Analysis of bladder cancer treatment plans generated for online adaptive radiotherapy and validation of an integrated independent dose calculation software

Abstract: Purpose: In 2019 Varian EthosTM linear accelerator (Varian Medical Systems, CA, USA) with integrated artificial intelligence (AI) was installed at Herlev Hospital, Copenhagen, Denmark, enabling online adaptive radiotherapy (oART) e.g for bladder cancer patients with a large margin reduction. The AI utilizes so called influencers for initial delineation of the targets and organs at risks (OAR) using reference planning computed tomography (CT) images as a guide. The system generates two plans for the treatment, of which one is re-calculated (scheduled plan) and the other is re-optimized (adapted plan) on the anatomy of the day. Moreover, this treatment approach makes traditional phantom based quality as- surance (QA) unavailable. This project was focused on three aims; the reduction of target volume achieved utilizing oART approach instead of conventional image-guided radiation therapy (IGRT); quality analysis of the adapted and scheduled plans; validation of a a com- mercial solution for an integrated independent dose calculation. Materials and Methods: In total 6 bladder cancer patients, treated with oART, were in- cluded in the study. Volumetric and dosimetric analysis were performed retrospectively for 102 adapted and scheduled plans in terms of target coverage and absorbed dose to OAR, such as Bowel Bag and Rectum. The results were compared to corresponding conventional IGRT plans with regular margins used in the clinic. Additionally, effect of bladder volume variations on absorbed dose to OAR was investigated. Conformity index (CI) and homo- geneity index (HI) were computed and compared for adapted and scheduled plans. Furthermore, 54 adapted plans were automatically transferred to Mobius Adapt (Varian Medical Systems, CA, USA) for verification. Global gamma indexes were calculated using the following criteria : 3%/3 mm, 3%/2 mm and 2%/2 mm, 10% threshold. Results from Mobius Adapt calculations were compared to measured local gamma indexes 3%/2 mm, 20% threshold with Delta4+ phantom (Scandidos, Upsalla, Sweden ). Results: oART enabled 13 %- 59 % planning target volume (PTV) reduction compared to the conventional approach for the studied patients. For scheduled plans, the results revealed an impact on absorbed dose to Bowel Bag (R2> 0.7) for cases with large bladder volume variations (median (Q2) >155 cm3, interquartile range (IQR) >60 cm3) during the course treatment, but not to Rectum (R2< 0.3). Whereas smaller bladder volume variations (Q2 < 100cm3, IQR< 30 cm3) did not affect the dose received neither to Bowel Bag or Rec- tum (R2< 0.4, R2< 0.2). A statistically significant improvement in homogeneity (p< 0.05, two patients p= 0.81, p= 0.42) and conformity (p< 0.05) was obtained in the adapted plans compared to the scheduled. No significant difference (p= 0.82, others p < 0.001) was ob- served between Delta4+ local gamma index 3%/2 mm and Mobius Adapt 3%/3 mm. Conclusion: Adaptation had significant impact on absorbed dose to OAR for all studied patients enabling target coverage, while reducing the PTV volume . An independent dose calculation with Mobius Adapt has been validated for dose verification during oART, where traditional phantom-based approaches are not available.