Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Guidance-base Diffusion Models for Improving Photoacoustic Image Quality

Tatsuhiro Eguchi, Shumpei Takezaki, Mihoko Shimano, Takayuki Yagi, Ryoma Bise

TL;DR

This work tackles the challenge of high imaging costs in photoacoustic imaging by enabling high-quality reconstructions from few single-shot acquisitions. It introduces a guidance-based diffusion model that conditions denoising diffusion on low-quality single-shot PA images and uses multi-shot information to steer the reverse process, augmented by a noise-mixing strategy weighted by light-distribution confidence maps. Empirical results on real PA data show that this approach outperforms baselines and ablations in PSNR and SSIM, demonstrating improved vessel fidelity and reduced noise. The method offers potential reductions in acquisition time and patient burden, with applicability to other denoising tasks beyond PA imaging.

Abstract

Photoacoustic(PA) imaging is a non-destructive and non-invasive technology for visualizing minute blood vessel structures in the body using ultrasonic sensors. In PA imaging, the image quality of a single-shot image is poor, and it is necessary to improve the image quality by averaging many single-shot images. Therefore, imaging the entire subject requires high imaging costs. In our study, we propose a method to improve the quality of PA images using diffusion models. In our method, we improve the reverse diffusion process using sensor information of PA imaging and introduce a guidance method using imaging condition information to generate high-quality images.

Guidance-base Diffusion Models for Improving Photoacoustic Image Quality

TL;DR

This work tackles the challenge of high imaging costs in photoacoustic imaging by enabling high-quality reconstructions from few single-shot acquisitions. It introduces a guidance-based diffusion model that conditions denoising diffusion on low-quality single-shot PA images and uses multi-shot information to steer the reverse process, augmented by a noise-mixing strategy weighted by light-distribution confidence maps. Empirical results on real PA data show that this approach outperforms baselines and ablations in PSNR and SSIM, demonstrating improved vessel fidelity and reduced noise. The method offers potential reductions in acquisition time and patient burden, with applicability to other denoising tasks beyond PA imaging.

Abstract

Photoacoustic(PA) imaging is a non-destructive and non-invasive technology for visualizing minute blood vessel structures in the body using ultrasonic sensors. In PA imaging, the image quality of a single-shot image is poor, and it is necessary to improve the image quality by averaging many single-shot images. Therefore, imaging the entire subject requires high imaging costs. In our study, we propose a method to improve the quality of PA images using diffusion models. In our method, we improve the reverse diffusion process using sensor information of PA imaging and introduce a guidance method using imaging condition information to generate high-quality images.

Paper Structure

This paper contains 12 sections, 5 equations, 6 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Prepare a paired dataset (low and high-quality images)
  4. Guidance-based diffusion model for improving photoacoustic image quality
  5. Denoising Diffusion Probabilistic Models
  6. Guidance toward higher quality
  7. Noise mix process of single-shot image-based models
  8. Experiments
  9. Correlation with Confidence Maps
  10. Quantitative evaluation
  11. Qualitative evaluation
  12. Discussion and Conclusion

Figures (6)

  • Figure 1: Left: Overall mechanisim of photoacoustic imaging. Heat map is the confidence map based on the position of light exposure. Right: The upper images are single-shot images (low-quality), and the bottom image is an averaged image (high-quality).
  • Figure 2: Overview of proposed method, which consists guidance toward higher quality images and Noise Mix Process with photoacoustic imaging condition
  • Figure 3: Preliminary experiments with confidence maps, Left: overview of evaluation method, Right: correlation of SSIM and confidence maps
  • Figure 4: Ablation experiments with guidance condition
  • Figure 5: guidance scale-wise SSIM
  • ...and 1 more figures