Image Processing Reference
In-Depth Information
The relative f MRI weighting was tested [146] in an MEG experiment and
conjoint f MRI/MEG analysis results were obtained similar to those reported
in previous f MRI, PET, MEG, and intracranial EEG studies. F. Babiloni et al.
[147] followed Dale et al. [146] in a high-resolution EEG and f MRI study to
incorporate nonthresholded f MRI activation maps with other factors. First of
all, W f MRI was reformulated to ( W f MRI ) ii
v 0 ) i / max , where i corre-
sponds to the relative change of the f MRI signal in the i th voxel, and max is
the maximal detected change. This way the relative E/MEMEG/f MRI scheme
is preserved and locations of stronger f MRI activations have higher prior
variance. Finally, the three available weighting factors were combined: f MRI
relative weighting, correlation structure obtained from f MRI described by the
matrix of correlation coefficients K S , and the gain normalization weighting
matrix W n (Section 8.2.3.2): . Although W f MRI alone
had improved EMSI localization, the incorporation of the K S led to finer local-
ization of neuronal activation associated with finger movement.
Although most of the previously discussed DECD methods are involved in
finding minimal norm solution, the fMRI-conditioned solution with minimal -
norm (regularization term in Equation 8.6, ) is shown to provide a sparser
activation map [148] with activity focalized to the seeded hotspot locations [143].
An fMRI-conditioned linear inverse is an appealing method due to its simplicity,
and the rich background of DECD linear inverse methods derived for the analysis
of E/MEG signals. Nonetheless, one should approach these methods with extreme
caution in a domain in which nonlinear coupling between BOLD and neural activity
is likely to overwhelm any linear approximation [141].
=
v 0
+
(1
WWWKWW
Q
=
12
/
12
/
12
/
12
/
|
R
|
n
s
n
|
R
|
L 2
L 1
f ( QQ
=
1
8.4.3.5
Beamforming
Lahaye et al. [149] suggest an iterative algorithm for conjoint analysis of EEG and
fMRI data acquired simultaneously during an event-related experiment. Their method
relies on iterated source localization by the LCMV beamformer (Equation 8.10),
which makes use of both EEG and fMRI data. The covariance used by the
beamformer is calculated anew each time step, using the previously estimated
sources. Although the original formulation is cumbersome, this method appears
promising as (a) it makes use of both spatial and temporal information available
from both modalities, and (b) it can account for silent BOLD sources using an
electrometabolic coupling constant that is estimated for each dipole.
C X
8.4.3.6
Bayesian Inference
During the last decade, Bayesian methods became dominant in probabilistic
signal analysis. The idea behind them is to use Bayes' rule to derive a posterior
probability of a given hypothesis having observed data D , which serves as evi-
dence to support the hypothesis
pD H pH
pD
(
|
)
(
)
pH D
(
|
)
=
(8.18)
()
Search WWH ::




Custom Search