Automatic seizure detection: going from sEEG to iEEG

Area of Science:

• Neuroscience
• Biomedical Engineering
• Signal Processing

Background:

• Automatic epileptic seizure detection algorithms exist for both scalp (sEEG) and intracranial (iEEG) electroencephalography.
• Assessing the optimal modality for seizure detection remains challenging.
• Existing algorithms developed for sEEG may not directly translate to iEEG without modification.

Purpose of the Study:

• To evaluate the performance improvement when adapting sEEG-based automatic seizure detection algorithms for iEEG data.
• To investigate the impact of feature extraction modifications on seizure detection accuracy in iEEG.

Main Methods:

• Collected over 24 hours of ictal and non-ictal iEEG data from 16 focal epilepsy patients.
• Utilized wavelet transformation (WT) for feature extraction and a support vector machine for classification.
• Adapted an sEEG algorithm for iEEG by including high-frequency WT features from lower levels.

Main Results:

• Achieved a high sensitivity of 96.4% for automatic seizure detection.
• Recorded a low false detection rate (FDR) of 0.20 per hour.
• Demonstrated significant performance enhancement by widening the frequency band in feature extraction for iEEG.

Conclusions:

• Algorithms designed for sEEG can be effectively adapted for iEEG with appropriate modifications.
• Frequency band widening in WT feature extraction is crucial for improving automatic seizure detection in iEEG.
• This study provides a method for optimizing existing sEEG algorithms for iEEG applications.