Segmentation and classification of EEG during epileptic seizures

Area of Science:

• Neuroscience
• Biomedical Engineering
• Signal Processing

Background:

• Epileptic seizures manifest as complex patterns in electroencephalogram (EEG) recordings.
• Accurate comparison and grouping of seizures are crucial for understanding epilepsy and developing effective treatments.
• Current methods may not fully capture the comprehensive morphological similarities between seizures.

Purpose of the Study:

• To develop an automated method for comparing epileptic seizures based on their overall patterns.
• To enable the grouping of seizures with similar morphologies from EEG data.
• To quantify the resemblance between two seizures, considering all channels.

Main Methods:

• EEG data segmentation into stationary characteristic segments.
• Feature extraction from each EEG segment.
• Clustering of segments across all channels to identify similar morphologies.
• Application of string matching algorithms to calculate seizure edit distance.

Main Results:

• Successful grouping of EEG segments based on morphology.
• Development of an edit distance metric to quantify seizure similarity.
• Demonstration of the method's ability to group seizures with similar morphology in 5 patients.
• Validation using both intracerebral and scalp EEG electrodes.

Conclusions:

• The proposed method effectively automates the comparison and grouping of epileptic seizures.
• The technique provides a quantitative measure of seizure similarity, aiding in clinical analysis.
• This approach holds promise for advancing the understanding and classification of epileptic seizure types.