Integrating Cardiac and Cardiovascular Simulations Using the HLA

1
Integrating Cardiac and Cardiovascular
Simulations Using the HLA

• 2002 Spring SIW
• Presented bySean Patrick Murphy

The work presented herein was supported by the
National Aeronautics and Space Administration
(NASA) through the National Space Biomedical
Research Institute (NSBRI) under NASA Cooperative
Agreement NCC9-58.
2
Project Participants (Phase 1)

• Principal Investigator
• James E. Coolahan, Ph.D. (JHU/APL)
• Collaborators / Participants
• JHU/APL
• Andrew Feldman, Ph.D. Cardiac modeling
• Robert R. Lutz Distributed sim design
• Sean P. Murphy Sim design development
• Randy Saunders Sim development
• Tim Gion Visualization
• Harvard/MIT
• Rama Mukkamala, Ph.D. Cardiovascular system
  modeling

3
The National Space Biomedical Research Institute
(NSBRI)
Organization Consortium of 12 institutions,
founded in 1997, funded by NASA.
Primary Mission Objective To ensure safe and
productive human exploration and development of
space beyond Earth orbit. See http//www.nsbri.or
g for more information.

• Research Teams
• Bone Loss
• Cardiovascular Alterations
• Human Performance Factors, Sleep and
  Chronobiology
• Immunology, Infection and Hematology
• Integrated Human Function
• Muscle Alterations and Atrophy
• Neurobehavioral and Psychosocial Factors
• Neurovestibular Adaptation
• Nutrition, Physical Fitness, and Rehabilitation
• Radiation Effects
• Smart Medical Systems
• Technology Development

• Consortium Members
• Baylor College of Medicine (lead)
• Brookhaven National Laboratory
• Harvard Medical School
• The JHU School of Medicine and APL
• MIT
• Morehouse School of Medicine
• Mount Sinai School of Medicine
• Rice University
• Texas AM University
• University of Arkansas for Medical Sciences
• University of Pennsylvania Health System
• University of Washington

4
NSBRI Integrated Human Function Long-Term Vision
A Critical Challenge for the Space Biomedical
Community
Intent Development of a digital human will
allow various countermeasures designed for
solving health problems in space to be evaluated
for their impact on the body. Integration will
allow researchers to
Integrated model of function from molecules
to cells to organs to humans
V I S I O N

• Predict potential problems
• Simulate health conditions
• Plan adequate responses for long-duration missions

5
Integrated Human FunctionDuring Long-Duration
Space Flight
Spacecraft Natural Environment Temperature Humidit
y Barometric pressure Atmospheric
composition Ambient light Microgravity

• Threats
• Radiation
• Coronal mass ejections
• Cosmic rays
• Airborne particulates
• Pathogens
• Airborne
• Waterborne
• Foodborne
• Wounds
• Decompression

The Human Body in Confined Space Flight in
Microgravity
Psychosocial factors
Neurovestibular adaptation
Circadian rhythms
Immunology / hematology
Other Issues Biomass (food) production Waste
management Air quality maintenance
Cardiovascular alterations
Countermeasures Shielding Exercise Artificial
gravity Pharmacology (Smart) medical care
Nutrition
Muscle atrophy
Bone loss
The System
6
Comparison of MS Levels
Warfighting MS Pyramid
Human Body MS Pyramid
PATRIOT-centric example
Cardiac-centric example
Human
Whole body
Gulf War
Campaign
Cardio-vascular
Air defense
System
Mission
Heart
Organ
Missile intercept
Engagement
Cell
Myocyte
Terminal guidance
Engineering
Ca
Molecule
7
HLA Advantagesin the Scientific Arena

• Simulation interoperation for synergistic results
• Increased interaction among research groups
• Localization of knowledge
• Reusable simulations
• Protection of intellectual property
• Inexpensive simulation performance increases
• Existing process model to aid interoperable
  simulation development

8
FEDEP Step 1 Federation Objectives CVVS
Federation Objectives

• Demonstrate the applicability of distributed
  simulation and HLA to the biomedical community
• Target pertinent NSBRI question
• Post-Flight Orthostatic Intolerance (PFOT)
• RCVSIM
• Arrhythmias

9
FEDEP Step 2 Federation Conceptual Model
Research Cardiovascular SIMulator

• Cardiovascular system modeled as a circuit
• Current Blood Flow
• Voltage Pressure Difference

Pulmonary Circulation
Left Ventricle
Right Ventricle
Systemic Circulation
10
FEDEP Step 2 Federation Conceptual Model
Contemporary Heart Modeling

• Current Paradigm
• Model ion fluxes through individual cardiac
  myocytes with a system of stiff ODEs
• Model current exchanges between myocytes based
  off of complex cardiac fibre orientation
• Model entire heart with approximately 1 million
  cells

Requires Enormous Computational Resources!
11
FEDEP Step 2 Federation Conceptual ModelA
Cardiac Electrical Activity Simulator
Depolarization (Cellular Automata model)
? In 2D, the tissue is represented as a regular
grid of square elements each with a randomized
lattice location (dots). ? Elements interact
with neighbors located inside an elliptical
region whose major axis parallels the local fiber
direction. ? The central element enters the
excited state when the current sourced by
excited elements in its neighborhood exceeds a
threshold.
Repolarization (Partial Differential Equation
Model)
Central Element
The action potential shape is determined by an
analytic function allowing for the exact
specification of the space-clamped action
potential durations (APDs). Also, this permits
the analytic specification of the steady-state
membrane I-V relation that will yield the desired
APD. APDs smoothing required in spatially
heterogeneous heart tissue is realized when the
monodomain (generalized cable equation) partial
differential equation representing the cardiac
syncytium is solved using a Crank-Nicholson
scheme.
Major benefit Computation time per beat is
decreased many orders of magnitude compared to
conventional approaches
12
FEDEP Step 2 Federation Conceptual ModelHCA
Model Parameters
Model implements the major characteristic
relations governing the propagation of cardiac
APs and can simulate cardiac electrical activity
under both normal and abnormal conditions.
A) Dispersion Relation (Conduction velocity
restitution) Plane wave propagation speed
versus local diastolic interval. B) Curvature
Relation Propagation speed versus wave
front curvature (a measure of the wave fronts
convexity or concavity) in resting tissue. C)
Action Potential Duration Restitution Dynamics
AP duration versus prior diastolic interval and
prior history of inter-beat intervals. D) Local
Anisotropy Ratios Local ratio of fast axis
to slow axis plane wave speeds.
Relations may be obtained directly from
experimental measurements of from a detailed
ion-channel-based model of cardiac myocytes.
HCA Simulation results for Relations (A) and (B)
when the models parameter values are determined
based on a Luo-Rudy-myocyte model of cardiac
tissue.
13
FEDEP Step 2 Federation Conceptual ModelHCA
Simulation of Acute Ischemia
Simulation of Action Potential Propagation in
Tissue with Acute Ischemia
Slowed Conduction and Action Potential Duration
Shortening
Post-Repolarization Refractoriness in Ischemic
Zone
Simulation of Infarction (necrosis) and
Macro-reentrant Arrhythmia
Unexcitable HCA elements (necrosis) surrounded by
a border zone of ischemic elements.
14
FEDEP Step 2 Federation Conceptual ModelThe
HCA Ventricle Simulator

• Hybrid Cellular Automata ventricle simulator
• Accurately models cardiac electrical waves
• Ventricle is modeled by a 700x700 2-D grid of HCA
  elements
• Grid topologically arranged as a cylinder
• Mechanical force generated by individual ring
  computed using Laplaces Law

Requires Modest Computational Resources!
15
FEDEP Step 3 Design FederationCVVS Conceptual
Overview
HCA Left Ventricle
HCA Right Ventricle
Systemic Capillary Bed
16
FEDEP Step 3 Design FederationCVVS Federation
Data Flow
Input Data File
LeftVentricle Federate
700 x 700 Array Vector Potential
Differences 700 Radii of Concentric Rings
Compliance dCdt
EDC ESC Intraventricular Pressure
Research CardioVascularSIMulator
ECG Federate
EDC ESC Intraventricular Pressure
700 x 700 Array Vector Potential
Differences 700 Radii of Concentric Rings
Compliance dCdt
Right Ventricle Federate
Input Data File
17
FEDEP Step 3 Design FederationComputer
Architecture
100 Mb Ethernet STD Subnet 14

• Dual P4, 1.7 GHz
• 2 GB, 400 MHz RDRAM
• Windows 2000

Right Ventricle Federate

• P3, 933 MHz
• 1.5 GB, 133 MHz SDRAM
• Red Hat Linux 7.1

• Dual P4, 1.7 GHz
• 2 GB, 400 MHz RDRAM
• Linux 7.1 SMP

RTI Exec
Left Ventricle Federate
RCVSIM Federate
18
FEDEP Step 4 Federation DevelopmentFederate
Software Architecture
Federate Wrapper Functionality

• The Local RTI Ambassador object provides
• An interface to the federation through RTI calls
  
• Simulation control flags
• Simulation state data

• Instantiate
• Local RTI Ambassador
• Federate Ambassador
• Simulation Interface
• Initialize Federation
• Run Simulation Loop
• Shut Down Federation
• Clean Up Variables

The Federate Ambassador object allows the
federation to interface with the federate.
The Simulation Interface provides a means of
transferring data and communication between the
original simulation and the rest of the
federation.

• Destroy/Exit Federation Execution
• Unpublish and Unsubscribe
• Delete Local Objects

• Create/Join Federation Execution
• Publish and Subscribe
• Register Local Objects

19
FEDEP Step 4 Federation DevelopmentIncorporati
ng Matlab-Based Federate
Compiled using EGCS
Compiled using MCC GCC
Dummy Matlab Functions
RCVSIM Federate Wrapper
Matlab Simulation
Shared Memory
C Interface Functions
Local RTI Ambassador
Output
Simulation IO Buffers
mxGetScalar()
RTI
mlfDoubleMatrix()
Input
Federate Ambassador
20
FEDEP Step 5 Test FederationBenchmarks I
Time to Complete 500 ms of Simulation
Time (minutes)
(a)
(b)
(c)
Integrated Legacy Simulation
CVVSFederation (DMSO RTI 1.3NG V4)
CVVSFederation (MAK RTI 1.3.5-ngc)
21
FEDEP Step 5 Test FederationBenchmarks II
Time to Complete 500 ms of FEDEX
Time (minutes)
No ECG Data
100 x 100 Elements
300 x 300 Elements
500 x 500 Elements
700 x 700 Elements
22
FEDEP Step 5 Test FederationBenchmark
Conclusions

• MAK RTI offers higher performance for this
  application
• Faster network would be most expedient method to
  decrease FEDEX times
• Data requirements of ECG federate greatly
  impacted federation execution times
• Temporal subsampling
• Spatial subsampling
• Entirely Linux-based federation offers greater
  performance (minus hardware constraints)

23
FEDEP Step 6 Federation Execution
ResultsHemodynamics During VT
Simulated ECG
Arterial Pressure LV Pressure
24
FEDEP Step 6 Federation Execution
ResultsRegulatory Responses to 3 Hz VT
Arterial Pressure
No Pulse During VT
Heart Rate Increases (but HCA heart model
ignores/blocks sinus beats during VT)
Venous Dead Volume Decreases
25
FEDEP Step 6 Federation Execution Results
Feedback between Federates
RCVSIM
End-Systolic Compliance
HCA Heart Model
Realized LV Elastance
26
FEDEP Lessons Learned

• FEDEP proved effective for CVVS Federation
  development
• Specific federates to be used in biomedical
  federations will typically be chosen in Step 1,
  not Step 3
• Development of federation conceptual model
  constrained by federate availability
• Lower fidelity medical federates can help reduce
  computational demands of biomedical federations
• Incorporating very diverse legacy simulations
  into biomedical federations will continue to be
  an obstacle
• Firewalls may present a problem when
  incorporating federates from disparate networks
  into a federation.

27
Potential Integration Pathways with other NSBRI
Teams
High Fidelity 3D electrical-mechanical cardiac
muscle models McCulloch, Winslow, Coolahan,
Detailed cardiac myocyte models Bers, Winslow
Analysis of effects of cardiac electrical and
mechanical alterations in microgravity
RCVSIM HCA Heart and Circulation Model (20
beats/hour on Pentium desktop computers)
Action potential characteristics Length-Tension
characteristics of cardiac muscle fibers
Blood Volume
Salt and Water Balance
Analysis of slow adaptation to microgravity
Analysis of the effects of muscle deconditioning
on cardiovascular function
Total Peripheral Resistance
Renin-Angiotensin Aldosterone System Mark,
Williams, Coolahan
Models of muscle mechanics and
energetics Kushmeric, Chase, Cabrera
Autoregulation Matching of local blood flow to
meet local metabolic demand (models of Groebe)
This collaboration is being explored
28
Summary

• Demonstrated applicability of HLA to biomedical
  simulations
• Integrated a simulation coded in MATLAB into an
  HLA federation
• Optimized CVVS Federation using data from simple
  benchmark FEDEXes
• Developed a meso-scale federation simulating
  cardiac-cardiovascular function
• Simulated system level response to a ventricular
  tachycardia