evaluating ica components...10 12 10 12 .01 0 'ampsort' , [0 0 10 12] same data: sorted by...

Evaluating ICA components

Novice EEGLAB Workshop, Sept 22, 2011, Mallorca, Spain: Julie Onton – IC Analysis Tools

Eye blink correction (remove IC)

Identify eye-blink

components:

Novice EEGLAB Workshop, Sept 22, 2011, Mallorca, Spain: Julie Onton – IC Analysis Tools

IC rejection/back-projection

0

0

0

0

0

0

0

0

0

0

0

0

0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

Novice EEGLAB Workshop, Sept 22, 2011, Mallorca, Spain: Julie Onton – IC Analysis Tools

Eye blink correction (remove IC)

Novice EEGLAB Workshop, Sept 22, 2011, Mallorca, Spain: Julie Onton – IC Analysis Tools

Eye blink correction (remove IC)

Novice EEGLAB Workshop, Sept 22, 2011, Mallorca, Spain: Julie Onton – IC Analysis Tools

Eye blink correction (remove IC)

Novice EEGLAB Workshop, Sept 22, 2011, Mallorca, Spain: Julie Onton – IC Analysis Tools

Eye blink correction (remove IC)

Novice EEGLAB Workshop, Sept 22, 2011, Mallorca, Spain: Julie Onton – IC Analysis Tools

Evaluating ICA components

Novice EEGLAB Workshop, Sept 22, 2011, Mallorca, Spain: Julie Onton – IC Analysis Tools

Extract epochs

Novice EEGLAB Workshop, Sept 22, 2011, Mallorca, Spain: Julie Onton – IC Analysis Tools

Extract epochs

Novice EEGLAB Workshop, Sept 22, 2011, Mallorca, Spain: Julie Onton – IC Analysis Tools

Reject data epochs (automatic)

>> EEG = pop_autorej(EEG, 'nogui', 'on', 'eegplot', 'on');

High enough

to keep

eye blinks

High standard

deviation,

multiples

passes

Novice EEGLAB Workshop, Sept 22, 2011, Mallorca, Spain: Julie Onton – IC Analysis Tools

Component ERP envelope

Novice EEGLAB Workshop, Sept 22, 2011, Mallorca, Spain: Julie Onton – IC Analysis Tools

Definition: The data envelope

Data (all channels)

Data envelope

(max/min traces)

Novice EEGLAB Workshop, Sept 22, 2011, Mallorca, Spain: Julie Onton – IC Analysis Tools

IC back-projection envelope

Novice EEGLAB Workshop, Sept 22, 2011, Mallorca, Spain: Julie Onton – IC Analysis Tools

IC back-projection envelope

IC envelopes

plotted for simplicity

(instead of all

back-projected

channels)

Novice EEGLAB Workshop, Sept 22, 2011, Mallorca, Spain: Julie Onton – IC Analysis Tools

IC contributions to ERP envelope

Novice EEGLAB Workshop, Sept 22, 2011, Mallorca, Spain: Julie Onton – IC Analysis Tools

Artifact

Components

IC contributions to ERP envelope

Novice EEGLAB Workshop, Sept 22, 2011, Mallorca, Spain: Julie Onton – IC Analysis Tools

Evaluating ICA components

Novice EEGLAB Workshop, Sept 22, 2011, Mallorca, Spain: Julie Onton – IC Analysis Tools

Component ERP image

Novice EEGLAB Workshop, Sept 22, 2011, Mallorca, Spain: Julie Onton – IC Analysis Tools

Component ERP Images

Novice EEGLAB Workshop, Sept 22, 2011, Mallorca, Spain: Julie Onton – IC Analysis Tools

Component ERP Images

10 12 0

Phase-sorted image

Novice EEGLAB Workshop, Sept 22, 2011, Mallorca, Spain: Julie Onton – IC Analysis Tools

Component ERP Images

10 12

10 12 .01

0

Phase-sorted alpha power

Novice EEGLAB Workshop, Sept 22, 2011, Mallorca, Spain: Julie Onton – IC Analysis Tools

Novice EEGLAB Workshop, Sept 22, 2011, Mallorca, Spain: Julie Onton – IC Analysis Tools

Component ERP Images

10 12

10 12 .01

0

'ampsort' , [0 0 10 12]

Same data: Sorted by alpha amplitude

Phase-sorted alpha power

>> help erpimage

'ampsort' = [center_ms, prcnt, freq, maxfreq] Sort epochs by amplitude.

Novice EEGLAB Workshop, Sept 22, 2011, Mallorca, Spain: Julie Onton – IC Analysis Tools

Component ERP Images

10 12

10 12 .01

0

'ampsort', [0 0 10 12]

Same sorting order: Amplitude vs. activations

Evaluating ICA components

Novice EEGLAB Workshop, Sept 22, 2011, Mallorca, Spain: Julie Onton – IC Analysis Tools

Stationary signals

0 0.5 1-1

-0.5

0

0.5

1

0 0.5 1-1

-0.5

0

0.5

1

0 0.5 1-1

-0.5

0

0.5

1

0 0.5 1-4

-2

0

2

4

Time Time

Time Time

Ma

gn

itu

de

Ma

gn

itu

de

M

ag

nit

ud

e

Ma

gn

itu

de

Slide courtesy of Petros Xanthopoulos, Univ. of Florida

10 Hz 2 Hz

0 0.5 1-1

-0.5

0

0.5

1

0 0.5 1-1

-0.5

0

0.5

1

0 0.5 1-1

-0.5

0

0.5

1

0 0.5 1-4

-2

0

2

4

2+10+20 Hz 20 Hz

0 0.2 0.4 0.6 0.8 1-3

-2

-1

0

1

2

3

0 5 10 15 20 250

100

200

300

400

500

600

Ma

gn

itu

de

Frequency (Hz)

Power spectrum

Novice EEGLAB Workshop, Sept 22, 2011, Mallorca, Spain: Julie Onton – IC Analysis Tools

Figure, courtesy of Ravi Ramamoorthi & Wolberg

Freq. decomp. Sum of freq. Time

domain

Time

Frequency

Frequency

domain

Summing stationary signals and decomposing

Novice EEGLAB Workshop, Sept 22, 2011, Mallorca, Spain: Julie Onton – IC Analysis Tools

Average of squared absolute values

Novice EEGLAB Workshop, Sept 22, 2011, Mallorca, Spain: Julie Onton – IC Analysis Tools

Average of squared amplitudes

Overlap 50%

Novice EEGLAB Workshop, Sept 22, 2011, Mallorca, Spain: Julie Onton – IC Analysis Tools

Novice EEGLAB Workshop, Sept 22, 2011, Mallorca, Spain: Julie Onton – IC Analysis Tools

0 ms 10 ms 20 ms 30 ms 40 ms 50 ms 60 ms

5 Hz

10 Hz

20 Hz

30 Hz

5 Hz

10 Hz

20 Hz

30 Hz

0 ms 10 ms 20 ms 30 ms 40 ms 50 ms 60 ms

Average of

squared

values

Spectrogram or ERSP

Absolute = ERS

Relative = ERSP (dB or %)

Absolute versus relative power

Novice EEGLAB Workshop, Sept 22, 2011, Mallorca, Spain: Julie Onton – IC Analysis Tools

FFT Wavelet

Fre

quency

Difference between FFT and wavelets

Novice EEGLAB Workshop, Sept 22, 2011, Mallorca, Spain: Julie Onton – IC Analysis Tools

Time-frequency resolution trade off

Novice EEGLAB Workshop, Sept 22, 2011, Mallorca, Spain: Julie Onton – IC Analysis Tools

FFT

In between

Pure wavelet

Novice EEGLAB Workshop, Sept 22, 2011, Mallorca, Spain: Julie Onton – IC Analysis Tools

Wavelet (0)= FFT Wavelet (1)

1Hz

2Hz

4Hz

6Hz

8Hz

10Hz

Wavelets factor

exact same number of

wavelets at all freqs

Novice EEGLAB Workshop, Sept 22, 2011, Mallorca, Spain: Julie Onton – IC Analysis Tools

Wavelet (0.8) Wavelet (0.5) Wavelet (0.2)

Modified wavelets

Scaled to require more wavelets at higher freqs (less than FFT though)

Novice EEGLAB Workshop, Sept 22, 2011, Mallorca, Spain: Julie Onton – IC Analysis Tools

Slide courtesy of Stefan Debener Novice EEGLAB Workshop, Sept 22, 2011, Mallorca, Spain: Julie Onton – IC Analysis Tools

Evaluating ICA components

Novice EEGLAB Workshop, Sept 22, 2011, Mallorca, Spain: Julie Onton – IC Analysis Tools

Novice EEGLAB Workshop, Sept 22, 2011, Mallorca, Spain: Julie Onton – IC Analysis Tools

Plot IC ERSP

Time points lost at

beginning and end of epoch!

Plot IC ERSP

Pure green denotes

non-significant points

Novice EEGLAB Workshop, Sept 22, 2011, Mallorca, Spain: Julie Onton – IC Analysis Tools

Novice EEGLAB Workshop, Sept 22, 2011, Mallorca, Spain: Julie Onton – IC Analysis Tools

Plot IC ERSP

Increase # freq bins

padratio = 1 padratio = 2

Evaluating ICA components

Novice EEGLAB Workshop, Sept 22, 2011, Mallorca, Spain: Julie Onton – IC Analysis Tools

Cortex

Two EEG channels

A

Scalp channel coherence includes source confounds!

B C

Scalp channel coherence confounds

Novice EEGLAB Workshop, Sept 22, 2011, Mallorca, Spain: Julie Onton – IC Analysis Tools

Is the observed

coherence from

C getting stronger?

Or B and C becoming

synchronous?

Cortex C

A B

MANY EEG channels

Separate out

independent EEG

components Measure their

synchronization

Scalp channel coherence confounds

Novice EEGLAB Workshop, Sept 22, 2011, Mallorca, Spain: Julie Onton – IC Analysis Tools

Cross-coherence amplitude and phase

Animal picture Distractor picture P

ha

se

(d

eg

ree

) A

mp

litu

de

(0

-1)

5 6

Novice EEGLAB Workshop, Sept 22, 2011, Mallorca, Spain: Julie Onton – IC Analysis Tools

IC cross coherence

Be sure to mask by

bootstrap significance limits

Novice EEGLAB Workshop, Sept 22, 2011, Mallorca, Spain: Julie Onton – IC Analysis Tools

IC cross coherence

Novice EEGLAB Workshop, Sept 22, 2011, Mallorca, Spain: Julie Onton – IC Analysis Tools

coherence

phase lag

Exercise

• ALL

- Load stern.set, epoch on Memorize letters, reject noisy epochs

• Novice

- From the GUI, plot component ERPs with maps

- Plot an IC ERP image; try sorting by RT or phase, is there any

effect of the time-locking event on the activation pattern?

- Try plotting power in a specified frequency band; how consistent

are any power changes across trials?

• Intermediate

- Plot ERSPs for an IC; for FFT, vary the 'winsize' and 'padratio';

for wavelets, vary number of 'cycles' and window size factor

- Compare FFT and wavelet methods; Do the results agree?

- Plot ERSPs with no baseline and with different baseline periods;

how might this affect your results/conclusions?

• Advanced

- Plot cross coherence between two selected ICs

> Compare this result with cross coherence between two

channels that are highly weighted in the respective ICs

Novice EEGLAB Workshop, Sept 22, 2011, Mallorca, Spain: Julie Onton – IC Analysis Tools

Supplementary lessons

Novice EEGLAB Workshop, Sept 22, 2011, Mallorca, Spain: Julie Onton – IC Analysis Tools

2 components, comparison on the same trials

Coherence amplitude 1

Phase coherence 0

Coherence amplitude 1

Phase coherence 90

Coherence amplitude 1

Phase coherence 180

Tri

al 1

T

rial

2

Tri

al 3

COHERENCE = mean(phase vector) Norm 0.33

Phase 90 degree

Cross-coherence amplitude and phase

Novice EEGLAB Workshop, Sept 22, 2011, Mallorca, Spain: Julie Onton – IC Analysis Tools

IC ERP difference

What is the IC ERP difference between these 2 conditions?

Novice EEGLAB Workshop, Sept 22, 2011, Mallorca, Spain: Julie Onton – IC Analysis Tools

IC ERP difference

Novice EEGLAB Workshop, Sept 22, 2011, Mallorca, Spain: Julie Onton – IC Analysis Tools

IC ERP difference

Novice EEGLAB Workshop, Sept 22, 2011, Mallorca, Spain: Julie Onton – IC Analysis Tools