ESTRO 2021 Abstract Book

S1533

ESTRO 2021

Fig.2: Calibration curve of cumulative predicted probability on training and test set. Error bars indicate the standard error. The final model showed good calibration in training and test set (Fig.2). Clinical predictors such as age or gender were not selected, suggesting that their impact on dyspnoea is minimal, thus e.g. high age does not seem to be a contraindication for RT in terms of risk of dyspnoea Conclusion Mean lung dose was predictive for radiation-induced pneumonitis in locally advanced NSCLC patients. Further studies are needed to develop the optimal assignment of local radiomics clusters to the patient. PO-1805 In silico tumour modelling accounting for tumour oxygenation and its dynamics F. Schiavo 1,2 , E. Kjellsson Lindblom 1,2 , I. Toma-Dasu 1,2 1 Stockholm University, Department of Physics, Stockholm, Sweden; 2 Karolinska Institutet, Department of Oncology and Pathology, Stockholm, Sweden Purpose or Objective The tumour oxygenation is one of the key parameters affecting the sensitivity to radiation. However, tumour hypoxia is seldomly accounted for in radiotherapy (RT) treatment planning, which is likely to compromise the outcome for patients with hypoxic tumours. This work presents the development of a novel three-dimensional (3D) in silico model of heterogeneous tumour oxygenation and its dynamics as part of a virtual testbed for radiotherapy treatment outcome simulation. Materials and Methods Complex networks of blood vessels simulating tumour vasculature are created in a lattice by means of fractal theory. The resulting distributions of the oxygen partial pressure (pO 2 ) are generated by solving the oxygen diffusion equation in three dimensions using a customized finite difference method. The choice of parameters determining the properties of the vascular network enables the simulation of chronic hypoxia by defining regions with different vascular density. In addition, acute hypoxia is simulated by randomly closing vessels, and anemic hypoxia by reducing the oxygen tension inside the vessels. The workflow for the model is presented in Figure 1. The resulting macrolevels of oxygen supply can be characterized by several model output quantities, such as the fractal dimension (FD), the microvascular density (MVD), and the hypoxic fraction (HF).