MultiScale Human Respiratory System Simulations

1
Multi-Scale Human Respiratory System Simulations
to Study the Health Effects of Aging, Disease
and Inhaled Substances Robert F. Kunz, Daniel C.
Haworth, Gulkiz Dogan The Pennsylvania State
University Andres Kriete Drexel University 59th
Annual Meeting of the APS Division of Fluid
Dynamics 21 November 2006 Sponsorship NIH
Grant 5-R01-ES014483-02
2
Motivation

• Goal 1 Develop, couple, apply, validate imaging
  and physics modeling of resolvable and
  sub-resolvable scales in human respiration
• Goal 2 Provide clinical-time-scale (hours)
  diagnostic information for disease and injury
  assessment.
• Step 1 Develop semi-automated end-to-end medical
  image through CFD analysis toolkit
• Step 2 CFD modeling
• Unsteady
• Multiphase/species
• Full resolution for upper branches
• Q1D for lower branches
• Bulk modeling in respiratory units for volume,
  aerodynamic loss, deposition, gas exchange

3
Medical Imaging

• CT scans of diseased patients
• Clinical image standard 3D sgi TIFF, DICOM
  header
• Pixel dimension 0.6 x 0.6 mm, slice separation
  0.8 mm.
• Amira for image visualization and image
  processing
• MEVISLab for DICOM headers for voxel sizes

4
Lobe Segmentation

• Lobes needed for volume filling branching
  algorithm.
• Fissures separating lung lobes are visible by a
  proper color map ? segment by following
  features/curves within 2-D transverse slices
  follow anatomic features
• Volume and surface of each lobe are calculated
• Smoothing, island removal, hole filling
• STL export

Left
Right
5
Airway Tree Segmentation
81 year old male patient

• Automated region growing based algorithm
• Threshold, colormap and interpolation features
• Maximum generation to be segmented accurately
  depends on CT scan quality and the lungs
  physical character for particular patient.
• STL export

6
Thinning

• Skeleton of the airways using distance map based
  thinning algorithm ?automated using an Amira
  script
• Output node coordinates of skeleton, connectivity
  and local diameters along each segment.
• _at_ one minute on a PC.

7
Partitioning

• In-house code to automatically partition airway
  tree
• Generation number of surface elements defined
  using ADT search to skeleton topology
• Allows truncation of airway tree at desired
  location.
• Enables automated assignment of model attributes
  (grid resolution, deposition efficiency) at
  different generations in grid generator and CFD
  code

8
Truncation

• Same in-house code automatically truncates the
  airway trees and assigns the boundary conditions
  to truncated airways for fully-resolved 3D CFD
  calculations.

After truncation at 0.85 through 5th generation
Before truncation
9
Automated Octree Based Grid Generation

• HARPOON for grid generation of both lobe and
  airway trees. Fully automated (batch mode) and
  quite fast 1 million cells _at_ 1 minute on a PC.
• Hexahedral dominant Cartesian meshes
• Hanging nodes ? requires CFD code support for
  arbitrary polyhedra
• Grid resolution is adapted by assigning different
  surface cell sizes to different generations
• Prism layers as postprocessing step

10
Automated Octree Based Grid Generation
11
Lower Branch Modeling

• Another in-house code to reconstruct the
  tracheobronchial tree from each truncated airway
  down to respiratory units in each lobe.
• A volume-filling algorithm has been devised such
  that the resulting spatial distribution of
  respiratory units is statistically uniform within
  each lobe.

12
Lower Branch Modeling

• Branching algorithm reproduces geometric
  statistics (diameters, lengths, branching angles)
  of a human tracheobronchial tree.
• Typical tracheobronchial trees represent 22-23
  generations, contain 70,000 - 100,000 branches
  _at_ 1/2 are terminal branches ending at
  a respiratory unit.

13
Summary of Geometry Work
14
CFD Approach

• Upper branches resolved _at_ O(105 - 106) elements
• Lower branches modeled as Q1D pipe sections _at_
  O(105) elements
• Volume, mass flow rates conserved, friction
  factors modeled
• Flow losses in lower branches due to pipe wall
  shear modeled using Q1D wall functions
• Flow losses in lower branches due to
  curvature/branching modeled using classical loss
  factors
• Unstructured finite-volume CFD code arbitrary
  polyhedral element support
• Interphase coupling for Eulerian multiphase
  modeling (particle transport)
• Parallelized for rapid execution on cluster
  computers (2-8 hours)

15
CFD Approach

• Ensemble averaged n-field mass and momentum
  conservation laws in Cartesian tensor form

16
Results