Abstract Book

S106

ESTRO 37

Position of head structures in the repeated setup procedures were also reproducible. Structures in neck and shoulder region which were comparably unstable may need better immobilization devices and careful setup to improve the position reproducibility. OC-0196 Reproducibility of the liver in stereotactic radiotherapy using abdominal compression. D. Zwiers 1 , S. Koekoek 1 , A. Bouwhuis-Scholten 1 , H. Piersma 1 , E. Van Dieren 1 1 Medisch Spectrum Twente, Radiotherapy, Enschede, The Netherlands Purpose or Objective In stereotactic treatment of a liver laesion, accurate positioning is essential. This is usually done using an abdominal compression band. However, patient comfort limits the amount of compression and residual movement and positioning error may occur. Hence, it is standard to use CBCT a priori and posteriori, to judge treatment positioning accuracy. Due to a limited field of view of the CBCT, the whole liver usually cannot be imaged. In such a case, the accuracy cannot be judged fully. The aim of this study is to develop a new method to judge the partially imaged liver and to determine the reproducibility of the liver positioning in abdominal compression. Material and Methods In total 24 patients treated in our institution were included in this retrospective study. All patients had been imaged a priori and a posteriori, using Cone Beam CT (CBCT) with standard abdominal settings (on a Varian TrueBeam), and an abdominal compression band (Orfit Industries) was applied. For this study, it was assumed that liver contour reproducibility was directly related to target reproducibility. For each patient, the reproducibility of the liver contour was determined by comparing the differences in center of mass (CMS), CMS-vector and DICE similarity coëfficiënt between the CBCT and planning CT. Using set limits, results were classified as poor, sufficient, or excellent. When liver was only partially imaged (75% of cases), contours of the plan CT were shortened to match CBCT length, allowing all patients to be analysed similarly. The suitability of this method was determined by comparing, for those patients for whom the entire liver was imaged, the DICE and CMS-vector between whole-liver and shortened-liver approach. To track intrafractional differences, the CMS and DICE were compared and the delta-CMS was determined. Finally, the reliability of SmartAdapt was determined by comparing the DICE, CMS and CMS-vector of the automatic contours of SmartAdapt with a manually drawn liver contour. Results The liver DICE, CMS and CMS-vector were sufficient to excellent, excluding one patient. CMS-vectors between both methods did not significantly differ (p=0.600). The DICE was significantly better for the shortened liver approach (p=0.039). For the DICE and CMS, no significant intrafractional differences were found (p>0.244). The correlation between the duration of the irradiation and the DICE was weak (r=0.053). The manual liver contour was significantly improved in terms of DICE, CMS and CMS-vector (p<0.0001). Conclusion The new method is valid and ensures reliable measurements of all liver stereotactic irradiations. The reproducibility of the liver contour in abdominal compression between fractions and during the treatment is sufficient to excellent, indicating that stereotactic treatment the tumor in the liver has been performed accurately.

Poster Viewing : Poster viewing 4: Planning applications and optimisation algorithms

PV-0197 Comparison of manual and two automated planning solutions for stereotactic brain radiosurgery M. Zamburlini 1 , J. Krayenbuehl 1 , A. Sonbadhar 1 , D. McDonnell 1 , M. Guckenberger 1 , N. Andratschke 1 1 University Hospital Zürich, Radiation Therapy, Zurich, Switzerland Purpose or Objective Stereotactic radiosurgery (SRS) is becoming increasingly more important in the management of brain metastases (BM). SRS treatment planning is challenging, because adequate beam geometry is patient dependent, which is particularly complex in the treatment of multiple BM (mBM) with a single isocenter. In this in-silico planning study, we compared two dedicated automatic treatment planning solutions (TPS) for BM with manual optimization. Material and Methods The planning study is based on 10 randomly selected patients treated with SRS at our hospital. For comparison purposes, the patients were classified in 3 groups: single BM (n=3); 3-5 BM (n=4) and >10 BM (n=3). This in-silico planning study compared manually generated treatment plans (Eclipse v13.6) with a fully automated planning solutions: Elements Multiple Brain Mets SRS (EL) (v1.5, Brainlab, Germany) and a semi-automated planning solution: Eclipse v15.5 with Hyperarc module (HA) (Varian Medical Systems, Palo Alto, USA). All plans were prepared for a TrueBeam linac with HD-MLC and a single isocenter. Plans were normalized to the PTV to have D100% > 99%. An inhomogeneous dose prescription was used with at least 113% of the prescribed dose (PD) covering 95% of each GTV and a Dmax of 123%-127% of the PD. Plans were compared based on target volume coverage, conformity index (CI) and dose to normal tissue. Results HA and CP achieved similar target coverage. However, EL was not able to achieve the requested dose inhomogeneity. EL plans were therefore re-normalized to achieve comparable GTV coverage. Table 1 and Figure 1 show the comparison in PTV and GTV coverage, as well as normal tissue sparing between the TPS. HA and CP show comparable PTV and GTV coverage. EL showed better PTV and GTV coverage, but less a favorable CI, due to the increased normalization to reach the necessary GTV mean dose. In the case of >10 BM, EL showed better brain sparing, probably due to the beam arrangement, which selected only a few targets to be treated for each given arc. Both EL and HA provided an advantage in terms of planning time with respect to CP. EL is fully automated and it required less than 5 minutes effective planning time. HA provides an automatic selection of the beam arrangement, which reduced planning time by 15 minutes or more for mBM depending on the number of lesions. Effective planning time remained at least 45 minutes for both HA and CP depending on the number of targets.