This dissertation is dedicated to the exploration and advancement of robotic vitreoretinal surgery (RVS) through the integration of optical coherence tomography (OCT) into surgical instruments. The aim is to optimize depth perception and precise instrument manipulation to enhance patient safety and improve surgical outcomes. Despite the clinical evaluation of instrument-integrated optical coherence tomography (iiOCT), its full potential remains untapped, highlighting a significant research gap and the necessity for the development of effective real-time guidance systems. The focus of this work is on the further development of iiOCT-integrated systems for RVS by utilizing intraoperative geometric retinal models. These models provide real-time support during complex vitreoretinal procedures and are based on 3D-reconstructed point clouds derived from instrument-integrated measurements. The models employ geometric assumptions, such as spherical or ellipsoidal shapes, to approximate the global eye geometry, while precise local modeling techniques describe the relevant area within the scanned region without geometric assumptions. The incorporation of registered preoperative data enhances data density and quality. Intraoperative retinal models serve as sensor safeguards and increase the robustness of distance measurements in the presence of disturbances. They enable supportive functions such as automatic distance control, virtual boundaries, and automated trajectory planning and execution. These functions contribute to consistently high-quality surgical outcomes, minimize the risk of unintended tissue damage, and enhance the efficiency of RVS procedures. Specifically, the contributions of this research encompass four areas: (1) Intraoperative Retinal Modeling: Development of methods for accurately describing the local and global retinal surface based on reconstructed 3D point clouds from sensorized instruments, improving visualization and understanding of the surgical environment. (2) Preoperative-Intraoperative Registration: Introduction of techniques to effectively transform registered preoperative OCT data into the robot's coordinate system, enabling precise navigation to preoperatively planned targets. (3) Model-Improved Sensing: Improvement of the reliability of sensor outputs in the presence of impairments such as blood through the use of geometric models, increasing the robustness of the entire robotic system. (4) Model-Based Robotic Assistance and Automation: Implementation of haptic shared control architectures that include virtual boundaries and automatic distance control to prevent unintended damage to sensitive ocular structures and enhance the safety of surgical procedures. Additionally, methods for model-based planning and automated execution of surgical steps are introduced, significantly reducing the cognitive load on surgeons and increasing efficiency. Comprehensive investigations are conducted in simulations, on model eyes, as well as on ex vivo pig and human eyes. This dissertation makes a significant contribution to the field of surgical assistance research and aims to improve the safety, efficiency, and user-friendliness of robotic systems in vitreoretinal surgery.
Marius Briel (Thu,) studied this question.