Abstract
Aims: The decision of treatment strategy made for atrial fibrillation (AF) patients is ideally based on individual atrial remodeling. Electrical remodeling is believed to influence the outcome of direct current cardioversion (DCC) of persistent AF. Characterization of AF from the surface electrocardiography (ECG) using spectral analysis is able to quantify atrial electrical remodeling. Applying spectral analysis of the surface ECG in patients with persistent AF undergoing DC cardioversion, the present study is aiming to (1) determine the feasibility of spectral analysis in estimating the dominant frequency (DF), median frequency (MF), and frequency bandwidth (FB) from the surface ECG; (2) assess and compare the QRS-T removal performance of two algorithms, average beat subtraction (ABS) and singular value decomposition (SVD); (3) detect if there are consistent changes of DF between leads V1 and V6, which may reflect the left-to-right atrial gradient; (4) explore the utility of this ECG spectral analysis technique for prediction of DCC outcome and the value of clinical and echocardiographic variables in predicting the outcome of DCC was evaluated as well.
Methods and results: This study cohort consisted of 11 consecutive patients (9 men, the mean age 64 ± 10 years) with persistent AF (the median duration 5 months, range 1 to 108 months) undergoing elective external cardioversion with three shock attempts at most. Shock results were observed. Three 10-second segments of the 12-lead digital ECG were obtained for each subject prior to cardioversion. After filtering, QRS-T complexes were removed from the ECG using both ABS and SVD. Frequency power spectra were generated by Fourier transformation of the remaining atrial signal ECG. The DF, MF and FB were determined in the corresponding power spectrum. The dominant rate (DR) was obtained from the conversion of DF. Due to the small population, no attempt was made to do statistical analysis and the data was generally described.
The atrial signal extraction performance was compared between these two algorithms by visually inspecting the number of leads that had residual truncated QRS-T waves. It was shown that although neither ABS nor SVD performed perfect with much residual ventricular activity in the remaining ECG, SVD caused less QRS-T-related residuals when compared with ABS in all subjects.
The mean value of DR obtained using SVD was slightly lower in people who were successfully converted to sinus rhythm (SR) than in those with final shock failure (392 ± 52 fpm vs. 404 ± 39 fpm). The mean DR of the subgroup where SR was initiated by one single shock was further lower when compared with the remaining subjects (358 ± 7 vs. 413 ± 44 fpm). A consistent pattern was observed for the DF of atrial activity to be faster in lead V1 than in lead V6, with a frequency difference between 0.4 and 2.0 Hz. The distinction between the DFs from leads V1 and V6 was obviously higher in subjects with successful DC cardioversion than in those with failed cardioversion (1.48 ± 0.47 vs. 0.15 ± 0.78 Hz). In addition, the three highest DF alternations existed in the subjects who were converted to SR by only one shock, which resulted in a further higher mean gradient in this group than in the remaining people who had at least one failed shock (1.77 ± 0.32 vs. 0.45 ± 0.77 Hz).
Conclusion: Spectral analysis of the surface ECG is feasible to non-invasively assess the DF, MF, and FB in patients with persistent AF. With less leads containing visual ventricular activity, the QRS-T subtraction performance of SVD might be superior to that of ABS. The DR converted from the DF would be useful in predicting shock results in patients with persistent AF. The atrial gradient could be reflected by the difference between the measurements of DF from leads V1 and V6. The atrial gradient detected from the ECG prior to cardioversion might be an important predictor for initial success of electrical cardioversion. This finding might be useful in identifying suitable candidates for DCC to avoid unnecessary cardioversion attempts and expect a higher likelihood of shock success. Due to the limitations of the present study, the results need to be verified by more investigations.