Article Text

Download PDFPDF

23 Evaluation of in-ear sensor systems for quantifying head impact exposure in youth football
  1. Stian Bahr Sandmo1,
  2. Andrew S McIntosh2,3,4,
  3. Thor Einar Andersen1,
  4. Inga K Koerte5,6,
  5. Roald Bahr1
  1. 1Department of Sports Medicine, Oslo Sports Trauma Research Center, Norwegian School of Sport Sciences, Norway
  2. 2Federation University Australia, Australia
  3. 3Monash University Accident Research Centre, Monash University, Australia
  4. 4McIntosh Consultancy and Research, Australia
  5. 5Department of Child and Adolescent Psychiatry, Psychosomatic and Psychotherapy, Ludwig-Maximilian University, Germany
  6. 6Department of Psychiatry, Pscyhiatry Neuroimaging Laboratory, Brigham and Women’s Hospital, Harvard Medical School, USA


Introduction Wearable sensor systems may be useful for measuring head-impact exposure. Here, we tested the validity of in-ear sensors developed to improve head coupling.

Methods First, the sensor was mounted to a Hybrid III headform (HIII) and impacted with a linear impactor or football. Peak linear acceleration (PLA), peak rotational acceleration (PRA) and peak rotational velocity (PRV) were obtained from both systems; random and systematic error were calculated using HIII as reference. Then, six youth football players wore sensors and performed a structured training protocol including heading and non-heading exercises; they also completed two regular football sessions. For each accelerative event recorded, PLA, PRA and PRV outputs were compared to video recordings. Receiver operating characteristic curves were used to determine the sensor’s discriminatory capacity in both on-field settings, determining cut-off values for predicting outcomes.

Results For the laboratory tests, the random error was 11% for PLA, 20% for PRA and 5% for PRV, respectively; the systematic error was 11%, 19% and 5%. For the structured training protocol, heading events yielded higher absolute values (PLA=15.6±11.8 g) than non-heading events (PLA=4.6±1.2 g); the area under the curve (AUC) was 0.98 for PLA. In regular training sessions, AUC was >0.99 for PLA. A 9 g cut-off value yielded a positive predictive value of 100% in the structured training protocol, compared to only 65% in regular football sessions.

Conclusion The sensor displayed systematic overestimation with considerable random error. Despite excellent on-field accuracy for discriminating head-impacts from other accelerative events, secondary means of verifying events are still necessary.

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.