PhySe-RPO: Physics and Semantics Guided Relative Policy Optimization for Diffusion-Based Surgical Smoke Removal

arXiv cs.AI / 3/25/2026

📰 NewsSignals & Early TrendsIdeas & Deep AnalysisModels & Research

共有:

Key Points

The paper introduces PhySe-RPO, a diffusion-based surgical smoke removal framework designed to handle limited paired supervision and real intraoperative variability.
Instead of using deterministic restoration, the method converts restoration into a stochastic policy and optimizes it via Physics- and Semantics-Guided Relative Policy Optimization.
A physics-guided reward enforces illumination and color consistency, while a CLIP-based semantic reward targets smoke-free restorations that preserve anatomical visual concepts.
The approach also adds a reference-free perceptual constraint to improve visual quality while maintaining physical consistency and clinical interpretability across synthetic and real robotic surgical datasets.
Results are reported as robust and principled for diffusion restoration with exploration/trajectory-level refinement and critic-free updates under constrained supervision.

Abstract

Surgical smoke severely degrades intraoperative video quality, obscuring anatomical structures and limiting surgical perception. Existing learning-based desmoking approaches rely on scarce paired supervision and deterministic restoration pipelines, making it difficult to perform exploration or reinforcement-driven refinement under real surgical conditions. We propose PhySe-RPO, a diffusion restoration framework optimized through Physics- and Semantics-Guided Relative Policy Optimization. The core idea is to transform deterministic restoration into a stochastic policy, enabling trajectory-level exploration and critic-free updates via group-relative optimization. A physics-guided reward imposes illumination and color consistency, while a visual-concept semantic reward learned from CLIP-based surgical concepts promotes smoke-free and anatomically coherent restoration. Together with a reference-free perceptual constraint, PhySe-RPO produces results that are physically consistent, semantically faithful, and clinically interpretable across synthetic and real robotic surgical datasets, providing a principled route to robust diffusion-based restoration under limited paired supervision.