Machine-learned-based prediction of lower extremity overuse injuries using pressure plates

Area of Science:

• Biomechanics
• Sports Medicine
• Machine Learning

Background:

• Running offers numerous physical and mental health benefits but carries a risk of lower extremity overuse injuries.
• These injuries have significant physical, psychological, and economic repercussions.
• Existing research often focuses on specific injuries or limited populations, lacking a comprehensive predictive approach.

Purpose of the Study:

• To predict the incidence of lower extremity overuse injuries in first-year students within a 6-month timeframe using foot pressure data.
• To identify specific biomechanical loading features from foot pressure measurements that are predictive of these injuries.

Main Methods:

• Development of a machine learning pipeline to analyze foot pressure data obtained from a pressure plate.
• Utilized foot pressure measurements from both male and female first-year students.
• Extracted predictive features including Fast Fourier Transform (FFT) coefficients and autoregressive (AR) process coefficients.

Main Results:

• The machine learning model achieved an Area Under the Curve (AUC) of 0.639 and a Brier score of 0.201 in predicting overuse injuries.
• Higher pressures on the forefoot were identified as the most significant predictors.
• Both lateral and medial foot areas, along with specific FFT and AR coefficients, were found to be predictive.

Conclusions:

• Foot pressure analysis using machine learning can predict lower extremity overuse injuries in student runners.
• High forefoot pressure is a key indicator, but the identified predictive features lack direct biomechanical interpretability.
• Further research is needed to enhance the practical application of these findings for effective injury prevention strategies.