ESTRO 2020 Abstract book

S898 ESTRO 2020

contrast. Clinical factors did not strengthen the prediction performance. Conclusion Multiparametric MRI radiomics holds the promise of improving medical imaging by providing insight to overall survival and prognosis in the era of personalised medicine. However, preliminary results should be interpreted carefully, as features show poor stability. In future work further normalization and quantization methods will be investigated in a larger cohort in order to test the stability and predictive power of radiomic features with regards to survival outcome. References 1 Collewet G. et al. Influence of MRI acquisition protocols and image intensity normalization methods on texture classification. Magnetic resonance imaging 2004. 2 Hatt M et al. IBSI: an international community radiomics standardization initiative. Journal of Nuclear Medicine 59.supplement 2018. 3 Whybra P et al. Assessing radiomic feature robustness to interpolation in 18 F-FDG PET imaging. Scientific reports 2019. 4 Kursa MB et al. Feature selection with the Boruta package. J Stat Softw 2010. PO-1566 NTCP modeling for radiation induced optic neuropathy in a high-risk proton therapy patient cohort A. KÖTHE 1,2 , P. Van Luijk 3,4 , S. Safai 1 , M. Kountouri 5 , D.C. Weber 1,6,7 , A.J. Lomax 1,2 , G. Fattori 1 1 Paul Scherrer Institute, Center for Proton Therapy, Villigen, Switzerland ; 2 ETH Zürich, Department of Physics, Zürich, Switzerland ; 3 University Medical Center Groningen, Department of Biomedical Sciences of Cells and Systems- Section Molecular Cell Biology, Groningen, The Netherlands ; 4 University Medical Center Groningen, Department of Radiation Oncology, Groningen, The Netherlands ; 5 University Hospital of Geneva, Department of Radiation Oncology, Geneva, Switzerland ; 6 University Hospital of Zürich, Department of Radiation Oncology, Zürich, Switzerland ; 7 Inselspital Universitätsspital Bern, Department of Radiation Oncology, Bern, Switzerland Purpose or Objective Radiation-induced optic neuropathy (RION) is a serious complication that severely decreases patients’ post- treatment quality of life and limits their activities of daily life. In this retrospective study, we expand on the commonly reported evidence of correlation with age and maximum dose by putting forward a Normal Tissue Complication Probability (NTCP) model that integrates phenomenological clinical variables and dosimetric parameters. delivered to the optical apparatus were selected from head and neck cancer adult patients treated at PSI between 1999 and 2014 with PBS proton therapy. Based on availability of follow-up (median 5.6 years), a cohort of 196 cases was assessed, 13 (6.5%) of which developed RION. Using DVHs, clinical and patient parameters, stepwise multivariable logistic regression analyses were performed to identify risk factors for the development of RION, from which a corresponding model could be derived. A wide range of possible combinations of clinical and dosimetric parameters has been considered to identify the best model. To obtain insight into the robustness of the selection, prediction performance and potential overfitting problems, Bootstrapping and Leave- One-Out cross-validation analyses were performed. Results I n univariable analysis, RION was significantly associated with age (p < 0.001), arterial hypertension (p = 0.002), tumor involvement in the optic apparatus (p = 0.03), and, surprisingly, D 99 (p = 0.004), a surrogate for minimum dose. Material and Methods 216 patients with D max ≥ 45 Gy RBE

PO-1565 Multiparametric MRI radiomics model to predict overall survival in Glioblastoma Multiforme E. Kolozsi 1 , J. Powell 2 , C. Piazzese 1 , S. Thomas 2 , J. Staffurth 2 , E. Spezi 1 1 Cardiff University- School of Engineering, Biomedical Engineering, Cardiff, United Kingdom ; 2 Velindre Cancer Centre, Department of Clinical Oncology, Cardiff, United Kingdom Purpose or Objective High-grade glioma (HGG) is the commonest primary brain tumour in adults, for which current treatment options are limited with a poor overall prognosis. Risks associated with brain biopsy depend on tumour location and radiomics may allow non-invasive tumour assessment to avoid complications. Radiomics refers to the extraction and analysis of quantitative imaging features from medical images and offers an innovative approach to address diagnostic and prognostic challenges in HGG. The present study aims to build a radiomics model from magnetic resonance imaging (MRI) to improve overall survival prediction in Glioblastoma Multiforme (GBM). Material and Methods A cohort of 32 patients with pre-treatment T1-weighted, T1 post-contrast, T2-weighted and FLAIR MRI were recruited, clinical variables included gender and age. Prior to feature extraction image intensities were normalized between μ±3σ 1 and then quantized to 32, between 1 and 2 32 . A total of 182 standardised features 2 were extracted from the gross tumour volume (GTV) using an in-house software, Spaarc Pipeline for Automated Analysis and Radiomic Computing (SPAARC) 3 . Intra-class correlation coefficient (ICC) was used to detect stable features among MRI sequences. In order to build a multivariate model which predicts overall survival in GBM patients, feature selection was required to strengthen reliability. The Boruta machine learning-based algorithm 4 was used to capture statistically significant features with respect to survival outcome variable.

Fig 1. MRI of GBM A. T1w B. T1+c C. T2w D. FLAIR Results Six shape-based features showed excellent reliability ICC ≥ 0.90, however, the Boruta algorithm confirmed these features as statistically not significant. The algorithm found 5 statistically relevant feature with poor ICC < 0.50 performance: one morphology feature in FLAIR, two GLRL matrix-based features in FLAIR, one GLCM matrix-based feature in FLAIR and one statistical feature in T1 post-