ESTRO 2022 - Abstract Book

S648

Abstract book

ESTRO 2022

A head and neck case was considered for illustration. It involved air, bone, teeth, and implants outside the CTV. Accelerated simultaneous boost was delivered to three PTVs (54, 60, and 70 Gy) in 33 fractions using VMAT technique. For simplicity, we focused on PTV70Gy and the mandible as organ-at-risk (OAR). Acuros XB and AAA algorithms were used to obtain Dm,m and Dw,w distributions, respectively. First, a plan was optimised to deliver uniform photon fluence to the PTV -as in Dw,w planning- and calculated with Dm,m. Second, another plan was optimised to achieve uniform Dm,m. It was recalculated in Dw,w to analyse the clinical differences with previous practice. Additionally, robustness was evaluated. Results Figure 1 shows the results of both planning strategies. Uniform fluence produced cold spots in bone and implants. Uniform Dm,m compensated them by locally increasing fluence, resulting in higher doses beyond Dw,w acceptability criteria.

Figure 2 shows the differences between homogeneous Dm,m and Dw,w. Doses were 1% higher for the target (muscle CTV), and up to +4% for the mandible, thus increasing toxicity risk. Robustness was impaired when the fluence increases and the heterogeneities did not match.

Conclusion Planning with Dm,m as with Dw,w can be disruptive.

There are systematic differences in dose prescription and constraints. For water-like tissues, 1% water/muscle difference is addressed in AAPM TG329. For bony tissues, either specific Dm,m constraints should be used, or the different Dm,m response should be switched off by recalculating in Dw,w or, more accurately, dose to reference-like medium (Dref,m*). Forcing homogeneous Dm,m distributions can introduce local fluence increases, deviating from previous practice. They increase NTCP if found in an OAR and impair robustness if their position mismatches the heterogeneity. Improving robustness implies using more homogeneous fluences while sacrificing PTV Dm,m homogeneity. Hence, Dm,m robust optimisation should be implemented. Homogeneous photon fluences resemble previous practice but can result in inhomogeneous Dm,m distributions difficult to evaluate. ICRU criteria should be adapted for target homogeneity, minimum, etc Alternatively, distributions can be recalculated in terms of Dw,w or Dref,m*. These issues should be considered in guidelines for Dm,m planning and evaluation, especially when consistency is critical, as in clinical trials and planning automation.

PD-0736 Evaluation of 3D dose distribution prediction for prostate VMAT based on deep learning

T. Fuangrod 1 , P. Kummanee 1

1 Chulabhorn Royal Academy, Princess Srisavangavadhana College of Medicine, Bangkok, Thailand