Abstract Book

ESTRO 37

S524

PO-0957 Dose Delivery Accuracy for Image-Guided Stereotactic Body Radiotherapy for Endometrial Cancer. S. Vieira 1 , M. Possanzini 1 , J. Stroom 1 , C. Greco 1 1 Fundação Champalimaud, Radiotherapy, Lisboa, Portugal Purpose or Objective Standard radiotherapy for intermediate and high risk endometrial Cancer (EC) is a pelvic irradiation followed by a brachytherapy boost. In our department replacement of the brachytherapy boost with image- guided stereotactic body radiotherapy (SBRT) is being investigated. This can be an alternative to deliver a high dose to the vaginal vault while reducing the volume of irradiated normal tissues and improving target coverage. The aim of this study is to assess SBRT boost dose delivery accuracy by calculating the dose in a CT representing treatment delivery anatomy. This is achieved by means of non-rigid deformation of the planning CT based on CBCT images acquired just prior treatment delivery. Material and Methods 10 EC patients were treated with 23 fractions (2Gy/fr) radiotherapy followed by a boost of 3 fractions imaged- guided SBRT. Two different boost dose levels were prescribed: (1) 3x5Gy to PTV1, CTV-PTV margin of 3mm, and (2) 3x7.5Gy to CTV2, without PTV (figure 1a). The boost VMAT treatment was delivered with 4 arcs using 10MV flattening filter free (FFF) beams. Boost treatment set-up comprises of a rectal balloon, a vaginal cylinder and a full bladder. Electromagnetic beacons (Calypso®) were inserted in the vaginal cylinder to track and limit beacon motion to <2mm. Repeated CBCT images were acquired prior to each fraction until a satisfactory set-up was achieved. After treatment, the planning CT was deformed to match final CBCT anatomy using a non-rigid structure guided multi-pass algorithm (Velocity® V3.2.0). Per patient, the CBCT with the most pronounced anatomical changes was used to perform deformation (i.e. worst case). To improve deformation accuracy, balloon and vaginal cylinder had to be delineated on the CBCT. Dose re-calculation (Eclipse TM ) was then performed on the reshaped planning CT using the original plan parameters to study the worst-case dosimetric effect of the deformations. Results Preliminary results reveal sub-millimeter beacon motion for all patients (not shown). The mean deformation vector magnitude for the reshaped CT in the relevant region of interest was 2.4±1.3 mm (SD). In figure 1b the deformation vectors are depicted for one patient. Planned and delivered doses are shown in Table1. Total delivered dose values are derived assuming worst case set-up for all 3 fractions. One patient was excluded due to deformation software problems. Doses of the organs at risk are within the constraints as defined in our treatment protocol. As expected, PTV1 and CTV2 (without PTV) presented less coverage than planned for all patients. The coverage of the CTV1 was adequate given the applied PTV margins.

time. Data was binned into 10 amplitude bins, setting the inclusion range to >95% while minimizing peak-to-peak range (Bones et al., R&O, 2017). For redundant data, the slice with minimum distance to the bin center was selected and missing data was interpolated. The reconstructed 4D-MRI scans were evaluated for sagittal smoothness (RMSE of a fitted parabola to the liver dome, Figure 1) and tumor visibility (G, relative gradient from liver to tumor tissue) in every bin. Results All scans were reconstructed within ~50 on a PC with E3- 1240 V2 CPU. The mean RMSE between the fitted parabola and liver dome over all scans and bins was 1.8±0.3mm (TSE) and 2.6±0.7mm (TFE), with minimum mean RMSE in exhale (TSE: 1.5±0.2mm, TFE: 2.3±0.4mm) and maximum RMSE in inhale (TSE: 1.8±0.1mm, TFE: 3.2±1.8mm). The mean G over all scans and bins was 113±35%/vox (TSE) and 63±24%/vox (TFE). For illustration, the images with the highest and lowest RMSE (Fig1) and G (Fig2) are shown.

Conclusion A 4D-MRI method was developed that allows flexibility in the selection of MRI sequences and provides clinically feasible acquisition-reconstruction times for daily imaging on an MR-Linac. The respiratory signal was extracted from the images and suitable for amplitude binning, resulting in sufficient liver smoothness in the TSE scans (mean RMSE <1voxel). The tumor was visible in the TSE and TFE 4D-MRI scans. The method needs to be validated in more patients. This research was sponsored by Elekta Oncology Systems AB.