ESTRO 38 Abstract book

S1132 ESTRO 38

Conclusion The first steps in this feasibility study, implemented to avoid the black box nature of DL, allowed to render useful and sufficient training and validation data, and to construct an accurate NN with optimized hyper- parameters. In a next step, the pixel intensity in the projection image corresponding to each ray will be used as output target value, to include the non-linearities into the model and to render realistic DRRs. EP-2057 Influence of implanted metals in new CT reconstruction algorithm for radiotherapy treatment planning T. Kamomae 1 , T. Nakaya 2 , F. Kawabata 2 , K. Okudaira 2 , M. Kumagai 1 , H. Oguchi 3 , Y. Itoh 1 , S. Naganawa 1 1 Nagoya University Graduate School of Medicine, Department of Radiology, Aichi, Japan ; 2 Nagoya University Hospital, Department of Radiological Technology, Aichi, Japan ; 3 Nagoya University Graduate School of Medicine, Department of Radiological and Medical Laboratory Sciences, Aichi, Japan Purpose or Objective Computed tomography (CT) imaging is used to delineate organs and calculate the dose distributions for treatment planning. To calculate the dose distributions, the CT numbers are converted to relative electron density (RED) or mass density by using the conversion curve that depends on the X-ray tube voltage. A recently developed algorithm directly reconstructs the CT images constructed from the RED values, which are independent of the tube voltage. This algorithm offers the freedom to choose the tube voltage to obtain higher contrast images. We have recently clarified the feasibility of using this algorithm for accurate dose calculations in the low- to middle-density regions correspond to the human body tissues. In this work, we evaluate this algorithm in the regions containing high-density implanted metals, which lead to the metal artifacts. Material and Methods Gammex phantom with various rods inserted was scanned to evaluate the CT number to RED conversion curves. These rods contained titanium and steel. All raw data were reconstructed with the standard filtered back- projection (FBP) and the novel DirectDensit TM (DD) algorithms, and additionally, the metal artifact reduction (MAR) algorithm was applied. Furthermore, the CT images of the experimental oral phantom that contained a mock metal crown and tooth were obtained with various tube voltages to evaluate the impact of metal artifacts and dose perturbations. Results For the rods correspond to the human body tissues (i.e., lungs, soft tissues, and bones), the mean differences in pixel values between the present and absent metal rods were − 6 ± 31, 2 ± 8, − 2 ± 25, and 3 ± 6 for 120 kV-FBP, 120 kV-FBP with MAR, 120 kV-DD, and 120 kV-DD with MAR, respectively. The maximum pixel value became saturated at about 2500 at the region of titanium and steel for DD images. The artifact indices at 10 mm apart from the metal in the experimental oral phantom were 464, 23, 318, and 8 for 120 kV-FBP, 120 kV-FBP with MAR, 120 kV- DD, and 120 kV-DD with MAR, respectively. For varied tube voltages in DD with MAR images, the artifact indices were 25, 23, 0, 8, and 13 for 70 kV, 80 kV, 100 kV, 120 kV, and 140 kV, respectively. The calculated dose differences between the 120 kV-FBP with MAR and DD with MAR images were in general less than 1% when the density override corrections for the remaining metal artifacts were performed.

Conclusion The metal artifacts and dose calculation results of the DD images were the approximation of standard FBP images, and the MAR was also beneficial in the DD algorithm. The restriction of the maximum pixel value in the DD image and the remaining metal artifact that tended to appear in the lower tube voltage need to be considered carefully. The DD algorithm can be used for treatment planning in the regions containing the high-density implanted metals, and lead to the benefit of an optimization of the tube voltage individually in the treatment planning process. EP-2058 Measuring eye deformation between planning and proton beam therapy position M. Jaarsma-Coes 1 , M.S. Schuurmans 2 , M.K.M.A. Hassan 2 , E. Astreinidou 3 , M. Marinkovic 4 , F.P. Peters 3 , J.W.M. Beenakker 1 1 LUMC, Radiology & Ophthalmology, Leiden, The Netherlands ; 2 LUMC, Radiology, Leiden, The Netherlands ; 3 LUMC, Radiotherapy, Leiden, The Netherlands ; 4 LUMC, Ophthalmology, Leiden, The Netherlands Purpose or Objective Uveal melanoma (UM) is the most common primary malignant intraocular tumour. Proton beam irradiation (PBT) is often the therapy of choice for large tumours. MRI is essential not only for diagnosis but also for the clinical target volume (CTV) definition (shape and extension of the tumour) and PBT treatment planning. However, while PBT is performed with the patient in sitting position, the acquisition of the MRI images are done with the patient in supine position. This change in gravity direction potentially changes the shape of the eye and tumour. As even small geometrical deformation can lead to over- or underdose to the tumour and surrounding healthy tissues, we used MRI to investigate and quantify the effect of different patient positions on the shape of the eye. Material and Methods Seven volunteers and one UM patient were scanned with closed eyes in two positions on a 3T Philips MRI scanner with a 47mm surface coil after giving informed consent. One set of images was acquired in supine position, while a second set was acquired mimicking the patient sitting for PBT (flexed position, figure 1). Additionally, two volunteers were scanned twice in the supine position to assess the reproducibility of the segmentation method.