Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Self-Supervised Training For Low Dose CT Reconstruction

Mehmet Ozan Unal, Metin Ertas, Isa Yildirim

TL;DR

This work tackles high radiation exposure in CT by enabling high-quality low-dose reconstructions without clean reference data through self-supervised training in the projection (sinogram) domain. The method jointly learns a differentiable CT forward/FBP pipeline with a learnable filter $FBP_\phi$ and a denoiser $f_\theta$, optimized via a self-supervised loss that respects the $J$-invariant principle. Three training modes are explored: single sinogram self-supervised, learned single shot, and learned self-supervised, with the latter yielding the strongest performance. Experiments on analytic phantoms and real CT images show improvements over FBP, SART, and BM3D variants in PSNR and SSIM, highlighting the approach’s practical potential and the ability to forego noise-free training data.

Abstract

Ionizing radiation has been the biggest concern in CT imaging. To reduce the dose level without compromising the image quality, low-dose CT reconstruction has been offered with the availability of compressed sensing based reconstruction methods. Recently, data-driven methods got attention with the rise of deep learning, the availability of high computational power, and big datasets. Deep learning based methods have also been used in low-dose CT reconstruction problem in different manners. Usually, the success of these methods depends on labeled data. However, recent studies showed that training can be achieved successfully with noisy datasets. In this study, we defined a training scheme to use low-dose sinograms as their own training targets. We applied the self-supervision principle in the projection domain where the noise is element-wise independent which is a requirement for self-supervised training methods. Using the self-supervised training, the filtering part of the FBP method and the parameters of a denoiser neural network are optimized. We demonstrate that our method outperforms both conventional and compressed sensing based iterative reconstruction methods qualitatively and quantitatively in the reconstruction of analytic CT phantoms and real-world CT images in low-dose CT reconstruction task.

Self-Supervised Training For Low Dose CT Reconstruction

TL;DR

This work tackles high radiation exposure in CT by enabling high-quality low-dose reconstructions without clean reference data through self-supervised training in the projection (sinogram) domain. The method jointly learns a differentiable CT forward/FBP pipeline with a learnable filter and a denoiser , optimized via a self-supervised loss that respects the -invariant principle. Three training modes are explored: single sinogram self-supervised, learned single shot, and learned self-supervised, with the latter yielding the strongest performance. Experiments on analytic phantoms and real CT images show improvements over FBP, SART, and BM3D variants in PSNR and SSIM, highlighting the approach’s practical potential and the ability to forego noise-free training data.

Abstract

Ionizing radiation has been the biggest concern in CT imaging. To reduce the dose level without compromising the image quality, low-dose CT reconstruction has been offered with the availability of compressed sensing based reconstruction methods. Recently, data-driven methods got attention with the rise of deep learning, the availability of high computational power, and big datasets. Deep learning based methods have also been used in low-dose CT reconstruction problem in different manners. Usually, the success of these methods depends on labeled data. However, recent studies showed that training can be achieved successfully with noisy datasets. In this study, we defined a training scheme to use low-dose sinograms as their own training targets. We applied the self-supervision principle in the projection domain where the noise is element-wise independent which is a requirement for self-supervised training methods. Using the self-supervised training, the filtering part of the FBP method and the parameters of a denoiser neural network are optimized. We demonstrate that our method outperforms both conventional and compressed sensing based iterative reconstruction methods qualitatively and quantitatively in the reconstruction of analytic CT phantoms and real-world CT images in low-dose CT reconstruction task.

Paper Structure

This paper contains 7 sections, 3 equations, 3 figures, 1 table.

Table of Contents

  1. Introduction
  2. Related Work
  3. Method
  4. Experiments
  5. Experiment Settings
  6. Results
  7. Conclusion

Figures (3)

  • Figure 1: Proposed working schema for self-supervised low-dose CT reconstruction.
  • Figure 2: From left to right: ground truth, FBP, SART sart1984, SART+TV sart_tv, SART+BM3D bm3D, the proposed method (learned self-supervised).
  • Figure 3: From top left to bottom right: ground truth, FBP, SART sart1984, SART+TV sart_tv, SART+BM3D bm3D, the proposed method (learned self-supervised).