Impact of MV treatment beam scatter

In order to quantitatively evaluate the impact of the MV treatment beam scatter on the CBCT image quality, CBCT scans of the Catphan 503 phantom were acquired with four different imaging protocols as summarized in Table ​Table11 using a gantry-mounted kV imager (XVI, Elekta Oncology Systems). The pretreatment CBCT imaging protocols used in this investigation are currently clinically used to examine patients with a lung target. It should be noted that, for the sake of simplicity, CBCT acquired without MV treatment beam delivery is denoted by pretreatment CBCT since, in general, it is acquired prior to treatment for patient setup. As an initial test, the in-treatment CBCT imaging protocols in Table ​Table11 were created by simply adding the parameters specifically required for in-treatment CBCT acquisition without adjusting the other acquisition parameters, such as the tube voltage, tube current, exposure time, and rotation angle.

Summary of four cone-beam CT imaging protocols used in this study

The CBCT acquired without MV treatment beam delivery is denoted by pretreatment CBCT as it is acquired prior to treatment. The results of an initial image quality test for the listed protocols are summarized

^aThe time for the in-treatment CBCT images was estimated as the beam-on time expected to deliver the VMAT plan for patient 1. In general, the time to acquire an in-treatment CBCT image varies depending on the VMAT plan, which is concurrently delivered

For the intrafraction CBCT acquisition, a VMAT plan for a lung cancer patient (patient 1) was delivered to the Catphan phantom. A two-arc VMAT plan was created for the lung SBRT using FFF 10 MV X-ray beams. Among the two arcs, the first arc plan, which was delivered during the CBCT acquisition, was created to deliver 2585.3 MU in 77 s along a full-arc gantry rotation (from − 179° to 179°). The corresponding gantry speed to deliver the first arc of the VMAT plan was estimated to be approximately 279°/min on average. The average planned dose rate was estimated to be 2014.5 MU/min.

For the CBCT acquisitions, two-dimensional projections were acquired and then reconstructed to multiple-phase three-dimensional image volumes. Prior to the CBCT reconstruction, the acquired projections were sorted into ten respiratory phases (0, 10, 20, …, 90) by an automatic phase-sorting algorithm [19] in XVI. Among the ten respiratory phases, 0-phase and 50-phase represent end-of-exhalation and end-of-inhalation phases, respectively, and the other eight phases represent intermediate phases between the end respiratory phases.

It is noteworthy to mention that different numbers of projections were acquired for the pretreatment and in-treatment CBCT protocols due to the concurrent delivery of MV treatment beams for the in-treatment CBCT as summarized in Table ​Table1.1. For the in-treatment CBCT acquisition, it is not necessary to define the gantry speed since the linac gantry rotates with time-varying angular speeds, which are optimized for radiotherapy treatment plan. Consequently, the number of projections acquired for the in-treatment CBCT could vary depending on the VMAT plans delivered with the CBCT acquisition. Instead of defining the gantry speed, an angular interval, so called the acquisition interval in XVI, was defined for acquiring the in-treatment CBCT; the acquisition interval is an acquisition parameter that defines the minimum gantry rotation, upon which kV projection acquisition is triggered (0.1° for this study as suggested in a previous study [20]). As summarized in Table ​Table1,1, more projections were acquired for the three-dimensional (3D) in-treatment CBCT than the 3D pretreatment CBCT: 495 vs. 333 projections. On the other hand, a larger number of projections was acquired for the 4D pretreatment CBCT (1010) than for the 4D in-treatment CBCT (266) as gantry rotated more slowly for 4D pretreatment CBCT. Since the CBCT image quality is largely affected by the number of projections acquired [21–23], it was not feasible to evaluate the effect of the MV scatter from the results in Table ​Table11.

In order to more appropriately evaluate the impact of the MV scatter on the CBCT image quality, the gantry speed of the 3D and 4D pretreatment CBCT imaging protocols was adjusted. For the 3D pretreatment CBCT acquisition, the gantry speed was reduced from 360 to 240°/min. For the 4D pretreatment CBCT acquisition, the gantry speed was adjusted from 67 to 254°/min. By adjusting the gantry speed, similar numbers of projections were acquired for the pretreatment and in-treatment CBCT protocols, leading to a fair comparison of the image quality between the pretreatment and in-treatment CBCTs. It should be noted that the aforementioned adjustments of the gantry speed, which largely affected the resulting CBCT image quality, were performed only for evaluating the MV scatter effect. For the other experiments, pretreatment CBCT images were acquired with the standard imaging protocols summarized in Table ​Table1.1. The kV projections acquired with the 3D and 4D imaging protocols were reconstructed into 3D CBCT images.

To quantitatively evaluate the image quality of the 3D-reconstructed CBCT images, several image quality metrics were calculated using DoseLab Pro 7.0 (Varian Medical Systems, Palo Alto, CA, USA), an image analysis software package. As illustrated in Fig. 2, the Catphan 503 phantom consisted of three modules, each of which was used to analyze different image quality metrics: the contrast and contrast-to-noise ratio (CNR) (CTP404 in Fig. 2a), and the overall and minimum uniformity (CTP486 in Fig. 2b). First, the CTP404 module has multiple inserts with various electron density values; 10 regions of interest (ROIs) detected by DoseLab are displayed in Fig. 2a. For this investigation, ROIs 1 (water-equivalent) and 2 (Delrin) were used to calculate contrast and CNR as follows:

where S₁ and S₂ represents the mean pixel values of ROIs 1 and 2, respectively, whereas σ₁ and σ₂ represents the standard deviations of the pixel values in ROIs 1 and 2, respectively. Second, overall and minimum uniformities were calculated for ROIs 19–23 using an equation that calculates the uniformity metric for a region:

where S₉₀ and S₁₀ represent the 90th and 10th percentile of the pixel values in an ROI, respectively. The overall uniformity was calculated as the uniformity calculated for all of the ROIs. The minimum uniformity was calculated as the lowest value among the uniformity values calculated for each of ROIs 19–23.

Axial cuts of two modules in the Catphan 503 phantom visualized with regions of interest used for quantitative evaluations of CBCT image quality. The two modules, a CTP404 and b CTP486, were used to calculate the contrast- and uniformity-related metrics, respectively