Autonomous surgical robots can improve consistency and safety in clinical outcomes, independent of surgeon expertise. Deep anterior lamellar keratoplasty (DALK), a partial-thickness corneal transplant that separates Descemet's membrane (DM) from the anterior stroma, demands high precision to avoid perforating the DM during pneumodissection. We introduce AUTO-DALK, a real-time optical coherence tomography (OCT) sensing robotic system for autonomous vertical needle insertion into deep stromal layers. Unlike conventional horizontal approaches that risk incomplete dissection or DM perforation, AUTO-DALK employs a sensor-guided vertical trajectory to achieve more uniform pneumodissection. The system integrates an OCT fiber optic sensor with an anatomical segmentation model to enable real-time depth perception and closed-loop needle control. AUTO-DALK was evaluated in both ex vivo and in vivo rabbit models against 1) freehand, 2) OCT-guided manual, and 3) teleoperated robotic insertions. Results show superior performance in needle placement, dissection accuracy, completion time, and big bubble formation, demonstrating the feasibility of OCT-guided autonomous robotic execution for partial-thickness keratoplasty.
Wang et al. (Thu,) studied this question.