Real-time topology-aware M-mode OCT segmentation for robotic deep anterior lamellar keratoplasty (DALK) guidance

2026-02-02

AI Summary

This work addresses the challenge of real-time, stable segmentation of corneal layer boundaries in M-mode OCT images during robot-assisted deep anterior lamellar keratoplasty (DALK), where noise, signal attenuation, and instrument shadowing often cause boundary ambiguity or discontinuity. To overcome this, the authors propose a lightweight, topology-aware end-to-end segmentation method built upon an enhanced UNeXt architecture, incorporating anatomical topology regularization for the first time to enforce correct layer ordering and continuity. The approach achieves real-time processing at over 80 Hz on a single GPU, demonstrating significantly improved boundary stability and anatomical plausibility on a rabbit eye dataset while maintaining high computational efficiency suitable for clinical deployment.

Technology Category

• Computer Vision: Segmentation
• Intelligent Robots: Multimodal Perception & Sensor Fusion
• Machine Learning: Calibration & Uncertainty Quantification

Application Category

• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• Security and Privacy: Data transparency and provenance
• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs

Abstract

Robotic deep anterior lamellar keratoplasty (DALK) requires accurate real time depth feedback to approach Descemet's membrane (DM) without perforation. M-mode intraoperative optical coherence tomography (OCT) provides high temporal resolution depth traces, but speckle noise, attenuation, and instrument induced shadowing often result in discontinuous or ambiguous layer interfaces that challenge anatomically consistent segmentation at deployment frame rates. We present a lightweight, topology aware M-mode segmentation pipeline based on UNeXt that incorporates anatomical topology regularization to stabilize boundary continuity and layer ordering under low signal to noise ratio conditions. The proposed system achieves end to end throughput exceeding 80 Hz measured over the complete preprocessing inference overlay pipeline on a single GPU, demonstrating practical real time guidance beyond model only timing. This operating margin provides temporal headroom to reject low quality or dropout frames while maintaining a stable effective depth update rate. Evaluation on a standard rabbit eye M-mode dataset using an established baseline protocol shows improved qualitative boundary stability compared with topology agnostic controls, while preserving deployable real time performance.

Problem

• real-time segmentation
• M-mode OCT
• topology-aware
• DALK
• Descemet's membrane

Innovation

• topology-aware segmentation
• M-mode OCT
• real-time guidance
• UNeXt
• anatomical regularization