Group differences in MEG-ICA derived resting state networks: Application to major depressive disorder

Abstract

Functional magnetic resonance imaging (fMRI) studies have revealed the existence of robust, interconnected brain networks exhibiting correlated low frequency fluctuations during rest, which can be derived by examining inherent spatio-temporal patterns in functional scans independent of any a priori model. In order to explore the electrophysiological underpinnings of these networks, analogous techniques have recently been applied to magnetoencephalography (MEG) data, revealing similar networks that exhibit correlated low frequency fluctuations in the power envelope of beta band (14-30Hz) power. However, studies to date using this technique have concentrated on healthy subjects, and no method has yet been presented for group comparisons. We extended the ICA resting state MEG method to enable group comparisons, and demonstrate the technique in a sample of subjects with major depressive disorder (MDD). We found that the intrinsic resting state networks evident in fMRI appeared to be disrupted in individuals with MDD compared to healthy participants, particularly in the subgenual cingulate, although the electrophysiological correlates of this are unknown. Networks extracted from a combined group of healthy and MDD participants were examined for differences between groups. Individuals with MDD showed reduced correlations between the subgenual anterior cingulate (sgACC) and hippocampus in a network with primary nodes in the precentral and middle frontal gyri. Individuals with MDD also showed increased correlations between insulo-temporal nodes and amygdala compared to healthy controls. To further support our methods and findings, we present test/re-test reliability on independent recordings acquired within the same session. Our results demonstrate that group analyses are possible with the resting state MEG-independent component analysis (ICA) technique, highlighting a new pathway for analysis and discovery. This study also provides the first evidence of altered sgACC connectivity with a motor network. This finding, reliable across multiple sessions, suggests that the sgACC may partially mediate the psychomotor symptoms of MDD via synchronized changes in beta-band power, and expands the idea of the sgACC as a hub region mediating cognitive and emotional symptomatic domains in MDD. Findings of increased connectivity between the amygdala and cortical nodes further support the role of amygdalar networks in mediated depressive symptomatology.

Clinical trials identifier: NCT00024635 (ZIA-MH002927-04).

Keywords: Depression; Magnetoencephalography (MEG); Network.

Conflict of interest statement

5. Conflict of Interest and Financial Disclosures

Funding for this work was supported by the Intramural Research Program of the National Institute of Mental Health, National Institutes of Health (IRP-NIMH-NIH; Clinical Trials Identifier: NCT00024635 (ZIA-MH002927-04), by a NARSAD Independent Investigator to CAZ, and by the Brain & Behavior Mood Disorders Research Award to CAZ. CAZ is listed as a co-inventor on a patent application for the use of ketamine and its metabolites in major depression; he has assigned his rights in the patent to the US Government but will share a percentage of any royalties that may be received by the Government. MLF is listed as a co-inventor on a patent application for the use of scopolamine to treat depression; she has assigned all rights to the US Government but will benefit financially from any royalties received by the US Government. The remaining authors declare no competing financial interests.

Published by Elsevier Inc.

Figures

Figure 1

Illustration of consistency among single participant independent component (IC) maps for two example networks: (A) Sensorimotor (Iq=0.72) and (B) Fronto-parietal, left lateralized (Iq=0.52). The group IC map is given on the left, while individual maps are shown for four arbitrarily chosen participants (the same four subjects are used to illustrate both networks). The sensorimotor network was highly robust across all participants. The fronto-parietal network was less robust in individual participants, although elements were visible in most datasets. All images are thresholded at a regression coefficient of 0.25.

Figure 2

(A) The combined group independent component (IC) with the primary node in the right precentral gyrus (crosshairs at peak in upper frame), with additional smaller nodes in left precentral gyrus and precuneus (crosshairs at peak in lower frame). (B) Image showing the IC map for all healthy controls and (C) patients, un-thresholded, with the crosshairs centered at the areas of significant differences between groups. (D) Areas of significantly decreased connectivity in MDD compared to control participants, thresholded at puncorrected <0.01, with only clusters significant at pcorr<0.05 shown. There is a cluster primarily in the subgenual cingulate cortex extending into the ventrolateral prefrontal cortex (pcorr<0.001, crosshairs centered on peak in left column), as well as a cluster in the hippocampus/parahippocampal gyrus (pcorr<0.001, crosshairs centered on peak in right column). Please note that the extrema of the color scales represent the approximate maximum value in the image.

Figure 3

(A) The combined group independent component (IC) with the primary node in the right insula/temporal cortex (crosshairs at peak in both frames). (B) Image showing the IC map for all healthy controls and (C) patients, un-thresholded, with the crosshairs centered at the areas of significant difference between groups. (D) Areas of significantly increased connectivity in MDD compared to control participants, thresholded at puncorrected <0.01, with only clusters significant at pcorr<0.05 shown. Crosshairs are centered at the peak of the large cluster that encompassed the right amygdala, thalamus, and dorsal cingulate (pcorr<0.001, both left and right columns). Please note that the extrema of the color scales represent the approximate maximum value in the image.

Figure 4

(A) The combined group independent component (IC) with the primary node in the left insula/temporal cortex (crosshairs at peak in both frames). (B) Image showing the IC map for all healthy controls and (C) patients, un-thresholded, with the crosshairs centered at the area of significant difference between groups (left column) and the peak in the vicinity of the amygdala (right column). (D) Image showing areas of significantly decreased connectivity in MDD compared to control participants, thresholded at puncorrected<0.01. The crosshairs are centered on the significant cluster in the left precentral and angular gyrus, shown in the left panel (pcorr<0.03). Note that we thresholded this image to additionally show the cluster containing the highest t-values in the left amygdala, although this cluster did not survive correction for multiple comparisons. Crosshairs are centered on this peak in the right column. Please note that the extrema of the color scales represent the approximate maximum value in the image.

Figure 5

Scatterplots showing connectivity between anatomically defined regions of interest (ROIs) and independent components (ICs) within regions known to be involved in the pathophysiology of MDD. (A) Connectivity between the precentral/precuneus IC and subgenual anterior cingulate (sgACC). (B) and (C) Connectivity between a right insulo-temporal IC and the right and left amygdala, respectively. (D) and (E) Connectivity between a left insulo-temporal IC and the right and left amygdala, respectively. All IC images are thresholded at β=0.25, with color scale maximum set to β=0.75.

Figure 6

Illustration of the concordance between voxel-wise group difference maps for Acquisition 1 (age/gender unbalanced sample overlapping with main study sample) and Acquisition 2 (fully independent recordings from similarly unbalanced sample). (A) Reduced connectivity to the precentral IC in MDD subjects was evident in both acquisitions. Two views (sagittal and axial) are shown to facilitate comparison of both the sgACC and hippocampus/parahippocampal gyrus clusters. Of particular note, compare the independent Acquisition 2 sample to the main finding in Figure 2D. (B) Image shows increased connectivity in the MDD subjects evident at identical statistical thresholds in both samples, analogous to the results in amygdala in Figure 3D. Please note that the extrema of the color scales represent the approximate maximum value in the image.