OCT-guided Robotic Subretinal Needle Injections: A Deep Learning-Based Registration Approach.

Area of Science:

• Ophthalmic surgery
• Robotics in medicine
• Medical imaging

Background:

• Subretinal injection (SI) is crucial for treating vitreoretinal disorders but faces challenges due to retinal fragility and surgeon limitations like hand tremor.
• Advancements in optical coherence tomography (OCT) provide micron-level visualization of retinal structures, essential for precise surgical guidance.

Purpose of the Study:

• To develop and evaluate a novel OCT-guided robotic steering framework for precise subretinal injections.
• To enable surgeons to plan and execute SI procedures with enhanced accuracy and control.

Main Methods:

• Integration of optical coherence tomography (OCT) imaging with robotic control for intraoperative registration.
• Utilizing affine transformations, robot kinematics, and a deep neural network for tool-tip localization within OCT volumes.
• Validation through cadaveric pig eye procedures and an aluminum target board experiment.

Main Results:

• The developed OCT-robot registration pipeline demonstrated successful integration of imaging and robotic control.
• Subretinal targeting in pig eyes yielded promising results with a mean Euclidean error of 23.8 micrometers.
• The framework facilitates automated execution of planned surgical trajectories.

Conclusions:

• The novel OCT-guided robotic steering framework offers a significant advancement for precise subretinal injections.
• This technology has the potential to improve outcomes for patients with vitreoretinal disorders.
• The system's accuracy suggests feasibility for clinical application in ophthalmic surgery.