A Deep Learning Approach for Automatic Seizure Detection in Children With Epilepsy.

Area of Science:

• Neurology
• Artificial Intelligence
• Biomedical Engineering

Background:

• Electroencephalogram (EEG) is crucial for diagnosing neurological disorders, especially epilepsy.
• Accurate seizure detection is essential due to the significant impact of epilepsy on patients' quality of life.
• Existing methods require extensive pre-processing, motivating the development of more efficient approaches.

Purpose of the Study:

• To propose a novel deep learning approach for detecting seizures in pediatric patients using minimally pre-processed EEG signals.
• To leverage automatic feature learning for improved classification accuracy between ictal and interictal states.
• To develop a unified system combining a 2D deep convolution autoencoder (2D-DCAE) and a neural network classifier.

Main Methods:

• A supervised deep convolutional autoencoder (SDCAE) model was developed, integrating a bidirectional long short-term memory (Bi-LSTM) classifier.
• Two models were evaluated using three EEG data segment lengths and a 10-fold cross-validation scheme.
• Performance was assessed using five standard evaluation metrics on a public pediatric EEG dataset.

Main Results:

• The best performing model, SDCAE with Bi-LSTM classifier and 4s EEG segments, achieved high accuracy (98.79%).
• The model demonstrated excellent performance across sensitivity (98.72%), specificity (98.86%), precision (98.86%), and F1-score (98.79%).
• Results were validated using the Children's Hospital Boston (CHB) and MIT dataset.

Conclusions:

• The proposed deep learning approach offers a highly effective method for pediatric seizure detection.
• The SDCAE with Bi-LSTM model significantly outperforms existing state-of-the-art methods on the same dataset.
• This method provides a promising tool for precise and efficient epilepsy diagnosis in children.