ESTRO 2020 Abstract Book

S1062 ESTRO 2020

enables margin reduction(1). However a soft tissue registration (like CTV-registration) is more challenging for RTTs, due to limited image quality of the 4D CBCT and anatomical changes such as tumor regression. The purpose of the observer study is to determine the feasibility of a CTV registration using a shaped region of interest. Material and Methods Six oesophageal cancer patients were included in this retrospective study. A 4D CT scan was made, from which a mid-position (midP) scan was derived for the purpose of treatment planning. The CTV was defined on the midP planning CT (pCT) by expansion of the GTV of 35 mm in the cranial-caudal direction and 5 mm in the transverse plane, and was extended to include all pathological lymph nodes. The CTV expansion into the gastric mucosa could be limited to 20 mm. Six 4D CBCTs were selected per patient, the first three fractions and weekly for the following three weeks. A shaped region of interest was created from the CTV, excluding fiducial markers. Two groups performed CTV registration, consisting of: 1) five specialized imaging RTTs, 2) five RTTs with our standard level of training. The registration was performed using automatic grey value registration (translation only). Manual adjustment of this automatic registration was recorded. To quantify inter-observer variability, first the standard deviation (SD) of the registration results in the Left-Right (LR), Cranial-Caudal (CC) and Anterior-Posterior (AP) was calculated per scan over all the observers. Secondly, the root-mean-square (RMS) of the SD over all fractions and patients was determined. As a reference, inter-observer variation was also calculated for the standard bone registration. Results Inter-observer variability of CTV registration ranged from 0.11 – 0.25 cm. The inter-observer variability of the specialized RTTs was a factor 1.6 – 1.9 lower compared to standardly trained RTTs, the chi-squared test showed a significant difference. The registration using a standard bone registration showed no significant difference, see table 1. In group 1 a total of 56.7% of the automatic registrations were manually adjusted, in group 2 this was the case in 41.1%, see figure 1.

optimized according to patient specific anatomy and adjusted in case of anatomical changes, which could improve the automatic registration.

1.

Voncken et al. Quantification of Esophageal Tumor Motion and Recommendations on Setup Verification Strategy During Image Guided Radiation Therapy.

PO-1904 ROI optimisation for surface guided radiation therapy T. Sauer 1 , D. Popp 1 , R. Fietkau 1 , C. Bert 1 1 Universitätsklinikum Erlangen- Friedrich-Alexander- Universität Erlangen-Nürnberg, Department of Radiation Oncology, Erlangen, Germany Purpose or Objective The goal of this study was to investigate how the choice of the surface ROI affects the registration results of surface imaging for breast cancer patient daily positioning. Further analysis was undertaken as to whether it is possible to find a generally valid optimal ROI that minimises the difference between surface imaging-derived and the clinically applied CBCT-shifts. Material and Methods The AlignRT system (VisionRT, London) and a CBCT (XVI, Elekta, Stockholm) installed on a Versa HD Linac (Elekta) were used in this study, which included 18 patients (13 Mamma-Ca, 5 thorax wall) of which 111 fractions were analysed. In the clinical workflow, the patients are pre- positioned with the AlignRT system and then shifted in 6DOF according to the CBCT. A new reference capture is taken immediately afterwards. For this study, an additional treatment capture was taken just before the CBCT. Retrospectively, this surface was registered to the reference capture with an offline tool (provided by VisionRT) that is based on the same algorithm as the AlignRT system. By varying the ROI, which determines the registered area of the surfaces, the registration results were optimised by minimising the deviation to the CBCT results. The ROIs consisted of two sets of surfaces: one obtained by applying a variable margin to the contoured breast, the other by combining manually contoured ROIs of various anatomical and geometrical structures including the sternum, contralateral breast, axilla etc. (see figure 1).

Results We observed a slight improvement in the registration results for increasing margin size in the first set of ROIs. For a margin of 5cm (compared to no margin) the median magnitude of the translational deviation (D transl ) decreased by 10% and the rotational deviation (D rot ) by 26%. For the second set, some combinations which led to significantly better results could be identified. Comparing the best to the worst performing ROI, D transl and D rot could be improved by 14% and 48%, respectively (see figure 2). A ROI comprising the breast surface and a belt caudal to the breasts (‘MaBE’ in figure 1) seemed to be the best compromise between registration result and frame rate, which is inversely proportional to the size of the ROI.

Conclusion Introducing a CTV registration using a shaped region of interest is possible but is associated with a larger inter- observer variation compared to a bone registration, which may have to be included in the margins. More training and experience with CTV registration is expected to reduce the inter-observer variation. Furthermore, the shaped region of interest could be