摘要
利用脑肿瘤核磁共振(Magnetic resonance,MR)图像提供的关于肿瘤和脑组织的形状、大小与定位等信息准确地分割出脑肿瘤区域,对监测脑肿瘤患者的肿瘤生长或缩小、制定手术或放化疗计划都起着重要的作用。探讨了脑肿瘤MR图像分割的背景与意义,整理了脑肿瘤分割方法中常用的评估指标以及实验数据库的发展过程。基于脑肿瘤MR图像的特点讨论了脑肿瘤分割的难点,并从MR成像缺陷、脑组织解剖结构以及脑肿瘤的复杂性等方面进行归纳。对脑肿瘤分割方法的分类以及常见的分割方法进行了研究,分析了基于图论的分割方法、基于可形变模型的分割方法以及基于机器学习的分割方法及其进展。最后,结合脑肿瘤分割中存在的问题对未来的研究工作进行展望。
By using the information about the brain tissues and tumors provided by the magnetic resonance(MR)images,the brain tumor regions can be segmented from the images,which plays an important role in monitoring the tumor growth or shrinking and formulating surgery or radiotherapy and chemotherapy plans.The background and significance of brain tumor MR image segmentation are introduced,and the common evaluation indicators of brain tumor segmentation methods and the development of experimental databases are sorted out.Based on the characteristics of brain tumor MR images,the difficulties of brain tumor segmentation are discussed and summarized from the aspects of MR imaging defects,brain tissue structure and the complexity of brain tumor.Then,the paper discusses the classification of brain tumor segmentation methods and studies common segmentation methods,focusing on the analysis of the graph-based methods,the deformable models-based methods and the machine learning-based methods as well as their progress.Finally,combined with the problems existing in brain tumor segmentation,the future research work is prospected.
作者
葛婷
詹天明
李勤丰
牟善祥
Ge Ting;Zhan Tianming;Li Qinfeng;Mu Shanxiang(School of Electronic and Optical Engineering,Nanjing University of Science and Technology,Nanjing 210094,China;School of Sciences,Jinling Institute of Technology,Nanjing 211169,China;School of Information Engineering,Nanjing Audit University,Nanjing 211815,China)
出处
《南京理工大学学报》
EI
CAS
CSCD
北大核心
2021年第2期179-188,共10页
Journal of Nanjing University of Science and Technology
关键词
脑肿瘤
核磁共振成像
脑肿瘤分割
医学图像分割
图论
水平集
模糊C-均值
人工神经网络
核方法
brain tumor
magnetic resonance image
brain tumor segmentation
medical image segmentation
brain tumor segmentation assessment
graph theory
level set
fuzzy C-means
artificial neural networks
kernel method
