Scribd
Upload
124 views38 pages

ITK Clinical Applications

The Insight Toolkit (ITK) is a open-source software toolkit for image segmentation and registration. ITK provides spatial objects that represent anatomical structures and can be used for tasks like segmentation, registration, surgical planning and guidance. Example applications include vascular registration, ultrasound annotation, and breast density estimation. ITK aims to establish common APIs for medical image analysis and provide educational resources to promote open science.

Uploaded by

mateyko
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
124 views38 pages

ITK Clinical Applications

The Insight Toolkit (ITK) is a open-source software toolkit for image segmentation and registration. ITK provides spatial objects that represent anatomical structures and can be used for tasks like segmentation, registration, surgical planning and guidance. Example applications include vascular registration, ultrasound annotation, and breast density estimation. ITK aims to establish common APIs for medical image analysis and provide educational resources to promote open science.

Uploaded by

mateyko
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 38

The Insight

Toolkit
Medical Image
Segmentation and
Registration

Stephen R. Aylward
Computer-Aided Diagnosis and
Display Laboratory (CADDLab)
Department of Radiology
UNC, Chapel Hill
Outline

z Spatial Objects

z Clinical Applications
Spatial Objects
Spatial Objects

z Goal of medical image analysis is object


analysis

z Segmentation, Registration, Surgical


Planning, Surgical Guidance, …
– Focus on forming and manipulating a scene of
objects
– Objects with non-image properties
z Elasticity, Viscosity, … for Registration
z Name, ID, text
z Haptic forces
Uses
z Data container independent
z Consistent API
z Wrapper to Image, Mesh, FEM,
SpatialFunctions and other concepts in ITK

z Why establish a common API?


– Marching cubes
– Registration
– Image generation
– Model-based segmentation
– Model-to-image registration
– Results of segmentation
– …
z All provided by basic API
Spatial Objects
z Scene
– List of root spatial objects

z Spatial Objects
– IsInside
– IsEvaluableAt
– ValueAt / DerivativeAt
– BoundingBox
– Resolution (of the data from which the instance was generated)
– AffineTransform (wrt Parent Object and World)
– Children
– Properties
Spatial Objects
z Group
z Blob
z Surface
z Line
z Landmark
z Ellipse
z Tube

z Derived from spatial object


– Implement:
IsInside, ComputeBoundingBox, and ValueAt
Specialized Access
z ComputeBoundingBox(Levels, Type)
Level: 0 = no children
1 = children
2 = children's children
Type: “Tube”
“Ellipse”
…

z e.g.,
obj.ValueAt( point, 2, “Tube” );
obj.ValueAt( point, SpatialObject::MaxDepth, NULL );
Examples
z Surgical Planning
– Blob = Liver
– Tubes = Vessels
– Surface = Surgical Cut Path
– Landmarks = Portal Vein Branch
z Intra-operative guidance
– Liver modeling from CT
– Ultrasound probe as an object
z Transform controlled by a tracker
– Model-to-image registration (Liver model to US data)
z Constrained by physical properties of liver’s parts

z Registration
– Rigid: example in Software Guide
– Non-Rigid: additional parameters
Example
z Examples/Registration/
Model-To-Image-Registration1.cxx
– ImageToSpatialObjectMetric
– OnePlusOneEvolutionaryOptimizer
– Registration Framework

z Spatial Objects: Group


(Easy to switch objects)
Ellipse Ellipse Ellipse

z Create Objects, SpatialObjectToImageFilter,


Rotate Objects, Re-register with image
Rendering Spatial Objects
typedef itk::SpatialObjectToImageFilter<GroupType,
ImageType> SpatialObjectToImageFilterType;
SpatialObjectToImageFilterType::Pointer imageGenerator =
SpatialObjectToImageFilterType::New();
imageGenerator->SetInput(group);
ImageType::SizeType size;
size[0]=200;
size[1]=200;
imageGenerator->SetSize(size);
imageGenerator->Update();

ImageType::Pointer img = imageGenerator->GetOutput();


SpatialObject To Image Registration
typedef itk::GroupSpatialObject<2> GroupType;
typedef itk::EllipseSpatialObject<2> EllipseType;

EllipseType::Pointer ellipse1 = EllipseType::New();


ellipse1->SetRadius(10);
offset[0]=100;
offset[1]=40;
ellipse1->GetTransform()->SetOffset(offset);
ellipse1->ComputeGlobalTransform();
…

GroupType::Pointer group = GroupType::New();


group->AddSpatialObject(ellipse1);
…

registration->SetFixedImage(imageFilter->GetOutput());
registration->SetMovingSpatialObject(group);
registration->SetTransform(transform);
registration->SetInterpolator(interpolator.GetPointer());
registration->SetOptimizer(optimizer);
registration->SetMetric(SpatialObjectToImageMetric);
Capabilities yet explored…
z Explicit multi-scale representation API

z Atlas of objects
– Common API for annotation
– Means, major modes of variation, etc.
– Instance-to-Atlas metrics

z Will be revised in itk v1.8


– Tree structure, objects, geometric
transforms as separate components
Clinical Applications
3D Ultrasound Augmentation
NCI R01: Sue Weeks MD & Julien Jomier
Vascular Registration

Vascular Registration Method

Fixed 3D Image
(US)
ImageToTube Gradient Descent
Metric Optimizer

Tri-linear
Interpolator

Euler 3D Rigid
Transformation
Moving Vasculature
(extracted from CT)
UNC vascular registration
Vascular Registration
Using 1 Processor Using 2 Processors

Performances:
Number of Processors running
Time/Iteration (sec)
on a dual Intel Total time
Pentium III 733Mhz
used
(512MB)
1 0.153 6.31
2 0.297 12.25
This is 1.94 time faster !
UNC vascular registration
Ultrasound Annotation

Vessel model to
image registration
in 2.56 seconds –
Multithreaded ITK
Image Registration
LandmarkInitializedMutualInformationRegistration.exe
Autonomic Nervous System
Assessment
Eric Whitsel, Erwann Rault
NVVLS
Dynein Arms via Models
J. Carson: EPA / Rare diseases

Quality Quality Quality


Of Fit Of Fit Of Fit
-- + ++
Quantification of Adipose
Tissue from Whole Body MRI scans
D. Metaxas (Rutgers), J. Udupa (UPenn),
C. Imielinska (Columbia)

Segmented Whole-body MRI scan Reconstructed 3D body


(St. Luke’s-Roosevelt Hospital Composition from
Obesity Research Center – Dr. Heymsfield) Segmented whole body MRI scan
Computer Assisted Segmentation
R. Whitaker (Univ. of Utah), P. Ratiu (Harvard / BWH)

Accuracy
Hand 96.86 (σ = 2.6)
Assisted 98.36 (σ = 0.3)

Time
Hand 271 minutes
Assisted 73 minutes
Uterine Fibroid Growth Study
NIEHS: Barbara Davis (WHL), Richard Semelka MD
Breast Density Estimation
WHI: Etta Pisano MD

Compensate for compression


Original

Global Threshold - Standard Auto threshold


Software Toolkit for
Image-Guided Surgery
NIH STTR: Kevin Cleary, Georgetown
CT to X-Ray Registration
Imperial and King’s College London

3D-2D
Registration X-ray

Verification of prosthesis placement


Haptics for Surgical Training
Microsoft: Mark Foskey
Small Animal Imaging

MicroPet / MicroCT
GE Corp Research
Histology to MR Registration
Benoit Paquier

MR Histology
DTI Analysis
Guido Gerig
Functional Image Analysis Tool
U of Iowa
ITK + Analyze
Mayo Clinic
Face Recognition
Julien Jomier
Education
• Methods in Image Analysis
BioEng 2630 (UPitt) / 16-725 (CMU)
G. Stetten, M.D., Ph.D.

Visualization and
Image
Analysis Laboratory

• Segmentation Methods Using ITK


Project Course (Rutgers)
D. Metaxas, Ph.D.

• Pattern Recognition and Scene Analysis


CompSci 254 (UNC)
S. Pizer, Ph.D.
Education
z Tutorials
– IEEE Vis 2002
– SPIE 2003
– MICCAI 2003
– RSNA 2003 InfoRad
– SPIE 2004…

z Books
– Software Guide (580 pages)
– Theory Book

z MedIA Special Issue


Foundation for Open Science
Dr. Terry Yoo (NLM)

z Open Source + Open Data = Open Science

z NCI/NIBIB CAD Workshop (Sept. 4 – 5, 2003)


– One solution for validation:
ITK with CAD Methods + Grid Computing (Data)

z ITK + NIH Open Data Initiative


– ITK Sets the standard for openness
– ITK Provides methods for validation
– ITK Provides I/O foundation for Open-data databases
– ITK Provides platform for method distribution
The Insight
Toolkit
Medical Image
Segmentation and
Registration

http://www.itk.org

Stephen R. Aylward
Computer-Aided Diagnosis and
Display Laboratory (CADDLab)
Department of Radiology
UNC, Chapel Hill

You might also like