Real-time tool to layer distance estimation for robotic subretinal injection using intraoperative 4D OCT

Area of Science:

• Ophthalmic microsurgery
• Medical robotics
• Medical image analysis

Background:

• Robotic systems offer enhanced precision for ophthalmic microsurgery, enabling procedures like targeted subretinal injections.
• Accurate distance measurement between surgical tools and retinal layers (ILM, RPE) is critical for robotic interventions.
• Intraoperative optical coherence tomography (iOCT) provides high-resolution, near real-time imaging for surgical guidance.

Purpose of the Study:

• To develop and validate a novel computational pipeline for precise distance estimation between an injection needle and retinal layers (ILM and RPE) using iOCT data.
• To enable real-time feedback for robotic-assisted ophthalmic microsurgery.

Main Methods:

• A convolutional neural network (CNN) was used to segment the needle tip and retinal boundaries (ILM, RPE) from iOCT B-scans.
• 3D surface point clouds were generated from segmentation maps for the tool and retinal layers.
• Minimum distances between the tool point cloud and layer point clouds were calculated.

Main Results:

• The method achieved average errors of 9.24 µm to the ILM and 8.61 µm to the RPE on ex-vivo porcine eyes.
• The system provided distance feedback at an average update rate of 15.66 Hz.
• The approach demonstrated robustness against noise in iOCT B-scans.

Conclusions:

• The developed pipeline accurately and efficiently estimates critical distances for robotic retinal interventions.
• This technology is suitable for real-time feedback in interventional robotic ophthalmic surgery.
• The findings support the advancement of robotic-assisted procedures in ophthalmology.