Transcript slides - CompuSteer

Image-Based Steering for
Integrative Biology
Lakshmi Sastry, Richard Wong, Helen Wright
with contributions from Ronald Fowler,
Sri Nagella and Anjan Pakhira
Acknowledgements
 Ken Brodlie and Jason Wood
 CompuSteer funding
 Integrative Biology project scientists
Image-based Steering
Integrated
display
Render
Visualization processing
Map
X

X
Filter
Simulate
Integrative Biology (IB)
 Grid technology to enable in-silico
experiments by computational
biologists
 Combined resources for computation,
data management, visualization and
data analysis
 Focus on fatal diseases – heart and
cancer
Example IB Applications
Modelling heart electrical activity during
arrhythmia:
 Tulane whole ventricular model –
epicardial potential distribution over
heart geometry during shock-induced
arrhythmia
 Fenton-Karma 4-variable model on
2D slice of tissue
An episode of arrhythmogenesis in ventricular model.
The arrhythmia is a figure of eight reentry with one
rotor on the anterior (left panel) and another on the
posterior (right panel) of the ventricles. The arrows
show wave propagation. The scale is saturated,
potentials above 20mV are shown in red and below 90mV are shown in blue.
Example IB Applications
 In vitro and in-silico models of
tumour growth during very early
stages
 Seamless secure access to very large
volumes of image data, processing,
simulation and interaction will
accelerate understanding of disease
process.
Steering for IB Applications
 Complex and compute intensive with
tens and hundreds of parameters
 Verification of models that continue to
be refined
 Computational exploration of
parameter space
 Expanding set of simulations and
visualization toolkits
Image-based Interaction
 Extrinsic parameters (scalars,
vectors) mimic widgets but minimise
context switching
 Parameters intrinsic to the solution
graphic, e.g. position specifications
 The IB interface provides a layer of
abstraction above the clientside
libraries for computational steering.
The Case for Server-side
Applications
 Application may already have steering
embedded
 Developing a steerable interface and other
scalable services for each application does not
scale
 Difficult to embed steering and other services
into certain visualization toolkits
 Users want continuity in their visualization
toolkits
 Minimises changes needed to application
software
Client-side Consequences
 Keep client generic – configure on
set-up to meet application
requirements
 Needs to handle various geometry
and image formats
 Application-specific activity e.g. to
resolve geometry elements or nodes,
takes place server-side
Client A
IB Interface
Visualisation &
interactors
panel
Control panel of
widgets
gViz
client
side
S
t
e
e
r
IB Server
Image & image based
parameter values from
coder/decoder
Visualisation toolkit
(e.g. Meshalyser)
Data
Simulation
(e.g. CARP)
V
i
e
w
gViz sim.
module
S
t
e
e
r
IB Interface
Visualisation &
interactors
panel
Client B
Control panel of
widgets
gViz
client
side
V
i
e
w
Collaborative gViz Overview
 Parameter changes are passed to all
collaborators for visibility (steer/view
arrows)
 Committed parameters are passed to all
collaborators and the simulation, locking
interactors
 Arrival of data unlocks interactors – implies
token-passing
 Data streams – not used here – separate
results from parameters
Demonstrator Elements
 Tumour modelling – growth of ductal
carcinoma in breast
 Results – time-varying tumour cell
counts in axial and radial direction of
duct
 Steering of nutrient consumption rate
and cell-to-duct-wall slip coefficient
 Utilises gViz rel.2 (collaborative) for
parameter passing, calling Fortran
Visualization Back-end
 IRIS Explorer, loosely coupled
 Simulation outputs file of results
(time step) which triggers
visualization
 Height-field plot varying in time
 height = cell numbers
 colour = pressure
Steering nutrient consumption and cell death
rate (6MB movie)
OpenGL Interactors
Experiences
 Hard to ‘wipe the slate clean’ before
starting again
 New collaboration helps
 Mode is ‘extreme collaboration’ (cf.
extreme programming)
 Needs dedicated time
 Trips - how long is ‘just long enough’?
Remaining Question Marks
 Token maintenance over the various
architecture pathways
 Recombination of 3rd party
geometries/images with interactors
 Anticipate little problem for extrinsics
 Intrinsics more difficult
 gViz and multiple simulations?
Remaining Question Marks
 How scalable is the architecture
really?
 Will scientists and steering libraries
ever really mix?
 What support do scientists need to
use steering libraries –
documentation, examples, GUI
builders?