Quantitative stability of CT numbers in novel cone-beam CT: An intra-institution cross-scanner and longitudinal...

Novel cone-beam computed tomography (CBCT) has become available on standard C-arm linacs, making it widely applicable in photon-based radiotherapy.

To evaluate CT number stability of novel CBCT systems installed on seven C-arm linacs within a single institution, across CBCT systems and over time, and to assess the CBCT-based dose calculation accuracy in head-and-neck, lung, and pelvic cancer patients.

A Gammex Advanced Electron Density phantom (Sun Nuclear) was scanned at installation of eleven CBCT imaging panels at seven TrueBeam C-arm linacs (Varian Medical Systems, A Siemens Healthineers Company), and after three and six months. Three CBCT protocols were evaluated, a head protocol (tube voltage of 100 kVp, iterative reconstruction, denoted iCBCT), a thorax protocol (tube voltage of 125 kVp; filtered back projection with Feldkamp-Davis-Kress (FDK) algorithm), and a pelvis protocol (tube voltage of 125 kVp; iCBCT). Fifteen tissue-equivalent phantom inserts were scanned individually, placed centrally in the phantom. The mean CT number was extracted, and CT number stability was assessed across imaging panels and timepoints. CT number stability was compared to a previous CBCT detector model. Moreover, routine measurements using four inserts were performed until 18 months after installation. Conversion curves for mass density estimation were generated, following a consensus guide. The spread of the curves due to CT number variation across imaging panels was assessed. Treatment plans created on planning CT (pCT) scans of 10 head-and-neck, 22 lung, and 15 pelvic cancer patients were recalculated on CBCT scans, selected to match the anatomy seen on the pCT. Dose-volume parameters for targets and organs-at-risk were compared between pCT and CBCT.

The CT numbers were consistent across the imaging panels and over time for iCBCT reconstruction, but to a lesser degree for FDK reconstruction. The CT number variation increased with the density of the phantom inserts. Still, for the high-density bone insert the range/interquartile range of the CT numbers across imaging panels were 51/29 HU (head protocol), 140/41 HU (thorax protocol), and 27/16 HU (pelvis protocol). For the longitudinal measurements, the median differences were 4 HU, -9 HU, and -8 HU after three months and -8 HU, -32 HU, and -26 HU after six months, for the head, thorax, and pelvis protocol, respectively. For the routine measurements, no clear time dependence was seen for the CT number differences. It was found that a single conversion curve per CBCT protocol could be used for all imaging panels. For iCBCT, median dose differences between pCT and CBCT were within 0.5% for head-and-neck and pelvis and 1.5% for lung, while FDK lung was within 2.5%. The largest deviations with iCBCT were -1.3% for head-and-neck, -1.1% for pelvis, and -3.8% for thorax, but -18.0% for thorax FDK.

The CT number stability was sufficient to allow for a single conversion curve per CBCT protocol to be applied across all CBCT imaging panels. A high dose calculation accuracy was found for 36 patients with iCBCT scans, while larger deviations were seen for eleven thorax FDK scans.