The University of SydneySlide 1 IMAGE PROCESSING Presented by
Paul Wong AMME4981/9981 Semester 1, 2015 Lecture 2
Slide 2
The University of SydneySlide 2 Required Inputs Geometry What
does it look like? Material properties What is it made of? Loads
What forces is it subjected to? Boundary conditions What is
happening at the system boundary?
Slide 3
The University of SydneySlide 3 Workflow for Biomedical
Problems 1. Data acquisition Scan region of interest Obtain
material properties for tissues and implants Estimate expected
loads 2. Solid modelling Convert image stacks into a virtual
replica Combine with CAD model of prosthesis 3. Finite element
analysis Generate appropriate mesh Characterise interaction between
anatomy and prosthesis Verify simulation results and prosthesis
design
Slide 4
The University of SydneySlide 4 SEGMENTATION &
RECONSTRUCTION
Slide 5
The University of SydneySlide 5 Objective To replicate the
geometry of the region of interest in silico Prosthesis is
typically modelled using CAD Anatomy can also be made in CAD, but
this is not ideal REGION OF INTEREST Prosthesis Anatomy
Slide 6
The University of SydneySlide 6 Organic Shapes Organic
structures typically have a complex geometry Irregular shape,
non-uniform curves No symmetry Multiple parts and interfaces
Differences between individuals Anatomy should be reconstructed
from scans Higher accuracy Enables patient-specific studies
Slide 7
The University of SydneySlide 7 CT Scan of Proximal Femur
Slide 8
The University of SydneySlide 8 Typical Workflow Import image
stack Separate into different tissue types or parts Apply smoothing
filter and/or fit curves Create surface model Generate solid
model
Slide 9
The University of SydneySlide 9 Commercial Software VSG Amira
User friendly interface Easy to learn Some very handy selection
tools are unique to Amira Excellent real-time visualisation options
Meshing often requires manual adjustment
Slide 10
The University of SydneySlide 10 Commercial Software
Materialise Mimics Market leader Highest degree of user control
Very powerful tools Not very user friendly Extremely expensive
Slide 11
The University of SydneySlide 11 Commercial Software Simpleware
ScanIP Student licences are available in S322 and S345 There are
only 20 licencesfirst come, first served How to learn the software:
Basic segmentation from CT: Proximal femur (Chapter 7, Simpleware
Tutorial Guide) Import and positioning of CAD implant: Proximal
femur (Chapter 8, Simpleware Tutorial Guide) Chapters 4-7,
Simpleware Reference Guide Just try something and see what
happens
Slide 12
The University of SydneySlide 12 Typical Workflow Import image
stack Separate into different tissue types or parts Apply smoothing
filter and/or fit curves Create surface model Generate solid
model
Slide 13
The University of SydneySlide 13 ScanIP User Interface
Slide 14
The University of SydneySlide 14 Segmentation
Slide 15
The University of SydneySlide 15 Selection Tools Selection
tools add voxels to masks NAMEFUNCTIONCOMMENTS BrushSelects voxels
manuallyOnly use this if all others are insufficient, or for
touch-ups ThresholdSelects voxels based on luminance Use if tissues
have significantly different grey values in the image Flood
fillSelects voxels based on connectivity There are a few
variations: Left click, inside volume Keeps connected voxels only
Left click, outside volume Fills gaps Right click, inside volume
Removes all connected voxels Right click, outside volume Removes
cavities
Slide 16
The University of SydneySlide 16 Smoothing Usually required
because: Imaging rarely distinguishes tissues perfectly Manual
slice-by-slice processing is difficult to align The dataset may
have image artefacts Organic structures are generally smooth
Misalignment creates jagged surfaces Not ideal for meshing Can
result in stress concentrations and singularities at FEA stage that
are not representative of real-world behaviour
Slide 17
The University of SydneySlide 17 Filters Filters are used to
manipulate a pre-existing mask TYPEFUNCTIONCOMMENTS
SmoothingRemoves jagged edges and surfaces Small features tend to
be lost, so test different amounts of smoothing (start small) to
balance smoothness with geometric accuracy (i.e. to retain
sufficient fine detail) MorphologicalManipulates size and shape
There are a few variations: Dilate Grows selected region Erode
Shrinks selected region Open Removes small islands (i.e. erode then
dilate) Close Removes small cavities (i.e. dilate then erode)
Slide 18
The University of SydneySlide 18 Surface vs Volume Models
Surface models Only define the outside shell of an object Most
systems achieve this using non-uniform rational B-splines (NURBS)
Not fully-defined Volume models Solid models define the entire
volume of an object e.g. Solidworks models Representation is
unambiguous
Slide 19
The University of SydneySlide 19 Dimensionality in Modelling
DIMENSIONGEOMETRYMESH 0PointNode 1CurveLine element 2SurfaceShell
element 3BodyVolume element
Slide 20
The University of SydneySlide 20 Visualisation Fast preview
Renders approximate volume in 3D viewer (using voxels from masks)
May not represent segmentation accurately Model preview Creates and
renders complete surface model (i.e. using triangles in 3D space)
Can be exported for volume meshing in other programs
Slide 21
The University of SydneySlide 21 Meshing Full model Generates
and renders volume model and/or (non-conformal) NURBS surfaces
Requires masks to be added to a model object Export options can be
accessed through Model setup ATTRIBUTE+FE GRID+FE FREE
RobustnessExcellent; can mesh virtually any geometry Good, but can
fail with complex geometries Mesh sizeVery largeMedium; depends on
refinement settings Mesh timeShortLonger FEA solution
timeLongShorter
Slide 22
The University of SydneySlide 22 Tips & Tricks Think in
terms of layers (a la Photoshop) Can perform Boolean operations Can
import multiple backgrounds Learn the keyboard and mouse shortcuts
There is an undo button (5 actions by default) Dont forget to
backup SIP files If in doubt, check the help guides
Slide 23
The University of SydneySlide 23 COMING UP
Slide 24
The University of SydneySlide 24 Next Week Group project
updates Each group to give an informal short talk about their
chosen topic Expect up to 2 minutes with 1-2 slides
Slide 25
The University of SydneySlide 25 Week 4 Guest lecture by Simon
Richards from Simpleware Introduction and Live Demo: 3D Image
Visualisation, Analysis and Model Generation with Simpleware 2-4pm
in Electrical Engineering Lecture Theatre 2 (Room 450)