Abstract

Info

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

Author(s):: FU Changyang; WANG Yu; XIAO Hongbing; XING Suxia; Beijing Key Laboratory of Big Data Technology for Food Safety, School of Computer and Information Engineering, Beijing Technology and Business University, Beijing 100048, China

Keywords:: Keywords: depression; functional brain network; multi-scale; feature fusion; support vector machine

Abstract:: Abstract: 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.