Estimation of Manual Wheelchair-Based Activities in the Free-Living Environment using a Neural Network Model with Inertial Body-Worn Sensors

This study aimed to develop a method to estimate daily activities of manual wheelchair (MWC) users using a sensor worn on the upper arm. The sensor data was used to classify activities into three categories: non-propulsion activity, MWC propulsion, and static time. A neural network model was developed and validated to accurately identify these activities, potentially helping to understand and prevent shoulder overuse in MWC users.

• 1
  The neural network model achieved high validity in the lab, with measures ≥0.87 for differentiating between non-propulsion activity, propulsion, and static time.
• 2
  A quasi-validation of field data showed acceptable validity (≥0.66) for identifying propulsion, suggesting the model's potential for real-world application.
• 3
  Participants' estimated mean daily time spent in non-propulsion activity, propulsion, and static time ranged from 158–409, 13–25, and 367–609 minutes, respectively.