- Email: [email protected]
Dosimetric feasibility of cone-beam CT-based treatment planning compared to CT-based treatment planning
Int. J. Radiation Oncology Biol. Phys., Vol. 66, No. 5, pp. 1553–1561, 2006 Copyright © 2006 Elsevier Inc. Printed in the USA. All rights reserved 0360-3016/06/$–see front matter
doi:10.1016/j.ijrobp.2006.08.031
PHYSICS CONTRIBUTION
DOSIMETRIC FEASIBILITY OF CONE-BEAM CT-BASED TREATMENT PLANNING COMPARED TO CT-BASED TREATMENT PLANNING SUA YOO, PH.D.,
AND
FANG-FANG YIN, PH.D.
Department of Radiation Oncology, Duke University Medical Center, Durham, NC Purpose: Cone-beam computed tomography (CBCT) images are currently used for positioning verification. However, it is yet unknown whether CBCT could be used in dose calculation for replanning in adaptive radiation therapy. This study investigates the dosimetric feasibility of CBCT-based treatment planning. Methods and Materials: Hounsfield unit (HU) values and profiles of Catphan, homogeneous/inhomogeneous phantoms, and various tissue regions of patients in CBCT images were compared to those in CT. The dosimetric consequence of the HU variation was investigated by comparing CBCT-based treatment plans to conventional CT-based plans for both phantoms and patients. Results: The maximum HU difference between CBCT and CT of Catphan was 34 HU in the Teflon. The differences in other materials were less than 10 HU. The profiles for the homogeneous phantoms in CBCT displayed reduced HU values up to 150 HU in the peripheral regions compared to those in CT. The scatter and artifacts in CBCT became severe surrounding inhomogeneous tissues with reduced HU values up to 200 HU. The MU/cGy differences were less than 1% for most phantom cases. The isodose distributions between CBCT-based and CT-based plans agreed very well. However, the discrepancy was larger when CBCT was scanned without a bowtie filter than with bowtie filter. Also, up to 3% dosimetric error was observed in the plans for the inhomogeneous phantom. In the patient studies, the discrepancies of isodose lines between CT-based and CBCT-based plans, both 3D and IMRT, were less than 2 mm. Again, larger discrepancy occurred for the lung cancer patients. Conclusion: This study demonstrated the feasibility of CBCT-based treatment planning. CBCT-based treatment plans were dosimetrically comparable to CT-based treatment plans. Dosimetric data in the inhomogeneous tissue regions should be carefully validated. © 2006 Elsevier Inc. Cone-beam CT, HU values, Treatment planning.
on the treatment couch. One potential approach to accomplishing the goal of replanning is to generate a modified plan based on CBCT that includes the latest information of patient positioning, target location and anatomy. Computed tomography-based treatment planning has been fundamentally validated for it has relatively longer history, and has been used as a gold standard for radiation therapy for the planning purpose (11–15). In contrast, CBCT is at its early stage in radiation therapy application and has not been evaluated for treatment planning. This study aimed to explore the dosimetric feasibility of CBCTbased treatment planning. CBCT and CT images of phantoms and patients were acquired with scanning techniques commonly used for routine clinic were used for the study. Hounsfield unit (HU) values and profiles were compared for CBCT and CT images. Dosimetric results were compared to evaluate CBCT-based plans relative to CT-based plans for both phantoms and patients.
INTRODUCTION Rapid development in image-guided radiation therapy (IGRT) has provided powerful tools for improving accuracy of patient positioning and target localization (1, 2). There have been many studies devoted to the understanding of set-up uncertainties for target localization (3– 6). Cone-beam CT (CBCT) offers a state-of-the-art technology to improve the target localization (1–2, 7). Patient positioning can be verified by matching soft tissues and/or bony structures in CBCT to those in planning CT images. However, organ motions, organ deformation, and changes of external body are often observed through the course of treatment for many patients. These target/organ variations cannot be resolved by simply repositioning the patient. One of the ideal solutions will include real-time replanning for such situations (8 –10). This replanning procedure could allow modification of the initial treatment plan to accommodate the variations while the patient is Reprint requests to: Sua Yoo, Ph.D., Department of Radiation Oncology, Duke University Medical Center, Box 3295, Durham, NC 27710. Tel: (919) 660-2179; Fax: (919) 681-7183; E-mail: [email protected]. Acknowledgments—The authors would like to acknowledge Peter
Munro, Ph.D. (Varian medical systems). This work is partially supported by a research grant from Varian Medical Systems. Received May 31, 2006, and in revised form Aug 14, 2006. Accepted for publication Aug 16, 2006. 1553
1554
I. J. Radiation Oncology
●
Biology
●
Physics
METHODS AND MATERIALS The OBI system The on-board imaging (OBI®, Varian Medical Systems, Palo Alto, CA) system consists of a kV X-ray source (KVS) and a kV amorphous-silicon digital imaging detector (KVD) mounted on the linear accelerator using robotic arms (Exact™), which are orthogonal to the electronic portal imaging device (aSi500, PortalVision™, Varian Medical systems). CBCT images were reconstructed using about 700 kV-projection images acquired over 360° rotation. When the center of the KVD is positioned at the isocenter in the longitudinal-lateral plan and 50 cm away from the isocenter in the vertical direction, the reconstructed field-of-view (FOV) is a circle of 24-cm diameter with a 15-cm length. This acquisition mode is called “full-fan” and is used for small anatomic sites such as the brain, head-and-neck, and a truncated part of larger sites. For larger sites, such as the pelvis, chest, and abdomen, the KVD is shifted by 14.8 cm laterally. In this mode, only part of the object is viewed in a half-fan projection. The other part of the object is viewed in the half-fan projection from the opposite direction. This acquisition mode is called “half-fan”. The FOV for the half-fan mode is a circle of 45-cm diameter with a 14-cm length. The effects of X-ray scatter and artifacts are larger in CBCT images than in CT images (16 –18). A bowtie filter mounted to the X-ray tube improved image quality because it reduces intensity variations across the detector and charge trapped in the detector (19, 20). Thus, there are 4 techniques available with the current CBCT acquisition: full-fan with (CBFFB), without a bowtie filter (CBFF), half-fan with (CBHFB), and without a bowtie filter (CBHF). However, the clinical protocols of CBCT should use a bowtie filter for better image quality and lower skin dose. This study included CBCT images with and without a bowtie filter for phantoms and CBCT images with a bowtie filter for patients. The user could determine the imaging geometry, the X-ray techniques, and the reconstruction parameters. For a default clinical full-fan mode, 24-cm diameter with 2.5-mm slice thickness was used, and 40-cm diameter with 2.5-mm slice thickness was used for a default half-fan mode. The acquisition X-ray technique was controlled by pre-determined parameters that were calibrated before clinical use. The techniques calibrated during installation of the OBI system were 125 kVp, 80 mA, and 25 ms with a bowtie and 125 kVp, 80 mA, and 9 ms without a bowtie. A GE Light-speed RT CT simulator (GE Medical Systems, Milwaukee, WI) was used for CT images. A “head-scan” for smaller sites such as head and neck uses 120 kVp and an automatically adjusted mA with 2.5 mm slice thickness. A “body-scan” for larger anatomic sites such as lung and pelvis uses 120 kVp and an automatically adjusted mA with 2.5 mm slice thickness. Although the automatically adjusted mA technique was used, mA during scanning was consistently found to be about 330 to 345 mA for a head scan and about 80 to 110 mA for a body scan.
HU comparisons Phantom cases. The CTP 404 in Catphan 504 (The Phantom Laboratory, Salem, NY) is designed to monitor HU values in CT images. It contains 1.2-cm diameter disks of various materials (e.g., Teflon, polystyrene, air, etc.) with different electron densities. HU values in CBCT images were calibrated using this phantom during installation. For this study, head-scan CT (CTHead), body-scan CT (CTBody), CBFFB, CBFF, CBHFB, and CBHF of Catphan were acquired. The region-of-interest (ROI) was selected
Volume 66, Number 5, 2006
for each material disk to cover a 0.7 cm ⫻ 0.7 cm square. The mean HU values in the ROIs were measured. The CIRS phantom Model 002HA6SN (CIRS Tissue Simulation and Phantom Technology, Norfolk, VA) is a homogeneous ellipsoidal phantom where a smaller cylindrical part could be separated out. The large ellipsoidal phantom, referred to as “body phantom” in this article, is 30 cm ⫻ 20 cm ⫻ 30 cm (width ⫻ height ⫻ length) and the small cylindrical phantom, referred to as “head phantom,” is 16 cm ⫻ 16 cm ⫻ 30 cm. CTHead, CBFFB, CBFF of the head phantom, and CTBody, CBHFB, and CBHF of the body phantom were acquired. HU profiles along a horizontal line across the phantoms were extracted from CT and CBCT images for comparison. The CIRS phantom Model 008 Dynamic Thorax Phantom (CIRS) was used without the motion controller. The phantom was the same size as the homogeneous body phantom, and included separate inserts representing the spine and lung containing a tumor. CTHead, CTBody, CBFFB, CBFF, CBHFB, and CBHF of the inhomogeneous phantom were acquired. Because CBFFB and CBFF could not encompass the entire phantom, the phantom was placed such that the spine and the right lung with the tumor were within the 24-cm diameter FOV. HU profiles along vertical lines crossing the lung with the tumor and along the vertical lines crossing the spine were extracted from CT and CBCT images for comparison. Patient cases. Two brain cancer patients (Patients A and B) were scanned to acquire CTHead and CBFFB. An ROI of 1.5 cm ⫻ 1.5 cm was selected to measure HU value in the brain region in CTHead and CBFFB. Two lung cancer patients (Patients C and D) were scanned to acquire CTBody and CBHFB. HU values in the lung and spine in CTBody and CBFFB were compared. Two prostate cancer patients (Patients E and F) were scanned to acquire CTBody and CBHFB. HU values in muscle, fat, and femoral head in CTBody and CBFFB were compared. All patient data were collected from routine clinical applications.
Dose comparison Phantom cases. All CT and CBCT images of the homogeneous phantoms (i.e., body and head phantoms) and the inhomogeneous phantom were imported into the ECLIPSE treatment planning system (ECLIPSE™, Varian Medical Systems). Plans based on CT and CBCT images were generated using one 10 cm ⫻ 10 cm photon beam. The plans for the homogeneous phantoms included a right lateral beam isocentered at the center of the phantoms. The beam in the plans for the head phantom had a 5-cm depth, and the beam in the plans for the body phantom had a 15-cm depth. The plans for the inhomogeneous phantom included an anterior beam isocentered at the center of the tumor passing through the right lung. The other plans for the same phantom included a posterior beam isocentered at the center of the phantom passing through the spine. The CT electron density calibration curve, which had been commissioned based on our simulation CT for the treatment planning purposes, was applied for all CT and CBCT images. Both 6MV and 15MV were used. Dose was computed with inhomogeneity correction using Modified Batho Method, which uses tissuemaximum ratio (21) and yielded the closest doses to measurements among available methods in Eclipse for both low (6MV) and high-energy (18MV) photons (22). For MU/cGy comparison, dose was normalized to the isocenter with a prescription of 100 cGy. The MU/cGy differences were obtained between CT-based and CBCT-based plans. For isodose distribution comparison, the normalization was adjusted so that the beam delivered 100 MU using
Treatment planning based on cone-beam CT
●
S. YOO
AND
F.-F. YIN
1555
the same plans. The resulting percentage isodose distributions were compared between CT- and CBCT-based plans. For the inhomogeneous phantoms, the right lung, tumor, and spine were contoured in each image and dose-volume-histograms (DVH) were compared as well. Patient cases. All CT and CBCT images of the Patients A to F were imported into ECLIPSE. 3D plans were generated for Patients A, C, and E, and IMRT plans were generated for Patients B, D, and F based on their CT images and treatment targets. The 3D plan for Patient A (brain case) included five 6 MV fields with wedges. The IMRT plan for Patient B (brain case) included six 6 MV fields. The 3D plan for Patient C (lung case) included two 15 MV fields. The IMRT plan for Patient D (lung case) included six 6 MV and fifteen MV fields. The 3D plan for Patient E (prostate case) included four 15 MV fields. The IMRT plan for Patient F (prostate case) included seven 15 MV fields. Verification plans were created based on CBCT, keeping the same MU values from the original CT-based plans. The resulting percentage isodose distributions were compared between CT-based and CBCT-based plans.
RESULTS HU comparison Phantom cases. Figure 1 shows comparison of measured mean HU values vs. expected HU values of the materials in the Catphan. The HU value differences between CBCT and CT were less than 10 HU, for all disks except for the Teflon. CBFF and CBHF displayed about 34 and 31 higher HU than CTHead and CTBody, respectively, for the Teflon. Note that the various materials in Catphan may not represent inhomogeneous tissues in actual human anatomy. Nevertheless, this study demonstrates that CBCT has the ability to generate images with HU values comparable to those in CT. Figure 2 displays the axial views of CT and CBCT images of the homogeneous and inhomogeneous phantoms. The window levels for the images were adjusted to display from ⫺1000 HU to 200 HU in grayscale. In the images of the homogeneous phantoms, shown in Fig. 2A to C, there was no noticeable difference between CBFFB, CBFF, and CTHead. When CBFF was zoomed-in, the ring artifact was discernable, as displayed in Fig. 2C-1. This ring artifact occurred due to a defect in some detector elements (17, 23).
Fig. 1. Hounsfield unit (HU) comparison in computed tomography (CT) and cone-beam computed tomography (CBCT) images of Catphan.
Fig. 2. Axial images cone-beam computed tomography; (CBCT). (A) CTHead, (B) CBFFB, (C) CBFF, (D) CTBody, (E) CBHFB, and (F) CBHF for the homogeneous phantoms. (C-1) is the magnified CBFF in the central region. (G) CTHead, (H) CBFFB, (I) CBFF, (J) CTBody, (K) CBHFB, and (L) CBHF of the inhomogeneous phantom.
However, the HU deviation in the ring artifact was less than ⫾20 HU in this study. CTBody showed uniform images whereas CBHFB displayed a dark circle beyond the head phantom as indicated by a black arrow in Fig. 2E, due to the ghosting effect (24). CBHF in Fig. 2F displayed a large dark region in the posterior region due to the cupping artifact (18, 24). The peripheral ring, indicated by the white arrow in Fig. 2H, occurred due to the normalization calibration. Normalization calibration was performed for a 25 cm diameter FOV, with a phantom covering a 24 cm diameter circle for the full-fan mode. CBCT images of the inhomogeneous phantom, Fig. 2H, I, K, and L, showed the streak artifact around the edges of the inhomogeneous objects, (i.e., spine and lung) due to beam hardening and scatter (16 –18, 25). Figure 3A and 3B shows the HU profiles extracted from the CT and CBCT axial images along the dotted lines shown in Figs. 2A and 2D. The vertical axis represents HU values scaled from ⫺200 to 200HU in Figs. 3A and 3B. The horizontal axis represents the spatial distance from the one end to the other end of the phantom: 16 cm for the head phantom and 30 cm for the body phantom. The HU profiles in the phantom periphery were reduced by 100 to 150HU in CBFFB and CBFF, compared with CTHead. The profiles of CBHFB and CBHF were more irregular than those of CBFFB and CBFF due to larger primary-to-scatter ratio with larger FOV and the larger phantom (18). The HU values in the profile of CBHFB varied from ⫺50 to 100 HU in the central region and peripheral region. The HU values in the profile of CBHF varied from ⫺150 to 0 HU with irregular distribution. The HU profiles extracted along the superior-inferior line in the frontal view of the images were
1556
I. J. Radiation Oncology
●
Biology
●
Physics
Fig. 3. Comparison of Hounsfield unit (HU) profiles. (A) The head homogeneous phantom in CTHead, CBFFB, and CBFF. (B) The body homogeneous phantom in CTBody, CBHFB, and CBHF. The inhomogeneous phantom through the tumor and right lung in (C) CTHead, CBFFB, and CBFF and in (D) CTBody, CBHFB, and CBHF. The inhomogeneous phantom through the spine in (E) CTHead, CBFFB, and CBFF and in (F) CTBody, CBHFB, and CBHF.
also compared. The same pattern of reduced HU appeared in the peripheral superior and inferior regions as observed in the axial view. Figures 3C and 3D shows the comparison of HU profiles along the vertical line cutting through the right lung and the tumor, as shown by the white dotted line in Fig. 2G. The profiles along the tumor and near the top and bottom of the body regions of CT images were close to 0 HU, whereas the profiles in the same regions of CBCT images varied from ⫺150 to 0 HU. The further reduction in HU values occurred in the periphery edge of the tumor and near the top and bottom body regions. The profiles within the lung of CBCT images showed reduced HU values ranging from 100 to 200 HU values compared with those of CT images. The nonuniformity and reduction of HU values became worse in CBCT images acquired without a bowtie filter than in CBCT images with a bowtie filter. The HU profiles extracted along the superior-inferior line in the frontal view of CBCT images illustrated reduced HU values. Figures 3E and 3F shows the comparison of HU profiles along the vertical line cutting through the body and the spine as shown in the black dotted line in Fig. 2G. The profiles of CBCT images showed reduced HU values rang-
Volume 66, Number 5, 2006
ing from ⫺50 to ⫺200 HU at the junctions between the body and the external air and between the body and the spine. The same pattern occurred in the profiles of the homogeneous phantoms as shown in Figs. 3A and 3B. The profiles of CBCT with a bowtie within the spine were close to those of CT, whereas the profiles of CBCT without a bowtie were about 100 HU lower. The posterior area between the spine and the external air showed reduced HU because of the streak artifact (18). The profiles in this posterior area showed interesting distributions; those of CBFFB/CBFF reached about ⫺200 HU whereas those of CBHFB/CBHF reached about ⫺100 HU. It was because the phantom was placed so that the posterior body was imaged near the edge of the FOV, where reduced HU values occurred more substantially in addition to the streak artifact. Patient cases. Table 1 summarizes the comparison of HU values in CT and CBCT images of patients for various tissue types. As much as possible, ROIs in CT and CBCT for each tissue type were selected so that they enclosed the same tissue type in the same location. All tissues, except the brain for Patient A and the fat for Patient F, showed lower HU values in CBCT than in CT. The HU differences were less than 70 HU values for all tissues except for the back muscle region in the vicinity of the spine. The HU differences in this back muscle between CT and CBCT were greater than 300 HU. As shown with the HU profiles of the posterior body region in Figs. 3E and 3F, the streak artifact contributed to the reduction of HU values in this area. Noticeable streak artifacts in small volumes also occurred in the CBCT images of the lung cancer patients because of breathing motion.
Table 1. Comparison of HUs in different tissues of patients Tissue type Brain cases Brain Lung cases Lung Spine Back muscle in the vicinity of the spine Prostate cases Muscle Fat Femoral head
HU ⫾ SD
HU ⫾ SD
CTHead CBFFB
Patient A 37 ⫾ 6.8 84 ⫾ 11.1
Patient B 60 ⫾ 33.2 54 ⫾ 18.4
CTBody CBHFB CTBody CBHFB CTBody CBHFB
Patient C ⫺965 ⫾ 40.2 ⫺978 ⫾ 38.4 184 ⫾ 46.8 110 ⫾ 50.6 2 ⫾ 46.2 ⫺337 ⫾ 74.6
Patient D ⫺887 ⫾ 191.6 ⫺910 ⫾ 82.9 226 ⫾ 112.5 218 ⫾ 81.1 ⫺10 ⫾ 42.8 ⫺326 ⫾ 52.0
CTBody CBHFB CTBody CBHFB CTBody CBHFB
Patient E 67 ⫾ 16.0 48 ⫾ 34.8 ⫺90 ⫾ 14.7 ⫺112 ⫾ 44.4 363 ⫾ 65.6 327 ⫾ 64.5
Patient F 65 ⫾ 25.1 61 ⫾ 54.1 ⫺104 ⫾ 16.3 ⫺90 ⫾ 43.3 414 ⫾ 95.5 352 ⫾ 90.1
Image
Abbreviations: CT ⫽ computed tomography; HU ⫽ Hounsfield unit; SD ⫽ standard deviation. Mean HU values ⫾ standard deviations were measured over 1.5 cm ⫻ 1.5 cm region of interest in each tissue area.
Treatment planning based on cone-beam CT
●
S. YOO
AND
F.-F. YIN
1557
Table 2. Comparison of the percent difference, ⌬, in MU/cGy between CT-based and CBCT-based plans Phantom
Homogeneous phantoms
Plan
(A) Lat beam
Inhomogeneous phantom (B) AP beam
(C) PA beam
Energy
6 MV
15 MV
6 MV
15 MV
6 MV
15 MV
CBFFB CBFF CBHFB CBHF
⫺0.58 ⫺0.41 0.67 ⫺2.55
⫺0.42 ⫺0.31 0.43 ⫺1.75
⫺2.30 ⫺3.07 ⫺2.62 ⫺2.45
⫺1.50 ⫺2.07 ⫺1.78 ⫺1.68
⫺2.38 ⫺3.44 ⫺0.39 ⫺2.01
⫺1.61 ⫺2.27 ⫺0.28 ⫺1.38
Abbreviations: CBCT ⫽ cone-beam computed tomography; CT ⫽ computed tomography; AP ⫽ anterior-posterior; PA ⫽ posterior-anterior. ⌬ ⫽ (MU/cGy of CBCT – MU/cGy of CT)/ MU/cGy of CT*100. (A) Comparison of the plans with the lateral beam isocentered at the center of the homogeneous phantom. (B) Comparison of the plans with the AP beam isocentered at the center of the tumor in the right lung for the inhomogeneous phantom. (C) Comparison of the plans with the PA beam isocenterd at the center of the phantom passing through the spine for the inhomogeneous phantom.
Dose comparison Phantom cases. Table 2(A) summarizes the percent difference in MU/cGy (⌬) between the CT-based and CBCTbased plans for the homogeneous phantoms. The negative ⌬ represents that the CBCT-based plan under-estimated MU/ cGy compared to the CT-based plan. All CBCT-based plans, except the CBHF-based plans, showed ⌬s less than 1%. The CBHF-based plans showed ⫺2.55% and ⫺1.75% of ⌬s for 6 MV and 15 MV, respectively. Relatively larger ⌬s variations in the CBHF-based plans were the results from the reduced HU values in the central posterior area as shown in Fig. 2F. The results using a 15 MV beam showed better agreement than those using a 6 MV beam. This is because tissue-air ratio or tissue maximum ratio of higher energy is less affected by the change of depth. Therefore,
Fig. 4. Comparison of isodose distributions of CT-based and CBCT-based plans using 6 MV photon for the homogeneous phantoms. Panels A and C are axial and frontal views along the central axis for plans based on CTHead, CBFFB, and CBFF for the head phantom. Panels B and D are for plans based on CTBody, CBHFB, and CBHF for the body phantom. The number in each line is in the relative dose (%) to the prescription of 100 cGy. CBCT ⫽ cone-beam computed tomography; CT ⫽ computed tomography.
higher energy photon beam is associated with lower tissueinhomogeneity correction factors (26). Figure 4 displays relative isodose distributions of the plans using 100 MU of 6 MV in the axial and frontal views along the central axis. The isodose lines from CBCT-based plans agreed well with those from CT-based plans except those from CBHF-based plans. Along the central axis, the discrepancies of 80% and 50% lines between CBHF-based and CT based plans were about 5 mm and 10 mm, respectively, which were equivalent to approximately 2% and 3% of differences in dose. These discrepancies increased in the posterior part where the cupping artifact was observed. The same trends were found in the frontal view. It was observed that plans using a 15 MV beam showed better agreement with CT-based plans than using a 6 MV beam. Table 2 (B) summarizes the ⌬s between the CT- and CBCT-based plans generated using the inhomogeneous phantom with the anterior beam isocentered at the center of the tumor in the right lung. The ⌬s ranged from ⫺1.50 to ⫺2.62% for plans based on CBCT with a bowtie filter and from ⫺1.6% to ⫺3.07% for plans based on CBCT without a bowtie filter. Based on the results shown in Figs. 3C and 3D, under-estimation of MU/cGy was expected for all CBCT-based plans. Plans based on CBCT with a bowtie filter showed better dosimetric agreement with plans based on CT than plans based on CBCT without a bowtie filter. Figure 5A–D displays relative isodose distributions of the plans using 100 MU of 6 MV in the axial and sagittal views along the central axis. The discrepancies of isodose lines between plans based on CTBody and CBHFB/CBHF were about 5 mm and 10 mm for 100% and 80% lines, respectively, which were equivalent to approximately 2% and 3% of differences in dose. The same discrepancies occurred between plans based on CTHead and CBFFB. The larger discrepancy occurred between the CTHead-based plan and CBFF-based plan, where the discrepancy in 100% and 80% isodose lines reached 7 mm and 17 mm, respectively, which were equivalent to 4% and 5% of difference in dose. Again, the isodose distributions in the plans using a 15 MV beam
1558
I. J. Radiation Oncology
●
Biology
●
Physics
Fig. 5. Comparison of CT-based and CBCT-based plans using 6 MV photon for the tumor in the right lung in the inhomogeneous phantom. Panels A and C are isodose distributions in the axial and sagittal views along the central axis for plans based on CTHead, CBFFB, and CBFF. Panels B and D are for plans based on CTBody, CBHFB, and CBHF. DVHs of plans based on (E) CTHead, CBFFB, and CBFF, and (F) CTBody, CBHFB, and CBHF. CBCT ⫽ conebeam computed tomography; CT ⫽ computed tomography.
showed better agreement than plans using a 6 MV beam. Figures 5E and 5F show the DVHs for the same plans. DVHs of CBFFB- and CBHFB/CBHF-based plans showed about 3% of dose-increase for the entire volume of the tumor and about 30 –70% volume of the right lung compared to the CT-based plans. CBFF-based plans showed about 5% of dose increment for the tumor and the right lung. Table 2 (C) summarizes the ⌬s between the CT-based and CBCT-based plans for the inhomogeneous phantom with the posterior beam isocentered at the center of the body passing through the spine. CBHFB-based plans showed very good agreement (⬍1%) with CTBody-based plans. Plans based on CBCT with a bowtie filter showed smaller ⌬s than plans based on CBCT without a bowtie. CBFF/ CBFFB-based plans for this case resulted in large ⌬s: ⫺1.61 to ⫺3.44%. These large ⌬s occurred not because of the spine but because of the posterior body region, which displayed further reduced HU values in the profiles shown in Fig. 3E. Again, the ⌬s were larger in 6 MV than in 15 MV. Figure 6A–D displays relative isodose distributions of the plans using 100 MU of 6 MV in the axial and sagittal views along the central axis. The isodose lines of the CBCT-based plans reached slightly further anteriorly than those of the CT-based plans. Isodose lines of CTBody- and CBHFB-
Volume 66, Number 5, 2006
based plans showed good agreement. The discrepancy between CTBody- and CBHF-based plans was 1 mm and 3 mm for 80% and 50% lines along the central axis, respectively, which were equivalent to about 1% of differences in dose. The discrepancies of isodose lines between CTHead- and CBFFB-based plans were 4 mm and 5 mm for 50% and 80% lines along the central axis, respectively, which were equivalent to about 1% of differences in dose. The discrepancies were increased between CTHead- and CBFF-based plans; 5 mm for 80% and 10 mm for 50% lines, equivalent to about 2% and 3% of differences in dose, respectively. The discrepancy of isodose lines between CTHead- and CBFFB/CBFF-based plans was larger than the discrepancy of those between CTBody- and CBHFB/CBHF-based plans. Figures 6E and 6F show comparison of DVHs from the same plans. DVHs of CBFFB/CBFF-based and CBHFbased plans showed about 1.5% and 1% of dose-increase for the spine, respectively, compared to CT-based plans. DVH of the spine in CBHFB-based plan overlapped with that in CTBody-based plan. The DVHs of the right lung from all plans overlapped as well. Patient cases. Figure 7 shows comparisons of the patient cases. The 100% dose line from CBCT-based plan covered a larger area than the same line from the CT-based plan for Patient A in Fig. 7C. However, the 99% line from the CBCT-based plan agreed with the 100% dose line from the
Fig. 6. Comparison of CT-based and CBCT-based plans using 6 MV photon for the spine in the inhomogeneous phantom. Panels A and C are isodose distribution in the axial and sagittal views along the central axis for plans based on CTHead, CBFFB, and CBFF. Panels B and D are for plans based on CTBody, CBHFB, and CBHF. DVHs of plans based on (E) CTHead, CBFFB, and CBFF, and (F) CTBody, CBHFB, and CBHF. CBCT ⫽ cone-beam computed tomography; CT ⫽ computed tomography.
Treatment planning based on cone-beam CT
●
S. YOO
AND
F.-F. YIN
1559
Also, there are small discrepancy in the 20% and 40% isodose lines due to set-up error between CT and CBCT. The CT of Patient F in Fig. 7P was scanned with a contrast material. Therefore, the bladder was contoured and assigned with 0 HU value for the CT-based plan. CBCT in Fig. 7Q appeared grainier than the CT in Fig. 7P because of increased scatter due to the large size of Patient F. The 45-cm diameter FOV did not enclose the patient body in CBCT. The external body was modified in CT such that the distance from one end to the other end in CT matched with that in CBCT. The 3D plans based on CT and CBCT of Patient E showed good agreement in most dose lines, as shown in Fig. 7O. The discrepancies occurred around the central region of 100% isodose line and the posterior end in 20% and 40% isodose lines. The CBCT-based plan delivered higher dose to the central region than the CT-based plan mainly because of the additional gas-filled in rectum. The 2-mm discrepancy around the posterior end happened because of the set-up error. The IMRT plans based on CT and CBCT for Patient F showed good agreement in their dose distributions, as shown in Fig. 7R, except in the posterior and left ends due to the set-up error. DISCUSSION
Fig. 7. Comparison of isodose distributions of CT-based and CBCT-based plans for (A) to (F) the brain cancer patients, (G) to (L) the lung cancer patients, and (M) to (R) the prostate cancer patients. (A), (D), (G), (L), (M), and (P) are the CT images, (B), (E), (H), (K), (N), and (Q) are the CBCT images, and (C), (F), (I), (L), (O), and (R) are the isodose distributions. The legend of dose lines is in Fig. 7(F). CBCT ⫽ cone-beam computed tomography; CT ⫽ computed tomography.
CT-based plan, indicating that the dose difference was only 1% in the high dose region for Patient A. Other isodose lines in Patient A and B showed good agreement In the CBCT images of the lung cancer patients shown in Figs. 7H and 7K, it is noticeable that the back muscle in the vicinity of the spine appeared darker (i.e., lower HU values) than other body regions. The streak artifact caused by breathing motion also appeared around the lungs and chest wall, but the size of the artifact was much smaller than that in the posterior region of the spine. The comparison of CT-based and CBCT-based plans showed good agreement between 20% to 80% isodose lines, but there were some discrepancy in the 105% isodose line. The discrepancy was larger in the 3D plan, shown in Fig. 7I, than in the IMRT plan in Fig. 7L. The increased dose in CBCT-based plans compared to CT-based plans was from the reduced HU values in the lungs and the posterior region of the spine.
This study investigated the dosimetric feasibility of CBCT-based treatment planning. The differences in HU values between CT and CBCT for the Catphan were less than 10 HU for most disk materials except the Teflon, which showed a maximum difference of 34 HU. The profiles of CBCT in the homogeneous phantoms displayed 100 to 150 lower HU at the periphery edge in the full-fan mode, and around the periphery edge and central regions in the half-fan mode, than the profiles of CT. Overall, the profiles of CBCT in the inhomogeneous phantom displayed 50 to 200 HU lower than those of CT. The HU values in CBCT of patients were also lower than those in CT. The back muscle in the vicinity of the spine showed about 300 HU lower in CBCT than in CT. The dosimetric consequence as the result of HU variation in CBCT was evaluated by comparing MU/cGy, isodose distributions, and DVHs. For the homogeneous phantoms, the discrepancy between the plans was small, with less than 1% of MU/cGy difference and good agreement in isodose lines when CBCT was scanned with a bowtie. The dosimetric discrepancy became larger for the inhomogeneous phantom when beams passed through the lung or the posterior body images at the edge of FOV. The differences were about 2–3% in MU/cGy and about 5–11 mm (equivalent to 2–3%) in isodose lines for the CBCT with a bowtie filter. In general, the 3D and IMRT plans based on CBCT for patients showed good agreement with the plans based on CT in general. About 2 mm discrepancy between 2 isodose lines occurred because of the set-up error. Discrepancy also occurred in the high-dose region of the 3D plan and the IMRT plan for the lung cancer patients. The regions enclosed by the 105% isodose lines were shaped differently between the
1560
I. J. Radiation Oncology
●
Biology
●
Physics
2 plans. The CBCT-based plans showed a larger region enclosed by 105% lines than CT-based plans. However, other areas enclosed by isodose lines below 90% matched very well. We set scanning techniques for CT and CBCT as close as possible. The current CBCT configuration uses 125 kVp whereas our CT scanner does not provide 125 kVp. Some investigations have sought to determine the impact of kVp to HU values and dose calculation. McCullough and Holmes stated that changing from 120 kVp to 140 kVp did not produce significant change in HU (14). Yet, Cozzi et al. reported that voltage variation from 100 to 140 kVp caused a maximum of 300 HU difference in high-density materials—a difference that resulted in a maximum of 4% dosimetric error (13). However, we found that CT images of Catphan using 120 kVp and 140 kVp produced a maximum difference of 10 HU. Thus, using 120 kVp instead of 125 kVp in CT would not produce considerable variations for this study. CBCT uses 80 mA and 9 ms without a bowtie filter, and 25 ms with a bowtie filter, whereas CT automatically adjusts mA. Cozzi et al. reported that exposure time (ms) and current (mA) did not affect the HU values (13). We also found that CT images of Catphan using automatically adjusted mA produced very small random differences in HU values (⬍ ⫾2 to ⫾5 HU) compared to CT images using fixed 80 mA. Therefore, the techniques used for CT acquisition were appropriate for this study. The difference in HU values in Catphan was smaller than the difference in the other phantoms and patients. The magnitude of scatter and artifacts in CBCT images are affected by imaging geometry, object size, and inhomogeneous tissues (18). Scatter and artifacts deteriorate image uniformity and mislead HU values in 3D reconstruction. An object like Catphan with small disks provides less scatter and artifacts than a large object with large inhomogeneous tissues (16, 18). The location of an object and neighboring tissue types/sizes also affects HU values. This study did not investigate HU variation caused by the various factors. However, the cases simulated in this study could represent common clinical situations. Some studies recommend that the relationship between electron densities and CT numbers (i.e., HU-ED curve) should be stored in the treatment planning system for accurate dose calculation (12, 15). It was reported that up to 2% of dosimetric differences were found by using different
Volume 66, Number 5, 2006
HU-ED curves for 250 to 300 HU differences (13, 27). However, for 2 reasons our study did not include a HU-ED generated specifically for CBCT. First, we found maximum of 34 HU difference in CBCT without a bowtie and 15 HU difference in CBCT with a bowtie filter, compared with CT. HU differences smaller than 34 HU values in a HU-ED curve will not affect dosimetric results. Second, there is no proper way to create a standard HU-ED curve of CBCT because the magnitude of scatter varies with patient-dependant factors such as size and location of patient body, inhomogeneous tissues, and neighboring tissues. The results for CBCT-based plans provided in this study were comparable to those for CT-based plans. However, there are some limitations in CBCT-based treatment planning. Due to the limited field size of the detector, only about 14 cm in-length could be scanned in CBCT whereas many of clinical cases require a larger view. Acquiring 2 sets of CBCT, 1 for a superior body and the other for an inferior body by shifting the couch, could provide a larger superiorinferior view. However, a user should understand that the periphery of CBCT image show reduced HU values and setting an isocenter in the superior or inferior edge of the image is not desirable. Thus, in the case of a larger treatment volume, CBCT-based planning could be limited. However, CBCT-based planning could be useful for realtime replanning to provide MU and dose calculation quickly while a patient is on the couch. It could be also used for verification planning to verify treatment delivery retrospectively even though it might not provide complete dosimetric data for the whole patient geometry. CONCLUSIONS This study investigated the potential feasibility of treatment planning based on CBCT. Being able to generate CBCT-based treatment plans close to CT-based plans promises a potential improvement in image-guided adaptive radiation therapy. CBCT images certainly include larger scatter and artifacts than CT images. Nevertheless, the dosimetric results in CBCT-based plans are comparable to the results in CT-based plans. We suggest that, if CBCT is used for the treatment planning purpose, CBCT should be scanned using a bowtie filter. Dosimetric data in the inhomogeneous tissue regions should be carefully validated.
REFERENCES 1. Jaffray D, Drake D, Moreau M, et al. A radiographic and tomographic imaging system integrated into a medical linear accelerator for localization of bone and soft-tissue targets. Int J Radiat Oncol Biol Phys 1999;45:773–789. 2. Mackie T, Kapatoes J, Ruchala K, et al. Image guidance for precise conformal radiotherapy. Int J Radiat Oncol Biol Phys. 2003;56:89 –105. 3. Goitein M. Organ and tumor motion: An overview. Semin Radiat Oncol 2004;14:2–9. 4. Hong T, Tome W, Chappell R, et al. The impact of daily setup variations on head-and-neck intensity-modulated ra-
diation therapy. Int J Radiat Oncol Biol Phys 2005;61:779 –788. 5. Langen K, Jones D. Organ motion and its management. Int J Radiat Oncol Biol Phys 2001;50:265–278. 6. Lattanzi J, McNeeley S, Donnelly S, et al. Ultrasound-based stereotactic guidance in prostate cancer– quantification of organ motion and set-up errors in external beam radiation therapy. Comput Aided Surg 2000;5:289 –295. 7. Letourneau D, Wong J, Oldham M, et al. Cone-beam-CT guided radiation therapy: Technical implementation. Radiother Oncol 2005;75:279 –286.
Treatment planning based on cone-beam CT
8. Mohan R, Zhang X, Wang 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 –1266. 9. Wu C, Jeraj R, Olivera G, et al. Re-optimization in adaptive radiotherapy. Phys Med Biol 2002;47:3181–3195. 10. Yan D, Vicini F, Wong J, et al. Adaptive radiation therapy. Phys Med Biol 1997;42:123–132. 11. Battista J, Rider W, Van Dyk J. Computed tomography for radiotherapy planning. Int J Radiat Oncol Biol Phys 1980;6:99 –107. 12. Constantinou C, Harrington J, DeWerd L. An electron density calibration phantom for CT-based treatment planning computers. Med Phys 1991;19:325–327. 13. Cozzi L, Fogliata A, Buffa F, et al. Dosimetric impact of computed tomography calibration on a commercial treatment planning system for external radiation therapy. Radiother Oncol 1998;48:335–338. 14. McCullough E, Holmes T. Acceptance testing computerized radiation therapy treatment planning systems: Direct utilization of CT scan data. Med Phys 1985;12:237–242. 15. Schneider U, Pedroni E, Lomax A. The calibration of CT hounsfield units for radiotherapy treatment planning. Phys Med Biol 1995;41:111–124. 16. Glover G, Pelc N. Nonlinear partial volume artifacts in x-ray computed tomography. Med Phys 1980;7:238 –248. 17. Hsieh J. Computed tomography: Principles, design, artifacts, and recent advances. Bellingham, WA: SPIE Press—The International Society for Optical Engineering, 2003. 18. Siewerdsen J, Jaffray D. Cone-beam computed tomography
19. 20. 21. 22.
23. 24. 25. 26. 27.
●
S. YOO
AND
F.-F. YIN
1561
with a flat-panel imager: Magnitude and effects of x-ray scatter. Med Phys 2001;28:220 –231. Bushberg J, Siebert J, Leidholdt E. The essential physics of medical imaging, 2nd ed. Philadelphia: Lippincott Williams and Wilkins, 2002. Whiting B, Massoumzadeh P, Sullivan J, et al. The influence of bowtie filters on X-ray CT signals. AAPM 47th Annual Meeting 2005 MO-D-I-611-07 (Seattle, WA). El-Khatib E, Battista J. Improved lung dose calculation using tissue-maximum ratios in the Batho correction. Med Phys 1984;11:279 –286. Yoo S, Yin F, Ajlouni M, et al. Commissioning of tissue inhomogeneity correction of Eclipse radiation treatment planning system using a lung phantom. AAPM 47th Annual Meeting 2005 SU-FF-T-186 (Seattle, WA). Barrett J, Keat N. Artifacts in CT: Recognition and avoidance. Radio Graphics 2004;24. Sharpe M. Imaging capabilities at delivery: KV-CT. ASTRO Image Guided Radiation Therapy (IGRT) Symposium 2006 (Las Vegas, Nevada). Endo M, Tsunoo T, Nakamori N. Effect of scatter radiation on image noise in cone beam CT. Proc SPIE 2000;3977: 514 –524. AAPM Report No 85. Tissue inhomogeneity corrections for megavoltage photon beams. Madison, WI: Medical Physics Publishing, 2004. Guan H, Yin F, Kim J. Accuracy of inhomogeneity correction in photon radiotherapy from CT scans with different settings. Phys Med Biol 2002;47:N223–231.
doi:10.1016/j.ijrobp.2006.08.031
PHYSICS CONTRIBUTION
DOSIMETRIC FEASIBILITY OF CONE-BEAM CT-BASED TREATMENT PLANNING COMPARED TO CT-BASED TREATMENT PLANNING SUA YOO, PH.D.,
AND
FANG-FANG YIN, PH.D.
Department of Radiation Oncology, Duke University Medical Center, Durham, NC Purpose: Cone-beam computed tomography (CBCT) images are currently used for positioning verification. However, it is yet unknown whether CBCT could be used in dose calculation for replanning in adaptive radiation therapy. This study investigates the dosimetric feasibility of CBCT-based treatment planning. Methods and Materials: Hounsfield unit (HU) values and profiles of Catphan, homogeneous/inhomogeneous phantoms, and various tissue regions of patients in CBCT images were compared to those in CT. The dosimetric consequence of the HU variation was investigated by comparing CBCT-based treatment plans to conventional CT-based plans for both phantoms and patients. Results: The maximum HU difference between CBCT and CT of Catphan was 34 HU in the Teflon. The differences in other materials were less than 10 HU. The profiles for the homogeneous phantoms in CBCT displayed reduced HU values up to 150 HU in the peripheral regions compared to those in CT. The scatter and artifacts in CBCT became severe surrounding inhomogeneous tissues with reduced HU values up to 200 HU. The MU/cGy differences were less than 1% for most phantom cases. The isodose distributions between CBCT-based and CT-based plans agreed very well. However, the discrepancy was larger when CBCT was scanned without a bowtie filter than with bowtie filter. Also, up to 3% dosimetric error was observed in the plans for the inhomogeneous phantom. In the patient studies, the discrepancies of isodose lines between CT-based and CBCT-based plans, both 3D and IMRT, were less than 2 mm. Again, larger discrepancy occurred for the lung cancer patients. Conclusion: This study demonstrated the feasibility of CBCT-based treatment planning. CBCT-based treatment plans were dosimetrically comparable to CT-based treatment plans. Dosimetric data in the inhomogeneous tissue regions should be carefully validated. © 2006 Elsevier Inc. Cone-beam CT, HU values, Treatment planning.
on the treatment couch. One potential approach to accomplishing the goal of replanning is to generate a modified plan based on CBCT that includes the latest information of patient positioning, target location and anatomy. Computed tomography-based treatment planning has been fundamentally validated for it has relatively longer history, and has been used as a gold standard for radiation therapy for the planning purpose (11–15). In contrast, CBCT is at its early stage in radiation therapy application and has not been evaluated for treatment planning. This study aimed to explore the dosimetric feasibility of CBCTbased treatment planning. CBCT and CT images of phantoms and patients were acquired with scanning techniques commonly used for routine clinic were used for the study. Hounsfield unit (HU) values and profiles were compared for CBCT and CT images. Dosimetric results were compared to evaluate CBCT-based plans relative to CT-based plans for both phantoms and patients.
INTRODUCTION Rapid development in image-guided radiation therapy (IGRT) has provided powerful tools for improving accuracy of patient positioning and target localization (1, 2). There have been many studies devoted to the understanding of set-up uncertainties for target localization (3– 6). Cone-beam CT (CBCT) offers a state-of-the-art technology to improve the target localization (1–2, 7). Patient positioning can be verified by matching soft tissues and/or bony structures in CBCT to those in planning CT images. However, organ motions, organ deformation, and changes of external body are often observed through the course of treatment for many patients. These target/organ variations cannot be resolved by simply repositioning the patient. One of the ideal solutions will include real-time replanning for such situations (8 –10). This replanning procedure could allow modification of the initial treatment plan to accommodate the variations while the patient is Reprint requests to: Sua Yoo, Ph.D., Department of Radiation Oncology, Duke University Medical Center, Box 3295, Durham, NC 27710. Tel: (919) 660-2179; Fax: (919) 681-7183; E-mail: [email protected]. Acknowledgments—The authors would like to acknowledge Peter
Munro, Ph.D. (Varian medical systems). This work is partially supported by a research grant from Varian Medical Systems. Received May 31, 2006, and in revised form Aug 14, 2006. Accepted for publication Aug 16, 2006. 1553
1554
I. J. Radiation Oncology
●
Biology
●
Physics
METHODS AND MATERIALS The OBI system The on-board imaging (OBI®, Varian Medical Systems, Palo Alto, CA) system consists of a kV X-ray source (KVS) and a kV amorphous-silicon digital imaging detector (KVD) mounted on the linear accelerator using robotic arms (Exact™), which are orthogonal to the electronic portal imaging device (aSi500, PortalVision™, Varian Medical systems). CBCT images were reconstructed using about 700 kV-projection images acquired over 360° rotation. When the center of the KVD is positioned at the isocenter in the longitudinal-lateral plan and 50 cm away from the isocenter in the vertical direction, the reconstructed field-of-view (FOV) is a circle of 24-cm diameter with a 15-cm length. This acquisition mode is called “full-fan” and is used for small anatomic sites such as the brain, head-and-neck, and a truncated part of larger sites. For larger sites, such as the pelvis, chest, and abdomen, the KVD is shifted by 14.8 cm laterally. In this mode, only part of the object is viewed in a half-fan projection. The other part of the object is viewed in the half-fan projection from the opposite direction. This acquisition mode is called “half-fan”. The FOV for the half-fan mode is a circle of 45-cm diameter with a 14-cm length. The effects of X-ray scatter and artifacts are larger in CBCT images than in CT images (16 –18). A bowtie filter mounted to the X-ray tube improved image quality because it reduces intensity variations across the detector and charge trapped in the detector (19, 20). Thus, there are 4 techniques available with the current CBCT acquisition: full-fan with (CBFFB), without a bowtie filter (CBFF), half-fan with (CBHFB), and without a bowtie filter (CBHF). However, the clinical protocols of CBCT should use a bowtie filter for better image quality and lower skin dose. This study included CBCT images with and without a bowtie filter for phantoms and CBCT images with a bowtie filter for patients. The user could determine the imaging geometry, the X-ray techniques, and the reconstruction parameters. For a default clinical full-fan mode, 24-cm diameter with 2.5-mm slice thickness was used, and 40-cm diameter with 2.5-mm slice thickness was used for a default half-fan mode. The acquisition X-ray technique was controlled by pre-determined parameters that were calibrated before clinical use. The techniques calibrated during installation of the OBI system were 125 kVp, 80 mA, and 25 ms with a bowtie and 125 kVp, 80 mA, and 9 ms without a bowtie. A GE Light-speed RT CT simulator (GE Medical Systems, Milwaukee, WI) was used for CT images. A “head-scan” for smaller sites such as head and neck uses 120 kVp and an automatically adjusted mA with 2.5 mm slice thickness. A “body-scan” for larger anatomic sites such as lung and pelvis uses 120 kVp and an automatically adjusted mA with 2.5 mm slice thickness. Although the automatically adjusted mA technique was used, mA during scanning was consistently found to be about 330 to 345 mA for a head scan and about 80 to 110 mA for a body scan.
HU comparisons Phantom cases. The CTP 404 in Catphan 504 (The Phantom Laboratory, Salem, NY) is designed to monitor HU values in CT images. It contains 1.2-cm diameter disks of various materials (e.g., Teflon, polystyrene, air, etc.) with different electron densities. HU values in CBCT images were calibrated using this phantom during installation. For this study, head-scan CT (CTHead), body-scan CT (CTBody), CBFFB, CBFF, CBHFB, and CBHF of Catphan were acquired. The region-of-interest (ROI) was selected
Volume 66, Number 5, 2006
for each material disk to cover a 0.7 cm ⫻ 0.7 cm square. The mean HU values in the ROIs were measured. The CIRS phantom Model 002HA6SN (CIRS Tissue Simulation and Phantom Technology, Norfolk, VA) is a homogeneous ellipsoidal phantom where a smaller cylindrical part could be separated out. The large ellipsoidal phantom, referred to as “body phantom” in this article, is 30 cm ⫻ 20 cm ⫻ 30 cm (width ⫻ height ⫻ length) and the small cylindrical phantom, referred to as “head phantom,” is 16 cm ⫻ 16 cm ⫻ 30 cm. CTHead, CBFFB, CBFF of the head phantom, and CTBody, CBHFB, and CBHF of the body phantom were acquired. HU profiles along a horizontal line across the phantoms were extracted from CT and CBCT images for comparison. The CIRS phantom Model 008 Dynamic Thorax Phantom (CIRS) was used without the motion controller. The phantom was the same size as the homogeneous body phantom, and included separate inserts representing the spine and lung containing a tumor. CTHead, CTBody, CBFFB, CBFF, CBHFB, and CBHF of the inhomogeneous phantom were acquired. Because CBFFB and CBFF could not encompass the entire phantom, the phantom was placed such that the spine and the right lung with the tumor were within the 24-cm diameter FOV. HU profiles along vertical lines crossing the lung with the tumor and along the vertical lines crossing the spine were extracted from CT and CBCT images for comparison. Patient cases. Two brain cancer patients (Patients A and B) were scanned to acquire CTHead and CBFFB. An ROI of 1.5 cm ⫻ 1.5 cm was selected to measure HU value in the brain region in CTHead and CBFFB. Two lung cancer patients (Patients C and D) were scanned to acquire CTBody and CBHFB. HU values in the lung and spine in CTBody and CBFFB were compared. Two prostate cancer patients (Patients E and F) were scanned to acquire CTBody and CBHFB. HU values in muscle, fat, and femoral head in CTBody and CBFFB were compared. All patient data were collected from routine clinical applications.
Dose comparison Phantom cases. All CT and CBCT images of the homogeneous phantoms (i.e., body and head phantoms) and the inhomogeneous phantom were imported into the ECLIPSE treatment planning system (ECLIPSE™, Varian Medical Systems). Plans based on CT and CBCT images were generated using one 10 cm ⫻ 10 cm photon beam. The plans for the homogeneous phantoms included a right lateral beam isocentered at the center of the phantoms. The beam in the plans for the head phantom had a 5-cm depth, and the beam in the plans for the body phantom had a 15-cm depth. The plans for the inhomogeneous phantom included an anterior beam isocentered at the center of the tumor passing through the right lung. The other plans for the same phantom included a posterior beam isocentered at the center of the phantom passing through the spine. The CT electron density calibration curve, which had been commissioned based on our simulation CT for the treatment planning purposes, was applied for all CT and CBCT images. Both 6MV and 15MV were used. Dose was computed with inhomogeneity correction using Modified Batho Method, which uses tissuemaximum ratio (21) and yielded the closest doses to measurements among available methods in Eclipse for both low (6MV) and high-energy (18MV) photons (22). For MU/cGy comparison, dose was normalized to the isocenter with a prescription of 100 cGy. The MU/cGy differences were obtained between CT-based and CBCT-based plans. For isodose distribution comparison, the normalization was adjusted so that the beam delivered 100 MU using
Treatment planning based on cone-beam CT
●
S. YOO
AND
F.-F. YIN
1555
the same plans. The resulting percentage isodose distributions were compared between CT- and CBCT-based plans. For the inhomogeneous phantoms, the right lung, tumor, and spine were contoured in each image and dose-volume-histograms (DVH) were compared as well. Patient cases. All CT and CBCT images of the Patients A to F were imported into ECLIPSE. 3D plans were generated for Patients A, C, and E, and IMRT plans were generated for Patients B, D, and F based on their CT images and treatment targets. The 3D plan for Patient A (brain case) included five 6 MV fields with wedges. The IMRT plan for Patient B (brain case) included six 6 MV fields. The 3D plan for Patient C (lung case) included two 15 MV fields. The IMRT plan for Patient D (lung case) included six 6 MV and fifteen MV fields. The 3D plan for Patient E (prostate case) included four 15 MV fields. The IMRT plan for Patient F (prostate case) included seven 15 MV fields. Verification plans were created based on CBCT, keeping the same MU values from the original CT-based plans. The resulting percentage isodose distributions were compared between CT-based and CBCT-based plans.
RESULTS HU comparison Phantom cases. Figure 1 shows comparison of measured mean HU values vs. expected HU values of the materials in the Catphan. The HU value differences between CBCT and CT were less than 10 HU, for all disks except for the Teflon. CBFF and CBHF displayed about 34 and 31 higher HU than CTHead and CTBody, respectively, for the Teflon. Note that the various materials in Catphan may not represent inhomogeneous tissues in actual human anatomy. Nevertheless, this study demonstrates that CBCT has the ability to generate images with HU values comparable to those in CT. Figure 2 displays the axial views of CT and CBCT images of the homogeneous and inhomogeneous phantoms. The window levels for the images were adjusted to display from ⫺1000 HU to 200 HU in grayscale. In the images of the homogeneous phantoms, shown in Fig. 2A to C, there was no noticeable difference between CBFFB, CBFF, and CTHead. When CBFF was zoomed-in, the ring artifact was discernable, as displayed in Fig. 2C-1. This ring artifact occurred due to a defect in some detector elements (17, 23).
Fig. 1. Hounsfield unit (HU) comparison in computed tomography (CT) and cone-beam computed tomography (CBCT) images of Catphan.
Fig. 2. Axial images cone-beam computed tomography; (CBCT). (A) CTHead, (B) CBFFB, (C) CBFF, (D) CTBody, (E) CBHFB, and (F) CBHF for the homogeneous phantoms. (C-1) is the magnified CBFF in the central region. (G) CTHead, (H) CBFFB, (I) CBFF, (J) CTBody, (K) CBHFB, and (L) CBHF of the inhomogeneous phantom.
However, the HU deviation in the ring artifact was less than ⫾20 HU in this study. CTBody showed uniform images whereas CBHFB displayed a dark circle beyond the head phantom as indicated by a black arrow in Fig. 2E, due to the ghosting effect (24). CBHF in Fig. 2F displayed a large dark region in the posterior region due to the cupping artifact (18, 24). The peripheral ring, indicated by the white arrow in Fig. 2H, occurred due to the normalization calibration. Normalization calibration was performed for a 25 cm diameter FOV, with a phantom covering a 24 cm diameter circle for the full-fan mode. CBCT images of the inhomogeneous phantom, Fig. 2H, I, K, and L, showed the streak artifact around the edges of the inhomogeneous objects, (i.e., spine and lung) due to beam hardening and scatter (16 –18, 25). Figure 3A and 3B shows the HU profiles extracted from the CT and CBCT axial images along the dotted lines shown in Figs. 2A and 2D. The vertical axis represents HU values scaled from ⫺200 to 200HU in Figs. 3A and 3B. The horizontal axis represents the spatial distance from the one end to the other end of the phantom: 16 cm for the head phantom and 30 cm for the body phantom. The HU profiles in the phantom periphery were reduced by 100 to 150HU in CBFFB and CBFF, compared with CTHead. The profiles of CBHFB and CBHF were more irregular than those of CBFFB and CBFF due to larger primary-to-scatter ratio with larger FOV and the larger phantom (18). The HU values in the profile of CBHFB varied from ⫺50 to 100 HU in the central region and peripheral region. The HU values in the profile of CBHF varied from ⫺150 to 0 HU with irregular distribution. The HU profiles extracted along the superior-inferior line in the frontal view of the images were
1556
I. J. Radiation Oncology
●
Biology
●
Physics
Fig. 3. Comparison of Hounsfield unit (HU) profiles. (A) The head homogeneous phantom in CTHead, CBFFB, and CBFF. (B) The body homogeneous phantom in CTBody, CBHFB, and CBHF. The inhomogeneous phantom through the tumor and right lung in (C) CTHead, CBFFB, and CBFF and in (D) CTBody, CBHFB, and CBHF. The inhomogeneous phantom through the spine in (E) CTHead, CBFFB, and CBFF and in (F) CTBody, CBHFB, and CBHF.
also compared. The same pattern of reduced HU appeared in the peripheral superior and inferior regions as observed in the axial view. Figures 3C and 3D shows the comparison of HU profiles along the vertical line cutting through the right lung and the tumor, as shown by the white dotted line in Fig. 2G. The profiles along the tumor and near the top and bottom of the body regions of CT images were close to 0 HU, whereas the profiles in the same regions of CBCT images varied from ⫺150 to 0 HU. The further reduction in HU values occurred in the periphery edge of the tumor and near the top and bottom body regions. The profiles within the lung of CBCT images showed reduced HU values ranging from 100 to 200 HU values compared with those of CT images. The nonuniformity and reduction of HU values became worse in CBCT images acquired without a bowtie filter than in CBCT images with a bowtie filter. The HU profiles extracted along the superior-inferior line in the frontal view of CBCT images illustrated reduced HU values. Figures 3E and 3F shows the comparison of HU profiles along the vertical line cutting through the body and the spine as shown in the black dotted line in Fig. 2G. The profiles of CBCT images showed reduced HU values rang-
Volume 66, Number 5, 2006
ing from ⫺50 to ⫺200 HU at the junctions between the body and the external air and between the body and the spine. The same pattern occurred in the profiles of the homogeneous phantoms as shown in Figs. 3A and 3B. The profiles of CBCT with a bowtie within the spine were close to those of CT, whereas the profiles of CBCT without a bowtie were about 100 HU lower. The posterior area between the spine and the external air showed reduced HU because of the streak artifact (18). The profiles in this posterior area showed interesting distributions; those of CBFFB/CBFF reached about ⫺200 HU whereas those of CBHFB/CBHF reached about ⫺100 HU. It was because the phantom was placed so that the posterior body was imaged near the edge of the FOV, where reduced HU values occurred more substantially in addition to the streak artifact. Patient cases. Table 1 summarizes the comparison of HU values in CT and CBCT images of patients for various tissue types. As much as possible, ROIs in CT and CBCT for each tissue type were selected so that they enclosed the same tissue type in the same location. All tissues, except the brain for Patient A and the fat for Patient F, showed lower HU values in CBCT than in CT. The HU differences were less than 70 HU values for all tissues except for the back muscle region in the vicinity of the spine. The HU differences in this back muscle between CT and CBCT were greater than 300 HU. As shown with the HU profiles of the posterior body region in Figs. 3E and 3F, the streak artifact contributed to the reduction of HU values in this area. Noticeable streak artifacts in small volumes also occurred in the CBCT images of the lung cancer patients because of breathing motion.
Table 1. Comparison of HUs in different tissues of patients Tissue type Brain cases Brain Lung cases Lung Spine Back muscle in the vicinity of the spine Prostate cases Muscle Fat Femoral head
HU ⫾ SD
HU ⫾ SD
CTHead CBFFB
Patient A 37 ⫾ 6.8 84 ⫾ 11.1
Patient B 60 ⫾ 33.2 54 ⫾ 18.4
CTBody CBHFB CTBody CBHFB CTBody CBHFB
Patient C ⫺965 ⫾ 40.2 ⫺978 ⫾ 38.4 184 ⫾ 46.8 110 ⫾ 50.6 2 ⫾ 46.2 ⫺337 ⫾ 74.6
Patient D ⫺887 ⫾ 191.6 ⫺910 ⫾ 82.9 226 ⫾ 112.5 218 ⫾ 81.1 ⫺10 ⫾ 42.8 ⫺326 ⫾ 52.0
CTBody CBHFB CTBody CBHFB CTBody CBHFB
Patient E 67 ⫾ 16.0 48 ⫾ 34.8 ⫺90 ⫾ 14.7 ⫺112 ⫾ 44.4 363 ⫾ 65.6 327 ⫾ 64.5
Patient F 65 ⫾ 25.1 61 ⫾ 54.1 ⫺104 ⫾ 16.3 ⫺90 ⫾ 43.3 414 ⫾ 95.5 352 ⫾ 90.1
Image
Abbreviations: CT ⫽ computed tomography; HU ⫽ Hounsfield unit; SD ⫽ standard deviation. Mean HU values ⫾ standard deviations were measured over 1.5 cm ⫻ 1.5 cm region of interest in each tissue area.
Treatment planning based on cone-beam CT
●
S. YOO
AND
F.-F. YIN
1557
Table 2. Comparison of the percent difference, ⌬, in MU/cGy between CT-based and CBCT-based plans Phantom
Homogeneous phantoms
Plan
(A) Lat beam
Inhomogeneous phantom (B) AP beam
(C) PA beam
Energy
6 MV
15 MV
6 MV
15 MV
6 MV
15 MV
CBFFB CBFF CBHFB CBHF
⫺0.58 ⫺0.41 0.67 ⫺2.55
⫺0.42 ⫺0.31 0.43 ⫺1.75
⫺2.30 ⫺3.07 ⫺2.62 ⫺2.45
⫺1.50 ⫺2.07 ⫺1.78 ⫺1.68
⫺2.38 ⫺3.44 ⫺0.39 ⫺2.01
⫺1.61 ⫺2.27 ⫺0.28 ⫺1.38
Abbreviations: CBCT ⫽ cone-beam computed tomography; CT ⫽ computed tomography; AP ⫽ anterior-posterior; PA ⫽ posterior-anterior. ⌬ ⫽ (MU/cGy of CBCT – MU/cGy of CT)/ MU/cGy of CT*100. (A) Comparison of the plans with the lateral beam isocentered at the center of the homogeneous phantom. (B) Comparison of the plans with the AP beam isocentered at the center of the tumor in the right lung for the inhomogeneous phantom. (C) Comparison of the plans with the PA beam isocenterd at the center of the phantom passing through the spine for the inhomogeneous phantom.
Dose comparison Phantom cases. Table 2(A) summarizes the percent difference in MU/cGy (⌬) between the CT-based and CBCTbased plans for the homogeneous phantoms. The negative ⌬ represents that the CBCT-based plan under-estimated MU/ cGy compared to the CT-based plan. All CBCT-based plans, except the CBHF-based plans, showed ⌬s less than 1%. The CBHF-based plans showed ⫺2.55% and ⫺1.75% of ⌬s for 6 MV and 15 MV, respectively. Relatively larger ⌬s variations in the CBHF-based plans were the results from the reduced HU values in the central posterior area as shown in Fig. 2F. The results using a 15 MV beam showed better agreement than those using a 6 MV beam. This is because tissue-air ratio or tissue maximum ratio of higher energy is less affected by the change of depth. Therefore,
Fig. 4. Comparison of isodose distributions of CT-based and CBCT-based plans using 6 MV photon for the homogeneous phantoms. Panels A and C are axial and frontal views along the central axis for plans based on CTHead, CBFFB, and CBFF for the head phantom. Panels B and D are for plans based on CTBody, CBHFB, and CBHF for the body phantom. The number in each line is in the relative dose (%) to the prescription of 100 cGy. CBCT ⫽ cone-beam computed tomography; CT ⫽ computed tomography.
higher energy photon beam is associated with lower tissueinhomogeneity correction factors (26). Figure 4 displays relative isodose distributions of the plans using 100 MU of 6 MV in the axial and frontal views along the central axis. The isodose lines from CBCT-based plans agreed well with those from CT-based plans except those from CBHF-based plans. Along the central axis, the discrepancies of 80% and 50% lines between CBHF-based and CT based plans were about 5 mm and 10 mm, respectively, which were equivalent to approximately 2% and 3% of differences in dose. These discrepancies increased in the posterior part where the cupping artifact was observed. The same trends were found in the frontal view. It was observed that plans using a 15 MV beam showed better agreement with CT-based plans than using a 6 MV beam. Table 2 (B) summarizes the ⌬s between the CT- and CBCT-based plans generated using the inhomogeneous phantom with the anterior beam isocentered at the center of the tumor in the right lung. The ⌬s ranged from ⫺1.50 to ⫺2.62% for plans based on CBCT with a bowtie filter and from ⫺1.6% to ⫺3.07% for plans based on CBCT without a bowtie filter. Based on the results shown in Figs. 3C and 3D, under-estimation of MU/cGy was expected for all CBCT-based plans. Plans based on CBCT with a bowtie filter showed better dosimetric agreement with plans based on CT than plans based on CBCT without a bowtie filter. Figure 5A–D displays relative isodose distributions of the plans using 100 MU of 6 MV in the axial and sagittal views along the central axis. The discrepancies of isodose lines between plans based on CTBody and CBHFB/CBHF were about 5 mm and 10 mm for 100% and 80% lines, respectively, which were equivalent to approximately 2% and 3% of differences in dose. The same discrepancies occurred between plans based on CTHead and CBFFB. The larger discrepancy occurred between the CTHead-based plan and CBFF-based plan, where the discrepancy in 100% and 80% isodose lines reached 7 mm and 17 mm, respectively, which were equivalent to 4% and 5% of difference in dose. Again, the isodose distributions in the plans using a 15 MV beam
1558
I. J. Radiation Oncology
●
Biology
●
Physics
Fig. 5. Comparison of CT-based and CBCT-based plans using 6 MV photon for the tumor in the right lung in the inhomogeneous phantom. Panels A and C are isodose distributions in the axial and sagittal views along the central axis for plans based on CTHead, CBFFB, and CBFF. Panels B and D are for plans based on CTBody, CBHFB, and CBHF. DVHs of plans based on (E) CTHead, CBFFB, and CBFF, and (F) CTBody, CBHFB, and CBHF. CBCT ⫽ conebeam computed tomography; CT ⫽ computed tomography.
showed better agreement than plans using a 6 MV beam. Figures 5E and 5F show the DVHs for the same plans. DVHs of CBFFB- and CBHFB/CBHF-based plans showed about 3% of dose-increase for the entire volume of the tumor and about 30 –70% volume of the right lung compared to the CT-based plans. CBFF-based plans showed about 5% of dose increment for the tumor and the right lung. Table 2 (C) summarizes the ⌬s between the CT-based and CBCT-based plans for the inhomogeneous phantom with the posterior beam isocentered at the center of the body passing through the spine. CBHFB-based plans showed very good agreement (⬍1%) with CTBody-based plans. Plans based on CBCT with a bowtie filter showed smaller ⌬s than plans based on CBCT without a bowtie. CBFF/ CBFFB-based plans for this case resulted in large ⌬s: ⫺1.61 to ⫺3.44%. These large ⌬s occurred not because of the spine but because of the posterior body region, which displayed further reduced HU values in the profiles shown in Fig. 3E. Again, the ⌬s were larger in 6 MV than in 15 MV. Figure 6A–D displays relative isodose distributions of the plans using 100 MU of 6 MV in the axial and sagittal views along the central axis. The isodose lines of the CBCT-based plans reached slightly further anteriorly than those of the CT-based plans. Isodose lines of CTBody- and CBHFB-
Volume 66, Number 5, 2006
based plans showed good agreement. The discrepancy between CTBody- and CBHF-based plans was 1 mm and 3 mm for 80% and 50% lines along the central axis, respectively, which were equivalent to about 1% of differences in dose. The discrepancies of isodose lines between CTHead- and CBFFB-based plans were 4 mm and 5 mm for 50% and 80% lines along the central axis, respectively, which were equivalent to about 1% of differences in dose. The discrepancies were increased between CTHead- and CBFF-based plans; 5 mm for 80% and 10 mm for 50% lines, equivalent to about 2% and 3% of differences in dose, respectively. The discrepancy of isodose lines between CTHead- and CBFFB/CBFF-based plans was larger than the discrepancy of those between CTBody- and CBHFB/CBHF-based plans. Figures 6E and 6F show comparison of DVHs from the same plans. DVHs of CBFFB/CBFF-based and CBHFbased plans showed about 1.5% and 1% of dose-increase for the spine, respectively, compared to CT-based plans. DVH of the spine in CBHFB-based plan overlapped with that in CTBody-based plan. The DVHs of the right lung from all plans overlapped as well. Patient cases. Figure 7 shows comparisons of the patient cases. The 100% dose line from CBCT-based plan covered a larger area than the same line from the CT-based plan for Patient A in Fig. 7C. However, the 99% line from the CBCT-based plan agreed with the 100% dose line from the
Fig. 6. Comparison of CT-based and CBCT-based plans using 6 MV photon for the spine in the inhomogeneous phantom. Panels A and C are isodose distribution in the axial and sagittal views along the central axis for plans based on CTHead, CBFFB, and CBFF. Panels B and D are for plans based on CTBody, CBHFB, and CBHF. DVHs of plans based on (E) CTHead, CBFFB, and CBFF, and (F) CTBody, CBHFB, and CBHF. CBCT ⫽ cone-beam computed tomography; CT ⫽ computed tomography.
Treatment planning based on cone-beam CT
●
S. YOO
AND
F.-F. YIN
1559
Also, there are small discrepancy in the 20% and 40% isodose lines due to set-up error between CT and CBCT. The CT of Patient F in Fig. 7P was scanned with a contrast material. Therefore, the bladder was contoured and assigned with 0 HU value for the CT-based plan. CBCT in Fig. 7Q appeared grainier than the CT in Fig. 7P because of increased scatter due to the large size of Patient F. The 45-cm diameter FOV did not enclose the patient body in CBCT. The external body was modified in CT such that the distance from one end to the other end in CT matched with that in CBCT. The 3D plans based on CT and CBCT of Patient E showed good agreement in most dose lines, as shown in Fig. 7O. The discrepancies occurred around the central region of 100% isodose line and the posterior end in 20% and 40% isodose lines. The CBCT-based plan delivered higher dose to the central region than the CT-based plan mainly because of the additional gas-filled in rectum. The 2-mm discrepancy around the posterior end happened because of the set-up error. The IMRT plans based on CT and CBCT for Patient F showed good agreement in their dose distributions, as shown in Fig. 7R, except in the posterior and left ends due to the set-up error. DISCUSSION
Fig. 7. Comparison of isodose distributions of CT-based and CBCT-based plans for (A) to (F) the brain cancer patients, (G) to (L) the lung cancer patients, and (M) to (R) the prostate cancer patients. (A), (D), (G), (L), (M), and (P) are the CT images, (B), (E), (H), (K), (N), and (Q) are the CBCT images, and (C), (F), (I), (L), (O), and (R) are the isodose distributions. The legend of dose lines is in Fig. 7(F). CBCT ⫽ cone-beam computed tomography; CT ⫽ computed tomography.
CT-based plan, indicating that the dose difference was only 1% in the high dose region for Patient A. Other isodose lines in Patient A and B showed good agreement In the CBCT images of the lung cancer patients shown in Figs. 7H and 7K, it is noticeable that the back muscle in the vicinity of the spine appeared darker (i.e., lower HU values) than other body regions. The streak artifact caused by breathing motion also appeared around the lungs and chest wall, but the size of the artifact was much smaller than that in the posterior region of the spine. The comparison of CT-based and CBCT-based plans showed good agreement between 20% to 80% isodose lines, but there were some discrepancy in the 105% isodose line. The discrepancy was larger in the 3D plan, shown in Fig. 7I, than in the IMRT plan in Fig. 7L. The increased dose in CBCT-based plans compared to CT-based plans was from the reduced HU values in the lungs and the posterior region of the spine.
This study investigated the dosimetric feasibility of CBCT-based treatment planning. The differences in HU values between CT and CBCT for the Catphan were less than 10 HU for most disk materials except the Teflon, which showed a maximum difference of 34 HU. The profiles of CBCT in the homogeneous phantoms displayed 100 to 150 lower HU at the periphery edge in the full-fan mode, and around the periphery edge and central regions in the half-fan mode, than the profiles of CT. Overall, the profiles of CBCT in the inhomogeneous phantom displayed 50 to 200 HU lower than those of CT. The HU values in CBCT of patients were also lower than those in CT. The back muscle in the vicinity of the spine showed about 300 HU lower in CBCT than in CT. The dosimetric consequence as the result of HU variation in CBCT was evaluated by comparing MU/cGy, isodose distributions, and DVHs. For the homogeneous phantoms, the discrepancy between the plans was small, with less than 1% of MU/cGy difference and good agreement in isodose lines when CBCT was scanned with a bowtie. The dosimetric discrepancy became larger for the inhomogeneous phantom when beams passed through the lung or the posterior body images at the edge of FOV. The differences were about 2–3% in MU/cGy and about 5–11 mm (equivalent to 2–3%) in isodose lines for the CBCT with a bowtie filter. In general, the 3D and IMRT plans based on CBCT for patients showed good agreement with the plans based on CT in general. About 2 mm discrepancy between 2 isodose lines occurred because of the set-up error. Discrepancy also occurred in the high-dose region of the 3D plan and the IMRT plan for the lung cancer patients. The regions enclosed by the 105% isodose lines were shaped differently between the
1560
I. J. Radiation Oncology
●
Biology
●
Physics
2 plans. The CBCT-based plans showed a larger region enclosed by 105% lines than CT-based plans. However, other areas enclosed by isodose lines below 90% matched very well. We set scanning techniques for CT and CBCT as close as possible. The current CBCT configuration uses 125 kVp whereas our CT scanner does not provide 125 kVp. Some investigations have sought to determine the impact of kVp to HU values and dose calculation. McCullough and Holmes stated that changing from 120 kVp to 140 kVp did not produce significant change in HU (14). Yet, Cozzi et al. reported that voltage variation from 100 to 140 kVp caused a maximum of 300 HU difference in high-density materials—a difference that resulted in a maximum of 4% dosimetric error (13). However, we found that CT images of Catphan using 120 kVp and 140 kVp produced a maximum difference of 10 HU. Thus, using 120 kVp instead of 125 kVp in CT would not produce considerable variations for this study. CBCT uses 80 mA and 9 ms without a bowtie filter, and 25 ms with a bowtie filter, whereas CT automatically adjusts mA. Cozzi et al. reported that exposure time (ms) and current (mA) did not affect the HU values (13). We also found that CT images of Catphan using automatically adjusted mA produced very small random differences in HU values (⬍ ⫾2 to ⫾5 HU) compared to CT images using fixed 80 mA. Therefore, the techniques used for CT acquisition were appropriate for this study. The difference in HU values in Catphan was smaller than the difference in the other phantoms and patients. The magnitude of scatter and artifacts in CBCT images are affected by imaging geometry, object size, and inhomogeneous tissues (18). Scatter and artifacts deteriorate image uniformity and mislead HU values in 3D reconstruction. An object like Catphan with small disks provides less scatter and artifacts than a large object with large inhomogeneous tissues (16, 18). The location of an object and neighboring tissue types/sizes also affects HU values. This study did not investigate HU variation caused by the various factors. However, the cases simulated in this study could represent common clinical situations. Some studies recommend that the relationship between electron densities and CT numbers (i.e., HU-ED curve) should be stored in the treatment planning system for accurate dose calculation (12, 15). It was reported that up to 2% of dosimetric differences were found by using different
Volume 66, Number 5, 2006
HU-ED curves for 250 to 300 HU differences (13, 27). However, for 2 reasons our study did not include a HU-ED generated specifically for CBCT. First, we found maximum of 34 HU difference in CBCT without a bowtie and 15 HU difference in CBCT with a bowtie filter, compared with CT. HU differences smaller than 34 HU values in a HU-ED curve will not affect dosimetric results. Second, there is no proper way to create a standard HU-ED curve of CBCT because the magnitude of scatter varies with patient-dependant factors such as size and location of patient body, inhomogeneous tissues, and neighboring tissues. The results for CBCT-based plans provided in this study were comparable to those for CT-based plans. However, there are some limitations in CBCT-based treatment planning. Due to the limited field size of the detector, only about 14 cm in-length could be scanned in CBCT whereas many of clinical cases require a larger view. Acquiring 2 sets of CBCT, 1 for a superior body and the other for an inferior body by shifting the couch, could provide a larger superiorinferior view. However, a user should understand that the periphery of CBCT image show reduced HU values and setting an isocenter in the superior or inferior edge of the image is not desirable. Thus, in the case of a larger treatment volume, CBCT-based planning could be limited. However, CBCT-based planning could be useful for realtime replanning to provide MU and dose calculation quickly while a patient is on the couch. It could be also used for verification planning to verify treatment delivery retrospectively even though it might not provide complete dosimetric data for the whole patient geometry. CONCLUSIONS This study investigated the potential feasibility of treatment planning based on CBCT. Being able to generate CBCT-based treatment plans close to CT-based plans promises a potential improvement in image-guided adaptive radiation therapy. CBCT images certainly include larger scatter and artifacts than CT images. Nevertheless, the dosimetric results in CBCT-based plans are comparable to the results in CT-based plans. We suggest that, if CBCT is used for the treatment planning purpose, CBCT should be scanned using a bowtie filter. Dosimetric data in the inhomogeneous tissue regions should be carefully validated.
REFERENCES 1. Jaffray D, Drake D, Moreau M, et al. A radiographic and tomographic imaging system integrated into a medical linear accelerator for localization of bone and soft-tissue targets. Int J Radiat Oncol Biol Phys 1999;45:773–789. 2. Mackie T, Kapatoes J, Ruchala K, et al. Image guidance for precise conformal radiotherapy. Int J Radiat Oncol Biol Phys. 2003;56:89 –105. 3. Goitein M. Organ and tumor motion: An overview. Semin Radiat Oncol 2004;14:2–9. 4. Hong T, Tome W, Chappell R, et al. The impact of daily setup variations on head-and-neck intensity-modulated ra-
diation therapy. Int J Radiat Oncol Biol Phys 2005;61:779 –788. 5. Langen K, Jones D. Organ motion and its management. Int J Radiat Oncol Biol Phys 2001;50:265–278. 6. Lattanzi J, McNeeley S, Donnelly S, et al. Ultrasound-based stereotactic guidance in prostate cancer– quantification of organ motion and set-up errors in external beam radiation therapy. Comput Aided Surg 2000;5:289 –295. 7. Letourneau D, Wong J, Oldham M, et al. Cone-beam-CT guided radiation therapy: Technical implementation. Radiother Oncol 2005;75:279 –286.
Treatment planning based on cone-beam CT
8. Mohan R, Zhang X, Wang 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 –1266. 9. Wu C, Jeraj R, Olivera G, et al. Re-optimization in adaptive radiotherapy. Phys Med Biol 2002;47:3181–3195. 10. Yan D, Vicini F, Wong J, et al. Adaptive radiation therapy. Phys Med Biol 1997;42:123–132. 11. Battista J, Rider W, Van Dyk J. Computed tomography for radiotherapy planning. Int J Radiat Oncol Biol Phys 1980;6:99 –107. 12. Constantinou C, Harrington J, DeWerd L. An electron density calibration phantom for CT-based treatment planning computers. Med Phys 1991;19:325–327. 13. Cozzi L, Fogliata A, Buffa F, et al. Dosimetric impact of computed tomography calibration on a commercial treatment planning system for external radiation therapy. Radiother Oncol 1998;48:335–338. 14. McCullough E, Holmes T. Acceptance testing computerized radiation therapy treatment planning systems: Direct utilization of CT scan data. Med Phys 1985;12:237–242. 15. Schneider U, Pedroni E, Lomax A. The calibration of CT hounsfield units for radiotherapy treatment planning. Phys Med Biol 1995;41:111–124. 16. Glover G, Pelc N. Nonlinear partial volume artifacts in x-ray computed tomography. Med Phys 1980;7:238 –248. 17. Hsieh J. Computed tomography: Principles, design, artifacts, and recent advances. Bellingham, WA: SPIE Press—The International Society for Optical Engineering, 2003. 18. Siewerdsen J, Jaffray D. Cone-beam computed tomography
19. 20. 21. 22.
23. 24. 25. 26. 27.
●
S. YOO
AND
F.-F. YIN
1561
with a flat-panel imager: Magnitude and effects of x-ray scatter. Med Phys 2001;28:220 –231. Bushberg J, Siebert J, Leidholdt E. The essential physics of medical imaging, 2nd ed. Philadelphia: Lippincott Williams and Wilkins, 2002. Whiting B, Massoumzadeh P, Sullivan J, et al. The influence of bowtie filters on X-ray CT signals. AAPM 47th Annual Meeting 2005 MO-D-I-611-07 (Seattle, WA). El-Khatib E, Battista J. Improved lung dose calculation using tissue-maximum ratios in the Batho correction. Med Phys 1984;11:279 –286. Yoo S, Yin F, Ajlouni M, et al. Commissioning of tissue inhomogeneity correction of Eclipse radiation treatment planning system using a lung phantom. AAPM 47th Annual Meeting 2005 SU-FF-T-186 (Seattle, WA). Barrett J, Keat N. Artifacts in CT: Recognition and avoidance. Radio Graphics 2004;24. Sharpe M. Imaging capabilities at delivery: KV-CT. ASTRO Image Guided Radiation Therapy (IGRT) Symposium 2006 (Las Vegas, Nevada). Endo M, Tsunoo T, Nakamori N. Effect of scatter radiation on image noise in cone beam CT. Proc SPIE 2000;3977: 514 –524. AAPM Report No 85. Tissue inhomogeneity corrections for megavoltage photon beams. Madison, WI: Medical Physics Publishing, 2004. Guan H, Yin F, Kim J. Accuracy of inhomogeneity correction in photon radiotherapy from CT scans with different settings. Phys Med Biol 2002;47:N223–231.