Introduction

Volumetric Modulated Arc Therapy (VMAT) is considered an established treatment modality for prostate cancer cases [ 1 ] and is also used in the management of prostate cancer to maximize the tumor dose while minimizing radiation exposure to nearby organs with the assumption of static patient geometry. However, organ motion, weight anatomic, and setup changes can be noticed during the treatment course. The geometrical variations during the treatment course could lead to dosimetric variations, such as overdosage or underdosage to critical organs of risks and treatment the targets since conformality index is a function of patient geometry [ 2 ]. As a result, in-room: Kilovoltage Cone Beam Computed Tomography (kVCBCT) was developed to verify whether the intended dose is precisely delivered to the treatment target and critical Organ at Risk (OAR). Radiation therapists can detect any differences between this imaging set and the planning Computed Tomography (pCT) using kVCBCT before treatment, resulting in a decision whether the treatment plan is progressing as scheduled or needs replanning, or if the repositioned patient. In addition to kVCBCT, Deformable Image Registration (DIR) is introduced in clinical radiotherapy to evaluate the dosimetric consequences of anatomic prostate tumor variations by registering the pCT with the kVCBCT in a process called Adaptive Radiation Therapy (ART) [ 2 ]. “ART is defined as a close-loop, iterative process, in which the treatment plan is modified based on feedback measurements performed during treatment” [ 3 ]. Accordingly, DIR as a fundamental tool in the radiotherapy field evaluates geometrical deformations of treatment targets and critical organs with day-to-day Image-guided Radiation Therapy (IGRT) [ 4 - 5 ]. For instance, the DIR system is used in ART to compute the accumulated dose through re-planning on daily kVCBCT images, which are acquired immediately before the treatment fraction. It is also used to translate the dose, propagate the contour, and compute the accumulated dose for intra-fraction radiotherapy using a Distance Vector Field (DVF) . In radiotherapy, deformable registration can spare normal tissue and improve target dose coverage. However, image artifacts are associated with all imaging modalities, due to the difference between the mathematical assumptions and the experimental setup in the algorithm of image reconstruction. The Feldkamp algorithm, a reconstruction algorithm for kVCBCT, is adapted from the Filtered Back Projection (FBP) algorithm in the reconstruction of fan beam CT. In addition, this algorithm for kVCBCT shows that image quality is a function of the distance between the central and peripheral planes. A larger Field of View (FOV) results in a larger quantity of detected scatter radiation, significantly degrading image quality. The use of kVCBCT images for dose calculation led to inaccurate kVCBCT-based dose calculation due to the artifact effects, resulting in inaccurate HU values. The hardware-based corrections, such as anti-scatter X-ray grids on the detector, collimators, bowtie filters, and deformable image registration are adapted to minimize image artifacts [ 7 ].

In this study, an open-source DIR software was used in combination with pCT, daily kVCBCT images, and the Monaco Treatment Planning System (TPS) from Elekta Centers for Medicare & Medicaid Services (CMS) in Maryland Heights, MO, USA. The present study aimed to evaluate the effects of anatomic variations during prostate VMAT on the delivered dose to both the treatment target and critical organs at risk.

Material and Methods

This experimental study evaluates the dosimetric influence of anatomical changes in prostate cancer treated with VMAT using daily kVCBCT.

Image acquisition

Eight patients, who received VMAT treatment for prostate cancer in 2021, participated in the study, and each patient had one pCT and one daily in-room kVCBCT, using X-ray Volume Imaging (XVI) on the Versa-HD Elekta X-ray volume imaging system (Elekta, Crawley, UK). The kVCBCT images were acquired during the first fraction using the XVI in a standard pelvic acquisition mode. Table 1 presents the details of the data acquisition for pCT and kVCBCT.

Design of radiotherapy plan

The designs of radiotherapy plan for pCT and deformed Computed tomography (dCT) are summarized in Table 2.

Deformable Image Registration

During the first treatment fraction, the pCT and the in-room kVCBCT were manually exported from the TPS. The original pCT included segmentations of the treatment targets and critical structures at risk. The data were then imported into an open-source deformable image registration software, which is called Deformable Image Registration and Adaptive Radiotherapy (DIRART) (https://github.com/krishprince/dirart). The voxel dimensions of the kVCBCT images and the pCT images were different; all image datasets were resampled using trilinear interpolation to achieve a uniform voxel dimension of 1×1×3 mm³.

In the DIRART software, the optical flow algorithm was verified for the registration of different imaging modalities and also used for deformable pCT and kVCBCT image registration using two-step registration methods [ 5 , 9 ]. The pCT and kVCBCT images were used as the moving and target images, respectively. The pCT image was then deformed to the kVCBCT image using the DIR system, involving a rigid registration, followed by a two-step deformable registration of the pCT to each kVCBCT while contouring the critical organs (Eq 1) [ 5 ]:

$\begin{array}{cc}I=I-I_{\mathrm{average}}+C& \mathrm{(1)}\end{array}$

where I_average and C represent the averaged image intensity and a constant value of 800 for the rectum and bladder, and 1200 for subcutaneous fat and the target organ.

First, the rigid registration was performed, and then the deformable vector field was acquired. The acquired DVF was used to deform and propagate the scan to the kVCBCT. Figure 1 shows the process of deformable registration for a patient. All dCT and kVCBCT corresponding structures were saved in DICOM format and imported into the TPS. The original treatment plan was then copied over, and the dose of the critical organs was evaluated on the dCT.

Figure 1. The overall workflow for the combined method: (1) importing the kVCBCT images and the pCT images into deformable image registration software, with different voxel dimensions; all image datasets were resampled using trilinear interpolation to achieve a uniform voxel 1×1×3 mm³. Treatment targets and critical organs at risk were contoured once resampling images. The voxels of these structures were then updated using the CT number of bones. Finally, the whole pCT image was deformed by using the Displacement Vector Field (DVF) from the first step. (CT: Computed Tomography, KVCBCT: kV Cone Beam Computed Tomography, pCT: planning Computed Tomography)

DIR evaluation

The registration accuracy was evaluated using the 95% Hausdorff Distance (HD95) and Dice Similarity Coefficient (DSC) indices according to TG132 [ 10 ]. HD95 is used to measure the distances between two subsets of a metric space. The 95^th percentile HD (HD95) is used instead of the 100^th percentile HD (HD100) to eliminate the subsets of outliers. A lower HD95 value indicates a higher correspondence between the deformed and kVCBCT-based structures. Jassim et al. (2023) recommended that the HD95 value should be within 0.3 cm for ART of the pelvis [ 8 ], which is calculated using the function defined in the Computation Environment for Radiotherapy Research (CERR) [ 11 ].

The DSC score is a measure of the overlap between two volumes (X, Y), i.e., 1 and 0 show the high and zero degrees overlapping [ 12 ]. The DSC score is ≥0.8 seems proper for adaptive radiotherapy of the pelvis [ 10 ], computed as follows (Eq 2):

$\begin{array}{cc}\mathrm{DSC}=\frac{2|X\cap Y|}{\left|X\right|+\left|Y\right|}& \mathrm{(2)}\end{array}$ [ 12 ].

Plan evaluation and statistical analysis

The prescription dose of the CT plan was 74 Gy delivered over 34 fractions, with the planning objective of distributing the dose over 95% of the PTV volume. The mean dose of the treatment target was evaluated for the bladder and rectum with the same evaluation for both the pCT and dCT plans. Statistical analysis was performed using the Excel Package software (Microsoft® Excel® 2016 MSO (Version 2305 Build 16.0.16501.20074) 64-bit), including descriptive and inferential analyses. Mean, percentage difference, and significant tests were calculated (P and Z values). Paired two-tailed t-tests were used to compare the pCT- and dCT-based planning at a statistical significance of P-value<0.05 and a confidence level of 95%. Mean absolute error (MAE) was also computed to measure the absolute average value of the error between the predicted and reference values. In the current study, the predicted values were mean doses of the dCT-based plan, and the reference values were mean doses of the pCT-based plan. When the MAE is low, the predicted value is considered more accurate. The MAE is computed using a specific equation (3):

$\begin{array}{cc}\mathrm{MAE}=\frac{{\sum }_{\mathrm{i=1}}^n\parallel {\mathrm{CT}}_{\mathrm{pre}}\left(x_i\right)-{\mathrm{CT}}_{\mathrm{ref}}\left(x_i\right)\parallel }{n}& \mathrm{(3)}\end{array}$

where n is the total number of patients (x_i), CT_pre is the parameter calculated in dCT, and CT_ref is the parameter calculated in the pCT image [ 11 ].

Results

Both HD95 and DSC values were computed for both rigid and deformable registration techniques. The rigid registration used pCT and kVCBCT images of the pelvis for the first fraction, while the deformable registration used dCT and kVCBCT images of the pelvis for the first fraction. The body, bladder, rectum, and prostate were contoured for both techniques. The average values of HD95 for the rigid and deformable registrations were 1.2 cm and 0.28 cm for the body, 1.3 cm and 0.27 cm for the bladder, 0.81 cm and 0.29 cm for the rectum, and 0.76 cm and 0.29 cm for the prostate. Figure 2 shows the average values of DSC for the rigid and deformable registrations. As depicted in Figure 2, the DSC values improved with DIR, with the average DSC values exceeding 0.8 for all critical organs.

Figure 2. Evaluation check and comparison of DSC for rigid and deformable image registration. In the left column, DIR improved registration when compared with rigid registration, with DSC values of more than 0.8. (DSC: Dice Similarity Coefficient, DIR: Deformable Image Registration)

Table 3 presents the volume measurements for the treatment target and OAR using both pCT and dCT plans. The extent of tumor shrinkage was not significant for these patients during the interval period between treatment planning and the first fraction of treatment, ranging of 0.05% to 0.064% with an average change of 0.04%. However, variable extents of critical organ variations were more observed in the dCT images relative to the pCT images, with the bladder experiencing a change range of 8.99% to 51.37% and an average change of 29.89%, and the rectum experiencing a change range of 4.50% to 69.01% and an average change of 26.85%. The critical value for the Z-value was 1.96. The volume measurements for the OAR and treatment target were reported in cubic millimeters (mm³).

Figure 3 displays the DVH metrics results versus the dose distributions, which are computed on the PCT anatomy and kVCBCT images respectively. Figure 3 presents mean dose distribution is different between the pCT- and dCT-based planning for all structures. The percentage differences in mean dose were 0.04%, 16.46%, and 24.19% for the target, rectum, and bladder, respectively.

Figure 3. Comparison of the percentage difference between the mean doses of pCT and dCT images for the test datasets of eight patients at the first fraction. (O-Blader: Original Bladder, O-Rectum: Original Rectum, O-Treatment Target: Original Treatment Target, pCT: planning Computed Tomography, dCT: deformed Computed Tomography, PTV: Planning Target Volume)

For the eight prostate cancer cases, the mean dose for the PTV and rectum had an average increase of 1% and 8%, respectively, while the bladder had an average decrease of 4% (P-value<0.05). Table 4 shows the dosimetric changes of the treatment target and critical organs at risk for eight patients as a function of volume changes using the pCT and dCT images with a 1.96 for the critical value of the Z-value.

Discussion

The use of kVCBCT has improved radiotherapy treatments in the context of Adaptive Radiotherapy (ART) and Image-Guided Radiotherapy (IGRT). Health professionals utilize kVCBCT to image the patient directly before treatment to assess changes in anatomy, such as patient weight loss and tumor shrinkage, and account for changes in the target position by adjusting the treatment. The re-simulated CT is used to account for the dosimetric effect of anatomical changes. However, using re-simulated CT is associated with several disadvantages, including increasing in complexity of the patient pathway, concomitant imaging dose to the patient, stressors on patients, waiting lists for patients, and healthcare costs [ 5 ]. Given extensively validating dose recalculation using kVCBCT images, this study aimed to evaluate the dosimetric consequences of pelvic anatomic changes using DIR and kVCBCT.

The calculation of dose distribution in complex irradiation procedures, such as VMAT provides the most direct quality assurance, due to the prediction of the treatment effectiveness and the potential for normal tissue complications. Accurate daily dose calculations lead to monitoring and reporting doses, particularly with dose-guided ART, and computing the dosimetric effect of significant changes in patients’ structures over the treatment course [ 4 , 7 , 13 ]. However, the kVCBCT images were used to calculate the dose and limitations, such as incorrect HU values and poor image quality reduced the strength of dose calculation [ 14 - 15 ]. Therefore, we utilized dCT images, which capture the features of the patient’s structures during treatment to recalculate the dose distribution. The optical flow algorithm applied in the DIRART software, verified during the registration of different imaging modalities, can be used for deformable pCT and kVCBCT image registration using a two-step registration method. Hence, the dosimetric consequences of anatomical change were assessed using kVCBCT and optical flow-based deformable image registration in image reconstruction through the DIRART software package developed using MATLAB.

The use of DIR improved image quality and accurately propagated the structures from pCT to dCT (Figure 2). The DIR procedure was reasonable for dose calculation from kVCBCT. However, this procedure has some disadvantages, as follows: the ROIs and new organ contours were required due to the poor quality of kVCBCT images, the methods are not suitable for online assessment of patient dose calculation and dose distribution changes, and the process takes a long time of approximately 30 minutes, leading to introducing uncertainty factors, such as patient position changes and organ volume variations, and making it suitable only for offline assessment of patient dose calculation and dose distribution changes. The improvement of the image quality of kVCBCT is recommended to reduce the procedure time of kVCBCT-based dose calculation.

In terms of volume changes in prostate cancer cases, six cases of the rectum and four cases of the bladder experienced a volume increase of critical OAR, and two cases of the rectum and four cases of bladder experienced a volume reduction of critical OAR throughout the treatment. As a result, these changes led to dosimetric differences between the pCT- and dCT-based plans (4).

The dosimetric impact of anatomic changes is evaluated using kVCBCT. The dosimetric effect of six-dimensional inter-fraction setup errors, including vertical, longitudinal, lateral, yaw, roll, and pitch errors, was reported using daily onboard kVCBCT in prostate cancer treated with intensity-modulated radiation therapy and VMAT [ 16 ]. However, setup changes, organ motion, weight loss, and structure deformation can occur during the treatment course, resulting in differing the actual dose to the target and OARs from the estimated values. In this study, the mean dose of the treatment target was similar between dCT and pCT plans, with a percentage difference of only 1%. However, for critical organs at risk, particularly the rectum and bladder, significant mean dose changes were observed for these patients, (Figure 3). The dCT plan had a lower mean dose for the rectum compared to the pCT plan, with a percentage difference of 4%. On the other hand, the dCT plan had a higher mean dose for the bladder compared to the pCT plan, with a percentage difference of 8%. Furthermore, visual and quantitative confirmation of the dose differences in the treatment target and critical organs indicated that the bladder had a greater volumetric change compared to the treatment target and rectum (Table 3).

However, kVCBCT has a limited FOV, leading to covering the complete treated anatomy region required for dose calculation, the treated anatomy region was adequately covered in the pelvis (prostate cancer cases). Therefore, only one consecutive kVCBCT scan was acquired with the couch at a specific position. kVCBCT is weak compared to pCT, resulting in not using dose calculation and contouring of tissues on these images. However, both contours and CT numbers can be propagated from the pCT to the kVCBCT images by applying for non-linear registration. Hence, the DIR will influence tissue contour correspondence, affecting accumulated dose, dose distribution, and dose calculation [ 1 , 5 , 6 ]. After importing pCT and kVCBCT, dCT was created with propagating scan, and the features of dCT image are that dCT has the CT number of pCT and geometry of kVCBCT. After producing dCT, it is transferred to the TPS. The dCT-based plan was created in deformed CT by copying pCT-based plan.

Some papers denote that Genito-urinary (GU) toxicity was related to the high mean dose, and rectal toxicity was related to V70 and high doses beyond the threshold doses. This study demonstrated that while the bladder mean dose in the dCT plan was lower than those in pCT plan, the rectum mean dose was higher than those in the pCT plan with a statistical significance. For the bladder, adaptive re-planning could decrease the dose to the rectum and other soft tissue an adaptive re-planning could decrease the dose to the bladder. The weight loss of patients will affect the bladder, rectum, and target volumes. Thus, the effect of body shape change can on the dose distribution needs more investigation [ 17 ].

Higher doses of radiotherapy increase the probability of tumor control and the risk of complications to organs at risk. Therefore, the daily fractional planning dose must be verified to improve the radiotherapy effect. The use of kVCBCT and DIR can enable precise computation of bladder, rectum, target, and normal tissue doses to enhance the fractioned planning dose. The decrease in optimization method time for re-contouring, dose calculation, and adaptive re-planning can cause DIR to become a valuable tool in the adaptive radiotherapy field.

Conclusion

The proposed framework is successfully used for dose guidance and provides the actual dose distribution of the day for VMAT-based delivery of radiation in prostate cancer cases. Therefore, DIR and kVCBCT are used as complementary methods to explore ART, which may improve the therapeutic ratio of this treatment.

Authors’ Contribution

N. Banaae provided a patient data set and revised the manuscript. HA. Nedaie contributed to the ideation and provided the infrastructure required for the work. Gh. Graily provided practical work. HH. Jassim contributed with ideation, editing, revising, and submitting the final version of the manuscript. All authors approved the final version of the manuscript.

Ethical Approval

This work was approved by the ethical committee of the Faculty of Associated of Medical Science, Tehran University of Medical Science, Tehran, Iran with this code: IR.TUMS.IKHC.REC.1400.290.

Informed Consent

All the participants signed the informed consent.

Conflict of Interest

None

References

1. Marin Anaya V, Fairfoul J. Assessing the feasibility of adaptive planning for prostate radiotherapy using Smartadapt deformable image registration. Med Eng Phys. 2019; 64:65-73. DOI | PubMed
2. Rigaud B, Simon A, Castelli J, Lafond C, Acosta O, Haigron P, et al. Deformable image registration for radiation therapy: principle, methods, applications and evaluation. Acta Oncol. 2019; 58(9):1225-37. DOI | PubMed
3. Yan D, Vicini F, Wong J, Martinez A. Adaptive radiation therapy. Phys Med Biol. 1997; 42(1):123-32. DOI | PubMed
4. Oh S, Kim S. Deformable image registration in radiation therapy. Radiat Oncol J. 2017; 35(2):101-11. Publisher Full Text | DOI | PubMed [ PMC Free Article ]
5. Yang D, Chaudhari SR, Goddu SM, Pratt D, Khullar D, Deasy JO, El Naqa I. Deformable registration of abdominal kilovoltage treatment planning CT and tomotherapy daily megavoltage CT for treatment adaptation. Med Phys. 2009; 36(2):329-38. Publisher Full Text | DOI | PubMed [ PMC Free Article ]
6. Qin A, Sun Y, Liang J, Yan D. Evaluation of online/offline image guidance/adaptation approaches for prostate cancer radiation therapy. Int J Radiat Oncol Biol Phys. 2015; 91(5):1026-33. DOI | PubMed
7. Giacometti V, Hounsell AR, McGarry CK. A review of dose calculation approaches with cone beam CT in photon and proton therapy. Phys Med. 2020; 76:243-76. DOI | PubMed
8. Jassim H, Nedaei HA, Geraily G, Banaee N, Kazemian A. The geometric and dosimetric accuracy of kilovoltage cone beam computed tomography images for adaptive treatment: a systematic review. BJR Open. 2023; 5(1):20220062. Publisher Full Text | DOI | PubMed [ PMC Free Article ]
9. Yang D, Brame S, El Naqa I, Aditya A, Wu Y, Goddu SM, et al. Technical note: DIRART--A software suite for deformable image registration and adaptive radiotherapy research. Med Phys. 2011; 38(1):67-77. Publisher Full Text | DOI | PubMed [ PMC Free Article ]
10. 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. Medical Physics. 2017; 44(7):e43-76. DOI
11. Deasy JO, Blanco AI, Clark VH. CERR: a computational environment for radiotherapy research. Med Phys. 2003; 30(5):979-85. DOI | PubMed
12. Dice LR. Measures of the amount of ecologic association between species. Ecology. 1945; 26(3):297-302. DOI
13. Giacometti V, King RB, Agnew CE, Irvine DM, Jain S, Hounsell AR, McGarry CK. An evaluation of techniques for dose calculation on cone beam computed tomography. Br J Radiol. 2019; 92(1096):20180383. Publisher Full Text | DOI | PubMed [ PMC Free Article ]
14. Yang Y, Schreibmann E, Li T, Wang C, Xing L. Evaluation of on-board kV cone beam CT (CBCT)-based dose calculation. Phys Med Biol. 2007; 52(3):685-705. DOI | PubMed
15. Ding GX, Duggan DM, Coffey CW, Deeley M, Hallahan DE, Cmelak A, Malcolm A. A study on adaptive IMRT treatment planning using kV cone-beam CT. Radiother Oncol. 2007; 85(1):116-25. DOI | PubMed
16. Shinde P, Jadhav A, Shankar V, Dhoble SJ. Evaluation of the dosimetric influence of interfractional 6D setup error in hypofractionated prostate cancer treated with IMRT and VMAT using daily kV-CBCT. J Med Imaging Radiat Sci. 2022; 53(4):693-703. DOI | PubMed
17. Laughlin BS, Lo S, Vargas CE, DeWees TA, Van Der Walt C, Tinnon K, et al. Clinical Practice Evolvement for Post-Operative Prostate Cancer Radiotherapy-Part 1: Consistent Organs at Risk Management with Advanced Image Guidance. Cancers (Basel). 2022; 15(1):16. DOI