The Effects of Contrast Agent Extravasation Perfusion Parameter Maps Derived from Dynamic Susceptibility Contrast Magnetic Resonance Imaging



The Effects of Contrast Agent Extravasation Perfusion Parameter Maps Derived from Dynamic Susceptibility Contrast Magnetic Resonance Imaging


Kathleen M. Schmainda

Eric S. Paulson

Jerrold L. Boxerman



Dynamic susceptibility contrast (DSC) magnetic resonance imaging (MRI) has been studied for over 20 years for the evaluation of brain tumor vascular morphology and function.1,2,3,4,5,6,7,8 The general principle underlying DSC methods is that the bolus administration of a gadolinium (Gd)-chelated contrast agent (the most commonly used MR contrast agent) induces a gradient of susceptibility or “magnetizability” between the contrast-containing vessel and tissue surrounding the vessel. The image signal intensity acquired during this administration transiently decreases.9 The signal change is converted into a relaxation rate change, which is proportional to the fraction of blood volume within each image pixel. From these curves, tracer kinetic principles can be applied in order to extract measures of cerebral blood flow (CBF), cerebral blood volume (CBV), and mean transit time. Figure 28.1 illustrates the steps necessary to collect and process these data for the determination of CBV. The CBV is often reported as a relative value, such that it can be qualitatively compared across the brain or reported as a value normalized to the relative CBV (rCBV) of normal-appearing brain tissue. As such CBV is often not reported in absolute units and instead referred to as rCBV. The rCBV is used most often for the assessment of brain tumors, which are notorious for contrast agent extravasation, and will therefore be the focus of this chapter. The effect of contrast agent extravasation on the DSC parameters, CBF, and mean transit time will also be discussed but not to a lesser extent because they have been used less often for the assessment of brain tumors.






FIGURE 28.1. Collection and processing of relative cerebral blood volume (rCBV) data. Shown in clock-wise order starting from the bottom left corner are (A) echo planar images being collected over time (B) the signal time course (S (t)) from one image voxel as contrast agent is bolus administered, resulting in a transient decrease in signal, (C) conversion of the S (t) into the T2 relaxation rate changes over time (ΔR2(t)) and (D) the final rCBV map shown to the right of the corresponding postcontrast T1-weighted image.

Determination of brain tumor rCBV from the contrast-enhanced MR signal can be complicated by a leaky blood–brain barrier (BBB), as is often the case with tumors. Under these conditions, Gd contrast agents extravasate or leak out of the vessels into the brain or tumor tissue, thereby diminishing the transient susceptibility effect (a smaller signal decrease) of the contrast agent passing through the vasculature. Owing to altered T1 or T2/T2* relaxation times after contrast accumulation in tissue, it also has the effect of altering the postbolus signal intensities, which of themselves can contribute uncertainties regarding the most accurate way to acquire and process these data. As a result of this issue, a variety of methods to acquire and analyze the DSC MRI data have been used. Consequently, although there exists a plethora of reports demonstrating the feasibility and utility of rCBV to evaluate brain tumors in patients,1,2,3,5,6,10,11,12,13,14,15,16,17,18 the lack of a commonly accepted method to most effectively handle these leakage effects has delayed a more widespread acceptance of rCBV in the clinic. It is therefore the goal of this chapter to clearly outline the issues of contrast agent extravasation and the various methods proposed to address them. Specifically, the following choices can affect the sensitivity of the DSC MRI to contrast agent leakage effects. These include choice of (a) contrast agent material, dose, and injection protocol; (b) pulse sequence, with the most common choices being gradient-echo, spin-echo, and multiecho time gradient echo or spin-echo and the parameter settings including flip angle, repetition time (TR), and echo time (TE); and (c) the postprocessing leakage-correction method. Although it is the combination of these choices that affect the degree to which contrast agent leakage affects rCBV, each aspect will be discussed separately.


Influence of Contrast Agent Material, Dose, and Injection Protocol

DSC images, which are used to create rCBV maps, rely on the compartmentalization of contrast agent for maximal effect. However, Gd-chelated contrast agents, which are
the most commonly used MR contrast agents, are small and can leak out of brain tumor and tissue vessels when the BBB is disrupted. This results in confounding leakage effects on the MRI signal and consequently can lead to inaccuracies in the determination of rCBV maps. However, these effects can be minimized by preloading the tissue with a dose of Gd contrast agent as described previously. Alternatively, using contrast agents that remain intravascular precludes contrast agent leakage effects entirely. However, use of these agents has their own practical and processing challenges, which are also described here.


Use of Contrast Agent Preload to Diminish T1 Leakage Effects

The preload or loading dose of a Gd contrast agent, which is typically 0.05 to 0.1 mmol/kg Gd, is usually half of or equal to the dose of contrast given during the DSC acquisition. This preload is administered to diminish any T1 changes that might subsequently occur with the bolus administration of contrast agent during the collection of DSC images. Specifically, it works by entering the interstitial space and decreasing the water T1 relaxation time. Then, subsequent doses of contrast agent, which also enter the interstitial space, will decrease the T1 relaxation time by a much smaller percentage, thus resulting in less of a competing T1 effect on the DSC signal during and after the bolus passage. The data shown in Figure 28.2 demonstrate this effect. Shown is gradient-echo echo-planar imaging (EPI) data continuously collected over time from a rat C6 glioma. After the first injection of Gd contrast agent, no discernable transient decrease in signal is observed. Rather, because of the leaky nature of this tumor, the signal immediately increases, demonstrating a strong T1 leakage effect. After administering the second dose of Gd contrast agent, a discernable signal transient becomes apparent. Thus, for voxels containing highly permeable vasculature, it would not be possible to compute rCBV without the loading dose of contrast agent. Yet, note that although there is now a signal transient from which rCBV can be computed, the postbolus signal still exceeds the baseline signal, suggesting that some leakage effects remain, and therefore further correction of these effects are necessary. This shows that it is not so much the magnitude of the effect of the contrast agent preload that matters, but rather that it results in a signal that can be processed to produce rCBV. This also provides the rationale for why additional postprocessing correction for leakage may still be necessary, as is described later.






FIGURE 28.2. Effect of loading dose of contrast agent in rat C6 glioma. Shown is the gradient-echo echo planar signal intensity collected every 1 second through the first (preload) and second injection of gadolinum contrast agent. Note that due to the leakiness of the C6 glioma there is no transient decrease in the T2*-weighted signal during the first passage. The T1 leakage effects dominate as evidenced by the rising signal. The second dose of contrast agent does result in a discernable transient decrease in the T2*-weighted signal (open arrow), followed by a continued signal increase suggesting remaining T1 leakage effects.






FIGURE 28.3. The permeability scaling factor as a function of loading dose T1, for both GE (gradient-echo) and SE (spin-echo) signals. This factor, which is equal to the permeability correction factor K2, can be minimized if the loading dose is sufficiently high. This result suggests that leakage effects can be made negligible by sufficiently saturating the tissue with contrast agent before the first pass susceptibility (PS) study is performed. (Adapted from Figure 6 in Donahue KM, et al. Utility of simultaneously acquired gradient-echo and spin-echo cerebral blood volume and morphology maps in brain tumor patients. Magn Reson Med. 2000;43:845–853.)

Figure 28.3 shows a simulation, previously published,3 that suggests the greater the loading dose, or the shorter the loading dose T1, the greater will be the T1 saturation. As a result, the leakage effect is minimized as indicated by a smaller permeability scaling or correction factor, K2. However, there are obviously limits to the amount of Gd contrast agent that can be administered. To address this issue, Hu et al.19 undertook a comprehensive study to determine the optimal loading dose and incubation time, defined as the time between the administration of the preload and the collection of DSC MRI data, which is necessary for the most accurate distinction between tumor and reference brain rCBV. They concluded that a “standard” loading dose of 0.1 mmole/kg with an incubation time of 6 minutes proved best. Even better distinction between tumor and posttreatment radiation effect
resulted when, in addition to using a contrast preload, a postprocessing correction scheme was also used.






FIGURE 28.4. Comparison of relative cerebral blood volume (rCBV) discrepancies (gadolinium-monocrystalline iron oxide nanocompounds [Gd-MION]/MION) for each correction scheme permutation (mean and 95% confidence interval). Although there is no statistically significant intrascheme or interscheme bias, mean discrepancy is closest to zero for P+C+ (−1.8%) followed by P+C− (+7.6%), P−C+ (+38.3%) and P−C− (−142.8%). The variance of rCBV discrepancies differed substantially between correction schemes, with P+C− (22-fold), P−C+ (32-fold), and P+C+ (267-fold) all statistically significantly lower compared to P−C−. The use of both correction techniques (P+C+) further significantly reduced the variance compared with that for each individually (12-fold vs. P+C−, 8-fold vs. P−C+). C, correction; P, preload. (Adapted from Figure 4 in Shen T. et al. Monocrystalline iron oxide nanocompounds (MION): Physiochemical properties. Magn Reson Med. 1993;29(5):599–694.)

Note that an important disadvantage of using a loading dose is the extra Gd contrast agent administered. Given the recent concerns regarding necrotizing systemic fibrosis20 related to use of Gd contrast agents in patients with compromised kidney function, use of extra contrast agent is often discouraged and has to be considered in the risk versus benefit ratio for patients. Also, when comparing postcontrast anatomical images, consistency of contrast agent dose should be maintained. One approach is to collect the standard postcontrast anatomic images after the administration of a standard dose preload, followed by the collection of the DSC data. In this way all postcontrast anatomic images will be consistently collected after a standard dose of contrast agent and will not be influenced by the extra contrast agent dose used for the DSC studies.

The importance of using a preload of contrast agent was further supported by a recent study, performed in the rat 9L gliosarcoma brain tumor model, where the determination of rCBV using a Gd contrast agent was compared with rCBV using monocrystalline iron oxide nanoparticles (MION),21 a contrast agent that because of its size usually remains intravascular.22 Figure 28.4, adapted from this study, shows that the difference between the Gd-rCBV and MION-rCBV was progressively diminished by either performing postprocessing correction (C+) or using a preload of contrast agent (P+) and is closest to zero when both the preload (P+) and postprocessing correction (C+) were performed. Similarly, in a more recent study performed in patients, the difference between the rCBV obtained from high-grade vascular tumors and reference brain was most clearly distinguished when the DSC data were obtained after a loading dose of contrast agent, a finding that was true for several different postprocessing methods.23 An example of the results described in this study are shown in Figure 28.5, showing a comparison of the rCBV results obtained with Acquisition Method C (with a preload) to those obtained using Acquisition Method A (without a preload).






FIGURE 28.5. Relative cerebral blood volume (rCBV) maps showing the effect of using a contrast agent predose. One slice of the rCBV image maps collected from a 21-year-old patient diagnosed with a malignant glioneuronal tumor. The top (A–F) and bottom rows (G–L) are maps created from data collected with Acquisition Methods A and C, respectively. Although both methods use a gradient-echo echo-planar imaging sequence with a 90-degree flip angle, Method A is collected during the administration of a standard dose, and Method C is collected after a predose and during administration of a double dose of contrast agent. Each column corresponds to the different analysis methods applied: Methods 1 (A, G), 2 (B, H), 3 (C, I), 4 (D, J), 5 (E, K), 6 (F, L). Note that when a predose of contrast agent is not used (top row), the variability in tumor blood volume is very dependent on the choice of analysis method with negative or zero tumor blood volume resulting in some cases. Collecting data after a predose (bottom row) decreases the dependence of the tumor blood volume on the chosen analysis method. (Adapted from Paulson ES, Schmainda KM. Comparison of dynamic susceptibility-weighted contrast-enhanced MR methods: Recommendations for measuring relative cerebral blood volume in brain tumors. Radiology. 2008;249(2):601–613.)


Use of Intravascular Contrast Agents

Although the preload method improves the estimation of tumor rCBV when Gd-based contrast agents are used, the degree to which the leakage effect is diminished will depend on the dose of contrast agent preload3 and the vascular permeability and leakage space (i.e., the extravascular–extracellular space) of the tissue.24 Also, as mentioned
previously and discussed in more detail later, postprocessing methods to correct for remaining leakage effects are also recommended. For these reasons, the use of intravascular contrast agents have been proposed as an alternative solution to the confounding issues owing to contrast agent leakage effects. In most cases, the intravascular contrast agents used have been iron oxide–based agents, such as MION21 or ultrasmall superparamagnetic iron oxide contrast agents.25,26 Many preclinical studies have used these agents to evaluate brain tumor blood volume and vessel size index.22,25,27,28,29,30 In addition, initial studies have been performed to demonstrate the feasibility of using iron oxide contrast agents for the determination of rCBV in patients.24,31 In these studies, rCBV was determined from the first passage of the iron oxide contrast agent, ferumoxytol, which is not only an intravascular agent but also a T2* agent in that it should minimally affect T1. The resulting rCBV maps were compared with those determined from images obtained during the first passage of Gd-based contrast material.24 From this study it was concluded that the iron oxide images resulted in more accurate maps of tumor blood volume, while use of Gd-based contrast material gave estimates of tumor blood volume that were lower than expected. However, the Gd-based rCBV maps were obtained from DSC data collected during the initial dose of contrast agent. Therefore, neither a contrast agent preload was used or postprocessing leakage correction applied. Consequently, the improvement in accuracy of iron-based rCBV over Gd-based rCBV measurements may be exaggerated given that neither a preload nor postprocessing method was used to correct for leakage effects and were therefore likely suboptimal. Although preloading may be practiced at some institutions, the use of leakage-corrected postprocessing is even more rare. So these results do represent what is most likely found in routine clinical practice. Yet, while postprocessing correction for leakage may no longer be necessary when using an intravascular agent for dynamic studies, changes in postbolus signal intensities remain because of the shortening of intravascular T2* relaxation time after contrast administration. Although these effects may no longer be a result of contrast agent leakage effects and shortening of tissue T1 relaxation times, they result from contrast agent recirculation effects in the blood plasma. The magnitude of this T2* shortening effect emanating from the vascular space will depend on blood volume and contrast dose and, therefore, must be evaluated and considered appropriately.






FIGURE 28.6. Steady-state perfusion parameter maps. (A) Gradient-echo (ΔR2*) and (B) spin-echo (ΔR2) cerebral blood volume (CBV) maps collected in the presence of steady-state distributions of ferumoxytol. Note the reduction of the impact of large vessels in the spin-echo CBV maps. Also shown are the (C) vessel radii and (D) vessel density maps computed from the ratio of ΔR2* to ΔR2. (Adapted from Christen T, et al. Combined spin and gradient echo imaging following injection of USPIOs in humans. In Proceedings of the International Society of Magnetic Resonance Medicine. Salt Lake City, Utah, 2013.)

Alternatively, if a sufficient dose of intravascular contrast agent can be given, steady-state blood volumes can be determined. This was demonstrated recently by Christen et al.32 who used a combined gradient-echo and spin-echo (SE) imaging sequence to collect images before and during the steady-state distribution of ferumoxytol (1.75 mg Fe/kg). From these images rCBV maps of total blood volume, derived from ΔR2* and microvascular blood volume, derived from ΔR2, were computed, as shown in Figure 28.6. A reduction of the impact on large vessels in the spin-echo CBV maps is apparent. Given that SE data collected under dynamic contrast conditions often suffers from poor signal to noise ratio (SNR), the ability to collect SE data under steady-state conditions would provide an improved ability to probe tissue microvasculature, with the potential of providing more specific information about the disease or treatments that affect tumor vasculature. Also shown are the vessel radius and density maps, which can be computed from the ratio of ΔR2* to ΔR2.

Although promising, only a few iron oxide agents have been approved for clinical use. However, this approval has been given only for liver imaging33 and the treatment of iron deficiency anemia in patients with chronic kidney disease.34 Although this limitation could eventually be overcome, what may be more challenging is the standard clinical practice of performing postcontrast imaging with Gd contrast agents, where lesions become brighter on T1-weighted imaging with uptake enabling their presence to be detected and treatment to be monitored.35 Iron oxide agents cannot replace this function because tumor uptake of these agents causes tumor tissue to become darker. Consequently, if iron oxide agents are used to obtain the rCBV
information, they may need to be administered in addition to a Gd-based contrast agent. Although one early study showed the feasibility and safety of a dual-contrast approach,31 incorporation of this practice into the clinic would require the performance of much larger clinical studies and Food and Drug Administration approval for clinical use. Therefore, if equally accurate information can be obtained using Gd agents for both anatomic and rCBV imaging, it may be less likely that we would move to a dual-contrast agent approach. Consequently, collection of rCBV information using iron oxide agents may remain in the preclinical setting only.

Only gold members can continue reading. Log In or Register to continue

Stay updated, free articles. Join our Telegram channel

Oct 7, 2018 | Posted by in OTOLARYNGOLOGY | Comments Off on The Effects of Contrast Agent Extravasation Perfusion Parameter Maps Derived from Dynamic Susceptibility Contrast Magnetic Resonance Imaging

Full access? Get Clinical Tree

Get Clinical Tree app for offline access