| |
Abstract:
A new method (called Fixed Point Analysis, FPA) is presented
to analyze multi-channel nonstationary signals, especially EEG and
MEG data from ERP experiments. This method is based on dynamical
systems theory and combines (i) a clustering algorithm to detect
functional segments in the timecourse of the signal trajectory and
(ii) a nonlinear analysis of the spatiotemporal dynamics of the
segments. The usefulness of this approach is demonstrated on
single-subject data from an auditory middle latency ERP experiment.
All known waves from 3 ms up to 40 ms are detected by FPC. Waves at
12ms and 30ms are described by a two-dimensional dynamical model
with a signal and dynamics fit of 98%, and it is possible to
describe the generating dynamical processes mathematically. In a
second application on group-averaged visual ERP data, a focused
investigation of the P300 component leads to a structuring into 3
subcomponents. This method offers a tool for detecting and modeling
functional components in high-dimensional brain signals.
Application of the algorithm to single-subject and single-trial
data is possible. The development of group statistics based on FPC
is in progress.
|
