1. Tuesday Seminar Deformable Image Registration
07th of January, 2014
Tuesday Seminar Deformable Image Registration
HyunSeok Lee

2. Contents What is Image Registration? Deformable Image Registration
Basic concept
Algorithm
Products
Open source
Debates about dose deform
TG132

3. What is image registration?
Image registration is the process of transforming different sets of data into one coordinate system.
Registration is necessary in order to be able to compare or integrate the data obtained from different measurements.

4. IR – Rigid Transformation
Rotation
Translation
Scale

5. IR – Affine Transformation
Rotation
Translation
Scale
Shear
No more preservation of lengths and angles
Parallel lines are preserved

6. IR – Perspective Transformation
(xo, yo, zo)  world coordinates
(xi, yi)  image coordinates

7. IR – Projective Transformation
(xp, yp)  Plane Coordinates
(xi, yi)  Image Coordinates
amn  coefficients from the equations of the scene and the image planes

8. IR – Non-Rigid Transformation
Needed for inter-subject registration and distortion correction
Non-linear
Many different parameterizations
Too much flexibility in the transformation can lead to undesirable results

9. Deformable Image Registration
Fundamental task in medical image processing
Typical uses
Longitudinal studies
where temporal structural or anatomical changes are investigated
Matching of images from different patients
Multi-modal registration
matching images of the same patient acquired by different imaging technologies
i) multi-modality fusion,
where information acquired by different imaging devices or protocols is fused to facilitate diagnosis and treatment planning;
Matching of images from different patients (inter-patient registration).
Multi-modal registration which means matching images of the same patient
acquired by different imaging technologies.
ii) longitudinal studies, where temporal structural or anatomical changes are investigated;
Tracking of deformable objects in a series of medical scans
iii) population modeling and statistical atlases used to study normal anatomical variability.

10. Matching Criteria (Objective Function)
DIR – Algorithm
Deformation Model
Optimization Method
Matching Criteria (Objective Function)

11. DIR – Products MIM Software Inc.
Intensity-based free-form deformable registration (VoxAlign)
PET/CT, MR/CT and 4D data sets deformable fusion
Atlas-based auto-contouring
Dose Accumulation
Adaptive re-contouring
Deformable registration QA and reporting

12. Intensity-based Methods
Intensity-based methods compare intensity patterns in images via correlation metrics
Sum of Squared Differences
Normalized Cross-Correlation
Mutual Information
Instead of specific features, only pixel intensity values are considered in order to find the transformation of interest.
Sum of Squared Differences
Only valid for same modality with properly normalized intensities in the case of MR.
Normalized Cross-Correlation
Allows for linear relationship between the intensities of the two images
Mutual Information
More general metric which maximizes the clustering of the joint histogram.

13. Feature-based Methods
Feature-based methods find correspondence between image features such as points, lines, and contours.
Distance between corresponding points
Similarity metric between feature values
e.g. curvature-based registration
Once corresponding points have been found, their locations in the two image can be used to reconstruct a spatial transformation.

14. Free-form Deformations
The general idea is to deform an image by manipulating a regular grid of control points that are distributed across the image at an arbitrary mesh resolution.
Control points can be moved and the position of individual pixels between the control points is computed from the positions of surrounding control points.
Apart from the smoothness properties that can be enforced using suitable basis functions, the control points can be placed at variable distances, giving a exible way of controlling deformation precision.
In addition, the concept of manipulating control points in order to deform an image can have an efficiency advantage over methods where deformations are computed on a per-pixel basis.

15. DIR – Products Mirada Medical
Algorithms are based on published algorithms but have been developed and optimized for particular RO use-cases and modality combinations.
e.g. CT-CT Optic Flow algorithm for PET/CT fusion
e.g. CT-MR multi-modal algorithm for MRI fusion
Multimodal deformable fusion
CT, PET, PET/CT, MRI and CBCT, including 4D data sets
Automatic contouring
using an atlas or previously contoured case
Dose warping and summation
Adaptive therapy

16. DIR – Products Velocity Medical Solutions
Multi-resolution modified basis spline algorithm
Multi-modality demons algorithm
Multi-modality deformable image registration
CT, MR, PET and SPECT
Atlas-based auto-contouring
Treatment response assessment
The paper, "Quantitative Evaluation of a Cone Beam Computed Tomography (CBCT)-CT Deformable Image Registration Method for Adaptive Radiation Therapy" published in the Journal of Applied Clinical Medical Physics (Vol 8(4): , 2007), evaluated the performance of Velocity's multi-modality B-Spline deformable registration algorithm for the fusion of treatment delivery CBCT volumes to CT planning volumes.

17. DIR – Open Source ITK Insight Segmentation and Registration Toolkit
An extensive suite of software tools for image analysis
Implemented in C++

18. DIR – Open Source DIRART
Deformable Image Registration and Adaptive Radiotherapy
It contains well implemented DIR algorithms and the essential functions for ART applications.
Implemented in MATLAB
Need CERR
free-form deformation (FFD)
Horn-Schunk optical flow (OF)
Demons
Technical Note: DIRART – A software suite for deformable image registration and adaptive radiotherapy research

19. DIR – Open Source Plastimatch
For high performance volumetric registration of medical images
ITK-based algorithms for translation, rigid, affine, demons, and B-spline registration
Some methods are GPU and multicore accelerated
Implemented in C++
Plugin to 3D Slicer
software package for visualization and medical image computing

20. Debate about dose deform

21. It is not appropriate to “deform” dose along with DIR in ART
For ART it is common to collect images of the patient throughout the course of therapy.
Because of temporal variations, it is usually necessary to deform images so as to merge them into a cohesive dataset.
This image registration makes the accurate merging of dose distributions difficult.
Some have decided to do this by “deforming” the dose distributions, somewhat analogous to deforming the images, but it has been suggested that this is not appropriate.

22. Erase Deformable IR is yet more error prone.
With deformed images (or the accompanying contours) we can at least choose to accept or modify them.
But when the dose is deformed we have nothing upon which we can base a visual evaluation.
The ultimate problem with deformed dose is our inability to measure it.
However, in the case of dose painting where target dose is heterogeneous, dose warping is necessary to ensure dose to corresponding spatial locations are accurately accumulated.
without accurately accumulating the dose over multiple images, it could be hazardous to adjust the treatment plan  we cannot
we should continue in a prospective manner to explore the usefulness of DIR using in silico clinical trials to determine which patient populations and clinical sites could benefit from DIR without actually changing current treatment plans and treatment paradigms.

23. Task Group No. 132
Use of Image Registration and Fusion Algorithms and Techniques in Radiotherapy
Emphasis the importance of acceptance testing, including end-to-end tests, phantom tests, and clinical data tests.
Describe the methods for validation and quality assurance of image registration techniques.
Describe techniques for patient specific validation.

24. Deformable PRESAGE® dosimeter
7th International Conference on 3D Radiation Dosimetry (IC3DDose)
A comparison between the VelocityAI dose distribution and the distribution from the dosimeter showed field displacements up to 7.3 mm and up to a 175% difference in field dimensions.
These results highlight the need for validating deformable dose mapping algorithms to ensure patient safety and quality of care.

25. Discussion & Question

26. Thank you for your attention

27. DIR – Algorithm
Deformation Model

28. DIR – Algorithm
Matching Criteria (Objective Function)

29. DIR – Algorithm
Optimization Method