An effective approach to classify epileptic EEG signal using local neighbor gradient pattern transformation methods

Area of Science:

• Neurology
• Biomedical Engineering
• Machine Learning

Background:

• Electroencephalographic (EEG) signals are crucial for diagnosing epileptic seizures.
• Manual analysis of non-stationary EEG signals for epilepsy diagnosis is inefficient and prone to errors.
• Developing automated, accurate, and efficient methods for epilepsy detection from EEG is a significant challenge.

Purpose of the Study:

• To propose and evaluate novel feature extraction techniques, local neighbor gradient pattern (LNGP) and symmetrically weighted local neighbor gradient pattern (SWLNGP), for automatic epilepsy detection.
• To assess the performance of these techniques when integrated with various machine learning classifiers, particularly artificial neural networks (ANN).
• To compare the proposed methods against existing techniques for epileptic seizure detection.

Main Methods:

• Feature extraction using Local Neighbor Gradient Pattern (LNGP) and Symmetrically Weighted Local Neighbor Gradient Pattern (SWLNGP).
• Classification of EEG signals using k-nearest neighbor (k-NN), quadratic linear discriminant analysis, support vector machine, ensemble classifier, and artificial neural network (ANN).
• Evaluation using 10-fold cross-validation across 17 different classification problems, including healthy-ictal, interictal-ictal, and healthy-interictal-ictal scenarios.

Main Results:

• The proposed LNGP and SWLNGP methods achieved higher classification accuracy when combined with the ANN classifier across all tested cases.
• The methods demonstrated superior performance compared to recently proposed techniques for epileptic detection in terms of both computational efficiency and classification accuracy.
• The study confirmed the effectiveness of LNGP and SWLNGP with ANN for various epilepsy classification tasks.

Conclusions:

• The proposed LNGP and SWLNGP feature extraction methods, coupled with an ANN classifier, offer a simple, fast, and reliable approach for real-time epilepsy detection.
• These methods provide a significant improvement over manual inspection and existing automated techniques for EEG-based epilepsy diagnosis.
• The developed approach holds promise for practical clinical application in identifying epileptic seizure activities.