Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Parallel Diffusion Model-based Sparse-view Cone-beam Breast CT

Wenjun Xia, Hsin Wu Tseng, Chuang Niu, Wenxiang Cong, Xiaohua Zhang, Shaohua Liu, Ruola Ning, Srinivasan Vedantham, Ge Wang

TL;DR

This work tackles high radiation dose in breast CT by introducing a parallel, dual-domain diffusion-model reconstruction framework for sparse-view cone-beam imaging. By splitting the workflow into projection- and image-domain DDPMs trained in parallel on sub-volumes, the method achieves high-resolution reconstructions from half- to one-third-dose data, outperforming U-Net and single-domain diffusion baselines. The approach addresses memory and compute barriers inherent to large 3D diffusion models and is designed for distributed/cloud deployment, with potential extension to photon-counting CT. Overall, the method demonstrates enhanced image quality under reduced dose, highlighting diffusion-type models as a viable path toward safer, more effective breast cancer screening.

Abstract

Breast cancer is the most prevalent cancer among women worldwide, and early detection is crucial for reducing its mortality rate and improving quality of life. Dedicated breast computed tomography (CT) scanners offer better image quality than mammography and tomosynthesis in general but at higher radiation dose. To enable breast CT for cancer screening, the challenge is to minimize the radiation dose without compromising image quality, according to the ALARA principle (as low as reasonably achievable). Over the past years, deep learning has shown remarkable successes in various tasks, including low-dose CT especially few-view CT. Currently, the diffusion model presents the state of the art for CT reconstruction. To develop the first diffusion model-based breast CT reconstruction method, here we report innovations to address the large memory requirement for breast cone-beam CT reconstruction and high computational cost of the diffusion model. Specifically, in this study we transform the cutting-edge Denoising Diffusion Probabilistic Model (DDPM) into a parallel framework for sub-volume-based sparse-view breast CT image reconstruction in projection and image domains. This novel approach involves the concurrent training of two distinct DDPM models dedicated to processing projection and image data synergistically in the dual domains. Our experimental findings reveal that this method delivers competitive reconstruction performance at half to one-third of the standard radiation doses. This advancement demonstrates an exciting potential of diffusion-type models for volumetric breast reconstruction at high-resolution with much-reduced radiation dose and as such hopefully redefines breast cancer screening and diagnosis.

Parallel Diffusion Model-based Sparse-view Cone-beam Breast CT

TL;DR

This work tackles high radiation dose in breast CT by introducing a parallel, dual-domain diffusion-model reconstruction framework for sparse-view cone-beam imaging. By splitting the workflow into projection- and image-domain DDPMs trained in parallel on sub-volumes, the method achieves high-resolution reconstructions from half- to one-third-dose data, outperforming U-Net and single-domain diffusion baselines. The approach addresses memory and compute barriers inherent to large 3D diffusion models and is designed for distributed/cloud deployment, with potential extension to photon-counting CT. Overall, the method demonstrates enhanced image quality under reduced dose, highlighting diffusion-type models as a viable path toward safer, more effective breast cancer screening.

Abstract

Breast cancer is the most prevalent cancer among women worldwide, and early detection is crucial for reducing its mortality rate and improving quality of life. Dedicated breast computed tomography (CT) scanners offer better image quality than mammography and tomosynthesis in general but at higher radiation dose. To enable breast CT for cancer screening, the challenge is to minimize the radiation dose without compromising image quality, according to the ALARA principle (as low as reasonably achievable). Over the past years, deep learning has shown remarkable successes in various tasks, including low-dose CT especially few-view CT. Currently, the diffusion model presents the state of the art for CT reconstruction. To develop the first diffusion model-based breast CT reconstruction method, here we report innovations to address the large memory requirement for breast cone-beam CT reconstruction and high computational cost of the diffusion model. Specifically, in this study we transform the cutting-edge Denoising Diffusion Probabilistic Model (DDPM) into a parallel framework for sub-volume-based sparse-view breast CT image reconstruction in projection and image domains. This novel approach involves the concurrent training of two distinct DDPM models dedicated to processing projection and image data synergistically in the dual domains. Our experimental findings reveal that this method delivers competitive reconstruction performance at half to one-third of the standard radiation doses. This advancement demonstrates an exciting potential of diffusion-type models for volumetric breast reconstruction at high-resolution with much-reduced radiation dose and as such hopefully redefines breast cancer screening and diagnosis.
Paper Structure (10 sections, 4 equations, 4 figures)

This paper contains 10 sections, 4 equations, 4 figures.

Table of Contents

  1. Introduction
  2. Results
  3. Half Dose Results
  4. One-third Dose Results
  5. Discussions and Conclusion
  6. Methods
  7. Conditional DDPM
  8. Parallel Pipeline
  9. Dataset
  10. Training and Inference

Figures (4)

  • Figure 1: Visualization of the three orthogonal planes through another 3D breast CT volume. The ground truth was reconstructed from fully-sampled projection data (300 views) using the FDK algorithm, and the remaining results were reconstructed from half dose data (150 views) using the competing methods respectively. The display window is set to [-100, 550] HU.
  • Figure 2: Visualization of the three orthogonal planes through a 3D breast CT volume. The ground truth was reconstructed from fully-sampled projection data (300 views) using the FDK algorithm, and the remaining results were reconstructed from one-third dose data (100 views) using the competing methods respectively. The display window is set to [-100, 550] HU.
  • Figure 3: Conditional dual-domain DDPM adapted for sparse-view breast CT.
  • Figure 4: Conditional dual-domain DDPM model for sparse-view breast CT image reconstruction.