ESTRO 2021 Abstract Book

S1525

ESTRO 2021

Results Out of the total eighty-eight radiomic features studied, 19 had a high ICC (>0.9), and these were further analysed to find whether these features showed a difference between BW and TB when using all time-points. Figure 1 illustrate examples of a feature of different distribution in TB and BW, Large Area Emphasis (LAE) and another of similar distribution in TB and BW, Large Area High Gray Level Emphasis (LAHGLE) . 8/19 features showed a difference in distribution between BW and TB. There was a greater distribution of the four features – gray level non-uniformity, gray level non-uniformity normalised, zone variance, large area emphasis in BW compared to TB (Figure 2a), and greater distribution of four others – sphericity, gray level variance, high gray level zone emphasis, small area high level emphasis in TB compared to BW as depicted in Figure 2b.

Conclusion This is the first study to identify robust radiomics features defining tumour bed and bladder wall in bladder cancer, taking into account inter-observer variability. Interestingly, features were different for the bladder wall and tumour bed. The identified radiomics features will be taken forward for further analysis. There is potential to identify changes in radiomics feature expression during treatment which may predict treatment response. PO-1799 Radiomic and dosiomic profiling of paediatric Medulloblastoma tumours treated with IMRT Cinzia Talamonti 1,2 , Stefano Piffer 2 , Leonardo Ubaldi 3 , Daniela Greto 4 , Francesco Laurina 3 , Antonio Ciccarone 5 , Piernicola Oliva 6 , Maria Evelina Fantacci 7 , Marzia Mortilla 8 , Stefania Pallotta 2,9 , Alessandra Retico 3 1 University of Florence, Dept. of Biomedical Experimental Clinical Science "Mario Serio", Florence, Italy; 2 National Institute of Nuclear Physics (INFN), Florence Unit, Florence, Italy; 3 National Institute of Nuclear Physics (INFN), Pisa Unit, Pisa, Italy; 4 Azienda Ospedaliero Universitaria Careggi, Radiotherapy Unit, Florence, Italy; 5 Meyer University Hospital, Health Physics Unit, Florence, Italy; 6 National Institute of Nuclear Physics (INFN), Cagliari Unit, Cagliari, Italy; 7 National Institute of Nuclear Physics (INFN), Pisa, Pisa, Italy; 8 Meyer University Hospital, Diagnostic Radiology Unit, Florence, Italy; 9 University of Florence, Dept. of Biomedical Experimental Clinical Science "Mario Serio" , Florence, Italy Purpose or Objective The purpose of this study is to apply a retrospective exploratory MR-based radiomics and dosiomic analysis based on machine-learning technologies and statistical analysis, to investigate imaging-based biomarkers of clinical outcomes in paediatric patients affected by medulloblastoma. This work was developed in the framework of the INFN-funded Artificial Intelligence in Medicine project. Materials and Methods A database of 50 paediatric patients who underwent surgery, chemotherapy and radiotherapy (RT) was retrospectively selected. It includes information on histology, prescribed drugs and planned dose distributions. Moreover, all MR and CT images acquired from pre-treatment to the end of follow-up are available. As a first step in a wider and deeper research, multiparametric data were considered, including: last MRI just before RT (T1w, T2w and FLAIR image sets) and dose distribution of the radiotherapy plan. Second order features from those images were extracted with PyRadiomics and analysed with two different programs, a homemade script and RadAR (Radiomics Analysis with R). Principal Component Analysis technique was exploited to decrease the number of variables involved while maintaining 90% of the variability of the data. Ten features were associated with radio-induced toxicity occurrence. Random Forest classifier was trained on different combinations of the available data. RadAR performs analysis of radiomic features, implementing multiple statistical methods. To compare multiple