Spatiotemporal dynamics of low frequency fluctuations in BOLD fMRI of the rat

Abstract

Purpose: To examine spatiotemporal dynamics of low frequency fluctuations in rat cortex.

Materials and methods: Gradient-echo echo-planar imaging images were acquired from anesthetized rats (repetition time = 100 ms). Power spectral analysis was performed to detect different frequency peaks. Functional connectivity maps were obtained for the frequency peaks of interest. The images in the filtered time-series were displayed as a movie to study spatiotemporal patterns in the data for frequency bands of interest.

Results: High temporal and spectral resolution allowed separation of primary components of physiological noise and visualization of spectral details. Two low frequency peaks with distinct characteristics were observed. Selective visualization of the second low frequency peak revealed waves of activity that typically began in the secondary somatosensory cortex and propagated to the primary motor cortex.

Conclusion: To date, analysis of these fluctuations has focused on the detection of functional networks assuming steady state conditions. These results suggest that detailed examination of the spatiotemporal dynamics of the low frequency fluctuations may provide more insight into brain function, and add a new perspective to the analysis of resting state fMRI data.

Figure 1

Location of seed on EPI image, power spectrum of the signal obtained from seed location (with time-course normalized to unit variance), and band-limited maps of temporal standard deviation for a rat with two clear low-frequency peaks (a), a rat with a single low-frequency peak (b), a dead rat (c), and a rat that exhibited significant cardiac and respiratory signal (d). a: Power spectrum for a cortical seed ROI shows two distinct peaks for this rat. Spatial maps for LF1 (column 3) and LF2 (column 4) contributions demonstrate high cortical specificity for both peaks, with highest signal magnitude near the sagittal sinus for LF1. b: Power spectrum from another rat exhibits only one clear peak (LF1). The peak map for LF1 contribution (column 3) shows high specificity to cortex, again with a focal increase in power near the sagittal sinus. The power in the LF2 range is much lower and primarily confined to the surface of the brain. c: Power spectrum from a dead rat showing no clear peaks. The 0-0.05 Hz (column 3) and 0.08-0.2 Hz (column 4) contributions are nonspecific. d: Power spectrum from a cortical seed placed near the draining veins exhibits respiratory and cardiac contributions. Spatial distribution map for cardiac contribution (column 3) shows high specificity to the area near draining veins, whereas respiratory contribution (column 4) is strongest around the sagittal sinus, draining veins, and ventricles. [Color figure can be viewed in the online issue, which is available at www.interscience.wiley.com.]

Figure 2

Connectivity maps created using time-courses filtered to retain LF1 (middle column) or LF2 (right column). The left column shows EPI images overlaid with seed ROIs. a: Connectivity maps from a rat with two well-defined spectral peaks. The seed locations are SI (first row) and SII (second row). Stronger correlation and less specificity is observed for LF1 compared with LF2. The correlation maps are less dependent upon the seed location for LF1. b: Connectivity maps from a rat with a single spectral peak. LF1 connectivity maps show strong correlation throughout the cortex that is relatively insensitive to the location of the seed, similar to a. The LF2 peak is more specific but less bilateral connectivity is observed than for the dataset with two peaks.

Figure 3

a: Average correlation in the region of interest (ROI) defined in the contralateral SI for a seed placed in SI. Rats 1-3 showed two clear peaks. Stronger correlation is observed for LF1 for all the datasets. b: Correlation matrix for average time-courses for different locations. Overall, stronger correlation is observed for LF1. c,d: Number of voxels with correlation coefficient > 0.5 for seeds placed in SI and SII: Pixels crossing the threshold are restricted to the somatosensory cortex for LF2, whereas pixels crossing the threshold are distributed among SI, SII, and CP for LF1. Also, the extent of the correlation in cortex and CP does not vary with the cortical seed location for LF1 (maximum difference < 2 voxels for seeds placed in SI and SII). e: Number of voxels with correlation coefficient > 0.5 for seed placed in the CP: No pixels in SI and SII show correlation > 0.5 for LF2, whereas the pixels crossing the threshold are distributed among SI, SII, and CP for LF1. Please note that only three datasets showing the two peaks were used for the results shown in Figure 3. Inclusion of the remaining three datasets yields qualitatively similar results. [Color figure can be viewed in the online issue, which is available at www.interscience.wiley.com]

Figure 4

Spatiotemporal dynamics for LF2 and LF1 peaks. a: Propagating waves were observed in the datasets filtered to retain LF2. A bilateral wave of low signal intensity in SI moves medially as areas of high intensity arise in SII (2.5-3 s after the start of the sequence). The wave of high intensity moves medially, and the cycle begins again with the appearance of bilateral areas of low intensity in SII (6.5 s). b: Waves observed in the LF2-filtered data in another rat. The pattern of propagation is similar to a. c: LF1 shows completely different propagation patterns. The most prominent pattern was a slow propagation of signal intensity from the surface of the brain inward. The pattern moves through layers and does not show functional specificity. No cortical waves similar to those shown in a and b were observed.

Figure 5

Temporal pattern of occurrence of the waves: Plot of the occurrence of waves (after filtering the data to retain LF2) for all three rats that exhibited a clear LF2 peak is shown. The waves appear to occur in groups. Alternate waves are colored differently to separate consecutive occurrences. The duration of the waves is defined as the time required for the high intensity to travel from SII to MI.

Figure 6

Spatiotemporal dynamics related to physiological or scanner noise. No patterns of cortical waves similar to those shown in Figure 4a and b were observed. a: Contribution from the primary cardiac peak. The only clear pattern forms around the draining veins along the surface of the cortex (pointed by white arrow), which alternately brighten and darken. b: Contribution from the primary respiratory peak. Periodic changes in signal intensity are apparent in the draining veins along the surface (white arrow) and the areas near the ventricles (red arrow). Interestingly, these changes are out of phase. c: Contribution of LF2 frequency range in the dead rat. No specific patterns can be detected, indicating that scanner noise is not the source of the waves. d,e: Time-courses for respiratory (d) and cardiac (e) contributions. Respiratory contributions from ventricles and draining veins have different phases. The timecourses were normalized to unit variance before plotting.