Oral Presentation 36th TROG Cancer Research Annual Scientific Meeting 2024

Whole breast knowledge-based planning predictions and transferability: a transcontinental consortia experience (#8)

Lorenzo Placidi 1 , Roberta Castriconi 2 , Peter Griffin 3 , Giovanna Benecchi 4 , Mark Burns 5 , Elisabetta Cagni 6 , Cathy Markham 7 , Valeria Landoni 8 , Eugenia Moretti 9 , Caterina Oliviero 10 , Giulia Rambaldi Guidasci 11 , Guenda Meffe 1 , Tiziana Rancati 12 , Alan Turner 13 , Alessandro Scaggion 14 , Alessia Tudda 2 , Vanessa Panettieri 7 , Karen McGoldrick 15 , Claudio Fiorino 2
  1. Fondazione Policlinico Universitario Agostino Gemelli IRCCS, Rome, Italy
  2. Medical Physics, IRCCS San Raffaele Scientific Institute, Milan, Italy
  3. Alfred Health Radiation Oncology, Melbourne, Vic, Australia
  4. Medical Physics, University Hospital of Parma AOUP, Parma, Italy
  5. Peter MacCallum Cancer Centre - Box Hill Campus, Box Hill, Vic, Australia
  6. Medical Physics Unit, Department of Advanced Technology, Azienda USL-IRCCS di Reggio Emilia, Reggio Emilia, Italy
  7. Peter MacCallum Cancer Centre - Parkville Campus, Parkville, Vic, Australia
  8. Department of Medical Physics, IRCSS Regina Elena National Cancer Institute, Rome, Italy
  9. Department of Medical Physics, University Hospital, Udine, Italy
  10. University Hospital ‘‘Federico II”, Napoli, Italy
  11. UOC di Radioterapia Oncologica, Fatebenefratelli Isola Tiberina – Gemelli Isola, Rome, Italy
  12. Data Science Unit, Fondazione IRCCS Istituto Nazionale dei Tumori, Milan, Italy
  13. Peter MacCallum Cancer Centre - Bendigo Campus, Bendigo, Vic, Australia
  14. Medical Physics Department, Veneto Institute of Oncology IOV–IRCCS, Padua, Italy
  15. Peter MacCallum Cancer Centre - Moorabbin Campus, Moorabbin, Vic, Australia

Purpose: Knowledge-based (KB) planning is a widely embraced method for ensuring high-quality and consistent treatment planning worldwide. KB models, with their transferability across institutions, promote collaboration [1]. This study builds upon previous KB model development within an Italian consortium, focusing on right whole breast (RWB) irradiation using tangential fields (TF), to evaluate the dependability and transferability of these models across international consortia.

Materials and methods: In Consortium A (Italy), ten institutes utilised RapidPlan (Varian Medical Systems, Inc.) to develop RWB-TF KB models, previously validated using common criteria [2]. Intra-consortium transferability was gauged by assessing the geometric Principal Component (PC1) overlap on 18 test patients from the other institutes. Cross-inter-consortium validation involved 20 breast patients from Consortium B (Australia), measuring transferability by ipsilateral lung PC1 within the 10th-90th percentiles. Dosimetric parameters, including ipsilateral lung mean dose and V20Gy, were calculated to assess prediction variability between consortia and compared with the clinical DVH. Predictions of ipsilateral lung DVH were categorised as optimal, suboptimal, improved, or failed based on the alignment of the clinical DVH with the predicted DVH band.

Results: In the cross-validation analysis, Consortium A exhibited high intra-institutional interchangeability (9 out of 10 models) with PC1 values within the 90th percentile, while in the international cross-validation cohort of Consortium B, 6 out of 10 models achieved the criterion. Notably, poor transferability was observed for Institute 6 within Consortium A (7 out of 18 patients) and Consortium B (7 out of 20 patients), as well as for Institute 10 within Consortium B (5 out of 20 patients). For the 20 external cross-validation patients in Consortium B, Consortium A KB models predicted ipsilateral lung DVHs as 76% optimal, 16% improved, 7% suboptimal, and 1% failed predictions. The predicted mean dose for ipsilateral lung using Consortium A KB models on the Consortium B dataset resulted in SDmean=1.1 Gy. In contrast, the clinical ipsilateral lung mean predicted dose standard deviation cSDmean=0.6 Gy, averaged across the Consortium B dataset. For ipsilateral lung V20Gy prediction, an average SDV20Gy of 2.7% was estimated, compared with the clinical Consortium B external cross-validation dataset providing a cSDV20Gy of 1.6%.

Conclusions: This study revealed consistent transferability and reliability of ipsilateral lung predictions using RWB-TF KB models. These results suggest the potential effectiveness of applying these models globally, beyond their national origin consortium, enhancing breast dosimetry and fostering international collaboration in radiation therapy planning.

  1. Panettieri V., Ball D., Chapman A., Cristofaro N., Gawthrop J., Griffin P., et al. Development of a multicentre automated model to reduce planning variability in radiotherapy of prostate cancer. Phys Imaging Radiat Oncol. 2019;11:34–40. doi: 10.1016/j.phro.2019.07.005.
  2. Tudda A et al., Knowledge-based multi-institution plan prediction of whole breast irradiation with tangential fields. Radiother Oncol. 2022 Oct;175:10-16. doi: 10.1016/j.radonc.2022.07.012. Epub 2022 Jul 19. PMID: 35868603.
  3. The current study was performed within the MIKAPOCo project, supported by AIRC Grant IG23150, 248/2021