Enhanced identification of atrial arrhythmic triggers using cardiac digital twins generated from imageless electrocardiographic imaging

Atrial arrhythmias, such as atrial fibrillation, are often triggered by ectopic beats. Conventional Electrocardiographic Imaging (ECGI) has proven effective for mapping ectopic foci on the epicardium; however, it relies on CT scans and often struggles to accurately identify the Earliest Activation Site (EAS), especially in the septal regions due to spatial constraints of ECGI.

This study aims to enhance the detection of the EAS for ectopic atrial beats by utilizing ECGI-Twins, generated through artificial intelligence to identify the cardiac digital twin that best matches the recorded imageless ECGI of a patient. This approach is designed to improve the accuracy of EAS detection, even in challenging regions such as the septum.

The ECGI-Twin technology was developed using a population of 17,220 cardiac digital twins generated from a set of 42 atrial and 17,220 torso geometries created through a statistical shape model. These geometries were used to simulate activations from 410 atrial stimulation points, covering a wide range of potential ectopic origins. To assess the accuracy of ECGI-Twin, the performance of both traditional ECGI and the new ECGI-Twin method was evaluated in localizing the EAS in a set of 2,870 simulated subjects, including stimulation points located in septal regions. The validation dataset was distinct from the training dataset, using different geometries and stimulation sites to ensure robust assessment. For each evaluated beat, the localization error was measured in centimetres by calculating the Euclidean distance between the actual stimulation site and the EAS position estimated by ECGI and ECGI-Twin.

Figure 1 illustrates an example where the ECGI-Twin approach more accurately replicates the propagation pattern and localizes the ectopic focus compared to traditional ECGI. The ECGI-only method showed a mean EAS localization error of 0.70 ± 0.48 cm, with 79.4% of ectopic beats being localized within 1 cm of the true origin (Figure 2). In contrast, the ECGI-Twin approach significantly improved performance, reducing the mean error to 0.46 ± 0.29 cm and achieving 95.6% accuracy within 1 cm (p<0.01). The improvement was especially pronounced for septal-originating ectopic beats, where the mean localization error decreased from 2.3 ± 1.03 cm with ECGI alone to 0.46 ± 0.24 cm using ECGI-Twin.

Integrating imageless ECGI with cardiac digital twins significantly enhances non-invasive EAS detection, particularly for septal-originating beats where conventional ECGI faces limitations. ECGI-Twins, capable of mapping from a single heartbeat, demonstrate strong potential for improving the precision of ablation procedures and identifying arrhythmogenic regions that standard mapping methods might overlook.

ECGI-Twin of a septal trigger