Related Experiment Video
Updated: Jan 7, 2026
Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography
Published on: November 30, 2022
Attention-aware network with lightness embedding and Hybrid Guided Embedding for laparoscopic image desmoking
Ziteng Liu1, Chenghong Zhang2, Dongdong He2
1School of Life Science and Technology, Harbin Institute of Technology, 2 Yikuang Str., Nangang District, Harbin 150080, China; State Key Laboratory of Robotics and System, Harbin Institute of Technology, 2 Yikuang Str., Nangang District, Harbin, 150080, China.
This study presents a novel desmoking network for computer-assisted surgery, improving laparoscopic image quality by explicitly using smoke distribution. The method effectively removes surgical smoke while preserving image details, outperforming existing techniques.
Area of Science:
- Computer-assisted surgery
- Medical imaging
- Image processing
Background:
- Surgical smoke degrades laparoscopic image quality in computer-assisted surgery.
- Existing smoke removal methods often cause over-desmoking artifacts.
Purpose of the Study:
- To introduce a novel desmoking network that reconstructs smoke-free images by explicitly utilizing smoke distribution information.
- To address limitations of current methods by preventing over-desmoking artifacts and preserving image details.
Main Methods:
- A desmoking network comprising a Smoke Attention Estimator (SAE) and a Hybrid Guided Embedding (HGE) module.
- SAE uses channel-aware position embedding with lightness prior for accurate smoke attention map generation.
- HGE employs convolutional layers and field transformation to generate residual terms, preserving fine details.
Main Results:
- Achieved at least 3.71% improvement in Peak Signal-to-Noise Ratio (PSNR) and 18.75% in Learned Perceptual Image Patch Similarity.
- Attained the lowest Perception-based Image Quality Evaluator score (24.55) on the Cholec80 dataset.
- Demonstrated real-time processing capability at ~174 frames per second with over 40 dB PSNR in smoke-free regions.
Conclusions:
- The proposed desmoking network effectively removes surgical smoke while preserving image details and color.
- The method offers superior performance compared to state-of-the-art techniques in both quantitative metrics and visual quality.
- The network shows strong potential for real-time applications in computer-assisted surgery.
