Correlated slow fluctuations in respiration, EEG, and BOLD fMRI
Introduction
Functional magnetic resonance imaging (fMRI) utilizes blood-oxygen-level-dependent (BOLD) contrast to study brain regions that respond to task-related activations (Bandettini et al., 1992, Kwong et al., 1992, Ogawa et al., 1992) or regions of spontaneous activity that are functionally connected to other regions (Biswal et al., 1995). The BOLD contrast, however, does not directly measure the neural activity but depends on the cerebral blood flow, blood volume, and oxygenation that are functionally coupled to the ongoing neural processes (Logothetis, 2008). Unfortunately, other more autonomic physiological processes such as cardiac pulsation and respiration can also affect the brain's BOLD response and thus cause confounding signal fluctuation (Biswal et al., 1996, Kruger and Glover, 2001, Weisskoff et al., 1993). These non-task-evoked BOLD signal fluctuations, of physiological origin, are generally considered as physiological noise (Kruger and Glover, 2001).
It is thus important to identify and characterize the sources of physiological noise in the BOLD fMRI signal so that noise reduction techniques can be developed and employed. So far several physiological noise sources have been identified and can be mainly characterized in two categories. The first type of physiological noises is quasi-periodic, time-locked to the cardiac or respiratory cycles. Cardiac-related artifacts in BOLD fMRI data without contribution of unspoiled spin-echoes (Zhao et al., 2000) are caused by pulsatility of blood and cerebrospinal fluid due to cardiovascular processes and tend to be localized near ventricles, sulci, and large vessels (Bhattacharyya and Lowe, 2004, Dagli et al., 1999). The respiration-related noises due to the bulk motion of the head as well as perturbation of the magnetic field by thoracic and abdominal movement are more spatially global but can also be localized similarly to cardiac fluctuations (Glover et al., 2000, Raj et al., 2001). Both cardiac and respiratory-related artifacts are composed of frequency components higher than those of BOLD signal (< 0.1 Hz) (Lowe et al., 1998), yet are usually temporally aliased due to undersampling. Accordingly, a number of methods have been developed to remove such undesired signal fluctuations, which can be operated as band-reject filtering (Biswal et al., 1996), subtraction in k-space (Hu et al., 1995, Wowk et al., 1997) or in image space (Chuang and Chen, 2001, Deckers et al., 2006, Glover et al., 2000), or more recently by utilizing external near-infrared spectroscopy (NIRS) measures of changes in the concentration of the oxyhemoglobin and deoxyhemoglobin (Strangman et al., 2002, Toronov et al., 2003). The second type of physiological noise features non-periodic low-frequency fluctuation that overlaps with the major components of BOLD signal. They include end-tidal CO2 fluctuation (Wise et al., 2004), residual movement artifacts not corrected by rigid body registration (Lund et al., 2005), and the slow changes of respiration depth/rate (Birn et al., 2006) and cardiac rate (Chang et al., 2009, de Munck et al., 2008, Shmueli et al., 2007). These fluctuations can account for a significant portion of variance in the BOLD signal (Bianciardi et al., 2009), and, more importantly, they affect widespread regions of the gray matter (Birn et al., 2006, Falahpour et al., 2013, Wise et al., 2004). Since the slow changes in pulse rate, respiration depth, and end-tidal CO2 have frequency components overlapping with the frequency range of BOLD signal, they are usually removed by using the approach of nuisance variable regression (Lund et al., 2006), which includes the fluctuation of physiological noise as nuisance regressors in a general linear model regression analysis (Birn et al., 2006, Chang et al., 2009, Falahpour et al., 2013, Shmueli et al., 2007). Recently the utility of NIRS signal simultaneously measured with fMRI was shown to be useful for partitioning the contributing factors of the low-frequency physiological noises in fMRI (Tong et al., 2011) as well as regressing out these noises (Cooper et al., 2012, Frederick et al., 2012). These results suggest that aliasing signals from cardiac pulse and respiration likely are not affecting BOLD low-frequency fluctuation, and that there is a strong correlation in slow BOLD fluctuation with global cerebral blood flow fluctuation.
The presence of physiological noise in the BOLD signal negatively affects the detection of neural activation in traditional task-related designs (Birn et al., 2009) and, more recently, has been found be to particularly problematic for studying resting-state functional connectivity at task-free conditions. In the latter type of analysis, specific functional networks, i.e. the resting-state networks (Biswal et al., 1995, Fox and Raichle, 2007), are inferred by measuring and mapping the temporal correlations of low-frequency BOLD signal fluctuation (< 0.1 Hz) between brain regions. Unfortunately, the slow BOLD fluctuation originating from physiological noise shares a common frequency range with neural-activity-related BOLD signal and may lead to spurious correlation in resting state networks (Cordes et al., 2001). Thus, the minimization of low-frequency physiological fluctuation is widely considered as a necessary procedure in analyzing resting state signals (Birn et al., 2006, Fox and Raichle, 2007, Shmueli et al., 2007, van Buuren et al., 2009, Wise et al., 2004).
However, although the low-frequency physiological fluctuation is considered a confounding source of noise in the BOLD signal the possibility of neural correlates for such physiological fluctuation has not been investigated in detail (Iacovellaa and Hasson, 2011). As mentioned above, there are two observations suggesting that there might be common neural activity underlying the fluctuation in physiological process and BOLD signals. First, their frequency distributions overlap in the low frequency range. In addition, for both cardiac and respiratory fluctuations correlates with BOLD signals were found to be largely present in the gray matter. Thus it is possible that there exist common neural activities related to both the low-frequency physiological fluctuation and BOLD signals which result in correlations between the two. However, to our knowledge the possible neural origin for the low frequency physiological fluctuations present in BOLD fMRI data has not yet been examined in detail.
Therefore, the goal of this study was to investigate whether the low-frequency fluctuation in physiological “noise” reflects neuronal activities. For this purpose we simultaneously acquired high-density macroscopic extracranial brain electroencephalography (EEG), BOLD fMRI data and cardiac and respiration waveforms in normal resting human subjects. Specifically, we quantified the respiratory fluctuation as the respiration volume per unit time as proposed by Birn et al. (2006) and the cardiac fluctuation as the across-beats variation in pulse rate. In addition, we characterized the neural signal temporal dynamics by quantifying the changes of EEG power in the alpha frequency band (8–13 Hz), as the alpha-band activity dominates the whole brain activity at rest and demonstrates fluctuation over the frequency range of < 0.1 Hz. Furthermore, a number of studies have revealed that fluctuation in the alpha band negatively correlates with resting-state BOLD signals at several brain regions that are recognized as belonging to some of the known resting state networks (Goldman et al., 2002, Laufs et al., 2003, Moosmann et al., 2003, Sadaghiani et al., 2010). Additionally, positive correlation between EEG alpha activity and endogenous fluctuation of occipital deoxyhemoglobin concentration (as measured by NIRS) has been found (Moosmann et al., 2003). These findings suggest an association between the neuronal synchronization (as reflected by alpha EEG) and hemodynamic activity (as reflected by BOLD-fMRI/NIRS). Therefore, a relationship between alpha-band EEG and physiological temporal fluctuation as measured in simultaneous fMRI and EEG experiments, if any, may suggest a mutual link between neuronal activity, physiological fluctuation, and BOLD signal. In this regard, we assessed the temporal cross-correlation between EEG power changes in the alpha band and respiratory/cardiac low frequency fluctuation concurrently measured during simultaneous fMRI and EEG experiments for eyes-closed and eyes-open resting conditions. Subsequent to and informed by the cross-correlation analysis, the temporal and spatial patterns of fMRI signal changes correlated to variations in the EEG and physiological signals were characterized using regression analysis.
Section snippets
Subjects and data acquisition
The study was conducted at the Laureate Institute for Brain Research and was approved by the University of Oklahoma Institutional Review Board (IRB). Nine healthy, right-handed subjects (mean age = 33 ± 10 years; one female) participated in the study. All participants provided written informed consent, and received financial compensation for their participation.
Simultaneous EEG, fMRI, cardiac, and respiratory data were collected in all subjects. High-density EEG signals from 126 channels were
Results
Examples of a single-session EEG alpha power fluctuation at the O1 electrode, respiratory fluctuation, and pulse rate fluctuation from a representative subject (subject #7) are illustrated in Figs. 1B, D and F, respectively. The power spectrum of the alpha GFP, RVT, and pulse rate fluctuations (without smoothing) averaged over all sessions and subjects is shown in Figs. 1G, H, and I, respectively. The power spectra reveal that the major energy of all the fluctuations is in the low-frequency
Discussion
In this work we investigated the hypothesis that the low-frequency fluctuation of the respiratory and cardiac physiological signals observed during fMRI at resting state may have contribution of neuronal origin. We have found strong correlation between the EEG power changes in the alpha frequency band as described by alpha global field power (GFP) and the low-frequency fluctuation of the respiration volume per time (RVT). Both alpha GFP and respiration fluctuations were found to be correlated
Conclusions
We have observed significant correlation between the low-frequency temporal fluctuation of respiration and changes of global alpha EEG power in human subjects at resting state, whereas much less correlation was found between alpha global power and cardiac fluctuation. In particular, this correlation between alpha EEG and respiration is much stronger in eyes-closed resting than in eyes-open resting, and is consistently observed from recordings inside and outside scanner. In addition, similar
Acknowledgments
This research was supported by the Laureate Institute for Brain Research and the William K. Warren Foundation. The authors are very grateful to anonymous reviewers for their constructive and helpful comments.
Conflicts of interest
All authors have no conflicts of interest or financial disclosures to declare.
References (68)
- et al.
Identification of EEG events in the MR scanner: the problem of pulse artifact and a method for its subtraction
Neuroimage
(1998) - et al.
A method for removing imaging artifact from continuous EEG recorded during functional MRI
Neuroimage
(2000) - et al.
Cardiac-induced physiologic noise in tissue is a direct observation of cardiac-induced fluctuations
Magn. Reson. Imaging
(2004) - et al.
Sources of functional magnetic resonance imaging signal fluctuations in the human brain at rest: a 7 T study
Magn. Reson. Imaging
(2009) - et al.
Separating respiratory-variation-related fluctuations from neuronal-activity-related fluctuations in fMRI
Neuroimage
(2006) - et al.
The respiration response function: the temporal dynamics of fMRI signal fluctuations related to changes in respiration
Neuroimage
(2008) - et al.
fMRI in the presence of task-correlated breathing variations
Neuroimage
(2009) - et al.
Mapping the MRI voxel volume in which thermal noise matches physiological noise—implications for fMRI
Neuroimage
(2007) - et al.
Relationship between respiration, end-tidal CO2, and BOLD signals in resting-state fMRI
Neuroimage
(2009) - et al.
Effects of model-based physiological noise correction on default mode network anti-correlations and correlations
Neuroimage
(2009)
Influence of heart rate on the BOLD signal: the cardiac response function
Neuroimage
The utility of near-infrared spectroscopy in the regression of low-frequency physiological noise from functional magnetic resonance imaging data
Neuroimage
AFNI: software for analysis and visualization of functional magnetic resonance neuroimages
Comput. Biomed. Res.
Localization of cardiac-induced signal change in fMRI
Neuroimage
The hemodynamic response of the alpha rhythm: an EEG/fMRI study
Neuroimage
A study of the brain's resting state based on alpha band power, heart rate and fMRI
Neuroimage
An adaptive filter for suppression of cardiac and respiratory noise in MRI time series data
Neuroimage
Subject specific BOLD fMRI respiratory and cardiac response functions obtained from global signal
Neuroimage
Physiological denoising of BOLD fMRI data using regressor interpolation at progressive time delays (RIPTiDe) processing of concurrent fMRI and near-infrared spectroscopy (NIRS)
Neuroimage
Physiological noise effects on the flip angle selection in BOLD fMRI
Neuroimage
The effect of hypercapnia on resting and stimulus induced MEG signals
Neuroimage
Functional connectivity in single and multislice echoplanar imaging using resting-state fluctuations
Neuroimage
Motion or activity: their role in intra- and inter-subject variation in fMRI
Neuroimage
Non-white noise in fMRI: does modelling have an impact?
Neuroimage
Correlates of alpha rhythm in functional magnetic resonance imaging and near infrared spectroscopy
Neuroimage
Brain correlates of autonomic modulation: combining heart rate variability with fMRI
Neuroimage
Event-related EEG/MEG synchronization and desynchronization: basic principles
Clin. Neurophysiol.
Low-frequency fluctuations in the cardiac rate as a source of variance in the resting-state fMRI BOLD signal
Neuroimage
A quantitative comparison of simultaneous BOLD fMRI and NIRS recordings during functional brain activation
Neuroimage
Testing for nonlinearity in time series: the method of surrogate data
Physica D
The roles of changes in deoxyhemoglobin concentration and regional cerebral blood volume in the fMRI BOLD signal
Neuroimage
Resting fluctuations in arterial carbon dioxide induce significant low frequency variations in BOLD signal
Neuroimage
The Berger rhythm: potential changes from the occipital lobes in man
Brain
Time course EPI of human brain function during task activation
Magn. Reson. Med.
Cited by (82)
A model of neurovascular coupling and its application to cortical spreading depolarization
2023, Journal of Theoretical BiologyVigilance associates with the low-dimensional structure of fMRI data
2023, NeuroImageCitation Excerpt :Changes in vigilance are known to be accompanied by changes in autonomic activity, reflected in peripheral measures such as breathing and heart rate. The latter can also influence BOLD fMRI signals, and notably, low-frequency (<0.15 Hz) changes in respiration volume and heart rate are found to covary with changes in arousal (Soon et al., 2021; Yuan et al., 2013). While the covariation between vigilance and low-frequency physiology renders it challenging to disentangle their influences on the BOLD signal, we nonetheless sought to estimate the contribution of vigilance-linked systemic hemodynamics on our low-dimensional fMRI state detection.
Interoceptive influences on resting-state fMRI
2023, Advances in Resting-State Functional MRI: Methods, Interpretation, and ApplicationsThe role of vigilance in resting-state functional MRI
2023, Advances in Resting-State Functional MRI: Methods, Interpretation, and ApplicationsPhysiological brain pulsations
2023, Advances in Resting-State Functional MRI: Methods, Interpretation, and ApplicationsAn orderly sequence of autonomic and neural events at transient arousal changes
2022, NeuroImageCitation Excerpt :With both HR and RV, significant positive correlations were observed in the SN and DMN at a time lag of 0 s, whereas negative correlations in the sensory/motor regions peaked at ∼7 s (Fig. 1, C-D and Fig. S3, C-D). These lag-specific rsfMRI correlations with the EEG alpha-power and the RV are somewhat consistent with what has been shown previously, i.e., more positive correlation at negative lags and the thalamus whereas more negative correlations at positive lags and the sensory/motor regions (Birn et al., 2006; Yuan et al., 2013). Overall, these correlation patterns were quite similar to those seen with the EEG alpha-power correlation (Fig. 1A), suggesting they were not independent but linked to the same underlying process, apparently related to arousal transitions indicated by EEG alpha-power modulations.