ESTRO 2021 Abstract Book

S508

ESTRO 2021

Conclusion Peritumoural density predicts LR and DM following lung SABR. The level of risk depends on dose variability away from the GTV. Dose around 1cm predicts LR in agreement with literature. Dose around 3cm predicts DM and was validated in an independent cohort. Different spatial interactions exist for different failure types. Location of failure must be reported identically across centres to validate results. In future, SABR plans could be personalised based on imaging biomarkers to improve outcome. OC-0641 Radiation pneumonitis in thoracic cancer patients: multi-center voxel-based analysis G. Palma 1 , S. Monti 1 , R. Pacelli 2 , Z. Liao 3 , J.O. Deasy 4 , R. Mohan 5 , L. Cella 1 1 Italian National Research Council, Institute of Biostructures and Bioimaging, Napoli, Italy; 2 University of Naples “Federico II”, Department of Advanced Biomedical Sciences, Napoli, Italy; 3 The University of Texas MD Anderson Cancer Center, Department of Radiation Oncology, Houston, USA; 4 Memorial Sloan Kettering Cancer Center, Department of Medical Physics, New York, USA; 5 The University of Texas MD Anderson Cancer Center, Department of Radiation Physics, Houston, USA Purpose or Objective To investigate dose response patterns associated with radiation pneumonitis (RP) in patients treated for thoracic malignancies with different radiation modalities. To this end voxel-based analysis (VBA) empowered by a novel strategy for the characterization of spatial properties of dose maps was applied. Materials and Methods Data from 382 lung cancer and mediastinal lymphoma patients from three institutions treated with different radiation therapy (RT) techniques were analyzed. The dose per fraction varied between 1.6 Gy and 18 Gy, the total physical dose between 30 Gy and 74 Gy. The incidence of CTCAE any grade RP on the entire dataset was 23.8%. Each planning CT and biologically effective dose (BED) map ( α / β =3 Gy) was spatially normalized on a common anatomical reference with segmented cardiac and pulmonary subregions. The VBA of local dose differences between patients with and without RP was performed and the clusters of voxels with dose differences that significantly correlated with RP at a p -level of 0.05 (S 0.05 ) were generated accordingly. The robustness of VBA inference was evaluated by a novel characterization for spatial properties of dose maps based on probabilistic independent component analysis (PICA) and connectograms [Irimia et al. Neuroimage Clin. 2012].