ESTRO 36 Abstract Book

S868 ESTRO 36 _______________________________________________________________________________________________

2 University Hospital Zurich and University of Zurich, Department of Radiation Oncology, Zurich, Switzerland 3 Princess Margaret Cancer Center, Department of Radiation Oncology- University of Toronto, Toronto, Canada 4 VU University Medical Center, Department of Radiation Oncology, Amsterdam, The Netherlands 5 VU University Medical Center, Department of Otolaryngology/Head and Neck Surgery, Amsterdam, The Netherlands 6 University Hospital Zurich and University of Zurich, Department of Pathology and Molecular Pathology, Zurich, Switzerland Purpose or Objective Oropharyngeal squamous cell carcinoma (OPSCC) is one of the fastest growing disease sites of head and neck cancers. HPV positive cancers have been shown to have better tumor control with radiotherapy and increased survival, which makes them interesting for de-escalation protocols. HPV is routinely tested using in situ hybridization for viral DNA, or immunohistochemistry for p16. However, an established, non-invasive, imaging biomarker of HPV status currently does not exist. Radiomics–the high- throughput extraction of large amounts of quantitative features from medical images–has already been shown to be of prognostic value for head and neck cancer. In this study we evaluate the use of a Radiomic approach to identify the HPV status of OPSCC patients. Material and Methods Three independent cohorts, with a total of 793 OPSCC patients were collected: C1 (N=543), C2 (N=159) and C3 (N=100). HPV status was determined by p16 and available for 686 patients. Patients underwent pre-treatment CT imaging and the tumor volume was manually delineated for treatment planning purposes. Images were visually assessed for the presence of CT artifacts (e.g. streak artifacts due to dental fillings) within the GTV, in which case they were excluded from further analysis. In total, 1378 Radiomic features were extracted, comprising: a) first-order statistics, b) shape, and c) (multiscale) texture (Laplacian of Gaussian and Wavelet). The model was learned on the C1 cohort and validated on the remaining cohorts. The Radiomic feature space was first reduced by selecting cluster medoids after hierarchical cluster analysis using correlation (ρ>0.9) as a distance measure. Multivariable logistic regression was performed using least absolute shrinkage and selection operator (LASSO) model selection (200 times 10-fold cross-validated). The area under the receiver operator curve was used to assess out- of-sample model performance in predicting HPV status. Results Out of the patients with known HPV scoring, we identified 337 (49%) patients without visible CT artifacts: C1 (N=206), C2 (N=88), C3 (N=43), of which 132, 20, and 18 were HPV positive, respectively. The modeling process resulted in a multivariable prediction model, with an AUC of 0.85. External validation in the C2 and C3 cohorts showed an AUC of 0.6 and 0.72, respectively. The receiver operator curves for training and validation are shown in Figure 1.

Conclusion Varying the RBE depending on end-point may strongly influence results when estimating carcinogenic risks from C-ion therapy and should be included in modelling risk of radiation-induced SC from C-ion therapy. EP-1608 Deriving HPV status from standard CT imaging: a radiomic approach with independent validation R. Leijenaar 1 , M. Nesteruk 2 , G. Feliciani 1 , F. Hoebers 1 , J. Van Timmeren 1 , W. Van Elmpt 1 , S. Walsh 1 , A. Jochems 1 , S. Huang 3 , B. Chan 3 , J. Waldron 3 , B. O'Sullivan 3 , D. Rietveld 4 , C. Leemans 5 , O. Riesterer 2 , K. Ikenberg 6 , P. Lambin 1 1 MAASTRO Clinic, Department of Radiation Oncology- GROW- School for Oncology and Developmental Biology- Maastricht University Medical Centre, Maastricht, The Netherlands