Application of Deep Learning to Predict Optimal Ablation Therapy for Atrial Fibrillation from Magnetic Resonance Imaging Data

The prevalence of atrial fibrillation (AF) is increasing in epidemic proportions: worldwide over 33 million individuals have AF (Chugh et al., 2014). The disease is associated with increased levels of morbidity and mortality, high risks of developing heart failure and stroke, and hence very high rates of patient hospitalizations. The overall economic burden of AF amounts to over 1% of total healthcare costs in the UK. Rhythm control strategies for maintaining sinus rhythm, such as antiarrhythmic drugs, can lead to significant improvements of cardiac output and quality of life. However, treatments of AF are complicated by its mechanisms for self-sustenance, such as the presence of AF-induced electrical and structural remodelling that generates more treatment-resistant arrhythmia (Nattel and Harada, 2014).

Radiofrequency catheter ablation (CA) therapy, which is aimed at destroying arrhythmogenic tissue areas in the atria via high energy delivery through a catheter, has become a first-line treatment for AF, and it is the only treatment with a proven long-term curative effect (Kirchhof et al., 2016). However, even advanced CA procedures have suboptimal long-term outcomes in patients with chronic forms of AF: over half of the patients return for additional treatment within three years (Calkins et al., 2017). This can be explained by the highly empirical nature of CA therapy, which targets "usual suspect" areas without knowledge of the underlying mechanisms. It is believed that ectopic electrical beats from the pulmonary veins (PV) can trigger AF, and that re-entrant drivers (RDs) generated by breakdown of such ectopic waves can sustain AF. However, empirical CA therapy based on electrical isolation of the pulmonary veins (PV) has low success rates in chronic AF patients, where extensive ablation of remodelled non-PV areas is commonly applied (Roten et al., 2012). This warrants the development of novel approaches that can improve the efficacy of CA therapy and clinical outcomes in a large patient population.