ESTRO 38 Abstract book

S218 ESTRO 38

the SLN was LR: 7 mm, CC: 9 mm and AP: 9 mm. When the table correction was based on the carina registration, the required GTV-PTV margin for the SLN was LR: 8 mm, CC: 12 mm and AP: 9 mm. OC-0419 Evaluation of Metabrain: a semi-automated delineation tool for edema surrounding brain metastasis Y. Pin 1 , D. Antoni 1 , A. Keller 1 , P. Truntzer 1 , J. Clavier 1 , O. Schaeffer 1 , M. Schmitt 1 , W. Waissi 1 , G. Noël 1 1 Centre de Lutte Contre le Cancer Paul Strauss, Radiation Oncology, Strasbourg, France Purpose or Objective The edema surrounding brain metastasis (BM) was the subject of several studies but the measurement methods remain controversial. The edema volume is an objective and relatively reproducible feature but its determination by contouring is time consuming which can limit its use. In this study, we propose and test a semi-automated tool to segment brain edema and measure its volume in patients with BM. Material and Methods We developed Metabrain, a threshold based automated segmentation tool to extract edema surrounding BMs on FLAIR sequence MRI. Developed in Python, the software imports image datasets and BMs GTV contours. Each edema, corresponding to a BM, is segmented and its volume measured based on a windowing (level and width) manually defined by the user. A contouring study was performed to assess the reliability and similarity of Metabrain contours. Ten BM edemas were contoured two times by 9 radiation oncologists: manually (MC) and with Metabrain (AC). A Dice index (DI) was calculated in the MC and AC groups, compared to the intersection of all contours in each group. The DI was also calculated between each manual and automated contour of each edema for each radiation oncologist. The volume of each edema and the total time for manual and automated contouring were recorded. DI, volumes and contouring time were compared using a paired Wilcoxon signed rank test. Contouring time and DI were compared between experimented physicians and residents. Results Ninety manual and semi-automated edemas were delineated. One edema on ten was removed because it was in contact with the skull and lead to an aberrant segmentation. In the AC group, the DI was significantly better than in the MC group (median DI of 0.95 vs 0.81, p < .001 ) and the contouring time significantly shorter (18 vs 40 minutes, p = 0.01 ). The median volume was significantly lower in the AC group than in the MC group (respectively a median volume of 1.89 and 2.16 mL, p < .001 ). The median DI between manual and Metabrain contours from a same radiation-oncologist was 0.80. No significant difference was found in DI and contouring time between experimented physicians and residents. Conclusion Metabrain is an efficient tool to measure the volume of edema surrounding brain metastasis. It performs contours more similar, accurate and time saving than manual contouring. Its use would make it easier to include the volume of cerebral edema in BM studies. OC-0420 How accurate is automatic determination of the Mid-Ventilation position and tumour motion? A. Van Nunen 1 , D. Schuring 1 1 Catharina hospital, Radiotherapy, Eindhoven, The Netherlands Purpose or Objective Different approaches can be applied to determine target position and PTV margins for lung cancer patients. In our hospital the Mid-Ventilation (MidV) approach is used. For quantifying the MidV position (MidV-P) and tumour motion,

value registration of Clipbox 3, the position of the SLN was visually checked. If necessary, the registration was manually adapted to make sure the position of the SLN was correct. Next, the residual errors of the SLN lymph nodes relative to the carina and vertebrae registration were analyzed. For both residuals the grand mean (M), systematic erros (Σ) and random errors (σ) were calculated. From these results the required GTV to PTV margins were derived using the margin recipe form van Herk et. al., taking delineation uncertainty and intra-fraction motion [1] in account as well.

References 1. A. Licup, Data mining in RT: Intrafraction motion and treatment time analysis for SBRT lung cancer patients (ESTRO37 2018) Results For all CBCT’s a successful SLN registration was performed, automatically or either after manual adaptation. The mean, systematic and random error (translations and rotations) for the SLN in relation to the correction based on the vertebrae or carina are summarized respectively in table 1. A and B.

Conclusion Positioning errors of the SLN lymph nodes were analyzed for locally advanced NSCLC patients using a thermoplastic mask. Based on the found results the initial GTV-PTV margin of 1-1.2 cm in our institute could be reduced as following: in case the table correction was based on the vertebrae registration, the required GTV-PTV margin for