Applied Radiology

Dr. Sasson is a third-year Resident in the Department of Radiology, Stanford University Medical Center, Stanford, CA. He received his MD in 1999 from Baylor College of Medicine, Houston, TX. He plans to begin a Neuroradiology Fellowship following completion of his residency.

As the use of multidetector-row computed tomography (MDCT) becomes more widespread, potential applications continue to develop. In patients with acute abdominal pain, MDCT has the potential to fill the roles currently served by the more invasive and time-consuming conventional digital subtraction angiography (DSA). Automated processing methods may also enable MDCT to image complex structures, such as the bowel in three dimensions, an advance with potential benefits not only in evaluating acute disease of the bowel, but also in overcoming the challenges posed by the large and complex MDCT data sets.

Technical advances in computed tomography (CT) have changed the landscape of abdominal imaging. Where digital subtraction angiography (DSA) once stood as the only diagnostic option for certain acute abdominal pathologies, such as mesenteric ischemia, multidetector-row CT (MDCT) has firmly established itself. The continuing development of this technology enables rapid and safe diagnosis of a wide variety of acute conditions. Selected applications that benefit from the particular advantages of MDCT over single-detector-row helical CT (SDCT) or DSA are described here, as are problems created by this technology and the potential solutions to those problems.

Single-detector CT is used routinely in a wide range of clinical emergencies in the United States and Europe. ¹ The advantages of CT evaluation of acute abdominal pain have been described at length. ^2,3 The limitations of SDCT, however, have precluded its routine use under certain conditions. ⁴ These limitations, including the relatively slow speed of scanning compared with breath-hold times or blood flow, relatively high contrast dose, and low z-axis resolution, can be overcome by MDCT. ^1,5 Therefore, MDCT has become an alternative to, or substitute for, DSA in carotid angiography, lower-extremity angiography, and intravenous urography, to name a few applications. ^6-11 The same advantages that enable MDCT to scan within a specified vascular phase--namely, rapid, high-resolution scanning in a single breath-hold--have also been used to create thin-collimation images and virtual three-dimensional (3D) representations of the colon. ¹² These same capabilities of rapid scanning and high z-axis resolution have great utility in the emergency evaluation of acute, nontraumatic abdominal pain. At the same time, they require the development and refinement of automated methods to display large, complex data sets in a rapid, reliable, and reproducible way. ¹³

Before exploring specific potential applications, it is helpful to briefly review the basic principles underlying MDCT. An excellent review by Fuchs et al ⁵ describes many of the technical details. Briefly, by using a two-dimensional (2D) array of X-ray beam detectors, multiple data channels may be obtained at once in a helical fashion. The number of data channels that can be sampled at once vary according to manufacturer, but are increasing. Depending on the slice profile and filter width, one can achieve effective slice thicknesses with sub-millimeter z-axis resolution. ¹³ Several different advantages result. Scan time can be decreased since a thicker section of the body can be imaged with each gantry rotation. This means that single breath-hold images of the abdomen and pelvis are feasible, even at thin collimation. For example, at our institution, scans from the dome of the diaphragm to the bottom of the pubic symphisis are routinely performed in approximately 25 seconds on an 8-detector scanner, at 5 mm collimation using a table speed of 13.5 mm/gantry rotation. Another way of exploiting the speed advantage of MDCT is in "multiphasic scanning," when the abdomen may be scanned rapidly in succession to acquire images during multiple phases of contrast enhancement. Additionally, as scan times decrease, so does the volume of contrast necessary for adequate arterial enhancement. Another advantage inherent to MDCT is the ability to retrospectively reconstruct the image data to thinner collimations, down to the collimation set for the smallest element in the detector array. This allows prospective reconstruction at thicker collimation to enable efficient data interrogation, but leaves as an option a more detailed, thinner-collimation examination of the areas of interest. For example, one may perform a CT angiogram of the abdomen at 5 mm collimation to look for acute pathology, such as retroperitoneal hematoma, while thinner, 1.25-mm reconstructions can be fed to a 3D postprocessing workstation for creation of high-resolution representations of the abdominal aorta.

There are disadvantages that arise with the use of MDCT. As the resolution of the study increases, so does the noise. This can be offset with increased tube current, but at the expense of increased dose. ⁵ Another issue is the size of individual studies. CT angiography (CTA) of the lower extremities, for example, can routinely generate thousands of images. The time to review each individual axial section can become prohibitive. Primary evaluation of 3D representations of the data set is possible. At the moment, however, physicians or specially trained technologists are required to construct these 3D representations, often taking hours at a time for complex cases. ¹³ Further, 3D images, by necessity, do not contain all of the information in the axial source images. For example, a 3D representation of the surface of the abdominal aorta, may not demonstrate an embolus within the lumen of the superior mesenteric artery, which would be evident on an axial image. This inevitably results in a trade-off between diagnostic accuracy and the time burden associated with interpretation of huge studies. The combination of computer automation and computer-assisted diagnosis that will likely evolve as the technologies mature, in response to the large datasets and 3D nature of volumetric scanning, may entail a fundamental change in the way we read and interpret these studies. ¹³

Acute, nontraumatic abdominal pain

The causes of acute abdominal pain are numerous and varied. ² CT has been well utilized in the diagnosis of many of them, but its use in the evaluation of vascular causes of abdominal pain has been limited. ⁹ One of the particular strengths of MDCT is the ability to perform CTA. Its use in lower-extremity angiography and aortic angiography has been described elsewhere. ^8,9 The application of MDCT to the diagnosis of ischemic bowel in the acute ill patient is a logical progression.

Mesenteric ischemia can result from either an acute or chronic reduction in arterial flow, or from venous thrombosis. ^14-16 CT findings in bowel ischemia are broad, and in many cases nonspecific. ¹⁴ In acute arterial ischemia, the small bowel may demonstrate four types of enhancement patterns: no enhancement; delayed enhancement with respect to other bowel loops; strong homogeneous enhancement with respect to other bowel loops; and a layered, or "target," pattern of bowel wall enhancement. Chen et al ¹⁵ found that nonenhancement and "target" enhancement were equally frequent, followed by strong homogenous enhancement and delayed enhancement. Secondary findings include bowel wall thickening, portal venous gas, pneumatosis intestinalis, arterial filling defects or occlusion, and solid organ parenchymal infarcts (Figure 1). ^15,16

The causes of bowel ischemia are varied, and include atherosclerosis, thromboembolic disease, vasculitides, and hypoperfusion resulting from low-flow states. ¹⁶ Mechanical occlusion as a result of herniation, intussusception, torsion, or direct invasion or encasement by tumor may also be responsible. ^17,18

Conventional helical CT has been used in the evaluation of ischemic bowel, and demonstrates certain findings that are typical of ischemia. ¹⁹ Specifically, in addition to the aforementioned enhancement patterns, bowel wall thickening and bowel dilatation may be seen. A specific vascular distribution, or later findings (such as pneumatosis intestinalis and frank obstruction), may suggest the diagnosis as well. ¹⁹ These findings are relatively nonspecific, and, in the setting of chronic mesenteric ischemia, the protean clinical symptoms may make the diagnosis difficult. Multidetector-row CT, however, enables exquisite visualization of the anatomy of the mesenteric arterial vasculature. With the aid of postprocessing, one can determine degree of vessel stenosis, detect mural calcification, and visualize thrombosis, or embolic occlusion in acute ischemia, and evaluate for the presence of collateral vessels in chronic ischemia (Figure 2).

Venous thrombosis accounts for 15% to 20% of cases of mesenteric ischemia. ¹⁷ The pathology can be demonstrated easily with venous phase contrast-enhanced CT. ¹⁸ Although SDCT is sufficient for the evaluation of large-vessel venous occlusion--at the level of the portal and superior mesenteric veins, for example--the higher resolution of MDCT enables depiction of occlusions in more-distal branch vessels (Figures 3 and 4). CT has particular advantages in the depiction of the portal venous anatomy, as it is difficult to access the portal system directly by conventional DSA. ²⁰ A preliminary study examining MDCT in acute ischemia was performed using 4-detector MDCT at 1.25-mm collimation imaged 25 seconds and 60 seconds after IV contrast infusion at 4 mL/sec. The results revealed a high sensitivity and specificity for acute arterial mesenteric ischemia (100% and 92%, respectively) using combined CT findings. ¹⁹ Multidetector-row CT also enables the detection of other, less common causes of mesenteric vascular injury ⁸ (Figure 5).

The ability to scan the entire abdomen and pelvis at once enables the depiction of other pathologies, distant from the mesenteric vascular axis, which may not have been identified in selective DSA. Also, while DSA is relatively safe, it has been noted to be susceptible to operator-dependent complications (reported rates of 1.8% to 14.7%), and the noninvasive nature of CT imaging enables faster, safer visualization of the vascular anatomy, with less potential for complication. ^21,22 These last two points are illustrated nicely in Figure 6, which demonstrates an unsuspected iatrogenic femoral pseudoaneurysm and active extravasation from the superficial femoral artery after coronary angiography.

MDCT and virtual endoscopy

Thin-collimation, rapid scanning made possible by MDCT has been used to depict the inner surface of hollow lumens, such as the colon and tracheo-bronchial tree. ^23,24 These techniques are most successful when the lumen is relatively large (tracheo-bronchial tree) or distended (colon). Extensive research on automated processing methods and computer-aided diagnosis are likely to be crucial to the successful use of multidetector-row technology in CT colonography. ¹³

Evaluation of the small bowel can also now be performed using such techniques as CT enteroclysis. ²⁵ While methylcellulose with barium has been used in traditional enteroclysis, low attenuation contrast, such as water or methylcellulose, can provide adequate luminal contrast in intravenous contrast-enhanced CT. ¹⁶ Using the techniques described by Maglinte et al, ²⁵ with a 14Fr 155-cm nasointestinal tube, low-attenuation contrast, such as water, can be used to provide luminal contrast in the small bowel. Additionally, the same tube can provide for gastric decompression, which, if used routinely, can allow for therapeutic decompression and diagnosis without the discomfort and difficulty of repeated nasogastric intubation. ²⁵ With adequate small-bowel distension, automated 3D processing techniques, such as center-path finding, can create simple representations of the complex anatomy of the bowel, as is done in CT colonography. ²⁴ In patients with acute abdominal pain, bowel obstruction often results in loops of distended bowel that transition to nonobstructed, nondistended bowel. It is possible, using the same 3D reconstructive techniques, to create 3D representations of the obstructed bowel lumen. With further advances in computer automated reconstructions, it may also be possible to "lay-out" the bowel in such a way as to accurately identify the point of obstruction.

MDCT and 3D representation

A common thread throughout this discussion is the necessity for advances in the representation and manipulation of large MDCT data sets. When a study comprises hundreds, or even thousands, of axial images, the efficiency of viewing each individual image is compromised. Several options besides viewing contiguous axial images are available for interpretation of MDCT examinations.

Various types of 3D and postprocessed images have been reviewed at length elsewhere, but will be briefly described here. ^7,13,26,27 The first, in terms of simplicity, is the multiplanar reformat (MPR). This is simply a reconstruction of the data set into another orthogonal plane, perpendicular to the original imaging plane. With thin-collimation examinations, the resolution of a MPR can approach and equal the resolution of the study in the original acquisition plane.

Curved planar reformations (CPR) are similar to MPR, except that the imaging plane is not orthogonal. Since few structures in the body line-up perfectly with the three orthogonal image planes (transverse, sagittal, and coronal) or single-plane obliques, typically, the CPR follows the contour or lumen of a blood vessel, bile duct, loop of bowel, or other structure, as it travels in and out of the orthogonal plane. The path of the vessel has traditionally been determined manually, by specifying the center of the vessel on contiguous axial images on a 3D workstation. Computer algorithms can reliably, reproducibly, and quickly find the center path down a blood vessel or other structure with little user input, and create a CPR. ²⁸

Curved planar reformations have pitfalls, however. For example, they can produce artifacts that either obscure or give the appearance of a stenosis or occlusion. If, for example, during the creation of a CPR, the path through the lumen of the vessel is off-center, some fraction of the caliber of the vessel will be projected, rather than the true diameter, resulting in an artificial stenosis. Automated algorithms for the creation of CPRs can prevent the creation of "false-postive" stenosis in this manner. To avoid missing a stenosis, two CPRs are usually created for any structure, at perpendicular planes. Automated techniques now enable dynamic, rotating CPRs to be visualized at any conceivable angle to the center path, in real time on distributed, Web-based workstations. ^28,29

On many 3D postprocessing platforms, using either MPR or CPR, a thickness of the imaged slab can be defined. The question then becomes, how to represent a 2D image of a 3D volume? Several options are available, including the maximum intensity projection (MIP); average intensity, or ray sum, projection (aIP); and minimum intensity projection (MinIP). For a slice of a particular thickness, the MIP will display the highest-attenuation pixel, whereas the average intensity or minimum intensity projections will display the average attenuation value, or the lowest-attenuation pixel.

Maximum intensity projections are typically used for contrast-enhanced vessels, and MinIPs for hollow-lumen, air- or water-containing structures (Figures 3C, 3D, and 7). Each can be affected by artifacts from either very-high- or low-attenuation structures outside the region of interest (such as bone or bowel gas) that obscure the relevant anatomy, but this problem can usually be overcome by specifying appropriate slab thicknesses or "cut-planes" to exclude irrelevant structures from the imaging plane. Raman et al ²⁹ have developed algorithms that create a curved planar slab, with the slab thickness automatically adjusted along its length to include the caliber of the vessels.

These postprocessing techniques are not exclusive; for example, to lay out the biliary tree, one may combine a CPR (to accommodate the complex, 3D structure) with a MinIP (as the biliary tree is typically lower in atten-uation that the surrounding liver parenchyma), and select a thin slab, or plane thickness of 10 mm, to prevent overlying bowel gas from obscuring the bile ducts in the MinIP calculations. The resulting image would be a curved-planar, thin-slab MinIP (Figure 7).

The preceding are all 2D representations of 3D structures. Volume rendering (VR) and shaded-surface display (SSD) are two types of 3D representations. Of the two, SSDs are less computationally intensive. In SSDs, a threshold value is specified by the user, applied, and used to define a surface that is then rendered with an imaginary source of illumination. This 3D structure can provide great detail of the surface anatomy of a specific structure, but one loses a large amount of information, as everything outside the threshold attenuation is thrown away.

Volume rendering is more complex and takes more computing power, but the 3D images it produces are superior. With VR, each particular 3D data point, or voxel, in the image is assigned an opacity level. This level is usually defined by attenuation so, for example, one could display the skeleton by assigning zero opacity or total transparency to attenuation values less than those of bone. The skeleton itself may be made completely opaque, resulting in an image equivalent to an SSD of the skeleton, or of varying transparency, enabling visualization of the internal architecture. More than a single opacity function can be used, enabling the display of vessels with one function, bone with another, and soft tissue with yet another. Each can be color-coded as well, and with an artificial external light source, depict stunning representations of complex anatomy (Figures 2A, 5D, and 6C);VR can be subject to artifacts, however. Opacity functions are based on setting threshold attenuation values, which can, for example, overestimate the degree of stenosis in a vessel (Figure 5D), or make small vessels translucent or invisible.

Combinations of these techniques, with automated processing methods, have successfully been used to create highly detailed, reproducible, and accurate angiograms of the lower extremities using MDCT ^27,30 (Figure 6A through C). As these methods are further developed and available to the radiologist at the time of interpretation, primary evaluation of these complex and large examinations can be performed using postprocessed reconstructions.

Conclusion

As MDCT becomes increasingly widespread, it will likely fill more and more roles that currently belong to diagnostic DSA. It has many potential applications in acutely ill patients, including techniques for the improved diagnosis of conditions that, heretofore, have required invasive and time-consuming examinations. The delineation of vascular abnormalities that result in bowel ischemia is just one example.

Other potential nonvascular applications of MDCT are being developed. The speed and higher resolution of MDCT enable the creation of 3D representations of complex data sets. As automated processing techniques mature, it may become routine to create a 3D representation of the bowel with a few mouse clicks; with a few more mouse clicks, a rotating image of the skeleton; and with a few more, a map of the abdominal vasculature. Although hurdles remain, MDCT will give the radiologist new tools to help quickly and safely diagnose a wide variety of disease, including acute abdominal pathology.