ARISE Beats Surgeon-Controlled Robots at Retinal Injections
- 19 hours ago
- 2 min read
Subretinal and retinal vein injections require targeting within a few dozen microns.
The retina does not tolerate correction after the fact.
An error of a fraction of a millimeter can cause hemorrhage, retinal detachment, or permanent vision loss.
In January 2026, researchers at the Institute of Automation, Chinese Academy of Sciences, published results from a robotic system that performed this task without a surgeon controlling the instrument.
The System
The system is named ARISE.
It was developed by a team led by researcher Gui-Bin Bian at the National Key Laboratory of Multimodal Artificial Intelligence Systems.
ARISE performed three procedures autonomously: subretinal injection, central retinal vein injection, and branch retinal vein injection.
The system relies on two mechanisms.
The first corrects for spatial misalignment between multiple camera views as the instrument moves through the eye.
The second combines data from multiple sensors, each with different error profiles, into a single positioning estimate.
Together, these allow the system to track instrument position throughout the intraocular space rather than at a single fixed entry point.
Trial Results
The results were published in Science Robotics.
In ex vivo porcine eyes, ARISE achieved 100% success across 20 subretinal injections, 20 central retinal vein injections, and 20 branch retinal vein injections.
In live animal eyes, it achieved 100% success across 16 procedures in each of the same three categories.
Positioning error decreased by 79.87% compared to manual injection.
Positioning error decreased by 54.61% compared to teleoperated robotic surgery, in which a surgeon remains at the controls.
What the Comparison Shows
The comparison to teleoperated surgery is the more significant of the two findings.
Manual technique is limited by tremor, fatigue, and the physical constraints of the human hand.
Teleoperated robotic surgery removes some of that variability but still depends on a surgeon's real-time judgment and control.
ARISE outperformed both.
Limitations
The results come from phantom, ex vivo, and animal models.
No human trial has been conducted.
Regulatory and clinical translation pathways remain unaddressed in this study.
What the data does establish is technical feasibility: a closed perception-to-action loop capable of executing a defined intraocular procedure with measured precision exceeding both manual and teleoperated alternatives.
Retinal disease is common globally. Surgeons trained to treat it are not evenly distributed.
A system with this level of demonstrated precision, if it clears the remaining stages of clinical validation, would address a structural gap rather than a marginal one.
Source: Science Robotics
