机器学习方法在fMRI数据中的应用研究

Application of machine learning method in fMRI data

文章通过机器学习方法对功能性核磁共振成像(fMRI)数据进行分类研究,旨在探讨精神分裂患者的差异性脑网络结构。在特征选择上,采用了复杂网络分析来度量脑网络的拓扑属性,并且以年龄为协变量通过双样本t检验构建显著差异性特征集;在分类算法上,采用粒子群优化支持向量机分类算法,在10折交叉验证下进行建模分析。结果表明精神分裂患者(SCH)组和正常被试(NC)组在全局属性和节点属性上存在显著差异,在分类模型下能够得到最佳准确率93%。

This paper classifies functional magnetic resonance imaging(fMRI)data by machine learning method in order to research the differential brain network structure of schizophrenics.In feature selections,complex network analysis is used to measure the topological attributes of the brain network,and the age is used as the covariate to construct the significant difference feature set through two-sample t-test.In the classification algorithm,the particle swarm optimization support vector machine algorithm is used for modeling and analysis under 1o fold cross validation.The results show that there are significant differences between patients with Schizophrenics(SCH)group and and Normal Subjects(NC)group in global attributes and node attributes,with an optimal classification accuracy of 93%.