Towards Deep Learning Guided Autonomous Eye Surgery Using Microscope and iOCT Images

Recent advancements in retinal surgery predict a modern operating room consisting of a surgical robot, a microscope, and intraoperative optical coherence tomography (iOCT). Integrating these tools raises the fundamental question of how to effectively combine them to enable surgical autonomy. In this work, we address this question by developing a unified framework that enables real-time autonomous surgical workflows utilizing these devices. To achieve this, we create a system consisting of the following features: (1) we develop a novel imaging system that integrates the microscope and iOCT in a real-time manner, accomplished by dynamically tracking the surgical instrument via a small iOCT scanning region (e.g. B-scan), enabling real-time integration of both sensors which was previously not possible; (2) we implement various convolutional neural networks (CNN) that automatically detect and segment task-relevant information for surgical autonomy; (3) we enable surgeons to intuitively select goal waypoints within both the microscope and iOCT views through simple mouse-click interactions; (4) and we integrate model predictive control (MPC) for trajectory generation while respecting kinematic constraints to ensure patient safety. We show the utility of our system by automating subretinal injection (SI), a challenging procedure that involves inserting a microneedle beneath the retina at a specific depth for targeted drug delivery. This task is challenging for surgeons due to the requirement for tens-of-micrometers of accuracy and precise depth perception. We validate our system by conducting 30 successful SI trials on pig eyes, achieving mean needle insertion accuracy of 26 micrometers to various subretinal goals and mean duration of 55 seconds. Project website is here: https://sites.google.com/view/eyesurgerymicroscopeoct/home