Abstract 12471: Optimal Prediction of Atrial Fibrillation Recurrence After Ablation: A Computational Anatomy Study

Background: Left atrial structural remodelling, assessed by left atrial (LA) sphericity or antero-posterior diameter, has been shown to predict recurrence after atrial fibrillation (AF) ablation. The study aimed to perform a computational shape analysis of the LA to quantitatively characterise the LA shape remodelling process and identify metrics that optimally predict recurrence. Methods: Pre-procedural bright-blood MRIs of the LA of patients undergoing AF ablation were segmented. Patient-specific smooth 3D meshes were fitted to the segmentations. A statistical shape model of the LA was created and the global features underpinning the observed shape variation extracted as principal components (PCs). PCs were optimally combined to create non-empirical atlas-based metrics using linear discriminant analysis. Meshes depicting mean and extreme recurrent and non-recurrent LA shapes were also synthetized. The capability of different metrics to predict recurrence was evaluated using the area under the ROC curve (AUC) of a leave 1 out cross validation test. Results: In total, 111 patients were included. At 12 months follow-up, LA sphericity was the best predictor of recurrence (AUC: 0.66) over novel atlas-based metrics (AUC: 0.65). At 24 months, atlas-based metrics were the best predictors of recurrence (AUC: 0.66), outperforming a combination of sphericity and volume (AUC: 0.64), sphericity alone (AUC: 0.63) and any other traditional metric. Conclusions: Novel atlas-based metrics improve the prediction of recurrence at 2 years post-AF ablation. They allow a more complete characterization of the LA shape remodelling process, for example by allowing the synthesis of recurrent and non-recurrent LA shapes, which may contribute to patient stratification for AF ablation.