Development of a Head Acceleration Event Classification Algorithm for Female Rugby Union.

Area of Science:

• Sports Medicine
• Biomechanics
• Machine Learning

Background:

• Instrumented mouthguards record head acceleration data in sports.
• Current methods require time-consuming video verification for data accuracy.
• Existing classification algorithms need sport- and sex-specific validation.

Purpose of the Study:

• To develop and validate the first algorithm for classifying head acceleration events in female rugby union players.
• To improve the automatic categorization of genuine and spurious head impacts.
• To enhance the reliability of head impact data collection in female contact sports.

Main Methods:

• 25 female rugby players wore instrumented mouthguards during 6 matches.
• Video recordings were used to label genuine and spurious impact events.
• Four machine learning algorithms were trained and tested on the collected kinematic data.

Main Results:

• A support vector machine algorithm achieved high performance, with an AUROC of 0.92 and AUPRC of 0.85.
• The developed algorithm accurately classified head acceleration events from female rugby players.
• This represents a significant advancement in head impact detection for female athletes.

Conclusions:

• The study presents a novel, validated algorithm for head impact classification in female rugby.
• This technology can contribute to safer participation by improving head impact data reliability.
• The findings support the development of sport-specific, sex-specific head impact monitoring systems.