ESTRO 33, 2014 holding period, longer inspiration period than expiration period similar to the patient breathing pattern. These patterns were generated by The QuasarTM Programmable respiratory motion phantom with lesion of 3 cm diameter. When target was moving in the amplitude of 1, 2 and 3 cm, the breath lengths were used: 2, 3 and 4s. All images obtained by each scan were delineated for each ITV lengths and volumes in Eclipse RTP system. Measured lengths and volumes were compared to the known values in which the lesion of known geometry moved a known distance. Results: In this study, for 2, 3, 4 cm amplitudes, known ITV length were 40 mm, 50 mm, 60mm and known volumes were 21.21 cc, 28.28 cc and 35.35 cc, respectively. For all patterns, when ITV lengths and volumes were compared with known values, 4DCT were the most similar, followed by CBCT and slow CT, respectively. Compared with each patterns, sinusoidal motion showed the smallest difference in 4D CT(2.25~-11.06%), slow CT (-14.5~22.78%) and CBCT (-9.75~13%). On the other hand, the difference of motion close to the patient breathing showed the largest in 4D CT (--4.83~20.25%), slow CT (-15.75~34.33%) and CBCT (-10.84~19.66%). There was also a distinct trend in difference of ITV length and volume with decreasing the breath length and increasing the amplitude. Conclusions: We evaluated the ITV length and volume between 4D CT, slow CT and CBCT using moving phantom for five breath patterns with various breath lengths and amplitudes. The difference of ITV lengths and volumes decreased with increasing the breath length and increased with increasing the amplitude for each breath pattern. This study is considered to be helpful to evaluate ITV using patient breath pattern in the future. EP-1705 Minimally interactive OAR and GTV segmentation in 4D FDG-18 PET/CT NSCLC: First clinical experience J. Dolz1, H. Kirisli1, M. Jurisic2, T. Fechter3, U. Christ4, S. Adebahr3, M. Mix3, W. Birkfellner2, L. Massoptier1, U. Nestle3 1 AQUILAB, Research Department, Lille, France 2 Medical University Vienna, Center for Medical Physics and Biomedical Engineering, Vienna, Austria 3 University Hospital Freiburg, Department of Radiation Oncology, Freiburg im Breisgau, Germany 4 University Hospital Freiburg, Department of Nuclear Medicine, Freiburg im Breisgau, Germany Purpose/Objective: Radiotherapy aims at delivering the highest possible dose to the GTV while minimizing the irradiation of surrounding healthy tissue. Hence, exact delineation of the tumor and its margins is a keystone in radiotherapy treatment planning. In case of lung cancer, respiratory movements observed on 4D NSCLC datasets make the GTV and OAR delineation a challenge. Furthermore, 4D annotation in clinical routine is an extremely tedious and time consuming task, and it suffers from high user variability. Here, a framework to segment 4D FDG-18 PET/CT NSCLC datasets with minimal user interaction is proposed. Materials and Methods: The proposed framework to segment OAR and GTV on 4D PET/CT NSCLC data follows a three-step strategy, and has been implemented using MITK platform. First, after providing few manual seeds using a 'paintbrush' tool, few OAR (lung, heart, aorta) are semi-automatically segmented in the first temporal 3D CT image of the 4D set. This is performed by using an energy minimization approach, involving watershed image transformation and graph-cut method. Second, non-rigid registration between 3D CT images is performed to compute the deformations along the 4D set.Thereby, the contours of the OAR segmented in the first time bin of the CT image are propagated in 4D. Last, the 4D lung contours are used as input to automatically segment the GTV in 4D PET images,using an adaptation of the Homburg algorithm. Both OAR and GTV segmentation methods were independently evaluated in previous works.Prior validation results (SPIE 2014. Medical imaging) indicate that both segmentation tools are accurate and robust. For this work, five FDG-18 PET/CT NSCLC datasets were used. Beside, for each 4D data, a clinical expert performed semi-automatic contouring at every time-point. These contours were used as reference standard to evaluate the accuracy of the deformed contours in step 2,by computing the overlap - Dice Similarity Coefficient (DSC) - between propagated and experts' contours at the same time bin. Results: Here,table 1 shows the quantitative results of the experiments. The curves correspond to the mean DSC computed between semiautomatic and propagated contours, for each OAR considered in this work, and for each of the 9 time bins in which the contours were propagated. Overall, the mean DSC was higher than 0.86, which demonstrates substantial agreement.
S249 Automatic GTV contours obtained in the step 3 showed a good performance of the proposed methodology.
Conclusions: A minimally interactive tool to segment OAR and GTV in 4D FDG-18 PET/CT NSCLC datasets has been proposed. Considering the results obtained, we may conclude that the preliminary clinical experiments provided promising results. Resulting contours may be used for various purposes, like ITV definition, mid ventilation volumes or aiding to treat moving targets. Consequently, further experiments will be carried out on larger databases to investigate its potential use in daily clinical practice. EP-1706 A pilot study examining deformable imaging in deriving a PET-based PTV for oesophageal cancer radiotherapy planning G. Ward1, S.M. Ramasamy2, J. Sykes1, F. Chowdhury3, A. Scarsbrook3, K. Harris2, P. Hatfield2, A. Crellin2, D. Sebag-Montefiore2, G. Radhakrishna2 1 Leeds Cancer Centre, Medical Physics and Engineering, Leeds, United Kingdom 2 Leeds Cancer Centre, Non Surgical Oncology, Leeds, United Kingdom 3 Leeds Cancer Centre, Clinical Radiology and Nuclear Medicine, Leeds, United Kingdom Purpose/Objective: Improving the delineation of the GTV remains a challenge in oesophageal cancer. Although PET-CT can identify the metabolically active gross tumour, there is uncertainty as to how it is best used. Detailed study of locoregional relapse is also important to determine the appropriate margins around the GTV for future studies of dose escalation. Materials and Methods: Data was collected for 16 patients treated between 2009 and 2010, 7 of which had infield recurrence. All patients were treated with chemoradiation receiving 3D conformal radiotherapy of 50Gy in 25 fractions to the PTV using the SCOPE 1 trial methodology. The PTV (PTVTP) uses the planning CT, diagnostic PET-CT and endoscopic ultrasound (EUS) without deformable image registration. In this study, the target volume was contoured on the PET-CT and deformable image registration (Mirada RTx, Mirada Medical, Oxford, UK) was used to transfer the volume to the planning CT. The entire circumference of the involved oesophagus was outlined as GTV (GTVPET) which was grown to form the PTVPET. This volume was compared to the PTVTP. For relapse patients, the relapse volume (GTVRELAPSE) was contoured on the relapse CT scan and deformable image registration was used to transfer the volume to the planning CT. The location of the relapse volume relative to the PTVTP and PTVPET was compared to determine its relationship to the high dose volume, i.e. the volume receiving 95% of the prescribed dose (figure 1).
Figure 1: For patient 6, the relapse volume is contained within both the PTVPET and the PTVTP. Results: The length of the PTVPET was shorter than the length of the PTVTP in 15/16 patients with an average reduction in the superior and inferior extensions of 1.8cm and 1.0cm respectively. Of the 7 relapse patients, one had small cell oesophageal cancer with an atypical pattern of relapse. This patient relapsed outside both the PTVPET and PTVTP. The remaining patients all relapsed in the high dose region (table 1). The GTVRELAPSE was contained within both the PTVPET and PTVTP for all but one patient. For this patient, the GTVRELAPSE extended beyond the inferior border of the PTVPET but was still contained within the PTVTP.