Retinal vein occlusion (RVO) is a serious eye condition that occurs when a vein in the retina—the light-sensitive tissue at the back of the eye—becomes blocked, leading to partial or complete vision loss. Current treatment options include periodic injections of anti-VEGF drugs to control abnormal blood vessel growth and steroids to reduce inflammation and swelling. However, these interventions do not directly remove the blockage.
Retinal Vein Cannulation Offers a Promising Alternative
A potential treatment for RVO is retinal vein cannulation (RVC), a highly delicate microsurgical procedure. During RVC, surgeons insert an extremely fine needle into the blocked retinal vein to deliver clot-dissolving drugs or medications that regulate abnormal blood vessel growth.
However, because retinal veins are as thin as a human hair, manually inserting a needle with sufficient precision poses major technical challenges. Even minor hand tremors can damage the retina, limiting the widespread use of this approach.
Robotics Could Overcome Human Precision Limits
To address these limitations, robotic systems may assist surgeons by enabling greater precision and stability during retinal microsurgery. By guiding needle placement with sub-millimetre accuracy, robots can potentially reduce the risk of retinal injury and improve procedural success.
Johns Hopkins Develops Autonomous Robotic System
As reported by medicalxpress, researchers at Johns Hopkins University have now developed an autonomous robotic system capable of reliably performing retinal vein cannulation. The system, described in a study published in Science Robotics, uses deep learning algorithms to guide surgical actions.
Specifically, the system analyses images from a surgical microscope along with optical coherence tomography (OCT) scans that provide cross-sectional views of eye tissue. Together, these data streams allow the robot to identify retinal veins and guide needle insertion with exceptional accuracy.
Deep Learning Enables Sub-100 Micron Precision
“This work builds on our long-standing interest in addressing the extreme precision and stability challenges of retinal microsurgery,” said Peiyao Zhang, first author of the study. He noted that retinal vein cannulation requires less than 100-micron accuracy, a level of precision beyond normal human physiological limits.
“The main goal was to demonstrate that combining robotic assistance with deep learning can enable an autonomous surgical workflow with precision and repeatability that are difficult to achieve manually,” Zhang explained.
How the Autonomous System Works
The system integrates two steady-hand eye robots, each designed for retinal microsurgery. One robot holds a microneedle, while the other manipulates a surgical tool. In parallel, three deep learning models track needle motion, interpret imaging data, and plan robotic actions in real time.
The researchers tested the system on ex vivo pig eyes, both stationary and moving vertically to simulate motion caused by breathing in live patients.
High Success Rates in Preclinical Testing
In these experiments, the robotic system successfully completed retinal vein cannulation in 90 per cent of stationary eyes and 83 per cent of moving eyes. Importantly, the system could also detect needle contact and entry into the vein, a critical step for safe drug delivery.
According to Zhang, embedding expert surgical knowledge into deep learning models could eventually allow clinicians without highly specialised training to perform robot-assisted procedures with outcomes comparable to experienced surgeons.
Implications for Future Eye Care
These findings suggest that robotic systems could play a meaningful role in treating RVO. Before clinical adoption, however, the system must undergo further validation in live animal studies and human clinical trials.
“Our results show that a highly delicate retinal surgical procedure can be partially automated in a safe, accurate, and repeatable manner,” Zhang said. “This approach could reduce surgeon workload and variability while improving precision during complex microsurgical tasks.”
Next Steps Toward Clinical Translation
Looking ahead, the research team plans to evaluate the system in live animal models. Ultimately, they aim to translate robot-assisted autonomous retinal vein cannulation into real-world surgical practice.
If successful, this technology could expand treatment options for patients with retinal vein occlusion and mark a major advance in robot-assisted ophthalmic surgery.