Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Laparoscopic Image Desmoking Using the U-Net with New Loss Function and Integrated Differentiable Wiener Filter

Chengyu Yang, Chengjun Liu

TL;DR

The paper tackles the problem of surgical smoke obscuring laparoscopic imagery and its impact on both surgeons and vision-based tools. It introduces ULW, a UNet-based pipeline that embeds a differentiable Wiener filter and uses a multi-objective loss $L = \alpha L_{MSE} + \beta L_{SSIM} + \gamma L_{perceptual}$ with $\alpha+\beta+\gamma=1$ to balance pixel fidelity and perceptual quality. Evaluations on 961 real paired laparoscopic images show ULW outperforms a baseline U-Net and Pix2Pix across metrics such as $SSIM$, $PSNR$, $MSE$, and $CIEDE-2000$, e.g., $SSIM=0.9907$, $PSNR=33.7061$, $MSE=0.0006$, $CIEDE-2000=1.8159$. The learnable Wiener layer enhances noise suppression and structural fidelity, while the combined loss improves texture realism, indicating potential for real-time smoke removal in surgical settings.

Abstract

Laparoscopic surgeries often suffer from reduced visual clarity due to the presence of surgical smoke originated by surgical instruments, which poses significant challenges for both surgeons and vision based computer-assisted technologies. In order to remove the surgical smoke, a novel U-Net deep learning with new loss function and integrated differentiable Wiener filter (ULW) method is presented. Specifically, the new loss function integrates the pixel, structural, and perceptual properties. Thus, the new loss function, which combines the structural similarity index measure loss, the perceptual loss, as well as the mean squared error loss, is able to enhance the quality and realism of the reconstructed images. Furthermore, the learnable Wiener filter is capable of effectively modelling the degradation process caused by the surgical smoke. The effectiveness of the proposed ULW method is evaluated using the publicly available paired laparoscopic smoke and smoke-free image dataset, which provides reliable benchmarking and quantitative comparisons. Experimental results show that the proposed ULW method excels in both visual clarity and metric-based evaluation. As a result, the proposed ULW method offers a promising solution for real-time enhancement of laparoscopic imagery. The code is available at https://github.com/chengyuyang-njit/ImageDesmoke.

Laparoscopic Image Desmoking Using the U-Net with New Loss Function and Integrated Differentiable Wiener Filter

TL;DR

The paper tackles the problem of surgical smoke obscuring laparoscopic imagery and its impact on both surgeons and vision-based tools. It introduces ULW, a UNet-based pipeline that embeds a differentiable Wiener filter and uses a multi-objective loss with to balance pixel fidelity and perceptual quality. Evaluations on 961 real paired laparoscopic images show ULW outperforms a baseline U-Net and Pix2Pix across metrics such as , , , and , e.g., , , , . The learnable Wiener layer enhances noise suppression and structural fidelity, while the combined loss improves texture realism, indicating potential for real-time smoke removal in surgical settings.

Abstract

Laparoscopic surgeries often suffer from reduced visual clarity due to the presence of surgical smoke originated by surgical instruments, which poses significant challenges for both surgeons and vision based computer-assisted technologies. In order to remove the surgical smoke, a novel U-Net deep learning with new loss function and integrated differentiable Wiener filter (ULW) method is presented. Specifically, the new loss function integrates the pixel, structural, and perceptual properties. Thus, the new loss function, which combines the structural similarity index measure loss, the perceptual loss, as well as the mean squared error loss, is able to enhance the quality and realism of the reconstructed images. Furthermore, the learnable Wiener filter is capable of effectively modelling the degradation process caused by the surgical smoke. The effectiveness of the proposed ULW method is evaluated using the publicly available paired laparoscopic smoke and smoke-free image dataset, which provides reliable benchmarking and quantitative comparisons. Experimental results show that the proposed ULW method excels in both visual clarity and metric-based evaluation. As a result, the proposed ULW method offers a promising solution for real-time enhancement of laparoscopic imagery. The code is available at https://github.com/chengyuyang-njit/ImageDesmoke.

Paper Structure

This paper contains 11 sections, 6 equations, 2 figures, 1 table.

Table of Contents

  1. Introduction
  2. Related Work
  3. A Novel U-Net Deep Learning with New Loss Function and Integrated Differentiable Wiener Filter (ULW) method for Laparoscopic Image Desmoking
  4. The U-Net Backbone
  5. Learnable Wiener Filter
  6. A New Loss Function that Integrates the Pixel, Structural, and Perceptual Properties
  7. Experiment
  8. Experiment Settings
  9. Experimental Results for the subjective evaluation
  10. Experimental Results for the objective evaluation
  11. Conclusion

Figures (2)

  • Figure 1: The system architecture of the proposed ULW method. The U-Net is used as the backbone while a differentiable Wiener filter layer is integrated after the input of the image with smoke.
  • Figure 2: The visual presentation of the desmoking results produced by the base model, the foundational model, and the proposed ULW method, respectively. The first two rows show the paired laparoscopic images with and without smoke, respectively. The third row displays the results of the base model, the fourth row reveals the results of the pix2pix model, and the last row shows the results of the proposed ULW method.