Radiol Clin N Am 40 (2002) 921 – 951
Technical developments in MR angiography Timothy J. Carroll, PhDa, Thomas M. Grist, MDb,* a
Department of Medical Physics, University of Wisconsin, Madison, WI 53792, USA b Department of Radiology, University of Wisconsin, Madison, WI 53792, USA
Catheter-based multistation digital subtraction angiography (DSA) is regarded as the gold standard for the evaluation of vascular disease. The high incidence of morbidity associated with radiographic DSA has placed emphasis on developing noninvasive methods of determining whether surgical intervention is indicated. Invasiveness, costs, and morbidity indicate the need for less invasive diagnostic tools. Magnetic resonance angiography (MRA) is emerging as an adjunct and, in some cases, an alternative to DSA. The growth of MRA is the result of improvements in image acquisition strategies and magnetic resonance (MR) contrast agents. Contrast agent-enhanced MRA (CE MRA) examinations result in images with highspatial resolution and fewer artifacts and require less time to acquire than noncontrast methods. This article reports on recent technical developments in this emergent technology. Many of these acquisition strategies attempt to optimize the use of the first pass of the bolus of contrast agent. To do so, they coordinate image acquisition with arrival of the contrast agent in the targeted anatomy. This may be accomplished given prior knowledge of the transit time of the bolus from injection site to the image volume or by triggering the acquisition when arrival of the contrast is detected. Triggered acquisitions are performed either by pulse sequences, which automatically detect the inflow of contrast agent, or by fluoroscopic acquisitions, which are triggered by an operator. CE MRA acquisitions rely on heavily T1 weighted, rapid images acquisition. The simplest way to speed up an acquisition is to reduce the repetition (TR). Recently, novel image reconstructions have been developed in which multicoil arrays share data to further * Corresponding author. E-mail address:
[email protected] (T.M. Grist).
reduce acquisition time. In addition, many groups are successfully using non-Cartesian k-space – based data acquisitions that sample k-space along spiral or radial trajectories. This article discusses some of the basic physics principles necessary to understand the limitations and challenges of CE MRA. A short technical description of some of the more recent technical advancements with several illustrative examples follows. The last section of this article describes currently available contrast agents and some agents that are in development or under review by the United States Food and Drug Administration.
Basics of MRA Localization of voxel signal in MR images is achieved by encoding the phase and frequency of spinning protons using magnetic field gradients. Therefore, the raw data acquired by MR scanners are actually a spatial-frequency representation of the image, commonly referred to as the ‘‘k-space’’ of the image. Images are produced when the acquired k-space data is reconstructed using Fourier transformation. An important feature of the k-space representation of an image is that low-spatial frequencies (ie, the central phase encoding values) contain the information necessary to reconstruct a low-resolution image. A low-spatial resolution image reflects the overall image contrast. As higher spatial frequencies are included in a reconstructed image, finer detail of the image begins to emerge. Fig. 1 demonstrates the effect of how low- and high-spatial frequencies contribute to an image. Unfortunately, in MRI, spatial resolution comes at a cost. Generally, for a given acquisition time, as the
0033-8389/02/$ – see front matter D 2002, Elsevier Science (USA). All rights reserved. PII: S 0 0 3 3 - 8 3 8 9 ( 0 2 ) 0 0 0 2 9 - 5
922
T.J. Carroll, T.M. Grist / Radiol Clin N Am 40 (2002) 921–951
spatial resolution increases, the signal-to-noise ratio (SNR) of the image decreases. The decrease in SNR is a result, in part, of the decrease in the voxel size, which results in fewer spins contributing to the signal from the voxel. The SNR of high-spatial resolution images may be increased by increasing scan time, either by increasing the scan TR or signal averaging. Signal averaging increases SNR by repeatedly acquiring an image and averaging the signal intensity of each voxel in successive images. As a rule of thumb, for a fixed spatial resolution, SNR increases in proportion to the square root of the imaging time. For example, doubling the imaging time results in roughly a 40% increase in SNR. The converse is also true: acquiring an image in half the time will decrease SNR by 40% relative to a standard acquisition. Low SNR is a serious technical challenge for all rapidimage acquisitions.
Contrast mechanisms MRA may be performed using time-of-flight (TOF) techniques, which depend on blood flow for vessel contrast [1]. In TOF imaging, the signal from stationary tissue within the imaging volume is suppressed via application of multiple radio-frequency (RF) pulses. When short repetition time (TR) values are chosen, the longitudinal magnetization of the tissue is not given sufficient time to regrow, resulting in very low signal. Blood flowing into the image volume, however, does not experience the same RF excitations, and therefore appears bright (Fig. 2). TOF examinations suffer from several sources of artifactual signal loss [2]. In TOF, slow flowing may become saturated, leading to a spurious thinning of the imaged vessel. This potentially could lead to overestimation of the severity of a stenosis. In addition,
Fig. 1. Image formation in MRI k-space. Images of an abdominal aortic anerysm demonstrate the correspondence between spatial frequencies and image resolution. (A) The spatial frequency, or ‘‘k-space,’’ representation of the aorta is Fourier transformed to form an image (B). (C) If only the central k-space is used in image formation, the resulting image (D) is a low-resolution version of the original. The Fourier transformation of the high-spatial frequencies (E) provides image detail such as delineation of edges (F). (Courtesy of Oliver Wieben, PhD, and Frank Korosec, PhD, University of Wisconsin.)
T.J. Carroll, T.M. Grist / Radiol Clin N Am 40 (2002) 921–951
923
Fig. 1 (continued ).
slice misregistration as a result of patient motion can be problematic in TOF examinations that are acquired over several minutes. Lengthy examination times and a sensitivity to in-plane and retrograde flow artifacts have prompted development of CE MRA acquisitions. By exploiting the T1-shortening effect of paramagnetic contrast agents to depict vessels, CE MRA overcomes many of the limitations inherent to TOF MRA [3]. CE MRA commonly is performed with short TR (< 10 milliseconds) gradient-recalled echo sequences. Fig. 3 shows how longitudinal magnetization approaches its equilibrium value as multiple short-TR RF pulses are applied to the sample. The time available for signal regrowth between RF pulses is TR, so that the amount of regrowth decreases as TR is reduced, resulting in less longitudinal magnetization (M0). Because signal intensity in MR is proportional to M0, short TR provides T1 contrast by suppressing the signal from tissues with slow regrowth of the longitudinal magnetization (long T1). In CE MRA, these rapid image acquisitions are combined with an intravenous bolus injection of a
paramagnetic contrast agent that produces transient shortening of the T1 of blood. The use of short TR sequences also allows for the rapid acquisition of high-resolution images. The time to acquire a three-dimensional (3D) volume is given by Time = Ny Nz TR, where Ny is the number of phases-encoding values, Nz is the number of acquired slices, and TR is the repetition time. For example, an acquisition with 12 phase encoding values, 32 slice partitions, and a TR of 6.0 milliseconds can be acquired in a 24-second breath hold. The use of heavily T1-weighted sequences introduces the problem of the background signal from fat. The T1 of fat is approximately 270 milliseconds, and therefore appears bright in CE MRA images. Typically, a precontrast mask image is acquired and used to subtract off the fat signal. Mask mode subtraction is used to increase the contrast of vessels by removing background signal but is sensitive to patient motion. In abdominal imaging, mask and contrast-enhanced images both must be acquired in a breath hold. For images of the peripheral vasculature, care must be taken to ensure that the
924
T.J. Carroll, T.M. Grist / Radiol Clin N Am 40 (2002) 921–951
Fig. 1 (continued ).
patient remains still between the acquisition of the mask and the contrast-enhanced image. Image contrast in CE MRA exams can be affected by several acquisition parameters. The signal enhancement provided by a bolus injection of contrast agent depends on the flip angle of the RF excitations, the sequence TR, and the T1 of the unenhanced blood. Signal enhancement and overall image quality depend strongly on intraarterial contrast concentration. Fig. 3 shows how T1 of the arterial blood depends on the intra-arterial gadolinium ([Gd]) concentration. As the concentration increases, blood T1 decreases. The resulting equilibrium signal enhancement for hypothetical background tissue (T1 = 700 milliseconds) and contrast-enhanced blood (T1 = 50 milliseconds) also is shown. The highest image contrast comes when intra-arterial [Gd] results in blood T1 values much less than the T1 of fat (T1(fat) = 270 milliseconds). The first pass intra-arterial contrast concentration depends, in turn, on injection rate and cardiac output [4 – 6]. The role of injection rate and cardiac output can be approximated from the relationship between
intra-arterial gadolinium concentration (IA[Gd]), injection rate (IR), cardiac output (CO), and the injected contrast-agent concentration ([Gd]Inj), IA½Gd ¼
IR CO
½GdInj :
ð1Þ
The blood T1, which results from this intra-arterial gadolinium concentration can be determined using 1 1 þ R1 IA½Gd : ¼ T1 1200ms
ð2Þ
where T1 is the resulting longitudinal relaxation rate of the intra-arterial blood, 1200 milliseconds is the initial T1, and R1 is the relaxivity of the particular gadolinium chelate (all values assumed for a 1.5 T field). Using equations 1 and 2 for a typical relativity of 4.5 mMolar1 sec1, a cardiac output of 5 L/min, and [Gd]Inj of 0.5 moles/L, as the injection rate goes from 1.5 ml/second to 0.5 ml/second the intra-arterial concentration decreases by factor of 3, but the blood T1
T.J. Carroll, T.M. Grist / Radiol Clin N Am 40 (2002) 921–951
925
Fig. 2. Time-of-flight MRA. (A) Blood flowing into the imaging volume is fully magnetized, and, therefore, gives a large signal before becoming saturated by repetitive radio-frequency pulses. (B) The high flow into the imaging volume at the circle of Willis gives high signal in the vessels, whereas background signal is suppressed as a result of saturation.
remains below 70 milliseconds for the slow injection. This results in an arterial signal intensity that is much brighter than the surrounding tissue. As injection rates decrease below 0.5 ml/second, however, arterial T1 increases rapidly; therefore, lower injection rates do not produce sufficient T1 shortening.
Centric view ordering Optimal intravascular signal enhancement occurs during peak arterial gadolinium concentration. Therefore, image acquisitions that acquire the central portion of k-space during peak gadolinium concen-
Fig. 3. Gadolinium-based contrast agents are used to decrease the spin-lattice relaxation time (T1) of blood. (A) As the intravascular contrast agent ([Gd]) increases, the T1 of arterial blood decreases. (B) Using T1-weighted pulse sequences in conjunction with gadolinium-based contrast agents, results in a transient increase in the intravascular signal. After steady state is been reached by application of multiple radio-frequency pulses, the contrast-enhanced (T1 = 50 milliseconds) blood has much higher signal than background signal (T1 = 700 milliseconds).
926
T.J. Carroll, T.M. Grist / Radiol Clin N Am 40 (2002) 921–951
tration have higher intravascular signal. For this reason, centric-view – ordered image acquisitions are used in widely CE MRA [7,8]. Centric view ordering is an acquisition strategy in which the lowest spatial frequency k-space lines are acquired firsthand; as the scan proceeds, higher spatial frequencies are acquired. Centric view ordering can be implemented in the phase-encoding direction only or in a truly centric manner in which k-space is sampled centrically in the slice- and phase-encoding direction [8]. This means of acquiring a 3D volume is referred to as
an ‘‘elliptical centric’’ acquisition because ky-kz lines are acquired from the center of k-spaced out in the shape of ellipses. Centric view ordering is beneficial in two ways. First, the portion of k-space that contributes to the intravascular signal is acquired at peak gadolinium concentration. Second, in anatomic regions where the arterial-to-venous time is rapid, centric view ordering provides excellent suppression of venous signal. In the renal and carotid arteries, arterial-to-venous transit time is typically less than ten seconds. By acquiring
Fig. 4. Carotid vascular disease on centric MRA. Coronal acquisition (center) and reformatted maximum intensity projection displays demonstrate bilateral atherosclerotic vascular disease. The high-spatial resolution in the sagitattal reformatted images is possible by prescribing a thin-slice acquisition, with many partitions. Despite the fact that this 3D image was acquired over 54 seconds, much longer that the arterial to venous transit time, the elliptical centric acquisition provides excellent suppression of venous signal. (Courtesy of Kevin Demarco, MD, Laurie Imaging Center.)
T.J. Carroll, T.M. Grist / Radiol Clin N Am 40 (2002) 921–951
927
Timing methods The use of rapid image acquisitions introduces some difficulties in the coordination of image acquisition and the arrival of the bolus of contrast agent. The collection of the central lines of k-space during peak arterial enhancement is key to the success of CE MRA examinations. If the central lines of k-space are acquired prior to the arrival of contrast, severe image artifacts can limit the diagnostic use of the image [9 – 11]. Alternatively, images acquired after the passage of the peak arterial contrast may be obscured by the enhancement of veins. Fig. 5 shows an example of an angiogram of the trifurcation vessels of a patient in which the central phase-encoding values are acquired
Fig. 4 (continued ).
the central-phase encoding values during peak arterial enhancement and high-spatial frequencies after venous opacification, centric-view – ordered acquisitions are able to images with high-arterial signal and minimal venous enhancement. Fig. 4 shows an example of a centrically-encoded examination of the carotid bifurcation. Despite the fact that this 3D image was acquired over 54 seconds, much longer than the arterial to venous transit time, the elliptical centric acquisition provides excellent suppression of venous signal.
Fig. 5. Images acquired 23, 56, and 91 seconds after the injection of a bolus of contrast agent. The image acquired at 23 seconds has insufficient contrast enhancement during the acquisition low-spatial frequency phase encodes resulting in banding in the blood vessels. The trifurcation vessels are well depicted in the properly timed image, acquired 56 seconds after the contrast injection. The arteries are obscured in the late phase image, acquired 96 seconds after the injection.
928
T.J. Carroll, T.M. Grist / Radiol Clin N Am 40 (2002) 921–951
Fig. 5 (continued ).
prior to the arrival of the bolus, a correctly timed examination, and an image acquired after venous opacification. In the image in which the central-phase encodes are acquired too early, severe banding artifact is evident. In the properly timed image acquisition, the trifurcation vessels are well depicted. When an image is acquired well after the bolus passes into the veins, the arteries are obscured by venous overlay.
Test bolus One means of ensuring proper timing of image acquisition is determining transit time of the contrast agent from the injection site to anatomic regions of interest with a small test bolus [12]. In test bolus acquisitions, a small volume of contrast agent is injected during the acquisition of a rapid two-dimensional (2D) T1-weighted acquisition. The acquisition is localized at the level of the anatomic region of interest. Ideally, the 2D images are acquired at a rate
of approximately one image per second. After the acquisition, the images are retrospectively inspected to determine the arrival time at the targeted anatomy. Knowledge of the arrival time is then used to synchronize the image acquisition with the arrival of the injection of the full bolus of contrast agent. Dose-timing acquisitions are successful in abdominal imaging. In this case, the operator must coordinate not only the acquisition of the image, but also the initiation of the breath hold. In the example in Fig. 6, a small region of interest is placed over the abdominal aorta, proximal to the renal arteries. A plot of signal intensity as a function image is produced to identify the arrival of the test bolus, 12 seconds after the injection. The patient is then instructed to breath hold for the acquisition of the 3D angiogram. High frame rate 2D dose-timing exams can also be used to capture clinically relevant contrast dynamics. Fig. 7 shows a 2D dose-timing examination that depicts the delayed filling of a dissection of the aortic arch in a patient. Recently, 2D images have
T.J. Carroll, T.M. Grist / Radiol Clin N Am 40 (2002) 921–951
929
930
T.J. Carroll, T.M. Grist / Radiol Clin N Am 40 (2002) 921–951
Fig. 6. Dose-timing images and data. Serial fast gradient-echo images are obtained at 1-second intervals during the intravenous administration of 1- to 2-cc gadolinium. A region of interest place on the vessel indicates the arrival time, which is used to determine the delay before the image acquisition.
been used to assess the severity and track the response to treatment of intracranial arteriovenous fistulae. Signal loss from intravoxel dephasing, however, limits these acquisitions to under 5 to 10 cm.
Fluoroscopic triggering Fluoroscopic triggering also uses rapid 2D image acquisition to determine when the contrast arrives in the vessels of interest [7]. Rather than injecting a
small dose of contrast to determine the arrival time, however, fluoroscopically triggered acquisitions use real-time images display to monitor the targeted anatomy. The appearance of the contrast agent in the 2D images cues an operator to initiate the acquisition of the high-resolution 3D volume. The rapid 2D fluoroscopic acquisition and 3D volume images are prescribed from a set of localizers, at which time prescan values for the 3D scan are determined. This reduces the latency between triggering of the 3D scan and acquisition of the central
Fig. 7. Delayed filling of false lumen in aortic dissection. Dose-timing scan indicates the delayed filling of the false lumen (arrow). The scan delay is selected to insure opacification of the false lumen (right).
T.J. Carroll, T.M. Grist / Radiol Clin N Am 40 (2002) 921–951
931
enhancement from the arrival of the contrast is viewed, the operator manually triggers the scanner to acquire the 3D volume image. In the case of abdominal imaging, the patient is coached to initiate a breath hold prior to acquisition of the 3D volume. In later implementations of this technique, 2D images are acquired and redisplayed in an interleaved fashion throughout the 3D acquisition [13]. This type of 3D acquisition embeds the acquisition and real time display of 2D images into acquisition of the highspatial resolution 3D image (Fig. 8). This means of embedding a fluoroscopic acquisition allows for simultaneous monitoring of bolus dynamics and acquisition of the 3D image data. It is also possible for the MR scanner to automatically detect the arrival of the bolus of contrast agent and trigger the acquisition of the 3D scan [14]. In automatic contrast detection, a small tracker volume is placed in a vessel of interest. The signal level of the tracker volume is repeatedly sampled during the injection of the contrast agent (Fig. 9). When the scanner detects signal level rise above a predetermined threshold, the acquisition of the 3D volume is triggered. An audible change in the scanning sequences allows the operator to initial a breath hold for abdominal scanning.
Moving table MRA
Fig. 7 (continued ).
phase encodes. The 2D scan is initiated and images are acquired and displayed on the scanner console at roughly one image per second. An operator then injects the full bolus of contrast agent. When signal
The detection of the bolus is combined with automated or manual table translation to allow for rapid full-body 3D MRA. These acquisitions are referred to as ‘‘bolus chase acquisitions,’’ because a single bolus is imaged at successive levels of the anatomy. Because these acquisitions ‘‘chase’’ the bolus as it flows from the abdomen to the distal runoff station, they are sometimes referred to as ‘‘bolus chase’’ acquisitions [15 – 19]. A major advantage of bolus chase acquisitions is that the full bolus of contrast agent is used to image multiple anatomic regions. This is advantageous in two respects. First, because a large volume, typically 0.3 mmol/kg body weight, is used at multiple imaging stations, the intra-arterial signal is very high. Typically, in multistation exams, the full dose must be split between stations, thus lowering the intra-vascular signal relative to the injection of the full contrast dose. Second, unlike acquisitions that use a separate injection to image each station, there is no residual intravascular gadolinium in the distal stations from prior injections. When mask mode subtraction is used, the residual contrast agent subtracts away from
932
T.J. Carroll, T.M. Grist / Radiol Clin N Am 40 (2002) 921–951
the arterial phase image, reducing intravascular image contrast. Bolus-chase MRA is performed typically with multiple passes of the patient through the magnet. After localization volumes are acquired, precontrast images volumes are acquired in a first pass through the magnet. A tracker volume is prescribed, typically in a proximal portion of the aorta and the bolus chase
scan is initiated. When the bolus arrives in the tracker volume, a breath hold is initiated and the abdominal image is acquired. After the abdominal station is acquired, the table automatically translates to a more distal station and second image station is acquired as the bolus fills the vasculature of the patient’s thighs. The final arterial image is acquired after the bolus arrives in the trifurcation vessels (Fig. 10). A third
Fig. 8. The view order for the ‘‘embedded fluoroscopy’’ technique (A) embeds a 2D real-time sequence within a high-resolution 3D CE MRA acquisition. There is a smooth transition from the high-temporal resolution real-time fluoroscopic triggering to the embedded fluoroscopy/3D acquisition at (C) 18 seconds. The subsequent 2D images are displayed in real-time with up to 1second temporal resolution. Images from before contrast arrival (5 seconds), the peak arterial phase (23 seconds), and peak venous phase (33 seconds) are shown (B). A coronal maximum intensity projection of the final high-resolution 3D result is shown in (C). Besides bilateral disease in the carotid bifurcations, there is an occlusion of the right vertebral and a severe stenosis at the origin of the left vertebral. (Courtesy of Sean Fain, PhD, University of Wisconsin.)
T.J. Carroll, T.M. Grist / Radiol Clin N Am 40 (2002) 921–951
933
that atherosclerotic disease is systemic, these examinations provide a means of rapidly performing fullbody screening.
Rapid imaging
Fig. 8 (continued ).
pass of the patient and image acquisition is possible if a late phase venous image is desired.
Angiosurf The idea of chasing a single bolus as it flows down the length of the body is extended to include full-body MRA from the cranial to the caudal stations [20]. By combining a short TR (TR = 2.1 milliseconds) 3D fast low angle shot (FLASH) sequence with a partial Fourier acquisition (see discussion later), a 3D image may be acquired in 12 seconds. These rapid acquisitions in combination with a dedicated moving table/ coil configuration allow five 3D volumes to be acquired in 72 seconds, providing craniocaudal coverage of 1.8 meters. These scans use the dose-timing technique to determine the arrival of contrast in a proximal region of the descending aorta. Then the full bolus of contrast is injected and 3D images are acquired at the level of the carotid bifurcation/aortic arch, abdominal aorta/renal arteries, femoral arteries, popliteal trifurcation, and pedal arch (Fig. 11). Given
The techniques introduced in the preceding section deal primarily with optimal acquisition of standard MR acquisitions. Although some of these techniques acquire multiple images, they do not repeatedly image the same anatomic region and require from 30 to 60 seconds to acquire the image. In recent years, there has been considerable effort to reduce the time to acquire 3D images. Rapid image acquisitions have obvious advantages in the abdomen and thorax where suspended respiration is required during image acquisition. In addition to improving the reliability for acquiring high-resolution arterial-phase images, high frame rate examinations are able to depict filling patterns in cases of dissections and arteriovenous shunts. Rapid acquisitions also open up the possibility of depicting complex contrast dynamics in 3D. The simplest way to acquire images rapidly is to reduce the TR of the pulse sequence. Short TR sequences are available through the development of high-performance gradient systems that achieve TRs below 2 milliseconds. The availability of very short TRs opens up the possibility of acquiring several 3D volume images in the same time that was previously required to acquire a single image. Many groups use acquisitions that do not acquire or ‘‘sample’’ some or part of the full complement of k-space points. This method of reducing scan time is called ‘‘undersampling,’’ and its result on image quality depends on how the undersampling is applied. In some cases the effects of undersampling can be minimized in the image reconstruction process.
Partial Fourier acquisitions Another approach to rapid imaging, which reduces the number of phase-encoding values, is partial Fourier acquisition [21]. In an attempt to maintain spatial resolution with rapid acquisitions, k-space partial Fourier scans acquire k-space data asymmetrically about the origin, undersampling one hemisphere or quadrant of k-space. As shown in Fig. 12, partial Fourier acquisitions are similar in principle to fractional echo acquisitions in that only high-spatial frequencies in one quadrant are acquired. Partial Fourier acquisition schemes have been implemented in the phase- and slice-encoding directions to achieve
934
T.J. Carroll, T.M. Grist / Radiol Clin N Am 40 (2002) 921–951
T.J. Carroll, T.M. Grist / Radiol Clin N Am 40 (2002) 921–951
935
Fig. 10. Bolus chase MRA acquires 3D contrast-enhanced MRA images at successive stations as the bolus of contrast agent travels from the abdomen to the distal run-off. Automated table translation is used to track the bolus similar to computed tomography bolus chase.
frame rates of 0.8 to 1.1 seconds. Despite the anisotropic spatial resolution, 3D projection imaging achieves much higher SNR and through-plane coverage than 2D projection acquisitions. As a result, partial Fourier acquisitions capture flow dynamics normally not seen in 3D acquisitions, at the expense of highly anisotropic spatial resolution.
Parallel acquisitions The time to acquire an image in MR depends on the number of phase-encoding steps that are required (Ny); therefore, reducing the number of phaseencoding steps reduces the scan time. Because the
highest k-space value sampled in any scan determines the spatial resolution, scan time reductions result in lower spatial resolution. Alternatively, one can reduce the number of k-space samples while maintaining high-spatial resolution by increasing the distance between k-space points, as in small field-of-view (FOV) imaging. Small FOV images may be acquired rapidly, but normally suffer from wrap-around artifact as a result of the undersampling. A novel approach to rapid MR images involves acquiring small FOV images with multi-coil arrays [22,23]. Because the individual coil sensitivity profiles depend on the position of a voxel, they contain information on the intensity of the signal that is wrapped into a voxel from outside the FOV; there-
Fig. 9. Automated detection of gadolinium arrival. In automatic contrast detection, a small tracker volume is placed in a vessel of interest. The signal level of the tracker volume is sampled repeatedly during the injection of the contrast agent. When the scanner detects signal level rise above a predetermined threshold, the acquisition of the 3D volume is triggered. (Courtesy of Martin Prince, MD, Cornell University.)
936
T.J. Carroll, T.M. Grist / Radiol Clin N Am 40 (2002) 921–951
the coil sensitivities are used to encode the position of a voxel. These acquisitions are called SMASH or SENSE. The reductions in scan time (ie, ‘‘speed-up factors’’) possible with these spatial encoding techniques depends on the number of coils used and how the sensitivities of the coils overlap. As the speed-up factor increases, however, the SNR of the resulting image decreases, limiting the amount of undersampling or speed-up that is possible. In practice, diagnostic quality images are acquired with speed-up factors of two to four. SENSE acquisition in the abdominal station is combined with manual table translation as part of a three-station single injection peripheral MRA examination. This technique, termed WakiTRAK, acquires a 3D fast field echo (FFE) dataset in the upper (aortoiliac) station with speed-up factors of 1.5 and 2.0 in the slice- and phase-encoding directions, respectively. This allows a 512 230 30 3D volume to be acquired in 11 seconds. This dramatically reduces upper-station scan time such that high-resolution exams of the middle, and particularly the lower, station arteries can be acquired sooner after contrast arrival in the aorta. This makes better use of the bolus and significantly reduces the incidence of venous enhancement. An example of a three-station WakiTRAK study is shown in Fig. 13. The lower station is acquired over 71 seconds, with a true spatial resolution of 0.9 0.9 1.0 mm and an elliptical centric acquisition beginning approximately 31 seconds after arrival of the contrast in the abdominal station. With properly designed coils, SENSE in theory can be applied in multiple stations.
Time-resolved acquisitions
Fig. 10 (continued ).
fore, using the individual coil sensitivity profiles, the wrap-around artifact can be eliminated. By forming linear combinations of coil sensitivities, nonaliased full FOV images can be produced. In effect,
An alternative method for acquiring angiograms is to rapidly acquire multiple 3D volumes. By acquiring images throughout the passage of the bolus of contrast agent, multiphase techniques are inherently insensitive to interpatient variability of contrast arrival [24 – 27]. An additional benefit of timeresolved acquisitions is the ability to depict pathologically delayed vessel filling. Another approach to time-resolved image acquisition relies on the fact that much of the information forming an MR image is present in the central region of k-space (see Fig. 1). By acquiring a multiphase examination in which the central phaseencoding values are acquired more often than the outer regions of kspace, a time series of 3D image may be reconstructed [28 – 30]. This technique (TRICKS) retrospectively combines the central
T.J. Carroll, T.M. Grist / Radiol Clin N Am 40 (2002) 921–951
937
Fig. 11. Angiosurf images: a 74-year-old male with peripheral vascular disease, coronary heart disease, renal insufficiency. The images demonstrate diffuse atherosclerotic disease, including a high-grade stenosis of the carotid bifurcation right side, high-grade stenosis common carotid artery left side, multiple wall irregularities entire aorta, occlusion of left renal artery, infrarenal abdominal aortic aneurysm (partly thrombosed), occlusion common iliac artery left side, and occlusions of the superficial femoral artery both sides. (Courtesy of Mathias Goyen, MD, University of Essen.)
938
T.J. Carroll, T.M. Grist / Radiol Clin N Am 40 (2002) 921–951
Fig. 11 (continued ).
phase-encoding values with high-spatial frequency data acquired later in time. In effect, TRICKS oversamples the central region of kspace relative to the sampling rate of the outer regions. In this way TRICKS is able to capture consistently an arterial time frame, free of venous overlay even in regions of rapid venous return, such as the carotid arteries [31]. In the distal extremities, where bolus chase techniques are shown to be sensitive to contrast arrival time and venous overlay, TRICKS is successful in acquiring diagnostic images in patients with severe pathology [32] (Fig. 14). A logical extension of the oversampling of the central phase encodes is to acquire the highest spatial frequencies only at the end of the contrast-enhanced scan, as in a ‘‘keyhole’’ acquisition. By acquiring the k-space points that contribute to edge depiction only once, at the end of the scan, the frame rate during the arterial phase is not compromised. Because recirculation of the initial bolus of contrast results in prolonged intravascular T1-shortening, sampling the highest k-space lines up to 3 to 4 minutes after first pass the contrast is possible. Fig. 15 shows a signal enhancement curve acquired in the femoral arteries of a volunteer after 0.1 mmol/kg of a gadoliniumbased contrast agent, showing that prolonged signal
enhancement persists for several minutes. The increased spatial resolution that is possible is demonstrated in an image of the carotid bifurcation.
Non-Cartesian acquisitions Despite the recent advances in rapid image acquisitions presented in the preceding sections, the use of conventional Fourier spin-warp imaging is limited in its ability to meet the demands of highresolution MRA. The dependence of spatial resolution on image acquisition time will ultimately limit the spatial resolution at which images are acquired. Novel approaches, however, to MR image acquisitions are beginning to show promise as a means to address some of the technical challenges facing the development of conventional MRA.
Undersampled PR Radial projections reconstruction (PR) k-space trajectories were introduced as the first technique for doing spatial localization in MR. Rather than sampling k-space on a rectilinear grid, as with conventional MR
T.J. Carroll, T.M. Grist / Radiol Clin N Am 40 (2002) 921–951
acquisitions, PR acquisitions sample k-space on radial trajectories. An example of radial k-space sampling is shown in Fig. 16. Unlike conventional MRA acquisitions, radial k-space trajectories sample the center and periphery of k-space on each echo. Reducing the image acquisition time with Fourier encoding can be achieved by acquiring fewer phaseencoding values. Undersampling in Fourier-encoded MRI usually requires the acquisition of low-spatial resolution images, small FOVs, or highly anisotropic spatial resolution. Undersampling in PR acquisitions is achieved by decreasing the number of angular samples
939
and does not decrease spatial resolution or FOV [33]. Rather, decreasing the number of radial samples results in a low intensity streak artifact, similar to that seen in computed tomographic (CT) examinations. The PR undersampling artifact has not proven to be less problematic in CE MRA, because blood vessels are the dominant signal source, unlike in CT, where artifact from bone can confound diagnosis. Fig. 17 shows a comparison between a conventional Fourier and undersampled PR contrastenhanced examination of a resolution phantom. In the PR image (center), small structures are visualized
Fig. 12. Partial Fourier acquisition. Image acquisitions can be accelerated by acquiring a portion of the full compliment of k-space points. (A) A k-space image prior to Fourier transformation with frequency (v) and phase-encoding (Ky) directions indicated. (B) Fractional-echo acquisitions can be used to reduce echo time as well as repetition time resulting in a more rapid acquisition of images. (C) Similarly, by acquiring a subset of the total number of phase-encoding values, a proportionate decrease in scanning time is achieved.
940
T.J. Carroll, T.M. Grist / Radiol Clin N Am 40 (2002) 921–951
be increased [34]. Using a combined PR-TRICKS acquisition, it is possible to acquire high resolution (0.5 mm 0.76 2.4 mm) 3D volumes with a frame rate equal to 2.5 seconds per frame. An example of a time frame from a PR-TRICKS examination of the trifurcation vessels in Fig. 10. In this example, the bolus of contrast arrives in the right leg several seconds after the left. The high frame rate afforded by PRTRICKS allows for optimal sampling of the central phase-encoding values to depict both legs.
PR-hyperTRICKS
Fig. 12 (continued ).
more accurately than the Fourier encoded image acquired in the same time (left). In order for Fourier encoding to acquire the same spatial resolution, four times the imaging time is required (right). This is demonstrated in vivo in a contrast-enhanced examination of the pulmonary arteries (Fig. 18). The many more distal branches of the pulmonary vasculature are visible in the undersampled PR acquisition as compared to the Fourier-encoded image acquired in the same time.
PR-TRICKS The short imaging times that are possible with undersampled PR acquisitions are ideal for timeresolved examinations. By combining an undersampled PR k-space trajectory to sample k-space in the kx-ky plane with a TRICKS-encoding in the slice direction, the frame rate of TRICKS examinations may
As described previously, TRICKS encoding is able to increase the frame rate over the normal multiphase examination by resampling the center of k-space at a higher rate than the higher spatial frequencies. The resampling of the center of k-space, however, results in TRICKS examinations having lower spatial resolution than single-image acquisitions acquired in the same time. In elliptical centric acquisitions of the renal and carotid arteries, the high-spatial frequency phaseencoding values are acquired several seconds after venous opacification. Motivated by the excellent venous suppression afforded by elliptical-centric examinations, further improvements in the spatial resolution of TRICKS have been achieved. By slightly modifying the TRICKS acquisition schedule, it is possible to double the slice resolution without sacrificing frame rate during the first pass of the bolus of contrast agent. These acquisitions, called hyperTRICKS acquisitions, perform a conventional TRICKS examination during the first pass of the contrast agent, then acquire the high-spatial frequencies slice encodes only at the end of the acquisition, similar in principle to keyhole imaging (Fig. 19). In this way, dynamic information on filling patterns and contrast arrival are available to the clinician. In addition, the spatial resolution of the arterial image is increased to recover the loss in spatial resolution normally associated with time-resolved acquisitions. Fig. 19 also shows an example of the improvements in spatial resolution that hyperTRICKS acquisitions are capable of in an extremely low-spatial resolution TRICKS examination. The spatial resolution of a single slice through the carotid bulb acquired at peak arterial enhancement is improved by including high-spatial frequencies acquired after venous opacification. As in elliptical-centrically encoded examinations, edge depiction of the artery is improved with minimal venous overlay. When hyperTRICKS is combined with undersampled PR (ie, PR-hyperTRICKS), further improvements in spa-
T.J. Carroll, T.M. Grist / Radiol Clin N Am 40 (2002) 921–951
941
Fig. 13. WakiTRACK peripheral MRA examination. A multicoil SENSE acquisition in the abdominal station is combined with manual table translation as part of a three-station single injection peripheral MRA examination. The third station ± 45° oblique view and coronal. Also, a zoomed image of the trifurcation vessels demonstrates diffuse occlusive disease. (Courtesy of Jeffrey Maki, MD, PhD, University of Washington.)
tial and temporal resolution are possible. In examinations of the distal extremities, PR-hyperTRICKS is used to improve the spatial resolution and coverage of these examinations.
VIPR Motivated by the scan time reductions achieved with undersampled PR, radial undersampling is
extended to 3D acquisitions [35]. Previous implementations of undersampling relied on traditional Fourier encoding in the slice direction. The VIPR (Vastly undersampled Isotropic Projection Reconstruction) acquisition samples k-space using 3D radial trajectories (Fig. 20). The low-intensity streaks associated with 2D undersampling appear as low intensity correlated noise distribution. There are several advantages to the VIPR sampling strategy. First, because radial sampling is per-
942
T.J. Carroll, T.M. Grist / Radiol Clin N Am 40 (2002) 921–951
Fig. 14. Time-resolved imaging of vascular malformation in 2-month-old child. Sequential 3D volumes were reconstructed every 2 seconds using the 3D TRICKS technique. Images show early arterial filling of the vascular malformation (A), followed by the blush (B), and then the venous drainage is demonstrated (C). Time-resolved imaging allows the use of a small 2-cc bolus in this infant, without a timing scan.
T.J. Carroll, T.M. Grist / Radiol Clin N Am 40 (2002) 921–951
943
944
T.J. Carroll, T.M. Grist / Radiol Clin N Am 40 (2002) 921–951
By interleaving sets of projection angles, it becomes possible to acquire high-spatial resolution and isotropic 3-D volumes. Time-resolved VIPR acquisition of 256 256 256 volumes in less than 4 seconds is possible. Fig. 20 shows a coronal view of an early time frame that depicts the pulmonary arteries. A coarctation of the aorta from a later frame in the same acquisition is best viewed in the sagittal plane. The large coverage and isotropic spatial resolution allow reprojection of these time-resolved acquisitions in any plane.
Contrast agents
Fig. 15. HyperTRICKS schematic. Time-resolved images are acquired during the first pass of contrast agent, then high-resolution steady state imaging is performed beginning 120 seconds after the start of the scan. The curve (A) shows signal enhancement in the femoral arteries of a volunteer after 0.1 mmol/kg of a gadolinium-based contrast agent, showing that prolonged signal enhancement persists for several minutes. The image resolution of the carotid study (B – D) increases (left to right) as high-spatial frequencies are acquired and used in the reconstruction.
formed in 3D, the spatial resolution and coverage of these examinations are isotropic. This is useful particularly in abdominal imaging. The isotropic 3D coverage eliminates the need for volumetric selection. In addition, reprojection at any angle of obliquity does not result in degraded spatial resolution. Second, the ability to undersample in 3D without a loss in coverage or spatial resolution allows for high-spatial resolution images to be acquired in a comfortable breath hold.
Gadolinium chelates have proved safe and effective for CE MRA in many clinical trials. These agents, however, are used primarily during the first pass of contrast-enhancement, as there is relatively high leakage of contrast through the vascular capillary network into the interstitial space. The distribution of standard gadolinium contrast agents results in enhancement of the soft tissues following intravenous injection of gadolinium contrast agent. Several gadolinium-based contrast agents are approved for use in the United States, and additional agents have met approval in Europe [36]. None of the three gadolinium-based contrast agents used in the United States, however, are in fact approved for use in CE MRA. MRA using gadolinium contrast agents is shown to be safe and effective in multiple clinical trials published in the literature. The pharmaceutical manufacturers, however, have not yet performed the clinical trials necessary for Food and Drug Administration approval of gadolinium agents for the MRA indication. Therefore, use of extracellular gadolinium contrast agents represents an ‘‘off-label’’ use of the pharmaceutical as established by clinical need under the guidance of a licensed physician. Many advances in the development of improved contrast agents aimed specifically at MRA have been made. Two general categories of contrast media are being developed, including extracellular and blood pool contrast agents. Extracellular contrast agents are similar to existing gadolinium chelates; however, some compounds are designed to enhance the T1 relaxivity of the gadolinium chelate [37]. For example, the weak protein binding of gadobenate dimeglumine (Multihance, Bracco, Milan, Italy) results in improved relaxivity (R1); therefore, injection of the contrast agent results in a shorter T1 relaxation time and improved signal intensity relative to standard gadolinium chelates [37,38]. In addition, formulations of gadobutrol (Gadovist, Schering, Berlin, Germany)
T.J. Carroll, T.M. Grist / Radiol Clin N Am 40 (2002) 921–951
945
Fig. 16. (A) Spin-warp MR image acquisitions sample the k-space representation of an image on a rectilinear grid. (B) Projection reconstruction acquisitions sample k-space on radial trajectories that pass through the center of k-space and resemble the spoke of a bicycle tire. These trajectories sample both the low-spatial frequencies (responsible for image contrast) and high-spatial frequencies (responsible for image detail) in every repetition time. Undersampling in the radial dimension does not cause wraparound artifact or reduced spatial resolution as in spin-warp imaging. (Courtesy of Karl Vigen, PhD, Stanford University.)
are prepared using a higher molar concentration of gadolinium per cubic centimeter (1.0 M versus the current standard of 0.5 M), which is shown to improve vessel SNR in one study of pelvic MRA [39].
The second general category of contrast agents includes blood pool contrast agents, which are intended to stay within the intravascular space. As described previously, one of the difficulties with
Fig. 17. Comparison between a conventional Fourier and undersampled projections reconstruction (PR) resolution. In the PR image (center), small structures are visualized more accurately than the Fourier encoded image acquired in the same time (left). In order for Fourier encoding to acquire the same spatial resolution, four times the imaging time is required (right). Acquiring a fewer number of projections than are required to full sample k-space results in a low-level streak artifact that emanates out from objects within an image. Undersampling radial projections, however, allows for rapid acquisition of high-spatial resolution images, where the streak artifacts are relatively minor. (Courtesy of Dana Peters, PhD, National Institutes of Health.)
946
T.J. Carroll, T.M. Grist / Radiol Clin N Am 40 (2002) 921–951
Fig. 18. (A) A Fourier-encoded examination of the pulmonary vasculature is compared with (B) an undersampled projections reconstruction (PR) examination in the same patient. The acquisitions time was the same for both examinations; however, the inplane spatial resolution of the undersampled PR image (0.7 mm 0.7 mm) is greater than the Fourier encoded image (0.7 mm 2.0 mm). (Courtesy of Dana C. Peters, PhD, National Institutes of Health.)
extracellular contrast agents is leakage of the contrast agent into the extracellular space, which results in reduced contrast between the intravascular signal and the soft tissues. In theory, blood pool contrast agents could reduce this leakage and thereby improve the contrast between the vascular phase and stationary tissues. More importantly, several of the blood pool contrast agents are designed so that there is prolonged intravascular signal associated with the contrast agent, thereby allowing the acquisition of high-resolution images [40]. Improved signal intensity and spatial resolution can be provided by the prolonged arterial phase of the blood pool contrast agents [41]. The prolonged arterial phase also comes at a disadvantage, however, because of the associated enhancement of venous structures surrounding the arteries. These techniques undoubtedly will require improved processing algorithms to segment the arteries or veins [42]. The success of these segmentation methods, however, needs to be demonstrated in large-scale clinical trials.
Several different blood pool contrast agents have been proposed. One agent, MS-325, uses a principal of strong protein binding to ensure that the gadoliniumbased contrast agent stays within the vascular phase [31,41,45]. MS-325 binds to serum albumin, which results in a prolonged blood half-life [40]. In addition, the protein binding results in improved (R1), thus leading to improved SNR at a lower dose. Gadomer-17 represents a dendrimer compound with multiple gadolinium atoms per molecule [43]. This contrast agent also is shown to have high relaxivity. The molecule is designed to ensure that the size is large enough to stay in the vessel, but small enough to be filtered renal filtration. The contrast agent has several potentially desirable properties for MRA, including high relaxivity and the fact that it stays within the blood pool. In addition, it has a relatively short blood-pool half-life, thus allowing repeated injections if clinically indicated. Finally, investigators have proposed the use of iron-based contrast agents to provide prolonged
Fig. 19. Projections reconstruction hypertricks scan of the distal lower extremities. Serial time frames demonstrate delayed filling of the right infrapopliteal vessels. The total examination time is 216, but images are reconstructed every 2 seconds (A – D). Image resolution is 0.7 0.7 2 mm. The high-resolution study provides excellent depiction of the small distal vessels. (Courtesy of Jiang Du, MS, University of Wisconsin.)
T.J. Carroll, T.M. Grist / Radiol Clin N Am 40 (2002) 921–951
947
948
T.J. Carroll, T.M. Grist / Radiol Clin N Am 40 (2002) 921–951
blood pool enhancement for MRA [44]. The ironbased agents effect T1 and T2, because the particulate compounds have an effect on R1 and R2 relaxivity. At low doses, however, the effect is primarily on the R1, resulting in T1 shortening for MRA. These agents have potential merit because of the long blood pool residence time, and the high relaxivity. Limitations of the iron-based agents, however, are related to
the fact that the also shortened T2 relaxivity, thereby lowering signal intensity at high doses.
Summary CE MRA has evolved rapidly since the early studies by Prince et al [3]. Whereas many of the
Fig. 20. VIPR (Vastly undersampled Isotropic Projection Reconstruction) acquisition samples k-space using in 3D radial trajectories (A). Pulmonary phase (B) and aortic phase (C) images may be reconstructed from a single breath-hold examination by weighting the k-space data differently. Image in (C) demonstrates coarctation of the aorta. High frame rates allow capture of an arterial phase, while simultaneously providing isotropic resolution and broad coverage. (Courtesy of Walter Block, PhD, University of Wisconsin.)
T.J. Carroll, T.M. Grist / Radiol Clin N Am 40 (2002) 921–951
949
Fig. 20 (continued ).
procedures in clinical use today rely heavily on the use of gadolinium contrast agents and standard Fourier transform acquisition techniques, advances will have a significant impact on MRA by shortening the acquisition time, improving the reproducibility of the image-acquisition techniques, and improving spatial resolution or SNR. From a technical basis, shorter acquisition times associated with fast gradients are likely to improve spatial resolution and allow for acquisition of MR images over large FOVs. In addition, alternative k-space sampling techniques, such as parallel imaging and PR, are expected to further reduce acquisition time, while maintaining or improving spatial resolution. The approval and subsequent use of new contrast agents will also have a beneficial impact on the image quality of contrastenhanced MRA applications. It is likely that these contrast agents will be coupled with advanced acquisition techniques to improve spatial resolution and technical success rates of MRA examinations.
References [1] Baum RA, Rutter CM, Sunshine JH, Bleba JS, Bleba J, Carpenter JP, et al. Multi-center trial to evaluate vascular magnetic resonance angiography of the lower extremity. JAMA 1995;274:875 – 80. [2] McCauley TR, Monib A, Dickey KW, et al. Peripheral vascular occlusive disease: accuracy and reliability of time-of-flight MR angiography. Radiology 1994;192: 351 – 7. [3] Prince MR, Yucel EK, Kaufman JA, Harrison DC, Geller SC. Dynamic gadolinium-enhanced threedimensional abdominal MR arteriography. J Magn Reson Imaging 1993;3:877 – 81.
[4] Carroll TJ, Korosec FR, Swan JS, Hany TF, Grist TM, Mistretta CA. The effect of injection rate on time resolved contrast-enhanced peripheral MRA. J Magn Reson Imaging 2001;14:401 – 10. [5] Leggett RW, Williams LR. A proposed blood circulation model for reference man. Health Phys 1995;69: 187 – 201. [6] Maki JH, Chenevert TL, Prince MR. Three-dimensional contrast-enhanced MR angiography. Top Magn Reson Imaging 1996;8:322 – 44. [7] Riederer SJ, Bernstein MA, Breen JF, et al. Threedimensional contrast-enhanced MR angiography with real-time fluoroscopic triggering: design specifications and technical reliability in 330 patient studies. Radiology 2000;215:584 – 93. [8] Wilman AH, Riederer SJ, Huston J, Wald JT, Debbins JP. Arterial phase carotid and vertebral artery imaging in 3D contrast-enhanced MR angiography by combining fluoroscopic triggering with an elliptical centric acquisition order. Magn Reson Med 1998;40:24 – 35. [9] Kim JK, Farb RI, Wright GA. Test bolus examination in the carotid artery at dynamic gadolinium-enhanced MR angiography. Radiology 1998;206:283 – 9. [10] Maki JH, Prince MR, Londy FJ, Chenevert TL. The effects of time varying intravascular signal intensity and k-space acquisition order on three-dimensional MR angiography image quality. J Magn Reson Imaging 1996;6:642 – 51. [11] Svensson J, Peterson J, Stahlberg F, Larsson E, Leander L, Olsson L. Image artifacts due to a time-varying contrast medium concentration in 3D contrast-enhanced MRA. J Magn Reson Imaging 1999;10:919 – 28. [12] Earls JP, Rofsky NM, DeCorato DR, Krinsky GA, Weinreb JC. Breath-hold single-dose gadoliniumenhanced three-dimensional MR aortography: usefulness of a timing examination and MR power injector. Radiology 1996;201:705 – 10. [13] Fain SB, Riederer SJ, Huston J 3rd, King BF. Embedded MR fluoroscopy: High temporal resolution real-
950
[14]
[15]
[16]
[17]
[18] [19]
[20]
[21]
[22]
[23]
[24]
[25]
[26]
[27]
T.J. Carroll, T.M. Grist / Radiol Clin N Am 40 (2002) 921–951 time imaging during high spatial resolution 3D MRA acquisitions. Magn Reson Med 2001;46:690 – 8. Foo TK, Saranathan M, Prince MR, Chenevert TL. Automated detection of bolus arrival and initiation of data acquisition in fast, three-dimensional MR angiography. Radiology 1997;203:275 – 80. Hany TF, Carroll TJ, Omary RA, Esparza-Coss E, Korosec FR, Mistretta CA, Grist TM. Aorta and runoff vessels: single-injection MR angiography with automated table movements compared with multiinjection time-resolved MR angiography – initial results. Radiology 2001;221:266 – 72. Ho KY, Leiner T, de Haan MW, Kessels AG, Kitslaar PJ, van Engelshoven JM. Peripheral vascular tree stenoses: evaluation with moving-bed infusion-tracking MR angiography. Radiology 1998;206:683 – 92. Ho VB, Choyke PL, Foo TK, et al. Automated bolus chase peripheral angiography: initial practical experiences and future directions of this work-in-progress. J Magn Reson Imaging 1999;10:376 – 88. Ho KY, Leiner T, van Engelshoven JM. MR angiography of run-off vessels. Euro Radiol 1999;9:1285 – 9. Meaney JF, Ridgeway JP, Chakraverty S, et al. Steppingtable gadolinium-enhanced digital subtraction MR angiography of the aorta and lower extremity arteries: preliminary experience. Radiology 1999;211:59 – 67. Ruehm SG, Goyen M, Barkhausen J, Kroger K, Bosk S, Ladd ME, Debatin JF. Rapid magnetic resonance angiography for detection of atherosclerosis. Lancet 2001;357:1086 – 91. Feinberg DA, Hale JD, Watts JC, Kaufman L, Mark A. Halving MR imaging time by conjugation: demonstration at 3.5 kG. Radiology 1986;161:527 – 31. Sodickson KD, Manning WJ. Simultaneous acquisition of spatial harmonics (SMASH): fast imaging with radiofrequency coil arrays. Magn Reson Med 1997;38: 591 – 603. Pruessmann KP, Weiger M, Scheidegger MB, Boesiger P. SENSE: sensitivity encoding for fast MRI. Magn Reson Med 1999;42:952 – 62. Schoenberg SO, Bock M, Knopp MV, et al. Renal arteries: optimization of three-dimensional gadolinium-enhanced MR angiography with bolus-timingindependent fast multiphase acquisition in a single breath hold. Radiology 1999;211:605 – 7. Wang Y, Johnston DL, Breen JF, et al. Dynamic MR digital subtraction angiography using contrast enhancement, fast data acquisition and complex subtraction. Magn Reson Med 1996;36:551 – 6. Wang Y, Winchester P, Khilnani M, et al. Contrastenhanced peripheral MR angiography from the abdominal aorta to the pedal arteries: combined dynamic two-dimensional and bolus-chase three-dimensional acquisitions. Invest Radiol 2001;36(3):170 – 7. Willig DS, Turski PA, Frayne R, et al. Contrast-enhanced 3D MR DSA of the carotid artery bifurcation: preliminary study of comparison with unenhanced 2D and 3D time-of-flight MR angiography. Radiology 1998;208:447 – 51.
[28] Korosec FR, Frayne R, Grist TM, Mistretta CA. Timeresolved contrast-enhanced 3D MR angiography. Magn Reson Med 1996;36:345 – 51. [29] Korosec FR, Turski PA, Carroll TJ, Mistretta CA, Grist TM. Contrast-enhanced MR angiography of the carotid bifurcation. J Magn Reson Imaging 1999;10: 317 – 25. [30] Mistretta CA, Grist TM, Korosec FR, et al. 3D timeresolved contrast-enhanced MR DSA: advantages and tradeoffs. Magn Reson Med 1998;40:571 – 81. [31] Carroll TJ, Korosec FR, Petermann GM, Grist TM, Turski PA. Carotid bifurcation: evaluation of timeresolved three-dimensional contrast-enhanced MR angiography. Radiology 2001;20:525 – 32. [32] Carroll TJ, Korosec FR, Swan JS, Grist TM, Frayne FR, Mistretta CA. A method for rapidly determining and reconstructing the peak arterial frame from a time resolved CE-MRA exam. Magn Reson Med 2000; 44:817 – 20. [33] Peters DC, Korosec FR, Grist TM, Block WF, Holden JE, Vigen KK, et al. Undersampled projection reconstruction applied to MR angiography. Magn Reson Med 2000;43:91 – 101. [34] Vigen KK, Peters DC, Grist TM, Block WF, Mistretta CA. Undersampled projection-reconstruction imaging for time-resolved contrast-enhanced imaging. Magn Reson Med 2000;43:170 – 6. [35] Barger AV, Block WF, Torpov Y, Grist TM, Mistretta CA. Time-resolved contrast-enhanced imaging with isotropic resolution and broad coverage using an undersampled 3D projection trajectory. Magn Reson Med 2002. [36] Knopp MV, von Tengg-Kobligk H, Floemer F, Schoenberg SO. Contrast agents for MRA. Future directions. J Magn Reson Imaging 1999;10:314 – 6. [37] Cavagna FM, Maggioni F, Castelli PM, Dapra M, Imperatori LG, Lorusso V, et al. Gadolinium chelates with weak binding to serum proteins. A new class of highefficiency general purpose contrast agents for magnetic resonance imaging. Invest Radiol 1997;32:780 – 96. [38] Volk M, Strotzer M, Lenhart M, Seitz J, Manke C, Feuerbach S, et al. Renal time-resolved MR angiography: Quantitative comparison of gadobenate dimeglumine and gadopentetate dimeglumine with different doses. Radiology 2001;220:484 – 8. [39] Goyen M, Lauenstein TC, Herborn CU, Debatin JF, Bosk S, Ruehm SG. 0.5 M Gd chelate (Magnevist) versus 1.0 M Gd chelate (Gadovist): dose-independent effect on image quality of pelvic three-dimensional MRangiography. J Magn Reson Imaging 2001;14:602 – 7. [40] Lauffer RB, Parmelee DJ, Dunham SU, Ouellet HS, Dolan RP, Witte S, et al. MS-325: albumin-targeted contrast agent for MR angiography. Radiology 1998; 207:529 – 38. [41] Grist TM, Korosec FR, Peters DC, Witte S, Walovitch RC, Dolan RP, et al. Steady-state and dynamic MR angiography with MS-325: initial experience in humans. Radiology 1998;207:539 – 44. [42] Mazaheri Y, Carroll TJ, Du J, Korosec FR, Block WF,
T.J. Carroll, T.M. Grist / Radiol Clin N Am 40 (2002) 921–951 Vigen KK, et al. Combined time-resolved and high spatial resolution 3D MRA using an extended adaptive acquisition. J Magn Reson Imag 2002;15:291 – 301. [43] Dong Q, Hurst DR, Weinmann HJ, Chenevert TL, Londy FJ, Prince MR. Magnetic resonance angiography with gadomer-17. An animal study original investigation. Invest Radiol 1998;33:699 – 708. [44] Taylor AM, Panting JR, Keegan J, Gatehouse PD,
951
Amin D, Jhooti P, et al. Safety and preliminary findings with the intravascular contrast agent NC100150 injection for MR coronary angiography. J Magn Reson Imaging 1999;9:220 – 7. [45] Bluemke DA, Stillman AE, Bis KG, Grist TM, Baum RA, D’Agostino R, et al. Carotid MR angiography: phase II study of safety and efficacy for MS-325. Radiology 2001;219:114 – 22.