ESTRO 2020 Abstract book

S8 ESTRO 2020

implementation. Collaboration with industry and scientific computing research groups is therefor of utmost importance to have access to the latest technology and scientific insights. A simplistic approach could be to team up with data science/data engineers and incorporate the tools from them and apply/interpret them, as this needs to come from the field itself. The success depends on both the quality (and amount) of the data and the ability to rephrase the clinical questions towards the data science engineers. This is the other aspect that artificial intelligence needs to deal with: access to clinical (patient) data. The hospital IT environment is the landscape is changing rapidly and medical physicists should play a role in this. Besides advice on the requirements for the IT frameworks, the medical physicist together with the physician is responsible for interpreting and develop the appropriate models: e.g. prediction models or more advanced solutions (cf. example of Babylon). Technology for mining data is only a single aspect of this work, but also challenges are present in aligning the different stakeholders (IT, physician, legal/patient privacy etc)., to strive for collaboration and keeping the common (clinical) goal. Medical physicist have a tradition to secure the implementation of new technology in clinical practice. The role of MP in this scenario and where MPs need to be involved in the right parts of the projects is currently under discussion as described above. Besides implementation, also development is needed where the industry needed to roll-out these new technologies wants to work with the innovators and early adopters, but these are not all of Medical Physics but only a subselection currently typically located in the larger academic hospitals. AI could be seen as just another medical device that needs training, commissioning and QA. Where the commissioning is not based on the classical dosimetry data but on patient data. Clinical physicists need training for these new developments and commissioning and QA programs need to be developed. The amount of training needed will depend on the specific implementations but as AI will be classified as a medical device the medical physicists should also approach these techniques as such with proper testing and validation (QA) programs in place. SP-0034 Medical physicists will substantially contribute to modeling biological effects in the era of personalized Radiation Oncology C. Fiorino 1 , R. Jeraj 2,3 1 San Raffaele Hospital Scientific Institute, Department Of Medical Physics, Milano, Italy ; 2 university Of Wisconsin, Department Of Medical Physics, Madison, Usa ; 3 university Of Ljubljana, Faculty Of Mathematics And Physics, Ljubljana, Slovenia Abstract text During the think-tank meeting in Budapest (October 24 th , 2019) on “the most provocative questions to Medical Physics (MP) in Radiation Oncology (RO)”, one of the four selected issues focused on the contribution of MP in modeling biological effects in the era of personalized RO. Several questions were suggested to the speakers/debaters to better orient the discussion: (1) Aren't even simple models validated on smaller but high quality datasets at least as interesting as the results obtained with the "new toys"? (2) How can MP contribute to model effects of concurrent treatment? (3) Can MP contribute in understanding the migration mechanisms of tumor cells? As for the other debated issues, four experts including out-of-field experts and one radiation oncologist were asked to debate the problems from their point of view. Promises, needs and priorities, suggesting visions and actions aimed to maximize the expected contribution of MP to the field were identified. Here we are summarizing the key points of discussion: 1) Medical physicists have fundamental physics skills to set up

SP-0032 Transforming CTV definition from an art to science T. Bortfeld 1 , C. Garibaldi 2 1 Mass. General Hospital, Radiation Oncology- Division Of Radiation Biophysics, Boston- Ma, Usa ; 2 istituto Europeo Di Oncologia, Medical Physics, Milan, Italy Abstract text Defining the clinical target volume (CTV) is a complex and challenging task that is traditionally owned by the radiation oncologists. Variations of CTVs drawn by different radiation oncologists for the same patient are substantial and vastly exceed the physical/technical uncertainties of targeting the radiation beam on a voxel in the patient. Input from physicists is needed to provide a more scientific basis for CTV definition. Research and development is required along at least three different directions. The first direction is the modeling of disease spread beyond the visible gross tumor volume, and infiltration of lymph nodes. Modeling the impact of multi- modality treatments including systemic therapies and immunotherapy, and their ability to take care of microscopic disease is also important. The second direction is to provide a probabilistic framework for CTV definition. The current binary CTV approach (100% tumor or 0% tumor) does not capture the inherent uncertainties of defining the CTV. The third and most immediate direction is to standardize current CTV definition practices based on existing clinical guidelines, by implementing them as computer algorithms in treatment planning systems. SP-0033 Medical physicists will drive the development and implementation of artificial intelligence in Radiation Oncology W. Van Elmpt 1 , N. Jornet 2 1 maastricht University, Department Of Radiation Oncology, Maastricht, The Netherlands ; 2 hospital De La Santa Creu I Sant Pau, Servei De Radiofísica I Radioprotecció, Barcelona, Spain - Artificial Intelligence follows distinctive (engineering) principles: development, implementation and QA that medical physicists have been doing for the past decades. Artificial Intelligence based solutions need to follow the same technology adoption in medical physics typically as other engineering principles in medicine. First a clinical unmet need or a problem is defined, next novel solutions are proposed based on the proposed requirements, followed by a phase validation or testing if the new solutions to validate they meet the requirements for use in clinical practice. Finally education of understanding and explaining the new technology to end users (e.g. clinicians, physicists, software engineers, RTTs etc) is important for successful implementation. Medical physicists need to work together with their surrounding disciplines to safely and efficiently introduce AI implementation into the current care path of patients. This starts with an understanding of the technology presently at hand. Although many medical physicists have by training strong analytical skills for problem solving, detailed knowledge about current machine learning, deep learning or other AI techniques is not ubiquitous. As for other technological advances, medical physicists do not necessarily have to be developers themselves of these solutions but could fulfill a coordinating role in this. However, playing such a central active role in this field would be attractive for some physicist to define the directions taken and secure the proper clinical Abstract text Key points: - Medical physicist needs to identify the clinical problems and is the key person to link the various technical disciplines (e.g. data scientists, IT) with clinical professionals (e.g. physicians, RTTs).