Personalized Modeling of Atrial Activation and P-waves: a Comparison Between Invasive and Non-invasive Cardiac Mapping

Biatrial personalized models incorporating functional and anatomical features are becoming a promising tool for planning therapy for patients with atrial fibrillation (AF). Conduction velocity (CV) is one of the main features to be matched during the process of functional personalization, as it can identify electrical abnormalities in the cardiac tissue. The spatial distribution of CV can be estimated from local activation times (LAT) maps from non-invasive electrocardiographic imaging (ECGI) or invasive electroanatomical mapping systems (EAMS). We investigated the effect of using either invasive LAT maps from EAMS or non-invasive LAT maps from ECGI to personalize two biatrial models by comparing the virtual P-waves obtained from these LAT maps with the measured P-waves from the surface electrocardiogram (ECG). For both modalities – ECGI and EAMS – we found a qualitative match between simulated and measured P-waves but observed quantitative differences. The root-mean-square error (RMSE) between measured and simulated signals for patient A was
${0.26}\pm 0.11\ mV$
and
$0.38\pm 0.31\ mV$
, while for patient B it was
$0.21\pm 0.09\ mV$
and
$0.14\pm 0.05\ mV$
for EAMS and ECGI, respectively. The correlation between measured and simulated signals from ECGI and EAMS was
${\it 0.69}\pm { 0.34}$
and
$0.71\pm 0.26$
for patient A and
$0.71\pm 0.18$
and
$0.72\pm 0.18$
for patient B. Our results suggest that LAT maps from ECGI and EAMS show differences, which are also reflected in the computed P-wave on the body surface.