Preprocessing for lessening the influence of eye artifacts in eeg analysis

Alejandro Villena,Lorenzo J. Tardon,Isabel Barbancho,Ana M. Barbancho,Elvira Brattico,Niels T. Haumann

DOI: https://doi.org/10.3390/app9091757

2024-11-19

Abstract:We dealt with the problem of artifacts in eeg signals in relation to the usage of lengthy trials. Specifically, we considered eye artifacts found in eeg signals,their influence in the analysis of the data and alternatives to diminish their impact on later studies of brain activity on lengthy tasks. We proposed a scheme of partial rejection on independent signal components, providesd a method to extract eeg signal components with diministhed influence of eye artifacts, and assess the importance of using artifact free signal excerpts to extract signal components in order to analyze brain activity in a musical context.

Signal Processing,Information Retrieval

What problem does this paper attempt to address?

The problem that this paper attempts to solve is: in long - term experiments, how to reduce the influence of eye artifacts on electroencephalogram (EEG) signal analysis. Specifically, the author focuses on the artifact problem in EEG signals caused by eye blinks and eyeball movements during long - term music listening experiments. These artifacts will interfere with the accurate analysis of brain activity, so effective methods need to be found to mitigate their influence.

### Problem Background

1. **Sources of Artifacts**
- Eye - movement artifacts (such as blinks, eyeball movements) are one of the most common artifacts in EEG signals.
- These artifacts will cause voltage fluctuations, and then contaminate the EEG signals recorded by scalp electrodes.

2. **Limitations of Existing Methods**
- For short - term experiments, it can be processed by simply removing the experimental segments containing artifacts.
- But for long - term experiments (such as music listening), this method is not feasible because subjects will inevitably blink or move their eyes during the experiment.

3. **Research Motivation**
- In long - term experiments, especially those involving real - world music listening, how to effectively reduce the influence of eye - movement artifacts on EEG signals is an urgent problem to be solved.

### Solutions

To meet this challenge, the author proposes two methods based on independent component analysis (ICA) to reduce the influence of eye - movement artifacts:

1. **Partial Rejection (PR)**
- By identifying signal segments related to eye - movement artifacts and applying attenuation to them without completely deleting these segments.
- Use a window function (such as the Blackman window) for a smooth transition to reduce sudden signal changes.

2. **Artifact - Diminished Unmixing Matrix (W0)**
- Before performing ICA, remove the samples marked as artifacts from all EEG channels to generate an unmixing matrix that is less affected by artifacts.
- In this way, a new set of independent components can be obtained, and these components do not contain specific sources of eye - movement artifacts.

### Experimental Design

- **Participants**: 5 right - handed native Spanish speakers participated in the experiment, aged between 20 and 29 years old.
- **Stimuli**: It includes four instrumental music pieces of different styles and a 1 - minute - 20 - second Spanish speech.
- **Equipment and Procedure**: Use OpenVibe software and TMSi Porti system to record EEG data, with a sampling rate of 2048 Hz, and then down - sample to 250 Hz.

### Results and Discussion

- **Artifact Detection**: Manually label eye - movement artifacts and count the number of artifacts and their characteristics for each participant.
- **Method Comparison**: Evaluate the effects of the two methods by calculating the negative correlation coefficient between EEG channels and EOG channels.
- **Conclusion**: Both the partial rejection method and the artifact - diminished unmixing matrix method can effectively reduce the influence of eye - movement artifacts on EEG signals, but each has its own advantages and disadvantages, and the specific choice depends on the application scenario.

### Formula Summary

1. **Independent Component Analysis (ICA) Formula**
\[
s(t)=W\cdot x(t)
\]
where \( s(t) \) is the independent component, \( x(t) \) is the original EEG signal, and \( W \) is the unmixing matrix.

2. **Negative Correlation Coefficient Calculation Formula**
\[
p(n, m)=\frac{\sum_{i = -\infty}^{\infty} EOG_m(i)S_{n,m}(i-\tau)}{\sqrt{\sum_{i = -\infty}^{\infty} EOG_m^2(i)}\sqrt{\sum_{i = -\infty}^{\infty} S_{n,m}^2(i-\tau)}}
\]
where \( EOG_m(i) \) is the signal of the EOG channel, and \( S_{n,m}(i) \) is the signal of the \( n \) - th independent component in the \( m \) - th experiment.