Epileptic seizure detection in EEGs signals using a fast weighted horizontal visibility algorithm.

Area of Science:

• Neuroscience
• Complex Systems
• Signal Processing

Background:

• Epileptic seizures are neurological disorders requiring accurate detection from electroencephalogram (EEG) signals.
• Traditional methods like Fast Fourier Transform (FFT) and sample entropy (SampEn) have limitations in speed and accuracy for seizure identification.
• Horizontal Visibility Graphs (HVGs) offer a novel approach to time series analysis by converting data into network structures.

Purpose of the Study:

• To propose and evaluate a fast weighted horizontal visibility graph algorithm (FWHVA) for efficient seizure detection in EEG signals.
• To compare the performance of FWHVA with established methods such as FFT and SampEn.
• To investigate the noise robustness and classification capabilities of graph features derived from FWHVA.

Main Methods:

• Development of a fast weighted horizontal visibility graph constructing algorithm (FWHVA).
• Application of FWHVA to EEG signals for time series analysis and feature extraction.
• Comparison of FWHVA performance against Fast Fourier Transform (FFT) and sample entropy (SampEn) methods.
• Investigation of noise-robustness using chaos signals and analysis of graph features (mean degree, mean strength) on EEG data.

Main Results:

• The FWHVA demonstrates significantly faster feature extraction compared to SampEn and FFT.
• The mean strength feature derived from FWHVA shows a significant increase in ictal EEG compared to healthy and inter-ictal states.
• Classification accuracy of 100% was achieved in distinguishing seizure from healthy EEG signals using FWHVA-based features.

Conclusions:

• The FWHVA is a computationally efficient and effective algorithm for seizure detection from EEG signals.
• Graph features derived from FWHVA, particularly mean strength, show high potential for identifying epileptic seizures.
• FWHVA-based methods outperform FFT and SampEn in terms of speed and classification accuracy for time series analysis in seizure detection.