E-Thesis 86 views 56 downloads
Instrumented Mouthguards: Advancing the Functionality and Reliability / DAVID POWELL
Swansea University Author: DAVID POWELL
DOI (Published version): 10.23889/SUThesis.68812
Abstract
Instrumented mouthguards (IMGs) are wearable devices designed to record kinematic data describing the head’s motion during potentially injurious impacts in sports. Already a popular device, IMGs are set for a surge in popularity following World Rugby’s mandate making them a requirement for all profe...
| Published: |
Swansea University, Wales, UK
2024
|
|---|---|
| Institution: | Swansea University |
| Degree level: | Doctoral |
| Degree name: | Ph.D |
| Supervisor: | Williams, E. M. P., and Arora, H. |
| URI: | https://cronfa.swan.ac.uk/Record/cronfa68812 |
| Abstract: |
Instrumented mouthguards (IMGs) are wearable devices designed to record kinematic data describing the head’s motion during potentially injurious impacts in sports. Already a popular device, IMGs are set for a surge in popularity following World Rugby’s mandate making them a requirement for all professional rugby players. Despite this widespread implementation, numerous aspects of the design may be improved. Specific improvements include stricter measures to ensure the validity of recorded data and the extraction of further information about the recorded data while avoiding time-consuming video review. Machine learning algorithms have been created to address the validation issue, but not for female sports specifically or rugby union. To address this, a dataset was collected from six women’s rugby union matches, resulting in 214 impacts from 480 minutes of play. After training, a machine learning algorithm yielded scores of 0.92 and 0.85 for the area under the receiver operator and precision-recall curves (AUROC/AUPRC) respectively, on test data. This advancement signifies a crucial step in female sports' head impact telemetry, enhancing safety and data reliability in contact sports for women. Secondly, a study used kinematic recordings to create algorithms predicting impact action (ball carrier vs. tackler) and impact type (direct head contact vs. secondary acceleration). Machine learning algorithms achieved 69.4%/0.721 and 65.4%/0.744 macro recall/AUROC scores. With further refinement, this may potentially automate impact analysis, aiding athlete protection. Lastly, methods to reliably report linear acceleration are not fullyunderstood. This was investigated by estimating the linear acceleration at the head’s centre of gravity with measurements from a cohort of 25 (11F) individuals. Substantial differences between maximum and minimum impact values were found. Given the variation in head shape and size between youth, adult males and adult females, this indicates a one-size-fits-all approach will not be appropriate and individualised measurements are required to estimate linear acceleration accurately. |
|---|---|
| Item Description: |
A selection of content is redacted or is partially redacted from this thesis to protect sensitive and personal information. |
| Keywords: |
Concussion, Instrumented Mouthguards, Machine Learning, Female Sport |
| College: |
Faculty of Science and Engineering |
| Funders: |
Zienkiewicz Centre for Computational Engineering grant |