Advanced atrial fibrillation (AF) patients have persistent AF, failed previous catheter ablation and/or an enlarged left atrium (LA), which is associated with a reduced success of AF ablation. Transthoracic echocardiography (TTE) and contrast enhanced magnetic resonance angiography (CE-MRA) are available to assess LA volume. However, it is unknown how these modalities relate in patients with advanced AF. We therefore compared the reproducibility of TTE and non-triggered CE-MRA in advanced AF patients and their ability to select patients with successful thoracoscopic AF ablation. Two independent observers measured LA volumes on 65 TTE and CE-MRA exams of advanced AF patients prior to AF ablation. Patients were followed after AF ablation with rhythm monitoring every 3 months for 1 year to determine AF recurrence. Inter-modality, inter- and intra-observer variability were determined using intraclass correlation coefficients (ICC). Receiver-operating characteristic (ROC) analysis was performed to determine sensitivity and specificity of TTE and CE-MRA volume and CE-MRA dimensions to identify patients with AF recurrence during follow-up. LA enlargement ≥ 34 ml/m was present in 60% of the patients. CE-MRA and TTE demonstrated a good correlation for LA volume assessment (intraclass correlation, ICC = 0.86; p < 0.001) with larger volumes consistently measured by CE-MRA. Major discrepancies were mostly attributed to TTE acquisition. Craniocaudal enlargement discriminated patients with AF recurrence (AUC 0.67 95% CI 0.55-0.85, p = 0.01). Non-triggered CE-MRA is a viable and reproducible 3D alternative for 2D TTE to assess LA volume in advanced AF patients. Craniocaudal enlargement was the only discriminator of AF recurrence after AF ablation.