Downloaded from www.ajronline.org by 117.253.209.72 on 11/09/15 from IP address 117.253.209.72. Copyright ARRS. For personal use only; all rights reserved
259
Review
.
,.
..
.
Edema and Tumor Perfusion: Quantitative 1H MR Imaging A. Grant
perfusion is of it has a therapy. Yet determining fused is difficuft without the variables that affect Quantitative MR Imaging of characterizing tumor because
perfusion
Characterization
within
central importance to the clinical oncoldirect effect on the success of cancer whether a tumor is well or poorly perthe use of invasive techniques, because tumor perfusion are poorly understood. of tumor edema may provide a means perfusion, of studying heterogeneity of
the tumor
mass,
or of monitoring
changes
in
after therapy. A combination of factors often resufts in production of a large amount of edema within cranial or extracranial tumors. Any tumor that is encapsulated, whether by a fibrous tumor capsule or by a structure such as the cranium, will have an elevation in interstitial fluid pressure because dissipation of fluid is hindered. Elevated pressure of interstitial fluid acts to occlude tumor capillaries, so edema can cause a striking reduction of tumor perfusion. Because MR imaging can potentially be used for quantitative imaging of tumor edema, it may provide a means of indirectly measuring tumor perfusion. A review of the Ifterature suggests that diffusion-weighted MR imaging may be befter than TI- or T2-weighted MR imaging for tumor
perfusion
quantitative
imaging
diftusion-weighted rather correlated
I hypothesize with
tumor
of tumor edema. I do not propose that imaging can measure perfusion directly; that a diffusion-weighted image can be edema.
Because
edema
indirectly
regulates
of interstitial fluid pressure, I propose an indirect correlation between the diffusion-weighted image and regional tissue perfusion. If the relationship between tumor perfusion and the pharmacokinetics of chemotherapeutic agents is better understood, MR imaging of tumor edema may even aid in predicting the delivery of drugs to a tumor. perfusion
by
Steen1
Tumor
ogist
Article
through
the mechanism
Received July 5, 1991 ; accepted after revision October 4. 1991. This work was supported in part by Biomedical Research Support Grant AA-05432 I Departments of Radiology (RC-05) and Bioengineering, University of Washington Steen. AJR 158:259-264,
February
1992 0361-803X/92/1582-0259
C American
Roentgen
Peritumoral edema is a widely recognized problem in the treatment of cranial malignancies [1], but it is increasingly apparent that edema is also a problem in extracranial tumors [2]. Edema acts to elevate interstitial pressure, which can produce a range of effects: Tumor perfusion may become inadequate and unpredictable; delivery of therapeutic agents to the tumor may be reduced; hypoxic foci of cells, which are inherently radioresistant [2], may develop; and mass effects of edema
in the cranium
can cause
neurologic
impairment
[1].
If a technique could be developed for quantitative MR imaging of tumor edema, it might enable researchers to characterize the relationship between edema and tumor perfusion by using a noninvasive technique.
Potential
Mechanisms
Generating
Edema
in Tumors
A variety of factors result in substantial edema within tumors at any site. Transcapillary filtration forces are slightly greater than absorptive forces in most tissues, so circulatory fluids tnansude into the interstitial space at a low rate [3]. The flux of water and protein from blood to interstitium is normally balanced by the flow oflymph, which carries water and protein back to the bloodstream. However, tumors lack a lymphatic system, so the normal pathway of drainage is absent [2, 4]. This tendency toward vasogenic edema in tumors is compounded by several other factors: (1) Tumor capillaries are often more permeable than normal capillaries [2, 5-7]. For
and American Cancer Society Institutional School of Medicine, Seattle, WA 98195. Ray Society
Grant IN-26-31. Address reprint requests
to A. G.
260
STEEN
example,
the capillary
adenocarcinoma
was
Downloaded from www.ajronline.org by 117.253.209.72 on 11/09/15 from IP address 117.253.209.72. Copyright ARRS. For personal use only; all rights reserved
normal tissue,
filtration
coefficient
1 0 to 1 000
times
and within
of a mammary higher
the range reported
than
that
of
for filtration
in
glomerular capillaries [7]. (2) Tumor capillaries may be temporarily occluded [8, 9], as a result of cell growth in a confined space [1 0]. This leads to an elevation of blood pressure on
the arterial side of the capillary tion [1 1]. (3) Overproduction polysaccharides by tumor
bed and increased
transuda-
ofextracellular collagen or mucocells can lead to accumulation of
polyanionic macromolecules in the matrix that become hydrated because of Donnan forces [1 2]. (4) Disaggregation or digestion of extracellular matrix by tumor cells [1 3] can increase the osmotic effect of matrix macromolecules [1 4]. (5) Increased intracapillary diffusion distances in tumors can lead to a compromised bioenergetic state and cellular hypoxia [15], so cells may be unable to maintain normal osmotic gradients and may release osmotically active ions such as K and (6) Cell death and lysis may result in release of many proteins and other osmotically active molecules normally sequestered in the cell [16, 17]. A combination ofthese factors is likely to lead to substantial edema within a tumor. For example, in a patient with a brain metastasis, it was estimated that fluid was produced within the tumor at a rate of 3.9 ml/hr, which led to an edematous volume in the brain of 87.3 cm3 [1 ]. Because edema production rate varies
from one tumor
to another,
whereas
resorption
rate by normal brain is probably relatively constant, ume of peritumoral edema in brain probably reflects of fluid production [1]. Effect
of Edema Normal Brain
on Perfusion
and
Bioenergetics
the volthe rate
in
of volume
equilibration
that,
in a constrained
volume
such as the cranium, must be compensated for by a reduction in other fluid and blood compartments [1 4]. When
this
compensation
is inadequate,
tissue
pressure
can
in-
crease, and studies of brain edema have shown a decrease of cerebral blood flow in edematous areas [1 8]. Because flow reduction can occur even with a relatively mild increase in intracranial
pressure
(ICP) and with adequate
tissue
perfusion
pressure, it is likely that edema directly compresses capillaries [14]. The effect of edema on tissue perfusion and bioenergetics has been studied in detail in brain, where cranial edema is associated with stroke, hypertension, trauma, inflammation, AIDS, and intracranial tumors. For example, 31P MR spectroscopy was used in a standard model of cranial hypertension to examine the effect of pressure ischemia on brain bioenergetics [1 9]. Graded increases in CSF pressure were used to induce pressure ischemia while 31P MR spectroscopy was
performed.
A spectrum
mm Hg, then infusing fluid
was acquired
at a resting
ICP of 0-5
ICP was raised in increments of 20 mm Hg by from a cisternal catheter. Cerebral perfusion
pressure (CPP) of blood declined as ICP was raised (CPP = mean arterial blood pressure ICP). Phosphocreatine was -
February
1992
generally maintained until the CPP reached 29 mm Hg and then declined rapidly to undetectable levels as CPP reached 0. Loss of adenosine tniphosphate began only when phosphocreatine was entirely exhausted [1 9]. This demonstrates that tissue pressure induced by edema can have a profound effect on brain
bioenergetics.
Interstitial
ogous to ICP and should perfusion. Edema
and Interstitial
In any tumor
that
fluid
have
pressure
similar
Fluid Pressure
(IFP) is anal-
effects
on tumor
in Tumors
is well encapsulated
(e.g.,
by a fibrous
capsule or within the cranium), and in which interstitial fluid movement is hindered, tumor edema can lead to a striking increase in interstitial pressure [1 4]. This can cause decreased tumor
perfusion
by causing
vascular
occlusion,
suggesting
that quantitative imaging of edema may provide direct insight into tumor perfusion. In general, IFP in tumors is significantly higher than in normal tissues, even in tumors not constrained within the cranium [2]. IFP was measured in normal rat tissue, in two tumors implanted in ovarian tissue (“tissue-isolated tumors”),
and in a tumor implanted subcutaneously [1 0]. IFP of normal muscle varied between 0.0 and 0.4 mm Hg, whereas IFP of dead tissue was 0 mm Hg. Mean IFP of tumors was 8.5 (±5.1) mm Hg, and ranged from 2.7 mm Hg in a 0.5i tumor to 22.8 mm Hg in a 5.0-g tumor.
sharp increase
in interstitial
lying the tumor and Within tissue-isolated
Vascular compression by edema fluid can have a profound effect on tissue blood flow and bioenergetics. Edema is a disturbance
AJA:158,
In subcutaneous
pressure
tumors,
the
began in the skin over-
plateaued at the tumors, pressures
skin-tumor interface. reached plateau val-
ues at a distance of 0.2-1 .1 mm into the tumor. IFP was elevated in 86% of all tumors studied. It has been estimated that an IFP of 10-i 5 mm Hg or more will cause vascular occlusion [10]. Preliminary data from human primary colorectal and breast carcinomas, and from metastases to the lung, liver, and lymph nodes, indicate that IFP in human tumors is significantly higher than
in normal
tissue
[20],
even
though
these
tumors
were
probably not well encapsulated. Mean human tumor IFP was 23.9 (±1 2.3) mm Hg in a series of 16 various tumors. Human tumor IFP ranged from 4 mm Hg in a small breast metastasis (1 9 cm3) to 50 mm Hg in a large lymph node metastasis (133 cm3) [20]. Elevated
IFP in tumors
is expected
to result
in a general
reduction in tumor blood flow and development of necrotic regions in the tumor [2, 6, 10]. In addition, elevated IFP should cause poor delivery of therapeutic agents, including hydrophilic
chemotherapeutic
biological
response
agents,
modifiers
therapeutic agents to tumors to the core of the tumor,
extravasation a convective
monoclonal
antibodies,
[2, 10]. IFP restricts
delivery
and
of
by reducing the access of blood reducing the driving force for
of fluid in the core of the tumor, and generating net flux of fluid toward the periphery of the tumor
[101.
Elevated reduction
IFP should
IFP in tumors is also expected to result in a in tumor bioenergetic state, whereas reduction of result in improved tumor bioenergetic state. Rebioenergetics, obtained by 31P MR spec-
cent data on tumor
Downloaded from www.ajronline.org by 117.253.209.72 on 11/09/15 from IP address 117.253.209.72. Copyright ARRS. For personal use only; all rights reserved
AJR:158,
February
1H MR OF EDEMA
1992
AND
troscopy of tumors, have shown that tumors typically show an increase in high-energy phosphates (adenosine tnphosphate and phosphocreatine) after therapy with cytotoxic drugs [21]. This phenomenon of tumor activation has been described in 9L gliosarcoma [22, 23], RIF-1 fibrosarcoma [24, 25], and 36B-1 0 glioma [26] and may be a general phenomenon in human tumors [21]. The mechanism behind tumor activation is unknown, but I propose that it results from a reduction of IFP in the tumor after treatment. Reduced IFP would permit an improvement in tumor perfusion, which would be expected to result in increased tumor oxygenation [23]. Changes in IFP in the tumor may be a general mechanism that accounts for reoxygenation of the tumor after therapy [26].
Potential
Techniques
for Quantitative
MR Imaging
of
Edema
Edema can cause contrast on MR images, and clinicians are often faced with the problem of discerning tumor from surrounding edema. Tissue Ti and T2 are generally dependent on water and protein composition, and pulse sequences sensitive to tissue relaxation characteristics can be used to differentiate tissues with abnormal water and protein composition [27]. A simple two-compartment, fast-exchange model predicts that relaxation rate is inversely proportional to the fractional water content of tissue [27]. In cold-injury edema and in osmotic edema of the brain, water protons have longer Ti and T2 values than in normal brain [28]. The long proton Ti and T2 generally measured in tumors may simply reflect the presence of edema around and within the tumors. Thus, prolonged Ti and T2 may not be characteristic of tumors, but may simply be indicative of local edema associated with tumors [28]. In fact, T2-weighted MR images usually cannot be used to distinguish between brain tumor and edema in “normal” brain surrounding the tumor [29]. If edema itself is largely responsible for contrast in brain tumor images, it is likely that Ti or T2-weighted images will not be helpful for discriminating tumor from edema. Yet edema alone is not sufficient to explain contrast on Ti and T2-weighted images. Ti and T2 values from normal and diseased tissue often overlap, and variations in Ti and T2 for the same tissue from different healthy subjects may be larger than the differences between normal and diseased tissues [30]. Although the findings on Ti -weighted MR images conrelate directly with extracellular water volume and total water content, and also correlate inversely with intracellular water content in several experimental tumors [31 321, conflicting evidence suggests that significant differences in Ti can occur without any change in water content. For example, in freshly excised mouse tissue, which varies in water content by less than 1 %, the variation in Ti may be as much as 5% [27]. For human brain tumors, Ti is proportional to edema [33]. However, a change in oxygen partial pressure is sufficient to alter Ti significantly [33], suggesting that Ti -weighted imaging will not be adequate for accurate quantification of tumor edema. -
-
,
TUMOR
PERFUSION
261
The findings on T2-weighted MR images also correlate directly with extracellular water volume and total water content, and inversely with intracellular water content in several tumors [31 , 32]. As mentioned before, T2-weighted MR images generally cannot be used to distinguish between brain tumor and edema in “normal” brain surrounding the tumor. This suggests that T2-weighted MR images are actually imaging edema. However, in vitro differences in T2 in a range of cell lines are not completely explained by cell hydration [34]. Diffusion-weighted MR imaging, or intravoxel incoherent motion (IVIM) imaging [35-38], may be a better method to quantify edema. Edema represents an increase in the free water compartment of tissue, so production of edema should result in an increase in the apparent diffusion coefficient (ADC) of water. Because diffusion-weighted imaging is a spin-echo technique, it generates a more intense echo from protons that are undergoing less translational motion [39, 40]. Regions of tumor in which proton diffusion is elevated are therefore expected to contribute less signal (Fig. 1). Factors that permit increased proton diffusion in tumor are unknown, but could include vasogenic edema leading to an increase in the extracellular water compartment, where diffusion is less hindered by cell membranes; increased convective movement of fluid through necrotic regions of tumor; or cellular swelling permitting an increase in the root mean square path length of proton diffusion within cells [40]. Diffusion-weighted imaging has been used to discriminate necrotic from living tumor cells, as necrotic tissue usually allows greater diffusion than living tissue does [41 ]. However, in one study, diffusion-weighted imaging showed a significant increase in signal intensity in ischemic rat brain within 14 mm of the onset of ischemia [42]. In contrast, on T2-weighted images, no changes in ischemic brain could be detected until 2 hr after the ischemic injury. Increased signal intensity in diffusion-weighted imaging of ischemic brain is indicative of a regional decrease in apparent proton diffusion. This was hypothesized to result from either a decrease in the extracellular water compartment due to cytotoxic edema or a decrease in the microscopic pulsatile movements of brain caused by blood perfusion [42]. Diffusion-weighted imaging can be used to generate an ADC image [43]. If diffusion-encoding gradients are weak, contrast in the ADC image will depend on magnetic susceptibility, perfusion, and restricted diffusion [41]. IVIM imaging [35-38] has been proposed as a way to measure capillary perfusion directly, but the method is equally sensitive to arterial blood flow, shunt flow, and convective fluid flow within the tissue of interest [44]. Thus, using IVIM imaging to measure tissue perfusion is considerably more difficult than simply measuring ADC. In ADC imaging, the strength of the diffusionsensitizing gradient is increased, so that image contrast becomes primarily dependent on diffusion rather than perfusion [4i]. Nevertheless, diffusion-weighted or ADC imaging is technically rather difficult [41 ]. Persistent eddy currents arising from large pulse gradients can contribute to signal intensity even in the absence of diffusion [41 J. Subject motion seriously
Downloaded from www.ajronline.org by 117.253.209.72 on 11/09/15 from IP address 117.253.209.72. Copyright ARRS. For personal use only; all rights reserved
262
STEEN
AJA:158,
February
1992
Fig. 1.-A-D, Series of proton images acquired from a rat bearing a flank-implanted 9L gliosarcoma (2 cm in diameter). The rat was anesthetized, wrapped with a Faraday shield to oxdude signal from body wall, and maintamed at 37#{176}c during Image acquisition. Images were acquired at 2 T with a diffusion-weighted spIn-echo Image sequence, 1000/80 [39], initially without diffusion-sensitizIng gradients to obtain a control Image (A). Then, soquential images were acquired with gradients applied (5 G/cm) along two
different axes. Slice thickness was 4 mm, with 400-kim in-plane resolution across a 50-mm field of view, and image acquisition took less than 5 mm. A difference is apparent between control image, acquired without gradients (A), and image acquired with a gradient applied (5 G/cm) (B). Diffusion-sensitive image (B) shows a loss of signal in certain regions of tumor that Indicates increased proton diffusion in these regions. Subtraction of control image (A) from diffusion-sensitive image (B) yields diffusion-weighted images (C and D), on which increased diffusion is apparent as dark areas. Diffusionweighted images were acquired with gradients applied along two different axes: along x axis for C and along z axis for D. Gross histologic examination of tumor showed that mass was formed of two nodules: a large nodule seen at bottom of images and a smaller nodule seen at top of the images. Diffusion-weighted
Images
(C
and
0)
show a region of relatively rapid diffusion in center of both nodules that may Indicate central necrosis of nodules. Comparison of subtraction images (C and 0) suggests that no significant anisotropy of diffusion is present within tumor.
degrades ADC images even when gross motion is controlled, as small-amplitude pulsatile motion is also significant for ADC [41 ]. Finally, ADC imaging may be no more successful than Ti- or T2-weighted imaging for separating tumor from sunrounding edema. Some of the new fast imaging techniques, such as echoplanar imaging, are expected to be quite useful for diffusion-weighted imaging, as these techniques can mmimize the motion artifacts that plague ADC imaging [45]. Correlation of Diffusion-Weighted Tumor Perfusion
MR Imaging
with
It has been hypothesized that tumor edema, with the associated increase in IFP, is the mechanism that limits tumor perfusion. Thus, quantitative MR imaging of edema may be correlated with tumor perfusion. This hypothesis is consistent with preliminary data, which show that large increases in the diffusion coefficient (D) of water can be associated with tumor progression and that D in tumors generally increases with increasing tumor malignancy [46]. For example, in normal
murine mammary tissue, D = 0.34 0.35 (cm2/sec x 10); in preneoplastic lesions, D = 0.44 0.67; and in mammary tumors, D = 0.70 [34]. In pure water, D = 2.38. In normal rat brain cortex, D = 0.78, and in intracranial 9L gliosarcoma, D = 1 .32 [46]. D was measured in 9L gliosarcomas during tumor progression and after treatment with several different dosages of carmustine (BCNU). Tumor growth was accompanied by an increase in D, consistent with increased formation of edema during tumor progression. No significant changes of D could be detected in 9L gliosarcoma 1 day after treatment with BCNU (i 3 mg/kg), but at 3 days after treatment with BCNU (i 0 mg/kg), a significant decrease in D was seen [46]. The bioenergetic changes in 9L gliosarcoma observed duning untreated tumor progression and after therapy with cytotoxic drugs suggest that tumor perfusion declines during tumor progression and that BCNU treatment induces an increase in tumor perfusion [22, 23, 47]. This conclusion is consistent with observed changes in D in 9L gliosarcomas [46]. The decrease in high-energy phosphates during un-
-
AJR:158,
February
1H MR OF EDEMA
1992
AND
tumor growth, caused by a progressive decline in perfusion [22], may result from a progressive increase in edema in the tumor [46]. The increase in high-energy phosphates in 9L gliosarcoma after BCNU therapy is associated
TUMOR
pressure
treated
Downloaded from www.ajronline.org by 117.253.209.72 on 11/09/15 from IP address 117.253.209.72. Copyright ARRS. For personal use only; all rights reserved
with
increased
tumor
perfusion
[23],
and could
result
from
a
decrease in tumor edema following BCNU therapy [46]. This strongly suggests that the increase in tumor bioenergetic state after therapy is a result of decreased tumor hydrostatic pressure, which results in improved blood flow to the tumor [2i, 23].
Because
tive imaging characterizing
14. 15.
17.
MR imaging potentially can be used for quantitaof tumor edema, it may provide a means of tumor perfusion. The potential ability of MR
imaging to give a qualitative or quantitative indication of tumor perfusion may enable clinicians to determine whether perfu-
sion within a tumor mass is heterogeneous. It may also become possible to monitor changes in perfusion associated with tumor reoxygenation between tumor perfusion
therapeutic
agents
after
therapy.
If the
imaging
of tumor
edema
eventually
delivery
of drugs to a tumor.
may
18.
19.
20.
relationship
and the pharmacokinetics of chemois better understood, quantitative MR aid in predicting
21.
the 22.
ACKNOWLEDGMENTS the diffusionand I am grateful for his contributions and his collegiality. Image analysis used Image 1 .27, which was provided by Wayne Rasband (Research Services Branch, National Institute of Mental Health). I thank Catherine Cunningham, Wil O’Loughlin, and three anonymous reviewers for help with the manuscript. Todd
weighted
Richards imaging
was instrumental
sequence
used
in implementing
to obtain
Figure
1
23.
,
REFERENCES 1 . Reulen H-J, Huber P. Ito U, Groger U. Peritumoral brain edema. Adv Neurol 1990:52:307-315 2. Jam AK. Delivery of novel therapeutic agents in tumors: physiological barriers and strategies. J Nat! Cancer Inst 1989:81:570-576 3. Granger HJ, Shepherd AP. Dynamics and control of the microcirculation. Adv Biomed Eng 1979;7: 1-63 4. Butler TP, Grantham FH, Guflino PM. Bulk transfer offluid in the interstitial compartment of mammary tumors. Cancer Res 1975;35:3084-3088 5. Shah-Yukich AA, Nelson AC. Characterization of solid tumor microvasculature: a three-dimensional analysis using the polymer casting technique. Lab Invest 1988;58:236-244 6. Jam AK. Determinants of tumor blood flow: a review. Cancer Res 1988;48:2641-2658 7. Sevick EM, Jam AK. Measurement of capillary filtration coefficient in a solid tumor. Cancer Res 1991;51 : 1352-1 355 8. Chaplin DJ, Olive PL, Durand RE. Intermittent blood flow in a munne tumor: radiobiological effects. Cancer Res 1987;47:597-601 9. Trotter MJ, Acker BD, Chaplin DJ. Histological evidence for nonperfused vasculature in a munne tumor following hydralazine administration. !nt J Radiat Oncol Biol Phys 1989:17:785-789 10. Boucher Y, Baxter LT, Jaln AK. Interstitial pressure gradients in tissueisolated and subcutaneous tumors: implications for therapy. Cancer Res 1990;50:4478-4484 1 1 . Kuroiwa T, Shibutani M, Tajima T, Hirasawa H, Okeda A. Hydrostatic
versus
osmotic
263
pressure
in the development
of vasogenic
brain
edema induced by cold injury. AdvNeurol 1990;52:11-19 1 2. Verdugo P. Goblet cell secretion and mucogenesis. Annu Rev Physio! 1990;52: 157-1 76 13. Koivunen E, Aistimaki A, ltknoen 0, Osman 5, Vuento M, Stenman U-H. Tumor-associated trypsin participates in cancer cell-mediated degradation
16.
Conclusions
PERFUSION
24.
25.
26. 27.
28. 29.
30.
31
32.
33.
34.
.
of extracellular matrix. Cancer Res 1991;51 :2107-2112 Marmarou A, Takagi H, Shulman K. Biomechanics of brain edema and effects on local cerebral blood flow. Adv Neuro! 1980;28:345-358 Groebe K, Vaupel P. Evaluation of oxygen diffusion distances in human breast cancer xenografts using tumor-specific in vivo data: role of various mechanisms in the development of tumor hypoxia. Int J Radiat Oncol Biol Phys 1988;15:691-697 Hatashita 5, Hoff JT, Salamat SM. An osmotic gradient in ischemic brain edema. Adv Neurol 1990;52:85-92 Kempski 0, Staub F, Jansen M, Baethmann A. Molecular mechanisms of glial cell swelling in acidosis. Adv Neurol 1990:52:39-45 Hatashita T, Ito M, Miyaoka M, Ishii S. Chronological afterations of regional cerebral blood flow, glucose utilization, and edema formation after focal ischemia in hypertensive and normotensive rats. Adv Neurol 1990:52: 29-37 Sutton LN, McLaughlin AC, Kemp W, et al. Effects of increased ICP on brain phosphocreatine and lactate, determined by simuftaneous ‘H and 31P NMA spectroscopy. J Neurosurg 1987;67:381-386 Boucher Y, Less JR, Posner MC, Wolmark N, Jam AK. Interstitial hypertension in human primary and metastatic tumors. In: Chapman JO, Dewey Wc, Whitmore GF, eds. Radiation research: a twentieth century perspective, vol. 1 . San Diego: Academic Press, 1991 :465 Steen AG. Response of solid tumors to chemotherapy monitored by in vivo 31P nuclear magnetic resonance spectroscopy: a review. Cancer Res 1989:49:4075-4085 Steen AG, Tamargo RJ, McGovern KA, et al. In vivo 31P nuclear magnetic resonance spectroscopy of subcutaneous 9L gliosarcoma: effects of tumor growth and treatment with 1 ,3-bis(2-chloroethyl)-1 -nitrosourea on tumor bioenergetics and histology. Cancer Res 1988;48:676-681 Steen AG, Graham MM. 31P magnetic resonance spectroscopy is sensitive to tumor hypoxia: perfusion and oxygenation of rat 9L gliosarcoma after treatment with BCNU. NMR Biomed 1991;4: 117-124 Li S-J, Webrle JP, Rajan SS, Steen AG, Glickson JO, Hilton J. Response of radiation-induced fibrosarcoma-1 in mice to cyclophosphamide monitored by in vivo 31P nuclear magnetic resonance spectroscopy. Cancer Res 1988;48:4736-4742 Rajan 55, Wehrle JP, U S-J, Steen AG, Glickson JO. 31P NMR spectroscopic study of bioenergetic changes in radiation-induced fibrosarcoma. 1 . Fibrosarcoma after radiation therapy. NMR Biomed 1989;2: 165-171 Steen AG. Characterization of tumor hypoxia by 31P MA spectroscopy. AIR 1991;157:243-248 Gore JC, Brown MS. Pathophysiological significance of relaxation. In: Pertain CL, Price AR, Patton JA, Kulkami MV, James AE, eds. Magnetic resonance imaging, vol. 2. Philadelphia: Saunders, 1988: 1070-1 074 Go KG, Edzes HT. Water in brain edema: observations by the pulsed nuclear magnetic resonance technique. Arch Neurol 1975;32:462-465 DeVrieS LS, Bydder GM. Tumors of the central nervous system. In: Partain CL, Price AR, PattonJA, Kulkami MV, James AE, eds. Magnetic resonance imaging, vol. 1 . Philadelphia: Saunders, 1988:144-168 Mathur-de Vre A. BiOmediCal implications of relaxation times of tissue water. In: Partaln CL, Price AR, Patton JA, Kulkami MV, James AE, eds. Magnetic resonance imaging, vol. 2. Philadelphia: Saunders, 1988: 1099-1108 Braunschwelger PG. Schiffer LM, Furmanski P. ‘H-NMR relaxation times and water compartmentalization in experimental tumor models. Magn Reson Imaging 1986:4:335-342 Braunschweiger PG. Reynolds K, Nelson TA, Maring E. The effect of dexamethasone on tissue water distribution and proton relaxation in PancO2 tumors. Magn Reson Imaging 1987:5:483-492 Lundbom N, Brown AD, Koenig SH, Lansen TA, Valsamis MP, Kasoff 55. Magnetic field dependence of l/T, of human brain tumors: correlations with histology. Invest Radiol 1990;25: 1 197-1 205 Beat PT, Hazlewood CF. Distinction of the normal, preneoplastic, and neoplastic states by water proton NMR relaxation times. In: Partain CL, Price AR, Patton JA, Kulkami MV, James AE, eds. Magnetic resonance
STEEN
264
35.
Downloaded from www.ajronline.org by 117.253.209.72 on 11/09/15 from IP address 117.253.209.72. Copyright ARRS. For personal use only; all rights reserved
36.
37.
38. 39.
40.
41.
imaging, vol. 2. Philadelphia: Saunders, 1988:1710-1734 Le Bihan 0, Breton E, Lallemand 0, Grenier P, Cabanis E, Laval-Jeantet M. MR imaging of intravoxal incoherent motions: application to diffusion and perfusion in neurologic disorders. Radiology 1986;161 :401-407 Le Bihan 0, Breton E, Lailemand 0, Aubin M-L, Vignaud J, Laval-Jeantet M. Separation of diffusion and perfusion in intravoxal incoherent motion MA imaging. Radiology 1988:168:497-505 Moonen CTW, van ZijI PCM, Frank JA, Bihan 0, Becker ED. Functional magnetic resonance imaging in medicine and physiology. Science 1990;250:53-61 Le Bihan 0. Magnetic resonance imaging of perfusion. Magn Reson Med 1990;14:283-292 Moseley ME, Cohen V. Kucharczyk J, et al. Diffusion-weighted MA imaging of anisotropic water diffusion in cat central nervous system. Radiology 1990;176:439-445 Moseley ME, Cohen Y, Mintorovitch J, et al. Early detection of regional cerebral ischemia in cats: comparison of diffusionand Trweighted MAI and spectroscopy. Magn Reson Med 1990;14:330-346 Henkelmann AM. Diffusion-weighted MR imaging: a useful adjunct to clinical diagnosis or a scientific cunousity? AiR 1990;155: 1066-1068
LIST OF BOOK
AJA:158,
42.
43.
44. 45. 46.
47.
epidermoid tumors. AiR 1990;155: 1059-1 065 Henkelman AM. Does IVIM measure classical perfusion? Magn Reson Med 1990;16:470-475 Haacke EM, Tkach JA. Fast MA imaging: techniques and clinical applications. MR 1990;155:951-964 Hooper J, Rajan 55, Rosa L, Le Bihan 0. Application of diffusion imaging to monitor tumor growth and response to chemotherapy. In: Book of abstracts, vol. 1 . Society of Magnetic Resonance in Medicine. Ninth annual meeting. Berkeley. CA: Society of Magnetic Resonance in Medicine, 1990:371 Steen AG, Tamargo AJ, Brem H, Glickson JO, Wehrle JP. In vivo 3P nuclear magnetic resonance spectroscopy of rat 9L gliosarcoma treated with BCNU: dose-response of spectral changes. Magn Reson Med
1989;1 1:258-266
REVIEWS
282
Head and Neck Imaging, 2nd ed. Som PM, Bergeron RT,
308
Stone Therapy
320
The Discipline of Curiosity.
eds
F, Miller K, Rassweller
Science in the World.
Groen
J, eds
J, Smit
E, Eijsvooge!
J, eds
326 346
352 368
388 398
412 426 442
DiagnostIc
Radiology 1990. Gooding CA, ed and Histologic Pathology of Nontumorous Diseases of Bones and Joints, vols. I and 2. Mllgram JW Errors in Chest Radiography. Viamonte M Jr Understanding MRI. Newhouse JH, Wiener JI Merrill’s Atlas of Radiographic Positions and Radiologic Procedures, vols. 1-3, 7th ed. Balllnger PW Nuclear Radiology (Fourth Series). Test and Syllabus. Thrall JH, section ed The Radiologic Clinics of North America. Radiographic Manifestions of Cornmon Congenital Anomalies. Oh KS, guested RSNA Today Video, Vol. 5, No. 1. The Radiological Society of North America RadiOIOgIC
Computers Medicine
eds
in Diagnostic Report
Radiology.
No 62.) Robertson
1992
Mintorovitch J, Moseley ME, Chileuitt L, Shimizu H, Cohen V. Weinstein PA. Comparison of diffusionand T2-weighted MRI for the early detection of cerebral ischemia and reperfusion in rats. Magn Reson Med 1991 18: 39-50 Tsuruda JS, Chew WM, Moseley ME, Norman D. Diffusion-weighted MA imaging of the brain: value of differentiating between extraaxial cysts and
AND VIDEOTAPE
in UrolOgy. Eisenberger
February
(The Institute
of Physical Sciences
J, Wankllng PF, Faulkner K, Holubinka
in MA,
259
Review
.
,.
..
.
Edema and Tumor Perfusion: Quantitative 1H MR Imaging A. Grant
perfusion is of it has a therapy. Yet determining fused is difficuft without the variables that affect Quantitative MR Imaging of characterizing tumor because
perfusion
Characterization
within
central importance to the clinical oncoldirect effect on the success of cancer whether a tumor is well or poorly perthe use of invasive techniques, because tumor perfusion are poorly understood. of tumor edema may provide a means perfusion, of studying heterogeneity of
the tumor
mass,
or of monitoring
changes
in
after therapy. A combination of factors often resufts in production of a large amount of edema within cranial or extracranial tumors. Any tumor that is encapsulated, whether by a fibrous tumor capsule or by a structure such as the cranium, will have an elevation in interstitial fluid pressure because dissipation of fluid is hindered. Elevated pressure of interstitial fluid acts to occlude tumor capillaries, so edema can cause a striking reduction of tumor perfusion. Because MR imaging can potentially be used for quantitative imaging of tumor edema, it may provide a means of indirectly measuring tumor perfusion. A review of the Ifterature suggests that diffusion-weighted MR imaging may be befter than TI- or T2-weighted MR imaging for tumor
perfusion
quantitative
imaging
diftusion-weighted rather correlated
I hypothesize with
tumor
of tumor edema. I do not propose that imaging can measure perfusion directly; that a diffusion-weighted image can be edema.
Because
edema
indirectly
regulates
of interstitial fluid pressure, I propose an indirect correlation between the diffusion-weighted image and regional tissue perfusion. If the relationship between tumor perfusion and the pharmacokinetics of chemotherapeutic agents is better understood, MR imaging of tumor edema may even aid in predicting the delivery of drugs to a tumor. perfusion
by
Steen1
Tumor
ogist
Article
through
the mechanism
Received July 5, 1991 ; accepted after revision October 4. 1991. This work was supported in part by Biomedical Research Support Grant AA-05432 I Departments of Radiology (RC-05) and Bioengineering, University of Washington Steen. AJR 158:259-264,
February
1992 0361-803X/92/1582-0259
C American
Roentgen
Peritumoral edema is a widely recognized problem in the treatment of cranial malignancies [1], but it is increasingly apparent that edema is also a problem in extracranial tumors [2]. Edema acts to elevate interstitial pressure, which can produce a range of effects: Tumor perfusion may become inadequate and unpredictable; delivery of therapeutic agents to the tumor may be reduced; hypoxic foci of cells, which are inherently radioresistant [2], may develop; and mass effects of edema
in the cranium
can cause
neurologic
impairment
[1].
If a technique could be developed for quantitative MR imaging of tumor edema, it might enable researchers to characterize the relationship between edema and tumor perfusion by using a noninvasive technique.
Potential
Mechanisms
Generating
Edema
in Tumors
A variety of factors result in substantial edema within tumors at any site. Transcapillary filtration forces are slightly greater than absorptive forces in most tissues, so circulatory fluids tnansude into the interstitial space at a low rate [3]. The flux of water and protein from blood to interstitium is normally balanced by the flow oflymph, which carries water and protein back to the bloodstream. However, tumors lack a lymphatic system, so the normal pathway of drainage is absent [2, 4]. This tendency toward vasogenic edema in tumors is compounded by several other factors: (1) Tumor capillaries are often more permeable than normal capillaries [2, 5-7]. For
and American Cancer Society Institutional School of Medicine, Seattle, WA 98195. Ray Society
Grant IN-26-31. Address reprint requests
to A. G.
260
STEEN
example,
the capillary
adenocarcinoma
was
Downloaded from www.ajronline.org by 117.253.209.72 on 11/09/15 from IP address 117.253.209.72. Copyright ARRS. For personal use only; all rights reserved
normal tissue,
filtration
coefficient
1 0 to 1 000
times
and within
of a mammary higher
the range reported
than
that
of
for filtration
in
glomerular capillaries [7]. (2) Tumor capillaries may be temporarily occluded [8, 9], as a result of cell growth in a confined space [1 0]. This leads to an elevation of blood pressure on
the arterial side of the capillary tion [1 1]. (3) Overproduction polysaccharides by tumor
bed and increased
transuda-
ofextracellular collagen or mucocells can lead to accumulation of
polyanionic macromolecules in the matrix that become hydrated because of Donnan forces [1 2]. (4) Disaggregation or digestion of extracellular matrix by tumor cells [1 3] can increase the osmotic effect of matrix macromolecules [1 4]. (5) Increased intracapillary diffusion distances in tumors can lead to a compromised bioenergetic state and cellular hypoxia [15], so cells may be unable to maintain normal osmotic gradients and may release osmotically active ions such as K and (6) Cell death and lysis may result in release of many proteins and other osmotically active molecules normally sequestered in the cell [16, 17]. A combination ofthese factors is likely to lead to substantial edema within a tumor. For example, in a patient with a brain metastasis, it was estimated that fluid was produced within the tumor at a rate of 3.9 ml/hr, which led to an edematous volume in the brain of 87.3 cm3 [1 ]. Because edema production rate varies
from one tumor
to another,
whereas
resorption
rate by normal brain is probably relatively constant, ume of peritumoral edema in brain probably reflects of fluid production [1]. Effect
of Edema Normal Brain
on Perfusion
and
Bioenergetics
the volthe rate
in
of volume
equilibration
that,
in a constrained
volume
such as the cranium, must be compensated for by a reduction in other fluid and blood compartments [1 4]. When
this
compensation
is inadequate,
tissue
pressure
can
in-
crease, and studies of brain edema have shown a decrease of cerebral blood flow in edematous areas [1 8]. Because flow reduction can occur even with a relatively mild increase in intracranial
pressure
(ICP) and with adequate
tissue
perfusion
pressure, it is likely that edema directly compresses capillaries [14]. The effect of edema on tissue perfusion and bioenergetics has been studied in detail in brain, where cranial edema is associated with stroke, hypertension, trauma, inflammation, AIDS, and intracranial tumors. For example, 31P MR spectroscopy was used in a standard model of cranial hypertension to examine the effect of pressure ischemia on brain bioenergetics [1 9]. Graded increases in CSF pressure were used to induce pressure ischemia while 31P MR spectroscopy was
performed.
A spectrum
mm Hg, then infusing fluid
was acquired
at a resting
ICP of 0-5
ICP was raised in increments of 20 mm Hg by from a cisternal catheter. Cerebral perfusion
pressure (CPP) of blood declined as ICP was raised (CPP = mean arterial blood pressure ICP). Phosphocreatine was -
February
1992
generally maintained until the CPP reached 29 mm Hg and then declined rapidly to undetectable levels as CPP reached 0. Loss of adenosine tniphosphate began only when phosphocreatine was entirely exhausted [1 9]. This demonstrates that tissue pressure induced by edema can have a profound effect on brain
bioenergetics.
Interstitial
ogous to ICP and should perfusion. Edema
and Interstitial
In any tumor
that
fluid
have
pressure
similar
Fluid Pressure
(IFP) is anal-
effects
on tumor
in Tumors
is well encapsulated
(e.g.,
by a fibrous
capsule or within the cranium), and in which interstitial fluid movement is hindered, tumor edema can lead to a striking increase in interstitial pressure [1 4]. This can cause decreased tumor
perfusion
by causing
vascular
occlusion,
suggesting
that quantitative imaging of edema may provide direct insight into tumor perfusion. In general, IFP in tumors is significantly higher than in normal tissues, even in tumors not constrained within the cranium [2]. IFP was measured in normal rat tissue, in two tumors implanted in ovarian tissue (“tissue-isolated tumors”),
and in a tumor implanted subcutaneously [1 0]. IFP of normal muscle varied between 0.0 and 0.4 mm Hg, whereas IFP of dead tissue was 0 mm Hg. Mean IFP of tumors was 8.5 (±5.1) mm Hg, and ranged from 2.7 mm Hg in a 0.5i tumor to 22.8 mm Hg in a 5.0-g tumor.
sharp increase
in interstitial
lying the tumor and Within tissue-isolated
Vascular compression by edema fluid can have a profound effect on tissue blood flow and bioenergetics. Edema is a disturbance
AJA:158,
In subcutaneous
pressure
tumors,
the
began in the skin over-
plateaued at the tumors, pressures
skin-tumor interface. reached plateau val-
ues at a distance of 0.2-1 .1 mm into the tumor. IFP was elevated in 86% of all tumors studied. It has been estimated that an IFP of 10-i 5 mm Hg or more will cause vascular occlusion [10]. Preliminary data from human primary colorectal and breast carcinomas, and from metastases to the lung, liver, and lymph nodes, indicate that IFP in human tumors is significantly higher than
in normal
tissue
[20],
even
though
these
tumors
were
probably not well encapsulated. Mean human tumor IFP was 23.9 (±1 2.3) mm Hg in a series of 16 various tumors. Human tumor IFP ranged from 4 mm Hg in a small breast metastasis (1 9 cm3) to 50 mm Hg in a large lymph node metastasis (133 cm3) [20]. Elevated
IFP in tumors
is expected
to result
in a general
reduction in tumor blood flow and development of necrotic regions in the tumor [2, 6, 10]. In addition, elevated IFP should cause poor delivery of therapeutic agents, including hydrophilic
chemotherapeutic
biological
response
agents,
modifiers
therapeutic agents to tumors to the core of the tumor,
extravasation a convective
monoclonal
antibodies,
[2, 10]. IFP restricts
delivery
and
of
by reducing the access of blood reducing the driving force for
of fluid in the core of the tumor, and generating net flux of fluid toward the periphery of the tumor
[101.
Elevated reduction
IFP should
IFP in tumors is also expected to result in a in tumor bioenergetic state, whereas reduction of result in improved tumor bioenergetic state. Rebioenergetics, obtained by 31P MR spec-
cent data on tumor
Downloaded from www.ajronline.org by 117.253.209.72 on 11/09/15 from IP address 117.253.209.72. Copyright ARRS. For personal use only; all rights reserved
AJR:158,
February
1H MR OF EDEMA
1992
AND
troscopy of tumors, have shown that tumors typically show an increase in high-energy phosphates (adenosine tnphosphate and phosphocreatine) after therapy with cytotoxic drugs [21]. This phenomenon of tumor activation has been described in 9L gliosarcoma [22, 23], RIF-1 fibrosarcoma [24, 25], and 36B-1 0 glioma [26] and may be a general phenomenon in human tumors [21]. The mechanism behind tumor activation is unknown, but I propose that it results from a reduction of IFP in the tumor after treatment. Reduced IFP would permit an improvement in tumor perfusion, which would be expected to result in increased tumor oxygenation [23]. Changes in IFP in the tumor may be a general mechanism that accounts for reoxygenation of the tumor after therapy [26].
Potential
Techniques
for Quantitative
MR Imaging
of
Edema
Edema can cause contrast on MR images, and clinicians are often faced with the problem of discerning tumor from surrounding edema. Tissue Ti and T2 are generally dependent on water and protein composition, and pulse sequences sensitive to tissue relaxation characteristics can be used to differentiate tissues with abnormal water and protein composition [27]. A simple two-compartment, fast-exchange model predicts that relaxation rate is inversely proportional to the fractional water content of tissue [27]. In cold-injury edema and in osmotic edema of the brain, water protons have longer Ti and T2 values than in normal brain [28]. The long proton Ti and T2 generally measured in tumors may simply reflect the presence of edema around and within the tumors. Thus, prolonged Ti and T2 may not be characteristic of tumors, but may simply be indicative of local edema associated with tumors [28]. In fact, T2-weighted MR images usually cannot be used to distinguish between brain tumor and edema in “normal” brain surrounding the tumor [29]. If edema itself is largely responsible for contrast in brain tumor images, it is likely that Ti or T2-weighted images will not be helpful for discriminating tumor from edema. Yet edema alone is not sufficient to explain contrast on Ti and T2-weighted images. Ti and T2 values from normal and diseased tissue often overlap, and variations in Ti and T2 for the same tissue from different healthy subjects may be larger than the differences between normal and diseased tissues [30]. Although the findings on Ti -weighted MR images conrelate directly with extracellular water volume and total water content, and also correlate inversely with intracellular water content in several experimental tumors [31 321, conflicting evidence suggests that significant differences in Ti can occur without any change in water content. For example, in freshly excised mouse tissue, which varies in water content by less than 1 %, the variation in Ti may be as much as 5% [27]. For human brain tumors, Ti is proportional to edema [33]. However, a change in oxygen partial pressure is sufficient to alter Ti significantly [33], suggesting that Ti -weighted imaging will not be adequate for accurate quantification of tumor edema. -
-
,
TUMOR
PERFUSION
261
The findings on T2-weighted MR images also correlate directly with extracellular water volume and total water content, and inversely with intracellular water content in several tumors [31 , 32]. As mentioned before, T2-weighted MR images generally cannot be used to distinguish between brain tumor and edema in “normal” brain surrounding the tumor. This suggests that T2-weighted MR images are actually imaging edema. However, in vitro differences in T2 in a range of cell lines are not completely explained by cell hydration [34]. Diffusion-weighted MR imaging, or intravoxel incoherent motion (IVIM) imaging [35-38], may be a better method to quantify edema. Edema represents an increase in the free water compartment of tissue, so production of edema should result in an increase in the apparent diffusion coefficient (ADC) of water. Because diffusion-weighted imaging is a spin-echo technique, it generates a more intense echo from protons that are undergoing less translational motion [39, 40]. Regions of tumor in which proton diffusion is elevated are therefore expected to contribute less signal (Fig. 1). Factors that permit increased proton diffusion in tumor are unknown, but could include vasogenic edema leading to an increase in the extracellular water compartment, where diffusion is less hindered by cell membranes; increased convective movement of fluid through necrotic regions of tumor; or cellular swelling permitting an increase in the root mean square path length of proton diffusion within cells [40]. Diffusion-weighted imaging has been used to discriminate necrotic from living tumor cells, as necrotic tissue usually allows greater diffusion than living tissue does [41 ]. However, in one study, diffusion-weighted imaging showed a significant increase in signal intensity in ischemic rat brain within 14 mm of the onset of ischemia [42]. In contrast, on T2-weighted images, no changes in ischemic brain could be detected until 2 hr after the ischemic injury. Increased signal intensity in diffusion-weighted imaging of ischemic brain is indicative of a regional decrease in apparent proton diffusion. This was hypothesized to result from either a decrease in the extracellular water compartment due to cytotoxic edema or a decrease in the microscopic pulsatile movements of brain caused by blood perfusion [42]. Diffusion-weighted imaging can be used to generate an ADC image [43]. If diffusion-encoding gradients are weak, contrast in the ADC image will depend on magnetic susceptibility, perfusion, and restricted diffusion [41]. IVIM imaging [35-38] has been proposed as a way to measure capillary perfusion directly, but the method is equally sensitive to arterial blood flow, shunt flow, and convective fluid flow within the tissue of interest [44]. Thus, using IVIM imaging to measure tissue perfusion is considerably more difficult than simply measuring ADC. In ADC imaging, the strength of the diffusionsensitizing gradient is increased, so that image contrast becomes primarily dependent on diffusion rather than perfusion [4i]. Nevertheless, diffusion-weighted or ADC imaging is technically rather difficult [41 ]. Persistent eddy currents arising from large pulse gradients can contribute to signal intensity even in the absence of diffusion [41 J. Subject motion seriously
Downloaded from www.ajronline.org by 117.253.209.72 on 11/09/15 from IP address 117.253.209.72. Copyright ARRS. For personal use only; all rights reserved
262
STEEN
AJA:158,
February
1992
Fig. 1.-A-D, Series of proton images acquired from a rat bearing a flank-implanted 9L gliosarcoma (2 cm in diameter). The rat was anesthetized, wrapped with a Faraday shield to oxdude signal from body wall, and maintamed at 37#{176}c during Image acquisition. Images were acquired at 2 T with a diffusion-weighted spIn-echo Image sequence, 1000/80 [39], initially without diffusion-sensitizIng gradients to obtain a control Image (A). Then, soquential images were acquired with gradients applied (5 G/cm) along two
different axes. Slice thickness was 4 mm, with 400-kim in-plane resolution across a 50-mm field of view, and image acquisition took less than 5 mm. A difference is apparent between control image, acquired without gradients (A), and image acquired with a gradient applied (5 G/cm) (B). Diffusion-sensitive image (B) shows a loss of signal in certain regions of tumor that Indicates increased proton diffusion in these regions. Subtraction of control image (A) from diffusion-sensitive image (B) yields diffusion-weighted images (C and D), on which increased diffusion is apparent as dark areas. Diffusionweighted images were acquired with gradients applied along two different axes: along x axis for C and along z axis for D. Gross histologic examination of tumor showed that mass was formed of two nodules: a large nodule seen at bottom of images and a smaller nodule seen at top of the images. Diffusion-weighted
Images
(C
and
0)
show a region of relatively rapid diffusion in center of both nodules that may Indicate central necrosis of nodules. Comparison of subtraction images (C and 0) suggests that no significant anisotropy of diffusion is present within tumor.
degrades ADC images even when gross motion is controlled, as small-amplitude pulsatile motion is also significant for ADC [41 ]. Finally, ADC imaging may be no more successful than Ti- or T2-weighted imaging for separating tumor from sunrounding edema. Some of the new fast imaging techniques, such as echoplanar imaging, are expected to be quite useful for diffusion-weighted imaging, as these techniques can mmimize the motion artifacts that plague ADC imaging [45]. Correlation of Diffusion-Weighted Tumor Perfusion
MR Imaging
with
It has been hypothesized that tumor edema, with the associated increase in IFP, is the mechanism that limits tumor perfusion. Thus, quantitative MR imaging of edema may be correlated with tumor perfusion. This hypothesis is consistent with preliminary data, which show that large increases in the diffusion coefficient (D) of water can be associated with tumor progression and that D in tumors generally increases with increasing tumor malignancy [46]. For example, in normal
murine mammary tissue, D = 0.34 0.35 (cm2/sec x 10); in preneoplastic lesions, D = 0.44 0.67; and in mammary tumors, D = 0.70 [34]. In pure water, D = 2.38. In normal rat brain cortex, D = 0.78, and in intracranial 9L gliosarcoma, D = 1 .32 [46]. D was measured in 9L gliosarcomas during tumor progression and after treatment with several different dosages of carmustine (BCNU). Tumor growth was accompanied by an increase in D, consistent with increased formation of edema during tumor progression. No significant changes of D could be detected in 9L gliosarcoma 1 day after treatment with BCNU (i 3 mg/kg), but at 3 days after treatment with BCNU (i 0 mg/kg), a significant decrease in D was seen [46]. The bioenergetic changes in 9L gliosarcoma observed duning untreated tumor progression and after therapy with cytotoxic drugs suggest that tumor perfusion declines during tumor progression and that BCNU treatment induces an increase in tumor perfusion [22, 23, 47]. This conclusion is consistent with observed changes in D in 9L gliosarcomas [46]. The decrease in high-energy phosphates during un-
-
AJR:158,
February
1H MR OF EDEMA
1992
AND
tumor growth, caused by a progressive decline in perfusion [22], may result from a progressive increase in edema in the tumor [46]. The increase in high-energy phosphates in 9L gliosarcoma after BCNU therapy is associated
TUMOR
pressure
treated
Downloaded from www.ajronline.org by 117.253.209.72 on 11/09/15 from IP address 117.253.209.72. Copyright ARRS. For personal use only; all rights reserved
with
increased
tumor
perfusion
[23],
and could
result
from
a
decrease in tumor edema following BCNU therapy [46]. This strongly suggests that the increase in tumor bioenergetic state after therapy is a result of decreased tumor hydrostatic pressure, which results in improved blood flow to the tumor [2i, 23].
Because
tive imaging characterizing
14. 15.
17.
MR imaging potentially can be used for quantitaof tumor edema, it may provide a means of tumor perfusion. The potential ability of MR
imaging to give a qualitative or quantitative indication of tumor perfusion may enable clinicians to determine whether perfu-
sion within a tumor mass is heterogeneous. It may also become possible to monitor changes in perfusion associated with tumor reoxygenation between tumor perfusion
therapeutic
agents
after
therapy.
If the
imaging
of tumor
edema
eventually
delivery
of drugs to a tumor.
may
18.
19.
20.
relationship
and the pharmacokinetics of chemois better understood, quantitative MR aid in predicting
21.
the 22.
ACKNOWLEDGMENTS the diffusionand I am grateful for his contributions and his collegiality. Image analysis used Image 1 .27, which was provided by Wayne Rasband (Research Services Branch, National Institute of Mental Health). I thank Catherine Cunningham, Wil O’Loughlin, and three anonymous reviewers for help with the manuscript. Todd
weighted
Richards imaging
was instrumental
sequence
used
in implementing
to obtain
Figure
1
23.
,
REFERENCES 1 . Reulen H-J, Huber P. Ito U, Groger U. Peritumoral brain edema. Adv Neurol 1990:52:307-315 2. Jam AK. Delivery of novel therapeutic agents in tumors: physiological barriers and strategies. J Nat! Cancer Inst 1989:81:570-576 3. Granger HJ, Shepherd AP. Dynamics and control of the microcirculation. Adv Biomed Eng 1979;7: 1-63 4. Butler TP, Grantham FH, Guflino PM. Bulk transfer offluid in the interstitial compartment of mammary tumors. Cancer Res 1975;35:3084-3088 5. Shah-Yukich AA, Nelson AC. Characterization of solid tumor microvasculature: a three-dimensional analysis using the polymer casting technique. Lab Invest 1988;58:236-244 6. Jam AK. Determinants of tumor blood flow: a review. Cancer Res 1988;48:2641-2658 7. Sevick EM, Jam AK. Measurement of capillary filtration coefficient in a solid tumor. Cancer Res 1991;51 : 1352-1 355 8. Chaplin DJ, Olive PL, Durand RE. Intermittent blood flow in a munne tumor: radiobiological effects. Cancer Res 1987;47:597-601 9. Trotter MJ, Acker BD, Chaplin DJ. Histological evidence for nonperfused vasculature in a munne tumor following hydralazine administration. !nt J Radiat Oncol Biol Phys 1989:17:785-789 10. Boucher Y, Baxter LT, Jaln AK. Interstitial pressure gradients in tissueisolated and subcutaneous tumors: implications for therapy. Cancer Res 1990;50:4478-4484 1 1 . Kuroiwa T, Shibutani M, Tajima T, Hirasawa H, Okeda A. Hydrostatic
versus
osmotic
263
pressure
in the development
of vasogenic
brain
edema induced by cold injury. AdvNeurol 1990;52:11-19 1 2. Verdugo P. Goblet cell secretion and mucogenesis. Annu Rev Physio! 1990;52: 157-1 76 13. Koivunen E, Aistimaki A, ltknoen 0, Osman 5, Vuento M, Stenman U-H. Tumor-associated trypsin participates in cancer cell-mediated degradation
16.
Conclusions
PERFUSION
24.
25.
26. 27.
28. 29.
30.
31
32.
33.
34.
.
of extracellular matrix. Cancer Res 1991;51 :2107-2112 Marmarou A, Takagi H, Shulman K. Biomechanics of brain edema and effects on local cerebral blood flow. Adv Neuro! 1980;28:345-358 Groebe K, Vaupel P. Evaluation of oxygen diffusion distances in human breast cancer xenografts using tumor-specific in vivo data: role of various mechanisms in the development of tumor hypoxia. Int J Radiat Oncol Biol Phys 1988;15:691-697 Hatashita 5, Hoff JT, Salamat SM. An osmotic gradient in ischemic brain edema. Adv Neurol 1990;52:85-92 Kempski 0, Staub F, Jansen M, Baethmann A. Molecular mechanisms of glial cell swelling in acidosis. Adv Neurol 1990:52:39-45 Hatashita T, Ito M, Miyaoka M, Ishii S. Chronological afterations of regional cerebral blood flow, glucose utilization, and edema formation after focal ischemia in hypertensive and normotensive rats. Adv Neurol 1990:52: 29-37 Sutton LN, McLaughlin AC, Kemp W, et al. Effects of increased ICP on brain phosphocreatine and lactate, determined by simuftaneous ‘H and 31P NMA spectroscopy. J Neurosurg 1987;67:381-386 Boucher Y, Less JR, Posner MC, Wolmark N, Jam AK. Interstitial hypertension in human primary and metastatic tumors. In: Chapman JO, Dewey Wc, Whitmore GF, eds. Radiation research: a twentieth century perspective, vol. 1 . San Diego: Academic Press, 1991 :465 Steen AG. Response of solid tumors to chemotherapy monitored by in vivo 31P nuclear magnetic resonance spectroscopy: a review. Cancer Res 1989:49:4075-4085 Steen AG, Tamargo RJ, McGovern KA, et al. In vivo 31P nuclear magnetic resonance spectroscopy of subcutaneous 9L gliosarcoma: effects of tumor growth and treatment with 1 ,3-bis(2-chloroethyl)-1 -nitrosourea on tumor bioenergetics and histology. Cancer Res 1988;48:676-681 Steen AG, Graham MM. 31P magnetic resonance spectroscopy is sensitive to tumor hypoxia: perfusion and oxygenation of rat 9L gliosarcoma after treatment with BCNU. NMR Biomed 1991;4: 117-124 Li S-J, Webrle JP, Rajan SS, Steen AG, Glickson JO, Hilton J. Response of radiation-induced fibrosarcoma-1 in mice to cyclophosphamide monitored by in vivo 31P nuclear magnetic resonance spectroscopy. Cancer Res 1988;48:4736-4742 Rajan 55, Wehrle JP, U S-J, Steen AG, Glickson JO. 31P NMR spectroscopic study of bioenergetic changes in radiation-induced fibrosarcoma. 1 . Fibrosarcoma after radiation therapy. NMR Biomed 1989;2: 165-171 Steen AG. Characterization of tumor hypoxia by 31P MA spectroscopy. AIR 1991;157:243-248 Gore JC, Brown MS. Pathophysiological significance of relaxation. In: Pertain CL, Price AR, Patton JA, Kulkami MV, James AE, eds. Magnetic resonance imaging, vol. 2. Philadelphia: Saunders, 1988: 1070-1 074 Go KG, Edzes HT. Water in brain edema: observations by the pulsed nuclear magnetic resonance technique. Arch Neurol 1975;32:462-465 DeVrieS LS, Bydder GM. Tumors of the central nervous system. In: Partain CL, Price AR, PattonJA, Kulkami MV, James AE, eds. Magnetic resonance imaging, vol. 1 . Philadelphia: Saunders, 1988:144-168 Mathur-de Vre A. BiOmediCal implications of relaxation times of tissue water. In: Partaln CL, Price AR, Patton JA, Kulkami MV, James AE, eds. Magnetic resonance imaging, vol. 2. Philadelphia: Saunders, 1988: 1099-1108 Braunschwelger PG. Schiffer LM, Furmanski P. ‘H-NMR relaxation times and water compartmentalization in experimental tumor models. Magn Reson Imaging 1986:4:335-342 Braunschweiger PG. Reynolds K, Nelson TA, Maring E. The effect of dexamethasone on tissue water distribution and proton relaxation in PancO2 tumors. Magn Reson Imaging 1987:5:483-492 Lundbom N, Brown AD, Koenig SH, Lansen TA, Valsamis MP, Kasoff 55. Magnetic field dependence of l/T, of human brain tumors: correlations with histology. Invest Radiol 1990;25: 1 197-1 205 Beat PT, Hazlewood CF. Distinction of the normal, preneoplastic, and neoplastic states by water proton NMR relaxation times. In: Partain CL, Price AR, Patton JA, Kulkami MV, James AE, eds. Magnetic resonance
STEEN
264
35.
Downloaded from www.ajronline.org by 117.253.209.72 on 11/09/15 from IP address 117.253.209.72. Copyright ARRS. For personal use only; all rights reserved
36.
37.
38. 39.
40.
41.
imaging, vol. 2. Philadelphia: Saunders, 1988:1710-1734 Le Bihan 0, Breton E, Lallemand 0, Grenier P, Cabanis E, Laval-Jeantet M. MR imaging of intravoxal incoherent motions: application to diffusion and perfusion in neurologic disorders. Radiology 1986;161 :401-407 Le Bihan 0, Breton E, Lailemand 0, Aubin M-L, Vignaud J, Laval-Jeantet M. Separation of diffusion and perfusion in intravoxal incoherent motion MA imaging. Radiology 1988:168:497-505 Moonen CTW, van ZijI PCM, Frank JA, Bihan 0, Becker ED. Functional magnetic resonance imaging in medicine and physiology. Science 1990;250:53-61 Le Bihan 0. Magnetic resonance imaging of perfusion. Magn Reson Med 1990;14:283-292 Moseley ME, Cohen V. Kucharczyk J, et al. Diffusion-weighted MA imaging of anisotropic water diffusion in cat central nervous system. Radiology 1990;176:439-445 Moseley ME, Cohen Y, Mintorovitch J, et al. Early detection of regional cerebral ischemia in cats: comparison of diffusionand Trweighted MAI and spectroscopy. Magn Reson Med 1990;14:330-346 Henkelmann AM. Diffusion-weighted MR imaging: a useful adjunct to clinical diagnosis or a scientific cunousity? AiR 1990;155: 1066-1068
LIST OF BOOK
AJA:158,
42.
43.
44. 45. 46.
47.
epidermoid tumors. AiR 1990;155: 1059-1 065 Henkelman AM. Does IVIM measure classical perfusion? Magn Reson Med 1990;16:470-475 Haacke EM, Tkach JA. Fast MA imaging: techniques and clinical applications. MR 1990;155:951-964 Hooper J, Rajan 55, Rosa L, Le Bihan 0. Application of diffusion imaging to monitor tumor growth and response to chemotherapy. In: Book of abstracts, vol. 1 . Society of Magnetic Resonance in Medicine. Ninth annual meeting. Berkeley. CA: Society of Magnetic Resonance in Medicine, 1990:371 Steen AG, Tamargo AJ, Brem H, Glickson JO, Wehrle JP. In vivo 3P nuclear magnetic resonance spectroscopy of rat 9L gliosarcoma treated with BCNU: dose-response of spectral changes. Magn Reson Med
1989;1 1:258-266
REVIEWS
282
Head and Neck Imaging, 2nd ed. Som PM, Bergeron RT,
308
Stone Therapy
320
The Discipline of Curiosity.
eds
F, Miller K, Rassweller
Science in the World.
Groen
J, eds
J, Smit
E, Eijsvooge!
J, eds
326 346
352 368
388 398
412 426 442
DiagnostIc
Radiology 1990. Gooding CA, ed and Histologic Pathology of Nontumorous Diseases of Bones and Joints, vols. I and 2. Mllgram JW Errors in Chest Radiography. Viamonte M Jr Understanding MRI. Newhouse JH, Wiener JI Merrill’s Atlas of Radiographic Positions and Radiologic Procedures, vols. 1-3, 7th ed. Balllnger PW Nuclear Radiology (Fourth Series). Test and Syllabus. Thrall JH, section ed The Radiologic Clinics of North America. Radiographic Manifestions of Cornmon Congenital Anomalies. Oh KS, guested RSNA Today Video, Vol. 5, No. 1. The Radiological Society of North America RadiOIOgIC
Computers Medicine
eds
in Diagnostic Report
Radiology.
No 62.) Robertson
1992
Mintorovitch J, Moseley ME, Chileuitt L, Shimizu H, Cohen V. Weinstein PA. Comparison of diffusionand T2-weighted MRI for the early detection of cerebral ischemia and reperfusion in rats. Magn Reson Med 1991 18: 39-50 Tsuruda JS, Chew WM, Moseley ME, Norman D. Diffusion-weighted MA imaging of the brain: value of differentiating between extraaxial cysts and
AND VIDEOTAPE
in UrolOgy. Eisenberger
February
(The Institute
of Physical Sciences
J, Wankllng PF, Faulkner K, Holubinka
in MA,
