Dr. MacEneaney is a Fellow and Dr. Dachman is an Associate Professor in the Department of Radiology, University of Chicago Hospitals, Chicago, IL.

Computed tomography angio-graphy (CTA) is a noninvasive method of assessing the vascular system; in some cases it has replaced conventional angiography. The development and widespread availability of helical CT and advanced post-processing techniques has made CTA a practical alternative. The rapid acquisition speed of helical CT allows a large volume to be imaged during the narrow temporal window of first pass peak vascular opacification during a single breath hold. The ongoing development of multidetector scanners will further increase CTA capabilities. The volumetric data sets generated are suitable for three-dimensional (3D) reconstruction. This process is becoming easier and faster with the advent of more powerful computer processors and software in commercially available workstations. Despite the rapid progress in MR angiography, the widespread use of helical scanners ensures the future of CTA.

CTA has several advantages over the other vascular imaging modalities. The examination is minimally invasive. High-rate/high-pressure delivery of contrast is essential but only venous access is required and the patient is spared the potential complications of conventional catheter angiography. Unlike MR, pacemakers and ferromagnetic aneurysm clips do not preclude CTA. Artifacts induced by turbulent blood flow through stenoses may cause over estimation of the degree of stenosis on sonography and MR. This does not influence CTA, although with suboptimal scanning protocols CTA can also overestimate stenoses. Important nonvascular pathology may be diagnosed. ¹ Renal impairment and contrast hypersensitivity may limit CTA in some patients.

Scan protocols

One of the relative disadvantages of spiral CT is poor z-axis resolution. This is responsible for partial volume effect and stairstep artifacts encountered on multiplanar and 3D reconstructions. Since many CTA applications require 3D and multiplanar reconstructions, optimization of z-axis resolution is vital. To minimize these problems and maximize vascular enhancement, careful attention should be paid to CTA scanning and reconstruction protocols. Performance is based on collimation, pitch, reconstruction spacing, and timing of scan with respect to contrast injection. ^2-4

Selection of appropriate collimation and pitch is influenced by several variables, including patient specific characteristics such as breath hold capacity, quality of venous access, and the examination specific characteristics, particularly the size of the vessel and region of interest. There is a trade-off between collimation, pitch, and the volume to be scanned. Narrow collimation and low pitch improve the z-axis resolution, but increase scan time and tube heating. First, choose a collimation less than the diameter of the vessel of interest and then increase the pitch to cover the distance required, ⁵ e.g., 3 mm for renal arteries, 7 to 10 mm for the aorta. If the desired scan is large and it is necessary to increase either the pitch or collimation, it is preferable to increase the pitch. ⁶ Increasing pitch from 1 to 2 increases the effective slice thickness by one third. A pitch of 2 is the practical upper limit.

Scanning at peak vascular enhancement is vital: as pitch increases, low-contrast resolution decreases and peak opacification simplifies 3D reconstructions. Appropriate timing optimizes contrast and allows either arteries or veins to be scanned. Scanning the abdominal aorta at 25 seconds after the start of the injection will usually give satisfactory images. In a healthy patient, empirical timing may suffice; however, time to peak enhancement varies particularly in patients with cardiac compromise. ⁷

There are two methods to ensure correct timing. The first uses a series of static scans at a single level of interest, i.e., the aortic arch for cases of suspected dissection and the level of the renal arteries for assessment of abdominal aortic aneurysms. Fifteen to 20 mL of contrast is injected at the anticipated injection rate and a scan performed every 3 seconds, beginning 10 seconds after injection. A region of interest is then placed on the vessel and the time to peak vascular enhancement is calculated. The second method is semi-automated. A series of scans is performed at the start of contrast injection. Image acquisition is triggered when a preset enhancement threshold is reached within a preselected region of interest. Visual assessment can be used to override the semi-automated software and commence scanning earlier.

Contrast is typically injected at 3 to 5 mL/sec. Higher rates of injection are preferred for identification of smaller vessels. Contrast volumes range from 100 to 180 mL. Non-ionic 300 mg/mL contrast is used, except for cases of suspected pulmonary embolism or dissection of the thoracic aorta. In these instances, lower concentrations of contrast media are used (150 to 240 mg/mL), as dense contrast medium may obscure low density thrombus or an intimal flap. Saline bolus chasers of 50 to 70 mL are employed to force the contrast through the tubing and the venous system to prolong the plateau phase of enhancement. ⁸

Axial resolution is approximately 1 mm, however, longitudinal resolution is poorer. Two parameters partially compensate for the relatively low longitudinal resolution, and so reduce partial volume artifacts and stairstep artifacts on multiplanar reformatting (MPR) and 3D reconstructions. One is the use of 180° rather than 360° interpolation algorithms (vide infra), and the other is reconstruction of overlapping images. Reconstructing 2 to 3 images per rotation is recommended. ² A useful rule of thumb is to reconstruct images at one-third the collimation. Highly overlapped reconstructions are required when the vessel of interest runs in the axial plane.

Initially, interpretation should be done on the axial source images. We recommend interactive reading at the console to facilitate use of overlapping reconstructions and minimize film waste. By displaying findings in a more familiar format, 3D reconstruction can help in difficult cases by revealing information present, but not readily evident on axial images alone.

Postprocessing techniques include MPR, maximum intensity projection (MIP), shaded-surface display (SSD), and volume rendering. MPR is the simplest technique and is an adjunct to viewing axial images for problem solving.

In SSD, a threshold attenuation level is selected by the radiologist. All voxels in the data set with a Hounsfield value within this threshold are displayed as a single structure. Relative depth is provided by shading from a computer-generated light source (figure 1). This model provides excellent anatomical detail and is popular among clinicians, but relative attenuation values are lost. Therefore, calcifications and stents cannot be identified on the basis of attenuation alone. SSD images must always be reviewed in conjunction with axial images to identify calcifications to prevent underestimation of degrees of stenosis due to adjacent mural calcifications. Selection of the threshold level is important, as inappropriate selection may artifactually generate or obscure pathology. ⁹ A threshold level of 160 HU is typical.

MIP views display the highest Hounsfield value voxel (contrast medium, calcium, bone)--along any plane. In contrast to SSD, MIP preserves relative attenuation values and lacks depth perception. This leads to difficulty in separating overlapping vessels in sites where several vessels are opacified (e.g., pulmonary artery and thoracic aorta) and in identifying small intravascular hypodense lesions (e.g., thrombus and intimal flaps). Calcified plaques may also obscure adjacent pathology. However, MIP views are better suited to reconstructing smaller vessels than SSD. Editing is necessary for both SSD and MIP images.

Clinical applications

In practice, CTA has replaced conventional catheter angiography in many diagnostic studies. It is well established in the assessment of aneurysms of the thoracic and abdominal aorta, of the pulmonary vasculature in suspected embolism, of the carotid and renal arteries for stenoses, and in the assessment of potential kidney donors.

Link et al ¹⁰ have demonstrated 100% agreement between CTA and digital subtraction angiography in cases of carotid stenosis >70%. Overall agreement between the two techniques was 89% using 2-mm collimation and a pitch of 1 (using only 100 mL of 270 mg/mL contrast). The overall agreement between Doppler sonography and angiography was 75%. While CTA outperformed Doppler sonography in this study, there was no significant difference between CTA and sonography in identifying patients whose stenoses required surgery. This lack of significance most likely reflects the relatively small study population, only 28 patients and 56 arteries. The addition of volume rendering to MIP images does not improve performance significantly. ¹¹ Carotid CTA is not used for primary screening at our institution; its role is limited to those patients who have abnormality demonstrated and who are unsuitable for MR angiography (MRA) or conventional catheter angiography. ⁵

In the thoracic aorta, CTA is valuable in the diagnosis aneurysms and suspected dissection. CTA performs as well as transesophageal echo and MR in assessment of dissection, although MR can demonstrate the entire length of the aorta (multidetector row CT also has this capability). ¹² Spiral CT is superior in the assessment of the supra-aortic branches (figure 2). As axial images have the highest resolution, these best demonstrate the intimal flap, however, difficulty may arise if the dissection extends into a horizontally oriented vessel. MR images may be of assistance in surgical planning. ¹³ In cases of suspected aortic dissection, a noncontrast-enhanced CT examination is performed initially to assess for media-stinal hematoma, followed by the enhanced scan. In these patients there may be significant cardiac compromise and a test injection should be used to calculate timing of scan delay. A protocol of 3-mm collimation with pitch of up to 2, or 5 mm with a pitch of 1, is recommended. ⁵

CTA is widely used for the diagnosis of pulmonary emboli (PE), though there is ongoing debate regarding the precise place in the hierarchy of investigations for those patients with suspected PE. ^14-16 The current resolution of CT allows detection of emboli to fourth order (segmental) vessels. The significance of emboli in more distal vessels is uncertain. The sensitivity and specificity of CT for central emboli has been reported as 91% and 78% respectively. ¹⁷ MPR is used as a problem-solving tool for exclusion of PE when the axial images are inconclusive. ¹⁸ A negative study effectively excludes pulmonary thromboembolism. ¹⁹ The examination is performed caudocranially to minimize respiratory motion artifact. Generally, 3 mm collimation and a pitch of 1.5 to 2 are used. Lower density contrast medium (150 to 240 mg/mL) helps ensure that small emboli are not obscured. In an animal study, CTA was more sensitive than 3D gadolinium-enhanced MRA for the detection of PE. ²⁰

CTA is an excellent screening tool of the renal arteries. Up to 10% of patients with hypertension have a renovascular etiology. CTA reliably identifies stenoses of the main renal arteries. As the renal arteries run axially, scan protocol is particularly important. Sensitivity of 100% has been reported for stenoses >50% using axial images with MPR and MIP reformats. ^21,22 Ostial and truncal stenoses can be differentiated (figure 3). A normal CTA virtually excludes the presence of a stenosis. ⁸ CTA with volume rendering has been reported faster and more accurate than CTA with MIP. ²³ CT is superior to MR and sonography in identification of accessory renal arteries. ²⁴

CTA has been proposed as the method of choice for investigating potential kidney donors. ^25-27 Its vascular capabilities include identification of accessory vessels, early branching of the renal artery, and aberrant venous anatomy with an overall accuracy of 95% compared with surgical findings. ²⁵ Axial images alone may be sufficient for evaluation, although 3D reconstructions (MIP and SSD) facilitate identification of important accessory renal vessels. The scan range extends to pelvic inlet to ensure identification of accessory arteries. Typical parameters: 3 mm collimation, pitch of 1.2 to 2, and 100 to 150 mL of contrast at >3 mL/sec. The enhanced scout/topographic image demonstrates calcifications and a delayed scout/topographic image after the CTA adequately depicts the collecting system and ureters and so obviates the need for an intravenous urogram.

CTA is ideal in the preoperative evaluation of abdominal aortic aneurysms, as it demonstrates their position, extent, and relationships to the renal and iliac arteries. ²⁸ Seven millimeter collimation may be required to cover the necessary volume; however, this runs the risk of missing renal artery stenoses and small accessory renal arteries. Narrower collimation should be used where possible. CTA is useful in assessment of aortic stent grafts after placement. MIP images are favored in this scenario, as the metallic component of the graft can be identified clearly. ²⁹

CTA has been employed to identify stenoses and occlusions of the splanchnic vessels. Its main role in this anatomic region is in assessing resectability of pancreatic tumors. Compared with axial images alone, CTA is more accurate in identifying unresectable tumors on the basis of vascular encasement. ³⁰ A multiphasic scan is performed: a standard nonenhanced scan to localize the pancreas, a second scan in the arterial phase (25 to 30 seconds after injection) of contrast enhancement (3-mm collimation, pitch 1 to 1. 5), and then a third scan in the portal venous phase (60 to 70 seconds after injection) of enhancement to include the abdomen. In most cases, only axial images are required.

The (near) future

The development of multidetector row CT has been described as an evolutionary advance perhaps as important as helical CT itself. ³¹ The primary limitation of CTA has been tube loading or heating, which limited potential scan volume. Multidetector row scanners have a wider beam and use a higher proportion of emitted photons than single-detector row CT scanners. The collimation of each detector row is unaltered. Technically, this implies that a greater volume of tissue can be scanned: it is possible to scan the entire aorta and femoral vessels in a single spiral scan (figure 4).

One hundred eighty degree linear interpolation algorithms have been used in helical CT. These reduce slice sensitivity profiles (and volume average artifacts); however, the images are relatively noisy. With multidetector row CT it is possible to use 360° algorithms by "combining" 180° data from two detector rows. This reduces noise without increasing slice sensitivity profile. The net result is improved z-axis resolution.

These technologic advances have implications for scan protocols, postprocessing (dedicated technologist), interpretation time, and patient radiation dose. With parallel software developments, there will be an expansion of the clinical applications of CTA, especially in intracranial and coronary circulations. The current reimbursement structure needs to be addressed to take these developments into account, as the additional physician time requirements of CTA are not acknowledged. AR