Measuring the dissimilarity between EEG recordings through a non-linear dynamical system approach

Area of Science:

• Neuroscience
• Biomedical Engineering
• Signal Processing

Background:

• Electroencephalogram (EEG) analysis is crucial for understanding brain activity.
• Accurate classification of EEG segments is essential for diagnosing neurological conditions.
• Existing methods for EEG dissimilarity measurement may have limitations.

Purpose of the Study:

• To introduce a new mathematical measure for quantifying dissimilarity between EEG segments.
• To evaluate the effectiveness of this novel measure in the automatic classification of EEG data.
• To assess the measure's performance on specific EEG activities like delta and alpha waves.

Main Methods:

• Developed a novel dissimilarity measure based on non-linear autoregressive functions and kernel estimators.
• Employed non-parametric regression for non-linear autoregressive estimation.
• Utilized multidimensional scaling and cluster analysis for EEG segment classification based on the new measure.

Main Results:

• The proposed dissimilarity measure demonstrated effectiveness in distinguishing between different EEG segments.
• Automatic classification using this measure showed high agreement with classifications made by human EEG specialists.
• The measure performed particularly well on EEG segments exhibiting delta and alpha wave activity.

Conclusions:

• The new EEG dissimilarity measure offers a promising approach for automated EEG analysis.
• This method has the potential to improve the accuracy and efficiency of EEG-based diagnostics.
• The findings support the utility of advanced mathematical techniques in neurophysiological signal processing.