity results provided a good match with the reference DTS data when the relative angle between the impact and the sensor locations was greater than 90º. However, filtered translational acceleration results provided a good match only when the relative angle between the impact and the sensor locations was 180º. The frequency distribution of the signal data at any given time was assessed using continuous wavelet transform (wavelet scalogram). There was a clear separation of the noise and signal when comparing the BT and DTS angular velocity wavelet scalogram for high ∠ϴ, allowing for efficient filtering. As the relative angle decreases, the noise region expands and overlaps with the signal, resulting in a lower match between the filtered and the reference data. There was complete overlap between the noise and signal in the translational acceleration wavelet scalograms. For the translational acceleration data, the first order Butterworth filter with 100 Hz cutoff frequency provided a better match with the DTS time history.
For the on-field evaluation of human headers, we found a great match between the Optotrak and BT angular displacement data, but a poor match for the translational displacements data in all nine headers. The translational displacement mismatch from the on-field tests arise from the error that accumulates from the double integration of the translational acceleration over time when the accelerations fall below the BT sensor sensitivity before the head impact. This prevents data capture while the player moves and takes a stance before heading the ball.
The study shows that this instrumented headband can accurately capture rotational kinematics for all impact locations, but can capture the translational kinematics only for the front or front-side impacts.
