基于图卷积网络的MDD静息态功能磁共振功能连接性别差异研究  

Gender Differences in Functional Connectivity of MDD Resting-State Functional Magnetic Resonance Based on Graph Convolutional Network

作　　者：万立 张仙华 高炬 张钰群 欧阳键 WAN Li;ZHANG Xian-hua;GAO Ju;ZHANG Yu-qun;OUYANG Jian(School of Communications and Information Engineering,Nanjing University of Posts and Telecommunications,Nanjing 210000,China;Department of Psychiatry,Suzhou Guangji Hospital,Suzhou 215131,China;School of Nursing,Nanjing University of Chinese Medicine,Nanjing 210000,China)

机构地区：[1]南京邮电大学通信与信息工程学院,南京210000 [2]苏州市广济医院精神科,苏州215131 [3]南京中医药大学护理学院,南京210000

出　　处：《中国临床心理学杂志》2024年第5期1141-1147,共7页Chinese Journal of Clinical Psychology

基　　金：国家自然科学基金青年项目(82001426);江苏高校青蓝工程优秀青年骨干教师。

摘　　要：目的:利用图卷积网络(GCN)模型探究不同性别抑郁症(MDD)患者的功能连接差异,为实现MDD的精准诊治提供依据。方法:基于Rest-meta-MDD项目大样本多中心静息态功能磁共振数据,使用GCN模型分别对男性和女性样本进行独立训练,获得不同性别高鉴别脑区的拓扑特征及其与临床症状的相关性。结果:与支持向量机相比,GCN模型的平均分类准确率提高了5%以上。其中,男性组MDD和HC分类的平均准确率为77.91%,且右侧前额叶、右侧顶叶和左侧下顶叶的节点度与汉密尔顿焦虑量表得分呈显著正相关。女性组的平均准确率为79.71%,且右侧枕叶的节点介数与汉密尔顿抑郁量表得分呈显著负相关。结论:GCN提升了男性和女性MDD的分类准确率,提示男性MDD可能为焦虑型,而女性MDD则可能为抑郁型。Objective:To investigate gender differences in functional connectivity among patients with major depressive disorder(MDD)using a graph convolutional network(GCN)model,providing insights for precision diagnosis and treatment of MDD.Methods:Leveraging large-scale,multi-center resting-state f MRI data from the Rest-meta-MDD project,GCN models were independently trained on male and female samples.The study identified gender-specific topological features of highly discriminative brain regions and their correlations with clinical symptoms.Results:The GCN model outperformed support vector machines,improving classification accuracy by over 5%.The average accuracy for distinguishing MDD from healthy controls was 77.91%in males,with significant positive correlations between nodal degrees in the right prefrontal cortex,right parietal lobe,and left inferior parietal lobule,and hamilton anxiety rating scale scores.In females,the average accuracy was 79.71%,with significant negative correlations between the betweenness centrality of the right occipital lobe and hamilton depression rating scale scores.Conclusion:The GCN-based model demonstrates superior efficacy in MDD classification.There are significant gender-specific differences in brain network characteristics,which are correlated with anxiety and depression symptoms in males and females,respectively.

关 键 词：抑郁症 静息态功能磁共振 图卷积网络 性别 

分 类 号：R395.1[哲学宗教—心理学]