ESTRO 38 Abstract book

S217 ESTRO 38

OC-0417 Impact of CT myelogram vs. MR imaging on spinal cord delineation in spine stereotactic radiosurgery C.J. Jin 1 , N. Tyagi 2 , E. Lis 3 , M. Patel 1 , J. Haseltine 1 , P. McLaughlin 4 , X. Cai 1 , X. Huang 1 , Y. Yamada 1 1 Memorial Sloan Kettering Cancer Center, Radiation Oncology, New York, USA; 2 Memorial Sloan Kettering Cancer Center, Medical Physics, New York, USA; 3 Memorial Sloan Kettering Cancer Center, Neuroradiology, New York, USA ; 4 Dana-Farber Cancer Institute- Harvard Medical School, Radiation Oncology, Boston, USA Purpose or Objective Accurate identification of spinal cord and cauda equina is paramount for spine stereotactic radiosurgery (SRS). This study investigates the inter-observer variability in contouring spinal cord (SC), cauda equina (CE) and thecal sac (TS) using CT myelogram vs. MRI T2 sequences. Material and Methods 7 radiation oncology (RO) observers (3 ROs and 4 RO residents) and 1 diagnostic neuroradiologist (DR) delineated SC, CE, and TS in the regions of interest for 6 spine SRS cases. Cases included 3 with CT myelogram pre- simulation and 3 with diagnostic MR spine fusion. Cases were evenly distributed between the junctional (C7-T1), mobile (L2), and semirigid (T3-T4) spine. Inter-observer variability was quantified using volumetric coefficient of variation (VCV), mean Hausdorff distance (HD), and the Dice coefficient (DSC). Segmentation variability in relation to imaging modality and spinal region was analyzed using the Wilcoxon rank sum test. Results In total, 96 contours were analyzed within the CT myelogram and MRI datasets. Across the 6 cases, no significant difference was found in mean VCV for CT myelogram vs MRI (10.7% vs 12.3% P= .675). The mean HD between RO and DR contours was significantly lower for CT vs MRI (0.32 vs 0.46 mm P= .008). The mean DSC agreement between RO and DR contours was 0.89 ± 0.04 for CT and 0.89 ± 0.11 for MRI (P= .063). Substantial differences in contour variability were seen in different spine locations. The mean VCV for SC/CE contours based on MRI varied across the junctional (21.1%), mobile (25.5%), and semirigid (10.1%) spine, while there was less variability based on CT myelogram (14.0%, 13.4%, and 16.3% respectively). TS volumes were less variable across all spine locations (mean VCV 6.3% TS vs 16.7% SC/CE, P= .036). Mean HD between RO and DR contours was significantly greater in the junctional vs semirigid spine for both SC (0.57 vs 0.19 mm, P= .002) and TS (0.49 vs 0.26 mm, P= .005), and this difference persisted across imaging modalities (CT: 0.37 vs 0.21 mm, P= .002; MR: 0.68 vs 0.24 mm, P= .003). Similarly, the DSC agreement was significantly lower in the junctional vs semirigid spine for SC (0.78 vs 0.92, P= .003) and TS (0.89 vs 0.93, P= .001), for both imaging modalities (CT: 0.88 vs 0.92, P= .001; MR: 0.79 vs 0.93, P= .003). Compared to RO volumes, DR volumes were larger for SC/CE (P= .036) and TS (P= .036), particularly for the MRI dataset (P = .036), with a similar trend for the CT myelogram dataset (P = .059). Figure: DSC agreement between RO and DR contours based on CT myelogram and MRI.

Conclusion Inter-observer variability differences in contouring SC, CE, and TS between CT myelogram and MRI were minor. CT myelogram produces slightly better agreement between RO and DR contours, but similar segmentation can be achieved from either imaging modality when there is no registration error. Contouring uncertainties that need to be considered include the greater inter-observer variability in the junctional spine, and the comparatively larger DR volumes. OC-0418 Residual misalignment of supraclavicular lymph nodes for NSCLC patients, to determine GTV- PTV margin J. Stam 1 , S. Gerrets 1 , R. De Haan 1 , J. Belderbos 1 , E. Damen 1 , P. Remeijer 1 1 Netherlands Cancer Institute, Department of Radiation Oncology, Amsterdam, The Netherlands Purpose or Objective For locally advanced non-small cell lung cancer (NSCLC) patients besides the primary tumor, the mediastinal lymph nodes and supraclavicular lymph nodes (SLN) are often involved. In case of SLN involvement a thermoplastic mask is used for immobilization. For the SLN a GTV-PTV margin of 1-1.2 cm in all directions is used in our institute. In this analysis the positioning-errors of the SLN were evaluated to determine the optimal GTV-PTV margin to cover involved SLN adequately. Material and Methods A total 720 CBCT’s of 30 patients with a fractionated treatment dose > 44Gy were analyzed. All patients were positioned on a thorax support and immobilized with a thermoplastic mask. The IMRT treatment plan included the primary tumor, mediastinal lymph nodes and SLN in one isocenter. For each fraction online CBCT registrations were performed using a 3D rectangular shaped region-of- interest (ROI) around the vertebrae (Clipbox 1) and a ROI around the carina (Clipbox 2) followed by an online table shift to correct for the carina misalignment. When the spinal cord dose was critical, the table shift was performed based on the vertebrae registration instead. To measure the residual set up error of the SLN, a grey value registration was performed using an additional ROI around the SLN (Clipbox 3), see figure 1. After the automatic grey