Journal of Medical Physics
 Home | Search | Ahead of print | Current Issue | Archives | Instructions | Subscription | Login  The official journal of AMPI, IOMP and AFOMP      
 Users online: 319  Home  EMail this page Print this page Decrease font size Default font size Increase font size 


 
 Table of Contents    
ORIGINAL ARTICLE
Year : 2022  |  Volume : 47  |  Issue : 4  |  Page : 344-351
 

Comparison of three commercial methods of cone-beam computed tomography-based dosimetric analysis of head-and-neck patients with weight loss


Department of Oncology, University of Alberta; Department of Medical Physics, Cross Cancer Institute, Edmonton, Alberta, Canada

Date of Submission28-Jan-2022
Date of Decision27-Aug-2022
Date of Acceptance02-Sep-2022
Date of Web Publication10-Jan-2023

Correspondence Address:
Dr. Satyapal Rathee
Department of Oncology, University of Alberta and Department of Medical Physics, Cross Cancer Institute, 11560 University Ave, Edmonton, Alberta T6G 1Z2
Canada
Login to access the Email id

Source of Support: None, Conflict of Interest: None


DOI: 10.4103/jmp.jmp_7_22

Rights and Permissions

 

   Abstract 

Purpose: This investigation compares three commercial methods of cone-beam computed tomography (CBCT)-based dosimetric analysis to a method based on repeat computed tomography (CT). Materials and Methods: Seventeen head-and-neck patients treated in 2020, and with a repeat CT, were included in the analyses. The planning CT was deformed to anatomy in repeat CT to generate a reference plan. Two of the CBCT-based methods generated test plans by deforming the planning CT to CBCT of fraction N using VelocityAI™ and SmartAdapt®. The third method compared directly calculated doses on the CBCT for fraction 1 and fraction N, using PerFraction™. Maximum dose to spinal cord (Cord_dmax) and dose to 95% volume (D95) of planning target volumes (PTVs) were used to assess “need to replan” criteria. Results: The VelocityAI™ method provided results that most accurately matched the reference plan in “need to replan” criteria using either Cord_dmax or PTV D95. SmartAdapt® method overestimated the change in Cord_dmax (6.77% vs. 3.85%, P < 0.01) and change in cord volume (9.56% vs. 0.67%, P < 0.01) resulting in increased false positives in “need to replan” criteria, and performed similarly to VelocityAI™ for D95, but yielded more false negatives. PerFraction™ method underestimated Cord_dmax, did not perform any volume deformation, and missed all “need to replan” cases based on cord dose. It also yielded high false negatives using the D95 PTV criteria. Conclusions: The VelocityAI™-based method using fraction N CBCT is most similar to the reference plan using repeat CT; the other two methods had significant differences.


Keywords: Cone-beam computed tomography, deformable registration, head and neck, weight loss


How to cite this article:
Rathee S, Burke B, Heikal A. Comparison of three commercial methods of cone-beam computed tomography-based dosimetric analysis of head-and-neck patients with weight loss. J Med Phys 2022;47:344-51

How to cite this URL:
Rathee S, Burke B, Heikal A. Comparison of three commercial methods of cone-beam computed tomography-based dosimetric analysis of head-and-neck patients with weight loss. J Med Phys [serial online] 2022 [cited 2023 Feb 1];47:344-51. Available from: https://www.jmp.org.in/text.asp?2022/47/4/344/367426



   Introduction Top


Conventional head-and-neck radiotherapy is generally delivered in 30 or more fractions. A significant fraction of these patients require ongoing surveillance of anatomical changes based on daily cone-beam computed tomography (CBCT) imaging. Significant anatomical changes include sensitive structures moving into high-dose regions or primary disease moving out of high-dose region resulting from poor immobilization and/or weight loss. Weight loss and/or poor immobilization may initiate a need analysis for adaptive replanning. Adaptive replanning or adaptive radiotherapy (ART) involves a duplication of the entire planning process including immobilization and computed tomography (CT) simulation, delineation of organs at risk (OARs) and target structures, plan optimization, and quality assurance. This process is obviously resource intensive; however, ART has been shown to improve tumor control and decrease toxicity.[1],[2]

Online CBCT imaging, used for accurate patient positioning prior to treatment, also provides 3D anatomical images to aid in the need analysis for ART. Although CT numbers in CBCT images are inaccurate due to the changing scattering volume compared to calibration phantom,[3],[4] patient-specific CT number to electron density calibration may be used.[5] In addition, the field of view (FOV) of CBCT is generally inadequate for imaging the entire “to-be-treated” volume as delineated in planning CT. MacFarlane et al.[6] assigned electron densities to CBCT data via slice-specific correlations of CT numbers between CBCT and planning CT after deforming the planning CT to CBCT. This is an attractive approach to overcome inaccuracies in CT numbers; the study did require in-house MATLAB coding (MATLAB 2019b, MathWorks Inc., Santa Clara, CA, USA) outside of clinical workflow and manual adjustment of contours for reporting dose to OARs and target volumes.

For “need to replan” analysis of ART, several studies deformed the planning CT and associated contours to the patient anatomy in CBCT, to create a synthetic CT in order to overcome the effect of inaccuracies in CBCT's CT numbers and limited FOV on dose estimation.[7],[8],[9],[10],[11] Except for Weppler et al.[11] the other studies used single-fraction CBCT to determine a change in dose to OAR or planning target volume (PTV) for need analysis of ART. The intent in these studies varied from simply to establish the feasibility of approach, to validate the anatomical change threshold, and more importantly to objectively determine the patients needing ART. Free-form deformable image registration (DIR) and clinically available systems were used in these studies. Clinical DIR algorithms have been evaluated in benchmarking studies involving a variety of anatomical sites,[12] and for head and neck.[12],[13],[14],[15] These studies applied known deformations to create synthetically deformed CT images and then validated the deformation maps generated by the commercial DIR algorithms against the known deformation maps using the AAPM's TG-132 metrics.[16] The results of these studies have shown that DIR algorithms can be very useful tools, but that care is still required in evaluating the results of DIR algorithms.

In addition to benchmarking evaluation of clinically available DIR methods, several studies have used DIR methods in CT-to-CT contour propagation.[17],[18] Ramadaan et al.[17] evaluated the accuracy of deformed structures compared to physician-drawn contours in CT-to-CT as well as in contrast CT to noncontrast CT cases. A physician-assessed ranking method similar to Hardcastle et al.[19] was used. The authors concluded that there were definite time savings and, in general, propagated contours required minimal editing, but do require a close review. Kumarasiri et al.[18] examined CT-to-CT contour propagation using four different commercial DIR products. Their evaluation involved both quantitative metrics (e.g., Dice similarity coefficient) and qualitative physician rankings. The results showed minimal differences in performance between the different algorithms based on the quantitative metrics, but a physician preference for VelocityAI (v3.0, Velocity Medical Systems, Atlanta, GA, USA) and SmartAdapt® (v11, Varian Medical Systems, Palo Alto, CA, USA). They also reported a 45-min time saving to their replanning process through the use of DIR contour propagation.

Recent studies that estimated dose to OARs and PTVs using CBCT images utilized either SmartAdapt (Eclipse Version 13 or higher, Varian Medical Systems, Palo Alto, CA, USA)[8],[11] or MIM[7],[10] (MIM version 6.4 or higher, MIM Software Inc., Cleveland, OH, USA) or NiftyReg,[9] an open-source registration package. At our institute, two other CBCT-based methods of dose changes in head-and-neck patients have become available: VelocityAI™ (Version 4.1, Varian Medical Systems, Palo Alto, CA, USA) and PerFraction™ (Sun Nuclear Corporation, Melbourne, FL, USA). Although PerFraction™ does not offer DIR, it does provide automation in dose change estimation and thus it is worth investigating.

The aim of this work is to study SmartAdapt®, VelocityAI™, and PerFraction™ in their ability to use CBCT images to estimate dose changes in head-and-neck cancer patients, and compare the results to those obtained using the repeat CT images. The study also evaluates the fidelity of the spinal cord deformation and its impact on the dosimetric analysis.


   Materials and Methods Top


Patients

We retrospectively selected 17 head-and-neck cancer patients treated from March to October 2020. All patients were replanned due to the gaps under the immobilization mask causing poor immobilization and/or dosimetric concerns. The final decision to replan was made by the treating radiation oncologist either based solely on the efficacy of immobilization (n = 10) or based on both the efficacy of immobilization and estimated increase in dose to OARs such as spinal cord or reduction in dose to PTV (n = 7). Dose estimates to OARs and PTV (i.e., dosimetric review), resulting from the change in patient anatomy as visible in CBCT, were carried out by a clinical physicist. Replanning involves making new immobilization, repeat CT simulation, treatment planning, and quality assurance. The selected patients had differing tumor sites: right tonsil (n = 7), base of tongue (n = 2), hypopharynx (n = 2), larynx (n = 2), oropharynx (n = 2), left tonsil (n = 1), and thyroid (n = 1). Initial CT simulation to treat days ranged from 10 to 34 (average = 19) and initial CT simulation to repeat CT simulation days ranged from 29 to 61 (average = 44). The factors that contributed to treatment delays were not exhaustively explored in this retrospective study. This study was determined to be of minimal risk and consistent with a quality improvement project using the institutional ethics screening tool and did not require further ethics board approval.

Treatment planning and treatment

All head-and-neck cancer patients are immobilized using a 5-point thermoplastic mask (Efficast, Orfit Industries, Wijnegem, Belgium) locking the patients' shoulder, face, and head onto a carbon fiber board (AIO board, ORFIT Industries, Wijnegem, Belgium). Patients were CT scanned from top of the head to carina using ≤600 mm axial FOV (512 × 512 pixel matrix), 3 mm thick contiguous slices, and 120 kVp (Brilliance Big Bore, Philips Healthcare, Cleveland, OH, USA). Volumetric modulated arc therapy plans containing two or three arcs and ACUROS (Eclipse version 13.6, Varian Medical Systems, Palo Alto, CA, USA) dose calculation algorithm are used for treatment planning. All patients except one had a low-dose PTV (PTV54: 95% of volume to 54 Gy) and a high-dose PTV (PTV60 or PTV66: 95% of volume to either 60 or 66 Gy) treated in 30 fractions. Plan optimization used constraints to facilitate sharp dose falloff away from the PTV as well as standard constraints on OAR, e.g., max dose to contoured spinal cord <45 Gy (Cord_dmax) and mean dose to each contoured parotid <26 Gy.

Head-and-neck cancer patients were treated on Varian's TrueBeam (version 2.7) platform equipped with kilovoltage (kV) CBCT and a 6 degrees-of-freedom (6DOF) couch. After initial patient setup based on reference tattoos and linear translations to isocenter, a CBCT is taken using the standard thorax protocol (axial FOV = 465 mm, 512 × 512 pixel matrix, contiguous 2 mm thick slices, 125 kV, and sup-inf FOV ≈ 170 mm). The FOV is generally insufficient to include the entirety of the contoured patient anatomy, with the anatomy being clipped inferiorly and portions of parotids superiorly being outside of the scanned volume. Although Varian's TrueBeam allows extended-length CBCT by stitching two separate CBCTs together to alleviate the issue of missing anatomy, this was not used clinically for head-and-neck cancer patients at our institute in 2020. A rigid registration between the planning CT and CBCT volumes, using 6DOF, is performed. If the linear shifts are <5 mm and the angular shifts are <2°, the shifts are applied and the patient is treated. Otherwise, a second CBCT is taken after shifting the patient, and the ensuing 6 DOF match is applied.

Radiation therapists monitor the immobilization for both effectiveness and gaps under the mask initiating a dosimetric review if needed.

Dosimetric review

A direct dose calculation on CBCT images is not accurate enough to provide adequate sensitivity to estimate the, usually small, changes in dose in these patients. Thus, we either use a difference of dose calculated on fraction 1 and fraction N (±1 day of repeat CT) CBCTs or a synthetic CT approach to utilize the CBCT for dose assessment.

Difference of dose on two cone-beam computed tomographies

This approach used the PerFraction™ software within SunCheck version 3.1.1.4723. PerFraction™ replaces the planning CT image data, within the FOV of CBCT, with the CBCT image data using the online 6DOF rigid registration. Each treatment unit is commissioned in SunCheck using the open-field 3D dose data for several square field sizes, reference calibration, and output factors. The CT number to electron density tables of each CBCT was measured using 30 cm diameter cylindrical solid water phantom, long enough to fill the longitudinal FOV of CBCT and containing plugs of known electron density (Model 467 Tissue Characterization Phantom, Gammex, Middleton, WI, USA). The software then calculates 3D dose and dose–volume histograms using the composite of CBCT and planning CT data, planning contours, and the log files of the treatment session.

Using the PerFraction™ software, 3D dose was calculated on CBCTs of fraction 1 and fraction N. The change in Cord_dmax and D95 of low- and high-dose PTVs was then determined between the two fractions. These changes in dose were then assumed to be changes in planning doses. This approach bypassed the errors due to inaccurate CT numbers in CBCT by assuming such inaccuracies to be similar in both fractions. Fraction 1 CBCT for all patients was reviewed and the anatomical change from planning CT was not significant. The PerFraction™ software uses the 6DOF online registration to position the CBCT image data within the planning CT image data, and as such contains errors due to the shortcomings of rigid registration, and differences in online registration between two fractions.

Synthetic computed tomography – Eclipse SmartAdapt®

Eclipse SmartAdapt® uses a modified demons-based DIR,[20] and it has been used in benchmarking studies,[12],[13],[14],[15] in contour propagation studies,[8],[18],[21],[22] and in weight loss analyses.[8],[11],[12] Eclipse SmartAdapt® was used to deform planning CT to the CBCT of fraction N. This process also deforms the OARs and PTVs' contours in planning CT, and creates a complete 3D “synthetic CT” with structure contours. However, a rigid registration of planning CT to CBCT must precede the deformable registration by SmartAdapt®. When we used the online 6DOF match for rigid registration, the resulting “synthetic CT” was not accepted by Eclipse for planning as the Digital Imaging and Communications in Medicine (DICOM) header containing rotations indicated nonaxial CT slices. In order to include rotations in online 6DOF match within the rigid registration prior to DIR, and allow dose calculation within Eclipse, we used an in-house MATLAB code. This code translated and rotated CBCT image data, as per online 6DOF match, without modifying the DICOM meta-data. The planning CT was first rigidly registered to this modified CBCT without rotations, followed by a deformable registration by SmartAdapt® to create the synthetic CT including the contours. The original treatment plan was then copied and isocenter anatomically aligned in the synthetic CT, and 3D radiation dose was calculated using the original MLC pattern and monitor units of each beam. The Cord_dmax and D95 of PTV54 and PTV60/PTV66 were recorded. In addition, the volumes of cord, PTV54, and PTV60/PTV66 were also recorded in the synthetic CT.

Synthetic computed tomography – VelocityAI™ navigator

VelocityAI™ is a stand-alone image archival and registration software with navigators designed for various tasks. The single-plan generation navigator (ACTOR) was used to deform the initial planning CT and structure sets (i.e., synthetic CT) to the patient anatomy in CBCT images. Initially, the ACTOR navigator used a rigid registration based on the online 6DOF registration. Rigid registration was followed by “deformable multi-pass registration” to create the synthetic CT, a deformed structure set, and a copy of original plan. The deformable multi-pass algorithm is a multi-resolution B-spline algorithm and has been validated in several studies.[13],[14],[15],[23],[24] This process created a copy of original plan with the isocenter aligned with the deformed anatomy, keeping the isocenter anatomically the same as in the original plan. Following a dose recalculation using the copy of original plan, similar to Eclipse SmartAdapt® method, all doses and volumes were recorded. The rotational component of online 6DOF registration was included in the VelocityAI™ process and Eclipse was able to calculate dose on the ensuing “synthetic CT,” so there was no need to run the MATLAB code in this case.

Repeat computed tomography reference

As this is a retrospective review with little control over the reproducibility of patient setup, we found significant differences in head rotation and neck curvature of patients between original planning CT and the repeat CT. In addition, there were significant differences in scan length, and the superior-inferior aspect of cord contour. The target volumes were recontoured by the radiation oncologist in the repeat CT, and these also differed from the initial planning CT. Thus, the original plan placed on the repeat CT did not create a good reference of dosimetric changes to compare with the CBCT-based methods. The VelocityAI™-based method, which used the repeat CT as a new anatomical reference, also created a reference synthetic CT by deforming the original CT and structures based on the new anatomical reference. The initial plan was anatomically aligned with the reference synthetic CT, and dosimetric parameters and volumes recorded similar to the CBCT-based methods.

Statistical analysis

The differences in dose parameters and volumes noted above were calculated for each of the aforementioned four methods from the original plan. We used a two-tailed Wilcoxon signed-rank test at a significance level α =0.05 to determine if the observed differences were statistically different from the original plan. The same test was used to determine the statistical significance of differences, if any, observed between the repeat CT reference method and the CBCT-based methods.


   Results Top


[Figure 1] shows the comparison of spinal cord volume between the original planning CT, repeat CT, and the other analysis methods. Repeat CT and VelocityAI™ CBCT show a mean difference of 0.67% and −1.07%, respectively, and both are not statistically significant (P > 0.05). The Eclipse SmartAdapt® method, on the other hand, shows a statistically significant (P = 0.0012) mean increase in volume of 9.56%. Sun Nuclear's PerFraction™ uses the same contours as the original planning CT, so the volume is unchanged as there was no DIR used. It is important to note that no change in spinal cord volume was expected between planning CT, repeat CT, and CBCT-based synthetic CT.
Figure 1: Change in spinal cord volume compared to original planning CT. Eclipse SmartAdapt® method showed a statistically significant expansion of spinal cord by 9.56% compared to original plan. CT: Computed tomography, CBCT: Cone-beam computed tomography

Click here to view


[Figure 2] shows the percentage difference in Cord_dmax for repeat CT and fraction N CBCT as evaluated by the different methods compared to the original plan. All changes were statistically significant (P < 0.005) with average differences of 3.85%, 6.77%, 4.09%, and 0.78% for repeat CT, Eclipse SmartAdapt®, VelocityAI™, and PerFraction™, respectively. Compared to the repeat CT, Eclipse SmartAdapt® overestimated Cord_dmax by an average of 2.76% (P < 0.01) while PerFraction™ underestimated by an average of 2.87% (P < 0.003). Meanwhile, there was no statistically significant difference in Cord_dmax between the repeat CT and VelocityAI™-based CBCT analyses (P = 0.904).
Figure 2: Change in maximum cord dose compared to the original planning CT. All changes were statistically significant with average increase of 3.85%, 6.77%, 4.09%, and 0.78%, respectively, for repeat CT-, Eclipse SmartAdapt®-, VelocityAI™-, and PerFraction™-based methods. CT: Computed tomography, CBCT: Cone-beam computed tomography

Click here to view


Analysis of the change in volume for both the high-dose PTV [Figure 3] and the low-dose PTV [Figure 4] shows a statistically significant (P < 0.02) reduction in volume for both structures between the planning CT and fraction N for each of the repeat CTs and CBCTs in Eclipse SmartAdapt® and in VelocityAI™ methods. Furthermore, Eclipse SmartAdapt® shows a statistically significant (P = 0.014) reduction in volume of high-dose PTV [Figure 3] compared to repeat CT while VelocityAI™ shows no statistically significant difference from the repeat CT (P = 0.53). For the low-dose PTV [Figure 4], the differences between all three methods were not statistically significant (P > 0.14); however, the average volume reduction is still larger in SmartAdapt®. Sun Nuclear's PerFraction™ does not register any reduction in volumes as it performed no DIR, and so used the original structures due to rigid registration.
Figure 3: Change in volume between original planning CT and fraction N for high-dose PTV. PerFraction™ uses the original contours, the other three methods showed a statistically significant reduction in PTV volumes consistent with weight loss. CT: Computed tomography, CBCT: Cone-beam computed tomography, PTV: Planning target volume

Click here to view
Figure 4: Change in volume between original planning CT and fraction N for low-dose PTV. PerFraction™ uses the original contours, the other three methods showed a statistically significant reduction in PTV volumes consistent with weight loss. CT: Computed tomography, CBCT: Cone-beam computed tomography, PTV: Planning target volume

Click here to view


All methods also show a statistically significant (P < 0.05) reduction in D95 of both PTV [Figure 5] and [Figure 6] structures between the planning CT and fraction N, although Sun Nuclear's PerFraction™ P value was 0.0414 compared to other methods P < 0.004. In addition, no statistically significant difference was found between the four methods for either PTV structure.
Figure 5: Change in dose to 95% of the PTV volume between original planning CT and fraction N for high-dose PTV. All methods showed a statistically significant reduction in PTV D95. CT: Computed tomography, CBCT: Cone-beam computed tomography, PTV: Planning target volume

Click here to view
Figure 6: Change in dose to 95% of the PTV volume between original planning CT and fraction N for low-dose PTV. All methods showed a statistically significant reduction in PTV D95. CT: Computed tomography, CBCT: Cone-beam computed tomography, PTV: Planning target volume

Click here to view


A “need to replan” analysis was carried out, using the repeat CT as the reference, separately using Cord_dmax, and D95 of PTVs with the following definitions of true positive, true negative, false positive, and false negative. When a reference plan based on repeat CT results in Cord_dmax ≥45 Gy, or D95 of any of the PTV being less than its prescribed dose, it was considered a true positive for needing a replan. When a reference plan based on repeat CT does not violate maximum dose to cord or prescribed dose to PTVs, then it was considered a true negative. When a reference plan base on repeat CT does not indicate a need for replan, but a CBCT-based method indicates a replan, it was considered a false positive and vice versa for false negative.

[Table 1] summarizes the need to replan, based Cord_dmax, indicated by the reference plan using repeat CT and corresponding CBCT dose analysis methods. In each case, maximum dose to spinal cord was rounded up if it is >44.5 Gy. [Table 2] and [Table 3] give the same data, separately, for PTV66 and PTV54, respectively.
Table 1: Need to replan based on maximum dose to cord being ≥45 Gy (n=17)

Click here to view
Table 2: Need to replan based on dose to 95% of planning target volume 66 being <66 Gy (n=10)

Click here to view
Table 3: Need to replan based on dose to 95% low-dose planning target volume being <54 Gy (n=16)

Click here to view



   Discussion Top


Several studies have investigated CBCT-based dosimetric analysis to determine the “need to replan” and compared it to dosimetric analysis performed using repeat CT simulation.[8],[9],[10],[22] In some of these studies, the repeat CT simulation, acquired midway through the course of treatment, is preplanned, while in others, the repeat CT was acquired due to anatomic or patient position changes. Most of these studies either repeat CT in the original immobilization or took great care in reproducing the patient position in the repeat CT simulation. Therefore, a repeat CT provided a ground truth to which the CBCT-based dosimetric analysis was compared. In the present retrospective study, all patients had a new immobilization and there was little attempt to reproduce the prior patient position since the decision to replan had already been taken, with or without a dosimetric review from medical physics. This study used DIR of the planning CT based on the repeated CT to create a ground truth. The intent here is to determine which approach of CBCT-based dosimetric analysis most closely resembles the one based on the repeat CT.

Fraction N CBCT was reviewed for all patients in offline review within Eclipse TPS. All patients in this group had weight loss that caused reductions in body contour ranging from 5 mm to 15 mm in CBCT images, compared to planning CT. Similarly, fraction 1 CBCT also reviewed to ensure no significant reduction in body contour has occurred. For example, the average change in PTV54 volume from CT simulation to fraction 1 was 0.2 cm3 (range: −5.6 to +6.6 cm3); this is not statistically different (P = 0.589) from no change. While the average change in PTV54 volume from CT simulation to fraction N was −3.5 cm3 (P < 0.05, range: −12.9–1.9 cm3). This indicates treatment-related weight loss. In a recent analysis, Nazari et al.[25] found that 79% of head-and-neck patients had treatment-related weight loss. The “need to replan,” assessed based on Cord_dmax, using the repeat CT and Velocity™-based CBCT analyses is quite similar; VelocityAI™-based CBCT analysis missed only one needed replan while creating one extra replan. Eclipse SmartAdapt®-based CBCT analysis missed one needed replan, and created three extra replans all of which result directly from nonphysical cord volume expansion [Figure 1]. This result contradicts Belshaw et al.[8] who did not find SmartAdapt® to create any false positive or false negative. Region of interest in SmartAdapt® and VelocityAI™ DIR in our study included the entire FOV of CBCT to include both the low- and high-dose PTVs; one or two slices suffering from cone artifact at the inferior end in CBCT were excluded. Belshaw et al.[8] may have only included the high-dose PTV, and only used rigid registration between planning and repeat CTs to create a ground truth reference assuming patient position being similar in two CT scans. SmartAdapt®-based DIR resulted in volume expansion by more than 10% in 13 out of 17 patients. While studies involving CBCT-based dose analysis using SmartAdapt®[8],[11],[22] have not focused on spinal cord volume changes, a benchmarking study[13] showed the lowest Dice similarity coefficient (DSC) for SmartAdapt® in Eclipse compared to other DIR methods. In a SmartAdapt® only CT-to-CT contour propagation study[17] [Table 2], spinal cord had the lowest DSC of all regions considered which is also supported in another study[18] [Table 2] and [Table 3]. Thus, there is indirect evidence for the poor performance of SmartAdapt® in deforming spinal cord resulting from changing neck curvature.

The perFraction™ method, as implemented and tested here, did not suggest any patient should be replanned based on Cord_dmax. PerFraction™ uses online registration to replace planning CT data with the CBCT data and calculates dose using the dedicated CT number to electron density calibration. It should be noted that the Cord_dmax was <44.5 Gy for fraction N itself for all patients. Thus, single-fraction dose calculation using PerFraction™ also did not indicate any “need to replan” in these patients. Considering poorer accuracy of CT numbers (compared to calibration phantoms) in patients, we decided to estimate the change in cord dose in fraction N from fraction 1; however, this approach suffers from differences in spinal cord position between two fractions, as shown in [Figure 7]. [Figure 7] also demonstrates the incorrect reduction in CT numbers of trabecular bone and spinal canal in the inferior region of CBCT due to shoulders. A simple region of interest analysis to determine the mean CT numbers of spinal canal and trabecular bone, in the inferior region of CBCT images, resulted in relative electron densities of 1.08 for trabecular bone (planning CT relative electron density = 1.2) and 0.87 for spinal canal (planning CT relative electron density of 1.05).
Figure 7: Typical example of spinal cord position in PerFraction™ (online match) during fraction 1 (left) and fraction N (right)

Click here to view


All methods showed a statistically significant reduction in D95 to the high-dose PTV with P < 0.05. However, there was no statistically significant difference between repeat CT- and CBCT-based methods (all P > 0.267). This similarity did not translate into similar “need to replan” as shown in [Table 2] where PerFraction™ method resulted in the most false negatives, followed by SmartAdapt®-based CBCT method, compared to repeat CT. The error in online registration and its difference between two fractions still plagues the PerFraction™ method in PTV66 registration. Moreover, PerFraction™ method does not account for volume reduction in the PTV due to weight loss. PerFraction™ does not determine new body contour resulting from weight loss, thus PTVs reaching near to patient surface in planning CT generally reside in air in CBCT, but PerFraction™ still deposits enough dose in this region without losing buildup. [Figure 3] shows the high-dose PTV volume reduced significantly (P < 0.002) in repeat CT and CBCT-based methods. In addition, the high-dose PTV volume reduced more in SmartAdapt®-based method nearly by 3% (P = 0.014) compared to repeat CT which may account for increased false negatives. Similar observations regarding the need to replan based on the lost coverage of PTV54 can be made from [Table 3].

Our ground truth reference using repeat CT found 8 out of 17 patients needing replan based on PTV66 which is significantly larger than found in Vickress et al.[10] Vickress et al. did not report a reduction in high-dose PTV volume. If the high-dose PTV does not reach near patient surface, i.e., interior PTV, moderate weight loss may not impact dose to the same extent as in our study. This study also did not report whether a new immobilization was used in repeat CT or if care was taken to reproduce the patient position from planning CT to repeat CT which may affect DIR-based contour propagation.


   Conclusions Top


CBCT-based dosimetric analysis using the VelocityAI™'s B-spline method provides similar results to a reference plan created using repeat CT in terms of change in maximum dose to spinal cord, and dose coverage of low- and high-dose PTVs for head-and-neck patients with weight loss. A similar method based on the demon's algorithm with trade name SmartAdapt® in Eclipse TPS gives much higher increase in maximum dose to cord due to its expansion of cord in deformable registration. It gives inferior match to repeat CT-based reference plan in terms of low- and high-dose PTV coverage. Direct CBCT-based dose calculation from fraction 1 to fraction N as provided by PerFraction™ method failed to identify “need to replan” based on increase in maximum dose to cord, possibly due to differences in online rigid registration between fraction 1 and fraction N.

Financial support and sponsorship

Nil.

Conflicts of interest

There are no conflicts of interest.

 
   References Top

1.
Wang L, Zhang S, Yuan K, Zhou L, Wang R. Dosimetric Study of the Deformable Image Registration Based Lung Cancer Adaptive Radiotherapy: 8th International Conference on Biomedical Engineering and Informatics; 2015. p. 402-6.  Back to cited text no. 1
    
2.
Mohan R, Zhang X, Wang H, Kang Y, Wang X, Liu H, et al. Use of deformed intensity distributions for on-line modification of image-guided IMRT to account for interfractional anatomic changes. Int J Radiat Oncol Biol Phys 2005;61:1258-66.   Back to cited text no. 2
    
3.
Hutton J, McCurdy B, Greer PB. Cone beam computerized tomography: The effect of calibration of the Hounsefield unit number to electron density on calculation accuracy for adaptive radiation therapy. Phys Med Biol 2009;54:N329-46.  Back to cited text no. 3
    
4.
Li J, Yao W, Xiao Y, Yu Y. Feasibility of improving cone-beam CT number consistency using a scatter correction algorithm. J Appl Clin Med Phys 2013;14:4346.  Back to cited text no. 4
    
5.
Chen S, Le Q, Mutaf Y, Lu W, Nichols EM, Yi BY, et al. Feasibility of CBCT-based dose with a patient-specific stepwise HU-to-density curve to determine time of replanning. J Appl Clin Med Phys 2017;18:64-9.  Back to cited text no. 5
    
6.
MacFarlane M, Wong D, Hoover DA, Wong E, Johnson C, Battista JJ, et al. Patient-specific calibration of cone-beam computed tomography data sets for radiotherapy dose calculations and treatment plan assessment. J Appl Clin Med Phys 2018;19:249-57.  Back to cited text no. 6
    
7.
Christian AH, Ulrik VE, Kenneth J, Cai G. Cone-beam computed tomography (CBCT) for adaptive image guided head and neck radiation therapy. Acta Oncol 2018;57:552-6.  Back to cited text no. 7
    
8.
Belshaw L, Agnew CE, Irvine DM, Rooney KP, McGarry CK. Adaptive radiotherapy for head and neck cancer reduces the requirement for rescans during treatment due to spinal cord dose. Radiat Oncol 2019;14:189.  Back to cited text no. 8
    
9.
Veiga C, McClelland J, Moinuddin S, Lourenco A, Ricketts K, Annkah J, et al. Toward adaptive radiotherapy for head and neck patients: Feasibility of using CT-to-CBCT deformable registration for “dose of the day” calculations. Med Phys 2014;41:031703-1 - 031703-12.   Back to cited text no. 9
    
10.
Vickress JR, Battista J, Barnett R, Yartsev S. Online daily assessment of dose change in head and neck radiotherapy without dose-recalculation. J Appl Clin Med Phys 2018;19:659-65.  Back to cited text no. 10
    
11.
Weppler S, Quon H, Banerjee R, Schinkel C, Smith W. Framework for the quantitative assessment of adaptive radiation therapy protocols. J Appl Clin Med Phys 2018;19:26-34.  Back to cited text no. 11
    
12.
Loi G, Fusella M, Lanzi E, Cagni E, Garibaldi C, Lacoviello G, et al. Performance of commercially available deformable image registration platforms for contour propagation using patient-based computational phantoms: A multi-institutional study. Med Phys. 2018;45:748-57. https://doi.org/10.1002/mp.12737.  Back to cited text no. 12
    
13.
Loi G, Fusella M, Vecchi C, Menna S, Rosica F, Gino E, et al. Computed tomography to cone beam computed tomography deformable image registration for contour propagation using head and neck, patient-based computational phantoms: A multicenter study. Practical Radiation Oncology. 2020;10:125-32.   Back to cited text no. 13
    
14.
Kubli A, Pukala J, Shah AP, Kelly P, Langen KM, Bova FJ, et al. Variability in commercially available deformable image registration: A multi-institution analysis using virtual head and neck phantoms. J Appl Clin Med Phys 2021;22:89-96.  Back to cited text no. 14
    
15.
Pukala J, Johnson PB, Shah AP, Langen KM, Bova FJ, Staton RJ, et al. Benchmarking of five commercial deformable image registration algorithms for head and neck patients. J Appl Clin Med Phys 2016;17:25-40.  Back to cited text no. 15
    
16.
Brock KK, Mutic S, McNutt TR, Li H, Kessler ML. Use of image registration and fusion algorithms and techniques in radiotherapy: Report of the AAPM Radiation Therapy Committee Task Group No. 132. Med Phys 2017;44:e42-76.  Back to cited text no. 16
    
17.
Ramadaan IS, Peick K, Hamilton DA, Evans J, Iupati D, Nicholson A, et al. Validation of Varian's SmartAdapt® deformable image registration algorithm for clinical application. Radiat Oncol 2015;10:73.  Back to cited text no. 17
    
18.
Kumarasiri A, Siddiqui F, Liu C, Yechieli R, Shah M, Pradhan D, et al. Deformable image registration based automatic CT-to-CT contour propagation for head and neck adaptive radiotherapy in the routine clinical setting. Med Phys. 2014;41:121712-1–121712-10.   Back to cited text no. 18
    
19.
Hardcastle N, Tomé WA, Cannon DM, Brouwer CL, Wittendorp PW, Dogan N, et al. A multi-institution evaluation of deformable image registration algorithms for automatic organ delineation in adaptive head and neck radiotherapy. Radiat Oncol 2012;7:90.  Back to cited text no. 19
    
20.
Wang H, Dong L, O'Daniel J, Mohan R, Garden AS, Ang KK, et al. Validation of an accelerated demons algorithm for deformable image registration in radiation therapy. Phys Med Biol. 2005;50:2887-905.   Back to cited text no. 20
    
21.
Thörnqvist S, Petersen JBB, Hoyer M, Bentzen LN, Muren LP. Propagation of target and organ at risk contours in radiotherapy of prostate cancer using deformable image registration. Acta Oncologica. 2010; 49:1023-32.   Back to cited text no. 21
    
22.
Mao W, Rozario T, Lu W, Gu X, Yan Y, Jia X, et al. Online dosimetric evaluation of larynx SBRT: A pilot study to assess the necessity of adaptive replanning. J Appl Clin Med Phys 2017;18:157-63.  Back to cited text no. 22
    
23.
Wu RY, Liu AY, Williamson TD, Yang J, Wisdom PG, Zhu XR, et al. Quantifying the accuracy of deformable image registration for cone-beam computed tomography with a physical phantom. J Appl Clin Med Phys 2019;20:92-100.  Back to cited text no. 23
    
24.
Kuznetsova S, Grendarova P, Roy S, Sinha R, Thind K, Ploquin N. Structure guided deformable image registration for treatment planning CT and post stereotactic body radiation therapy (SBRT) Primovist® (Gd-EOB-DTPA) enhanced MRI. J Appl Clin Med Phys 2019;20:109-18.  Back to cited text no. 24
    
25.
Nazari V, Pashaki AS, Hasanzadeh E. The reliable predictors of severe weight loss during the radiotherapy of Head and Neck Cancer. Cancer Treat Res Commun 2021;26:100281.  Back to cited text no. 25
    


    Figures

  [Figure 1], [Figure 2], [Figure 3], [Figure 4], [Figure 5], [Figure 6], [Figure 7]
 
 
    Tables

  [Table 1], [Table 2], [Table 3]



 

Top
Print this article  Email this article
  

    

 
   Search
 
  
    Similar in PUBMED
   Search Pubmed for
   Search in Google Scholar for
 Related articles
    Article in PDF (807 KB)
    Citation Manager
    Access Statistics
    Reader Comments
    Email Alert *
    Add to My List *
* Registration required (free)  


    Abstract
   Introduction
    Materials and Me...
   Results
   Discussion
   Conclusions
    References
    Article Figures
    Article Tables

 Article Access Statistics
    Viewed180    
    Printed2    
    Emailed0    
    PDF Downloaded39    
    Comments [Add]    

Recommend this journal