This Article Abstract Full Text (PDF) CME: Take the course for this article: Use of H215O-PET and DCE-MRI to Measure Tumor Blood Flow eLetters: Submit a response to this article Alert me when this article is cited Alert me when eLetters are posted Alert me if a correction is posted Citation Map Services E-mail this article link to a friend Similar articles in this journal Similar articles in PubMed Alert me to new issues of the journal Download to citation manager Reprints/Permissions Google Scholar Articles by de Langen, A. J. Articles by Lubberink, M. PubMed PubMed Citation Articles by de Langen, A. J. Articles by Lubberink, M.

Cancer Imaging

Use of H₂¹⁵O-PET and DCE-MRI to Measure Tumor Blood Flow

Departments of ^aPulmonary Diseases, ^cPhysics and Medical Technology, and ^dNuclear Medicine & PET Research, VU University Medical Center, Amsterdam, The Netherlands; ^bDepartment of Pulmonary Diseases, Maastricht University Hospital, Maastricht, The Netherlands

Key Words. Positron emission tomography • Magnetic resonance imaging • Blood flow • Perfusion • Angiogenesis • Oncology

Correspondence: Mark Lubberink, Ph.D., Department of Nuclear Medicine & PET Research, VU University Medical Center, P.O. Box 7057, 1007 MB Amsterdam, The Netherlands. Telephone: 31-20-4444346; Fax: 31-20-4443090; e-mail: mark.lubberink{at}vumc.nl

Received November 29, 2007; accepted for publication April 9, 2008.

Disclosure: No potential conflicts of interest were reported by the authors, planners, reviewers, or staff managers of this article.

	Learning objectives

Top Learning objectives Abstract Introduction Blood Flow Measurements Validation and Reproducibility Monitoring Response to... Monitoring Response to... Methodological Considerations-... Conclusion Author Contributions Acknowledgments References

 
After completing this course, the reader will be able to:

1. Discuss the principles of perfusion imaging with H₂¹⁵O-PET and DCE-MRI.
   
2. Compare the differences between and the limitations of the two methods.
   
3. Critically review publications on the use of both methods in monitoring response to anticancer therapy.
   

	ABSTRACT

Top Learning objectives Abstract Introduction Blood Flow Measurements Validation and Reproducibility Monitoring Response to... Monitoring Response to... Methodological Considerations-... Conclusion Author Contributions Acknowledgments References

 
Positron emission tomography (PET) with H₂¹⁵O and dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) provide noninvasive measurements of tumor blood flow. Both tools offer the ability to monitor the direct target of antiangiogenic treatment, and their use is increasingly being studied in trials evaluating such drugs. Antiangiogenic therapy offers great potential and, to an increasing extent, benefit for oncological patients in a variety of palliative and curative settings. Because this type of targeted therapy frequently results in consolidation of the tumor mass instead of regression, monitoring treatment response with the standard volumetric approach (Response Evaluation Criteria in Solid Tumors) leads to underestimation of the response rate. Monitoring direct targets of anticancer therapy might be superior to indirect size changes. In addition, measures of tumor blood flow contribute to a better understanding of tumor biology.

This review shows that DCE-MRI and H₂¹⁵O-PET provide reliable measures of tumor perfusion, provided that a certain level of standardization is applied. Heterogeneity in scan acquisition and data analysis complicates the interpretation of study results. Also, limitations inherent to both techniques must be considered when interpreting DCE-MRI and H₂¹⁵O-PET results. This review focuses on the technical and physiological aspects of both techniques and aims to provide the essential information necessary to critically evaluate the use of DCE-MRI and H₂¹⁵O-PET in an oncological setting.

	INTRODUCTION

Top Learning objectives Abstract Introduction Blood Flow Measurements Validation and Reproducibility Monitoring Response to... Monitoring Response to... Methodological Considerations-... Conclusion Author Contributions Acknowledgments References

 
Angiogenesis, the formation of new blood vessels from the endothelium of pre-existing vasculature, plays a central role in tumor growth and metastasis [1].

Once a tumor grows beyond 1–2 mm in size, passive diffusion of nutrients and oxygen is insufficient and neovascularization becomes necessary [2]. These newly formed blood vessels are highly abnormal and heterogeneous, even in tumors of equal histology and grade [3–5]. Areas of dilated, tortuous, and leaky vessels exist together with less abnormal areas [6, 7]. Interstitial hypertension, caused by vascular hyperpermeability and mechanical stress of lymphatics by tumor cell growth, can lead to blood flow stasis or even retrograde flow in tumors [8, 9]. This results in hypoxia and acidosis, which in turn is a major cause of resistance to radiation and chemotherapy and associated with a poorer prognosis [10].

Recently, drugs have been designed to specifically inhibit angiogenesis. Their effect typically results in consolidation of the tumor mass instead of regression.

Therefore, the standard volumetric approach with computed tomography (CT) (Response Evaluation Criteria in Solid Tumors) may not be sufficient to assess treatment response. Measuring change in tumor blood flow (perfusion) during antiangiogenic therapy might allow for the discrimination of responders from nonresponders irrespective of volumetric response.

Dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) and positron emission tomography using radiolabeled water (H₂¹⁵O-PET) are sensitive techniques that aim to noninvasively study and quantify the physiology of tumor microcirculation. Both techniques are based on the continuous acquisition of two-dimensional (2D) or 3D images during the uptake and clearance of an administered tracer. DCE-MRI makes use of paramagnetic tracers, mostly consisting of a low-molecular-weight gadolinium (Gd)-based agent. PET uses positron-emitting tracers, of which H₂¹⁵O can be used to study tumor blood flow. Both techniques are now being used in several phase I, II, and III clinical trials evaluating tumor vascular response to antiangiogenic drugs. However, acquisition protocols and study designs are far from uniform and may affect results. This complicates the comparison of imaging studies and possibly results in an underestimation of the abilities of DCE-MRI and H₂¹⁵O-PET. Both PET and MRI have specific advantages and disadvantages. Some limitations are inherent to the techniques, but others can be overcome by study design. These issues need to be addressed when designing a study and must be kept in mind when interpreting the results of such studies.

Mutual understanding between those using H₂¹⁵O-PET and those using DCE-MRI in oncology is generally very limited. Both techniques have great potential, but limitations still exist and must not be neglected. This is the first review to discuss both H₂¹⁵O-PET and DCE-MRI in monitoring tumor vascular response side-by-side. We focus on the technical and physiological aspects of both techniques and aim to supply the reader with essential information necessary to critically evaluate the use of DCE-MRI and H₂¹⁵O-PET in an oncological setting. First, the technical and physiological background of PET and MRI are discussed. Then, the methodology of flow measurements with both techniques is presented, and their validation and reproducibility are addressed. Finally, the application of DCE-MRI and H₂¹⁵O-PET in monitoring response to anticancer treatment and the methodological considerations influencing quantitative measurements of tumor perfusion are discussed.

	BLOOD FLOW MEASUREMENTS

Top Learning objectives Abstract Introduction Blood Flow Measurements Validation and Reproducibility Monitoring Response to... Monitoring Response to... Methodological Considerations-... Conclusion Author Contributions Acknowledgments References

 
Background

Technical Background of PET/MRI Signal
There is a fundamental difference between imaging an MRI contrast agent and imaging a PET tracer. Paramagnetic contrast agents are not by themselves detectable with MRI, but are visible because they shorten T1 and T2 relaxation times of the nearby hydrogen nuclei. In clinical practice, Gd compounds are most commonly used. The relation between signal intensity and Gd concentration is complicated. Gd concentration cannot be measured directly, because the signal enhancement (T1 relaxation) is a characteristic of the tissue being studied. There is no linear relationship between the degree of enhancement and the Gd concentration, particularly at high concentrations. Therefore, absolute quantification with DCE-MRI is difficult, and most often the proportional change in signal enhancement between the baseline and post-treatment measurements is used.

In contrast, H₂¹⁵O concentration shows a linear relation with signal intensity as measured with PET. Therefore, this technique is able to absolutely quantify tumor perfusion.

Physiological Background of PET/MRI Signal
H₂¹⁵O is a freely diffusible positron-emitting tracer. Therefore, regional uptake directly and specifically reflects tissue perfusion. Gd-based contrast agents are never freely diffusible. Therefore, the degree of signal enhancement with DCE-MRI depends on several physiological and physical characteristics, including contrast concentration, tissue perfusion, permeability, and volume of the extracellular extravascular space (EES) [11]. Thus, H₂¹⁵O-PET and DCE-MRI are both able to monitor tumor microvasculature, but the first specifically measures tissue perfusion whereas the latter measures a combination of processes, mainly perfusion and permeability.

Modeling Signal Intensity

General Model
PET and DCE-MRI perfusion measurements are both based on the Fick principle (Fick, 1870) and analyzed according to the Kety-Schmidt model [12], although the nomenclatures used in PET and DCE-MRI are slightly different. In a dynamic scan, dC_tissue(t)/dt (the change in tissue concentration of a flow tracer or contrast agent at a certain time point) is equal to K₁ (the plasma to tissue transport rate constant) times C_plasma(t) (the tracer concentration) minus k₂ (the tissue to plasma rate constant) times the tracer concentration in tissue at that point in time:

(1)

The compartment model corresponding to this equation is shown in Figure 1. K₁ equals flow F multiplied by extraction fraction E, which in its turn is a function of F and the permeability surface product PS (the latter only being relevant in the case of nonfreely diffusible tracers):

(2)

In H₂¹⁵O-PET studies, K₁ is equal to F and thus represents true blood flow, whereas in DCE-MRI studies, K₁ is a function of blood flow (F), permeability (P), and the vascular surface (S) and is also called K^trans or K^ps. The term K^trans is generally used to describe the kinetics of contrast agents with a high PS product, thus mainly representing perfusion, whereas K^ps is used for contrast agents with a lower PS product and thus represents a mixed effect of perfusion and permeability as shown in equation 2 [13, 14].

View larger version (8K): [in this window] [in a new window]   Figure 1. Single–tissue compartment model. Abbreviations: Cblood, concentration of tracer in arterial whole blood; Ctissue, concentration of tracer in tissue; K1, rate constant from plasma to tissue; k2, rate constant for tissue clearance.

For a freely diffusible tracer with 100% extraction, such as water in the range of flow values relevant for tumor imaging, equation 1 corresponds to:

(3)

Here, V_T is the tissue to plasma concentration ratio in equilibrium, which corresponds to the partition coefficient of water. V_T is equal to one if there is full exchange of tracer between blood and tissue. The solution of this equation, including a fractional blood volume v_b, is:

(4)

A nonlinear least-squares fit of this function to the measured time–activity curve in the tissue of interest yields the perfusion F and distribution volume of water V_T [15]. Therefore, this method also allows for the quantification of tissue perfusion when V_T does not equal one.

In the case of MRI, a different nomenclature is used. Low-molecular-weight Gd is not freely diffusible, but partially leaks through the vessel wall into the interstitial space and leaks more through pathological tumor vessels as a result of greater permeability. DCE-MRI applying Gd contrast agents exploits this hyperpermeable nature of tumor vessels. The same kinetic model as above can be applied to fit DCE-MRI data [16]. Tofts et al. [11] proposed the terms K^trans, k_ep, fPV, and v_e as outcome parameters derived from this single–tissue compartment model. The equation is similar to equation 3:

(5)

This equation describes the change in contrast agent concentration in the interstitium of the tumor tissue C_tissue(t), where C_plasma(t) is the concentration of contrast agent in the plasma space of the tumor tissue. K^trans is the endothelial transfer coefficient surface area product, which is the transport rate constant from plasma to tumor tissue. k_ep is the reflux coefficient or the transfer of contrast agent from tissue back to the blood. The solution of this equation is similar to the solution given for PET:

(6)

The two compartments within the tumor are the blood plasma and the EES. fPV is the fraction of plasma volume related to whole tissue volume. v_e, which equals K^trans/k_ep, is a measure for the EES fraction. Note that the definitions of F and K^trans, and V_T and v_e, are not similar because water is freely diffused in tissue, including cells, whereas Gd is only transferred into the extracellular space. Table 1 lists all the parameters involved in modeling PET and MRI signaling.

Specific Aspects of H₂¹⁵O-PET
A dynamic PET scan following a bolus injection of H₂¹⁵O consists of a series of consecutive short scans (frames), with typical durations starting with 5 seconds per frame and increasing to 30 seconds per frame (totaling 10 minutes), and starts simultaneously with the bolus injection. Although frame durations shorter than 5 seconds are possible, they would result in very noisy images because of the low number of counts per frame and this would not contribute to more reliable flow measurements. The amount of injected H₂¹⁵O is about 1 GBq and depends on the high count rate capabilities of the scanner. Prior to the emission scan, a short (several minutes) transmission scan using rotating ⁶⁸Ge rod sources, analogous to a CT scan but acquired with 511 keV gamma radiation instead of x-rays, has to be performed to correct for photon attenuation in tissue. In the case of combined PET-CT scanners, a low-dose (20–50 mAs) CT scan can be used for this purpose. A conventional PET scan after administration of 1 GBq H₂¹⁵O results in an effective radiation dose of approximately 1 mSv, or about half the average annual radiation load per person from the environment [17]. A combined PET-CT scan with 500 MBq H₂¹⁵O and a 30-mAs low-dose CT of, for example, the thorax results in a higher effective dose of 1.4 mSv. In the latter case, the lower amount of administered H₂¹⁵O is feasible because of the improved signal-to-noise ratio of state-of-the-art scanners. The spatial resolution of PET images is on the order of 6 mm in all three directions in clinical PET scanners. Because of the short radioactive half-life of ¹⁵O (2 minutes), a scan can be repeated within 10 minutes of the previous scan, which allows for the measurement of direct short-term therapy effects within one imaging session. Also, a scan can be followed by a second scan with a different tracer to study other aspects of tumor biology. Because of the short half-life, H₂¹⁵O-PET studies require the availability of an on-site cyclotron for the production of ¹⁵O.

Specific Aspects of DCE-MRI
For DCE-MRI, a clinical MRI scanner can be used with field strength usually in the range of 1–3 Tesla. The scanning protocol starts with high-resolution 3D imaging of the tumor and its environment, with both T1- and T2-weighted sequences. These images are used to delineate the tumor volume. The protocol continues with five precontrast T1-weighted measurements with different flip angles to determine the T1 relaxation time in the blood and tissue before contrast arrival. This T1 value is required for the model-based quantification [16]. Then, the contrast is given by i.v. injection. Upon injection, the dynamic acquisition starts using the same 3D T1-weighted pulse sequence and the same slice positions, but with a fixed flip angle (e.g., 35 degrees), containing 30–35 scans of about 1 second each. Thus, the temporal resolution is 1 second. The spatial resolution in-plane is about 1.5 x 1.5 mm, and through-plane is about 10 mm. The use of thick image planes, however, leads to a degradation of in-plane resolution, depending on heterogeneity in the axial direction, because each plane represents an average image of a 10-mm thick volume. For all tumors that are subject to respiratory motion, the patient is asked to hold his breath as long as possible. Renal clearance of contrast occurs with an elimination phase half-life of 100 minutes [18]. Because of the relatively slow clearance of contrast agent, DCE-MRI scans cannot be repeated within the same imaging session. Figure 2 shows images of a dynamic MR sequence following a contrast bolus injection.

View larger version (64K): [in this window] [in a new window]   Figure 2. Representative dynamic contrast-enhanced magnetic resonance images from a patient with non-small cell lung cancer. The sequential images show the tumor at different time points after contrast bolus administration. The signal intensity curve over time was measured in the tumor, the aorta, and the pulmonary artery (AP).

Parametric Images
In principle, it is possible to apply the above equations on a pixel-by-pixel basis to obtain parametric images, a graphical representation showing F, V_T, K^trans, or v_e, for each pixel [19]. Parametric images retain the original image resolution, allowing for better assessment of heterogeneity than region of interest (ROI)-based methods. Heterogeneity analyses can add to ROI analyses, and DCE-MRI seems especially promising because of its high spatial and temporal resolution. Nonlinear least-squares fitting of equations 4 or 6 on a pixel-by-pixel basis is very time-consuming, which is not a limiting factor for MRI because of the small number of image planes, but is impractical for the number of voxels of about one million in a typical clinical PET image volume. In addition, it yields noisy parametric images. Although a subregion of the total imaged volume can be selected in order to limit the computation time, faster methods are available and preferable. Basically, there are two fast options for creating parametric images. First, the differential equation 1 can be integrated on both sides, resulting in the following equation:

(7)

When evaluating this equation and plotting, for each frame, the left-hand side of the equation versus the integral ratio on the right-hand part, this equation describes a straight line, intercepting the y-axis at K₁ (F or K^trans) and the x-axis at K₁/k₂ (V_T or v_e). Using this linearization, F and V_T (or K^trans and v_e) can be obtained for each pixel using a simple and fast linear least-squares fit [20].

The second option to create parametric images is by the use of basis functions:

(8)

A set of basis functions exp(–β_it) C_plasma(t) is created by convolution of the plasma input function with a set of single exponential functions, with exponential clearance rate constants β_i, for example, between 0.01 and 1 min^–1. For each voxel, the basis function that, multiplied by K₁ (F or K^trans), best fits the measured data is selected. Each of these iterations involves a simple linear fit procedure, which is much faster than a nonlinear fit and limits the time necessary to construct parametric images from several hours to several minutes. In addition, this method reduces image noise while allowing for inclusion of a blood volume compartment and retaining the quantitative accuracy of the resulting parametric images. The basis function method also allows for fixation of V_T or v_e to further reduce noise [19, 21]. Figure 3 shows an H₂¹⁵O perfusion image calculated using the basis function method, compared with typical images of fluorine-18 fluorothymidine (¹⁸FLT), a proliferation tracer, and images of H₂¹⁵O radioactivity concentration. Figure 4 shows a DCE-MRI perfusion image calculated by nonlinear least-squares fits of equation 6 to each pixel's tissue enhancement curve.

View larger version (24K): [in this window] [in a new window]   Figure 3. Coronal images of H215O uptake (A), perfusion (B), and, for reference, fluorine-18 fluorothymidine (18F-FLT) uptake (C) of a patient with a mediastinal relapse of non-small cell lung cancer. The perfusion image was calculated using the basis function method, with the distribution volume VT fixed to 1, and corrected for both arterial and venous blood volume.

View larger version (23K): [in this window] [in a new window]   Figure 4. T1-weighted magnetic resonance image (left) and dynamic contrast-enhanced magnetic resonance imaging Ktrans perfusion image (right), also shown as overlay in the T1 image, of a patient with non-small cell lung cancer, calculated by least-squares fits to equation 6. Ktrans is the volume transfer constant between plasma and the extravascular extracellular space.

	VALIDATION AND REPRODUCIBILITY

Top Learning objectives Abstract Introduction Blood Flow Measurements Validation and Reproducibility Monitoring Response to... Monitoring Response to... Methodological Considerations-... Conclusion Author Contributions Acknowledgments References

Validation
Dynamic H₂¹⁵O-PET has been well validated in the brain and myocardium [22–27]. Some validation studies have been performed in other tissues, including the lung [28, 29], kidney [30], and skeletal muscle [31, 32]. To date, no validation studies have been performed in tumors. In theory, the model should also fit tumor perfusion measurements because H₂¹⁵O is freely diffusible throughout the body and the range of reported values (0.15–1.29 ml blood/ml tissue per minute) lies well within the range of validated perfusion values [33, 34]. On the other hand, typical vascular abnormalities in tumor tissue (vascular shunts, large vessels situated within a lesion) can result in overestimation of perfusion if models are used without correction for these effects [35, 36]. Therefore, validation studies in tumors are still needed and awaited.

Reproducibility
In brain and myocardium [37–40], as well as in some other tissues [34, 41, 42], H₂¹⁵O-PET has been shown to be reproducible with within-subject coefficients of variation (wCV) in the range of 9%–14%. The reproducibility value (defined as the range within which 95% of measurements fall) was in the range of 22%–34%. Two oncological studies have been performed [15, 34]. Although the study groups were small, reproducibility results were in the same range as those observed in the brain and myocardium, with wCV of 11% for abdominal tumors [34] and errors up to 10% reported for breast tumors [15].

DCE-MRI

Validation
The first approach for validating DCE-MRI perfusion is to verify whether an MRI-derived parameter does indeed deliver absolute perfusion measurements. No such validation study has been performed in tumors. In the human myocardium, K^trans was compared with myocardial perfusion as measured using H₂¹⁵O-PET [43]. There was a strong correlation for dipyridamole-induced flow (r = 0.70; p = .001) and a moderate correlation for myocardial perfusion reserve (r = 0.48; p = .04) between MRI and PET. In a second approach, it was validated against immunohistochemistry as the gold standard. Correlation of DCE-MRI signal enhancement with microvascular density (MVD) was found to be inconsistent [44]. One reason for this may be that tissue enhancement is not only determined by MVD but also by the status of the capillary–interstitial space barrier (vessel wall permeability), which is not accounted for by immunohistochemistry. In breast cancer lesions, tissue expression of vascular endothelial growth factor (VEGF) was closely correlated to k_ep [45]. Because VEGF expression shows a spatial association with vessel permeability [46, 47], this finding further supports the theory that enhancement of the DCE-MRI signal partially reflects vessel permeability.

Because DCE-MRI measures a combination of flow and permeability, it is difficult to validate. Attempts have shown that signal enhancement reflects vascular status, but validation studies in tumors, as have been done for the myocardium, are still awaited.

Reproducibility
Two studies have evaluated reproducibility in an oncological setting. The first evaluated the reproducibility of quantitative and semiquantitative kinetic parameters in 21 patients with solid tumors [48]. AUGC and v_e were found to be reproducible with wCVs of 12% and 9%, respectively. K^trans and k_ep showed greater variability (with wCVs of 29% and 24%, respectively). The latter are more sensitive to changes in the AIF, which could explain the lower reproducibility. Comparable results were found in a study of 11 patients with advanced cancers [49]. AUGC and K^trans showed wCVs of 16% and 19%, respectively. The reproducibility value was 30% for AUGC and 36% for K^trans. With serial imaging of patients, a decrease in K^trans of >40%–45% can be addressed as a treatment effect [48–51].

	MONITORING RESPONSE TO ANTICANCER TREATMENT

Top Learning objectives Abstract Introduction Blood Flow Measurements Validation and Reproducibility Monitoring Response to... Monitoring Response to... Methodological Considerations-... Conclusion Author Contributions Acknowledgments References

 
Predicting response to therapy as early as possible creates the opportunity to customize patient treatment. There is a need for this "customization" because response rates to intensive treatment with serious, and sometimes lethal, side effects are often low. To continue treatment with limited or no effect only provides the patient with side effects, while depriving the patient of potential benefit from other treatment. The ultimate goal is to customize treatment prior to initiation. Encouraging results have been published on gene-expression profiling, where correlations between gene expression and drug sensitivity were found [52–54]. Some success has been achieved in measuring the expression level of tumor markers to which the drug is targeted [55]. However, before these tools can be used in a clinical setting, further research is needed. Until then, predicting response early in treatment is still a major step forward when compared with size-based response evaluation after several cycles of systemic therapy.

	MONITORING RESPONSE TO ANTIANGIOGENIC THERAPY

Top Learning objectives Abstract Introduction Blood Flow Measurements Validation and Reproducibility Monitoring Response to... Monitoring Response to... Methodological Considerations-... Conclusion Author Contributions Acknowledgments References

 
Several components of tumor vasculature have been targeted by systemic drugs. The main differentiation can be made between antiangiogenic and antivascular drugs. The first inhibit neovascularization. The most striking inhibition has been accomplished by targeting VEGF, via either monoclonal antibodies (MAbs) directed against circulating VEGF or by targeting the VEGF receptor (VEGFR) with tyrosine kinase inhibitors (TKIs). Bevacizumab (a VEGF MAb) has been shown to decrease tumor perfusion, vascular permeability, MVD, interstitial fluid pressure (IFP), and circulating endothelial and progenitor cells [56]. This suggests a direct antivascular effect. Receptor TKIs generally target not only the tyrosine kinase of the VEGFR but also the tyrosine kinases of other receptors, such as the platelet-derived growth factor receptor, fibroblast growth factor receptor, and epidermal growth factor receptor (EGFR), which suggests additional effects when compared with MAbs. Examples are imatinib, sorafenib, and sunitinib. Selective blocking of EGFR is another inhibitory pathway of angiogenesis [57]. Examples of EGFR TKIs are erlotinib and gefitinib. Antivascular or vascular disruptive agents (VDAs) damage existing blood vessels, aiming at depriving the tumor from oxygen and nutrients. Combretastatin, the major VDA under investigation, is able to rapidly shut down blood vessels and can result in pronounced tumor necrosis in patients with solid tumors [58, 59].

Knowledge of the nature of a drug action, and its influence on PET and MRI derived parameters, can aid in choosing the best imaging tool and time interval to monitor response. H₂¹⁵O-PET measures perfusion and V_T. Because water is freely diffusible, a change in vascular permeability cannot be monitored accurately, although this could be measured with PET using other tracers such as [⁶⁸Ga]transferrin or [¹¹C]methylalbumin [23, 60]. A decrease in IFP (resulting from a decrease in vascular permeability or lowering of mechanical stress by reducing tumor cell density) results in an increase in perfusion. A decrease in MVD results in a decrease in perfusion. V_T (the ratio of H₂¹⁵O in tissue to that in plasma at equilibrium) is expected to be stable with antiangiogenic therapy and to decrease with antivascular therapy because of induced necrosis or increased intercapillary distance.

DCE-MRI–derived K^trans has several physiological interpretations, depending on the balance between capillary permeability and blood flow. In situations in which capillary permeability is very high, the flux of contrast agent into the EES is flow limited and K^trans will approximate tissue perfusion. On the other hand, when permeability is low and blood flow is high, K^trans gives an indication of capillary permeability. However K^trans frequently indicates a combination of flow and capillary permeability [11]. A decrease in vascular permeability and/or MVD results in a decrease in K^trans because of a lower extraction rate and lower blood vessel surface area, respectively. A decrease in IFP, however, results in an increase in K^trans because of lower interstitial resistance.

As indicated, antiangiogenic therapy promotes several morphological and functional changes in a tumor with sometimes opposite effects on MRI- and PET-derived parameters. The net effect is measured, and therefore it is of no surprise that the relation between treatment effects and imaging parameters is not straightforward. In spite of the above limitations, H₂¹⁵O-PET and DCE-MRI have shown their potential in response assessment early in anticancer treatment. In renal cell cancer, paired fluorine-18 fluorodeoxyglucose (¹⁸FDG)- and H₂¹⁵O-PET scans were performed in five patients before and after SU5416 (a VEGFR TKI) and interferon-. ¹⁸FDG and H₂¹⁵O-PET both showed a decrease in one patient with stable disease, an increase in one with progressive disease, and no change in three patients [61]. Razoxane (a cytostatic and antiangiogenic agent) was shown to reduce tumor perfusion in patients with advanced renal cell cancer [62]. Statistical significance, however, was not reached, and perfusion did not correlate with tumor progression, possibly because of the small sample size. Response monitoring to endostatin and combretastatin therapy revealed a reduction in tumor perfusion, as measured with H₂¹⁵O-PET [59, 63]. Both studies showed a dose-dependent reduction in perfusion, with a nonlinear relation in the endostatin study. Combretastatin caused a rapid perfusion reduction in solid tumors after 30 minutes, which remained significant after 24 hours. Interestingly V_T also was determined, which showed a reduction 30 minutes after injection with recovery after 24 hours. This trend might be a result of partial reversal of the vascular response, indicating that part of the tissue was targeted but still viable.

DCE-MRI has been incorporated into several phase I and II trials. PTK/ZK (a VEGFR TKI) has been tested in a phase I trial in patients with advanced cancers [50, 51, 64]. DCE-MRI was performed at baseline, day 2, and the end of each 28-day cycle. K^trans decreased in a dose-dependent way. A reduction of >40% in the baseline value at day 2 was predictive of stable disease. In a recent phase II study, patients with advanced breast cancer were treated with bevacizumab monotherapy for one cycle, followed by six cycles of bevacizumab, doxorubicin, and docetaxel. Significant reductions in K^trans, k_ep, and v_e were seen after one cycle, which persisted when cytotoxic chemotherapy was added. The decreases in K^trans, k_ep, and v_e were correlated with increased apoptosis and a reduction in VEGFR expression [65]. Several targeted agents have been shown to decrease DCE-MRI–derived parameters in a dose-dependent manner, making DCE-MRI a useful indicator in drug pharmacology [66, 67]. A recent study of DCE-MRI in patients with progressive multiple myeloma treated with thalidomide found amplitude A, which is proportional to the relative signal enhancement, to be a good prognostic indicator of progression-free survival [68].

Most PET and MRI results are derived from small studies. Together with heterogeneity in acquisition protocols and study design, these factors might limit their value. Therefore, prospective phase III trials with predefined cutoff values for PET and MRI results are awaited.

Potential Synergistic Value of DCE-MRI and H₂¹⁵O-PET
Because PET measures true perfusion and MRI measures a combination of perfusion and vascular permeability, the two modalities can complement each other. In theory, both perfusion and permeability can be isolated when F and K^trans are known. An alternative is the use of a PET tracer such as [⁶⁸Ga]transferrin or [¹¹C]methylalbumin to measure permeability directly. So far, no oncological study has been performed that incorporated both DCE-MRI and H₂¹⁵O-PET.

Combining Information on Tumor Vasculature and Metabolism
Tumor metabolism can be measured with ¹⁸FDG-PET. ¹⁸FDG uptake is the product of perfusion and extraction. It can be hypothesized that poorly perfused and hypometabolic areas reflect necrosis, while poorly perfused and hypermetabolic areas reflect hypoxia (because of anaerobic glucose hypermetabolism under hypoxic circumstances). It is possible to measure oxygen consumption directly using ¹⁵O₂ and PET. Knowledge of the regional tumor microenvironment can supply relevant information on mechanisms of treatment failure, but can also aid in treatment planning (e.g., radiotherapy). Recently, the synergistic value of H₂¹⁵O-PET and ¹⁸FDG-PET was shown in breast cancer, where a pretherapeutic low ratio of ¹⁸FDG uptake to perfusion was the best predictor of complete response to neoadjuvant chemotherapy and also predictive of disease-free survival [69]. ¹⁸FDG metabolism and perfusion were positively correlated, although highly variable. Zasadny et al. [70] also found a positive correlation between ¹⁸FDG metabolism and perfusion, with the additional finding of a higher slope of the curve at lower flow rates, possibly indicating hypoxia in areas with low perfusion and high metabolism. In contrast, ¹⁸FDG metabolism was found to be negatively correlated with perfusion in non-small cell lung cancer (NSCLC), liver tumors, and head and neck cancer. Perfusion in the latter two was assessed with contrast-enhanced CT [71–73]. In NSCLC and head and neck cancer, the ratio was found to positively correlate with tumor size, suggesting the presence of hypoxia in larger tumors where tumor growth exceeds angiogenesis [71, 73]. In contrast, a high tumor perfusion rate predicted poor response to radiotherapy in head and neck cancer [74]. This reflects the complexity of tumor biology. High tumor blood flow does not immediately indicate sufficient oxygenation and thus favorable radiotherapy outcome. Hypoxia stimulates angiogenesis via upregulation of specific transcription and growth factors, like VEGF [75]. This results in a more complex, but abnormal, chaotic vascular network with increased blood flow, but inadequate oxygen and nutrient supply. Thus, hypoxic areas can exist irrespective of high blood flow, a finding confirmed in a PET study in brain tumors using fluorine-18 fluoromisonidazole (a hypoxia tracer) and H₂¹⁵O [76].

Timing of Response Evaluation
Jain [77, 78] introduced the term "normalization" to describe the (transient) reversal of vessel abnormalities his group observed after antiangiogenic treatment. The time span of normalization, the "normalization window," describes the period in which abnormal vessels either regress or normalize. In this period, the nutrient, drug, and oxygen supply is thought to be improved [77, 78]. Eventually, all tumors show resistance, after which vessels become abnormal again. The time span of this window is unknown for most drugs and, in addition to drug characteristics, may depend on the tumor type and possibly on individual tumor and patient characteristics. Because of these time-dependent tumor changes, it seems extremely important to know when to scan in order to adequately measure treatment effect.

In most protocols, time to response evaluation is based on CT response protocols instead of considerations of tumor biology. This might be one of the reasons why H₂¹⁵O-PET and DCE-MRI failed to demonstrate a correlation with outcome in some trials.

The reported normalization windows after treatment with VEGF TKIs or MAbs have been heterogeneous [79, 80], which makes designing future trials difficult but challenging. Serial measurements can aid in the definition of this time window for new drugs [79]. Therefore, if financially and logistically possible, serial PET and MRI measurements should be planned for new drug trials in order to assess the best time point for response monitoring after antiangiogenic treatment. Although most patients in such trials have a limited life expectancy, this strategy can result in a high cumulative radiation dose, especially with combined PET-CT scanners.

	METHODOLOGICAL CONSIDERATIONS—FACTORS INFLUENCING QUANTITATIVE MEASUREMENTS OF TUMOR PERFUSION AND PERMEABILITY

Top Learning objectives Abstract Introduction Blood Flow Measurements Validation and Reproducibility Monitoring Response to... Monitoring Response to... Methodological Considerations-... Conclusion Author Contributions Acknowledgments References

 
AIF Definition
Independently of the parameter used, several considerations have to be made about the ROI and AIF. The latter describes the concentration of MRI contrast or PET tracer in plasma over time, which has a great influence on measurements of (semi)quantitative parameters. It is generally accepted that the AIF varies between patients and between visits within each patient because of variations in cardiac output, renal function, and vascular tone. In principle, the AIF should be accurately measured for each scan. With PET, accurate measurement of the AIF can be obtained by arterial blood sampling with an online system, measuring arterial radioactivity concentration independently of tumor site. Because arterial sampling is invasive, an image-derived AIF may be more convenient. However, in many cases it is not possible to measure an image-derived AIF because of the absence of a suitable artery within the imaging field of view.

With MRI, image-derived measurement of the AIF is the only option, because measurement of the contrast agent concentration by arterial sampling is not possible. AIF measurements have shown that DCE-MRI reproducibility is worse with suboptimal AIF definition than with using a standardized AIF [49, 81]. In addition, the relation between tracer concentration and signal intensity becomes progressively nonlinear with increasing tracer concentration. This is especially the case in the feeding artery during the first passage of the administered bolus. Therefore, it has been proposed that the AIF should be measured with a small prebolus, and then the true AIF for the main bolus can be reconstructed from this [82]. Several other methods have been developed to meet this problem. Most of them are based on reference region models or a functional form of a population-derived AIF [81, 83]. Whichever technique is being used, it has a great impact on the results.

Tissue ROI Definition
The tissue ROI should be a well-defined part of the tumor, consistent throughout follow-up and preferably throughout the whole study to avoid sampling errors. This could be the entire tumor or a subregion. ROI can be defined on CT, MRI, PET emission, or PET transmission images. DCE-MRI obviously uses the high-resolution MR images, but PET has no well-defined anatomic landmarks available. CT can be used, but coregistration incorporates spatial error. Integrated PET-CT overcomes much of this problem, although respiratory motion still challenges accurate coregistration in the thorax and upper abdomen. Definition of anatomic ROIs using CT and MRI has its own disadvantages. Metabolically inactive areas (reflecting necrosis) are difficult to identify and might be incorporated, causing underestimation of the perfusion in viable tumor tissue. In addition, atelectasis and adjacent infectious consolidation complicate ROI definition.

With PET, ROI can be defined either on the H₂¹⁵O image or on a subsequently performed PET scan with a tissue-accumulating tracer. Because of the limited spatial resolution of PET and the "noisy" data produced by a diffusible tracer like H₂¹⁵O, ROI definition on an H₂¹⁵O-PET image is difficult (Fig. 3A). Peritumoral areas of high flow (resulting from infection or adjacent large vessels) and small or highly irregular tumors complicate delineation. Parametric images, depicting flow instead of radioactivity concentration, show better tumor-to-background contrast [19], but are still in need of improvement (Figs. 3B and 4). In the case of heterogeneity, areas of low or absent flow are not accounted for (not visible) and only perfused areas are selected. Therefore, the ROIs defined on parametric images probably do not reflect the true tumor mass. Adding a tissue-accumulating tracer (allowing for better tumor-to-background contrast, e.g., ¹⁸FDG or ¹⁸FLT) to the scan protocol creates the opportunity to define an ROI reflecting metabolically active tissue, which can then be copied to the perfusion image (Fig. 3).

Preferably, a 3D ROI is constructed to ensure all tumor tissue is incorporated. When 2D regions are used, it is recommended to have a minimum of three sections through the tumor in order to account for spatial heterogeneity within a lesion [84]. ROI definition is not easy and straightforward, but should be based on the desired information and knowledge of treatment effects, and, most importantly, it should be performed in a standardized, reproducible manner [85].

Organ and Tumor Position
Quantification of tumor perfusion is more complicated in the lung and upper abdominal lesions. In the liver, AIF definition is difficult because of the dual hepatic blood supply (portal vein and hepatic artery). The hepatic artery accounts for the greatest part of the blood supply to liver tumors, but the portal vein contributes to a small amount in small metastases and at the surface of large metastases [86–89]. A model including the dual hepatic blood supply, addressing these issues, has been suggested [90, 91].

Tumor motion influences delineation and quantification. Thoracic lesions at the base of the lung and close to the thoracic wall, as well as upper abdominal lesions, are most vulnerable to respiratory motion [92]. A dynamic DCE-MRI sequence can be performed within a single breathhold, thereby avoiding substantial motion error. PET, with its lower temporal resolution, is especially affected. Pulmonary gated PET, using only the data required during a certain phase of the respiratory motion cycle, is a promising tool to compensate for this type of error [93, 94], although difficult to implement for dynamic H₂¹⁵O scans because of limited count rates.

Spatial and Temporal Resolution
Heterogeneity occurs at the microscopic level. DCE-MRI with its high spatial resolution, typically 1.5 x 1.5 mm in-plane and 10 mm through-plane, offers detailed pictures, whereas PET has a spatial resolution of approximately 6 mm, resulting in an averaged perfusion picture [95]. MRI also has better temporal resolution, which is on the order of 0.1 seconds for a 2D acquisition and 1 second for a 3D acquisition. This ensures that the upslope of the contrast enhancement can be measured with adequate temporal resolution, mostly within one breathhold. This allows for detailed heterogeneity analyses that might add to perfusion measures. Tumor vasculature is highly heterogeneous throughout lesions, both in place and time. Microvascular blood flow fluctuates markedly within seconds, even without treatment [96, 97]. Excellent spatial and temporal resolution only give a quick snapshot and might not be representative of the overall vascular status within a given tumor region. Therefore, it can be hypothesized that an average picture (time and place) might provide more practical information than a detailed one.

Partial Volume Effects
Partial volume effects typically occur in small lesions. This effect refers to two distinct phenomena that negatively influence quantitative measurements, image blurring and averaging signal at the voxel level [98]. Because of the finite resolution, the activity of a small source is projected as a larger but less active source (part of the signal spills out). Averaging at the voxel level occurs with all image modalities and is explained by the fact that every image is built up by pixels (2D) or voxels (volumetric pixels, 3D). Lesion boundaries do not conform themselves to this tight framework and thus voxels can exhibit both tumor and adjacent tissue, resulting in an over- or underestimation of the true pixel/voxel activity.

In this way, spill in can also be explained. Nearby large vessels, obviously with high blood flow, can cause spill in of activity that can be accounted for in compartment models (equation 4). In a similar way, hypoperfused lesions surrounded by physiologically highly perfused tissue (e.g., liver or spleen) are also affected.

Response Definition in Clinical Trials
Response evaluation is usually done by calculating the percentage change between baseline and post-treatment values. This delta function substantially depends on the value of the divisor. With a small divisor, a small difference results in an unrealistic response rate. At the other end of the spectrum, a response could be missed because of a relatively large divisor. Thus, simply calculating the percentage change might not be useful for very small or large tumors. Reproducibility studies can shed light on this subject [99]. Quantification is highly dependent on both hardware and software. Image resolution, reconstruction methods, ROI strategy, and quality control (i.e., calibration) all account for fluctuations in results. Differences of up to 40% are described in the literature for standardized uptake values (SUVs) with ¹⁸FDG-PET [100]. SUV measurements, however, require crosscalibration of a dose calibrator and the PET scanner, whereas flow measurements using H₂¹⁵O do not, removing uncertainties resulting from calibration errors in the case of flow studies. In general, most errors can be overcome when only one scanner is used, together with uniform reconstruction methods and data analysis. However, today most clinical studies are multicenter designed. If absolute values are necessary for within-study evaluation and when general recommendations for cutoff values are desired, standardization is required. Until that time, it is recommended that at least each patient be scanned on the same scanner pre- and post-treatment with equal methods of scan acquisition, data reconstruction, and analysis. In this way, the proportional change in flow values can be used safely.

	CONCLUSION

Top Learning objectives Abstract Introduction Blood Flow Measurements Validation and Reproducibility Monitoring Response to... Monitoring Response to... Methodological Considerations-... Conclusion Author Contributions Acknowledgments References

 
DCE-MRI and H₂¹⁵O-PET are able to reliably measure tumor perfusion, provided that a certain level of standardization of both the scan acquisition and data analysis is applied. Their value and potential have been shown in several phase I, II, and III drug trials. Both tools can be used to (a) assess anticancer drug activity early in drug development, (b) evaluate response to anticancer drugs early in treatment, and (c) study biological processes in tumors, which increases knowledge on the tumor microenvironment and might lead to a better understanding of drug failure patterns.

There are still some issues that need to be reconsidered to allow for routine clinical use of DCE-MRI and H₂¹⁵O-PET. H₂¹⁵O-PET studies require the availability of an on-site cyclotron, while DCE-MRI data analysis is not straightforward, and can be complex and time-consuming. In addition, the predictive value of H₂¹⁵O-PET and DCE-MRI has to be confirmed in phase III trials with predefined cutoff values for response definition before recommendations on clinical use can be made.

However, their noninvasive nature in combination with promising results must be regarded as a stimulus to incorporate DCE-MRI and H₂¹⁵O-PET in future trials. This will contribute to a better understanding of the value of both techniques and might one day serve patients in the clinical setting.

	AUTHOR CONTRIBUTIONS

Top Learning objectives Abstract Introduction Blood Flow Measurements Validation and Reproducibility Monitoring Response to... Monitoring Response to... Methodological Considerations-... Conclusion Author Contributions Acknowledgments References

 
Conception/design: Adrianus J. de Langen, J. Tim Marcus, Mark Lubberink

Provision of study materials or patients: Adrianus J. de Langen, Vivian E. M. van den Boogaart, Mark Lubberink

Manuscript writing: Adrianus J. de Langen, Vivian E. M. van den Boogaart, J. Tim Marcus, Mark Lubberink

Final approval of manuscript: Adrianus J. de Langen, Vivian E. M. van den Boogaart, J. Tim Marcus, Mark Lubberink

	ACKNOWLEDGMENTS

Top Learning objectives Abstract Introduction Blood Flow Measurements Validation and Reproducibility Monitoring Response to... Monitoring Response to... Methodological Considerations-... Conclusion Author Contributions Acknowledgments References

 
The authors are grateful to Dr. E. F. Smit and Dr. A. C. Dingemans for their comments and to Dr. A. A. Lammertsma for critically reviewing the manuscript.

	REFERENCES

Top Learning objectives Abstract Introduction Blood Flow Measurements Validation and Reproducibility Monitoring Response to... Monitoring Response to... Methodological Considerations-... Conclusion Author Contributions Acknowledgments References

 

1. Folkman J. Role of angiogenesis in tumor growth and metastasis. Semin Oncol 2002;29(suppl 16):15–18.[Medline]
2. Folkman J. Tumor angiogenesis: Therapeutic implications. N Engl J Med 1971;285:1182–1186.[Medline]
3. Mineura K, Sasajima T, Itoh Y et al. Blood flow and metabolism of central neurocytoma: A positron emission tomography study. Cancer 1995;76:1224–1232.[CrossRef][Medline]
4. Mineura K, Shioya H, Kowada M et al. Blood flow and metabolism of oligodendrogliomas: A positron emission tomography study with kinetic analysis of 18F-fluorodeoxyglucose. J Neurooncol 1999;43:49–57.[CrossRef][Medline]
5. Lassen U, Andersen P, Daugaard G et al. Metabolic and hemodynamic evaluation of brain metastases from small cell lung cancer with positron emission tomography. Clin Cancer Res 1998;4:2591–2597.[Abstract]
6. Jain RK. Determinants of tumor blood flow: A review. Cancer Res 1988;48:2641–2658.[Abstract/Free Full Text]
7. Hobbs SK, Monsky WL, Yuan F et al. Regulation of transport pathways in tumor vessels: Role of tumor type and microenvironment. Proc Natl Acad Sci U S A 1998;95:4607–4612.[Abstract/Free Full Text]
8. Padera TP, Stoll BR, Tooredman JB et al. Pathology: Cancer cells compress intratumour vessels. Nature 2004;427:695.[CrossRef][Medline]
9. Jain RK, Tong RT, Munn LL. Effect of vascular normalization by antiangiogenic therapy on interstitial hypertension, peritumor edema, and lymphatic metastasis: Insights from a mathematical model. Cancer Res 2007;67:2729–2735.[Abstract/Free Full Text]
10. Fukumura D, Jain RK. Tumor microenvironment abnormalities: Causes, consequences, and strategies to normalize. J Cell Biochem 2007;101:937–949.[CrossRef][Medline]
11. Tofts PS, Brix G, Buckley DL et al. Estimating kinetic parameters from dynamic contrast-enhanced T(1)-weighted MRI of a diffusable tracer: Standardized quantities and symbols. J Magn Reson Imaging 1999;10:223–232.[CrossRef][Medline]
12. Kety SS. Measurement of local contribution within the brain by means of inert, diffusible tracers; examination of the theory, assumptions and possible sources of error. Acta Neurol Scand Suppl 1965;14:20–23.[Medline]
13. Padhani AR, Hayes C, Landau S et al. Reproducibility of quantitative dynamic MRI of normal human tissues. NMR Biomed 2002;15:143–153.[CrossRef][Medline]
14. Taylor JS, Tofts PS, Port R et al. MR imaging of tumor microcirculation: Promise for the new millennium. J Magn Reson Imaging 1999;10:903–907.[CrossRef][Medline]
15. Wilson CB, Lammertsma AA, McKenzie CG et al. Measurements of blood flow and exchanging water space in breast tumors using positron emission tomography: A rapid and noninvasive dynamic method. Cancer Res 1992;52:1592–1597.[Abstract/Free Full Text]
16. de Lussanet QG, Backes WH, Griffioen AW et al. Gadopentetate dimeglumine versus ultrasmall superparamagnetic iron oxide for dynamic contrast-enhanced MR imaging of tumor angiogenesis in human colon carcinoma in mice. Radiology 2003;229:429–438.[Abstract/Free Full Text]
17. Smith T, Tong C, Lammertsma AA et al. Dosimetry of intravenously administered oxygen-15 labelled water in man: A model based on experimental human data from 21 subjects. Eur J Nucl Med 1994;21:1126–1134.[Medline]
18. Brix G, Semmler W, Port R et al. Pharmacokinetic parameters in CNS Gd-DTPA enhanced MR imaging. J Comput Assist Tomogr 1991;15:621–628.[Medline]
19. Lodge MA, Carson RE, Carrasquillo JA et al. Parametric images of blood flow in oncology PET studies using [¹⁵O]water. J Nucl Med 2000;41:1784–1792.[Abstract/Free Full Text]
20. Blomqvist G. On the construction of functional maps in positron emission tomography. J Cereb Blood Flow Metab 1984;4:629–632.[Medline]
21. Boellaard R, Knaapen P, Rijbroek A et al. Evaluation of basis function and linear least squares methods for generating parametric blood flow images using ¹⁵O-water and positron emission tomography. Mol Imaging Biol 2005;7:273–285.[CrossRef][Medline]
22. Raichle ME, Martin WR, Herscovitch P et al. Brain blood flow measured with intravenous H2(15)O. II. Implementation and validation. J Nucl Med 1983;24:790–798.[Abstract/Free Full Text]
23. Brooks DJ, Frackowiak RS, Lammertsma AA et al. A comparison between regional cerebral blood flow measurements obtained in human subjects using 11C-methylalbumin microspheres, the C¹⁵O₂ steady-state method, and positron emission tomography. Acta Neurol Scand 1986;73:415–422.[Medline]
24. Bergmann SR, Fox KA, Rand AL et al. Quantification of regional myocardial blood flow in vivo with H₂¹⁵O. Circulation 1984;70:724–733.[Abstract/Free Full Text]
25. Araujo LI, Lammertsma AA, Rhodes CG et al. Noninvasive quantification of regional myocardial blood flow in coronary artery disease with oxygen-15-labeled carbon dioxide inhalation and positron emission tomography. Circulation 1991;83:875–885.[Abstract/Free Full Text]
26. Schafers KP, Spinks TJ, Camici PG et al. Absolute quantification of myocardial blood flow with H(2)(15)O and 3-dimensional PET: An experimental validation. J Nucl Med 2002;43:1031–1040.[Abstract/Free Full Text]
27. Rimoldi O, Schafers KP, Boellaard R et al. Quantification of subendocardial and subepicardial blood flow using ¹⁵O-labeled water and PET: Experimental validation. J Nucl Med 2006;47:163–172.[Abstract/Free Full Text]
28. Richard JC, Janier M, Decailliot F et al. Comparison of PET with radioactive microspheres to assess pulmonary blood flow. J Nucl Med 2002;43:1063–1071.[Abstract/Free Full Text]
29. Mintun MA, Ter Pogossian MM, Green MA et al. Quantitative measurement of regional pulmonary blood flow with positron emission tomography. J Appl Physiol 1986;60:317–326.[Abstract/Free Full Text]
30. Juillard L, Janier MF, Fouque D et al. Renal blood flow measurement by positron emission tomography using ¹⁵O-labeled water. Kidney Int 2000;57:2511–2518.[CrossRef][Medline]
31. Fischman AJ, Hsu H, Carter EA et al. Regional measurement of canine skeletal muscle blood flow by positron emission tomography with H2(15)O. J Appl Physiol 2002;92:1709–1716.[Abstract/Free Full Text]
32. Ruotsalainen U, Raitakari M, Nuutila P et al. Quantitative blood flow measurement of skeletal muscle using oxygen-15-water and PET. J Nucl Med 1997;38:314–319.[Abstract/Free Full Text]
33. Hoekstra CJ, Stroobants SG, Hoekstra OS et al. Measurement of perfusion in stage IIIA-N2 non-small cell lung cancer using H(2)(15)O and positron emission tomography. Clin Cancer Res 2002;8:2109–2115.[Abstract/Free Full Text]
34. Wells P, Jones T, Price P. Assessment of inter- and intrapatient variability in C¹⁵O₂ positron emission tomography measurements of blood flow in patients with intra-abdominal cancers. Clin Cancer Res 2003;9:6350–6356.[Abstract/Free Full Text]
35. Eskey CJ, Wolmark N, McDowell CL et al. Residence time distributions of various tracers in tumors: Implications for drug delivery and blood flow measurement. J Natl Cancer Inst 1994;86:293–299.[Abstract/Free Full Text]
36. Ponto LL, Madsen MT, Hichwa RD et al. Assessment of blood flow in solid tumors using PET. Clin Positron Imaging 1998;1:117–121.[CrossRef][Medline]
37. Lammertsma AA, Jones T. Low oxygen extraction fraction in tumours measured with the oxygen-15 steady state technique: Effect of tissue heterogeneity. Br J Radiol 1992;65:697–700.[Abstract]
38. Coles JP, Fryer TD, Bradley PG et al. Intersubject variability and reproducibility of ¹⁵O PET studies. J Cereb Blood Flow Metab 2006;26:48–57.[CrossRef][Medline]
39. Carroll TJ, Teneggi V, Jobin M et al. Absolute