Upload
others
View
7
Download
0
Embed Size (px)
Citation preview
4D Treatment Planning Workshop 2015 - 27.11.2015 - Oncoray - Dresden, Germany - Hugo Furtado
Furtado H.1,3, Renner A.2,3, Daniel M. 2,3, Buschmann M. 2,3,
Seppenwoolde Y.2,3 , Georg D.2,3, Birkfellner W.1,3
1 Center for Medical Physics and Biomedical Engineering, Medical University Vienna, Austria 2 Department of Radiation Oncology, Medical University Vienna, Austria
3 Christian Doppler Laboratory for Medical Radiation Research for Radiation Oncology, Medical University Vienna, Austria
New developments and improvements in Deformable Image Registration
4D Treatment Planning Workshop 2015 - 27.11.2015 - Oncoray - Dresden, Germany - Hugo Furtado 2
Outline
• Introduction
Motivation and objectives
• Methodology
Common framework
Examples
• Ongoing projects at the MUW
Autocontouring
Featurelets
• New trends / Ideas
• Conclusion
4D Treatment Planning Workshop 2015 - 27.11.2015 - Oncoray - Dresden, Germany - Hugo Furtado 3
Motivation - Why the strong need for DIR
Patient anatomy
changes over time
Pla
nn
ing
Treatment
Bladder
Uterus
4D Treatment Planning Workshop 2015 - 27.11.2015 - Oncoray - Dresden, Germany - Hugo Furtado 4
Motivation - Why the strong need for DIR
• With deformable image registration we can
Dose accumulation
Auto contouring (atlas based)
Contour propagation (Planning CT to weekly CT)
Multi modal image manual contour guidance
4D Treatment Planning Workshop 2015 - 27.11.2015 - Oncoray - Dresden, Germany - Hugo Furtado 5
Motivation - Typical challenges to solve
• Different image content between modalities
• Very big deformations
• Appearing / Disappearing tissue
• Validation! (very important, see next talk)
4D Treatment Planning Workshop 2015 - 27.11.2015 - Oncoray - Dresden, Germany - Hugo Furtado 6
Outline
• Introduction
Motivation and objectives
• Methodology
Common framework
Examples
• Ongoing projects at the MUW
Autocontouring
Featurelets
• New trends / Ideas
• Conclusion
4D Treatment Planning Workshop 2015 - 27.11.2015 - Oncoray - Dresden, Germany - Hugo Furtado 7
How does it work - basic framework
A. Sotiras, C. Davatzikos, and N. Paragios, “Deformable medical image registration: A survey,” IEEE Transactions on Medical Imaging 32(7), pp. 1153–1190, 2013.
4D Treatment Planning Workshop 2015 - 27.11.2015 - Oncoray - Dresden, Germany - Hugo Furtado 8
How does it work
A. Sotiras, C. Davatzikos, and N. Paragios, “Deformable medical image registration: A survey,” IEEE Transactions on Medical Imaging 32(7), pp. 1153–1190, 2013.
4D Treatment Planning Workshop 2015 - 27.11.2015 - Oncoray - Dresden, Germany - Hugo Furtado 9
How does it work
A. Sotiras, C. Davatzikos, and N. Paragios, “Deformable medical image registration: A survey,” IEEE Transactions on Medical Imaging 32(7), pp. 1153–1190, 2013.
4D Treatment Planning Workshop 2015 - 27.11.2015 - Oncoray - Dresden, Germany - Hugo Furtado 10
How does it work
A. Sotiras, C. Davatzikos, and N. Paragios, “Deformable medical image registration: A survey,” IEEE Transactions on Medical Imaging 32(7), pp. 1153–1190, 2013.
4D Treatment Planning Workshop 2015 - 27.11.2015 - Oncoray - Dresden, Germany - Hugo Furtado 11
How does it work
A. Sotiras, C. Davatzikos, and N. Paragios, “Deformable medical image registration: A survey,” IEEE Transactions on Medical Imaging 32(7), pp. 1153–1190, 2013.
4D Treatment Planning Workshop 2015 - 27.11.2015 - Oncoray - Dresden, Germany - Hugo Furtado 12
How does it work
A. Sotiras, C. Davatzikos, and N. Paragios, “Deformable medical image registration: A survey,” IEEE Transactions on Medical Imaging 32(7), pp. 1153–1190, 2013.
4D Treatment Planning Workshop 2015 - 27.11.2015 - Oncoray - Dresden, Germany - Hugo Furtado 13
How does it work
A. Sotiras, C. Davatzikos, and N. Paragios, “Deformable medical image registration: A survey,” IEEE Transactions on Medical Imaging 32(7), pp. 1153–1190, 2013.
4D Treatment Planning Workshop 2015 - 27.11.2015 - Oncoray - Dresden, Germany - Hugo Furtado 14
Simple example
B-spline + Image based similarity
4D Treatment Planning Workshop 2015 - 27.11.2015 - Oncoray - Dresden, Germany - Hugo Furtado 15
Simple example
B-spline + Image based similarity
4D Treatment Planning Workshop 2015 - 27.11.2015 - Oncoray - Dresden, Germany - Hugo Furtado 16
Making it more complex
• Clinical research version software from commercial vendors
• DIR software containing some of the latest improvements
• Combined approach between image and model based
Image similarity measure
Grid regularization
Shape based regularization
Controlling structures are used
O. Weistrand and S. Svensson, “The anaconda algorithm for deformable image registration in radiotherapy,” Medical Physics 42(1), pp. 40–53, 2015.
4D Treatment Planning Workshop 2015 - 27.11.2015 - Oncoray - Dresden, Germany - Hugo Furtado 17
Making it more complex - objective function
𝒇 𝒗 = 𝜶𝑪 𝒗 + 𝜷𝑯 𝒗 + 𝜸𝑺 𝒗 + 𝜹𝑫(𝒗)
Image similarity
(correlation coefficient)
Regularization of DF
Controlling structures
Smoothness
Invertibility
Penalty for
„non-intuitive“
deformations
ROI POI
O. Weistrand and S. Svensson, “The anaconda algorithm for deformable image registration in radiotherapy,” Medical Physics 42(1), pp. 40–53, 2015.
4D Treatment Planning Workshop 2015 - 27.11.2015 - Oncoray - Dresden, Germany - Hugo Furtado 18
Regularization of the deformation field
Shackleford et. al., “Analytic regularization of uniform cubic b-spline deformation fields.,” Medical image computing and computer-assisted intervention : MICCAI ... International
Conference on Medical Image Computing and Computer-Assisted Intervention 15(Pt 2), pp. 122–129, 2012.
Johansson et. al., "Globally optimal displacement fields using local tensor metric," in Image Processing (ICIP), 2012 19th IEEE International Conference on , vol., no.,
pp.2957-2960, Sept. 30 2012-Oct. 3 2012
4D Treatment Planning Workshop 2015 - 27.11.2015 - Oncoray - Dresden, Germany - Hugo Furtado 19
Regularization of the deformation field
Penalize „non-intuitive deformations“
O. Weistrand and S. Svensson, “The anaconda algorithm for deformable image registration in radiotherapy,” Medical Physics 42(1), pp. 40–53, 2015.
4D Treatment Planning Workshop 2015 - 27.11.2015 - Oncoray - Dresden, Germany - Hugo Furtado 20
Guiding structures
O. Weistrand and S. Svensson, “The anaconda algorithm for deformable image registration in radiotherapy,” Medical Physics 42(1), pp. 40–53, 2015.
4D Treatment Planning Workshop 2015 - 27.11.2015 - Oncoray - Dresden, Germany - Hugo Furtado 21
Outline
• Introduction
Motivation and objectives
• Methodology
Common framework
Examples
• Ongoing projects at the MUW
Autocontouring
Featurelets
• New trends / Ideas
• Conclusion
4D Treatment Planning Workshop 2015 - 27.11.2015 - Oncoray - Dresden, Germany - Hugo Furtado 22
Autocontouring for cervical cancer patients
• We used as atlas to contour CTs or CBCTs:
Planning CT
One weekly CT scan with comparable bladder volume
One weekly CT scans with comparable uterus position
Several atlases “stapled” together (using several comparable structures)
Same with pre delineation to guide DIR
4D Treatment Planning Workshop 2015 - 27.11.2015 - Oncoray - Dresden, Germany - Hugo Furtado 23
Autocontouring - Results
• Atlases with comparable bladder volume work better
• Effect less pronounced for the target
Figure B. Auto-contouring using the Dice Similarity Index (DSI) showing that using only the planning
CT as atlas did not result in convincing contour quality for the bladder and the target volume. Using a
scan with a comparable bladder volume the results improved.
(Data (x-axis) in chronological order).
4D Treatment Planning Workshop 2015 - 27.11.2015 - Oncoray - Dresden, Germany - Hugo Furtado 24
Autocontouring - Results
• Always bellow the intra-observer variability
• Both evaluated methods quite similar
4D Treatment Planning Workshop 2015 - 27.11.2015 - Oncoray - Dresden, Germany - Hugo Furtado 25
Contours in real life
4D Treatment Planning Workshop 2015 - 27.11.2015 - Oncoray - Dresden, Germany - Hugo Furtado 26
Contours in real life
4D Treatment Planning Workshop 2015 - 27.11.2015 - Oncoray - Dresden, Germany - Hugo Furtado 27
Head and Neck
CT week 0 CT week 4 - DIR
4D Treatment Planning Workshop 2015 - 27.11.2015 - Oncoray - Dresden, Germany - Hugo Furtado 28
Head and Neck
CT week 0 CT week 4
4D Treatment Planning Workshop 2015 - 27.11.2015 - Oncoray - Dresden, Germany - Hugo Furtado 29
Conclusion
• Results are reasonable…
• But still there are a lot of issues
Structures need correcting
DICE scores are relatively low
• Method is more elaborate and leads theoretically to better DFs but…
Cannot control optimization parameters
Cannot assess the influence of all the parts of the objective function
Method still a “grey” box…
4D Treatment Planning Workshop 2015 - 27.11.2015 - Oncoray - Dresden, Germany - Hugo Furtado 30
New developments - Featurelets
• As with other methods
Model independent - purely intensity based
The original idea contains minimization of elastic energy, though
• Piecewise rigid registration method
Image is divided in sub-regions (featurelets)
Each featurelet is registered on the target image
Deformation field constructed from the displacements
4D Treatment Planning Workshop 2015 - 27.11.2015 - Oncoray - Dresden, Germany - Hugo Furtado 31
New developments - Featurelets
Söhn et. al., “Model-independent, multimodality deformable image registration by local matching of anatomical features and minimization of elastic energy,”
Medical Physics 35(3), pp. 866–878, 2008.
4D Treatment Planning Workshop 2015 - 27.11.2015 - Oncoray - Dresden, Germany - Hugo Furtado 32
New developments - Featurelets
4D Treatment Planning Workshop 2015 - 27.11.2015 - Oncoray - Dresden, Germany - Hugo Furtado 33
Featurelets - Previous results
• Applied on 2 phantoms, lung and pelvic cases
• Comparison with commercial software
Fabri et. al., “A quantitative comparison of the performance of three deformable registration algorithms in radiotherapy,” Zeitschrift fur Medizinische Physik 23(4), pp. 279–290, 2013.
Zambrano et. al., “Performance validation of deformable image registration in the pelvic region,” Journal of Radiation Research 54(SUPPL.1), pp. i120–i128, 2013.
4D Treatment Planning Workshop 2015 - 27.11.2015 - Oncoray - Dresden, Germany - Hugo Furtado 34
Featurelets - Influence of different featurelet size
103 voxel 153 voxel 353 voxel
4D Treatment Planning Workshop 2015 - 27.11.2015 - Oncoray - Dresden, Germany - Hugo Furtado 35
Featurelets - Conclusion
• Simple method
• Results are in line with similar approaches
• Future plans:
GPU implementation
DF regularization
Fixing structures
Etc...
4D Treatment Planning Workshop 2015 - 27.11.2015 - Oncoray - Dresden, Germany - Hugo Furtado 36
Outline
• Introduction
Motivation and objectives
• Methodology
Common framework
Examples
• Ongoing projects at the MUW
Autocontouring
Featurelets
• New trends / Ideas
• Conclusion
4D Treatment Planning Workshop 2015 - 27.11.2015 - Oncoray - Dresden, Germany - Hugo Furtado 37
New trends / Ideas
• Initial Guess
Manual: how to do it?
Atlas based - find the image that most resembles your target
With automatic anatomical landmark annotation
Werner et. al., “Assessing accuracy of non-linear registration in 4d image data using automatically detected landmark correspondences,” Progress in Biomedical
Optics and Imaging - Proceedings of SPIE 8669, 2013.
4D Treatment Planning Workshop 2015 - 27.11.2015 - Oncoray - Dresden, Germany - Hugo Furtado 38
New trends / Ideas
Regional constraints: bone should not deform or should only piecewise deform
König et. al. OC-0409: Deformable image registration with guaranteed local rigidity, Radiotherapy and Oncology, Volume 115 , S197 - S198
4D Treatment Planning Workshop 2015 - 27.11.2015 - Oncoray - Dresden, Germany - Hugo Furtado 39
New trends / Ideas
Regional constraints: sliding organs or regions
Berendsen et. al., “Registration of organs with sliding interfaces and changing topologies,” Progress in Biomedical Optics and Imaging - Proceedings of SPIE 9034, 2014.
4D Treatment Planning Workshop 2015 - 27.11.2015 - Oncoray - Dresden, Germany - Hugo Furtado 40
New trends / Ideas
• Modeling the way humans deal with “invisible” organ boundaries
We make a lot of assumptions based on prior knowledge and we “interpolate”
4D Treatment Planning Workshop 2015 - 27.11.2015 - Oncoray - Dresden, Germany - Hugo Furtado 41
New trends / Ideas
“Regmentation”: joint optimization of segmentation and registration metrics
Blaffert et. al., “Bladder segmentation in mr images with watershed segmentation and graph cut algorithm,” Progress in Biomedical Optics and Imaging - Proceedings of SPIE 9034, 2014
𝒇 𝒗 = 𝜶𝑪 𝒗 + 𝜷𝑯 𝒗 + 𝜸𝑺 𝒗 + 𝜹𝑫(𝒗)
Controlling structures
ROI
4D Treatment Planning Workshop 2015 - 27.11.2015 - Oncoray - Dresden, Germany - Hugo Furtado 42
Discussion / Conclusions
• Strong need for DIR
• Many new improvements
• Still a bit disappointing…
• Many things to do in the future
4D Treatment Planning Workshop 2015 - 27.11.2015 - Oncoray - Dresden, Germany - Hugo Furtado 43
Thank you for your attention! Funding from the Austrian Federal Ministry of Science, Research
and Economy and the National Foundation for Research,
Technology and Development is greatly acknowledged.
http://www.meduniwien.ac.at/hp/radonc/
Acknowledgements