Background
Daily repeatability of patient positioning is vital to the successful treatment outcome in radiotherapy. The megavoltage computed tomographic (MVCT) image quality from helical TomoTherapy (HT) has been proved sufficient for image-guided radiotherapy (IGRT) including tumor identification and setup verification [
1]. Moreover, considering that MVCT has a reliable CT number to electron density calibration curve, MVCT has also been proved accurate to be used for calculating adaptive daily dose distribution which is an assurance of adaptive radiotherapy (ART) [
2]. HT MVCT consisted of advantages in lower absorption dose and a larger imaging capacity (theoretically 40 cm reconstruction field of view (FOV) × 160 cm longitudinal scanning length) than C-arm based cone beam CT (CBCT) technology [
3,
4]; thus making HT daily MVCT imaging safe in assessing patient treatment location and make on-line adaptive re-planning possible using the same imaging data sets.
In the process of MVCT image acquisition on TomoTherapy Operator Station (version 5.0, Accuray, Sunnyvale, CA), there are options for ‘acquisition pitch’ (AP) and ‘reconstruction interval’ (RI). AP is directly proportional to couch speed, scan duration and especially correlated to the patient absorption dose [
5]. In the meanwhile, RI is much related to 3D reconstruction with varied slice thickness which will impact MVCT image qualities especially in the craniocaudal resolution.
There are still debates regarding to the effect of MVCT scan options on setup verification. Studies argued that there is no noticeable difference in image quality among different APs [
6] or the issue could be ignored [
7], while some studies concluded that pitch variation and longitudinal MVCT image resolution make statistically significant contribution to the MVCT- planning kilovoltage CT (KVCT) registration process [
8,
9]. To the best of our knowledge, there is no study published on the effect of MVCT scan options on adaptive dose calculation. In order to investigate the quantitative effect of MVCT scan and reconstruction options on IGRT and ART, an experiment with two different phantoms was performed and the clinical effects were analyzed and presented in terms of image qualities, the accuracy for setup verification and dosimetry discrepancies.
Discussion
IGRT is an efficient tool to improve the accuracy of targeting and is necessary to ART [
15]. In head and neck and prostate cancer patients, setup error > 5 mm can occur in 10% [
16] and 20% [
4] fractions respectively when volume imaging was performed every second day. Li found that the interfractional variations in patient setup and in shapes, sizes, and positions of both targets and normal structures are site specific and may be used to determine the site-specific margins, which emphasized the importance of ART [
17]. Duma [
18] recommends daily MVCT scan for patients treated with doses close to the tolerance dose of the critical ROI (such as spinal cord). Furthermore, random setup errors can only be eliminated through the use of daily image guidance, whilst the dose from MVCT scan is critical to evaluate the treatment outcome. Quantitatively, Shah [
4] proved that the typical HT MVCT imaging dose is approximately 1.5 cGy per image and the uniform MVCT dose delivered using HT is greatest when the anatomic thickness is the smallest and the pitch is set to the lowest value. Since reduplicative MVCT scans on human body is unethical, an anthropomorphic phantom based experiment was created for this study, which aims to give a reasonable MVCT scan protocol by investigating the effect of scanning setting and technique during setup verification and adaptive dose calculation in TomoTherapy system.
Since option of AP is directly proportional to scan time and creating additional irradiation dose to patients [
5], it is recommended to use pitch as large as possible if there is no impact on IGRT imaging quality and ART accuracy. In term of adaptive dose calculation, our study shown that MVCT from any AP and RI combinations has no significant effect on these MVCT calculated dose distributions. Langen [
2] has indicated that the MVCT pitch ratio has little effect on dose recalculation. The robustness of the recalculations is impressed, given the coarse sagittal reconstructions that were used [
2]. With the conclusion, then the use of AP of ‘coarse’ may be tempting in an effort to reduce the imaging dose, which a patient has to receive due to the MVCT acquisition. Yue [
19] also argued that different APs have no significant difference on CT number (
p = 1.000) and image noise (
p = 0.667) among MVCT imaging technical settings. But there is no study, to the best of our knowledge, has ever reported the effect of MVCT RIs. According to our study, we have concluded and recommended the fastest and harmless scan option in case of adaptive planning: AP of ‘coarse’ with RI of 6 mm. In term of ART, the image stability is another critical issue. As Langen [
20] resulted that, during a 9-month period, an average variation of 20 Hounsfield units was seen (by a CT-to-electron density phantom) for the fan-beam MVCT imaging system. However, Langen [
2] also argued that, this variation does not affect the dose calculation result significantly. An agreement of 0.5% was found between the planned and recalculated dose received by 95% of the target volume. Therefore, it was concluded that the fan-beam MVCT dose calculation accuracy is similar to that of the initial diagnostic KVCT dose calculation. And in another perspective, to approve the stability, MVCT image quality is one of the most important content in the Tomotherapy monthly QA protocol according to report TG-148, where image noise, uniformity, spatial resolution, contrast, and the MVCT dose are monitored. Furthermore, ‘CT number calibration’ is a mandatory weekly QA procedure on TomoTherapy operation station, which will compare the average measured HU to expected HU for both homogeneity water equivalent cheese phantom and air. If the absolute difference is smaller than 70 HU, TomoTherapy system will automatically correct the CT numbers in order to ensure consistency of the CT number values. So far, based on the arguments above, we can be confident to the MVCT image’s stability for accurate dose calculation in adaptive radiotherapy.
However, in term of IGRT, our study shows that large AP (coarse) at KVCT-MVCT registration would lead to large error, which was coincident to a previous study which found that 4 and 2 mm inter-slice spacings generally have lower residual error than 6 mm spacing [
21] Therefore, to tradeoff between IGRT and ART requirement, AP of ‘normal’ with RI of 2 mm is recommended in daily IGRT and ART clinical application based on our research results. However, in case of stereotactic body radiation therapy (SBRT) with early stage lung tumor (< 4 cm) [
22], since AP of ‘normal’ and ‘coarse’ have larger step size per rotation (8 mm/rotation and 16 mm/rotation, respectively) which may lose critical target anatomy information, tumor density [
23] and texture features [
24] which may be used to predict therapeutical effect, AP of ‘fine’ (4 mm/rotation) with RI 1 mm setting is recommended.
Smaller RI may improve the longitudinal display resolution, with neither impact on scan time, patient dose nor impact on ART dose recalculation as well as IGRT registration accuracy. Furthermore, larger reconstruction resolution may miss smaller tumor or lymph nodes in adaptive delineation and replanning process. Levegrun [
8] also reminded that even with the finest pitch, the MVCT images of a thin object are blurred longitudinally to limit the visibility of small anatomic structures. Therefore, no matter which AP is selected, the smaller RI is recommended to provide relative higher resolution in daily IGRT and ART applications. However, since the maximum number of slices allowed for an MVCT scan is 300, thus the ‘coarse’ setting and larger RI is a compromise if the scan is necessary to cover as long as possible. In this case, ‘full image’ registration technique should be used to remedy the disadvantage of large AP that might generate large registration errors. A reasonable explanation is that registration with ‘full image’ technique enrolled more voxels from MVCT, which provided closer axial resolution to planning CT. Our recent studies also proved that, if two series of images have a closer axial resolution, the registration will be matched better and will deliver higher registration accuracy [
12,
25].
Compared to lateral and ventrodorsal directions, our study shows that registration error in craniocaudal direction is significantly larger, which is consistent with previous published data and our previous study from both KVCT-MVCT and KVCT-CBCT registration techniques [
8,
26‐
29]. This phenomenon may be caused by the trade-off against imaging speed and increase with the pitch of the MVCT scan [
30]. Regarding to the craniocaudal registration errors, we found that it decreased in the pelvic phantom than in the chest phantom and we assumed it may come from the lower craniocaudal resolution in the chest but rather higher one in pelvis because of the complicated pelvic bony structures. Considering that the bony structures of the chest are similar in between transversal slices, the CT-MVCT registration protocol of craniospinal irradiation should probably be different from the other anatomic site such as head and neck and pelvic etc. Therefore, we suggest the craniocaudal scanning length should encompass at least one entire vertebra (including thoracic spine or lumbar spine) in craniospinal irradiation in order to correct the body roll direction and this claim should be substantiated by further research. Furthermore, the automatic registration algorithm could potentially eliminate inter-observer’s deviation altogether but also can impair local matches highly weighted by individual observers and may introduce a procedure-dependent new error [
8]. Therefore, manually perform a double check and adjustment to the automatic registration result by radiation oncologist should be part of the strict CT-MVCT registration protocol in craniospinal irradiation, lest low craniocaudal resolution of MVCT may bring large registration error just as one vertebra is incorrectly matched to the adjacent one [
30,
31].
MVCT scan takes approximate 10 s for one gantry rotation regardless of AP options, which will include 2~ 5 respiratory cycles for chest imaging. With the highest pitch, AP of ‘coarse’ will cover craniocaudal12mm per rotation. Considering that each fraction of scan should encompass the motion envelope of the target for accurate IGRT and ART, inspiratory motion and resultant imaging artifacts cannot be avoided by breath holding in TomoTherapy MVCT serials. A phantom based experimental study has proved that motion artifacts in TomoTherapy MVCT or planning CT studies changed the accuracy of the automatic registration process by less than 2.0% [
21]. Based on this result, effects of AP and RI on IGRT and ART accuracy were investigated by static phantoms in this study, and we propose that it is possible to apply present findings to the moving targets or organs. However, this conclusion relies on two assumptions: first, moving target is much smaller than the total volume of the thorax imaged; second, respiratory movement of the organs like ribcage, diaphragm does not affect the registration process. If they cannot be met, gross motion may bring unexpected impact to IGRT and ART accuracy, which should be investigated by further research.
One of the probable resolution of gross respiratory motion and target shape change is deformable registration. As mentioned above, since manually performed double check is critical to accurate IGRT, a clinician may modify the automatic registration result based on their judgement on the alignment of the target. This may bring a new tradeoff between alignment of target and surrounded organs. Deformable registration can enroll more voxels and multi scales of mutual information during the match, which may meet the requirements from both target and organs, and bring more accurate registration in real situations. If so, we assume that AP of ‘fine’ with smaller RI may bring better deformable registration result and fewer residual errors because of its more sufficient imaging information. However, this assumption should also be substantiated by further research.
Beside of the translational corrections, TomoTherapy IGRT system can use the gantry angle to correct for the roll (along with the craniocaudal axis) adjustment automatically. TomoTherapy software allows a roll correction of any angle bigger than 0.1° (0.1° is the limiting value of gantry’s mechanical precision) prior to treatment. Rotational adjustment in one direction results in gantry rotation in the opposite direction to achieve the correct angular relationship between the radiation source and patient. However, roll correction was not enrolled in this study. We considered that if the image volume is rotated clockwise or counterclockwise, lateral and vertical adjustments become linked, and future lateral or vertical adjustments will proportionally affect each other. Since this study predominantly aimed to reveal the residual errors on 3 translational directions, rotational disagreement was eliminated to avoid its translational impacts in the experiment. However, if deformable registration method can be used in TomoTherapy IGRT module, rotational disagreement value and the corresponding correction may bring higher accuracy, which should also be substantiated by further research.