Program Information
Automatic Contour Propagation for Diagnostic and Intraoperative MR-Images Acquired During MR-Guided HDR Prostate Brachytherapy
M Borot de Battisti1*, B Denis de Senneville2,3 , M Peters1 , J R N van der Voort van Zyp1 , J J W Lagendijk1 , M A Moerland1 , (1) University Medical Center Utrecht, Department of Radiotherapy, Utrecht, The Netherlands (2) University of Bordeaux/CNRS UMR 5251, Talence, France (3) University Medical Center Utrecht, Imaging Division, Utrecht, The Netherlands
Presentations
WE-G-201-3 (Wednesday, August 2, 2017) 4:30 PM - 6:00 PM Room: 201
Purpose: Organ motions occurring during the course of MR-guided HDR prostate brachytherapy may lead to deviations between the delivered and planned dose distribution. In current practice, patient’s anatomy is assessed by manually delineating organs and tumor on MR-images acquired before and during the procedure which is time-consuming and operator-dependent. This study aims to evaluate a deformable registration tool and automatic delineation by contour propagation for diagnostic and intraoperative MR-images of patients receiving MR-guided HDR prostate brachytherapy.
Methods: The algorithm for deformable registration maximizes edge alignment between images and is based only on contrasts between tissues, which are assumed to be elastic. The method is robust to structures entering/exiting the image, e.g. a catheter, which makes it suitable for registration of diagnostic and intra-operative MR images. To evaluate the proposed tool, diagnostic and intraoperative MR-images of 6 patients (with prostate, bladder, rectum, urethra, CTV manually delineated by an experienced radiation oncologist) were registered: (I) The delineations of diagnostic MRI were propagated to pre-treatment MRI (acquired directly after catheters’ placement) and (II) the delineations of pre-treatment MRI were propagated to post-treatment MRI (acquired just before removal of catheters). In both situations, the propagated and the manual delineations of the destination image (set as references) were compared by evaluating the Dice Similarity Coefficient (DSC). The Euclidian distance from each propagated contour point to the nearest manual contour point was also calculated.
Results: The median DSC of [prostate,bladder,rectum,urethra,CTV] for all patients was [0.78,0.80,0.70,0.18,0.68] and [0.82,0.88,0.89,0.65,0.89] when propagating contours from diagnostic to pre-treatment MRI and from pre- to post-treatment MRI, respectively. Moreover, 54% and 82% of all propagated contour points were within 3mm from the manual contours, respectively.
Conclusion: The proposed registration tool shows good promises for automatically propagating contour during MR-guided HDR prostate brachytherapy.
Contact Email: