EpilepIndex: A novel feature engineering tool to detect epilepsy using EEG signals

Area of Science:

• Neurology
• Machine Learning
• Biomedical Signal Processing

Background:

• Epilepsy is a common neurological disorder characterized by seizures, posing risks of fatal accidents due to potential loss of consciousness.
• Electroencephalogram (EEG) analysis for epilepsy detection is challenging due to the non-linear and complex nature of brain signals, making manual interpretation subjective and difficult.
• Automatic detection of epileptic seizures using machine learning offers a promising alternative to overcome the limitations of traditional visual EEG examination.

Purpose of the Study:

• To develop and evaluate a novel machine learning framework for the automatic classification of epileptic seizures from EEG data.
• To explore a comprehensive set of 22 clinically significant classification cases, combining binary and multi-class problems, for robust epilepsy detection.
• To introduce a new integrated epilepsy detection index (EpilepIndex) for improved differentiation between epileptic and normal EEG signals.

Main Methods:

• Utilized a standard EEG database comprising five distinct datasets, including epileptic EEG recordings.
• Extracted 11 statistically significant non-linear entropy features from the non-linear EEG signals.
• Tested 10 different machine learning classifiers across 22 classification cases using 10-fold cross-validation.

Main Results:

• Random Forest and Optimized Forest classifiers demonstrated classification accuracies exceeding 90% for all 22 evaluated cases.
• The proposed method achieved highly competitive accuracies compared to existing literature, addressing a wider range of classification scenarios.
• The novel EpilepIndex achieved 100% accuracy in distinguishing between epileptic and normal EEG signals.

Conclusions:

• The developed machine learning approach, particularly using Random Forest and Optimized Forest, shows significant potential for accurate and reliable automatic epilepsy detection from EEG.
• The comprehensive evaluation across 22 diverse classification cases represents a novel contribution to the field, offering a more thorough assessment of detection algorithms.
• The EpilepIndex provides a highly effective tool for epilepsy diagnosis, complementing existing methods and potentially improving patient outcomes.