ESTRO 38 Abstract book

S65 ESTRO 38

useful for a more comprehensive characterization of image phenotypes of the tumor. Radiomics analysis has recently emerged as a promising tool for the definition of new imaging biomarkers by high-throughput extraction of quantitative image features. Several studies have shown that texture features extracted from PET, CT or MRI images may have a higher predictive and/or prognostic power than simpler, more “standard” metrics (SUV, tumor shape, tumor volume, etc.). Notwithstanding, such great promises, radiomics analysis presents several challenges that need to be carefully addressed. Standardization (features definition, features computation, features dependences etc ..) is one of the most important aspects which need to be defined to ensure robustness, reproducibility and dissemination of the obtained results among centers. In conclusion, all previously described topics find application when using HIS. HIS have an enormous potential in the generation of complementary information (images and other data type) that, once integrated in a unique model, may allow a more comprehensive understanding of the tumor pathology as well as a more accurate definition of the target for RT treatment. In this scenario the perspective of an NM physicist is to be active in this complex technological and methodological puzzle, working in strong collaboration with all the other people involved in the clinical and research work (physicists, engineers, physicians, technicians, etc.) having as a common objective the best treatment for each patient. SP-0133 Working in radiotherapy from the perspective of an MRI physicist L.E. Olsson 1 1 Lund University, Department of Medical Radiation Physics, Malmö, Sweden Abstract text There is a substantial increase of imaging in the recent development in radiotherapy. The on-going process of installing dedicated radiotherapy MR-scanners in the oncology/radiotherapy clinics is an obvious proof of this development. MRI is a much more complex modality and applications extend from diagnosis to specific imaging for target identification, treatment planning, treatment follow-up and adaptive regimes. An in-house MR-scanner also brings along many safety issues, both for personnel and patients. Altogether, there is an increasing need for MRI expertise in the radiotherapy department. The demand on accurate geometry and quantification of the data is considerably larger in therapy than diagnostic MRI applications. That means that what can easily be accepted as a well know harmless artefacts in the diagnostic setting can have detrimental negative effects if used for treatment guidance in radiotherapy. Therefore, the MRI-physicist needs specific training in how the image data will be used in the radiotherapy workflow. The MRI-expertise can be an in-house MRI-physicist working with radiotherapy or as a service provided by diagnostic radiology physicists. In any case, it is important that the MRI-physicist will be specifically trained and dedicated to radiotherapy. Not all MRI-physicists from radiology can have the knowledge needed in radiotherapy. Similarly, an MRI-physicist working in radiotherapy cannot be an expert on all MRI techniques. An ideal model is to have a dedicated MRI-physicist in-house in radiotherapy, which will act as a link to the other MRI-physicists in radiology, and thereby facilitate the discussion and knowledge transfer in both directions. The present question will be even more important to discuss when hybrid modalities, such as MR-linacs, will be implemented in the clinics. It is obvious that the MR- scanner is a part of the treatment machine. Therefore, the installation of MR-linacs should be combined with recruitment of in-house MRI-expertise. Traditionally, there have been different physicists

Joint Symposium: ESTRO-EFOMP: Multi-disciplinary working in Radiotherapy

SP-0132 Working for radiotherapy applications: The perspective of a nuclear medicine physicist in the era of Hybrid Imaging Systems V. Bettinardi 1 , M.G. Cattaneo 2 1 IRCCS San Raffaele Hospital, Nuclear Medicine, Milano, Italy; 2 IRCCS San Raffaele Hospital, Medical Physics, Milano, Italy Abstract text Radiation therapy (RT) is a cornerstone of modern cancer therapy used in approximately 50% of all cancer patients during their clinical course of illness. A common feature of RT treatment with curative intent is to have very sharp dose gradients between the target and adjacent normal tissues. The introduction of more tailored RT planning/delivery techniques using intensity-modulated beams (IMRT), stereotactic radiotherapy (SBRT), tracking with gating or robotic radiosurgery, and precise image- guidance (IGRT) have largely contributed to improve the outcome both in terms of tumor control and in reducing toxicities. In this regard, multimodal imaging and in particular that of the latest generation of hybrid imaging systems (HIS) such as: PET/CT, SPECT/CT and more recently also PET/MRI, can play a crucial role to further exploit the potentialities of modern RT technology. Besides the great clinical value of HIS, technical aspects make these systems also particularly useful for RT applications. In fact, HIS allow the acquisition of anatomical and functional images with the patient being in the same position during the whole study session. This characteristic is further improved in fully integrated PET/MRI systems due to the possibility of truly simultaneous acquisition (spatial and temporal) of the two studies. This “hardware” alignment is very important for an accurate target volume delineation (TVD) as errors introduced by image co-registration are minimized. Furthermore, the integration of anatomical, functional and molecular information allows the visualization of various pathophysiological aspects of the tumors which may be important for assessing individual parameters related to cancer cure procedure. In this highly technological scenario the perspective of a nuclear medicine (NM) physicist, who has access to HIS is to work on topics aiming to produce useful and accurate information (images and other type of data) for an optimal and personalized RT. A first of these topics is the set-up of a comprehensive quality control protocol, to guarantee stability/reproducibility of the systems' performance with particular reference to the need of RT applications. Definition and implementation of acquisition, reconstruction and post-processing protocols accounting for different tracers (e.g. 18F-FDG, 18F-FAZA, 11C- Methionine, etc.) and different targets (e.g. brain, lung, liver, etc.) are also important topics of a unique pipeline which aims at the most accurate TVD. Delineation of the target volume, and in particular of the biological target volume, is still an open topic that needs to be improved. Several methods based on different image processing techniques (e.g. thresholding, probabilistic, clustering, etc.) have been proposed over the years but none of them have been accepted as standard for clinical use. This has left, to the human operator, the role of gold standard in the TVD. However, new techniques based on a machine learning/deep learning, have shown to overtake human performance in computer vision, image recognition, and image interpretation challenges. Furthermore, the same techniques have shown very promising results also when applied to automatic segmentation for TVD, in particular when a multimodal and/or a multiparametric imaging approach has been used. An accurate TVD also allows a more accurate extraction of information that can be