计算机工程与应用 ›› 2022, Vol. 58 ›› Issue (16): 18-30.DOI: 10.3778/j.issn.1002-8331.2203-0229

• 热点与综述 • 上一篇    下一篇

生成对抗网络在肝脏肿瘤图像分割中的应用综述

张颖,仇大伟,刘静   

  1. 山东中医药大学 智能与信息工程学院,济南 250355
  • 出版日期:2022-08-15 发布日期:2022-08-15

Review on Application of Generative Adversarial Network in Liver Tumor Image Segmentation

ZHANG Ying, QIU Dawei, LIU Jing   

  1. College of Intelligence and Information Engineering, Shandong University of Traditional Chinese Medicine, Jinan 250355, China
  • Online:2022-08-15 Published:2022-08-15

摘要: 由于肝脏肿瘤图像复杂多样且肝脏肿瘤图像数据集获取困难等问题,快速准确地诊断肝脏肿瘤疾病面临着诸多挑战,尤其是肝脏肿瘤的分割是其中的关键研究内容。生成对抗网络在半监督学习领域具有强大的优越性,因此其在医学图像处理中得到广泛应用。为了分析肝脏肿瘤图像在分割领域的现状以及未来发展,针对应用GAN的肝脏肿瘤图像分割方法进行研究,介绍GAN模型的网络结构与衍生模型,重点总结并分析生成对抗网络在肝脏肿瘤图像分割中的应用,包括基于网络结构改进的GAN方法、基于生成器或判别器改进的GAN方法和基于GAN的其他改进方法。最后在已有的研究进展和基础之上,对GAN在肝脏肿瘤图像分割中的应用进行总结,讨论GAN在肝脏肿瘤图像分割上所面临的挑战,并对其未来发展进行展望。

关键词: 生成对抗网络(GAN), 图像分割, 肝脏肿瘤

Abstract: Due to the complexity and diversity of liver tumor images and the difficulty in obtaining the data set of liver tumor images, the rapid and accurate diagnosis of liver tumor diseases faces many challenges, especially the segmentation of liver tumors is the key research content. Generative adversarial network(GAN) is widely used in medical image processing because of its strong superiority in semi-supervised learning. In order to analyze the present situation and future development of liver tumor image segmentation, the segmentation methods of liver tumor image using GAN are studied, the network structure and derivative model of GAN model are introduced, and the application of generative adversarial network in liver tumor image segmentation is mainly summarized and analyzed, including the segmentation method based on improved network structure, improved GAN method based on generator or discriminator and other improved methods based on GAN. Finally, on the basis of the existing research progress, this paper summarizes the application of GAN in liver tumor image segmentation, discusses the challenges that GAN faces in liver tumor image segmentation, and looks forward to its future development.

Key words: generative adversarial network(GAN), image segmentation, liver tumor