| |
Abstract:
Cortical current sources of event-related potential
components can be estimated for individual subjects by combining
high-density EEG recordings with structural MRI scans. An unsolved
problem of this approach is whether a computed solution deviates
systematically from a baseline distribution of current sources. To
test this question we adapted the permutation test of Blair and
Karniski (1994). Current sources were estimated with a minimum-norm
algorithm from potential maps recorded from 124 electrodes in a
picture-word priming paradigm. For data reduction the estimated
sources of each solution were projected to a "normalized brain" as
described by Talairach und Tournoux (1998) with 864 volume
elements. Solutions were calculated for twelve time- points taken
from a prestimulus baseline and a poststimulus "epoch of interest".
Permutations of a given result (6 pre, 6 post source patterns) were
created by forming all possible pairwise comparisons. For each
permutation a t-statistic was calculated for each volume element
across the six replications and summed across all volume elements.
The distribution of the resulting t-sum statistic can be used to
determine the probability of the given result. A significant
difference in the cortical source distribution of the two time
ranges could be proved for each individual subject although the
shift of sources from the baseline to the epoch-of-interest
concerned not more than about 5% of the volume elements.
|
