ESTRO 38 Abstract book

S515 ESTRO 38

were determined by recursive feature elimination. The radiomic features were not normalized on any data sets. Multivariable logistic regression, SVM, and random forest, were applied for training classifier. The classification performance of HPV status was assessed by the area under the receiver operator curve (AUC). The Wilcoxon test was used to assess significance between AUCs and random (AUC=0.5). Results Out of 255 patients, 174 (68.2%) patients were with HPV positive and the rest 81 (31.8%) patients were with negative. After feature selection, 3 radiomic features, (i) original_shape_Sphericity, (ii) original_firstorder_ Entropy, and (iii) log-sigma-3-0-mm- 3D_glszm_SizeZoneNonUniformityNormalized, were selected to develop models. The multivariable logistic regression classifier achieved the highest AUC on both training and test sets, yielding mean AUCs of 0.79 (95% CI: 0.78 - 0.79, p-value < 10 -4 Wilcoxon test) and 0.72 (95% CI: 0.71 - 0.72, p-value < 10 -4 Wilcoxon test), respectively. The AUCs are shown in Figure 1 . The performance of HPV status prediction of the three classifiers are shown in Table 1 .

(ROO) were used to make radiomics and clinical data extracted in the hospitals intelligible for models. We validated the model distributedly using the Varian Learning Portal securing local patient data.

Results We found a significant split in both training (log-rank test p=0.009) and validation cohorts (log-rank test p=0.03). Conclusion The original model was reproduced with the significant data split. This way we showed that imaging prediction models can be made in a distributed fashion without data- sharing. PO-0952 CT-based Radiomics Predicting HPV Status in Head and Neck Squamous Cell Carcinoma Z. Shi 1 , C. Zhang 2 , M. Welch 3 , P. Kalendralis 1 , W. Leonard 1 , A. Dekker 1 1 GROW – School for Oncology and Development Biology- Maastricht University Medical Centre, Department of Radiation Oncology Maastro Clinic, Maastricht, The Netherlands ; 2 Maastricht University, Department of Data Science and Knowledge Engineering, Maastricht, The Netherlands ; 3 University of Toronto, Department of Medical Biophysics, Toronto, Canada Purpose or Objective Human papillomavirus (HPV) testing is an important prognostic factor for oropharyngeal squamous cell carcinoma (OPSCC). HPV-related OPSCC is now considered to be a separate tumour type from non-HPV related tumours with differential cancer prognosis. Conventionally, HPV status is determined via a minimally invasive needle biopsy. The aim of this study was to investigate whether CT image-derived radiomics are able to predict HPV status of patients diagnosed with primary OPSCC in a non-invasive approach. Material and Methods Six independent cohorts, 255 patients in total were collected in this study, in which patients were treated with radiation only or chemo-radiation therapy as part of their treatment. HPV positive was defined as expression of the p16 gene variant. CT scans with visible artifacts (e.g., metallic dental fillings) within the GTV were excluded from further analysis. The data was randomly split into training (n=142), tuning (n=48) and test (n=65) sets. CT images were resampled to isotropic voxels of 2 mm via linear interpolation. A total of 1105 radiomic features, consisting of histogram statistics, shape, texture and features by Wavelet and Laplacian of Gaussian filtering, were extracted from the GTV via an open-source radiomics package O-RAW that is an extension wrapper of PyRadiomics. Feature selection was based on: (i) feature stability ranking, (ii) Kolmogorov-Smirnov test of each feature between HPV positive and negative, and (iii) correction between the remaining features. For the model, a parsimonious yet optimally predictive set of features from the remaining candidates after selection

Conclusion It is possible to classify HPV status for HNSCC patients using CT image-derived features, which could lead to better precision treatment decision. However, the results should be further validated on larger and external datasets. PO-0953 Are quality assurance phantoms useful to assess radiomics reproducibility? A multi-center study A. Traverso 1 , I. Zhovannik 2 , Z. Shi 1 , P. Kalendralis 1 , R. Monshouwer 3 , M. Starmans 4 , S. Klein 5 , E. Pfaehler 6 , R. Boellaard 6 , A. Dekker 7 , L. Wee 1 1 Maastricht Radiation Oncology Maastro Clinic, Radiotherapy, Maastricht, The Netherlands ; 2 Radboud University Medical Centre, Radiation Oncology, Nijmegen, The Netherlands ; 3 Radboud University Medical Center, Radiation Oncology, Nijmegen, The Netherlands ; 4 Erasmus Medical Center, Medical Physics, Rotterdam, The Netherlands ; 5 Erasmus Medical Centre, Medical Physics, Rotterdam, The Netherlands ; 6 University Medical Centre Groningen, Medical Physics, Groningen, The Netherlands ; 7 Maastricht Radiation