Accurate and reproducible measurement of three-dimensional shoulder kinematics would contribute to better understanding shoulder mechanics, and therefore to better diagnosing and treating shoulder pathologies. Current techniques of 3D kinematics analysis use external markers (acromial cluster or scapula locator) or medical imaging (MRI or CT-Scan). However those methods present some drawbacks such as skin movements for external markers or cost and irradiation for imaging techniques. The EOS low dose biplanar X-Rays system can be used to track the scapula, humerus and thorax for different arm elevation positions. The aim of this study is to propose a novel method to study scapulo-thoracic kinematics from biplanar X-rays and to assess its reliability during abduction in the scapular plane.

This study is based on the EOS™ system (EOS Imaging, Paris, France), which allows acquisition of 2 calibrated, low dose, orthogonal radiographs with the subject standing at 30 to 40° angle of coronal rotation to the plane of one of the X-ray beams, in order to limit superimposition with the ribcage and spine. Seven abduction positions in the scapular plane were maintained by the subjects for 10 seconds, during X-ray acquisition. Between two positions, the subjects returned at rest position. Arm elevations were approximately 0, 10, 20, 30, 60, 90 and 150° (position 1 to 7). Six subjects were enrolled to perform a reproducibility study based on the 3D reconstructions of 2 experienced observers three times each. For each subject, a personalised 3D reconstruction of the scapula was created. The observer digitises clearly visible anatomical landmarks on both stereoradiographs for each arm position. These landmarks are used to make a first adjustment of a parameterised 3D model of the scapula. This provides a pre-personalised model of the subject's scapula which is then rigidly registered on each pair of X-rays until its retroprojection fits best on the contours that are visible on the X-rays. The thorax coordinate system (CS) was built following the ISB (International Society of Biomechanics) recommendations. The CS associated to the scapula was a glenoid centred CS based on the ellipse which fit on the glenoid rim on the 3D model of scapula. Scapular CS orientation and translation in the thorax CS was calculated following a Y,X,Z angle sequence for each position.

Each 3D reconstruction of the scapula was performed in approximately 30 minutes. The most reproducible rotation was upward/downward rotation (along X axis) with a 95% confidence interval (95% CI) from 2.71° to 3.61°. Internal/external rotation and anterior/posterior tilting were comprised respectively between 5.18° to 8.01° and 5.50° to 7.23° (CI 95%). The most reproducible translation was superior-inferior translation (along Y axis) with a 95% CI from 1.22mm to 2.46mm. Translation along X axis (antero-posterior) and Z axis (medio-lateral) were comprised respectively between 2.49mm to 4.26mm and 2.47mm to 3.30mm (CI 95%).

We presented a new technique for 3D functional quantitative analysis of the scapulo-thoracic joint. This technique can be used with confidence; uncertainty of the measures seems acceptable compared to the literature. Main advantages of this technique are the very low dose irradiation compared to the CT-Scan and the possibility to study arm elevation above 120°.