Dysconnection topography in schizophrenia revealed with state-space analysis of EEG

Abstract

Background: The dysconnection hypothesis has been proposed to account for pathophysiological mechanisms underlying schizophrenia. Widespread structural changes suggesting abnormal connectivity in schizophrenia have been imaged. A functional counterpart of the structural maps would be the EEG synchronization maps. However, due to the limits of currently used bivariate methods, functional correlates of dysconnection are limited to the isolated measurements of synchronization between preselected pairs of EEG signals.

Methods/results: To reveal a whole-head synchronization topography in schizophrenia, we applied a new method of multivariate synchronization analysis called S-estimator to the resting dense-array (128 channels) EEG obtained from 14 patients and 14 controls. This method determines synchronization from the embedding dimension in a state-space domain based on the theoretical consequence of the cooperative behavior of simultaneous time series-the shrinking of the state-space embedding dimension. The S-estimator imaging revealed a specific synchronization landscape in schizophrenia patients. Its main features included bilaterally increased synchronization over temporal brain regions and decreased synchronization over the postcentral/parietal region neighboring the midline. The synchronization topography was stable over the course of several months and correlated with the severity of schizophrenia symptoms. In particular, direct correlations linked positive, negative, and general psychopathological symptoms to the hyper-synchronized temporal clusters over both hemispheres. Along with these correlations, general psychopathological symptoms inversely correlated within the hypo-synchronized postcentral midline region. While being similar to the structural maps of cortical changes in schizophrenia, the S-maps go beyond the topography limits, demonstrating a novel aspect of the abnormalities of functional cooperation: namely, regionally reduced or enhanced connectivity.

Conclusion/significance: The new method of multivariate synchronization significantly boosts the potential of EEG as an imaging technique compatible with other imaging modalities. Its application to schizophrenia research shows that schizophrenia can be explained within the concept of neural dysconnection across and within large-scale brain networks.

Figure 1. Head diagram of the EEG sensor positions and labeling.

The diagram shows the correspondence between the high-density 129-channel Sensor Net (EGI, Inc.) and the International 10–10 System. The Sensor Net locations that match the positions of International 10–10 system are labeled. The 10–10 System names are followed by the numbers of the Sensor Net. The sensors corresponding to the 10–20 System (presented in all the maps hereafter) are in bold. The gray background highlights all the sensors included in the analyses. The sensor locations encircled in green exemplify the first and second neighborhoods for the sensor encircled in brown (sensor 73), i.e., the territory considered in the calculation of a single value of S-estimator.

Figure 2. Whole-head S-estimator maps for SZ patients and normal controls.

Group-averaged maps for the broad-band resting EEG (1–70 Hz) are shown for patients (A) and controls (B). C: Difference map, Patients vs. Controls. Here and hereafter the clusters of sensors with S-estimator significantly higher or lower in patients than in controls are in red or blue, respectively. There are no significant differences in the gray regions.

Figure 3. Spectral breakdown of the S-estimator data into conventional EEG bands.

Group-averaged difference S-maps (Patients vs. Controls) are given for the following EEG bands: δ: 1–3 Hz; θ: 3–7 Hz; α: 7–13 Hz; β: 13–30 Hz; and γ: 30–70 Hz.

Figure 4. Temporal stability of the S-estimator topography in SZ.

Group-averaged difference maps (Patients vs. Controls) for the broad-band EEG from ten patients who participated in the first (A) and second (B) EEG sessions. C: Difference map between the first EEG vs. second EEG of patients. The regions where the S-estimator in the patients' first EEG was significantly higher or lower than that in the second EEG are in red or in blue, respectively. There are no significant differences in the gray regions. D: Map of the likelihood that the synchronization estimations from the first and second EEG can be considered stationary according to Kolmogorov-Smirnov test. The color is inversely related to the probability: the lighter, the more probable.

Figure 5. Correlation between synchronization and SZ symptoms.

The topographies of correlations between the S-estimator changes in patients and their symptoms as measured by the Positive Symptom Scale (PS), Negative Symptom Scale (NS), and General Psychopathology Scale (GP) are shown. The regions where the significant correlations are direct or inverse are marked in brown or turquoise, respectively. There are no significant correlations in the gray regions.

Figure 6. Whole-head power maps for SZ patients and normal controls.

The group-averaged maps of absolute power for the broad-band resting EEG are shown for patients (A) and for controls (B). C: Difference map (Patients vs. Controls). In the blue regions the absolute power in patients is significantly lower than that in controls. There are no significant differences in the gray regions.

Figure 7. Relationship between synchronization and power.

A: Correlation map between the relative changes of S-estimator and the relative changes of power for patients. The correspondence between color and correlation strength is shown by the scale bar. B: Significance map for the correlations shown in A; color convention is as in Fig. 5.