Machine learning-based estimation of structure-specific load around the ankle and knee joint during running using IMU...

Area of Science:

• Biomechanics
• Sports Medicine
• Machine Learning

Background:

• Running generates repetitive musculoskeletal loads, increasing the risk of overuse injuries.
• Accurate structure-specific load (SSL) estimation is crucial for injury prevention and rehabilitation.
• Current SSL estimation methods are time-consuming and computationally intensive.

Purpose of the Study:

• To develop and validate a machine learning model for estimating SSL on the Achilles and patellar tendons, and knee and ankle joints during running.
• To assess the feasibility of using data from one or two inertial measurement units (IMUs) for SSL estimation.

Main Methods:

• A long-short-term-memory (LSTM) based model was developed.
• The model was trained and evaluated using data from 43 participants during running.
• Structure-specific loads were estimated using time-series data from IMUs.

Main Results:

• The LSTM model achieved high R-squared scores (0.93 for Achilles tendon force, 0.84 for patellar tendon force, 0.89 for ankle contact force, and 0.86 for knee contact force).
• SSL estimation was found to be practical and reliable using data from a single pelvis-mounted IMU.
• The model accurately estimated SSL during the stance phase of running.

Conclusions:

• Machine learning models applied to IMU data provide a viable solution for practical and reliable SSL monitoring in runners.
• This approach has significant potential for real-world applications in injury prevention and rehabilitation for runners.
• The findings support the use of IMU-based technology for personalized load management in running.