Poster CC4 - Integrating Edge Computing and AI with Wearable Technology for Health Optimization in Division 1 Female Soccer Athletes

Introduction: Female athletes encounter distinct physiological challenges and elevated injury risks, necessitating tailored methodologies for health monitoring and performance optimization. Wearable technology, such as the WHOOP wrist sensor, facilitates continuous acquisition of real-time physiological data, thereby offering intricate insights into an athlete's condition beyond traditional metrics like heart rate and pulse. This study aims to integrate wearable technology with edge computing to enhance health analytics for Division 1 female soccer players, addressing the current deficit in comprehensive data collection and analysis within women's sports. Utilizing advanced data analytics and machine learning, this research seeks to improve injury prevention, optimize performance, and enhance overall health monitoring for these athletes.

Materials and

Methods: This IRB-approved study (IRB #2113291-3) equipped thirteen Division 1 female soccer athletes with WHOOP wrist sensors from August 2023 to January 2024, generating 11,277 data points. The study targeted physiological metrics including heart rate (HR), resting heart rate (RHR), and strain. Data analysis was conducted using R-Studio and Python, applying Pearson correlation coefficients and linear regression to analyze relationships between training loads and recovery metrics. Generalized logistic models (GLM) and confusion matrices were utilized to construct predictive models for athlete strain and recovery. Durbin-Watson tests assessed residual autocorrelation, and the area under the curve (AUC) measured model accuracy.

Results, Conclusions, and Discussions: The data analysis revealed significant correlations (p < 0.01) between calories burned and strain (0.85), calories and heart rate (0.53), and strain and heart rate (0.5), indicating the substantial impact of training loads on recovery metrics. The GLM identified heart rate and resting heart rate as significant predictors of high strain, with an AIC of 1196.9, suggesting robust model fit. The Durbin-Watson statistic of 1.65 indicated minimal residual autocorrelation. The model achieved an AUC of 0.875, demonstrating high accuracy in differentiating strain levels. The confusion matrix indicated 78.9% accuracy and 77.9% precision, underscoring the model's predictive reliability for athlete strain.

The integration of wearable technology and machine learning offers a transformative approach to health and performance management in women's soccer. This study highlights the potential of wearable sensors to provide comprehensive health analytics that extend beyond traditional metrics, facilitating proactive and personalized athlete management. By employing advanced data analysis techniques, we can develop predictive models that enhance injury prevention and optimize personal training regimens. The findings support a shift towards data-driven strategies in sports science, emphasizing the role of technology in advancing the health and performance of female athletes.

This study demonstrates the feasibility and efficacy of using wearable technology and edge computing for health analytics in women's soccer. The significant correlations and predictive capabilities of the models highlight the importance of incorporating comprehensive data collection and advanced analytics in sports science. Personalized rehabilitation and training protocols, informed by real-time data and predictive models, can significantly improve athlete outcomes. Future research will focus on refining these models, expanding their application to diverse populations, and exploring additional physiological metrics to further enhance the understanding of athlete health and performance. This approach represents a strategic shift in monitoring and managing athlete wellbeing, advocating for the integration of technology in sports to optimize health and performance outcomes.

Acknowledgements (Optional): The authors would like to acknowledge the Lehigh University Department of Athletics, Lehigh University Department of Sports Medicine, and Lehigh University Strength and Conditioning for their continued support.