Beyond Blind Averaging Analyzing Event-Related Brain Dynamics

Transcription

1 Beyond Blind Averaging Analyzing Event-Related Brain Dynamics Scott Makeig Swartz Center for Computational Neuroscience Institute for Neural Computation University of California San Diego La Jolla, CA sccn.ucsd.edu

2 The adequacy of blind response averaging IF. If equivalent stimuli (passively) evoke the same macro field responses (with fixed latencies and polarities or phase) in all trials and If all the REST of the EEG can be considered to be Gaussian noise sources that are not affected by the stimuli.. THEN The stimulus-locked average contains all the meaningful event-related EEG/MEG brain dynamics.

3 The inadequacy of blind response averaging EEG data ERP mean + EEG NOISEh (Highly) questionable assumptions:? The living brain produces passive responses?????????????????????? Ongoing EEG processes are not perturbed by events without transient phase locking no ERP contributions from ongoing EEG processes.? Evoked response processes are spatially segregated from ongoing EEG processes.? The true response baseline is flat.? Equivalent stimulus events evoke equivalent brain responses eventrelated brain dynamics are stationary from trial to trial. EEG1 EEG2 EEG4

4 Monkey LOOK see Monkey Do Do Monkey see Monkey do

5 ERP EEG? EEG? ERP EEG? EEG? EEG? EEG? ERP EEG? ERP C20-50 ERP

6 Electrodes EEG Local Synchrony Cortex Domains of synchrony Local Synchrony Spatial Source Filtering Scalp sensors average the dynamics of cortical (and non-brain) sources Skin Skull

7 2 nd -Order Beamforming Uses a forward head model (approximation) applied first, to separate cortical signals. Assumes the signals at every pre-defined brain voxel are uncorrelated. Minimizes 2 nd -order (pairwise) correlations. Spatial source filtering by beamforming and ICA Blind Beamforming by ICA Does not require a forward head model Can separate both cortical and non-cortical sources. A head model can be applied last. Assumes the signals from both cortical and non-brain sources are near independent with no source geometry constraint. Minimizes all orders of source interaction components have minimum mutual information.

8 ICA blind EEG source separation Unmixes scalp channel mixing (spatial averaging) EEG Cocktail Party

9 The recorded channel data = source mixtures Two-back working memory experiment 9 (of 100) channels and independent components Note the large difference between the continuous signal for this component (IC5) and its (repeated) ERP average! The ERP fails to capture its regular alpha bursting! Separated nearindependent source activities Julie Onton & S. Makeig (in press)

10 Simultaneously active dipolar independent components Equivalent dipoles Single dipole model Dual-symmetric dipole model Julie Onton & S. Makeig (in press)

11 ERP-Image Plotting µv, green is 0! " #$ %&' ("# )*"+,!--

12 What produces event-related potential averages (ERPs)? Inter-trial Coherence (ITC) ( phase-locking factor ) Significant consistency of local phase of a physiological waveform across experimental trials. EVENTS given given frequency EVENT delay LOCAL PHASE EVENT LOCAL PHASE EVENT LOCAL PHASE PHASE LOCKING

13 Single trials ERP-image plot of simulated data µv Average ERP Mean of 400 simulated trials NB: NO amplitude increase! p = 0.02 Inter-trial coherence (ITC) p = 0.02

14 Quasi-pure independent component ERP Trials ordered by phase/latency ITC

15 Quasi-pure scalp channel phase locking Phase -Ordereed Trials µv Stimulus Hz High Alpha, Non-target ERP µv ITC ITC p = Time (ms) Makeig et al., Science 2002

16 Collections of single trials, even at the source level, are regular, but in multiple ways so they appear noisy! The ERP Image Stim RT Sorted Trials Time (ms µv Time (ms) 43.7

17 ERP image of alpha power in single trials Load 7 Load 5 db Load 3 Onton & Makeig, in prep.

18 So, no single measure captures the event-related brain dynamics ERP Is a given ERP feature a true ERP, or phase resetting, or both (ITC)!? Does it coincide with an EEG power increase or decrease (ERSP)? ITC No amplitude effects (ERSP & ERP)! ERSP Does not show phase statistics (ITC). Is a given power increase also in the ERP, or not? But, together are these measures yet enough?? No way! For one, all these measures are still blind averages of all available good trials But are all the trial activities the same??

19 Multiple trial modes from an component cluster trials time-locked to letter onsets in a memory task Onton et al., NeuroImage 05

20 No single measure can capture all the event-related brain dynamic features of the recorded data! " # " $ %%%%

21 EEGLAB An open-source EEG/MEG signal processing environment for Matlab