Predicting vertical ground reaction force characteristics during running with machine learning.

Area of Science:

• Biomechanics
• Sports Medicine
• Machine Learning

Background:

• Running-related injuries (RRIs) are common and often stem from musculoskeletal load imbalances.
• Ground reaction forces (GRFs) estimate biomechanical load but are typically measured in labs.
• Portable sensors and machine learning offer potential for real-world injury risk monitoring.

Purpose of the Study:

• To develop and evaluate a machine learning model that predicts key vertical GRF characteristics from sacral acceleration.
• To assess the model's accuracy in estimating parameters relevant to running-related injuries.

Main Methods:

• Utilized three-dimensional sacral acceleration data from runners.
• Trained machine learning models to predict vertical GRF parameters: contact time, active peak, impact peak, and impulse.
• Compared model performance against baseline and established methods.

Main Results:

• The models accurately predicted active peak (0.080 BW RMSE), impact peak (0.198 BW RMSE), impulse (0.0073 BWs RMSE), and contact time (0.0101s RMSE).
• The proposed machine learning approach demonstrated superior performance compared to existing methods.
• Achieved a root-mean-squared error of 0.080 BW for active peak, 0.198 BW for impact peak, 0.0073 BWs for impulse, and 0.0101s for contact time.

Conclusions:

• Sacral acceleration can be effectively used with machine learning to predict important GRF metrics during running.
• This method shows promise for non-invasive, real-time monitoring of biomechanical load to help prevent running-related injuries.
• The findings suggest a valuable tool for athletes and clinicians to assess injury risk factors in dynamic running environments.