Eyeworld

38 | EYEWORLD | SUMMER 2026 ATARACT C tool for vitreoretinal microsurgery and features active tremor cancellation and force-controlled membrane peeling. Preceyes (Zeiss) allows precise control of surgical instruments inside the eye through a teleoperated robotic arm. The Steady-Hand Eye Robot (Johns Hopkins) and MYNUTIA (KU Leuven) are cooperative systems where the surgeon and robot share control. Dr. Huang discussed the Polaris (Horizon Surgical Systems), which was originally de- signed for partially automated cataract surgery with two robotic arms and multiple automated steps. However, in the first in-human procedure, the system was implemented as a one-handed teleoperated platform controlled by the sur- geon, without automation, highlighting the gap between concept and clinical reality. He also spoke about ForSight Robotics as another player in the cataract robotics space. This system is a teleoperated platform in which the surgeon operates remotely using hand con- trollers, and based on available demonstrations, it appears to adopt a bimanual surgical ap- proach, closer to how surgeons perform cataract surgery today. What's next? Dr. Huang highlighted Xensur and the use of real surgical instruments and high-resolution motion tracking. This enables training of cataract surgery skills, both training surgeons and training robots how expert sur- geons operate. High-quality data is key to the future, he added. Anshu Arundhati, MD, spoke about simu- lator training surgery in her presentation. The old way of learning at the patient's expense is no longer the way forward, she said. Simula- tion-based medical education prioritizes patient safety and improves trainees' skills and surgical outcomes. The COVID-19 pandemic perma- nently accelerated adoption of digital education platforms and the need for high fidelity learn- ings, she said, adding that there's also the work- force shortage to consider. AR/VR is a critical tool for scaling education efficiently. Dr. Anshu said simulation allows for quan- tifiable success, and there have been several papers to show that VR simulation improves basic cataract surgical skills and enhances surgi- cal efficiency. It also helps with error reduction, and studies show reduced intraoperative errors and complication rates and improved surgical confidence, especially in novices. are most important for prediction and capturing complex interactions between many variables. For example, in IOL calculation, axial length might be more significant when considered together with corneal power in certain scenari- os, while in others, its relationship with anterior chamber depth could be more relevant. The attention mechanism automatically learns these contextual relationships. Dr. Whang went on to discuss significant pa- rameters for ELP, noting that since the contribu- tion of each parameter to ELP changes depend- ing on the axial length range, it is challenging to define a single equation that applies across the entire spectrum of axial length. We need an alternative, he said, and one alternative is structural equation modeling. However, even with this approach, the optimal combination of variables and the best-fitting equations change depending on axial length. The attention mechanism is ideal for understanding the relationship between vari- ous ocular parameters and finding the optimal combination. By combining with existing deep learning approaches, we can implement it for IOL power calculations, he said. Dr. Whang went on to describe a study to compare the performance of standard deep learning models (5 types) with models enhanced by various attention mechanisms (30 combinations) in ELP prediction, and to evaluate whether attention-enhanced models demonstrate overfitting tendency. In the study, 1,120 patients underwent conventional cataract surgery at Yeouido St. Mary's Hospital. Axial length, corneal power, anterior chamber depth, lens thickness, and WTW (corneal diameter) were measured by swept-source OCT, and ELP was calculated by vergence formula. Among the 35 deep learning models, 20 showed better results than traditional multiple regression anal- ysis, and of the 30 deep learning models com- bined with attention modules, 22 demonstrated improved performance compared to standard deep learning models. Yu-Hsuan (Alex) Huang, MD, PhD, pre- sented on advances in robotic cataract surgery. Robotic surgery can be classified by the level of control, including robot-assisted tools, teleoper- ated systems, cooperative systems, and partially or fully autonomous systems. He mentioned several products in the robotics space. Micron (Carnegie Mellon University) is a robot-assisted continued from page 37 About the physicians (continued) Vance Thompson, MD Vance Thompson Vision Sioux Falls, South Dakota Woong-Joo Whang, MD The Catholic University of Korea Yeouido St. Mary's Hospital Seoul, South Korea

• Blank Page
• Blank Page (1)