A calibration-free workflow for image-based mixed reality navigation of total shoulder arthroplasty

Abstract

ABSTRACT Accurate pin placement to guide renaming of the glenoid surface in total shoulder arthroplasty (TSA) is a critical step to restore range of motion in the glenohumeral joint. Achieving proper pin position with free-hand is complicated due to inadequate intra-operative availability of pre-operative planning data. Mixed reality provides a new modality of surgical navigation for arthroplasty without distracting surgeon from the patient anatomy. However, achieving accurate alignment between the pre-operative plan and the intra-operative anatomy is challenging, especially if workflow-compliant solutions are desired. In this paper, we present a calibration-free mixed reality navigation workflow for glenoid pin placement in TSA that contributes to solving some open challenges. We use a visible light stereo camera on a surgical support arm combined with a transformer-based disparity estimation algorithm to accurately reconstruct the 3D anatomy despite surgical lighting. Besides, we introduce a slide-on marker calibrated by design to overcome perceptual difficulties that arise when trying to align real to virtual objects. In a user study on a semi-anthropomorphic phantom, we assess the end-to-end pin placement accuracy using both our system and a clinical reference standard solution that relies on patient-specific drill guides. Averaging over 15 drilled trajectories, our mixed reality system achieved a pin placement accuracy of ( mm, ).