COMPUTATIONAL LUNG MODELING

1
DISTRIBUTION OF AIRWAY CONSTRICTION IN ASTHMA
QUANTIFIED VIA IMAGE ASSISTED ANATOMIC MODELING
Nora T. Tgavalekos1, B. Suki1, J. Venegas2 and
K.R. Lutchen1 1Department of Biomedical
Engineering, Boston University, Boston, MA 2Dept.
of Anesthesia and Medicine , Mass General
Hospital, Boston, MA
COMPUTATIONAL LUNG MODELING
SENSITIVITY ANALYSIS
IMAGE ASSISTED STRUCTURE VS. FUNCTION
Mean

• Previous morphometric lung models have been
• used to explore the frequency dependence of RL
  and EL
• after homogeneous and heterogeneous broncho-
• constriction and ASM shortening are imposed on
  the airways.
• Models suggest a relationship between the
  pattern of
• constriction and the impact on mechanical
  function.
• Shapes are consistent with measured RL and EL
  in asthma.

SD
Figure 2 Sensitivity of the lung model to
random constriction patterns. Simulation were
all run with a mean (m) constriction of 50
and standard deviation (SD) of 70 applied to the
entire lung (dlt2mm.) Monte Carlo analysis of 20
runs showed that for any given random draw, the
change in RL and EL will be no more than 10 of
the mean.
Figure 5 The ventilation images were obtained
using Positron Emission Tomographic (PET)
scanning techniques. PET scans were taken on a
mild-moderate asthmatic at baseline (left) and
after a methacholine challenge (right) at a dose
equal to the patients PC10. Each set of
ventilation images contain 15 slices of the lung,
the top left is the apex of the lung and the
bottom right the base of the lung. The images
were taken in the cranio-caudal direction.
Regions of high ventilation correspond to light
colors and regions of low ventilation correspond
to dark colors. At baseline, ventilation is
homogenously distributed throughout the lung.
After a methacholine challenge there are dark
patches in the upper and dependent regions of the
lung. These regions have low ventilation, which
may be a consequence of gas trapping.
Figure 1 Simulations of RL and EL versus
frequency for baseline (black) and following
homogeneous and heterogeneous broncho-constriction
. Peripheral airways were constricted with a
m20 and CV50 (Heterogeneous) and m20,
CV0(Homogeneous).
These models are consistent with the
statistical features of lengths and diameters
in a real lung, however they do not provide a
mapping between specific airways in the tree and
specific locations in the lung.
c
GOAL
Figure 3 Impact of Localized Constriction on
Lung Mechanics. Simulations were performed after
imposing a heterogeneous constriction pattern
(m50 , SD70) on only specific regions of the
lung a) apex b) middle c) lower d) upper half of
the lung. On the right is the impact of dynamic
RL and EL for each of these cases. In all cases
affecting only a specific region has a relatively
minor impact on mechanics as opposed to affecting
the entire lung as shown in Figure 2. Color bar
in upper left corner describes the degree of
constriction in each branch i.e. white represents
airways diameters at baseline FRC and red
represents to airways that are closed or nearly
closed.
To advance 3D airway tree models to predict
function from structure particularly when
constriction patterns are imposed heterogeneously
on the tree in specific anatomic locations.
ADVANCING 3D LUNG MODELS
Kitaoka advanced algorithms to create
geometrically 3D space filling airway trees.
The algorithm utilizes a deterministic
space-filling algorithm to create a branching
structure which is physiologically consistent
in the amount of fluid delivered to each branch
and the spatial arrangement of branches within
the lung.
Figure 6 The mechanical impact of the lung model
at baseline and after applying different
constrictions on specific locations in the airway
tree model is shown with corresponding lung
cartoons above. There is very little impact on
mechanics when only the cranial-dorsal region is
severely constricted (a). It is not until the
remaining portions of the tree have mild
constriction when the mechanical response will
begin to changes dramatically. Also shown is a
general attempt to match subject specific
mechanical and PET measurements. By applying a
high degree of heterogeneity to only the upper
dorsal regions of the lung (regions shown by PET
to have under ventilation) (dlt2 mm.) and
constriction pattern with the same mean but lower
heterogeneity to the rest of the tree (dlt 2 mm.),
we are able to match the subjects mechanical
data and still remain consistent with the PET
images (b).
The airway tree has a total 50,400 branches.
There are 28 generations each with its own
distribution of diameters and lengths.
SUMMARY
We have advanced 3D models that are
morphometrically and anatomically consistent
with a real human lung and in which prediction
of lung function is possible Preliminary
analysis suggests that airway constriction in
asthma produces gross heterogeneity of
constriction with localized areas experiencing
closure and substantial constriction throughout
the lung periphery When synthesized with
imaging data sensitive to function, these models
may be used to provide new insights into
structure-function relationships in healthy
versus asthmatic lungs

• The mechanical response of a single airway can
  be modeled with a transmission line model
  with resistive and inertial forces. The airways
  terminate into alveoli tissue elements, which
  have viscoelastic properties.

Figure 4 Impact of Variable Constriction
Pattern. Here we examined varying degrees of
heterogeneity and mean constriction while
impacting the entire lung. Baseline is in black.
In Case A, the mean and degree of constriction
decreases as one moves from the apex to the base
(apex m50, SD70 , middle m40 SD40 ,
and base m30, SD30). In Case B the mean
constriction remains the same but the SD
decreases as moving from the apex to the base
(apex m50, SD70 , middle m50 SD40 ,
and base m50, SD30). In order to get a high
degree of heterogeneity in function a significant
fraction of the lung must have a high mean and
spread in constriction