An instrumented manual wheelchair as a alternative to video analysis for validating locomotion task detection tools

Abstract

Manual wheelchair (MWC) is a constraining form of locomotion for the musculoskeletal system and many MWC users are concerned by musculoskeletal disorders (MSD). Detecting and quantifying MWC related activities are of high interest for the day-to-day follow-up of patients using a MWC as a primary mode of locomotion. Indeed, it could assist clinician in early identification of MSD occurrence or optimizing the user environment (furnishing arrangements, etc). Since few years, several tools combining sensors and data processing algorithms were developed. However, the validation of these tools is difficult, in particular because video analysis used as a reference is both highly time consuming and sometimes inaccurate due to the cameras field of view. Hence, there is a need of an alternative reference method to facilitate the validation of such tools. As an alternative to video analysis, we propose to use an instrumented MWC, initially dedicated for investigating MWC biomechanics, as a new reference method for tasks identifications. This MWC was equipped with handrim dynamometers to measure handrim forces and two angular potentiometers fixed on rear wheels to determine the MWC trajectory. Data processing included data aggregation followed by multichannel symbolic representation based on force and velocity thresholds. Currently, the method was implemented for flat horizontal displacement and allowed distinguishing resting: start-up, propulsion, braking and turning motion various curvature radii. The method was tested on 18 able-bodied subjects during actual displacements. Results show that for most subjects, the method allowed identifying start-up, propulsion, braking and turning motion tasks. However, a specific adaptation of some thresholds was required for 2 subjects. Since the first subject kept his hands on the handrims during the recovery phase, making difficult the separation of the different tasks and the other subject used a desynchronized propulsion pattern between right and left upper-limbs (close to an alternative push strategy), which resulted in a failure of the identification of propulsion cycle. Even if a real validation of this method against video analysis is still required, the results of the method are encouraging. Some works are still needed to include displacements on more various inclined surfaces and uneven terrain.