Background: Complex fractionated atrial electrograms (CFAE) have become targets for catheter ablation of atrial fibrillation (AF). Frequency components of AF signals have also become important markers for identifying potential mechanisms of AF, yet inaccuracies exist, particularly in standard dominant frequency (SDF) calculations especially at CFAE sites. We developed new methodology to improve accuracy of AF rate determinations at such recording sites.
Objective: To develop optimal methods for estimating activation rates in paroxysmal and persistent AF.
Methods: Electrograms were obtained from one right atrial, coronary sinus, and 6 left atrial (LA) endocardial regions manifesting CFAEs in paroxysmal (N = 7) and persistent (N = 7) AF patients. SDF was measured from 8.4 s intervals and compared to (1) optimized DF (ODF) calculated by optimizing the filter coefficients which maximized dominant frequency power, (2) autocorrelation (AC), with the rate estimated as the inverse of the signal phase shift generating the largest autocorrelation coefficient, and (3) ensemble average (EA), with the rate estimated by summing successive signal segments and selecting segment length yielding maximum power. Rate measurements were compared between groups, at baseline and with additive interference, having similar frequency content to the electrograms, to test the robustness of the different methods.
Results: From pooled data (N = 168 recording sites), a significantly higher LA dominant frequency was found in persistent versus paroxysmal patients using each method (P < 0.001), with a mean value for all methods of 6.23 +/- 0.08 Hz versus 5.32 +/- 0.10 Hz, respectively. At the highest additive interference level, the rate measurement error was significantly greater in SDF as compared with EA (P = 0.010) and ODF (P = 0.035), and at all interference levels SDF had the largest error of any method.
Conclusions: SDF appears less robust to additive interference, compared to the ODF and EA methods of estimating the activation rate at CFAE sites in this small group of patients. Use of optimized filter coefficients for DF measurement, or use of correlative methods such as EA, that reinforce the signal rather than filtering the noise, may improve calculation of activation rates.