Bidirectional connectivity alterations in schizophrenia: a multivariate, machine-learning approach

Abstract

Objective: It is well known that altered functional connectivity is a robust neuroimaging marker of schizophrenia. However, there is inconsistency in the direction of alterations, i.e., increased or decreased connectivity. In this study, we aimed to determine the direction of the connectivity alteration associated with schizophrenia using a multivariate, data-driven approach.

Methods: Resting-state functional magnetic resonance imaging data were acquired from 109 individuals with schizophrenia and 120 controls across two openly available datasets. A whole-brain resting-state functional connectivity (rsFC) matrix was computed for each individual. A modified connectome-based predictive model (CPM) with a support vector machine (SVM) was used to classify patients and controls. We conducted a series of multivariate classification analyses using three different feature sets, increased, decreased, and both increased and decreased rsFC.

Results: For both datasets, combining information from both increased and decreased rsFC substantially improved prediction accuracy (Dataset 1: accuracy = 70.2%, permutation p = 0.001; Dataset 2: accuracy = 64.4%, permutation p = 0.003). When tested across datasets, the prediction model using decreased rsFC performed best. The identified predictive features of decreased rsFC were distributed mostly in the motor network for both datasets.

Conclusion: These findings suggest that bidirectional alterations in rsFC are distributed in schizophrenia patients, with the pattern of decreased rsFC being more similar across different populations.

Keywords: connectome-based predictive modeling; machine learning; multivariate analysis; resting state functional connectivity; schizophrenia.

Figure 1

Schematic of multivariate classification using resting-state functional connectivity (rsFC). CPM, connectome-based predictive modeling, SVM, support vector machine.

Figure 2

Results of the classification analysis. Panel A shows the results of the within-dataset leave-one-out classification. For both datasets, CPM-SVM classification using both increased and decreased edges showed the highest classification accuracy with a permutation test p < 0.01 (last two columns). Panel B shows the results of the external dataset classification. When Dataset 1 was used as the training set, CPM-SVM classification using the decreased matrix and both matrices could predict subjects in Dataset 2 above chance level. (* in the figure denotes both the permutation test and sign test p < 0.02). Panel C (UCLA dataset) and D (COBRE dataset) shows the confusion matrix of the within-dataset classification result. Blue boxes indicate individuals correctly identified by the model.

Figure 3

The edges selected in CPM-SVM prediction between datasets (upper) and the ‘consensus edges’ selected during the of the leave-one-out process (lower). The numbers denote the number of edges selected. Those edges that decreased in the schizophrenia patients linked the motor to other networks (blue matrix plots). Meanwhile, the edges that increased in the patient group did not have a similar pattern between the two datasets (red matrix plots). CBL, cerebellum; SC, subcortical; SAL, salience; Vas, visual association; VI, visual A; VII, visual B; Mot, motor; DMN, default mode network; FP, frontoparietal; MF, medial frontal.