Blood Pool Agent Strongly Improves 3D Magnetic Resonance Coronary Angiography Using an Inversion Pre-Pulse

Transcription

1 Magnetic Resonance in Medicine 41: (1999) Blood Pool Agent Strongly Improves 3D Magnetic Resonance Coronary Angiography Using an Inversion Pre-Pulse Mark B. M. Hofman, 1 * Robert E. Henson, 1 Sándor J. Kovács, 1 Stefan E. Fischer, 1,2 Randall B. Lauffer, 2 Kofi Adzamli, 3 Jan De Becker, 4 Samuel A. Wickline, 1 and Christine H. Lorenz 1 * The ability of a blood pool contrast agent to enhance MR coronary angiography was defined. The proximal coronary vessels of pigs were imaged before and after administration of Gd-DTPA bound covalently to bovine serum albumin (0.2 mmol/ kg). The contrast agent resulted in a reduction of the blood T 1 value to 33 5 msec, as determined in vivo with a Look-Locker technique. Both 2D and 3D imaging techniques were performed. An inversion pulse suppressed the signal of nonblood tissue postcontrast. After contrast agent administration, in the 3D data set the signal-to-noise ratio (SNR) of blood and contrast-tonoise ratio (CNR) of blood to myocardium were improved by factors of and 15 8, respectively (P F 0.05). Postcontrast, the 3D acquisition was superior to the 2D technique in terms of spatial resolution, SNR of blood, and CNR of blood to myocardium. The high contrast of the 3D data set allowed for direct and rapid display of coronary arteries using a closest vessel projection. Magn Reson Med 41: , Wiley-Liss, Inc. Key words: magnetic resonance imaging; blood pool contrast agent; coronary angiography; swine INTRODUCTION The first clinical comparison of MR coronary angiography with conventional X-ray contrast angiography for detection of coronary artery lesions with more than 50% obstruction exhibited a sensitivity and specificity of 90% and 92%, respectively (1). Subsequent studies revealed lower values, in the range of 60 80% sensitivity (2, 3), indicating that the technique requires improvement before it can achieve widespread clinical application. The previous studies each used a 2D breath-hold technique (4, 5). Furthermore, improved respiratory gating using navigator echoes to monitor the respiratory motion in real time now permit imaging strategies without breathholding (6). Oshinski et al. showed that 2D imaging techniques manifested no significant difference in image quality between the breath-hold and respiratory gating approaches (7). The use of respiratory gating has opened 1 Center for Cardiovascular MR, Cardiovascular Division, Barnes-Jewish Hospital, St. Louis, Missouri. 2 EPIX Medical Inc., Cambridge, Massachusetts. 3 Mallinckrodt Inc., St. Louis, Missouri. 4 Philips Medical Systems, Best, the Netherlands. *Correspondence to: Mark B. M. Hofman, Ph.D., or Christine H. Lorenz, Ph.D., Center for Cardiovascular Magnetic Resonance, Cardiovascular Division, Barnes-Jewish Hospital, 216 South Kingshighway Blvd., St.Louis, MO chl@ccmr.wustl.edu Received 14 October 1997; revised 6 May 1998; accepted 9 July Wiley-Liss, Inc. 360 the field for 3D coronary MR imaging techniques, allowing a 3D data set to be obtained in a single acquisition (6,8). However, the 3D imaging technique suffers from a lower contrast-to-noise ratio (CNR) of blood to surrounding tissue compared with the 2D technique, owing to a lower spin refreshment by flow (6), which results in reduced sensitivity and specificity for detection of coronary artery disease (9). On the other hand, when compared with the 2D technique, 3D techniques allow higher spatial resolution and are less operator dependent. By applying thin-slab 3D acquisitions, the inflow signal enhancement can be improved (10). By using thin slabs, however, the slab positioning is more critical, and the contrast between blood and myocardium is still limited. A significant improvement of MR coronary angiography is expected with the use of a blood pool MR contrast agent. Such an agent shortens the T 1 relaxation time of blood, thereby increasing the MR signal of blood. These agents are administered intravenously and thus are minimally invasive. At present, the only clinically approved contrast agents for MRI that enhance the blood signal are small gadolinium chelates, for example, Gd-DTPA. However, current formulations of these agents diffuse rapidly from the vascular space, resulting in almost no improvement of contrast between blood and myocardium for coronary angiography except for approximately the first 40 sec after injection, which can be considered an intravascular phase. One study has reported that MR angiography of the aorta and its major branch vessels is most successful when performed during this intravascular phase (11). In the case of coronary angiography, because of the necessary cardiac gating due to the extensive motion of the coronary arteries within the cardiac cycle, the effective duty cycle of MR coronary angiography is limited to about 10% of the cardiac cycle. Therefore, routine MR coronary angiography, with its requirement for high spatial resolution and extensive spatial coverage, remains a challenge to accomplish in such a short period of time. Accordingly, we characterized the enhancement of MR coronary angiography with the use of a blood pool MR contrast agent, Gd-DTPA bound covalently to bovine serum albumin, or BSA-(Gd-DTPA) n (12, 13), which remains in the vascular space for an extended period of time. We sought to demonstrate that a 3D imaging technique would be feasible for MR coronary angiography with such an agent, thus providing a high-resolution technique that is minimally operator dependent.

2 MR Coronary Angiography with Blood Pool Agent 361 METHODS Animal Preparation Eight farm pigs weighing kg were studied. Anesthesia was induced with 4.4 mg/kg tiletamine, 2.2 mg/kg zolazepam, and 2.2 mg/kg ketamine, delivered intramuscularly. Central venous access was obtained through an indwelling double-lumen femoral catheter. Sedation was maintained with propafol ( mg/hr i.v.). The animals were ventilated at a rate of min 1. Initial administrations of BSA resulted in a urticarial skin reaction. To limit potential hypersensitivity reactions to the agent, subsequent animals were pretreated with diphenhydramine and prednisone. The average heart rate was 89 7 beats/min. In one animal the left circumflex coronary artery was ligated surgically as part of another protocol. ECG electrodes were attached to the chest for scan synchronization. A 20-cm-diameter radio frequency (RF) receiver coil was placed over the chest to optimize the signal-tonoise ratio (SNR). MR imaging was performed on a 1.5- tesla whole-body system (Gyroscan S15 ACS-NT, Philips Medical Systems). The Washington University Animal Studies Committee approved the protocol. Angiography Protocols Before contrast delivery, MR coronary angiography was performed with both 2D and 3D techniques. The 2D technique was an ECG-triggered fast gradient echo method with flow compensation (TR of 12 msec, TE of 4.4 msec, of 15 60, 110-Hz bandwidth per pixel) (4). Eight phaseencoding steps were performed for each heart cycle, with data acquisition in mid-diastole. Fat suppression was achieved by a fat frequency-selective RF pulse. The image resolution was 1 1 4mm 3, and the field of view was mm 2. The number of signal averages (NSA) was set to two. The acquisition was respiratory gated using a navigator image from the diaphragm (gating efficiency approximately 30%) (6, 7). Typically, 20 transverse slices overlapping by 1 mm were acquired in 27 min, covering the proximal portions of the coronary arteries. The pulse sequence of the 3D acquisition was similar to the 2D experiment except that the TR was 14.3 msec, the NSA was equal to one, and the image resolution was 1 1 2mm 3. The excitation angle varied from 16 to 23 to optimize the SNR of blood with a constant signal over k-space (14). The 3D transverse slab had a thickness of 6 cm, covering the proximal coronary arteries. The typical acquisition time was 20 min. BSA-(Gd-DTPA) n was delivered i.v. at a dose of 0.2 mmol/kg (gadolinium) for six animals and at a dose of 0.15 and 0.17 mmol/kg for the remaining two studies. The BSA concentration in the agent was mm, assuming a molar absorption at 280 nm of 40 mm -1 cm -1. The gadolinium concentration was determined in vitro using a spectrometer at 20 MHz, assuming an R1 value of 15 mm -1 s -1 (13). There was an average of 25 gadolinium ions per BSA molecule (range for the different batches). The average volume of the administered agent was ml, depending on the pig s weight and gadolinium concentration. The agent was administered over a period of about 20 to 40 min, and the agent was split to obtain T 1 measurements at different doses of the agent. Imaging was started 5 min after completion of the injection. Following contrast administration, 2D MR coronary angiography was performed on a small number of slices around the origin of the left coronary arteries, and then a 3D acquisition was performed. The 2D and 3D techniques differed with respect to the precontrast studies in three aspects. First, the MR signal of the myocardium and other tissues was suppressed by an RF inversion pulse before the imaging pulses. The delay between the inversion pulse and the imaging pulse at the center of k-space was set such that the myocardial signal was nulled (T msec). The delay was calculated using the measured myocardium T 1 value and taking into consideration the residual saturation from previous heartbeats. Second, the 3D acquisition was performed with an isotropic spatial resolution of mm 3, with a slab thickness of 6 cm, whereas the spatial resolution didn t change for the 2D acquisition. Third, the variable excitation angle optimized for the low T 1 value of blood ranged from approximately 31 to 54 for the 3D technique and from 29 to 50 for the 2D technique. The exact excitation angles were calculated for each study using the in vivo determined blood T 1 value such that the signal of blood was maximized and constant over the eight k-space steps, assuming a spoiled gradient echo (14). Owing to the short T 1 relaxation time of blood, no spin refreshment by flow was assumed. In five animals an extra 3D acquisition was taken, using a steady-state imaging technique to obtain a normal T 1 - weighted image. Instead of the inversion pre-pulse and fat saturation pulse, 22 RF pre-pulses, equal to the imaging RF pulses, were performed to create a steady state in the magnetization. The fixed excitation angle was set to 50, the Ernst angle for blood. In a theoretical calculation for a spoiled GE sequence, this angle yielded the best CNR for the expected T 1 values of blood, fat, and myocardium. Relaxation Time Measurements T 1 values of blood and myocardium were measured in vivo using a Look-Locker (LL) technique (15, 16). With this technique, the relaxation curve was sampled, after an initial nonselective inversion RF pulse, with a series of 40 imaging RF pulses at a time interval of 10 msec each. The imaging RF excitation angle was small (5 ), to limit signal variation induced by time-of-flight effects. The acquisition was triggered to every other R-wave, resulting in a TR of approximately 1.3 sec. The measurement was performed at the midventricular level. The T 1 value was determined by a two-parameter nonlinear least-squares fit, taking the imaging pulses into consideration (16). The T 1 measurements were performed 5 min after contrast administration, and in six of eight studies additional measurements were taken after 1 hr. The total contrast volume was split, and T 1 measurements were obtained at different stages of agent injection. The TR and the sample interval for the LL acquisitions at lower doses of the contrast agent were increased to maintain coverage of the full blood relaxation curve. In four animals, the T 2 * of blood was determined at a dose of 0.2 mmol/kg (gadolinium) using a gradient echo pulse sequence with a series of eight different echo times (TE of 5 to 20 msec), obtained in an interleaved fashion

3 362 Hofman et al. (17). This acquisition was gated to the respiration and triggered on the ECG to mid-diastole to limit motion artifacts. Signal to Noise/Contrast to Noise From the MR coronary angiographic images, the SNR of blood was calculated as times the signal in the proximal left anterior descending (LAD) coronary artery divided by the standard deviation in the background [SD(B)]. The CNR was determined as S/SD(B), where S is the difference in signal magnitude between blood and myocardium or blood and fat. The fat signal was determined at the subcutaneous layer, since the amount of epicardial fat in these pigs was small. The myocardial signal was measured in the left ventricle anterior to the outflow tract. Because of the use of a surface coil, there was variation in the absolute signal from study to study. Therefore, ratios of SNR and CNR before and after contrast administration (i.e., SNR post /SNR pre ) were calculated and averaged over the different animals. These ratios were tested for a significant difference from one, using the two-sided Student s t test. For these ratios, the SNR and CNR values of the postcontrast 3D acquisitions were corrected to a slab with 30 2-mm slices, equal to the precontrast 3D acquisitions. Data are presented as the mean SD, and differences were considered significant at P Theoretical SNR Calculations To characterize the blood SNR contrast-dose behavior, a theoretical calculation of the relative blood SNR was performed (14). The same 3D spoiled GE pulse sequence (TR of 14.3 msec, TE of 4.4 msec) was assumed, with eight excitations per heartbeat and a heart rate of 89 beats/min. A thick-slab 3D acquisition was considered, and therefore no signal refreshment was assumed. At each dose, the T 1 and T 2 * of blood and the T 1 of myocardium were determined, assuming a linear relationship between relaxation rate and contrast dose. At zero dose, for blood a T 1 of 1.2 sec and T 2 * of 241 msec and for myocardium a T 1 of 1 sec were specified (18, 17). For the agent, dose-dependent changes in relaxation rate from the in vivo determined values were applied. At each dose, the optimum flip angle sequence was determined, and the MR signal was calculated for the situation with and without a myocardial suppression inversion pre-pulse. For comparison with the steadystate technique, the relative blood signal was calculated with the excitation angle equal to the Ernst angle. The 3D coverage per unit of time of 3D pulse sequence is limited; in this study 1 mm/min was obtained. We made a theoretical calculation of the dose dependency of this 3D acquisition technique in comparison with two faster 3D acquisition techniques: a turbo GE and a segmented echo planar imaging (EPI) technique. For the turbo GE technique, we used the parameters described by Goldfarb et al.: a TR of 3 msec, a TE of 1.2 msec, and a receiver bandwidth per pixel ( f) of 1064 Hz (19). For the segmented EPI, a pulse sequence as described by Wielopolski et al. was assumed: a TR of 8 msec, a TE of 3.5 msec, f of 1280 Hz, and an EPI factor of 6 (20, 21). For the EPI sequence, the echo time was assumed to be 3.5 msec, around the middle of the echo train, to take into consideration signal saturation for short T 2 * values. For all three techniques, an acquisition window within the cardiac cycle of 115 msec and a heart rate of 60 beats/min were assumed. The T 1 and T 2 * values of blood and myocardium at different doses were applied, and the optimal flip angle sequence was calculated for each technique. The calculated MR signal was corrected for differences in receiver bandwidth and number of echoes per unit of time relative to the 3D sequence, the fast GE technique, used in this study (TR, 14.3 msec; TE, 4.4 msec; f, 110 Hz). RESULTS Relaxation Times The administration of the contrast agent BSA-(Gd-DTPA) n at a dose of 0.2 mmol/kg (gadolinium) reduced the T 1 relaxation time of blood to a value of 33 5 msec and reduced the myocardial T 1 value to msec. At this dose, the in vivo T 2 * value of blood was 15 2 msec. There was no significant difference in T 1 of blood and myocardium directly after injection and an hour later, showing that the agent remained intravascular for a prolonged period of time. The blood inverse T 1 relaxation time as a function of the contrast dose is shown in Fig. 1. Linear regression revealed a slope of sec 1 mmol 1 kg and sec 1 mmol 1 kg for blood and myocardium, respectively. For the T 2 * value of blood, a dose dependency of the inverse T 2 * relaxation time of 313 sec 1 mmol 1 kg was estimated, as determined using a precontrast literature value (17). These values were applied in the theoretical SNR calculation. Signal to Noise/Contrast to Noise Before administration of the blood pool agent, the 2D technique showed fair contrast, but differentiation between neighboring coronary veins and arteries was difficult owing to the limited spatial resolution, especially for the left side of the coronary tree. The coronary angiograms obtained with the 3D acquisition showed a very low FIG. 1. The inverse T 1 relaxation time of blood ( ) and myocardium ( ) as a function of the BSA-(Gd-DTPA) n dose, as determined in vivo. Error bars show one standard deviation. The plotted lines are the result of a linear regression (for blood T 1 : slope 170 9sec 1 mmol 1 kg, intercept sec 1, r 0.97; for myocardium T 1 : slope sec 1 mmol 1 kg, intercept sec 1, r 0.93).

4 MR Coronary Angiography with Blood Pool Agent 363 contrast of blood to myocardium, with a CNR ratio of approximately two (Fig. 2). The blood SNR and the CNR of blood to fat were not significantly different between the 2D a nd 3D techniques, whereas the CNR of blood to myocardium was significantly better in the 2D technique (see Fig. 2 and Table 1). After administration of BSA-(Gd-DTPA) n, the results obtained by both the 2D and 3D techniques improved with respect to the precontrast images. The SNR improved by a factor of (P 0.001) for the 3D technique, whereas it actually decreased slightly for the 2D technique (a factor of ). Both 2D and 3D techniques showed a dramatic improvement in contrast of blood to myocardium, as can be observed in Fig. 2 and Table 1. In the postcontrast images, high signal is obtained only from the blood pool and bone marrow, while the MR signal of all other tissues, such as epicardial fat, myocardium, and epicardial fluid, was suppressed. The signal of both coronary arteries and veins were enhanced. The CNR between blood and myocardium improved by a factor of for the 2D versus a factor of 15 8 for the 3D technique. After administration of the blood pool agent, the contrast between blood and myocardium or blood and fat was not significantly different between the 3D and the 2D techniques, with a higher CNR for the 3D technique due to the higher blood SNR. The inversion pulse applied after the contrast suppressed the signal of myocardium well, and in the 3D data the myocardial SNR postcontrast was reduced to % of the precontrast value. The 3D acquisition technique with pre-pulses for fat and myocardial suppression was superior to the steady-state imaging technique. The SNR of blood was not significantly Table 1 SNR and CNR Ratios Pre- and Postcontrast Delivery SNR blood CNR bloodmyocardium CNR blood-fat 2D post/pre ** ** D post/pre ** 15 8* * 3D/2D pre ** D/2D post ** ** ** 3D post/2d pre ** ** ** Data are the mean standard deviation (*P value 0.05, **P value 0.01). For these ratios the SNR and CNR values of the postcontrast 3D acquisitions were corrected to a slab with 30 2-mm slices, equal to the precontrast 3D acquisitions. Only in the postcontrast acquisitions was an inversion pulse applied to suppress the myocardial signal. SNR, signal-to-noise ratio; CNR, contrast-to-noise ratio. different between the two techniques (ratio of ), but the CNR of both blood to myocardium and blood to fat were significantly higher using the pre-pulse technique, by factors of and , respectively. This difference in contrast was clearly visible in the images obtained. Theoretical SNR Calculation The calculated dose dependency of the relative MR signal of blood is presented in Fig. 3. The dose of 0.2 mmol/kg (gadolinium) used in this study appeared to be optimal for this pulse sequence. At zero dose, the highest blood signal was obtained without an inversion pre-pulse, but at a dose of 0.15 mmol/kg or higher no signal loss of blood is induced by the inversion pulse. The blood MR signal obtained with the steady-state technique was lower than that obtained using the variable FIG. 2. Axial images showing the aortic root plus a segment of the left anterior descending coronary artery and the right coronary artery in cross-section (arrows). All panels show the same anatomical slice with different imaging techniques, pre- and postcontrast. The slice thickness is not equal for the different images (4 mm for the 2D technique, 2 mm for the precontrast 3D acquisition, and 1 mm for the postcontrast 3D acquisition). FIG. 3. Theoretical dependency of the relative MR signal of blood on the contrast agent dose. A spoiled GE sequence (TR, 14.3 msec; TE, 4.4 msec) was assumed, and optimized excitation angles were determined. The MR signal is expressed relative to the signal in the case of no T 1 saturation and no T 2 * signal decay. The blood signal was calculated for three cases: without a myocardial saturation pre-pulse, with a myocardial saturation pre-pulse, and using a steady-state imaging technique. A second x-axis with blood T 1 and T 2 * values is shown for reference.

5 364 Hofman et al. excitation angle approach, but at higher doses this difference disappeared. Figure 4 shows the result of the theoretical comparison of SNR of blood per unit of time for three different 3D acquisition techniques: the fast GE used in this study, the turbo GE described by Goldfarb et al. (19), and the segmented EPI described by Wielopolski et al. (21). In terms of blood signal per unit of time, the fast GE sequence appears to be the optimal method of these three. However, compared with this fast GE sequence, the 3D coverage per unit of time is 5 and 10.5 times higher for the turbo GE and the segmented EPI, respectively. The optimum contrast dose is comparable for the fast GE and the EPI techniques, whereas it is about 6.5 times higher for the turbo GE technique owing to its short TR and TE values. Postprocessing The 3D postcontrast data set showed a high CNR ratio and an isotropic 1-mm spatial resolution, which substantially improved postprocessing. The maximum intensity projection (MIP) was inappropriate for these data sets, as shown in Fig. 5 (left panel). Large vascular structures such as heart chambers and large blood vessels project over the coronary arteries, making them invisible on the MIP image. Better visualization of the coronary arteries can be obtained with a closest vessel projection (CVP), using an algorithm similar to one proposed by Siebert et al. (22). This CVP, implemented on an EasyVision workstation (Philips Medical Systems), was applied on all postcontrast 3D data sets (Fig. 5, right panel). Before the CVP was applied, the tissue around the heart was removed manually by cutting out a cylinder in the 3D data set. To obtain a good differentiation of the coronary arteries in front of the heart chambers, the CVP threshold had to be set at 58 8% of the blood signal FIG. 5. Different projection techniques of the same 3D postcontrast data set. The left panel shows a maximum intensity projection, and the right panel shows a closest vessel projection in the same view. Projections were performed after removing tissue outside the heart. The proximal right coronary artery (RCA), left anterior descending (LAD) coronary artery, and great cardiac vein (GCV) can be seen. in the aortic root. This threshold permitted rapid visualization of the coronary arteries with limited user interaction. In some cases, the proximal right coronary artery (RCA) and left circumflex coronary artery (LCX) were obscured by overlapping parts of the left and right atria. In those cases, a local CVP of a slab around this vessel segment was used to reveal the artery. This slab position was less sensitive to positioning compared with a local slab using MIP. Another method of image processing improved with the use of a blood pool agent is curved planar reformatting (CPR) (23). With this method, an image is reformatted along a trajectory that runs through the vessel of interest. Figure 6 shows an example of such a CPR image obtained from a 3D postcontrast data set with a path that was manually placed along the RCA and the LAD coronary artery. DISCUSSION The blood pool contrast agent BSA-(Gd-DTPA) n at a dose of 0.2 mmol/kg (gadolinium) resulted in a very short value for the T 1 relaxation times of whole blood, similar to those found previously (24). The observed T 1 values in blood and myocardium are consistent with a true blood pool distribution. This dramatic reduction of the T 1 of blood resulted in two significant improvements in MR coronary angiography: an increase in SNR using the 3D technique and an increase in the contrast of blood to myocardium in both 2D and 3D techniques. FIG. 4. Theoretical dependency of the relative MR signal of blood on the contrast agent dose for three different acquisition techniques: a fast GE sequence (TR, 14.3 msec; TE, 4.4 msec; f, 110 Hz) used in this study; turbo GE (TR, 3 msec; TE, 1.2 msec; f, 1064 Hz), described by Goldfarb et al. (19); and a segmented EPI sequence (TR, 8 msec; TE, 3.5 msec; f, 1280 Hz, EPI factor, 6) described by Wielopolski et al. (21). The straight lines show the signal behavior for a sequence with a myocardial suppression inversion pulse, whereas the dotted lines represent the case without such a pre-pulse. The MR signal is expressed relative to the signal in case of no T 1 saturation and no T 2 * signal decay for the fast GE and for the other two techniques scaled to correct for differences in receiver bandwidth ( f) and number of echoes read per cardiac cycle. FIG. 6. Curved planar reformatted image of a path along the right coronary artery (RCA) through the aortic root (AR), and along the left anterior descending (LAD) coronary artery, obtained from a postcontrast 3D data set.

6 MR Coronary Angiography with Blood Pool Agent 365 Signal to Noise/Contrast to Noise The SNR of blood in the 3D image increased, whereas no improvement was found using the 2D technique. This can be explained by the precontrast difference in MR signal saturation due to lower flow-induced spin refreshment in the thick 3D slab compared with the thin 2D image slice. Postcontrast, the short T 1 results in minimum saturation, independent of flow effects. A similar increase in SNR of the 3D technique using an iron oxide based blood pool agent has been reported previously (25). Using the 2D technique, the actual SNR of blood in the LAD coronary artery decreased according to our data. This may have resulted from the fact that for the postcontrast excitation angle calculation we did not assume significant spin refreshment with such a short T 1, although this assumption may not be accurate based on our results. In the aorta, with a lower amount of through-plane flow at mid-diastole, no difference in blood SNR with the 2D technique was found between pre- and postcontrast angiograms. There is no improvement in blood SNR by the blood pool agent for a 2D technique, owing to the high inflow refreshment in the coronary artery. However, one might expect some improvement in blood SNR in the 2D technique for situations where a long arterial segment is located in-plane or when the flow is impaired owing to obstructive vessel lesions. The amount of blood SNR improvement observed in coronary 3D acquisitions likely will be less than that obtained with 3D acquisitions in other parts of the body. Before contrast administration, the amount of spin saturation within the blood pool is smaller for coronary angiography owing to the low duty cycle (about 10%) of the MR acquisition, a consequence of cardiac gating in middiastole. The MR blood signal recovers in the remaining time, which results in a higher blood signal for this kind of gated acquisition in the situation with no contrast agent compared with the blood signal with a nongated (full duty cycle) 3D acquisition. For a blood T 1 of 1200 and a 3D acquisition as applied in this study, the blood SNR improvement due to this recovery time is about a factor of 2.4. This effect can be observed in Fig. 3 at zero dose, where the signal by the steady-state technique (with no signal recovery time) is 2.4 times lower than the signal for the no pre-pulse technique. The second major improvement obtained with a blood pool agent is better contrast between blood and myocardium. Without a blood pool agent, the contrast in the 3D acquisition is mostly proton density weighted with additional fat suppression, revealing a low contrast between blood and myocardium. The contrast improvement is obtained by the inversion pre-pulse in combination with the short blood T 1 relaxation time to suppress the myocardium selectively. Others have used an inversion pre-pulse for coronary angiography, although previous reported methods were based upon inflow effects of the blood (26, 27). With the use of an intravascular contrast agent, this dependency on flow effects is removed. A similar application of pre-pulsing for 2D MR projection imaging after contrast administration has been published for noncoronary applications (28). The use of this pre-pulse has no impact on the overall scan duration for coronary imaging because of the small acquisition window within the cardiac cycle. A high dose [ 0.15 mmol/kg (gadolinium)] is necessary to suppress the myocardium without reducing the blood SNR, as shown in Fig. 3. At lower doses, the suppression of the myocardium has an impact on image quality in terms of reduction of blood SNR. Two other techniques have been mentioned in the literature as suppressing the MR signal of myocardium: magnetization transfer saturation (MTS), which results in a reduction of the myocardial signal of 30% (8), and T 2 magnetization preparation, which improves the CNR at the cost of a significant reduction of blood SNR value (29). Both these techniques are less efficient in selective suppression of the myocardial signal compared with the combination of the contrast agent and the pre-inversion pulse used in this study, which resulted in a reduction of the myocardial signal by 74%. After contrast administration, the amount of MTS is expected to decrease owing to a shorter myocardial T 1 relaxation time (30). T 2 magnetization preparation is not expected to be as effective, owing to the simultaneous reduction of the blood T 2 value by the agent. The 3D acquisition postcontrast had a higher spatial resolution, but the data in Table 1 are corrected for this difference. The excitation angles pre- and postcontrast were different, but this is a reasonable compensation to account for the differences in T 1 relaxation times. The main difference between the pre- and postcontrast acquisitions was the application of the inversion pulse to suppress the myocardial signal. When applied precontrast, this inversion pulse would be expected to suppress the MR signal of blood, rendering the imaging technique useless. This myocardial pre-pulse is feasible only postcontrast. Without the myocardial suppression, the myocardial signal postcontrast will also approximately double, according to theory. Therefore, to obtain an improvement in contrast of blood to myocardium, an inversion pre-pulse or other myocardial suppression technique is required. The improved SNR and CNR of the 3D technique postcontrast renders it superior to 2D imaging. The blood pool contrast agent allows for 3D imaging with an isotropic 1-mm resolution and an SNR similar to the value in the 2D technique, but with a through-plane spatial resolution four times better. This improved spatial resolution may be important for stenosis characterization. Another advantage of the 3D technique is reduced operator dependence due to the relatively thick imaging slab. The major limitation of the high-resolution 3D scanning is the increased scan duration for the same coronary artery coverage that can be attained with typical 2D methods. Image Postprocessing The improved CNR and the ability to obtain a 3D volume with a high isotropic spatial resolution allows for improved image postprocessing. This feature is critical, because MR coronary angiography will likely require postprocessing in a clinical setting to provide an overview of the arteries for grading stenoses. Interpretation of the images based on reviewing stacks of images is time-consuming and requires much experience. The curved planar reconstruction reduces the view of a tortuous vessel to one image, but generation of these reconstructions requires manual interaction (approximately min). MIP is currently the most widely employed method of MR angio-

7 366 Hofman et al. gram postprocessing for display of vessels in other parts of the body. The presence of many other vascular tissues in and around the heart compromises MIP for coronary visualization, because the relatively small coronary vessels will be obscured by the other, more prominent vascular structures. Removal of these other structures prevents this overprojection, but it currently requires time-consuming manual editing of the data set (31). A more robust approach is the closest vessel projection (CVP). With this technique, small vessels can be visualized in front of large vascular structures. The CVP is fast and not as user dependent as the manual segmentation techniques reported previously (32). The type of images produced by this technique could potentially be used to guide more elaborate segmentation of specific vessels to obtain absolute vessel diameters, because the rough segmentation performed within the CVP is still too crude for actual diameter calculations. A disadvantage of the contrast agent is that both arteries and veins are enhanced. However, coronary veins are also clearly visualized with the 2D precontrast acquisitions. The high spatial resolution that can be obtained in the 3D postcontrast acquisition potentially allows for better differentiation, by applying algorithms like vessel tracking. Another major improvement afforded by using a blood pool agent for MR coronary angiography is that the MR signal of blood no longer depends in any significant way on flow effects. This improvement is expected to result in more accurate stenosis grading, because overestimation of stenosis severity is a common problem with the current flow-dependent MR techniques that do not use a contrast agent (2). Choice of 3D Acquisition For the 3D technique used in this study, a standard fast GE pulse sequence with a low receiver bandwidth was applied to generate pulse sequences similar to those with the 2D technique. The 3D coverage per unit of time of this 3D pulse sequence is limited: in this study a rate of 1 mm/min was obtained. Faster 3D MR techniques for coronary angiography have been described recently, by applying segmented EPI or GE acquisitions with very short TR values (19 21). The theoretical comparison of SNR per unit of time in Fig. 4 shows that the fast GE used in this study appeared to be the optimal method among the three in terms of SNR; however, it has the lowest 3D coverage per unit of time. From this comparison, it seems that segmented EPI would represent the most promising technique. However, the sensitivity of this EPI technique to flow artifacts and signal variation over k-space due to short T 2 * values is higher. The ideal technique will depend on required SNR, volume coverage, and available scan time. Further studies will be necessary to determine which method is preferable for clinical use. Human Studies Because we studied the effect of a blood pool contrast agent in pig coronary vessels, the results cannot be immediately extrapolated to human beings for the following reasons. First, the proximal arteries of the pigs in this study are smaller than those of humans 2 3 mm in diameter as opposed to 3 5 mm. Second, there is less pericardial fat tissue around the coronary vessels in pigs, which increases the importance of myocardial suppression compared with fat suppression. The contrast agent used in this study, BSA-(Gd-DTPA) n, is not in development for human use, presumably because of safety concerns (immunogenicity, anaphylaxis, gadolinium retention, dosage) (33, 34). However, this agent clearly demonstrates the maximal potential of gadolinium-based blood pool agents for coronary angiography. At present, one small-molecule gadolinium-based agent, MS-325, is being employed in clinical trials as an angiographic agent. This agent is converted to a blood pool agent in vivo via reversible albumin binding (35, 36). In addition, an iron oxide based agent, AMI-227, has been described to have angiographic capabilities (25). It remains to be seen whether signal-decreasing effects (T 2 * effects) at higher concentrations of iron agents will limit the achievable signal enhancement. Conclusion The advantages of a blood pool agent for MR coronary angiography are short T 1 of blood, facilitating highresolution 3D imaging with less operator dependence than 2D imaging; optimal suppression of nonvascular tissue, facilitating postprocessing; and independence of the MR signal from flow effects, which may improve detection and absolute grading of stenoses. ACKNOWLEDGMENTS The authors thank Mr. David Guarraia and Mr. John S. Allen for assistance with the animal preparation and Dr. S. Nema and Mr. Randy M. Setser for assistance in the contrast agent preparation. This study was supported in part by the Barnes-Jewish Hospital Foundation; the Wolff Charitable Trust; Philips Medical Systems, Best, the Netherlands; Mallinckrodt, Inc., St. Louis Missouri; and EPIX Medical Inc., Cambridge Massachusetts. REFERENCES 1. Manning WJ, Li W, Edelman RR. A preliminary report comparing magnetic resonance coronary angiography with conventional angiography. N Engl J Med 1993;328: Duerinckx AJ, Urman MK. Two-dimensional coronary MR angiography: analysis of initial clinical results. Radiology 1994;193: Post JC, van Rossum AC, Hofman MBM, Valk J, Visser CA. Clinical utility of two-dimensional breath-hold MR angiography in coronary artery disease. In: Proceedings of the SMR/ESMRMB Joint Meeting, Nice, p Edelman RR, Manning WJ, Burstein D, Paulin S. Coronary arteries: breath-hold MR angiography. Radiology 1991;181: 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: Hofman MBM, Paschal CB, Li D, Haacke EM, van Rossum AC, Sprenger M. MRI of coronary arteries: 2D breath-hold vs 3D respiratory-gated acquisition. J Comput Assist Tomogr 1995;19: Oshinski JN, Hofland L, Mukundan S, Dixon WT, Parks WJ, Pettigrew RI. Respiratory gated coronary magnetic resonance angiography compares favorably with breath-hold imaging. In: Proceedings of the SMR/ESMRMB Joint Meeting, Nice, p. 22.

8 MR Coronary Angiography with Blood Pool Agent Li D, Paschal CB, Haacke EM, Adler LP. Coronary arteries: threedimensional MR imaging with fat saturation and magnetization transfer contrast. Radiology 1993;187: Post JC, van Rossum AC, Hofman MBM, Valk J, Visser CA. Threedimensional respiratory-gated MR angiography of coronary arteries: comparison with conventional coronary angiography. AJR Am J Roentgenol. 1996;166: Li D, Kaushikkar S, Haacke EM, et al. Coronary arteries: threedimensional MR imaging with retrospective respiratory gating. Radiology 1996;201: Prince MR. Gadolinium-enhanced MR aortography. Radiology 1994;191: Lauffer RB, Brady TJ. Preparation and water relaxation properties of proteins labeled with paramagnetic metal chelates. Magn Reson Imaging 1985;3: Ogan MD, Schmiedl U, Moseley ME, Grodd W, Paajanen H, Brasch, RC. Albumin labeled with Gd-DTPA An intravascular contrast-enhancing agent for magnetic resonance blood pool imaging: preparation and characterization. Invest Radiol 1987;22: Stehling MK. Improved signal in snapshot FLASH by variable flip angles. Magn Reson Imaging 1992;10: Look DC, Locker DR. Time saving in measurement of NMR and EPR relaxation times. Rev Sci Instrum. 1970;41: Brix G, Schad LR, Deimling M, Lorenz WJ. Fast and precise T 1 imaging using a TOMROP sequence. Magn Reson Imaging 1990;8: Chien D, Levin DL, Anderson CM. MR gradient echo imaging of intravascular blood oxygenation: T 2 * determination in the presence of flow. Magn Reson Med 1994;32: Tadamura E, Hatabu H, Li W, Prasad PV, Edelman RR. Effect of oxygen inhalation on relaxation times in various tissues. J.Magn Reson Imaging 1997;7: Goldfarb JW, Li W, Griswold MA, Prasad PV, Edelman RR. Contrast enhanced breath-hold 3D coronary magnetic resonance angiography. Proceedings of the International Society for Magnetic Resonance in Medicine, Fifth Scientific Meeting and Exhibition; 1997; New York. Berkeley, CA: International Society for Magnetic Resonance in Medicine. 20. Wielopolski PA, Manning WJ, Edelman RR. Single breath-hold volumetric imaging of the heart using magnetization-prepared 3-dimensional segmented echo planar imaging. J Magn Reson Imaging 1995;4: Wielopolski PA, de Feyter P, de Jaegere P, de Bruin HG, Oudkerk M. Single breathhold three-dimensional MR coronary angiography. Proceedings of the International Society for Magnetic Resonance in Medicine, Fourth Scientific Meeting and Exhibition; 1996; Vancouver. Berkeley, CA: International Society for Magnetic Resonance in Medicine. 22. Siebert JE, Rosenbaum TL, Pernicone JR. Automated segmentation and presentation algorithms for 3D MR angiography. Book of Abstracts, Society of Magnetic Resonance in Medicine, Tenth Annual Scientific Meeting and Exhibition; 1991; San Francisco. Berkeley, CA: Society of Magnetic Resonance in Medicine. 23. Scheidegger MB, Stuber M, Boesiger P. Coronary artery display from 3D MR data sets with curved planar reformatting ( vessel tracking reformat ). Proceedings of the International Society for Magnetic Resonance in Medicine, Fifth Scientific Meeting and Exhibition; 1997; New York. Berkeley, CA: International Society for Magnetic Resonance in Medicine. 24. Moseley ME, White DL, Wang SC, et al. Vascular mapping using albumin-(gd-dtpa), an intravascular MR contrast agent, and projection MR imaging. J Comput Assist Tomogr 1989;13: Stillman AE, Wilke N, Li D, Haacke M, McLachlan S. Ultrasmall superparamagnetic iron oxide to enhance MRA of the renal and coronary arteries: studies in human patients. J Comput Assist Tomogr 1996;20, Wang SJ, Hu BS, Macovski A, Nishimura DG. Coronary angiography using fast selective inversion recovery. Magn Reson Med 1991;18: Atalar E, McVeigh E, Zerhouni E. Coronary artery blood flow imaging using pre-saturation segmentation k-space data acquisition. Proceedings of the Society Magnetic Resonance in Medicine Twelfth Annual Scientific Meeting; 1993; New York. Berkeley, CA: Society of Magnetic Resonance in Medicine. 28. Marchal G, Bosmans H, van Hecke P, Jiang YB, Aerts P, Bauer H. Experimental Gd-DTPA polylysine enhanced MR angiography: sequence optimization. J Comput Assist Tomogr 1991;15: Brittain JH, Hu BS, Wright GA, Meyer CH, Macovski A, Nishimura DG. Coronary angiography with magnetization-prepared T2 contrast. Magn Reson Med 1995;33: Jones RA, Haraldseth O, Schjott J, et al. Effect of Gd-DTPA-BMA on magnetization transfer: application to rapid imaging of cardiac ischemia. J Magn Reson Imaging 1993;3: Edelman RR, Manning WJ, Pearlman J, Li W. Human coronary arteries: projection angiograms reconstructed from breath-hold two-dimensional MR images. Radiology 1993;187: Doyle M, Scheidegger MB, de Graaf RG, Vermeulen J, Pohost GM. Coronary artery imaging in multiple 1-sec breath holds. Magn Reson Imaging 1993;11: Baxter AB, Melnikoff S, Stites DP, Brasch RC. Immunogenicity of gadolinium-based contrast agents for magnetic resonance imaging induction and characterization of antibodies in animals. Invest Radiol 1991;26: Moseley ME, White DL, Wang SC, et al. Vascular mapping using albumin-(gd-dtpa), an intravascular MR contrast agent, and projection MR imaging. J Comput Assist Tomogr 1989;13: Lauffer RB, Parmelee DJ, Ouellet HS, et al. MS-325: a small-molecule vascular imaging agent for magnetic resonance imaging. Acad Radiol.1996;3:S356 S Grist TM, Korosec FR, Peters DC, Walovitch RC, Yucel EK, Mistrettal CA Steady state and dynamic MR angiographic imaging with MS-325: initial experience in humans. Proceedings of the International Society for Magnetic Resonance in Medicine, Fifth Scientific Meeting and Exhibition; 1997; New York. Berkeley, CA: International Society for Magnetic Resonance in Medicine.