Quantitative QRS voltage variability for differentiating fast atrial fibrillation from supraventricular tachycardia: a large ECG dataset study

Background Differentiating fast atrial fibrillation (AF) from supraventricular tachycardia (SVT) is challenging when ventricular rates are high and QRS complexes are narrow. Traditional visual assessment of rhythm irregularity is subjective and can lead to misdiagnosis. Objective amplitude variability metrics may improve diagnostic accuracy. Purpose To evaluate the utility of beat-by-beat QRS voltage variability as a quantitative marker to distinguish fast AF from SVT using open-source ECG recordings and AI analysis. Methods A retrospective case-control study analyzed 600 ECG recordings (300 fast AF, 300 SVT) sourced from PhysioNet open databases. High-quality lead II signals were selected. QRS peak-to-peak amplitudes were measured beat-by-beat across 10-second windows to compute a QRS Voltage Variability Index (QRS-VI) = (SD/mean ×100). AI algorithms were used for automated amplitude extraction and variability computation. ROC analysis assessed diagnostic performance, including optimal cutoff values, sensitivity, specificity, and area under the curve (AUC). Results The QRS-VI was significantly higher in fast AF than SVT (17.8 ± 6.2 % vs 4.9 ± 1.7 %; p < 0.001). ROC analysis demonstrated excellent discriminatory power (AUC = 0.96). An optimal cutoff of ≥10.5 % yielded sensitivity 92 % and specificity 89 % for identifying fast AF. Conclusion Quantitative QRS voltage variability is a robust, reproducible, and objective marker for differentiating fast AF from SVT. This metric has potential clinical application for automated ECG interpretation, telemetric monitoring, and wearable cardiac devices.