Preprocedure Application of Machine Learning and Mechanistic Simulations Predicts Likelihood of Paroxysmal Atrial Fibrillation Recurrence Following Pulmonary Vein Isolation. (July 2020)

• Record Type: Journal Article
• Authors: Shade, Julie K.; Ali, Rheeda L.; Basile, Dante; Popescu, Dan; Akhtar, Tauseef; Marine, Joseph E.; Spragg, David D.; Calkins, Hugh; Trayanova, Natalia A.
• Is Part Of: Circulation. Volume 13:Number 7(2020)
• Publication Date: 2020-07
• Subjects: atrial fibrillation -- computational biology -- heart atria -- machine learning -- magnetic resonance imaging; Arrhythmia -- Periodicals; Heart -- Electric properties -- Periodicals; 616.128
• DOI: 10.1161/CIRCEP.119.008213
• Languages: English
• ISSNs: 1941-3149

Abstract

Abstract : Background: Pulmonary vein isolation (PVI) is an effective treatment strategy for patients with atrial fibrillation (AF), but many experience AF recurrence and require repeat ablation procedures. The goal of this study was to develop and evaluate a methodology that combines machine learning (ML) and personalized computational modeling to predict, before PVI, which patients are most likely to experience AF recurrence after PVI. Methods: This single-center retrospective proof-of-concept study included 32 patients with documented paroxysmal AF who underwent PVI and had preprocedural late gadolinium enhanced magnetic resonance imaging. For each patient, a personalized computational model of the left atrium simulated AF induction via rapid pacing. Features were derived from pre-PVI late gadolinium enhanced magnetic resonance images and from results of simulations of AF induction. The most predictive features were used as input to a quadratic discriminant analysis ML classifier, which was trained, optimized, and evaluated with 10-fold nested cross-validation to predict the probability of AF recurrence post-PVI. Results: In our cohort, the ML classifier predicted probability of AF recurrence with an average validation sensitivity and specificity of 82% and 89%, respectively, and a validation area under the curve of 0.82. Dissecting the relative contributions of simulations of AF induction and raw images to the predictive capability of the ML classifier, we found that when only features from simulations of AF induction were used to train the ML classifier, its performance remained similar (validation area under the curve, 0.81). However, when only features extracted from raw images were used for training, the validation area under the curve significantly decreased (0.47). Conclusions: ML and personalized computational modeling can be used together to accurately predict, using only pre-PVI late gadolinium enhanced magnetic resonance imaging scans as input, whether a patient is likely to experience AF recurrence following PVI, even when the patient cohort is small. Abstract : Supplemental Digital Content is available in the text.