Automated and Expert-Level Identification of Interictal Epileptiform Discharges with AI-Powered methods.

Area of Science:

• Neuroscience
• Artificial Intelligence
• Biomedical Engineering

Background:

• Focal epilepsy diagnosis relies on manual identification of interictal epileptiform discharges (IEDs) from EEG, a process that is time-consuming and prone to variability.
• Automated detection of IEDs is crucial for improving epilepsy care, but current methods face challenges in accuracy and efficiency.

Purpose of the Study:

• To develop and evaluate a novel deep learning approach for accurate and efficient detection of interictal epileptiform discharges (IEDs) in focal epilepsy.
• To integrate feature extraction, dataset balancing, latent space compression, and feature pruning for an optimized classification model.

Main Methods:

• A deep learning model combining a Convolutional Autoencoder (CAE) for feature extraction and a Kolmogorov-Arnold Network (KAN) for classification was developed.
• The methodology included candidate selection for dataset balancing, latent space compression for feature optimization, and feature pruning for enhanced interpretability.
• The model's performance was evaluated based on precision, sensitivity, and accuracy, with a focus on the impact of feature pruning.

Main Results:

• The proposed deep learning model achieved 100% precision, 71% sensitivity, and 83.78% accuracy in classifying IEDs.
• Feature pruning reduced the model's complexity to 23 inputs while maintaining a high accuracy of 87%.
• The KAN-based approach demonstrated superior classification efficiency with fewer parameters compared to conventional deep learning models.

Conclusions:

• The developed deep learning approach offers a scalable and clinically viable solution for automated EEG monitoring and seizure detection in epilepsy care.
• Feature pruning enhances model interpretability and efficiency, making it suitable for real-time applications and wearable seizure detection devices.
• This novel method has the potential to improve early diagnosis, treatment planning, and patient outcomes in epilepsy management.