Abstract Book

S233

ESTRO 37

oncology (e.g., tumor characterization for diagnostic purposes). However, there are substantial challenges to comparing and reproducing results across sites and studies for several reasons. We will begin with a discussion of some the causes of variability including sensitivity to segmentation, preprocessing methods, and methodology and implementation of the radiomics feature calculations. We will then describe collaborative, multi-institutional efforts intended to increase the reproducibility of radiomics research. This will include our work in standardizing the terminology used to describe imaging features, allowing direct comparisons between features collected by different investigators or in different datasets or even domains. We will then describe phantoms, both digital and hardware, that can be used to standardize the calculation of radiomics features. Finally, we will describe publicly available datasets with segmentations and reference values for the most commonly used features. These publicly available resources should help drive consensus in the community and increase reproducibility in radiomics research. SP-0447 Radiomics – Clinical challenges and opportunities – Will radiomics change our practice in the next years? M. Sollini 1 1 Sollini Martina, Dep. Biomedical Sciences- Humanitas University, Milan, Italy Abstract text Hippocrates (460 - 370 BC), the father of the modern medicine, stated that “It’s more important to know what sort of person has a disease than to know what sort of disease a person has” and personalized (or precision) medicine is still a challenge in patients management in the third millennium. In the past years, several attempts have been made using different approaches to reach the goal of personalized medicine in all fields of medicine including oncology. In the setting of personalized medicine, radiomics may be consider as "unique". In fact radiomics extracting quantitative features from medical images using mathematical algorithms, is able to capture the heterogeneity of the texture of a lesion. In recent years there has been emerging evidence that image heterogeneity within primary tumor can permit in vivo lesion characterization and provide predictive information. Radiomics has been applied to different image modalities (ultrasound, CT, PET/CT and MRI), tumor types and different clinical setting from diagnosis to prognosis passing through tumor response assesment. Another challenge of the medicine of the third millennium ìs to conciliate patients and computerization. Many instruments have been develop to assist medical doctor in clinical practice including artificial intelligence approaches and question answering computer systems. Currently it is not possible to predict if medical doctor will be substitute by robot or intelligent computer systems in the next future. However, it is reasonable to speculate that the role of computer technology will increase in medicine and that radiomics will be part of the medical challenge in the era of the evidence-based medicine. Therefore it is crucial to evaluate radiomics opportunities and limitations in order to use this tool in the best way to be really useful to persionalized medicine. SP-0448Radiomics – How does artificial intelligence shape the future of medical imaging? F. Maes 1 1 KU Leuven, ESAT / PSI, Heverlee, Belgium Abstract text Medical imaging, by its ability to acquire detailed information about the internal organs of the human body and possible pathologies in a non-invasive way, plays a

crucial role in all stages of the medical decision process, not only for early patient diagnosis and individualized therapy planning, but also for therapy outcome assessment and prediction, for population screening, and also in translational pre-clinical and clinical research. In order to optimally exploit all available imaging data and to support the effective use of ‘big data’ in the context of more personalised and predictive medicine, reliable computer-aided image analysis is indispensable to extract and quantify the relevant information from the imaging data, to fuse complementary information and to support the interpretation thereof. The analysis of medical imaging data is complicated by ambiguity that is induced by the intrinsic limitations of the imaging itself, by the variable appearance of the objects of interest in the images, and by the heterogeneous nature of the data to be analyzed (multi-dimensional, multi-modal, multi- temporal, multi-parametric, multi-subject, multi- center…). Medical image computing, which is a branch of scientific computing at the intersection of medical imaging, computer vision and machine learning, aims at developing computational strategies for medical image analysis that can cope with the complexity of medical imaging data to enable (semi-)automated analysis with sufficient accuracy and robustness. Such strategies rely on mathematical models that incorporate prior knowledge about the typical appearance of the objects of interest in the images, including photometric properties, geometric properties and context. A natural and powerful strategy is to construct suitable models from the data itself by analysis of previously analysed images. Recent advances in supervised machine learning of models from training data, especially deep learning based on convolutional neural networks, have shown great promise for many problems in computer vision, including image classification, object recognition and segmentation, provided that large amounts of data are available for the network to converge to a stable solution with good generalization ability. This requirement is usually not met in medical image analysis where the availability of training data is limited, which poses additional challenges. Moreover, while deep learning seems to eliminate the assumptions that are implicitly made when relying on handcrafted image features, plenty of heuristics are embedded in the actual implementation of the chosen neural network architecture, in the optimization strategy and in the sampling of the training data presented to the network, which complicates the interpretation of the model and the optimization of its performance. SP-0449 Responses to DNA damage in cancer pathogenesis and treatment J. Bartek 1 1 Karolinska Institutet, Department of Molecular Biochemistry and Biophysics- Science for Life Laboratory, Stockholm, Sweden Abstract text The lecture will summarize our recent data in the following areas: 1. Molecular mechanisms of signalling and repair of DNA double strand breaks, with emphasis on the chromatin response and DNA repair pathway choice; 2. Mechanisms of replication stress, cellular responses to such stress and involvement of oncogene-triggered replication stress in pathogenesis of human cancer; 3. Involvement of stress response pathways and genome Symposium: Genomic Instability and DNA repair in cancer