- Email: [email protected]
Multi-slice CT contrast enhancement regimens
Clinical Radiology (2004) 59, 1051–1060
REVIEW
Multi-slice CT contrast enhancement regimens P. Dawson* Department of Imaging, UCL Hospitals, London, UK Received 21 April 2004; received in revised form 14 July 2004; accepted 14 July 2004
KEYWORDS Computed tomography (CT); Contrast media; Infusion protocols
Historically, the development of progressively faster Computed Tomography (CT) technology has dictated a recurrent need to re-examine intravenous contrast agent enhancement regimens. The most recently introduced development, the very fast, multi-slice helical/spiral systems, have raised the same issue yet again. It is possible, exploiting the technology to its maximum potential as regards speed, to perform an examination many times faster (depending on the number of detector rings from 4 to 64) even than with earlier single slice spiral instruments. In order to optimise image quality, such maximal speed gains will not usually be sought but, nevertheless, imaging time will generally be substantially reduced. It is natural that the question of a possible need to modify contrast agent enhancement protocols designed for an earlier generation of slower machines should again be considered. Using as a basis known contrast agent pharmacokinetics and results of modelling techniques, the matter is tackled in this paper. q 2004 The Royal College of Radiologists. Published by Elsevier Ltd. All rights reserved.
Introduction The history of CT development has been marked by a number of stepwise increases in scanning speed. The biggest such step in recent years was taken when spiral/helical technology replaced incremental slice technology. Now multi-slice spiral systems have given further impetus to this process. As explained elsewhere,1 a 4-detector row multi-slice system with a 0.5 s rotation time may be eight times as fast as a single-slice system with a typical 1 s rotation time and 16-slice or 64-machines may, of course, be even faster. Two caveats might usefully be entered here. First, in the interests of optimization of image quality, exploitation of such maximum possible gains in image acquisition speed should not be routine. Secondly, the most useful application of these systems will not always be in the fast coverage of large anatomical volumes in * Guarantor and correspondent: P. Dawson, Department of Imaging, UCL Hospitals, The Middlesex Hospital, Mortimer Street, London WIT 3AA, UK; fax: 00-44-207-380-9068. E-mail address: [email protected].
shorter times but, rather, will frequently be in the coverage of modest volumes in reasonable times with high Z-axis (isotropic) spatial resolution1 during a single breath-hold. These points having been made, this technology again raises the sort of question occurring repeatedly throughout the history of CT development: even if we consider that, in practical terms, the typical speed increase is usually only by a more modest factor of, say, three rather than eight or more, does this call for any radical modification in contrast agent administration protocols? At first sight it might seem reasonable to suggest that faster contrast agent injections would be appropriate to the shorter scan times. Perhaps different concentration formulations of contrast agents from those typically used might be appropriate. The ideal outcome of such a re-examination for both imaging department staff and patients would be a conclusion that the faster delivery of smaller total loads of contrast should optimize enhancement in the shorter timescales involved. It is the purpose of this article to examine these and related questions.
0009-9260/$ - see front matter q 2004 The Royal College of Radiologists. Published by Elsevier Ltd. All rights reserved. doi:10.1016/j.crad.2004.07.009
1052
The ideal approach to questions of practical clinical importance would be a practical clinical method, that is a clinical trial. However, since there are several variables to be considered, some under control (e.g. concentration, total dose, delivery rate of agent) and some not (patient body weight, cardiac output), either a very large trial or a whole series of comparative trials would have to be undertaken. In view of this it seems reasonable to suggest that pharmacokinetic modelling might be a way of obtaining some of the information required. Several authors have demonstrated the efficacy of such modelling.2–8
P. Dawson
6.
7.
8.
Pharmacokinetics of contrast agents This subject and its application to the modelling of contrast enhancement in CT has been described in detail elsewhere.2–8 A brief outline is given in Appendix 1. It will be helpful for the following analysis and discussion to set out some general points of principle established by such modelling and by clinical experience. 1. Intra-vascular contrast agents are sometimes helpful simply to ‘label’ blood vessels to help in image interpretation, as oral or rectal contrast may be used to label bowel.8 For this purpose, only modest blood levels of contrast agent are necessary and precise timing is not critical. 2. For CT angiography (CTA) proper, higher blood levels maintained throughout the scan are, of course, required and timing is critical. This will be discussed further below. 3. The principal reason for intra-vascular (IV) contrast agent administration is to increase the contrast between structures to render them more visible and, in particular, to enhance the difference between normal and abnormal tissue CT numbers, so as to render lesions more conspicuous. It has been shown both in clinical practice and theoretically, that the conspicuity of typical relatively hypovascular liver lesions increases as the normal liver tissue CT number increases. This suggests that not too great an emphasis on economy in contrast agent use would be best policy. 4. Even though image acquisition may now be very fast, we cannot argue that as a consequence the imaging time window may now be narrow. A narrow window may easily be missed, and a reasonably wide time window for imaging remains desirable. 5. The magnitude of arterial enhancement achieved is a function of contrast
9.
10.
11.
12.
concentration, infusion rate and total time of infusion. The magnitude of hepatic enhancement is a function of total iodine dose administered, and may be adjusted by changes in concentration of agent administered, total volume or both. The liver must be scanned in its entirety before onset of the ‘equilibrium phase’, the time of onset of which is itself to a significant degree a function of infusion duration. This argument is valid for any organ but is critical for liver imaging since equilibrium is established very early in the liver.8,9 Before the introduction of helical CT, infusion rates were sometimes prolonged and experiments with two-phase injections were made in an effort to delay the onset of equilibrium (Figs. 1a–d, 2). Additionally, portal venous phase scanning was often started much too early in terms of contrast enhancement in order to beat this deadline. The result was images of the liver which were sub-optimal in enhancement for the early slices and probably, in spite of all efforts, in the equilibrium phase for the later ones. The speed of the scanner determines its ability to record image data during the most advantageous time period, e.g. maximized arterial enhancement, the pre-equilibrium phase in the liver or individual phases in true multiphasic imaging. Cardiac output has a significant influence on timing of arrival and peak of contrast agent concentration in the arteries (relevant to CTA). However, although there is a delay in peak enhancement in the liver too in cases of low cardiac output, this will be little affected except in extreme cases, being much more dependent on total dose delivered. A reduction in cardiac output results in an increase in peak arterial enhancement. The fundamental explanation for this is that with a lower cardiac output less blood is ejected from the heart to dilute the contrast agent during transit, and in practice this over-rides the effect of a prolonged transit time working in the direction of reducing peak enhancement. Of course, all this is only true up to a point— zero cardiac output delivers no contrast agent at all to the arteries. For a given contrast load, arterial and organ enhancement are inversely proportional to body mass, so the contrast total load to body mass ratio is a key determinant of enhancement.
Multi-slice CT contrast enhancement regimens
1053
Figure 1 The plasma and interstitial concentrations (expressed as CT number) for infusion rates of (a) 1 ml per s (NZ1), (b) 2 ml per s (NZ2), (c) 3 ml per s (NZ3) and (d) 5 ml per s (NZ5). The curves cross at the equilibrium point and, the faster the injection rate, the earlier comes the point. 13. Some factors are invariants to contrast agent injection protocol changes. (a) The venous to arterial circulation time is a function of the patient and is not under the radiologist’s control (b) The time of arrival of contrast agent in
the portal vein is also a function of the patient. (c) The cortical nephrogram is ‘immediate’ (subject to circulation time) but the pyelogram is delayed some 3–5 minutes after IV injection.
1054
P. Dawson
Fig. 1. (continued)
Multi-slice CT contrast enhancement regimens
1055
Figure 2 Modelled plasma contrast concentration (P(t)) and interstitium concentration (Cis(t)) curves, expressed as CT number for a monophasic infusion regimen at 3 ml per s (NZ3), and a biphasic infusion regimen at 5 ml per s for 10 s followed by 1 ml per s for 100 s (NZ5.1, respectively). The equilibrium point may be seen to be delayed by some 40 s in the two-phase regimen as opposed to the one-phase regimen.
The factors to be considered in enhanced CT are vascular enhancement (blood contrast agent concentrations) and tissue/organ ROI enhancement (a weighted sum of micro-vascular and interstitium contrast agent concentrations).8,9 In CT
angiography the former alone is of interest; in any other examination the latter is of greater interest.8,10 Some other useful results of pharmacokinetic modelling8 might usefully be summarized here.
Figure 3 The blood concentration curves for four different injection rates, at 1, 2, 3 and 5 ml/s. As the rate increases higher peaks are obtained, but the levels are not sustained.
1056
P. Dawson
Figure 4 The blood concentrations achieved in a two-phase injection (here 5 ml/s for 10 s followed by 1 ml/s for 100 s) produces blood levels comparable with those achieved by a one-phase 3 ml/s injection but somewhat better sustained. Such manoeuvres are unnecessary with faster multi-slice scanners.
Detailed pharmacokinetic analysis predicts to some extent what seems obvious, as follows. (a) Faster injections of the same total dose give higher peak blood flow values at earlier times, but the high values are sustained for a shorter period. This may be seen from Fig. 3 in which modelled curves are presented for infusion rates of 1, 2, 3 and 5 ml/s. The effect of a biphasic injection is shown in Fig. 4. (b) Varying the infusion rates over the range typically used clinically has different effects on arteries and organs. The magnitude of arterial concentrations is proportional to the flux of agent, i.e. to flow rate!concentration. Conversely, the portal venous phase enhancement of the liver is quite insensitive to changes in contrast administration flux over this range but depends more simply on total load administered. Therefore, although perhaps some reduction in contrast dose in CTA may be possible with multi-slice scanning (see below), significant reductions in dose for hepatic imaging, and therefore in effect for abdominal imaging in general, are unlikely because of total dose dependency dominance. (c) On the other hand, where the hepatic arterial phase of liver enhancement is of interest, this is dependent on injection rate. As regards faster injections in general, it is important to note
that there is little to be gained by pushing the infusion rate beyond about 5 ml/s since the right heart and lungs represent a buffer preventing achievement of higher and higher concentrations by faster and faster injections.11
Discussion and illustrative examples CT angiography This may be seen as a special case. In all other CT examinations it is the combination of vascular and extravascular contrast that effects the enhancement; in the case of CT angiography it is, by definition, the intravascular agent that contributes. Here we may state some basic principles. 1. An adequate blood concentration of contrast must be achieved and sustained during the time of the data acquisition. 2. Cephalo-caudad data acquisition during a single breath-hold (where a breath-hold is necessary) of around 20 s should be the aim. 3. Accurate timing of the start of data acquisition should be achieved using bolus timing techniques. 4. On the basis of (3), (1) may be achieved by
Multi-slice CT contrast enhancement regimens
1057
approximately matching agent infusion time to necessary scanning time. This in turn may be seen to determine total dose required.
disastrous), and the 5 ml/s infusion rate would dictate a total dose of some 200 ml. This is unacceptable and the choices would be between:
It is more convenient in CTA to think in terms of flux of contrast agent, i.e. rate of infusion! concentration. This may be increased to achieve higher blood concentrations and consequent improved signal-to-noise ratio either by increasing the rate of injection, or by increasing the concentration of the contrast preparation, or both. Some would argue that in this application a higher concentration formulation of the agent is useful.12 As has been noted earlier, however, there is little to be gained by pushing the infusion rate beyond about 5 ml/s. The total dose to be delivered is then dictated by the desirability of approximately matching the infusion time to the scan time as suggested above. One misunderstanding should be cleared up. There is a widespread perception that a prolonged infusion achieves a plateau of contrast concentration in blood vessels. This is not the case: concentrations continue to increase, albeit at an ever slower rate, and become asymptotic to a plateau never reached. The application and consequences of these principles are illustrated by a few examples.
† reducing the injection rate to approximately 3 ml/s; † accepting a lower blood concentration and reducing contrast load to 120 ml; † accepting a reduced z-axis resolution by scanning at 4!2.5 mm collimation, to halve the scan and infusion times; † or using an 8-, 16- or 64-slice machine capable of covering the region at the same high z-axis resolution in a significantly shorter time.
Aorto-iliac CTA Such a study might be done for suspected abdominal aortic aneurysm. The extent of the region of the abdominal aorta and the iliac vessels to the level of the proximal femoral vessels is some 40 cm. Using a 4-slice machine, 4!1.25 mm collimation, a pitch of 1.5 and a gantry rotation time of 0.5 s (dictating a table speed of 15 mm/s), this can be covered in about 25 s, suitable for an average breath-hold. If an infusion rate of 5 ml/s is chosen to obtain high levels of contrast, then about 125 ml of contrast would give a match of acquisition and infusion times. To study the smaller volume needed for renal artery assessment (no more than 20 cm), the acquisition time is approximately halved so the total contrast dose required is halved. This in turn has a consequence discussed below under Shorttime scans. Thoraco-lumbar aortography This is an anatomical block of greater length, approximately 60 cm. Were the above parameters to be used with a 4-slice machine, the scan time would be 40 s. This is not an easy breath-hold (although with a cephalo-caudad scan breathing during the later pelvic portion would not be
Aorto-bifemoral CTA The distance from the distal infrarenal abdominal aorta to the ankles is some 120 cm. With a 4-slice machine, 4!2.5 mm collimation, a pitch of 1.5 and a gantry rotation time of 0.5 s (dictating a table speed of 3 cm/s) this would be covered in about 40 s. With an infusion rate of 4 ml/s the infusion total load would be rather large at 160 ml—and this for a not very high z-axis resolution. The answer here is clearly to use an 8-, 16- or 64-slice machine, since even dropping the infusion rate to 3 ml/s would be of little help in limiting contrast dose if 4!1.25 mm acquisition was decided upon. The essential point concerning total doses in CTA is not so much that they can be reduced with faster and faster machines (see 8-slice and 16-slice examples above) but, rather, that good CTA could not previously have been achieved at all. It would be unreasonable to be too prescriptive, and the above are examples to illustrate general principles. Design of protocols may be seen as an art. It is simply best to base some artistic endeavours on immutable scientific principles. So, no protocol may be considered in some absolute sense “right” but many may be deemed “wrong”. Very-short-time scans There is an important point to be made about short acquisition time scans taking less than about 20 s and the need ideally to match scan and injection times. Two problems arise. First, timing becomes even more critical because of the narrow timewindow; and, secondly, the contrast concentrations reached in blood after a short infusion time matched to a short data acquisition time is relatively low. As regards the latter problem, two strategies may be adopted. The contrast flux may be increased; this is achieved by use of a higher concentration formulation rather than by an increase in rate of injection since the latter, as
1058
P. Dawson
Figure 5 Blood and liver contrast enhancement curves for a region of interest (ROI) in the liver. It may be seen that the liver enhancement peaks at 60–70 s. A typical modest image acquisition time (much shorter is possible) for a multislice CT machine is indicated, demonstrating that it can be completed during a high-enhancement period and, even if begun as late as 60 s, will be completed before 100 s and therefore before equilibrium.
previously discussed, is of little avail above approximately 5 ml/s. Alternatively, a longer infusion may be carried out and the scanning commenced later when blood levels are higher. Can total doses be reduced in CTA with multislice CT? In one sense this is a non-question, in that only now with multi-slice systems can CTA be done well and so the relative dose issue is rather meaningless. What can be said is that savings in contrast dose are not easy to make because certain minimum blood concentrations must be reached. When superfast systems are used to reduce the scan time, the injection time, which needs to be matched to it, is reduced. If the injection time is short, only low blood levels will be achieved unless the contrast flux during the short time period of injection is very high. Since injection rates higher than about 5 ml/s are not much more effective, the only options are to increase contrast concentration and/or to prolong the injection time to deliver more contrast and to delay imaging until the higher levels have been achieved. As far as CTA is concerned, the scope for contrast agent savings with faster scanners would seem, therefore, to be limited by fundamentals.
CT pulmonary angiography This is, of course, a specific example of CT angiography. There is, however, an argument about concentrations of contrast agent to be used and vessel enhancement to be ideally achieved. In this examination the search is for intraluminal filling defects that may actually be hidden by too dense a contrast column. Some authorities recommend a dilute contrast infusion, in which case the above arguments do not apply.
Brain imaging Here there are no complex issues. 50–100 ml suffice and, functional studies aside, there is no need for fast delivery. No modification of long-established protocols is necessary.
Abdominal imaging In virtually all abdominal imaging the liver will also be imaged, and should be imaged optimally. On this assumption we conclude that the delivery of contrast has to be tailored to this end with other organs being, perhaps, imaged earlier or later in the evolution of the distribution of contrast agent as
Multi-slice CT contrast enhancement regimens
appropriate (discussed below). Therefore, liver imaging will be considered first.
The liver The liver poses unique challenges because of its dual blood supply and because of the need to complete imaging before onset of a very early equilibrium phase.9 A key determinant of this is the injection duration.8 The issues surrounding the two blood supplies of the liver are well enough understood. We are most often interested in the portal venous phase. As discussed above, this does not occur maximally until some 60–70 seconds after commencement of infusion (Fig. 5). Fast multi-slice systems allow the commencement of scanning in this phase to be delayed until some such time with the confidence that the whole liver may be completely covered, even at highest z-axis resolution, before the onset of equilibrium.9 It is no longer important with fast scanning to be concerned about the fact that the faster the infusion rate the sooner the equilibrium point arrives and the equilibrium phase begins (Fig. 1a–d). Furthermore, it has been shown that the conspicuity of hypovascular liver lesions, the usual issue, increases as the normal liver enhancement increases in the PV phase. Leaving aside the question of timing, the enhancement of the liver by its principal blood supply is a function of the total dose given.8 It therefore follows that it is unlikely we can achieve more than a marginal reduction in total dose administered without sacrificing information. It is worth noting here that there is a divergence between mainstream US and European practice here, in that a typical European total dose for a liver study would be 120 ml of an agent 300 mgI/ml concentration, whereas a typical US dose would be 150–180 ml.
Hepatic arterial phase imaging In the context of a contrast administration regimen tailored to optimize liver portal venous (PV) phase enhancement, any other imaging can be incorporated: for example, the hepatic arterial (HA) phase of the liver, the cortical nephrogram phase of the kidneys and the arterial phase of the pancreas may be obtained in an early scan or scans. Since an element of optimization of these is clearly appropriate, a higher injection rate of the agent may be used (4–5 ml/s rather than 3 ml/s) but with a total dose aimed at the PV phase. The parenchymal phase of the pancreas may be obtained at about 35 s and a portal venous phase scan incorporated in the
1059
PV liver scan will demonstrate a good parenchymogram, now with useful demonstration of venous drainage. A “delayed phase” of the kidneys may be obtained simply by a third scan. If, unusually, only the kidneys, pancreas or spleen, etc., were to be imaged, this could indeed be achieved with a significantly smaller total dose of agent.
Functional/physiological imaging This important field will not be discussed here. No change of contrast agent administration regimen is called for when multi-slice as opposed to singleslice or incremental CT is used.
Summary At first sight it would seem that shorter infusion times of consequently lower total doses of contrast agent are possible in CTA with faster machines. However, this is to ignore the obvious that a shorter injection will only achieve lower blood concentrations. This cannot be effectively offset by faster injection rates, as rates higher than some 5 ml/s achieve little and increase the risk of extravasation significantly. The use of higher concentration formulations of contrast may help to a limited extent, but adequate levels will probably only be achieved when this is combined with a prolongation of injection and a delay of commencement of the short scan sequence until higher blood levels are reached. Obviously, these manoeuvres increase the contrast dose. As regards all other CT organ imaging, physiological and pharmacological fundamentals dictate that only a marginal reduction in total dose of contrast medium may be achieved by using faster infusions and exploiting the faster scanning capacity of multi-slice CT. Fast multi-slice CT obviates the need for biphasic contrast agent infusion regimens promulgated to mitigate the early liver equilibrium phase problem. In short, there is no clear need, and little to be achieved, by modifying for multi-slice systems protocols developed for single -slice systems. The same conclusion would be essentially valid for 16slice machines (or more) when these become available.
Appendix A It may be shown8 that for an infusion of contrast agent at N mg Iodine per s, the blood iodine
1060
P. Dawson
concentration, B(t), at time t is given by: BðtÞ Z
h i KKt N ð1 K HctÞ 1 K e Vp K
nature and timing of the ‘equilibrium point’ and the ‘equilibrium phase’.9
References where K is the rate constant for leak of agent from the plasma to the interstitial space and Hct is the haematocrit. Since contrast agents do not to any significant degree enter cells in the circulation, the plasma concentration, P(t), and the blood concentration, B(t), are related by B(t)Z(1KHct)P(t). Interstitial concentrations of contrast, which cannot be monitored directly, are given approximately by the difference equation: Cis ðt C DtÞ zCis ðtÞ C ½ðK12 PðtÞÞ=Vis K ðK21 Cis ðtÞÞ=Vis D where K12 and K21 are rate constants describing transfer from plasma to interstitium and interstitium to plasma, respectively, and Vis is the interstitial volume. The smaller the increments, Dt, chosen, the more precise the calculation. A tissue region of interest (ROI) will, at any time, have a net concentration, C(t), given by a weighted sum of the blood and interstitial concentration at that time: CðtÞ Z :B BðtÞ C :is Cis ðtÞ The weighting factors, :B and :is, are the fractional blood and interstitial volumes in the ROI. These equations may be used to generate model curves very close to those observed in clinical practice,8 and they may further be used to cast light on such important issues as the
1. Dawson P, Lees WR. Multislice technology in computed tomography. Clin Radiol 2001;56:302–30. 2. Dawson P, Blomley M. Contrast agent pharmacokinetics revisited. I. Reformulation. Acad Radiol 1996;3:S261–S2. 3. Blomley M, Dawson P. Contrast agent pharmacokinetics revisited. II. Computer-aided analysis. Acad Radiol 1996;3: S264–S7. 4. Krause W. Application of pharmacokinetics to computed tomography. Injection rates: mono-, bi- or multi-phasic. Invest Radiol 1996;31:91–100. 5. Bae TB, Heiken JP, Brink JA. Aortic and hepatic peak enhancement at CT: effect of contrast medium injection rate—pharmacokinetic analysis and experimental porcine model. Radiology 1998;206:455–64. 6. Bae TB, Heiken JP, Brink JA. Aortic and hepatic peak enhancement at CT. Part I. Prediction with a computer model. Radiology 1998;207:647–55. 7. Bae TB, Heiken JP, Brink JA. Aortic and hepatic peak enhancement at CT. Part II. Effect of reduced cardiac output in a porcine model. Radiology 1998;207:657–62. 8. Dawson P, Blomley M. The value of mathematical modelling in understanding contrast enhancement in CT with particular reference to the detection of hypovascular liver metastases. Eur J Radiol 2002;41:222–36. 9. Dawson P, Morgan J. The meaning and significance of the equilibrium phase in enhanced computed tomography of the liver. Br J Radiol 1999;72:438–42. 10. Kormano M, Dean P. Extra-vascular contrast material: the major component of contrast enhancement. Radiology 1976; 121:379–82. 11. Blomley MJK, Dawson P. Bolus dynamics: theoretical and experimental aspects. Br J Radiol 1997;70:351–9. 12. Foley WD, Karcaaltincaba M. Computed tomography angiography: principles and clinical applications. J Comput Assist Tomogr 2003;27:S23–S30.
REVIEW
Multi-slice CT contrast enhancement regimens P. Dawson* Department of Imaging, UCL Hospitals, London, UK Received 21 April 2004; received in revised form 14 July 2004; accepted 14 July 2004
KEYWORDS Computed tomography (CT); Contrast media; Infusion protocols
Historically, the development of progressively faster Computed Tomography (CT) technology has dictated a recurrent need to re-examine intravenous contrast agent enhancement regimens. The most recently introduced development, the very fast, multi-slice helical/spiral systems, have raised the same issue yet again. It is possible, exploiting the technology to its maximum potential as regards speed, to perform an examination many times faster (depending on the number of detector rings from 4 to 64) even than with earlier single slice spiral instruments. In order to optimise image quality, such maximal speed gains will not usually be sought but, nevertheless, imaging time will generally be substantially reduced. It is natural that the question of a possible need to modify contrast agent enhancement protocols designed for an earlier generation of slower machines should again be considered. Using as a basis known contrast agent pharmacokinetics and results of modelling techniques, the matter is tackled in this paper. q 2004 The Royal College of Radiologists. Published by Elsevier Ltd. All rights reserved.
Introduction The history of CT development has been marked by a number of stepwise increases in scanning speed. The biggest such step in recent years was taken when spiral/helical technology replaced incremental slice technology. Now multi-slice spiral systems have given further impetus to this process. As explained elsewhere,1 a 4-detector row multi-slice system with a 0.5 s rotation time may be eight times as fast as a single-slice system with a typical 1 s rotation time and 16-slice or 64-machines may, of course, be even faster. Two caveats might usefully be entered here. First, in the interests of optimization of image quality, exploitation of such maximum possible gains in image acquisition speed should not be routine. Secondly, the most useful application of these systems will not always be in the fast coverage of large anatomical volumes in * Guarantor and correspondent: P. Dawson, Department of Imaging, UCL Hospitals, The Middlesex Hospital, Mortimer Street, London WIT 3AA, UK; fax: 00-44-207-380-9068. E-mail address: [email protected].
shorter times but, rather, will frequently be in the coverage of modest volumes in reasonable times with high Z-axis (isotropic) spatial resolution1 during a single breath-hold. These points having been made, this technology again raises the sort of question occurring repeatedly throughout the history of CT development: even if we consider that, in practical terms, the typical speed increase is usually only by a more modest factor of, say, three rather than eight or more, does this call for any radical modification in contrast agent administration protocols? At first sight it might seem reasonable to suggest that faster contrast agent injections would be appropriate to the shorter scan times. Perhaps different concentration formulations of contrast agents from those typically used might be appropriate. The ideal outcome of such a re-examination for both imaging department staff and patients would be a conclusion that the faster delivery of smaller total loads of contrast should optimize enhancement in the shorter timescales involved. It is the purpose of this article to examine these and related questions.
0009-9260/$ - see front matter q 2004 The Royal College of Radiologists. Published by Elsevier Ltd. All rights reserved. doi:10.1016/j.crad.2004.07.009
1052
The ideal approach to questions of practical clinical importance would be a practical clinical method, that is a clinical trial. However, since there are several variables to be considered, some under control (e.g. concentration, total dose, delivery rate of agent) and some not (patient body weight, cardiac output), either a very large trial or a whole series of comparative trials would have to be undertaken. In view of this it seems reasonable to suggest that pharmacokinetic modelling might be a way of obtaining some of the information required. Several authors have demonstrated the efficacy of such modelling.2–8
P. Dawson
6.
7.
8.
Pharmacokinetics of contrast agents This subject and its application to the modelling of contrast enhancement in CT has been described in detail elsewhere.2–8 A brief outline is given in Appendix 1. It will be helpful for the following analysis and discussion to set out some general points of principle established by such modelling and by clinical experience. 1. Intra-vascular contrast agents are sometimes helpful simply to ‘label’ blood vessels to help in image interpretation, as oral or rectal contrast may be used to label bowel.8 For this purpose, only modest blood levels of contrast agent are necessary and precise timing is not critical. 2. For CT angiography (CTA) proper, higher blood levels maintained throughout the scan are, of course, required and timing is critical. This will be discussed further below. 3. The principal reason for intra-vascular (IV) contrast agent administration is to increase the contrast between structures to render them more visible and, in particular, to enhance the difference between normal and abnormal tissue CT numbers, so as to render lesions more conspicuous. It has been shown both in clinical practice and theoretically, that the conspicuity of typical relatively hypovascular liver lesions increases as the normal liver tissue CT number increases. This suggests that not too great an emphasis on economy in contrast agent use would be best policy. 4. Even though image acquisition may now be very fast, we cannot argue that as a consequence the imaging time window may now be narrow. A narrow window may easily be missed, and a reasonably wide time window for imaging remains desirable. 5. The magnitude of arterial enhancement achieved is a function of contrast
9.
10.
11.
12.
concentration, infusion rate and total time of infusion. The magnitude of hepatic enhancement is a function of total iodine dose administered, and may be adjusted by changes in concentration of agent administered, total volume or both. The liver must be scanned in its entirety before onset of the ‘equilibrium phase’, the time of onset of which is itself to a significant degree a function of infusion duration. This argument is valid for any organ but is critical for liver imaging since equilibrium is established very early in the liver.8,9 Before the introduction of helical CT, infusion rates were sometimes prolonged and experiments with two-phase injections were made in an effort to delay the onset of equilibrium (Figs. 1a–d, 2). Additionally, portal venous phase scanning was often started much too early in terms of contrast enhancement in order to beat this deadline. The result was images of the liver which were sub-optimal in enhancement for the early slices and probably, in spite of all efforts, in the equilibrium phase for the later ones. The speed of the scanner determines its ability to record image data during the most advantageous time period, e.g. maximized arterial enhancement, the pre-equilibrium phase in the liver or individual phases in true multiphasic imaging. Cardiac output has a significant influence on timing of arrival and peak of contrast agent concentration in the arteries (relevant to CTA). However, although there is a delay in peak enhancement in the liver too in cases of low cardiac output, this will be little affected except in extreme cases, being much more dependent on total dose delivered. A reduction in cardiac output results in an increase in peak arterial enhancement. The fundamental explanation for this is that with a lower cardiac output less blood is ejected from the heart to dilute the contrast agent during transit, and in practice this over-rides the effect of a prolonged transit time working in the direction of reducing peak enhancement. Of course, all this is only true up to a point— zero cardiac output delivers no contrast agent at all to the arteries. For a given contrast load, arterial and organ enhancement are inversely proportional to body mass, so the contrast total load to body mass ratio is a key determinant of enhancement.
Multi-slice CT contrast enhancement regimens
1053
Figure 1 The plasma and interstitial concentrations (expressed as CT number) for infusion rates of (a) 1 ml per s (NZ1), (b) 2 ml per s (NZ2), (c) 3 ml per s (NZ3) and (d) 5 ml per s (NZ5). The curves cross at the equilibrium point and, the faster the injection rate, the earlier comes the point. 13. Some factors are invariants to contrast agent injection protocol changes. (a) The venous to arterial circulation time is a function of the patient and is not under the radiologist’s control (b) The time of arrival of contrast agent in
the portal vein is also a function of the patient. (c) The cortical nephrogram is ‘immediate’ (subject to circulation time) but the pyelogram is delayed some 3–5 minutes after IV injection.
1054
P. Dawson
Fig. 1. (continued)
Multi-slice CT contrast enhancement regimens
1055
Figure 2 Modelled plasma contrast concentration (P(t)) and interstitium concentration (Cis(t)) curves, expressed as CT number for a monophasic infusion regimen at 3 ml per s (NZ3), and a biphasic infusion regimen at 5 ml per s for 10 s followed by 1 ml per s for 100 s (NZ5.1, respectively). The equilibrium point may be seen to be delayed by some 40 s in the two-phase regimen as opposed to the one-phase regimen.
The factors to be considered in enhanced CT are vascular enhancement (blood contrast agent concentrations) and tissue/organ ROI enhancement (a weighted sum of micro-vascular and interstitium contrast agent concentrations).8,9 In CT
angiography the former alone is of interest; in any other examination the latter is of greater interest.8,10 Some other useful results of pharmacokinetic modelling8 might usefully be summarized here.
Figure 3 The blood concentration curves for four different injection rates, at 1, 2, 3 and 5 ml/s. As the rate increases higher peaks are obtained, but the levels are not sustained.
1056
P. Dawson
Figure 4 The blood concentrations achieved in a two-phase injection (here 5 ml/s for 10 s followed by 1 ml/s for 100 s) produces blood levels comparable with those achieved by a one-phase 3 ml/s injection but somewhat better sustained. Such manoeuvres are unnecessary with faster multi-slice scanners.
Detailed pharmacokinetic analysis predicts to some extent what seems obvious, as follows. (a) Faster injections of the same total dose give higher peak blood flow values at earlier times, but the high values are sustained for a shorter period. This may be seen from Fig. 3 in which modelled curves are presented for infusion rates of 1, 2, 3 and 5 ml/s. The effect of a biphasic injection is shown in Fig. 4. (b) Varying the infusion rates over the range typically used clinically has different effects on arteries and organs. The magnitude of arterial concentrations is proportional to the flux of agent, i.e. to flow rate!concentration. Conversely, the portal venous phase enhancement of the liver is quite insensitive to changes in contrast administration flux over this range but depends more simply on total load administered. Therefore, although perhaps some reduction in contrast dose in CTA may be possible with multi-slice scanning (see below), significant reductions in dose for hepatic imaging, and therefore in effect for abdominal imaging in general, are unlikely because of total dose dependency dominance. (c) On the other hand, where the hepatic arterial phase of liver enhancement is of interest, this is dependent on injection rate. As regards faster injections in general, it is important to note
that there is little to be gained by pushing the infusion rate beyond about 5 ml/s since the right heart and lungs represent a buffer preventing achievement of higher and higher concentrations by faster and faster injections.11
Discussion and illustrative examples CT angiography This may be seen as a special case. In all other CT examinations it is the combination of vascular and extravascular contrast that effects the enhancement; in the case of CT angiography it is, by definition, the intravascular agent that contributes. Here we may state some basic principles. 1. An adequate blood concentration of contrast must be achieved and sustained during the time of the data acquisition. 2. Cephalo-caudad data acquisition during a single breath-hold (where a breath-hold is necessary) of around 20 s should be the aim. 3. Accurate timing of the start of data acquisition should be achieved using bolus timing techniques. 4. On the basis of (3), (1) may be achieved by
Multi-slice CT contrast enhancement regimens
1057
approximately matching agent infusion time to necessary scanning time. This in turn may be seen to determine total dose required.
disastrous), and the 5 ml/s infusion rate would dictate a total dose of some 200 ml. This is unacceptable and the choices would be between:
It is more convenient in CTA to think in terms of flux of contrast agent, i.e. rate of infusion! concentration. This may be increased to achieve higher blood concentrations and consequent improved signal-to-noise ratio either by increasing the rate of injection, or by increasing the concentration of the contrast preparation, or both. Some would argue that in this application a higher concentration formulation of the agent is useful.12 As has been noted earlier, however, there is little to be gained by pushing the infusion rate beyond about 5 ml/s. The total dose to be delivered is then dictated by the desirability of approximately matching the infusion time to the scan time as suggested above. One misunderstanding should be cleared up. There is a widespread perception that a prolonged infusion achieves a plateau of contrast concentration in blood vessels. This is not the case: concentrations continue to increase, albeit at an ever slower rate, and become asymptotic to a plateau never reached. The application and consequences of these principles are illustrated by a few examples.
† reducing the injection rate to approximately 3 ml/s; † accepting a lower blood concentration and reducing contrast load to 120 ml; † accepting a reduced z-axis resolution by scanning at 4!2.5 mm collimation, to halve the scan and infusion times; † or using an 8-, 16- or 64-slice machine capable of covering the region at the same high z-axis resolution in a significantly shorter time.
Aorto-iliac CTA Such a study might be done for suspected abdominal aortic aneurysm. The extent of the region of the abdominal aorta and the iliac vessels to the level of the proximal femoral vessels is some 40 cm. Using a 4-slice machine, 4!1.25 mm collimation, a pitch of 1.5 and a gantry rotation time of 0.5 s (dictating a table speed of 15 mm/s), this can be covered in about 25 s, suitable for an average breath-hold. If an infusion rate of 5 ml/s is chosen to obtain high levels of contrast, then about 125 ml of contrast would give a match of acquisition and infusion times. To study the smaller volume needed for renal artery assessment (no more than 20 cm), the acquisition time is approximately halved so the total contrast dose required is halved. This in turn has a consequence discussed below under Shorttime scans. Thoraco-lumbar aortography This is an anatomical block of greater length, approximately 60 cm. Were the above parameters to be used with a 4-slice machine, the scan time would be 40 s. This is not an easy breath-hold (although with a cephalo-caudad scan breathing during the later pelvic portion would not be
Aorto-bifemoral CTA The distance from the distal infrarenal abdominal aorta to the ankles is some 120 cm. With a 4-slice machine, 4!2.5 mm collimation, a pitch of 1.5 and a gantry rotation time of 0.5 s (dictating a table speed of 3 cm/s) this would be covered in about 40 s. With an infusion rate of 4 ml/s the infusion total load would be rather large at 160 ml—and this for a not very high z-axis resolution. The answer here is clearly to use an 8-, 16- or 64-slice machine, since even dropping the infusion rate to 3 ml/s would be of little help in limiting contrast dose if 4!1.25 mm acquisition was decided upon. The essential point concerning total doses in CTA is not so much that they can be reduced with faster and faster machines (see 8-slice and 16-slice examples above) but, rather, that good CTA could not previously have been achieved at all. It would be unreasonable to be too prescriptive, and the above are examples to illustrate general principles. Design of protocols may be seen as an art. It is simply best to base some artistic endeavours on immutable scientific principles. So, no protocol may be considered in some absolute sense “right” but many may be deemed “wrong”. Very-short-time scans There is an important point to be made about short acquisition time scans taking less than about 20 s and the need ideally to match scan and injection times. Two problems arise. First, timing becomes even more critical because of the narrow timewindow; and, secondly, the contrast concentrations reached in blood after a short infusion time matched to a short data acquisition time is relatively low. As regards the latter problem, two strategies may be adopted. The contrast flux may be increased; this is achieved by use of a higher concentration formulation rather than by an increase in rate of injection since the latter, as
1058
P. Dawson
Figure 5 Blood and liver contrast enhancement curves for a region of interest (ROI) in the liver. It may be seen that the liver enhancement peaks at 60–70 s. A typical modest image acquisition time (much shorter is possible) for a multislice CT machine is indicated, demonstrating that it can be completed during a high-enhancement period and, even if begun as late as 60 s, will be completed before 100 s and therefore before equilibrium.
previously discussed, is of little avail above approximately 5 ml/s. Alternatively, a longer infusion may be carried out and the scanning commenced later when blood levels are higher. Can total doses be reduced in CTA with multislice CT? In one sense this is a non-question, in that only now with multi-slice systems can CTA be done well and so the relative dose issue is rather meaningless. What can be said is that savings in contrast dose are not easy to make because certain minimum blood concentrations must be reached. When superfast systems are used to reduce the scan time, the injection time, which needs to be matched to it, is reduced. If the injection time is short, only low blood levels will be achieved unless the contrast flux during the short time period of injection is very high. Since injection rates higher than about 5 ml/s are not much more effective, the only options are to increase contrast concentration and/or to prolong the injection time to deliver more contrast and to delay imaging until the higher levels have been achieved. As far as CTA is concerned, the scope for contrast agent savings with faster scanners would seem, therefore, to be limited by fundamentals.
CT pulmonary angiography This is, of course, a specific example of CT angiography. There is, however, an argument about concentrations of contrast agent to be used and vessel enhancement to be ideally achieved. In this examination the search is for intraluminal filling defects that may actually be hidden by too dense a contrast column. Some authorities recommend a dilute contrast infusion, in which case the above arguments do not apply.
Brain imaging Here there are no complex issues. 50–100 ml suffice and, functional studies aside, there is no need for fast delivery. No modification of long-established protocols is necessary.
Abdominal imaging In virtually all abdominal imaging the liver will also be imaged, and should be imaged optimally. On this assumption we conclude that the delivery of contrast has to be tailored to this end with other organs being, perhaps, imaged earlier or later in the evolution of the distribution of contrast agent as
Multi-slice CT contrast enhancement regimens
appropriate (discussed below). Therefore, liver imaging will be considered first.
The liver The liver poses unique challenges because of its dual blood supply and because of the need to complete imaging before onset of a very early equilibrium phase.9 A key determinant of this is the injection duration.8 The issues surrounding the two blood supplies of the liver are well enough understood. We are most often interested in the portal venous phase. As discussed above, this does not occur maximally until some 60–70 seconds after commencement of infusion (Fig. 5). Fast multi-slice systems allow the commencement of scanning in this phase to be delayed until some such time with the confidence that the whole liver may be completely covered, even at highest z-axis resolution, before the onset of equilibrium.9 It is no longer important with fast scanning to be concerned about the fact that the faster the infusion rate the sooner the equilibrium point arrives and the equilibrium phase begins (Fig. 1a–d). Furthermore, it has been shown that the conspicuity of hypovascular liver lesions, the usual issue, increases as the normal liver enhancement increases in the PV phase. Leaving aside the question of timing, the enhancement of the liver by its principal blood supply is a function of the total dose given.8 It therefore follows that it is unlikely we can achieve more than a marginal reduction in total dose administered without sacrificing information. It is worth noting here that there is a divergence between mainstream US and European practice here, in that a typical European total dose for a liver study would be 120 ml of an agent 300 mgI/ml concentration, whereas a typical US dose would be 150–180 ml.
Hepatic arterial phase imaging In the context of a contrast administration regimen tailored to optimize liver portal venous (PV) phase enhancement, any other imaging can be incorporated: for example, the hepatic arterial (HA) phase of the liver, the cortical nephrogram phase of the kidneys and the arterial phase of the pancreas may be obtained in an early scan or scans. Since an element of optimization of these is clearly appropriate, a higher injection rate of the agent may be used (4–5 ml/s rather than 3 ml/s) but with a total dose aimed at the PV phase. The parenchymal phase of the pancreas may be obtained at about 35 s and a portal venous phase scan incorporated in the
1059
PV liver scan will demonstrate a good parenchymogram, now with useful demonstration of venous drainage. A “delayed phase” of the kidneys may be obtained simply by a third scan. If, unusually, only the kidneys, pancreas or spleen, etc., were to be imaged, this could indeed be achieved with a significantly smaller total dose of agent.
Functional/physiological imaging This important field will not be discussed here. No change of contrast agent administration regimen is called for when multi-slice as opposed to singleslice or incremental CT is used.
Summary At first sight it would seem that shorter infusion times of consequently lower total doses of contrast agent are possible in CTA with faster machines. However, this is to ignore the obvious that a shorter injection will only achieve lower blood concentrations. This cannot be effectively offset by faster injection rates, as rates higher than some 5 ml/s achieve little and increase the risk of extravasation significantly. The use of higher concentration formulations of contrast may help to a limited extent, but adequate levels will probably only be achieved when this is combined with a prolongation of injection and a delay of commencement of the short scan sequence until higher blood levels are reached. Obviously, these manoeuvres increase the contrast dose. As regards all other CT organ imaging, physiological and pharmacological fundamentals dictate that only a marginal reduction in total dose of contrast medium may be achieved by using faster infusions and exploiting the faster scanning capacity of multi-slice CT. Fast multi-slice CT obviates the need for biphasic contrast agent infusion regimens promulgated to mitigate the early liver equilibrium phase problem. In short, there is no clear need, and little to be achieved, by modifying for multi-slice systems protocols developed for single -slice systems. The same conclusion would be essentially valid for 16slice machines (or more) when these become available.
Appendix A It may be shown8 that for an infusion of contrast agent at N mg Iodine per s, the blood iodine
1060
P. Dawson
concentration, B(t), at time t is given by: BðtÞ Z
h i KKt N ð1 K HctÞ 1 K e Vp K
nature and timing of the ‘equilibrium point’ and the ‘equilibrium phase’.9
References where K is the rate constant for leak of agent from the plasma to the interstitial space and Hct is the haematocrit. Since contrast agents do not to any significant degree enter cells in the circulation, the plasma concentration, P(t), and the blood concentration, B(t), are related by B(t)Z(1KHct)P(t). Interstitial concentrations of contrast, which cannot be monitored directly, are given approximately by the difference equation: Cis ðt C DtÞ zCis ðtÞ C ½ðK12 PðtÞÞ=Vis K ðK21 Cis ðtÞÞ=Vis D where K12 and K21 are rate constants describing transfer from plasma to interstitium and interstitium to plasma, respectively, and Vis is the interstitial volume. The smaller the increments, Dt, chosen, the more precise the calculation. A tissue region of interest (ROI) will, at any time, have a net concentration, C(t), given by a weighted sum of the blood and interstitial concentration at that time: CðtÞ Z :B BðtÞ C :is Cis ðtÞ The weighting factors, :B and :is, are the fractional blood and interstitial volumes in the ROI. These equations may be used to generate model curves very close to those observed in clinical practice,8 and they may further be used to cast light on such important issues as the
1. Dawson P, Lees WR. Multislice technology in computed tomography. Clin Radiol 2001;56:302–30. 2. Dawson P, Blomley M. Contrast agent pharmacokinetics revisited. I. Reformulation. Acad Radiol 1996;3:S261–S2. 3. Blomley M, Dawson P. Contrast agent pharmacokinetics revisited. II. Computer-aided analysis. Acad Radiol 1996;3: S264–S7. 4. Krause W. Application of pharmacokinetics to computed tomography. Injection rates: mono-, bi- or multi-phasic. Invest Radiol 1996;31:91–100. 5. Bae TB, Heiken JP, Brink JA. Aortic and hepatic peak enhancement at CT: effect of contrast medium injection rate—pharmacokinetic analysis and experimental porcine model. Radiology 1998;206:455–64. 6. Bae TB, Heiken JP, Brink JA. Aortic and hepatic peak enhancement at CT. Part I. Prediction with a computer model. Radiology 1998;207:647–55. 7. Bae TB, Heiken JP, Brink JA. Aortic and hepatic peak enhancement at CT. Part II. Effect of reduced cardiac output in a porcine model. Radiology 1998;207:657–62. 8. Dawson P, Blomley M. The value of mathematical modelling in understanding contrast enhancement in CT with particular reference to the detection of hypovascular liver metastases. Eur J Radiol 2002;41:222–36. 9. Dawson P, Morgan J. The meaning and significance of the equilibrium phase in enhanced computed tomography of the liver. Br J Radiol 1999;72:438–42. 10. Kormano M, Dean P. Extra-vascular contrast material: the major component of contrast enhancement. Radiology 1976; 121:379–82. 11. Blomley MJK, Dawson P. Bolus dynamics: theoretical and experimental aspects. Br J Radiol 1997;70:351–9. 12. Foley WD, Karcaaltincaba M. Computed tomography angiography: principles and clinical applications. J Comput Assist Tomogr 2003;27:S23–S30.