Prediction of the outcome in cardiac arrest patients undergoing hypothermia using EEG wavelet entropy

Cardiac arrest (CA) is the leading cause of death in the United States. Induction of hypothermia has been found to improve the functional recovery of CA patients after resuscitation. However, there is no clear guideline for the clinicians yet to determine the prognosis of the CA when patients are treated with hypothermia. The present work aimed at the development of a prognostic marker for the CA patients undergoing hypothermia. A quantitative measure of the complexity of Electroencephalogram (EEG) signals, called wavelet sub-band entropy, was employed to predict the patients' outcomes. We hypothesized that the EEG signals of the patients who survived would demonstrate more complexity and consequently higher values of wavelet sub-band entropies. A dataset of 16-channel EEG signals collected from CA patients undergoing hypothermia at Long Beach Memorial Medical Center was used to test the hypothesis. Following preprocessing of the signals and implementation of the wavelet transform, the wavelet sub-band entropies were calculated for different frequency bands and EEG channels. Then the values of wavelet sub-band entropies were compared among two groups of patients: survived vs. non-survived. Our results revealed that the brain high frequency oscillations (between 64100 Hz) captured from the inferior frontal lobes are significantly more complex in the CA patients who survived (p-value <; 0.02). Given that the non-invasive measurement of EEG is part of the standard clinical assessment for CA patients, the results of this study can enhance the management of the CA patients treated with hypothermia.