Comparison of countershock prediction features based on autoregressive and fourier transformed spectral analysis

Abstract

Spectral analysis of ventricular fibrillation (VF) ECG has been used for predicting countershock success, where the Fast Fourier Transformation (FFT) is the standard spectral estimator. Autoregressive (AR) spectral estimation should compute the spectrum with less computation time. This study compares the predictive power and computational performance of features based on FFT and AR methods. In an animal model of VF, 41 shocks were delivered in 25 swine. Two new AR based prediction features were developed in this study. For a proof of concept a microcontroller program was implemented. Calculating the area under the receiver operating characteristic (ROC) curve (AUC), the results of the features using AR modeling (85 %; 89 %) are better than common parameters based on FFT (72 %; 78 %). In the calculation time comparison the AR based parameters yield better results (nearly 2.5 times faster) than FFT based parameters. Summing up, AR spectral estimators are an attractive option compared to FFT due to the computational speed and the better prediction outcome.