Deep Multi-View Feature Learning for EEG-Based Epileptic Seizure Detection.

Area of Science:

• Neurology
• Biomedical Engineering
• Artificial Intelligence

Background:

• Epilepsy is a neurological disorder characterized by abnormal brain neuron discharge, leading to seizures that can be life-threatening.
• Electroencephalogram (EEG) signal analysis is crucial for monitoring epilepsy patients and enabling timely seizure detection and intervention.
• Effective feature identification in EEG signals is paramount for accurate epilepsy seizure detection.

Purpose of the Study:

• To propose a multi-view deep feature extraction method for improved epilepsy seizure detection from EEG signals.
• To enhance the seizure identification ability of EEG features through deep learning.
• To develop an interpretable and generalizable classification model for epilepsy seizure detection.

Main Methods:

• Utilized Fast Fourier Transform (FFT) and Wavelet Packet Decomposition (WPD) to create initial multi-view features from EEG signals.
• Employed Convolutional Neural Networks (CNN) for automatic deep feature learning, reducing dimensionality and improving seizure identification capabilities.
• Integrated a Multi-View Takagi-Sugeno-Kang Fuzzy System (MV-TSK-FS) for constructing an interpretable, rule-based classification model.

Main Results:

• The proposed multi-view deep feature extraction method achieved at least 1% higher classification accuracy than conventional methods (PCA, FFT, WPD).
• The method demonstrated at least 4% higher accuracy compared to single-view deep feature approaches.
• The MV-TSK-FS classifier, built on deep multi-view features, exhibited strong generalizability.

Conclusions:

• The multi-view deep feature extraction method offers superior performance for epilepsy seizure detection using EEG signals.
• Deep learning combined with multi-view fuzzy systems provides an effective strategy for enhancing diagnostic accuracy in epilepsy.
• This approach holds promise for improving patient monitoring and clinical intervention in epilepsy management.