Recommendations of megavoltage computed tomography settings for the implementation of adaptive radiotherapy on helical tomotherapy units*

Abstract Megavoltage computed tomography (MVCT) image quality metrics were evaluated on an Accuray Radixact unit to recommend scan settings for the implementation of a consistent adaptive radiotherapy program. Megavoltage computed tomography image quality was evaluated and compared to a kilovoltage CT (kVCT) simulator using a commercial cone beam computed tomography image quality phantom. Megavoltage computed tomographies were acquired on the Accuray Radixact using fine, normal, and coarse pitches, with all available reconstruction slice thicknesses, each of which were reconstructed using standard and iterative reconstruction (IR). Image quality metrics (IQM) were evaluated using DoseLab: automatically and manually calculated spatial resolution, subject contrast, and contrast‐to‐noise ratio (CNR). Scanning time was 15.6 s/cm for fine, 8.1 s/cm for normal, and 5.6 s/cm for coarse pitch. Automatically evaluated spatial resolutions ranged from 0.39, 0.41, to 0.42 lp/mm for standard reconstruction and from 0.24, 0.21, to 0.18 lp/mm for soft‐tissue IR, respectively, with general IR yielding values in between these. Spatial resolution for kVCT was measured to be at least 0.42 lp/mm. Contrast was consistent across MVCT settings with 8.1 ± 0.2%, while kVCT contrast was 10.27 ± 0.05%. CNR was calculated to be 3.3 ± 0.4 for standard reconstruction, 7.4 ± 0.4 for general IR, and 12.0 ± 1.9 for soft‐tissue IR. It was found that increasing reconstruction slice thickness for a given pitch does not improve IQMs. Based on the consistency of contrast metrics across pitch values and the only slightly reduced spatial resolution using normal compared to fine pitch, we recommend the use of normal pitch with 2 mm slice thickness to maximize image quality for ART while limiting scanning time. Only for sites for which improved CNR is required and reduced spatial resolution is acceptable, soft‐tissue IR is recommended.


| INTRODUCTION
Adaptive radiotherapy (ART) holds the promise that the on-or offline adaptation of treatment plans will improve the efficacy of cancer treatments by improving dose delivery to the tumor and minimizing normal tissue doses based on the most recent anatomy of a patient. 1 In general, adaptation of a treatment plan due to variations in organ filling, weight gain or loss as well as tumor response requires volumetric imaging of the patient. 2 Commonly, this is performed by acquiring a repeat simulation computed tomography (CT) scan followed by manual replanning. Automated techniques using deformable image registration (DIR) and automated segmentation of contours or contour propagation, followed by automated replanning, can make ART more feasible for use across all body sites and cancer types. These automated techniques, especially DIR and auto segmentation, require volumetric imaging with sufficient image quality to produce correct results. [3][4][5][6] Recently, linear accelerator platforms that include kilovoltage or megavoltage computed tomography [kilovoltage (kV)/megavoltage computed tomography (MVCT)] for daily patient setup have come into common use in clinics around the world. This volumetric imaging data could serve as the input data for automated ART. Helical tomotherapy units can perform MVCTs over the entire volume-of-interest (VOI) length. These MVCTs can then be used as deformation targets for the original planning kVCT, followed by structure propagation and ultimately, automated plan re-optimization. 3

| MATERIALS AND METHODS
A commercial (cone beam) computed tomography (CT/CBCT) image quality phantom, CatPhan 504 (The Phantom Laboratory, Greenwich, NY), was used to evaluate MVCT image quality on an Accuray Radixact (Accuray, Sunnyvale, CA) radiation therapy system. This phantom contains geometry and sensitometry, high-resolution, low-contrast, and uniformity test modules.7 Other image quality phantoms with test modules for spatial resolution, contrast, and contrast-to-noise could also have been employed for this study.
After acquiring a kilovoltage computed tomography (kVCT) using the department's CT simulator with 2.5 mm slice thickness and 120 kVp/43 mAs technique, a phantom plan was created on the Accuray Precision treatment planning system. Megavoltage computed tomographies were then acquired on the Radixact (Accuray, Sunnyvale, CA) using fine, normal, and coarse pitches with all available reconstruction slice thicknesses: fine (1, 2 mm), normal (2, 4 mm), and coarse (3, 6 mm). Each MVCT was reconstructed using standard and iterative reconstruction (IR) techniques. 8 Employing DoseLab v6.8 (Varian Medical Systems, Palo Alto, CA), several image quality metrics (IQM) were evaluated. Image quality metrics used in the comparison were automatically and manually evaluated spatial resolution, contrast, and contrast-to-noise ratio (CNR). Additionally, MVCT scanning speed in time per scan length was calculated to evaluate feasibility for daily use of MVCT imaging. Spatial resolution is given by the spatial frequency (in lp/mm) at which the value of the normalized modulation transfer function (MTF) is 50%, where the modulation values are normalized by the maximum modulation which corresponds to the lowest resolution bar pattern. Modulation is calculated as S90ÀS10 S90þS10 from the 90th (S 90 ) and 10th (S 10 ) percentile signal levels in the region-of-interest (ROI) defined in DoseLab. Contrast between two regions is calculated as where S 2 and S 1 are the mean pixel values in the regions of larger and lower signals, respectively. CNR is calculated from the contrast between two regions, C 2,1 , divided by the noise present, , which is the ratio of the quadratic sums of standard deviations and signals in the ROIs.
To assess image quality outside of the axial image acquisition plane, the CBCT phantom was positioned in line with the table movement direction. Subsequently, MVCTs were acquired using all pitch values and reconstruction techniques, using the lowest available slice thickness for each. Spatial resolution and contrast metrics were evaluated from sagittal reconstructions of these images employing ImageJ following the calculation methodologies outlined above. 9

| RESULTS
Upon visual inspection, MVCT images exhibit more noise and lower spatial resolution than kVCT (cf. Fig. 1, left column, rows 1-4). While it is possible to identify the slice thickness and low-contrast circles (inner ring) on the kVCT (cf. Fig. 1, right column, row 1), they are absent on the MVCT (cf. Fig. 1, right column, rows 2-4). The use of iterative reconstruction techniques also leads to a noticeable decrease in spatial resolution and noise at a given pitch, whereas subject contrast appears to be consistent (cf. Fig. 2).
Axial in-plane image quality metrics for kVCT and MVCT are tab-

| DISCUSSION
In this study, we have evaluated a number of image quality metrics (IQMs) for MVCT to determine which combination of slice thickness, pitch, and reconstruction algorithm will maximize the image quality at a reasonable scanning time allowing for an efficient implementation of daily imaging for ART. We found that contrast metrics were consistent across pitch values, and therefore the remaining primary IQMs for comparison were scanning speed and spatial resolution.
The highest spatial and axial resolutions were found for fine pitch (using manual evaluation); however, the use of a fine pitch increases the scanning time twofold compared to the use of a normal pitch. show that a normal pitch with 2 mm slice thickness maximizes image quality while limiting scanning time and should therefore be the preferred option for implementation of ART into the clinic (cf. Table 1).
The choice of reconstruction technique depends primarily on the required soft-tissue contrast for daily setup and DIR algorithms. In cases where bony anatomy is most adequate for registration, we recommend using standard reconstruction, which yields the highest spatial resolution with the lowest CNRs. Soft-tissue IR is recommended if soft-tissue contrast, that is, CNR, is tantamount and reduced spatial resolution is acceptable. General iterative reconstruction offers a middle ground between standard and soft-tissue iterative reconstruction having both adequate spatial resolution and CNR.
Spatial resolution was found to be the most sensitive IQM for sagittal and coronal reconstructions with respect to pitch value; however, it was found to be consistent among reconstruction techniques. This is likely due to the already small spatial resolution for In general, the reconstruction technique should be kept the same throughout the treatment to yield consistent results in adaptive dose accumulation.

CONFLI CT OF INTEREST
This work has been supported in part by a research grant from Accuray Inc. CV has received meeting attendance support from Accuray, Inc.; RB has received research support from Accuray Inc.; MG and WT have received research grants and research support from Accuray, Inc.
T A B L E 2 Summary of sagittal plane image quality metrics acquired using the kV-CBCT phantom.