Transcript Contrasts & Inference

Contrasts & Inference - EEG &
MEG
Outi Tuomainen & Rimona Weil
17.5.2005
mfd
Outline
• ERPs/ERFs in SPM: a revision
• A short introduction to the "conventional"
quantification of ERPs
• Contrasts and inference in M/EEG vs. fMRI
• How to do it in SPM + things to bear in mind
How does SPM/EEG work?
Preprocessing
Raw
M/EEG data
Projection
SPM5-stats
2D - scalp
SPM{t}
SPM{F}
Control of FWE
mass-univariate
analysis
Single trials
Epoching
Artefacts
Filtering
Averaging, etc.
3D-source
space
Kiebel, S. 2005
Revision: ERPs/ERFs
Average ERPs as an estimate of
event-related EEG activity?
Assumption 1: Detected signal
should have stable characteristics in
each single trial
- Instead: multiple components whose
amplitude and latency can vary
independently (e.g. latency jitter)
- So: the averaged ERP may present
only gross picture of the neural
processes elicited by the event of
interest
Assumption 2: Background EEG is
random and uncorrelated with ERP signal
- Instead: EEG is not entirely uncorrelated
with event-related activity
Kiebel & Friston 2004
Data (at each voxel)
Single subject
Multiple subjects
Trial type 1
...
...
Subject j
...
...
Trial type i
Subject 1
Trial type n
Subject m
Kiebel, S. 2005
Revision: ERPs/ERFs
• "ERPs are signal-averaged epochs of EEG that are
time-locked to the onset of stimulus"
• So a waveform can be seen as a time series that plots
scalp voltage (µV, T) over time (ms)
• ERPs are usually recorder at multiple scalp electrode
sites
spatial parameter to complement the
temporal and frequency information
• Quantifying ERPs: can be organised into three
categories: temporal, spatial and spatiotemporal
Quantification of ERPs
Cond1
A) Temporal:
- how waveforms recorded at
individual sites vary over time
across experimental conditions
- amplitude and latency as a
function of condition
B) Spatial:
- topographic mapping:
quantifying variation in voltage
across the scalp electrode array at
single time point or time window
Cond2
Cond3
C) Spatiotemporal:
- how scalp topographic patterns
vary across time (correlation of
successive topographic maps)
Quantification of ERPs
Effect-Specific Hypothesis vs. EffectUnspecific Hypothesis
- for example: component should be
present at Cond1 not in Cond2 -> a priori
restriction to a set of electrode sites and
time window
- Quantifying the waveform:
A) peak amplitude (max/min), mean
amplitude (typically arithmetic average),
peak-to-peak amplitude, mean area
amplitude
B) latency measures: max/min point in a
time window (peak-picking), onset/offset
latencies
Mean amplitude (µV), peak amplitude
and latency measures (µV, ms)
Statistics: (ANOVA/MANOVA and
appropriate corrections* and follow-up
tests); e.g. group-electrode-condition
Mauchly’s test for sphericity ≤ 0.05;
Greenhouse-Geisser and Huynh-Feldt
corrections
Quantification of ERPs
- Mean amplitude by Condition (2,
within-subject factor) & Group (2)
at Fz (electrode, within-subject
factor)
- Electrode factor?
-> e.g. if Left Hemisphere
electrodes are likely to be
systematically different from Right
Hemisphere electrodes
- In high-density montages it is a
good idea to divide electrodes into
averaged regions (anteriorposterior, left-right, ventral-dorsal)
Outline
• ERPs/ERFs in SPM: a revision
• A short introduction to the "conventional"
quantification of ERPs
• Contrasts and inference in M/EEG vs. fMRI
• How to do it in SPM + things to bear in mind
References in the end ….