Deep learning for robust detection of interictal epileptiform discharges

Area of Science:

• Neuroscience
• Artificial Intelligence
• Medical Technology

Background:

• Accurate detection of interictal epileptiform discharges (IEDs) is crucial for epilepsy diagnosis and seizure prediction.
• Existing methods for IED detection may lack speed, reliability, or robustness.

Purpose of the Study:

• To develop and validate a novel deep learning approach for automated IED detection from intracranial EEG (iEEG).
• To enhance the performance of deep learning models using data augmentation techniques.

Main Methods:

• Developed IEDnet, a deep learning model utilizing a long short-term memory network and an auxiliary classifier generative adversarial network (AC-GAN).
• Trained IEDnet on expert-annotated and AC-GAN augmented iEEG data from epilepsy patients.
• Compared IEDnet performance against Support Vector Machine (SVM) and Random Forest (RF) classifiers.

Main Results:

• IEDnet demonstrated superior sensitivity and specificity in detecting IEDs compared to SVM and RF.
• AC-GAN data augmentation significantly improved IEDnet's detection performance.
• IEDnet exhibited robustness to variations in sampling frequency and noise, and showed cross-institutional generalization.

Conclusions:

• IEDnet offers a highly effective and robust method for automated interictal spike detection in iEEG.
• AC-GAN augmentation is a valuable technique for enhancing supervised deep learning models in neuroscience applications.