Stanford Instrumented Mouthguard MiG2.0 for Measuring Head Kinematics

Liu and team (2021) measured head impact kinematics to calculate brain strain and strain rate in American football. They used a wearable device (instrumented mouthguards) to collect data related to head impact kinematics measurements during a specific period of time. Finite element head models were used to calculate brain strain and ascertain the risk of mild traumatic brain injury (mTBI) — a type of brain injury that causes cognitive deficits and changes in neurological function. The rapid rotation of the skull and the rotation of the brain causes a large magnitude of brain strain, leading to tissue pathology.

In this and other similar studies, the link between head kinematics and deformation of the brain is used as a predictor of risk of brain injury. Measuring head kinematics on-field is possible through the use of accelerometers attached to the helmet to obtain measurements of head movement with six degrees of freedom (DoF), such as the Head Impact Telemetry System (HITS), sensor-mounted helmets, skin-attached devices, and ear-mounted devices. To negate concerns on skin softness that impedes the accuracy of skull movement with these devices, researchers prefer instrumented mouthguards in the current study as the upper portion of the mouthguard is rigidly fixed to the skull to detect accurate head movement. Gyroscope and accelerometer equipped mouthguards to obtain on-field head impact data.

A sensor in the wearable device is triggered when the absolute value of the head acceleration crosses a set threshold at the time of a head impact. Six degree of freedom kinematics between the pre and post trigger time of the sensor are recorded. The time window between pre-trigger time and post-trigger time should be long enough to save enough data and identify peak brain strain in an impact. An ideal time window may be achieved as most devices allow adjustments in the device by the manufacturer or the user. Consequently, data is saved, but this requires several milliseconds and only after the commit is the sensor reset and ready to record the data for the next impact.

Certain limitations exist in sensor-based wearable devices that record head kinematics during head impact. First, false positive data obtained from mouthguards (i.e. when they are triggered even though there is no head impact) needs to be eliminated by comparing head impact data from mouthguard to recorded video. Algorithms were used to identify real head impacts. However, the length of time window for measurement of head impact kinematics is not clear. Further, there is much variance between pre-trigger and post-trigger times in wearable devices.

Stanford University Football Team Head Impact Kinematics Study

In the study described in this article, 118 head impacts were recorded by video and compared to those collected by Stanford Instrumented Mouthguard. MiG2.0 devices were used to collect the data from training and games of the Stanford University football team. All 118 impacts were found to be subconcussive (i.e. when no concussion is recorded or diagnosed after the impact). The calculations were obtained in the following manner:

• At the accelerometer, MiG2.0 devices recorded angular velocities and linear accelerations.
• These parameters were filtered by Butterworth low-pass filters.
• Cut-off frequency used was 160 Hertz.
• Angular acceleration was calculated using five-point derivative of angular velocity
• Linear acceleration was transformed from accelerometer to Head CoG
• Parameters for MiG2.0 triggering, pre-trigger time and post-trigger time were t=0, t<0, and t>0.

Data was confirmed using video analysis and neural network classifier. Peaks of head kinematics were recorded by the devices, i.e. angular acceleration and linear acceleration recorded at the head center of gravity (CoG). In addition, peak occurrence times were also extracted. The Kungliga Tekniska Högskolan (KTH) finite element (FE), previously validated by head impact and cadaver experiments was used to measure brain strain. The model consists of the following features:

• The scalp, the skull
• The brain, the meninges
• The cerebrospinal fluid
• Eleven pairs of the largest parasagittal bridging veins
• Head is rigid and moves according to on-field head kinematics
• Brain to be deformed by inertial forces
• Peak values were calculated using the double precision solver

The peak values are calculated using the KTH finite element head model simulations:

• 95th percentile maximal principal strain (95% MPS) peak
• 95th percentile maximal principal strain rate (95% MPSR) peak

Based on these principles, different head individual models were developed with differing head geometries to represent different brain strains:

• Six representative brains were used from the WU-Minn Human Connectome Project (WUM HCP) to understand brain geometry.
• Brain strain is decided by rotation and not translation.
• The coordinates of the KTH model were scaled according to moment of inertia and not the brain mass.
• The scaling was done so that the same moment of inertia was achieved at the X axis (posterior to anterior), Z axis (superior to inferior), and Y axis (left to right) with the brain’s CoG as the origin.

The instrumented mouthguard was triggered due to linear acceleration, collected by the accelerometer, and head kinematics were recorded as an impact event in a time window. In MiG2.0, the accelerometer is located at the incisor so as to avoid the influence of jaw slamming. In devices where the accelerometer was located at the molar, the triggering event is connected to a different set of conditions. It was observed that although pre-trigger and post-trigger times shifted, the length of the time window remained the same in all devices. Researchers adjusted pre-trigger times and post-trigger times and obtained the required results:

• Peak values of angular velocity, angular acceleration, and linear acceleration at CoG were plotted against the peak times.
• Kernel density estimations were plotted at the right and top.
• 95% peak and 5% and 95% peak time were also plotted.
• Peak time for angular and linear acceleration at CoG was close to triggering.
• Angular velocity was dispersive between t=0 and t=50.
• Kinematics of some cases reached peaks much before or after triggering.
• Values of 95% MPS and 95% MPSR of the head impacts, when input into the seven head models, and their waveforms derived, showed considerable variance (multiple spikes, variance among head models, peaks occurred far before or after triggering). The variances could be due to different brain geometry, brain size, and inertia. Collecting additional data and adjusting parameters can incorporate missed injury cases in future.

The MiG2.0 mouthguard designed by the Stanford University research team incorporates most parameters required to accurately measure sub-concussive brain strain and concussions in head impacts. The use of the instrumented mouthguard fitted with an accelerometer is capable of using a sufficient history and modeling impact of head impact in football player heading instances.

References

Liu, Y., Domel, A. G., Cecchi, N. J., Rice, E., Callan, A. A., Raymond, S. J., Zhou, Z., Zhan, X., Zeineh, M., Grant, G., & Camarillo, D. B. (2021). Time Window of Head Impact Kinematics Measurement for Calculation of Brain Strain and Strain Rate in American Football. Annals of Biomedical Engineering, 49, 2791–2804. https://doi.org/10.1007/s10439-021-02821-z