The validation of inertial measurement units in upper limb movement during wheelchair activity

Abstract

Introduction: 3D motion capture is accepted to be the gold standard approach to all data collection for the production of accurate data, however, in recent literature the ecological validity of 3D systems has come into question. This has brought about the use of Inertial measurement units (IMU’s) and the investigation into their accuracy in enabling accurate real life scenario data collection. The depth of research on IMU’s in wheelchair data collection is limited, and at current there is limited attempts at a validation in comparison to an established data collection software. This research aims to validate basic upper body motions seen within wheelchair propulsion in an attempt to identify the current limitations of IMU’s. Methods: numerous IMU placements and calibration stances were investigated for the collection of elbow flexion and shoulder flexion. IMU’s for elbow flexion were placed at the wrist and central on the forearm. For shoulder flexion the IMU was placed 1 cm above the elbow joint. During the calibration phase, standing anatomical position with thumbs forwards, palms outwards, bent elbow and straight arms outwards at 90° for elbow and shoulder flexion, respectively. For both shoulder and elbow flexion, the participant started at a neutral position and moved through to 90° of flexion and returned to starting position. Quintic was used as the validated software. Results: the wrist placed IMU determined elbow flexion more accurately than the forearm placed IMU. Range of motion for both shoulder and elbow flexion were well calculated within several degrees when using the anatomical thumbs forwards calibration pose; with 99.97 and 89.69°, respectively compared to 88.48 and 97.32°. However, the degree of elbow flexion was overestimated with a difference of 27.78°, with the IMU being 91.11° and quintic being 63.33°. Similar results were also seen for the prediction of elbow flexion during the starting phase with a difference of 26.58°. Discussion: with range of motion accurately calculated in comparison to quintic, and shoulder flexion maximum and minimum values also being similar when segment angle was calculated. Then the differences are likely due to error in the calculation of joint angle using a calculation of global coordinate system from the IMU coordinate system during data processing. Therefore, future research should target alternate approaches to data processing in order to reduce the errors seen.