7th ICHNO Abstract book

7th ICHNO 7 th ICHNO Conference International Conference on innovative approaches in Head and Neck Oncology 14 – 16 March 2019 Barcelona, Spain __________________________________________________________________________________________ page 61

pre-treatment 18F-FDG PET/CT (PETpre), and MRI/CT for RT planning purposes in our department. We focused on patients who experienced a recurrence at the site of primary tumour. All of them did an 18F-FDG PET/CT (PETrec) at the time of failure. In recurrent patients, the GTV-PETpre and the GTV- PETpost were contoured by means of an adaptive thresholding algorithm implemented on the dedicated iTaRT workstation (Tecnologie Avanzate, Italy). Both GTV- PETpre and the GTV-PETrec were transferred on the original planning CT scans by means of deformable co- registration of PETrec on PETpre in the Ray-Station treatment planning system. The overlapping volume of the pre-treatment volume and failure volume was generated: “GTV-PETpre ∩ GTV- PETrec”. The dose delivered to the 99% of a volume (D99) was measured within GTV-PREpre ∩ GTV-PETrec and GTV- PETrec. The recurrent volume was defined as: ‘‘In-Field (IF)’’, ‘‘Extending Outside the Field (EF)’’ or ‘‘Out-of- Field (OF)’’ if it had received >95%, 20–95% or <20% of the prescribed dose, respectively. Results We found 10/87 (11.5%) recurrences at primary site (2 oral cavity, 2 nasopharynx, 2 oropharynx, 3 hypopharynx and 1 larynx). The mean GTV-PETpre was 13.1 cc (4.6-37.4), while the mean GTV-PETrec was 4.3 cc (1.1-12.7). Mean D99 of GTV-PETpre ∩ GTV-PETrec was 68.1Gy, [66.5- 69.2], considering a prescription dose of 70 Gy to the PTV. Two recurrences were 100% inside GTV-PETpre, 4 recurrences were mostly inside (61-91%) and 4 recurrences were marginal to GTV-PETpre (33-1%). Six recurrences (60%) were defined as IF, 3 (30%) as EOF and one (10%) as OF. Conclusion In all 10 patients an overlap existed between the planning 18F-FDG PET and the recurrence scan, which indicates a high probability of the recurrence to originate from the GTV-PETpre volume. Furthermore 60% of recurrences were IF while 10% were OF. Our study indicates, even though not conclusive, that the recurrence may come from the strongest FDG-signal These results support the hypothesis of an intensification of the dose on these volumes. PO-120 Up-front F18-FDG PET/CT in suspected salivary gland carcinoma M. Westergaard-Nielsen 1 1 Odense University Hospital, ENT - Head and Neck Surgery, Odense C, Denmark Purpose or Objective To investigate whether a 18F-FDG-PET/CT (PET/CT) based diagnostic strategy adds decisive new information compared to conventional imaging in the evaluation of salivary gland lesions and the detection of cervical lymph node metastases, distant metastases, and synchronous cancer in patients with salivary gland carcinoma. Material and Methods The study was a blinded prospective cohort study. Data were collected consecutively through almost three years. All patients underwent conventional imaging - magnetic resonance imaging (MRI) and chest X-ray (CXR) - in addition to PET/CT prior to surgery. Final diagnosis was obtained by histopathology. MRI/CXR and PET/CT were interpreted separately by experienced radiologists and nuclear medicine physicians. Interpretation included evaluation of tumour site, cervical lymph node metastases, distant metastases, and synchronous cancer.

Results Ninety-one patients were included in the study. Thirty- three had primary salivary gland carcinoma with cervical lymph node metastases in eight. With PET/CT, the sensitivity and specificity regarding the detection of these two conditions were 92% and 29%, and 100% and 68%, respectively. With MRI/CXR they were 90% and 26%, and 50% and 88%, respectively. PET/CT diagnosed distant metastases in five patients, while MRI/CXR detected these in two patients. Finally, PET/CT diagnosed two synchronous cancers, whereas MRI/CXR did not detect any synchronous cancers. Conclusion Compared with MRI/CXR PET/CT did not improve discrimination of benign from malignant salivary gland lesions. However, PET/CT may be advantageous in primary staging and in the detection of distant metastases and synchronous cancers. PO-121Use of radiomics in the recurrence patterns after IMRT for Head and Neck cancer: a preliminary study S. Li 1 , Z. Hou 1 , K. Wang 1 , J. Yang 1 , W. Ren 1 , S. Gao 1 , F. Meng 1 , P. Wu 1 , B. Liu 1 , J. Liu 1 , J. Yan 1 1 The Comprehensive Cancer Centre of Drum Tower Hospital- Medical School of Nanjing University & Clinical Cancer Institute of Nanjing University, The Comprehensive Cancer Centre of Drum Tower Hospital, Nanjing, China Purpose or Objective To analyze the recurrence patterns and reasons in patients with head and neck cancer (HNC) treated with intensity- modulated radiotherapy (IMRT) and to investigate the feasibility of radiomics for analysis of nasopharyngeal carcinoma (NPC) radioresistance. Material and Methods We analyzed 504 HNC patients treated with IMRT from Jul- 2009 to Aug-2016, 26 of whom developed with recurrence. For the HNCs with recurrence, CT, MR or PET/CT images of recurrent disease were registered with the primary planning CT for dosimetry analysis. The recurrences were defined as in-field, marginal or out-of-field, according to dose-volume histogram (DVH) of the recurrence volume. To explore the predictive power of radiomics for NPCs with in-field recurrences (NPC-IFR), 16 NPCs with non-progression disease (NPC-NPD) were used for comparison. For these NPC-IFRs and NPC-NPDs, 1117 radiomic features were quantified from the tumor region using pre-treatment spectral attenuated inversion- recovery T2-weighted (SPAIR T2W) magnetic resonance imaging (MRI). Intraclass correlation coefficients (ICC) and Pearson correlation coefficient (PCC) was calculated to identify influential feature subset. Kruskal-Wallis test and receiver operating characteristic (ROC) analysis were employed to assess the capability of each feature on NPC- IFR prediction. Principal component analysis (PCA) was performed for feature reduction. Artificial neural network (ANN), k-nearest neighbor (KNN) and support vector machine (SVM) models were trained and validated by using stratified 10-fold cross validation.