摘要
传统的医学图像分割中特征提取算法的设计复杂性与应用局限性、稳定性以及特定的特征提取算法与特定的分类器结合的多样性制约着医学图像分割技术的发展,而深度学习是机器学习领域中使用多重非线性变换对数据进行多层抽象的热门算法,其多被应用于医学图像的分类和识别中。在肺组织分割中,针对肺部组织纹理复杂,且胸部CT图像数据的随机噪声大,采用相对成熟的传统分割算法对CT图像进行预处理,再结合深度学习的理论,设计一个合理的神经网络模型,利用已经标记好的多组肺部CT图像进行训练,使其能够准确地分割出肺部组织。基于U-net神经网络的深度学习方法对肺实质的分割进行研究与实现,并针对临床扫描胸部CT图像进行了实验验证,能够较为准确快速地分割出肺实质。
The design complexity and application limitation,stability,and the diversity of specific feature extraction algorithms combined with specific classifiers limit the development of image processing technology in traditional image processing techniques.In the field of machine learning,deep learning is an algorithm that attempts to use multiple nonlinear transformation to abstract data.On the basis of the relatively mature traditional medical CT image segmentation algorithm,combined with the theory of deep learning,a reasonable neural network model is designed to preprocess and train the tagged CT image so that it can accurately segment the lung tissue.With the lung parenchyma is taken as an example based on U-Net network and the clinical scanning chest CT images are used to verify that the lung parenchyma can be segmented more accurately and quickly.
作者
程立英
高宣爽
申海
黄丹阳
覃文军
CHENG Liying;GAO Xuanshuang;SHEN Hai;HUANG Danyang;TAN Wenjun(College of Physical Science and Technology, Shenyang Normal University, Shenyang 110034, China;Key Laboratory of Intelligent Computing in Medical Image, Ministry of Education, Northeastern University, Shenyang 110189, China)
出处
《沈阳师范大学学报（自然科学版）》
CAS
2020年第3期278-282,共5页
Journal of Shenyang Normal University:Natural Science Edition
基金
辽宁省科技厅自然科学基金资助项目(2019-ZD-0480)
国家自然科学基金面上项目(61971118)。
