Seizure Detection of EEG Signals Based on Multi-Channel Long- and Short-Term Memory-Like Spiking Neural Model.

Area of Science:

• Neurology
• Biomedical Engineering
• Signal Processing

Background:

• Seizures are a common neurological disorder with significant health impacts.
• Early detection and diagnosis of seizures are crucial for effective patient management.
• Existing methods for seizure detection require improvement in efficiency and accuracy.

Purpose of the Study:

• To develop a novel and efficient seizure detection method.
• To enhance the accuracy of early seizure diagnosis.
• To leverage advanced signal processing and neural network techniques for epilepsy detection.

Main Methods:

• Discrete Wavelet Transform (DWT) for signal decomposition into 5 levels.
• Extraction of time-frequency features from wavelet coefficients.
• Training a multi-channel Long- and Short-Term Memory-like Spiking Neural Network (LSTM-SNP) model for seizure detection.

Main Results:

• High seizure detection accuracy of 98.25% achieved on the CHB-MIT dataset.
• Specificity of 98.22% and sensitivity of 97.59% demonstrate robust performance.
• The proposed method shows competitive detection capabilities for epilepsy.

Conclusions:

• The DWT and LSTM-SNP based method offers a promising approach for accurate and efficient seizure detection.
• This technique can significantly aid in the early diagnosis and management of epilepsy.
• The study highlights the potential of integrating advanced signal processing with neural networks in neurological disorder detection.