Evaluating the Window Size's Role in Automatic EEG Epilepsy Detection.

Area of Science:

• Neuroscience
• Computational Biology
• Medical Informatics

Background:

• Electroencephalography (EEG) is crucial for brain condition assessment and epilepsy diagnosis.
• Analyzing EEG signals, especially epilepsy-specific waveforms, is complex for human interpretation due to signal variability and random occurrences.
• Automated EEG analysis offers significant potential for improving epilepsy diagnosis accuracy.

Purpose of the Study:

• To investigate the influence of varying window sizes on the classification accuracy of EEG signals for epilepsy detection.
• To compare the performance of different machine learning classifiers and training algorithms in analyzing EEG data.

Main Methods:

• Utilized the University of Bonn dataset comprising EEG data.
• Processed EEG data into epochs with 50% overlap and window lengths from 1 to 24 seconds.
• Extracted statistical and spectral features to train four machine learning classifiers: a neural network (with Broyden-Fletcher-Goldfarb-Shanno, multistart, and genetic algorithm training) and k-nearest neighbours.

Main Results:

• Classification accuracy was positively impacted by larger window sizes, particularly around 21 seconds.
• The study compared the effectiveness of different window lengths across multiple machine learning models.
• Optimal window size selection is critical for accurate automated EEG analysis.

Conclusions:

• Larger window sizes (approx. 21s) enhance EEG signal classification accuracy for epilepsy detection.
• Machine learning models, when trained on appropriately sized EEG windows, can significantly aid in epilepsy diagnosis.
• Automated analysis of EEG data shows promise for improving diagnostic efficiency and accuracy.