ESTRO 38 Abstract book

S1127 ESTRO 38

reconstruction seems to be more accurate, but it has the disadvantage that the attenuation map must be removed from the system after every RT measurement to prevent an incorrect reconstruction of patient data that were not scanned in RT setup. Alternatively, offline reconstruction can be implemented at any PC via e7tools, and the reconstruction could be automated, thereby diminishing human error. The cause of the systematic signal difference in online and offline reconstruction needs to be investigated further. [1] Carney et al. doi: 10.1118/1.2174132 Table 1: Comparison of measured PET activities for online and offline reconstruction of the RT and the reference scans.

L. TaeuberT 1,2,3 , A. Pfaffenberger 1,3 , Y. Berker 4 , B. Beuthien-Baumann 5,6 , A.L. Hoffmann 7,8,9 , E. Troost 6,7,8,9,10 , S.A. Koerber 3,6,11 , M. Kachelrieß 4 , C. Gillmann 1,3 1 German Cancer Research Center, Division of Medical Physics in Radiation Oncology, Heidelberg, Germany ; 2 University of Heidelberg, Faculty for Physics and Astronomy, Heidelberg, Germany ; 3 Heidelberg Institute for Radiation Oncology, National Center for Radiation Research in Oncology, Heidelberg, Germany ; 4 German Cancer Research Center, Division of X-Ray Imaging and Computed Tomography, Heidelberg, Germany ; 5 German Cancer Research Center, Division of Radiology, Heidelberg, Germany ; 6 National Center for Tumor Diseases, Partner Site Dresden, Dresden, Germany ; 7 Helmholtz-Zentrum Dresden-Rossendorf, Institute of Radiooncology – OncoRay, Dresden, Germany ; 8 Technische Universität Dresden, Faculty of Medicine and University Hospital Carl Gustav Carus Dresden- Department for Radiotherapy and Radiation Oncology, Dresden, Germany ; 9 OncoRay, National Center for Radiation Research in Oncology, Dresden, Germany ; 10 German Cancer Consortium, Partner Site Dresden, Dresden, Germany ; 11 Heidelberg University Hospital, Department of Radiation Oncology, Heidelberg, Germany Purpose or Objective Integrating PET/MR hybrid imaging into radiation treatment (RT) planning has great potential to improve tumor delineation and dose prescription. Since these scans must be acquired under treatment conditions, attenuation correction of RT positioning devices is necessary. Attenuation maps can be implemented either online (directly at the PET/MRI scanner) or offline (at another PC). In this study, we compare both methods and assess their impact on PET image quality using a CT-based user- generated attenuation map of an RT table overlay. Material and Methods CT images of an RT table overlay (in-house construction) were acquired on a stand-alone CT (Somatom Definition Flash, Siemens Healthineers, Erlangen, Germany) at 120 kV and 360 mAs. Based on the CT images, an attenuation map of the RT table overlay was calculated via the bilinear approach [1]. The RT table overlay was then mounted onto the patient table of the PET/MRI (Biograph mMR, Siemens Healthineers) and two sets of PET-measurements were taken using an active 68 Ge phantom (32 MBq, 10 min scan time). The phantom was scanned with the RT table overlay (RT scan), and without the RT table overlay (reference scan). PET reconstructions of the phantom scans were performed online at the PET/MRI scanner and offline using the e7tools (Version VA20, Siemens Healthineers) with identical reconstruction parameters. For the PET- reconstructions of the RT scan, the attenuation map of the RT table overlay was implemented. Attenuation correction accuracy was evaluated by comparing PET activities between RT and reference scans in 10 ROIs placed every 10 slices along the phantom in longitudinal direction, both for the online and the offline reconstruction methods. Results The RT table overlay attenuation map was successfully added to the hardware attenuation maps produced online and offline. Table 1 compares measured PET activities. For the online reconstruction, a mean percentage difference of 0.7% was found between the reference and the RT scan. For the offline reconstruction, a mean percentage difference of 1.4% was found. A systematic difference of around 500 Bq/ml was found between the online and offline reconstructions. Conclusion For the integration of PET/MRI in RT planning, attenuation correction of RT positioning devices is viable. The online

EP-2051 A comparative analysis of MR signal normalization methods during proton therapy treatment G. Buizza 1 , C. Paganelli 1 , G. Fontana 2 , A. Franconeri 3 , M.V. Raciti 3 , A. Pella 2 , L. Anemoni 4 , A. Iannalfi 5 , L. Preda 3 , F. Valvo 5 , G. Baroni 1 1 Politecnico di Milano, DEIB - Department of Electronics Information and Bioengineering, Milano, Italy ; 2 National Centre of Oncological Hadrontherapy CNAO, Clinical Bioengineering Unit, Pavia, Italy ; 3 National Centre of Oncological Hadrontherapy CNAO, Diagnostic Imaging Unit, Pavia, Italy ; 4 National Centre of Oncological Hadrontherapy CNAO, Medical Radiology Technicians Unit, Pavia, Italy ; 5 National Centre of Oncological Hadrontherapy CNAO, Radiation Oncology Unit, Pavia, Italy Purpose or Objective To compare typical normalization strategies applied to T1- and T2-weighted MR images in patients affected by meningioma for evaluating the MR signal intensity (SI) during proton therapy (PT). Material and Methods 55 exams consisting of contrast-free T1w (VIBE, 3D GRE, TR=5.16 ms, TE=2.09 ms, flip angle=9°, resolution=1.02x1.02x1.02 mm) and T2w (TSE, 2D SE, TR=11.9-14.5 s; TE=104 ms, flip angle=80°, resolution=0.47x0.47x3 mm) MRI were acquired on a 3T scanner during PT for 17 patients. Gross tumour volumes (GTV) were manually delineated on T2w images, whereas white matter (WM) contours were obtained by an automatic algorithm, post-processed and visually checked. T1w images were rigidly registered on T2w images, to allow their evaluation on the same regions. Three normalization methods based on histogram matching (M1,[1]), WM-driven standardization (M2,[2]) and min-max (M3) were implemented. Two metrics computed in WM areas were considered for the evaluation: the Jeffreys’ divergence between any combination of mode-aligned histograms (JD) and the ratio between interquartile range and median computed on the median SI values (npCV). Medians and percentiles (5 th -95 th ) from WM and GTV regions were also compared. Results For T1w-images, JD (median[IQR]) was 0.029[0.023], 0.035[0.041] and 0.056[0.074] for M1, M2 and M3, with respect to 0.045[0.045] derived from raw ones. According to JD and npCV (Tab. 1), M1 was the best-performing