Electrocardiogram-based pulse prediction during cardiopulmonary resuscitation

Area of Science:

• Cardiology
• Medical Technology
• Signal Processing

Background:

• Resuscitation requires frequent interruptions of cardiopulmonary resuscitation (CPR) to assess pulse and rhythm.
• Minimizing these interruptions is crucial for improving patient outcomes during cardiac arrest events.

Purpose of the Study:

• To develop and validate a method for predicting real-time pulse status using electrocardiogram (ECG) segments during CPR.
• To assess the accuracy of ECG-based pulse prediction with and without CPR artifact.

Main Methods:

• A cohort of patients with ventricular fibrillation arrest undergoing resuscitation was analyzed.
• Wavelet-based ECG features were extracted from audio-supplemented defibrillator recordings.
• A logistic model combining three ECG features was developed to predict pulse status, validated against manual annotations.

Main Results:

• The logistic model achieved an area under the receiver operator characteristic curve (AUC) of 0.84 for pulse prediction during CPR.
• Individual ECG features showed moderate predictive performance (AUC 0.67-0.79).
• The model's performance was consistent regardless of the presence of CPR artifact.

Conclusions:

• Wavelet-based ECG analysis can predict pulse status with moderate accuracy during CPR.
• Further research is needed to integrate real-time pulse prediction into clinical practice to guide resuscitation decisions and minimize interruptions.