Feasibility of image-based augmented reality guidance of total shoulder arthroplasty using microsoft HoloLens 1

Wenhao Gu, Kinjal Shah, Jonathan Knopf, Nassir Navab, Mathias Unberath
Oct 2020

Abstract

Total Shoulder Arthroplasty (TSA) is a shoulder replacement procedure to treat severe rotator cuff deficiency, primarily caused by osteoarthritis in elderly patients. One of the critical factors in reducing postoperative complications is accurate drilling of a centring hole on the glenoid surface at a precise position and orientation. While the drilling path is planned pre-operatively on 3D diagnostic images, the absence of visual guidance during surgery can lead to low reproducibility. In this paper, we present the design and feasibility analysis of a marker-less image-based registration pipeline using the Microsoft HoloLens 1 and its built-in sensors to guide glenoid drilling during TSA. Our solution intra-operatively registers the pre-operative 3D scan to the exposed glenoid surface both with and without occlusion. Our results provide a breakdown of the sources contributing to registration error. In addition to the commonly discussed errors (SLAM-based head tracking, partial overlap etc.), we find that the poor performance of the depth sensing camera becomes a major source of error. We further find that partial overlap between the source and target remains a large concern for registration in high occlusion scenarios. This work begins to characterise the depth sensor error and suggests future work towards image-based augmented reality guidance.