MENU

Click here to

×

Are you sure ?

Yes, do it No, cancel

Feasibility of An MRI-Only Workflow for Post-Implant Dosimetry of LDR Brachytherapy in Prostate Patients

R Nosrati1,2*, M Wronski3 , W Song4 , A Pejovic-Milic1 , G Morton3,5 , G Stanisz1,2,5 , (1) Ryerson University, Toronto, Canada (2) Sunnybrook Research Institute, Toronto, Canada (3) Sunnybrook Health Sciences Centre, Toronto, Canada (4) Virginia Commonwealth University, Richmond, USA (5) University of Toronto, Toronto, Canada

Presentations

(Thursday, 7/18/2019) 7:30 AM - 9:30 AM

Room: 303

Purpose: To investigate the feasibility of an MRI-based workflow for post-implant dosimetry of prostate Low-Dose-Rate (LDR) brachytherapy in ten prostate patients with permanent seed brachytherapy.

Methods: Ten prostate cancer patients with a total of 726 implanted I-125 LDR seeds participated in this study. All patients were scanned on a 3T MR scanner (Philips Achieva) within 30 days following implantation, using a 3D multi-echo gradient echo sequence with the following parameters: TE1/∆TE/TR=2.3/2.3/8.6ms, Bandwidth=723Hz/pixel, FOV=250x250x100mm³ and 1.5mm³ isotropic resolution. The post-processing algorithm included the following steps: 1) Field map estimation and image distortions correction. 2) Quantitative Susceptibility Mapping (QSM) to generate positive contrast for the seeds. 3) Prostate segmentation on T2-weighted using fully Convolutional Neural Network (CNN- UNET). 4) Applying spatial clustering algorithms (DBSCAN and K-medoids) for seed localization within the segmented prostate volume plus 10mm margin. 5) Validation of the seed visualization and localization with the clinical CT-based approach.

Results: The calculated distortion maps showed an average distortion of 1.1mm around the seeds. Fig.1 shows sample results for five patients. The proposed MRI-based pipeline created positive contrast for all seeds with high spatial resolution. Calcifications are diamagnetic with similar magnetic susceptibility to soft tissue which makes them differentiable from paramagnetic seeds on QSM; the arrows in Fig.1, patient 5, indicate the calcification on CT that was invisible on QSM. The automated prostate segmentation using CNN-UNET resulted in an average Dice score of 0.84. All implanted seeds were correctly identified; The maximum and average distance between MR and CT-derived seed positions were 2.8mm and 1.3±1.4mm respectively. The spatial accuracy of the QSM-based seed localization was improved by 1.3±0.9mm by applying the distortion correction prior to QSM.

Conclusion: Seed depiction on QSM is comparable to CT thus QSM has great potential to exclude CT from the current standard CT-only or CT/MR fusion practices.

Keywords

Not Applicable / None Entered.

Taxonomy

Not Applicable / None Entered.

Contact Email