ESTRO 2020 Abstract book

S479 ESTRO 2020

Conclusion Early mortality after RT/C-RT for HNSCC decreased significantly between 2000-2017, but not in recurrence- free patients. The lack of a specific high-risk period questions the use of a 90-day period as a standard metric for treatment-associated toxicity. PO-0799 Target delineation uncertainties using MRI for H&N cancer patients R. Zukauskaite 1,2 , C.N. Rumley 3,4,5 , C.R. Hansen 2,6,7 , M. Jameson 3,5,8 , Y. Trada 3,5,8 , J. Johansen 1 , N. Gyldenkerne 1 , J.G. Eriksen 9 , F. Aly 3,5,8 , M. Heinke 8 , R.L. Christensen 2,7 , M. Lee 5,8 , C. Brink 2,7 , L. Holloway 3,5,8 1 Odense University Hospital, Department of Oncology, Odense, Denmark ; 2 Institute of Clinical Research, University of Southern Denmark, Odense, Denmark ; 3 Ingham Institute for Applied Medical Research, Ingham Institute for Applied Medical Research, Sydney, Australia ; 4 Townsville Hospital and Health Service, Department of Radiation Oncology, Townsville, Australia ; 5 University of New South Wales, South Western Clinical School, Sydney, Australia ; 6 Aarhus University Hospital, Danish Center of Particle Therapy, Aarhus, Denmark ; 7 Odense University Hospital, Laboratory of Radiation Physics, Odense, Denmark ; 8 Liverpool and Macarthur Cancer Therapy Centre, Department of Radiation Oncology, Sydney, Australia ; 9 Aarhus University Hospital, Department of Experimental Clinical Oncology, Aarhus, Denmark Purpose or Objective MR imaging is often a part of radiotherapy planning in patients with head and neck (H&N) cancer. With the introduction of MR-guided adaptive radiotherapy, MRI will become central to volume definition. However, the utilization of MRI instead of PET/CT in radiotherapy planning is not well established; additionally, the inter- observer variation (IOV) of target delineation using MRI alone is unclear. We aimed to analyse IOV in delineating GTV (gross tumour volume) on MR images for patients with H&N cancer. Material and Methods The cohort consisted of 15 and 17 H&N patients from two independent departments. Four clinical oncologists from Denmark and four radiation oncologists from Australia, with a range of clinical expertise, independently contoured primary tumour and nodal GTVs on T2-weighted MR images performed at baseline of treatment planning (Figure 1). The oncologists were provided with a summary of clinical and imaging reports, and a MRI contouring protocol. Contouring was performed using MIM or Pinnacle software. Simultaneous truth and performance level estimation (STAPLE) reference volumes were generated for each structure using all observer contours. IOV was analysed using the Dice similarity coefficient (DSC) for volume overlap measurement, and a mean Hausdorff distance for measurement of distance between surfaces of contours.

Results The majority of the 32 carcinomas were in oropharynx (n=20), mainly advanced, (T3-T4 or N+, 20 and 26, respectively (TNM classification 7 th Ed)). There were 32 GTV-Ts and 68 GTV-Ns contoured per each observer. The mean GTV-T volume across all patients was 12.1 (range of 0.68 - 44.1) cm 3 and average std across observers for all patients 3.2 cm 3 ; the average GTV-Ns volume across all patients was 4.8 (range 0.09 - 46.9) and average std 1.8 cm 3 . Mean DSC relative to a STAPLE volume for GTV-Ts and GTV-Ns across all patients were 0.72±0.12 (Figure 2) and 0.76±0.11, respectively. Mean Hausdorff distance for GTV- Ts and GTV-Ns across all patients were 1.19±0.1 and 0.11±0.14 cm, respectively.

Conclusion Knowledge of inter-observer variations in identifying treatment target is important heading to a MRI-guided adaptive radiotherapy era. Contouring GTVs in H&N on MRI shows good average agreement for delineation of GTV-T and GTV-N among eight observers from two separate radiation departments; however, there are still specific areas in tumours that are difficult to resolve as either malignant or oedema.