Epileptic State Prediction using Phase Space Domain and Machine Learning Algorithms.

Area of Science:

• Neuroscience
• Biomedical Engineering
• Data Science

Background:

• Epilepsy affects millions globally, necessitating advanced diagnostic tools.
• Current epilepsy diagnosis relies on neuroimaging and electroencephalogram (EEG) analysis, which can be complex.
• There is a need for predictive systems to improve the detection of epileptic states using novel features.

Purpose of the Study:

• To investigate the prediction and detection of epileptic states.
• To explore the efficacy of transforming 2-dimensional EEG time series data into the phase space domain for feature extraction.
• To develop a predictive model for epilepsy detection.

Main Methods:

• EEG time series data was transformed into the phase space domain.
• Angular distance and probability density functions were computed to extract phase space features.
• Renyi and Tsallis complex features were extracted and used to train various machine learning models.
• Leave-one-subject-out cross-validation was employed to evaluate model performance.

Main Results:

• Probabilistic models trained on phase domain complex features achieved 91.5% accuracy.
• This accuracy surpassed other evaluated learning algorithms.
• The phase space domain proved effective for extracting discriminating features.

Conclusions:

• The phase space domain offers a promising approach for epilepsy detection and prediction.
• The developed method demonstrates high accuracy in identifying epileptic states.
• This technique holds potential for improving diagnostic capabilities in epilepsy management.