A skin mobility-based estimation of the scapular orientation and its accuracy verification

Abstract

Although the motion of the shoulder joint should be analyzed in the motion analysis of the upper extremity, its detailed analysis is quite difficult, because the shoulder is a compound joint. Particularly, the motion of the scapula, which slides on the thorax, cannot easily be measured by using skin markers secured on the characteristic points on the scapula. To solve this problem, we developed a neural network-based method capable of estimating the scapular orientation from the 3D position of the skin markers measured by means of a motion capture system. As training data given to the network, we used the scapular orientation computed by a 2D/3D registration using a 3D model of subject’s scapula reconstructed from CT images and its fluoroscopic image as well as the 3D position of the skin markers measured simultaneously. Accuracy verification experiments employing 5 healthy volunteers demonstrated that the scapular orientation in a lateral elevation of the upper arm can be estimated in an accuracy of less than 3.0 deg in term of average RMS error.