[1]洪启帆,玄祖兴,李雅馨.基于全卷积神经网络的低剂量CT去噪算法[J].中国医学物理学杂志,2023,40(6):695-700.[doi:DOI:10.3969/j.issn.1005-202X.2023.06.005]
　HONG Qifan,XUAN Zuxing,LI Yaxin.Fully convolutional neural network based algorithm for low-dose CT image denoising[J].Chinese Journal of Medical Physics,2023,40(6):695-700.[doi:DOI:10.3969/j.issn.1005-202X.2023.06.005]
点击复制

基于全卷积神经网络的低剂量CT去噪算法()
分享到：

《中国医学物理学杂志》[ISSN:1005-202X/CN:44-1351/R]

卷:
40卷
期数:
2023年第6期
页码:
695-700
栏目:
医学影像物理
出版日期:
2023-06-27

文章信息/Info

Title:
Fully convolutional neural network based algorithm for low-dose CT image denoising
文章编号:
1005-202X（2023）06-0695-06
作者:
洪启帆1; 玄祖兴2; 李雅馨1
1.北京联合大学智慧城市学院， 北京 100101； 2.北京联合大学数理与交叉科学研究院， 北京 100101
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
关键词:
低剂量CT; 全卷积神经网络; 噪声; 注意力机制; 特征融合
Keywords:
Keywords: low-dose computed tomography full convolutional neural network noise attention mechanism feature fusion
分类号:
R318；TP391.4
DOI:
DOI:10.3969/j.issn.1005-202X.2023.06.005
文献标志码:
A
摘要:
针对低剂量医学CT图像因减少辐射剂量而引入大量噪声，导致图像质量显著下降的问题，提出一种基于残差注意力机制和自适应特征融合的低剂量CT图像去噪算法，该算法使用全卷积神经网络来完成低剂量CT图像去噪。在网络框架中引入残差注意力机制和选择性内核特征融合模块，以过滤噪声信息，提取有效特征并自适应地融合图像特征，避免重建过程中的细节损失，提高图像质量，使去噪后的图像在感知上更接近原始图像。定性和定量实验表明，与现有的算法相比，在真实的临床数据集上，所提出的算法能够有效地抑制噪声，并恢复低剂量CT图像中更详细的纹理。与低剂量CT图像相比，所提出的算法将峰值信噪比提高14.94%，结构相似度提高4.68%，均方根误差降低40.11%，可以满足医学成像的诊断要求。
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.

相似文献/References:

[1]钱姗姗,边兆英,路利军,等.两种基于Anscombe变换域滤波的低剂量CT重建方法讨论[J].中国医学物理学杂志,2013,30(02):4023.[doi:10.3969/j.issn.1005-202X.2013.02.015]
[2]陈伟国,贺捷新,莫天澜,等. 低剂量CT心肌灌注鲁棒去卷积参数成像方法[J].中国医学物理学杂志,2017,34(11):1131.[doi:DOI:10.3969/j.issn.1005-202X.2017.11.010]
　CHEN Weiguo,HE Jiexin,MO Tianlan,et al. Robust deconvolution for low-dose CT myocardial perfusion imaging[J].Chinese Journal of Medical Physics,2017,34(6):1131.[doi:DOI:10.3969/j.issn.1005-202X.2017.11.010]
[3]刘海坤,王健,杨嵩,等. 基于深度学习和二维高斯拟合的视网膜血管管径测量方法[J].中国医学物理学杂志,2019,36(2):171.[doi:DOI:10.3969/j.issn.1005-202X.2019.02.010]
　LIU Haikun,WANG Jian,YANG Song,et al. Retinal vessel diameter measurement based on depth learning and two-dimensional Gaussian fitting[J].Chinese Journal of Medical Physics,2019,36(6):171.[doi:DOI:10.3969/j.issn.1005-202X.2019.02.010]
[4]谢丰雪,杨帆,冯维,等.基于深度学习的低剂量CT图像去噪[J].中国医学物理学杂志,2022,39(5):547.[doi:DOI:10.3969/j.issn.1005-202X.2022.05.004]
　XIE Fengxue,YANG Fan,FENG Wei,et al.Low-dose CT image denoising based on deep learning[J].Chinese Journal of Medical Physics,2022,39(6):547.[doi:DOI:10.3969/j.issn.1005-202X.2022.05.004]
[5]崔奥宇,许幸芬,田苗,等.基于全卷积神经网络的冠状动脉中心线提取方法[J].中国医学物理学杂志,2023,40(4):429.[doi:DOI:10.3969/j.issn.1005-202X.2023.04.006]
　CUI Aoyu,XU Xingfen,TIAN Miao,et al.Coronary artery centerline extraction method based on fully convolutional network[J].Chinese Journal of Medical Physics,2023,40(6):429.[doi:DOI:10.3969/j.issn.1005-202X.2023.04.006]

备注/Memo

备注/Memo:
【收稿日期】2022-12-23 【基金项目】北京市优秀人才培养资助青年拔尖个人项目（2018000026833ZK57）；北京联合大学人才强校优选计划（BPHR2020EZ01）；北京联合大学学科团队一体化建设计划项目（ZB10202001） 【作者简介】洪启帆，硕士，研究方向：图像处理、深度学习等，E-mail: 18810536635@163.com 【通信作者】玄祖兴，博士，教授，研究方向：深度学习、应用数学，E-mail: zuxingxuan@163.com
更新日期/Last Update: 2023-06-28