Transcript Epidemiological Simulation and Social Networks

Exploratory Visualization of
Infectious Disease Propagation
Ben Houston, Neuralsoft
Zack Jacobson, Health Canada
NX-Workshop on Social Network Analysis
and Visualization for Public Safety
18 – 19 October 2005
Introduction
• Goal: Deliver a disease propagation
modeling application based roughly on the
VITA NetViz information visualization
component
• Focused on application development.
• Theory is an important, but secondary, aid.
Approach
• Rapid application development
– Evolutionary, circular development model.
– Regular client/customer demonstration w/
feedback.
– Don’t over think problems, focus on
presenting possible solution.
– Often the effective solutions are only apparent
once your client sees aspects of it in front of
them.
Technology
• Language: Microsoft C#
• Graphics: GDI+ or OpenGL or Direct3D
• Mathematics: custom library
Simulation Components
Agents – one agent per “individual”, intrinsic
characteristics: activity level, sociability, resiliency,
dynamic attributes: health, energy level, position,
disease state.
Environment – infinite 2-dimensional plane. Agents are
free to move. Agent-agent proximity is used to model
disease exposure and transmission.
Disease Model – a state transition graph. Interaction
between the agent’s intrinsic and dynamic attributes
determines the progression of the disease (the speed
and path through disease’s state transition graph.)
Simulation Initialization
– Population model used to initialize the
intrinsic characteristics of each of the agents.
Specifies “distribution” of characteristics.
– Disease model – a general disease model
which via parameters specification can be
used to represent diverse specific diseases.
Running the Simulation
• All data produced during the simulation is
captured in an efficient compressed form
in real time.
Interactive Visualization
• 3 separate synchronized views which each
allow intuitive access to different data
dimensions.
• Currently, only one view can be displayed
on the screen at a time, but simultaneous
views may be in the future.
Interactive Visualization - 1
Population Snapshot View– Overhead view of
a 2D plane. The user can scroll through time to inspect
specific times of interest.
Interactive Visualization - 2
Agent-Agent Transmission Graph – The
implicit social network created by tracking agent-agent
infections.
Interactive Visualization - 3
Population Attribute vs. Time Charts -Population statistics. Useful for finding minimal, maximal
points or slopes.
Demonstration and Discussion
– How do features of interest in one view appear on the
other views?
– Allow exploratory modification of the simulation to
allow testing of containment strategies.
– Currently we are just sampling one population from
the population distribution – it may be useful to
automatically sample many representative
populations to find a mean.
– Increase complexity of the environment (barriers,
transmission modifiers), and agent behavior.
– Can this tool (or a further evolution of it) be useful in a
war room scenario to help responders? If so how? If
not why?