1. Presented by Biomedical Modeling and Simulation Richard C. Ward Modeling and Simulation Group Computational Sciences and Engineering Division Research supported by the Department of Energy’s Office of Science Office of Advanced Scientific Computing Research

2. 2 Ward_BioModelSim_0611 Biomedical modeling and simulation at ORNL  Three-dimensional organ and tissue modeling using CT or other imagery (pulmonary, arterial, musculoskeletal)  Integration of models at multiple temporal and spatial scales  Biokinetic and biotransport modeling  Prediction of outcomes based on biomedical models  Computational environments (data repositories, search tools, visualization, etc.) in support of biomedical and medical applications  Design of middleware to address interoperability

3. Geometry models using imaging data 3 Ward_BioModelSim_0611 X-ray CT data (example: National Library of Medicine Visible Human) NURBS (nonuniform rational B-spline) model from visible human CT data Finite element analysis (FEA) from NURBS

4. 4 Ward_BioModelSim_0611 CT scans used to construct geometrical model of AAA Numerical simulations give wall mechanical stress distribution Models predict AAA rupture site from stress distribution CT scans used to construct geometrical model of AAA Numerical simulations give wall mechanical stress distribution Models predict AAA rupture site from stress distribution Vascular systems modeling: Predicting rupture of abdominal aortic aneurysm Collaboration with University of Tennessee Medical Center Department of Surgery and Vascular Research Laboratory

5. Hyperelastic model of AAA modifies stress analysis Produces higher stress concentrations at same location 5 Ward_BioModelSim_0611 Hyperelastic: 0.61 N/cm 2 Linear elastic: 0.49 N/cm 2

6. 6 Ward_BioModelSim_0611 Using high-performance computing resources for pulmonary flow modeling  Finite element problem-solving environment  Computational fluid dynamics  Fluid-structure interactions  Equation formulator  Java GUI on user’s desktop computer  Automatic mesh partitioning  Computations routed to high- performance computer using NetSolve  Results returned to user’s desktop computer  Links to client-server visualization software  Automated archiving of scientific data sets Collaboration with A.J. Baker, UT, and Shawn Ericson, UT/ORNL JICS

7. 7 Ward_BioModelSim_0611 Deposit of particulates related to complexity of flow revealed Rotational flow in airways visualized Comen, Kleinstreuer, and Zhang ( J Fluid Mech, 435, pp. 25-52, 2001) Airway model

8. 8 Ward_BioModelSim_0611 Species pulmonary flow modeling PICMSS (Parallel Interoperable Mechanics System Simulator) used to generate species flow using the airway model Comen, Kleinstreuer, and Zhang ( J Fluid Mech, 435, pp. 25-52, 2001) Airway model Image courtesy of Shawn Ericson, JICS

9. 9 Ward_BioModelSim_0611 Cardiovascular modeling environments High-performance computing resources ConnectIntegrate Models Computations Visualization Predictions

10. 10 Ward_BioModelSim_0611 Modeling toxic exposure: Inhalation of Hg vapor Model developed by R. W. Leggett, K. F. Eckerman, and N. B. Munro Life Sciences Division Promptly exhaled Hg 0 Promptly exhaled Hg 0 Hg 0 exhaled after conversion from Hg++ Respiratory tract model Red blood cells Brain Long-term Other Long-term Liver Long-term Kidneys Long-term Plasma Hg 0 Diffusible Non- diffusible Urinary bladder UrineFeces GI tract model

11. 11 Ward_BioModelSim_0611 Goal: Predict migration of smooth muscle cells from media to intima due to inflammatory response after injury Model for predicting vascular disease Predictive multiscale modeling  Spatial modeling of cell migration  Kinetic modeling of biochemicals  Result: A multi-scale hybrid continuous-discrete predictive model for tissue pathology Atherosclerotic artery MMP3 proMMP9 TIMP1 proMMP9 MMP3 proMMP9 TIMP1 TIMP3 TIMP2 MMP9 Collagen IV MMP3 Inhibited Activation of MMP-9 Inhibition of MMP-9 MMP-9-induced collagenolysis ACTIVE Matrix metalloproteinases (MMPs)

12. 12 Ward_BioModelSim_0611 Support provided by Defense Advanced Research Projects Agency (DARPA) Program Manager: Rick Satava Virtual Soldier Project 12 Ward_BioModelSim_0611

13. Post-wounding Preparation ORNL contributes to DARPA Virtual Soldier Build computer model of “generic” patient Store records on “dog tags” Post-wounding information Pre-wounding information Computer model provides total informational awareness for forward medical team 13 Ward_BioModelSim_0611 Assemble detailed individual medical records Use pre- and post-wounding individual data to create predictive model of specific patient ORNL involved

14. High-level integrative physiological models Computations performed by University of Washington Cardiovascular/ pulmonary flow Circuit models describe blood flow and arterial and venous pressures Airway mechanics + - - + System circulation Four-Chamber heart model Pulmonary system 14 Ward_BioModelSim_0611

15. Finite-element heart simulations  Computations combine biomechanical, electrophysiology, and biochemistry models  Simulations conducted on two 105-node dual Opteron Dell Linux clusters  Typically used only up to 32 nodes per simulation  Overall, obtained substantial speedups by combining new algorithms and high-performance computing  Used pre-computation and interpolation to allow team to develop real-time models for 2 h worth of heartbeats 15 Ward_BioModelSim_0611 Conducted by Andrew McCulloch’s Cardiac Mechanics Research Group (University of California in San Diego)

16. 16 Ward_BioModelSim_0611 Computational speed up for finite-element simulations 2002200320042005200620072008 Year 10 -6 10 -5 10 -4 10 -3 10 -2 10 -1 10 -0 Computational Speed (beats/second) 300 MHz SGI Origin 2100 2 ODE model 1 CPU 833 MHz Pentium 3 2 ODE model 1 CPU 2.0 GHz Pentium 4 21 ODE model 1 CPU 2.3 GHz Pentium 4 21 ODE model 16 dual CPU nodes of Linux cluster 2.3 GHz Pentium 4 76 ODE model 96 dual CPU nodes of Linux cluster 78 hours/beat 10 minutes/beat Data courtesy of the Cardiac Mechanics Research Group, UCSD

17. 17 Ward_BioModelSim_0611 ORNL developed middleware architecture WS = Web services Prediction software Prediction software Wound trajectory database 3D segmented anatomy model Experimental data Data repository Simulation Results Taxonomy Results Ontology VSP middleware An early plan WS

18. ORNL HotBox integrates all the DARPA Virtual Soldier windows HotBox interface Anatomical ontology: Foundational model of anatomy Anatomical ontology: Foundational model of anatomy Predicted location of wound SCIRun Net Physiology display Geometry window with thorax model 18 Ward_BioModelSim_0611

19. ORNL solves biomedical problems  Convert CT slice data to finite-element mesh  Abdominal aneurysms  Prediction of wounds  Data repositories  Parallel computations  Computational tools for toxicants  Agent technologies  Ontologies and informatics 19 Ward_BioModelSim_0611

20. 20 Ward_BioModelSim_0611 Contacts Barbara Beckerman Program Manager, Biomedical Engineering Computational Sciences and Engineering Division (865) 576-2681 beckermanbg@ornl.gov Richard Ward Senior Research Scientist Computational Sciences and Engineering Division (865) 574-5449 wardrc@ornl.gov 20 Ward_BioModelSim_0611