Multivariate classification of social anxiety disorder using whole brain functional connectivity

• Published in: Brain Structure and Function Vol. 220; no. 1; pp. 101 - 115
• Main Authors: Liu, Feng, Guo, Wenbin, Fouche, Jean-Paul, Wang, Yifeng, Wang, Wenqin, Ding, Jurong, Zeng, Ling, Qiu, Changjian, Gong, Qiyong, Zhang, Wei, Chen, Huafu
• Format: Journal Article
• Language: English
• Published: Berlin/Heidelberg Springer Berlin Heidelberg 01.01.2015; Springer Nature B.V
• Subjects: Adolescent; Adult; Anatomy & physiology; Anxiety; Anxiety Disorders - classification; Anxiety Disorders - pathology; Biomedical and Life Sciences; Biomedicine; Brain; Brain - blood supply; Brain - pathology; Brain Mapping; Case-Control Studies; Cell Biology; Female; Humans; Imaging, Three-Dimensional; Magnetic Resonance Imaging; Male; Multivariate Analysis; Neural Pathways - blood supply; Neural Pathways - pathology; Neurology; Neuropsychology; Neurosciences; NMR; Nonlinear Dynamics; Nuclear magnetic resonance; Original Article; Psychiatric Status Rating Scales; Regression Analysis; ROC Curve; Social Behavior; Support Vector Machine; Young Adult; Functional connectivity; Resting-state fMRI; Consensus features; Multivariate pattern analysis; Support vector machine; Social anxiety disorder/social phobia
• ISSN: 1863-2653; 1863-2661; 1863-2661; 0340-2061
• DOI: 10.1007/s00429-013-0641-4

Recent research has shown that social anxiety disorder (SAD) is accompanied by abnormalities in brain functional connections. However, these findings are based on group comparisons, and, therefore, little is known about whether functional connections could be used in the diagnosis of an individual patient with SAD. Here, we explored the potential of the functional connectivity to be used for SAD diagnosis. Twenty patients with SAD and 20 healthy controls were scanned using resting-state functional magnetic resonance imaging. The whole brain was divided into 116 regions based on automated anatomical labeling atlas. The functional connectivity between each pair of regions was computed using Pearson’s correlation coefficient and used as classification feature. Multivariate pattern analysis was then used to classify patients from healthy controls. The pattern classifier was designed using linear support vector machine. Experimental results showed a correct classification rate of 82.5 % ( p  < 0.001) with sensitivity of 85.0 % and specificity of 80.0 %, using a leave-one-out cross-validation method. It was found that the consensus connections used to distinguish SAD were largely located within or across the default mode network, visual network, sensory-motor network, affective network, and cerebellar regions. Specifically, the right orbitofrontal region exhibited the highest weight in classification. The current study demonstrated that functional connectivity had good diagnostic potential for SAD, thus providing evidence for the possible use of whole brain functional connectivity as a complementary tool in clinical diagnosis. In addition, this study confirmed previous work and described novel pathophysiological mechanisms of SAD.