Training low dose CT denoising network without high quality reference data

Abstract

Objective. Currently, the field of low-dose CT (LDCT) denoising is dominated by supervised learning based methods, which need perfectly registered pairs of LDCT and its corresponding clean reference image (normal-dose CT). However, training without clean labels is more practically feasible and significant, since it is clinically impossible to acquire a large amount of these paired samples. In this paper, a self-supervised denoising method is proposed for LDCT imaging. Approach. The proposed method does not require any clean images. In addition, the perceptual loss is used to achieve data consistency in feature domain during the denoising process. Attention blocks used in decoding phase can help further improve the image quality. Main results. In the experiments, we validate the effectiveness of our proposed self-supervised framework and compare our method with several state-of-the-art supervised and unsupervised methods. The results show that our proposed model achieves competitive performance in both qualitative and quantitative aspects to other methods. Significance. Our framework can be directly applied to most denoising scenarios without collecting pairs of training data, which is more flexible for real clinical scenario.

MM-Net: Multiframe and Multimask-Based Unsupervised Deep Denoising for Low-Dose Computed Tomography

“…The attention gate was applied to image restoration tasks, such as denoising [68]- [70], superresolution [71], inpainting [72], dehazing [73], and image enhancement [74]. It improved the reconstruction performance by employing useful features and ignoring irrelevant features, and it effectively suppressed noise when the attention gate was included [69].…”

CT image denoising methods for image quality improvement and radiation dose reduction

“…The predominant DL models for CT denoising are GANs and CNNs. As shown in Figure 2a, out of 99 publications examined, 61 studies use the models based on CNN, 59–119 while 30 studies are based on GAN 120–149 . Additionally, two studies adopt Transformer‐based approaches 150,151 .…”

“…The majority of research focusing on the objective image quality evaluations of DL algorithms has consistently demonstrated remarkable noise reduction compared to FBP and IR at equivalent or lower radiation dose levels. 74,77,79,82,90,92,93,95,103,104,113,114,147 Additionally, the implementation of DL for metal artifact reduction demonstrates superior results when compared to IR. 62,86,119,121 CT image denoising approaches show promising potential, but are not widely accepted in routine clinical practice.…”

“…Self‐supervised learning is an unsupervised method that doesn't rely on external labels 70,82,94,95,105,140,152 . In self‐supervised learning, the model creates its own pseudo‐labels or supervisory signals by designing pretext tasks that require understanding and processing of the data.…”

Dose reduction and image enhancement in micro‐CT using deep learning

“…15,16 Given their ability to learn high-level features from pixel data through hierarchical networks, supervised CNN algorithms attempt to find a mapping function that reduces noise in low dose images (or also for limited angle tomography scans) from matching high dose images (or full angle acquisitions). Unsupervised DL methods for CT image restoration have also been explored 17,18 and most derive from the deep image prior proposed by Ulyanov et al 19 who showed that an image could be enhanced without requiring any prior training data other than the image itself.…”