Abstract Book

ESTRO 37

S519

Purpose or Objective In modern-day radiotherapy, cone-beam imaging protocols are often used to position esophageal cancer patients prior to treatment. Due to the limited soft-tissue contrast on the CBCT images, these patients are often aligned to the vertebrae close to the target volume. To account for residual positioning errors, wide ranging CTV- to-PTV margins are currently applied. The aim of this work is to assess the residual interfraction tumor motion based on longitudinal series of MR images in a cohort of 15 esophageal cancer patients. Material and Methods As part of an ongoing imaging study in our institute, 15 esophageal cancer patients underwent six T2-weighted MRI scans (1 prior to treatment and 5 during neo- adjuvant chemoradiation (23x1.8 Gy) at weekly intervals). The tumor was carefully delineated at all scans. Follow-up scans were registered to the first (reference) scan with the Elastix toolbox (Klein 2010). The registration consisted of a two-step process: First, a rigid translation based on bony anatomy in the proximity of the tumor was performed to mimic an online CBCT setup strategy. Secondly, a non-rigid B-Spline registration was performed where tumor masks and grey values were simultaneously registered using both a kappa statistic and a mutual information metric, respectively (Fig. 1). The resulting deformation vector field within the tumor was analysed and used as a surrogate for the local residual tumor motion over the entire course of treatment. Based on population statistics of the systematic (Σ) and random (σ) components of these residual errors, CTV-to-PTV margins were calculated using the van-Herk-recipe (van Herk 2000) for an online bony anatomy based CBCT setup strategy.

Results The table summarizes results obtained for the five patients including the motion amplitude, the RMS error when using the rigid or the piece-wise rigid method, and the error as a percentage of the motion. There is significant improvement in results when using the piece- wise approach for four of the patients where the final error is equal or smaller than 10% of the motion amplitude. In all cases, the final RMS registration error is <1.6 mm. The figure shows the displacement results of both methods compared to the annotated tumor motion for a representative patient, which clearly illustrates underestimated displacement by the rigid registration. The figure also shows two x-rays and DRRs at different time points of the breathing cycle showing how the tumor and diaphragm can move relative to the ribs.

Conclusion A piece-wise rigid 2D/3D registration method, where the ribs are rendered separately from the soft-tissue, improved registration accuracy for the investigated patient data set. This approach is a potential alternative to dual-energy fluoroscopy as it can be used with standard treatment machines.

Poster: Physics track: Inter-fraction motion management (excl. adaptive radiotherapy)

PO-0951 What CTV-to-PTV margins are required for esophageal cancer radiotherapy? M. Boekhoff 1 , A. Kotte 1 , S. Mook 1 , A. Borggreve 1 , L. Goense 1 , S. Heethuis 1 , P. Van Rossum 1 , A. Van Lier 1 , J. Lagendijk 1 , G. Meijer 1 1 UMC Utrecht, Radiotherapy, Utrecht, The Netherlands

Results The median 3D tumor motion with respect to the vertebrae was 6mm (Fig. 2). This interfraction motion was most pronounced in the cranial-caudal direction for both the systemic and random components (3.3 mm and 4.7 mm respectively) (Table 1). Especially, in tumors