Real-time incremental estimation of retinal surface using laser aiming beam

Abstract

Vitreoretinal surgery procedures demand high precision and have to be performed with limited visualization and access. Using virtual fixtures in conjunction with robotic surgery has the potential to improve the safety and accuracy of these procedures. A cornerstone of many of these virtual fixtures is reconstruction of the retinal surface with respect to the surgical tool, which is difficult to obtain due to the inadequacy of traditional stereo vision techniques in the intact eye. A structured light technique applied using an actuated handheld instrument has been proposed to combat this problem, but it only provides a reconstruction at the start of the procedure; it cannot update it as the eye moves during surgery. We propose updating the initial estimate of the retinal plane a single point at a time, by continued detection of a laser aiming beam in each camera frame, as in the initial structured-light approach. This paper presents the technique and demonstrates it via experiment.