EMD-WOG-2DCNN based EEG signal processing for Rolandic seizure classification.

Area of Science:

• Neurology and Artificial Intelligence
• Medical Signal Processing

Background:

• Epilepsy affects 65 million worldwide, with accurate diagnosis crucial for treatment and prognosis.
• Childhood benign epilepsy with centrotemporal spikes (BECTS), or benign Rolandic epilepsy, is the most common childhood focal epilepsy.
• Current EEG interpretation relies on manual analysis by neurologists, a time-consuming process.

Purpose of the Study:

• To introduce the most common childhood benign epilepsy type (BECTS).
• To propose a novel machine learning model for accurate distinction of Rolandic EEG patterns from normal signals.
• To address the clinical need for efficient and accurate epilepsy diagnosis.

Main Methods:

• Developed a machine learning model utilizing empirical mode decomposition (EMD) for EEG feature extraction.
• Employed weighted overlook graph (WOG) to represent EMD-decomposed intrinsic mode functions (IMFs).
• Utilized a two-dimensional convolutional neural network (2DCNN) for classification of EEG signals.

Main Results:

• The proposed model achieved an accuracy exceeding 97.6% on the Rolandic epilepsy dataset.
• Performance was superior to other representative machine learning models.
• The model demonstrated comparable or superior performance on the Bonn public EEG dataset.

Conclusions:

• The developed machine learning model accurately distinguishes benign Rolandic epilepsy EEG patterns from normal signals.
• The model meets real clinical needs for efficient and precise epilepsy diagnosis.
• Future work aims to extend the model for classifying other epilepsy types and hospital system implementation.