Committee on Credible Practice of Modeling & Simulation in Healthcare https://simtk.org/home/cpms COMMON PRACTICE GUIDELINES: A SIGNIFICANT GAP IN COMPUTATIONAL MODELING AND SIMULATION IN HEALTHCARE https://simtk.org/home/cpms Can’t hear us? Select Audio -> Integrated VoIP -> Join Conference Presenter: Jeff Webb Moderator: Joy Ku
32
Embed
Common Practice Guidelines: A Significant Gap in Computational Modeling and Simulation in Healthcare
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Committee on Credible Practice of Modeling & Simulation in Healthcarehttps://simtk.org/home/cpms
COMMON PRACTICE GUIDELINES: A SIGNIFICANT GAP IN COMPUTATIONAL MODELING AND SIMULATION IN HEALTHCARE
• Program Background• Users, Integrators, and Collaborators• Implementation and Deployment• Keys to Success• Conclusion
Agenda
Cop
yrig
ht 2
015.
All
right
s re
serv
ed. A
pplie
d R
esea
rch
Ass
ocia
tes,
Inc.
7
PROGRAM BACKGROUND
Cop
yrig
ht 2
015.
All
right
s re
serv
ed. A
pplie
d R
esea
rch
Ass
ocia
tes,
Inc.
8
• Organization: Applied Research Associates, Inc. (ARA)
• Telemedicine & Advanced Technology Research Center (TATRC) Award #: W81XWH-13-2-0068
• Principal Investigator: Mr. Jeff Webb
• Amount: $6,959,593
• Period of Performance: Sept 2013 – Sept 2018
• Disclaimer: This work is supported by the US Army Medical Research and Materiel Command. The views, opinions and/or findings contained in this report are those of the author(s) and should not be construed as an official Department of the Army position, policy, or decision unless so designated by other documentation.
Project Information
Cop
yrig
ht 2
015.
All
right
s re
serv
ed. A
pplie
d R
esea
rch
Ass
ocia
tes,
Inc.
9
High Level Objectives• Create a publicly available physiology research platform that
enables accurate and consistent simulated physiology across training applications
• Lower the barrier to create medical training content• Engage the community to develop and extend physiology models• Meet the training needs of the military• Expand the body of knowledge regarding the use of simulated
physiology for medical education
Cop
yrig
ht 2
015.
All
right
s re
serv
ed. A
pplie
d R
esea
rch
Ass
ocia
tes,
Inc.
10
• Powerful backbone for applications• Powers immersive training content
and other M&S tools• Extensible standardized data model
with leverageable tools• Patient variability focused on
homeostasis with conditions, insults, and interventions for pathophysiology
What BioGears IS What BioGears IS NOT• Not a game or fancy application• Not (currently) a predictive model• Not (currently) designed to
directly ingest patient records
BioGears Conceptual System Architecture
BioGears Powered Applications
Cop
yrig
ht 2
015.
All
right
s re
serv
ed. A
pplie
d R
esea
rch
Ass
ocia
tes,
Inc.
11
The Team
Past Contributors:Rachel Clipp, PhDJerry HeneghanYeshitila Gebremichael, PhDZack SwarmPat RusslerBeth SmithPaul RutledgeFederico MenozziAlex SomersKatie CarterCassidy LimerUNC Eshelman School of PharmacyDr. Brett Talbot
Government Side:JPC-1 Director: Jan Harris, PhD, RNJPC-1 Portfolio Manager: Kevin Kunkler, MDTATRC Grants Officer Representative: Harvey MageeTATRC Research Scientist: Geoff Miller, MS
Cop
yrig
ht 2
015.
All
right
s re
serv
ed. A
pplie
d R
esea
rch
Ass
ocia
tes,
Inc.
12
Timeline / Milestones2013
2014
2016
2017
Milestone: Project Kick Off (Sept 15, 2013)Creation of Common Data ModelRe-architecture of Physiology EngineBase System DevelopmentBase System ValidationMilestone: Mini Build Release and Website Launch (Oct 2014)Base System Additions & ImprovementsSystem/Feature Development & ValidationCommunity Outreach BioGears Conference PlanningMilestone: Beta Build Release and Users’ Group Conference (Fall 2015)System/Feature Development & ValidationUser Community SupportContinued Community OutreachPublications & Conference Presentations
(9/2017 – 9/2018) Website Maintenance Only
Current
2018
2015
Cop
yrig
ht 2
015.
All
right
s re
serv
ed. A
pplie
d R
esea
rch
Ass
ocia
tes,
Inc.
13
USERS, INTEGRATORS, AND COLLABORATORS
Cop
yrig
ht 2
015.
All
right
s re
serv
ed. A
pplie
d R
esea
rch
Ass
ocia
tes,
Inc.
14
• USAMRMC: Advanced Modular Manikin (4 teams for phase 1; 1 for phase 2) – Patient simulation hardware and software
• RDECOM: Combat Medic – UnReal serious game (early version of BioGears)
• TATRC: HumanSim Sedation and Airway – clinician anesthesia training (pre-BioGears physiology)
• PEOSTRI: Medical Simulation Training Architecture RIF - requires open framework with direct reference to BioGears
• DHP: Warrior Health Avatar SBIR references BioGears directly
1 State-Based Cause and effect using state machines and discrete-event systems Whole Body Proctor Driven
Manikins +++
2 Statistical Fitting of data Whole Body Epidemiology modeling ++3 0-D Physics Based Lumped component model Whole Body Biomedical curriculum +4 1-D CFD Larger geometries with laminar flow Regions of Interest Aortic flow analysis -
5 3-D CFD Specific geometries where flow direction and stresses are important Regions of Interest Aneurysm or stenosis
analysis ---
3) Lumped element modeling:• Simplifies spatially distributed physical
systems into a topology of discrete entities• Reduces state space to a finite dimension
– PDEs of the continuous (infinite-dimensional) time and space model into ODEs with a finite number of parameters
Cop
yrig
ht 2
015.
All
right
s re
serv
ed. A
pplie
d R
esea
rch
Ass
ocia
tes,
Inc.
19
Translating Needs to Specifications
• Key decisions:• Mechanistic where possible/appropriate, phenomenological otherwise• Precision and Accuracy – Rigor and Reproducibility
• Correlation does not mean causation• Correlation for validation purposes• Causation for implementation
• Design challenges – how to balance fidelity• Equal parts biology, math, science, engineering, and black arts
Modeling Approach:• Middle-out
Body Systems Organs Tissues Cells Proteins Genes
Cop
yrig
ht 2
015.
All
right
s re
serv
ed. A
pplie
d R
esea
rch
Ass
ocia
tes,
Inc.
20
• How we choose (with an example system):• We always start with a literature search• Some we inherited general approach (e.g., cardiovascular)• Some we extended from existing (e.g., renal)• Some are derived from an external source (e.g., drugs/PK-PD)• Some are a combination of external sources (e.g., environment & acid-base balance)• Some are developed from scratch (e.g., energy)• Some are re-implemented fresh (e.g., anesthesia machine)• Some are mainly just warehouses of information (e.g., blood chemistry)• All to meet stated contractual requirements
• We are both a consumer and creator of models• Typical criteria for leveraging:
• Typically peer reviewed• Validated with, or derived from empirical data• Sound physics – conservation• Mechanistic/feedback based
• Creating to fill gaps:• Often developed from math/relationships presented in text books or published papers• We employ a methodical design review approach
• Parent models are often only as good as their worst child model (garbage in is garbage out, even when scaled) – high complexity
Model Selection
Cop
yrig
ht 2
015.
All
right
s re
serv
ed. A
pplie
d R
esea
rch
Ass
ocia
tes,
Inc.
21
• Guidelines for credible practice of modeling & simulation in healthcare
• Did not intentionally follow rules during design and implementation, but our approach fits within this context
Verification and Validation1. Verification: Unit tests ensure correct
implementation and sound physics principles for all tools
2. System Level Validation: All major systems (cardiovascular, respiratory, blood chemistry, etc.) are validated for clinical output level data
3. Compartment Level Calibration: Individual organs (kidney, liver, etc.) or functional units (trachea, alveoli, etc.) are validated wherever possible
4. Scenario Calibration & Validation: Every insult, intervention, and assessment includes a matrix with validation data for whole body combined effects from multiple systems
5. Combined Scenario Validation: All four showcases and several other scenarios validated for combined effects – heavily leveraged SME consultants Bryan Bergeron MD, Nicholas Moss PhD, and Stephen Mangum PharmD
1,886 output parameters validated!138 unit tests for verification not shownVerification and methodology waveform validation not shown – see methodology reports
Resting Physiology
System Data TypeValidation Measures in Deviation Category
• Many publications, presentations, seminars, etc.• Our forums include an authentication system so that
people can log-in, post questions and/or discussion topics, and get alerts
• Extremely beneficial to use standard terms for both inputs (e.g., physicochemical properties) and outputs (e.g., assessments, vitals, compartments)
• What users care about:• Most clinicians do not necessarily care what goes on under the
hood, as long as the results are correct and meets the training objectives
• Most researchers/educators care immensely what models are employed and how
• Interactions are/were very helpful to shift focus and reorder tasks
• Determined future needs:• Serialization (saved states for instantaneous loading)• Vitals override• “Try before you buy” graphical user interface• Faster, less functional version• New acute injuries and other conditions
Community Outreach
Cop
yrig
ht 2
015.
All
right
s re
serv
ed. A
pplie
d R
esea
rch
Ass
ocia
tes,
Inc.
28
CONCLUSION
Cop
yrig
ht 2
015.
All
right
s re
serv
ed. A
pplie
d R
esea
rch
Ass
ocia
tes,
Inc.
29
• Use the website: information on downloading, running, and integrating
• BioGears is designed for model extensions and improvements to be integrated by the user base over time
• Let us know how we can help
• We would love to work with CPMS as they define a certification process
How to Use BioGears
Cop
yrig
ht 2
015.
All
right
s re
serv
ed. A
pplie
d R
esea
rch
Ass
ocia
tes,
Inc.
30
• Use the software for any and all applications (please let us know)
• Report problems• Post and respond to Forums (biogearsengine.com/forums)• Submit code
• Currently just email us (biogearsengine.com/workwithus)• Moving to a public repository – GitHub/BitBucket hosted