ESTRO 38 Abstract book

S1037 ESTRO 38

obtained from a total of 26 single fraction clinical plans created for solitary lesions ranging in diameter (equivalent sphere) from 0.5cm to 3cm. Results For the 26 plans investigated, the model predicted the V 12Gy to within 0.67cc on average with a standard deviation of 0.53cc. Predictions were found to be most sensitive to the sphericity and size of the target, where small changes in gradient and conformity had the largest impact on the final NTV volume. Conclusion This knowledge-based method for NTV prediction in intracranial SRS could be used as a guide for deciding the prescription dose to targets prior to treatment planning in a busy clinical environment. EP-1909 Delta-radiomics signature predicts outcomes after preoperative chemoradiotherapy in rectal cancer C. Song 1 , J. Seung Hyuck 1 , K. Bohyoung 2 , K. Jae-Sung 1 1 Seoul National Univ. Bundang Hospital, Radiation Oncology, Seongnam- Gyeonggi-Do, Korea Republic of ; 2 Hankuk University of Foreign Studies, Division of Biomedical Engineering, Yongin, Korea Republic of Purpose or Objective To develop and compare delta-radiomics signatures from 2- (2D) and 3-dimensional (3D) features that predict treatment outcomes following preoperative chemoradiotherapy (CCRT) and surgery for locally advanced rectal cancer. Material and Methods In total, 101 patients (training cohort, n = 67; validation cohort, n = 34) with locally advanced rectal adenocarcinoma between 2008 and 2015 were included. We extracted 55 features from T2-weighted magnetic resonance imaging (MRI) scans. Delta-radiomics feature was defined as the difference in radiomics feature before and after CCRT. Signatures were developed to predict local recurrence (LR), distant metastasis (DM), and disease-free survival (DFS) from 2D and 3D features. The least absolute shrinkage and selection operator regression was used to select features and build signatures. The delta-radiomics signatures and clinical factors were integrated into Cox regression analysis to determine if the signatures were independent prognostic factors. Results The radiomics signatures for LR, DM, and DFS were developed and validated using both 2D and 3D features. Outcomes were significantly different in the low- and high-risk patients dichotomized by optimal cutoff in both the training and validation cohorts. In multivariate analysis, the signatures were independent prognostic factors even when considering the clinical parameters. There were no significant differences in C-index from 2D vs. 3D signatures. Conclusion Delta-radiomics signatures from both 2D and 3D features successfully predicted the outcomes and were independent prognostic factors irrespective of other conventional clinicopathologic factors. External validation is warranted to ensure their performance. EP-1910 CT image standardization is superior to larger but heterogeneous datasets for robust radiomic models D. Vuong 1 , M. Bogowicz 1 , J. Unkelbach 1 , R. Foerster 1 , S. Denzler 1 , A. Xyrafas 2 , M. Pless 2 , S. Thierstein 2 , S. Peters 3 , M. Guckenberger 1 , S. Tanadini-Lang 1 1 University Hospital Zurich and University of Zurich and Swiss Group for Clinical Cancer Research SAKK, Radiation Oncology, Zürich, Switzerland ; 2 Swiss Group for Clinical Cancer Research SAKK, Coordinating Center, Bern, Switzerland ; 3 Centre Hospitalier Universitaire Vaudois CHUV, Oncology, Lausanne, Switzerland

the discretization method based on relative range outperformed those based on bins of fixed width in units of SUV and was then adopted. The less robust families considering repeatability and discrimination were the Gray Level Run Length Matrix (GLRLM) and the Neighborhood Gray Tone Difference (NGTD) while the Grey Level Cooccurrence Matrix (GLCM) and the SUV first order RF were the most suitable. In Figure 1, the red bars indicate the most robust RF considering acquisition/processing, 14 of first and 20 of higher order. The resulting RF suitable for radiomic studies shown in Figure 1 were 11 of first and only 5 of higher order. Relaxing the limit for delineation agreement to ICC≥0.60, the numbers increased to 12 and 8 respectively.

Conclusion Based on a large set of phantom experiments, a list of 34 FDG-PET RF (from 72) was suggested as sufficiently robust against both repeatability and pattern discrimination. Delineation uncertainty for Pca is quite large, reducing the number of robust RF to 16/34, increased to 20 for a lower ICC threshold (0.60) for contouring agreement. EP-1908 A Guide For Predicting Normal Tissue Dose in Stereotactic Radiosurgery D. Cummins 1 , C. Skourou 1 , S. O'Sullivan 2 , P. Davenport 1 , D. Fitzpatrick 2 , C. Faul 2 , M. Javadpour 3 , M. Dunne 4 1 St.Luke's Radiation Oncology Network, Radiotherapy Physics, Dublin, Ireland ; 2 St.Luke's Radiation Oncology Network, Radiation Oncology, Dublin, Ireland ; 3 St.Luke's Radiation Oncology Network, Beaumont Hospital National Centre for Neurosurgery, Dublin, Ireland ; 4 St.Luke's Radiation Oncology Network, Clinical Trials, Dublin, Ireland Purpose or Objective The defining factors for selecting a prescription dose for intracranial metastases in stereotactic radiosurgery (SRS) are the size of the target and consequently the dose received by surrounding normal brain tissue. Prescription dose is recursively adjusted [24Gy-18Gy] following completion of a plan, until the normal brain tissue dose constraints are met. The availability of a nomogram to inform dose selection from the time of diagnosis would reduce planning time by eliminating the recursive adjustment of the prescription. This paper describes the development of such a guide using a knowledge-based method for predicting normal tissue dose as a function of Data from 50 previous SRS treatment plans (completed on iPlan with non-co-planar dynamic conformal arcs) was used to extract the terms for the key planning metrics (conformity index and gradient index) and define them as a function of target volume. The relationship between the measured target diameter and the dose to normal tissue volume (NTV) was established by approximating a spherical target volume covered by the prescription dose which could inform the expected V 40% . A scaling parameter described by a modeled non-linear fall-off of dose beyond the target was then used to scale V 40% in order to provide a first order approximation of the resulting NTV The predictive model was retrospectively validated against calculated NTV (chosen as V 12Gy for this study) target diameter in SRS. Material and Methods