摘要
自2006年深度学习这一概念提出以来,各研究领域对于深度学习技术的研究热度一直高居不下。深度学习的出现,对计算机视觉领域的发展起到了重要推动作用。计算机视觉的主要研究任务是对图像、视频等进行目标的检测、识别以及分割等,目前已经广泛应用于医疗、金融和工业领域中。其中最常见的应用场景是医学图像处理。图像分割是医学图像处理任务中一个重要的研究方向,目前已经出现了很多图像分割方法,其中包含传统的分割方法和基于深度学习模型的分割方法。首先介绍了阈值分割法、区域生长法以及图割法等传统的图像分割方法;其次总结了FCN、U-Net、U-Net++、SegNet以及DeepLab系列的网络架构,并对其优缺点进行了分析;同时,着重阐述了图像分割方法在医疗图像处理中的应用;最后讨论了未来基于深度学习的医学图像分析将要面临的挑战和发展机遇。
Since the concept of deep learning was proposed in 2006,research enthusiasm for deep learning technology in various fields has been high.The emergence of deep learning has played an important role in promoting the development of computer vision.The main task of computer vision is to detect,identify or segment objects in images and videos.It has been widely used in medical,financial and industrial fields,and the most common application scenario is medical image processing.Image segmentation is an important research direction in medical image processing tasks.Current image segmentation methods include methods of traditional and methods based on deep learning.Firstly,we introduce traditional image segmentation methods such as threshold segmentation,region growing and graph cutting.Secondly,we summarize the network architecture of FCN,U-Net,U-Net++,SegNet and DeepLab,and analyze their advantages and disadvantages.Next,we focus on the application of image segmentation method in medical image processing.Finally,we discuss future challenges and development opportunities for medical image analysis based on deep learning.
作者
孔令军
王茜雯
包云超
李华康
KONG Lingjun;WANG Qianwen;BAO Yunchao;LI Huakang(Jinling Institute of Technology,Nanjing 211169,China;Nanjing University of Posts and Telecommunications,Nanjing 210003,China;Xi’an Jiaotong-Liverpool University,Suzhou 215123,China)
出处
《无线电通信技术》
2021年第2期121-130,共10页
Radio Communications Technology
基金
中国博士后科学基金资助项目(2020M671595)
江苏省博士后科研资助计划项目(2020Z198)。
关键词
人工智能
深度学习
医疗图像
图像分割
artificial intelligence
deep learning
medical image
image segmentation
