Valid detection of self-propelled wheelchair driving with two accelerometers

Area of Science:

• Rehabilitation Engineering
• Biomedical Instrumentation
• Movement Science

Background:

• Accurate monitoring of wheelchair mobility is crucial for individuals with spinal cord injury (SCI).
• Existing methods for detecting wheelchair propulsion may lack precision or practicality.
• Developing a reliable, non-intrusive method for assessing self-propelled wheelchair driving is needed.

Purpose of the Study:

• To assess the validity of a novel method for detecting self-propelled wheelchair driving.
• To evaluate the effectiveness of using two accelerometers and a custom algorithm for activity classification.

Main Methods:

• Ten participants with SCI performed daily activities, including self-propelled wheelchair driving, in a rehabilitation setting.
• Two ActiGraph GT3X+ accelerometers were used: one on the wrist, one on the wheelchair wheel.
• A custom MatLab algorithm analyzed accelerometer data to differentiate propulsion from other activities, validated against video recordings.

Main Results:

• The accelerometer-based method achieved 85% overall agreement, 88% sensitivity, and 83% specificity in detecting self-propelled wheelchair driving.
• Discrepancies were noted during low-speed treadmill driving, sloped treadmill driving, and when being pushed with excessive arm movements.

Conclusions:

• A simple method using two accelerometers and an algorithm can validly detect self-propelled wheelchair driving.
• The system shows high accuracy, with limitations observed in specific atypical daily activities.