Fig. 4: ROC curves from the support vector machine (SVM) method using all topological properties to classify three groups.

Nodal properties (non-hubs of Ne, Dc, Bc and their corresponding hubs), and connections (rich-club, feeder, and local) are selected as features. Combining pairwise comparisons A MDDSI vs CN, B MDDNSI vs CN, C MDDSI vs MDDNSI, ROC curves reveal feeder connections have a better classification effect for all indicators. CN, cognitively normal; MDDSI, major depression disorder with suicidal ideation; MDDNSI, major depression disorder with non-suicidal ideation; AUC, area under curve; ROC, receiver operating characteristic curve; Ne, nodal efficiency; Bc, betweeness centrality; Dc, degree centrality.