Linear and non-linear feature extraction from rat electrocorticograms for seizure detection by support vector machine

Area of Science:

• Neuroscience
• Biomedical Engineering
• Signal Processing

Background:

• Epilepsy is a neurological disorder characterized by seizures, necessitating accurate detection for effective treatment.
• Electrocorticogram (ECoG) signals offer a direct measure of brain activity for seizure monitoring.
• Automated seizure detection systems are crucial for improving patient care and research.

Purpose of the Study:

• To develop and evaluate an automated system for detecting seizures using ECoG signals.
• To compare the effectiveness of linear and non-linear features in seizure detection.
• To assess the system's suitability for online implementation by minimizing computational time.

Main Methods:

• ECoG signals were recorded from animals before and after pentylenetetrazol injection.
• Signals were segmented and labeled as ictal or non-ictal, with 24 features extracted.
• A support vector machine (SVM) classifier was trained and tested using these features.

Main Results:

• A combination of linear and non-linear features yielded the highest accuracy of 95.3% for seizure detection.
• Some linear features individually or in combination outperformed non-linear features.
• The system demonstrated robustness by classifying signals without artifact removal, reducing computational load.

Conclusions:

• The developed SVM-based system effectively detects seizures from ECoG signals with high accuracy.
• Combining diverse signal features enhances detection performance.
• The system's efficiency and robustness support its potential for online seizure detection applications.