ESTRO 38 Abstract book

S1036 ESTRO 38

on three CBCTs acquired on the first, middle (12 th fraction) and last fraction of the treatment. The image intensity levels were normalized using the mean value of a region of interest in the bladder, and a voxel re-sampling was performed to obtain isotropic voxel spacing. To denoise images and enhance image characteristics all the CBCTs were convolved with different kernels (Gaussian, Laplacian of Gaussian, and Median). Radiomic features were extracted from 3D tumor regions and the delta- features, i.e. the relative feature values change were evaluated between the mid and last fraction of the RT (Delta1), the first and mid-fraction of the CT- RT (Delta2), and the first and last fraction of CT-RT (Delta3). Wilcoxon signed-rank test was used to evaluate if delta-features in patients with pCR after CT-RT were significantly different from patients with partial or no response. Results Nine patients had pCR after CT-RT. Eleven Delta1, five Delta2, and twelve Delta3 features were significantly different (p < 0.05) in patients with pCR. Conclusion These preliminary results show the potential of delta- radiomics for predicting pCR of rectal cancer patients early during the CT-RT. EP-1907 Which FDG-PET features are robust enough for Radiomic studies in pancreatic cancer patients? L. Presotto 1 , M. Mori 2 , P. Passoni 3 , M.L. Belli 4 , S. Broggi 2 , G.M. Cattaneo 2 , M. Picchio 1 , N.G. Di Muzio 3 , V. Bettinardi 1 , C. Fiorino 2 1 IRCCS San Raffaele Scientific Institute, Nuclear Medicine, Milano, Italy ; 2 IRCCS San Raffaele Scientific Institute, Medical Physics, Milano, Italy ; 3 IRCCS San Raffaele Scientific Institute, Radiotherapy, Milano, Italy ; 4 IRST Istituto Scientifico Romagnolo per lo Studio e la Cura dei tumori, Medical Physics, Meldola, Italy Purpose or Objective FDG-PET Radiomics is promising for the characterization of pancreatic cancer (Pca). However, uncertainties due to delineation/segmentation and to acquisition/processing may affect its reliability. Aim of this study was to assess robust radiomic features (RF) based on the impact of delineation uncertainty and of parameters affecting image Twenty-five Pca patients previously treated with IMRT were considered. Four PET Pca contours were available: 2 manual, 1 semi-automatic (based on SUV maximum gradient: PET_Edge) and 1 automatic (40%SUV_max thresholds): 72 RF were extracted with CGITA software (v. 1.4). The stability of RF against inter-observer variability was quantified by Intra-Class Correlation Coefficient (ICC); the robustness of RF for PET_Edge and 40%SUV was assessed against manual delineation. The impact of acquisition/ processing on RF was tested on uniform and purpose built heterogeneous phantoms. Statistical effect size indicators were used to determine: i) the impact of each nuisance factor (discretization, acquisition statistics, reconstruction algorithm, filtering), and ii) the discriminating power. Based on the robustness with respect to all factors, we categorized RF based on repeatability and on heterogeneous PET patterns discrimination, leading to the definition of a list of robust RF. Among them, the ones showing an ICC≥0.80 (for “inter-observer variability”) in the delineation study were finally defined as suitable for Pca Radiomics. Results Inter-observer agreement was moderate (median DICE: 0.73); 35 (47%) RF showed an ICC<0.80, mostly in the Voxel-Alignment (VA) matrix and in the Intensity-Size Zone (ISZ) matrix families. The number of RF with ICC<0.80 for PET_Edge and SUV40% (considering the worst ICC value against observers) increased to 44 and 54 respectively. Regarding image acquisition/processing, acquisition/processing. Material and Methods

dosiomics features, of nasopharyngeal carcinoma (NPC) treated by Intensity Modulated Radiation Therapy (IMRT). Material and Methods Patients diagnosed with NPC treated with RT were included in this study. Clinical and instrumental follow-up was performed every month for the first 2 years, then every 6 months. Pre-treatment PET and CT-scans were collected as well as the three dimensional dose distribution calculated on the CT. The CT, PET images and the calculated dose distribution were pre-processed with re-sampling and 3-D filtering using Gaussian, Laplacian of Gaussian, and Median filters. 728 radiomic shape, size, histogram-based and textural features were calculated from the filtered and unfiltered images and dose in the gross target volume, which was contoured using CT and PET. Sequential feature selection was used to identify a subset of features that best predict the data and remove redundant or not significant predictors. An ensemble learning classifier with adaptive boosting was trained on the patient dataset for prediction of local control (positive classifier for appearance of disease in the treated site during follow-up, negative otherwise). The predictive power of the model was assessed using sensitivity (probability that test is positive on patients with recurrence) and specificity in five-fold cross validation. The area under ROC curve (AUC) was used to investigate correlation of features with recurrence. Results After a median follow-up of 31.4 (95%CI 3.8-86.7) months, 49 out of 60 (82.6%) patients were free from local recurrence. The features selected were 1 shape (solidity), 1 CT (Low Gray level zone emphasis from GLSZM), 1 PET (Second measure of information correlation of GLCM) and 3 dose features (Percentile area 90, Strength of NGTDM, Small Zone high grey level emphasis of GLSZM). The classifier scored sensitivity 72.7% and specificity 89.8% in the cross validation. The single most predictive feature for recurrence was PET (AUC=0.712). The category with highest AUC was dose, as the combined features had AUC of 0.735. Conclusion These findings show that PET radiomic and dosiomic variables are correlated with local control of NPC. The model incorporating radiomic features from imaging and dose can predict local control in this type of head and neck cancer. EP-1906 CBCT delta-radiomics for predicting complete pathological response of rectal cancer after CT-RT G. Pirrone 1 , E. Palazzari 2 , F. Navarria 2 , R. Innocente 2 , J. Stancanello 3 , G. Fanetti 2 , G. Franchin 2 , C. Cappelletto 1 , A. De Paoli 2 , G. Sartor 1 , M. Avanzo 1 1 Centro di Riferimento Oncologico di Aviano CRO IRCCS, Medical Physics Department, Aviano, Italy ; 2 Centro di Riferimento Oncologico di Aviano CRO IRCCS, Radiation Oncology Department, Aviano, Italy ; 3 Oncoradiomics SA, Oncoradiomics SA, Liege, Belgium Purpose or Objective The study of the variation of radiomic features extracted from cone-beam CT (CBCT) acquired for Image-Guided Radiotherapy (IGRT) opens a promising scenario for assessing tumor response during chemo-radiotherapy (CT- RT). The objective of this study is to determine whether variation of radiomic features (delta-radiomics) predicts complete pathological response (pCR) of patients treated with RT for rectal cancer. Material and Methods The analysis was conducted on daily-setup imaging data of a total of 19 patients diagnosed with rectal cancer who received preoperative volumetric arc therapy (VMAT) with a prescribed dose of 54 Gy in 25 fractions. All the CBCT images were acquired with 125 kVp and 80 mA with a Varian Trilogy LinAc. An expert radiation oncologist contoured the gross volume target (GTV) for each patient