Abstract 267: Photoplethysmography-Based Algorithm Uninterruptedly Supports Pulse Detection in Porcine Automatic 30:2 Cardiopulmonary Resuscitation Data by Handling Transitions Between Compressions and Pauses

• Published in: Circulation (New York, N.Y.) Vol. 138; no. Suppl_2
• Main Authors: Wijshoff, Ralph, Van de Laar, Jakob, Muehlsteff, Jens, Noordergraaf, Gerrit Jan
• Format: Journal Article
• Language: English
• Published: 06.11.2018
• ISSN: 0009-7322; 1524-4539
• DOI: 10.1161/circ.138.suppl_2.267

Abstract only Introduction: Pulse checks by manual palpation during CPR are challenging, time-consuming and interrupt chest compressions. To support pulse checks during 30:2 CPR, we developed an algorithm to uninterruptedly detect pulse in a photoplethysmography (PPG) signal by handling transitions between compressions and ventilation pauses. We retrospectively evaluated the algorithm on porcine automated-CPR (ACPR) data. Methods: In 10 anesthetised pigs, VF was induced, followed by 20 min of 30:2 ACPR with 2.5 s ventilation pauses (Phase 1). Next, up to four defibrillation attempts were made, each shock followed by 2 min of 30:2 ACPR (Phase 2). Infrared nasal PPG and thoracic impedance (TI) signals were measured, with aortic blood pressure (ABP) as reference. The developed algorithm detected pulse rate (PR) in the PPG signal, by estimating the main frequencies present in the signal and ignoring compression frequencies, where the compression rate (CR) was determined from the TI waveform. The PPG signal was analysed by sliding a 5 s window by 1 s shifts. When analysing a signal with an abrupt change due to the start/end of a compression series, frequency estimates will be compromised. Therefore, the algorithm identified the start/end of compressions in the TI signal and analysed in the 5 s window (1) a compression-free interval of at least 2 s or (2) a compression interval of at least 2.8 s. The window was ignored if no such interval was found. Pulse detection performance was quantified by comparison to pulse presence annotations determined from the protocol and visual inspection of ABP and PPG signals. Results: Table I gives sensitivity for pulse presence (SNV), specificity for arrest (SPC) and positive and negative predictive values (PPV/NPV). Conclusions: Overall, SPC, PPV and NPV are high. In Phase 2, SNV is reasonable in pauses, but compromised during compressions when PR and CR are close. In 30:2 CPR the algorithm can provide frequent detections from 2 s pauses. Table I: detection results.