Comparative Assessment of Liver Tumor Motion Using Cine–Magnetic Resonance Imaging Versus 4-Dimensional Computed Tomography

International Journal of

Radiation Oncology biology

physics

www.redjournal.org

Physics Contribution

Comparative Assessment of Liver Tumor Motion Using CineeMagnetic Resonance Imaging Versus 4-Dimensional Computed Tomography Annemarie T. Fernandes, MD,* Smith Apisarnthanarax, MD,y Lingshu Yin, PhD,* Wei Zou, PhD,z Mark Rosen, MD, PhD,x John P. Plastaras, MD, PhD,* Edgar Ben-Josef, MD,* James M. Metz, MD,* and Boon-Keng Teo, PhD* Departments of *Radiation Oncology and xRadiology, University of Pennsylvania, Philadelphia, Pennsylvania; yDepartment of Radiation Oncology, University of Washington, Seattle, Washington; and zRutgers Cancer Institute of New Jersey, New Brunswick, New Jersey Received Aug 28, 2014, and in revised form Dec 19, 2014. Accepted for publication Dec 29, 2014.

Summary Accurate radiation therapy targeting of hepatic tumors is critically dependent on organ and tumor motion assessment. The extent of tumor motion measured using 4DCT and cine-MRI in patients with hepatic tumors treated with radiation therapy was evaluated. CineMRI detected larger differences in hepatic intrafraction tumor motion when compared with 4DCT, most notably in the superior/inferior direction, and may be useful when treating without respiratory management,

Purpose: To compare the extent of tumor motion between 4-dimensional CT (4DCT) and cine-MRI in patients with hepatic tumors treated with radiation therapy. Methods and Materials: Patients with liver tumors who underwent 4DCT and 2-dimensional biplanar cine-MRI scans during simulation were retrospectively reviewed to determine the extent of target motion in the superioreinferior, anteriore posterior, and lateral directions. Cine-MRI was performed over 5 minutes. Tumor motion from MRI was determined by tracking the centroid of the gross tumor volume using deformable image registration. Motion estimates from 4DCT were performed by evaluation of the fiducial, residual contrast (or liver contour) positions in each CT phase. Results: Sixteen patients with hepatocellular carcinoma (nZ11), cholangiocarcinoma (nZ3), and liver metastasis (nZ2) were reviewed. Cine-MRI motion was larger than 4DCT for the superioreinferior direction in 50% of patients by a median of 3.0 mm (range, 1.5-7 mm), the anterioreposterior direction in 44% of patients by a median of 2.5 mm (range, 1-5.5 mm), and laterally in 63% of patients by a median of 1.1 mm (range, 0.2-4.5 mm). Conclusions: Cine-MRI frequently detects larger differences in hepatic intrafraction tumor motion when compared with 4DCT most notably in the superioreinferior direction, and may be useful when assessing the need for or treating without respiratory management, particularly in patients with unreliable 4DCT imaging. Margins wider than the internal target volume as defined by 4DCT were required to encompass

Reprint requests to: Boon-Keng Teo, PhD, Department of Radiation Oncology, University of Pennsylvania, 3400 Civic Center Boulevard, TRC 2 West, Philadelphia, PA, 19104; Tel: 215-662-2428; E-mail: kevin.teo@ uphs.upenn.edu Conflict of interest: none. Int J Radiation Oncol Biol Phys, Vol. 91, No. 5, pp. 1034e1040, 2015 0360-3016/$ - see front matter Ó 2015 Elsevier Inc. All rights reserved. http://dx.doi.org/10.1016/j.ijrobp.2014.12.048

Supplementary material for this article can be found at www.redjournal.org. AcknowledgmentdThe data in this paper were presented at the 55th Annual Meeting of the American Society for Radiation Oncology, September 22-25, 2013, Atlanta, GA.

Volume 91  Number 5  2015

particularly in patients with unreliable 4DCT imaging.

Liver tumor motion: Cine-MRI versus 4DCT 1035

nearly all the motion detected by cine-MRI for some of the patients in this study. Ó 2015 Elsevier Inc. All rights reserved.

Introduction The accuracy of radiation treatment targeting is affected by organ and tumor motion, particularly when treating organs that are sensitive to respiratory motion, such as the liver (1). Methods to assess motion when treating hepatic malignancies include fluoroscopy, respiratory-correlated CT (4DCT), cone-beam CT (3-dimensional, gated, or 4D), and cine-MRI. Fluoroscopy and 4DCT are modalities that capture snapshots of motion but are limited to short acquisition durations due to radiation exposure. In addition, these imaging modalities often require placement of radioopaque fiducial markers to appropriately track motion in the region of the hepatic tumor, because the tumor frequently cannot be visualized directly. High-quality 4DCT images require a consistent and regular breathing pattern during acquisition. Irregular breathing results in phase sorting errors that manifest as image artifacts that may compromise the accuracy of motion assessment in radiation therapy (2, 3). In addition to the superior softtissue image contrast offered by MRI compared with CT, cine-MRI does not expose patients to radiation, thus allowing for longer evaluation that may capture irregular motion that may be present during beam-on time (4). Previous studies have compared fiducial marker motion using 4DCT versus fluoroscopy or respiratory tracking (5, 6), liver tumor motion using cine-MRI with or without abdominal compression (7), and liver tumor motion using cine-MRI versus diaphragm motion with fluoroscopy (8). Although 4DCT has been compared with cine-MRI in patients with lung tumors where fiducial markers are not necessary (9), to our knowledge no studies exist that compare 4DCT with cine-MRI in the evaluation of hepatic tumor motion. The purpose of this study was to compare the differences in tumor motion estimation between 4DCT and cine-MRI in patients treated with radiation therapy for hepatic malignancies.

Methods and Materials

therapy (nZ3). The average tumor volume was 209.3 cm3. The majority (nZ11) of tumors were located in the medial division (liver segments 1, 4, 5, 8) compared with the left lateral (nZ3, segments 2 or 3) or right lateral (nZ2, segments 6 or 7) divisions. There were 6 patients with tumors in the superior portion of the liver. Few patients had ascites (nZ4) or Child-Pugh B or C cirrhosis (nZ3).

4DCT motion analysis 4DCT simulation procedure All patients underwent a low pitch (0.1) spiral 4DCT scan in a free-breathing, uncoached state on a Siemens Sensation Open scanner (Siemens Medical Solutions, Malvern, PA). The 4DCT data were sorted into 8 respiratory phase bins using the real-time position management system (Varian Medical Systems, Palo Alto, CA). The voxel size was 0.97 mm  0.97 mm  3 mm. The total 4DCT acquisition time varied from 60 seconds to 120 seconds. The time to acquire data in the area of interest (eg, fiducial marker) was approximately 2-3 breathing cycles and varied between 6 and 15 seconds. 4DCT motion analysis The 4DCT data were analyzed in the Eclipse treatment planning system, version 11 (Varian Medical Systems). Thirteen patients had fiducial or surrogate marker placement suitable for tracking on 4DCT: 8 had Visicoils placed percutaneously next to the tumor before simulation, 1 patient had surgical clips close to the tumor, and 4 patients had residual enhancing lipiodol (Fig. 1A) from prior transarterial chemoembolization close to the tumor. In these patients the most visible marker was delineated on each phase of the 4DCT. The most extreme positions of the center of the fiducial contour were tracked in 3 directionsdsuperioreinferior (S/I), anterioreposterior (A/P), and lefteright (L/R)dto determine the range of tumor motion. In the 3 patients without trackable markers, identifiable features of the liver contour (eg, inferior border) near the tumor (identified from MRI) were used to evaluate tumor motion (Fig. 1B-D) on 4DCT.

Patients Cine-MRI motion analysis Sixteen patients with hepatic tumors who completed radiation therapy between October 2011 and April 2013 were included in this study under an institutional review boardeapproved retrospective protocol. Patients had hepatocellular carcinoma (nZ11), cholangiocarcinoma (nZ3), or liver metastases (nZ2). Patients were treated with conventionally fractionated proton radiation therapy (nZ8), conventionally fractionated intensity modulated radiation therapy (nZ5), or stereotactic body radiation

Cine-MRI acquisition procedure All patients underwent cine-MRI in the treatment position on the same day approximately 1 hour after the 4DCT scan. Biplanar 2-dimensional (2D) images in the sagittal and coronal planes were obtained through the centroid of the dominant liver tumor on a Siemens Espree 1.5 T scanner (Siemens Medical Solutions) using a balanced steady state free precession (bSSFP) sequence with an

1036 Fernandes et al.

International Journal of Radiation Oncology  Biology  Physics

Fig. 1. Representative examples of 4-dimensional CT motion evaluation. (A) Representative motion evaluation in a patient with residual lipiodol. (B) T1-weighted MRI was used to help identify the target on exhale phase of 4-dimensional CT. (C, D) Liver and target contours in inhale (C) and exhale (D) phases used for evaluation of motion without markers. in-plane voxel size of 2.81 mm  2.81 mm, slice thickness of 9.6 mm, field of view of 36 cm  36 cm, flip angle of 30 , echo time of 1.09 ms, and repetition time of 300 ms. The torso and spine coils were used. Each sagittal and coronal image was acquired in an alternating fashion every 0.3 seconds for a total of 1000 images over 5 minutes. Cine-MRI tumor tracking and motion analysis The tumor was contoured directly in 2D on the first sagittal and coronal bSSFP images. In some cases, T1-weighted images were used to help demarcate the tumor boundaries if they were less visible on the bSSFP images (Fig. 2). The contours were then propagated to all subsequent images using rigid image registration of the liver outline, followed by deformable image registration (MIM Software Inc, Cleveland, OH). Minor editing was made to the contours when necessary. For each patient, the tumor centroid positions were graphed over time and the histogram and cumulative distribution functions generated. The range of the tumor motion was calculated by subtracting the tumor position at the 90th percentile from the tumor position at the 10th percentile. This number represents the range where the border of the tumor in each axial direction can be found 90% of the time. A similar analysis was performed on each 1minute block of cine-MRI subdivided from the 5-minute data, to evaluate the impact of irregular breathing and variations in motion estimates acquired at different time points.

Comparative analysis A direct comparison of motion amplitudes from 4DCT versus cine-MRI can never be made given the different

imaging times used (several seconds vs 5 minutes). If 100% of the tumor contours are used to generate a cine-MRI internal target volume (ITV), very large motion envelopes encompassing 100% of all tumor positions will result because some of the rare but deeper irregular inhale/exhale events are captured during the 5-minute scan. The cineMRI motion amplitude that was generated using a 10% cutoff of the deepest inhale and exhale positions is equivalent to the motion margin such that for 80% of the time, there is 100% volume coverage with some loss in coverage 20% of the time (10% of time in each direction). Determination of overall target coverage is a much more complex calculation, because it depends on the distribution of positions and extent of tumor excursions outside the 90th percentile. We assume that a 90th percentile cutoff is likely to exclude most of the highly irregular respiratory cycles. The 4DCT range of motion was subtracted from the cine-MRI range of motion for each direction. Positive values indicate larger cine-MRI motion compared with 4DCT, and negative values indicate smaller cine-MRI motion compared with 4DCT. Clinically significant changes in motion were defined as motion difference of 3 mm, because this cutoff has been used in previous studies evaluating the impact of hepatic tumor motion. This value represented either the S/I resolution of the 4DCT (5) or the threshold for repositioning in image-guided radiation therapy (7). Statistical analysis was performed using STATA 13 IC software (College Station, TX) using descriptive statistics and box plots to display all motion data. The Wilcoxon rank sum test, paired t test, or c2 test was used to investigate the relationship between patient, tumor, and treatment

Volume 91  Number 5  2015

Liver tumor motion: Cine-MRI versus 4DCT 1037

Fig. 2. Magnetic resonance imaging liver tumor contouring. Liver tumor indicated by red arrows on the balanced steady state free precession image on the coronal (A) and sagittal (B) views. In (C), the tumor boundary was less visible on the balanced steady state free precession image, and a T1-weighted image (D) was used to help demarcate the tumor boundary. A color version of this figure is available at www.redjournal.org. characteristics with motion in the S/I direction. For analysis, S/I direction tumor motion using cine-MRI was dichotomized into <1 cm or 1 cm. The factors tested were diagnosis, tumor volume, liver volume, tumor location, presence of cirrhosis, presence of ascites, type of imaging marker (fiducial vs lipidol), and treatment modality. A second analysis was performed to assess motion after excluding the 3 patients without trackable markers on 4DCT.

Results A representative S/I displacement time graph is shown in Figure 3A, whereas the corresponding frequency histogram and cumulative distribution function is shown in Figure 3B. The mean, median, and range of hepatic tumor motion estimates in the S/I, A/P, and L/R directions using 4DCT and cine-MRI are displayed in Table 1 and Figure e1 (available online at www.redjournal.com). The largest hepatic tumor motion was seen in the S/I direction, whereas the smallest was seen in the L/R direction. The median hepatic tumor motion differences between cine-MRI and 4DCT were 1.8 mm (6.0 to 7.0 mm), 0.8 mm (2.0 to 5.5 mm), and 0.8 mm (2.0 to 4.5 mm) in the S/I, A/P, and L/R directions, respectively. No patient, tumor, or treatment characteristics significantly correlated with motion in the

S/I direction. Pairwise differences between 4DCT and cine-MRI motion estimates in all 3 directions are displayed in Table 2 (see also Fig. e1, available online at www .redjournal.com). Cine-MRI motion was larger compared with 4DCT for the S/I direction in 50% of patients by a median of 3.0 mm (range, 1.5-7 mm), the A/P direction in 44% of patients by a median of 2.5 mm (range, 1-5.5 mm), and the lateral margin in 63% of patients by a median of 1.1 mm (range, 0.2-4.5 mm). To include motion detected by both cine-MRI and 4DCT, we estimate that additional margins of 3 mm, 2 mm, and 1 mm in the S/I, A/P, and L/R directions, respectively, to the 4DCT, would be necessary in 90% (10) of the patients in this study where neither coaching nor abdominal compression were used. The regularity of respiration during the cine-MRI session was evaluated by comparing the tumor S/I motion measurements in 1-minute intervals versus the entire 5minute duration. Figure 4 indicates that the range in tumor motion varied by at least 4 mm between the different 1minute data intervals in 5 subjects (1, 2, 7, 9, and 12). In 4 subjects (1, 6, 8, and 11), the motion estimate from 4DCT differed from the average 1-minute cine data by more than 3 mm. The difference between 4DCT S/I motion estimates and the motion estimates derived from individual 1-minute cine motion data was statistically significant (PZ.029). However, the difference was not significant

1038 Fernandes et al.

International Journal of Radiation Oncology  Biology  Physics

Fig. 3. Liver tumor motion by cine-MRI in one representative patient. (A) The 500 data points of vertical displacement normalized to the median tumor position were plotted over time (300 seconds). (B) Data points from (A) were plotted in histogram form to display frequency of tumor position. To determine the tumor motion by cine-MRI, the tumor position from the 10th percentile was subtracted from the tumor position at the 90th percentile. (PZ.156) when the comparison was performed using the average of the 1-minute data subsets for each patient. This is an indication that the regularity of respiration plays an important role in obtaining an accurate motion estimate.

Discussion Cine-MRI can be a complementary imaging modality to 4DCT for assessing tumor motion because patients can be imaged over longer periods of time. To our knowledge, this is the first study comparing liver tumor motion between 4DCT and cine-MRI. We demonstrate that cine-MRI detects differences in hepatic intrafraction tumor motion when compared with 4DCT, most notably in the S/I direction. Because not all patients can tolerate active breathing control (11, 12), when treating without breathhold, caution should be taken: differences in tumor

motion detected between 4DCT and cine-MRI may be clinically significant. The magnitude of liver tumor motion in this study is comparable to liver tumor motion reported by Eccles et al (7), despite differences in methods used to assess motion. Eccles et al evaluated tumor edge motion in 4 regions of interest on cine-MRI and determined the range of tumor motion as the 95th percentile minus the 5th percentile. They reported amplitudes of 9.4 to 11.7 mm and 4.9 to 5.6 mm in the S/I and A/P directions, respectively (7). Our study evaluated the range of motion of the liver tumor centroid position as the 90th percentile minus the 10th percentile and found a range of 9.2 mm and 5.5 mm in the S/I and A/P directions, respectively. Kirilova et al (8) found more dramatic liver tumor motion of 15.5 mm, 10 mm, and 7.5 mm in the S/I, A/P, and L/R directions, respectively. This study also evaluated tumor edge by cine-MRI but used the maximal tumor edge differences rather than frequency

Volume 91  Number 5  2015 Table 1

Liver tumor motion: Cine-MRI versus 4DCT 1039

Liver tumor motion using 4DCT and cine-MRI

Motion direction

Mean/median 4DCT motion (range), mm

Mean/median cine-MRI motion (range), mm

Superioreinferior Anterioreposterior Lefteright

8.0/7.5 (4-16) 4.8/4.0 (2-10) 2.1/2.0 (0-6)

9.2/9.3 (3-18) 5.5/4.8 (2.0-15.5) 2.8/2.8 (1-4.5)

Abbreviations: 4D Z 4-dimensional; CT Z computed tomography; MRI Z magnetic resonance imaging.

percentiles, which likely included irregularities in breathing amplitude. Gierga et al (13) evaluated abdominal tumor motion with fluoroscopy and found S/I motion of 7.4 mm and A/P motion of 3.8 mm. Because tumor motion was evaluated by fluoroscopy over 30 seconds, this motion is likely an underestimation of true liver tumor motion. A comparative study between 4DCT and 30-second cineMRI was recently performed for liver tumors. Even with the shorter 30-second study, larger ITVs were observed with MRI than with 4DCT (14). Heerkens et al (10) evaluated the motion of the pancreas using cine-MRI for the purposes of margin determination, but the data were only limited to 2 60-second scans. Estimation of tumor motion from 4DCT relies entirely on the reproducibility and consistency of respiration at the time of simulation and during treatment. Analysis of the consecutive 1-minute datasets from the 5-minute cine-MRI data showed that irregular breathing can lead to variations in motion estimates of at least 4 mm in 31% of patients. Such variations may be due to patient anxiety while inside the MRI but may also be indicative of the naturally variable breathing pattern. These results indicate the difficulty in getting a highly accurate estimation of tumor motion from a single 4DCT scan, which captures tumor motion over a few breathing cycles in a single session. Monitoring of respiratory pattern before, during, and after the 4DCT session may be necessary to determine whether patient respiration during the 4DCT scan is representative of longer-term respiratory pattern. If necessary, repeat 4DCT imaging may be conducted if the initial 4DCT scan is deemed unreliable owing to inconsistency in patient breathing. The beam-on time is approximately 2 to 4 minutes for Table 2

conventionally fractionated liver intensity modulated radiation therapy or proton therapy and approximately 4 minutes for stereotactic body radiation therapy. Therefore, a 5-minute cine-MRI motion may offer better insight into actual tumor motion during treatment compared with a 4DCT scan acquired over only 2 to 3 breathing cycles. In our study, we found that median differences between 4DCT and cine-MRI measures of liver tumor were 3.0 mm, 2.5 mm, and 1.1 mm in 50%, 43.8%, and 62.5% of patients in the in the S/I, A/P, and L/R directions, respectively. In the S/I direction, 31.3% of patients had clinically significant larger liver tumor using cine-MRI compared with 4DCT, whereas fewer patients in the A/P (18.8%) and lateralemedial (6.1%) directions had clinically significant tumor margin differences. Ge et al (5) compared 4DCT with daily fluoroscopy and found comparable tumor margin underestimation by 4DCT, with a median of 3.9 mm, 3.0 mm, and 1.7 mm in the S/I, A/P, and L/R directions, respectively (5). Inclusion of nearly all the motion detected by both cine-MRI and 4DCT would require additional margins of 3 mm, 2 mm, and 1 mm in the S/I, A/P, and L/R directions, respectively, to the 4DCT-derived ITV in 90% of the cases in this study. Our study is limited by relatively small patient numbers and the reliance of fiducial markers for measuring liver tumor motion on 4DCT. Because of the discrepancy between fiducial movement and tumor movement, possible migration of fiducial markers, and whole-liver hysteresis, fiducial marker motion may not be an accurate surrogate for hepatic tumor motion. In the 3 patients without fiducial markers, liver motion was used as a tumor motion surrogate. Although liver motion has been shown to be a poor surrogate for tumor motion (8), a secondary analysis excluding these 3 patients did not alter our results significantly, and our conclusions remain the same. We consider this effect to be small compared with the discrepancy in motion amplitudes measured between 4DCT and cine-MRI, which is the focus of this work. Additionally, our study is limited by the difference in methods for evaluating liver tumor motion by 4DCT and cine-MRI. Because we are unable to identify the hepatic tumor on 4DCT and we cannot identify the fiducial marker on MRI, we used 2 different motion identification

Pairwise differences between 4DCT and cine-MRI motion estimates in the S/I, A/P, and L/R directions Parameter

S/I

A/P

L/R

Patients with larger cine-MRI motion compared with 4DCT Differences in subgroup (mm) Patients with larger cine-MRI motion compared with 4DCT that is clinically significant (>3 mm) Patients with smaller cine-MRI motion compared with 4DCT Differences in subgroup (mm) Patients with smaller cine-MRI motion compared with 4DCT that is clinically significant (>3 mm)

8 (50) 3.0 (1.5-7) 5 (31)

7 (44) 2.5 (1-5.5) 3 (19)

10 (63) 1.1 (0.2-4.5) 1 (6)

4 (25) 0.75 (0.5-6) 1 (6)

8 (50) 1.0 (0.1-2) 0

4 (25) 0.5 (0.5-2) 0

Abbreviations: A/P Z anterioreposterior; L/R Z lefteright; S/I Z superioreinferior. Values are number (percentage) or median (range).

1040 Fernandes et al.

International Journal of Radiation Oncology  Biology  Physics

effect is minimal owing to the relatively large slice thickness of 9.6 mm.

Conclusions Cine-MRI is a helpful tool to accurately assess hepatic tumor motion, to determine the need for respiratory management, and to determine appropriate tumor margins in settings where respiratory management is not available or feasible. For the patients in this study, inclusion of additional margins to the internal target volume defined by 4DCT may be needed to account for tumor motion detected by cine-MRI, particularly for patients with irregular breathing. Fig. 4. Comparison of 4-dimensional CT (4DCT, red triangles) with cine-MRI superioreinferior motion over entire 5 minutes (blue circles) and cine-MRI superioreinferior motion over 1-minute intervals (x). A color version of this figure is available at www.redjournal.org. methods in this study. With the 4DCT, we looked at tumor motion extremes on the 8 respiratory-phase images. When determining tumor motion with the cineMRI, we had 500 images to use, so we looked at the 90% confidence interval for tumor motion in each direction. Although 2 different methods for assessing tumor motion were used, this difference augments our conclusion, because the discrepancy would bias our results to favor underestimation of tumor motion with the cineMRI, where 20% of data points were not considered. Although both scans were acquired on the same day, the patient’s breathing pattern may be different between the scans. However, the main advantage of the cine-MRI is the long imaging time that may offer insight into potential large intrafraction amplitude variations during treatment. An often overlooked effect in 4DCT imaging is motion blurring caused by the finite X-ray tube rotation time. The typical X-ray rotation time in 4DCT scans is between 0.5 and 1.0 seconds, which implies that each CT phase image is time-averaged over this interval. Because the acquisition time for each time point in the cine-MRI is 0.3 seconds, the cine-MRI has better time resolution and is better able to capture the extreme positions of the tumor motion than 4DCT. This may account for some portion of the larger motion amplitude measured on cine-MRI compared with 4DCT. One other confounding factor in the 2D MRI motion estimate is the effect of tumor motion in and out of the imaging plane, which can lead to an apparent shift in the contour if the target has a curvature perpendicular to the imaging plane. In our study, this

References 1. Rosu M, Dawson LA, Balter JM, et al. Alterations in normal liver doses due to organ motion. Int J Radiat Oncol Biol Phys 2003;57: 1472-1479. 2. Balter JM, Ten Haken RK, Lawrence TS, et al. Uncertainties in CTbased radiation therapy treatment planning associated with patient breathing. Int J Radiat Oncol Biol Phys 1996;36:167-174. 3. Chen GT, Kung JH, Beaudette KP. Artifacts in computed tomography scanning of moving objects. Semin Radiat Oncol 2004;14:19-26. 4. Shimizu S, Shirato H, Aoyama H, et al. High-speed magnetic resonance imaging for four-dimensional treatment planning of conformal radiotherapy of moving body tumors. Int J Radiat Oncol Biol Phys 2000;48:471-474. 5. Ge J, Santanam L, Noel C, et al. Planning 4-dimensional computed tomography (4DCT) cannot adequately represent daily intrafractional motion of abdominal tumors. Int J Radiat Oncol Biol Phys 2013;85: 999-1005. 6. Minn AY, Schellenberg D, Maxim P, et al. Pancreatic tumor motion on a single planning 4D-CT does not correlate with intrafraction tumor motion during treatment. Am J Clin Oncol 2009;32:364-368. 7. Eccles CL, Patel R, Simeonov AK, et al. Comparison of liver tumor motion with and without abdominal compression using cine-magnetic resonance imaging. Int J Radiat Oncol Biol Phys 2011;79:602-608. 8. Kirilova A, Lockwood G, Choi P, et al. Three-dimensional motion of liver tumors using cine-magnetic resonance imaging. Int J Radiat Oncol Biol Phys 2008;71:1189-1195. 9. Cai J, McLawhorn R, Read PW, et al. Effects of breathing variation on gating window internal target volume in respiratory gated radiation therapy. Med Phys 2010;37:3927-3934. 10. Heerkens HD, Van Vulpen M, Van den Berg CA, et al. MRI-based tumor motion characterization and gating schemes for radiation therapy of pancreatic cancer. Radiother Oncol 2014;111:252-257. 11. Dawson LA, Brock KK, Kazanjian S, et al. The reproducibility of organ position using active breathing control (ABC) during liver radiotherapy. Int J Radiat Oncol Biol Phys 2001;51:1410-1421. 12. Eccles C, Brock KK, Bissonnette JP, et al. Reproducibility of liver position using active breathing coordinator for liver cancer radiotherapy. Int J Radiat Oncol Biol Phys 2006;64:751-759. 13. Gierga DP, Chen GT, Kung JH, et al. Quantification of respirationinduced abdominal tumor motion and its impact on IMRT dose distributions. Int J Radiat Oncol Biol Phys 2004;58:1584-1595. 14. Akino Y, Oh RJ, Masai N, et al. Evaluation of potential internal target volume of liver tumors using cine-MRI. Med Phys 2014;41:111704.