Frequency-specific electrocorticographic correlates of working memory delay period fMRI activity

Abstract

Electrocorticography (ECoG) and functional MRI (BOLD-fMRI) have been used previously to measure brain activity during working memory delay periods. These studies have separately reported oscillation changes in the theta (4-8 Hz) band and BOLD-fMRI increases during delay periods when information is maintained in memory. However, it is not known how intracranial cortical field potential (CFP) changes relate to BOLD-fMRI responses during delay periods. To answer this question, fMRI was obtained from six epilepsy patients during a visual working memory task. Then, following subdural macroelectrode implant, continuous ECoG was used to record CFPs during the same task. Time-frequency analyses showed delay period gamma band oscillation amplitude increases on electrodes located near fMRI activity, while in the theta band changes were higher for electrodes located away from fMRI activation. The amplitude of the ECoG gamma band response was significantly positively correlated with the fMRI response, while a negative correlation was found for the theta band. The findings are consistent with previous reports of local field potential (LFP) coupling in the gamma band with BOLD-fMRI responses during visual stimulation in monkeys, but are novel in that the relationship reported here persists after the disappearance of visual stimuli while information is being maintained in memory. We conclude that there is a relationship between BOLD-fMRI increases and human working memory delay period gamma oscillation increases and theta decreases. The spectral profile change provides a basis for comparison of working memory delay period BOLD-fMRI with field potential recordings in animals and other human intracranial EEG studies.

Figure 1. Localization of intracranial macroelectrodes and classification of their relationship to fMRI delay activity in a single patient

Electrode locations were localized and confirmed by intraoperative photography (a) by co-registering a high-resolution CT volume (b) to the pre-surgical T1-weighted MRI that was used to reconstruct the cortical surface (c). Electrodes proximal (c, white spheres) to delay period fMRI activity increases (c, orange activity) were categorized separately from electrodes located distal (c, red spheres) to fMRI activity increases or decreases relative to baseline. Other electrodes (c, black spheres) were not considered because they did not meet criteria. An example electrode (superior lateral frontal #4 – SLF4) is highlighted with a yellow circle (a-c). The average BOLD signal from the cluster of fMRI activity near the SLF4 electrode is shown (d, white line) next to the boxcar regressor (d, white line) used to model the eight delay periods within the run (136 TRs, 272 sec total scan time). The time-frequency difference spectrum (delay minus control) is shown (e) for the SLF4 electrode.

Figure 2. Distribution of electrode locations proximal and distal to fMRI delay activity

Intracranial macroelectrode positions are initially modeled as spheres on cortical surface reconstructions in native space. To aide visualization of locations across patients, each electrode sphere is transformed to standard space and displayed here in relation to a gray/white matter border surface reconstruction of the standard space (MNI) N27 brain. White spheres mark electrodes located proximal to fMRI delay activity, and red spheres indicate electrodes located distal to fMRI delay activity.

Figure 3. Individual patient time-frequency difference spectra

Each patient's delay period difference spectrum (proximal minus distal electrodes) is shown in a separate panel (a through f). To the right of each patient's spectrum, a quantification of the positive (red bar, proximal>distal) and negative (blue bar, proximal

Figure 4. Group analysis time-frequency difference spectrum

Each patient's delay period difference spectrum (proximal minus distal) was averaged to produce a single time-frequency spectrum (a) where green-to-red colors indicate greater change for proximal relative to distal electrodes and green-to-blue indicates greater change for distal relative to proximal electrodes. In each band, positive (b, red, proximal>distal) and negative (b, blue, proximal

Figure 5. Relationship between amplitude of ECoG and fMRI delay activity

For all 118 proximal and distal electrodes, the amplitude of delay period ECoG activity was plotted against the fMRI activity. A significant negative correlation was found in the theta band (a, r=-0.22, p=0.02) and a positive correlation was found in the mid-gamma band (b, r=0.19, p=0.04).

Figure 6. Theta band power spectral density is shifted lower for cortical potentials recorded near fMRI activity

Power spectral density across the delay interval was computed from the periodogram of the ECoG signals in each patient separately for the set of electrodes proximal to fMRI activity (white) and for the set of electrodes distal to any fMRI activity (red). A consistent and significant (p=0.02) shift in the spectrum was noted in five out of six patients indicating that total power in the theta band is relatively higher for cortical potentials recorded away from fMRI delay activity.

Figure 7. Frequency-Specific Delay Period Power Increases Irrespective of BOLD-fMRI Activity

A total of 397 electrodes across the 6 patients were evaluated for significant (> 2 std. change) in the delay period relative to baseline. Of the total 397 electrodes, 253 (63.7%) were sorted into three categories according to sustained increases in either the gamma band (a, 45 or 11.3% of electrodes), the theta band (b, 112 or 28.2% of electrodes) or both the gamma and theta bands (c, 96 or 24.2% of electrodes).