ARUP; ISBN: 978-0-9562121-5-3 - CMBBE 2012 - Cardiff University

catheterization mean measurements (e.g. [4, 8, 9]). In [10], a multistep procedure has
been devised to construct a lumped-parameter model representative of the right and left
pulmonary trees based on measured flow waveforms at the inlet and two outlets and
pressure data at a location inside the 3D domain and in the left atrium. In [11], an
involved multiscale automatic tuning procedure has been constructed to match typical
pressure waveform features at the abdominal aortic inlet and features from the measured
infrarenal flow waveform. In [12], a simple automatic tuning algorithm has been
proposed to construct Windkessel models to represent the pulmonary tree downstream
of each pulmonary artery outlet, based on the measured inlet flow data, pulmonary flow
split (fig. 1) and transpulmonary pressure gradient, and on morphometric data. It was
further assumed that on average flow to each small outlet was proportional to its crosssection
with a power law. This is the method used for the results section.
However, such algorithms have to be devised depending on where the measurements
are done and “trusted” [13]. When a surrogate of a pressure measurement is done (such
as wedge pressures), then to localize the pressure measurement in the 3D model is not
straightforward. In such cases, boundary conditions have to be constructed in coherence
with these data. Moreover, if a first subset of measured hemodynamics data gives rise
for example to a simulated pressure loss in the 3D model that is incoherent with the
other measured data, then an iterative process between clinicians and modelers is
necessary to devise boundary conditions that are representative of the patient’s state.
4. RESULTS
Patient-specific 3D model of the PAs and SVC were constructed from MRI data on five
pre-stage 3 (Glenn) patients (fig. 1). Velocity was prescribed at the inlet based on the
measured flow and Windkessel outlet boundary conditions were constructed as
described in the methods section [12]. The resulting total left and right resistances were
not the same (e.g 1658*10 5 kg.s -1 .m -4 vs 671*10 5 kg.s -1 .m -4 for the last patient of fig.1).
The resistances were even more different at the all the outlets [14]. 3D pulsatile
simulations resulted in low wall shear stress (potentially deleterious for these patients)
and complex pressure waveform at the SVC which dampens after the anastomosis.
Although the five patients anatomies varied largely, energy loss correlated strongly with
cardiac index (linear correlation, R 2 =0.94), and was in low (a few mW). The complex
flow observed until the first bifurcations can explain why imaging flow in such
locations is challenging (fig. 1). The results confirmed that pressure measurements
would need to be much more accurate than currently to be prescribed at the different
outlets, since pressure losses in the 3D domain were around 1mmHg. We highlighted
common features among the five patients, but the results demonstrated clearly
differences in the main indicators presented above and in 3D maps (fig. 1), motivating
the search for virtual surgical designs specific to each patient.
Subsequently, for the five patients above, virtual surgical designs of the stage 3 Fontan
palliation were explored [14]: the artificial graft sutured on one end to the IVC can be
attached to the Glenn anastomosis at its other end in different ways. Typical “Tjunctions”,
or with an offset towards the left lung, and several non-conventional “Yjunctions”
were compared in terms of energy, wall shear stress, venous pressure and
hepatic factors distribution between the right and left lungs (linked to pulmonary
arteriovenous malformations) – see fig. 1. Overall the “Y” design led to better
performances, but design should be customized for individual patients. In fact, for each