Applied Radiology

Dr. Niendorf is a third-year Radiology Resident at Beth Israel-Deaconess Medical Center, Boston, MA. Prior to residency, he completed his MD, PhD at the University of Wisconsin, Madison, WI. His PhD dissertation developed a magnetic resonance imaging method to measure single kidney glomerular filtration rate. He plans to begin an MRI fellowship after he completes his residency.

Angiogenesis is essential for tumor development. Assessing the response of tumors to new cytostatic treatments, such as anti-angiogenesis agents, requires functional evaluation of the tumor vasculature, in addition to the traditional measures of tumor size. Magnetic resonance perfusion imaging (MRPI) can provide exogenous and endogenous tracer methods to generate information about tumor vascular characteristics such as permeability, perfusion, and blood volume. As a result, MRPI is a critical tool for investigating the efficacy of new therapies and, specifically, for monitoring tumor response to therapy.

Angiogenesis has emerged as a compelling physiologic model in the description of tumor vascular development. Angiogenesis refers to the stimulation of new vessel growth, or neovascularization. This process occurs as part of normal development and during normal remodeling processes, such as in wound healing and in the heart in response to myocardial ischemia. ¹ Increased angiogenic activity is implicated in a variety of disease states as well, including diabetic retinopathy, endometriosis, and certain dermatological disorders such as psoriasis. The influence of angiogenesis on cancer has stimulated the development of anti-angiogenesis agents as a major focus of contemporary research.

Assessing the response of tumors to anti-angiogenic treatment presents challenges. Traditional imaging methods have typically employed size measurements as an index of response to tumoricidal agents. ² Size measurements alone will be less effective when using treatments geared toward a cytostatic effect. At present, the use of clinical parameters, such as quality of life assessments or meaningful impacts on longevity, as measures of response has limited value and is difficult to standardize. Therefore "functional" or "physiologic" imaging techniques have been developed to assist in the evaluation of in vivo response to anti-angiogenic therapy.

Recognizing the increased importance of anti-angiogenesis treatments for tumors throughout the body, this article is aimed toward developing 1) an understanding of angiogenesis and its role in tumor development, 2) an understanding of the problems associated with monitoring the response of tumors to anti-angiogenesis therapy, and 3) an appreciation for the role of magnetic resonance perfusion imaging (MRPI) techniques as they pertain to evaluating neoplasms.

Angiogenesis

In the early stages of development, neoplastic cells rely on the diffusion of sustaining substances from the existing host vasculature to survive. ³ The efficacy of this process limits the size of tumors to several cubic millimeters. In order for tumors to continue to grow, neovascularization is critical. Typically by the time a tumor can be resolved via conventional screening methods, it has expanded into the neovascularity phase. ⁴

Angiogenesis has been described using a four-stage model: stimulation, growth, vessel formation, and stabilization. Cell neovascularization is typically controlled by a balance between angiogenic and anti-angiogenic factors. ⁵ A combination of various stimulating factors can influence angiogenic pathways and weight the balance to upregulate angiogenesis--ie, stimulate growth of endothelial cells in existing microvessels. Stimulating factors can include those related to the microenvironment, such as hypoxia, in-creased lactic acid, and decreased pH. Stimulated endothelial cells then break down the vessel basement membrane via secretion of proteases in order to allow growth toward the tumor. ⁶ The newly sprouting cells form a capillary loop and canalize to form a new vessel. Stabilization of the new vessel is the final step in neovascularization. However, this step may be defective in tumors, allowing prolonged increased permeability, as has been demonstrated in many cerebral neoplasms. ⁷

Since Folkman first suggested the angiogenesis model in tumor development in 1971, increasingly complex pathways for stimulating angiogenesis have been discovered. ⁸ Many different factors affect the balance between promotion and inhibition of neovascularization. The process may be heterogeneous within individual tumors, with multiple coexistent "stages" resulting in the tumor's use of numerous angiogenesis-stimulating factors for continued growth. ⁹ In addition to allowing neoplastic growth, angiogenesis may propagate metastasis by providing prolonged access to circulation via the increased permeability of the neovasculature or by allowing tumor cells to escape into the circulation during the proteolytic stages of angiogenesis. ¹⁰

One of the most well-studied angiogenic factors is vascular endothelial growth factor (VEGF), which regulates vascular permeability and activates endothelial cells. Modulators of VEGF and other angiogenic factors have been proposed for controlling the neoplastic process. A variety of agents are being developed to target distinct stages of angiogenesis. A few mechanisms include inhibiting the initial stimulation of angiogenesis, blocking proteolytic activity, and affecting endothelial cell survival or proliferation. ¹¹

There are several proposed clinical uses of these agents: prophylaxis in patients with a family history of cancer or those harboring premalignant lesions; adjuvant therapy in combination with traditional chemotherapy, radiation therapy, or surgery; and induction of sustained cancer remission. ¹² While the possibility for implementation of single-agent anti-angiogenesis treatment has been raised, the success of preliminary trials with this approach has been limited. This has prompted continued development of new and combined anti-angiogenic therapies. ^13,14

MR techniques

Standard approaches to assess tumors capitalize on the outstanding soft-tissue contrast and sensitivity to contrast media inherent to MR imaging to determine size, enhancement characteristics, and soft-tissue interfaces. MRI is also being employed to noninvasively obtain complementary functional information. In this regard, MR methods directed toward the evaluation of tumor vascularity and angiogenesis represent an important focus of active research. For this purpose, several MR techniques are capable of providing physiologic information, including: MR perfusion imaging, MR spectroscopy, diffusion imaging, and BOLD (blood-oxygenation-level­dependent) imaging. ¹⁵

MR perfusion imaging is a general class of techniques that can be used to evaluate tumor vascular characteristics such as blood flow, plasma volume, permeability, and mean transit time. ¹⁶ The information provided by MRPI is often related to a combination of several of the above tumor characteristics and is influenced by the particular MRPI technique being used, the qualitative tumor vascular characteristics, and the properties of the contrast mechanism (endogenous or exogenous). Understanding the nature of the information provided by MRPI is itself an area of active investigation using various tumor models, contrast agents, pharmacokinetic models, and MR imaging techniques. Nonetheless, MRPI data appear to at least provide useful surrogate markers of tumor biologic activity and will likely play an expanding role in tumor assessments. ¹⁷ For this reason, MRPI data are being used to investigate pathways of angiogenesis, help monitor the response of tumors to anti-angiogenic therapy, determine anti-angiogenic therapeutic dose during early phase trials, aid in the development of combination therapy regimens, provide a prognosis based on predicting tumor grade, and assist in the selection of biopsy sites within heterogeneous lesions. ^18,19 Thus, MRPI can contribute to the preclinical research, clinical research, and clinical application phases of anti-angiogenic therapy development. ³

MR perfusion imaging represents microvascular flow through tissue--ie, the capillaries and venules. When tracers that distribute extravascularly, such as gadopentetate dimeglumine (Gd-DTPA), are used, MRPI clearly depends on the inflow or outflow of blood, but overall enhancement of tumors depends on several additional variables, including vascular permeability, extravascular extracellular volume fraction, and local blood pressure or impedance. ²⁰

The perfusion or blood flow (BF) through a tissue, typically given as a volume flow rate normalized to tissue mass, is related to the intravascular blood volume and the mean transit time (MTT) of blood through the tissue, as follows:

BF [ mL · min ^-1 · g ^-1 ] · MTT [min] = BV [ mL · g ^-1 ]

(Equation 1)where BV represents the microvascular blood volume. MR perfusion imaging methods are intended to derive BF and BV using pharmacokinetic models, and thus allow estimation of MTT . It has already been shown that microvascular blood volume (BV) reflects the micro-vascular density (MVD) of the tumor, the latter being a useful parameter for determining prognosis. ^19,21,22

Blood flow can show variability among different tumor types, as well as within a particular lesion. This variability in flow is most obvious when considering tumors with arteriovenous shunting and extremely slow flow. ^23-25 Vascular permeability may also be highly variable, depending on tumor type and neovascular maturity. The spectra of both microvascular volume flow rates and permeability present challenges to modeling tumor enhancement; not only may both coexist within a single heterogeneous lesion, but both may also vary in time in response to therapy. ²⁶ For example, a tumor may exhibit paradoxical response with reduced permeability but increased blood flow, and thus demonstrate little, if any, change in overall enhancement. Therefore, it is important to derive values for individual components of the process.

MR perfusion imaging techniques utilize exogenous or endogenous tracers to reflect perfusion. The most commonly used exogenous tracers are low-molecular-weight contrast agents such as Gd-DTPA. However, macromolecular, "intravascular," and receptor/targeted exogenous agents are being investigated to provide further insight and flexibility in evaluating perfusion and MVD. ²⁷ Endogenous methods typically utilize water protons within blood as a tracer, and thus rely on mechanisms intrinsic to MR technology. Relevant examples of endogenous MRPI methods are arterial spin labeling (ASL) and BOLD imaging.

Exogenous MRI techniques

Exogenous, or dynamic contrast-enhanced (DCE) MR imaging methods rely on the pharmacokinetics of contrast agent uptake within the tumor. In preliminary studies, exogenous methods have been used to demonstrate the efficacy of anti-angiogenesis agents in a sarcoma model ²⁸ and in breast cancer. ²⁹ Imaging sequences with T1-, T2-, and T2*-weighting can be used. While low-molecular-weight contrast agents have found the most widespread clinical use, macromolecular, super-paramagnetic and targeted receptor contrast agents are being developed. Quido et al ³⁰ have compared the use of clinically approved low-molecular-weight gadolinium chelate agents with various ultra-small super-paramagnetic iron oxide (USPIO) agents using a T1-weighted technique in a colon carcinoma model and found that clinically approved gadolinium chelate agents may be recommended due to T2* effects using the USPIO agents. However, Bremer et al ³¹ found the use of USPIO agents feasible in a variety of tumors using a proton-density imaging sequence. Kobayashi et al ³² also found the use of macromolecular contrast agents to be useful in several tumor models.

T1 imaging methods can be used at lower contrast agent concentration, while T2* effects predominate at higher concentration and cause rapid signal dephasing. Figure 1 demonstrates the initial increase in signal intensity corresponding to an increase in Gd-DTPA concentration when using a T1-weighted sequence with a variety of flip angles. Signal differences between the curves depend on relative T1-weighting due to selected flip angle. This figure also demonstrates a peak and then decline in signal intensity at higher Gd-DTPA concentrations. Commonly used T1-weighted imaging methods include inversion prepped "fast spoiled gradient-recalled echo" (FSPGR) and "fast low-angle shot" (FLASH) techniques. Echoplanar imaging (EPI) methods are commonly employed to exploit T2 or T2* effects during perfusion imaging.

Many pharmacokinetic models have been proposed in order to interpret the signal intensity changes observed following contrast agent administration. It is important to note that each pharmacokinetic model has certain assumptions that may introduce errors into the analyses. Nonetheless, information can be ascertained that at least provides surrogate markers of the physiology, such as estimates of perfusion and permeability. ^17,33 For example, MRPI was recently used to correlate vascular permeability to biologic aggressiveness in ovarian carcinomas. ³⁴ Two commonly used models are briefly described to demonstrate contrast-enhancement pharmacokinetics: one- and two-compartment models.

The most basic model, the one-compartment model, assumes that the intravenous contrast agent immediately achieves equilibrium among all accessible compartments, such as the intravascular, intracellular, and extravascular extracellular compartments (Figure 2). The concentration of the contrast agent within the tumor or region-of-interest (ROI), C(t) , is then described by simple exponential decay as it is eliminated:

C(t) = C ₀ · e ^{-K _e (t)}

(Equation 2)where C ₀ is the initial concentration and K _e is the elimination rate constant. ¹⁷ This model can be used in the two theoretical extremes of first-pass contrast agent behavior: complete permeability, with 100% extraction of the tracer; or with no extraction of the tracer (ie, purely intravascular agents). These extremes would allow straightforward determination of vascular permeability ( PS * ) and BF, respectively. However, commonly used low-molecular-weight agents have variable extraction in tumors, thus complicating pharmacokinetic modeling. ^35,36 For this reason, other models such as the two-compartment model are often employed.

The two-compartment model accounts for distribution of the contrast agent within the intravascular space and within a second, readily available space, such as the extravascular extracellular space (Figure 3). In this case, the permeability represents the net exchange rate between the capillary and extravascular extracellular compartments ( PS * ). Two equations represent the time rate of change within the intravascular and extravascular extracellular compartments:

Intravascular compartment:

v * dCp = F * (C _a ­ C _p ) ­ PS * (Cp ­ C _e )

dt

(Equation 3)Extravascular extracellular compartment: v _e dCe = PS * (C _p ­ C _e )

dt

(Equation 4)where dC _p /dt and dC _e /dt represent time rate of change of concentration within the capillary and extravascular extracellular space respectively; C _a , C _p , and C _e are the arterial, capillary, and extravascular extracellular space time-dependent concentrations, respectively; F * is the perfusion ( BF ); PS * is the permeability; v * is the fractional blood volume ( BV ) within a voxel; and v _e is the fractional extravascular extracellular volume. ^17,37 Since it is not possible to spatially resolve the capillary and extravascular extracellular space using MRI, the enhancement of a single voxel or ROI overlying a tumor represents the combined effects of enhancement in both of these compartments. By measuring arterial and tumor enhancement, it is possible to solve Equations 3 and 4 to obtain the two unknowns of capillary and interstitial concentration ( C _p and C _e ).

The challenges in designing an MRPI sequence and the ability to use a particular model depend on the temporal resolution of the imaging sequence, the ability to image an adequate region or volume of interest, the signal-to-noise ratio of the data, and the ability to measure the arterial input function ( C _a ). ^17,26 For example, using the two-compartment model, DCE MR imaging methods can be used to reflect permeability ( PS * ), perfusion ( F * or BF ), and blood volume ( BV or v * ). ³⁸ However, signal-to-noise ratio limitations often make the two-compartment model difficult to fit. Also, the two-compartment model requires measurement of the arterial input function. This requirement may necessitate the acquisition of multiple slices in those circumstances in which it is not feasible to prescribe a single slice containing both the tumor and a major artery. Adequate temporal resolution is needed to resolve rapid signal changes within the artery following bolus contrast administration. If any of these conditions cannot be satisfied, a simpler model, such as the single-compartment model, must be used at the sacrifice of informational content.

Common DCE MR methods provide temporal resolution on the order of 1 to 2 seconds while imaging 1 to 2 slices. Contemporary multislice techniques can sustain satisfactory temporal resolution while improving volume coverage. For example, Hussain et al ³⁹ explored the use of a T1-weighted sensitivity encoding (SENSE) time-resolved DCE method that provided 7-second temporal resolution while covering 24 image slices. Trade-offs between adequate temporal resolution and sufficient tumor sampling must be planned carefully.

Example of exogenous contrast imaging-- The following patient with known metastatic renal cell carcinoma presented with a suspicious posterior mediastinal mass and was evaluated using a T1-weighted DCE MR technique. Three coronal images of the posterior thorax from the temporal series of images acquired following Gd-DTPA administration demonstrate successive enhancement of the pulmonary vasculature, aorta, and tumor, respectively (Figure 4). An ROI analysis was applied to the aorta and tumor to provide arterial and tumor signal intensity data (Figure 5). The signal in the aorta demonstrates the bolus technique, or arterial input function ( C _a ). The signal in the tumor increases shortly after the arterial phase, and delayed enhancement is due to vascular permeability. The signal intensity data were then curve-fit using Equations 3 and 4 to provide estimates of permeability and perfusion.

Endogenous MRI techniques

Endogenous, non­contrast-enhanced, or intrinsic MR contrast methods can reveal information about the tissue vasculature without exogenous contrast administration. While endogenous methods generally suffer from lower contrast-to-noise ratio and motion sensitivity compared with exogenous methods, advantages include insensitivity to vessel permeability, lack of IV contrast administration, ²⁰ and the ability to perform repeat studies. Although it is possible to give a second dose of gadolinium when using exogenous methods to accomplish repeat or additional studies, each additional contrast bolus greatly complicates pharmacokinetic modeling. In contradistinction, ASL has no cumulative residual effect. Thus, ASL offers the possibility for an unlimited number of sequential studies to replace poor-quality data (eg, poor breath-holding) and to assess response to a variety of physiologic challenges (eg, administering O ₂ or CO ₂ ).

Two examples of endogenous methods are ASL and BOLD imaging. ⁴⁰ Arterial spin labeling relies on movement or inflow of water molecules with a different radiofrequency (RF) pulse history from the surrounding stationary tissue. Since ASL does not utilize a bolus contrast administration, no arterial input function is required. In general, ASL methods utilize RF pulses to create signal contrast between inflowing spins and surrounding stationary tissue. There are two general classes of ASL: continuous and pulsed methods.

Continuous methods are used predominantly in cerebral imaging and involve a continuously applied spatially selective "tagging" RF pulse inferior to the slice of interest in the brain, relying on inflow of "tagged" spins into the ROI. These images are then subtracted from a control image in which no RF "tagging" pulse is applied to obtain the final images (Figure 6A). Subtraction is performed to suppress the signal of background stationary tissue because the relative signal changes due to inflow of tagged spins is small (on the order of 1% to 2%). ^20,41 It is often not possible to use continuous ASL techniques in body imaging due to high RF amplifier requirements, ²⁰ and this has prompted the use of pulsed ASL techniques for body tumor assessments. Pulsed ASL suffers from reduced signal-to-noise ratio compared with continuous techniques.

Pulsed ASL also involves subtracting a control image from a "tagged" image, with two commonly used techniques: EPISTAR (echoplanar imaging and signal targeting with alternating radiofrequency) and FAIR (flow-sensitive alternating inversion recovery). In EPISTAR, the "tagged" image is acquired following a "tagging" RF pulse over a region upstream to the imaging slice. The control image is obtained following a similar "tagging" pulse in a downstream location where "tagged" spins are unable to enter the image slice. The control is acquired in this fashion to allow subtraction of magnetization transfer effects by the initial "tagging" pulse. ⁴² Again, the final image is obtained by subtracting the control from the "tagged" slice (Figure 6B). Alternatively, FAIR ASL is performed by subtracting an image acquired following a nonselective RF tagging pulse from an image acquired following a selective inversion centered on the imaged slice (Figure 6C). ²⁰

Blood-oxygenation-level­dependent imaging methods rely on contrast related to the paramagnetic properties of deoxyhemoglobin. Deoxyhemoglobin enhances the rate of energy transitions between hydrogen nuclei ("spins") in its immediate vicinity, thus decreasing T2 (or T2*) and creating more rapid signal loss (ie, dephasing of the transverse magnetization). Intuitively, BOLD imaging is appealing for evaluating angiogenesis, since regional hypoxia is a possible stimulus of angiogenesis. However, application for evaluating neoplasms is still poorly understood. ⁴⁰ Blood-oxygenation-level­dependent contrast may depict changes in blood volume or oxygenation by comparing images acquired at baseline to images acquired during oxygen administration (hyperoxia). ⁴⁰ Tumors demonstrated high signal enhancement in response to hyperoxia. ⁴³ Blood-oxygenation-level­dependent im-aging has also been applied to map changes in blood volume fraction and vascular functionality associated with angiogenesis. ⁴⁴

Example of endogenous contrast imaging-- The following patient presented with retroperitoneal lymph-adenopathy related to metastastic renal cell carcinoma and was evaluated using a FAIR ASL technique (Figure 7). Figure 7D is an axial T2-weighted image provided to demonstrate retroperitoneal lymphadenopathy. Figures 7A and 7B represent ASL images acquired in the same axial imaging plane with nonselective and selective RF tagging pulses, respectively. The final subtracted ASL image (Figure 7C) demonstrates enhancement of multiple retroperitoneal lymph nodes, with relative suppression of surrounding background tissue.

Conclusion

Angiogenesis, a crucial process involved in tumor growth, consists of interactions among numerous stimulating and inhibiting factors to influence the proliferation of blood vessels. Although many new anti-angiogenic therapies have been developed, the evaluation of tumor response to these new therapies presents challenges. MR perfusion imaging techniques can respond to these challenges by noninvasively providing surrogate markers of tumor vascular function. The expanding role of MRPI in evaluating tumors provides opportunities to improve our understanding of certain aspects of tumor physiology and to improve the care of patients undergoing antiangiogenesis treatments.

Acknowledgments

I would like to express much appreciation to Dr. Neil Rofsky for guidance and manuscript review. I would also like to express appreciation to Dr. Cedric deBazelaire and Dr. David Alsop for both discussions regarding many of the tracer kinetic modeling and MR concepts within this paper, and for allowing use of data and example images in Figures 1, 4, 5, and 7. Thanks also to Dr. Jonathan Kruskal for recommending angiogenesis reference material, and to Michael Larson for assistance with preparing figures.