부산대학교 도서관

Robotic ophthalmic surgery is an emerging technology to facilitate high-precision interventions such as subretinal injection and removing swinging tissues in retinal detachment using microscopy and iOCT. However, locating the instrument tip outside iOCT's range-limited ROI is challenging, especially at the initial target-approaching stage. Meanwhile, due to 2D perspective projection and the lack of depth perception with the required micron precision, current image-based methods cannot effectively navigate the instrument tip's trajectory towards both intra-retinal and above-retinal target points. To address this limitation, we propose using shadows of the instrument tip and target to estimate their relative depth position and optimize the instrument tip's insertion trajectory until it approaches targets within the iOCT's scanning area. Our method achieves a mean depth error of 0.0127 mm for above-retinal targets and 0.3473 mm for intra-retinal targets in the surgical simulator without damaging the retina, triggering subsequent iOCT-dominant micron-precision manipulations. This method also succeeds in the experiment of target approaching on a retina model.