Signal Processing for Functional Brain Imaging: ICA Lab Exercise - - PowerPoint PPT Presentation

Signal Processing for Functional Brain Imaging: ICA Lab Exercise

Dimitri Van De Ville Medical Image Processing Lab, EPFL/UniGE dimitri.vandeville@epfl.ch

April 10, 2014

Lab exercise

2

n Fusion ICA Toolbox [Vince Calhoun and colleagues]

n

http://mialab.mrn.org/software/fit/docs/v2.0a_FIT_Manual.pdf

n Download and install the toolbox and data from

the course’s website

n http://miplab.epfl.ch/teaching/micro-513/

n Add path of toolbox

n addpath(genpath(‘/path/to/toolbox’)); n addpath(genpath(‘C:/path/to/toolbox’));

n Make new directory

n ! mkdir xxx_analysis n cd xxx_analysis

n Launch toolbox

n fusion

Symmetric fusion

n Different features “at same level” in decomposition

n Joint ICA (blind) n Parallel ICA (semi-blind) 3

Data (X)

feature 1 controls

=

Common mixing matrix

x

Joint independent components (S)

feature 2 patients feature 1 feature 2

Data (X)

feature 1 controls

=

mixing matrix 1

x

ICs (S1)

patients feature 1

Data (X)

controls

=

mixing matrix 2

x

ICs (S2)

feature 2 patients feature 2 maximize correlation

4

Joint ICA

Dataset #1: fMRI + EEG ERP

n Combine (non-simultaneously acquired) EEG & fMRI

n Auditory oddball experiment

n Data

n For each subject

n fMRI feature = peak of “target”-locked fMRI map n EEG feature = electrode Cz of “target”-locked EEG signal

5

[Calhoun et al., 2006]

Joint ICA: Input

n Group average of fMRI peak map & ERP

6

[Calhoun et al., 2006]

Joint ICA

n After launching fusion,

select joint ICA button.

n Clicking on Setup

Analysis button, a window will open to select the directory where all the analysis information will be stored.

n Set the initial

parameter.

7

Have You Selected The Data Files?

n Select the data in setup analysis window.

n One group, two features (fMRI & EEG). n Do not use “file selection window” for data selection

but use pattern at root data directory.

n Enter the file patterns for both features. n Select the data directory. FMRI beta maps & EEG ERP of all

subjects are in the subdirectory erp_fmri.

n Run Analysis

8

Visualize

n Click on Display button. n Create EPR-FMRI movie.

n In “Utilities” menu in Display

GUI window.

9

[Calhoun et al., 2006]

11

Components with ERP load

[Calhoun et al., 2006]

Maps at specific ERP timepoints

12

[Calhoun et al., 2006] N1 - primary auditory cortex N2 - sensory integration, secondary aud. cortex P3a - frontal and parietal lobes

ERPs at specific positions in the brain

13

[Calhoun et al., 2006]

14

Parallel ICA

Dataset #2: fMRI + SNP

n Setup analysis

n Parallel ICA n Two groups (HC & SZ), two features (fMRI & Gene) n Do not use “file selection window” for data selection,

but use pattern at root data directory

n data not organized in one group folder n select: fmri_gene_sampled/Healthy n select: fmri_gene_sampled/SZ

n Run analysis n Visualize

n Use “sorting” to get features with most correlation between

mixing coefficients

15