CBCT image quality QA: Establishing a quantitative program

Abstract Purpose Routine quality assurance (QA) of cone‐beam computed tomography (CBCT) scans used for image‐guided radiotherapy is prescribed by the American Association of Physicists in Medicine Task Group (TG)‐142 report. For CBCT image quality, TG‐142 recommends using clinically established baseline values as QA tolerances. This work examined how image quality parameters vary both across machines of the same model and across different CBCT techniques. Additionally, this work investigated how image quality values are affected by imager recalibration and repeated exposures during routine QA. Methods Cone‐beam computed tomography scans of the Catphan 604 phantom were taken on four TrueBeam® and one Edge™ linear accelerator using four manufacturer‐provided techniques. TG‐142 image quality parameters were calculated for each CBCT scan using SunCHECK Machine™. The variability of each parameter with machine and technique was evaluated using a two‐way ANOVA test on a dataset consisting of 200 CBCT scans. The impact of imager calibration on image quality parameters was examined for a subset of three machines using an unpaired Student’s t‐test. The effect of artifacts appearing on CBCTs taken in rapid succession was characterized and an approach to reduce their appearance was evaluated. Additionally, a set of baselines and tolerances for all image quality metrics was presented. Results All imaging parameters except geometric distortion varied with technique (P < 0.05) and all imaging parameters except slice thickness varied with machine (P < 0.05). Imager calibration can change the expected value of all imaging parameters, though it does not consistently do so. While changes are statistically significant, they may not be clinically significant. Finally, rapid acquisition of CBCT scans can introduce image artifacts that degrade CBCT uniformity. Conclusions This work characterized the variability of acquired CBCT data across machines and CBCT techniques along with the impact of imager calibration and rapid CBCT acquisition on image quality.


| INTRODUCTION
Cone-beam computed tomography (CBCT) scans are often performed using on-board imaging (OBI) as part of image-guided radiation therapy (IGRT) with linear accelerators. 1 In contrast to diagnostic CT imaging, which often incorporates advanced image processing techniques to aid radiologists in diagnosis, [2][3][4] linear accelerator-based CBCT imaging typically uses filtered backprojection reconstruction techniques with the aim of providing an image with adequate soft tissue visualization that can be used to verify or improve patient alignment prior to treatment. 5,6 Cone-beam computed tomography images can also be used to estimate the physical dose that patients receive during treatment, 7 and are being explored for use in treatment planning 8,9 and adaptive radiotherapy. 10 All of these uses require CBCT images to have adequate and consistent image quality: from adequate contrast for soft tissue visualization to consistent Hounsfield unit (HU) to attenuation coefficient mapping for dosimetric calculations. 11 The increasing complexity and utilization of CBCT images require consistent CBCT performance, and thus a more stringent QA program.
Routine quality assurance (QA) of CBCT scanners tests for the consistency and adequacy of image quality metrics and helps ensure high-quality, clinical CBCT scans. QA of CBCT systems on linear accelerators, including safety, mechanicals, and image quality, is outlined in the American Association of Physicists in Medicine (AAPM) Task Group (TG)-142 12 with further recommendations for CT-based IGRT systems outlined in TG-179. 13 According to the guidelines of TG-142, image quality metrics/parameters for CBCT images performed on a monthly basis include: geometric distortion, spatial resolution, contrast, HU constancy, uniformity, and noise. 12 TG-142 recommends a tolerance of "baseline" for all but geometrically measured metrics ( Table 1)indicating that image QA should evaluate for changes in image quality relative to what was measured during machine commissioning and are based on the individual institution's data. In addition to TG-142, linear accelerator vendors provide specifications for the image quality of their CBCT scans that should be met by scanners in clinical use (Table 1). Ideally, monthly CBCT QA would test for both changes in image quality and for meeting manufacturer specifications by the machine.
Establishing a monthly CBCT image quality QA protocol that aligns with TG-142 requires determining a baseline value of each QA metric and the acceptable tolerances for measurements that are different from baseline. Image quality metrics will fluctuate between scans 14 and the tolerance would ideally be set to accept these fluctuations while still being sensitive to actual changes in the image quality. 15 Cone-beam computed tomography QA procedures and expected CBCT image parameter fluctuations have been discussed by several authors. Yoo et al. 16  Images were analyzed using SunCHECK Machine™ manufactured by Sun Nuclear Corporation (Melbourne, FL). Figure 1 shows the Catphan ® 604 modules along with the regions of interest used for image analysis in the software. Table 3 shows a list of image quality parameters that are measured on a monthly basis as per the recommendations of the TG-142 protocol. 12 Calculation of image quality parameters in SunCHECK Machine™ follows recommendations from the Catphan ® manufacturer. 14 Geometric distortion is determined using four center holes that are 50 mm apart, and is calculated as the largest difference in absolute value between measurements of the center holes. Spatial resolution is measured by generating a modulation transfer function (MTF) by calculating the modulation for each spatial frequency region of interest (ROI) [ Fig. 1(b)]. Uniformity is calculated using five ROIs, each approximately 2.9 cm in diameter, placed at the center and at the 3, 6, 9, and 12 o'clock positions around the periphery of the phantom [ Fig. 1(c)]. The mean HU for each of the five ROIs is measured and the maximum difference between the mean value of each peripheral ROI and the central ROI is determined. The largest difference is reported as uniformity. Contrast, C, is calculated using Eq. (1): where A is the mean gray scale value of the  (2), is used.
For determination of all image quality parameters, CBCTs were imported into SunCHECK Machine™, manually registered to the correct module, and ROIs were placed. All data were stored in SQL databases and analysis was completed using Excel (Microsoft Corporations, Redmond WA), and SPSS (IBM, Armonk NY).

2.A | Machine and technique dependence
Two hundred CBCTs were taken across five linear accelerators and using four CBCT techniques over the course of 12 months. These CBCTs were taken as part of routine monthly QA procedures with additional CBCTs taken specifically for this work. Ten CBCTs were taken with each technique on each accelerator. No image calibrations were done on any of the linear accelerators during the period the CBCTs were taken. All monthly QA CBCTs taken during the year time frame were included in this analysis except for nine scans that did not follow institutional protocols for acquisition (ie, not aligned to isocenter or wrong CBCT technique was used). This set of CBCT acquisitions were used to determine the variability of each image quality parameter as a function of machine and technique. The difference in the mean values of each image quality parameter across machines and techniques was evaluated using two-way ANOVA tests.

2.B | CBCT calibration
The TrueBeam ® 2.7 MR3 imaging system undergoes a calibration by the user before clinical use, after major service or part replacement,

2.C | Artifacts
Initial data analysis showed that some CBCTs taken for routine machine QA exhibited a cylindrical artifact of lower HU values in the center of the phantom (Fig. 2). It was noticed that the artifact appeared when CBCTs were acquired in rapid succession, and presented on CBCT scans that were taken immediately after either a Pelvis or Spotlight scan (the two higher dose scans in this study) was taken. Scans taken using the Head and Pelvis techniques primarily showed the artifact. This work investigated what QA procedures induced the artifact and the effects of the artifact on image quality parameters, mainly uniformity.  Table 3.

2.D | Baselines and tolerances
T A B L E 3 Image quality parameters, units, and slice used for analysis.

3.A | Machine and technique dependence
Image quality parameters display a range of trends when evaluated graphically as a function of technique and machine (Fig. 3). As expected, parameters, such as noise, slice thickness, and spatial reso-

3.B | CBCT calibration
Cone-beam computed tomographies were acquired pre-and postimager calibration with a total of 185 CBCTs in the dataset and a range from 5 to 12 CBCT image sets in each group (machine and technique). For each image quality parameter studied, a two-tailed, unpaired Student's t-test was used to test the hypothesis that the mean value of that image quality parameter was different between data collected before and after the CBCT calibration procedure. Differences were considered significant if P < 0.05. The differences in the means between the pre-and post-calibration data are shown in The change in the expected value of the image quality parameters after calibration was compared to the machine-to-machine variability shown in Fig. 3. Figure 4 shows, for each technique, the range in the average value for a given image quality parameter across all machines. Also plotted is the largest change in the parameter due to recalibrating the CBCT technique reported in Table 4.
With the exception of the Teflon, Delrin and 50% bone HU plugs for the Spotlight technique, the change in a given machine's expected HU value is within the range seen across machines. For the Teflon, Delrin and 50% bone HU plugs measured with the Spotlight technique, calibration-induced changes to the expected value of the plug were less than the machine-to-machine variability.

3.C | Artifacts
The artifact displayed in Fig. 2

was primarily produced in Head and
Pelvis CBCTs that were completed immediately after a high-dose scan and when the fan type was changed between the two scans. To  This artifact was not seen when a maximum of two CBCT scans were acquired in one session with the CBCTs having the same fan type (full or half fan). In-house QA procedures were adjusted to follow this two scan, same fan-type limit in order to minimize the presence of this artifact. Any CBCTs with this artifact were excluded from the analysis completed in other sections of this work if it was verified that the CBCTs with the artifact had been acquired rapidly after other CBCT scans.

3.D | Baselines and tolerances
The CBCT dataset analyzed in Section 3.A was used to generate baselines and tolerances. With the exception of some high-Z HU constancy tests that included the Teflon and 50% bone materials, the differences in image quality metrics between machines evaluated in Section 3.A fell within the guidelines of the AAPM TG-142 protocol and the manufacturer's specifications (Table 2) even though the differences were statistically significant. Additionally, as shown in  (Table 5).
Baselines for geometric distortion and uniformity were set to zero as this is the ideal value for these parameters and it was desired for tolerances to be relative to this value. Baseline values were defined to be technique-specific because various CBCT acquisition settings, including kV, mAs, fan type, and trajectory, would be expected to affect image quality parameters, such as noise, slice thickness, and spatial resolution, as seen in Fig. 3.
The spread and standard deviation from Section 3.A were used to determine institutional tolerances, which are presented in Supplemental Table S6. Where specific tolerances were defined in either TG-142 or the specifications of the Varian TrueBeam ® , the institution adopted these tolerances. These included geometric distortion (1 mm), uniformity (40 HU), and HU constancy for non-high-Z material (50 HU). Two standard deviations (95% confidence interval) of the institutional data were used to set the results for spatial resolution, contrast, noise, and slice thickness based on the analysis found in Stanley et al. 18 For HU constancy tests of high-Z material, the tolerance was set to either 50 HU (manufacturer specification) or two standard deviations of the dataset, whichever was more lenient.

| DISCUSSION
During the course of investigation, it was found that certain QA procedures affected the variability of CBCT image quality metrics. First, acquiring CBCTs in rapid succession during QA has the potential to T A B L E 4 Difference in the mean between the pre-calibration and post-calibration data on three TrueBeam ® machines (TB1, TB2, and TB3). Items in bold are statistically significant (P < 0.05).

Parameter
Head Spotlight  Thorax  Pelvis   TB1  TB2  TB3  TB1  TB2  TB3  TB1  TB2  TB3  TB1  TB2  It was also found that image quality measurements pre-and post-CBCT calibrations were statistically different. As baselines and tolerances are often set using CBCT data, the user could potentially run into an issue where current baselines and tolerances do not appropriately match the measurements that are performed using a new CBCT calibration. To prevent this, the user would need to characterize any changes in image quality metrics and adjust baselines and tolerances accordingly after completing a CBCT calibration.
There has been limited guidance on establishing CBCT QA programs across multiple machines and techniques of the same type.
Results from the two-way ANOVA test (Section 3.A) indicated that T A B L E 5 Measurements of uniformity for cone-beam computed tomographies (CBCTs) acquired immediately after a high-dose scan and CBCTs acquired with at least a 10-min interval between scans for the Head and Pelvis techniques.

Scan number
Post This work began as a departmental initiative to determine datadriven baselines and tolerances in order to improve our Linac QA

| CONCLUSION
This work focused on the standardization of a CBCT image QA program in an institution with multiple machines and multiple CBCT F I G . 5. Catphan ® uniformity module for cone-beam computed tomographies (CBCTs) acquired in rapid succession and post 10-min interval. The calculated uniformity value is presented for each CBCT.

SUPPORTING IN FORMATION
Additional supporting information may be found online in the Supporting Information section at the end of the article.