- Email: [email protected]
Evaluation of Tumor Position and PTV Margins Using Image Guidance and Respiratory Gating
Int. J. Radiation Oncology Biol. Phys., Vol. 76, No. 5, pp. 1578–1585, 2010 Copyright Ó 2010 Elsevier Inc. Printed in the USA. All rights reserved 0360-3016/10/$–see front matter
doi:10.1016/j.ijrobp.2009.08.002
PHYSICS CONTRIBUTION
EVALUATION OF TUMOR POSITION AND PTV MARGINS USING IMAGE GUIDANCE AND RESPIRATORY GATING CHRISTOPHER NELSON, PH.D.,* PETER BALTER, PH.D.,* RODOLFO C. MORICE, M.D.,y KARA BUCCI, M.D.,z LEI DONG, PH.D.,* SUSAN TUCKER, PH.D.,{ SASTRY VEDAM, PH.D.,* JOE Y. CHANG, M.D., PH.D.,z AND GEORGE STARKSCHALL, PH.D.* From the Departments of *Radiation Physics, yPulmonary Medicine, zRadiation Oncology, and {Bioinformatics & Computational Biology, The University of Texas M. D. Anderson Cancer Center, Houston, TX Purpose: To evaluate the margins currently used to generate the planning target volume for lung tumors and to determine whether image-guided patient setup or respiratory gating is more effective in reducing uncertainties in tumor position. Methods and Materials: Lung tumors in 7 patients were contoured on serial four-dimensional computed tomography (4DCT) data sets (4–8 4DCTs/patient; 50 total) obtained throughout the course of treatment. Simulations were performed to determine the tumor position when the patient was aligned using skin marks, image-guided setup based on vertebral bodies, fiducials implanted near the tumor, and the actual tumor volume under various scenarios of respiratory gating. Results: Because of the presence of setup uncertainties, the reduction in overall margin needed to completely encompass the tumor was observed to be larger for imaged-guided patient setup than for a simple respiratorygated treatment. Without respiratory gating and image-guided patient setup, margins ranged from 0.9 cm to 3.1 cm to completely encompass the tumor. These were reduced to 0.7–1.7 cm when image-guided patient setup was simulated and further reduced with respiratory gating. Conclusions: Our results indicate that if respiratory motion management is used, it should be used in conjunction with image-guided patient setup in order to reduce the overall treatment margin effectively. Ó 2010 Elsevier Inc. Image-guided patient setup, Respiratory gating, 4DCT, Lung cancer, Target volume.
The general concept of the planning target volume (PTV) is to create a structure that accounts for the geometric uncertainties in aligning the clinical target volume (CTV) with the coordinates of the treatment machine. These uncertainties are due to internal motion of the target as well as uncertainties in patient setup. The PTV is designed based on the tumor location at the time of simulation (1) and the statistically determined daily setup uncertainties of the patients. Larger PTV margins ensure adequate coverage of the tumor, but also increase the amount of healthy tissue irradiated (2). PTV margins must be adequate to ensure tumor coverage and yet must be kept small enough to prevent excess radiation-induced toxicity of normal tissue, which is often the limiting factor when escalating the dose. In the radiation treatment of lung tumors, the amount of healthy tissue irradiated limits dose escalation strategies, which have been shown to increase tumor control (3–5).
Two techniques are typically used to reduce the uncertainty in tumor location during treatment delivery, particularly for lung tumors and others subject to respirationinduced motion: respiratory-correlated radiation therapy and image-guided radiation therapy (IGRT). In the case of lung tumors, respiration-induced movement (6–8) is greatest for those located near the diaphragm. Respiratory-controlled radiation therapy reduces the effects of respiratory motion (that is, intrafractional uncertainty) by linking beam delivery with respiration, either by treating during a breath hold or by gating the delivery of the treatment beam. IGRT, conversely, is not used to account for respiratory motion but to adjust for daily (interfractional) changes in tumor position (prior to treatment delivery). Daily IGRT reduces uncertainties in tumor position by allowing the radiation therapist to realign the tumor with its initial position at the time of simulation, thereby reducing the likelihood of a geometric miss (9). IGRT can be implemented using different alignment
Reprint requests to: Christopher Nelson, Ph.D., Department of Radiation Physics, Unit 525, The University of Texas M. D. Anderson Cancer Center, 1515 Holcombe Blvd., Houston, TX 77030. Tel: (713) 563-2471; Fax: (713) 563-6949; E-mail: chnelson@ mdanderson.org
Conflict of interest: none. Acknowledgment—The authors would like to acknowledge Keith Britton for his contribution to this project. Received Feb 12, 2009, and in revised form Aug 3, 2009. Accepted for publication Aug 4, 2009.
INTRODUCTION
1578
Reduction of Setup Uncertainties in Lung Cancer d C. NELSON et al.
techniques; commonly based on bony landmarks, fiducials implanted in the periphery of the tumor, or tumor itself. Different alignment techniques are sometimes limited by the imaging capabilities of the machine whether it has twodimensional or three-dimensional imaging (10, 11). Although these techniques are routinely used to reduce the two types of uncertainty in tumor position, questions remain about their optimal application. For instance, image-guided positioning can be accomplished by means of various strategies, but it is unclear which of these strategies optimizes the relationship between the CTV and PTV. In addition, the question remains whether a respiratory-gated treatment should be delivered unless the patient has first been aligned by means of IGRT. The purpose of this study is to evaluate the margins currently used to generate the PTV for lung tumors and to determine whether image-guided patient setup or respiratory gating is more effective in reducing uncertainties in tumor position. To achieve these aims, we used actual patient data in simulations of four different imageguided setup techniques and four degrees of respiratory gating to examine the location of the tumor throughout the course of treatment and evaluated the fraction of tumor outside of the standard PTV to compared image-guided setup with respiratory gating. METHODS AND MATERIALS Patient selection/image acquisition Data from 7 patients with primary lung tumors who had been treated on investigational protocols at our institution were chosen for the current analysis. Five patients had undergone bronchoscopic implantation of gold fiducials (NMPE, Inc., Lynwood, WA) in the periphery of their lung tumors (12) after giving informed consent in the institutional review board–approved protocol ID03-0208. In addition, we examined data for 2 patients who had been enrolled on institutional review board–approved protocol ID2003-0962 and subjected to imaging in accordance with the protocol. Patients eligible for both protocols were to be treated definitively for non– small-cell lung cancer with radiation for a minimum of 6 weeks. Concurrent chemotherapy was allowed, but patients could not have received prior radiation treatments. Conventional, four-dimensional computed tomography (4DCT) simulation procedures were used for all patients accrued in this study. For each 4DCT, the patient’s respiration was monitored by tracking the vertical displacement of the abdomen (RPM, Varian Medical Systems, Palo Alto, CA). The respiratory signal was used to retrospectively sort images acquired in cine mode into the 4DCT data set (13). At the time of this study, all patients with thoracic tumors who were treated at our institution were immobilized for 4DCT simulation using a vacuum bag (Vac-Loc) placed on a wing-board/T-bar (both from MedTec, Orange City, IA). Three tattoos were placed on each patient’s skin at the level of the carina for daily alignment with the lasers in the treatment room. BBs were placed on the skin marks to make the marks’ locations visible during each imaging session. The first 4DCT scan was acquired for treatment planning and to assess tumor motion. Weekly 4DCT data sets were then acquired throughout the remainder of the patient’s treatment. For each 4DCT data set acquired, BBs were placed on the patient’s skin marks and the patient was immobilized using the simulation procedures.
1579
Fig. 1. Illustration showing how fractional volumes (%FV) is computed from the tumor contours and the planning target volume (PTV) generated from internal target volume on each phase of the first four-dimensional computed tomography data set (ITVsim).
Tumor delineation For each 4D data set, the gross tumor volume (GTV) was contoured on the 50% phase (although not always the case, we chose the 50% phase as end expiration). Using a rigid contour propagation software system (14), we propagated the GTV contours to the remaining nine phases of the 4DCT data set for review by the physician. With GTV contours on each phase of each 4DCT data set, we then performed four alignments to simulate different patient setup scenarios. Before aligning the computed tomography (CT) data sets, all contours from each data set were copied to the 50% phase so that when the image alignment was performed using a single phase (50%), the contours on the remaining nine phases would translate with the 50% phase that was being aligned. After the alignment was completed, each GTV contour (total = # 4DCT data sets 10) was uniformly expanded by 8 mm to generate the CTV, in accordance with the standard of practice in our institution and based on the work of Giraud et al. (15).
Simulation of patient setup scenarios Four separate alignments were performed to simulate different daily patient setup scenarios: conventional patient setup in which the patient is aligned using external skin marks, image-guided patient setup in which the patient is aligned based on the location of vertebral bodies (bony landmarks), image-guided patient setup in which the patient is aligned based on fiducials implanted near the periphery of the tumor, and image-guided patient setup in which the patient is aligned based on the location of the tumor itself. The first alignment was performed to simulate patient setup without image guidance (i.e., aligning the skin marks of the patient with the lasers in the treatment room). To perform this alignment, an isocenter was identified on the 50% phase of the 4DCT data set from each week as determined by the BBs placed at the time of simulation. The isocenters from each week were aligned to the isocenter point on the 50% phase of the first 4DCT data set acquired at simulation, effectively simulating patient alignment by skin marks. The remaining three CT alignments were performed to simulate the use of image guidance to set the patient up before treatment delivery. To align the data sets based on vertebral bodies, the same image registration technique (14) was used to align the vertebral bodies in the 50% phase of each subsequent 4DCT data set to
I. J. Radiation Oncology d Biology d Physics
1580
the vertebral bodies in the 50% phase of the first 4DCT data set acquired at simulation. The vertebral bodies that were used for the alignment were chosen by contouring the vertebral body on the simulation CT data set that also contained the contours of the tumor; thus, the vertebral body used for the registration was in the same superoinferior plane as the tumor. The image registration method is based on the image intensities within the region of interest for the contoured vertebral body in the simulation CT data set. For simulation of alignment based on implanted fiducial location, the CT data sets were aligned by registering the location of the fiducial on the 50% phase of each subsequent CT data set to the location in the 50% phase at simulation. Finally, image alignment based on the location of the tumor was done. To minimize the effects of changing tumor size/shape/contouring uncertainties from week to week, the treatment planning system was used to identify the center of volume (COV) of the GTV contour on the 50% phase of each 4DCT data set. The CT alignments were then performed so that the GTV COV in the 50% phase from each subsequent week was aligned with the GTV COV in the 50% phase of the initial 4DCT acquired at simulation. The alignments based on tumor and fiducial location simulated one in which a gated image was acquired and the alignment was based on the gated image. After each of the alignments, we had a series of tumor contours acquired during free breathing (10 phases) simulating patient alignment using external skin marks, as well as simulating image guidance aligning to fiducials implanted near the tumor, the vertebral bodies, and the tumor.
Simulation of gated treatment The goal of respiratory gating is to allow beam delivery in a patient’s treatment only during a small portion of the respiratory cycle. Given tumor contours for each phase of multiple 4DCT data sets representing the patient during normal respiration (100% duty cycle, no gating), respiratory-gated treatment can easily be simulated by simply removing particular phases from the set of tumor contours. We generated three different gated treatment scenarios, representing treatment delivery with 50%, 30%, and 10% duty cycles at end expiration. To simulate a 50% duty cycle, the tumor contours from the five phases surrounding end inspiration (i.e., the 0% phase) were removed leaving only the contours from the 30%, 40%, 50%, 60%, and 70% phases of the 4DCT data set. To simulate a 30% duty cycle, we used only the contours from the 40%, 50%, Table 1. Baseline tumor volumes during the simulation 4DCT, the number of fiducials implanted, and the total number of 4DCTs used in this study Tumor volumes (cm3) Tumor GTVmip ITVsim PTV7mm PTV3+7mm # Fiducials # 4DCTs 1 2 3 4a 4b 5 6 7
137.3 43.9 235.6 2.7 3.0 14.6 94.4 136.1
305.4 125.4 500.4 23.4 24.0 65.3 258.4 326.5
524.1 248.1 821.5 70.2 71.0 140.8 472.7 565.8
649.0 323.0 996.1 103.2 104.3 197.7 593.1 700.9
2 1 1 4 2 0 0
8 6 8 7 4 9 8
Abbreviations: 4DCT = four-dimensional computed tomography; GTVmip = union of the 10 gross tumor volumes; ITVsim = internal target volume on each phase of the first four-dimensional computed tomography data set; PTV = planning target volume.
Volume 76, Number 5, 2010
Table 2. %FV 1s for all computed tomography alignments in which the clinical target volume was not completely contained within PTV3+7mm Duty cycle Tumor # (data set alignment) 2 (skin) 3 (skin) 3 (Vert) 5 (skin) 6 (skin) 7 (skin) 7 (Vert)
100%
50%
30%
10%
0.3 0.5 0.8 1.5 0.4 0.8 1.9 3.4 0.1 0.3 0.1 0.3 0.2 0.4
0.2 0.5 0.3 0.9 0.1 0.4 0.6 2.0 0.1 0.1 0.0 0.1 0.1 0.2
0.1 0.4 0.2 0.7 0.1 0.3 0.2 0.9 0.0 0.1 0.0 0.1 0.0 0.2
0.1 0.2 0.1 0.4 0.0 0.2 0.0 0.2 0.0 0.0 0.0 0.1 0.0 0.1
Abbreviations: %FV = fractional volumes; PTV = planning target volume. and 60% phases. Finally, to simulate a 10% duty cycle, only the 50% phase tumor contours were used. Four different simulations of patient setup were examined for each of four different magnitudes of respiratory gating to generate a total of 16 treatment scenario simulations for both GTVs and CTVs.
Metrics of evaluating different patient alignments/gating duty cycles The fractional volume of tumor, either GTV or CTV, outside of the PTV was the metric used to indicate how well current PTV margins encompass the tumor during treatment (when the term tumor is used, we refer to both the GTV and the CTV in separate analysis). Two PTVs were generated using the tumor contours from the first 4DCT data set (simulation data set). The first PTV that was generated, denoted by PTV7mm, was generated by uniformly expanding the union of the CTVs (or the internal target volume [ITV]) on each phase of the first 4DCT data set (ITVsim) by 7 mm to account for patient setup uncertainties. The second PTV that was generated for evaluation, denoted by PTV3+7mm was generated by first expanding ITVsim by 3 mm to account for interfractional uncertainty in tumor motion, and then expanding this by an additional 7 mm to account for patient setup uncertainties. Both methods of PTV generation are used in our clinic. Then, the fractional volume of the tumor (both GTV and CTV) outside PTV7mm and PTV3+7mm was determined for each phase of each 4DCT data set. The average fractional volume was determined by averaging the fractional volume outside the PTV on each phase over the total number of phases used (Fig. 1). All fractional volumes reported here are expressed as percentages and denoted by %FV. This calculation was done for all four alignment techniques and for all four degrees of respiratory gating. For each alignment strategy and gating combination, the unions of all the tumor contours were generated; these are denoted by GTVtotal and CTVtotal for the GTVs and CTVs, respectively. To determine if respiratory gating is more beneficial than image-guided patient alignment for tumor localization, we determined the margin needed for a volume to completely encompass the tumor (GTVtotal/ CTVtotal) for each of the patient alignment techniques and respiratory gating strategies. Reference tumor volumes were iteratively generated with margins increasing in 0.5-mm increments until less than 0.1 cm3 of the tumor volume was outside of the reference tumor volume. For GTV evaluation, the starting point from reference volumes were expanded was the union of the 10 GTVs (GTVmip)
Reduction of Setup Uncertainties in Lung Cancer d C. NELSON et al.
Table 3. %FV 1s for all computed tomography alignments in which the clinical target volume was not completely contained within PTV7mm
Table 5. Margins needed to completely encompass tumor when the patients were aligned using the vertebral bodies for each degree of respiratory gating simulated Margins (cm) needed to encompass (GTVtotal, CTVtotal)
Duty cycle Tumor # (data set alignment)
Duty cycle 100%
50%
30%
10%
0.3 0.7 2.0 3.1 0.4 1.2 0.1 0.2 1.7 2.4 0.8 1.5 0.2 0.3 0.1 0.3 0.6 1.6 0.5 0.9 0.4 0.8 0.5 0.9 0.2 0.7 0.6 1.5 5.6 8.8 0.6 0.8 0.1 0.3 0.1 0.2 0.6 1.0 0.6 1.2
0.1 0.4 1.7 3.1 0.4 1.2 0.0 0.1 0.7 1.6 0.3 0.9 0.1 0.2 0.1 0.2 0.5 1.5 0.4 0.8 0.2 0.6 0.2 0.8 0.1 0.6 0.3 1.2 2.2 5.7 0.3 0.6 0.1 0.2 0.0 0.2 0.2 0.7 0.3 0.7
0.0 0.2 1.1 2.7 0.3 1.0 0.0 0.0 0.4 1.3 0.2 0.7 0.1 0.2 0.0 0.1 0.3 1.4 0.2 0.7 0.1 0.4 0.1 0.7 0.1 0.5 0.2 0.9 1.1 3.6 0.2 0.5 0.0 0.1 0.0 0.1 0.2 0.6 0.2 0.6
0.0 0.1 0.4 1.7 0.1 0.6 0.0 0.0 0.1 0.7 0.1 0.4 0.0 0.1 0.0 0.1 0.1 0.8 0.0 0.2 0.0 0.0 0.0 0.1 0.0 0.0 0.0 0.2 0.3 1.5 0.1 0.3 0.0 0.0 0.0 0.0 0.1 0.3 0.1 0.3
Tumor # 1 (skin) 2 (skin) 2 (vert) 2 (fiducial) 3 (skin) 3 (vert) 3 (fiducial) 3 (tumor) 4a (skin) 4a (vert) 4a (fiducial 2) 4a [fiducial 3) 4a (fiducial 4) 4b (skin) 5 (skin) 6 (skin) 6 (vert) 6 (tumor) 7 (skin) 7 (vert)
Abbreviations: %FV = fractional volumes; PTV = planning target volume.
on each phase of the simulation 4DCT data set, and for CTV evaluation, the starting point for expansion was ITVsim. In the treatment planning system, the three-dimensional region of interest is generated from a series of two-dimensional curves. After each reference volume was generated, all curves in the volume with an area less than 0.1 cm2 were removed using tools in the treatment planning system. The margin resulting from this iterative sequence of expansion and removal was used for the expansion we used to evaluate tumor localization. A t-test was used to determine statistical significance between the different degrees of respiratory gating and nogating and between the various alignment techniques.
1 2 3 4a 4b 5 6 7
Table 4. Margins needed to completely encompass tumor when the patients were aligned using skin marks for each degree of respiratory gating simulated Margins (cm) needed to encompass (GTVtotal, CTVtotal) Duty cycle
1 2 3 4a 4b 5 6 7
100%
50%
30%
10%
(1.15, 1.15) (1.30, 1.45) (3.10, 3.10) (0.95, 1.10) (0.85, 0.90) (1.65, 1.70) (1.50, 1.65) (1.70, 1.75)
(1.05, 1.10) (1.30, 1.45) (2.05, 2.30) (0.95, 1.10) (0.85, 0.90) (1.55, 1.65) (1.30, 1.45) (1.60, 1.60)
(0.95, 1.05) (1.25, 1.40) (1.90, 2.20) (0.95, 1.10) (0.80, 0.90) (1.35, 1.45) (1.25, 1.35) (1.60, 1.60)
(0.95, 0.95) (1.20, 1.40) (1.70, 2.00) (0.85, 1.00) (0.60, 0.75) (1.15, 1.20) (1.15, 1.30) (1.50, 1.50)
Abbreviations: GTV = gross tumor volume; CTV = clinical target volume.
100%
50%
30%
10%
(0.70, 0.75) (1.00, 1.00) (2.85, 2.75) (0.85, 0.90) (0.40, 0.45) (0.80, 0.85) (1.05, 1.15) (1.70, 1.75)
(0.60, 0.75) (1.00, 1.00) (1.75, 2.00) (0.90, 0.90) (0.35, 0.45) (0.80, 0.85) (0.95, 1.10) (1.55, 1.60)
(0.55, 0.60) (0.95, 0.95) (1.65, 1.95) (0.90, 0.90) (0.40, 0.45) (0.70, 0.75) (0.90, 1.05) (1.55, 1.60)
(0.65, 0.70) (0.95, 0.95) (1.40, 1.70) (0.70, 0.70) (0.15, 0.15) (0.60, 0.65) (0.85, 0.85) (1.45, 1.50)
Abbreviations: GTV = gross tumor volume; CTV = clinical target volume.
RESULTS Table 1 shows the tumor volumes for GTVmip, ITVsim, PTV7mm, and PTV3+7mm; the number of fiducials implanted near the tumor; and the total number of 4DCT data sets used in this study for each patient. Patient 1 had two fiducials, one of which (fiducial 1) dislodged during the third week of treatment; therefore, only five 4DCTs with this fiducial alignment were used. Patient 4 had two tumors, denoted by 4a (upper lobe) and 4b (lower lobe). This patient also had four fiducials that remained in place during the course of treatment; fiducials 1 and 2 were implanted near tumor 4a and Table 6. Margins needed to completely encompass tumor when the patients were aligned using the fiducials implanted in the periphery of the tumor for each degree of respiratory gating simulated Margins (cm) needed to encompass (GTVtotal, CTVtotal) Duty cycle Tumor # fiducial #
Tumor #
1581
1-1 1-2 2 3 4a-1 4a-2 4a-3 4a-4 4b-1 4b-2 4b-3 4b-4 5-1 5-2 6 7
100%
50%
30%
10%
(0.50, 0.55) (0.45, 0.75) (0.75, 1.05) (1.60, 1.60) (0.65, 0.65) (0.90, 0.90) (0.80, 0.95) (0.70, 0.90) (0.40, 0.45) (0.45, 0.45) (0.20, 0.40) (0.40, 0.45) (0.35, 0.45) (0.65, 0.75) NA* NA*
(0.45, 0.45) (0.45, 0.55) (0.65, 0.75) (1.35, 1.35) (0.65, 0.65) (0.90, 0.90) (0.70, 0.90) (0.70, 0.90) (0.30, 0.35) (0.35, 0.45) (0.20, 0.25) (0.40, 0.45) (0.25, 0.35) (0.65, 0.75)
(0.45, 0.45) (0.45, 0.55) (0.65, 0.70) (1.30, 1.35) (0.65, 0.65) (0.90, 0.90) (0.60, 0.90) (0.70, 0.90) (0.20, 0.35) (0.20, 0.40) (0.15, 0.20) (0.40, 0.40) (0.15, 0.25) (0.60, 0.70)
(0.30, 0.40) (0.45, 0.50) (0.65, 0.70) (1.00, 1.05) (0.15, 0.30) (0.30, 0.40) (0.50, 0.75) (0.65, 0.65) (0.20, 0.35) (0.15, 0.20) (0.15, 0.15) (0.15, 0.15) (0.15, 0.20) (0.60, 0.70)
Abbreviations: GTV = gross tumor volume; CTV = clinical target volume. * Patients did not have fiducials implanted near the tumor.
I. J. Radiation Oncology d Biology d Physics
1582
Table 7. Margins needed to completely encompass tumor when the patients were aligned using the tumor center of volume for each degree of respiratory gating simulated. Margins (cm) needed to encompass (GTVtotal, CTVtotal) Duty cycle Tumor # 1 2 3 4a 4b 5 6 7
100%
50%
30%
10%
(0.55, 0.70) (0.40, 0.50) (1.70, 1.70) (0.65, 0.65) (0.15, 0.40) (0.40, 0.50) (1.05, 1.20) (0.70, 0.70)
(0.55, 0.70) (0.30, 0.45) (1.00, 1.15) (0.65, 0.65) (0.15, 0.40) (0.35, 0.40) (0.95, 1.15) (0.65, 0.65)
(0.55, 0.60) (0.25, 0.40) (0.90, 1.00) (0.65, 0.65) (0.30, 0.40) (0.25, 0.30) (0.85, 1.10) (0.65, 0.65)
(0.55, 0.60) (0.05, 0.40) (0.70, 0.90) (0.10, 0.30) (0.05, 0.15) (0.20, 0.25) (0.70, 0.70) (0.65, 0.60)
Abbreviations: GTV = gross tumor volume; CTV = clinical target volume.
fiducials 3 and 4 near tumor 4b. Patients 6 and 7 were not enrolled on the implanted-fiducial protocol. We first evaluated how well our PTV margins that are used clinically encompass the tumor in the presence of both setup uncertainties and tumor motion. When PTV3+7mm was evaluated to see how well it encompassed the GTV when patients were aligned using skin marks, seven of eight tumors were completely encompassed within the PTV. The only non-zero %FV was small, at 0.1% (Patient 3). When a gated treatment was simulated and the patient was aligned using the vertebral bodies, implanted fiducials, or the tumor COV, the %FV was always 0%. When the %FV of the CTV was evaluated using PTV3+7mm and skin marks were used for alignment, three of eight tumors were completely encompassed within the PTV. When the CT data sets were aligned using the vertebral bodies,
Volume 76, Number 5, 2010
the CTV was completely contained within PTV3+7mm for seven of the eight tumors. When the CT data sets were aligned using implanted fiducials or the tumor COV, the CTVs of all eight tumors were completely contained within PTV3+7mm. Table 2 shows the %FV for all patients whose CTVs were not completely encompassed within PTV3+7mm. The largest %FV was 1.9%, with the remaining non-zero %FVs at less than 1%. When the clinical PTV was generated without accounting for a 3-mm uncertainty in tumor motion, GTVs for six of the eight tumors were completely contained within the PTV when the data sets were aligned using skin marks. For the two tumors (#3 and #5) that were not contained completely within the PTV, the %FV was less than 0.2%. When alignment was based on vertebral bodies, the GTV was completely contained within the PTV for all patients except Patient 3, who nevertheless has a very small %FV value at 0.1%. When the CT data sets were aligned using implanted fiducials or tumor COV, the GTV was always completely contained within PTV7mm. When PTV7mm was evaluated for coverage of the CTV, none of the tumors examined were completely contained within the PTV when the data sets were aligned by skin marks and only three of the eight tumors were completely contained within the PTV when the data sets were aligned using vertebral bodies. Because three of five patients with implanted fiducials had multiple fiducials (Patient 4 had two tumors each aligned by four fiducials), 9 of the 14 alignments based on fiducial location resulted in the CTVs being completely contained within the PTV. When the alignment was based on the tumor COV, six of eight tumors were completely contained within the PTV. Table 3 shows the %FV for all patients/alignments in which PTV7mm did not completely encompass the CTV for each respiratory gating duty cycle simulated.
Fig. 2. Average margins needed to completely encompass unions of all the tumor contours for the gross tumor volume (GTVtotal) and clinical target volume (CTVtotal).
Reduction of Setup Uncertainties in Lung Cancer d C. NELSON et al.
1583
Fig. 3. Average difference (1s) in margin between nongated treatment and gated treatment, and between patient setup using skin marks and image-guided patient setup. Zero on this graph represents nongated treatment (left side) and skin mark alignment (right side).
The data were then grouped based on the alignment technique (nongated simulation) and the averages standard deviations of %FV were determined, yielding values of 1.5% 1.8, 0.3% 0.3, 0.1% 0.2, and 0.0% 0.1 for alignments based on skin marks, vertebral bodies, implanted fiducials, and the tumor COV, respectively. Thus, COV alignments were most accurate, followed closely by alignments using implanted fiducials and vertebral bodies, and finally alignments by skin marks. The second goal of this study was to determine whether image-guided patient setup or respiratory gating is more effective in reducing uncertainties in tumor position. This was evaluated by generating volumes from the simulation 4DCT data set until this completely encompassed the tumor when the CT data sets were aligned based on skin marks, vertebral bodies, implanted fiducials, and the tumor COV in the presence of various respiratory gating simulations. The margins needed to completely encompass the tumor (both the GTV and CTV) when the data sets were aligned using skin marks, vertebral bodies, implanted fiducials, and the tumor COV are shown in Table 4, Table 5, Table 6, and Table 7, respectively, for each degree of respiratory gating simulated. To better interpret these data, we determined for each alignment technique and respiratory gating simulation an average (and standard deviation) for the eight tumors analyzed in this study (Fig. 2). Qualitatively, this figure shows that both respiratory gating and image-guided patient setup enabled a reduction in margins. The largest margins were observed when the patient was aligned using skin marks, with no respiratory gating (average margin 1.5 0.7 and 1.6 0.7 cm, respectively, for GTV and CTV). The margins were reduced with respiratory gating (1.1 0.4 and 1.3 0.4 cm, respectively, for GTV and CTV) and further reduced when the patient was
aligned using image guidance (1.2 0.8 and 1.2 0.7 cm for the GTV and CTV when aligned using vertebral bodies, 0.6 0.3 and 0.7 0.3 cm for the GTV and CTV when aligned using implanted fiducials, and 0.7 0.5 and 0.8 0.4 cm for the GTV and CTV when aligned using the tumor COV). The margins for each of the image-guided patient setups were further reduced with gating. To better quantify the amount of reduction possible, for each alignment technique, we subtracted the margin during a gated treatment from the margin calculated without respiratory gating and then determined an average and standard deviation. Then, for the nongated simulations (duty cycle = 100%), the margin needed to adequately encompass the tumor in simulations of image-guided
Fig. 4. Illustration showing unions of all the tumor contours for the gross tumor volume (GTVtotal) extending beyond planning target volume (PTV) PTV7mm and PTV3+7mm for Patient 3 when the computed tomography data sets were aligned using skin marks.
1584
I. J. Radiation Oncology d Biology d Physics
patient setup was subtracted from the margin needed when the CT data sets were aligned using skin marks, and again the averages and standard deviations were computed. By these means, we were able to extract the differences between a nongated treatment and a gated treatment and the differences between daily alignment using skin marks and image-guided setup based on vertebral bodies, implanted fiducials, and the tumor COV, with p values determined using a t-test (Fig. 3). DISCUSSION From our data, we found that only a small portion of the CTV was located outside of the PTV; in most cases, the amount of tumor outside the PTV was clinically insignificant. This indicates that the PTV margins used clinically are sufficient using our current immobilization techniques. In addition, we found that tumor localization using image-guided patient setup surrogates for tumor position (vertebral bodies) allowed a larger reduction in margin than if a respiratorygated treatment was delivered. This margin could be reduced even more if the patient was aligned using either the tumor or fiducials implanted near the tumor. Based on these data, it is safe to say that if one intends to deliver a respiratory-gated treatment, it should be done in conjunction with an imageguided setup to ensure that the patient is aligned properly before treatment. This study was not without its limitations. One limitation of our analysis is that we assumed that the CTV edge was always a uniform 8 mm from the GTV. In clinical practice, the CTV cannot always accommodate a uniform 8-mm expansion. The CTV is usually edited so that it does not meet the chest wall or pass into the vertebral bodies. However, our blind expansion of GTVs into CTVs was conservative in that the CTV is usually no more than an 8-mm uniform expansion. Also, we iteratively expanded our PTV margins so that the GTVtotal and CTVtotal were completely encompassed within the PTV. Routinely, margins of approximately 1 cm were needed to completely encompass the tumors when alignment was done using skin marks. However, Patient 3 needed a margin of nearly 3.5 cm to completely encompass the tumor. Figure 4 shows an axial slice from the maximum intensity projection data set generated from the 4DCT set acquired at simulation for Patient 3. The large uniform expansion was needed to cover a small region of GTVtotal protruding from the PTV on the posterior right side. Opposite this small region was a large region of PTV space unoccupied by tumor. Well-defined tumors that
Volume 76, Number 5, 2010
were spherical and isolated in the lung were more easily delineated, and the smaller uncertainties in tumor contour translated into smaller PTV margins. In clinical practice, the goal is to deliver at least 100% of the prescribed dose to 95% of the PTV volume. In this study, we determined PTV margins that were needed to encompass 100% of the tumor volume. The more ideal study would be to determine the margin need to encompass the volume that is occupied by the tumor 95% of the time, but requires constructing regions of interest based on tumor location probability and thus adding complexity to the problem. We suspect that this type of analysis would lead to smaller margins because it removes any outliers in tumor position. Another complexity of this study is that end expiration (50% phase) does not always correspond to the end point of tumor motion (i.e., a phase error). Although in most cases the 50% phase did correspond to the end exhalation point of the tumor, there were several instances in which this did not occur. The ultimate goal is to deliver a lethal tumor dose while minimizing dose to healthy tissues. The use of smaller margins delivers less dose to the healthy tissue; however, such a reduction in margin should not be implemented without fully understanding what is being done to warrant the reduction in margin. These smaller margins may look good in the planning stages; however, it is the execution of the treatment that matters. If the PTV margins are not sufficient to contain the tumor, the prescribed dose will not be delivered to the target. The design of the PTV starts with imaging at simulation. Although 4DCT acquisition captures the tumor motion during simulation (13, 16), the use of mid-ventilation CT has been shown as a valuable imaging method for reducing the size of treatment portals for delivery, thus reducing the amount of healthy tissue irradiated (17). Before treatment delivery, IGRT can be employed to reduce uncertainties in tumor position. IGRT techniques used for daily alignment are imaging with a CT-on-rails system mounted inside the treatment room, cone-beam CT imaging, megavoltage portal imaging, and kilovoltage on-board imaging (18–21). Image contrast is sufficient to accurately align based on the tumor when three-dimensional CT data sets, such as those obtained by CT-on-rails or cone-beam CT systems are used. However, when image quality is not adequate or the tumor is not visible, the patient must be aligned using surrogates for the tumor. Using MV portal imaging and kV imaging, vertebral bodies and fiducials implanted near the tumor can typically be visualized however these images often lack the contrast to display the tumor.
REFERENCES 1. ICRU. ICRU Report 62, Prescribing, recording and reporting photon beam therapy (Supplement to ICRU Report 50). International Commission on Radiation Units and Measurements. Bethesda, MD: ICRU; 1999. 2. Beckmann GK, Kolbl O, Krieger T, et al. How can we further improve radiotherapy for stage-III non-small-cell lung cancer? Lung Cancer 2004;45(Suppl 2):S125–132.
3. Belderbos JS, De Jaeger K, Heemsbergen WD, et al. First results of a phase I/II dose escalation trial in non-small cell lung cancer using three-dimensional conformal radiotherapy. Radiother Oncol 2003;66:119–126. 4. Grills IS, Yan D, Martinez AA, et al. Potential for reduced toxicity and dose escalation in the treatment of inoperable non-small-cell lung cancer: A comparison of intensity-
Reduction of Setup Uncertainties in Lung Cancer d C. NELSON et al.
5.
6.
7. 8. 9. 10. 11. 12. 13.
modulated radiation therapy (IMRT), 3D conformal radiation, and elective nodal irradiation. Int J Radiat Oncol Biol Phys 2003;57:875–890. Rosenzweig KE, Mychalczak B, Fuks Z, et al. Final report of the 70.2-Gy and 75.6-Gy dose levels of a phase I dose escalation study using three-dimensional conformal radiotherapy in the treatment of inoperable non-small cell lung cancer. Cancer J 2000;6:82–87. Stevens CW, Munden RF, Forster KM, et al. Respiratory-driven lung tumor motion is independent of tumor size, tumor location, and pulmonary function. Int J Radiat Oncol Biol Phys 2001;51: 62–68. Shirato H, Seppenwoolde Y, Kitamura K, et al. Intrafractional tumor motion: Lung and liver. Semin Radiat Oncol 2004;14:10–18. Ross CS, Hussey DH, Pennington EC, et al. Analysis of movement of intrathoracic neoplasms using ultrafast computerized tomography. Int J Radiat Oncol Biol Phys 1990;18:671–677. Xing L, Thorndyke B, Schreibmann E, et al. Overview of image-guided radiation therapy. Med Dosim 2006;31:91–112. Bissonnette JP, Purdie TG, Higgins JA, et al. Cone-beam computed tomographic image guidance for lung cancer radiation therapy. Int J Radiat Oncol Biol Phys 2009;73:927–934. Chang JY, Dong L, Liu H, et al. Image-guided radiation therapy for non-small cell lung cancer. J Thorac Oncol 2008;3:177–186. Morice R, Jimenez CA, Eapen GA, et al. Bronchoscopic implantation of gold fiducials for estimating lung tumor motion during gated radiation therapy. Chest 2005;128:163S. Pan T, Lee TY, Rietzel E, et al. 4D-CT imaging of a volume influenced by respiratory motion on multi-slice CT. Med Phys 2004;31:333–340.
1585
14. Zhang L, Dong L, Court L, et al. Validation of CT-assisted targeting (cat) software for soft tissue and bony target localization. Med Phys 2005;32:2106. 15. Giraud P, Antoine M, Larrouy A, et al. Evaluation of microscopic tumor extension in non-small-cell lung cancer for three-dimensional conformal radiotherapy planning. Int J Radiat Oncol Biol Phys 2000;48:1015–1024. 16. Keall PJ, Starkschall G, Shukla H, et al. Acquiring 4D thoracic CT scans using a multislice helical method. Phys Med Biol 2004;49:2053–2067. 17. Wolthaus JW, Schneider C, Sonke JJ, et al. Mid-ventilation CT scan construction from four-dimensional respiration-correlated CT scans for radiotherapy planning of lung cancer patients. Int J Radiat Oncol Biol Phys 2006;65:1560–1571. 18. Zhang L, Garden AS, Lo J, et al. Multiple regions-of-interest analysis of setup uncertainties for head-and-neck cancer radiotherapy. Int J Radiat Oncol Biol Phys 2006;64: 1559–1569. 19. Wong JR, Grimm L, Uematsu M, et al. Image-guided radiotherapy for prostate cancer by CT-linear accelerator combination: prostate movements and dosimetric considerations. Int J Radiat Oncol Biol Phys 2005;61:561–569. 20. Thilmann C, Nill S, Tucking T, et al. Correction of patient positioning errors based on in-line cone beam CTs: Clinical implementation and first experiences. Radiat Oncol 2006; 1:16. 21. Pisani L, Lockman D, Jaffray D, et al. Setup error in radiotherapy: On-line correction using electronic kilovoltage and megavoltage radiographs. Int J Radiat Oncol Biol Phys 2000; 47:825–839.
doi:10.1016/j.ijrobp.2009.08.002
PHYSICS CONTRIBUTION
EVALUATION OF TUMOR POSITION AND PTV MARGINS USING IMAGE GUIDANCE AND RESPIRATORY GATING CHRISTOPHER NELSON, PH.D.,* PETER BALTER, PH.D.,* RODOLFO C. MORICE, M.D.,y KARA BUCCI, M.D.,z LEI DONG, PH.D.,* SUSAN TUCKER, PH.D.,{ SASTRY VEDAM, PH.D.,* JOE Y. CHANG, M.D., PH.D.,z AND GEORGE STARKSCHALL, PH.D.* From the Departments of *Radiation Physics, yPulmonary Medicine, zRadiation Oncology, and {Bioinformatics & Computational Biology, The University of Texas M. D. Anderson Cancer Center, Houston, TX Purpose: To evaluate the margins currently used to generate the planning target volume for lung tumors and to determine whether image-guided patient setup or respiratory gating is more effective in reducing uncertainties in tumor position. Methods and Materials: Lung tumors in 7 patients were contoured on serial four-dimensional computed tomography (4DCT) data sets (4–8 4DCTs/patient; 50 total) obtained throughout the course of treatment. Simulations were performed to determine the tumor position when the patient was aligned using skin marks, image-guided setup based on vertebral bodies, fiducials implanted near the tumor, and the actual tumor volume under various scenarios of respiratory gating. Results: Because of the presence of setup uncertainties, the reduction in overall margin needed to completely encompass the tumor was observed to be larger for imaged-guided patient setup than for a simple respiratorygated treatment. Without respiratory gating and image-guided patient setup, margins ranged from 0.9 cm to 3.1 cm to completely encompass the tumor. These were reduced to 0.7–1.7 cm when image-guided patient setup was simulated and further reduced with respiratory gating. Conclusions: Our results indicate that if respiratory motion management is used, it should be used in conjunction with image-guided patient setup in order to reduce the overall treatment margin effectively. Ó 2010 Elsevier Inc. Image-guided patient setup, Respiratory gating, 4DCT, Lung cancer, Target volume.
The general concept of the planning target volume (PTV) is to create a structure that accounts for the geometric uncertainties in aligning the clinical target volume (CTV) with the coordinates of the treatment machine. These uncertainties are due to internal motion of the target as well as uncertainties in patient setup. The PTV is designed based on the tumor location at the time of simulation (1) and the statistically determined daily setup uncertainties of the patients. Larger PTV margins ensure adequate coverage of the tumor, but also increase the amount of healthy tissue irradiated (2). PTV margins must be adequate to ensure tumor coverage and yet must be kept small enough to prevent excess radiation-induced toxicity of normal tissue, which is often the limiting factor when escalating the dose. In the radiation treatment of lung tumors, the amount of healthy tissue irradiated limits dose escalation strategies, which have been shown to increase tumor control (3–5).
Two techniques are typically used to reduce the uncertainty in tumor location during treatment delivery, particularly for lung tumors and others subject to respirationinduced motion: respiratory-correlated radiation therapy and image-guided radiation therapy (IGRT). In the case of lung tumors, respiration-induced movement (6–8) is greatest for those located near the diaphragm. Respiratory-controlled radiation therapy reduces the effects of respiratory motion (that is, intrafractional uncertainty) by linking beam delivery with respiration, either by treating during a breath hold or by gating the delivery of the treatment beam. IGRT, conversely, is not used to account for respiratory motion but to adjust for daily (interfractional) changes in tumor position (prior to treatment delivery). Daily IGRT reduces uncertainties in tumor position by allowing the radiation therapist to realign the tumor with its initial position at the time of simulation, thereby reducing the likelihood of a geometric miss (9). IGRT can be implemented using different alignment
Reprint requests to: Christopher Nelson, Ph.D., Department of Radiation Physics, Unit 525, The University of Texas M. D. Anderson Cancer Center, 1515 Holcombe Blvd., Houston, TX 77030. Tel: (713) 563-2471; Fax: (713) 563-6949; E-mail: chnelson@ mdanderson.org
Conflict of interest: none. Acknowledgment—The authors would like to acknowledge Keith Britton for his contribution to this project. Received Feb 12, 2009, and in revised form Aug 3, 2009. Accepted for publication Aug 4, 2009.
INTRODUCTION
1578
Reduction of Setup Uncertainties in Lung Cancer d C. NELSON et al.
techniques; commonly based on bony landmarks, fiducials implanted in the periphery of the tumor, or tumor itself. Different alignment techniques are sometimes limited by the imaging capabilities of the machine whether it has twodimensional or three-dimensional imaging (10, 11). Although these techniques are routinely used to reduce the two types of uncertainty in tumor position, questions remain about their optimal application. For instance, image-guided positioning can be accomplished by means of various strategies, but it is unclear which of these strategies optimizes the relationship between the CTV and PTV. In addition, the question remains whether a respiratory-gated treatment should be delivered unless the patient has first been aligned by means of IGRT. The purpose of this study is to evaluate the margins currently used to generate the PTV for lung tumors and to determine whether image-guided patient setup or respiratory gating is more effective in reducing uncertainties in tumor position. To achieve these aims, we used actual patient data in simulations of four different imageguided setup techniques and four degrees of respiratory gating to examine the location of the tumor throughout the course of treatment and evaluated the fraction of tumor outside of the standard PTV to compared image-guided setup with respiratory gating. METHODS AND MATERIALS Patient selection/image acquisition Data from 7 patients with primary lung tumors who had been treated on investigational protocols at our institution were chosen for the current analysis. Five patients had undergone bronchoscopic implantation of gold fiducials (NMPE, Inc., Lynwood, WA) in the periphery of their lung tumors (12) after giving informed consent in the institutional review board–approved protocol ID03-0208. In addition, we examined data for 2 patients who had been enrolled on institutional review board–approved protocol ID2003-0962 and subjected to imaging in accordance with the protocol. Patients eligible for both protocols were to be treated definitively for non– small-cell lung cancer with radiation for a minimum of 6 weeks. Concurrent chemotherapy was allowed, but patients could not have received prior radiation treatments. Conventional, four-dimensional computed tomography (4DCT) simulation procedures were used for all patients accrued in this study. For each 4DCT, the patient’s respiration was monitored by tracking the vertical displacement of the abdomen (RPM, Varian Medical Systems, Palo Alto, CA). The respiratory signal was used to retrospectively sort images acquired in cine mode into the 4DCT data set (13). At the time of this study, all patients with thoracic tumors who were treated at our institution were immobilized for 4DCT simulation using a vacuum bag (Vac-Loc) placed on a wing-board/T-bar (both from MedTec, Orange City, IA). Three tattoos were placed on each patient’s skin at the level of the carina for daily alignment with the lasers in the treatment room. BBs were placed on the skin marks to make the marks’ locations visible during each imaging session. The first 4DCT scan was acquired for treatment planning and to assess tumor motion. Weekly 4DCT data sets were then acquired throughout the remainder of the patient’s treatment. For each 4DCT data set acquired, BBs were placed on the patient’s skin marks and the patient was immobilized using the simulation procedures.
1579
Fig. 1. Illustration showing how fractional volumes (%FV) is computed from the tumor contours and the planning target volume (PTV) generated from internal target volume on each phase of the first four-dimensional computed tomography data set (ITVsim).
Tumor delineation For each 4D data set, the gross tumor volume (GTV) was contoured on the 50% phase (although not always the case, we chose the 50% phase as end expiration). Using a rigid contour propagation software system (14), we propagated the GTV contours to the remaining nine phases of the 4DCT data set for review by the physician. With GTV contours on each phase of each 4DCT data set, we then performed four alignments to simulate different patient setup scenarios. Before aligning the computed tomography (CT) data sets, all contours from each data set were copied to the 50% phase so that when the image alignment was performed using a single phase (50%), the contours on the remaining nine phases would translate with the 50% phase that was being aligned. After the alignment was completed, each GTV contour (total = # 4DCT data sets 10) was uniformly expanded by 8 mm to generate the CTV, in accordance with the standard of practice in our institution and based on the work of Giraud et al. (15).
Simulation of patient setup scenarios Four separate alignments were performed to simulate different daily patient setup scenarios: conventional patient setup in which the patient is aligned using external skin marks, image-guided patient setup in which the patient is aligned based on the location of vertebral bodies (bony landmarks), image-guided patient setup in which the patient is aligned based on fiducials implanted near the periphery of the tumor, and image-guided patient setup in which the patient is aligned based on the location of the tumor itself. The first alignment was performed to simulate patient setup without image guidance (i.e., aligning the skin marks of the patient with the lasers in the treatment room). To perform this alignment, an isocenter was identified on the 50% phase of the 4DCT data set from each week as determined by the BBs placed at the time of simulation. The isocenters from each week were aligned to the isocenter point on the 50% phase of the first 4DCT data set acquired at simulation, effectively simulating patient alignment by skin marks. The remaining three CT alignments were performed to simulate the use of image guidance to set the patient up before treatment delivery. To align the data sets based on vertebral bodies, the same image registration technique (14) was used to align the vertebral bodies in the 50% phase of each subsequent 4DCT data set to
I. J. Radiation Oncology d Biology d Physics
1580
the vertebral bodies in the 50% phase of the first 4DCT data set acquired at simulation. The vertebral bodies that were used for the alignment were chosen by contouring the vertebral body on the simulation CT data set that also contained the contours of the tumor; thus, the vertebral body used for the registration was in the same superoinferior plane as the tumor. The image registration method is based on the image intensities within the region of interest for the contoured vertebral body in the simulation CT data set. For simulation of alignment based on implanted fiducial location, the CT data sets were aligned by registering the location of the fiducial on the 50% phase of each subsequent CT data set to the location in the 50% phase at simulation. Finally, image alignment based on the location of the tumor was done. To minimize the effects of changing tumor size/shape/contouring uncertainties from week to week, the treatment planning system was used to identify the center of volume (COV) of the GTV contour on the 50% phase of each 4DCT data set. The CT alignments were then performed so that the GTV COV in the 50% phase from each subsequent week was aligned with the GTV COV in the 50% phase of the initial 4DCT acquired at simulation. The alignments based on tumor and fiducial location simulated one in which a gated image was acquired and the alignment was based on the gated image. After each of the alignments, we had a series of tumor contours acquired during free breathing (10 phases) simulating patient alignment using external skin marks, as well as simulating image guidance aligning to fiducials implanted near the tumor, the vertebral bodies, and the tumor.
Simulation of gated treatment The goal of respiratory gating is to allow beam delivery in a patient’s treatment only during a small portion of the respiratory cycle. Given tumor contours for each phase of multiple 4DCT data sets representing the patient during normal respiration (100% duty cycle, no gating), respiratory-gated treatment can easily be simulated by simply removing particular phases from the set of tumor contours. We generated three different gated treatment scenarios, representing treatment delivery with 50%, 30%, and 10% duty cycles at end expiration. To simulate a 50% duty cycle, the tumor contours from the five phases surrounding end inspiration (i.e., the 0% phase) were removed leaving only the contours from the 30%, 40%, 50%, 60%, and 70% phases of the 4DCT data set. To simulate a 30% duty cycle, we used only the contours from the 40%, 50%, Table 1. Baseline tumor volumes during the simulation 4DCT, the number of fiducials implanted, and the total number of 4DCTs used in this study Tumor volumes (cm3) Tumor GTVmip ITVsim PTV7mm PTV3+7mm # Fiducials # 4DCTs 1 2 3 4a 4b 5 6 7
137.3 43.9 235.6 2.7 3.0 14.6 94.4 136.1
305.4 125.4 500.4 23.4 24.0 65.3 258.4 326.5
524.1 248.1 821.5 70.2 71.0 140.8 472.7 565.8
649.0 323.0 996.1 103.2 104.3 197.7 593.1 700.9
2 1 1 4 2 0 0
8 6 8 7 4 9 8
Abbreviations: 4DCT = four-dimensional computed tomography; GTVmip = union of the 10 gross tumor volumes; ITVsim = internal target volume on each phase of the first four-dimensional computed tomography data set; PTV = planning target volume.
Volume 76, Number 5, 2010
Table 2. %FV 1s for all computed tomography alignments in which the clinical target volume was not completely contained within PTV3+7mm Duty cycle Tumor # (data set alignment) 2 (skin) 3 (skin) 3 (Vert) 5 (skin) 6 (skin) 7 (skin) 7 (Vert)
100%
50%
30%
10%
0.3 0.5 0.8 1.5 0.4 0.8 1.9 3.4 0.1 0.3 0.1 0.3 0.2 0.4
0.2 0.5 0.3 0.9 0.1 0.4 0.6 2.0 0.1 0.1 0.0 0.1 0.1 0.2
0.1 0.4 0.2 0.7 0.1 0.3 0.2 0.9 0.0 0.1 0.0 0.1 0.0 0.2
0.1 0.2 0.1 0.4 0.0 0.2 0.0 0.2 0.0 0.0 0.0 0.1 0.0 0.1
Abbreviations: %FV = fractional volumes; PTV = planning target volume. and 60% phases. Finally, to simulate a 10% duty cycle, only the 50% phase tumor contours were used. Four different simulations of patient setup were examined for each of four different magnitudes of respiratory gating to generate a total of 16 treatment scenario simulations for both GTVs and CTVs.
Metrics of evaluating different patient alignments/gating duty cycles The fractional volume of tumor, either GTV or CTV, outside of the PTV was the metric used to indicate how well current PTV margins encompass the tumor during treatment (when the term tumor is used, we refer to both the GTV and the CTV in separate analysis). Two PTVs were generated using the tumor contours from the first 4DCT data set (simulation data set). The first PTV that was generated, denoted by PTV7mm, was generated by uniformly expanding the union of the CTVs (or the internal target volume [ITV]) on each phase of the first 4DCT data set (ITVsim) by 7 mm to account for patient setup uncertainties. The second PTV that was generated for evaluation, denoted by PTV3+7mm was generated by first expanding ITVsim by 3 mm to account for interfractional uncertainty in tumor motion, and then expanding this by an additional 7 mm to account for patient setup uncertainties. Both methods of PTV generation are used in our clinic. Then, the fractional volume of the tumor (both GTV and CTV) outside PTV7mm and PTV3+7mm was determined for each phase of each 4DCT data set. The average fractional volume was determined by averaging the fractional volume outside the PTV on each phase over the total number of phases used (Fig. 1). All fractional volumes reported here are expressed as percentages and denoted by %FV. This calculation was done for all four alignment techniques and for all four degrees of respiratory gating. For each alignment strategy and gating combination, the unions of all the tumor contours were generated; these are denoted by GTVtotal and CTVtotal for the GTVs and CTVs, respectively. To determine if respiratory gating is more beneficial than image-guided patient alignment for tumor localization, we determined the margin needed for a volume to completely encompass the tumor (GTVtotal/ CTVtotal) for each of the patient alignment techniques and respiratory gating strategies. Reference tumor volumes were iteratively generated with margins increasing in 0.5-mm increments until less than 0.1 cm3 of the tumor volume was outside of the reference tumor volume. For GTV evaluation, the starting point from reference volumes were expanded was the union of the 10 GTVs (GTVmip)
Reduction of Setup Uncertainties in Lung Cancer d C. NELSON et al.
Table 3. %FV 1s for all computed tomography alignments in which the clinical target volume was not completely contained within PTV7mm
Table 5. Margins needed to completely encompass tumor when the patients were aligned using the vertebral bodies for each degree of respiratory gating simulated Margins (cm) needed to encompass (GTVtotal, CTVtotal)
Duty cycle Tumor # (data set alignment)
Duty cycle 100%
50%
30%
10%
0.3 0.7 2.0 3.1 0.4 1.2 0.1 0.2 1.7 2.4 0.8 1.5 0.2 0.3 0.1 0.3 0.6 1.6 0.5 0.9 0.4 0.8 0.5 0.9 0.2 0.7 0.6 1.5 5.6 8.8 0.6 0.8 0.1 0.3 0.1 0.2 0.6 1.0 0.6 1.2
0.1 0.4 1.7 3.1 0.4 1.2 0.0 0.1 0.7 1.6 0.3 0.9 0.1 0.2 0.1 0.2 0.5 1.5 0.4 0.8 0.2 0.6 0.2 0.8 0.1 0.6 0.3 1.2 2.2 5.7 0.3 0.6 0.1 0.2 0.0 0.2 0.2 0.7 0.3 0.7
0.0 0.2 1.1 2.7 0.3 1.0 0.0 0.0 0.4 1.3 0.2 0.7 0.1 0.2 0.0 0.1 0.3 1.4 0.2 0.7 0.1 0.4 0.1 0.7 0.1 0.5 0.2 0.9 1.1 3.6 0.2 0.5 0.0 0.1 0.0 0.1 0.2 0.6 0.2 0.6
0.0 0.1 0.4 1.7 0.1 0.6 0.0 0.0 0.1 0.7 0.1 0.4 0.0 0.1 0.0 0.1 0.1 0.8 0.0 0.2 0.0 0.0 0.0 0.1 0.0 0.0 0.0 0.2 0.3 1.5 0.1 0.3 0.0 0.0 0.0 0.0 0.1 0.3 0.1 0.3
Tumor # 1 (skin) 2 (skin) 2 (vert) 2 (fiducial) 3 (skin) 3 (vert) 3 (fiducial) 3 (tumor) 4a (skin) 4a (vert) 4a (fiducial 2) 4a [fiducial 3) 4a (fiducial 4) 4b (skin) 5 (skin) 6 (skin) 6 (vert) 6 (tumor) 7 (skin) 7 (vert)
Abbreviations: %FV = fractional volumes; PTV = planning target volume.
on each phase of the simulation 4DCT data set, and for CTV evaluation, the starting point for expansion was ITVsim. In the treatment planning system, the three-dimensional region of interest is generated from a series of two-dimensional curves. After each reference volume was generated, all curves in the volume with an area less than 0.1 cm2 were removed using tools in the treatment planning system. The margin resulting from this iterative sequence of expansion and removal was used for the expansion we used to evaluate tumor localization. A t-test was used to determine statistical significance between the different degrees of respiratory gating and nogating and between the various alignment techniques.
1 2 3 4a 4b 5 6 7
Table 4. Margins needed to completely encompass tumor when the patients were aligned using skin marks for each degree of respiratory gating simulated Margins (cm) needed to encompass (GTVtotal, CTVtotal) Duty cycle
1 2 3 4a 4b 5 6 7
100%
50%
30%
10%
(1.15, 1.15) (1.30, 1.45) (3.10, 3.10) (0.95, 1.10) (0.85, 0.90) (1.65, 1.70) (1.50, 1.65) (1.70, 1.75)
(1.05, 1.10) (1.30, 1.45) (2.05, 2.30) (0.95, 1.10) (0.85, 0.90) (1.55, 1.65) (1.30, 1.45) (1.60, 1.60)
(0.95, 1.05) (1.25, 1.40) (1.90, 2.20) (0.95, 1.10) (0.80, 0.90) (1.35, 1.45) (1.25, 1.35) (1.60, 1.60)
(0.95, 0.95) (1.20, 1.40) (1.70, 2.00) (0.85, 1.00) (0.60, 0.75) (1.15, 1.20) (1.15, 1.30) (1.50, 1.50)
Abbreviations: GTV = gross tumor volume; CTV = clinical target volume.
100%
50%
30%
10%
(0.70, 0.75) (1.00, 1.00) (2.85, 2.75) (0.85, 0.90) (0.40, 0.45) (0.80, 0.85) (1.05, 1.15) (1.70, 1.75)
(0.60, 0.75) (1.00, 1.00) (1.75, 2.00) (0.90, 0.90) (0.35, 0.45) (0.80, 0.85) (0.95, 1.10) (1.55, 1.60)
(0.55, 0.60) (0.95, 0.95) (1.65, 1.95) (0.90, 0.90) (0.40, 0.45) (0.70, 0.75) (0.90, 1.05) (1.55, 1.60)
(0.65, 0.70) (0.95, 0.95) (1.40, 1.70) (0.70, 0.70) (0.15, 0.15) (0.60, 0.65) (0.85, 0.85) (1.45, 1.50)
Abbreviations: GTV = gross tumor volume; CTV = clinical target volume.
RESULTS Table 1 shows the tumor volumes for GTVmip, ITVsim, PTV7mm, and PTV3+7mm; the number of fiducials implanted near the tumor; and the total number of 4DCT data sets used in this study for each patient. Patient 1 had two fiducials, one of which (fiducial 1) dislodged during the third week of treatment; therefore, only five 4DCTs with this fiducial alignment were used. Patient 4 had two tumors, denoted by 4a (upper lobe) and 4b (lower lobe). This patient also had four fiducials that remained in place during the course of treatment; fiducials 1 and 2 were implanted near tumor 4a and Table 6. Margins needed to completely encompass tumor when the patients were aligned using the fiducials implanted in the periphery of the tumor for each degree of respiratory gating simulated Margins (cm) needed to encompass (GTVtotal, CTVtotal) Duty cycle Tumor # fiducial #
Tumor #
1581
1-1 1-2 2 3 4a-1 4a-2 4a-3 4a-4 4b-1 4b-2 4b-3 4b-4 5-1 5-2 6 7
100%
50%
30%
10%
(0.50, 0.55) (0.45, 0.75) (0.75, 1.05) (1.60, 1.60) (0.65, 0.65) (0.90, 0.90) (0.80, 0.95) (0.70, 0.90) (0.40, 0.45) (0.45, 0.45) (0.20, 0.40) (0.40, 0.45) (0.35, 0.45) (0.65, 0.75) NA* NA*
(0.45, 0.45) (0.45, 0.55) (0.65, 0.75) (1.35, 1.35) (0.65, 0.65) (0.90, 0.90) (0.70, 0.90) (0.70, 0.90) (0.30, 0.35) (0.35, 0.45) (0.20, 0.25) (0.40, 0.45) (0.25, 0.35) (0.65, 0.75)
(0.45, 0.45) (0.45, 0.55) (0.65, 0.70) (1.30, 1.35) (0.65, 0.65) (0.90, 0.90) (0.60, 0.90) (0.70, 0.90) (0.20, 0.35) (0.20, 0.40) (0.15, 0.20) (0.40, 0.40) (0.15, 0.25) (0.60, 0.70)
(0.30, 0.40) (0.45, 0.50) (0.65, 0.70) (1.00, 1.05) (0.15, 0.30) (0.30, 0.40) (0.50, 0.75) (0.65, 0.65) (0.20, 0.35) (0.15, 0.20) (0.15, 0.15) (0.15, 0.15) (0.15, 0.20) (0.60, 0.70)
Abbreviations: GTV = gross tumor volume; CTV = clinical target volume. * Patients did not have fiducials implanted near the tumor.
I. J. Radiation Oncology d Biology d Physics
1582
Table 7. Margins needed to completely encompass tumor when the patients were aligned using the tumor center of volume for each degree of respiratory gating simulated. Margins (cm) needed to encompass (GTVtotal, CTVtotal) Duty cycle Tumor # 1 2 3 4a 4b 5 6 7
100%
50%
30%
10%
(0.55, 0.70) (0.40, 0.50) (1.70, 1.70) (0.65, 0.65) (0.15, 0.40) (0.40, 0.50) (1.05, 1.20) (0.70, 0.70)
(0.55, 0.70) (0.30, 0.45) (1.00, 1.15) (0.65, 0.65) (0.15, 0.40) (0.35, 0.40) (0.95, 1.15) (0.65, 0.65)
(0.55, 0.60) (0.25, 0.40) (0.90, 1.00) (0.65, 0.65) (0.30, 0.40) (0.25, 0.30) (0.85, 1.10) (0.65, 0.65)
(0.55, 0.60) (0.05, 0.40) (0.70, 0.90) (0.10, 0.30) (0.05, 0.15) (0.20, 0.25) (0.70, 0.70) (0.65, 0.60)
Abbreviations: GTV = gross tumor volume; CTV = clinical target volume.
fiducials 3 and 4 near tumor 4b. Patients 6 and 7 were not enrolled on the implanted-fiducial protocol. We first evaluated how well our PTV margins that are used clinically encompass the tumor in the presence of both setup uncertainties and tumor motion. When PTV3+7mm was evaluated to see how well it encompassed the GTV when patients were aligned using skin marks, seven of eight tumors were completely encompassed within the PTV. The only non-zero %FV was small, at 0.1% (Patient 3). When a gated treatment was simulated and the patient was aligned using the vertebral bodies, implanted fiducials, or the tumor COV, the %FV was always 0%. When the %FV of the CTV was evaluated using PTV3+7mm and skin marks were used for alignment, three of eight tumors were completely encompassed within the PTV. When the CT data sets were aligned using the vertebral bodies,
Volume 76, Number 5, 2010
the CTV was completely contained within PTV3+7mm for seven of the eight tumors. When the CT data sets were aligned using implanted fiducials or the tumor COV, the CTVs of all eight tumors were completely contained within PTV3+7mm. Table 2 shows the %FV for all patients whose CTVs were not completely encompassed within PTV3+7mm. The largest %FV was 1.9%, with the remaining non-zero %FVs at less than 1%. When the clinical PTV was generated without accounting for a 3-mm uncertainty in tumor motion, GTVs for six of the eight tumors were completely contained within the PTV when the data sets were aligned using skin marks. For the two tumors (#3 and #5) that were not contained completely within the PTV, the %FV was less than 0.2%. When alignment was based on vertebral bodies, the GTV was completely contained within the PTV for all patients except Patient 3, who nevertheless has a very small %FV value at 0.1%. When the CT data sets were aligned using implanted fiducials or tumor COV, the GTV was always completely contained within PTV7mm. When PTV7mm was evaluated for coverage of the CTV, none of the tumors examined were completely contained within the PTV when the data sets were aligned by skin marks and only three of the eight tumors were completely contained within the PTV when the data sets were aligned using vertebral bodies. Because three of five patients with implanted fiducials had multiple fiducials (Patient 4 had two tumors each aligned by four fiducials), 9 of the 14 alignments based on fiducial location resulted in the CTVs being completely contained within the PTV. When the alignment was based on the tumor COV, six of eight tumors were completely contained within the PTV. Table 3 shows the %FV for all patients/alignments in which PTV7mm did not completely encompass the CTV for each respiratory gating duty cycle simulated.
Fig. 2. Average margins needed to completely encompass unions of all the tumor contours for the gross tumor volume (GTVtotal) and clinical target volume (CTVtotal).
Reduction of Setup Uncertainties in Lung Cancer d C. NELSON et al.
1583
Fig. 3. Average difference (1s) in margin between nongated treatment and gated treatment, and between patient setup using skin marks and image-guided patient setup. Zero on this graph represents nongated treatment (left side) and skin mark alignment (right side).
The data were then grouped based on the alignment technique (nongated simulation) and the averages standard deviations of %FV were determined, yielding values of 1.5% 1.8, 0.3% 0.3, 0.1% 0.2, and 0.0% 0.1 for alignments based on skin marks, vertebral bodies, implanted fiducials, and the tumor COV, respectively. Thus, COV alignments were most accurate, followed closely by alignments using implanted fiducials and vertebral bodies, and finally alignments by skin marks. The second goal of this study was to determine whether image-guided patient setup or respiratory gating is more effective in reducing uncertainties in tumor position. This was evaluated by generating volumes from the simulation 4DCT data set until this completely encompassed the tumor when the CT data sets were aligned based on skin marks, vertebral bodies, implanted fiducials, and the tumor COV in the presence of various respiratory gating simulations. The margins needed to completely encompass the tumor (both the GTV and CTV) when the data sets were aligned using skin marks, vertebral bodies, implanted fiducials, and the tumor COV are shown in Table 4, Table 5, Table 6, and Table 7, respectively, for each degree of respiratory gating simulated. To better interpret these data, we determined for each alignment technique and respiratory gating simulation an average (and standard deviation) for the eight tumors analyzed in this study (Fig. 2). Qualitatively, this figure shows that both respiratory gating and image-guided patient setup enabled a reduction in margins. The largest margins were observed when the patient was aligned using skin marks, with no respiratory gating (average margin 1.5 0.7 and 1.6 0.7 cm, respectively, for GTV and CTV). The margins were reduced with respiratory gating (1.1 0.4 and 1.3 0.4 cm, respectively, for GTV and CTV) and further reduced when the patient was
aligned using image guidance (1.2 0.8 and 1.2 0.7 cm for the GTV and CTV when aligned using vertebral bodies, 0.6 0.3 and 0.7 0.3 cm for the GTV and CTV when aligned using implanted fiducials, and 0.7 0.5 and 0.8 0.4 cm for the GTV and CTV when aligned using the tumor COV). The margins for each of the image-guided patient setups were further reduced with gating. To better quantify the amount of reduction possible, for each alignment technique, we subtracted the margin during a gated treatment from the margin calculated without respiratory gating and then determined an average and standard deviation. Then, for the nongated simulations (duty cycle = 100%), the margin needed to adequately encompass the tumor in simulations of image-guided
Fig. 4. Illustration showing unions of all the tumor contours for the gross tumor volume (GTVtotal) extending beyond planning target volume (PTV) PTV7mm and PTV3+7mm for Patient 3 when the computed tomography data sets were aligned using skin marks.
1584
I. J. Radiation Oncology d Biology d Physics
patient setup was subtracted from the margin needed when the CT data sets were aligned using skin marks, and again the averages and standard deviations were computed. By these means, we were able to extract the differences between a nongated treatment and a gated treatment and the differences between daily alignment using skin marks and image-guided setup based on vertebral bodies, implanted fiducials, and the tumor COV, with p values determined using a t-test (Fig. 3). DISCUSSION From our data, we found that only a small portion of the CTV was located outside of the PTV; in most cases, the amount of tumor outside the PTV was clinically insignificant. This indicates that the PTV margins used clinically are sufficient using our current immobilization techniques. In addition, we found that tumor localization using image-guided patient setup surrogates for tumor position (vertebral bodies) allowed a larger reduction in margin than if a respiratorygated treatment was delivered. This margin could be reduced even more if the patient was aligned using either the tumor or fiducials implanted near the tumor. Based on these data, it is safe to say that if one intends to deliver a respiratory-gated treatment, it should be done in conjunction with an imageguided setup to ensure that the patient is aligned properly before treatment. This study was not without its limitations. One limitation of our analysis is that we assumed that the CTV edge was always a uniform 8 mm from the GTV. In clinical practice, the CTV cannot always accommodate a uniform 8-mm expansion. The CTV is usually edited so that it does not meet the chest wall or pass into the vertebral bodies. However, our blind expansion of GTVs into CTVs was conservative in that the CTV is usually no more than an 8-mm uniform expansion. Also, we iteratively expanded our PTV margins so that the GTVtotal and CTVtotal were completely encompassed within the PTV. Routinely, margins of approximately 1 cm were needed to completely encompass the tumors when alignment was done using skin marks. However, Patient 3 needed a margin of nearly 3.5 cm to completely encompass the tumor. Figure 4 shows an axial slice from the maximum intensity projection data set generated from the 4DCT set acquired at simulation for Patient 3. The large uniform expansion was needed to cover a small region of GTVtotal protruding from the PTV on the posterior right side. Opposite this small region was a large region of PTV space unoccupied by tumor. Well-defined tumors that
Volume 76, Number 5, 2010
were spherical and isolated in the lung were more easily delineated, and the smaller uncertainties in tumor contour translated into smaller PTV margins. In clinical practice, the goal is to deliver at least 100% of the prescribed dose to 95% of the PTV volume. In this study, we determined PTV margins that were needed to encompass 100% of the tumor volume. The more ideal study would be to determine the margin need to encompass the volume that is occupied by the tumor 95% of the time, but requires constructing regions of interest based on tumor location probability and thus adding complexity to the problem. We suspect that this type of analysis would lead to smaller margins because it removes any outliers in tumor position. Another complexity of this study is that end expiration (50% phase) does not always correspond to the end point of tumor motion (i.e., a phase error). Although in most cases the 50% phase did correspond to the end exhalation point of the tumor, there were several instances in which this did not occur. The ultimate goal is to deliver a lethal tumor dose while minimizing dose to healthy tissues. The use of smaller margins delivers less dose to the healthy tissue; however, such a reduction in margin should not be implemented without fully understanding what is being done to warrant the reduction in margin. These smaller margins may look good in the planning stages; however, it is the execution of the treatment that matters. If the PTV margins are not sufficient to contain the tumor, the prescribed dose will not be delivered to the target. The design of the PTV starts with imaging at simulation. Although 4DCT acquisition captures the tumor motion during simulation (13, 16), the use of mid-ventilation CT has been shown as a valuable imaging method for reducing the size of treatment portals for delivery, thus reducing the amount of healthy tissue irradiated (17). Before treatment delivery, IGRT can be employed to reduce uncertainties in tumor position. IGRT techniques used for daily alignment are imaging with a CT-on-rails system mounted inside the treatment room, cone-beam CT imaging, megavoltage portal imaging, and kilovoltage on-board imaging (18–21). Image contrast is sufficient to accurately align based on the tumor when three-dimensional CT data sets, such as those obtained by CT-on-rails or cone-beam CT systems are used. However, when image quality is not adequate or the tumor is not visible, the patient must be aligned using surrogates for the tumor. Using MV portal imaging and kV imaging, vertebral bodies and fiducials implanted near the tumor can typically be visualized however these images often lack the contrast to display the tumor.
REFERENCES 1. ICRU. ICRU Report 62, Prescribing, recording and reporting photon beam therapy (Supplement to ICRU Report 50). International Commission on Radiation Units and Measurements. Bethesda, MD: ICRU; 1999. 2. Beckmann GK, Kolbl O, Krieger T, et al. How can we further improve radiotherapy for stage-III non-small-cell lung cancer? Lung Cancer 2004;45(Suppl 2):S125–132.
3. Belderbos JS, De Jaeger K, Heemsbergen WD, et al. First results of a phase I/II dose escalation trial in non-small cell lung cancer using three-dimensional conformal radiotherapy. Radiother Oncol 2003;66:119–126. 4. Grills IS, Yan D, Martinez AA, et al. Potential for reduced toxicity and dose escalation in the treatment of inoperable non-small-cell lung cancer: A comparison of intensity-
Reduction of Setup Uncertainties in Lung Cancer d C. NELSON et al.
5.
6.
7. 8. 9. 10. 11. 12. 13.
modulated radiation therapy (IMRT), 3D conformal radiation, and elective nodal irradiation. Int J Radiat Oncol Biol Phys 2003;57:875–890. Rosenzweig KE, Mychalczak B, Fuks Z, et al. Final report of the 70.2-Gy and 75.6-Gy dose levels of a phase I dose escalation study using three-dimensional conformal radiotherapy in the treatment of inoperable non-small cell lung cancer. Cancer J 2000;6:82–87. Stevens CW, Munden RF, Forster KM, et al. Respiratory-driven lung tumor motion is independent of tumor size, tumor location, and pulmonary function. Int J Radiat Oncol Biol Phys 2001;51: 62–68. Shirato H, Seppenwoolde Y, Kitamura K, et al. Intrafractional tumor motion: Lung and liver. Semin Radiat Oncol 2004;14:10–18. Ross CS, Hussey DH, Pennington EC, et al. Analysis of movement of intrathoracic neoplasms using ultrafast computerized tomography. Int J Radiat Oncol Biol Phys 1990;18:671–677. Xing L, Thorndyke B, Schreibmann E, et al. Overview of image-guided radiation therapy. Med Dosim 2006;31:91–112. Bissonnette JP, Purdie TG, Higgins JA, et al. Cone-beam computed tomographic image guidance for lung cancer radiation therapy. Int J Radiat Oncol Biol Phys 2009;73:927–934. Chang JY, Dong L, Liu H, et al. Image-guided radiation therapy for non-small cell lung cancer. J Thorac Oncol 2008;3:177–186. Morice R, Jimenez CA, Eapen GA, et al. Bronchoscopic implantation of gold fiducials for estimating lung tumor motion during gated radiation therapy. Chest 2005;128:163S. Pan T, Lee TY, Rietzel E, et al. 4D-CT imaging of a volume influenced by respiratory motion on multi-slice CT. Med Phys 2004;31:333–340.
1585
14. Zhang L, Dong L, Court L, et al. Validation of CT-assisted targeting (cat) software for soft tissue and bony target localization. Med Phys 2005;32:2106. 15. Giraud P, Antoine M, Larrouy A, et al. Evaluation of microscopic tumor extension in non-small-cell lung cancer for three-dimensional conformal radiotherapy planning. Int J Radiat Oncol Biol Phys 2000;48:1015–1024. 16. Keall PJ, Starkschall G, Shukla H, et al. Acquiring 4D thoracic CT scans using a multislice helical method. Phys Med Biol 2004;49:2053–2067. 17. Wolthaus JW, Schneider C, Sonke JJ, et al. Mid-ventilation CT scan construction from four-dimensional respiration-correlated CT scans for radiotherapy planning of lung cancer patients. Int J Radiat Oncol Biol Phys 2006;65:1560–1571. 18. Zhang L, Garden AS, Lo J, et al. Multiple regions-of-interest analysis of setup uncertainties for head-and-neck cancer radiotherapy. Int J Radiat Oncol Biol Phys 2006;64: 1559–1569. 19. Wong JR, Grimm L, Uematsu M, et al. Image-guided radiotherapy for prostate cancer by CT-linear accelerator combination: prostate movements and dosimetric considerations. Int J Radiat Oncol Biol Phys 2005;61:561–569. 20. Thilmann C, Nill S, Tucking T, et al. Correction of patient positioning errors based on in-line cone beam CTs: Clinical implementation and first experiences. Radiat Oncol 2006; 1:16. 21. Pisani L, Lockman D, Jaffray D, et al. Setup error in radiotherapy: On-line correction using electronic kilovoltage and megavoltage radiographs. Int J Radiat Oncol Biol Phys 2000; 47:825–839.