Determination of Intra-Atrial Fibrillatory Cycle Length From the Electrocardiogram Using Auto-Correlation Analysis

The cycle length (CL) of atrial fibrillation (AF) may have clinical utility, yet reliable CL measurements typically require invasive electrophysiologic study (EPS). We hypothesized that autocorrelation analysis of the ECG accurately reflects intra-atrial CL in AF, and is more robust than spectral analysis from QRS-subtracted ECGs. Methods We studied 39 patients undergoing AF ablation. For ECG leads V5, aVF and V1, an atrial sample was cross-correlated to its ECG at successive timepoints. Resulting waveforms were auto-correlated; the time of first maximum (>0) was identified as AF CL. For comparison, we subtracted mean QRS complexes from each ECG, then determined CL as the dominant frequency between 4-10 Hz. ECG CL estimates were compared to mean AF CL from the coronary sinus. Results ECGs in n=18 patients had QRS ectopy or aberrancy preventing QRS subtraction. In n=21 patients, auto-correlation (p<0.0002) and spectral (p<0.009) analyses predicted intraatrial AF CL. Auto-correlation better reflected intra-atrial CL (regression slope = 0.80, intercept 42 ms) than spectral analysis (slope 0.50, intercept 81 ms). In ECGs with QRS ectopy (n=18), autocorrelation remained predictive of AF CL (slope = 1.17, intercept 26 ms, r=0.87; p=0.001). Conclusion Intra-atrial CL in AF is accurately determined by autocorrelation analysis of the surface ECG. This approach may have wider clinical applicability than traditional ECG spectral analysis. Keywords—Atrial Fibrillation, Correlation, Cycle length, ECG, Electrophysiologic Study, Fourier Analysis.