Does Monte Carlo Propagation improve the precision of ventricular tachycardia ablation target prediction compared to deterministic models in patient-specific cardiac models?
Applying Monte Carlo Propagation to patient-specific cardiac models significantly improves the precision of VT ablation target prediction by quantifying uncertainty and reducing false positives.
Ventricular tachycardia (VT) occurs when scarred myocardial tissue creates self-sustaining reentrant circuits disrupting normal ventricular rhythm. Catheter ablation treats VT by focusing on critical circuit sites, or ablation targets. However, the models used in clinical spaces are deterministic - relying on singular simulation runs that can overlook critical ablation sites and lead to increased procedural costs or other potential complications. This study applied Monte Carlo Propagation (MCP) to a clinically-standard Eikonal Cardiac Model to quantify uncertainty in ablation target prediction. A computational pipeline processed data from the Evaluation of Myocardial Infarction from Delayed-Enhancement Cardiac MRI (EMIDEC) dataset: MRI scans were segmented, then converted into meshes using GMSH, reformatted into FEniCS format for finite element analysis, and used for electrophysiological simulation. The uncertainty framework then identified targets that were stable across various parameters and assigned confidence scores for decision support. Across nine patients, the MCP model achieved 33.7% target stability, with 21.0 stable nodes per case. Compared to its deterministic counterpart, MCP was able to improve top-10 precision by 40.0 percentage points (0.967 MCP vs. 0.567 deterministic) and was able to flag 36.8% of the deterministic predictions as low-confidence false positives. In terms of Receiver Operating Characteristic (ROC), MCP had an AUC (area under the curve) of 0.966 while the deterministic baseline had an AUC of 0.500, demonstrating the model’s ability to differentiate high and low confidence points. These results demonstrate that using uncertainty quantification in patient-specific cardiac models allows for a more accurate prioritization of ablation targets while reducing false positives. By providing predictions weighted on confidence, the framework supports risk-stratified clinical decision-making. While larger validation is needed, these results establish proof-of-concept for uncertainty-aware cardiac modeling.
Sudipta Barua (Tue,) studied this question.