skip to main content
OSTI.GOV title logo U.S. Department of Energy
Office of Scientific and Technical Information

Title: SU-F-J-97: A Joint Registration and Segmentation Approach for Large Bladder Deformations in Adaptive Radiotherapy

Abstract

Purpose: To improve results of deformable image registration (DIR) in adaptive radiotherapy for large bladder deformations in CT/CBCT pelvis imaging. Methods: A variational multi-modal DIR algorithm is incorporated in a joint iterative scheme, alternating between segmentation based bladder matching and registration. Using an initial DIR to propagate the bladder contour to the CBCT, in a segmentation step the contour is improved by discrete image gradient sampling along all surface normals and adapting the delineation to match the location of each maximum (with a search range of +−5/2mm at the superior/inferior bladder side and step size of 0.5mm). An additional graph-cut based constraint limits the maximum difference between neighboring points. This improved contour is utilized in a subsequent DIR with a surface matching constraint. By calculating an euclidean distance map of the improved contour surface, the new constraint enforces the DIR to map each point of the original contour onto the improved contour. The resulting deformation is then used as a starting guess to compute a deformation update, which can again be used for the next segmentation step. The result is a dense deformation, able to capture much larger bladder deformations. The new method is evaluated on ten CT/CBCT male pelvismore » datasets, calculating Dice similarity coefficients (DSC) between the final propagated bladder contour and a manually delineated gold standard on the CBCT image. Results: Over all ten cases, an average DSC of 0.93±0.03 is achieved on the bladder. Compared with the initial DIR (0.88±0.05), the DSC is equal (2 cases) or improved (8 cases). Additionally, DSC accuracy of femoral bones (0.94±0.02) was not affected. Conclusion: The new approach shows that using the presented alternating segmentation/registration approach, the results of bladder DIR in the pelvis region can be greatly improved, especially for cases with large variations in bladder volume. Fraunhofer MEVIS received funding from a research grant by Varian Medical Systems.« less

Authors:
; ;  [1];  [2]
  1. Fraunhofer MEVIS, Luebeck (Germany)
  2. Fraunhofer MEVIS, Bremen (Germany)
Publication Date:
OSTI Identifier:
22634706
Resource Type:
Journal Article
Resource Relation:
Journal Name: Medical Physics; Journal Volume: 43; Journal Issue: 6; Other Information: (c) 2016 American Association of Physicists in Medicine; Country of input: International Atomic Energy Agency (IAEA)
Country of Publication:
United States
Language:
English
Subject:
60 APPLIED LIFE SCIENCES; 61 RADIATION PROTECTION AND DOSIMETRY; ACCURACY; ALGORITHMS; BIOMEDICAL RADIOGRAPHY; BLADDER; CALORIMETRY; COMPUTERIZED TOMOGRAPHY; DATASETS; EUCLIDEAN SPACE; IMAGES; ITERATIVE METHODS; PELVIS; RADIOTHERAPY; SKELETON

Citation Formats

Derksen, A, Koenig, L, Heldmann, S, and Meine, H. SU-F-J-97: A Joint Registration and Segmentation Approach for Large Bladder Deformations in Adaptive Radiotherapy. United States: N. p., 2016. Web. doi:10.1118/1.4956005.
Derksen, A, Koenig, L, Heldmann, S, & Meine, H. SU-F-J-97: A Joint Registration and Segmentation Approach for Large Bladder Deformations in Adaptive Radiotherapy. United States. doi:10.1118/1.4956005.
Derksen, A, Koenig, L, Heldmann, S, and Meine, H. Wed . "SU-F-J-97: A Joint Registration and Segmentation Approach for Large Bladder Deformations in Adaptive Radiotherapy". United States. doi:10.1118/1.4956005.
@article{osti_22634706,
title = {SU-F-J-97: A Joint Registration and Segmentation Approach for Large Bladder Deformations in Adaptive Radiotherapy},
author = {Derksen, A and Koenig, L and Heldmann, S and Meine, H},
abstractNote = {Purpose: To improve results of deformable image registration (DIR) in adaptive radiotherapy for large bladder deformations in CT/CBCT pelvis imaging. Methods: A variational multi-modal DIR algorithm is incorporated in a joint iterative scheme, alternating between segmentation based bladder matching and registration. Using an initial DIR to propagate the bladder contour to the CBCT, in a segmentation step the contour is improved by discrete image gradient sampling along all surface normals and adapting the delineation to match the location of each maximum (with a search range of +−5/2mm at the superior/inferior bladder side and step size of 0.5mm). An additional graph-cut based constraint limits the maximum difference between neighboring points. This improved contour is utilized in a subsequent DIR with a surface matching constraint. By calculating an euclidean distance map of the improved contour surface, the new constraint enforces the DIR to map each point of the original contour onto the improved contour. The resulting deformation is then used as a starting guess to compute a deformation update, which can again be used for the next segmentation step. The result is a dense deformation, able to capture much larger bladder deformations. The new method is evaluated on ten CT/CBCT male pelvis datasets, calculating Dice similarity coefficients (DSC) between the final propagated bladder contour and a manually delineated gold standard on the CBCT image. Results: Over all ten cases, an average DSC of 0.93±0.03 is achieved on the bladder. Compared with the initial DIR (0.88±0.05), the DSC is equal (2 cases) or improved (8 cases). Additionally, DSC accuracy of femoral bones (0.94±0.02) was not affected. Conclusion: The new approach shows that using the presented alternating segmentation/registration approach, the results of bladder DIR in the pelvis region can be greatly improved, especially for cases with large variations in bladder volume. Fraunhofer MEVIS received funding from a research grant by Varian Medical Systems.},
doi = {10.1118/1.4956005},
journal = {Medical Physics},
number = 6,
volume = 43,
place = {United States},
year = {Wed Jun 15 00:00:00 EDT 2016},
month = {Wed Jun 15 00:00:00 EDT 2016}
}