Development of a Wheelchair Propulsion Monitoring Device

Principal Investigator: Dan Ding, Ph.D.

Co-Investigators: Shivayogi Hiremath, MS; Annmarie Kelleher, MS, OTR/L, ATP

Funding Source: Paralyzed Veterans of America

2007-2009

The goal of this project is to develop and validate an unobtrusive wheelchair propulsion monitoring device along with a context-aware software package capable of recognizing different wheelchair propulsion patterns and quantifying independent mobility through self-propulsion and upper limb usage for wheelchair propulsion and activities of daily living (ADL).

To date, we have made the following progress.

· The device development was completed. The device integrates three sub-devices including an e-watch worn by the user that monitors upper limb movements, a wheel rotation datalogger clipped on the wheel that collects mobility information, and a vibration datalogger attached underneath the wheelchair seat that monitors wheelchair acceleration.

· The context-aware software package was also partially completed. The software was able to classify wheelchair population patterns with two machine learning algorithms and we were able to achieve average classification accuracies in the range of 60-90% depending on surface type (high resistance surface such as carpet and low resistance surface such as tile). the results from three testing pilots showed he classification accuracy over 80%. The software also can distinguish self-propulsion from external pushing based on the e-watch on the wrist and the wheelchair mounted accelerometer. The result across the three subjects was relatively consistent with average accuracies above 80% for both surfaces.

· We developed an algorithm that can estimate temporal parameters of wheelchair propulsion including stroke frequency, propulsion time, and recovery time based on hand acceleration. This was not part of the original proposal, but we discovered that wearable sensors can actually provide us more valuable information. The estimates of the temporal parameters based on hand acceleration were compared with those based on the SMART^Wheel. The results revealed high intraclass correlation coefficients (ICC) for two surfaces tested (>0.75).

· We are currently in the process of modifying the IRB protocol which will allow us to collect data from subjects in the National Veterans Wheelchair Games (NVWG) in July.

The purpose of the subject testing at the NVWG is to validate the software for activity recognition. Subjects will perform a series of consecutive activities including some daily activities of living such as transfer, preparing meal, doing laundry, computer work etc., and wheelchair propulsion on different terrains including ramps and cross-slopes.

We conducted subject testing in the National Veterans Wheelchair Games in July 2009.

The follow-up study on using the activity monitor to evaluate wheelchair propulsion and training was funded through the RERC on Spinal Cord Injury and will start next year. In addition, we will use the device to collect real-life upper extremity usage data and examine the relationship with upper extremity pain and injury.