Page 211 - Artificial Intelligence for Computational Modeling of the Heart

184 Chapter 6 Additional clinical applications

Importantly, it has been shown that the position of the electrode
can significantly impact response to the therapy [391].
Personalized computational models have been used to com-
plement clinical and observational studies to derive a deeper un-
derstanding of the disease and the effect of therapy [9]. For in-
stance, Crozier et al. suggested that the location of the leads of
the CRT device has a more significant impact than the presence
of conduction blocks [407]. In [408], the authors concluded that
the optimal left ventricular (LV) pacing site in left bundle branch
block (LBBB) cases without scar should be most distant from the
right ventricular (RV) lead. They further hypothesized that in cases
with scar the LV lead needs to be placed remotely from both RV
lead and scar. Willemen et al. validated in canine experiments the
observations from computational models, indicating a different
response of RV and LV function to changes in interventricular (VV)
pacing delay [409]. In another study, Lee et al. used modeling to
explore the long term effects of atrioventricular (AV) and VV de-
lay optimization, suggesting decreased benefits over time [410].
Niederer et al. compared the benefits of multi vs single site LV pac-
ing on virtual hearts, pointing out potential advantages in cases
with posterolateral scar [411]. Finally, Costa et al. traced back the
increased risk of VT in cases with scar to observations of increased
repolarization dispersion around scar tissue [412], a situation that
is becoming more and more noticeable clinically.
Such multi-scale models of the heart could ultimately repre-
sent a powerful tool for decision support, by enabling a physi-
cian to perform virtual CRT beforehand and assess the chances
of response a patient has. Towards this aim, a cardiac model-
ing pipeline for CRT guidance based on pre-operative and non-
invasive data is presented. Early results on a small cohort of pa-
tients are reported.

6.1.2 Methods
6.1.2.1 Data acquisition
The study comprised ten patients of age 18 and older that suf-
fered from either dilated (90%) or ischemic (10%) cardiomyopathy
and received CRT treatment following international recommen-
dations at the Bordeaux University Hospital between January 2016
and October 2017 [413]. For each patient the following data acqui-
sition protocol was applied.
At baseline, a 12-lead electrocardiogram (ECG) was recorded.
Siemens Magnetom Aera with 1.5 Tesla (Erlangen, Germany) was
used to acquire 2-chamber, 3-chamber, 4-chamber, and short axis
cine images as well as delayed enhancement sequences, which