Toward a computationally efficient numerical approach to simulate the left atrial appendage occlusion in patient-specific anatomies.
TL;DR
The proposed pipeline enables scenario-based, patient-specific LAAO simulation and substantially reduces computational cost while preserving accuracy, whereas neglecting hooks (BFM) is not acceptable.
Abstract
Background
AND
Objectives
Left atrial appendage occlusion (LAAO) is increasingly used to prevent stroke in atrial fibrillation, but device selection and positioning remain challenging due to complex, patient-specific anatomy and uncertain wall mechanics. Finite element (FE) simulations could support pre-interventional planning, provided they are both accurate and computationally efficient. This study extends a previously validated Watchman FLX FE model to patient-specific anatomies and evaluates model simplifications to reduce runtime without compromising clinically relevant predictions.
Methods
A detailed Watchman FLX model (Complete Model, CM) and two simplified variants (No Fabric Model, NFM, in which the proximal fabric membrane is omitted; Bare Frame Model, BFM, consisting solely of the metallic frame and excluding both the fabric and the anchoring hooks) were implemented in Abaqus/Explicit. Three patient-specific LA-LAA anatomies were reconstructed from CT. For each anatomy, four scenarios were simulated (optimal/suboptimal catheter positioning × low/high wall stiffness), for a total of 12 scenarios and 36 simulations. Deployment, anchoring and release were modeled. Outcomes included LAA sealing (gap classification), device protrusion into the LA, intrusion into the LAA, and CPU time.
Results
With the CM, procedural outcome strongly depended on positioning and wall stiffness, even for a fixed device size. For all anatomies, at least one scenario achieved complete sealing and acceptable alignment. The NFM reproduced CM gap classification in all cases and intrusion/protrusion within ±1 mm in 10/12 scenarios, correctly identifying poor outcomes in the remaining two. The BFM showed unstable deployment in 2/12 scenarios and mispredicted gaps and protrusion in others. Average runtimes were 5.6 h (CM), 3.8 h (NFM), and 3.2 h (BFM).
Conclusions
The proposed pipeline enables scenario-based, patient-specific LAAO simulation. Neglecting the fabric (NFM) substantially reduces computational cost while preserving accuracy, whereas neglecting hooks (BFM) is not acceptable.