Gait Phase Identification During Level, Incline and Decline Ambulation Tasks Using Portable Sonomyographic Sensing

Area of Science:

• Biomechanics and Robotics
• Medical Imaging and Signal Processing

Background:

• Clinical use of powered lower-limb assistive devices depends on reliable control strategies.
• Accurate identification of gait phases (stance and swing) is crucial for effective device control.
• Existing sensing methods may add complexity to assistive device systems.

Purpose of the Study:

• To evaluate ultrasound (US) imaging as a novel sensing modality for intuitive control of lower-limb assistive devices.
• To develop and validate a task-independent method for identifying gait phases using US muscle imaging.
• To assess the accuracy and efficiency of US-based gait phase identification.

Main Methods:

• Collected ultrasound images of rectus femoris and vastus intermedius muscles during level, incline, and decline walking.
• Extracted static (e.g., correlation coefficient, structural similarity) and temporal (time-derivatives) features from US images.
• Trained support vector machine classifiers using these features for task-independent gait phase identification.

Main Results:

• Static features alone achieved 88.3% accuracy in task-independent gait phase identification.
• Incorporating temporal features significantly improved accuracy to 92.8% (p < 0.01).
• The developed algorithm operated efficiently at speeds exceeding 100 Hz.

Conclusions:

• Ultrasound imaging provides a viable, non-invasive method for continuous gait phase estimation across various terrains.
• Task-independent approaches using US muscle imaging can reliably identify stance and swing phases.
• This technology offers potential for simpler, more intuitive control strategies, enhancing the clinical relevance of assistive devices.