Sequential graph convolutional network and DeepRNN based hybrid framework for epileptic seizure detection from EEG...

Area of Science:

• Health Informatics
• Computational Neuroscience
• Biomedical Engineering

Background:

• Epilepsy is a serious brain condition characterized by recurrent seizures due to abnormal brain electrical activity.
• Electroencephalogram (EEG) signals are crucial for seizure identification, but their low amplitude and non-stationary nature pose challenges for traditional deep learning models.
• Existing deep learning (DL) models struggle with the irregular structures of physiological recordings, hindering consistent diagnostic outcomes.

Purpose of the Study:

• To develop a novel hybrid deep learning framework for accurate and reliable automated epileptic seizure detection.
• To address the limitations of existing DL models in handling the complex nature of EEG signals.
• To improve diagnostic consistency and outcomes in epilepsy management.

Main Methods:

• A novel hybrid framework combining a sequential graph convolutional network (SGCN) and a deep recurrent neural network (DeepRNN) was proposed.
• DeepRNN was developed by fusing a gated recurrent unit (GRU) with a traditional RNN to overcome the vanishing gradient problem.
• Features such as line length, auto-covariance, auto-correlation, and periodogram were extracted from raw EEG signals and processed in the time-frequency domain for SGCN input.

Main Results:

• The proposed SGCN-DeepRNN model achieved an accuracy of 99.007% on the CHB-MIT and TUH datasets.
• The model demonstrated high sensitivity and specificity in epileptic seizure detection.
• The framework effectively extracts both spatial and temporal information from EEG signals, leading to improved performance.

Conclusions:

• The SGCN-DeepRNN hybrid framework represents a significant advancement in automated epileptic seizure detection.
• This model outperforms existing deep learning approaches, offering a more accurate and reliable diagnostic tool.
• The developed framework holds promise for enhancing epilepsy diagnosis and patient care.