Graph Convolutional Networks Reveal Network-Level Functional Dysconnectivity in Schizophrenia

Abstract

Background and hypothesis: Schizophrenia is increasingly understood as a disorder of brain dysconnectivity. Recently, graph-based approaches such as graph convolutional network (GCN) have been leveraged to explore complex pairwise similarities in imaging features among brain regions, which can reveal abstract and complex relationships within brain networks.

Study design: We used GCN to investigate topological abnormalities of functional brain networks in schizophrenia. Resting-state functional magnetic resonance imaging data were acquired from 505 individuals with schizophrenia and 907 controls across 6 sites. Whole-brain functional connectivity matrix was extracted for each individual. We examined the performance of GCN relative to support vector machine (SVM), extracted the most salient regions contributing to both classification models, investigated the topological profiles of identified salient regions, and explored correlation between nodal topological properties of each salient region and severity of symptom.

Study results: GCN enabled nominally higher classification accuracy (85.8%) compared with SVM (80.9%). Based on the saliency map, the most discriminative brain regions were located in a distributed network including striatal areas (ie, putamen, pallidum, and caudate) and the amygdala. Significant differences in the nodal efficiency of bilateral putamen and pallidum between patients and controls and its correlations with negative symptoms were detected in post hoc analysis.

Conclusions: The present study demonstrates that GCN allows classification of schizophrenia at the individual level with high accuracy, indicating a promising direction for detection of individual patients with schizophrenia. Functional topological deficits of striatal areas may represent a focal neural deficit of negative symptomatology in schizophrenia.

Keywords: connectome; graph analysis; machine learning; magnetic resonance imaging; neuroimaging; psychosis.

Fig. 1.

The overall pipeline of graph convolutional network model. (A) Graph construction for each individual using resting-state functional connectivity. (B) The architecture and implementation of graph convolutional network. Note: GAP, global average pooling; HC, healthy controls; ReLu, Rectified Linear Unit; SCZ, schizophrenia.

Fig. 2.

Top 10 salient regions contributing to GCN and SVM classification. The size of each region indicates the magnitude of contribution. Bar plots are used to illustrate statistically significant differences in topological characteristics between patients with schizophrenia and controls. Note: AMYG, amygdale; ANG, angular gyrus; CAU, caudate; GCN, graph convolutional network; MTG, middle temporal gyrus; ORBsup, orbitofrontal gyrus, superior part; ORBsupmed, orbitofrontal gyrus, superior medial part; PAL, pallidum; PUT, putamen; REC, rectus; SFGdor, superior frontal gyrus, dorsal part; SFGmed, superior frontal gyrus, medial part; SVM, support vector machine; TPOmid, temporal pole, middle part; TPOsup, temporal pole, superior part.

Fig. 3.

Correlation between nodal efficiency in the bilateral putamen and pallidum with severity of negative symptoms in individuals with schizophrenia.