Biomedical Engineering Reference
In-Depth Information
4
2
Insula
0
Thalamus
2
−
−
1.5
1
Alpha
0.5
0
4
Cingulate
Sup. temp.
inf. front.
Occipital
2
0
−
2
−
0
50
100
150
200
250
Time(s)
Figure 12.13
(a) Time course of the average MRI percentage signal change for regions of interest in which
the BOLD signal was positively (top) and negatively (bottom) correlated with alpha rhythm for a subject. At the
center is the alpha power time course convolved with a hemodynamic response function, which was used as
the independent response model to create the tomographic map of alpha activity. (b) Regions where the MR
signal increased and decreased with elevations in alpha power. The bottom bar shows the Pearson correlation
value between the signal intensity and alpha power modulation. (
From:
[50]. © 2002 Lippincott Williams &
Wilkins. Reprinted with permission.)
weighted and superimposed activity of multiple neuronal current generators, which
might be functionally and spatially separate. To study the fMRI correlates of the
alpha rhythm arising from a particular region of interest (or as a result of a particu-
lar activity, such as attention modulation), it is necessary to account for, and reverse,
this superposition and smearing effect as much as possible. Statistical methods such
as ICA are proving to be promising in separating out these functionally independent
activations, followed by extraction of the time course of their spectral power [76].
Alternatively, simple spatial filters such as the Laplacian or large Laplacian are able
to somewhat localize the activity of the EEG to that originating from the cortical
areas directly underneath the respective electrodes.
More recently, an alternative approach to this correlation analysis has been pro-
posed, that of using the frequency band of interest as a regressor in the statistical
model used for generating fMRI images.
Traditionally, fMRI images are based on statistical parametric mapping (SPM)
and employ a combination of classical statistics and topological inference, describ-
ing voxel-specific responses to experimental conditions. A detailed overview of the
physical and mathematical basis of fMRI is beyond the scope of this text and can be
found in [77]. Briefly, fMRI data is first spatially processed and registered onto a
common anatomical space. The responses in this space are characterized using the
general linear model (GLM). The GLM serves to describe the responses using a con-
volution model of the standard hemodynamic response function (HRF). This tenta-
tively explains the fact that BOLD signals are mathematically represented as the
delayed vascular response to a neural excitation function.
Search WWH ::
Custom Search