Characterization of atrial arrhythmias in body surface potential mapping: A computational study

PURPOSE Atrial tachycardia (AT), flutter (AFL) and fibrillation (AF) are very common cardiac arrhythmias and are driven by localized sources that can be ablation targets. Non-invasive body surface potential mapping (BSPM) can be useful for early diagnosis and ablation planning. We aimed to characterize and differentiate the arrhythmic mechanisms behind AT, AFL and AF from the BSPM perspective using basic features reflecting their electrophysiology. METHODS 19 simulations of 567-lead BSPMs were used to obtain dominant frequency (DF) maps and estimate the atrial driving frequencies using the highest DF (HDF). Regions with |DF-HDF|≤1Hz were segmented and characterized (size, area); the spatial distribution of the differences |DF-atrialHDFestimate| was qualitatively analyzed. Phase singularity points (SPs) were detected on maps generated with Hilbert transform after band-pass filtering around the HDF (±1Hz). Connected SPs along time (filaments) and their histogram (heatmaps) were used for rotational activity characterization (duration, spatiotemporal stability). Results were reproduced in clinical layouts (252 to 12 leads) and with different rotations and translations of the atria within the torso, and compared with the original 567-lead outcomes using structural similarity index (SSIM) between maps, sensitivity and precision in SP detection and direct feature comparison. Random forest and least-square based algorithms were used to classify the arrhythmias and their mechanisms' location, respectively, based on the obtained features. RESULTS Frequency and phase analyses revealed distinct behavior between arrhythmias. AT and AFL presented uniform DF maps with low variance, while AF maps were more heterogeneous. Lower differences from the atrial HDF regions correlated with the driver location. Rotational activity was most stable in AFL, followed by AT and AF. Features were robust to lower spatial resolution layouts and modifications in the atrial geometry; DF and heatmaps presented decreasing SSIM along the layouts. The classification of the arrhythmias and their mechanisms' location achieved balanced accuracy of 72.0% and 73.9%, respectively. CONCLUSION Non-invasive characterization of AT, AFL and AF based on realistic models highlights intrinsic differences between the arrhythmias, enhancing the BSPM utility as an auxiliary clinical tool.