This page outlines the work I completed for my Honours
dissertation in 2002.
Virus and other disease outbreaks are attracting more and more media
attention. Apart from the obvious health effects of infection, the
economic impact can be substantial as illustrated by the avian flu
outbreak in Hong Kong during 1997 and the foot-and-mouth disease (FMD)
outbreak in Britain in 2000. In order to better understand, predict, and
eventually contain epidemic spread, accurate models need to be devised.
Current epidemiological models are dominated by statistical and
deterministic methods that do not take into account spatial
variables. Considering that natural landscapes are typically far from
homogeneous and uniform, it makes sense to incorporate spatial
heterogeneity parameters to produce a better model. Cellular automata
(CA) is a
paradigm that is conducive for modelling space and my project
focusses on investigating the applicability of CA to disease spread
simulation.
SimDemic
SimDemic is the epidemic spread simulation that I wrote
as part of my Honours research in 2002. It is based on the SIR epidemic
model and uses a 2-D cellular automaton to emulate the spatial
behaviour of a spreading infection.
My epidemic model is far from comprehensive and is intended only to
provide preliminary evidence of CA's usefulness as a epidemic
modelling paradigm.
Below are some pre-synthesized scenarios that show some of the
capabilities of a CA epidemic model. Each applet runs the same set of
CA rules, but has a different starting state for the cell grid. You
will need Java Web Start
installed on your browser for these applets to work.
The left panel of the interface depicts the extent of infection; the
right panel shows the population density map. Both maps are of the same
grid and updated synchronously. Note that just one infective host will
turn a square red, whilst the saturation of the blue indicates how
close to its carrying capacity a cell has become.
Further details can be found in my Honours dissertation provided below.
Homogeneous scenario
Click here to see a scenario
that uses "pure" CA where each cell can only contain one host.
Consequently there are no empty spaces so hosts cannot move
even if the movement parameter is non-zero.
Click here to use a world with
a uniform host distribution where each cell contains 100
susceptibles. Each cell has a maximum capacity of 200.
Varied population density
Click here to start
with a world with a top to bottom density gradient of high to low. It
is observed that the outbreak spreads rapidly at first from the edge
source at the top of the grid, but gradually slows down.
Click here to see the low to
high density scenario.
Corridors of spread
Click here to see a scenario
that shows how epidemics are likely to spread along what could
be geographical or cultural features than open space.
Barriers to spread
Click here to see the the
scenario that investigates the effect of barriers to host movement and
(hopefully) virus spread.
Realism in Epidemic Models
Shih Ching Fu
A brief literature review discussing some existing
epidemic models and outlining the epidemic spread parameters they address.
Incubation Time Does Not Mean Idle Time Shih Ching Fu
An experimental paper I wrote that describing how I tried to
reproduce the results of some existing epidemic models.
A copy of my Honours seminar abstract can be found
here.
Epidemic Modelling Using Cellular Automata
Shih Ching Fu and George Milne
First Australian Conference on Artificial Life (ACAL2003)
Canberra, Australia, 6-7 December 2003
A Flexible Automata Model for Disease Simulation
Shih Ching Fu and George Milne
Sixth International Conference on Cellular Automata for Research and
Industry (ACRI2004)
Amsterdam, The Netherlands, 25-27 October 2004
