Functional connectivity of auditory and visual areas during rest and stimulus presentation

1
Linearity of the BOLD response A comparison
between fMRI and MEG C.M.Stevenson1,
M.J.Brookes1, G.R.Barnes2, A.Hillebrand2,
S.T.Francis1 and P.G.Morris1 1Sir Peter
Mansfield Magnetic Resonance Centre, School of
Physics and Astronomy, The University of
Nottingham, Nottinghamshire, United
Kingdom. 2WELLCOME Trust Laboratory for MEG
Studies, Aston University, Birmingham, West
Midlands, United Kingdom.
2
Linearity of the BOLD response fMRI/MEG a
comparison
INTRODUCTION For fMRI to reach its full
potential and provide a quantitative measure of
brain activity, it is necessary to understand the
neural basis of the BOLD response.1
CBV STIMULUS NEURAL RESPONSE CBF
BOLD CMRO2
Simultaneous microelectrode and fMRI BOLD
measurements have shown that the BOLD response is
best characterised by local field potentials
(LFPs), which are synaptic in origin. 2 Since
synaptic processes are largely thought to be the
basis of MEG signals, one would expect a good
correlation between BOLD and MEG responses.3
Implementation of a parallel fMRI/MEG experiment
allows investigation into whether non-linearities
in the BOLD response are a result of
non-linearities in the underlying neural activity
or are a consequence of haemodynamic coupling.
References 1 Ogawa, S. et al., Magn. Reson.
Med., 14, 68-78 (1990). 2 Logothetis, N.K. et
al. Nature,12, 150-157 (2001). 3
Kaufman, L. and Lu, Z.L, Magnetic Source Imaging
of the Human Brain, Lawrence Erlbaum
Associates Inc., 2003.
3
Linearity of the BOLD response fMRI/MEG a
comparison
AIMS In this work, we investigate the degree to
which the BOLD response reflects underlying
neuronal activity. Here, the induced oscillatory
response in the ß-band (15-30Hz) is measured by
MEG and the haemodynamic response is measured by
fMRI, for stimuli of varying durations. These are
compared to determine whether the non-linearity
in the BOLD response reflects a non-linearity in
activity, as determined by MEG, or is a
consequence of haemodynamic coupling.
4
Methods
MEG A 275 channel CTF system running a third
order gradiometer configuration was used,
with a sample rate of 600Hz.
Co-registration to anatomical MRI was performed
using head digitisation (Polhemus
Isotrack). fMRI Data were acquired using a
Philips Achieva 3T system running EPI
(TR2000ms TE45ms 3mm3 voxels
192mm FOV). 18 axial slices were acquired
covering the motor cortex.

• PARADIGM
• Four healthy subjects took part in the study.
• The paradigm comprised visually cued brisk
  abductions of the right hand index finger at 4Hz.
• Four stimulus durations of 1, 2, 4 and 6
  seconds were presented randomly.
• MEG Experiment A single trial contained a 2s
  pre-stimulus rest period, the finger movement and
  a
• post stimulus rest period, giving a total trial
  length of 12secs. 20 trials were carried out for
  each
• duration.
• fMRI Experiment No pre-stimulus rest period
  was included. The total trial length was extended
  to
• 30secs to allow the haemodynamic response
  function to return to baseline. Due to the
  increased signal to
• noise 8 trials were carried out for each duration.

5
Data Analysis
MEG Analysis of MEG data was carried out using
Synthetic Aperture Magnetometry (SAM). For all
subjects, large variations in cortical
oscillatory power were observed in the ß-band
(15-30Hz). Localisation of these beta band
effects was achieved by comparing the oscillatory
power during two time windows termed the active
and passive windows.
Hilbert transforms were applied to timecourses of
electrical activity in the beta-band to obtain
the envelope of oscillatory activity. Linearity
of the MEG response was assessed by integration
of the Hilbert envelope of oscillatory power for
the duration of the stimulus, and also for the
post-movement rebound. The baseline for
integration was taken to be an average of the
points in the pre-stimulus rest period.
6
Data Analysis

• fMRI
• In order to identify areas of significant (p0.05
  corrected) BOLD contrast, standard
• regression techniques were employed using
  SPM.
• BOLD time courses were extracted from fMRI data
  using regions of interest defined
• from the BOLD statistical maps.
• Linearity of the BOLD response was assessed by
  calculating the area under the
• BOLD timecourses. An average signal value
  was calculated within the window
• (25slttlt30s) and used to give a baseline for
  integration.

7
Results Spatial Localisation
MEG beta band 15-30Hz response. ERD. Tstat
1.1 MEG beta band 15-30Hz
response. ERS. Tstat 1.1
BOLD Response T-stat 5.1
8
MEG Results Group Average
Hilbert Transforms of Virtual Electrode traces
for 4 subjects at peak beta-band activity.
1s
4s
2s
6s
9
RESULTS MEG Linearity
Figure 4 a) Hilbert Transform of peak
beta-band activity for 4s duration. Red lines
show time period used for integration. Green line
depicts baseline for integration. Pale blue
shaded area shows integrated area of power loss
used to assess linearity. b) A plot of the
corresponding integrated area for each stimulus
duration and for each subject. Note the linear
trend.
b)
Figure 5 a) Hilbert Transform of peak
beta-band activity for 4s duration. Pink shaded
area shows integrated area of rebound used to
assess linearity. b) A plot of the integrated
area of rebound for each stimulus duration, in
each subject. This is significantly less linear
than the power loss response.
10
Discussion and Conclusion

• The good spatial correlation between ß-band
  activity and the BOLD response suggests that the
  two measurements are closely related. Both
  localise to the contra-lateral primary motor
  cortex with the post-movement rebound being found
  slightly anterior to the movement related
  desynchronisation.
• The BOLD responses behave as predicted, with
  timecourses for stimuli of longer durations
  peaking later. However, the responses are
  non-linear, with the responses for shorter
  durations being underestimated by a simple linear
  model.
• The ß-band power losses increase roughly linearly
  with stimulus duration and do not show the same
  non-linearity as the BOLD response. However, no
  trend is apparent in the same results for the
  post-movement rebound.
• This would imply that either non-linearities in
  BOLD are primarily haemodynamic in origin, or
  that they reflect non-linearity in other aspects
  of the neuromagnetic response such as the beta
  rebound, or phase locked transient or sustained
  effects.