Movement-Based Low Back Pain Subgroups Using Motion Tape Strain Data with Biomechanical and Causal Feature...

Area of Science:

• Biomedical Engineering
• Rehabilitation Science
• Data Science

Background:

• Low back pain (LBP) is a global health issue causing movement impairments.
• Current smart technologies for LBP are often costly and impractical for clinical settings.
• Existing machine learning (ML) methods primarily differentiate LBP patients from healthy individuals, neglecting subgroup identification within the LBP population.

Purpose of the Study:

• To explore the utility of Motion Tape (MT) wearable strain sensors for capturing detailed movement data in individuals with LBP.
• To develop and evaluate a novel feature engineering approach using biomechanical and causal discovery methods for analyzing MT data.
• To identify potential subgroups within the LBP population based on distinct movement and coordination patterns.

Main Methods:

• Utilized six MT sensors on the lower back to collect movement data from 10 LBP participants.
• Engineered features based on biomechanical properties and time-series causal discovery to extract inter-segment coordination patterns.
• Developed a subgroup discovery pipeline using clustering on diverse movement tasks and analyzed clinical characteristics of identified subgroups.

Main Results:

• MT sensor data effectively characterized movement patterns in LBP participants.
• Causal coordination features provided insights into motor control, including asymmetries and restrictions, surpassing amplitude-based methods.
• Preliminary analysis identified three potential LBP subgroups based on biomechanical and coordination patterns, suggesting varied motor control strategies.

Conclusions:

• The developed framework using MT sensors and causal discovery shows promise for identifying meaningful LBP subgroups.
• This approach offers improved interpretability compared to time-series foundation models for LBP subgrouping.
• The methodology has the potential to generalize to other sensor modalities and larger cohorts for enhanced LBP management.