Presentation is loading. Please wait.

Presentation is loading. Please wait.

Wellcome Dept. of Imaging Neuroscience University College London

Similar presentations


Presentation on theme: "Wellcome Dept. of Imaging Neuroscience University College London"— Presentation transcript:

1 Wellcome Dept. of Imaging Neuroscience University College London
SPM for EEG/MEG Stefan Kiebel Wellcome Dept. of Imaging Neuroscience University College London

2 Overview: SPM5 for EEG/MEG
Statistical Parametric Mapping voxel-based approach Conventional analysis Localisation of effects Evoked responses and power Spatial forward modelling/ Source reconstruction Forward model important for source reconstruction and DCM Source reconstruction localises activity in brain space t-contrast: tests for single dimension in parameter space F-contrast: tests for multiple dimensions inference at first or second level (fixed or random effects) over conditions or groups: main effect, difference, interaction: average over time window parametric modulation with extrinsic variable power comparison in time-frequency domain Dynamic Causal Modelling Models ERP/ERF as network activity. Explains differences between evoked responses as modulation of connectivity.

3 EEG and MEG EEG MEG - ~1929 (Hans Berger) - Neurophysiologists
In the same way like SPM for fMRI Use SPM2‘s methods! However, model needs to be adapted to make proper inference Comparison with fMRI analysis to aid illustration Also, conventional model (to be shown) in the same framework - ~1929 (Hans Berger) - Neurophysiologists From clinical system to 64, 127, 256 sensors - Potential V: ~10 µV - ~1968 (David Cohen) - Physicists From ~ 30 to more than sensors - Magnetic field B: ~10-13 T

4 275 sensor axial gradiometer MEG system supplied by VSM medtech. VSM medtech says  Designed for unprecedented measurement accuracy, the combination of up to 275 optimum-length axial gradiometers and unique noise cancellation technology ensures the highest possible performance in some of today's most demanding urban magnetic environments. In the same way like SPM for fMRI Use SPM2‘s methods! However, model needs to be adapted to make proper inference Comparison with fMRI analysis to aid illustration Also, conventional model (to be shown) in the same framework

5 MEG data Example: MEG study of finger somatotopy
[Meunier 2001] 400 stimulations of each finger right In the same way like SPM for fMRI Use SPM2‘s methods! However, model needs to be adapted to make proper inference Comparison with fMRI analysis to aid illustration Also, conventional model (to be shown) in the same framework ~ 50 ms left Index f Little f

6 event-related potential/field (ERP/ERF)
single trials . . . average event-related potential/field (ERP/ERF)

7 Single trial/evoked response
Voxel spaces SPM 2D Single trial/evoked response sensor data SPM 3D

8 . . . . . . . . . . . . Data (at each voxel) Single subject
Multiple subjects Subject 1 Trial type 1 . . . . . . Subject j Trial type i . . . . . . Subject m Trial type n

9 Mass univariate Time Time model specification parameter estimation
hypothesis statistic single voxel time series Intensity SPM

10 How does SPM/EEG work? Preprocessing Projection SPM5-stats SPM{t}
SPM{F} Control of FWE Raw M/EEG data 2D - scalp mass-univariate analysis Single trials Epoching Artefacts Filtering Averaging, etc. In the same way like SPM for fMRI Use SPM2‘s methods! However, model needs to be adapted to make proper inference Comparison with fMRI analysis to aid illustration Also, conventional model (to be shown) in the same framework 3D-source space

11 2 level hierarchical model
SPM for M/EEG Time and time-frequency contrasts M/EEG data Design matrices 2 level hierarchical model 2D- or 3D- M/EEG data SPM{t} SPM{F} Preprocessing fMRI/ sMRI data Covariance constraints Corrected p-values

12 Conventional analysis: example
Average between 150 and 190 ms Example: difference in N170 component between trial types PST [ms] s1 a1 s2 Trial type 1 a2 s3 a3 General linear model (here: 2-sample t-test) s1 a4 s2 a5 Trial type 2 s3 a6

13 Summary statistics approach
Example: difference between trial types Contrast: average between 150 and 190 ms 2nd level contrast -1 1 . . . = = + Identity matrix second level first level

14 Gaussian Random Fields
t59 Control of Family-wise error Worsley et al., Human Brain Mapping, 1996 p = 0.05 Cluster Gaussian 10mm FWHM (2mm pixels) Search volume

15 Summary Conventional preprocessing in sensor space.
After preprocessing, convert to voxel-space. Adjustment of p-values! Analysis of power or time data. t-contrast: tests for single dimension in parameter space F-contrast: tests for multiple dimensions inference at first or second level (fixed or random effects) over conditions or groups: main effect, difference, interaction: average over time window parametric modulation with extrinsic variable power comparison in time-frequency domain SPM needed to get to the DCM bit. Cool source reconstruction.


Download ppt "Wellcome Dept. of Imaging Neuroscience University College London"

Similar presentations


Ads by Google