Frequency analysis of heart rate variability may predict seizure in the early post-cardiac arrest phase

Abstract

Introduction: Seizure is frequently encountered in the post-cardiac arrest phase and is an important predictor of poor prognosis. Early detection of seizure with in-time effective treatment is clinically challenging, especially in this era of target temperature management. Heart rate variability (HRV) signals has been shown the potentials to predict epilepsy. However, few focused on seizure prediction by continuous electroencephalography (EEG) and electrocardiogram (ECG) recordings. Objectives/Aims: To verify the roles of HRV in predicting seizure episodes in the early post-cardiac arrest phase. Methods: This is a prospective cohort study conducted in the Emergency Intensive Care Unit of National Taiwan University Hospital from January 2018 to December 2018. Adult non-traumatic cardiac arrest patients receiving CPR with sustained return of spontaneous circulation were recruited. ECG and EEG were continuously recorded (IntelliVue MX800®, Philips) in the first 72 h post-cardiac arrest. EEG signals around seizure episodes were decomposed by empirical mode decomposition (EMD). Approximate entropy (ApEn) of the intrinsic mode function (IMF) 4 was calculated. HRV was processed from ECG recorded 5 minutes before onset of candidate seizure using Pan-Tompkin’s algorithm, Fourier transformation and sample entropy. A set algorithm including support vector machine (SVM) classification and leave-one-out cross validation was launched to clarify the association of HRV and subsequent seizures. Results: A total of 27 candidate events from 15 patients (male: 7, mean age 67.4 years old) were analysed. Nine (33%) were confirmed seizure attacks. The results showed that power of high frequency (pHF), low-to-high frequency ratio (LF/HF), and sample entropy were the best parameters to match the prediction model, with sensitivity of 66.7% and specificity of 83.3%. The overall prediction accuracy was 77.8%. Conclusion: The frequency parameters of HRV may predict seizures in the early post-cardiac arrest phase. Further validation is needed to clarify its roles in clinical application.