No date

Session Item

Sunday

August 29

08:45 - 10:00

N104

Normal tissue effects of combined therapies

Chair: , Denmark;

Chair: Navita Somaiah, United Kingdom

Overview: Recent years have seen an increasing use of radiotherapy in combination with novel treatment modalities such as molecular targeted agents. However, little is known about the differential response of multiple treatment modalities on normal tissue and the impact this might have on the development of early and late radiation-related side effects. This symposium offers an overview of: • A selection of preclinical studies focusing on the combination of radiation with molecular therapies such as anti-angiogenic drugs and DNA damage response inhibitors. • The development of in vitro and in vivo normal tissue models, with an emphasis on lung, to evaluate both tumour control and normal tissue damage.

Session Type: Symposium

Track: Radiobiology

Journey:

Commissioning of a GPU-based multi-criteria optimisation algorithm for HDR brachytherapy

Presentation Number: PP-0150

Abstract

Abstract Title:

Commissioning of a GPU-based multi-criteria optimisation algorithm for HDR brachytherapy

Authors:

Cédric Bélanger¹, Éric Poulin², Sylviane Aubin³, J. Adam M. Cunha⁴, Luc Beaulieu¹

¹CHU de Québec - Université Laval, 1.Département de Physique, de génie physique et d'optique, Université Laval, Québec, QC, Canada, 2.Département de Radio-Oncologie, Centre de Recherche du CHU de Québec – Université Laval, Québec, QC, Canada, Québec, Canada; ²CHU de Québec - Université Laval, 2.Département de Radio-Oncologie, Centre de Recherche du CHU de Québec – Université Laval, Québec, QC, Canada, Québec, Canada; ³CHU de Québec - Université Laval, 2.Département de Radio-Oncologie, Centre de Recherche du CHU de Québec – Université Laval, Québec, QC, Canada, Québec, Canada; ⁴University of California in San Francisco, 3.Department of Radiation Oncology - Division of Physics, UCSF, SF, CA, USA, San Francisco, USA

Show Affiliations

Purpose or Objective

In our previous study, we evaluated the impact of combining our graphics processing unit (GPU)-based multi-criteria optimisation (gMCO) algorithm with an interactive graphical user interface (gMCO-GUI) for high-dose-rate (HDR) brachytherapy. gMCO-GUI allows the planner to navigate in real-time through gMCO plans pool of Pareto-optimal plans. In this work, we report on the commissioning of this algorithm using a commercial clinically-validated treatment planning system (TPS) as our baseline.

Material and Methods

gMCO algorithm implements on GPU the TG43-2D formalism, L-BFGS optimizer, DVH curves and 3D dose matrices computations. Hence, gMCO allows the generation of 1000 Pareto-optimal treatment plans within 10 s. Our MCO workflow was commissioned against Oncentra Prostate v4.2.2 (OcP) (Elekta, Veenendaal, The Netherlands) by using 14 HDR brachytherapy prostate cancer patients. The target, bladder, rectum and urethra structures were delineated from ultrasound images and used for DVH evaluation. For the DVH computations, the random sampling point method as described in the OcP physics manual was implemented in gMCO. In both gMCO and OcP, 50000 points were sampled in each structure, stored in histograms containing 1000 bins with a maximum dose of 400% of the prescribed dose. With gMCO, 1000 plans/case were generated and a single treatment plan (highest target coverage while meeting clinical goals) was selected. The optimized dwell times of the selected plan were exported in the DICOM-RTPLAN format from gMCO-GUI. The final dosimetry of gMCO plans was calculated in OcP. Paired two-sided Wilcoxon signed-ranked tests were used to compare the dosimetric results between gMCO and OcP.

Results

When comparing the TG43 calculations, differences less than 0.1% were observed between gMCO and OcP for different radius (0.1 cm to 10.0 cm) and angles (0, 45, 90, 135, and 180). Furthermore, when comparing the dose to a point with 2 catheters and 2 dwell positions/catheter, a difference less than 0.01% was obtained. Over all gMCO and OcP calculated structure volumes, linear regressions provided a R²=1 with slopes ranging from 0.99 to 1.00 (mean differences within 0.3%). Fig. 1 illustrates the DVH curves calculated with gMCO and OcP for a random case.

When looking into dosimetric indices, no statistically significant differences (p>0.05) were observed in the target V₁₀₀, the urethra D₁₀, the bladder V₇₅ and rectum V₇₅. To validate the 3D dose calculation, a physicist carefully reviewed the isodose lines slice by slice in gMCO and OcP for one case. Fig. 2 illustrates the isodose lines for one gMCO plan imported in OcP.

Conclusion

The GPU-based dose calculation algorithms (TG43, DVH, 3D dose) implemented in gMCO algorithm were evaluated and agreed with the ones integrated in OcP. While slight differences were observed, integrating these tools in the clinical TPS would allow faster calculations for the plan generation and generalize their usage in the clinical workflow by supporting GPU-based MCO algorithms.