180 likes | 354 Views
ICA clustering vs. CORRMAP for EEG pre-processing using EEGLAB. Amir Omidvarnia Nov. 23, 2009. Outline. EEGLAB Guideline tips of Using ICA for EEG pre-processing Introduction to CORRMAP Results of ICA Clustering on continues data. EEGLAB Tips.
E N D
ICA clustering vs. CORRMAP forEEG pre-processing using EEGLAB Amir Omidvarnia Nov. 23, 2009
Outline • EEGLAB Guideline tips of Using ICA for EEG pre-processing • Introduction to CORRMAP • Results of ICA Clustering on continues data
EEGLAB Tips • “Finding N components (from N-channel data) typically requires more thankN^2 data sample points (at each channel), where N^2 is the number of weights in the unmixing matrix and k is a multiplier. In our experience, the value of k increases as the number of channels increases.” From EEGLAB Website • It means that we should apply ICA on continues data, not after extracting epochs and applying ICA on each condition separately.
EEGLAB Tips • “We do not recommend this approach (concatenating datasets of all subjects together), since it tacitly assumes that the very same set of brain and non-brain source locations and, moreover, orientations, plus the very same electrode montage exist in each session and/or subject.” From EEGLAB Website • “We recommend performing one ICA decomposition on all the data collected in each data collection session, even when task several conditions are involved.” From EEGLAB Website
CORRMAP • CORRMAP is a plug-in for EEGLAB toolbox which gets an IC template as its input and finds all other similar IC’s within several datasets based on a correlation-like criterion. • CORRMAP needs at least two datasets and it finds at most three similar IC’s from each dataset. Therefore, we need to record more than one trial per session from each subject, or split one dataset into two or more, if we want to use CORRMAP.
CORRMAP • In fact, CORRMAP constructs a cluster of similar IC’s in compare to the input template. So, CORRMAP can be used along with ICA Clustering (It has been recommended by CORRMAP creators).
CORRMAP • Standard form of Pattern Reversal VEP From “Visual Evoked Potentials Standard (2004)”
CORRMAP • Template selection is heuristic and needs experience. • A typical proper template (in my mind!)
CORRMAP • A typical proper template (in my mind!) ERSP (A presentation of average PSD in time-frequency plane) ITC (A presentation of Alpha phase coherence)
CORRMAP • Typical output of CORRMAP as a part of EEGLAB (Of course, I have fed two identical datasets into it)
ICA Clustering on continues data (before extracting epochs) • Because of memory shortage, I’m not still able to apply ICA on most continues datasets. • Best cluster for the VEP dataset of Subject SS
ICA Clustering on continues data (before extracting epochs) Before After
ICA Clustering on continues data (before extracting epochs) Before After
ICA Clustering on continues data (before extracting epochs) Before After
ICA Clustering on continues data (before extracting epochs) Before After
ICA Clustering on continues data (before extracting epochs) Before After