Abstract

Info

Title:: Fully convolutional neural network based algorithm for low-dose CT image denoising

Author(s):: HONG Qifan1; XUAN Zuxing2; LI Yaxin1; 1. Smart City College, Beijing Union University, Beijing 100101, China 2. Institute of Fundamental and Interdisciplinary Sciences, Beijing Union University, Beijing 100101, China

Keywords:: Keywords: low-dose computed tomography full convolutional neural network noise attention mechanism feature fusion

Abstract:: Abstract: A low-dose CT image denoising algorithm based on a residual attention mechanism and adaptive feature fusion is proposed to address the problem of the decline of image quality caused by a large amount of noise introduced by low-dose CT images due to reduced radiation dose. The algorithm uses a fully convolutional neural network to accomplish low-dose CT image denoising. A residual attention mechanism and a selective kernel feature fusion module are introduced into the network framework to remove noise, extract effective features and adaptively fuse image features to avoid detail loss during reconstruction, thereby improving image quality and making the denoised images perceptually closer to the original images. The qualitative and quantitative experiments show that the proposed algorithm can effectively suppress noise and recover more detailed textures in low-dose CT images as compared with existing algorithms on real clinical datasets. Compared with low-dose CT images, the proposed algorithm increases the peak signal-to-noise ratio by 14.94%, improves the structural similarity by 4.68%, and reduces the root mean square error by 40.11%, meeting the diagnostic requirements.