RGA-Net: A Vision Enhancement Framework for Robotic Surgical Systems Using Reciprocal Attention Mechanisms
Quanjun Li, Weixuan Li, Han Xia, Junhua Zhou, Chi Man Pun, Chen Xuhang
AI summary
Problem
Surgical smoke severely degrades endoscopic video feeds, increasing cognitive load and injury risk, while existing dehazing methods struggle with dense, non-homogeneous smoke and lack paired training data.
Approach
The framework employs a hierarchical encoder-decoder network with a Dual-Stream Hybrid Attention module for local-global feature capture and an Axis-Decomposed Attention module for efficient multi-scale processing, connected by reciprocal cross-gating blocks for bidirectional feature modulation.
Key results
- Achieves state-of-the-art PSNR (30.258 dB) and SSIM (0.945) on DesmokeData
- Sets new benchmarks on LSD3K with PSNR 25.820 dB and SSIM 0.855
- Outperforms general dehazing models and specialized surgical smoke removal networks across all metrics
- Effectively restores fine textures and structural details in dense, non-homogeneous smoke conditions
Why it matters
Enables consistently clear surgical visualization, reducing cognitive burden and iatrogenic injury risks for robotic minimally invasive procedures.
Abstract
Robotic surgical systems rely heavily on high- quality visual feedback for precise teleoperation; yet, surgical smoke from energy-based devices significantly degrades endo- scopic video feeds, compromising the human-robot interface and surgical outcomes. This paper presents RGA-Net (Re- ciprocal Gating and Attention-fusion Network), a novel deep learning framework specifically designed for smoke removal in robotic surgery workflows. Our approach addresses the unique challenges of surgical smoke-including dense, non-homogeneous distribution and complex light scattering-through a hierarchical encoder-decoder architecture featuring two key innovations: (1) a Dual-Stream Hybrid Attention (DHA) module that combines shifted window attention with frequency-domain processing to capture both local surgical details and global illumination changes, and (2) an Axis-Decomposed Attention (ADA) module that efficiently processes multi-scale features through factorized attention mechanisms. These components are connected via reciprocal cross-gating blocks that enable bidirectional feature modulation between encoder and decoder pathways. Extensive experiments on the DesmokeData and LSD3K surgical datasets demonstrate that RGA-Net achieves superior performance in restoring visual clarity suitable for robotic surgery integration. Our method enhances the surgeon-robot interface by providing consistently clear visualization, laying a technical foundation for alleviating surgeons’ cognitive burden, optimizing operation workflows, and reducing iatrogenic injury risks in minimally invasive procedures. These practical benefits could be fur- ther validated through future clinical trials involving surgeon usability assessments. The proposed framework represents a significant step toward more reliable and safer robotic surgical systems through computational vision enhancement.