Time lag dependent multimodal processing of concurrent fMRI and near-infrared spectroscopy (NIRS) data suggests a global circulatory origin for low-frequency oscillation signals in human brain

Abstract

Low frequency oscillations (LFOs), characterized by frequencies in the range 0.01-0.1 Hz are commonly observed in blood-related brain functional measurements such as near-infrared spectroscopy (NIRS) and functional magnetic resonance imaging (fMRI). While their physiological origin and implications are not fully understood, these signals are believed to reflect some types of neuronal signaling, systemic hemodynamics, and/or cerebral vascular auto-regulation processes. Here, we examine a new method of integrated processing of concurrent NIRS and fMRI data collected on six human subjects during a whole brain resting state acquisition. The method combines the high spatial resolution offered by fMRI (approximately 3mm) and the high temporal resolution offered by NIRS (approximately 80 ms) to allow for the quantitative assessment of temporal relationships between the LFOs observed at different spatial locations in fMRI data. This temporal relationship allowed us to infer that the origin of a large proportion of the LFOs is independent of the baseline neural activity. The spatio-temporal pattern of LFOs detected by NIRS and fMRI evolves temporally through the brain in a way that resembles cerebral blood flow dynamics. Our results suggest that a major component of the LFOs arise from fluctuations in the blood flow and hemoglobin oxygenation at a global circulatory system level.

Copyright 2010 Elsevier Inc. All rights reserved.

Figures

Figure 1

3-D reconstructions of Subject 5’s forehead from the structural scan using OsiriX (Rosset et al., 2004). The positions of the detectors and sources are clearly marked by the MR-visible markers and some imprints of the source and detector fibers. The distances between paired sources and detectors are 3 cm (C1 and D2) and 1 cm (B1).

Figure 2

Temporal trace of data collected over path C1 for Subject 3. (a) Original Δ[HbO] (blue) and its resampled data (red). (b) Enlarged section of original and resampled Δ[HbO] indicated by the black block in (a). (c) 21 regressors of various time shifting (separated by 0.72 s) used in the study from Δ[HbO]. (d), (e) and (f) are the corresponding graphs of (a), (b) and (c) with Δ[Hb] calculated from data collected over the same path.

Figure 3

The left panel (a, c, e, g, i, k) are the power spectra of Δ[HbO] (calculated from path C1), oxygenation and respiration waves obtained by the simultaneous recording of the subject during the experiment through the finger-tip oximeter and respiration belt for the subject 1–6. The right panel (b, d, f, h, j, l) are the corresponding spectra of Δ[HbO] for the subject 1–6, calculated from path C1 and their resampled data after high-pass filtered by FEAT at 0.2 Hz. Note after resampling, the main component of the signal is the LFO.

Figure 4

z-statistic maps of the brain (subject 3) using NIRS Δ[HbO] (path C1) as regressors that shifted from −7.2 to 7.2 s seconds in 0.72 second steps. The number on the upper left corner of each graph indicates the regressor’s time shift for that analysis. The z-statistic map with the green boundary is the map in which the NIRS regressor has not been shifted. The red circle calls out the position of one of the markers, which is closer to the source.

Figure 5

Illustration of the procedure for calculating the travel time of the LFO from one voxel to the other. (a) Sagittal view of Subject 3 with z-statistic map at 0 time shift overlaid; two voxels on the superior sagittal sinus are chosen, as shown in purple and green circles. (b) plot of z-value vs. time shift for the two voxels in (a) with their maxima marked. (c) two z-statistic maps of different time shifts were picked from Figure 4 to select two voxels appeared at beginning and end of the wave’s passage. (d) plot of z-value v.s. time shift for the two voxels in (c) with their maxima marked.

Figure 6

Comparison of the 3-D rendered images of the subject 3’s head using different methods. (a) 3-D rendering structural images of the subject 3’s head with the markers on it. (b) 3-D reconstructions of phase contrast images showing the main blood vessels. (c) 3-D reconstructions using the maximum z-value over time lag maps of the NIRS Δ[HbO] (z > 4).

Figure 7

z-statistic maps of the brain (Subject 3) using NIRS −Δ[Hb] (path C1) as shifted from −7.2 to 7.2 s seconds in 0.72 second steps. The number on the upper left corner of each graph indicates the regressor’s time shift for that analysis. The z-statistic map with the green boundary is the map, in which the NIRS regressor has not been shifted. The red circle calls out the position of one of the markers, which is closer to the source.

Figure 8

z-statistic maps from a midline sagittal slice (Subjects 1–6), using NIRS Δ[HbO] (path C1) data as regressors shifted from −7.2 to 7.2 s seconds in 0.72 second steps. The z-statistic maps with the green boundaries are the maps, in which the NIRS regressor has not been shifted.

Figure 9

z-statistic maps from an axial slice parallel to the AC-PC plane through the lateral ventricles (Subjects 1–6), using NIRS Δ[HbO] (path C1) data as regressors shifted from −7.2 to 7.2 s seconds in 0.72 second steps. The z-statistic maps with the green boundaries are the maps, in which the NIRS regressor has not been shifted.