Detection of preictal state in epileptic seizures using ensemble classifier

Area of Science:

• Neuroscience
• Medical Technology
• Artificial Intelligence

Background:

• Epilepsy affects a significant patient population, with approximately 30% experiencing seizures resistant to conventional medication or surgery.
• Early prediction of seizures is crucial for timely therapeutic interventions in refractory epilepsy cases.

Purpose of the Study:

• To develop and validate a novel ensemble classifier for accurate detection of the preictal state from electroencephalogram (EEG) signals.
• To address the challenge of accurately identifying the onset of the preictal state, a critical precursor to seizures.

Main Methods:

• A novel ensemble classifier was developed, integrating deep learning and handcrafted features from EEG signals.
• The method utilizes a comprehensive feature set as input and combines three distinct classifiers for preictal state detection.
• Preprocessing techniques were employed for effective noise removal from the EEG data.

Main Results:

• The proposed method was evaluated on the publicly available CHBMIT scalp EEG dataset, comprising data from 22 subjects.
• Achieved an average accuracy of 94.31%.
• Demonstrated high performance with a sensitivity of 94.73% and specificity of 93.72%.

Conclusions:

• The developed ensemble classifier effectively detects the start of the preictal state by combining diverse feature sets and advanced classification techniques.
• The study highlights the method's superior performance in accuracy, sensitivity, and specificity compared to existing approaches.
• This approach offers a promising tool for improving seizure prediction and management in epilepsy patients.