Surgical smoke removal via residual Swin transformer network

Area of Science:

• Medical Imaging
• Computer Vision
• Robotic Surgery

Background:

• Surgical smoke degrades visual quality in robot-assisted minimally invasive surgery (RMIS).
• Maintaining a clear operating field is crucial for surgical precision and patient safety.

Purpose of the Study:

• To develop an effective method for removing surgical smoke from endoscopic videos.
• To improve visual clarity during RMIS procedures.

Main Methods:

• A novel method utilizing Swin transformer for surgical smoke removal.
• Employing convolutional neural networks for shallow feature extraction.
• Utilizing Swin transformer blocks for deep feature extraction and smoke-free image generation.

Main Results:

• The proposed method demonstrates strong performance in quantitative and qualitative experiments.
• Validation in surgical instrument segmentation tasks confirms desmoking effectiveness.
• Outperforms existing state-of-the-art surgical smoke removal techniques on synthetic and real datasets.

Conclusions:

• The method successfully removes surgical smoke, enhancing endoscopic video quality.
• Improves surgical safety by reducing risks associated with poor visualization.
• Facilitates subsequent computer vision tasks like instrument segmentation and surgical automation.