ESTRO 2020 Abstract Book

S277 ESTRO 2020

M.L.K. Chua 1 1 National Cancer Center Singapore, Radiation Oncology, Singapore, Singapore Abstract text The treatment landscape of nasopharyngeal carcinoma (NPC) has evolved substantially in recent years, and thus deserves a timely update. In the 90s, improvement in survival rates of locally advanced (stage 3 and 4) NPC were driven by both advances in radiotherapy and the combination of radiotherapy with systemic chemotherapy. This has led to a change in the natural history of the disease, with distant metastasis dominating as the primary pattern of failure. This clinical phenomenon has provided the scientific rationale for the design of more contemporary studies testing the clinical efficacy of different timings, sequencing, and intensities of chemotherapy in combination with radiotherapy. We now have new data from several randomised clinical trials supporting the recommendation of induction chemotherapy with chemo-radiotherapy as an option for first-line standard of care in these high-risk patients. Nonetheless, these protocols of systemic intensification also result in incremental acute toxicities, and thus future directions will involve improvements in disease risk stratification to optimise selection of patients for such intensive treatment protocols. In this space, a well- validated liquid biopsy biomarker - EBV DNA may have a promising role, and prospective trials are now under way to employ this molecular test as a method of treatment personalisation in NPC. The potential of immunotherapy in the treatment of locally advanced high-risk disease will also be discussed. SP-0492 Validation and commissioning of AI contouring tools A. Green 1 1 The University of Manchester c/o The Christie NHS Foundation Trust, Department 58- Radiotherapy Related Research, Manchester, United Kingdom Abstract text Artificial Intelligence (AI) has become ubiquitous in modern life, and has inevitably found uses in the clinic. The recent approval of several commercial tools for OAR segmentation has accelerated the uptake of AI contouring tools, and their validation, commissioning and ongoing QA is now of great interest and importance. In this lecture, we will explore the fundamental process by which an AI contouring tool produces contours, and investigate the implications on validation and commissioning. We will compare the commissioning of an AI tool to that of a standard registration-based tool with which the community is already familiar and highlight particular pitfalls specific to AI tools that may not have been encountered when commissioning other tools. As a concrete example, we will look at work done by colleagues at The Christie Hospital, UK during their commissioning of an AI contouring workflow. Teaching Lecture: Validation and commissioning of AI contouring tools

Teaching Lecture: 4D imaging for radiation therapy using MRI and PET

SP-0493 4D imaging for radiation therapy using MRI and PET D. Thorwarth 1 1 University Hospital Tübingen, Section for Biomedical Physikcs, Tübingen, Germany Abstract text In this teaching lecture, an overview will be given on current 4D imaging techniques in magnetic resonance imaging (MRI) and positron emission tomography (PET). In contrast to other fields of medical imaging, radiotherapy planning, delivery and adaptation require dedicated 4D imaging techniques to inform about the motion excursion of tumors in order to adequately take this information into account for treatment planning. In a first part of the lecture, current 4D PET imaging techniques such as phase binning will be introduced. Additionally, consequences for radiotherapy target volume delineation will be discussed. In a second part, state-of- the-art as well as emerging 4D MRI techniques will be reviewed including implications for target and organ at risk delineation as well as challenges and opportunities for offline and online radiotherapy plan adaptation. SP-0494 Quantitative Digital Pathology Biomarkers of Neoadjuvant Therapy Response in Breast Cancer W. Tran 1 , F. Lu 2 , S. Tabbarah 3 , A. Lagree 3 , D. Dodington 4 , K. Jerzak 5 , S. Gandhi 5 , E. Rakovitch 3 , A. Shenfield 6 1 university Of Toronto, Department Of Radiation Oncology, Toronto, Canada ; 2 sunnybrook Health Sciences Centre, Anatomic Pathology, Toronto, Canada ; 3 sunnybrook Health Sciences Centre, Department Of Radiation Oncology, Toronto, Canada ; 4 sunnybrook Health Sceinces Centre, Anatomic Pathology, Toronto, Canada ; 5 sunnybrook Health Sciences Centre, Division Of Medical Oncology, Toronto, Canada ; 6 sheffield Hallam University, Department Of Engineering And Mathematics, Sheffield, United Kingdom Abstract text Neoadjuvant (i.e., pre-operative) chemotherapy (NAC) is indicated for high-risk cancer patients, in part, to reduce the tumor’s size with the desired outcome to kill tumor cells completely (known as a pathological complete response (pCR)). Patients who achieve pCR have a significantly lower risk of breast cancer recurrence and longer survival rate. However, a significant proportion of women who are treated with NAC do not achieve a pCR and we currently are unable to ascertain how effectively women’s tumors will respond to NAC on an individual level. We present research on the development of a digital pathology platform, utilizing artificial intelligence (AI), that enables the identification of biomarkers from digitized tumor core biopsy pathology specimens. AI- driven digital pathology biomarkers were studied to develop prediction models that are associated with the likelihood of NAC response. Predicting which patients who are unlikely to respond to NAC will assist in developing adaptive treatments or administering additional Teaching Lecture: Quantitative Digital Pathology Biomarkers of Neoadjuvant Therapy Response in Breast Cancer