A robust seizure detection and prediction method with feature selection and spatio-temporal casual neural network...

Area of Science:

• Neurology
• Biomedical Engineering
• Artificial Intelligence in Medicine

Background:

• Epilepsy is a common neurological disorder characterized by recurrent seizures, posing significant safety risks and impacting patient quality of life.
• Real-time electroencephalogram (EEG) diagnosis is crucial for epilepsy management, but conventional methods suffer from high computational costs and low portability due to extensive feature requirements.
• There is a need for efficient, lightweight, and robust algorithms for seizure detection and prediction in epilepsy patients.

Purpose of the Study:

• To develop an efficient, lightweight, and robust algorithm for seizure detection and prediction in epilepsy patients.
• To address the limitations of conventional methods by reducing computational cost and improving portability.
• To enhance the accuracy and sensitivity of real-time EEG-based epilepsy diagnosis.

Main Methods:

• The proposed algorithm utilizes an interpretative feature selection method to reduce model complexity and training difficulties.
• A spatial-temporal causal neural network (STCNN) is employed to capture both spatial and temporal information for dynamic feature tracking and diagnosis.
• Experiments were conducted using leave-one-out cross-validation (LOOCV) and cross-patient validation (CPV) on multiple datasets (CHB-MIT, Siena, Kaggle).

Main Results:

• The algorithm demonstrated improved detection accuracy and prediction sensitivity in the LOOCV-based approach.
• Significant improvements were also observed in the CPV-based method, indicating robustness across different patient groups.
• The STCNN model effectively tracked and diagnosed changing features by integrating spatial-temporal data.

Conclusions:

• The developed algorithm shows superior performance and robustness in seizure detection and prediction.
• This approach offers enhanced capability for managing diverse and complex clinical epilepsy scenarios.
• The findings suggest a promising advancement in real-time EEG analysis for epilepsy care.