Re-UNet: A Novel Multi-scale Reverse U-shaped Network Architecture for Low-dose CT Image Reconstruction

Abstract

In recent years, the growing awareness of public health has brought attention to low-dose computed tomography(LDCT) scans. However, the CT image generated in this way contains a lot of noise or artifacts, which make increasing researchers to investigate methods to enhance image quality. The advancement of deep learning technology has provided researchers with novel approaches to enhance the quality of LDCT images. In the past, numerous studies based on convolutional neural networks(CNN) have yielded remarkable results in LDCT image reconstruction. Nonetheless, they were unsuccessful in devising a more rational model framework, ultimately leading to an escalation in model complexity. In this paper, we propose a novel reverse U-shaped network architecture for LDCT image reconstruction, in which we design a novel multi-scale feature extractor and edge enhancement module that yields a positive impact on CT images to exhibit strong structural characteristics. Evaluated on a public dataset, the experimental results demonstrate that the proposed model outperforms the compared algorithms based on traditional U-shaped architecture in terms of preserving texture details and reducing noise, as demonstrated by achieving the highest peak signal-to-noise ratio (PSNR), structural similarity (SSIM) and RMSE value. This study may shed light on the reverse U-shaped network architecuture for CT image reconstruction, and could be investigate the potential on other medical image processing.

Note:
Funding Information: This work was supported in part by the National Natural Science Foundation of China under Grant No.62076209, and in part by the NHC Key Laboratory of Nuclear Technology Medical Transformation (MIANYANG CENTRAL HOSPI- TAL) under Grant No.2022HYX012.

Conflict of Interests: None to declare.

Keywords: Low-dose CT; UNet ;deep learning; image reconstruction.

JEL Classification: I19