摘要
脑群体图基于受试者的神经影像、非成像信息构建,可从全局研究角度出发探究脑疾病间的潜在关联性和发病机理。图卷积神经网络(graph convolutional neural networks,GCN)可较好处理不规则图数据,近年来在脑群体图研究中得到较多应用,用于分类、预测脑疾病。本文首先分别从频域和空间域介绍GCN算法的基本原理和典型模型;其次具体阐述群体图的构建流程,分别从单一、多群体图两个角度介绍GCN在该领域的应用;最后,讨论了GCN在脑群体图分析方面存在的问题,并对未来发展方向进行展望。
The brain population graph is constructed based on the subject’s neuroimaging and non-imaging information,which can explore the potential correlation and pathogenesis of brain diseases from a global research perspective.Graph convolutional neural networks(GCN)can process irregular graph data efficiently.In recent years,it has been widely used in brain population graphs to classify or predict brain diseases.Firstly,the basic principle and typical models of the GCN algorithm from the spectrum domain and the spatial domain.Secondly,it specifically explains the construction process of the population graph and introduces the application of GCN in this field from the perspective of single and multi-population graphs.Finally,the existing problems of GCN in brain population graph analysis are discussed and the future development direction is prospected.
作者
张格
林岚
吴水才
ZHANG Ge;LIN Lan;WU Shuicai(Intelligent Physiological Measurement and Clinical Translation,Beijing International Base for Scientific and Technological Cooperation,College of Life Science and Bioengineering of Beijing University of Technology,Beijing 100124,China)
出处
《生命科学仪器》
2021年第4期23-30,共8页
Life Science Instruments
基金
国家自然科学基金(81971683)
北京市自然科学基金-海淀原始创新联合基金(L182010)
关键词
图卷积神经网络
群体图
脑疾病
机器学习
神经影像
graph convolutional neural networks
population graph
brain disease
machine learning
neuroimaging
