Automatic gait EVENT detection in older adults during perturbed walking

Area of Science:

• Biomechanics
• Gerontology
• Machine Learning

Background:

• Accurate gait event detection is crucial for assessing balance and fall risk in older adults.
• Traditional methods struggle with perturbed walking, limiting gait analysis.
• Automated detection for perturbed walking remains an unmet need.

Purpose of the Study:

• To develop and validate an automated method for gait event detection.
• To address limitations of existing methods in perturbed walking scenarios.
• To improve the assessment of balance control and fall risk in older adults.

Main Methods:

• Utilized a bidirectional gated recurrent unit (Bi-GRU) model.
• Integrated ground reaction force, joint angles, and marker data.
• Tested on 307 healthy older adults during regular and perturbed walking.

Main Results:

• Achieved over 97% accuracy in detecting touchdown and liftoff events.
• Demonstrated a mean error of less than 14 ms for gait event detection.
• Validated the model for both regular and perturbed walking conditions.

Conclusions:

• Kinematic approaches are effective for automated gait event detection.
• The developed method shows potential for real-time gait monitoring in clinical settings.
• Enables personalized fall prevention programs through precise gait analysis.