Coronary MR angiography (MRA) is able to depict the

Transcription

1 Improved Coronary Artery Definition With T2-Weighted, Free-Breathing, Three-Dimensional Coronary MRA René M. Botnar, PhD; Matthias Stuber, PhD; Peter G. Danias, MD, PhD; Kraig V. Kissinger, BS, RT; Warren J. Manning, MD Background Three-dimensional (3D) navigator-gated and prospectively corrected free-breathing coronary magnetic resonance angiography (MRA) allows for submillimeter image resolution but suffers from poor contrast between coronary blood and myocardium. Data collected over 00 ms/heart beat are also susceptible to bulk cardiac and respiratory motion. To address these problems, we examined the effect of a T2 preparation prepulse (T2prep) for myocardial suppression and a shortened acquisition window on coronary definition. Methods and Results Eight healthy adult subjects and 5 patients with confirmed coronary artery disease (CAD) underwent free-breathing 3D MRA with and without T2prep and with 120- and 60-ms data-acquisition windows. The T2prep resulted in a 123% (P 0.001) increase in contrast-to-noise ratio (CNR). Coronary edge definition was improved by 33% (P 0.001). Acquisition window shortening from 120 to 60 ms resulted in better vessel definition (11%; P 0.001). Among patients with CAD, there was a good correspondence with disease. Conclusions Free-breathing, T2prep, 3D coronary MRA with a shorter acquisition window resulted in improved CNR and better coronary artery definition, allowing the assessment of coronary disease. This approach offers the potential for free-breathing, noninvasive assessment of the major coronary arteries. (Circulation. 1999;99: ) Key Words: imaging angiography contrast media vessels Coronary MR angiography (MRA) is able to depict the major coronary arteries. 1 6 Several impediments, however, limit its clinical utility. These obstacles include compensation for cardiac and respiratory motion, need for submillimeter spatial resolution, and suppression of signal from adjacent epicardial fat and myocardium. Initial approaches used 2-dimensional (2D) gradient-echo strategies 1 3,5,7,8 to take advantage of unsaturated blood inflow and breath holding to minimize respiratory motion. In contrast to breath holding, respiratory bellows 9 or MR navigator echoes that assess diaphragmatic or cardiac position can be used for respiratory gating with or without slice correction to allow for free-breathing coronary MRA By eliminating the time constraint of the breath hold, navigators allow for submillimeter spatial resolution and minimize registration errors. Three-dimensional (3D) MRA techniques offer favorable signal-to-noise ratios (SNR) and are also well suited for navigator approaches. 13,17 A disadvantage of 3D versus 2D MRA is lower contrast between coronary blood and myocardium. To overcome this problem, contrast agents 18 or prepulses 4 may be useful. We chose to examine the impact of a T2 preparation prepulse (T2prep), which has been demonstrated to enhance contrast in 2D MRA applications. 7 Because image quality may also be affected by respiratory motion during the data-acquisition window, we also varied the acquisition window between 60 and 120 ms. To avoid bias in image assessment, we developed an objective tool to determine vessel border sharpness. We hypothesized that the combination of 3D MRA with T2prep and a shorter acquisition window would allow for both improved contrast-to-noise ratio (CNR) and coronary edge definition. Methods Subjects Eight healthy adults (mean age, 24 years; range, 18 to 37; 5 men) without a history of cardiovascular disease and 5 patients (mean age, 58 years; range, 42 to 71; 4 men) with radiographically confirmed coronary disease (CAD) were studied. No breath holds or respiratory coaching were performed. Written informed consent was obtained from all participants, and the protocol was approved by the hospital s Committee on Clinical Investigations. Imaging Procedure All subjects were examined supine with a commercial, 1.5-T Gyroscan ACS-NT (Philips Medical Systems) scanner with Power- Trak 3000 gradients (15 mt/m, 50 mt m ms slew rate), a cardiac software patch, and a 5-element phased-array research cardiac coil. Received December 27, 1998; revision received March 24, 1999; accepted April 9, From the Department of Medicine, Cardiovascular Division (R.M.B., M.S., P.G.D., K.V.K., W.J.M.) and Department of Radiology (W.J.M.), Beth Israel Deaconess Medical Center and Harvard Medical School, Boston, Mass; and Philips Medical Systems (R.M.B., M.S.), Best, Netherlands. Reprint requests to René Botnar, PhD, Beth Israel Deaconess Medical Center, Cardiovascular Division, Cardiac MR, 330 Brookline Ave, Boston, MA rbotnar@caregroup.harvard.edu 1999 American Heart Association, Inc. Circulation is available at

2 3140 T2-Weighted, Free-Breathing, 3D Coronary MRA Figure 1. Coronal scout of heart and diaphragm. This image was used to identify origin of LM and to position navigator on free wall of LV. Localizer Scans Two scout scans were performed to localize the left main (LM) coronary artery and to position the cardiac navigator. An ECGtriggered multislice, multishot 2D turbo field echo scan with thoracic transverse, sagittal, and coronal images was used to identify the heart, followed by a coronal scout with left ventricular (LV) navigator gating and prospective slice correction. 14 The coronal scout images were used to localize the LM and position the LV navigator (Figure 1). 3D Coronary MRA With T2prep The coronal scout position of the LM (Figure 1) was used as the center slice position of the 3D MRA scan. The MR pulse sequence consisted of 4 blocks (Figure 2) repeated with every heart cycle, with a temporal relationship chosen to minimize the time interval between the navigator and the imaging sequence. Figure 2. Schematic of pulse sequence. Image acquisition was performed with a trigger delay so as to image during mid diastole, a period of minimal cardiac motion. Imaging block was preceded by T2prep for muscle suppression, 2D selective realtime navigator pulse for respiratory motion compensation, and frequency-selective fat suppression prepulse (Fat Sat). Depending on length of acquisition window (60 or 120 ms), 8 or 16 echoes were acquired per shot. These sequence blocks were repeated with every heart cycle. TFE indicates turbo field echo. T2prep Flow-independent coronary MRA with T2prep was first described by Brittain et al. 7 It uses a 90 radiofrequency (RF) pulse to flip the M z magnetization vector into the M xy plane and is most beneficial in sequences with low inflow contrast. T2prep allows suppression of tissues with short T2 relaxation times, eg, cardiac muscle (T2 50 ms), 19 cardiac veins with deoxygenated blood (20% O 2 saturation, T2 35 ms), 20 and epicardial fat. Tissues with long T2 relaxation times, eg, arterial blood (T2 250 ms), are minimally influenced. The CNR between arterial blood and myocardium can be adjusted by varying the time between the 90 and the tip-up pulse. Flow sensitivity of T2prep can be minimized by nonselective RF pulses and use of an even number of 180 refocusing pulses. We applied 4 refocusing pulses 7 and used an MLEV (Malcom-Levitt) weighted T2prep as well as composite 180 RF pulses (90 x 180 y 90 x ) to correct for pulse imperfections due to B 0 and B 1 field inhomogeneities. CNR Optimization A Bloch equation 21 simulation, taking into account arterial blood and myocardium T1 and T2 (at 1.5 T) of 1200 and 250 ms and 850 and 50 ms, 19,22 respectively, was used. This demonstrated maximal CNR with 90-ms T spacing. However, SNR drops as T increases (Figure 3). On the basis of in vivo experiments, the threshold for SNR of arterial blood was set to 8, corresponding to a T of 50 ms, a reasonable compromise between SNR and CNR (Figure 3). Navigator-Guided Prospective Slice Correction A cranial-caudal oriented 2D selective navigator through the basal LV was used to compensate for respiratory motion. 16,23 Navigator gating with prospective slice correction was used with a 3-mm gating window and slice correction in the craniocaudal direction (principal respiratory motion). For motion correction in slice-selection direction, the measured displacement was translated in a frequency offset of the RF excitation pulse of the imaging sequence. No scaling of the measured displacement was performed. Scan efficiency was defined as accepted shots/total heart beats. Imaging Sequence A frequency-selective fat-suppression pulse to suppress signal from epicardial fat preceded the imaging acquisition. The field-of-view

3 Botnar et al June 22, Figure 3. Results of simulation of Bloch equations. SNR of blood and CNR between blood and cardiac muscle were calculated for different delay T of the 90 tip-up pulse (x axis). Due to the short T2 of cardiac muscle compared with that of arterial blood, relative (rel) signal level of cardiac muscle decreases faster than that of blood. CNR maximum can be observed for T of 90 ms. With increasing T between 90 RF pulses, however, SNR of blood decreases. Therefore, as a compromise between CNR and SNR, T of 50 ms was used in all studies. x axis: tip-up delay T (in ms). Primary y axis: relative signal level. Secondary y axis: CNR between blood and myocardium. was mm with an image matrix of (in-plane resolution, mm). The 3D slab had 10 slices with a3mm slice thickness, interpolated to 20 slices with a 1.5 mm thickness. Each 3D volume slice was acquired in 38 shots by use of a centric k-space ordering scheme with a 7.4-ms repetition time and 2.5-ms time echo. To examine the impact of the data-acquisition window, imaging was performed with 120-ms (16 echoes/shot) and 60-ms (8 echoes/shot) acquisition windows. All data were acquired at mid diastole. Study Protocol Three coronary scans were performed in all healthy subjects. To investigate the impact of the T2prep on image quality, scans with and without T2prep were performed in random order. In all other aspects, imaging parameters were maintained, including an acquisition window of 60 ms. The third scan used T2prep with a 120-ms acquisition window. SNR and CNR Evaluation SNR of coronary blood was determined from a region-of-interest in the ascending aorta at the level of the LM origin. Mean blood signal (S blood ) and SD (N blood ) were calculated, with SNR defined as S blood /N blood. Contrast between coronary blood and cardiac muscle was calculated as (1) CNR S blood S muscle 0.5 N blood N muscle Mean muscle signal (S muscle ) was determined at the anterobasal LV. CNR was also calculated between proximal left anterior descending coronary artery (LAD; arterial) and great cardiac vein (venous) blood. Curved-Plane Vessel Reconstruction All data were analyzed on an EasyVision workstation (Philips Medical Systems). The user navigates through the 3D data set and interactively marks the vessel of interest. The course of the LAD and left circumflex artery (LCx) are then reconstructed (Figure 4) and displayed in a single image plane, from which the contiguous length of the LAD and LCx was then determined. Vessel Edge Detection Vessel tracing and edge detection were performed on gray-scale images of the raw 3D data sets with the LAD and LCx slices closest to the LM bifurcation. Edge detection included a centerline and edge-detection procedure (Figure 5; Appendix). Vessel sharpness was defined as the average edge value along the calculated vessel border. Higher edge values correspond to better vessel delineation. Patient Protocol All patients with radiographically confirmed CAD underwent T2prep 3D MRA with the short (60 ms) acquisition window. The 3D Figure 4. LAD and LCx images were reformatted for length quantification. Path of coronary artery tree was interactively defined on 3 different views of 3D data set. Images A through C show reformatted images where LM coronary artery, LAD, and LCx were visible. Length of LAD or LCx was measured along dashed lines shown here.

4 3142 T2-Weighted, Free-Breathing, 3D Coronary MRA Figure 5. Vessel border-detection algorithm applied on a raw image (A) of 1 subject. In a first run, centerline and corresponding intensity profiles are calculated. Intensity profiles (B) across the vessel allow for vessel border approximation with FWHM criteria. In a second step, true vessel border is detected on edge image (C), which is defined as maximum value found within interval defined by FWHM border lines. AA indicates ascending aorta; RVOT, right ventricular outflow tract. Figure 6. Baseline image (no prepulses) (A) and T2prep image (B) from 3D coronary MRA data set. Application of T2prep (B) improved contrast between LV cavity and surrounding LV wall. Blood-to-muscle contrast enhancement helped in this representative case to visualize a distal part of LAD, which is not well seen in baseline image. Respiratory motion artifacts are also reduced in T2 contrastenhanced image, whereas ghosts of chest wall can be observed in baseline image.

5 Botnar et al June 22, Figure 7. Baseline images (A, C) and T2prep-enhanced images (B, D) of 3D coronary data set. Application of T2prep (B, D) helped to suppress cardiac muscle as well as skeletal muscle in chest and back. Due to enhanced blood-to-muscle contrast in T2prep images (B, D), visualization of LAD and LCx is improved compared with reference images (A, C). Suppression of venous blood additionally helps to differentiate between great cardiac vein (GVC), LAD, and LCx, which is difficult in reference image (A). Another advantage of T2prep is reduction of respiratory motion and flow-related artifacts (B, D). MRA data sets were reformatted along the coronary vessel path and compared with the corresponding x-ray angiogram. Statistics All data are expressed as mean SD. Continuous variables were compared with a 2-tailed paired Student s t test, with significance as P Results Scanning was completed without incident in all subjects. Total imaging time for each acquisition varied from 10 to 20 minutes depending on heart rate and navigator efficiency. T2prep: Healthy Subjects In Figure 6, the CNR impact of T2prep is demonstrated between the ventricular cavity blood and the myocardium. In the absence of T2prep (Figure 6A), there is little contrast between the myocardium and blood pool. With T2prep (Figure 6B), there is suppression of myocardium, with improved LAD visualization. In Figure 7, a similar comparison is shown for the LAD and the LCx. In addition to myocardial suppression, T2prep also suppressed signal from the great cardiac vein (Figure 7B), and fewer respiratory artifacts were observed (Figure 6B and Figure 7B and D). Overall, T2prep improved CNR 123% (P 0.001) between coronary blood and myocardium (Table 1). There was a trend toward improved CNR between arterial and venous blood and a trend toward reduced SNR (Table 1). Application of the T2prep did not adversely affect navigator performance (Table 1). The use of T2prep allowed for a 33% improvement in LAD and LCx vessel definition (P 0.001; Table 2). The objectively measured proximal vessel diameter was not T2prep dependent. The high image quality of all coronary scans allowed visualization of the entire LM and the

6 3144 T2-Weighted, Free-Breathing, 3D Coronary MRA TABLE 1. Impact of T2prep and Data-Acquisition Window on SNR and CNR and Scan Efficiency in 8 Healthy Subjects No T2prep (60 ms) T2prep (60 ms) P T2prep (120 ms) P vs T2prep at 60 ms CNR blood-muscle NS SNR blood NS CNR artery-vein NS NS Scan efficiency, % NS NS proximal and mid LAD (Figures 4 and 8, top) in all subjects. The visible length of the LCx (Figure 8, bottom) was longer (121%; P 0.003) with T2prep (Table 2), and there was a trend toward longer visualization of the LAD (Table 2). Acquisition Window Reduction of the acquisition window resulted in a smaller improvement (11%; P 0.001) in LAD and LCx definition (Table 2). Most notably, the LCx was longer (48%; P 0.05) with the shorter (60 ms) acquisition window (Table 2). There was also a trend (P 0.07) toward a longer visible length of the LAD (Table 2) and a better CNR between arterial and venous blood (Table 1). There was no significant difference in CNR, SNR, or measured vessel diameter (Tables 1 and 2). Patient Study Good agreement between x-ray angiography and coronary 3D MRA regarding vessel anatomy can be observed in a representative sample shown in Figure 9. This case shows the potential of coronary MRA to assess the more distal parts of the coronary system. Muscle and fat suppression were sufficient for good delineation of the mid LAD from myocardium and epicardial fat. An example from a patient with diffuse coronary disease is shown in Figure 10. Discussion A noninvasive method to evaluate the native coronary arteries would be a major clinical advance. In this objective study comparing free-breathing 3D coronary MRA with and without T2prep, we demonstrated that T2prep improved CNR and definition of the LAD and LCx but made no significant difference in visible LAD length. The latter was primarily influenced by imaging volume. In contrast, visualization of the LCx appeared to be related to its anatomic location. The improved visualization of the LCx with T2prep is likely due to improved suppression of myocardium and veins. As previously discussed by Brittain et al, 7 T2prep should suppress deoxygenated venous blood (Figure 7B) owing to its shorter T2. We observed a trend toward an improved CNR between arterial and venous blood. This may be particularly important, because the paths of the great cardiac vein and LCx are quite close. This suppression might also be helpful for individuals in whom the vein crosses the LAD (Figure 7A and 7B). Coronary diameters were T2prep independent. T2prep also reduced respiratory motion artifacts, likely owing to suppression of chest wall muscle signal. Reduction of the acquisition window from 120 to 60 ms led to an improvement in LAD delineation. However, compared with T2prep, the absolute magnitude was small and came at the expense of a doubled scan time. This time might be better used to increase the volume coverage or use of averaging. T2prep in combination with free-breathing coronary MRA resulted in quantitatively better definition of the LAD and LCx. Compared with previous studies, 3 5 our findings are based on objective quantitative analyses of vessel-wall sharpness. Application of automated vesseledge algorithms with a standardized vessel sharpness score might allow for more objective comparisons of different imaging strategies. Clinical Considerations T2prep 3D coronary MRA showed good agreement with x-ray angiography for assessment of vessel anatomy as well as CAD. Application of T2prep suppressed the signal from the myocardium and therefore allowed for improved TABLE 2. Impact of T2prep and Data-Acquisition Window on Vessel Sharpness, Contiguous Length, and Coronary Diameter in 8 Healthy Subjects No T2prep (60 ms) T2prep (60 ms) P T2prep (120 ms) P vs T2prep at 60 ms LAD Vessel sharpness, % Average diameter, mm NS NS Vessel length, mm LCx Vessel sharpness, % Average diameter, mm NS NS Vessel length, mm

7 Botnar et al June 22, Figure 8. Reformatted images of LAD (top) and LCx (bottom) of all 8 healthy subjects with T2prep used with 60-ms acquisition window. One important observation of this study is the robust image quality of the presented technique, which allows for visualization of LAD and LCx in all examined cases. blood-to-myocardium contrast. Both diffuse and focal coronary disease could be visualized. The accuracy of the methodology remains to be examined in large, multicenter studies. With the introduction of intravascular contrast agents, 24,25 the question arises whether T2prep would offer a synergistic effect. According to our preliminary experiences, such intravascular contrast agents reduce the differ-

8 3146 T2-Weighted, Free-Breathing, 3D Coronary MRA Figure 9. Assessment of LM and LAD by 3D coronary MRA (A, B) and x-ray angiography (C) in 42-year-old patient with chest pain. Good suppression of signal from myocardium allows assessment of proximal (prox) as well as more distal parts of LAD (A). Reformatted sagittal view (B) corresponds well to x-ray angiographic view (C) with good vessel definition for proximal as well as more distal regions (144 mm of LM/LAD is visualized). Figure 10. Assessment of diffuse coronary disease in proximal LAD in 71-year-old patient. Due to good muscle suppression, diffusely diseased LAD can clearly be depicted in MRA (A) and corresponds well to findings in contrast x-ray angiogram (B).

9 Botnar et al June 22, ential in T2 between blood and myocardium, thus decreasing the beneficial effect of the T2prep. Technical Considerations Navigator correction was used to compensate for respiratory motion in the foot-to-head direction. The navigator was applied in the vicinity of the LM with a constant correction factor. This study was not intended to examine the impact of the correction factor. It is likely that the correction factor for the more distal coronary segments would be closer to Conclusions The combined approach of free-breathing navigator-gated and slice-tracked 3D coronary MRA together with a T2prep and a shorter acquisition window resulted in an improved CNR between coronary blood and myocardium and thereby allowed for better definition of the coronary vessels. The technique has been successfully applied to healthy volunteers and a small patient group. Visually assessed vessel anatomy as well as diffuse and focal coronary artery disease correlated well with x-ray angiographic findings. This approach offers the potential for noninvasive assessment of the major coronary arteries. Appendix Vessel tracing and edge detection were performed on gray-scale images of the 3D raw data sets. The user defines the vessel of interest by drawing points p [k] along the vessel path, whereby k is the index of each point. Each of these points is a vector, with a unity direction d [k] and a unity normal nˆ[k] vector defined by Equation 2: (2) d k p k p k p k p k,n k d k 0 The position of these points is then updated with the center of gravity p c [k], representing the vessel centerline. The center of gravity was calculated along the intensity profile I[k, r ], defined by the normal vector nˆ[k] and a predefined window, w (Equation 3). (3) p c k w/ 2 I k, n r k, n n w/ 2 w/ 2 I k, n n w/ 2 (4) I k, n I k, r k, n, with r k, n p k n n k, n w/ 2, w/ 2 The centerline point p c [k] is then regridded with cubic spline interpolation and equidistant spacing between adjacent points P cs [i] and p cs [i]. The index of the regridded centerline points p cs is i. The direction d cs [i], the normal n cs [i] vectors, and the intensity profiles I cs [i, r ] are updated (Equations 2 and 4). After the centerline of the vessel has been defined, vessel border detection is performed. The vessel border location is roughly estimated with full-width halfmaximum (FWHM) criteria (Equation 5). (5) I max cs i 2 max i I cs i, p FWHM I cs 2 (6) p FWHM i p cs i d cs i 0 p FWHM i p cs i d cs i 0 The search window is increased until the criterion of Equation 5 is fulfilled. The orientation of the vessel border points p FWHM [i] with respect to the centerline is determined by the cross product (Equation 6). The final vessel-border detection is done with the Deriche edgedetection filter, 27 a combination of a low-pass filter for noise reduction and a high-pass filter for edge detection. The filter parameter was empirically set to 2.5. Applied on a step function, the Deriche filter returns the step size of the function at the position of the step (edge). The vessel border is identified as the maximum value of the edge profile in the interval [0, p FWHM [i]] and [0, p FWHM [i]]. The edge points are denoted p Deriche [i]. Vessel delineation (sharpness) is defined as the average edge value along the entire vessel border (Equation 7). N Edge i, p Deriche (7) sharpness i N N is the number of edge points. Higher values identify better vessel delineation. The mean difference between the vessel border points p Deriche [i] and p Deriche [i] is defined as the vessel diameter. References 1. Edelman RR, Manning WJ, Burstein D, Paulin S. Coronary arteries: breath-hold MR angiography. Radiology. 1991;181: Meyer CH, Hu BS, Nishimura DG, Macovski A. Fast spiral coronary artery imaging. Magn Reson Med. 1992;28: Manning WJ, Li W, Edelman RR. A preliminary report comparing magnetic resonance coronary angiography with conventional angiography. N Engl J Med. 1993;328: Li D, Paschal CB, Haacke EM, Adler LP. Coronary arteries: threedimensional MR imaging with fat saturation and magnetization transfer contrast. Radiology. 1993;187: Pennell DJ, Keegan J, Firmin DN, Gatehouse PD, Underwood SR, Longmore DB. Magnetic resonance imaging of coronary arteries: technique and preliminary results. Br Heart J. 1993;70: Post JC, van Rossum AC, Hofman MB, Valk J, Visser CA. Protocol for two-dimensional magnetic resonance coronary angiography studied in three-dimensional magnetic resonance data sets. Am Heart J. 1995;130: Brittain JH, Hu BS, Wright GA, Meyer CH, Macovski A, Nishimura DG. Coronary angiography with magnetization-prepared T2 contrast. Magn Reson Med. 1995;33: Pennell DJ, Bogren HG, Keegan J, Firmin DN, Underwood SR. Assessment of coronary artery stenosis by magnetic resonance imaging. Heart. 1996;75: Oshinski JN, Hofland L, Mukundan S, Jr, Dixon WT, Parks WJ, Pettigrew RI. Two-dimensional coronary MR angiography without breath holding. Radiology. 1996;201: Ehman RL, Felmlee JP. Adaptive technique for high-definition MR imaging of moving structures. Radiology. 1989;173: 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: Wang Y, Rossman PJ, Grimm RC, Riederer SJ, Ehman RL. Navigatorecho-based real-time respiratory gating and triggering for reduction of respiration effects in three-dimensional coronary MR angiography. Radiology. 1996;198: Li D, Kaushikkar S, Haacke EM, Woodard PK, Dhawale PJ, Kroeker RM, Laub G, Kuginuki Y, Gutierrez FR. Coronary arteries: threedimensional MR imaging with retrospective respiratory gating. Radiology. 1996;201: McConnell MV, Khasgiwala VC, Savord BJ, Chen MH, Chuang ML, Edelman RR, Manning WJ. Comparison of respiratory suppression methods and navigator locations for MR coronary angiography. AJR Am J Roentgenol. 1997;168: Danias PG, McConnell MV, Khasgiwala VC, Chuang ML, Edelman RR, Manning WJ. Prospective navigator correction of image position for coronary MR angiography. Radiology. 1997;203:

10 3148 T2-Weighted, Free-Breathing, 3D Coronary MRA 16. Stuber M, Botnar RM, Danias PG, Kissinger KV, Manning WJ. Submillimeter 3D coronary MRA using real-time navigator correction: comparison of navigator locations. Radiology. In press. 17. Wang Y, Grimm RC, Rossman PJ, Debbins JP, Riederer SJ, Ehman RL. 3D coronary MR angiography in multiple breath-holds using a respiratory feedback monitor. Magn Reson Med. 1995;34: Goldfarb JW, Edelman RR. Coronary arteries: breath-hold, gadoliniumenhanced, three-dimensional MR angiography. Radiology. 1998;206: Harrison R, Bronskill MJ, Henkelman RM. Magnetization transfer and T2 relaxation components in tissue. Magn Reson Med. 1995;33: Gomori JM, Grossman RI, Yu-Ip C, Asakura T. NMR relaxation times of blood: dependence on field strength, oxidation state, and cell integrity. J Comput Assist Tomogr. 1987;11: Bloch F, Hansen WW, Packard M. Nuclear induction. Physiol Rev. 1946;70: Bottomley PA, Hardy CJ, Argersinger RE, Allen-Moore G. A review of 1H nuclear magnetic resonance relaxation in pathology: are T1 and T2 diagnostic? Med Phys. 1987;14: McConnell MV, Khasgiwala VC, Savord BJ, Chen MH, Chuang ML, Edelman RR, Manning WJ. Prospective adaptive navigator correction for breath-hold MR coronary angiography. Magn Reson Med. 1997;37: Li D, Dolan RP, Walovitch RC, Lauffer RB. Three-dimensional MRI of coronary arteries using an intravascular contrast agent. Magn Reson Med. 1998;39: Hofman MBM, Henson RE, Kovacs SJ, Fischer SE, Lauffer RB, Adzamli K, De Becker J, Wickline SA, Lorenz CH. Blood pool agent strongly improves 3D magnetic resonance coronary angiography using an inversion pre-pulse. Magn Reson Med. 1999;41: 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: Deriche R. Fast algorithms for low-level vision. IEEE Trans PAMI. 1990;12:78 87.