ESTRO 38 Abstract book

S501 ESTRO 38

figure 2, suggesting definite modes of tumour shrinkage can be identified from our approach.

Conclusion For patients treated in the cranium region, MC simulations of the use of a thin, energy-layer specific multi-leaf collimator have shown to lower doses to normal tissues in comparison to grid and un-collimated contour scanning, especially the brainstem. Best dose results are obtained when combining collimation with pencil beam placements following the target contour. Further improvements are expected either by improving the model in the optimization and/or by using different contour expansions within the same field, or even optimizing the expansion and beam placement itself. PO-0932 Identification of modes of tumour changes in NSCLC during radiotherapy L.M. Amugongo 1 , E. Vasquez Osorio 1 , A. Green 1 , D. Cobben 1 , V.H. Marcel 1 , A.M. Alan 1 1 University of Manchester, Division of Cancer Sciences- Faculty of Biology- Medicine and Health- University of Manchester, Manchester, United Kingdom Purpose or Objective Cone-bean computed tomography (CBCT) is routinely used to set up and verify patient position during radiotherapy (RT) of non-small cell lung cancer (NSCLC) patients. CBCT images can be used to assess tumour changes during RT. This information is currently used qualitatively to inform treatment adaption. However, uncertainty exists regarding the mode of tumour change (elastic or non- elastic) and therefore the safety of adapting treatment. This study introduces a novel approach based on shells around the tumour boundary to automatically distinguish different modes of tumour change. Material and Methods CBCT images of 80 NSCLC patients were collected and registered to the planning CT scan using a two-step alignment process. (1) Rigid registration of the bones to remove gross translation and rotations; (2) Localised soft tissue registration, translation only, on a slightly expanded gross tumour volume (GTV). Intensity normalisation was then applied to the CBCTs. Eight shells, 1mm thick and extending 4 mm inwards and outwards from the planning CT GTV surface, were created. The external shells were constrained to the lung/tumour boundary. Typically, shells 1-3 describe gross target volume changes, 4-5 tumour/lung boundary changes and 6-8 show changes to the local lung tissue. From each shell, the mean intensity was extracted across all normalized CBCTs. Linear fits of the mean intensities for each shell were plotted. Affinity propagation, an unsupervised clustering method, was used to explore whether modes of tumour change could be identified based on the gradients of these fits. Results All CBCTs were successfully registered to the planning CT scan and shells were visually verified to ensure they avoided mediastinium/chest wall. Distinct modes of tumour changes could be seen in the intensity gradients of the CBCT. The majority of patients showed small density changes during the course of treatment (figure 1a), but some showed large changes (figure 1b). Affinity propagation clustered the patients into 7 groups, shown in

Conclusion Our novel methodology automatically identified distinct modes of tumour changes for lung cancer patients from routine on-treatment CBCT imaging. Future work will include more patients, to optimise the clusters, identify and validate the mode of change for each group, e.g., related to histopathology. PO-0933 Single isocenter SRS for multiple brain metastases: dosimetric comparison of DCAT and VMAT