Classification of Epileptic EEG Signals Using Synchrosqueezing Transform and Machine Learning.

Area of Science:

• Neurology
• Signal Processing
• Machine Learning

Background:

• Epilepsy is a common neurological disorder affecting millions globally.
• Electroencephalography (EEG) signals are crucial for identifying epileptic seizure segments.
• Accurate seizure detection is vital for patient management and treatment.

Purpose of the Study:

• To develop and evaluate a high-resolution time-frequency (TF) representation, Synchrosqueezing Transform (SST), for enhanced epileptic seizure detection from EEG data.
• To compare the performance of the SST-based method against traditional approaches like the short-time Fourier transform (STFT).
• To assess the efficacy of various machine learning classifiers when applied to SST-derived EEG features.

Main Methods:

• Utilized Synchrosqueezing Transform (SST) for high-resolution time-frequency analysis of EEG signals.
• Extracted novel features, including higher-order joint TF (HOJ-TF) moments and gray-level co-occurrence matrix (GLCM)-based features, from SST representations.
• Employed diverse machine learning algorithms (kNN, LR, NB, SVM, BT, S-kNN) for classifying seizure and non-seizure EEG segments.

Main Results:

• The SST-based method demonstrated high performance on both the IKCU and CHB-MIT EEG datasets.
• Achieved excellent accuracy (ACC), precision (PRE), and recall (REC) rates, exceeding 95% in several metrics for the IKCU dataset.
• Showcased superior performance compared to STFT-based methods, with accuracy over 95% in most cases.

Conclusions:

• The proposed SST-based approach offers a robust and effective method for epileptic seizure detection using EEG signals.
• The novel TF features derived from SST significantly improve seizure detection accuracy.
• This method shows promise for clinical application and compares favorably with existing state-of-the-art techniques.