Experimental study on the effects of physical training on the defibrillation threshold

Abstract

Background: Chest compression artifacts during cardiopulmonary resuscitation (CPR) deteriorate the rhythm diagnosis of automated external defibrillators (AED). Cardiopulmonary resuscitation must therefore be interrupted for a reliable shock/no-shock decision. However, these hands-off intervals adversely affect the defibrillation success, and, in addition, pauses in chest compressions compromise circulation. An accurate diagnosis of the rhythmwhile performing CPR is therefore needed to minimize these hands-off intervals. Methods: The characteristics of the CPR artifact are very variable, and the artifact presents an important spectral overlap with human cardiac arrest rhythms. Consequently, elaborate adaptive signal processing techniques are needed to filter the CPR artifact and reconstruct the underlying artifact-free ECG. Following an additive noise model, several methods were initially tested by artificially mixing human ventricular fibrillation samples and CPR artifacts recorded from pigs in asystole. The performance of the filters was evaluated for different levels of corruption, that is, signal-to-noise ratio. These studies showed satisfactory results after the filters were applied, both in terms of the improved signal-to-noise ratio and the sensitivity of the AED, which exceeded the 90% performance goal set by the American Heart Association (AHA). A posterior study reported similar results using a mixture of human ventricular fibrillation and human CPR artifacts extracted from real out-of-hospital interventions. Results: The first filtering method tested on shockable and nonshockable cardiac arrest rhythms from of out-of-hospital interventions was published in 2004. The artifact was estimated using up to 4 additional reference channels, acquired using a modified version of a commercial AED. The sensitivity after filtering (96.7%) exceeded AHA goals; the method, however, failed to accurately identify nonshockable rhythms because the specificity (79.9%) fell below the 95% AHA performance goal. Later, efforts focused on simplifying the filtering methods by either analyzing the ECG alone or using only 1 reference channel (2-channel methods). The ECG alone is not sufficient to estimate the artifact; however, comparable results were reported for multichannel and 2-channel methods on out-of-hospital rhythms. The possibility of identifying the rhythm by directly analyzing the corrupted ECG has also been proposed, although the reported specificity is also unsatisfactory. Conclusion: Currently, a reliable rhythm analysis during CPR is not possible because the specificity is too low. Two-channel filtering methods combined with rhythm identification applied to the filtered ECG should be further investigated to improve the detection of nonshockable rhythms.