A Complex Gaussian Process Framework for Coherent ECGI Mapping and Isthmus Identification in Reentrant VT

Carlos Fambuena Santos1, María Correas García2, Andrea Cano Cabañero2, Jana Reventós Presmanes3, Jean-Baptiste Guichard4, Andreu Porta4, Ivo Roca Luque5, Alfred Peris1, Andreu M. Climent6, Maria de la Salud Guillem Sánchez6
1UPV, 2ITACA Institute, Universitat Politècnica de València, 3Arrhythmias Department, Hospital Clínic de Barcelona, 4Hospital Clinic de Barcelona, 5Institut Clínic Cardiovascular, Hospital Clínic de Barcelona, Catalonia, Spain, 6Universitat Politècnica de València


Abstract

Non-invasive electrocardiographic imaging (ECGI) for reentrant ventricu-lar tachycardia (VT) suffers from spatial incoherence due to circular activa-tion, hindering isthmus localization. We introduce a Complex Gaussian Pro-cess (CGP) framework to model local activation time (LAT) circular topolo-gy, improving spatial consistency. We compute conduction velocity (CV) and curl from high-resolution CGP-LAT maps, hypothesizing the isthmus exhibits slow conduction (high gradient) and low curl. The method was eval-uated on simulated (N=4, ground truth) and patient (N=1, clinical electroana-tomical mapping) VT data against standard deflection-based LAT maps. CGP-LAT maps showed significantly improved spatial coherence. In simula-tions, CGP yielded superior propagation direction accuracy (mean cosine 0.81 vs 0.64) and lower vector field error (RMSE 0.91 vs 1.08 ms/mm). Ap-plying CV/curl criteria to CGP maps allowed more precise simulated isthmus localization. In patients, CGP analysis revealed clearer CV/curl patterns, identifying channels concordant with clinical annotations and EAM. Addressing reentry's topology, CGP combined with wavefront analysis of-fers a promising non-invasive tool for improved VT isthmus localization.