Abstract Book

S98

ESTRO 37

respectively, while all OAR had a mean Dice Coefficient of 0.8 ± 0.1. In 13-29% of the FxCTs, depending on the OAR, the V35Gy dose constraint was violated by more than 1 ml (Table 1). Through PCA, the first three principal components mainly revealed anterior-posterior deformations for the stomach and the duodenum whereas the bowel mostly deforms in the cranio-caudal direction (Fig 1). Thirty-three modes were required to describe the 90% of the variance.

Conclusion Strict control of rectum filling status using CBCT imaging ensured that doses delivered were similar to the doses that were planned. Patients with toxicity tended to have higher doses in the inferior part of the rectum. OC-0184 Principal component analysis of daily changes in organ-at-risk anatomy in pancreatic cancer patients A. Magallon-Baro 1 , P.V. Granton 1 , M. Loi 1 , A.G. Zolnay 1 , M.T.W. Milder 1 , J.J. Nuyttens 1 , M.S. Hoogeman 1 1 Erasmus MC - Cancer Institute, Radiation Oncology, Rotterdam, The Netherlands Purpose or Objective Integration of chemotherapy with SBRT emerged as a valuable option for patients with Locally Advanced Pancreatic Carcinoma (LAPC). At our institution, patients that respond to chemotherapy receive a hypofractionated course of 5 x 8 Gy on the Cyberknife with tumor tracking accounting for respiratory motion via fiducial markers. A recurrent problem is that the prescribed dose to the tumor is impeded due to tight dose-volume constraints (V35Gy<1ml) of the surrounding healthy organs (stomach, duodenum, bowel). Therefore, day-to-day anatomical variations of these organs might be of a major concern when treating or could rather be exploited by daily adaptation of the treatment plan. With this in mind, we have explored a population-based statistical approach of these Organs-at-Risk (OARs) to characterize and describe the daily anatomical changes through principal component analysis (PCA). Material and Methods A total of a 100 high-resolution CT scans were collected for 25 LAPC patients, including the planning CT scan (pCT) and three subsequent in-room CTs (FxCT) acquired prior to fraction delivery. FxCT contours were propagated to the pCT using fiducial marker alignment. Quantitative metrics including surface-to-surface distance, Dice coefficient, volumetric differences and differences in the V35Gy dose constraint were abstracted. Daily variations of the OAR were characterized based on the deformation vector fields (DVFs) obtained from deformable registrations between the FxCT and pCT contours. Subsequently, an average organ mesh of the three OAR was created by averaging the prior DVFs of each subject. A second contour-based non-rigid registration was used to map the average DVF organs of each individual to the average DVF organs of a reference patient in order to create a population-based combined statistical PCA of the three contour sets. PCA was used to extract the eigenvectors which describe the most dominant directions of motion of each OAR. Results The range of motion of the three OAR with respect to the pCT had an average surface-to-surface and center-of- mass displacement between 6 - 10 mm and 8 - 12 mm,

Conclusion Analysis of delineated OAR in high-quality daily CTs shows the significance of OAR motion even when target tracking is used. A PCA model was successfully built and might provide valuable information to clinicians to improve management of daily OAR motion. To this end, PCA is now being used to explore planning restrictive volume margins, adaptive plan-of-the-day strategies and probabilistic treatment planning. OC-0185 SBRT using MR-guided, video-assisted gated treatment delivery during patient breath holds. J.R. Van Sornsen de Koste 1 , M.A. Palacios 2 , A.M.E. Bruynzeel 1 , B.J. Slotman 1 , S. Senan 1 , F.J. Lagerwaard 1 1 VU University Medical Center, Department of Radiation Oncology, Amsterdam, The Netherlands