Spectral characterization of ECG in out-of-hospital cardiac arrest patients

This study investigates ECG and demographic data from patients suffering from out-of-hospital cardiac arrest. We analyse the ability of 4 spectral features to discriminate between pre-shock ECG segments corresponding to successful versus unsuccessful shocks. We achieve poorer performance compared to what has been reported in a previous study on 2 of the features. We demonstrate improved performance by combining, decorrelating and reducing the number of features. This study supports the hypothesis that ECG from out-of-hospital patients contains information predictive of the outcome of shock therapy. Further, it demonstrates how pattern recognition methodology readily applies to this important problem