Cardiac magnetic resonance imaging-derived atrial fibrosis in patients with pre-atrial fibrillation.

INTRODUCTION: Atrial fibrosis identified on cardiac magnetic resonance (CMR) imaging has been proposed as a preprocedural imaging biomarker for patient selection for rhythm control interventions in patients with atrial fibrillation (AF). Whether atrial fibrosis is present in patients considered as 'pre-AF' is unknown. METHODS AND RESULTS: We prospectively recruited 12 participants with pre-AF as defined by the Future Innovations in Novel Detection for Atrial Fibrillation (FIND-AF machine learning algorithm, without AF diagnosed during AF screening, and compared them to 25 participants with confirmed AF. All participants underwent CMR using a 3T system with left atrial fibrosis quantification and ADAS-3D left atrial image postprocessing software. Participants with pre-AF had smaller left atrial end-systolic (33.6±9.8 vs 43.0±17.0, p=0.003) and end-diastolic (16.5±8.7 vs 28.2±14.4, p=0.007) volumes, and higher left atrial ejection fraction (59.6±14.6 vs 40.7±17.5, p=0.005) than participants with AF. The extent of atrial fibrosis was not different between those with pre-AF and AF (borderzone (%) 5.2±5.0 vs 2.9±6.9, p=0.772; borderzone fibrosis (cm) 6.2±5.8 vs 6.8±10.7, p=0.927). CONCLUSION: CMR identifies atrial fibrosis before manifest AF in patients with pre-AF as defined by a machine learning algorithm.