ECG Imaging (ECGI) provides non-invasive, single beat panoramic assessment of cardiac electrophysiological parameters, which makes it a promising tool in different clinical settings. Reproducibility is a key aspect for any methodology with potential for clinical translation, but it remains undetermined for ECGI. This study aimed to assess clinically significant ECGI reproducibility impacting ventricular epicardial mapping.
Ten (n=10) patients underwent ECGI during left-ventricular epicardial pacing delivered from cardiac resynchronization therapy devices. Two experts performed ECGI (CardioInsight, Medtronic, MN), using the same cardiac computed tomography and body-surface ECG recordings (matched QRS complexes), but independently performing semi-automatic cardiac segmentation and identification of 252 body-surface electrodes. The closest epicardial sites on the two cardiac geometries were paired. Similarity of reconstructed unipolar electrograms (UEG) was assessed measuring the pairwise correlation coefficient between them (rUEG) and the absolute difference of their peak-to-peak amplitudes (|ΔAUEG|). Agreement between ECGI activation time (AT) maps was assessed using the correlation coefficient (rAT) and absolute percentage difference (|ΔAT|).
The median number of ECGI epicardial sites was 1828 (1768, 1929) and 2028 (1812, 2447) for operator 1 and 2, respectively, with median number of paired sites equal to 1790 (1762, 1906). The median and maximum (95%) distance between them was 2.9 (2.5, 3.1) mm and 10.0 (8.6, 11.1) mm, respectively. Reconstructed UEG were similar, with median rUEG= 0.97 (0.94, 0.98) and median |ΔAUEG| = 10.3% (8.5%, 12.5%). Local AT were also similar, with rAT = 0.85 (0.74, 0.88) and median |ΔAT| = 2.5 (2.0, 3.0) ms, while the maximum (95%) |ΔAT| = 40.5 (27.6, 62.0) ms.
In conclusion, the morphology of UEGs was not significantly impacted by inter-operator variability in cardiac segmentation and electrode identification, but local AT may differ in a small number of cardiac sites.