Observer-based SLAM in robot-assisted eye surgery

Vitreoretinal eye surgery requires ultimate surgeon´s precision and skills. Significant variation exists in the approaches between different surgeons. Procedures are often indicated as difficult to perform. The surgeon´s personal skill level is considered critical, including hand motion stability and left/right hand dexterity. Precision of instrument manipulation is a key requirement and a limiting factor in such procedures. The collaboration between high-precision research at the Eindhoven University of Technology and a spin-off company PRECEYES has produced a working prototype for teleoperation robot-assistance during vitreoretinal surgery. This PRECEYES Surgical System enables the surgeon to achieve a higher level of precision and decrease the time span of procedures. Incorporating feedback to the surgeon in the PRECEYES Surgical System will further exploit the benefits of robotassistance. For this purpose, information about the environment is required. On the basis of detected vein junctions from camera images, a mapping method has been developed for mapping and tracking the retina during surgery. Simultaneous Localization And Mapping (SLAM) is a well-known method in mobile robotic applications. It is used to navigate through an unknown environment, while at the same time building a map of the environment. This method has been adapted to be used on a stationary sensor, the camera, in a dynamical environment, the eye. This is called observer-based SLAM. It decouples relative movement of individual vein junctions from movement of the eye. In this way, higher accuracy can be achieved, and an estimate can be provided for parts of the eye that have not been updated recently. The benefits and potential of this method are demonstrated in simulations as well as experiments.