Complex relationship between BOLD signal and synchronization/desynchronization of human brain MEG oscillations

Abstract

Functional magnetic resonance imaging (fMRI) depends on the coupling of cerebral blood flow, energy demand, and neural activity. The precise nature of this interaction, however, is poorly understood. A positive correlation between BOLD-response and cortically generated local field potentials, which reflect the weighted average of synchronized dentrosomatic components of pyramidal synaptic signals, has been demonstrated. Likewise, positive BOLD-responses have been reported in conjunction with scalp-recorded synchronized electromagnetic activity by a number of groups. However, it is not yet clear how the opposite electromagnetic pattern, i.e. cortical desynchronization, is related to the BOLD signal. To address this question, we conducted a combined event-related fMRI and 275 sensor whole-head MEG study during identical visual two-choice reaction time task conditions in 10 human subjects. We found complex sequences of MEG-synchronization and desynchronization across a wide frequency range in the visual and motor area in close correspondence with "locales" of positive BOLD-responses. These results indicate that a correspondence of positive BOLD-responses is not exclusively found for cortical synchronization but also for desynchronization, suggesting that the relationship between BOLD signals and electromagnetic activity might be more complex than previously thought.

Figure 1

Sensor array consisting of 275 radial first‐order gradiometers uniformly distributed over both hemispheres, giving an average inter‐channel spacing of 2.2 cm.

Figure 2

Time‐frequency spectrogram across the entire group of subjects (n = 10), i.e., the time‐frequency analysis was done on the separate time series of each study subject (raw data) and the time‐frequency representations were then averaged. The plots are an average of all sensors across all subjects. Top: time–frequency analysis is referenced to the visual stimuli. Bottom: time–frequency analysis referenced to the motor reaction.

Figure 3

Time‐frequency spectrogram before and after averaging of data across the entire group of subjects (n = 10). Top: time–frequency plots in occipital (visual) cortex (green in Fig. 1) are referenced to the visual stimuli. Bottom: time–frequency plots in central (motor) cortex area (purple in Fig. 1) referenced to the motor reaction (also see Fig. 1). Left side: time–frequency plots of raw data (before averaging). Right side: time–frequency plots of averaging data (i.e., averaged event‐related magnetic fields).

Figure 4

A surface‐rendered group‐averaged (n = 10) overview is provided on the task‐related BOLD‐response (upper left and lower left). Note that convolving the hemodynamic response function with stimuli and responses gave identical results. The motor response‐related θ‐ERS (upper right) and the stimulus‐related α‐ERD (lower right) are also depicted.

Figure 5

Group MEG SAM analysis and event‐related fMRI analysis (n = 10 subjects). In the SAM analysis, post‐stimulus activity (sliding time‐window searching for maximum activity change) is compared to 300 ms prestimulus in the α frequency band (combined analysis for left and right stimulus). Blue indicates a significant decrease in power (P < 0.05) associated with desynchronization (ERD). In the fMRI group analysis, the BOLD‐response is thresholded at a Z value of 2.8 (P = 0.05, two‐tailed). α‐Desynchronization is seen in the left and right visual cortex area and in the parietal area where the MEG‐activity change is stronger on the right side whereas the BOLD‐response is stronger on the left side.

Figure 6

Group MEG SAM analysis and event‐related fMRI analysis (n = 10 subjects). In the SAM analysis, post‐stimulus activity (sliding time‐window searching for maximum activity change) is compared to 300 ms pre‐stimulus in the β (left) and γ (right) frequency bands (combined analysis for left and right stimulus). Blue indicates a significant decrease in power (P < 0.05) associated with event‐related desynchronization (ERD). In the fMRI group analysis, the BOLD‐response is thresholded at a Z value of 2.8 (P = 0.05, two‐tailed). β‐ERD and γ‐ERD are both seen in the left and right motor area. The MEG‐activity change are relatively symmetrical in both hemispheres whereas the BOLD‐response is stronger on the left side.

Figure 7

Group MEG SAM analysis and event‐related fMRI analysis (n = 10 subjects). In the SAM analysis, response‐locked activity (sliding time‐window searching for maximum activity change) is compared to 300 ms preresponse in the θ‐frequency band (combined analysis for left and right stimulus). Red indicates a significant increase in power (P < 0.05) associated with synchronization (ERS). θ‐ERS is seen in the motor area and ACC. In the fMRI group analysis, the BOLD‐response in the motor area is thresholded at a Z value of 2.0 (P = 0.05, two‐tailed). The BOLD‐response in the ACC is thresholded at a Z value of 1.7 (P = 0.05, one‐tailed). The MEG‐activity changes are relatively asymmetrical with θ in the right motor area and right ACC being stronger than on the left side.

Figure 8

MEG event‐related SAM (SAMerf) analysis of stimulus‐related activity (1–30 Hz) changes (n = 10 subjects). The virtual channel time series for each stimulus were averaged (i.e., averaging across single trials) and a sliding window ratio of post‐ to pre‐stimulus power was used to generate functional images. 300 ms sliding time window (center peak at 150 ms post‐stimulus) was compared to a fixed 300 ms pre‐stimulus window, with log power ratios calculated for each voxel. Individual volumes were normalized to Z‐scores and a t‐statistic was calculated for group averages (P < 0.05, two‐tailed).

Figure 9

Correlation of θ activity (ERS) and motor response time thresholded at Spearman's R = −0.7.