This publication was supported by an educational grant from Amersham Health, Princeton, NJ. The opinions expressed in this publication are those of the authors and not necessarily those of Amersham Health.

Dr. Bae reports relationships with Tyco Healthcare and Mallinckrodt through patent agreements and as a consultant. Dr. Fishman reports relationships with Siemens Medical Solutions and Amersham Health as a consultant. Dr. Foley reports a relationship with GE Medical Systems through an investigator agreement. Dr. Naidich reports a relationship with Siemens Medical Solutions through its Advisory Board and as a consultant. Dr. Saini reports a relationship with GE Medical Systems through research support. Dr. Becker, Dr. Sahani, Dr. Siegel, Dr. Tahktani, and Dr. Zinreich report that no such relationships exist.

Dr. Saini i s the Director of Computed Tomography at Massachusetts General Hospital and is a Professor of Radiology at Harvard Medical School, Boston, MA. Dr. Sahani is an Instructor of Radiology at Harvard Medical School and is a Staff Radiologist at Massachusetts General Hospital.

CT angiography (CTA) of the liver is invariably performed as a preoperative study in patients with metastatic or primary malignant liver disease for whom hepatic resection or placement of a chemotherapy infusion pump are treatment options. ^1,2 In this setting, preoperative knowledge of any variations from normal hepatic arterial anatomy is critical for optimal surgical treatment. ^3,4 Although the primary goal of hepatic CTA is to visualize the main branches of the common hepatic artery, depiction of the hepatic venous anatomy is also important for segmental localization of liver tumors. The traditional goals of liver imaging, hepatic lesion detection, and characterization are less important in these patients since these tasks will have already been accomplished during an initial diagnostic CT.

A more technically challenging but less common application of CTA of the liver is the evaluation of individuals who may donate part of their liver for living-related hepatic transplantation. ⁵ In such cases, it is important to visualize all three of the hepatic vascular systems: the arterial system, the portal venous system, and the hepatic venous system. ⁶ Of these, the most challenging is the hepatic arterial system, since it is necessary to demonstrate not only the main trunk hepatic artery and its branches, but also the arterial supply to various hepatic segments, particularly segment 4.

Technical issues associated with CTA of the liver can be divided into image acquisition protocol and contrast administration protocol. Key factors in image acquisition protocol include scan range, table speed (which is determined by gantry rotation speed, pitch, number of data-channels, and effective detector configuration), and the reconstructed slice thickness. Although it is seldom discussed, selection of kVp is important in determining vascular density. ⁷ Along with mA, kVp also determines radiation dose, which should be minimized in living-related donor candidates. Key factors in contrast administration protocol are iodine dose, which is determined by the concentration and administered volume of contrast medium, and the rate of contrast media injection. Finally, scan delay or timing is the crucial link between image acquisition techniques and contrast administration protocol, and this coordination must be done precisely due to the short scan times associated with multidetector CTA.

Image acquisition protocol

For high-quality CTA of the liver, breath-hold whole-liver acquisitions are necessary to avoid anatomic gaps. Hence, thin-slice scanning at sufficiently rapid table speed is critical. For example, a z-axis coverage of 200 mm in a 20-second breath-hold requires a table speed of at least 10 mm/sec. Obviously, faster table speeds will allow acquisitions in shorter breath-holds. Table 1 lists the typical image acquisition parameters for 16- and 4-slice CT scanners. The advantages of the 16-slice CT acquisition over the 4-slice acquisition are a shorter breath-hold and better quality three-dimensional (3D) CTA reformations due to the availability of submillimeter slices. Although data are reconstructed at the thinnest possible slices with 50% overlap for 3D reformations, for axial viewing, slightly thicker slices (1.25 to 1.5 mm) are adequate.

Thin-slice scanning enables exquisite visualization of the main trunk branches of the common hepatic artery (Figure 1). Within the liver, it is also possible to routinely visualize the segmental branches, which is necessary in living-related donor work-up (Figure 2). Anatomic variants, such as replaced or accessory hepatic arteries, are also easily depicted (Figure 2). On portal-phase images, the hepatic venous (Figure 3) and portal venous (Figure 4) structures are readily seen. Since these structures are comparatively larger than hepatic arteries, this phase can be reconstructed at 1.25 mm thickness. Typical scan delay for this phase is 70 seconds.

Contrast administration protocol

Figure 5 presents simulated contrast enhancement curves for the aorta and liver based on administration of 150 mL of 300 mgI/mL concentration contrast media injected at 4 mL/sec and assuming normal cardiac output. ⁸ Since arterial-phase scans should be obtained prior to significant portal venous enhancement, Figure 5 also shows that the optimal period for hepatic CTA imaging is approximately 20 to 40 seconds after contrast administration. Although aortic CTA studies can be performed at much lower total iodine dose, liver CTA requires a total iodine dose of 45 g in order to permit imaging of the hepatic venous and portal venous systems. Hence, with a 300 mgI/mL concentration, 150 mL is injected at 5 mL/sec, while with 370 mgI/mL iodine concentration, it is possible to inject the 45-g iodine dose with a volume of 120 mL and inject it at a rate of 4 mL/sec. With both these contrast administration protocols, the rate of iodine infusion is identical.

For accurate timing of image acquisition, standard delays are commonly used because this simplifies workflow. With this approach, a scan delay of 20 seconds is appropriate for the hepatic arterial system, followed by imaging of the venous systems at 70 seconds. However, use of bolus tracking techniques is preferred, because it permits customization of scan delay for individuals with reduced cardiac output. ⁹ For hepatic CTA, bolus-tracking scans that monitor aortic attenuation are acquired every 2 seconds beginning at 10 seconds after start of injection, and imaging is triggered at aortic enhancement of 150 HU with the portal/hepatic venous phase imaging performed at 70 seconds. Figure 6 illustrates the value of using bolus tracking to precisely time image acquisition due to variations in cardiac status. Note, in patients with poor cardiac output, the aortic enhancement is delayed and bolus tracking allows initiation of scanning with an optimized longer delay.

Radiographic parameters

The targeted arterial density during CTA is 250 to 350 HU. It is possible, however, to achieve higher levels of enhancement or lower iodine dose by using a kVp of 80, at which X-ray photon energy is closer to the k-edge of iodine, which is 33 keV. According to Huda et al, ⁷ iodine density is 50% lower at 140 kVp, compared with 80 kVp. Lowering kVp, however, requires increasing the mAs so that image noise is constant. Unfortunately, while an 80-kVp acquisition would be optimal for attenuation of X-rays by iodine, the mA requirements for overcoming image noise are so great that they cannot be met by current generation CT scanners. Hence, we are currently undertaking CTA scanning with kVp of 100 to 120 with a slight increase in mAs compared with the 140 kVp technique. This approach also permits reduction of radiation dose, although the images are slightly noisier. Initial experience suggests that the higher image noise can be tolerated in CTA studies, due to the high soft-tissue contrast between enhancing vessels and background liver. This is especially true for submillimeter slices, since a reduction in slice thickness also increases image noise, which provides another need for increasing tube current. Therefore, axial viewing is undertaken at a slice thickness of 1.25 to 1.5 mm with thinner slices used for 3D reformations. For CT scanners with automatic exposure control, a noise index of 20 HU appears to be suitable for hepatic CTA. However, considerable research needs to done in this arena for optimizing image quality and radiation dose.

Conclusion

In order to perform high-quality hepatic CTA, it is essential that optimal image acquisition techniques and contrast administration protocols are utilized. In addition, X-ray parameters, including peak tube-potential (kVp) and tube-current (mA), should be chosen carefully, since they have an important effect on image noise, vascular enhancement, and radiation dose. Today, all preoperative vascular evaluation of the liver is undertaken with multidetector CT, which makes catheter angiog-raphy unnecessary.

Figure Captions

FIGURE 1. Hepatic CTA demonstrating normal hepatic arterial anatomy.

FIGURE 2. Hepatic CTA demonstrating variation in hepatic arterial anatomy. In this case, there is a replaced right hepatic artery. Note the clear visualization of the intrahepatic branches, such as the segment 4 branch that arises from the left hepatic artery.

FIGURE 3. Hepatic CTA of the portal venous system showing trifurcation of the main portal vein.

FIGURE 4. Hepatic CTA of the hepatic venous system showing a variation in hepatic venous anatomy. In this case, there is an accessory right hepatic vein. Recognition of this variation is especially important in living-related liver donors.

FIGURE 5. Simulated enhancement curves of the aorta and liver after injection of 150 mL of 300 mgI/mL concentration at a rate of 4 mL/sec with normal cardiac output. Note that the optimal period for hepatic arterial imaging is approximately 20 to 40 seconds after contrast administration, before significant portal venous enhancement occurs. (Image courtesy of Kyongtae T. Bae, MD, Mallinckrodt Institute of Radiology, Washington University School of Medicine, St. Louis, MO.)

FIGURE 6. Bolus tracking allows precise timing of image acquisition in patients with (A) normal or (B) reduced cardiac output. Note the arterial enhancement would not have been adequate in the latter case if a fixed delay of 20 seconds had been used since the aortic attenuation of even 150 HU is not achieved until 30 seconds.

Discussion

ELLIOT K. FISHMAN, MD: Thank you. That's an interesting comment about the lower kVp. Depending upon how long your acquisition is, and how high you adjust the mA, is tube heating going to be any limitation?

SANJAY SAINI, MD: We tried 80 kVp on a patient for an aortic CTA study. The mAs were fixed at 190, which the tube could support possibly because the scan did not extend into the legs. Although the image was a bit noisy, it was more than adequate for visualization of the aorta. Another concern with a lower kVp is a higher absorbed X-ray dose. Hence the lower mAs may be appropriate for this reason as well.

FISHMAN: In terms of some of the 3D rendering, as you were showing the hepatic arteries, can you comment on the impact that the concentration or the contrast would have depending on which rendering techniques you use? For example, would the use of volume rendering versus MIP make a significant difference?

SAINI: Until recently, we were scanning with 300 mgI/mL concentration and have since used higher concentrations as well. There is no visible difference in the images with the higher concentration. It seems that the advantage of the higher concentration contrast media is that you can use a lower injection volume and a lower injection rate.

FISHMAN: Contrast volume makes some impact on volume rendering, and would probably make a much bigger impact on the MIP images, but I don't know how much impact.

KYONGTAE T. BAE, MD, PhD: In the pediatric population, we would use 80 kVp, correct?

MARILYN SIEGEL, MD: Yes.

BAE: But I think the low kVp can be problematic in large patients for scanning the pelvis, where you have to penetrate the large bony structure.

SAINI: Correct. However, I understand that CT scanners are being introduced this year that will allow mAs up to 1000.

DAVID P. NAIDICH, MD: I have a question about technique. I was curious about looking at really small vessels, especially in the liver and in small pulmonary arteries. Are you using a specific reconstruction algorithm?

SAINI: Standard soft-tissue algorithm.

NAIDICH: You're using a standard algorithm. You haven't played around with alternate algorithms?

SAINI: No.

W. DENNIS FOLEY, MD: The vascular-phase image was an odd mixture of the hepatic artery and the portal vein in some cases. In other cases, it's the hepatic artery without the portal vein. Now that you are going to dispel this tracking technique, are you separating out hepatic artery visualization from the portal vein? Or do you like to see both together?

SAINI: I don't necessarily think we like to see both together. In all studies, we get a portal-venous-phase scan as well. I haven't looked critically enough to know that we are better off today than we were yesterday, so I don't know the answer to that.

FOLEY: One thing we've done is to deliberately separate out the two. So you can have what might be called an early
arterial phase, in which you simply define arteries. Then we scan in the late arterial phase, which actually turns out to be the best timing to detect hypervascular lesions and generate a vascular map of the portal vein.

But we would do that only if the patient is a candidate for surgery, transplantation, or chemoembolization.

SAINI: Is this early arterial phase at peak aorta plus 10 seconds?

FOLEY: Well, actually, no, the other arterial phase begins at peak aortic. That's determined from a mini-bolus. But for the late arterial phase, it's aortic peak plus 15 seconds.

SAINI: The early arterial phase imaging would be at 20 seconds?

FOLEY: It varies, somewhere be-tween 12 seconds with very fast circulation and up to 30 seconds in patients with very slow circulation.