Premium
SU‐F‐BRF‐14: Increasing the Accuracy of Dose Calculation On Cone‐Beam Imaging Using Deformable Image Registration in the Case of Prostate Translation
Author(s) -
Fillion O,
Gingras L,
Archambault L
Publication year - 2014
Publication title -
medical physics
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.473
H-Index - 180
eISSN - 2473-4209
pISSN - 0094-2405
DOI - 10.1118/1.4889083
Subject(s) - cone beam computed tomography , image registration , imaging phantom , cone beam ct , nuclear medicine , translation (biology) , prostate , medicine , prostate cancer , computer science , radiation treatment planning , centroid , medical imaging , fiducial marker , computed tomography , artificial intelligence , radiology , radiation therapy , image (mathematics) , cancer , biochemistry , chemistry , messenger rna , gene
Purpose: Artifacts can reduce the quality of dose re‐calculations on CBCT scans during a treatment. The aim of this project is to correct the CBCT images in order to allow for more accurate and exact dose calculations in the case of a translation of the tumor in prostate cancer. Methods: Our approach is to develop strategies based on deformable image registration algorithms using the elastix software (Klein et al., 2010) to register the treatment planning CT on a daily CBCT scan taken during treatment. Sets of images are provided by a 3D deformable phantom and comprise two CT and two CBCT scans: one of both with the reference anatomy and the others with known deformations (i.e. translations of the prostate). The reference CT is registered onto the deformed CBCT and the deformed CT serves as the control for dose calculation accuracy. The planned treatment used for the evaluation of dose calculation is a 2‐Gy fraction prescribed at the location of the reference prostate and assigned to 7 rectangular fields. Results: For a realistic 0.5‐cm translation of the prostate, the relative dose discrepancy between the CBCT and the CT control scan at the prostate's centroid is 8.9 ± 0.8 % while dose discrepancy between the registered CT and the control scan lessens to −2.4 ± 0.8 %. For a 2‐cm translation, clinical indices like the V90 and the D100 are more accurate by 0.7 ± 0.3 % and 8.0 ± 0.5 cGy respectively when using registered CT than when using CBCT for dose calculation. Conclusion: The results show that this strategy gives doses in agreement within a few percents with those from calculations on actual CT scans. In the future, various deformations of the phantom anatomy will allow a thorough characterization of the registration strategies needed for more complex anatomies.