A method to remove CPR artefacts from human ECG using only the recorded ECG

Abstract

To show the possibility of using cardiopulmonary resuscitation (CPR) artefact suppression methods that do not need additional reference signals to model CPR artefacts. A CPR suppression method based on a Kalman filter was designed. The artefact was modelled using the fundamental frequency of the compressions, estimated from the spectral analysis of the ECG. Artificial mixtures of human shockable rhythms and CPR artefacts were used to design the algorithm that was then tested on samples obtained from real out-of-hospital cardiac arrest episodes. The shock/no-shock decision of an automated external defibrillator (AED) was evaluated before and after CPR suppression for 131 shockable and 347 non-shockable samples. The sensitivity improved from 56% (95% CI, 47-64%) to 90% (95% CI, 84-94%). However, the specificity decreased from 91% (95% CI, 87-93%) to 80% (95% CI, 76-84%). CPR artefacts can be suppressed using methods based on the analysis of the ECG alone. The hardware of current AEDs does not need to be replaced, although better artefact suppression methods exist for modified AEDs with additional reference channels.