A novel local senary pattern based epilepsy diagnosis system using EEG signals

Area of Science:

• Neurology
• Biomedical Engineering
• Signal Processing

Background:

• Epilepsy is a prevalent neurological disorder requiring accurate diagnosis.
• Electroencephalogram (EEG) signals are crucial for epilepsy diagnosis.
• Effective feature extraction from EEG signals is essential for accurate diagnosis.

Purpose of the Study:

• To propose a novel texture descriptor for distinctive EEG feature extraction.
• To develop an accurate EEG recognition method for epilepsy diagnosis.
• To evaluate the performance of the proposed method using a standard EEG dataset.

Main Methods:

• A novel Local Senary Pattern (LSP) texture descriptor was developed for feature extraction from 5x5 EEG blocks.
• Features were extracted using a ternary function with 10 threshold values derived from standard deviation, resulting in 15,360 features.
• Feature concatenation, Neighborhood Component Analysis (NCA) for reduction, and classifiers (SVM, KNN, QDA, LDA) were employed.
• The Bonn University EEG database was used for validation, with 7 cases defined for testing.

Main Results:

• The proposed LSP method achieved 93.0% classification accuracy for the 5-class case using the Bonn University EEG database.
• Extensive feature extraction using LSP and thresholding generated a large feature set.
• Neighborhood Component Analysis effectively reduced the high-dimensional feature space for classification.

Conclusions:

• The novel Local Senary Pattern (LSP) based method demonstrates significant success in EEG-based epilepsy diagnosis.
• The proposed feature extraction technique is effective in capturing distinctive patterns in EEG signals.
• The method offers a promising approach for improving the accuracy of automated epilepsy detection systems.