Squeezing the Juice Out of the Rhythm Strip to Predict How Much is Needed to Defibrillate

© 2023 HMP Global. All Rights Reserved.

Any views and opinions expressed are those of the author(s) and/or participants and do not necessarily reflect the views, policy, or position of EP Lab Digest or HMP Global, their employees, and affiliates. 

EP LAB DIGEST. 2023;23(7):4.

Dear Readers,

When defibrillation testing was routinely performed at the time of implantable cardioverter-defibrillator (ICD) insertion, it became clear that the appearance of the rhythm-induced ventricular fibrillation (VF) subjectively predicted whether the ICD shock would be successful. When fine, low-amplitude VF was induced, a fear instinctively developed that the shock would fail. On the contrary, reassurance was felt when relatively slow, large-amplitude VF was induced, and in fact, the somewhat organized ventricular arrhythmias would terminate spontaneously at times.

Similarly, when a patient has an out-of-hospital cardiac arrest (OHCA), it seems that the prognosis is poor when the initial rhythm is low-amplitude, fine-appearing VF. It has also been well demonstrated that successful defibrillation in the field is highly time-dependent and that the appearance of the VF becomes much finer over time, before the development of what is often referred to as an agonal rhythm, before the culmination in asystole. It would make sense that careful analysis of the electrocardiogram (ECG) waveforms, when the patient is initially in VF, might be predictive and able to tailor resuscitation efforts.

Related to this topic, a study was recently published by Coult et al using artificial intelligence (AI) to predict shock-refractory VF during resuscitation of OHCA patients. They looked at the digital external defibrillator rhythm strips in 1376 patients who had an OHCA due to VF in Seattle—a city that has been on the cutting edge of resuscitation research for decades. Impressively, 43% of patients achieved functional neurological survival. The investigators processed 3-second segments of the ECG rhythm strip immediately before and 1 minute after the initial external shock during cardiopulmonary resuscitation (CPR) to predict a poor clinical outcome. They showed that the likelihood of a poor clinical outcome correlated very well with the number of shocks that the patient received and used failure of ≥3 shocks as a surrogate endpoint of refractory VF. They then developed an algorithm to predict which patients would require ≥3 shocks. The machine learning (ML) algorithm strongly predicted refractory VF with a specificity of 91%, sensitivity of 63%, and positive likelihood ratio of 6.7.

What did the ML algorithm identify as characteristics of refractory VF? Interested readers should refer to the complex ML algorithm that was developed that processed the electrograms into a scalogram, which was projected onto an Eigenscalogram library. The representative ECG examples in the publication showed that characteristics of the pre-shock ECG that predicted failure were the absence of low-frequency signals, with additional differences in the post-shock ECG showing a greater disparity of frequencies in the patients who were refractory to shocks.

There is justifiably significant enthusiasm for the application of AI and ML in cardiology and electrophysiology. This study by Coult et al showing that the development of algorithms to analyze the rhythm strips at the time of an OHCA, immediately before and after the first shock is delivered, demonstrates the power of using AI to quickly process this readily available information to predict which patients will be difficult to defibrillate. Of course, the clinical applicability of this information remains to be determined, but one could envision a more customized approach to patients who are likely going to be difficult to defibrillate. 

Disclosures: Dr Knight has served as a paid consultant to Medtronic and was an investigator in the PULSED AF trial. In addition, he has served as a consultant, speaker, investigator, and/or has received EP fellowship grant support from Abbott, AltaThera, AtriCure, Baylis Medical, Biosense Webster, Biotronik, Boston Scientific, CVRx, Philips, and Sanofi; he has no equity or ownership in any of these companies.

Reference

1. Coult J, Yang BY, Kwok H, et al. Prediction of shock-refractory ventricular fibrillation during resuscitation of out-of-hospital cardiac arrest. Circulation. 2023 June 2. doi:10.1161/CIRCULATIONAHA.122.063651