A Data-Driven Approach to Estimate Changes in Peak Knee Contact Force With Exoskeleton Assistance

Area of Science:

• Biomechanics
• Rehabilitation Engineering
• Wearable Robotics

Background:

• Knee osteoarthritis (OA) affects millions, causing pain and mobility loss.
• Lower-limb exoskeletons show promise in reducing knee loading for OA management.
• Accurate, real-time estimation of knee joint forces is crucial for developing effective exoskeletons but remains a significant challenge.

Purpose of the Study:

• To develop and validate data-driven models for estimating peak knee contact forces during walking.
• To assess the feasibility of using readily available sensor data (EMG, GRF, knee angle) for this estimation.
• To support the design and optimization of load-reducing exoskeletons for knee OA.

Main Methods:

• Two data-driven models were created to estimate peak knee contact forces in early and late stance phases.
• Models utilized features from electromyography (EMG), ground reaction force (GRF), and knee angle.
• Training data came from healthy adults (N=6) under various exoskeleton assistance conditions; peak forces were derived from EMG-informed musculoskeletal simulations (OpenSim Moco).

Main Results:

• Models achieved 90% accuracy for early-stance and 79% accuracy for late-stance peak force directional changes (>0.1 BW).
• Ground reaction force and knee angle were key predictors in both models.
• Electromyography features captured phase-specific muscle contributions: quadriceps (early stance), plantar flexors (late stance), and hamstrings (both phases).

Conclusions:

• A practical, data-driven method for rapidly estimating changes in peak knee contact force was developed.
• This approach can facilitate rapid intervention testing and human-in-the-loop optimization of exoskeleton assistance.
• The findings contribute to advancing exoskeleton technology for knee load reduction in conditions like knee OA.