Automatic detection and classification of artifacts in single-channel EEG
Summary
This study introduces an automated algorithm to detect and classify artifacts in ambulatory electroencephalography (EEG) recordings. The developed system achieves high accuracy, improving the reliability of mobile EEG diagnostics.
Area of Science:
- Biomedical Engineering
- Signal Processing
- Neuroscience
Background:
- Ambulatory electroencephalography (EEG) offers valuable diagnostic insights without hospitalization.
- Ambulatory EEG signals are susceptible to noise and artifacts, compromising data quality.
- Automated artifact management is crucial for reliable ambulatory EEG analysis.
Purpose of the Study:
- To develop and validate an automatic algorithm for artifact detection and classification in single-channel ambulatory EEG.
- To enhance the utility of ambulatory EEG monitoring by improving signal quality.
- To investigate both subject-specific and generic model implementations for artifact handling.
Main Methods:
- Feature extraction from EEG signals and wavelet subbands.
- Application of a feature selection algorithm to identify discriminating features.
- Utilizing a non-linear support vector machine for artifact classification.
- Validation using single-channel (Fp1-F7) EEG recordings from healthy subjects and a reference dataset.
Main Results:
- The artifact detection algorithm achieved average sensitivity and specificity exceeding 95% for both subject-specific and generic models.
- The classification algorithm demonstrated mean accuracies of 78% for subject-specific models and 64% for generic models.
- The classification model's performance was confirmed on an independent reference dataset.
Conclusions:
- The proposed algorithm effectively detects and classifies artifacts in ambulatory EEG signals.
- Both subject-specific and generic implementations show promising results, enhancing the reliability of ambulatory EEG.
- This automated approach supports more accurate diagnostic information from mobile EEG monitoring.
