In on-field studies considering impact to athletes, biomechanical parameters are measured to quantify the severity of a head impact and in some cases estimate the risk of head and brain injury. Wearable sensors play an important role in such studies because they are devices that measure head kinematics. Kinematics can be used as inputs to numerical models of the head-brain that estimate tissue strain. These strains have been proposed as metrics on which risk and severity of brain injury can be inferred. Kinematics from wearable sensors have systematic and random errors – they measure head motions that differ from the actual motion of the athlete's head. Kinematic errors will undoubtedly lead to brain strain estimates that differ from strains estimated based on the actual motions of the head. The difference in strains estimated from wearable sensor kinematics, and the kinematics describing the actual head motion, can be considered as strain error. It is not known which kinematic errors explain strain errors best. Knowledge of which kinematic errors explain strain errors is important because:​

1. This knowledge could inform the ongoing improvement of wearable sensors;​ specifically wearable sensors could be refined to minimize the kinematic errors that lead to strain errors;​ and
2. This knowledge could inform researchers conducting on-field studies;​ specifically by quantifying the extent of strain errors that result from imperfect wearable sensors.

While brain injury researchers are relying more and more on finite element brain models, it is essential to try and understand which kinematic errors lead to strain errors.

Football helmet impacts were simulated in laboratory-based experiments using the Hybrid III head and GforceTracker (GFT) mounted football helmets. Impact kinematics from both the Hybrid III and the GFT sensor were collected and used in a finite element brain model (the Simulated Injury Monitor (SIMon)) to calculate the corresponding brain strain response. Errors in brain strain response between the Hybrid III and the GFT data were compared with corresponding input kinematic errors using regression analysis to determine the input error that has the highest coefficient of determination (R2) with the output error. Maximum principal strain (MPS) from both rotationally transformed, and linear and rotationally transformed GFT kinematics to the Hybrid III reference frame were also compared to determine the effect on brain strain calculation. In addition, the distribution of strains predicted by Hybrid III and GFT was examined.

The overarching findings from this study were:​ (1) errors in resultant angular velocity are most explanatory of strain errors;​ and (2) errors in component directions of angular velocity affect the magnitude and the spatial distribution of strains throughout the brain. The results of this study also suggest that linear accelerations do not contribute to SIMon predicted brain strains. Therefore, the complex kinematic transforms that re-express linear accelerations measured on the helmet to a co-ordinate system with the origin at the head center, may be unneeded. Also, through regression of MPS with 99.9%, 99% and 95%-ile strains, it was found that variations in any of the 95%, 99%, and 99.9%-ile strains could explain over 96% of the variations in MPS.

The primary interpretation of these findings is that the accurate measurement of both resultant and component rotational velocity is neccesary to obtain accurate estimates of strain magnitude and strain distribution.