Session: Brachytherapy - Salle de bal et foyer

le 24 juin 2022 from 11h00 EST to 12h00 EST

Scientific Session 1 – Brachytherapy
Friday, June 24, 2022, 11:00-12:00

Scientific Session 1:Brachytherapy – Presentation 1

Towards clinical patient-specific Monte Carlo dose calculations for brachytherapy: egs_brachy developments

Narjes Moghadam, Martin Martinov, Shannon Jarvis, Claire Zhang, Samuel Ouellet, Michael Thibodeau, Joanna Cygler, Michelle Hilts, Luc Beaulieu, Rowan Thomson
Carleton University, The University of British Columbia Okanagan, Université Laval, The Ottawa Hospital Cancer Centre

Purpose: We present egs_brachy developments towards clinical implementation of advanced TG186 model-based dose calculations for permanent-implant prostate and breast brachytherapy, to overcome TG43 limitations.

Methods: A graphical user interface (GUI) for egs_brachy, eb_gui, is developed using the Qt5 C++ toolkit.  The eb_gui enables users to convert clinical DICOM (CT, RTSTRUCT, RTPLAN) files to EGSnrc-compatible formats, generate full-tissue virtual patient models, and execute egs_brachy treatment simulations that include previously-benchmarked seed models.  Furthermore, detailed 3D dose distributions, dose volume histograms, and clinical metrics are generated.  Following the recommended TG-186 commissioning process, idealized and patient-specific test cases were designed. Dose distributions and dosimetric parameters for these cases were then calculated and validated. 

Results: The eb_gui provides an effective bridge that enables data in clinical format to be readily used for advanced Monte Carlo simulations with egs_brachy. Idealized test cases plus prostate and breast patient-based simulations were successfully developed, tested and validated based on TG186 recommendations. The reliability of eb_gui is demonstrated through comparison of MC results with clinical treatment planning system (TPS) where good agreement is illustrated for both breast and prostate cases.

Conclusion: We have successfully implemented and tested eb_gui, an interface for bridging egs_brachy with clinical brachytherapy DICOM data.  The eb_gui code has been released as free open-source software, offering a readily-accessible package to improve the accuracy of clinical dose calculations via more accurate MC dose calculations compared to the TG43 algorithm.  Future work includes a retrospective study of LDR dosimetry from patient plans across Canadian centers.


Scientific Session 1:Brachytherapy – Presentation 2

A post-implant dosimetry simulator for permanent breast seed implant brachytherapy

Claire Zhang, Deidre Batchelar, Jeff Andrews, Juanita Crook, Michelle Hilts
BC Cancer, UBC Okanagan

Purpose: Permanent breast seed implant (PBSI) brachytherapy is a promising treatment that has the potential to be widely utilized with increased standardization, optimization, and robustness. In this study we develop and validate a PBSI post-implant dosimetry simulator.  The simulator is based on previous seed displacement data that was utilized to quantify plan robustness to implant uncertainties. This simulator will enable evaluation of optimal PBSI plans and facilitate future studies on PBSI planning optimization.

Methods: The PBSI simulator was developed in Matlab (2020) by sampling displacement from previous clinical data, applying simulated displacement to planned seeds position, calculating target dose using TG43 formulism, and computing a range of possible post-implant dosimetry outcomes under various seed displacement scenarios. Dose calculation and dose-volume histogram (DVH) computation was validated by comparing to published results and MIM Symphony (v6.9). Simulated dosimetry of evaluated target volume was compared to clinical post-implant dosimetry for 10 cases as validation.

Results:  Dose calculation uncertainty at distance ≥1cm was < 2%. Mean DVH computation uncertainty was <1% for V90 and V100, and < 4% for V150 and V200. On average, simulated median target dose showed <1% deviation for V90 and V100, and < 4% deviation for V150 and V200 from clinical post-implant dosimetry for 10 cases.

Conclusions: A PBSI post-implant dosimetry simulator was developed and validated as a tool to simulate post-implant dosimetry from PBSI plans. This simulator can be used to evaluate plan robustness to seed displacement and will facilitate future research in improving PBSI planning methods.


Scientific Session 1:Brachytherapy – Presentation 3

Radiobiological modeling of Pd-103 permanent breast seed implant brachytherapy using Monte Carlo dose calculations

Michael Thibodeau, Michelle Hilts, Deidre Batchelar, Juanita Crook, Rowan Thomson
Carleton University, BC Cancer, University of British Columbia

Purpose: Assess tumour control probability (TCP) and normal tissue complication probability (NTCP) for Pd-103 permanent breast seed implant (PBSI) brachytherapy by coupling radiobiological models to dose evaluations carried out with TG43 assumptions and advanced patient-specific TG186 models.

Methods: This preliminary study considers PBSI treatments for 3 patients. Virtual patient and treatment models are developed from DICOM data (CT, RTSTRUCT, RTPLAN) under different assumptions: (a) “TG43”: all water, no interseed attenuation; (b) “TG186”: realistic tissues/seeds modelled. Monte Carlo simulations using the EGSnrc application egs_brachy then generate two 3D dose distributions for each patient; dose volume histograms and metrics are extracted. Various radiobiological models are then coupled to the dose distributions to evaluate TCP and NTCP.

Results: Dose distributions from both formalisms are considerably different, leading to noticeable differences in radiobiological indices.  For each of the three patients, TG43 models overestimate the dose delivered to the target, as seen from D90= 80.1 Gy in one case, compared to the TG186 model where D90= 71.9 Gy for the same patient. Normal tissues such as the ipsilateral lung, heart, and skin, display the opposite trend: TG43 models predict lower doses than TG186. These trends are reflected in the calculations of TCP for the TG43 models which are 18.7% higher in one case, and NTCP lower by 15.1% for the heart.

Conclusion: This preliminary study demonstrates discrepancies in radiobiological indices computed using traditional TG43 and advanced TG186 models.  Future work aims to investigate the impact of patient-specific dose calculations using a full cohort of patients.


Scientific Session 1:Brachytherapy – Presentation 4

Characterization of patient specific template-based hybrid brachytherapy for treatment of gynecological cancers

Michael Kudla, Felipe Castro, Jocelyn Moore, Deidre Batchelar, Francois Bachand
The University of British Columbia, BC Cancer - Kelowna

Purpose: Interstitial vaginal and pelvic brachytherapy is standard of care in locally advanced vaginal cancers and vaginal recurrences. We have developed an in-house design and 3D printing workflow for the creation of patient-specific cylindrical templates (PSCTs) which facilitate IC/IS HDRBT. Geometric and dosimetric reproducibility of preplans is evaluated.

Methods: Seven patients with primary or recurrent vaginal malignancies were treated with IC/IS HDRBT facilitated by a PSCT.  Pre-plans were exported, and 3D modeled using in-house written software. After treatment, needle insertion and dosimetry statistics were compared to pre-insertion plans.

Results: The treated median HRCTV D90 (%), V100(%), V150(%), V200(%), Bladder, Rectum, Sigmoid and Small Bowel D2cm3, were 110.8%, 96.7%, 52.8%, 28.0%, 18.0 Gy, 15.4 Gy, 11.0 Gy, and 12.9 Gy respectively. The median differences of the final treated plan (post-plan) to the preplan were 1.1%, -0.5%, 13.1%, 4.1%, 1.0 Gy, 0.8 Gy, 0.5 Gy, and -0.8 Gy respectively. The median insertion displacement of dwell positions for straight needles, curved needles, and overall was 1.3mm, 4.0mm, and 2.3mm, respectively. 

Conclusions: 3D printed patient-specific templates for the guidance of interstitial high dose rate brachytherapy of the vagina can be used to preplan and deliver highly conformal IC/IS Vaginal HDRBT treatments. These templates can be used to create higher quality treatment plans as compared to state of the art techniques, particularly for deep, localized, and top-of-vault lesions, improving target coverage, reducing OAR dose, reduce interstitial needle length and trauma in non-tumor tissue and improve confidence in needle insertion without intraoperative imaging.


Scientific Session 1:Brachytherapy – Presentation 5

Results of a Prospective Applicator and Hybrid Interstitial Needle Selection Machine Learning Model for High-Dose-Rate Cervical Brachytherapy

Kailyn Stenhouse, Michael Roumeliotis, Robyn Banerjee, Kevin Martell, Philip McGeachy
University of Calgary - Tom Baker Cancer Centre

Purpose: To validate a trained machine learning (ML) model as a decision-support tool for applicator and needle arrangement for cervical brachytherapy.

Methods: Six patients receiving high-dose-rate cervical brachytherapy were enrolled in this prospective study. For each patient, the primary radiation oncologist (RO) contoured the high-risk clinical target volume (HR-CTV) on a pre-brachytherapy MRI, as well as indicated the expected approximate location of the ring plane and tandem axis post-applicator insertion. HR-CTV geometry metrics were features used in a previously trained ML model. Patients were treated using the applicator selected by the RO, and this decision was compared to the model prediction. For all patients, applicator selection, needle number, and arrangement were compared. In patients with applicator selection disagreement, a retrospective treatment plan was created to compare dosimetry.

Results: There was agreement in applicator selection between the RO and ML model predictions in five of the six patient cases. For patient cases where a hybrid interstitial applicator was used (4,5,6), needle number and arrangement were similar. Patient four had full agreement between the predicted positions while predictions for patients five and six differed by a single needle translation or addition. For the disagreement, replanning indicated that using the ML-predicted hybrid interstitial applicator and needle arrangement reduced bowel D2cc (2.20 Gy BED), permitted increased dose to the HR-CTV (1.35 Gy BED), and maintained comparable dose to other volumes.

Conclusions: Preliminary results show good agreement between RO and ML predictions. Prospective implementation as a decision-support tool may improve dosimetric quality of brachytherapy implants.


Scientific Session 1:Brachytherapy – Presentation 6

Systematic Monte Carlo dose recalculation adapted to the dosimetry data pipeline: a use-case in brachytherapy

Samuel Ouellet, Yannick Lemaréchal, Francisco Murillo, Martin Martinov, Marie-Claude Lavallée, Rowan Thomson, Philippe Després, Luc Beaulieu
Université Laval, Carleton University, CHU de Québec

Purpose: Monte Carlo simulations are a key component to study the relationship between the tissue response and dose delivered. A systematic MC dose recalculation would provide a statistically significant sample allowing a reassessment of the dose-response relationship. The goal of this work is to build a systematic Monte Carlo dose recalculation adapted to the dosimetric data pipeline, in this case, applied to brachytherapy.

Methods: The MC dose recalculation automatically extracts the information stored in every DICOM RT instances to launch dose calculation with the TOPAS MC code or with egs_brachy. The parameters of the Monte Carlo simulation, the dose distributions, the uncertainty and organ segmentation are stored into the DICOM using appropriate tags or containers. The whole process is encapsulated into a docker image and is tested in the data streamline of an existing numerical infrastructure.

Results: The integration of the Monte Carlo dose recalculation into the dosimetric data pipeline allowed a weekly production rate of 80 dose distributions (1e9 photons/patient at 0.8 · 1e6 histories/min·core) without any human intervention. The automation of the MC dose recalculation does not introduce errors in the simulations because both manually and automatically generated dose maps agree within their uncertainties.

Conclusions: The systematic Monte Carlo dose recalculation built for brachytherapy stores all the produced information and the details of the process that leads to it in the DICOM format, ensuring the integrity of the data. This self-contained data can potentially allow reassessments of the dose-response relationship and machine learning training.


Scientific Session 1:Brachytherapy – Presentation 7

Radiobiological dose evaluation for uveal melanoma plaque brachytherapy: influence of the alpha/beta ratio

Geetha Menon, Ezekiel Weis, Albert Murtha, Matthew Larocque
Cross Cancer Institute, University of Alberta

Purpose: To evaluate the magnitude of dosimetric variation produced by differences in the selection of alpha/beta ratios during biological effective dose (BED) calculation in I-125 plaque brachytherapy (PBT) for uveal melanoma. 

Methods: Tumour and organs-at-risk (OARs; lens, optic nerve head (ONH), fovea) BEDs were calculated using the formalism for low-dose-rate temporary implant, employing both conventional alpha/beta ratios (BEDconv; in Gy; 10 for tumour and 3 for OARs) and those reported in literature for the eye (BEDeye; 11.5=tumour, 1.2=lens, 1.75=ONH, 2.58=fovea). Differences in BED were evaluated based on implant days (4 (BED4days) vs 7 (BED7days); removal times based on patient location/operating room availability, and not on tumour-specific characteristics; n=165 in each group) and further by tumour height (£5mm vs >5mm). Descriptive statistics were performed.

Results: All 330 patients included in the study, were prescribed 70 Gy (absolute dose) to the tumour apex (mean height: 5.6±2.4mm). Significant differences (p<0.05) were observed between BEDconv and BEDeye for tumour and ONH for both implant durations. The BED7days/BED4days for tumour was on average 0.88. Compared to 4 days, a 7 day implant reduced ONH and fovea BEDs by almost half. For both implant durations, when separated by tumour height, significant BED differences were observed only for the lens; higher lens doses for taller tumours.

Conclusions: The calculation of PBT BEDs, identified as a predictor of tumour control and OAR toxicity, was seen to be strongly dependent on the choice of the alpha/beta ratio, warranting the need for accurate radiobiological modeling of individual eye tissues.