基于图神经网络的自闭症辅助诊断研究综述

自闭症谱系障碍是一种复杂的神经系统发展障碍疾病，截至目前其病因尚不明确。图神经网络作为非欧几里得空间深度学习的重要分支，在处理图结构数据的相关任务中取得优异表现，为医学领域的成像和非成像模式的集成提供了可能，因此利用图神经网络进行自闭症等脑部疾病神经成像诊断逐渐成为研究热点。阐述传统机器学习方法在自闭症疾病预测中应用，介绍图神经网络的基本分类，按照图中节点与边关系的建模方法，从基于人群图和基于个体图两个角度对图神经网络在自闭症辅助诊断中的应用进行梳理和分析，并归纳现有诊断方法的优劣势。根据目前基于图神经网络的自闭症神经成像诊断的研究现状，总结了脑神经科学领域辅助诊断技术面临的主要挑战和未来研究方向，对于自闭症等脑部疾病辅助诊断的进一步研究具有指导意义和参考价值。

Review of Research on Auxiliary Diagnosis of Autism Based on Graph Neural Networks

Autistic Spectrum Disorder(ASD) is a complex neurological developmental disorder;however, its etiology remains unclear.Graph Neural Network(GNN), a branch of Deep Learning(DL) in non-Euclidean space, has excelled in a variety of tasks dealing with graph-structured data.GNN provides a universal approach for the integration of imaging and non-imaging modalities in medicine.With their development, the use of GNNs for neuroimaging diagnosis of autism and other brain diseases has gradually become a research hotspot.This paper expounds the application of traditional machine learning methods in autism disease prediction, and introduces the basic classification of GNNs.Based on the modeling approach of node-edge relationships in graphs, the application of GNNs in the auxiliary diagnosis of autism is categorized and analyzed from two perspectives:population-based and individual-based graphs, and the advantages and disadvantages of existing methods are summarized.Finally, according to the existing studies of GNN-based neuroimaging diagnosis of autism, the main challenges hampering the auxiliary diagnosis in brain neuroscience are concluded, which displays a certain guiding significance and reference value for further research on auxiliary diagnosis of autism and other brain diseases.