A machine learning approach to epileptic seizure prediction using Electroencephalogram (EEG) Signal

Area of Science:

• Neuroscience
• Medical Technology

Background:

• Epileptic seizures pose diagnostic challenges.
• Accurate detection of seizures from brain electrical activity is crucial for patient care.

Purpose of the Study:

• To investigate brain electrical activity properties for improved seizure detection.
• To explore optimal methods for identifying patterns in epileptic Electroencephalogram (EEG) signals.

Main Methods:

• Utilized 100 single-channel EEG recordings from healthy volunteers and epilepsy patients.
• Applied feature extraction (Butterworth, Fourier, Wavelet Transforms) and selection (T-test, SFFS).
• Classified preprocessed EEG signals using Support Vector Machine (SVM) and K-Nearest Neighbors (KNN) algorithms.

Main Results:

• Evaluated classification performance using Accuracy, Sensitivity, and Specificity.
• Demonstrated that SVM achieved slightly superior performance compared to KNN.
• Identified effective feature engineering techniques for EEG signal analysis.

Conclusions:

• The study provides a robust framework for seizure detection using EEG analysis.
• SVM algorithm shows promise for accurate and timely identification of epileptic seizures.
• Findings are valuable for neurophysiologists in clinical practice.