Bayesian Framework for Atrial LAT Estimation in ECGI

Local activation time (LAT) mapping computed on ECGI signals offers a comprehensive representation of the electrical propagation in the heart. However, traditional methods for LAT estimation based on invasive mapping systems produce artefacts when applied on ECGI signals. This study aims to introduce and evaluate a novel Bayesian framework for estimating LATs using ECGI. This method was compared against the traditional LAT method (-dV/dt) and the spatiotemporal gradient (STG) optimization method. We used in-silico pacing models and two patients in the evaluation, one in sinus rhythm and another during pacing post-cavotricuspid isthmus (CTI) ablation. In simulations, the Bayesian approach yielded LAT maps with Pearson correlation coefficients exceeding 0.91 across all noise settings, significantly outperforming the -dV/dt and the STG methods, whose correlations ranged from 0.75 to 0.81 and from 0.78 to 0.86, respectively. Additionally, the error in locating the earliest and latest activation sites was reduced by up to 0.8 cm in low-noise scenarios. Clinically, the Bayesian framework identified and quantified real lines of block (LoB) without introducing artifacts, significantly decreasing the presence of artificial LoB across simulation data. The novel Bayesian framework for LAT estimation using ECGI represents a significant advancement over conventional methods, offering a non-invasive, accurate, and efficient alternative for mapping cardiac arrhythmias.