Optimising Beat Selection and Averaging for ECGI to Enhance EGM Reconstruction Fidelity

Electrocardiographic Imaging (ECGI) offers spatially resolved information on cardiac electrophysiology by solving an inverse problem. By strategically placing electrodes on the torso, ECGI captures body surface potentials (BSP) to reconstruct the underlying electrical activity as electrograms (EGM) on the epicardial surface. Single-beat BSP are typically used to solve the inverse problem. However, the presence of noise and artifacts within the chosen beat can significantly distort the reconstructed EGM. This paper introduces a novel approach to mitigate the impact of noise and artifacts on ECGI solutions by generating single averaged BSP beats. Our approach automatically identifies BSP segments with minimum RR-variability and selects consecutive artifact-free beats using entropy-based criteria. We compare the proposed approach with alternatives and our results demonstrate the efficacy of this approach in reducing amplitude variability across selected beats before averaging, thereby enhancing the fidelity of reconstructed EGMs. Furthermore, our study establishes the optimal number of beats required during averaging. Specifically, we found that at least 10 beats are needed to compute ventricular repolarisation time, and 50 beats for atrial repolarisation time. The proposed method offers a more efficient, reliable, and objective approach alternative to manual beat selection in ECGI.