ESTRO 36 Abstract Book

S655 ESTRO 36 2017 _______________________________________________________________________________________________

nodes.

above structures was reduced from 18.1±4.8 to 13.2±5.5 minutes (p<0.01), mainly dominated by the reduction in time needed for the mediastinal delineation and heart. Conclusion Besides a reduction in contouring time, the inter-observer variation is also reduced if an atlas-based segmentation approach is used as the initial starting point for delineations. Especially for the larger structures such as the heart and mediastinum the impact on time gain and increase of quality is significant. EP-1226 Stereotactic robotic body radiotherapy for patients with pulmonary oligometastases P. Berkovic 1 , A. Gulyban 1 , L. Swenen 1 , D. Dechambre 1 , P. Viet Nguyen 1 , N. Jansen 1 , C. Mievis 1 , N. Bartelemy 1 , P. Lovinfosse 1 , M. Baré 1 , F. Lakosi 2 , L. Janvary 3 , P.A. Coucke 1 1 C.H.U. - Sart Tilman, Radiotherapy department, Liège, Belgium 2 Health Science Center- University of Kaposvar, Radiation Oncology, Kaposvar, Hungary 3 University of Debrecen - Medical Center, Onco logy Clinic, Debrecen, Hungary Purpose or Objective To analyse local control (LC), pulmonary and distant progression free survival (pulmonary PFS, DFS), overall survival (OS) and toxicity in a cohort of patients treated by stereotactic body radiotherapy (SBRT) for oligometastatic pulmonary lesions. To evaluate the potential influence of age, histology, controlled primary, performance status, biological effective dose (BED) and other parameters on the obtained results. Material and Methods Consecutive patients with up to 3 synchronous lung metastases were included in this study for Cyberknife at the Liege University Hospital. All patients were referred for stereotactic treatment after a full staging including baseline registration of the pulmonary function, chest and abdominal diagnostic computed tomography (CT) and [18F]-fluorodeoxyglucose (FDG) positron emission tomography (PET)-CT imaging confirming the presence or absence of tumoral activity at the primary tumour site and extra-pulmonary metastases. The intended prescription dose was 60 Gy in 3 fractions, prescribed on the 80% isodose line and adapted based on clinical risk-factors. Local control (LC), lung and distant progression free survival (lung and distant PFS) and overall survival (OS) of patients were generated using Kaplan-Meier survival curves. Age, gender, performance status (PS), primary histology, controlled primary as patient specific, while total BED10Gy (a/b = 10) prescribed dose as treatment related factors were analysed using log-rank test to determine their impact on outcome. Results Between 05/2010 and 03/2016, 131 patients with 164 lesions were irradiated. Treatments were delivered 3x/week in a median of three fractions. According to the RECIST criteria a complete or partial response were observed in 86 and 27 lesions, while 12 remained stable. After mean follow-up of 14 months, the 1 and 2-year LC/lung PFS/DPFS/OS were 85.0/62.2/82.6/91.3% and 69.0/44.8/69.8% and 77.9% respectively. Age (>65 years) and controlled primary tumour influenced DPFS (p=0.017) and OS (p=0.02) respectively, while LC and OS differed significantly for BED10Gy (>120 vs. <=120 Gy, p<0.001 and p =0.016) and primary histology (adenocarcinoma or others, p=0.003 and p=0.006) (Figure 1 and 2). Grade 1/2/3/4 fatigue, chest pain and dyspnoea were present in 77/3/0/0, 20/0/0/0 and 26/1/1/0 treatments as acute, while 22/0/0/0, 14/37/0/0 and 18/2/3/1 as late toxicity. One patient died due to RT-induced pulmonary haemorrhage.

Conclusion Chemoradiotherapy using AHF may achieve a higher pathological therapeutic effect than chemoradiotherapy using CF for squamous cell lung cancer in primary tumors. EP-1225 Atlas-based segmentation reduces inter- observer variation and delineation time for OAR in NSCLC W. Van Elmpt 1 , J. Van der Stoep 1 , J. Van Soest 1 , T. Lustberg 1 , M. Gooding 2 , A. Dekker 1 1 MAASTRO Clinic, Department of Radiation Oncology, Maastricht, The Netherlands 2 Mirada Medical Ltd, Science and Medical Technology, Oxford, United Kingdom Purpose or Objective Tumor and organs-at-risk (OAR) delineations are considered a major uncertainty in radiotherapy. Automatic segmentation methods are currently available that may guide the delineations of OAR. However, the inter-observer variability in OAR delineations are rarely studied and the effect of automated methods on delineation variability has not yet been performed. In this study we systematically quantified the (reduction of) inter-observer variation by providing the delineation expert with an atlas-based generated automatic contour including time spent on delineations. Material and Methods Atlas-based automatic delineations were performed using commercial available software with an atlas derived for 10 stage I NSCLC patients using institutional delineation guidelines with minimal anatomical distortions. In a next step, 20 consecutive prospective stage I-III NSCLC patients were selected from clinical routine. For these patients, 3 experienced radiation technologists independently created delineations for heart, mediastinum, spinal cord, esophagus and brachial plexus according to the institutional standards. Time taken was also recorded. Next, the automatic atlas-based contour was provided as a starting point for a second round of delineations (blinded for the initial contour). The proposed contour was allowed to be adapted (or discarded) and modified into a clinical acceptable contour. The inter-observer variation was quantified as the non-overlapping volume of the 3 observers for both the initial contours and the adapted contours. Results are expressed as mean±SD, p-values calculated using a Wilcoxon test. Results Comparing the initial contours with the proposed atlas- generated contour, the inter-observer variation volumes reduced significantly for the mediastinum: 253±93 cm 3 to 168±103 cm 3 (p<0.01), spinal cord: 32±10 cm 3 to 17±3 cm 3 (p<0.01) and heart: 211±69cm 3 to 136±72 cm 3 (p<0.01). For the esophagus there was no reduction inter-observer variation volume (p=0.601), also no clinically significant differences for brachial plexus were observed: 12.9±5.4 cm 3 vs 12.2±5.1cm 3 . The average delineation time for the

Figure 1 : Kaplan-Meier curves and log-rank test for LC