A Machine Learning Approach to Estimate Hip and Knee Joint Loading Using a Mobile Phone-Embedded IMU

Area of Science:

• Biomechanics
• Medical Engineering
• Machine Learning

Background:

• Hip osteoarthritis alters joint loading, impacting patient mobility and exercise prescription.
• Current joint loading measurement methods are lab-dependent, limiting clinical application.
• Objective monitoring of joint loading is crucial for personalized rehabilitation strategies.

Purpose of the Study:

• To develop and evaluate a machine learning model for estimating hip joint loading using only mobile phone sensor data.
• To assess the feasibility of non-invasive, accessible joint loading monitoring outside of clinical laboratory settings.
• To investigate the potential of inertial measurement unit (IMU) data for quantifying joint biomechanics in osteoarthritis.

Main Methods:

• Collected data from 10 hip osteoarthritis patients performing nine exercises.
• Simultaneously recorded 3D motion capture, ground reaction force, and mobile phone IMU data.
• Utilized musculoskeletal modeling to compute ground truth joint loading and trained a machine learning model on phone sensor data.

Main Results:

• The machine learning pipeline achieved a mean absolute error of 29% for the left hip and 36% for the right hip in estimating joint loading for unseen patients.
• Demonstrated the potential of using accelerometer and gyroscope data from a single mobile phone for joint loading estimation.
• Highlighted the current limitations in accuracy for direct clinical application.

Conclusions:

• Estimating hip joint loading using mobile phone sensors is a feasible, albeit currently imprecise, approach for remote patient monitoring.
• Further research, potentially incorporating data from multiple sensor locations, is needed to achieve clinical-grade accuracy.
• This work represents a step towards accessible, real-world biomechanical assessment for osteoarthritis management.