Synopsis

Keywords: Psychiatric Disorders, fMRI (resting state), Stereoencephalography (sEEG)

Here, we obtained pre-operative fMRI and intracranial recordings from two subjects with treatment-refractory depression undergoing stereoencephalography as part of a DBS study. Intracranial recordings were obtained from leads bilaterally implanted in the orbitofrontal cortex, subgenual cingulate, ventral striatum, hippocampus, and amygdala. fMRI-based resting-state functional connectivity (RSFC) was calculated for ROIs defined by the recording sites. We identified significant spatial correlations between power bandpassed in canonical frequency bands and fMRI RSFC. Correlations between both modalities were highest within the beta band. These data suggest that it is possible to map between networks defined by fMRI and intracranial electrophysiology.

Introduction

Mental health disorders, such as major depressive disorder, are widely believed to be mediated by dysfunctional brain networks. Non-invasive resting-state fMRI (rsfMRI) and intracranial electrophysiological recordings represent two methods for characterizing pathological brain networks for targeting therapeutic neuromodulation. While fMRI is spatially comprehensive, intracranial recordings are a more direct measure of brain electrophysiological activity and can resolve oscillatory brain activity due to its high temporal resolution. Previous studies using invasive electrophysiological recordings in clinical populations with depression and anxiety have identified neural oscillatory activity in the beta band (14-30 Hz) across corticolimbic structures as a correlate of symptom severity^1-3. Here, we aim to characterize the relationship between resting state fMRI and electrophysiological functional connectivity across low and high frequencies in cortico-striatal-limbic networks for two subjects with treatment resistant depression (TRD).

Methods

Two subjects with severe TRD underwent stereoencephalography (sEEG) monitoring to inform the implantation of a deep brain stimulation (DBS) device as a part of the PRESIDIO trial at the University of California San Francisco. Pre-operative rsfMRI data were collected prior to implantation of bilateral intracranial leads in the orbitofrontal cortex, subgenual cingulate, ventral striatum, hippocampus, and amygdala. We acquired gradient-echo rsfMRI data using a 3T GE scanner with a 32-channel head coil, TR/TE=2s/30ms, voxel size=3.75x3.75x4cm, flip angle=86, and FOV=24cm.

The fMRI data were preprocessed with fMRIprep. Specific preprocessing steps included nonsteady state volume removal, slice-time correction, motion correction, normalization to MNI space, regression of cerebrospinal fluid and white matter global signal in addition to temporal derivatives and quadratic terms, bandpass filtering between 0.008 and 0.1 Hz, spatial smoothing with a 6mm FWHM kernel, and co-registration to the T1 weighted image. Spherical ROIs with 6 mm radii were generated at the electrode contact sites (N= 160) and the time series was extracted and averaged in each ROI using the AFNI toolbox. RSFC was then calculated as the correlation between mean timeseries for all possible combinations of electrode ROIs.

For the sEEG data, high gamma (70-150 Hz), low gamma (30-70 Hz), beta (14-30 Hz), alpha (8-14 Hz), theta (4-8 Hz), delta (1-4 Hz) power envelopes were extracted using the Hilbert transform. Power envelopes were normalized across each channel and averaged across 1s bins. Electrophysiological FC was measured as the correlation of the envelope of each frequency band between recording sites.

Spatial correlation between seed-based electrophysiological FC for each frequency range and fMRI RSFC was calculated with a (Fisher r-to-z transformed) Pearson correlation. We then compared the correlation between FC_fMRI-FC_sEEG for each frequency range.

Results

We studied two subjects with treatment resistant depression who had intracranial electrodes implanted in limbic networks (Fig. 1A). We first derived functional connectivity maps for fMRI fMRI FC and electrophysiological FC at each frequency range (Fig. 1B, C). In both subjects, sEEG FC in each frequency band was spatially correlated to fMRI RSFC in all subjects. Correlations between sEEG and fMRI were highest for the beta band [S1: mean r = 0.49, P < 0.0001, S2: mean r = 0.46, P < 0.0001] compared to other bands across pairs of electrode contacts/ROIs (Fig. 2A). There were significant correlation between fMRI-based FC and sEEG-based FC in multiple regions including the amygdala, hippocampus, and orbitofrontal cortex (Fig. 2B).

Discussion

Here, we show a strong spatial correlation between electrophysiological FC and fMRI RSFC in a unique cohort of subjects with TRD implanted with electrodes in the corticolimbic networks. Interestingly, electrophysiological FC in the beta band correlated best with fMRI RSFC in both subjects. This relationship was consistent in several corticolimbic structures, including the amygdala, hippocampus, and orbitofrontal cortex.

Our results replicate findings from previous studies showing correspondence between rsfMRI and electophysiological FC across multiple frequency ranges^4,5. Recent studies have associated beta frequency activity with anxious thinking and depression^1-3. It is therefore possible that beta activity is more robust in limbic networks of individuals with TRD.

Conclusion

Our results thus far suggest a mapping between depression networks identified using invasive and non-invasive methods for understanding the spatial organizations of brain networks.

Acknowledgements

These studies were generously funded by the Ray and Dagmar Dolby Family Fund and the Foundation for OCD Research.