Elsevier

NeuroImage

Volume 49, Issue 2, 15 January 2010, Pages 1479-1489
NeuroImage

Relating BOLD fMRI and neural oscillations through convolution and optimal linear weighting

https://doi.org/10.1016/j.neuroimage.2009.09.020Get rights and content

Abstract

The exact relationship between neural activity and BOLD fMRI is unknown. However, several recent findings, recorded invasively in both humans and monkeys, show a positive correlation of BOLD to high-frequency (30–150 Hz) oscillatory power changes and a negative correlation to low-frequency (8–30 Hz) power changes arising from cortical areas. In this study, we computed the time series correlation between BOLD GE-EPI fMRI at 7 T and neural activity measures from noninvasive MEG, using a time–frequency beam former for source localisation. A sinusoidal drifting grating was presented visually for 4 s followed by a 20 s rest period in both recording modalities. The MEG time series were convolved with either a measured or canonical haemodynamic response function (HRF) for comparison with the measured BOLD data, and the BOLD data were deconvolved with either a measured or a canonical HRF for comparison with the measured MEG. In the visual cortex, the higher frequencies (mid-gamma = 52–75 Hz and high-gamma = 75–98 Hz) were positively correlated with BOLD whilst the lower frequencies (alpha = 8–12 Hz and beta = 12–25 Hz) were negatively correlated with BOLD. Furthermore, regression including all frequency bands predicted BOLD better than stimulus timing alone, although no individual frequency band predicted BOLD as well as stimulus timing. For this paradigm, there was, in general, no difference between using the SPM canonical HRF compared to the subject-specific measured HRF. In conclusion, MEG replicates findings from invasive recordings with regard to time series correlations with BOLD data. Conversely, deconvolution of BOLD data provides a neural estimate which correlates well with measured neural effects as a function of neural oscillation frequency.

Introduction

The exact relationship between neuronal events and haemodynamic changes measured with blood-oxygenation-level-dependent (BOLD) functional magnetic resonance imaging (fMRI) is still unknown, although many recent studies, including those described below, have provided a qualitative correspondence. BOLD fMRI lacks a direct relationship to neural properties, although a correspondence to neural amplitude or duration has been found in specific paradigms (Bellgowan et al., 2003).

The local neuromagnetic fields generated by current flow in the dendrites of cortical pyramidal cells can be measured noninvasively above the scalp with magnetoencephalography (MEG). The neural current flow generating these fields is related to the local field potentials (LFPs) measured invasively (Hamalainen et al., 1993) and reflects input and intracortical processing. Local neural potentials/fields have also been shown to overlap spatially with gradient-echo (GE) BOLD data (Brookes et al., 2005, Martuzzi et al., 2009, Singh et al., 2002, Winterer et al., 2007) and to be the cause of the majority of metabolic demand which drives the BOLD response (Attwell and Laughlin, 2001). Logothetis et al. (2001) convolved LFPs and single- and multiunit spikes recorded in monkeys with a haemodynamic response function (HRF) and found LFPs to correlate with BOLD contrast better than spike output. Using invasive neural recordings in humans and comparing to fMRI obtained in different subjects, Mukamel et al. (2005) band pass-filtered the LFPs at multiple frequency bands and showed a negative correlation of BOLD with low-frequency LFPs and a positive correlation with higher frequencies.

BOLD responses are often modelled as a convolution of neural events with an HRF; however, the neural events are usually approximated by the stimulus timing rather than being directly measured. The shape of the HRF is often approximated by a single or sum of two Gamma functions. Deconvolving the fMRI response with the HRF is important in understanding the dynamics of the underlying neural activity (Glover, 1999, Gitelman et al., 2003). However, large variability has been shown across individual subjects' measured HRFs and this variability can affect parameter estimation of significant changes and detection of significant voxels (Handwerker et al., 2004, Lu et al., 2006).

Previous studies have compared MEG or electroencephalography (EEG) data to fMRI in a variety of ways. Studies have used EEG sensor-level data (Mart et al., 2004, Parkes et al., 2006, Tyvaert et al., 2008) or MEG broadband power from dipole fits (Nangini et al., 2008) to convolve with the HRF to improve detection of BOLD changes. In experiments to study the resting state in which no stimulus was presented, EEG data recorded concurrently with fMRI were the only available information with which to investigate BOLD correlates (Gonçalves et al., 2006). De Munck et al. (2007) recorded simultaneous EEG and fMRI and deconvolved the BOLD with the EEG alpha power to obtain an ‘alpha response function’ rather than assuming the traditional HRF. They found a negative correlation in visual cortex and positive correlation in thalamus, as reported previously (Laufs et al., 2003, Gonçalves et al., 2006).

Here, we extend previous work by applying a time–frequency beam former (Dalal et al., 2008) on MEG data to extract a time–frequency spectrogram at every voxel location in the brain and compare with BOLD fMRI data. Analysis was focussed on voxels within a visual cortex region of interest. Use of the subject-specific HRF was compared to the double-Gamma HRF (provided in SPM, www.fil.ion.ucl.ac.uk/spm). Two spatial–temporal–spectral comparisons were made: (1) MEG convolved with the HRF to predict BOLD and (2) fMRI deconvolved with the HRF to predict MEG. We show excellent agreement across the frequency spectrum between BOLD correlated with MEG, and BOLD correlated with invasive neuroelectrical recordings (Mukamel et al., 2005).

Section snippets

Acquisition

Six healthy subjects gave informed consent and participated in the study. Ethical approval was obtained prior to data collection from the University of Nottingham Medical School Ethics Committee. A sinusoidal drifting grating (drifting at 8 Hz) was presented in a circular window with a radius of 2.5° of visual angle, and with centre of the circular window at 4° from fixation, in the lower left quadrant of the visual field. Three Michelson contrasts (0.25, 0.5 and 1) were presented

Results

Both MEG and fMRI results showed responses in visual cortex consistent with previous findings for this stimulus (Stevenson et al., 2008). Changes in neural oscillatory power in motor cortex were observed due to the subjects' button press, but as the fMRI slice coverage did not include motor cortex, these results were not further analysed.

Discussion

Both MEG and fMRI activity is observed in primary visual cortex as well as surrounding areas, indicating a general correspondence of underlying activity between the methods. Lower frequencies in MEG source estimates are negatively temporally correlated with fMRI whilst higher frequencies are positively correlated. These results are consistent whether using forward convolution or deconvolution and whether using a subject-specific HRF or standard SPM HRF. Furthermore, MEG time–frequency beam

Conclusion

In conclusion, the results here support the trend that lower neural oscillation frequencies negatively correlate with BOLD fMRI and higher frequencies positively correlate. Both convolution and deconvolution can be used to assess this correlation. Using a time–frequency beam former on MEG data helped elucidate these findings. However, the robustness of these findings remains to be tested on different paradigms.

Acknowledgments

The authors would like to thank Dr. Sarang S. Dalal for help with NUTMEG and the time–frequency beam former. Funding was provided by the Medical Research Council, the Engineering and Physical Sciences Research Council, the Wellcome Trust and the University of Nottingham. JMZ is funded by the Whitaker International Scholar program.

References (79)

  • FrimanO. et al.

    Adaptive analysis of fMRI data

    NeuroImage

    (2003)
  • FristonK.J. et al.

    Nonlinear responses in fMRI: the Balloon model, Volterra kernels, and other hemodynamics

    NeuroImage

    (2000)
  • FujimakiN. et al.

    An fMRI-constrained MEG source analysis with procedures for dividing and grouping activation

    NeuroImage

    (2002)
  • GitelmanD.R. et al.

    Modeling regional and psychophysiologic interactions in fMRI: the importance of hemodynamic deconvolution

    NeuroImage

    (2003)
  • GloverG.H.

    Deconvolution of impulse response in event-related BOLD fMRI

    NeuroImage

    (1999)
  • GonçalvesS.I. et al.

    Correlating the alpha rhythm to BOLD using simultaneous EEG/fMRI: inter-subject variability

    NeuroImage

    (2006)
  • HallS.D. et al.

    The missing link: analogous human and primate cortical gamma oscillations

    NeuroImage

    (2005)
  • HandwerkerD.A. et al.

    Variation of BOLD hemodynamic responses across subjects and brain regions and their effects on statistical analyses

    NeuroImage

    (2004)
  • HulvershornJ. et al.

    Spatial sensitivity and temporal response of spin echo and gradient echo bold contrast at 3 T using peak hemodynamic activation time

    NeuroImage

    (2005)
  • KilnerJ.M. et al.

    Hemodynamic correlates of EEG: a heuristic

    NeuroImage

    (2005)
  • LaufsH. et al.

    EEG-correlated fMRI of human alpha activity

    NeuroImage

    (2003)
  • LuY. et al.

    Using voxel-specific hemodynamic response function in EEG–fMRI data analysis

    NeuroImage

    (2006)
  • MakniS. et al.

    Bayesian deconvolution fMRI data using bilinear dynamical systems

    NeuroImage

    (2008)
  • Martinez-MontesE. et al.

    Concurrent EEG/fMRI analysis by multiway partial least squares

    NeuroImage

    (2004)
  • MoradiF. et al.

    Consistent and precise localization of brain activity in human primary visual cortex by MEG and fMRI

    NeuroImage

    (2003)
  • MuthukumaraswamyS.D. et al.

    Spatiotemporal frequency tuning of BOLD and gamma band MEG responses compared in primary visual cortex

    NeuroImage

    (2008)
  • NirY. et al.

    Coupling between neuronal firing rate, gamma LFP, and BOLD fMRI is related to interneuronal correlations

    Curr. Biol.

    (2007)
  • ParkesL.M. et al.

    Combining EEG and fMRI to investigate the post-movement beta rebound

    NeuroImage

    (2006)
  • PrammerM.G. et al.

    A new approach to automatic shimming

    J. Magn. Reson.

    (1988)
  • RobsonM.D. et al.

    Measurements of the temporal fMRI response of the human auditory cortex to trains of tones

    NeuroImage

    (1998)
  • SalmelinR. et al.

    Spatiotemporal characteristics of sensorimotor neuromagnetic rhythms related to thumb movement

    Neuroscience

    (1994)
  • SatoM.-a. et al.

    Hierarchical Bayesian estimation for MEG inverse problem

    NeuroImage

    (2004)
  • SinghK.D. et al.

    Task-related changes in cortical synchronization are spatially coincident with the hemodynamic response

    NeuroImage

    (2002)
  • SoltysikD.A. et al.

    Comparison of hemodynamic response nonlinearity across primary cortical areas

    NeuroImage

    (2004)
  • StancákA. et al.

    Desynchronization and recovery of beta rhythms during brisk and slow self-paced finger movements in man

    Neurosci. Lett.

    (1995)
  • StevensonC.M. et al.

    Neuromagnetic correlates of the fMRI BOLD response

    Int. Congr. Ser.

    (2007)
  • TuunanenP.I. et al.

    Comparison of BOLD fMRI and MEG characteristics to vibrotactile stimulation

    NeuroImage

    (2003)
  • TyvaertL. et al.

    Effects of fluctuating physiological rhythms during prolonged EEG-fMRI studies

    Clin. Neurophysiol.

    (2008)
  • Van der ZwaagW. et al.

    fMRI at 1.5, 3, and 7 T: characterising BOLD signal changes

    NeuroImage

    (2009)
  • Cited by (64)

    • Late dominance of the right hemisphere during narrative comprehension

      2022, NeuroImage
      Citation Excerpt :

      Several lines of evidence show that high gamma activity is strongly correlated to the firing rates of local neuronal populations (Manning et al., 2009; Nir et al., 2007; Ray and Maunsell, 2011). Both stimulus-related gamma-power increments and beta-power decrements are also strongly correlated to local blood flow and the blood oxygen level-dependent (BOLD) signal (Hall et al., 2014; Muthukumaraswamy and Singh, 2008; Zumer et al., 2010), whereas a similar correspondence between blood flow and ERPs is not always found (Brovelli et al., 2005; Foucher et al., 2003; Logothetis et al., 2001). High gamma-band activity arising outside primary sensory or motor cortices is less readily detectable in M/EEG, very likely due to signal loss caused by the mixing of out-of-phase rhythms at the scalp electrode/MEG sensor (Pfurtscheller and Cooper, 1975) and the presence of a substantial myogenic noise floor at these frequencies (Jerbi et al., 2009; Whitham et al., 2007).

    View all citing articles on Scopus
    View full text