Wearable Sensor Data to Track Subject-Specific Movement Patterns Related to Clinical Outcomes Using a Machine...

Area of Science:

• Biomechanics
• Machine Learning
• Wearable Sensor Technology

Background:

• Wearable sensors offer rich human movement data, but clinical integration is limited.
• Objective assessment of gait changes in response to therapy is crucial for knee osteoarthritis (OA) management.

Purpose of the Study:

• To develop and validate a machine learning approach using wearable sensor data to detect subject-specific gait alterations.
• To correlate these gait changes with improvements in clinical outcomes for knee OA patients.

Main Methods:

• Utilized 3D linear acceleration data from wearable sensors on the lower back, thigh, and shank.
• Employed a one-class support vector machine (OCSVM) to establish baseline gait patterns from pre-intervention trials.
• Quantified gait changes by the percentage of outlier strides post-intervention.

Main Results:

• A significant positive correlation (Spearman's ρ = 0.78) was found between the percentage of outlier strides and improved self-reported clinical outcomes.
• The OCSVM successfully identified individual-specific deviations in gait patterns post-intervention.

Conclusions:

• This study presents a novel, proof-of-concept methodology integrating machine learning and wearable sensors for gait analysis.
• The approach offers an objective measure to assess clinically meaningful changes in human movement following exercise therapy in knee OA.