RGA-Net: A Vision Enhancement Framework for Robotic Surgical Systems Using Reciprocal Attention Mechanisms

TL;DR

RGA-Net tackles the challenging problem of surgical smoke removal in robotic surgery by introducing a domain-specific, encoder–decoder network that fuses multi-scale features through two novel attention modules. The Dual-Stream Hybrid Attention combines shifted window self-attention with a frequency-domain pathway, while Axis-Decomposed Attention factorizes attention along block and grid axes for efficient reconstruction. Cross-Gating enables bidirectional feature modulation between encoder and decoder, yielding superior multi-scale fusion. Extensive experiments on DesmokeData and LSD3K demonstrate state-of-the-art restoration of visual clarity, with ablations confirming the complementary contributions of DHA, ADA, and Cross-Gating to performance and robustness in non-uniform smoke conditions.

Abstract

Robotic surgical systems rely heavily on high-quality visual feedback for precise teleoperation; yet, surgical smoke from energy-based devices significantly degrades endoscopic video feeds, compromising the human-robot interface and surgical outcomes. This paper presents RGA-Net (Reciprocal 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 further 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.

Figures (4)

• Figure 1: Overview of RGA-Net.
• Figure 2: Comparison of different methods on the DesmokeData dataset.
• Figure 3: Comparison of different methods on the LSD3K dataset.
• Figure 4: Ablation comparisons sampled from the DesmokeData and LSD3K datasets. The first two rows are from DesmokeData and the last two rows are from LSD3K. (settings: case1 = ADA+CrossGating, case2 = DHA+CrossGating, case3 = DHA+ADA).