A Deep Learning-Driven Autonomous System for Retinal Vein Cannulation: Validation Using a Chicken Embryo Model.

Area of Science:

• Ophthalmology
• Robotics
• Medical Imaging

Background:

• Retinal vein occlusion (RVO) causes vision loss, and its treatment, retinal vein cannulation (RVC), is technically challenging due to delicate retinal veins.
• Microsurgical RVC requires high precision and tremor-free manipulation, often necessitating advanced tools for improved outcomes.

Purpose of the Study:

• To develop and evaluate an automated robotic system for precise retinal vein cannulation.
• To enhance the accuracy and stability of RVC procedures through robotic assistance and advanced imaging.

Main Methods:

• An automated robotic system was developed, incorporating a top-down microscope and B-scan optical coherence tomography (OCT) for depth sensing.
• Deep learning models were utilized for real-time needle navigation, contact detection, and puncture recognition.
• A chicken embryo model served as a surrogate for human retinal veins to test the system's efficacy.

Main Results:

• The automated system achieved 85% accuracy in detecting needle position and puncture events.
• Significant reductions in navigation and puncture times were observed compared to manual RVC methods.
• The system demonstrated reliable performance in autonomous microsurgical tasks.

Conclusions:

• The integration of advanced imaging and deep learning in robotic systems offers a promising approach for automating microsurgical tasks.
• This technology has the potential to make retinal vein cannulation safer, more precise, and reproducible.
• The developed system paves the way for improved treatments for retinal vein occlusion.