EEG signal analysis using classification techniques: Logistic regression, artificial neural networks, support vector...

Area of Science:

• Neurology
• Biomedical Engineering
• Signal Processing

Background:

• Epilepsy is a neurological disorder characterized by seizures, often diagnosed through qualitative analysis of electroencephalogram (EEG) signals.
• Current diagnostic methods rely on visual interpretation of EEG patterns, which can be subjective.
• Quantitative analysis of EEG signals offers a more objective approach to epilepsy diagnosis.

Purpose of the Study:

• To quantitatively analyze EEG signals for epilepsy detection using Fourier analysis.
• To compare the performance of various machine learning classification techniques for identifying epileptic patients.
• To determine the most effective classification method for EEG-based epilepsy diagnosis.

Main Methods:

• EEG data from epileptic and non-epileptic patients were analyzed using Fourier transform to extract frequency domain features.
• Classification models including logistic regression, support vector machines, artificial neural networks, and convolutional neural networks were trained and evaluated.
• Performance metrics were used to compare the effectiveness of each classification technique.

Main Results:

• Fourier analysis enabled quantitative identification of patterns distinguishing epileptic from non-epileptic individuals.
• Artificial neural networks demonstrated the highest classification accuracy at 86%.
• The study successfully extracted distinguishing frequency band features from EEG signals.

Conclusions:

• Quantitative analysis of EEG signals via Fourier transform is a viable method for epilepsy detection.
• Artificial neural networks are the most effective classification technique for characterizing epileptic patients based on EEG frequency data.
• This approach enhances objective diagnosis and characterization of epilepsy.