文章摘要

文章信息/Info

Title:: Research progress in neuropsychiatric diseases based on graph neural network

作者:: 王海源1; 吴凯1; 2; 3; 4; 5; 陈小怡1; 彭润霖1; 梁丽琴1; 周静2; 3; 4; 5; 6△; （1.华南理工大学生物医学科学与工程学院，广州 511442；2.华南理工大学国家人体组织功能重建工程技术研究中心，广州 510006；3.广东省精神疾病转化医学工程技术研究中心, 广州 510370；4.广东省老年痴呆诊断与康复工程技术研究中心, 广州 510500；5.华南理工大学广东省生物医学工程重点实验室，广州，510006；6.华南理工大学材料科学与工程学院, 广州 510006）

Author(s):: WANG Haiyuan1; WU Kai1; 2; 3; 4; 5; CHEN Xiaoyi1; PENG Runlin1; LIANG Liqin1; ZHOU Jing2; 3; 4; 5; 6; （1.School of Biomedical Sciences and Engineering, South China University of Technology, Guangzhou 511442, China; 2.National Engineering Research Center for Tissue Restoration and Reconstruction, South China University of Technology, Guangzhou 510006; 3. Guangdong Engineering Technology Research Center for Translational Medicine of Mental Disorders, Guangzhou 510370; 4. Guangdong Engineering Technology Research Center for Diagnosis and Rehabilitation of Dementia, Guangzhou 510500; 5. Guangdong Province Key Laboratory of Biomedical Engineering, South China University of Technology, Guangzhou 510006; 6. School of Material Science and Engineering, South China University of Technology, Guangzhou 510006）

Keywords:: Magnetic resonance imaging; Neuropsychiatric diseases; Brain network; Automatic classification; Graph neural network; Disease diagnosis

摘要:: 神经精神疾病严重影响患者脑解剖结构、神经系统功能及心理健康，其早期识别与诊断对患者的治疗及康复具有重要意义。基于神经影像数据构建复杂的脑网络，可用于定量化分析神经精神疾病患者的脑结构及功能异常，为研究神经精神疾病的神经影像生物标记物提供重要参考。近年来，图神经网络具有处理非欧几里得数据、能充分利用节点与连边的拓扑结构和属性等优势，被广泛应用于神经精神疾病的辅助诊断研究。本文对图卷积神经网络的基本原理和神经精神疾病的最新研究进展进行了总结和分析，并展望了动态脑网络、大样本与多中心、可视化与可解释性等研究热点。

Abstract:: Neuropsychiatric diseases seriously affect the anatomical structure of the brain, nervous system function, and mental health of patients. Early identification and diagnosis are of great significance for the treatment and rehabilitation of patients with neuropsychiatric diseases. The construction of complex brain networks based on neuroimage data can be used to quantitatively analyze the brain structure and function abnormalities in patients with neuropsychiatric diseases, and provide an important reference for the development of neuroimaging biomarkers for neuropsychiatric diseases. In recent years, graph neural network has been widely used in the diagnosis of neuropsychiatric diseases because of their advantages of processing non-Euclidean data and making full use of the topological structure and attributes of nodes and connected edges. We summarize the basic principles of graph convolutional network and the latest research progress in neuropsychiatric diseases, and look forward to research hotspots such as dynamic brain network, large sample and multi center, visualization and interpretability.