7th ICHNO Abstract book

page 62 7 th ICHNO Conference International Conference on innovative approaches in Head and Neck Oncology 14 – 16 March 2019 Barcelona, Spain __________________________________________________________________________________________ 7th ICHNO

Conclusion In-field and high-dose region relapse were the main recurrence patterns which may be due to the radioresistance. After integration in the clinical workflow, radiomic analysis can be served as imaging biomarkers to facilitate early salvage for NPC patients who are at risk of in-field recurrence. PO-122 CT /PET based dosiomics and radiomics model predicts local control of nasopharyngeal carcinoma G. Fanetti 1 , M. Avanzo 2 , G. Pirrone 2 , C. Avigo 3 , J. Stancanello 4 , A. De Paoli 1 , E. Palazzari 1 , A. Drigo 2 , C. Gobitti 1 , E. Vaccher 5 , T. Baresic 6 , C. Bampo 6 , E. Borsatti 6 , G. Sartor 2 , G. Franchin 1 1 IRCCS Centro di Riferimento Oncologico CRO National Cancer Institute, Division of Radiation Oncology, Aviano, Italy; 2 IRCCS Centro di Riferimento Oncologico CRO National Cancer Institute, Division of Medical Physics, Aviano, Italy; 3 S. Martino Hospital, Division of Medical Physics, Belluno, Italy; 4 Maastricht University Medical Centre, The D-Lab: Decision Support for Precision Medicine- GROW–School for Oncology and Developmental Biology, Maastricht, The Netherlands; 5 IRCCS Centro di Riferimento Oncologico CRO National Cancer Institute, Division of Medical Oncology, Aviano, Italy ; 6 IRCCS Centro di Riferimento Oncologico CRO National Cancer Institute, Division of Nuclear Medicine, Aviano, Italy Purpose or Objective To develop and validate a model predictive for local control, based on CT-PET radiomics and planning CT 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 (F-UP) was performed every 3 months 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, and area under ROC curve 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

Results The median follow up was 26 (range 3-65) months. 13/26 (50%) occurred in the primary tumor, 8/26 (31%) occurred in regional lymph nodes, and 5/26 (19%) patients developed a primary and regional failure. Dosimetric and target volume analysis of the recurrence indicated that there were 24 in-field, and 1 marginal as well as 1 out-of- field recurrence. Among the HNCs with recurrence, 20 NPCs developed in-field failure (NPC-IFR). With pre- therapeutic SPAIR T2W MRI images available, 11 NPC-IFRs (11 of 20 NPC-IFRs who had available pre-therapeutic MRI) and 16 NPC-NPDs were subsequently employed for radiomic analysis. Results showed that NPC-IFRs versus NPC-NPDs could be differentiated by 8 features (AUCs: 0.727-0.835). The classification models showed potential in prediction of NPC-IFR with higher accuracies (ANN: 0.812, KNN: 0.775, SVM: 0.732).