MLWAVE: A novel algorithm to classify primary versus secondary asphyxia-associated ventricular fibrillation.

Area of Science:

• Biomedical engineering
• Cardiovascular research
• Machine learning in medicine

Background:

• Cardiopulmonary resuscitation (CPR) success hinges on rapid identification of cardiac arrest causes.
• Differentiating primary ventricular fibrillation (VF) from asphyxia-associated VF is crucial for effective treatment.
• Current methods for VF waveform analysis may lack the precision needed for personalized resuscitation strategies.

Purpose of the Study:

• To develop and validate a novel technique using wavelet synchrosqueezed transform (WSST) and a decision-tree classifier.
• To discriminate between primary VF and secondary asphyxia-associated VF using ECG signal analysis.
• To create a machine learning algorithm (MLWAVE) for classifying VF types and compare its performance against established methods.

Main Methods:

• Analysis of ECG data from swine models with induced primary VF (n=18) or asphyxia-associated VF (n=12).
• Application of WSST to the initial 35 seconds of VF ECG signals to extract differentiating features.
• Development of the MLWAVE algorithm and assessment using classification accuracy and AUCROC; comparison with the Amplitude Spectrum Area (AMSA) technique.

Main Results:

• The MLWAVE technique achieved 100% accuracy and an AUCROC of 1.00 in classifying VF types within the first 35 seconds.
• MLWAVE outperformed the AMSA technique (100% vs. 85.7% accuracy; 1.00 vs. 0.82 AUCROC).
• While MLWAVE showed superior performance, differences were not statistically significant due to small sample sizes (p=0.24).

Conclusions:

• The MLWAVE signal processing method demonstrates high accuracy (100%) in classifying VF waveforms.
• Accurate VF classification can guide personalized resuscitation, optimizing interventions like defibrillation timing.
• This approach holds potential for improving cardiac arrest survival rates through tailored CPR strategies.