ESTRO 2022 - Abstract Book

S1537

Abstract book

ESTRO 2022

This work presents an automated approach for treatment planning using a dose painting strategy (DP) based on the combined information on the tumour clonogenic cell number (CCN) and on the tumour oxygen distribution derived from PET images. Materials and Methods The treatment planning workflow was created in RayStation (v10, RaySearchLaboratories) via scripting. The automated treatment plan pipeline consists of 3 stages (fig. 1): Dose prescription: Normalized uptake of a PET tracer dedicated to hypoxia is converted by means of a non-linear function into oxygen partial pressure maps (pO 2 ). Dose modifying factors to counteract radioresistance in the hypoxic areas are calculated based on the obtained pO 2 values. The CCN is derived from FDG-PET images as follows: the normalized uptake of FDG is converted into CCN by means of a linear conversion function, whose curve origin and slope were derived from the patient dataset. The DP strategy aims at 95% tumor control probability (TCP) in the CTV. Three levels of uniform dose are assigned to the hypoxic target volume (HTV), to the GTV-HTV and to the CTV-GTV. Plan optimization: The algebra combination of the volumes of interest (VOIs) is automatically performed prior to optimization. The whole optimization process is composed of three different steps: 1) two optimizations where targets are prioritized; 2) an optimization of previous solution where the dose to organs at risk (OARs) is reduced to meet the clinical dose constraints; 3) two minimax robust optimizations that mimicked both the OAR doses retrieved from (2) and the prescribed doses to the targets. Plan Evaluation: A dosimetric evaluation of the nominal plan is performed accounting for target coverage and OAR constraints. The target TCP is calculated by considering the underlying radiosensitivity and CCN derived from the PET images in the initial stage.

Results The presented treatment planning pipeline was tested on a head and neck cancer case imaged with FMISO as hypoxia tracer. The automated treatment planning pipeline was proven able to render a dose distribution that successfully met the clinical goals (Fig.2) with the following plan specifications: the total dose delivered in 35 fractions with an integrated boost. The minimax robust optimization considering ±3mm setup errors (7 scenarios) was performed. The whole optimization process required about 45 minutes to be performed.

Conclusion The presented automatic treatment planning workflow has shown to be feasible and it can be readily applied to different treatment sites and modalities supported by the treatment planning system.