Distinguishing Epileptiform Discharges From Normal Electroencephalograms Using Scale-Dependent Lyapunov Exponent

Area of Science:

• Neuroscience
• Computational Neuroscience
• Biomedical Engineering

Background:

• Epileptiform discharges are crucial for understanding epilepsy physiology.
• Accurate identification of epileptiform discharges in electroencephalogram (EEG) is vital for clinical epilepsy management.
• Existing automated methods often struggle with noise and lack explainability.

Purpose of the Study:

• To develop an explainable, automated computer program for distinguishing epileptiform discharges from normal EEG.
• To introduce and validate a multiscale complexity measure, the scale-dependent Lyapunov exponent (SDLE), for EEG analysis.
• To achieve high accuracy in classifying EEG segments with potential for clinical application.

Main Methods:

• Analysis of 640 multi-channel EEG segments (540 epileptiform discharges, 100 controls).
• Application of the scale-dependent Lyapunov exponent (SDLE) as a feature extraction method.
• Classification using Random Forest Classifier (RF) and Support Vector Machines (SVM).

Main Results:

• Features derived from SDLE effectively distinguished epileptiform discharges from normal EEG.
• The proposed approach achieved a robust accuracy exceeding 99% using RF and SVM.
• A novel parameter quantifying EEG signal regularity (ratio of spectral energy to SDLE) was introduced and found to be significantly higher in epileptiform discharges.

Conclusions:

• The SDLE-based approach offers a highly accurate and explainable method for identifying epileptiform discharges.
• The introduced regularity parameter provides insight into the classification accuracy.
• This method shows significant potential for widespread clinical adoption in epilepsy diagnosis and management.