Low-frequency coherent fluctuations in BOLD activity: a preliminary report

Background
Low frequency (<0.1Hz) fluctuations originating from blood flow and oxygenation have been observed in the brain by different groups (Golanov et al. 1994; Biswal.et al. 1995). The goal of this study was to use BOLD fMRI to characterize coherent fluctuations in those low frequencies between spatially distant brain regions.

Methods
Thirteen right-handed subjects were studied using blocked-design BOLD fMRI at rest and as they performed sequential finger movements at a slow rate (~0.5 Hz) with their right hand. Serial acquisitions of EPI images were obtained at 3T using a single-shot 2D gradient-echo echo-planar imaging sequence. Data were processed and analyzed using standard procedures implemented in the statistical parametric mapping software (SPM2). Temporal profile of brain activity in 6 predefined regions (left S1M1, right S1M1, SMA, left thalamus, right cerebellum and CSF) were extracted on a subject-by-subject basis using the VOI tool in SPM2. After deconvolution (Gitelman et al. 2003), time series data representing movement and resting conditions were concatenated to create 2 within-condition time-series. After subtracting the mean over all epochs from each epoch, the (complex) coherency was calculated in the 5 lowest frequency bins with a frequency resolution of 1/17.5Hz (0, 0.05, 0.11, 0.17, 0.23 Hz). Real and imaginary parts of coherency, representing correlation and correlation of phase-shifted signals, respectively, were analyzed separately. Coherency was computed independently for each region pairs and each subject. Significance was defined as p<0.05 Bonferroni corrected for multiple comparisons.

Results
The main finding was the presence of coherent fluctuations in BOLD signal mainly in the lowest frequencies for (almost) all regions. The real part of coherency was equally pronounced during the movement and rest conditions. The only exception was a larger coherence during rest than during the task condition in the lowest frequencies between left and right S1M1. We could not find any significant imaginary part of coherency indicating that the time delays between dependent neural activations are negligible compared to the inverse frequencies under study. Partialling out the data in any of the regions did not have any significant impact on coherence map. Importantly, there was no evidence of coherent activity between any of the brain regions and CSF in any of the frequency bands.

Discussion
BOLD signal recorded during different behavioral steady-states showed very similar coherent fluctuations for all regions pairs studied mainly in the lowest frequencies. Those results are in good agreement with electrophysiological recordings in monkeys in which high coherence in band-limited power of local field potential signals have been reported at very low frequencies (<0.1Hz) (Leopold et al. 2003). In that study, coherence patterns were also highly similar under distinctly different behavioral states. Task-independent coherence in lower frequencies may be related to whole brain slow synchronous oscillations whose significance remains to be elucidated.

References
Golanov et al. (1994). Am J Physiol. 266; R204-214.
Biswal et al. (1995). Magn Reson Med. 34; 537-541.
Gitelman et al. (2003). NeuroImage. 19; 200-207.
Leopold. et al. (2003). Cereb Cortex. 13; 422-433.