Image quality evaluation of intra‐irradiation cone‐beam computed tomography acquired during one‐ and two‐arc prostate volumetric‐modulated arc therapy delivery: A phantom study

Abstract Purpose To evaluate (a) the effects of megavoltage (MV)‐scatter on concurrent kilovoltage (kV) projections (P MVkV) acquired during rotational delivery, and (b) the image quality of intra‐irradiation cone‐beam computed tomography (ii‐CBCT) images acquired during prostate volumetric‐modulated arc therapy (VMAT) delivery. Methods Experiment (1): P MVkVs were acquired with various MV beam parameters using a cylindrical phantom: field size (FS), MV energy (6 or 15 MV), dose rate (DR), and gantry speed. The average pixel values were calculated in a region on each P MVkV which were extracted at eight equally spaced gantry angles. Experiment (2): 11 one‐arc and seven two‐arc 15 MV prostate VMAT plans were used along with a pelvis phantom. One plan was selected from each of arc plans and its MV energy was changed to 6 MV. After P MVkVs were acquired, projections consisting of MV‐scatter only (P MVS) were acquired with closing kV blades and subtracted from P MVkV (P MVScorr). Projections by kV beams only were acquired (P kV). The corresponding CBCT images were reconstructed (CBCTMVkV, CBCTMVScorr, and CBCTkV). The root‐mean‐square errors (RMSEs) were calculated in prostate region and 3D gamma analysis was conducted, in which the CBCT‐number was used instead of doses between ii‐CBCT images and CBCTkV (30 HU/1 mm). Results Experiment (1): The MV‐scatters were dependent on the FSs, MV energies, and DRs. Experiment (2): The median RMSEs for CBCTMVScorr were decreased by 107.5 HU (1‐arc) and 42.9 HU (2‐arc) compared to those for CBCTMVkV. The median GPRs for CBCTMVScorr were 94.7% (1‐arc) and 93.4% (2‐arc), while those for CBCTMVkV were 61.1% and 79.9%, respectively. GPRs for 6 MV plans were smaller than those for 15 MV plans. Conclusions The number of MV‐scatters increased with larger FSs and DRs, and smaller MV energy. The MV‐scatters were corrected on the CBCTMVScorr regardless of the number of arcs.


| INTRODUCTION
Image-guided radiotherapy has been developed extensively in the past 2 decades. 1 One such method is cone-beam computed tomography (CBCT) image acquisition by a linear accelerator (linac)-mounted kilovoltage (kV) imaging subsystem. With the advent of CBCT, it has become possible to confirm the location of internal organs in the treatment position prior to megavoltage (MV) beam irradiation. 2 Subsequently, the demand for the monitoring of the target or internal organs during MV beam irradiation has increased. Poulsen et al. proposed kilovoltage intrafraction monitoring (KIM), which is a three-dimensional (3D) target position estimation method during MV beam irradiation for the prostate region. [3][4][5] The authors estimated the 3D positions of implanted radiopaque markers with a monoscopic view by using spatial probability density. KIM has been used for the prostate and liver with volumetric-modulated arc therapy (VMAT). [6][7][8][9][10] However, KIM only extracts 3D point positions and cannot generate 3D volume images.
In routine clinical practice, it is desirable to obtain 3D images of the actual delivered dose distributions for adaptive radiotherapy and accurate prognosis prediction. To calculate the distributions, 3D volume images should be acquired during MV beam irradiation, which can reflect the actual positions of the target or internal organs inside the patient body. Although CBCT images that are established prior to MV beam irradiation can be used for this purpose, the internal organ positions may differ during the setup and irradiation. CBCT acquisition methods during rotational therapy such as VMAT delivery, or intra-irradiation CBCT (ii-CBCT) acquisition, have been presented to acquire 3D volume images during MV beam irradiation. [11][12][13][14] A major issue related to ii-CBCT acquisition is scattered X-rays of MV beams from a patient (MV-scatters), which are incident on the flat panel detector (FPD) of the linac-mounted kV imaging subsystem. The authors in the above studies used a Catphan phantom (Phantom Laboratory, Salem, NY, USA) to evaluate the image quality of the ii-CBCT images, which did not mimic human anatomy. Boylan et al. demonstrated correction methods for MVscatters using an anthropomorphic phantom and prostate VMAT patients. 13 The authors subtracted MV-scatter maps from the kV projections acquired during 15 MV beam delivery, as follows: (a) 2dimensional (2D) MV-scatter maps of the phantom, (b) mean 2D maps of the phantom, (c) 2D maps of another phantom, and (d) maps estimated using an analytical model. It was revealed that subtracting the MV-scatter map acquired by the same object was appropriate for the correction. However, only a qualitative visual evaluation was conducted for the patient ii-CBCT images.
In recent years, in-treatment magnetic resonance (MR) images acquired using MR-cobalt or MR-linac machines, which have been introduced in radiotherapy treatment, have provided superior softtissue contrast over in-treatment CBCT images. 15 However, these machines are not commonly installed globally and cannot perform VMAT or non-coplanar MV beam deliveries at present. From this perspective, ii-CBCT acquisition is a preferable option because linacs with kV imaging subsystems are in widespread use.
The purpose of this study was (a) to investigate the basic characteristics of MV-scatters relating to various MV beam parameters [field size (FS), MV energy, dose rate (DR), and gantry speed] on concurrent kV imaging projections, and (b) to evaluate the qualities of MV-scatter-contaminated and MV-scatter-corrected ii-CBCT images using a pelvis phantom, which were acquired during 1-arc or 2-arc 15 MV prostate VMAT deliveries.    16 .The target was local prostate only, lymph nodes were not irradiated. The plan details are described in Table 3. Each plan was created and optimized by the The MV-scatter correction method is summarized in Fig. 2. Firstly, the subject was imaged using a kV imaging subsystem during rotational MV beam irradiation (P MVkV ). In this case, kV projections consisting of MV-scatter only (P MVS ) had to be generated to correct the MV-scatters on P MVkV . Thus, concurrent kV imaging was conducted using closing kV blades, following which P MVS was acquired. For 2-arc plans, P MVS were acquired by each arc. Thereafter, P MVS was subtracted from the corresponding P MVkV angle wise (P MVScorr ) and the subtraction was demonstrated pixel wise. The correction could be expressed as:

| MATERIALS AND METHODS
where σ and θ are the image noise and projection angle, respectively.
Furthermore, the subject was scanned using a kV beam only (P kV ). T A B L E 3 Details of prostate volumetric-modulated arc therapy plans used in this study (Experiment 2). The CBCT images were reconstructed from P MVkV , P MVScorr , and

2.B.2 | Image quality evaluation
The root-mean-square error (RMSE) was calculated in prostate region for the ii-CBCT images compared to CBCT kV for the quantita- inside right (left) ilium) was defined in lesser pelvis region. The RMSE of CBCT i (i = MVkV or MVScorr) was defined as where HU u, v;s ð Þ is the CT number at pixel position u, v ð Þ in slice s.
Moreover, N u and N v , and N s are the total number of pixels along the u and v directions, and the total number of slices, respectively.
Furthermore, 3D gamma analysis was applied for each ii-CBCT compared to CBCT kV . VOI for the analysis was the entire range of the image, and different from the VOI in lesser pelvis region defined above. Low        The difference of GPR between 6 MV and 15 MV VMAT is shown in Fig. 8. GPRs of CBCT MVkV s acquired during 6 MV VMAT for the 1-arc and 2-arc plans were 26.3% and 53.5%, while those acquired during 15 MV VMAT were 48.3% and 71.6%, respectively.
The image quality of CBCT MVkV acquired during 6 MV VMAT was degraded compared to that acquired during 15 MV VMAT since more MV-scatters were generated, which was supported by the results of the Experiment 1 [ Fig. 3(a)]. The improvements in the median RMSEs of CBCT MVScorr from CBCT MVkV for the 2-arc plan were slightly smaller than those for the 1-arc plan, although the number of MV-scatters for the 2-arc plan was smaller than that of the 1-arc plan, as discussed above. However, the GPRs of CBCT MVScorr for the 2-arc plans, which considered the CBCT-number difference and distance-to-agreement simultaneously, were comparable with those for the 1-arc plan.

| CONCLUSION S
In this study, the effects of MV-scatters were evaluated by varying the MV beam parameters. Moreover, to the best of our knowledge, this is the first study to evaluate the image quality of ii-CBCT images of 1-arc and 2-arc prostate VMAT deliveries quantitatively using the RMSE and 3D gamma analysis. The number of MV-scatters increased with larger FSs, higher DRs, and smaller MV energy. Although the effects of the MV-scatters resulted in cupping artifacts on the ii-CBCT images, the MV-scatters were corrected in the MV-scatter-corrected ii-CBCT images by MV-scatter correction, which exhibited >90% GPRs regardless of whether the 1-arc or 2-arc plan was used.

ACKNOWLEDG MENTS
HI and AK conducted the phantom study and analysis, and drafted the manuscript. MN and TM conceived the study, participated in its design and coordination, and helped to draft the manuscript. All authors read and approved the final manuscript.

CONFLI CT OF INTEREST
The authors of this publication have no conflict of interest to declare.