Yanyan Ma, Jinzi Liu, Cheng Yang, Qiaoman Huang, Yixu Lin, Shuiqing Gui, Tangxing Jiang
Ventricular fibrillation (VF) and pulseless ventricular tachycardia are major shockable rhythms in cardiac arrest. Although prompt defibrillation is essential, the probability of shock success changes as myocardial ischemia, energy depletion, coronary perfusion, and cardiopulmonary resuscitation (CPR) quality alter the VF waveform. This narrative review examines the physiological basis, waveform measures, and clinical evidence for VF waveform analysis. It distinguishes animal experiments, retrospective clinical cohorts, and prospective intervention trials. Time-domain, frequency-domain, time-frequency, nonlinear, and combined measures provide complementary information, but most are less well validated than amplitude spectrum area (AMSA). Higher AMSA is consistently associated with defibrillation success; however, reported thresholds vary with patient population, device, signal processing, measurement timing, and outcome definition. Prospective trials have not established that AMSA-guided treatment improves survival. Real-time electrocardiogram analysis and artifact suppression may shorten interruptions in chest compressions, while artificial intelligence may combine waveform and physiological information. These approaches remain investigational, lack adequate prospective validation, and are not recommended by current guidelines for deciding when to deliver a shock in routine care. The review also considers refractory VF, including acute coronary occlusion, vector-change defibrillation, double sequential external defibrillation, and extracorporeal cardiopulmonary resuscitation. No waveform or artificial intelligence threshold has been validated to decide who should receive these interventions. VF waveform analysis is promising as an adjunct to guideline-based resuscitation, but future trials must show improved CPR delivery and patient-centered outcomes. An indicated shock should not be delayed or withheld on the basis of waveform analysis alone.