Filtering the cardiopulmonary resuscitation artifact: Influence of the signal-to-noise-ratio on the accuracy of the shock advice algorithm

A reliable diagnosis by automated external defibrillators (AED) during cardiopulmonary resuscitation (CPR) would reduce hands-off time, thus increasing the resuscitation success. Several filtering techniques have been proposed to remove the artifact induced on the ECG by chest compressions. The improvement in the signal-to-noise ratio (SNR) has been widely used to test the performance of the filter, using artificial mixtures of ECG signals and CPR artifacts. In this work, we analyzed the influence of the SNR, estimated from corrupted out-of-hospital cardiac arrest episodes, on the AED diagnostic accuracy before and after artifact removal. Filtering improved the sensitivity for records with low SNR, however the specificity was largely independent of the SNR. Moreover, the total specificity decreased after filtering due to misclassified asystole records.