《计算机应用》唯一官方网站 ›› 2023, Vol. 43 ›› Issue (5): 1612-1619.DOI: 10.11772/j.issn.1001-9081.2022040620

• 多媒体计算与计算机仿真 • 上一篇    

基于特征金字塔网络和密集网络的肺部CT图像超分辨率重建

申利华(), 李波   

  1. 武汉科技大学 计算机科学与技术学院,武汉 430081
  • 收稿日期:2022-05-07 修回日期:2022-07-18 接受日期:2022-07-22 发布日期:2022-08-18 出版日期:2023-05-10
  • 通讯作者: 申利华
  • 作者简介:申利华(1999—),女,湖北恩施人,硕士研究生,主要研究方向:计算机视觉、医学图像处理 2469366101@qq.com
    李波(1975—),男,湖北武汉人,教授,博士,主要研究方向:机器学习、智能计算。

Super-resolution reconstruction of lung CT images based on feature pyramid network and dense network

Lihua SHEN(), Bo LI   

  1. School of Computer Science and Technology,Wuhan University of Science and Technology,Wuhan Hubei 430081,China
  • Received:2022-05-07 Revised:2022-07-18 Accepted:2022-07-22 Online:2022-08-18 Published:2023-05-10
  • Contact: Lihua SHEN
  • About author:SHEN Lihua, born in 1999, M. S. candidate. Her research interests include computer vision, medical image processing.
    LI Bo, born in 1975, Ph. D., professor. His research interests include machine learning, intelligent computing.

摘要:

针对肺部计算机断层扫描(CT)图像的超分辨率(SR)重建中需要加大对肺结节的关注度、满足重建后的特征具有客观存在性等问题,提出一种基于特征金字塔网络(FPN)和密集网络的肺部图像SR重建方法。首先,在特征提取层利用FPN提取特征;其次,在特征映射层设计基于残差网络的局部结构,再用特殊的密集网络连接此类局部结构;再次,在特征重建层利用卷积神经网络(CNN)将不同深度的卷积层逐渐降为图像大小;最后,利用残差网络融合初始低分辨率(LR)特征与重建的高分辨率(HR)特征,形成最终的SR图像。对比实验显示,FPN中2次特征融合和特征映射中5个局部结构连接的深度学习网络效果更佳。所提出的网络相较于超分辨率卷积神经网络(SRCNN)等经典网络重建SR图像的峰值信噪比(PSNR)更高,并且可以获得更好的视觉质量。

关键词: 肺部计算机断层扫描图像, 超分辨率重建, 特征金字塔网络, 密集网络, 残差网络

Abstract:

To pay more attention to pulmonary nodules and satisfy the objective existence of reconstructed features in lung Computed Tomography (CT) image Super-Resolution (SR) reconstruction, a lung image SR reconstruction method based on Feature Pyramid Network (FPN) and dense network was proposed. Firstly, at the feature extraction layer, FPN was used to extract features. Secondly, the local structure based on residual network was designed at the feature mapping layer, and then the special dense network was used to connect the local structure. Thirdly, at the feature reconstruction layer, Convolution Neural Network (CNN) was used to gradually reduce the convolution layers with different depths to the image size. Finally, the residual network was used to integrate the initial Low-Resolution (LR) features and the reconstructed High-Resolution (HR) features to form the final SR image. In comparison experiments, the deep learning network with two feature fusion in FPN and five local structure connections in feature mapping has better effect. Compared with classic networks such as Super-Resolution Convolutional Neural Network (SRCNN), the proposed network has higher Peak Signal-to-Noise Ratio (PSNR) and better visual quality of the reconstructed SR images.

Key words: lung Computed Tomography (CT) image, Super-Resolution (SR) reconstruction, Feature Pyramid Network (FPN), dense network, residual network

中图分类号: