ESTRO 38 Abstract book

S613 ESTRO 38

Conclusion Multi‐atlas compared to single‐atlas achieved significantly better segmentation accuracy and time efficiency for H&N OARs. Extended atlas did not improve accuracy, however clinically relevant time sparing was observed. PO-1103 Introducing contrast-delayed magnetic resonance imaging in radiosurgery treatment of glioblastoma. F. Padelli 1 , E. De Martin 2 , Y. Mardor 3 , D. Last 3 , V. Pinzi 4 , M.G. Bruzzone 1 , L. Fariselli 4 , V. Cuccarini 1 , D. Aquino 1 1 Fondazione IRCCS Istituto Neurologico Carlo Besta, Neuroradiology Unit, Milan, Italy ; 2 Fondazione IRCCS Istituto Neurologico Carlo Besta, Health Department, Milan, Italy ; 3 Sheba Medical Center, Advanced Technology Center, Tel Hashomer, Israel ; 4 Fondazione IRCCS Istituto Neurologico Carlo Besta, Radiotherapy Unit, Milan, Italy Purpose or Objective Stereotactic radiosurgery (SRS) of intracranial lesions is based on the accurate delivery of very high doses of radiation in one or a few fractions to a well‐defined volume, thus effectively sparing adjacent organs at risk (OARs). Milestone of this technique is therefore the acquisition of appropriate imaging for a really optimised treatment planning. Conventional magnetic resonance imaging (MRI) may not be able to differentiate tumor / non‐tumor enhancing tissues. In this study we introduce treatment response assessment maps (TRAMs) [Zach et al. Neuro Oncol. 2015], based on the concept of delayed contrast extravasation MRI, into a radiosurgery treatment planning system (TPS) for target identification purposes. Material and Methods Five patients presenting with disease progression of gliobastoma multiforme (GBM), previously treated according to our clinical practice, were enrolled in this study. For all patients an irradiation theoretical target volume was defined by contouring the enhancement area on a 3D T1‐weighted MRI sequence (1‐mm slice thickness, contiguous slices) acquired after contrast agent intravenous injection. TRAMs were obtained subtracting the post‐contrast 3D T1‐ weighted images from the same MRI sequence acquired about 75 minutes after. Tumor burden was then also identified and outlined on TRAMs images, specifically processed to be imported in the TPS (Figure 1). Maintaining the target coverage maximization as primary objective (prescription dose to 95% of the target) for comparison purposes, plan optimization tests were performed in two ways for each patient: considering only the conventionally delineated target or considering only the TRAMs delineated target. The plans obtained for each patient were compared in terms of target volume and dose volume histogram (DVH) data.

Results For all patients the target volume contoured on the TRAMs images was smaller than the one contoured on conventional MRI, with a fraction of approximately ½ in one case. Non‐overlapping areas were also identified. Consequently, the percentage of healthy brain volume receiving 12 Gy was always in favour of the TRAM target case, with a reduction from 1.5 to 5%. Doses to other OARs adjacent to the target were also reduced. Conclusion The addition of delayed contrast extravasation MRI information in the identification of the radiosurgery treatment target can affect the planning optimisation process in the re‐irradiation scenario of relapsing gliobastoma multiforme. Further investigations on the actual clinical impact of this imaging modality seems appropriate. PO-1104 Implementing an automated target delineation service in multi-institutional environment in Finland J. Heikkilä 1 , H. Virsunen 1 , L. Voutilainen 1 , K. Vuolukka 1 , A. Nevantaus 2 , M. Haatanen 2 , L. Sailas 3 , J. Seppälä 1 1 Kuopio University Hospital, Centre of Oncology, Kuopio, Finland ; 2 Central Finland Central Hospital, Department of Radiotherapy, Jyväskylä, Finland ; 3 North Karelia Central Hospital, Department of Radiotherapy, Joensuu, Finland Purpose or Objective According to many publications there are remarkable variations in radiotherapy (RT) target delineation. These publications conclude that interobserver variability in target delineation is the biggest source of uncertainty in RT process, potentially inducing systematic errors in radiation dose delivery. In some cases, this uncertainty could have an impact on clinical outcome of individual patients. [1] Different tools to reduce variability in target volume delineation has been studied, such as training, guidelines, and autosegmentation methods. In this study we implemented fully automated target delineation service into multi‐institutional environment, which should decrease the interobserver variability. Material and Methods The developed delineation service consists of the following components (Fig 1): pseudononymization gateway in a hospital network, Ensemble integration platform between the hospital and delineation service,