- Email: [email protected]
Three-Dimensional Motion of Liver Tumors Using Cine-Magnetic Resonance Imaging
Int. J. Radiation Oncology Biol. Phys., Vol. 71, No. 4, pp. 1189–1195, 2008 Copyright Ó 2008 Elsevier Inc. Printed in the USA. All rights reserved 0360-3016/08/$–see front matter
doi:10.1016/j.ijrobp.2007.11.026
CLINICAL INVESTIGATION
Liver
THREE-DIMENSIONAL MOTION OF LIVER TUMORS USING CINE-MAGNETIC RESONANCE IMAGING ANNA KIRILOVA, B.SC.,* GINA LOCKWOOD, M.MATH.,y PERRY CHOI, M.D.,z NEELUFER BANA, B.SC. (HON),z MASOOM A. HAIDER, M.D.,x KRISTY K. BROCK, PH.D.,* CYNTHIA ECCLES, B.SC.,z z AND LAURA A. DAWSON, M.D. Departments of *Radiation Physics, y Biostatistics, z Radiation Oncology, and x Medical Imaging, Princess Margaret Hospital, University of Toronto, Toronto, ON, Canada Purpose: To measure the three-dimensional motion of liver tumors using cine-magnetic resonance imaging (MRI) and compare it to the liver motion assessed using fluoroscopy. Methods and Materials: Liver and liver tumor motion were investigated in the first 36 patients with primary (n = 20) and metastatic (n = 16) liver cancer accrued to our Phase I stereotactic radiotherapy study. At simulation, all patients underwent anteroposterior fluoroscopy, and the maximal diaphragm excursion in the craniocaudal (CC) direction was observed. Cine-MRI using T2-weighted single shot fast spin echo sequences were acquired in three orthogonal planes during free breathing through the centroid of the most dominant liver tumor. ImageJ software was used to measure the maximal motion of the tumor edges in each plane. The intra- and interobserver reproducibility was also quantified. Results: The average CC motion of the liver at fluoroscopy was 15 mm (range, 5–41). On cine-MRI, the average CC tumor motion was 15.5 mm (range, 6.9–35.4), the anteroposterior motion was 10 mm (range, 3.7–21.6), and the mediolateral motion was 7.5 mm (range, 3.8–14.8). The fluoroscopic CC diaphragm motion did not correlate well with the MRI CC tumor motion (r = 0.25). The mean intraobserver error was <2 mm in the CC, anteroposterior, and mediolateral directions, and 90% of measurements between observers were within 3 mm. Conclusions: The results of our study have shown that cine-MRI can be used to directly assess liver tumor motion in three dimensions. Tumor motion did not correlate well with the diaphragm motion measured using kilovoltage fluoroscopy. The tumor motion data from cine-MRI can be used to facilitate individualized planning target volume margins to account for breathing motion. Ó 2008 Elsevier Inc. Cine-MRI, Hepatobiliary cancer, Liver metastases, Organ motion.
Detailed information about liver and liver tumor motion is crucial for establishing the smallest and safest margin for irradiation without compromising the therapeutic dose necessary for optimal tumor coverage. Liver motion is mostly related to breathing and is largest in the craniocaudal (CC) direction, with a range of 5–50 mm (1–3). Its adverse effect on radiotherapy (RT) planning and treatment include the introduction of artifacts on the planning computed tomography (CT) scans, leading to inaccurate tumor and normal tissue volumes (4–6), altered dosimetry using a static plan (7), an increased volume of normal tissue exposure (8), increased
required planning target volume (PTV) margins (9), and a greater risk of toxicity. Before liver tumor motion can be considered in planning and appropriate methods of its management applied, the magnitude of tumor motion during free breathing must be measured. Diaphragm CC motion measured during kilovoltage (kV) fluoroscopy can be used as a surrogate for liver tumor motion. However, liver motion might not necessarily correlate with the CC diaphragm motion, because it is more complex, consisting of translations, rotations, and hysteresis (10). Earlier studies have investigated liver motion with different imaging modalities under various breathing conditions, including a scintillation camera after administration
Reprint requests to: Anna Kirilova, B.Sc., Department of Radiation Physics, Princess Margaret Hospital, University of Toronto, 610 University Ave., Toronto, ON M5G 2M9 Canada. Tel: (416) 946-7805; Fax: (416) 946-6566; E-mail: anna.kirilova@rmp. uhn.on.ca Supported in part by a grant from Varian Medical Systems and by the generosity of the Susan Grange family. Patients treated in this study were treated on research protocols funded in part by Elekta Oncology Systems, National Cancer
Institute of Canada Grant 018207, and the Canadian Cancer Society. L. Dawson was a recipient of an American Society for Clinical Oncology career development award. Presented in part at the European Society of Radiation Oncology (ESTRO) 2004 Annual Scientific Meeting, Amsterdam, October 2004. Conflict of interest: none. Received Aug 8, 2007, and in revised form Nov 16, 2007. Accepted for publication Nov 16, 2007.
INTRODUCTION
1189
1190
I. J. Radiation Oncology d Biology d Physics
of 99Tc (11, 12), ultrasonography (1, 3), fluoroscopy, and computed tomography (5, 7). Fiducial markers placed in, or close to, the tumor have been used by other investigators to better estimate the motion in relation to the diaphragm using fluoroscopy and CT. The reported limitations with this method include possible migration of the markers during the RT course and the lack of three-dimensional motion information (13). Repeat CT scans acquired during exhale and inhale breath hold have been used to estimate the extremes of liver position resulting from breathing motion (14). More recently, respiratory sorted ‘‘four-dimensional’’ (4D) CT has been used to quantify lung and liver motion during the period of image acquisition (15, 16). However, liver tumors are not well visualized without the use of intravenous contrast, which is generally not used during respiratory-sorted CT scans (17). At present, no consensus has been reached regarding the best imaging strategies for organ motion and its management for RT planning and delivery. However, regardless of the intended strategy to consider and/or reduce the adverse effects of motion for RT planning and treatment, the breathing motion needs to be estimated. Because liver tumor motion varies substantially among patients, there is motivation to develop strategies that allow repeated, low-risk measurements of patients’ liver tumor motion due to breathing. Magnetic resonance imaging (MRI) can be used for this purpose. It has the advantage of improved soft tissue visualization compared with CT (17, 18). Also, the recently available high-speed MRI sequences (cine-MRI) permit direct visualization of the respiratory motion of the liver tumor itself, without the concern of the dose exposure to the patient. Shimizu et al. (19) first reported the feasibility of high-speed MRI to assess liver tumor motion in 1999. The primary goal of this study was to describe the three-dimensional (3D) motion of liver tumors using cine-MRI in the first 36 patients undergoing treatment planning for stereotactic body RT on our Phase I-II protocol. The secondary goals were to compare the 3D tumor motion with the diaphragm motion assessed at anteroposterior (AP) fluoroscopy and to measure the inter- and intraobserver variability in motion assessment using cine-MRI. METHODS AND MATERIALS Patients The patients included in this study had provided written informed consent for an institutional research ethics board–approved Phase III protocol of stereotactic body RT for liver metastases or hepatobiliary carcinoma. The eligibility criteria included no contraindications for MRI, unresectable liver cancer, a Child-Pugh A liver function score, and a life expectancy >12 weeks. Patients whose disease burden was not in the liver and those with a Karnofsky performance status of <60 were excluded. All patients accrued to the clinical study underwent imaging with fluoroscopy and MRI, in addition to a triphasic planning CT scan at simulation.
Fluoroscopy The patients were placed in the supine position either on a chest board (MedTech, Orange City, IA) or an evacuated immobilization
Volume 71, Number 4, 2008
bag (Vac Lok, Bionix, Toledo, OH), with their arms above their head and a leg immobilizer under their knees. AP fluoroscopy imaging was done on a conventional simulator (Nucletron, Veenendaal, The Netherlands) to measure the CC range of the diaphragm during free breathing for 60 s at a 100-cm source-to-axis distance set to the approximate mid-plane of the patient at the level of the mid-right costal margin. The fluoroscopic images were recorded on VHS or DVD for off-line analysis of each fluoroscopic session. The maximal CC excursion of the dome of the diaphragm was measured. The simulator’s graticule was calibrated daily as a part of the simulator quality assurance procedure. At the analysis, the magnification of the graticule was determined using a standard metal ruler to measure the distance between the indexes on the gradicule as seen on the video screen, and a magnification factor determined by dividing the measured value over the true value. This factor was then applied to all measurements.
Cine-MRI On the same day of fluoroscopy, cine-MRI was done to measure the three-dimensional tumor motion during free breathing. The patients were placed in the same position as for fluoroscopy, either on a flat table top with an evacuated immobilization bag (Vac Lok) that fit in the magnet bore or only on a flat table top when the patient was too large to fit in the magnet bore with the immobilization bag. Single 5-mm slices through the centroid of the dominant liver tumor were acquired manually in axial, sagittal, and coronal planes on a 1.5T General Electric Signa Twin Speed (Milwaukee, WI) scanner, using the General Electric four-channel torso coil and a T2-weighted single shot fast spin echo sequence with the following parameters (echo time, 90 ms; repetition time, 1,300 ms; slice thickness, 5 mm; field of view, 32–48 cm; matrix, 256 192). The in-plane resolution was 1.6–2.5 mm, and the temporal resolution was one slice/s during a 60-s period for each orthogonal plane. If the frequency of imaging appeared to be somewhat in phase with the respiratory frequency, a short pause (e.g., 0.5 s) was introduced to provide a noticeable phase difference to ensure that the maximal excursion of motion was more likely to be measured.
Statistical analysis The maximal diaphragm excursion in the CC direction as measured on AP kV fluoroscopy was recorded prospectively at simulation by a radiation therapist and verified off-line by a radiation oncologist (Fig. 1). All cine-MRI series were exported to a Digital Imaging and Communications in Medicine server and evaluated using ImageJ software, version 1.31 (National Institutes of Health, Bethesda, MD). ImageJ is a public domain Java image-processing program that can be used to track a point or object through a series of images. In our study, the position of the tumor edge in each image was determined, so that for each imaging series in one plane, the maximal motion of the tumor edges could be evaluated. The 12 measurements obtained for each patient were performed twice by two observers in 36 patients, for a total of 1,728 measurements (Fig. 2). The four observer measurements were averaged for each tumor edge in each plane, and the maximal motion was calculated for the superior, inferior, anterior, posterior, medial, and lateral edges of the tumor and for the CC, AP and mediolateral (ML) directions overall. The 3D coordinates for every measurement was saved in Excel and checked by a third observer. Box plots were used to display all motion data. The maximal motion of the superior and inferior, anterior and posterior, and medial and lateral edges of the tumor and the CC motion in the coronal and sagittal planes, AP motion in the sagittal and
3D motion of liver tumors using Cine-MRI d A. KIRILOVA et al.
1191
Fig. 1. (a) Maximal inhale diaphragm position, and (b) maximal exhale diaphragm position, 20 mm in craniocaudal direction, measured with anteroposterior fluoroscopy for 30 s of normal breathing. The interobserver error was determined by averaging the absolute differences between the mean measurements of the two observers for each plane and summarizing for all patients using the mean values and ranges. Box plots are used to display the reproducibility data.
axial planes, and ML motion in the coronal and axial planes were compared using Mann-Whitney U tests. The motion between patients who did and did not have the immobilization device at the time of MRI were also compared using Mann-Whitney U tests. The fluoroscopic CC motion was correlated with the cine-MRI CC tumor motion using a Pearson correlation coefficient. Chisquare tests were used to investigate the relationship between selected disease characteristics and the motion in all directions, which were dichotomized into #10 and >10 mm. The factors tested were diagnosis, tumor size, and tumor location.
RESULTS All image data were acquired between July 2003 and October 2004. In this study, 36 patients with metastasis (n = 16), hepatocellular carcinoma (n = 14), and intrahepatic cholangiocarcinoma (n = 6) were imaged with fluoroscopy, cineMRI, and tri-phasic CT for the purpose of RT planning. These patients were subsequently treated with stereotactic body RT. On cine-MRI, the average maximal motion of the 36 liver tumors assessed was 15.5 mm (range, 6.9–35.4 mm) in the CC, 10.1 mm (range, 3.7–21.6 mm) in the AP, and 7.5 mm
Intra- and interobserver reproducibility The intraobserver error was quantified for each observer. The absolute differences in the repeat measurements of the maximal excursions were calculated for each tumor edge in each plane. The error for each plane was calculated for each observer separately and for the observers combined by averaging the relevant differences. The overall intraobserver variability was calculated similarly using the eight relevant measures.
a)
20.5 20
mm
19.5 19
18.5 1
11
21 Sec
31
41
51
b) 30.5
mm
30 29.5 29 1
11
21 Sec
31
41
51
c) Fig. 2. (a) Cranial (arrow) and caudal (arrowhead) edges of tumor on cine-magnetic resonance imaging, with graphs showing change in (b) cranial and (c) caudal tumor edge position from breathing, based on 50 s of cine-magnetic resonance imaging.
(range, 3.8–14.8 mm) in the ML directions (Table 1 and Fig. 3). The tumor motion observed was predominantly in the CC and AP directions and the least in the ML direction, as expected. No significant differences were found in the maximal motion between the superior and inferior edges of the tumor, anterior and posterior edges, or medial and lateral edges. Table 2 lists the motion separated by the edge evaluated. No differences were found when the motion was compared between planes, except for the AP motion, which was significantly greater in the axial plane (mean, 9.5 mm; median, 8.3 mm) compared with the the motion in the sagittal plane (mean, 7.7 mm; median, 6.1 mm). We also investigated the tumor motion in all directions (#10 mm vs. >10 mm) according to potential explanatory factors such as diagnosis, tumor size, and tumor location. None of these factors were significantly associated with the amplitude of motion. Of the 36 patients, 15 underwent simulation and treatment on a chest board or flat tabletop and 21 underwent simulation and treatment in the immobilization device (Vac Lok); however, only 15 patients underwent MRI with the Vac Lok. No significant differences were found in the diaphragm/liver tumor motion according to the use of the immobilization device at MRI. The mean intraobserver error in MRI tumor motion measurements for Observer 1 was 1.2 mm (range, 0.4–2.1 mm) in the CC direction, 0.9 mm (range, 0.3–2.7 mm) in the AP direction, and 0.7 mm (range, 0.3–1.4 mm) in the ML direction. For Observer 2, the error was 1.7 mm (range, 0.4–2.9 mm) in the CC direction, 1.5 mm (range, 0–4.3 mm) in the AP direction, and 1.7 mm (range, 0–4.0 mm) in the ML direction. The overall intraobserver error (mean of the two observers) was 1.4 mm (range, 0.4–2.3 mm), 1.2 mm (range, 0.4–3.1 mm), and 1.1 mm (range, 0.3–2.5) in the CC, AP, and ML direction, respectively (Fig. 4a). These were one-eleventh (CC), one-eighth (AP), and one-seventh (ML) the size of the calculated mean MRI motion. The overall interobserver error was 2.3 mm (range, 0.3– 10.5 mm) in the CC direction, 2.5 mm (range, 0.4–7.3 mm) in the AP direction, and 1.9 mm (range, 0.2–4.4) in the ML direction (Fig. 4b). These errors were one-seventh (CC) and onequarter (AP and ML) the size of the measured MRI motion. Anteroposterior fluoroscopy (n = 35) revealed that the average diaphragm motion in the CC direction was 15 7 mm (range, 5–41 mm). The fluoroscopic CC motion did not correlate well with the cine-MRI CC tumor motion (r = 0.25;
Volume 71, Number 4, 2008 40
Maximum motion, mm
I. J. Radiation Oncology d Biology d Physics
1192
30
20
10
0
AP
CC
ML
Fig. 3. Box plot of cine-magnetic resonance imaging directional maximal motion. Also shown are 50% interquartile range, median, range, and outliers. Median motion in each direction: craniocaudal, 13.3 mm, anteroposterior, 9.2 mm, and mediolateral, 6.9 mm.
Fig. 5). The mean absolute value of the difference between the fluoroscopic diaphragm CC motion and the cine-MRI CC tumor motion (n = 35) was 6.5 mm (range, 0.1– 0.6 mm). In 16 cases, the difference was >5 mm (45.7%) and in 8 cases was >10 mm (22.9%). Fluoroscopy overestimated the tumor motion in 51.4% of cases and underestimated it in 48.6% of cases (Table 3). DISCUSSION Knowledge of organ motion is important for determining accurate PTV margins for delivery of stereotactic body RT and to know whether strategies to reduce motion are warranted. Because liver tumor motion has been substantial in most of our patients, we investigated active breathing control to temporarily immobilize the liver with repeat breath holds during simulation imaging and RT. However, approximately one-third of patients are not suitable for this technique because of poor end-exhale breath-hold reproducibility or intolerance (20). Treatment based on using the average freebreathing position or exhale breath-hold position is another strategy that can reduce the normal tissue volume required to be treated (7). Respiratory gating and synchronization of the radiation delivery with patients’ breathing cycle are other
Table 1. Maximal liver tumor motion on fluoroscopy of diaphragm and cine-magnetic resonance imaging Table 2. Cine-MRI directional liver tumor edge motion Fluoroscopy of diaphragm (mm)
Cine-MRI of tumor (mm)
Variable
CC
CC
AP
ML
Average SD Minimum Maximum
15.2 7.3 5.0 41.0
15.5 6.7 6.9 35.4
10.1 4.7 3.7 21.6
7.5 2.7 3.8 14.8
Abbreviations: MRI = magnetic resonance imaging; CC = craniocaudal; AP = anteroposterior; ML = mediolateral.
Variable
Mean (mm)
Median (mm)
Minimum (mm)
Maximum (mm)
CC cranial CC caudal AP anterior AP posterior ML right ML left
14.9 14.1 7.6 9.3 6.1 7.2
13.2 12.2 6.6 8.2 5.5 6.2
6.7 6.2 3.1 3.2 3.4 3.8
27.5 35.4 21.6 20.6 13.1 14.8
Abbreviations as in Table 1.
3D motion of liver tumors using Cine-MRI d A. KIRILOVA et al.
11 10 9 8 7 6 mm 5 4 3 2 1 0
11 10 9 8 7 6 mm 5 4 3 2 1 0
a) Intra-observer Error
Table 3. Absolute differences between fluoroscopy and MRI craniocaudal motion
CC
AP
b) Inter-observer Error
CC
AP
ML
methods to reduce the adverse effects of breathing motion (21). Because of the diaphragm’s visibility on fluoroscopy, it is often used for evaluation of respiratory motion, and the diaphragm has been used as a surrogate for liver tumor motion assessment at simulation for liver cancer RT planning. Our AP fluoroscopy data showed an average amplitude of diaphragm motion in the CC direction of 15.2 mm 7.3 mm 45 40 r=0.25
35
MR (mm)
30 25 20 15 10 5 0
5
10
15
20
25
30
35
Tumor motion
n
%
>5 mm
>10 mm
Fluoroscopy less than cine-MRI Fluoroscopy greater than cine-MRI All
17
48.6
9 (52.9)
5 (29.4)
18
51.4
7 (38.9)
3 (16.7)
35
100
16 (45.7)
8 (22.9)
Abbreviation: MRI = magnetic resonance imaging. Data in parentheses are percentages.
ML
Fig. 4. Box plots of cine-magnetic resonance imaging directional intra- and interobserver error (n = 36). Also shown are 50% interquartile range, median, range, and outliers. (a) Intraobserver error median values for craniocaudal direction, 1.5 mm; anteroposterior, 1.0 mm; and mediolateral, 1.0 mm. (b) Interobserver error median values for craniocaudal direction, 1.8 mm; anteroposterior, 2.1 mm; and mediolateral, 1.5 mm.
0
1193
40
45
Fluoroscopy (mm)
Fig. 5. Correlation between craniocaudal motion according to diaphragm motion on fluoroscopy and tumor motion on cine-magnetic resonance imaging (n = 35).
(range, 5–41), comparable to results reported by other investigators, obtained under the same patient conditions (i.e., supine position with normal breathing) (11, 22). Diaphragm motion has been shown to correlate with the movement of fiducial markers implanted in the liver; however, the correlation was worse for tumors distal from the diaphragm, and the discrepancies ranged from 0.9 to 3.9 mm in the CC direction (2). Inserted fiducial markers have also been used for fluoroscopic real-time liver tumor tracking during RT, which can reduce the liver volume irradiated (23). A disadvantage of fiducial-based techniques is the invasiveness of the marker insertion procedure and their possible migration during the RT course (13). In addition, patients with unresectable liver tumors often have other co-morbidity factors that could make them more prone to bleeding and infection. In the present study, we have demonstrated the feasibility of using cine-MRI to measure liver tumor motion for RT planning. Our results have shown that AP fluoroscopy evaluation alone is not representative of liver tumor motion as assessed using cine-MRI. Fluoroscopy underestimated the actual tumor motion in approximately one-half of the cases. The possible contributing factors to this difference, during the acquisition of the cine-MRI scans include claustrophobia in the MRI unit, leading to increased motion; sampling only every second, leading to underestimation of motion; and the potential for rotations, deformations, and out of plane changes, leading to overestimation of the motion assessed at MRI with the single shot fast spin echo imaging technique we used. The temporal resolution (1 slice/s) of cine-MRI was the best available to us at the data acquisition of this study, and it was a limitation. MRI techniques are constantly evolving, but the trade-off among resolution, acquisition speed, and signal/noise ratio still prevents a detailed full-volumetric examination of respiratory motion in real time. In this study, to achieve high-temporal resolution, the imaging parameters of the single shot fast spin echo sequence were optimized by reducing the phase-encoding steps to 192, resulting in lower in-plane resolution (range, 1.6–2.5 mm), a trade-off we found acceptable because the observed liver tumor motion was much larger (mean CC, 15.5 mm; mean AP, 10.5 mm; and mean ML, 7.9 mm). Despite this sampling error, approximately 50% of the time, MRI revealed tumor motion that was greater than the diaphragm motion at fluoroscopy. The limitations of AP fluoroscopy to measure diaphragm motion include that the angle of the beam is not perpendicular
1194
I. J. Radiation Oncology d Biology d Physics
to the dome of the diaphragm and rotations of the liver cannot be measured, leading to possible underestimation of the CC motion. Also, the CC motion of the apex of the liver, which can rotate, is measured (i.e., fluoroscopy and MRI measured the motion of different structures). Nonetheless, the magnitude of differences in liver tumor motion on cine-MRI compared with the diaphragm motion on fluoroscopy was larger than we expected, demonstrating that the motion varies depending on how it is measured and/or is variable within 1 or 2 h (the interval between the kV fluoroscopy and MRI motion assessments). This suggests the need for additional studies to investigate the extent of the variability in motion from day to day. Because of the presented results, we have initiated studies to investigate the reproducibility of organ motion with serial scanning sessions using the same imaging techniques during a course of RT. A limitation of the cine-MRI motion assessment used in the present study was that three consecutive two-dimensional imaging series in each orthogonal plane through the centroid of the tumor were evaluated, rather than the three-dimensional motion of the entire tumor volume. This is especially relevant for tumors with an irregular shape and larger size, because a single slice might not reflect the motion of the entire tumor and in- and out-of-plane changes can occur from rotations and deformations, which could result in differences in the motion profile of the cranial and caudal tumor edge (Fig. 2). The CC liver tumor motion was not significantly different whether it was assessed in the sagittal or coronal plane. Greater differences were observed when AP motion was compared between the axial and sagittal planes, likely related to the rotations and deformations, which were observed mostly in the AP plane. The CC motion of the diaphragm on AP fluoroscopy represents a two-dimensional projection of full liver translations, rotations, and deformations resulting from breathing. In contrast, the cine-MRI shows motion through the centroid of the tumor. The comparison of AP kV fluoroscopy to cine-MRI tumor was not the primary goal of this study. However it was of interest, because it emphasized that CC motion on fluoroscopy might not represent the CC motion of the tumor using other imaging types. Whether this resulted from the reproducibility of breathing or the different imaging modalities is not known. The diaphragm motion at these cine-MRI slices did not represent the full diaphragm or maximal diaphragm motion as assessed using kV fluoroscopy; hence, the diaphragm motion in the planes investigated was not measured on cineMRI. In our current practice, at simulation for liver cancer RT planning, we continue to use both AP kV fluoroscopy to assess diaphragm motion and cine-MRI to assess liver tumor motion, because the two imaging techniques appear complimentary. We also evaluate liver motion at radiation delivery to confirm that it is no larger than that predicted at simulation. To our knowledge, only a few reports with a limited number of cases have described the use of dynamic MRI for the evaluation of liver tumor motion. Shimizu et al. (19) reported that 3D movement of a spherical liver tumor in 1 patient was detectable using high-speed MRI and had the potential to improve the accuracy of the PTV determination. The same
Volume 71, Number 4, 2008
group imaged three moving liver tumors with cine-MRI and found a CC, AP, and ML motion of 10.6 7 mm, 4.6 1.6 mm, and 5.2 1.8 mm, respectively. The investigators concluded that MRI can be used to establish the PTV of moving liver tumors more precisely than can CT (24). Korin et al. (25) used cine-MRI in 15 volunteers and measured an average CC displacement of the diaphragm and liver during normal breathing of 15 mm, which corresponds with our results. In our analysis, although the intra- and interobserver error was measurable, it was far less than the differences observed between cine-MRI and fluoroscopy. Time-resolved 4D-CT data can also be used to assess liver motion due to breathing; however, visualization of the tumor on CT is challenging unless fudicial markers are implanted or contrast media is administered (16). In contrast to 4D-CT, cine-MRI provides excellent soft tissue contrast, direct visualization of the tumor motion without fiducial markers or contrast enhancement and allows greater flexibility than CT in selecting image orientation. More recent fast imaging employing steady-state acquisition (FIESTA) sequence in combination with parallel imaging technique can deliver increased temporal (three slices each second) and contrast resolution compared to the single shot fast spin echo sequence used in this study. The strength of the single shot fast spin echo sequence, used for this analysis, is that the spatial distortion is less than with gradient echo methods such as FIESTA. Future improvements in MRI gradient speed and image reconstruction have the potential to capture the behavior of the entire liver during free breathing in time-resolved volumetric MRI images (4D sequences), allowing the volumetric tumor motion to be more accurately quantified (26). The motion analysis in this study was manual, because automatic tracking tools were not available when this project was conducted. Automated methods to track liver motion on multiple slices are now available to facilitate motion assessment, saving time and facilitating analysis of large volumes of data, and, possibly, reducing observer error (27). Manual measurements of motion measured using cine-MRI have been found to correlate highly with automated methods to measure motion; therefore, we believe that the manual measurements used in the present study did not introduce a substantial error in the data analysis (27). Cine-MRI can also be used to evaluate changes in liver tumor motion patterns over time and/or with different immobilization strategies such as abdominal compression or gating. CONCLUSIONS Cine-MRI is a feasible, noninvasive imaging technique that can be used to measure liver tumor motion with high precision. These data can be considered in the creation of individualized PTV margins to account for breathing motion. In most patients, the tumor motion assessed on cine-MRI did not appear to correlate well with the diaphragm motion measured using kV fluoroscopy, providing a rationale for additional studies to evaluate the reproducibility of motion due to breathing.
3D motion of liver tumors using Cine-MRI d A. KIRILOVA et al.
1195
REFERENCES 1. Davies SC, Hill AL, Holmes RB, et al. Ultrasound quantitation of respiratory organ motion in the upper abdomen. Br J Radiol 1994;67:1096–1102. 2. Balter JM, Dawson LA, Kazanjian S, et al. Determination of ventilatory liver movement via radiographic evaluation of diaphragm position. Int J Radiat Oncol Biol Phys 2001;51: 267–270. 3. Suramo I, Paivansalo M, Myllyla V. Cranio-caudal movements of the liver, pancreas and kidneys in respiration. Acta Radiol Diagn 1984;25:129–131. 4. Chen GT, Kung JH, Beaudette KP. Artifacts in computed tomography scanning of moving objects. Semin Radiat Oncol 2004;14:19–26. 5. Balter JM, Ten Haken RK, Lawrence TS, et al. Uncertainties in CT-based radiation therapy treatment planning associated with patient breathing. Int J Radiat Oncol Biol Phys 1996;36: 167–174. 6. Aruga T, Itami J, Aruga M, et al. Target volume definition for upper abdominal irradiation using CT scans obtained during inhale and exhale phases. Int J Radiat Oncol Biol Phys 2000;48: 465–469. 7. Balter JM, Lam KL, McGinn CJ, et al. Improvement of CTbased treatment-planning models of abdominal targets using static exhale imaging. Int J Radiat Oncol Biol Phys 1998;41: 939–943. 8. Ten Haken RK, Balter JM, Marsh LH, et al. Potential benefits of eliminating planning target volume expansions for patient breathing in the treatment of liver tumors. Int J Radiat Oncol Biol Phys 1997;38:613–617. 9. 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. 10. Shirato H, Seppenwoolde Y, Kitamura K, et al. Intrafractional tumor motion: lung and liver. Semin Radiat Oncol 2004;14: 10–18. 11. Weiss PH, Baker JM, Potchen EJ. Assessment of hepatic respiratory excursion. J Nucl Med 1972;13:758–759. 12. Harauz G, Bronskill MJ. Comparison of the liver’s respiratory motion in the supine and upright positions: Concise communication. J Nucl Med 1979;20:733–735. 13. Shirato H, Harada T, Harabayashi T, et al. Feasibility of insertion/implantation of 2.0-mm-diameter gold internal fiducial markers for precise setup and real-time tumor tracking in radiotherapy. Int J Radiat Oncol Biol Phys 2003;56:240–247. 14. Brock KM, Balter JM, Dawson LA, et al. Automated generation of a four-dimensional model of the liver using warping and mutual information. Med Phys 2003;30:1128–1133.
15. Rietzel E, Chen GT, Choi NC, et al. Four-dimensional imagebased treatment planning: Target volume segmentation and dose calculation in the presence of respiratory motion. Int J Radiat Oncol Biol Phys 2005;61:1535–1550. 16. Beddar AS, Kainz K, Briere TM, et al. Correlation between internal fiducial tumor motion and external marker motion for liver tumors imaged with 4D-CT. Int J Radiat Oncol Biol Phys 2007;67:630–638. 17. Hussain SM, Semelka RC. Hepatic imaging: comparison of modalities. Radiol Clin North Am 2005;43:929–947. 18. Voroney JP, Brock KK, Eccles C, et al. Prospective comparison of computed tomography and magnetic resonance imaging for liver cancer delineation using deformable image registration. Int J Radiat Oncol Biol Phys 2006;66:780–791. 19. Shimizu S, Shirato H, Xo B, et al. Three-dimensional movement of a liver tumor detected by high-speed magnetic resonance imaging. Radiother Oncol 1999;50:367–370. 20. 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. 21. Wagman R, Yorke E, Ford E, et al. Respiratory gating for liver tumors: use in dose escalation. Int J Radiat Oncol Biol Phys 2003;55:659–668. 22. Hanley J, Debois MM, Mah D, et al. Deep inspiration breathhold technique for lung tumors: The potential value of target immobilization and reduced lung density in dose escalation. Int J Radiat Oncol Biol Phys 1999;45:603–611. 23. Shirato H, Shimizu S, Kitamura K, et al. Four-dimensional treatment planning and fluoroscopic real-time tumor tracking radiotherapy for moving tumor. Int J Radiat Oncol Biol Phys 2000;48:435–442. 24. 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. 25. Korin HW, Ehman RL, Riederer SJ, et al. Respiratory kinematics of the upper abdominal organs: a quantitative study. Magn Reson Med 1992;23:172–178. 26. von Siebenthal M, Szekely G, Gamper U, et al. 4D MR imaging of respiratory organ motion and its variability. Phys Med Biol 2007;52:1547–1564. 27. Prakash V, Stainsby VA, Craig T, et al. Validation of a supervised automated algorithm for fast quantitative evaluation of organ motion on MRI. Int J Radiat Oncol Biol Phys. In press.
doi:10.1016/j.ijrobp.2007.11.026
CLINICAL INVESTIGATION
Liver
THREE-DIMENSIONAL MOTION OF LIVER TUMORS USING CINE-MAGNETIC RESONANCE IMAGING ANNA KIRILOVA, B.SC.,* GINA LOCKWOOD, M.MATH.,y PERRY CHOI, M.D.,z NEELUFER BANA, B.SC. (HON),z MASOOM A. HAIDER, M.D.,x KRISTY K. BROCK, PH.D.,* CYNTHIA ECCLES, B.SC.,z z AND LAURA A. DAWSON, M.D. Departments of *Radiation Physics, y Biostatistics, z Radiation Oncology, and x Medical Imaging, Princess Margaret Hospital, University of Toronto, Toronto, ON, Canada Purpose: To measure the three-dimensional motion of liver tumors using cine-magnetic resonance imaging (MRI) and compare it to the liver motion assessed using fluoroscopy. Methods and Materials: Liver and liver tumor motion were investigated in the first 36 patients with primary (n = 20) and metastatic (n = 16) liver cancer accrued to our Phase I stereotactic radiotherapy study. At simulation, all patients underwent anteroposterior fluoroscopy, and the maximal diaphragm excursion in the craniocaudal (CC) direction was observed. Cine-MRI using T2-weighted single shot fast spin echo sequences were acquired in three orthogonal planes during free breathing through the centroid of the most dominant liver tumor. ImageJ software was used to measure the maximal motion of the tumor edges in each plane. The intra- and interobserver reproducibility was also quantified. Results: The average CC motion of the liver at fluoroscopy was 15 mm (range, 5–41). On cine-MRI, the average CC tumor motion was 15.5 mm (range, 6.9–35.4), the anteroposterior motion was 10 mm (range, 3.7–21.6), and the mediolateral motion was 7.5 mm (range, 3.8–14.8). The fluoroscopic CC diaphragm motion did not correlate well with the MRI CC tumor motion (r = 0.25). The mean intraobserver error was <2 mm in the CC, anteroposterior, and mediolateral directions, and 90% of measurements between observers were within 3 mm. Conclusions: The results of our study have shown that cine-MRI can be used to directly assess liver tumor motion in three dimensions. Tumor motion did not correlate well with the diaphragm motion measured using kilovoltage fluoroscopy. The tumor motion data from cine-MRI can be used to facilitate individualized planning target volume margins to account for breathing motion. Ó 2008 Elsevier Inc. Cine-MRI, Hepatobiliary cancer, Liver metastases, Organ motion.
Detailed information about liver and liver tumor motion is crucial for establishing the smallest and safest margin for irradiation without compromising the therapeutic dose necessary for optimal tumor coverage. Liver motion is mostly related to breathing and is largest in the craniocaudal (CC) direction, with a range of 5–50 mm (1–3). Its adverse effect on radiotherapy (RT) planning and treatment include the introduction of artifacts on the planning computed tomography (CT) scans, leading to inaccurate tumor and normal tissue volumes (4–6), altered dosimetry using a static plan (7), an increased volume of normal tissue exposure (8), increased
required planning target volume (PTV) margins (9), and a greater risk of toxicity. Before liver tumor motion can be considered in planning and appropriate methods of its management applied, the magnitude of tumor motion during free breathing must be measured. Diaphragm CC motion measured during kilovoltage (kV) fluoroscopy can be used as a surrogate for liver tumor motion. However, liver motion might not necessarily correlate with the CC diaphragm motion, because it is more complex, consisting of translations, rotations, and hysteresis (10). Earlier studies have investigated liver motion with different imaging modalities under various breathing conditions, including a scintillation camera after administration
Reprint requests to: Anna Kirilova, B.Sc., Department of Radiation Physics, Princess Margaret Hospital, University of Toronto, 610 University Ave., Toronto, ON M5G 2M9 Canada. Tel: (416) 946-7805; Fax: (416) 946-6566; E-mail: anna.kirilova@rmp. uhn.on.ca Supported in part by a grant from Varian Medical Systems and by the generosity of the Susan Grange family. Patients treated in this study were treated on research protocols funded in part by Elekta Oncology Systems, National Cancer
Institute of Canada Grant 018207, and the Canadian Cancer Society. L. Dawson was a recipient of an American Society for Clinical Oncology career development award. Presented in part at the European Society of Radiation Oncology (ESTRO) 2004 Annual Scientific Meeting, Amsterdam, October 2004. Conflict of interest: none. Received Aug 8, 2007, and in revised form Nov 16, 2007. Accepted for publication Nov 16, 2007.
INTRODUCTION
1189
1190
I. J. Radiation Oncology d Biology d Physics
of 99Tc (11, 12), ultrasonography (1, 3), fluoroscopy, and computed tomography (5, 7). Fiducial markers placed in, or close to, the tumor have been used by other investigators to better estimate the motion in relation to the diaphragm using fluoroscopy and CT. The reported limitations with this method include possible migration of the markers during the RT course and the lack of three-dimensional motion information (13). Repeat CT scans acquired during exhale and inhale breath hold have been used to estimate the extremes of liver position resulting from breathing motion (14). More recently, respiratory sorted ‘‘four-dimensional’’ (4D) CT has been used to quantify lung and liver motion during the period of image acquisition (15, 16). However, liver tumors are not well visualized without the use of intravenous contrast, which is generally not used during respiratory-sorted CT scans (17). At present, no consensus has been reached regarding the best imaging strategies for organ motion and its management for RT planning and delivery. However, regardless of the intended strategy to consider and/or reduce the adverse effects of motion for RT planning and treatment, the breathing motion needs to be estimated. Because liver tumor motion varies substantially among patients, there is motivation to develop strategies that allow repeated, low-risk measurements of patients’ liver tumor motion due to breathing. Magnetic resonance imaging (MRI) can be used for this purpose. It has the advantage of improved soft tissue visualization compared with CT (17, 18). Also, the recently available high-speed MRI sequences (cine-MRI) permit direct visualization of the respiratory motion of the liver tumor itself, without the concern of the dose exposure to the patient. Shimizu et al. (19) first reported the feasibility of high-speed MRI to assess liver tumor motion in 1999. The primary goal of this study was to describe the three-dimensional (3D) motion of liver tumors using cine-MRI in the first 36 patients undergoing treatment planning for stereotactic body RT on our Phase I-II protocol. The secondary goals were to compare the 3D tumor motion with the diaphragm motion assessed at anteroposterior (AP) fluoroscopy and to measure the inter- and intraobserver variability in motion assessment using cine-MRI. METHODS AND MATERIALS Patients The patients included in this study had provided written informed consent for an institutional research ethics board–approved Phase III protocol of stereotactic body RT for liver metastases or hepatobiliary carcinoma. The eligibility criteria included no contraindications for MRI, unresectable liver cancer, a Child-Pugh A liver function score, and a life expectancy >12 weeks. Patients whose disease burden was not in the liver and those with a Karnofsky performance status of <60 were excluded. All patients accrued to the clinical study underwent imaging with fluoroscopy and MRI, in addition to a triphasic planning CT scan at simulation.
Fluoroscopy The patients were placed in the supine position either on a chest board (MedTech, Orange City, IA) or an evacuated immobilization
Volume 71, Number 4, 2008
bag (Vac Lok, Bionix, Toledo, OH), with their arms above their head and a leg immobilizer under their knees. AP fluoroscopy imaging was done on a conventional simulator (Nucletron, Veenendaal, The Netherlands) to measure the CC range of the diaphragm during free breathing for 60 s at a 100-cm source-to-axis distance set to the approximate mid-plane of the patient at the level of the mid-right costal margin. The fluoroscopic images were recorded on VHS or DVD for off-line analysis of each fluoroscopic session. The maximal CC excursion of the dome of the diaphragm was measured. The simulator’s graticule was calibrated daily as a part of the simulator quality assurance procedure. At the analysis, the magnification of the graticule was determined using a standard metal ruler to measure the distance between the indexes on the gradicule as seen on the video screen, and a magnification factor determined by dividing the measured value over the true value. This factor was then applied to all measurements.
Cine-MRI On the same day of fluoroscopy, cine-MRI was done to measure the three-dimensional tumor motion during free breathing. The patients were placed in the same position as for fluoroscopy, either on a flat table top with an evacuated immobilization bag (Vac Lok) that fit in the magnet bore or only on a flat table top when the patient was too large to fit in the magnet bore with the immobilization bag. Single 5-mm slices through the centroid of the dominant liver tumor were acquired manually in axial, sagittal, and coronal planes on a 1.5T General Electric Signa Twin Speed (Milwaukee, WI) scanner, using the General Electric four-channel torso coil and a T2-weighted single shot fast spin echo sequence with the following parameters (echo time, 90 ms; repetition time, 1,300 ms; slice thickness, 5 mm; field of view, 32–48 cm; matrix, 256 192). The in-plane resolution was 1.6–2.5 mm, and the temporal resolution was one slice/s during a 60-s period for each orthogonal plane. If the frequency of imaging appeared to be somewhat in phase with the respiratory frequency, a short pause (e.g., 0.5 s) was introduced to provide a noticeable phase difference to ensure that the maximal excursion of motion was more likely to be measured.
Statistical analysis The maximal diaphragm excursion in the CC direction as measured on AP kV fluoroscopy was recorded prospectively at simulation by a radiation therapist and verified off-line by a radiation oncologist (Fig. 1). All cine-MRI series were exported to a Digital Imaging and Communications in Medicine server and evaluated using ImageJ software, version 1.31 (National Institutes of Health, Bethesda, MD). ImageJ is a public domain Java image-processing program that can be used to track a point or object through a series of images. In our study, the position of the tumor edge in each image was determined, so that for each imaging series in one plane, the maximal motion of the tumor edges could be evaluated. The 12 measurements obtained for each patient were performed twice by two observers in 36 patients, for a total of 1,728 measurements (Fig. 2). The four observer measurements were averaged for each tumor edge in each plane, and the maximal motion was calculated for the superior, inferior, anterior, posterior, medial, and lateral edges of the tumor and for the CC, AP and mediolateral (ML) directions overall. The 3D coordinates for every measurement was saved in Excel and checked by a third observer. Box plots were used to display all motion data. The maximal motion of the superior and inferior, anterior and posterior, and medial and lateral edges of the tumor and the CC motion in the coronal and sagittal planes, AP motion in the sagittal and
3D motion of liver tumors using Cine-MRI d A. KIRILOVA et al.
1191
Fig. 1. (a) Maximal inhale diaphragm position, and (b) maximal exhale diaphragm position, 20 mm in craniocaudal direction, measured with anteroposterior fluoroscopy for 30 s of normal breathing. The interobserver error was determined by averaging the absolute differences between the mean measurements of the two observers for each plane and summarizing for all patients using the mean values and ranges. Box plots are used to display the reproducibility data.
axial planes, and ML motion in the coronal and axial planes were compared using Mann-Whitney U tests. The motion between patients who did and did not have the immobilization device at the time of MRI were also compared using Mann-Whitney U tests. The fluoroscopic CC motion was correlated with the cine-MRI CC tumor motion using a Pearson correlation coefficient. Chisquare tests were used to investigate the relationship between selected disease characteristics and the motion in all directions, which were dichotomized into #10 and >10 mm. The factors tested were diagnosis, tumor size, and tumor location.
RESULTS All image data were acquired between July 2003 and October 2004. In this study, 36 patients with metastasis (n = 16), hepatocellular carcinoma (n = 14), and intrahepatic cholangiocarcinoma (n = 6) were imaged with fluoroscopy, cineMRI, and tri-phasic CT for the purpose of RT planning. These patients were subsequently treated with stereotactic body RT. On cine-MRI, the average maximal motion of the 36 liver tumors assessed was 15.5 mm (range, 6.9–35.4 mm) in the CC, 10.1 mm (range, 3.7–21.6 mm) in the AP, and 7.5 mm
Intra- and interobserver reproducibility The intraobserver error was quantified for each observer. The absolute differences in the repeat measurements of the maximal excursions were calculated for each tumor edge in each plane. The error for each plane was calculated for each observer separately and for the observers combined by averaging the relevant differences. The overall intraobserver variability was calculated similarly using the eight relevant measures.
a)
20.5 20
mm
19.5 19
18.5 1
11
21 Sec
31
41
51
b) 30.5
mm
30 29.5 29 1
11
21 Sec
31
41
51
c) Fig. 2. (a) Cranial (arrow) and caudal (arrowhead) edges of tumor on cine-magnetic resonance imaging, with graphs showing change in (b) cranial and (c) caudal tumor edge position from breathing, based on 50 s of cine-magnetic resonance imaging.
(range, 3.8–14.8 mm) in the ML directions (Table 1 and Fig. 3). The tumor motion observed was predominantly in the CC and AP directions and the least in the ML direction, as expected. No significant differences were found in the maximal motion between the superior and inferior edges of the tumor, anterior and posterior edges, or medial and lateral edges. Table 2 lists the motion separated by the edge evaluated. No differences were found when the motion was compared between planes, except for the AP motion, which was significantly greater in the axial plane (mean, 9.5 mm; median, 8.3 mm) compared with the the motion in the sagittal plane (mean, 7.7 mm; median, 6.1 mm). We also investigated the tumor motion in all directions (#10 mm vs. >10 mm) according to potential explanatory factors such as diagnosis, tumor size, and tumor location. None of these factors were significantly associated with the amplitude of motion. Of the 36 patients, 15 underwent simulation and treatment on a chest board or flat tabletop and 21 underwent simulation and treatment in the immobilization device (Vac Lok); however, only 15 patients underwent MRI with the Vac Lok. No significant differences were found in the diaphragm/liver tumor motion according to the use of the immobilization device at MRI. The mean intraobserver error in MRI tumor motion measurements for Observer 1 was 1.2 mm (range, 0.4–2.1 mm) in the CC direction, 0.9 mm (range, 0.3–2.7 mm) in the AP direction, and 0.7 mm (range, 0.3–1.4 mm) in the ML direction. For Observer 2, the error was 1.7 mm (range, 0.4–2.9 mm) in the CC direction, 1.5 mm (range, 0–4.3 mm) in the AP direction, and 1.7 mm (range, 0–4.0 mm) in the ML direction. The overall intraobserver error (mean of the two observers) was 1.4 mm (range, 0.4–2.3 mm), 1.2 mm (range, 0.4–3.1 mm), and 1.1 mm (range, 0.3–2.5) in the CC, AP, and ML direction, respectively (Fig. 4a). These were one-eleventh (CC), one-eighth (AP), and one-seventh (ML) the size of the calculated mean MRI motion. The overall interobserver error was 2.3 mm (range, 0.3– 10.5 mm) in the CC direction, 2.5 mm (range, 0.4–7.3 mm) in the AP direction, and 1.9 mm (range, 0.2–4.4) in the ML direction (Fig. 4b). These errors were one-seventh (CC) and onequarter (AP and ML) the size of the measured MRI motion. Anteroposterior fluoroscopy (n = 35) revealed that the average diaphragm motion in the CC direction was 15 7 mm (range, 5–41 mm). The fluoroscopic CC motion did not correlate well with the cine-MRI CC tumor motion (r = 0.25;
Volume 71, Number 4, 2008 40
Maximum motion, mm
I. J. Radiation Oncology d Biology d Physics
1192
30
20
10
0
AP
CC
ML
Fig. 3. Box plot of cine-magnetic resonance imaging directional maximal motion. Also shown are 50% interquartile range, median, range, and outliers. Median motion in each direction: craniocaudal, 13.3 mm, anteroposterior, 9.2 mm, and mediolateral, 6.9 mm.
Fig. 5). The mean absolute value of the difference between the fluoroscopic diaphragm CC motion and the cine-MRI CC tumor motion (n = 35) was 6.5 mm (range, 0.1– 0.6 mm). In 16 cases, the difference was >5 mm (45.7%) and in 8 cases was >10 mm (22.9%). Fluoroscopy overestimated the tumor motion in 51.4% of cases and underestimated it in 48.6% of cases (Table 3). DISCUSSION Knowledge of organ motion is important for determining accurate PTV margins for delivery of stereotactic body RT and to know whether strategies to reduce motion are warranted. Because liver tumor motion has been substantial in most of our patients, we investigated active breathing control to temporarily immobilize the liver with repeat breath holds during simulation imaging and RT. However, approximately one-third of patients are not suitable for this technique because of poor end-exhale breath-hold reproducibility or intolerance (20). Treatment based on using the average freebreathing position or exhale breath-hold position is another strategy that can reduce the normal tissue volume required to be treated (7). Respiratory gating and synchronization of the radiation delivery with patients’ breathing cycle are other
Table 1. Maximal liver tumor motion on fluoroscopy of diaphragm and cine-magnetic resonance imaging Table 2. Cine-MRI directional liver tumor edge motion Fluoroscopy of diaphragm (mm)
Cine-MRI of tumor (mm)
Variable
CC
CC
AP
ML
Average SD Minimum Maximum
15.2 7.3 5.0 41.0
15.5 6.7 6.9 35.4
10.1 4.7 3.7 21.6
7.5 2.7 3.8 14.8
Abbreviations: MRI = magnetic resonance imaging; CC = craniocaudal; AP = anteroposterior; ML = mediolateral.
Variable
Mean (mm)
Median (mm)
Minimum (mm)
Maximum (mm)
CC cranial CC caudal AP anterior AP posterior ML right ML left
14.9 14.1 7.6 9.3 6.1 7.2
13.2 12.2 6.6 8.2 5.5 6.2
6.7 6.2 3.1 3.2 3.4 3.8
27.5 35.4 21.6 20.6 13.1 14.8
Abbreviations as in Table 1.
3D motion of liver tumors using Cine-MRI d A. KIRILOVA et al.
11 10 9 8 7 6 mm 5 4 3 2 1 0
11 10 9 8 7 6 mm 5 4 3 2 1 0
a) Intra-observer Error
Table 3. Absolute differences between fluoroscopy and MRI craniocaudal motion
CC
AP
b) Inter-observer Error
CC
AP
ML
methods to reduce the adverse effects of breathing motion (21). Because of the diaphragm’s visibility on fluoroscopy, it is often used for evaluation of respiratory motion, and the diaphragm has been used as a surrogate for liver tumor motion assessment at simulation for liver cancer RT planning. Our AP fluoroscopy data showed an average amplitude of diaphragm motion in the CC direction of 15.2 mm 7.3 mm 45 40 r=0.25
35
MR (mm)
30 25 20 15 10 5 0
5
10
15
20
25
30
35
Tumor motion
n
%
>5 mm
>10 mm
Fluoroscopy less than cine-MRI Fluoroscopy greater than cine-MRI All
17
48.6
9 (52.9)
5 (29.4)
18
51.4
7 (38.9)
3 (16.7)
35
100
16 (45.7)
8 (22.9)
Abbreviation: MRI = magnetic resonance imaging. Data in parentheses are percentages.
ML
Fig. 4. Box plots of cine-magnetic resonance imaging directional intra- and interobserver error (n = 36). Also shown are 50% interquartile range, median, range, and outliers. (a) Intraobserver error median values for craniocaudal direction, 1.5 mm; anteroposterior, 1.0 mm; and mediolateral, 1.0 mm. (b) Interobserver error median values for craniocaudal direction, 1.8 mm; anteroposterior, 2.1 mm; and mediolateral, 1.5 mm.
0
1193
40
45
Fluoroscopy (mm)
Fig. 5. Correlation between craniocaudal motion according to diaphragm motion on fluoroscopy and tumor motion on cine-magnetic resonance imaging (n = 35).
(range, 5–41), comparable to results reported by other investigators, obtained under the same patient conditions (i.e., supine position with normal breathing) (11, 22). Diaphragm motion has been shown to correlate with the movement of fiducial markers implanted in the liver; however, the correlation was worse for tumors distal from the diaphragm, and the discrepancies ranged from 0.9 to 3.9 mm in the CC direction (2). Inserted fiducial markers have also been used for fluoroscopic real-time liver tumor tracking during RT, which can reduce the liver volume irradiated (23). A disadvantage of fiducial-based techniques is the invasiveness of the marker insertion procedure and their possible migration during the RT course (13). In addition, patients with unresectable liver tumors often have other co-morbidity factors that could make them more prone to bleeding and infection. In the present study, we have demonstrated the feasibility of using cine-MRI to measure liver tumor motion for RT planning. Our results have shown that AP fluoroscopy evaluation alone is not representative of liver tumor motion as assessed using cine-MRI. Fluoroscopy underestimated the actual tumor motion in approximately one-half of the cases. The possible contributing factors to this difference, during the acquisition of the cine-MRI scans include claustrophobia in the MRI unit, leading to increased motion; sampling only every second, leading to underestimation of motion; and the potential for rotations, deformations, and out of plane changes, leading to overestimation of the motion assessed at MRI with the single shot fast spin echo imaging technique we used. The temporal resolution (1 slice/s) of cine-MRI was the best available to us at the data acquisition of this study, and it was a limitation. MRI techniques are constantly evolving, but the trade-off among resolution, acquisition speed, and signal/noise ratio still prevents a detailed full-volumetric examination of respiratory motion in real time. In this study, to achieve high-temporal resolution, the imaging parameters of the single shot fast spin echo sequence were optimized by reducing the phase-encoding steps to 192, resulting in lower in-plane resolution (range, 1.6–2.5 mm), a trade-off we found acceptable because the observed liver tumor motion was much larger (mean CC, 15.5 mm; mean AP, 10.5 mm; and mean ML, 7.9 mm). Despite this sampling error, approximately 50% of the time, MRI revealed tumor motion that was greater than the diaphragm motion at fluoroscopy. The limitations of AP fluoroscopy to measure diaphragm motion include that the angle of the beam is not perpendicular
1194
I. J. Radiation Oncology d Biology d Physics
to the dome of the diaphragm and rotations of the liver cannot be measured, leading to possible underestimation of the CC motion. Also, the CC motion of the apex of the liver, which can rotate, is measured (i.e., fluoroscopy and MRI measured the motion of different structures). Nonetheless, the magnitude of differences in liver tumor motion on cine-MRI compared with the diaphragm motion on fluoroscopy was larger than we expected, demonstrating that the motion varies depending on how it is measured and/or is variable within 1 or 2 h (the interval between the kV fluoroscopy and MRI motion assessments). This suggests the need for additional studies to investigate the extent of the variability in motion from day to day. Because of the presented results, we have initiated studies to investigate the reproducibility of organ motion with serial scanning sessions using the same imaging techniques during a course of RT. A limitation of the cine-MRI motion assessment used in the present study was that three consecutive two-dimensional imaging series in each orthogonal plane through the centroid of the tumor were evaluated, rather than the three-dimensional motion of the entire tumor volume. This is especially relevant for tumors with an irregular shape and larger size, because a single slice might not reflect the motion of the entire tumor and in- and out-of-plane changes can occur from rotations and deformations, which could result in differences in the motion profile of the cranial and caudal tumor edge (Fig. 2). The CC liver tumor motion was not significantly different whether it was assessed in the sagittal or coronal plane. Greater differences were observed when AP motion was compared between the axial and sagittal planes, likely related to the rotations and deformations, which were observed mostly in the AP plane. The CC motion of the diaphragm on AP fluoroscopy represents a two-dimensional projection of full liver translations, rotations, and deformations resulting from breathing. In contrast, the cine-MRI shows motion through the centroid of the tumor. The comparison of AP kV fluoroscopy to cine-MRI tumor was not the primary goal of this study. However it was of interest, because it emphasized that CC motion on fluoroscopy might not represent the CC motion of the tumor using other imaging types. Whether this resulted from the reproducibility of breathing or the different imaging modalities is not known. The diaphragm motion at these cine-MRI slices did not represent the full diaphragm or maximal diaphragm motion as assessed using kV fluoroscopy; hence, the diaphragm motion in the planes investigated was not measured on cineMRI. In our current practice, at simulation for liver cancer RT planning, we continue to use both AP kV fluoroscopy to assess diaphragm motion and cine-MRI to assess liver tumor motion, because the two imaging techniques appear complimentary. We also evaluate liver motion at radiation delivery to confirm that it is no larger than that predicted at simulation. To our knowledge, only a few reports with a limited number of cases have described the use of dynamic MRI for the evaluation of liver tumor motion. Shimizu et al. (19) reported that 3D movement of a spherical liver tumor in 1 patient was detectable using high-speed MRI and had the potential to improve the accuracy of the PTV determination. The same
Volume 71, Number 4, 2008
group imaged three moving liver tumors with cine-MRI and found a CC, AP, and ML motion of 10.6 7 mm, 4.6 1.6 mm, and 5.2 1.8 mm, respectively. The investigators concluded that MRI can be used to establish the PTV of moving liver tumors more precisely than can CT (24). Korin et al. (25) used cine-MRI in 15 volunteers and measured an average CC displacement of the diaphragm and liver during normal breathing of 15 mm, which corresponds with our results. In our analysis, although the intra- and interobserver error was measurable, it was far less than the differences observed between cine-MRI and fluoroscopy. Time-resolved 4D-CT data can also be used to assess liver motion due to breathing; however, visualization of the tumor on CT is challenging unless fudicial markers are implanted or contrast media is administered (16). In contrast to 4D-CT, cine-MRI provides excellent soft tissue contrast, direct visualization of the tumor motion without fiducial markers or contrast enhancement and allows greater flexibility than CT in selecting image orientation. More recent fast imaging employing steady-state acquisition (FIESTA) sequence in combination with parallel imaging technique can deliver increased temporal (three slices each second) and contrast resolution compared to the single shot fast spin echo sequence used in this study. The strength of the single shot fast spin echo sequence, used for this analysis, is that the spatial distortion is less than with gradient echo methods such as FIESTA. Future improvements in MRI gradient speed and image reconstruction have the potential to capture the behavior of the entire liver during free breathing in time-resolved volumetric MRI images (4D sequences), allowing the volumetric tumor motion to be more accurately quantified (26). The motion analysis in this study was manual, because automatic tracking tools were not available when this project was conducted. Automated methods to track liver motion on multiple slices are now available to facilitate motion assessment, saving time and facilitating analysis of large volumes of data, and, possibly, reducing observer error (27). Manual measurements of motion measured using cine-MRI have been found to correlate highly with automated methods to measure motion; therefore, we believe that the manual measurements used in the present study did not introduce a substantial error in the data analysis (27). Cine-MRI can also be used to evaluate changes in liver tumor motion patterns over time and/or with different immobilization strategies such as abdominal compression or gating. CONCLUSIONS Cine-MRI is a feasible, noninvasive imaging technique that can be used to measure liver tumor motion with high precision. These data can be considered in the creation of individualized PTV margins to account for breathing motion. In most patients, the tumor motion assessed on cine-MRI did not appear to correlate well with the diaphragm motion measured using kV fluoroscopy, providing a rationale for additional studies to evaluate the reproducibility of motion due to breathing.
3D motion of liver tumors using Cine-MRI d A. KIRILOVA et al.
1195
REFERENCES 1. Davies SC, Hill AL, Holmes RB, et al. Ultrasound quantitation of respiratory organ motion in the upper abdomen. Br J Radiol 1994;67:1096–1102. 2. Balter JM, Dawson LA, Kazanjian S, et al. Determination of ventilatory liver movement via radiographic evaluation of diaphragm position. Int J Radiat Oncol Biol Phys 2001;51: 267–270. 3. Suramo I, Paivansalo M, Myllyla V. Cranio-caudal movements of the liver, pancreas and kidneys in respiration. Acta Radiol Diagn 1984;25:129–131. 4. Chen GT, Kung JH, Beaudette KP. Artifacts in computed tomography scanning of moving objects. Semin Radiat Oncol 2004;14:19–26. 5. Balter JM, Ten Haken RK, Lawrence TS, et al. Uncertainties in CT-based radiation therapy treatment planning associated with patient breathing. Int J Radiat Oncol Biol Phys 1996;36: 167–174. 6. Aruga T, Itami J, Aruga M, et al. Target volume definition for upper abdominal irradiation using CT scans obtained during inhale and exhale phases. Int J Radiat Oncol Biol Phys 2000;48: 465–469. 7. Balter JM, Lam KL, McGinn CJ, et al. Improvement of CTbased treatment-planning models of abdominal targets using static exhale imaging. Int J Radiat Oncol Biol Phys 1998;41: 939–943. 8. Ten Haken RK, Balter JM, Marsh LH, et al. Potential benefits of eliminating planning target volume expansions for patient breathing in the treatment of liver tumors. Int J Radiat Oncol Biol Phys 1997;38:613–617. 9. 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. 10. Shirato H, Seppenwoolde Y, Kitamura K, et al. Intrafractional tumor motion: lung and liver. Semin Radiat Oncol 2004;14: 10–18. 11. Weiss PH, Baker JM, Potchen EJ. Assessment of hepatic respiratory excursion. J Nucl Med 1972;13:758–759. 12. Harauz G, Bronskill MJ. Comparison of the liver’s respiratory motion in the supine and upright positions: Concise communication. J Nucl Med 1979;20:733–735. 13. Shirato H, Harada T, Harabayashi T, et al. Feasibility of insertion/implantation of 2.0-mm-diameter gold internal fiducial markers for precise setup and real-time tumor tracking in radiotherapy. Int J Radiat Oncol Biol Phys 2003;56:240–247. 14. Brock KM, Balter JM, Dawson LA, et al. Automated generation of a four-dimensional model of the liver using warping and mutual information. Med Phys 2003;30:1128–1133.
15. Rietzel E, Chen GT, Choi NC, et al. Four-dimensional imagebased treatment planning: Target volume segmentation and dose calculation in the presence of respiratory motion. Int J Radiat Oncol Biol Phys 2005;61:1535–1550. 16. Beddar AS, Kainz K, Briere TM, et al. Correlation between internal fiducial tumor motion and external marker motion for liver tumors imaged with 4D-CT. Int J Radiat Oncol Biol Phys 2007;67:630–638. 17. Hussain SM, Semelka RC. Hepatic imaging: comparison of modalities. Radiol Clin North Am 2005;43:929–947. 18. Voroney JP, Brock KK, Eccles C, et al. Prospective comparison of computed tomography and magnetic resonance imaging for liver cancer delineation using deformable image registration. Int J Radiat Oncol Biol Phys 2006;66:780–791. 19. Shimizu S, Shirato H, Xo B, et al. Three-dimensional movement of a liver tumor detected by high-speed magnetic resonance imaging. Radiother Oncol 1999;50:367–370. 20. 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. 21. Wagman R, Yorke E, Ford E, et al. Respiratory gating for liver tumors: use in dose escalation. Int J Radiat Oncol Biol Phys 2003;55:659–668. 22. Hanley J, Debois MM, Mah D, et al. Deep inspiration breathhold technique for lung tumors: The potential value of target immobilization and reduced lung density in dose escalation. Int J Radiat Oncol Biol Phys 1999;45:603–611. 23. Shirato H, Shimizu S, Kitamura K, et al. Four-dimensional treatment planning and fluoroscopic real-time tumor tracking radiotherapy for moving tumor. Int J Radiat Oncol Biol Phys 2000;48:435–442. 24. 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. 25. Korin HW, Ehman RL, Riederer SJ, et al. Respiratory kinematics of the upper abdominal organs: a quantitative study. Magn Reson Med 1992;23:172–178. 26. von Siebenthal M, Szekely G, Gamper U, et al. 4D MR imaging of respiratory organ motion and its variability. Phys Med Biol 2007;52:1547–1564. 27. Prakash V, Stainsby VA, Craig T, et al. Validation of a supervised automated algorithm for fast quantitative evaluation of organ motion on MRI. Int J Radiat Oncol Biol Phys. In press.