基于脑功能超网络的多特征融合分类方法

针对在超网络上提取局部脑区指标作为特征，忽视了全局的拓扑信息，继而影响网络拓扑的评估，降低分类器性能的问题，提出了一种基于脑功能超网络的多特征融合分类方法，该方法首先在抑郁症数据集上构建超网络，其次将局部脑区特征和子图特征进行融合。最后采用基于多核的SVM分类器进行分类。为了验证所提方法的有效性，选取28例正常被试和38例抑郁症患者进行实验，结果表明，该方法获得了令人满意的分类准确率，平均可达91.60%。获得的异常区域包括左侧舌回、左侧尾状核、左侧丘脑等重要的抑郁症病发区域。故而该基于脑功能超网络的多特征融合分类方法可以有效地用于分类正常人和抑郁症患者。

Machine learning classification method combining multiple features of brain function hyper-network

Focused on the issue that local brain region properties is extracted as features on hyper-network, ignoring the global topology information, which affects the evaluation of network topology and reduces the performance of the classifier. Machine learning classification method combining multiple features of brain function hyper-network is proposed. Firstly, hyper-networks are constructed on major depression disorder dataset. Secondly, brain region features and subgraph features are combined as features. Finally, multi-kernel SVM is adopted to classify. To certify the proposed method, 28 normal control subjects and 38 major depression disorder patients are selected for experiment. The experimental results show that the proposed method achieves satisfactory accuracy, with an average of 91.60%. The abnormal brain regions include left Lingual gyrus, left Caudate nucleus, left Thalamus and so on important brain regions of major depression disorder. Machine learning classification method combining multiple features of brain function hyper-network can effectively classify normal control subjects and major depression disorder patients.