ESTRO 38 Abstract book

S1117 ESTRO 38

0.68 pitch, smooth filter, and gantry rotation time 1.5 sec, respectively. In the phantom model of this study, optimal Contrast Detail Values were determined to be: 1% contrast, 2mm; 0.5% contrast, 4mm; 0.3% contrast, 7mm. Additional measurements were made to confirm the prediction error model is justified and the results are validated. Conclusion Protocol was improved comparing to those using the standard clinic protocol.CT image quality can be improved with the protocol created in this study, to provide better soft tissue contrast, which would be beneficial for RT contorting for SBRT and stereotactic radiosurgery in which accurate delineation of small-sized, low-contrast regions are important. EP-2036 Visualization of prostate fiducial markers using phase-cycled bSSFP MRI Y. Shcherbakova 1 , S. Mandija 1 , L.W. Bartels 1 , L.G.W. Kerkmeijer 2 , J.R.N. Van der Voort van Zyp 2 , C.A.T. Van den Berg 2 1 UMC Utrecht, Center for Image Sciences/Imaging Division, Utrecht, The Netherlands ; 2 UMC Utrecht, Department of Radiotherapy, Utrecht, The Netherlands Purpose or Objective The existing manual techniques for visualization of gold fiducial markers (FMs) are mostly based on spoiled gradient-echo imaging (SPGR). FMs appear as signal voids in magnitude images, which makes it difficult to distinguish them from hemorrhages/calcifications. Being able to distinguish FMs in MR images is crucial for MR-only RT, where CT images are not available. Automatic FMs detection methods are available, but require special software which is not available at the MR console. Here we investigated a new method for distinctive FMs visualization in the prostate for MR-only RT which facilitates FMs detection at the MR console. Material and Methods A dynamic bSSFP scan is performed with RF phase cycling, where for each dynamic acquisition the phase of the RF excitation pulse is increased stepwise according to a certain incrementing scheme. Complex phase-cycled bSSFP images were acquired on a phantom (agar phantom with 4 gold FMs GM1054 implanted) and on 8 patients at 3T: TR 9.2 ms, TE 4.6 ms, FA 20˚, FOV (phantom) 200x200x80mm 3 , FOV (patients) 320x320x60 mm 3 , acquisition voxel size 2x2x2 mm 3 , reconstruction voxel size 1x1x1 mm 3 for both setups, 6 dynamics acquisitions with RF phase increment of , scan time 01:48 min (phantom) and 02:12 min (patients). On each dynamic acquisition the manifestations of the magnetic field distortions around the FMs (the artifacts) are different due to different frequency offset related to the corresponding RF phase. Thus, the contrast from FMs in the images appears different for each dynamic acquisition. CT scans of the phantom and patients were acquired. The distances between the centers of the markers were measured both on MRI and CT images by 1 observer. The accuracy of detecting FMs by was assessed by comparing the measured distances between FMs on MRI and CT images.

In this work, we investigated the performance of a free web-based software to quantitatively analyze this test. Material and Methods A Varian Clinac 2100 CD equipped with the Millennium 120 MLC and the aSi-500 Portal Vision (EPID) was used (Varian Medical Systems, Palo Alto, CA). The EPID was placed with a source-detector-distance of 180 cm (0.4 mm/pixel at the isocenter plane). A tungsten ball (5 mm-diameter) was used as a target in the 2x2 cm 2 MLC-based WL test performed in our department. One hundred fifty portal images (WL images) were retrieved from the Aria system to be analyzed using a web- based application ("Winston-Lutz-Automatic Analyzer", http://winston-lutz.herokuapp.com/), and also with a FDA (U.S. Food and Drug Administration) accredited software (DoseLab Pro v. 6.40, Mobius Medical Systems, LP, Houston, TX). DoseLab was used as reference. Both softwares calculate on each portal image the distance between the centroid of the tungsten ball shadow and the radiation field center ("delta"). Different image processing tools and algorithms are implemented in each software. Delta values given by both softwares were compared, and agreement between both softwares was assessed using the Bland-Altman method. Results An average difference (bias) in "delta" metric of -0.01 mm (SD: 0.13 mm) was found. The 95% limits of agreement between both softwares were from -0.28 mm to 0.25 mm. Conclusion The 95% limits of agreement found were comparable to the pixel size of the portal images analyzed (0.4 mm). Therefore, the results given by Winston-Lutz-Automatic Analyzer are comparable to those reported by DoseLab Pro for WL test analysis. EP-2035 Robust optimization of CT reconstruction and scanning parameters S. Alani 1 , E. Gez 1 , J. Zidan 1 1 ziv medical center, radiation oncology, zefad, Israel Purpose or Objective CT simulation has become an integral component of modern RT planning and therefore needs to be continually optimized. We improved the detection of small and low‐ contrast regions in images obtained during CT simulation by optimization of CT reconstruction and scanning parameters. For potential applications involving detection of low-contrast tumor structures Material and Methods A CT phantom containing a contrast detail modulus for detection of low‐contrast structures was used to optimize the CT protocol. The parameters (A) Pitch, (B) Reconstruction Filter, and (C) Rotation Time type were varied for assessment of image quality. Three factors, three levels, and nine experiments were identified. According to the Taguchi approach an L9 orthogonal array was selected. The reconstruction parameters of the CT scanner Toshiba aquillion Pitch, Reconstruction Filter type, and Rotation Time, were iteratively scanned according to the orthogonal array. A Catphan 604 CT phantom was used to characterize low‐contrast resolution (CPT730 module). All CT scan images were analyzed by IMAGE‐OWL software. The objective of the study was to identify parameters that maximize the low‐contrast resolution of the images. The ANOVA and F‐tests were used to analyze results using JMP 14.1.0 statistical software. Results The optimal settings and predicted optimal values for low‐ contrast resolution were determined. The ANOVA was used to determine the optimum combination of process parameters more accurately by investigating the relative importance of each process parameter. We determined that Pitch (61.3%) had the most significant influence on low contrast resolution, followed by the Reconstruction Filter type (31.3%). The optimal setting level is A1‐B1‐C3,