Separating respiratory-variation-related fluctuations from neuronal-activity-related fluctuations in fMRI
Introduction
The central challenge in functional magnetic resonance imaging (fMRI) is the detection of relatively small neuronal-activation-induced blood oxygenation changes in the presence of various other signal fluctuations. In addition to thermal noise, scanner related variations, and bulk subject movement, signal fluctuations can be caused by several physiological processes. Pulsations of the blood induced by the heart beat, for example, result in signal changes mostly in voxels containing a high proportion of blood and/or cerebral spinal fluid (CSF). In addition, movement of the chest during respiration causes magnetic field changes that can cause a shifting of the brain image. Respiration can affect the fMRI time series in another way—by changing the arterial level of CO2, a potent vasodilator. This is perhaps most clearly seen in response to a breath-holding challenge, where a breath-hold of 30 s duration results in an average signal increase of 3–5% (Kastrup et al., 1999a, Kastrup et al., 1999b, Kastrup et al., 1999c, Li et al., 1999, Stillman et al., 1995). More recently, studies have shown that small fluctuations in end-tidal CO2 at a frequency of about 0.03 Hz occur naturally during normal breathing at rest and are significantly correlated with the blood-oxygenation-level-dependent (BOLD) fMRI signal fluctuations (Wise et al., 2004). These CO2 fluctuations are hypothesized to be due to subtle changes in the depth and rate of breathing during the scan. Existing physiologic correction techniques reduce MR signal changes occurring in synchrony with the respiratory cycle, corresponding to the movement associated with respiration, but they do not correct for variations in the inspired volume. This, of course, raises the concern that changes in breathing patterns during an fMRI experiment, which may be task or stimulus correlated, may lead to artifactual signal changes or mask existing function. Determining the impact of these fluctuations on fMRI studies and developing methods to correct for these variations are therefore of critical importance.
Low-frequency fluctuations induced by changes in breathing are particularly problematic for resting-state functional connectivity analysis, a technique that infers the connections of neuronal networks by measuring the correlation of low-frequency (<0.1 Hz) BOLD-fMRI signal fluctuations between and within brain regions (Biswal et al., 1995, Lowe et al., 1998). These low-frequency BOLD fluctuations are hypothesized to result from oscillations in neuronal activity synchronized within and across brain regions. There are, however, other factors that can cause signals in different parts of the brain to be correlated. Cardiac pulsations are often present and synchronized in regions of high blood volume or CSF. Respiration fluctuations often appear near edges in the image. Techniques have therefore been implemented to remove fluctuations at these cardiac and respiration frequencies (and their harmonics) prior to the functional connectivity analysis (Biswal et al., 1996, Chuang and Chen, 2001, Glover et al., 2000, Hu et al., 1995). The variations in respiration depth from breath to breath, however, occur at much lower frequencies which are not filtered out by typical physiological correction routines. Furthermore, the frequency of these respiratory changes (∼0.03 Hz) overlaps with the frequencies of fluctuations believed to result from varying brain activity at rest (<0.1 Hz). Therefore, there is still the danger that the functional network identified by this correlation method in fact represents areas where the fMRI signal has similar respiratory-change-related fluctuations, such as large vessels.
The goals of this study were: (1) to characterize the temporal and spatial patterns of respiration-variation-induced fMRI signal changes; (2) to investigate the impact of these respiration fluctuations on fMRI time series, both for task-related activation and for resting-state functional connectivity analysis; and (3) to evaluate potential methods to reduce this artifact. A method is presented by which these variations in respiration can be estimated from a respiration belt placed around the subject's chest without the need for a separate monitor for end-tidal CO2. The paradigm used in this study was a lexical decision task. In addition to activations in motor and language areas, this task commonly results in deactivations in the anterior and posterior cingulate. Such deactivations are observed across a wide range of tasks and are believed to reflect brain regions that are more active during rest, therefore referred to as the “default mode network” (Raichle et al., 2001). Recently, Greicius et al. showed that brain regions within this default mode network are also correlated at rest, supporting the notion that the default mode network reflects the activity of the brain when not performing a specific cognitive task (Greicius et al., 2003). We chose to use this system to investigate the impact of respiration changes on resting-state functional connectivity analysis for two primary reasons. First, the deactivations can define a starting point for determining which brain regions are correlated. Secondly, the regions within the default mode network overlap with many of the regions that are strongly affected by respiration changes (i.e. where the fMRI signal is significantly correlated with fluctuations in end-tidal CO2). While there may indeed be true differences between task-related deactivations and resting-state connectivity in the default mode network, our hypothesis is that the additional respiration fluctuations lead to errors in defining the network of neurons correlated during rest and that reducing the variability of breathing or regressing out these fluctuations leads to an improved ability to detect function and a better agreement between deactivations and regions correlated with these areas at rest.
Section snippets
Subjects and imaging parameters
Ten normal, healthy, right-handed volunteers were scanned under an Institutional Review Board (IRB) approved protocol after obtaining informed consent. Time series of T2*-weighted echo-planar MR images were acquired on a 3 T General Electric (GE) Signa MRI scanner (Waukesha, WI, USA) using an 8-channel GE receive coil with whole body RF excitation. Whole brain coverage was achieved using 27–28 sagittal 5 mm thick slices (TR: 2000 ms, TE: 30 ms, FOV: 24 cm, slice thickness: 5 mm, matrix:
Results
Cardiac fluctuations were most prominent in regions with CSF and with large vessels, such as the sagittal sinus and the Circle of Willis. Signal changes correlated with the respiration-related chest movement at 0.3 Hz were minimal and primarily located at the edges of the image in the phase encoding direction, consistent with a magnetic field shift in synchrony with the chest movement. The depth of breathing (divided by the breath-to-breath period) varied on average by about 28.1% ± 11.8%
Discussion
There is a growing interest in low-frequency fluctuations of the fMRI signal at rest, driven by the theory that these fluctuations reflect variations in brain activity during rest. Biswal and colleagues, for example, first noticed correlations in low-frequency fMRI signal fluctuations between the left and right motor cortices even when subjects were not explicitly performing a motor task (Biswal et al., 1995). Since then, studies have shown correlated fMRI signals in a number of other systems
Conclusion
Detection of task-related BOLD signal changes is improved when changes in respiration volume per unit time are regressed out of the signal. This improvement was largest in regions with a high density of blood vessels. Including these respiration fluctuations as regressors in the analysis, however, should be done with caution in cases where the breathing pattern is likely to vary in synchrony with the task.
The natural variation in breathing depth and rate during rest has a particularly
Acknowledgments
Support for this research was provided by the National Institute of Mental Health Intramural Research Program.
References (35)
AFNI: software for analysis and visualization of functional magnetic resonance neuroimages
Comput. Biomed. Res.
(1996)- et al.
Regional variability of cerebral blood oxygenation response to hypercapnia
NeuroImage
(1999) - et al.
Independent component analysis of nondeterministic fMRI signal sources
NeuroImage
(2003) - et al.
Comparison of methods for detecting nondeterministic BOLD fluctuation in fMRI
Magn. Reson. Imaging
(2004) - et al.
Functional MRI of human brain during breath holding by BOLD and FAIR techniques
NeuroImage
(1999) - et al.
Functional connectivity in single and multislice echoplanar imaging using resting-state fluctuations
NeuroImage
(1998) - et al.
Non-white noise in fMRI: does modelling have an impact?
NeuroImage
(2006) - et al.
Regional differences in the CBF and BOLD responses to hypercapnia: a combined PET and fMRI study
NeuroImage
(2000) - et al.
Functional MRI of brain during breath holding at 4 T
Magn. Reson. Imaging
(1995) - et al.
Breath holding reveals differences in fMRI BOLD signal in children and adults
NeuroImage
(2005)
Resting fluctuations in arterial carbon dioxide induce significant low frequency variations in BOLD signal
NeuroImage
A hypercapnia-based normalization method for improved spatial localization of human brain activation with fMRI
NMR Biomed.
Functional connectivity in the motor cortex of resting human brain using echo-planar MRI
Magn. Reson. Med.
Reduction of physiological fluctuations in fMRI using digital filters
Magn. Reson. Med.
IMPACT: image-based physiological artifacts estimation and correction technique for functional MRI
Magn. Reson. Med.
Mapping functionally related regions of brain with functional connectivity MR imaging
AJNR Am. J. Neuroradiol.
Calibrated functional MRI: mapping the dynamics of oxidative metabolism
Proc. Natl. Acad. Sci. U. S. A.
Cited by (1121)
Meditation attenuates default-mode activity: A pilot study using ultra-high field 7 Tesla MRI
2023, Brain Research BulletinLongitudinal multiparametric MRI of traumatic spinal cord injury in animal models
2023, Magnetic Resonance Imaging