Dr. Lohan is a Diagnostic Cardiovascular Imaging Fellow, Dr. Saleh is a Research Fellow, Dr. Nael is a Radiology Resident, Dr. Krishnam is an Assistant Professor of Radiology, and Dr. Finn is a Professor of Radiology and Medicine, Chief of Diagnostic Cardiovascular Imaging, and Director of Magnetic Resonance Research, Department of Radiology, David Geffen School of Medicine at UCLA, Los Angeles, CA.
Abstract
Magnetic resonance angiography (MRA) continues to evolve rapidly, having grown from a variety of flow-dependent techniques through contrast-enhanced techniques, each with its own specific advantages and requirements. Recent technical developments such as the introduction of 3T scanners, the increasing sophistication of radiofrequency (RF) architecture, and the dissemination of parallel acquisition techniques have greatly increased the power and practicality of contrast-enhanced MRA, such that it is now highly competitive with computed tomographic angiography (CTA) in most vascular territories. Nonetheless, non–contrast-enhanced MRA is still widely used, particularly in neurovascular imaging. As a result, it is crucial that interpreting physicians are familiar with the basic principles, relative advantages, and limitations of each of these methods, including associated potential diagnostic pitfalls. The authors present the progression of nonenhanced and enhanced MRA to date, illustrating key characteristics as well as discussing the current applications of each in diagnostic imaging.
Conventional catheter angiography (CCA) has long been regarded as the gold standard for anatomic vascular imaging. However, CCA is invasive and expensive and exposes the patient to radiation and iodinated contrast agents.
Magnetic resonance angiography (MRA) has long held promise as an effective noninvasive alternative to CCA for many vascular territories. Early attempts at vessel analysis made use of time-of-flight (TOF) and phase-contrast (PC) techniques, while more recently contrast-enhanced (CE) MRA has gained wide acceptance, in many cases relegating conventional angiography to a secondary role during image-guided vascular intervention.
In this article, we discuss the evolving roles of CE-MRA relative to non–CE-MRA in clinical practice, drawing on our own experience and that of many others. We emphasize that there is a wide spectrum of hardware and software platforms on which MRA techniques are implemented and that the performance, quality, and reliability of the results are influenced by local platform configuration and operator skill.
Time-of-flight MRA
Time-of-flight (TOF) MRA takes advantage of through-plane blood flow to highlight patent blood vessels against a relatively suppressed stationary background.1-3 Thus, the requirement for intravenous contrast administration is obviated. Furthermore, presaturation can be used effectively to eliminate signal (flow enhancement) from a specific direction, generating selective arteriograms or venograms in the brain and body.4,5
Time-of-flight MRA can be implemented in a 2-dimensional (2D) (sequential multislice) or a 3-dimensional (3D) (volume slab) format, depending on the target vascular territory. For the evaluation of detailed arterial anatomy, the 3D version is preferred, because it allows for isotropic voxels (by virtue of very thin digital slices, known as partitions). Three-dimensional TOF-MRA is time-consuming, taking approximately 5 to 10 minutes, depending on coverage and desired spatial resolution. For this reason, its applications are limited to structures such as the head and neck, which can be immobilized during the acquisition. Because of the widespread acceptance of neurovascular MR imaging (MRI), 3D TOF-MRA of the intracranial arteries is the single most widely used MRA technique of any kind. It is also commonly used for imaging the cervical carotid arteries.
The dependence on adequate flow enhancement imposes certain constraints on TOF-MRA. If the repetition time (TR) is too short, there will be insufficient time for fresh spins to enter the slice and vascular contrast will suffer. Also, if a vessel is oriented largely within the imaging slab or slice, the blood may become saturated (an effect known as in-plane saturation) even if flow is not severely compromised (Figures 1 through 9). For this reason, TR is generally chosen to balance acquisition time and flow enhancement. For TOF-MRA, TRs of approximately 20 to 40 msec are typical. Also, for a given TR, the flip angle is chosen in the range of 20˚ to 40˚ to optimize the blood signal while suppressing background. It was recognized early in the development of TOF-MRA that loss of signal within blood vessels can occur because of flow-induced phase dispersion, mimicking or exaggerating ste-noses. This effect is more pronounced at longer echo times (TE), and it can be partially reversed by a technique known as gradient motion rephrasing.2
Outside the head and neck, breathing motion artifact generally renders 3D TOF-MRA useless. For this reason, TOF-MRA of the chest, abdomen, and pelvis has generally been implemented as a 2D, sequential breath-hold technique.9-11 Two-dimensional TOF-MRA differs from its 3D counterpart in several important ways. With 3D TOF-MRA, the entire imaged volume is exposed to all the radiofrequency (RF) pulses, so that some degree of saturation is common and slowly flowing spins are poorly visualized. With 2D TOF-MRA, only a single slice is excited, so that through-plane flow sensitivity is heightened, and 2D TOF-MRA has been used to good effect for venous imaging,12-14 where flow sensitivity is more important than spatial resolution. Slice thickness with 2D TOF-MRA is generally greater than partition thickness in 3D TOF-MRA. The acquisition order of the slices is sequential, such that each slice is independent of the others. For this reason, each slice can be acquired in an independent breath-hold, a useful property for abdominal and thoracic imaging. It should be pointed out that sequential 2D TOF-MRA is also commonly performed for the extracranial carotid arteries.15,16
Multiple overlapping thin-slab acquisition (MOTSA) MRA17 represents a hybrid between sequential 2D MRA and single-slab 3D MRA. As the name suggests, with MOTSA, the thickness of each 3D slab is decreased in order to decrease saturation effects. Each overlapping sub-volume is acquired sequentially and is finally fused with the others to form the complete 3D volume. Multiple overlapping thin-slab MRA is widely employed for imaging both the intracranial arteries and the extracranial carotids.18,19
The vulnerability of the blood to saturation during short TR 3D acquisition can be completely offset by the T1-shortening effect of paramagnetic contrast agents, as described in the early work of Prince et al20 and Maki et al.21 When the center of k-space is made to coincide with the peak intravascular concentration of a contrast bolus, the increase in blood signal is maximized. In recent years, developments in CE-MRA have benefited from steady improvements in RF architecture, gradient technology, parallel acquisition,22-26 and the proliferation of 3T whole-body MR systems.27-32
Clinical applications
Three-dimensional TOF-MRA has long been the technique of choice for noninvasive imaging of the head and neck arteries.33-37 For several reasons, the carotid circulation represents the most receptive anatomic region for 3D TOF-MRA. On average, the carotids have a predominantly superoinferior orientation and moderately high flow rates, making axial multislab acquisition suitable in most cases. It is practical to ask patients to keep their heads still for the several minutes required for a high-resolution study. Also, dedicated arrays of head-neck surface coils are widely available to maximize the signal-to-noise ratio (SNR) and to support parallel acquisition.
Indeed, when using an optimized technique on a cooperative patient, TOF-MRA allows superb visualization of the circle of Willis, which provides information about vascular patency, caliber, and the presence and location of intracranial aneurysms. This technique has shown particular value in the diagnosis, localization, and follow-up of patients with moyamoya disease and other intracranial vasculopathies. The majority of protocols use a MOTSA TOF technique to avoid the saturation effects of 3D TOF and to facilitate peripheral vessel depiction. However, knowledge of the limitations of TOF-MRA is important so that artifactual stenoses and occlusions are recognized. Particularly in the region of the skull base and sphenoid sinus, susceptibility artifact due to bone and air may cause artifactual signal loss in the adjacent internal carotid artery. Intravoxel dephasing due to disturbed flow through the carotid siphon may also simulate stenosis, with a further reduction in the specificity of this technique if unrecognized. The importance of a complete data review, particularly of the source data images, has been repeatedly stressed in the literature,38 so that these artifacts are not misinterpreted.
Recently, hardware technology has evolved to the point at which image quality with CE-MRA may rival or exceed that of TOF-MRA in the supra-aortic circulation. Several studies have addressed the usefulness of CE-MRA at 1.5T in the carotid39 and vertebrobasilar circulations.40 The increased SNR and gadolinium (Gd) sensitivity at 3T can be used to support aggressive parallel acquisition strategies during the first passage of a Gd bolus, resulting in even higher spatial resolution.29,32 It could be argued that, with advanced machine hardware, CE-MRA may supersede TOF-MRA for most applications in the head and neck.
Phase-contrast MRA
Phase-contrast MRA (PC-MRA) involves the application of bipolar phase-encoding gradient pairs, to encode velocity in the direction of the gradient.41 As a result, spins moving along the direction of the gradient field undergo a phase shift proportional to their velocity. Conversely, stationary tissue accumulates zero net phase (ideally). Subtraction of these flow-sensitive data sets from reference images allows visual representation of vascular flow and, thus, angiographic depiction. Importantly, as the phase shift experienced is proportional to the velocity of moving spins, phase-contrast imaging allows quantitative assessment of flow velocities (phase-contrast flow quantification—Figure 10).
Phase can have values only between +180˚and –180˚. Typically, one half of the phase spectrum is assigned to flow in one direction and the other half to flow in the opposite direction. The flow sensitivity is determined by the gradient settings and is expressed by the velocity encoding value (Venc). The Venc is defined as the flow velocity that will result in a phase shift of 180˚. Ideally, the Venc is set for exactly the highest velocity flow expected in the vessel of interest. For slow flow, the Venc is set lower (eg, the portal vein may be 30 cm/sec), whereas for faster flow the Venc is set higher (eg, the aorta or just distal to a stenosis may be >200 cm/sec).
If the Venc chosen is too low, then velocity aliasing (wraparound) will occur, whereas if the Venc chosen is too high, the vascular contrast-to-noise ratio (CNR) will be low. Given that the greatest difference achievable between the flow-compensated and flow-sensitive data sets acquired during application of the bipolar gradient pairs occurs in opposite directions (ie, 180˚), phase shifts greater than the preselected Venc factor will be misrepresented in the image obtained. Therefore, this parameter must be carefully chosen so that artifactual reductions in signal intensity are not experienced, using higher Venc factors (>100 cm/sec) for regions of rapid arterial flow and lower Venc factors (20 to 30 cm/sec) for venous applications.42
Although PC-MRA can be made less vulnerable to saturation than TOF-MRA by choosing a small flip angle, there is still some T1 dependence, and the overall CNR can be increased by contrast injection. Gadolinium injection also makes it possible to use very short TRs, which decreases the acquisition time.
As is the case with TOF-MRA, PC-MRA may be applied as either a 2D projectional technique or a 3D technique. However, this latter volumetric technique is a time-consuming process, often requiring >20 minutes for image acquisition. Furthermore, PC-MRA is vulnerable to intravoxel dephasing, as is TOF-MRA. However, PC-MRA has the advantage of allowing interrogation of flow direction and improved visualization of slow-flow vessels even in regions with complex flow patterns.
Clinical applications
The potential application of PC-MRA to a variety of vascular territories has been evaluated, with generally satisfactory results.43-45 The strength of this method lies in its ability to depict directional flow, as well as allowing quantitative analysis of flow characteristics. This is of particular value in the assessment of arteriovenous malformations, for which data acquisition using carefully selected Venc factors provide both arterial-and venous-phase images on sequential measurements with different Venc values. With the recent development of rapid high-resolution volumetric contrast-enhanced MRA techniques, however, PC-MRA is often reserved for research and “problem-solving” applications (Figures 11 through 13).
Steady-state free-precession MRA
Steady-state free precession (SSFP) exploits the coherent magnetization recoverable when TR <<T2. Steady-state free-precession MRA may generate rapid, bright-blood images without the requirement for Gd administration. With this technique, the images are not dependent on flow to generate vascular contrast, which is reflected more in the T2/T1 ratio of the blood. Also, SSFP techniques operate at very short TR (~3 msec) and large flip angles, generating bright-blood 2D images in <1 second. For this reason, breath-holding is not required for single-shot SSFP, and it may be used effectively for the evaluation of abdominal veins and the abdominal aorta. For the evaluation of the thoracic aorta (eg, to rule out dissection or aneurysm), electrocardiographic (EKG)-gated “segmented” SSFP cine has become a cornerstone technique.47 Cine images can be acquired in a few seconds, and serious disease of the aorta can be ruled out or confirmed in as little as a few minutes.48 Three-dimensional implementations of SSFP have been used successfully for imaging the coronary arteries, both in breath-hold49 and free-breathing formats.50 When combined with navigator-gating, free-breathing nonenhanced MRA in regions such as the chest becomes clinically feasible. However, SSFP-MRA is sensitive to off-resonance artifacts, which manifest as dark bands and ghosts that can be exacerbated by flow and motion. Off-resonance artifacts are more troublesome at higher field strengths, though a number of approaches have been described that address this challenging problem.51
Clinical applications
As mentioned above, EKG-gated 3D SSFP-MRA has shown promise for coronary artery imaging, and CE-MRA of the coronary artery is challenging because of the conflicting requirements of first pass, rapid acquisition, and volume coverage.49,51,52 Moreover, free-breathing angiography has performed well in a number of clinical scenarios, including aortic,53 renal,54 carotid,55 and peripheral arterial imaging56 (Figures 14 through 16).
Contrast-enhanced MRA (CE-MRA)
Contrast-enhanced MRA has become one of the most powerful and useful applications of MRI in recent years. It involves the intravenous administration of a bolus of a Gd-based T1-shortening contrast agent, which dramatically increases the signal intensity of blood on the first pass.57 The blood signal greatly exceeds that of nonvascular background tissue on 3D T1-weighted ultrafast gradient-echo pulse sequences, such that high CNR angiograms are generated.
Contrast-enhanced MRA is based on the presence of Gd in the blood, independent of the flow waveform and flow velocity in the vessels. It does not rely upon the physiologic flow of in-phase spins for signal generation and is free from many of the limitations of TOF-MRA and PC-MRA that were considered above.6 The basic pulse sequence for CE-MRA is an ultra-short TR/TE 3D spoiled gradient-echo sequence and has a relatively simple structure. The TR is made as short as possible to speed up image acquisition and capture the first pass of contrast. The presence of Gd makes the blood immune to saturation, so the short TR tends to saturate only background tissue. In general, the highest flip angle achievable by the MRI machine is used, which is consistent with RF-specific absorption rate (SAR) limits that are mandated by the Food and Drug Administration. The TE is also made as short as possible—ideally approximately 1.0 msec, so as to limit any flow-induced dephasing effects. For this reason, vessel diameters measured on high-quality CE-MRA studies tend to have far less exaggeration of stenosis due to turbulent flow. Several examples are shown below.
However, CE-MRA is highly dependent on accurate timing of the contrast bolus. Attention must be paid to the following factors: 1) Coordinating image acquisition with the arrival of the contrast bolus within the vessels of interest; 2) Filling the center of k-space during the peak period of contrast enhancement; 3) Using the smallest Gd dosage possible that will produce diagnostic image quality; 4) Obtaining high-resolution arterial data acquisition while minimizing detrimental venous enhancement; and 5) Preventing patient motion during acquisition, particularly respiratory motion during chest and abdominal imaging.
A number of methods are used to time the bolus, including the use of a fixed timing delay, a test bolus, repeated multiphase scanning, and real-time fluoroscopic detection of contrast arrival.58 Our experience has been that the injection of a small test bolus yields the most consistent results, confirms the integrity of the intravenous access site, and facilitates planning.
The development of parallel acquisition techniques59,60 has provided a spring-board for immense advances in the performance of CE-MRA applications. Concurrent technical advances in RF hardware have facilitated parallel data acquisition, which involves undersampling of k-space, and coil sensitivity profiles are used to calculate the missing k-space data.61 Parallel acquisition invokes a penalty in SNR, a limitation that may be mitigated by imaging at higher field strengths such as 3T. As a result, high spatial resolution imaging over large fields of view during a fraction of the time required for full k-space sampling has become a routine tool, greatly advancing the clinical utility of CE-MRA. Indeed, the high sensitivity and specificity of this technique has been convincingly shown when imaging the carotid arteries, the renal arteries, and the thoracic and abdominal aorta in both children and adults.62-64 Furthermore, the development of complex RF surface coil arrays with up to 32 receiver channels (or more) has resulted in high-performance MRA from the carotid arteries to the calf trifurcation vessels.65
A specific implementation of CE-MRA worthy of special mention is time-resolved MRA, during which a fast gradient-echo 3D sequence can clarify the temporal sequence of vascular and cardiac chamber enhancement.66,67 Three-dimensional data sets can be acquired within 1 to 2 seconds and repeated for as long as required to image dynamic changes. The performance of 3D time-resolved MRA can be increased by the use of parallel acquisition and temporal data sharing. Time-resolved echo-shared angiographic technique (TREAT)68 or time-resolved imaging of contrast kinetics (TRICKS) combining view sharing69 and parallel imaging allows dynamic imaging at high spatial and temporal resolutions following the administration of contrast doses as low as 0.05 mmol/kg.70 Such high temporal resolution allows confident separation of the arterial and venous phases of contrast enhancement and can provide information regarding directional flow and visceral perfusion as well as quantify curve parameters such as time-to-peak (TTP) signal intensity, maximal upslope of the curve (MUS), maximal signal intensity (MSI), and mean transit time (MTT).70 Furthermore, when used at 3T, this technique provides both functional and anatomic information, allowing a reduction in the dose of Gd administered. The results of recent research offers promise for very high performance in time-resolved MRA using vastly undersampled projection reconstruction imaging.71,72
Clinical applications
The spatial resolution of modern CE-MRA continues to increase to such a degree that confident visualization of peripheral vessels measuring only 2 to 3 mm is now possible. As a result, CE-MRA has replaced CCA in many applications, allowing an alternative, less invasive, and clinically viable approach to vascular imaging. For this reason, CE-MRA has been successfully applied to every anatomic region, such that, in the absence of specific contraindications, its use should always be considered (Figures 17 through 19).
Established applications for CE-MRA include: 1) Atherosclerotic arterial occlusive disease that involves the carotid, abdominal aortic, renal, mesenteric, and peripheral extremity vessels; 2) Follow-up of the patency of surgically created vascular bypass grafts; 3) Assessment of pulmonary arterial patency in the presence of suspected acute or chronic pulmonary thromboembolic disease; 4) Exclusion of acute arterial insult, such as intramural hematoma, rupture, or dissection; 5) Diagnosis and follow-up of systemic vasculitis; 6) Investigation of the presence and extent of arteriovenous malformations, including distribution, soft tissue involvement, circulations
in-volved, and degree of shunting. 7) Preoperative planning for a variety of surgical techniques, including donor assessment, considerations of thoracic and abdominal aortic
aneurysms for endovascular repair, and evaluation of lower extremity vasculature prior to fibular free-flap creation; and 8) Diagnosis of central congenital vascular anomalies, including coarctation of the aorta and anomalous vessel origins and courses.
Discussion
Over the past several years, MRA has obviated the need for conventional catheter angiography in a variety of clinical situations. The advent of parallel imaging has helped to consolidate the role of CE-MRA as a practical diagnostic tool. Nonetheless, the versatility of non-enhanced techniques (including TOF-MRA for the intracranial circulation and 3D SSFP-MRA for much of the remaining vasculature) renders them viable alternatives to CE-MRA in patients in whom a contrast agent is contraindicated (eg, in pregnant patients), refused, or undesirable.
Recent reports that link Gd administration to a rare disorder—nephrogenic systemic fibrosis—in patients with renal failure has generated concern about the use of CE-MRA in this patient population. Most cases to date have occurred in the context of dialysis-dependent patients who received high-dose Gd.73,74 At the time of this writing, it is unclear what the ultimate implications of the association will be regarding the use of Gd contrast agents in this patient population. In the meantime, it would seem appropriate to minimize the dosage of Gd in susceptible patients. In instances in which the clinical indications for CE-MRA are strong, imaging at higher field strengths (such as 3T) permits reduction in the dosage of Gd contrast agent administered. Although much research remains to be done in this area, it may be that the requirement for dose reduction generates an additional impetus to perform CE-MRA at 3T, if available.
As illustrated in Table 1, each MRA technique has specific strengths, clinical indications, and limitations, and, as a result, no single technique is optimal for all applications. Despite recent concerns regarding the use of Gd-based contrast in patients with severe renal impairment, these agents enjoy an impressive safety record, even compared with iodinated X-ray contrast agents.
As MR technology continues to evolve, newer and more powerful implementations of both contrast-enhanced and non–contrast-enhanced MRA will undoubtedly emerge.
REFERENCES