Dr. Mong is a third-year Resident in Diagnostic Radiology at the Hospital of the University of Pennsylvania, Philadelphia, PA. He received his BA in Biological Sciences from Swarthmore College, Swarthmore, PA in 1995 and his MD from the University of Pennsylvania School of Medicine, Philadelphia, in 1999. He plans to begin a Fellowship in Pediatric Radiology at the Children's Hospital of Philadelphia in July 2004.

Recent advances in computed tomographic angiography (CTA) and magnetic resonance angiography (MRA) enable comprehensive evaluation of the supra-aortic vascular tree, including cervical and intracranial vessels. Both CTA and contrast-enhanced MRA provide highly accurate assessment of severe carotid artery stenosis, and are an improvement over traditional time-of-flight MRA in the depiction of great-vessel origins and the cervical vertebral arteries. Both show promise in acute stroke.

Comprehensive radiologic examination of both the intracranial and extracranial supra-aortic vessels is a valuable part of the evaluation of stenotic or occlusive disease, under a variety of clinical conditions. In an acute cerebrovascular accident, for example, pertinent information can be found not only at the site of arterial thromboembolism but also at its source. Occlusions at some sites, such as the cervical internal carotid artery, are not amenable to intra-arterial thrombolysis. ¹ At other sites, including the distal intracranial vessels, occlusions may be more likely to benefit from intravenous (IV) tissue plasminogen activator than would occlusions in the proximal intracranial vessels. ²

Arterial dissection is a common cause of stroke in young patients. A comprehensive examination of the supra-aortic vessels can detect dissection, sometimes at multiple levels of both the carotid and vertebral arteries. ³ In patients under consideration for carotid endarterectomy, determination of the percent of stenosis is important, as this finding correlates with surgical benefit according to the North American Symptomatic Carotid Endarterectomy Trial (NASCET). ⁴ In addition, tandem lesions along the course of the internal carotid artery (ICA) may be a contraindication to surgery. Atherosclerosis at the origin of the great vessels may change therapy in stroke patients as well. ⁵ For all of these reasons, it is essential to image the entire length of the carotid and vertebral vasculature, from the aortic arch to the circle of Willis.

Digital subtraction angiography (DSA) remains the gold standard for the evaluation of steno-occlusive disease of the supra-aortic vasculature, and has the added advantage of enabling directed interventions, including intra-arterial thrombolysis of an identified occlusion, or in some cases, angioplasty and stenting of a stenosis. However, DSA has inherent limitations; it is costly, time consuming, and involves opacification of the vessel lumen only, without providing information about the vessel wall or perivascular soft tissue. In addition, the procedure presents a risk of neurologic deficit, with 0.4% risk of permanent deficit and 2.3% risk of any neurologic complication. ⁶

Advances in both computed tomographic angiography (CTA) and magnetic resonance angiography (MRA) have sought to address the limitations of DSA and provide a noninvasive alternative for the comprehensive evaluation of the supra-aortic vessels. The purpose of this review is to discuss recent developments in CTA and MRA and the ability of these technologies to depict stenoses and occlusions. There are inherent contraindications to each of the techniques, including, for CTA, contrast-related nephrotoxicity and allergy, and for MRA, incompatibility with metal and electronics implanted in the patient's body. Many patients, however, will be eligible for both, leaving the clinician and radiologist to select the most appropriate study.

Technique: CTA

CT angiography begins with the IV administration of iodinated contrast material, typically 100 to 150 mL of iodinated contrast material (300 mg I/mL) at 3 to 4 mL/sec. Adequate timing of the bolus is critical to obtain maximum contrast between the vessel lumen and surrounding tissues during data acquisition. This may be accomplished with an empiric "best-guess" delay, a test-bolus technique, or automatic triggering.

The empiric strategy is based on the patient's age, weight, and cardiac status; this strategy is the least successful in optimizing contrast between tissues, although it is still adequate in many cases and is often used with a 30-second delay when only the intracranial vessels are to be imaged. ⁷ The test-bolus technique involves the injection of a small amount of contrast, typically 20 mL, prior to diagnostic image acquisition, while measuring the change in attenuation within the vessel of interest over time. The diagnostic examination may then be performed to coincide with the calculated peak contrast enhancement. With the automatic triggering technique, a threshold value of contrast enhancement within the vessel of interest is predetermined, typically 335 to 355 HU. Data acquisition begins with repeated scanning over the vessel of interest. When the threshold of enhancement is reached, diagnostic image acquisition is triggered manually or automatically by the computer.

The recent development of multi-detector scanners has revolutionized CT data acquisition. The limited heat load capacity of the X-ray tube precluded optimal examination of both extracranial and intracranial circulation during the same examination for single-row detector systems. ¹ Multidetector scanners and increases in gantry rotation speed have translated into faster data acquisition. This allows coverage of both the extracranial and intracranial circulation during the same examination without overheating the X-ray tube. This increase in detector number has also allowed for better resolution. Sixteen-detector scanners have the capability of acquiring data while maintaining near-isotropic voxel dimensions. Acquisition of near-isotropic voxels enables reformatting of data and, therefore, display of vessels in any dimension without significant loss of resolution, regardless of the plane of acquisition. ⁸ Typical parameters for CTA of the cervical vasculature include a pitch of 6 and rotation time of 0.8 seconds, with mA 220, kV 120, slice thickness of 1.25 mm, and reconstruction interval of 1.0 mm.

Images may be reformatted to display the entire length of vessels and their anatomic relationships. Current techniques include multiplanar reformatting, curved planar reformation, limited- and full-volume maximum intensity projection (MIP), and shaded-surface display. Limited and whole volume MIPs are most often used in standard practice. They clearly present the data in a format analogous to conventional angiography for referring clinicians. Although such reconstructions may be valuable visual aids, they do not include all data inherent in the source images. It becomes imperative, therefore, to view the original axial images.

Technique: MRA

Traditionally, MR examination of the supra-aortic vessels has been performed using two-dimensional (2D) time-of-flight (TOF) and three-dimentional (3D) TOF sequences for the cervical vasculature, and a 3D TOF sequence for the intracranial vasculature. These techniques rely on flow-related enhancement, rather than the injection of IV contrast. Specifically, gradient-echo images are obtained in TOF angiography, which contrasts dark saturated stationary tissue with inflowing bright unsaturated intravascular blood. Image acquisition may include an entire volume of tissue (3D) or contiguous sections (2D) This technique has been limited by long acquisition times, which can lead to motion artifact, as well as saturation-related signal loss due to in-plane flow on 2D imaging. Disordered or turbulent flow also presents problems, including dephasing of spins, which leads to a decrease in signal intensity. As stenosis within a vessel increases, there is progressive loss of signal within the post-stenotic segment, which leads to exaggeration of severity and extent of the stenosis. ⁹

Three-dimensional contrast-enhanced (CE) MRA involves the injection of paramagnetic contrast material, which shortens the T1 relaxation time of tissue within its vicinity. This technique greatly shortens examination time and decreases the problem of motion inherent in TOF sequences. As with CTA, precise timing of the bolus is critical for achieving adequate contrast and may be accomplished with an empiric-timing technique, a test bolus, or automatic threshold triggering. Parameters for 3D contrast MRA of the carotid arteries might include repetition time <5 msec, echo time <2 msec, flip angle 30š to 45š, field of view 280 * 280 * 80 mm, and matrix 512 * 256 * 48, obtained after injection of gadolinium, typically 0.1 to 0.3 mmol/kg, usually >2 to 3 mL/sec. ⁷

Time-resolved MRA is an alternative to methods that depend on the precisely timed delivery of a contrast bolus. This technique, which involves rapid continuous sampling of frames of data after contrast administration, with each frame acquired over 2 to 3 seconds. This technique produces multiple 3D data sets. It is then possible to retrospectively reconstruct the data set containing peak arterial enhancement. As an added benefit, temporal resolution may provide information on the hemo-dynamic significance of a stenosis, such as in subclavian steal phenomenon.

Time-resolved techniques have inherently less spatial resolution, however. One time-resolved technique that seeks to address the trade-off between resolution and speed of acquisition is time-resolved imaging of contrast kinetics (TRICKS). Time-resolved imaging of contrast kinetics samples the center of k-space, which contains phase-encoding contrast information, earlier and more often than the periphery of k-space, which contains frequency-encoded spatial information. Sampling predominantly the center of k-space during peak arterial enhancement maximizes contrast by allowing acquisition of phase-encoded contrast information to coincide with peak arterial enhancement. Because only a portion of k-space is being sampled at one time, faster frame rates can be generated, which virtually guarantees that one of the captured frames will contain optimal enhancement that is free of venous contamination. Frequency-encoded spatial information can be acquired later in the scan, after peak arterial enhancement has passed, and then combined with the phase-encoded data to increase resolution. ⁷ Regardless of the bolus-timing technique, contrast-enhanced data are often acquired in the coronal plane using a head-and-neck coil to assess the entire length of the carotid and vertebral systems. Three-dimensional gradient-echo pulse sequences with low repetition and echo times are used. Scanning can be accomplished in a 20- to 40-second breath-hold. As with CTA, post-processing of MRA images may involve the production of MIPs or volume-renderings, with MIPs most often useful to define the length of the vessels and provide an image analogous to conventional angiography for referring physicians.

Evaluation: Cervical carotid arteries

Both multidetector CTA (MDCTA) and CE MRA represent an improvement over 2D TOF MRA in the evaluation of the anterior extracranial vasculature. Specifically, both techniques can depict the origins of the great vessels more accurately, which are not optimally imaged with TOF, as a result of motion artifact.

More attention has been paid to stenosis of the carotid bifurcation. NASCET data demonstrated that patients with at least a 70% (severe) symptomatic stenosis of the ICA on DSA benefited from carotid endarterectomy. ⁴ With a normal lumen of the ICA measuring approximately 7 mm, a high-grade (70%) stenosis would measure only 2.1 mm. ¹⁰ Resolution on this scale must therefore be demonstrated by any modality prior to its use as a screening tool.

While not the focus of this review, 2D ultrasound has demonstrated efficacy in the detection of severe carotid stenosis, though it is often difficult to mentally reconstruct 3D plaque from 2D data. Newer 3D ultrasound appears particularly promising, with 90% sensitivity and 92% sensitivity for detection of severe carotid stenosis. ¹¹ More attention has been focused in the literature, however, on the role of CTA and MRA, perhaps because these modalities are not operator-dependent.

Both MDCTA and CE MRA demonstrate high accuracy in the detection of severe carotid artery stenosis (>70%), as compared with DSA. Recently presented data show that MDCTA has a sensitivity of 98% to 100% and a specificity of 96% to 100% for the detection of severe stenosis, while CE MRA demonstrates a sensitivity of 93% to 98% and a specificity of 96% to 100%. ^12-15 Figures 1 and 2 compare the depiction of stenoses by CTA, CE MRA, and traditional TOF MRA.

In addition to stenosis severity, accurate description of plaque morphology provides valuable information, as plaques that are more prone to disruption, fracture, or fissure are more likely to cause symptoms. ¹⁶ Initial data suggest that MDCTA may be more sensitive to abnormalities in plaque morphology than DSA or CE MRA. In a recent study of 22 patients who underwent MDCTA, CE MRA, and DSA, all 11 plaque irregularities described on DSA were seen on MDCTA, plus an additional 5. By comparison, CE MRA detected only 9. MDCTA may also be more sensitive than CE MRA for the detection of ulcerated plaque (sensitivities, 96% and 89%, respectively). ¹²

Both CTA and MRA have demonstrated high accuracy for the detection of carotid artery dissection, a significant cause of stenosis in young patients. This is of particular importance when DSA reveals only acute occlusion of the carotid artery, a finding that has a number of causes. CT angiography has demonstrated an accuracy of 100% for the detection of dissection, as compared with DSA. ¹⁷ Contrast-enhanced MRA may have little to offer over TOF MRA, which has a sensitivity of 95% and a specificity of 99%. Fat-saturated T1-weighted images or black-blood fast-spin-echo images on MR are also extremely useful, demonstrating high-signal eccentric hematoma around a dark flow void. ¹⁸

Accurate assessment of carotid artery stenosis is necessary not only in surgical planning but in follow-up studies of patients who have undergone implantation of an endovascular stent, as these devices are subject to subsequent intimal hyperplasia and thrombus formation. With CTA, an MIP is often inadequate for the assessment of stent stenosis, as the stent and the intravascular contrast are too similar in attenuation to be distinguished. Thin MIPs or oblique reformats may be helpful, as shown in Figure 3. A volume-rendering postprocessing algorithm may be used, with a maximal opacity assigned to the contrast material and a minimal opacity assigned to the stent. This technique leads to an overestimation of stent wall thickness, probably as a result of partial volume effects, but has nonetheless been shown to be accurate in the small number of patients in whom its use has been reported. ¹⁹

Stent patency has traditionally not been assessed with 2D TOF MRA, as stent geometry and metal composition create significant susceptibility artifact and radiofrequency shielding. Stent patency is often merely inferred on the basis of the signal proximal and distal to the artifact. Contrast-enhanced MRA is currently under investigation for the assessment of stent patency and may be possible with the use of higher flip angles (75° to 150°), to increase the conspicuity of blood, and a shorter echo time (<2 msec), which may reduce artifact. ⁷

Many stents demonstrate an artificial narrowing and a reduction in signal intensity, despite contrast injection. Even when it is possible to demonstrate the vessel lumen, a band-like artifact can be observed at the ends of the stent. A recent in vitro analysis of several nitinol stents and a tantalum stent demonstrated good performance for the nitinol Passager Stent graft (Boston Scientific, Natick, MA), the nitinol Memotherm iliac stent (Angiomed Bard, Karlsruhe, Germany), and the tantalum Strecker stent (Boston Scientific). The reduction in signal intensity and lumen width within the stents ranged from 0% to 34%. ²⁰

Evaluation: cervical vertebral arteries

Normal anatomical variations, such as kinking and coiling of the vertebral arteries, make the posterior circulation difficult to assess by all modalities. CT angiography performs as well as, or perhaps better than, DSA in depicting the ostia of the vertebral arteries, and has a high accuracy for the detection of stenosis. ²¹ By comparison, TOF MRA either fails to demonstrate the ostia of the vertebral arteries, or overestimates the degree of stenosis in this area. Often there is loss of signal in the vertebral arteries at the level of the C1­C2 disk space, as the arteries flow within the imaging plane. In the remainder of the cervical vertebral arteries, 2D TOF MRA demonstrates a sensitivity of 92% to 100% and a specificity of 90% to 96% in evaluating severe stenosis or occlusion. ²²

As with CTA, CE MRA generally can eliminate the artifact at the vertebral ostia that is seen on TOF MRA. ⁷ Contrast-enhanced MRA demonstrates excellent sensitivity and specificity in the cervical vertebral arteries, detecting >50% stenoses with a sensitivity of 100% and a specificity of 96%, and detecting complete occlusions with a sensitivity and specificity of 100%. ²³

Vertebral artery dissections have not been well studied with newer techniques. Fat-saturated T1 dark-blood sequences often fail to show the eccentric hematoma that is characteristic of dissection, probably because of the small diameter of the vessel and the high signal of the surrounding normal tissue. Initial reports with MDCTA or a combination of CE MRA and TOF MRA are promising for the detection of long-segment "pearl-and-string stenosis," but DSA remains important for diagnosis. ²⁴

Evaluation of the vertebral arteries is often of clinical concern when subclavian steal is suspected. The steal phenomenon arises from occlusion in the proximal subclavian artery, with arterial blood to the affected arm supplied by the ipsilateral vertebral artery distal to the stenosis. This results in reversal of flow in the vertebral artery and subsequent vertebrobasilar symptoms during greater demand of blood to the extremity, such as during exercise. MR angiography appears especially suited to the diagnosis of this entity, with contrast-enhanced techniques providing accurate degree of stenosis within the subclavian artery, and flow techniques, such as from time-resolved MRA, demonstrating direction of flow within the vertebral artery. ²⁵ CT angiography, which does not provide information on directional flow, may therefore yield false-negative results. ²⁶

Evaluation: intracranial vasculature

Stenotic disease of the intracranial vasculature has not been studied extensively, perhaps because there is little to offer in the way of intervention. Time-of-flight MRA depicts moderate-to-severe stenosis in the ICA and MCA with a sensitivity of 100% ²⁷ (Figure 4). More attention has been given to acute occlusive disease, however. Resistance to the use of MRA or CTA prior to DSA in the evaluation of acute cerebrovascular accident is often based on the assumption that valuable time will be lost in delivering definitive treatment. Any analysis of the role of CTA and MRA must, therefore, include an assessment of the feasibility of their use in acutely ill patients. However, information provided by CTA or MRA on the site of occlusion may enable quicker and more accurate targeting of the affected vessel during subsequent DSA. ²⁸

Excluding postprocessing time, CTA adds minutes to a standard non-contrast CT examination, with one author reporting <5 minutes for injection and scanning time combined. ²⁹ Time-of-flight MRA sequences require 5 to 10 minutes, and CE MRA takes even less time. Practically, however, any MR examination will follow at least a noncontrast CT examination (NCCT), as a NCCT is the standard first-line diagnostic tool of the emergency room used to exclude hemorrhage. An average time of 40 minutes is reported from patient arrival to initial CT exam, and 1.7 hours from CT to MR exam. This delay between examinations can be reduced over time with additional experience and training. ^{28, 30} However, as MR is not viewed in many institutions as critical to the management of stroke in the acute stage, there is movement toward assessment with CTA combined with perfusion CT in the emergency room.

Examination of the accuracy of CTA and MRA for acute occlusion is complicated by possible lysis or extension of the clot during the delay between the noninvasive imaging examination and subsequent DSA. With this caveat, MDCTA has demonstrated occlusions to the level of the M3 branches, while simultaneously depicting occlusions, stenoses, or floating thrombi in the extracranial circulation. In the examination of large-vessel occlusion, MDCTA has demonstrated a sensitivity of 91% to 100% and a specificity of 93.7% to 100%. ^28,31 These values should be tempered by initial data from Mullins et al, ³² who performed a retrospective analysis of 479 consecutive patients with acute stroke who underwent CTA and CT perfusion imaging. While the investigators report a high specificity (93%), the sensitivity of MDCTA in diagnosing acute infarction was considerably lower (57%). ³²

Traditionally, studies of MRA in acutely ill patients have focused on TOF sequences rather than contrast enhancement, with TOF demonstrating a sensitivity of 80% to 100% and a specificity that is comparable to that of DSA. ^33-35 When combined with diffusion-weighted imaging, TOF MRA has a diagnostic accuracy of 94% in determining stroke subtype (ie, large-vessel occlusion, cardioembolic disease, small-vessel disease, or stroke of other etiology), as gauged by the patient's final diagnosis at hospital discharge. ³⁶ Contrast-enhanced MRA, by comparison, requires further evaluation. Isoda et al ³⁷ reported acceptable viewing of the arterial phase of enhancement without venous contamination only 77.6% of the time, suggesting that further improvement is necessary before CE MRA can replace TOF sequences. Evaluation of the intracranial vertebral arteries is problematic, as both CTA and MRA have difficulty in distinguishing high-grade stenosis from hypoplasia. In a small number of patients, the accuracy of CTA as compared with DSA was only 66% for stenosis >50%, and only 50% for total occlusion. This same series demonstrated 100% accuracy for detecting stenosis and occlusion of the basilar artery, however. ³⁸

Similarly, intracranial vertebral arteries are difficult to assess with TOF MRA (sensitivity, 84%; specificity, 82% to 93%). The reduced accuracy of TOF MRA in this region may be due to the use of separate head-and-neck coils, an approach that places the intracranial vertebral arteries at the junction of two separate examinations for the head and the neck. In the evaluation of the basilar artery, TOF MRA demonstrates a sensitivity of 100% and a specificity of 78% to 88%. ²² Evaluation of the intracranial posterior circulation with CE MRA remains to be elucidated, but accurate visualization of the intracranial portions of the vertebral arteries is possible in most patients in a single examination of the supra-aortic vessels. ³⁹

Conclusion

Advances in both CTA and CE MRA enable assessment of the entire supra-aortic arterial tree, from the aortic arch to the circle of Willis. Each of these techniques is extremely valuable in the evaluation of steno-occlusive disease. Each demonstrates high accuracy in the evaluation of the carotid bifurcation, for example, although CTA has a slightly better resolution and often more accurately depicts plaque. Both CTA and CE MRA deserve to be considered before DSA, which could be reserved as a problem-solving tool.

Although CTA and CE MRA more accurately evaluate the ostia and the extracranial portions of the vertebral arteries, representing a marked improvement over TOF MRA, they demonstrate mixed results in evaluation of the posterior circulation. In acute intra-arterial occlusion, both technical and diagnostic issues have limited the use of CTA and CE MRA prior to DSA. Quick CTA scanning, however, is an attractive method of stroke evaluation prior to DSA or the initiation of therapy, as it adds little time to a noncontrast CT scan, and is gaining in popularity as physicians realize the strides that have been made in terms of resolution and accuracy.

Acknowledgment

The author thanks Ron Wolf, MD for reviewing this manuscript and for providing CTA and MRA images.