Prediction of Sinus Rhythm Maintenance After Electrical Cardioversion Using Spectral and Vector Cardiographic ECG Analysis

Introduction

Electrical cardioversion (ECV) remains a treatment option for atrial fibrillation (AF). The study aimed to find predictors of SR maintenance after ECV using spectral and vector cardiographic (VCG) analysis of ECGs.

Methods

Consecutive patients with AF referred for elective ECV were prospectively enrolled. A digital ECG recording was obtained before the ECV and was analyzed using spectral and VCG analysis. AF activity was analyzed using spectral analysis to determine the dominant frequency (DF), RI (regularity index), and OI (organizational index). QRS complexes were analyzed using vectorcardiography to determine the dXmean, dYmean, and dZmean (derivation of VCG signals). We used Lasso Logistic Regression (LLR) in five-fold cross-validation for feature selection and to build combined predictive models of SR maintenance. For model training and evaluation, data were split in a 60%–40% ratio for training and testing, respectively.

Results

A total of 80 patients were enrolled (age 70.2 ± 10.6 years, 49 (61%) were men, BMI 29.7 kg/m²). At the 3-month follow-up, AF recurrence was present in 36 patients (45%). The best single VCG parameter to predict SR maintenance was dZMean (OR 0.18, 95% CI 0.06–0.51, p < 0.001). VCG-domain parameters combined into the LLR model showed an area under the curve (AUC) of 0.78. From the spectral analysis domain, the best predictor was DF (OR 3.54, 95% CI 1.28–10.25), p = 0.006; spectral features led to an AUC of 0.76 when combined in the LLR model. Clinical features did not form a model since no features passed feature selection. Combining VCG and spectral analysis features led to an LLR model with an AUC of 0.79.

Conclusion

The combination of spectral analysis of AF activity and VCG analysis of ventricular activity provided more accurate predictive information than either analysis alone.