基于多尺度功能脑网络融合特征的抑郁症分类算法

Classification of depression using fusion features based on multi-scale functional brain network

提出一种多尺度功能脑网络融合特征的抑郁症分类方法,具体思想包括:首先通过精细化脑区,建立4种不同尺度的脑网络;然后对每种尺度的脑网络分别提取局部特征和全局特征,并将多种尺度脑网络的特征进行有效融合并降维;最后使用支持向量机对患者脑部功能磁共振影像进行分类。试验结果表明,分别提取局部特征和全局特征,并进行有效融合,可以提升识别效果;空间尺度减小会得到更多有效特征,进而能够有效提升分类结果;多尺度特征融合也可以在很大程度上对分类结果起到积极作用。与传统单一大尺度脑网络方法相比,本研究提出的方法获得了更加优秀的试验结果,识别率可达88.67%,充分验证了本研究提出方法的有效性和可行性,并为抑郁症患者的临床诊断与治疗提供生物学依据。

A novel method for the classification of depression is proposed based on a multi-scale functional brain network and fusion features. After 4 different scales of brain networks are established by refining the brain region, the local and global features are extracted from each scale of brain network, and the features of multi-scale brain networks are effectively fused and the dimensionality is reduced. Finally, support vector machine is used to classify the functional magnetic resonance images of the brain. The experimental results show that it is effective to improve the recognition effect by the fusion of separately extracted local and global features. More effective features can be obtained by reducing the spatial scale, which can remarkably improve the classification results. Multi-scale feature fusion can also greatly promote the accuracy and recall rate of classification. Compared with the traditional method with single large-scale brain networks, the proposed method achieves better performances. The classification accuracy by the proposed method reaches 88.67%, which fully verifies the effectiveness and feasibility of the proposed method, and also provides a biological basis for the clinical diagnosis and treatment of depression.