Translational Research: Multi-Scale Models of the Pulmonary Circulation in Health and Disease - Multiscale Computer Modeling in Biomechanics and Biomedical Engineering - page 261

Fig. 2 The generation of an anatomically based model of the pulmonary airways and
vasculature. From left to right the figure shows: a single high resolution CT image of a lung
in the supine posture; rendering of volumetric MDCT of the lung and automatic placement of
surface points on each lobe; initial arteries (red), veins (blue), and airways (white) used for model
generation; volume-filling airways (right lung) and blood vessels (left lung)
branching structure in perfusion distribution and ventilation-perfusion ( V/ Q)
matching [ 12 - 16 ]. In particular, the fractal model raised questions regarding the
significance of gravity (e.g. West's zonal model [ 11 ]) on function [ 15 , 16 ]. With
this debate in mind Burrowes et al. constructed an anatomically-based finite ele-
ment model of the pulmonary circulation, based on MDCT (multi-detector com-
puted tomography) imaging which aimed to capture individual lung shape and
blood vessel distribution [ 22 ].
The anatomically-based model of the pulmonary vasculature followed the spirit
of previous models of the pulmonary airways developed by Tawhai et al. [ 47 , 48 ].
An illustration of the generation of geometric models of the pulmonary vasculature
and airways is given in Fig. 2 . In each geometric model (airway or vascular), the
lungs or lobes and the pulmonary trunk (central blood vessels and airways) are
segmented from imaging data (in this case MDCT) using image-processing soft-
ware. The segmented airways and blood vessels are represented in a finite element
mesh by vectors describing their centerlines, and spatial coordinates for the
location of branching points. The distal airways and their accompanying blood
vessels are then generated via a volume-filling branching (VFB) algorithm [ 48 ],
which generates individual representations of airways and blood vessels to a user-
defined level, which is typically the level of the pulmonary acinus (a gas exchange
unit). The lung or lobe volume that is segmented from the imaging data acts as a
'host volume', which is filled with a uniform grid of seed points that represent the
location of the approximately 32,000 acinar units in the lung. Then, the following
steps are repeated until there are no seed points remaining in the set:
1. Seed points are grouped into sets with each seed point being associated with the
closest terminal (parent) vessel in the current tree.
2. The vector in the direction of the parent vessel and the coordinates of the center
of mass of the seed points associated with it are used to define a splitting plane.