Mapping Between fMRI and Electrophysiological Networks Encoding Depressive Symptoms

2614 

Abstract Submission 

John Alvarez¹, Niki Sabetfakhri¹, Joline Fan¹, Natalya Slepneva¹, Julian Motzkin¹, Melanie Morrison¹, Leo Sugrue¹, Kristin Sellers¹, Edward Chang¹, Andrew Krystal¹, A Moses Lee¹

¹University of California, San Francisco, San Francisco, CA

John Alvarez  
University of California, San Francisco
San Francisco, CA

Niki Sabetfakhri  
University of California, San Francisco
San Francisco, CA

Joline Fan  
University of California, San Francisco
San Francisco, CA

Natalya Slepneva  
University of California, San Francisco
San Francisco, CA

Julian Motzkin  
University of California, San Francisco
San Francisco, CA

Melanie Morrison  
University of California, San Francisco
San Francisco, CA

Leo Sugrue  
University of California, San Francisco
San Francisco, CA

Kristin Sellers  
University of California, San Francisco
San Francisco, CA

Edward Chang  
University of California, San Francisco
San Francisco, CA

Andrew Krystal  
University of California, San Francisco
San Francisco, CA

A Moses Lee  
University of California, San Francisco
San Francisco, CA

Multiple prior imaging studies have implicated a role for network abnormalities in depression. Non-invasive methods like resting-state functional magnetic resonance imaging (rs-fMRI) are able to comprehensively characterize the spatial organization of functional connectivity (FC) networks while invasive intracranial electroencephalography (iEEG) can characterize the dynamics of networks with unparalleled temporal resolution across varying time scales. Here, we leverage the complementary spatiotemporal strengths of these modalities, which provide a unique opportunity to investigate the electrophysiological basis of canonical fMRI networks while characterizing their relationship to depressive symptoms.

Two participants with severe treatment-resistant depression underwent a brain mapping protocol, involving extensive fMRI followed by 10-day iEEG monitoring to identify personalized targets for therapeutic deep brain stimulation. iEEG leads recorded from bilateral contacts in the orbitofrontal cortex, subgenual cingulate, ventral striatum, hippocampus, and amygdala (Figure 1A). iEEG data was preprocessed and segmented into 5-minute recordings sessions corresponding with subjects' symptom self-reports (Figure 1B-C).

We computed mean iEEG-FC for each recording session across canonical frequency bands by calculating mean phase-locking values (PLV) and band-limited power (BLP) correlations between electrode pairs (Figure 1B-C). rs-fMRI data was preprocessed using fMRIprep. We then used spherical regions of interest derived from electrode locations to calculate fMRI-FC (Figure 1D). Pearson correlation was calculated between fMRI-FC and iEEG-FC.

We then trained a support vector machine (SVM) model on parcellated networks' cortical and subcortical FC patterns to predict network affiliations of iEEG electrodes (Figure 1E). We quantified correspondence between spatial organization of iEEG and fMRI networks using mutual information analysis and permutation testing. To characterize iEEG network dynamics, we calculated temporal correlations between z-scored intranetwork electrode pairs and determined if networks exhibited significantly correlated dynamics based upon permutation testing. Pearson correlations were calculated between self-reported depression scores and iEEG network FC. We then ran permutation tests to identify networks across frequency bands that significantly correlated with depressive symptoms.

We demonstrate that there is a significant spatial correspondence between iEEG functional connectivity (FC) and fMRI BOLD-FC corresponding regions-of-interest near the electrodes. These spatial correlations between BOLD-FC and iEEG phase-locking value (PLV) and band-limited power (BLP) were observed across frequency bands after correcting for distance (p<.0001, permutation test). We then demonstrate that electrophysiological networks share a similar topological organization with each individual's canonical fMRI networks based upon a mutual information metric (p<.001, permutation test). These electrophysiological networks have correlated intranetwork dynamics within specific frequency bands. In turn, fluctuations in the intranetwork FC within subcomponents of the default mode network (DMN) correlated with depression (VAS-D/HAMD6) over the course of days within the two subjects (p<.05, permutation test).

Together, these results support a model in which synchronous, correlated electrophysiological activity gives rise to the spontaneous spatial fluctuations in BOLD characterizing canonical fMRI networks. Further, these results suggest that the dynamics of DMN subnetworks encode variation in depressive symptoms over time.

Deep Brain Stimulation

Psychiatric (eg. Depression, Anxiety, Schizophrenia)

EEG/MEG Modeling and Analysis

fMRI Connectivity and Network Modeling ²

Neurophysiology of Imaging Signals ¹

Affective Disorders

Computational Neuroscience

DISORDERS

ELECTROPHYSIOLOGY

FUNCTIONAL MRI