Heart Rate Variability and Clinical Features as Predictors of Atrial Fibrillation Recurrence After Catheter Ablation: A Pilot Study

Figures

• 

  FIGURE 1. Flow diagram describing the source population.

• 

  FIGURE 2. Example of available implantable cardiac monitor data: (A) whole flashback, (B) first 300 beats of the flashback, (C) last 100 beats of the flashback, and (D) first 2 min of the last atrial fibrillation (AF) episode recorded before catheter ablation, illustrating the short-term changes in the R–R intervals (ms) prior to the AF onset and the differences in R–R intervals between the flashback (A) and the AF episode (D).

• 

  FIGURE 3. Schematic representation of the overall method. NR, no recurrence; R, recurrence; SFFS, sequential forward floating search.

• 

  FIGURE 4. (A) Mean accuracy of the single classifiers on the validation set, plotted as a function of the number of selected features. (B) Mean and standard deviation of the accuracy of the multiple classifiers (line and colored area, respectively) for each subset of features on the validation set. AWV, accuracy weighted voting; CART, classification and regression trees; KNN, K-nearest neighbors; OWV, optimum weighted voting; SVM, support vector machine; SVMg, support vector machine Gaussian kernel; SVMp, support vector machine polynomial kernel.

• 

  FIGURE 5. Mean and standard deviation of the performance metrics on the test set of the different classification methods. The table shows the mean values. AWV, accuracy weighted voting; CART, classification and regression trees; KNN, K-nearest neighbors; OWV, optimum weighted voting; SVM, support vector machine; SVMg, support vector machine Gaussian kernel; SVMp, support vector machine polynomial kernel.

• 

  FIGURE 6. (A) Receiver operating characteristic (ROC) curves of the single classifiers and area under the curve (AUC). (B) ROC curves of the ensemble classifiers and the AUC. AWV, accuracy weighted voting; CART, classification and regression trees; KNN, K-nearest neighbors; OWV, optimum weighted voting; SVM, support vector machine; SVMg, support vector machine Gaussian kernel; SVMp, support vector machine polynomial kernel.

• 

  FIGURE 7. Percentage of features from each feature group used by the different classification methods. AF, atrial fibrillation; AWV, accuracy weighted voting; CART, classification and regression trees; FB, flashback; KNN: K-nearest neighbors; L100, last 100 beats; OWV, optimum weighted voting; SVM, support vector machine; SVMg, support vector machine Gaussian kernel; SVMp, support vector machine polynomial kernel.