Muscle artifact removal from human sleep EEG by using independent component analysis
Maite Crespo-Garcia1, Mercedes Atienza, Jose L Cantero
1Laboratory of Functional Neuroscience, University Pablo de Olavide, Ctra. de Utrera, Km. 1, 41013, Seville, Spain.
Annals of Biomedical Engineering
|January 30, 2008
Summary
This study found AMUSE, a fast algorithm, effectively removes muscle artifacts from sleep EEG recordings. It outperforms other methods, especially for EEG arousals, improving sleep data quality.
Area of Science:
- Neuroscience and Biomedical Engineering
- Sleep Medicine and Electroencephalography
Background:
- Muscle artifacts are common in sleep EEG, complicating analysis and diagnosis.
- Existing EEG artifact correction methods often overlook muscle activity, necessitating specialized techniques.
Purpose of the Study:
- To evaluate four Independent Component Analysis (ICA) algorithms for myogenic artifact removal from sleep EEG.
- To identify the optimal ICA method for minimizing muscle artifact contamination in sleep recordings.
Main Methods:
- Comparison of AMUSE, SOBI, Infomax, and JADE ICA algorithms.
- Assessment of artifact elimination efficacy, signal-to-noise ratio independence, and computational speed.
- Validation on a real case with sleep EEG arousals across different sleep stages.
Main Results:
- AMUSE, Infomax, and SOBI significantly outperformed JADE in removing temporal muscle artifacts.
- AMUSE demonstrated independence from signal-to-noise ratio in non-temporal regions and superior speed.
- AMUSE successfully separated muscle artifacts from EEG arousals in a clinical sleep recording.
Conclusions:
- AMUSE is a highly effective and computationally efficient ICA algorithm for removing muscle artifacts from human sleep EEG.
- The findings support AMUSE as a valuable tool for improving the accuracy of quantitative EEG analysis during sleep.
