Filtering mechanical chest compression artefacts from out-of-hospital cardiac arrest data.

AIM

Filtering techniques to remove manual compression artefacts from the ECG have not been incorporated to defibrillators to diagnose the rhythm during cardiopulmonary resuscitation. Mechanical and manual compression artefacts may be very different. The aim of this study is to characterize the compression artefact caused by the LUCAS 2 device and to evaluate whether filtering the LUCAS 2 artefact results in an accurate rhythm analysis.

METHODS

A dataset of 1045 segments were obtained from 230 out-of-hospital cardiac arrest (OHCA) patients after LUCAS 2 activation. Rhythms were 201 shockable, 270 asystole and 574 organized. Segments during asystole were used to characterize the artefact in time and frequency domains. Three filtering methods, a comb filter and two adaptive filters, were used to remove the mechanical compression artefact. The filtered ECG was then diagnosed with a shock decision algorithm from a defibrillator.

RESULTS

When compared to the manual compression artefact, the LUCAS 2 artefact presented a similar amplitude (1.2 mV, p-value 0.26), fixed frequency (101.7 min(-1)), more harmonic components, smaller spectral dispersion, and a more regular waveform (p-val <3 × 10(-7)). The sensitivity (SE) and specificity (SP) before filtering the LUCAS 2 artefact were 52.8% (90% low CI, 46.0%) and 81.5% (79.0%), respectively. For the best filter, SE and SP after filtering were 97.9% (95.7%) and 84.1% (82.0%), respectively. Optimal filters require more harmonics and smaller bandwidths than for manual compressions.

CONCLUSION

Filtering resulted in a large increase in SE and small increase in SP. Despite differences in artefact characteristics between manual and mechanical compressions, filtering the LUCAS 2 compression artefact results in SE/SP values comparable to those obtained for manual compression artefacts. The SP is still below the 95% recommended by the American Heart Association.