Posicionamento de eletrodos em gêmeos digitais cardíacos: desenvolvimento e avaliação de um novo software

Abstract

The generation of electrocardiograms (ECGs) from personalized computational models has emerged as a promising tool for investigating cardiac electrophysiology. However, its accuracy critically depends on the correct placement of electrodes on the patient's torso. This work presents the development of a new software tool for mannal electrode positioning on patient-specific models and performs a comparative evaluation between this approach and an automatic method based on deformable registration and statistical torso models. Four patients with magnetic resonance imaging (MRI) data were analyzed, one of whom also had computed tomography (CT) imaging, enabling the realistic inclusion of bony structures. For each patient, the positions of the precordial electrodes were obtained using both methods, and pseudo-ECGs were subsequently simulated using the MonoAlg3D cardiac simulator. Quantitative evaluation included the spatial distance between corresponding electrodes, the Pearson correlation coefficient (PCC), and the relative root mean square error (rRMSE) between the simulated signals. The results indicate that electrodes V1-V4 exhibit small spatial discrepancies between methods, whereas V5 and V6 show larger variations, reaching up to 8em in some patients. Despite these geometric differences, the pseudo-ECGs maintained a high degree of morphological similarity, with PCC values frequently exceeding 0.999. The most relevant differences were observed in signal amplitude, reflected by higher rRMSE values in the lateral leads, highlighting the increased sensitivity of these regions to electrode positioning errors. The inclusion of CT data reduced geometric uncertainties and yielded the most consistent signals between methods, emphasizing the importance of combining multiple imaging modalities to construct higher-fidelity digital cardiac twins. Study limitations include the lack of a complete body field of view for marking limb leads, the low resolution of MRI data for accurate bone segmentation, and the small sample size. Nevertheless, the results demonstrate that both approaches produce clinically coherent pseudo-ECGs, while underscoring the need for improved geometric precision in lateral regions and validation in a larger cohort. Future work will focus on incorporating additional imaging modalities, validating electrode positions against real ECG recordings, and investigating the impact of these positioning errors on the electrocardiographic inverse problem.