The Neurophysiological Biomarker Toolbox (NBT)

ICA - Independent Component Analysis

Version: June 2017

Approximate time needed to complete the tutorial: 25 min.

An EEG signal is a mixture of signals from many sources. Some of these sources are not from the brain and produce so-called artifacts. Typical sources of artifacts are eye-blinks, breathing, heartbeat, or 50 Hz line noise. We want to clean our EEG signal from artifacts and, therefore, need a way of filtering these artifacts away. Independent component analysis (ICA) is a method that can extract artifacts from an EEG signal.

ICA will find, as the name indicates, independent components (i.e., different components / sources) in the EEG signal. We can then select which of these components we want to reject and remove them from the EEG signal using a mathematical procedure called “projection”.

The ICA components are signals that have been combined from several EEG channels, this means that they act as spatial filters on the data. What the ICA algorithm does is to put different weights on each EEG channel such that each component is “independent” from all other components.

a more technical explanation

Common ICA components

This document contains frequency and time patterns of common ICA components.

Note that these data are relative to 129-channel recordings in eyes-closed rest condition (only for heart beating the recordings are from a system with fewer electrodes). Some artifacts are more evident for specific channels location (i.e. eye channels, sensori-motor channels).

Common ICA components

How do I run ICA?

For this tutorial we will use a 1 minute sample signal. It is the same signal you used during the How to remove bad channels and transient artifacts tutorial!

In the following tutorial we assume you have rejected bad channels and transient artifacts as described in How to remove bad channels and transient artifacts. If something goes wrong during this tutorial, you can reload the file using the method shows in Getting started with EEG analysis in NBT.

  1. Go to Pre-processing | ICA | Run ICA on good channels only. ICA only gives good results if the signal is relatively stationary. We, therefore, apply a 0.5 Hz high-pass filter, this happens automatically (and is only done for the ICA, your signal will not be filtered).
  2. NBT will ask you how many components you want. Answer 10.
  3. Wait for the ICA algorithm to run; it usually takes between 100 and 300 steps (if it takes much longer contact a tutor).

Rejecting ICA components

You will get different components and numbers from the ones below, because ICA is not an analytical process. The components also depend on which channels you have marked as bad, and which transient artifacts you have removed.

We will now identify brain components and artifactual components.

  1. Go to Pre-processing | ICA | Visualize ICA | Plot component activations

we inspect component activations, i.e., how the components change over time. Go to Pre-processing | ICA | Visualize ICA | Plot component activations (if no clear components exist, it might help to remove more transient artifacts). If the component activation traces are crossing a lot, you may want to change the scaling of the y-axis using the ' ' sign in the lower right corner of the scroll window.

Try to identify which components are brain components? And which are artifact? Artifact components are characterized by sharp changes, or huge variations in voltage, or in case of the present file sharp non-oscillatory transitions that originate from eye sacades. Brain components would, e.g., have a clear alpha rhythm (~ 10 Hz).
Which components would you like to get rid of? (discuss your choice with the neighboring student)

We see that components 4 and 8 have clear alpha rhythms (between 8-13 Hz). These will often have a so-called dipolar topography as you can verify by going to Pre-processing | ICA | Visualize ICA | Reject component by map, click Ok. A window will pop up asking “Click OK to return to NBT”, do NOT press OK until you finish visualizing the component power spectrum (detailed below in the steps). Eye-blinks or sacades will show a topography around the eye-region (frontal electrodes). On the ICA maps, you will also find topographies that are very localized. These reflect bad channels, because brain sources will always produce signal at multiple electrodes when working with high-density EEG (~100 channels). Bad channels should not be removed by rejecting components by map (see below how to remove these single-channel ICA components. Do this first and then re-run the ICA).

Before we mark artifact components as bad we look at the power spectrum and topographic plots. Click on the buttons for components that you believe to reflect sources of artifacts. A power spectrum with a map pops up. If you still think this component is an artifact click 'accept' (see Figures below), and click Ok (else just click Ok). It often helps to view components activations, power spectrum and topographic plots together, because it is the combined information of scalp topography and temporal profile of the signal that safely allows distinguishing between neuronal and artifact sources.

Component 4 clearly comes from the brain, because of the ~10 Hz peak in the power spectrum, and the typical dipolar activation on the topographic map.

On the other hand, we classify component 6 as an artifact component, because of the typical eye-region topographic map. The component also does not have (a strong) ~10 Hz peak in the power spectrum.

Only remove maximal 3-4 components! And only remove components that are clear artifacts.

The components you marked as bad will be removed from the signal automatically before computing biomarkers. The signal without the bad components is not saved in the file since we do not want to change the raw signal. However, the indices of the bad components are now stored in the SignalInfo in the info-file, which means that ICA does not have to be run when biomarkers are computed at a later time point.

Which components did you select? [The most prominent artifacts that should be removed were components 2, 5 and 6 (based on the second image on this page).]

Go to next tutorial: Compute Biomarkers

Advanced: Removing single-channel ICA components

We could refine the result by removing components that appear only to represent a single channel. To identify the channels that are causing these single-channel ICA components and to set them as a bad channels, do this:

First, select Pre-processing|ICA|Visualize ICA|Reject component by map and identify components with single-channel topography.

From the Figure above it seems that components 1 and 7 represent single channels.

We remove them by returning to the function Pre-processing|Artifact Removal|Mark transient artifacts for current NBT signal

First, identify or estimate which electrodes / channels the single-channel ICA components on the map represent.

You can do this by comparing the ICA map with an EEG channel map.

Then, return to marking transient artifacts (Pre-processing|Artifact Removal|Mark transient artifacts for current NBT signal), to manually detect and mark the bad channels.

Alternatively, you could select Pre-processing|ICA|Mark ICA components as bad and write 1 7. NBT will then try to estimate which channels underlie these components and add them to the list of bad channels. However, you should be careful with this function! It is still experimental, and it might sometimes be better not to remove single channel ICA components.

You will also get single channel ICA components if you try to calculate more components than there are sources in the signal

Take Home Message. What are the characteristics of a normal neurophysiological component? I.e. what should you NOT reject?
- Scalp maps show dipoles.
- Spectral peaks at typical EEG frequencies (e.g., alpha rhythms)
- Regular ERP-image plots (the colored lines in the top-right image of the pop_prop show that activity is NOT peaking only in a few trials).

Click here if you want to read the EEGLAB tutorial on how to deal with ICA!

Go to next tutorial: Biomarker computation

If you are interested in a more advanced tutorial on Biomarker Computation, follow this link: Compute Biomarkers

tutorial/how_to_use_ica_to_remove_artifacts.txt · Last modified: 2017/06/13 19:06 by Mia Thomaidou
The NBTwiki platform - version 2.8 - 9 May 2013
Copyright (C) 2008-2015