Laboratory Evaluation of Low-Cost Wearable Sensors for Measuring Head Impacts in Sports

Abstract

Advances in low-cost wearable head impact sensor technology provide potential benefits regarding sports safety for both consumers and researchers. However, previous laboratory evaluations are not directly comparable and do not incorporate test conditions representative of unhelmeted impacts. This study addresses those limitations. The xPatch by X2 Biosystems and the SIM-G by Triax Technologies were placed on a National Operating Committee on Standards for Athletic Equipment (NOCSAE) headform with a Hybrid III neck which underwent impact tests using a pendulum. Impact conditions included helmeted, padded impactor to bare head, and rigid impactor to bare head to represent long- and short-duration impacts seen in helmeted and unhelmeted sports. The wearable sensors were evaluated on their kinematic accuracy by comparing results to reference sensors located at the headform center of gravity. Statistical tests for equivalence were performed on the slope of the linear regression between wearable sensors and reference. The xPatch gave equivalent measurements to the reference in select longer-duration impacts, whereas the SIM-G had large variance leading to no equivalence. For the short-duration impacts, both wearable sensors underpredicted the reference. This error can be improved with increases in sampling rate from 1 to 1.5 kHz. Follow-up evaluations should be performed on the field to identify error in vivo.

Laboratory evaluation of a wearable head impact sensor for use in water polo and land sports

“…This finding is consistent with previous evaluations of the SIM-G and other inertial head impact sensors. 15,21,24 Limitations common to inertial sensor technology may contribute to SIM-G inaccuracies reported here. For example, the SIM-G underreported values of PLA and PRV in both headgears, a pattern that is consistent with the insufficiently low sampling rates (41 kHz) used by many head impact sensors.…”

“…Separate linear regressions were computed to quantify the peak kinematic value registered by the SIM-G as a function of reference values from the headform in each headgear. The tolerance for practical equivalence between the SIM-G and reference peak values was set to 0.2, a liberal threshold recommended by a previous evaluation of the SIM-G. 21 SIM-G kinematic values were considered practically equivalent if the 95% CI of the regression slope (B) for the interaction term remained between 0.8 and 1.2. In other words, this analysis sought to determine whether there was a 95% probability that the SIM-G would record a value within 20% of the reference kinematic value.…”

“…For example, the SIM-G underreported values of PLA and PRV in both headgears, a pattern that is consistent with the insufficiently low sampling rates (41 kHz) used by many head impact sensors. 21,25 Noticeable improvements in head impact sensor performance can be attained by increasing sampling rates above 1 kHz. 21 On the other hand, overestimations in PRA by the SIM-G could be due to signal processing limitations: the SIM-G directly measures linear acceleration and rotational velocity, but rotational acceleration values are calculated by differentiating rotational velocity.…”

“…21,25 Noticeable improvements in head impact sensor performance can be attained by increasing sampling rates above 1 kHz. 21 On the other hand, overestimations in PRA by the SIM-G could be due to signal processing limitations: the SIM-G directly measures linear acceleration and rotational velocity, but rotational acceleration values are calculated by differentiating rotational velocity. Therefore, any noise in the recorded rotational velocity signal will inherently be amplified.…”

A Comparison of Youth Flag and Tackle Football Head Impact Biomechanics

“…The SIM-G validation data are inconsistent, with investigations reporting optimal impact detection (100% of non-helmeted impacts detected and 80% of helmeted impacts detected in one study, 21 99% of helmeted impacts detected in another study 22 ) and linear acceleration validity (no difference between headform and SIM-G linear acceleration at low and medium energy impacts in one study, 23 approximately 14.25% error on average in another study 22 ) while a separate investigation reported less than optimal acceleration validity (underpredicted linear and rotational acceleration in helmeted and non-helmeted condition). 24 No on-field head impact biomechanics data collection system is perfect. The SIM-G was chosen for this study because it has the ability to record head impact data from both helmeted and non-helmeted participants.…”