2024 Program with Abstracts

Session 2B: Brachytherapy - Campania A

le 6 juin 2024 from 15h00 CDT to 16h00 CDT

Scientific Session – Brachytherapy 
Thursday, June 6, 2024, 15:00-16:00


Scientific Session 2B: Brachytherapy – Presentation 1 

Establishment of a spectral measurement protocol for determination of the inter- and intra source variability of the Xoft electronic brachytherapy source.

Azin Esmaelbeigi, Nada Tomic, Shirin A. Enger
McGill University

Purpose: Discrepancies in the material composition and geometry of the Xoft electronic brachytherapy due to the manual manufacturing of the sources, cause varialbility in the source output. In a previous study we developed a Monte Carlo-based software, E-Brachy, to characterize the Xoft source and quantify the discrepancies in its photon spectrum and dosimetric properties. This study aimed to establish a measurement protocol for the spectrometry of these sources, and compare the results with simulations to detect the intra-model discrepancies.

Methods: WA CdTe gamma and X-ray detector was purchased and used for the measurements. The detector was calibrated using 137Cs, 152Eu, and 57Co characteristic peaks. A source holder was designed and 3D-printed to stabilize the source on the optical table. The detector was mounted 15.5 cm from the source tip. These measurements were performed for 5 different sources of S7500 model both at the tip and the side of the source.

Results: An average difference of 6.3% with a range of 0 to 44% in the intensity of the measured spectra was observed between the sources at the tip. This value decreased to 6% with a range of 0 to 30% at the side of the source. The differences are more prominent in the characteristic energy peaks of the anode and target materials.

Conclusion: A measurement protocol was developed to be combined with a Monte Carlo-based software to investigate the inter-and-intra-source variability for the Xoft sources. The x-ray spectra emitted within the five Xoft sources of the same model show a significant discrepancy, which confirms the intra-model variability in the source material composition and construction.


Scientific Session 2B: Brachytherapy  – Presentation 2

Development of a dose optimization protocol for the clinical translation of rectal intensity-modulated brachytherapy.

Jonathan Kalinowski, Alana Thibodeau-Antonacci, Té Vuong, Shirin, Enger
McGill University, Jewish General Hospital

Purpose: We have designed a novel applicator and delivery system for high dose-rate endorectal intensity-modulated brachytherapy (IMBT), featuring a dynamically rotating 180° tungsten shield. A retrospective treatment planning investigation showed that, compared to current endorectal brachytherapy, rectal IMBT decreases dose to organs at risk for equivalent tumour dose. We aim to develop a clinically viable, IMBT-specific dose optimization protocol to take advantage of the rotational degree of freedom to maximize contralateral sparing.

Methods: Retrospective rectal IMBT dosimetry was performed with the Monte Carlo-based treatment planning system, RapidBrachyMCTPS. One fraction of a patient who received endorectal brachytherapy at our institution was selected. Four dose optimization scenarios were considered: “Clinical”, using only contours created during clinical treatment planning; “Anatomical”, using anatomical contours; “ContralateralBox”, replacing the anatomical contours with a contralateral dose spill region adjacent to the applicator; and “ContralateralLateralBoxes”, adding in two dose spill regions lateral to the applicator to the previous setup.

Results: Compared to the Clinical optimization method, relevant OAR dose metrics were reduced by 20%-60% for the other three methods evaluated. Treatment time was also reduced by about 15%. The Anatomical and ContralateralBox schemes provided comparable sparing, while the ContralateralLateralBox method sacrificed contralateral sparing for lateral rectal sparing.

Conclusions: Dose constraints on the contralateral rectum are necessary to provide optimal OAR sparing with IMBT. Performance of these four optimization methods will be evaluated on a larger cohort of 10 patients.

Scientific Session 2B: Brachytherapy  – Presentation 3

Generation of 3D reference dosimetric datasets towards adoption of model-based dose calculations for permanent implant prostate brachytherapy.

Fatemeh Akbari, Samuel Ouellet, Narjes Moghadam, Vasiliki Peppa, Sandra Oliver, Vicent Gimenez-Alventosa, Luc Beaulieu, Javier Vijande, Rowan M. Thomson
Carleton University, University of Athens, Universitat Politècnica de València, Université Laval

Purpose: To develop a foundational set of test cases for prostate Low-Dose Rate (LDR) brachytherapy to support the commissioning of Model-Based Dose Calculation Algorithms (MBDCAs), as per the recommendations of the joint Working Group of the American Association of Physicists in Medicine (AAPM), the European Society for Radiotherapy and Oncology (ESTRO), and the Australasian Brachytherapy Group (ABG) on MBDCAs in Brachytherapy and TG 186.

Methods: This study comprises four test cases: (1) a single seed in water, (2) fifty-eight seeds in water (TG43 calculation), (3) fifty-eight seeds within a realistic virtual patient model, and (4) similar to (3) without prostate calcifications, all using 125I OncoSeed6711. Simulations are carried out with four Monte Carlo (MC) codes: EGSnrc with eb_gui (egs_brachy graphical user interface), MCNP, TOPAS, and penRed/PENELOPE. Doses are compared on a voxel-by-voxel basis.

Results and Discussions: The air-kerma strength per history factor ratios between MCNP and EGSnrc and between penRed/PENELOPE and EGSnrc are 0.994±0.007 and 0.998±0.007, respectively, signifying good agreement among the simulations. Simulations with all codes have more than 90% of voxels with a global difference under ±2%, indicating that any minor difference reported with respect to the reference dataset would not be clinically significant. Local differences are seen due to the various cross-section libraries and radiation transport algorithms utilized.

Conclusions: This work provides preliminary results of the test cases created to compare 125I prostate LDR treatment between different MBDCAs. The data will be integrated into the Brachytherapy Source Registry (IROC-Houston).


Scientific Session 2B: Brachytherapy  – Presentation 4 

Modeling the 224Ra Alpha DaRT Source to Investigate Nuclei Desorption.

Ali Asghar Mowlavi, Yuji Kamio, Bijan Jia, Reza Shamsabadi, Jean-François Carrier
CRCHUM, Montréal, Québec, Canada & Department of Physics, Hakim Sabzevari University, Sabzevar, Iran & Département de physique, Université de Montréal

Purpose: In this study, we performed an analysis of the exact geometry of the Alpha DaRT source using GEANT4. Two distributions of 224Ra have been incorporated: a thin layer of 224Ra beneath a 1nm steel layer, and a half-Gaussian function with a standard deviation (σ) of 1.4nm to achieve 45% desorption for 220Rn.

Methods: By considering 224Ra radioactive nuclei as a source, we evaluated the release fractions for daughter nuclei. By increasing the thickness of a thin steel layer to 0.05 nm on the radium layer and adjusting the σ value of the half-Gaussian, calculations were carried out to achieve a released fraction for 220Rn of approximately 0.45 for the source. The resulting steel layer on radium have a thickness of 1nm, and σ=1.4nm respectively.

Results: The results of fraction released from the source are obtained for 220Rn, 216Po, 212Pb, 208Tl and 208Pb as 0.45588, 0.14202, 0.075917, 0.01486 and 0.03645 for 1nm steel, and 0.45314, 0.14104, 0.07616, 0.01489 and 0.03667, for the half-Gaussian distribution respectively. As recoil causes some atoms to undergo desorption from the source, the remaining nuclei penetrate deeper into the steel. The average depths of recoil daughter nuclei increase from 6.79nm for 220Rn to 19.63nm for 208Pb, with intermediate values of 11.29nm for 216Po and 18.46nm for 208Tl.

Conclusions: The results indicate that the released fraction of daughter nuclei from the source is the same for both geometries, with a negligible difference. We also computed the integrated activity of each radionuclide within the source and the tumor.


Scientific Session 2B: Brachytherapy – Presentation 5

High-Risk Clinical Target Volume Auto-Contouring on MRI for Cervical Cancer Brachytherapy Using 3D U-Net Transfer Learning.

Fletcher Barrett, Sarah Quirk, Roberto Souza, Kailyn Stenhouse, Kevin Martell, Michael Roumeliotis, Philip McGeachy
University of Calgary, Brigham and Women's Hospital, Johns Hopkins University

Purpose: To develop an MRI-based tool that contours the high-risk clinical target volume (HR-CTV) for cervical cancer patients prior to high-dose-rate (HDR) brachytherapy applicator insertion using transfer learning.

Methods: Pre-insertion T2-weighted MRIs from patients treated with ring and tandem-based HDR cervical brachytherapy between 2015 and 2020 formed a retrospective cervical dataset. Transfer learning models were pretrained on T2-weighted MRIs from the 2023 Brain Tumour Segmentation (BraTS) challenge using original or modified 3D U-Net architectures – modified architectures used fewer layers and channels than the original U-Net. Models were trained to maximize the Dice Similarity Coefficient (DSC). Models with the highest validation DSC were fine-tuned on the cervical dataset using clinical HR-CTV contours. Models were compared based on their average validation and testing DSC.
A model without transfer learning built using the original U-Net was compared to the best transfer learning model using Wilcoxon signed-rank testing.

Results: The retrospective dataset included 80 patients with cervical cancer and the BraTS dataset consisted of 1000 patients with meningioma.
Transfer learning models built with the original U-Net architecture performed better than the modified architectures. The best transfer learning model yielded significantly higher (p < 0.001) training/validation/testing DSCs (0.71±0.04, 0.70±0.03, 0.64±0.03) compared to the model without transfer learning (0.55±0.06, 0.59±0.05, 0.55±0.05).

Conclusions: HR-CTV auto-contouring on MRI was significantly improved using transfer learning. This demonstrates the benefit of transfer learning for the limited number of patients typical in brachytherapy datasets. Future work will involve pretraining models on a pelvic MRI dataset to achieve more clinically acceptable results.

Scientific Session 2B: Brachytherapy  – Presentation 6

Dosimetric evaluation of TG-43 formalism in intraoperative radiotherapy of glioblastoma.

David Santiago Ayala Alvarez, Peter G F Watson, Marija,Popovic, Veng Jean Heng, Michael D C Evans, Valerie Panet-Raymond, Jan Seuntjens,
McGill University, University Health Network

Purpose: To evaluate the TG-43 formalism as an accurate and clinically accessible alternative to the vendor-provided TARGIT method for dosimetry in intraoperative radiotherapy (IORT) of glioblastoma, with the potential to improve treatment planning and reduce side effects.

Methods: We retrospectively analysed IORT plans of 20 glioblastoma patients (including hypothetical IORT plans for the control group), comparing INTRABEAM dose calculations using TG-43, TARGIT, MC dose to water (MCw), and MC dose to heterogeneous media (MChet). Dose-volume histograms (DVHs) were generated for organs at risk (OARs), and combined IORT and external beam radiotherapy (EBRT) doses were assessed against dose constraints. Statistical analyses employed Wilcoxon signed-rank tests.

Results: TG-43 dosimetry yielded OAR doses closely matching the MCw reference, demonstrating excellent agreement with water-based MC simulations. In contrast, TARGIT consistently underestimated doses across OARs (p < 0.05) compared to TG-43. MChet, accounting for tissue heterogeneity, showed higher doses for certain OARs, likely due to dose variations within bony regions. EBRT combined with TG-43 or TARGIT remained within dose constraints, while EBRT with MChet exceeded them in some instances, highlighting the importance of accurate dose assessment.

Conclusions: Our findings support the use of the TG-43 formalism for IORT dosimetry with the INTRABEAM system, offering improved accuracy compared to TARGIT. While tissue heterogeneity in specific OARs may warrant MC simulations, the TG-43 approach provides a valuable and efficient alternative in many clinical settings, potentially optimizing IORT treatment planning and reducing side effects in glioblastoma patients.