ESTRO 36 Abstract Book

S932 ESTRO 36 _______________________________________________________________________________________________

hence not be placed at the bottom, but in the center of the water container. More than 90 % of the holes within the recognized beam parts are correctly located by the algorithm (Fig. 2). The main reason for detection failures are the connection pins, which are often filled with air bubbles (no MR signal). The detection rate could be improved using prior knowledge of the shape of the beam parts, available from the CAD model.

Electronic Poster: Physics track: Images and analyses

EP-1703 Rapid prototyping phantom using LEGO® for MRI distortion correction in MR guided radiation therapy S. Neppl 1,2 , M. Reiner 1 , M. Peller 3 , C. Belka 1 , K. Parodi 2 , F. Kamp 1 1 LMU Munich, Department of Radiation Oncology, Munich, Germany 2 LMU Munich, Department of Medical Physics, Garching b. München, Germany 3 LMU Munich, Department for Clinical Radiology, Munich, Germany Purpose or Objective An accurate geometry is critical for the use of MR images for dose calculation in MR guided radiation therapy. Therefore, we designed an easily adjustable distortion phantom based on LEGO® technic parts to detect and correct geometrical inaccuracies caused by magnetic field inhomogeneities. Material and Methods The designed phantom consists of LEGO® technic beam parts with 13 holes, rectangular beam parts with 3+4 holes ("L shape" for stability) and pins to connect 2 or 3 beam parts. The holes within the beam parts have a diameter of 4.85 mm and a center distance of 7.99 mm. The phantom has a size of 24x24x22 cm³ and is placed in a container filled with water (Fig. 1). The LEGO® parts do not give a signal and are therefore not visible on the MR image in contrast to their water-filled holes. The MR images were acquired on a SIEMENS Magnetom® Aera with a clinically used T1 weighted MPRAGE (Magnetization Prepared RApid Gradient Echo) sequence with an isotropic voxel size of 1.5 mm, TE = 1.98 ms, TR = 1900 ms, TI = 900 ms and a flip angle of 8°. An automatic hole detection of the Lego parts was developed with the Image Processing Toolbox™ of Matlab® 2016a. First the beam parts are segmented with an adaptive threshold and then a search for circular structures is performed within the beam mask. The centers of the recognized circles are iteratively corrected to the brightest position within the circle. The exact reference hole positions are imported from the CAD model of the phantom.

Figure 2: An exemplary snippet of a phantom MR slice. The blue circles visualize the recognized holes on the MR image and the red dots show the exact positions given by the CAD model. Conclusion A cheap and extremely customizable phantom was designed using LEGO® technic parts to correct MR images for geometric distortion. For fast prototyping, the size of the phantom can be easily adapted. The hole positions can be extracted correctly and with a high detection rate from the MR images. The shift from their nominal position to the detected position can be used to create a distortion map. EP-1704 Breast tumour bed contouring: influence of surgical clips assessed on the same imaging. D. Ciardo 1 , M. Leonardi 1 , A. Morra 1 , G. Fanetti 1 , D. Damaris 1 , S. Dicuonzo 1 , V. Dell'Acqua 1 , R. Ricotti 1 , F. Cattani 2 , R. Cambria 2 , G. Baroni 3 , R. Orecchia 4 , B. Jereczek-Fossa 1 1 European Institute of Oncology, Department of Radiation Oncology, Milan, Italy 2 European Institute of Oncology, Unit of Medical Physics, Milan, Italy 3 Politecnico di Milano, Dipartimento di Elettronica- Informazione e Bioingegneria, Milan, Italy 4 European Institute of Oncology, Department of Medical Imaging and Radiation Science, Milan, Italy Purpose or Objective To evaluate the inter-observer variability in the contouring of tumour bed after lumpectomy in breast- conservative surgery patients. Material and Methods We retrospectively analysed the planning computed tomography (CT) of 47 patients who underwent external radiotherapy (ERT) post-lumpectomy and intra-operative radiotherapy (IORT). Three to 5 surgical clips were positioned to delineate the excision cavity. The CT acquired for ERT planning was modified to obtain a virtual CT scan with hidden clips: in particular, clips were blurred using a fully automated MATLAB script that replaces each voxel of the clip with a value obtained from a normal pseudo-random distribution with mean value and variance

Figure 1: Photo of the designed LEGO® phantom in water. This phantom consists of 270 "beam 13 parts" (grey), 85 "beam 3+4 parts" (black), long pins (blue) and short pins (black). Results A very customizable MRI distortion correction phantom was developed. The beam parts are correctly detected in the inner part of the images. At the edges of the phantom the distinction to the air outside the container and a low contrast leads to undetected holes. The phantom should