CNN-Informer: A hybrid deep learning model for seizure detection on long-term EEG

The CNN-Informer model shows strong generalization ability across two EEG datasets.

• Timely detection of epileptic seizures can significantly reduce injuries for patients.
• Deep learning models have successfully detected seizures but face challenges with computational complexity and overfitting.
• The CNN-Informer model utilizes a Convolutional Neural Network to extract local features from multi-channel EEG signals.
• The model effectively captures long-range dependencies while maintaining low computational complexity and memory usage.
• Raw EEG signals are filtered using Discrete Wavelet Transform to improve feature extraction and classification.
• Post-processing operations enhance the accuracy of detection results.

AI simplified