Location via proxy:   [ UP ]  
[Report a bug]   [Manage cookies]                
Scribd
Upload
58 views

From Raw Eeg Data To Erp: Eva A.M. Van Poppel, MSC

This document provides an overview of electroencephalography (EEG), event-related potentials (ERPs), and how to process raw EEG data in MATLAB and EEGLAB to produce ERPs. It describes what EEG and ERPs are, introduces basic MATLAB functions and syntax, and outlines the steps to import raw EEG data, filter, epoch, and analyze it in EEGLAB to create ERPs, including re-referencing, artifact rejection, and statistics.

Uploaded by

Abhinandan Chatterjee
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
58 views

From Raw Eeg Data To Erp: Eva A.M. Van Poppel, MSC

This document provides an overview of electroencephalography (EEG), event-related potentials (ERPs), and how to process raw EEG data in MATLAB and EEGLAB to produce ERPs. It describes what EEG and ERPs are, introduces basic MATLAB functions and syntax, and outlines the steps to import raw EEG data, filter, epoch, and analyze it in EEGLAB to create ERPs, including re-referencing, artifact rejection, and statistics.

Uploaded by

Abhinandan Chatterjee
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 30

From raw EEG data to ERP

Eva A.M. van Poppel, MSc


Overview
• What is EEG?
• What is an ERP?
• Introduction to MATLAB
• Break
• From raw EEG data to ERP in EEGLAB

2
What is EEG?

EEG stands for Electro


Encephalo Graphy
Picture by Saint Luke's Health System

It is the firing of neuronal pyromodal cells,


measured by electrodes on the outside of the brain
(scalp) in microvolts (µV).

3
What is EEG?

Picture by Saint Luke's Health System


Can we measure the activity of inner brainstructures,
like the hippocampus, with EEG?

No, we can only measure the outer layer of neurons,


the neocortex. To measure deeper brain structures,
you need other brain imaging techniques, like MRI.
But compared to MRI, EEG has a better timing.
4
Frequencies
What is an ERP?
ERP stands for
Event Related
Potential.
It is the evoked
brain response
after a stimulus.

The start of the


stimulus is
normally at time
point 0 ms.

6
Introduction to MATLAB
• MATLAB stands for Matrix Laboratory
• First of all, it is a calculator. For example, type 5 +
8*5 and press Enter
• A Matrix is an array (series) of data, stored in rows
and columns (2D). For example, type [1:5;6:10] and
press Enter.
• Variable is the way MATLAB stores data in the
working memory (workspace). Press arrow up and
make it A= [1:5;6:10]
Variable A now contains your matrix. Note that
variable ans will be overwritten.
7
• Now, we want to know what is in row 2, column 3 of
Variable A. We type: A(2,3) and press Enter
• What happens when you forget one of the dimensions
(row or column) and type A(4)?
• Type B=4 (Enter) C=5 (Enter). Type B+C (Enter)
• Sometimes, you don’t want to see all internal
calculations (for example when you load in a complete
EEG data set). In this case, you should use the
semicolon ;
• Type D=ones(1,10);
• How does D look like? What happens when you make
D=ones(10,1) ?
• This is called a vector, which is just one row or column
containing data (1D)

8
Transpose and Text
• Type E=D’
• Transpose ‘
• T = ‘Text’;
• Price = ‘10CHF’
• randn function makes a vector with normally
distributed random numbers, which are the same on
each Matlab start up.
• X=randn(5)
• Y=[1:2:10]
• Y = [1:2:10;2:3:15]

9
Loops
• Press the button «New script» in the left upper
corner.
• Type:
for i = 1:length(Y)
F(i) = Y(1,i)+1;
end
• What did you do? How does F look?

10
FOR & IF Loop
for j = 1:length(Y)
if F(j) > Y(1,j)
G(j) = Y(1,j)
end
end

11
MATLAB tricks
• Don’t show calculations ;
• Transpose ‘
• Text ‘’
• Comments %
• Paragraph %%
• Run full script F5
• Ctrl + C abort run
• Compare scripts
• Clear G; Clear all

12
Break

13
EEGLAB
• Download on
https://sccn.ucsd.edu/eeglab/downloadtoolbox.php
• Extract for example in Documents
• Click Set Path in Matlab  Add folder  eeglab
• Save the path, but don’t save this on a shared
server e.g. Rekenbeest
• Type eeglab in the command window and press
Enter

14
Memory settings

15
Import data
• Go to File  Manage EEGLAB extensions
 Data processing extensions

• Choose the amplifier / EEG data recording


software you use and install

• Go to File  Import data  Using EEGLAB


functions and plugins
and choose your data type

16
Names
Edit  Channel
locations.
Use a template
with XYZ
coordinates and
channel names for
the EEG cap you
used.

17
Reference
• The ERP
value at
electrode X is
the value of
electrode X
with respect
to the
location of
the reference
electrode.

• Therefore, we re-reference the recorded data to the


average of the mastoid electrodes behind the ears.
18
Re-reference
• In EEGLAB, click Tools  Re-reference  Re-
reference data to channel(s): M1 M2
• Exclude channels containing non-EEG data, like
EMG, EOG or trigger channels.
• Add current reference channel back to the data
when you want to restore the data of the physical
reference electrode. You cannot use this when you
recorded with average reference (TMSi & ANT).

19
Filter
• A high-pass filter passes signals with a frequency
higher than a certain cutoff frequency. For EEG
data, you should use a high-pass filter of at least
0.1 Hz to remove the lower drift noise.
• A low-pass filter passes signals with a frequency
lower than a certain cutoff frequency. For ERP
analysis, we typically use a low-pass band from 35
Hz.
• A filter shifts the signal. Therefore, we first run a
high-pass filter, and afterwards run the low-pass
filter to shift it back.

20
Filter
• A filter never cuts off at the given band exactly.
• The slope is dependent of the filter order.
• EEGLAB uses a basic FIR filter as a default
(Finite Impulse
Response)
• The line noise
(electricity network in
Europe) is at 50 Hz.
You need a notch filter
of at least 48-52Hz to
filter this out. 21
Filter
• Because of filter artifacts at the end and beginning
of the data, it is important you filter the data before
epoching (cutting)
• In EEGLAB, Tools  Filter the data  Basic FIR filt
 Lower edge of the frequency pass band  0.1
Hz
• In EEGLAB, Tools  Filter the data  Basic FIR filt
 Higher edge of the frequency pass band  35
Hz

22
Epoch
• Now, we want to aline our data, to make sure all
trials start at 0 ms at the start of the stimulus.
Therefore we “cut” our data -1000 ms before the
stimulus and lasting 2500 ms after.
• In EEGLAB, Tools  Extract epochs  Epoch limits
in seconds -1 2.5

23
Baseline correction
• Baseline correction is the procedure of relativizing
the brain signal of interest (evoked response) with
respect to a control (baseline) signal
• In wake data, you normally use -200 to 0 ms before
the stimulus onset as the baseline
• In (deep) sleep data, we use -1000 to 0 ms as the
baseline
• EEGLAB automatically asks for the baseline period
in ms after the epoch settings

24
Artefact rejection
• Plot  Channel data scroll
• Settings  Time range to display  1 Epoch
• In wake data, EEG values > 75µV are considered
artefacts and those trials are marked for rejection.
In (deep) sleep data, this rule does not apply
• Mark trials containing movements and artefacts for
rejection, write the trial number in an Excel
• You need at least 25 correct trials per subject for a
nice ERP without noise
• Normally, 5 -10% of trials contain artefacts
25
Interpolation
• Some channels are noisy. In this case, we don’t
reject the whole trial, but interpolate the channel
affected
• Interpolation is an artificial reconstruction of the
channel, using the data of the surrounding
channels
• Run the script Eva_Interpolation.m to interpolate
certain channels per trial
• When you want to interpolate a bad channel in all
trials, you have to do this before the epoching using
Tools  Interpolate electrodes
26
Final datasets
• When you interpolated bad channels, you can
delete the trials containing artefacts
• In EEGLAB: Edit  Select data  Epoch range fill in
trials to reject e.g. [1 10] Click remove these and ok
• Save as P5_Final.set

27
Create ERP
• File  Create study  Simple ERP study
• Load in all final datasets
• One row means the same subject, so two files on
the same row means a within subject design,
where each subject on a seperate row means a
between subject design

28
ERP Statistics
• Download the Fieldtrip toolbox and save it in your
path
• Or go to File  Manage EEGLAB extensions  Data
processing extensions  Fieldtrip-lite
• Choose the “Montecarlo/Permutation based”
statistics with “Cluster correction (CC)”

29
Sources
Discovering Statistics Using SPSS, Andy Field, 3rd edition, 2009,
SAGE Publications Ltd.

Discovering Statistics Using R,


Andy Field, 1st edition, 2012, SAGE Publications Ltd.

https://ch.mathworks.com/help/matlab/matlab_prog/loop-
control-statements.html

https://sccn.ucsd.edu/eeglab/downloadtoolbox.php

http://www.fieldtriptoolbox.org/download

You might also like