Clinical tool to measure shoulder joint kinematics in an objective and accurate manner using inertial measurement units

The shoulder is the most mobile joint of the body, but also the second most affected articulation by musculoskeletal disorders [1]. With a choice of over 40 different clinical tests, there is little consensus regarding the shoulder evaluation protocol, mostly due to their inherent subjectivity and lack of sensitivity [2,3]. This study aims to develop and study the validation of an objective and accurate tool, based on wearable inertial measurement units (IMUs), that allows a measurement of the humerus orientation. Twenty young healthy participants were recruited. Five participants with any orthopedic shoulder pathology were additionally recruited to evaluate the protocol's feasibility. Each subject was equipped with a set of five IMUs (Physilog, MindMaze SA, CH) installed on the humerus, wrists and the thorax, as well as twenty-three reflective markers on the upper limbs (Qualisys AB, SW). The protocol consisted of calibration, analytical and functional tasks. Three IMU-based models were developed to answer different application needs in terms of complexity and precision (Fig. 1). The kinematics of the humerus segment was measured first with a single humerus IMU (model A). In model B, an additional IMU was placed on the trunk to assess the kinematics of the humerothoracic joint. Finally, the last model (C) made use of the wrist and humerus-mounted IMUs and was based on the segment trajectories in 3D space. All models were compared against the optical-based reference system for shoulder movements in the three anatomical planes. In addition, the B-B score, an asymmetry power-metric score based on two functional movements was computed for all participants [4]. Humeral elevation (abduction-adduction and flexion-extension, average range of motion (ROM): 214°) showed good accuracy with an average error lower than 4° for models A and C, while model B presented larger limits of agreement of the error (Fig. 1). The measurement of axial rotation (average ROM: 141°) was lower in models A and B, with range of motion differences up to 50°. The addition of a wrist sensor (model C) was closer to the reference system for axial rotation with limits of agreement within 5°. B-B scores were 108% ±24% (mean ±std) for the healthy participants and 39% ±33% for the symptomatic participants. Fig. 1: Schematics of the models and corresponding differences ESMAC_HCL_fig.pngDownload : Download high-res image (180KB)Download : Download full-size image In a clinical context where the trunk can be maintained statically, a single humerus sensor is sufficient to measure elevation (model A and C). Model C showed the potential to eliminate soft tissue artifacts for axial rotation measurements. Higher errors in model B might result from the drift of the relative orientation. The B-B score was lower in the symptomatic group, although additional symptomatic participants will be recruited to further assess reliability and sensitivity of the methods.