Aspects of simulator cone-beam CT for radiotherapy treatment planning

Background and purpose: Following a recent major upgrade in cone-beam computed tomography (CBCT) software and functionality, we have reassessed aspects of our Varian Acuity simulator performance for use in treatment planning. The feasibility of using CBCT for treatment planning has been assessed and here we report specifically on Hounsfield number (HN) accuracy and related dose errors, and digitally reconstructed radiograph (DRR) image quality. Methods: Using a Catphan (R) 600 CT phantom, HN accuracy and uniformity were investigated for a range of CBCT imaging modes. This included the variation in HNs with scan length and phantom position. Results were compared with those acquired from conventional CT. Treatment plans for three sites were generated using the Rando phantom, and results from CBCT-based data were compared to that from CT-based data using a gamma analysis. Image quality of DRRs based on CBCT data were compared with those from CT data both quantitatively, by calculating the modulation transfer function (MTF) and qualitatively, by counting the number of line pairs visible on a phantom. Results and conclusions: Catphan data showed that for certain cases, the HN calibration of the Acuity CBCT was out of tolerance and could lead to errors in dose calculation of >2%. HNs were only acceptable for scan lengths >10 cm. In multi-scan mode, geometric shifts and differences in HNs were seen on CT slices on either side of the interface between the two acquisitions. However, comparisons between treatment plans calculated using CBCT data and conventional CT data from Rando phantoms showed that head, pelvis and thorax plans were acceptable. CBCT DRR image quality compared favourably with a conventional CT scanner in some respects; however, image uniformity and low contrast resolution were poorer due to the 'cupping' artefact obtained with CBCT scans.