ESTRO 2023 - Abstract Book

S846

Tuesday 16 May 2023

ESTRO 2023

based response adaptive MR-guided radiotherapy will be shown. Here, a radioresistant subvolume inside the GTV, defined by a band of ADC values, will be dose escalated once weekly using a hybrid MR-Linac.

SP-1048 Delta-radiomics in MR guided RT D. Cusumano 1 1 Mater Olbia Hospital, UO Fisica Medica e Radioprotezione, Olbia, Italy

Abstract Text The introduction of Magnetic Resonance guided Radiotherapy (MR guided RT) in clinical practice is opening innovative perspectives for image-based modelling and decisional support systems: on board MR images represent indeed an incredible amount of data for which extensive research is needed, with the aim to explore their potentialities and fully take advantage of this unprecedented content of knowledge [1]. In the modern context of personalised Radiotherapy, being able to find image-based biomarkers extract from on-board MR images able to early predict the treatment response is of utmost importance to support the clinician in the identification of the optimal therapeutic dose. The identification of image-based biomarkers is today increasingly obtained using radiomic approaches, which consist in extracting quantitative features from defined regions of interest (i.e., tumor) and so using these parameters to generate models able to predict clinical outcomes. The use of MRI-based biomarkers precedes the advent of hybrid MR-Linac systems and is not limited to these: due to its better soft-tissue contrast over traditional CT, MRI has appeared as one of the most promising imaging modalities in Radiomics. Furthermore, the evaluation of the radiomics features trend and their variation throughout the treatment course (i.e., “delta radiomics”) has recently reported very promising results, in several cases superior to those obtained using the standard radiomic approach. Considering the analysis of images acquired during treatment, delta radiomics approach allows to model the patient’s treatment response to therapy, so including in the elaboration of predictive model the concept of patient sensitivity to the radiation treatment. In recent years delta radiomics in MR-guided RT has reported significant results in studies involving patients affected by pelvic (rectal and cervical cancer) and abdominal lesions (pancreas and liver). As for rectal cancer, different delta radiomics features calculated from MR images acquired during MRgRT have been identified as potential predictors of clinical and pathological complete response (pCR) after neoadjuvant chemoradiotherapy (nCRT)[2,3]. In this context one of the most significant proposals was the Early Regression Index (ERITCP), a radiobiological parameter able to predict pCR in LARC patients undergoing long-course nCRT, by combining the volumetric information of the Gross Tumor Volume delineated on the staging and mid-therapy MR images. Initially developed on 1.5 T staging MR images, ERITCP was then externally validated considering 0.35 T MR images and used as key-parameter to guide a dedicated MRgRT clinical trial [4–6]. Further studies also demonstrated that ERITCP could be considered as a good predictor in other clinical outcomes related to rectal cancer (long-term disease-free survival) and in other pathologies (cervix)[7,8]. Besides pelvis, interesting generating-hypothesis studies have been reported in the prediction of local control for patients affected by pancreatic and liver lesions [9–11]. In conclusion, although it is a discipline in its infancy, Delta Radiomics in MR-guided RT represents a continuously growing field, having already reported significant results on a clinical point of view. [1] Cusumano D et al Phys Med 2021;85:175–91. https://doi.org/10.1016/j.ejmp.2021.05.010. [2] Cusumano D et al Phys Med 2021;84:186–91. https://doi.org/10.1016/j.ejmp.2021.03.038. [3] Boldrini L et al Radiol Med 2019;124:145–53. https://doi.org/10.1007/s11547-018-0951-y. [4] Fiorino C et al Radiother Oncol 2018;128:564–8. https://doi.org/10.1016/j.radonc.2018.06.019. [5] Cusumano D et al Int J Radiat Oncol Biol Phys 2020;108:1347–56. https://doi.org/10.1016/j.ijrobp.2020.07.2323. [6] Chiloiro G et al BMC Cancer 2022;22:67. https://doi.org/10.1186/s12885-021-09158-9. [7] Cusumano D et al Appl Sci 2020;10:1–10. https://doi.org/10.3390/app10228001. [8] Cusumano D et al Radiother Oncol 2022;174:30–6 https://doi.org/10.1016/j.radonc.2022.07.001. [9] Cusumano D et al Diagnostics 2021;11:72. https://doi.org/10.3390/diagnostics11010072. [10] Jin WH et al Sci Rep 2022;12:18631. https://doi.org/10.1038/s41598-022-22826-5. [11] Simpson G et al Med Phys 2020;47:3682–90. https://doi.org/10.1002/mp.14200. Abstract Text Medical images are routinely used to evaluate treatment response in oncology and in clinical trials, typically by measuring tumor changes before and after treatment (RECIST). Since manual measurement of tumor is prone to errors and to inter- and intra-observer variability, ongoing research aims for automation of (volumetric) RECIST assessment. Amongst others, convolutional neural networks have been used to automatically evaluate RECIST scores. Whereas change in tumor or volume is a simple yet effective method to evaluate response to treatment, response patterns may be heterogeneous and complex. Therefore, methods to effectively extract dynamic information from medical images should be explored. Personalized medicine could be improved by capturing the longitudinal information in medical images for, amongst others, disease monitoring, treatment evaluation, or outcome prediction. Besides the use of longitudinal image analysis during or after treatment, accurate outcome prediction prior to treatment could allow for the choice of the right treatment or even a watch-and-wait approach. Compared to 1D signal analysis, there are additional challenges when evaluating longitudinal 2D and 3D imaging, including registration, reliable segmentation, inconsistent imaging intervals, and sparse data. Longitudinal image analysis could be performed by modelling hand-crafted features, e.g. delta-radiomics or longitudinal radiomics. This requires accurate and SP-1049 Analyses of longitudinal imaging data for outcome prediction Janita van Timmeren Radboud University Medical Center, Radiotherapy, Nijmegen, The Netherlands