Abstract Information

P-41

Identifying propulsion and non-propulsion activity in manual wheelchair users with spinal cord injury in their free-living environments using wearable inertial sensors

Fortune E, Cloud B, Madansingh S, Murphree D, Zhao K, Morrow M
Mayo Clinic, Rochester, Minnesota, United states

Title: Identifying propulsion and non-propulsion activity in manual wheelchair users with spinal cord injury in their free-living environments using wearable inertial sensors

Objective: Shoulder pain is the most common site of musculoskeletal pain in manual wheelchair (MWC) users and can significantly limit function and quality of life. Overuse is thought to be a major causal factor. However, little is known regarding MWC users’ cumulative exposure to activities of daily living (ADLs) which pose impingement risk. To better understand the link between MWC use and shoulder impingement, it is vital to objectively classify the type and quantify the frequency of ADLs performed in MWC users’ everyday life. This abstract presents preliminary data illustrating how inertial sensors can be used to identify propulsion and non-propulsion activity in MWC users in their natural environment.

Design/Methods: MWC users with SCI were recruited to this IRB-approved study. For lab-based validation, participants performed multiple trials of MWC-related ADLs: 1. reaching for an object at counter height and overhead, 2. cross-body backpack lifting, 3. transfers, 4. propulsion on a ramp (15 propulsion cycles) and level ground (2 minutes). Acceleration data were acquired at 128 Hz from one Opal inertial measurement unit (IMU; APDM Inc) secured to the lateral left upper arm. Video data were acquired at 60 Hz. Activity and peak detection algorithms, previously developed for able-bodied participants [1], were applied to the acceleration data. Upper arm activity time for all activities and peak detection during propulsion were validated by comparison to video. Activity cycles were defined as the acceleration data between consecutive peaks within activity segments. A neural network model, based on activity cycle feature extraction, was developed and validated from the lab data to classify activity as propulsion or non-propulsion. Participants also wore the upper arm IMU for 1 to 3 days in their natural environments. The lab-validated algorithm was used to estimate participants’ daily arm activity and propulsion time.

Results: 4 MWC users participated (29±6yr, injury levels T3-T6, 1 F). The mean (SD) accuracy of detected active seconds during propulsion and non-propulsion activities was 97 (2) and 83(6)%. Propulsion peak detection accuracy was 100%. Cross-validation yielded an AUC of 0.97 for differentiating between propulsion and non-propulsion activity. The neural network model’s sensitivity, specificity and accuracy were ≥0.95, similar to or greater than previous studies [2]. Participants spent 148 (36) mins of time active each day, with 36 (24) active mins of propulsion which is slightly less time than reported from wheel-mounted devices [3]. However, measuring propulsion activity at the arm is more directly applicable for shoulder injury risk estimation.

Conclusion: Time spent in propulsion and non-propulsion activity can be measured in MWC users in their natural environment using one upper arm IMU. This allows for shoulder overuse tracking to estimate shoulder injury risk.

Support: Mayo Clinic RMRC, on behalf of the Craig H. Neilsen Fund for Spinal Cord Injury Care and Research Honoring Robert D. Brown Jr., M.D. and the NIH (R01 HD84423).


Content

You Must Be Logged In


This Reminders tab will display for your scheduled lectures 15 minutes before start time and for new messages waiting.
Top