ESTRO 38 Abstract book

S556 ESTRO 38

the lowest imaging and dosimetric uncertainties. The DLM appears particularly attractive due to its accuracy and the very fast calculation time (<1 min). PO-1008 Image quality characterisation of a proton gantry-mounted cone-beam CT system K. Andersson 1 , C. Vallhagen Dahlgren 1 , A. Dasu 1 1 The Skandion Clinic, Department of Medical Physics, Uppsala, Sweden Purpose or Objective Cone-beam computed tomography (CBCT) is an import tool for precise patient positioning and tracking of anatomical changes during the course of radiotherapy. For proton therapy, it is crucial to reduce such uncertainties due to the sharp distal dose fall-off. Proton gantry- mounted CBCT generally differ from linac systems, e.g., the source-to-isocenter distance (SID) may be larger and they may lack bowtie filters. Organ doses for a proton- gantry CBCT have recently been investigated by Ardenfors et al [1]. They concluded that posterior scans result in lower organ doses than anterior scans and should therefore be preferred if image quality could be preserved. The aim of the current project was to evaluate image quality for several scan protocols of the same proton gantry-mounted CBCT system. Material and Methods A Catphan® 600 phantom (The Phantom Laboratory, Salem, NY) was scanned with the CBCT system of a proton pencil-beam scanning gantry (Ion Beam Applications, Belgium) at the Skandion Clinic, Sweden. The system has SID=261.8 cm and no bowtie filter applied. Images were acquired for the head, thorax and pelvic protocols with rotations of 360°, 190° anterior or 190° posterior. Image quality was assessed several times over several months. Images were analyzed using ImageJ software (NIH, Bethesda, MD) using the image quality parameters recommended by the EFOMP-ESTRO-IAEA protocol [2]. The tolerance levels used for geometrical accuracy, CT number accuracy and uniformity were the same as during system acceptance. Contrast-to-noise ratios (CNR) and resolution were also evaluated. Results Geometrical accuracy was within tolerance for every protocol, while uniformity exceeded the tolerance for all series (see Table 1). CT number accuracy was generally within tolerance, but in some cases higher deviations were seen for a few contrast targets. CNR:s were considerably low for head scans, and were highest for pelvis protocols. Resolution was considered to be acceptable for all scans. Generally no relevant image quality differences between posterior and anterior scans were found.

bulk density method (BDM) for prostate MRI-only radiotherapy. Material and Methods Thirty-nine patients received VMAT for prostate cancer (78 Gy in 39 fractions). T2-weighted MR images were acquired in addition to the planning CT images. pCT were generated from MRI by four methods: a DLM, a PBM, an ABM and a BDM (water-air-bone density assignment). The DLM was a generative adversarial network (GAN) using a perceptual loss. The PBM was performed with feature extraction and approximate nearest neighbour search. DLM and PBM were trained with a cohort of 25 patients. The four methods were compared in a validation cohort of 14 patients. Imaging endpoints were mean absolute error (MAE) and mean error (ME) of Hounsfield units (HU) from voxel-wise comparisons between pCT and reference CT. Dose uncertainties of the methods were defined as the absolute mean differences between DVH parameters for the organs at risk and PTV calculated from the reference CT and from the pCTs for each method. 3D gamma index analyses (local, 1%/1mm) were also performed. The Wilcoxon test was used to compare the uncertainty of the DLM to those of the three other methods. Results In the whole pelvis, the DLM showed significantly lower MAE (mean value of 37 HU) compared to the PBM (41 HU), ABM (43 HU and) and BDM (99 HU). The ME obtained from the PBM (-1 HU) was lower compared to those of the DLM (-9 HU), ABM (-8 HU) and BDM (-18 HU). The table shows the dose uncertainty of each pCT generation method for each volume-of-interest. The figure shows the dose uncertainty of each method along the whole DVH for the rectum. Significant differences are displayed with the use of the symbol *. DVH differences were significantly lower when using the DLM and the PBM, than the ABM and BDM. All the mean gamma values were significantly lower with the DLM or the PBM, compared to the ABM and BDM.

Conclusion While geometrical accuracy was found to be acceptable, tests of CT number accuracy, uniformity and low resolution showed deviations. Therefore image quality based on these aspects needs to be further investigated. Posterior scans resulted in equal image quality as anterior scans, and could hence preferably be used since organ doses are reduced. The stability of the system will be assessed based on measurements repeated periodically for a longer time period.

Conclusion In order to generate pCT from MRI for dose calculation, the four assessed methods provide clinically acceptable uncertainties (<1%). The DLM and PBM provide however