ESTRO 2021 Abstract Book

S1480

ESTRO 2021

Conclusion The addition of a boost to prostate SBRT treatment plans can be achieved with adequate target coverage. However, this can lead to exceeding OAR tolerances for some patients. The dosimetric impact of intrafraction motion was minimal based on the RayPilot data. Intrafraction motion using CBCT data from the same treatment dataset showed a greater impact on plan dosimetry, but these results may have been influenced by user bias or inherent error due to voxel size in the images. Further investigation should be carried out before implementing this treatment technique clinically. References : [1]Trainer et al, PO-1598: Investigating the use of RayPilot for motion management during prostate SBRT: initial experience, Radiother Oncol , Vol 152, Sup 1 Nov 2020 p S869-S870 [2]Trainer et al, Analysis of the Intra-Fractional Motion of the Prostate During SBRT Using an EM Transmitter, Int. J. Radiat. Oncol. Biol. Phys. 2020;108(3):e340 PO-1755 What is the best reference image for IGRT using 4D CBCT? M. van Herk 1 , A. Bryce-Atkinson 1 , J. Lindsay 1 , C. Faivre-Finn 1 , C. Eccles 2 1 The University of Manchester, Radiotherapy Related Research, Manchester, United Kingdom; 2 The Christie Hospital NHS Foundation Trust, Radiotherapy, Manchester, United Kingdom Purpose or Objective 4D CBCT for IGRT has been available for over a decade, with IGRT using 4D MRI coming soon. 4D IGRT is known to be of most benefit for tumours with amplitude over 1cm. Typically, each frame of a 4D localisation scan is registered to a 3D reference scan and the resulting registrations combined to derive a setup correction. There is, however, an issue on how to derive this 3D reference scan from 4D CT, and different methods are used based on software availability and workflows. The purpose of this work is to compare the accuracy of 4D CBCT patient setup corrections using a reference image generated with several common methods to create a 3D scan from 4D CT and derive practical recommendations for clinical workflows. Materials and Methods 10 4D CBCT scans of 5 lung SABR patients with substantial tumour motion were collected (4D CT range 1.1 - 1.9 cm, SI direction). Tumour length in SI direction ranged from 0.5-3 cm. 4D CBCT was acquired and processed on the Elekta XVI system (V5.03). Scans were acquired with 2-4 minutes scan time, ~650-1300 projection images, 120 kV, 0.32 mAs per image. Average, MIP, Mid-ventilation (~30%) and Mid-position reference scans were generated from 4D CT using in-house software. A dual registration workflow was used: CBCT is registered on selected vertebral bodies first; next a 4D soft tissue registration is performed, using ITV+5mm as region of interest, edited to remove anatomy outside the lung. Manual registration was not performed. To evaluate consistency, 4D registration was repeated 100 times, adding random shifts (SD=0.3cm in x, y and z) after bone registration to simulate variable baseline shifts. Derived table shifts were compared. Failure rate, defined as number of residual errors> 0.5cm, is also reported. Results Automatic 4D registration was reproducible (table shift SD < 0.1cm) and reliable (0-14% failure rate) for most reference images (Table 2, failures excluded in measurements). Failures occurred mostly for MIP when the tumour was close the diaphragm (85% registrations failed), or when the baseline shift exceeded the tumour size (all images). Systematic offsets of up to 0.3cm were found compared to the mid-position image - for the mid-ventilation due to hysteresis, artifacts and phase selection; for MIP and AVG due to shape differences. Visually, using a MIP reference was the most challenging (Fig. 1D).