Elsevier

NeuroImage

Volume 65, 15 January 2013, Pages 499-510
NeuroImage

The impact of “physiological correction” on functional connectivity analysis of pharmacological resting state fMRI

https://doi.org/10.1016/j.neuroimage.2012.09.044Get rights and content

Abstract

Growing interest in pharmacological resting state fMRI (RSfMRI) necessitates developing standardized and robust analytical approaches that are insensitive to spurious correlated physiological signals. However, in pharmacological experiments physiological variations constitute an important aspect of the pharmacodynamic/pharmacokinetic profile of drug action; therefore retrospective corrective methods that discard physiological signals as noise may not be suitable. Previously, we have shown that template-based dual regression analysis is a sensitive method for model-free and objective detection of drug-specific effects on functional brain connectivity. In the current study, the robustness of this standard approach to physiological variations in a placebo controlled, repeated measures pharmacological RSfMRI study of morphine and alcohol in 12 healthy young men is tested. The impact of physiology-related variations on statistical inferences has been studied by: 1) modeling average physiological rates in higher level group analysis; 2) Regressing out the instantaneous respiration variation (RV); 3) applying retrospective image correction (RETROICOR) in the preprocessing stage; and 4) performing combined RV and heart rate correction (RVHRCOR) by regressing out physiological pulses convolved with canonical respiratory and cardiac hemodynamic response functions. Results indicate regional sensitivity of the BOLD signal to physiological variations, especially in the vicinity of large vessels, plus certain brain structures that are reported to be involved in physiological regulation, such as posterior cingulate, precuneus, medial prefrontal and insular cortices, as well as the thalamus, cerebellum and the brainstem. The largest impact of “correction” on final statistical test outcomes resulted from including the average respiration frequency and heart rate in the higher-level group analysis. Overall, the template-based dual regression method seems robust against physical noise that is corrected by RV regression or RETROICOR. However, convolving the RV and HR with canonical hemodynamic response functions caused a notable change in the BOLD signal variance, and in resting state connectivity estimates. The impact of RVHRCOR on statistical tests was limited to elimination of both morphine and alcohol effects related to the somatosensory network that consists of insula and cingulate cortex—important structures for autonomic regulation. Although our data do not warrant speculations about neuronal or vascular origins of these effects, these observations raise caution about the implications of physiological ‘noise’ and the risks of introducing false positives (e.g. increased white matter connectivity) by using generalized physiological correction methods in pharmacological studies. The obvious sensitivity of the posterior part of the default mode network to different correction schemes, underlines the importance of controlling for physiological fluctuations in seed-based functional connectivity analyses.

Highlights

► The BOLD signal is sensitive to physiological noise near large vessels. ► Higher sensitivity to physiological signal in structures linked to autonomic control. ► Template-based dual-regression analysis, robust to effects of physiological noise. ► Group-level modeling of physiological rates changed the extent of morphine effects. ► Canonical cardiac and respiratory HRF filters not suitable in pharmacological tests.

Introduction

In most pharmacological resting state fMRI studies, changes in heart rate or respiration following drug administration are conspicuous enough to raise concern about the interpretation of statistical test outcomes. Some drugs (like opiodergic substances) change respiration patterns that lead to hypercapnic conditions that affect vascular reactivity. Many drugs increase heart rate or blood pressure and thus increase pressure wave pulsatility in cerebral arteries and result in motion-related spurious effects in brain regions that are in the proximity of major arteries (Dagli et al., 1999). The main concern is that global physiological responses reflected in circulatory or respiratory alterations would be associated with nonspecific changes in vascular response and cerebral perfusion that would either obscure important neuronal central nervous system (CNS) drug effects, or mislead the interpretation of vascular artifacts as true neuronal responses.

Generally, respiration- or heart-related variations are considered as nuisance factors confounding the blood oxygen level dependent (BOLD) response. They might reduce the sensitivity to detecting smaller-scale neuronal effects, or might exacerbate vascular-modulated effects (Beckmann et al., 2005, Birn et al., 2006, Birn et al., 2008, Chang and Glover, 2009a, Chang and Glover, 2009b, Frank et al., 2001, Harvey et al., 2008, Shmueli et al., 2007, Teichert et al., 2010, van Buuren et al., 2009). For instance, some effects of respiration, like correlated bulk motion or change in B0 field homogeneity due to abdominal movement (Brosch et al., 2002, Glover and Pauly, 1992, Van de Moortele et al., 2002), are purely artifactual. Motion- or B0-related variations due to respiration lead to a reduction of signal to noise ratio (SNR). Various corrective actions during acquisition (Glover and Pauly, 1992, Lee et al., 2006, Noll et al., 1998, Pfeuffer et al., 2002) and in post-processing have been proposed (Brooks et al., 2008, Chang and Glover, 2009a, Deckers et al., 2006, Glover et al., 2000, Harvey et al., 2008, Jo et al., 2010, Lund et al., 2006, Shmueli et al., 2007, Starck et al., 2010). In principle, these methods apply either a band pass filter, isolate and remove an independent noise component, or examine functional connectivity with regression models that include either heart and respiration pulses (sometimes convolved with a presumed hemodynamic response function), or average CSF and deep white matter signal as a covariate of no interest.

Recently, it has been argued that physiological signals should not be necessarily treated as artifactual nuisance (Iacovella and Hasson, 2011). Respiration and heart rate are regulated on the one hand, by an interplay between brain areas that integrate autonomic, somatomotor, and proprioception; and on the other hand by cerebrovascular autoregulation that adaptively maintains oxygen levels via arterial inflow (Gray et al., 2009, Napadow et al., 2008, Pattinson et al., 2009b). In cases where an experimental condition causes a significant change in respiration patterns (e.g. in an opiodergic drug experiment), an adaptive response in different brain regions (Evans et al., 1999, McKay et al., 2008, McKay et al., 2010, Pattinson et al., 2009a) or a global physiological change in total brain perfusion might result (Kastrup et al., 1999, Noth et al., 2008). Particularly, a drug-induced change in physiology also interacts with the central nervous system through responses of autonomic brain centers (such as hypothalamus, amygdala, or the cingulate cortex) or alterations in perception of bodily state (e.g. insula or somatosensory cortex) (McKay et al., 2008, McKay et al., 2010, Pattinson et al., 2009a, Pattinson et al., 2009b). These effects constitute parts of the pharmacodynamics of drug actions, which (whether of neuronal or vascular origin) are indispensible and should not be obscured by noise-correction.

Previously, we have shown that template-based dual regression analysis is a sensitive method for model-free and objective detection of drug-specific effects on functional connectivity (Khalili-Mahani et al., 2012). The aim of the current analysis is to examine the robustness of our proposed method for studying drug-induced changes in resting state functional brain connectivity to various approaches for physiological noise correction. We will examine four aspects of physiological confounds: average change in physiological rates (which are modeled in higher-level group analysis); breathing-related chest movement (which we have used as a covariate in connectivity estimations); physical noise related to physiological motion (which was corrected using RETROICOR); and finally, hemodynamic response to systemic physiological variations (which was addressed using RVHRCOR). Furthermore, we will investigate the within- and between-subject variations in the extent and topography of physiology-related BOLD variations to investigate systemic effects on resting-state BOLD fluctuations.

Section snippets

Subjects

Twelve healthy male subjects (age range 18–40; BMI 18–26 kg/m2) volunteered for a randomized double dummy, double blind, placebo-controlled study involving three visits (each one week apart). The Medical Ethics Review Board of Leiden University Medical Center approved the study. Exclusion criteria included MRI-contraindications such as implants, pacemakers, or prostheses; any account of medical disorders that could pose a risk to subject's health (e.g. opioid allergy, positive hepatitis B, C or

Effects of pharmacological manipulations on physiology-related variations

Fig. 2 summarizes drug by time interactions with physiological variables and their impact on statistical maps, comparing effects of each drug to placebo.

Discussion

Growing interest in the pharmacological application of functional neuroimaging raises three concerns about the confounding effects of physiological signals: first, whether physiological variables should be treated as sources of noise or as signals of interest related to central adaptive responses of the brain to drug effects. Next, how to distinguish the neuronal from the cerebrovascular changes that are caused by variations in respiratory and heart rates. And importantly, how robust the

Conflicts of Interest

Authors have no conflict of interest to declare.

Acknowledgments

This study was supported by The Netherlands Organization for Scientific Research (NWO, VIDI grant 91786368 to SR). We thank Dr. Remco Zoethout for study design and Evelinda Baerends for her role in study execution; and Jordan Gross and Rene Post (CHDR) for assisting with the data acquisition.

References (55)

  • H.J. Jo et al.

    Mapping sources of correlation in resting state FMRI, with artifact detection and removal

    Neuroimage

    (2010)
  • Y. Kong et al.

    Assessment of physiological noise modelling methods for functional imaging of the spinal cord

    Neuroimage

    (2012)
  • T.E. Lund et al.

    Non-white noise in fMRI: does modelling have an impact?

    Neuroimage

    (2006)
  • F. Luo et al.

    Differential responses in CBF and CBV to cocaine as measured by fMRI: implications for pharmacological MRI signals derived oxygen metabolism assessment

    J. Psychiatr. Res.

    (2009)
  • C. Madjar et al.

    Task-related BOLD responses and resting-state functional connectivity during physiological clamping of end-tidal CO(2)

    Neuroimage

    (2012)
  • L.C. McKay et al.

    A bilateral cortico-bulbar network associated with breath holding in humans, determined by functional magnetic resonance imaging

    Neuroimage

    (2008)
  • L.C. McKay et al.

    Sub-cortical and brainstem sites associated with chemo-stimulated increases in ventilation in humans

    Neuroimage

    (2010)
  • K. Murphy et al.

    Robustly measuring vascular reactivity differences with breath-hold: normalising stimulus-evoked and resting state BOLD fMRI data

    Neuroimage

    (2011)
  • V. Napadow et al.

    Brain correlates of autonomic modulation: combining heart rate variability with fMRI

    Neuroimage

    (2008)
  • K.T. Pattinson et al.

    Determination of the human brainstem respiratory control network and its cortical connections in vivo using functional and structural imaging

    Neuroimage

    (2009)
  • J.E. Perthen et al.

    Caffeine-induced uncoupling of cerebral blood flow and oxygen metabolism: a calibrated BOLD fMRI study

    Neuroimage

    (2008)
  • K. Shmueli et al.

    Low-frequency fluctuations in the cardiac rate as a source of variance in the resting-state fMRI BOLD signal

    Neuroimage

    (2007)
  • T. Starck et al.

    Correction of low-frequency physiological noise from the resting state BOLD fMRI—effect on ICA default mode analysis at 1.5 T

    J. Neurosci. Methods

    (2010)
  • T. Teichert et al.

    Effects of heartbeat and respiration on macaque fMRI: implications for functional connectivity

    Neuropsychologia

    (2010)
  • R.G. Wise et al.

    Resting fluctuations in arterial carbon dioxide induce significant low frequency variations in BOLD signal

    Neuroimage

    (2004)
  • X.N. Zuo et al.

    Reliable intrinsic connectivity networks: test-retest evaluation using ICA and dual regression approach

    Neuroimage

    (2010)
  • C.F. Beckmann et al.

    Investigations into resting-state connectivity using independent component analysis

    Philos. Trans. R. Soc. Lond. B Biol. Sci.

    (2005)
  • Cited by (0)

    View full text