Abstract:

The existing classification model of mental diseases uses the static index of brain network as the characteristic while ignoring the spatial dynamic information of brain network, which will result in an inferior classification performance. To overcome this limitation and improve the performance of the classification model, a classification method based on the spatial dynamic of resting-state functional brain connections is proposed. The spatial dynamic characteristics of brain connections are extracted by analyzing the brain connections with high-dimensional templates. By selecting characteristics through the statistical analysis, a classification model based on resting-state functional brain connections can be constructed. The conducted experiments distinguish between depression patients and normal subjects and the results show that the classification accuracy of model utilized spatial dynamic characteristics （83.0%） is 5.2 percentage points higher than that with static index （77.8%）.

Key words: spatial dynamics, functional magnetic resonance imaging, support vector machine, depression

摘要：

现有的精神疾病分类模型仅采用脑网络的静态指标作为特征，忽略了脑网络的空间动态信息，导致分类性能不高。为克服这一局限性，提升分类模型的性能，提出了基于功能脑连接空间动态的分类方法。通过高维模板对脑连接进行空间动态分析，提取脑连接空间动态特征。利用统计分析进行特征选择，构建基于静息态功能脑连接的分类模型。通过对抑郁症患者与正常被试的分类实验结果表明，脑连接空间动态特征的分类准确率（83.0%）比传统采用脑网络的静态指标特征的分类准确率（77.8%）高5.2个百分点。

关键词: 空间动态, 功能磁共振成像, 支持向量机, 抑郁症