Chaymae Acherki

 

This thesis falls within the field of Human-Computer Interaction (HCI) and focuses on interaction with head-mounted Augmented Reality (AR) displays. AR is a promising solution in surgery, overlaying real-time guidance information directly within the surgeon’s field of view. The application domain of this thesis is arthroscopic surgery, a minimally invasive technique widely used to treat joint pathologies such as anterior cruciate ligament injuries. Conventional arthroscopic techniques rely on mechanical guides and impose constraints such as complex gestures, restricted visibility, and non-intuitive hand-eye coordination, which can limit accuracy and increase surgical difficulty. Although prior work has demonstrated the technical feasibility of AR-assisted surgery, most studies have focused on system accuracy, with limited attention to interface design and user needs. This thesis adopts a human-centered approach to AR surgical guidance, starting from the needs of the surgeons who would use it. To this end, semi-structured interviews and a survey were first conducted with surgeons to identify their information needs, the perceived limitations of conventional mechanical guides, and their expectations of an AR guidance system. The finding that surgeons disagreed about where guidance information should be placed within the surgical scene, together with the fact that limited studies in the literature have addressed the placement of AR guidance information, also referred to as spatial anchoring, led us to investigate different AR guidance placement strategies. We therefore conducted an empirical study of three spatial anchoring strategies to examine the effect of AR guidance placement on task accuracy, completion time, and user preferences. From this study, we derive practical design guidelines for AR guidance systems. Building on these results and the preferences expressed in the survey, an AR guidance prototype was then evaluated against the conventional mechanical guide, comparing objective performance measures, such as task accuracy and completion time, and subjective measures such as cognitive load. This comparative study demonstrates the benefit of AR guidance, particularly for novice surgeons. Like most AR guidance systems, this prototype relies on tracking the surgical instruments. However, this tracking can introduce small errors whose nature and magnitude are unpredictable and depend on many factors. Rather than seeking to eliminate these errors, the final part of this work investigates four approaches to visualizing tracking uncertainty, showing that making uncertainty visible can improve task performance, user trust, and user confidence. In conclusion, this thesis contributes to the design of AR surgical interfaces by addressing both the human factors and technical performance, which together shape their clinical usability. It provides empirical findings and concrete design recommendations to improve the usability of AR guidance in surgical procedures, in order to increase the accuracy of the surgical gesture.