AI-Derived Pericardial Effusion Volume and Long-Term Mortality After Transcatheter Aortic Valve Implantation.
Authors
Affiliations (2)
Affiliations (2)
- Department of Radiology, Paracelsus Medical University, 5020 Salzburg, Austria.
- Division of Cardiology, Department of Internal Medicine II, Paracelsus Medical University, 5020 Salzburg, Austria.
Abstract
<b>Background:</b> Pericardial effusion is a common finding on pre-procedural computed tomography (CT) in patients undergoing transcatheter aortic valve implantation (TAVI). However, its clinical significance remains uncertain. Using artificial intelligence (AI)-based image analysis, we sought to determine whether pericardial effusion represents an incidental imaging finding, a marker of hemodynamic burden, or a prognostically relevant phenotype. <b>Methods:</b> This retrospective single-center study included 470 consecutive patients undergoing TAVI for severe aortic stenosis. Pericardial effusion volume was quantified using an AI-based CT segmentation workflow and categorized as no effusion (0 mL), trace effusion (>0-<5 mL), small effusion (5-<30 mL), or larger effusion (≥30 mL). Associations with baseline clinical and echocardiographic characteristics were assessed. Long-term mortality was evaluated using Kaplan-Meier analysis, Cox regression models, and restricted cubic spline analyses. <b>Results:</b> Detectable pericardial effusion was present in 276 patients (58.7%), although larger effusions were uncommon (5.1%). Increasing pericardial effusion volume was associated with a higher prevalence of atrial fibrillation (<i>p</i> = 0.001), higher systolic pulmonary artery pressure (<i>p</i> = 0.004), and lower TAPSE/sPAP ratios (<i>p</i> = 0.007). In contrast, left ventricular ejection fraction and transvalvular gradients did not differ across effusion categories. During long-term follow-up, no significant differences in mortality were observed between pericardial effusion groups (log-rank <i>p</i> = 0.46). Pericardial effusion volume was not associated with mortality when analyzed as a continuous variable (HR 1.00, 95% CI 0.996-1.01; <i>p</i> = 0.842), after logarithmic transformation (adjusted HR 0.91, 95% CI 0.76-1.08; <i>p</i> = 0.293), or in restricted cubic spline analyses. <b>Conclusions:</b> AI-derived pericardial effusion volume identifies a hemodynamic phenotype characterized by atrial fibrillation, pulmonary hypertension, and impaired right ventricular-pulmonary arterial coupling. Despite these associations, pericardial effusion volume does not independently predict long-term mortality after TAVI, suggesting that it reflects cardiovascular congestion and remodeling rather than a prognostically relevant phenotype.