Aims
Atrial cardiomyopathy (ACM) is associated with the progression from paroxysmal through persistent to permanent atrial fibrillation (AF) and increased arrhythmia recurrence rates after pulmonary vein isolation (PVI). We compare the most common left atrial (LA) late gadolinium enhancement magnetic resonance imaging (LGE-MRI)-methods (Utah-method and image-intensity-ratio (IIR)-methods) and endocardial voltage mapping for ACM-detection and outcome prediction after PVI for AF.  

Methods
In this prospective study, 37 ablation-naive patients (66±9 years, 84% male) with persistent AF were included and electrically cardioverted into sinus rhtyhm. Subsequently, they underwent LA-LGE-MRI and high-definition voltage and activation mapping (2129±484 sites) in sinus rhythm prior to PVI. MRI-post-processing-analyses were performed by two independent expert laboratories (Marrek for the Utah segmentation as describers of this methodology and Adas 3D medical as developers of the Adas software for the IIR-methods) which were blinded to any clinical data. Arrhythmia recurrence was recorded within 12 months following PVI.

Results
The global ACM-extent was highly variable: median LA low-voltage substrate (LA-LVS) was 12.9% at <1.0mV and 2.7% at <0.5mV cut-off; median LA-LGE-extent using the Utah-method was 18.3% and 0.03%-93.1% using the IIR-methods. LA activation time was significantly correlated with LA-LVS (r=0.76 at <0.5mV and r=0.82 at <1.0mV, both p<0.0001), but not with LA-LGE-extent.

The highest regional matching between LA-LVS <0.5mV and LA-LGE was found for the anterior wall in 57% of patients using the Utah-method and in 59% using IIR 1.20. The corresponding values for the posterior wall were 19% and 38%, respectively.

Arrhythmia recurrence occurred in 15 (41%) patients. Freedom from arrhythmia was significantly lower in those with relevant LA-LVS (≥2cm² at 0.5mV) but not in those with relevant LGE (Utah-stages III&IV): 43% versus 81%, p=0.009 and 50% versus 67%, p=0.338, respectively. Furthermore, relevant LA-LVS was the only predictor for arrhythmia recurrence in multivariate regression analysis. 

Conclusion
The different LA-LGE-MRI-methods have large discrepancies regarding extent and distribution of ACM and are different to the extent and regional distribution of LA-LVS as assessed in endocardial high-definition voltage mapping in sinus rhythm. Further improvements of the LA-LGE-MRI-methods are required to enable correct diagnosis of ACM and for future evaluation of MRI-guided ablation protocols.