基于fMRI动态功能连接的抑郁症患者分类研究*

针对当前抑郁症诊断正确率偏低、误诊率偏高的问题，利用fMRI动态功能连接研究了抑郁症辅助诊断问题。首先采用滑动时间窗技术研究功能连接及其网络拓扑特性的动态变化，然后基于这些动态特征应用多元模式分析方法对22名抑郁症患者和27名健康被试进行分类。采用动态分析方法能够增加样本数量，从而更加有利于一些分类算法的应用。实验结果表明以动态功能连接和网络拓扑特性为特征的分类正确率均为93.88%，明显优于对应非动态特征81.63%和85.71%的结果。进一步分析表明,具有高辨别力的特征所对应的脑区主要分布在默认网络、情感网络、视觉皮层区等，动态功能连接可能为抑郁症的辅助诊断提供新的手段。

Classifying Patients with Depression based on fMRI Dynamic Functional Connectivity*

This paper investigated a computer aided diagnosis model of depression using dynamic functional connectivity, the aim of which was to improve the accuracy of depression diagnosis. Firstly, it studied the dynamic change and topologic characteristic of fMRI based brain functional connectivity by sliding time window technique. Then, it employed a multivariate pattern analysis method to classify 22 patients with depression from 27 healthy volunteers. Dynamic functional connectivity analysis could increase the number of samples, which made more classification algorithms practicable for this problem. The best performance based on dynamic features was 93.88%, which was much better than non-dynamic features based results 81.63% and 85.71%. Further analysis demonstrated that the brain areas corresponding to the most discriminating features were mainly located in default mode network, affective network and visual cortical areas. Dynamic functional connectivity might be a potential measure for the diagnosis of depression.