Radiotherapy dose calculation on KV cone-beam CT image for lung tumor using the CIRS calibration

Abstract

On-board kilovoltage (KV) cone-beam computed tomography (CBCT) images are used predominantly for the setup of patients' positioning. The image data can also potentially be used for dose calculation with the precise calibration of Hounsfield units (HU) to electron density (HU-density). CBCT calibration was analyzed in this study. A clinical treatment planning system was employed for CT and KV CBCT image to dose calculations and subsequent comparisons. Two HU-density tables were generated using the Computerized Imaging Reference Systems (CIRS) phantom. The results showed that a maximum ∼4% dose discrepancy was observed for inserts. The single field isodose curves were very close. The lung clinical patient study indicated that the volume of lung tumor that achieved the prescribed dose in CBCT was lower than in the CT plan. Our study showed that the dosimetric accuracy of CBCT-based dose calculation for lung tumor is acceptable only for the purpose of dosimetric checks with calibration applied. KV CBCT images cannot replace traditional CT images for dose calculation accuracy.

Introduction

Image-guided radiation therapy (IGRT) is the use of frequent imaging during a course of radiation therapy to improve the precision and accuracy of the delivery of treatment.¹ Cone-beam computed tomography (CBCT) systems mounted on the linear accelerator have become available for IGRT. The role of kilovoltage (KV) CBCT is to help reduce interfractional motion and to try to assess patient status (tumor evaluation, adaptive planning).² KV CBCT imaging has shown enough soft tissue contrast and spatial resolution for soft-tissue based setup, but the image quality is affected by the acquisition parameters. In principle, the KV CBCT data set can be used to calculate dose distribution, which means that the planned dose distribution can be evaluated and verified on each treatment day. Methods for calibrating conventional fan-beam CT to electron density have been widely used in clinical dose calculation.³ As a result, the calibration of KV CBCT images for dose calculation is an active area of research.

Previous work has indicated that the imperfect CBCT image quality has minimal impact (3%) on the dosimetric accuracy, even if the intrafraction organ motion is small.⁴ Therefore, dosimetric discrepancy, with respect to the lung CT-based plan, exists for large size subjects and respiratory motion because of the significant image artifacts and Hounsfield units (HU) number inaccuracy. The variation in the Hounsfield units and electron density (HU-density) curve could induce a non-negligible dose error.⁵

The purpose of this study is to assess the feasibility of using the Computerized Imaging Reference Systems (CIRS) phantom to calibrate kV CBCT images for dose calculation in lung tumor patients.

Materials and methods

3D image acquisition

In our study, parameter settings for acquiring CT images (Philips Big core) were 130 kV, 200mA, and 25 mS. A default 3.0 mm slice thickness was applied for image and reconstruction image pixel 512*512. Varian's Trilogy™ system (Varian Medical Systems, Salt Lake City, UT) has two modes of image acquisition: full fan with full bowtie (FFFB) mode for small objects and half fan with half bowtie (HFHB) mode for large objects. For the lung density phantoms (LP) we chose the HFHB mode, which is suitable for large size objects with a large scan window. Chest-on model and default parameter settings for acquiring KV CBCT images were 120 kV, 25mA, and pulse time 40 mS. A default 2.5 mm slice thickness was applied for image reconstruction and reconstruction image pixel 512*512.

Phantom specification

The CIRS phantom (Computerized Imaging Reference Systems, Inc., Norfolk, VA) was used with a CT scanner to obtain a CT HU-density table for a baseline dose calculation and stability comparison. Table 1 shows the density specification of the phantom and 17 density inserts. It is composed of lung equivalent material (relative electron density (RED = 0.190) and bone equivalent material (RED = 1.117) surrounded by tissue at the periphery. This phantom was scanned in the HFHB mode for testing dose calculation in the thorax region. The results are shown in Figure 1. The phantom was used for testing large objects.

Table 1.

Specification of Computerized Imaging Reference Systems (CIRS) electron density phantom (model 062)

	Quantity	Electron density (per mL*1023)	RED
Syringe H2O	1	3.340	1.000
Lung (inhale)	2	0.634	0.190
Lung (exhale)	2	1.632	0.489
Adipose	2	3.170	0.949
Breast (50/50)	2	3.261	0.976
Muscle	2	3.483	1.043
Liver	2	3.516	1.052
Trabecular bone	2	3.730	1.117
Dense bone	2	5.052	1.512

Figure 1.

Axial computed tomography (CT) and cone-beam (CB) CT slices of the Computerized Imaging Reference Systems (CIRS) phantom.

Treatment planning system

The collapsed cone convolution (CCC) algorithm in Pinnacle TPS (Philips Radiation Oncology Systems, Fitchburg, WI) was used for dosimetric calculations in this study. It provided a reliable dose calculation accuracy for photon beams with energy of 6 megavolts (MV) and 15 MV. The Pinnacle 8.0 version has the capacity to store multiple calibration tables and allows users to select before dose calculation, even with a plan already attached to the image set.

Cone-beam computed tomography (CBCT) Hounsfield unit (HU)-density tables

For the planning of CT image-based treatment, a well-characterized HU-density calibration table is a prerequisite for accurate dose calculation. Similarly, for CBCT images, the stability of HU numbers and the HU-density table needs to be evaluated. The HU-density calibration curve from the vertical radiation planning CBCT scan was used as a reference. The oblique of 45° radiation planning CBCT scan was evaluated.

Feasibility of CBCT images for dose computation

All CBCT images used in this section were acquired using the default settings and imported via DICOM into the Pinnacle TPS. The pre-commissioned CT HU-density curve was applied to all planning CT images. The CBCT from the lung density phantom was applied to the CBCT images of the corresponding test phantom (CIRS). A plan with a single anteroposterior field was used to investigate the effects of HU calibration curves in this extreme heterogeneous situation. Meanwhile, five different insert region mean doses were evaluated. To evaluate the feasibility of CBCT images for treatment planning, all planning parameters were kept the same, including field size, beam arrangement, dose fluence map, and planned monitor units. The Pinnacle 8.0 version makes it possible to generate a plan on new image set (CBCT image) as long as the two image sets (planning CT and CBCT) are registered.

Patients study

Five lung tumor patients were selected for the evaluation study of CBCT-based dose calculation. For the lung tumor cases, the targets included the clinical target volume (CTV) and planning target volume (PTV), and the organs at risk included the left lung, the right lung, total lung, heart, and spinal cord. In the CT and CBCT plans, the field shape of each beam was the same. We compared the dose volume histogram (DVH) curve between the CT and CBCT plan.

Results

Figure 2 shows a different situation for large objects. The RED profiles were converted from the CBCT HU profile using HU-density conversion generated from the lung density phantom. The solid blue line obtained from the CT image was used as a baseline. For the CT image-based treatment plan, a well-characterized HU-density calibration table is a prerequisite for accurate dose calculation. Similarly, for CBCT images, the HU-density table needs to be evaluated.

Figure 2.

Hounsfield units (HU)-density curves from the planning computed tomography (CT) and cone-beam (CB) CT scans of the phantom.

For Table 2 a discrepancy of 729.6 HU was observed for the dense bone from the expected value compared with CBCT with vertical radiation for the same insert in the lung DP. For the oblique of 45° radiation the HU discrepancy was 138.6 HU for the dense bone. The calibration curve was obtained from the CBCT HU calibration in the following dosimetric calculations because of a severe cupping artifact in this area. Therefore, as shown in the RED profile comparisons in Figure 2, the discrepancy resulting from the cupping artifact still has not been eliminated after the HU-density conversion. The impact of this artifact on the dose calculation accuracy will be addressed further.

Table 2.

Hounsfield unit (HU) numbers of eight inserts from cone-beam computed tomography (CBCT) images of Computerized Imaging Reference Systems (CRIS) phantom, and their discrepancies from the expected mean HU values in planning CT images

	CT	CBCT1	CBCT-CT Discrepancy	CBCT2	CBCT2-CT Discrepancy	CBCT2-CBCT1 Discrepancy
Liver	52.7	48.4	−4.3	77.4	24.7	29
Bone 200	239.9	37.8	−202.1	165.5	−74.4	127.7
Dense bone 800	876.4	146.8	−729.6	285.4	−591	138.6
Lung (inhale)	−815.5	−551.6	263.9	−555.2	260.3	−3.6
Muscle	44.5	2.8	−41.7	−8.6	−53.1	−11.4
Lung (exhale)	−486.4	−388.4	98	−327.3	159.1	61.1
Adipose	−68.5	−168.2	−99.7	−86.9	−18.4	81.3
Breast (50/50)	−31.5	−293.4	−261.9	−240.9	−209.4	52.5

CBCT, cone-beam computed tomography; CT, computed tomography.

In Figure 3, two identical 1-field plans with 6 MV and 15 MV energy were performed on planning CT and CBCT images, respectively, for comparison. We found that isodose lines were all similar and there was no significant difference. Tables 3 and 4 list the comparison of mean doses between the CT, CBCT1 (CBCT with vertical radiation), and CBCT2 plans (CBCT with the oblique of 45° radiation) for five inserts. We concluded that the average dose discrepancy range was 0–3.95%. Thus, a maximum ∼4% dose discrepancy was observed for inserts.

Figure 3.

Isodose lines were compared at five levels, including 100%, 90%, 70%, 50%, and 30% for a 1-field plan both 6 megavolts (MV) and 15 MV energy x-ray applied on the planning computed tomography (CT) with cone beam (CB) CT image registration of the density phantom. Note: the coarse lines represent the CT plan. The fine lines represent the CBCT plans using different color lines.

Table 3.

Five inserts mean dose comparisons between computed tomography (CT) plan and cone-beam (CB) CT1 plans by 6 MV and 15 MV energy x-ray using two different calibrations for the body phantom

		Insert 1	Insert 2	Insert 3	Insert 4	Insert 5
6MV	CT	432.3	314.6	222.1	163.3	122.3
	CBCT1	445.3	318.4	222.5	160.9	118.6
Discrepancy		3.01%	1.21%	0.18%	1.47%	3.03%
15MV	CT	357.4	294.9	222.1	173.4	137.5
	CBCT1	370.8	297.7	222.1	171.7	134.7
Discrepancy		3.75%	0.95%	0%	0.98%	2.04%

CBCT, cone-beam computed tomography; CT, computed tomography; MV, megavolts.

Table 4.

Five inserts mean dose comparisons between computed tomography (CT) plan and cone-beam (CB) CT2 plan by 6 MV and 15 MV energy x-ray using two different calibrations for the body phantom

		Insert 1	Insert 2	Insert 3	Insert 4	Insert 5
6MV	CT	414.5	305.9	222.5	169.0	130.5
	CBCT2	430.9	310.7	222.3	172.0	132.1
Discrepancy		3.95%	1.57%	0.09%	1.77%	1.22%
15MV	CT	367.2	295.7	222.1	178.3	148.1
	CBCT2	375.8	301.2	222.4	180.9	150.3
Discrepancy		2.34%	1.86%	0.13%	1.45%	1.48%

CBCT, cone-beam computed tomography; CT, computed tomography; MV, megavolts.

As shown in Figure 4, the red lines represent the CTV and the green lines represent the PTV. Results show that the CBCT-based dose calculation was underestimated for the targets. For organs at risk, the yellow lines represent the left lung and the lavender lines represent the total lung. Results show that the CBCT-based dose calculation was underestimated for the lungs. The purple lines represent the spinal cord, the blue lines represent the right lung, and the sky-blue lines represent the heart. Results show that the dose calculation on a CBCT image has a high agreement with that of a planning CT image for spinal cord, right lung, and heart. As a result, we concluded that the dose calculation on a CBCT image has a high agreement with that of a planning CT image for homogeneous tissue, but not for heterogene structures.

Figure 4.

The dose volume histogram (DVH) of a lung tumor patient comparison of computed tomography (CT) and cone beam (CB) CT image calibrated by lung density phantom. Note: the dotted lines represent the planning CT plan. The solid lines represent the CBCT plans using different color lines.

Discussion

CBCT volumetric imaging integrated with a medical linear accelerator opens two important applications: patient setup and dose reconstruction or verification.⁶ By imaging the patient routinely during a course of radiation therapy, the accuracy of patient setup can be improved. The question is whether current KV CBCT images could be used directly for radiation dose calculation.⁷ In reality, the quality of the CBCT image is influenced by many factors, such as scatter, beam hardening, and intra-scanning organ motion.⁸ In our study, we used the CIRS phantom and CRT of lung tumors to investigate the KV CBCT to calculate dose.

Dose distributions based on CT and CBCT represent the dose delivered to the patient. In reality, dose calculation on CT imaging has been studied by the phantom and widely used by physicists. In general, the difference between the planning CT and KV CBCT reconstructed dose distribution arises from two factors: (i) relative electron density variation in CBCT images; and, (ii) patient positioning error and organ deformation or displacement. Because CT image scanning is rapid within a breathing cycle, the CBCT image scanning is slow out of a breathing cycle.⁷ In this investigation, we used the static phantom and did not consider the breathing motion. Our static phantom studies indicate that there are small discrepancies between the doses computed using CT and KV CBCT images. The lung tumor case suggests that there are large discrepancies between the doses computed using CT and KV CBCT images in the inhomogeneous tissue. The significant differences are in the DVHs of PTV and gross tumor volume (GTV). The discrepancies between planned and reconstructed doses are lower for the organs at risk.

The KV CBCT-based doses agree with the CT-based calculation to within ∼4% in most situations for the phantom. Therefore, CBCT images can be used for dosimetric validation calibrated by the CIRS phantom for homogeneous tissue. The on-board KV CBCT produces images with HU comparable to the planning CT in homogenous phantoms, yet there are large discrepancies for inhomogeneous materials.⁹ Replacing the conventional planning CT with CBCT for the purpose of radiation treatment planning is not recommended. In addition to the inferior image quality, which may affect the physician's ability to delineate the tumor and sensitive structures, the limited field of view (FOV) of CBCT can be solved by scanning twice in different places and viewing the CBCT images altogether.

Although the discrepancy of doses between the surface and centre is small compared with the uncertainty in therapy dose delivery, efforts are still needed to reduce this imaging dose as much as possible, as its long-term biological impact on normal tissue outside the high-dose area remains unknown. Our studies indicate that the stability of the HU-density conversion is very important and must be taken into account in the treatment planning system (TPS) for accurate dose calculation. Our results suggest that the CBCT image must be calibrated by HU numbers, field angle, and material inhomogeneity. These variations result in the instability of the HU-density table, which, consequently, affect the dose accuracy.

There are other variables that have not been addressed in this study that could cause deviations in HU numbers or instability of HU-density calibrations. A previous study reported that different HU-density conversions corresponding to different KV would result in dosimetric variations.¹⁰ Several groups have stated that the level of acceptable dose agreement for single fields is about 3%, or 3 mm for dose accuracy.¹¹ In this study, our result is higher than this principle.

Overall, a universal HU-density table is insufficient for reaching the level of precision required for treatment planning based on KV CBCT image. A higher degree of tissue heterogeneity results in higher dose discrepancies.¹² For higher accuracy in dose calculation, more effective scatter correction methods need to be applied on the CBCT imaging system, other than using only the scatter grid. This study was conducted assuming that the results would be equally valid for HU-density calibration for any KV CBCT system because the underlying physics of the relation between density and HU number is independent of manufacturer.

Conclusion

This study investigated the discrepancy of doses between CT and CBCT imaging. We found that there was a discernable impact on the HU-density calibration on the CBCT image, which makes it unfeasible to reduce the imaging dose to patients without compromising dose calculation accuracy. Small scatter contamination results in less HU number variation for the same density inserts located at different distances from the image center, thus, less variation in the derived HU-density conversion. Moreover, the field angle and its density composition had a significant impact on HU-density tables and, thus, on dose calculation. Our results proved that using a site-specific calibration curve gives a maximum 4% dose agreement with CT plans in phantom studies. The site-specific calibration is necessary for dose calculation on KV CBCT images to obtain high accuracy. The phantom and lung patient studies showed that the dosimetric accuracy of CBCT-based dose calculation is acceptable for the purpose of dosimetric checks. However, the direct use of CBCT for patient dose calculation instead of traditional planning CT is not recommended.

Disclosure

No authors report any conflict of interest.