DOI: 10.1148/radiol.2203010132
(Radiology. 2001;220:810-815.)
© RSNA, 2001
Free-breathing Cardiac MR Imaging: Study of Implications of Respiratory Motion—Initial Results1
Kay Nehrke, PhD,
Peter Börnert, PhD,
Dirk Manke, MS and
Johannes C. Böck, MD
1 From the Department of Magnetic Resonance, Technical Systems Research Division, Philips Research Laboratories, Roentgenstrasse 24-26, D-22315 Hamburg, Germany (K.N., P.B.); Institute of Biomedical Engineering, University of Karlsruhe, Germany (D.M.); and Department of Radiology, München-Solln am Martha-Maria Krankenhaus, Munich, Germany (J.C.B.). Received December 13, 2000; revision requested January 18, 2001; revision received February 22; accepted March 23. Supported in part by project grant BO 866/3-1 from the Deutsche Forschungsgemeinschaft. Address correspondence to K.N. (e-mail: kay.nehrke@philips.com).
 |
ABSTRACT
|
|---|
The respiratory motion of several anatomic regions (right hemidiaphragm, left ventricle of the heart, chest wall, abdominal wall) was investigated during free breathing in 10 healthy volunteers by using multinavigator technology and real-time magnetic resonance (MR) imaging. The respiratory motion shows hysteretic effects, which are strongly subject dependent and might have some effect on the quality of cardiac MR images.
Index terms: Coronary vessels, MR, 54.121412, 54.12144 • Diaphragm, MR, 54.121412, 569.12144 • Heart, MR, 54.121412, 51.12144
|
INTRODUCTION
|
|---|
Respiratory motion is one of the major problems in coronary magnetic resonance (MR) angiography. To avoid image degradation from severe artifacts, motion-compensation strategies, such as breath holding (1) or respiratory gating (2), have to be applied. Most gating approaches involve use of navigator echoes (3,4) to monitor the feet-to-head motion of the right hemidiaphragm dome during MR data acquisition. This information is used to accept or reject the data measured with respect to a chosen reference motion state. Furthermore, the navigator information may be used to perform prospective motion correction, such as section tracking (5), to further improve image quality within the chosen gating window. For this approach, use of a motion model is required to properly map the diaphragmatic motion onto the respiratory motion of the heart.
A prerequisite for such a model is that use of a single parameter, such as the diaphragm position, is sufficient to fully characterize the respiratory motion in the region of interest. In fact, Wang et al (6) reported that the respiratory motion of the heart is approximately a global translation dominated by feet-to-head motion that is linearly related to the feet-to-head motion of the diaphragm. For the right coronary artery root and the left anterior descending coronary artery, the mean correction factors were 0.57 ± 0.26 (SD) and 0.7 ± 0.18, respectively (6). These data have subsequently been used to confirm improvements in image quality in case of section tracking (5). On the other hand, there is a wide variation in correction factors between individuals, which necessitates patient-dependent calibration (7,8).
A potential limitation of these studies is that the anatomy of the thorax was measured during breath holding, which might not properly reflect the ongoing dynamic processes during continuous breathing (6). In principle, real-time imaging techniques allow the study of respiratory motion during free breathing. Findings in such experiments have also indicated strong inconsistencies among individual subjects (9,10). The processing effort required for manual extraction of the displacement information from the image data is high, however, and makes coverage of more than a few respiratory cycles impractical.
Multinavigator pulses were recently introduced by Sachs et al (11). Owing to the high spatial and temporal resolutions and automatic evaluation of the navigator information, this technique allows detailed study of spatial and temporal correlations of respiratory motion.
The purpose of this study was to investigate respiration-induced displacement of the heart during free breathing to potentially improve the accuracy of gating and motion-correction techniques in cardiac MR imaging.
|
Materials and Methods
|
|---|
Subjects
Ten healthy adult male volunteers (age range, 25–45 years; mean age, 34 years ± 5), whose hearts were in sinus rhythm and who did not have contraindications to MR imaging, were examined. All participants gave their written informed consent, and the research protocol was approved by our institutional review board.
Multinavigator Pulses
In vivo experiments were performed with a 1.5-T whole-body MR imager (Gyroscan ACS NT; Philips Medical Systems, Best, the Netherlands) with self-shielded gradients (23 mT/m in 0.2 msec). The imaging software was extended to provide as many as four independent pencil-beam navigator pulses, which can be positioned and angulated freely in space (Fig 1). Two-dimensional (2D) spatially selective radio-frequency pulses based on spiral gradient waveforms are used to excite a pencil-beam–shaped region (12). The pencil-beam magnetization is acquired in the presence of a readout gradient applied in the pencil-beam direction (refocused gradient echo; repetition time msec/echo time msec, 20/2; pixel bandwidth, 210 Hz). The Fourier transform of this signal yields the projection of the pencil-beam magnetization onto the beam axis, the so-called navigator profile. By means of a cross-correlation analysis (13) between the current profile and a prestored reference profile, displacement of the anatomic region under consideration (eg, diaphragm) along the pencil beam is determined in real time and stored for subsequent analysis.
View larger version (43K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 1. Sagittal (left) and coronal (middle) 2D gradient-echo MR images indicate navigator positions (white boxes). Right: Navigator profile images indicate the positions of the heart and diaphragm (arrows on left and middle) as a function of time. In the navigator profile image of the heart (top right), a contribution from cardiac motion is superimposed.
|
|
The total duration of the navigator sequence, including excitation, acquisition, and evaluation of the navigator signal, was 20 msec. To investigate spatial and temporal correlations of respiratory motion, a pure navigator sequence was implemented in which the different navigators were applied repeatedly in cyclic order. This multinavigator sequence was used to record the respiratory motion of several anatomic regions (left ventricle of the heart, dome of the right hemidiaphragm, chest wall, abdominal wall) (Fig 1) during 10 minutes. The diameter of the pencil beams was between 25 and 35 mm. The spatial resolution of the navigator profiles was 1 mm. With spatial interpolation of the correlation function, an accuracy of the displacement data in the range of 0.2 mm was achieved. To attain quasi-simultaneous data for each of the navigators, the delay of 20 msec between succeeding navigators was corrected afterward by interpolating neighboring data points of the breathing curves. In addition, the contribution of cardiac motion in the oversampled data was carefully filtered out (Fig 2).
The spatial and temporal correlations between the different navigators were depicted in 2D histogram plots, in which the navigator displacements were plotted against each other. Correction factors for the relationship between the diaphragm and heart were obtained by means of the linear least squares method that was fit to the corresponding data. To check the data-evaluation method, additional 2D histograms were prepared, in which, instead of cardiac motion being averaged out, distinct heart phases (systole and diastole) were selected. For this purpose, only those navigator measurements were taken in which the superimposed high-frequency contribution from cardiac displacement was maximal or minimal (Fig 2).
Real-time MR Imaging
Real-time MR images of the thoracicoabdominal region were acquired in the electrocardiography-triggered single-shot mode (2D gradient-echo sequence, 4.0/1.8, 128 x 90 image matrix, half Fourier, 70% k-space coverage, 300-mm field of view, 150-msec acquisition window in middiastole, during 200 heart cycles) with continuous breathing by the volunteers. The imaging plane was double oblique and covered the proximal part of the right coronary artery. Accurate geometric parameters were obtained on a high-spatial-resolution, contrast-prepared coronary image (the protocol is described in reference 14) of the corresponding volunteer.
The diaphragm position was monitored during the acquisition by using leading and trailing navigator echoes. In a subsequent postprocessing step, two types of cine images were prepared. For the first type, the real-time images were sorted with respect to the diaphragm position, regardless of the respiratory phase. For the second type, the images were also separated into inspiration and expiration. The cine images were independently examined by three authors (K.N., P.B., D.M.). The continuity and smoothness of the heart displacement on the cine images was considered to be a measure of the correlation between the motions of the diaphragm and heart.
|
Results
|
|---|
Two-dimensional Histograms
In Figure 3, the correlation between the feet-to-head motion of the right hemidiaphragm and that of the left ventricle is shown in 2D histograms for four volunteers. Well-defined trajectories without much scattering were found for all volunteers. This indicates a good correlation between the motions of the diaphragm and heart. However, the histograms also show hysteretic loops with different paths for expiration and inspiration. In hysteresis, a certain diaphragm position corresponds to different heart positions for inspiration and expiration. This hysteretic behavior is strongly subject dependent; its variability is roughly reflected by the four examples in Figure 3.
View larger version (62K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 3. Four 2D histogram plots of cardiac and diaphragmatic breathing curves for four volunteers. Dark squares indicate many counts, and lighter squares indicate fewer counts. The straight lines are linear fits to the data (cf = correction factor [slope], r = correlation coefficient). Bottom right: The direction of the hysteresis is always counterclockwise. Bottom left: The dotted lines depict the mapping of a 5-mm diaphragmatic gating window (horizontal double-headed arrow) onto the corresponding cardiac displacement range (vertical double-headed arrow). The offset of the plots is relative to the arbitrarily chosen reference navigator.
|
|
Among the histograms for the 10 volunteers, those for two show a linear relationship; for four, a moderate hysteresis; for two, a marked hysteresis; and for two, a strong hysteresis. The slope of the linear fits varied between 0.46 and 0.99, comparable to the range of correction factors given by Taylor et al (7). However, the hysteretic behavior led to poor fits to the linear model, as indicated by smaller correlation coefficients.
For the limited number of volunteers in this initial study, the same direction of the hysteretic loop was always observed, which corresponds to a delay of the respiration-induced motion of the heart with respect to the motion of the diaphragm. The hysteretic loops could also be described quantitatively by means of a simple delay. By using subject-dependent delay times as long as 0.6 second for all volunteers, the gap could be decreased below the scattering width of the trajectories. This is demonstrated in Figure 4, left, for one volunteer, who exhibited the strongest hysteretic effect (compare with Fig 3, bottom right).
View larger version (25K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 4. Two-dimensional histogram plots of cardiac and diaphragmatic breathing curves for one volunteer (see Fig 3, bottom right). Dark squares indicate many counts, and lighter squares indicate fewer counts. Left: An appropriate shift ( = 0.6-second delay of the heart with respect to the diaphragm) between the two breathing curves closes the hysteretic loop. Right: The hysteresis is shown for two heart phases.
|
|
The hysteretic effect was found to be independent of the phase in the cardiac cycle. This is illustrated in Figure 4, right, in which 2D histograms are shown for systole and diastole. Although the histograms are coarser, owing to fewer data points, the hysteresis is clearly present in both heart phases. This indicates that the hysteretic effect is independent of cardiac motion. In addition, this effect is also independent of the data-evaluation method.
The hysteretic behavior was also observed for the correlation between chest wall motion and abdominal wall motion in an anterior-to-posterior direction (Fig 5, left), which further indicates the intrinsic nature of the effect. Again, the hysteresis could be removed by introducing an appropriate subject-dependent delay (Fig 5, right). The hysteretic loops were less pronounced (delay times as long as 0.2 second), however, and could be resolved for only six of the volunteers. For the remaining subjects, the gap size was below the scattering width of the trajectories.
View larger version (21K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 5. Two-dimensional histogram plots of breathing curves of abdominal wall and chest wall for one volunteer (see Fig 3, bottom right). Dark squares indicate many counts, and lighter squares indicate fewer counts. Left: The histogram shows a hysteretic loop. Right: The loop can be closed with an appropriate delay of = 0.2 second.
|
|
Real-time Cine Images
The respiratory-phase–ordered cine images showed a smooth motion of the heart during the respiratory cycle for all volunteers. In contrast, discontinuities were observed on the cine images obtained without phase ordering for the four volunteers who exhibited a marked or strong hysteresis. Comparison of the corresponding frames of the cine images showed that the diaphragm was shaped differently during inspiration and expiration. An example is shown in Figure 6, in which images were obtained with the navigator at the same feet-to-head position of the diaphragm but in different respiratory phases (midinspiration, midexpiration).
View larger version (65K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 6. Free-breathing, navigator-gated, three-dimensional, T2-weighted, contrast-enhanced, segmented-k-space, gradient-echo MR images were obtained in the same volunteer as in Left, middle: Single-shot gradient-echo MR images of the right coronary artery were acquired with a fixed feet-to-head diaphragm position but in different respiratory phases. In the left image, the dotted lines indicate the contours of the diaphragm and heart in the expiration phase (middle image), the black line indicates the position of the right coronary artery (RCA in the right image), the black circle in the left image indicates the aorta, and the white circle in the right image indicates the intersection with the left anterior descending coronary artery (LAD in the right image).
|
|
These low-spatial-resolution, single-shot images provide mainly a survey of the heart. The position of the right and left anterior descending coronary arteries, sketched in these images, were obtained from a high-spatial-resolution contrast-enhanced coronary image acquired in the same volunteer (Fig 6, right). The images indicate a small dent in the diaphragm during inspiration in comparison with a rather convex shape during expiration. This led to a tilting of the heart in this plane and, consequently, also of the coronary arteries. Tilting can have a serious effect on the quality of high-spatial-resolution MR images of the coronary arteries.
|
Discussion
|
|---|
The multinavigator results indicate that the respiratory motion of the heart in the feet-to-head direction is well correlated to the motion of the right hemidiaphragm in the same direction. However, the relationship is only roughly linear. Different spatial correlations are observed during inspiration and expiration that manifest as hysteretic loops in the 2D histograms. Similar hysteretic loops were observed in the correlation between the chest and abdominal walls. This phenomenon cannot be studied systematically by analyzing breath-hold images. Instead, free-breathing images have to be acquired. This might be the reason that this hysteretic behavior has not been reported previously, to our knowledge.
The hysteretic loops can be described quantitatively by using a simple delay in imaging between the different anatomic regions. Since respiration is a nearly periodic motion, this delay can also be regarded as a phase shift. Hence, one important conclusion is that not only the amplitude but also the phase of respiratory motion depends on the spatial position in the human body. Phase shifts as large as 40°, a size that is considerable, were found in the present study. One explanation for the phase shift may be that the respiration-induced motion of the heart is retarded as a result of the viscoelastic properties of the lungs. In fact, delayed periodic motion is a typical property of forced dampened oscillation.
Another reason for the hysteresis may be that respiration is a rather complex process in which different groups of muscles contract during inspiration (diaphragm, external intercostal muscle) and expiration (internal intercostal muscle, transverse muscle of the thorax) (15). Obviously, the interplay between these muscles changes during the respiratory cycle, which may qualitatively explain the hysteretic behavior. This hypothesis is supported by findings on the respiratory cine images, in which a differently shaped diaphragm was observed during inspiration and expiration. From this point of view, the phase shift is a pure phenomenologic model to describe the experimental results. One might further speculate that the different respiratory muscles involved do not contract or relax simultaneously, however, but have different phases in the respiratory cycle. This would naturally explain the phase shifts between different spatial regions in the body.
As indicated by the respiratory cine images of the right coronary artery plane, the hysteretic effect may have a serious effect on the accuracy and reliability of gating and motion-correction schemes used for MR coronary angiography. In a gated acquisition, a certain range of displacement of the diaphragm around the most frequent motion state, which is usually at end expiration, is chosen as the acceptance window (Fig 3, bottom left). In this example, the relative size of the gating window (5 mm) was only 25% of the displacement range of the whole diaphragm (20 mm). However, the corresponding cardiac displacements are located in a window of 9 mm, which is 60% of the displacement range for the whole heart (15 mm). As a consequence, the effective gating window was much broader than expected on the basis of the linear model proposed by Wang et al (6). Furthermore, prospective motion-correction schemes, such as section tracking (5), may lead to considerable errors, because the same correction is applied for the two respiratory motion states of the heart, inspiration and expiration.
The most direct approach to circumvent these problems is to place the navigator directly through the heart, as discussed previously (9,10). However, destructive interference between the pencil-beam navigator and the imaging volume may affect image quality and navigator accuracy.
Alternatively, one could combine the standard diaphragmatic navigator, which is robust from the experimental point of view, with a better motion model that takes the hysteretic effect into account. Of course, such a model requires patient-specific calibration in the preparation phase before imaging. Multinavigators are a potential technique for such preimaging. As in this study, displacements of heart and diaphragm could be measured for several respiratory cycles by using multiple navigators in the preparation phase. The spatial correlations could be analyzed, and a function table or an analytic expression could be derived from the data to map the diaphragmatic motion onto the respiratory motion of the heart. During imaging, respiratory-phase–adapted gating could be performed, in which the respiratory phase is considered in addition to the diaphragm position to estimate the heart position for hysteretic motion patterns.
For the limited number of volunteers in this study, the hysteresis could always be described quantitatively by using a simple delay of the heart with respect to the diaphragm. Hence, it would also be interesting to introduce an appropriate delay between the diaphragmatic navigator and cardiac imaging sequences to correct for the hysteresis. This delay would allow standard gating to be performed on the basis of only the diaphragm position.
In summary, respiration-induced displacements of the diaphragm and heart show hysteretic effects, which are strongly subject dependent and might have some effect on the quality of cardiac MR images. The multinavigator technique is a powerful tool with which to study these complex motion patterns. Advanced gating techniques and motion-compensation approaches that address the hysteretic effect might improve image quality in free-breathing cardiac MR imaging.
|
FOOTNOTES
|
|---|
Abbreviation: 2D = two-dimensional
Author contributions: Guarantors of integrity of entire study, K.N., P.B.; study concepts and design, K.N., P.B., J.C.B.; literature research, K.N., P.B.; experimental studies, K.N., D.M.; data acquisition, K.N., D.M.; data analysis/interpretation, K.N., P.B.; manuscript preparation, K.N.; manuscript definition of intellectual content and editing, K.N., P.B.; manuscript revision/review and final version approval, all authors.
|
REFERENCES
|
|---|
-
Paling MR, Brookeman JR. Respiration artifacts in MR imaging: reduction by breath holding. J Comput Assist Tomogr 1986; 10:1080-1082.[Medline]
-
Prato FS, Nicholson RL, King M, Knill RL, Reese L, Wilkins K. Abolition of respiration movement markedly improved NMR images of the thorax and upper abdomen (abstr) In: Book of abstracts: Society of Magnetic Resonance in Medicine 1983. Berkeley, Calif: Society of Magnetic Resonance, 1983; 284.
-
Ehman RL, Felmlee JP. Adaptive technique for high-definition MR imaging of moving structures. Radiology 1989; 173:255-263.[Abstract/Free Full Text]
-
Sachs TS, Meyer CH, Hu BS, Kohli J, Nishimura DG, Macovski A. Real-time motion detection in spiral MRI using navigators. Magn Reson Med 1994; 32:639-645.[Medline]
-
McConnell MV, Khasgiwala VC, Savord BJ, et al. Prospective adaptive navigator correction for breath-hold MR coronary angiography. Magn Reson Med 1997; 37:148-152.[Medline]
-
Wang Y, Riederer SJ, Ehman RL. Respiratory motion of the heart: kinematics and the implications for the spatial resolution in coronary imaging. Magn Reson Med 1995; 33:713-719.[Medline]
-
Taylor AM, Jhooti P, Keegan J, Firmin DN, Pennel DJ. Improved MR coronary angiography using real-time navigator echoes and a subject specific calculated adaptive motion correction factor (abstr) In: Proceedings of the Sixth Meeting of the International Society for Magnetic Resonance in Medicine. Berkeley, Calif: International Society for Magnetic Resonance in Medicine, 1998; 322.
-
Nagel E, Bornstedt A, Schnackenburg B, Hug J, Oswald H, Fleck E. Optimization of realtime adaptive navigator correction for 3D magnetic resonance coronary angiography. Magn Reson Med 1999; 42:408-411.[CrossRef][Medline]
-
Danias PG, Stuber M, Botnar RM, Kissinger KV, Edelman RR, Manning WJ. Relationship between motion of coronary arteries and diaphragm during free breathing: lessons from real-time MR imaging. AJR Am J Roentgenol 1999; 172:1061-1065.[Abstract/Free Full Text]
-
Johansson LO, Fischer SE, Smink J, Hofman MBM, Lorenz CH. Using real-time imaging to determine the correlation between the position of the heart and a navigator echo positioned on the diaphragm (abstr) In: Proceedings of the Sixth Meeting of the International Society for Magnetic Resonance in Medicine. Berkeley, Calif: International Society for Magnetic Resonance in Medicine, 1998; 855.
-
Sachs TS, Meyer CH, Pauly JM, Hu BS, Nishimura DG, Macovski A. The real-time interactive 3-D-DVA for robust coronary MRA. IEEE Trans Med Imaging 2000; 19:73-79.[CrossRef][Medline]
-
Nehrke K, Börnert P, Groen J, Smink J, Böck JC. On the performance and accuracy of 2D navigator pulses. Magn Reson Imaging 1999; 17:1173-1181.[CrossRef][Medline]
-
Wang Y, Grimm RC, Felmlee JP, Riederer SJ, Ehman RL. Algorithms for extracting motion information from navigator echoes. Magn Reson Med 1996; 36:117-123.[Medline]
-
Stuber M, Botnar RM, Danias PG, et al. Double-oblique free-breathing high resolution three-dimensional coronary magnetic resonance angiography. J Am Coll Cardiol 1999; 34:524-531.[Abstract/Free Full Text]
-
Pschyrembel W, ed. Klinisches wörterbuch 258th ed. Berlin, Germany: de Gruyter, 1998; 131.
TOP
ABSTRACT
INTRODUCTION
) and left ventricle (
). Low-pass filtering (fc = 1 Hz, solid line) removes the superimposed cardiac motion.