Abstract Book

S1138

ESTRO 37

against dosimetric variances induced by uncorrected inter-fractional anatomical changes. Material and Methods Data of 18 prostate cancer patients with seminal vesicles (SV) invasion were included, each consisting of a planning CT scan and 8-13 follow-up CT scans. On each scan the prostate+SV (CTV), rectum, and bladder were delineated. PTV margins used were 4 mm (prostate) and 7 mm (prostate+SV). For each planning CT a 9-beam IMRT (19x3.4Gy) MR-Linac treatment plan was made using Monaco (research). To simulate anatomical variance, the delineations from the follow-up CTs were copied to the planning CT after prostate registration. To simulate setup errors 40 adapted plans were made from each reference plan, each accounting for a random <1cm isocenter shift. The plans were adapted using segment aperture morphing [Ahunbay et al., Med. Phys., 2008], followed by segment weight optimization (SWO) only, or SWO + segment shape optimization (SWO+SSO). Following our current clinical guidelines on plan quality we evaluated the anal sphincter and bladder D50%, and rectum D35% for three scenarios: A) Reference plan, evaluated on follow-up delineations. B) Adapted plans, evaluated on the reference delineation. C) Adapted plans, evaluated on follow-up delineations. The three groups allow for an estimation of the dosimetric variations due to anatomical changes that are currently clinically accepted (A), plan adaptation induced changes (B), and compound changes (C). Scenarios B and C were used to study correlations and the impact of DVH- based thresholds on the balance of plan rejection rate and quality. Results Scenario B deviates the least from the reference plan, resulting in a lower apparent dose to OARs on average (fig. 1). Scenario A shows a large spread in doses, especially the bladder, and the variation is comparable in size to that of scenario C. CTV coverage was always maintained (not shown). The dosimetric difference between SSO+SWO and SWO was small. The mean plan adaptation times were 7 min (SWO+SSO) and 3 min (SWO). In fig. 2 we show as a function of dosimetric thresholds, the plan rejection rate for adapted plans (left), and which fraction of plans under scenario C have the same reject/accept status as in scenario B (right). Conclusion For the first time, dosimetric variations induced by daily online plan adaptation on the MR-Linac were investigated in the context of uncorrected variations induced by anatomical changes. Dosimetric tolerances for online plan adaptation should be tailored to the dosimetric variations induced by uncorrected variability. Anatomical dosimetric variance is likely to decrease once delineations based on daily MRI images are used for plan adaptation.

Electronic Poster: Physics track: CT Imaging for treatment preparation

EP-2075 Evaluation of image fusion for computed tomography of head and neck patients with dental implants C.H. LEE 1 , C. Kim 2 , J. Kwak 1 , H. Song 1 , G. Back 1 1 ASAN MEDICAL CENTER, radiation oncology, seoul, Korea Republic of 2 KAIST, KAIST, Daejeon, Korea Republic of Purpose or Objective Computed tomography (CT) is useful for planning radiation treatment of patients with head and neck cancer. However, dental implants cause metal artifacts that degrade CT image quality. Many methods for reducing metal artifacts have been proposed. Here we use a gantry-tilted scan followed by image fusion for this purpose. This study aimed to reduce metal artifacts in CT and improve image quality.

Material and Methods We obtained two computed tomographic images, one with a superior 20° gantry tilt, and one with a normal setup (0° tilt) from each of ten patients with dental implants who had head and neck cancer. The gantry tilt moved the metal artifacts to a superior position. We then reconstructed the gantry-tilted images as normal-scan images. In a second step, image fusion, we combined the normal image excepting the parts containing the metal artifacts with the reconstructed, gantry-tilt image that had a reduced metal artifact. The fusion process was confined to the vicinity of the dental implants. The result was a third image, the fusion image. We selected regions of interest (ROIs) in which to compare the normal-scan image with the fusion image. Then we evaluated the CT number in Hounsfield units (HU) in each of these ROIs to allow quantitative comparison.