Contrasts & Inference - EEG & MEG

Himn Sabir

1

Topics

• 1st level analysis• 2nd level analysis• Space-Time SPMs• Time-frequency analysis• Conclusion

2

Voxel Space

3

(revisited)

2D scalp projection

(interpolation in sensor space)

3D source reconstruction

(brain space)

2/3D images over peri-stimulus time bins

Data ready to be analysed

M/EEG modelling and statistics

4

Epoched time-series data

Data is analysed using the General Linear model at each voxel and Random Field Theory to adjust the p-values for multiple comparisons.

Typically one wants to analyse multiple subjects’ data acquired under multiple conditions

2-Level Model

Time

Intensity

Tim

e

Single voxel time series

Model specification

Parameter

estimation

Hypothesis

Statistic

SPM

1st Level Analysis

5

Epoched time-series data

At the 1st level, we select periods or time points in peri-stimulous time that we would like to analyse. Choice made a priori.

Example: if we were interested in the N170 component, one could average the data between 150 and 190 milliseconds.

Time is treated as an experimental factor and we form weighted-sums over peri-stimulus time to provide input to the 2nd level

0

1

• Similar to fMRI analysis. The aim of the 1st level is to compute contrast images that provide the input to the second level.

• Difference: here we are not modelling the data at 1st level, but simply forming weighted sums of data over time

1st Level Analysis

6

Epoched time-series data

Example: EEG data / 8 subjects / 2 conditions

1. Choose Specify 1st-level

2. Select 2D images

3. Specify M/EEG matfile

4. Specify Time Interval

For each subject

5. Click Compute

Timing information

SPM output:

2 contrast images

average_con_0001.img

2nd Level Analysis

7

Epoched time-series data

Given the contrast images from the 1st level (weighted sums), we can now test for differences between conditions or between subjects.

1Tc =

2X

2

+ 2

second levelsecond level

-1 1

2nd level contrast 2nd level model = used in fMRI

SPM output:

Voxel map, where each voxel contains

one statistical value

The associated p-value is adjusted

for multiple comparisons

2nd Level Analysis

8

Epoched time-series data

Example: EEG data / 8 subjects / 2 conditions

1. Specify 2nd-level

2. Specify Design

SPM output:

Design Matrix

2nd Level Analysis

9

Epoched time-series data

Example: EEG data / 8 subjects / 2 conditions

3. Click Estimate

4. Click Results

5. Define Contrasts

Output: Ignore brain outline:

“Regions” within the 2D map in

which the difference

between the two conditions is significant

Space-Time SPMs (Sensor Maps over Time)

10

Time as another dimension of a Random Field

Advantages:

• If we had no a priori knowledge where and when the difference between two conditions would emerge

• Especially useful for time-frequency power analysis

Both approaches available: choice depends on the data

We can treat time as another dimension and construct

3D images (2D space + 1D peri-stimulus time)

We can test for activations in space and time

Disadvantages:

• not possible to make inferences about the temporal extent of evoked responses

Space-Time SPMs (Sensor Maps over Time)

11

How this is done in SMP8

Example: EEG data / 1 subject / 2 conditions (344 trials)

2. Choose options

32x32x161 images for

each trial / condition

3. Statistical Analysis

(test across trials)

4. Estimate + Results

5. Create contrasts

1. Choose 2D-to-3D image on the SPM8 menu and epoched data: e_eeg.mat

Space-Time SPMs (Sensor Maps over Time)

12

How this is done in SMP8

Example: EEG data / 1 subject / 2 conditions (344 trials)

Ignore brain outline!!!

More than 1 subject:

• Same procedure with averaged ERP data for each subject

• Specify contrasts and take them to the 2nd level analysis

Overlay with EEG image:

Time-Frequency analysis

13

Transform data into time-frequency domain

Not phase-locked to the stimulus onset – not revealed with classical averaging methods

[Tallon-Baudry et. al. 1999]

Useful for evoked responses and induced responses:

SPM uses the Morlet Wavelet Transform

Wavelets: mathematical functions that can break a signal into different frequency components.

The transform is a convolution

The Power and Phase Angle can be computed from the wavelet coefficients:

Time-Frequency analysis

14

How this is done in SPM8:

1. Choose time-frequency on the SPM8 menu and epoched data: e_meg.mat

2. Choose options

t1_e_eeg.mat and t2_e_eeg.mat power at each frequency, time and channel (t1*); phase angles (t2*)

3. Average

4. Display

mt1_e_eeg.mat and

mt2_e_eeg.mat

Example: MEG data / 1 subject / 2 conditions (86 trials)

5. 2D Time-Frequency SPMs

Summary

15

(2D interpolation or 3D source reconstruction)

1st Level Analysis

(create weighted sums of the data over time)

(contrast images = input to the 2nd level)2nd Level Analysis

(test for differences between conditions or groups)

(similar to fMRI analysis)Time-Space SPMs

(time as a dimension of the measured response variable)

Time-Frequency Analysis(induced responses)

Projection to voxel space

References

• S. J. Kiebel: 10 November 2005. ppt-slides on ERP analysis at http://www.fil.ion.ucl.ac.uk/spm/course/spm5_tutorials/SPM5Tutorials.htm

• S.J. Kiebel and K.J. Friston. Statistical Parametric Mapping for Event-Related Potentials I: Generic Considerations. NeuroImage, 22(2):492-502, 2004.

• S.J. Kiebel and K.J. Friston. Statistical Parametric Mapping for Event-Related Potentials II: A Hierarchical Temporal Model. NeuroImage, 22(2):503-520, 2004.

• Todd, C. Handy (ed.). 2005. Event-Related Potentials: A Methods Handbook. MIT

• Luck, S. J. (2005). An Introduction to the Event-Related Potential Technique. MIT Press.

16

Thank You!

17

For difficult questions:j.kilner@fil.ion.ucl.ac.uk

(James Kilner)