Session 7
3D MRA Visualization and Artery-Vein Separation
Using Blood-Pool Contrast Agent MS-3251
Tianhu Lei, PhD, Jayaram K. Udupa, PhD, Punam K. Saha, PhD, Dewey Odhner, MS
Richard Baum, MD, Satish K. Tadikonda, PhD, E. Kent Yucel, MD
RATIONALE AND OBJECTIVES
Magnetic resonance angiography (MRA) is established
as an important complementary technique to conventional
angiography, and contrast– enhanced MRA (CE-MRA)
offers even higher contrast between the vascular lumen
and surrounding structures. MS-325 is a gadolinium-based
MR contrast agent designed specifically for blood-pool
imaging, or MRA, and is the only gadolinium-based intravascular contrast agent undergoing trials in humans. MS325 provides excellent vascular and selective arterial enhancement during dynamic MRA. The long blood residence time also allows acquisition of steady-state images
of the arteries and veins with excellent spatial resolution
(1,2).
With the increasing use of CE-MRA, venous contamination of arterial images becomes a common concern (ie,
venous enhancement may confound the visualization of
arteries). Currently available viewing techniques, such as
targeted maximum intensity projection, multiplanar reformation, and “fly through” used in virtual endoscopy, can
be used only to minimize this problem but not to solve it
(3,4). These techniques are also time intensive and require
more operators. Although the ability to acquire dynamic
images may facilitate artery-vein separation by providing
an artery mask that can be applied to steady-state images,
this approach requires motion correction and image registration between dynamic and steady-state images.
Acad Radiol 2002; 9(suppl 1):S127–S133
1From the Department of Radiology, University of Pennsylvania, 4th Floor,
Blockley Hall, 423 Guardian Dr, Philadelphia, Pa 19104-6021 (T.L., J.K.U.,
P.K.S., D.O., R.B.), and EPIX Medical, Inc, Cambridge, Mass (S.K.T.,
E.K.Y.). Supported by a contract from EPIX Medical, Inc. Address correspondence to J.K.U.
©
AUR, 2002
The separation of artery and vein is of significant importance to correctly diagnose and treat peripheral vascular diseases. The strategies for artery-vein separation include both acquisition methods and postprocessing techniques. Among the current developments, acquisition
methods include phase-contrast and time-resolved acquisition approaches (5–7), and postprocessing techniques
cover correlation analysis and graph searching methods
(8 –10). The shortcomings of these approaches are the
limitations of their applications and the costs. For instance, phase-contrast acquisition approaches (5) are limited to cases in which the blood flow directions in artery
and vein are opposite to each other. In the time-resolved
acquisition approaches (6,7), the image must be acquired
during the first pass of a contrast agent or accomplished
with cardiac gating. Correlation analysis (8) requires
seven or eight MRA data sets in a single breath hold for
a three-dimensional (3D) angiogram of the lung. In graph
searching approaches (9), the node costs require 3D edge
strength and a model of preferred branch direction. The
enhanced artery visualization method (10) is limited to
the segmentation of a small number of the main overlapping veins in the peripheral vasculature. Clearly, a more
general approach for artery-vein separation is desirable.
MATERIALS AND METHODS
Principle
We have developed a segmentation approach for separating artery and vein and a shell-rendering method for
visualizing vascular structures (11,12). The anatomic separation uses fuzzy connected object delineation principles
and algorithms (13,14). In fuzzy connected object definition, the notion of an “object” is captured by how the
image elements “hang together” relative to each other in
S127
LEI ET AL
the image. “Hanging-togetherness” is described formally
through the fuzzy topologic notion of fuzzy connectedness. To every possible pair (v1, v2) of image elements, a
“strength of connectedness,” which is a number between
0 and 1, is assigned. This strength is determined to be the
largest of the strengths of all possible connecting paths (a
“path” is simply a sequence of nearby image elements)
between v1 and v2. The strength of a particular path is
defined as the smallest “affinity” along the path between
pairwise nearby image elements. Affinity between two
image elements is expressed as a function of the distance
between the elements and the degree of similarity of a
number of image intensity properties (such as actual intensity, its rate of change) estimated at the two elements.
In short, a fuzzy connected object is defined based on
both intensity criteria and topologic criteria, both of
which are handled in a fuzzy setting. Practical algorithms
based on dynamic programming have been developed for
extracting fuzzy connected objects in 2-, 3-, and higherdimensional images. Their effectiveness has been proved
in several medical applications, such as routine quantification of multiple sclerosis lesions via multiprotocol MR
imagery (15–20), MRA (14), independent clutter-free display of hard- and soft-tissue structures (including detailed
neurovascular structures) from computed tomography
(CT) data for craniomaxillofacial surgery (21), and, recently, in mammographic density quantification (22).
Method
The first step of this separation process is the segmentation of the entire vessel structure from the background
and other clutter of the CE-MRA image via absolute
fuzzy connectedness (22). The object of interest here is
the entire vessel, and one set of “seed” image elements is
required. In the earlier stage of this research, we used
several simple intensity-based features (original intensity,
gradient magnitude) locally computed at every image element. Given the feature vector associated with each of a
pair of image elements, their affinity is determined as a
function of these two vector variables. Multivariate
Gaussian and ramp functions were used. It is clear that
this formulation is defined on fixed neighborhood structure to capture affinity between two image elements.
Therefore, the affinity is really independent of local structure size and can become sensitive to noise. We use a
simple but effective method that formulates fuzzy affinity
based on the local structure size—“scale.” By using scalebased affinity (23), local hanging-togetherness is defined
S128
Academic Radiology, Vol 9, Suppl 1, 2002
over a neighborhood the size of which varies with the
size of the local structure.
The second step is to separate artery from vein within
this entire vessel structure by relative fuzzy connectedness
(24). Let artery and vein be the object of interest and the
co-object in the image, respectively. Two sets of “seed”
image elements are specified inside each of these objects.
The strength of connectedness of every image element
with respect to each of these “seed” image elements is
then determined using a different affinity relation for each
of the objects. An image element is considered to belong
to that object (or co-object) for which the strength of connectedness with respect to its “seed” image elements is
the highest. The image domain is thus partitioned into
two regions, each representing an object (or a co-object).
In the fuzzy description of each object, outside the respective region, the image elements are assigned a zero
value, and inside, either the original image intensity or
the winning strength of connectedness is assigned. Scalebased affinity is also used in this relative connectedness
analysis.
The separation of artery-vein structures is performed in
an iterative fashion. The small regions of the bigger aspects of artery and vein are separated in the initial stage.
Some of the image elements in these already separated
regions are then used as “seeds” in a second stage for
separation of finer aspects of the vessels, and so on. Further regions are added in subsequent stages so that a
fairly complete separation up to the level of fine vessels
is achieved.
Procedure
The implementation of the previously described twostep process is based on 3DVIEWNIX—an open, transportable, multidimensional, multimodality, multiparametric imaging software system—and its library of 3D imaging functions, designed, developed, and maintained by the
Medical Image Processing Group, University of Pennsylvania (25).
In the step of vessel segmentation, “seed” image elements are specified on some typical segments of the vessels in the CE-MRA images. Then a map of absolute
fuzzy connectedness that indicates how the image elements in the image hang together to form vessels is
formed. By thresholding this connectedness map, a binary
image is generated. By using this binary image to mask
the original CE-MRA images, an object (vessel-only) image is created.
Academic Radiology, Vol 9, Suppl 1, 2002
Figure 1.
3D MRA VISUALIZATION AND ARTERY-VEIN SEPARATION
Selection of “seed” points from the shell-rendered image.
Figure 2. Paths created by central line
algorithm.
In the step of artery-vein separation, after a set of
“seed” image elements is created inside each artery and
vein, respectively, two individual maps of relative fuzzy
connectedness that are the binary images for artery and
vein are formed. The iteration is stopped once there is no
further change of image elements in these two maps.
Then, by using these two binary images to mask the vessel-only image, the object and co-object (artery and vein)
images are generated.
“Seed” selection in vessel segmentation and artery-vein
separation is the only human interaction procedure required by this entire approach. In vessel segmentation, a
human operator “paints” a few “seed” image elements. In
artery-vein separation, a pair of “seed” image elements
(the start and the end) for each branch to be included is
specified by the 3D shell-rendered image (Fig 1). Then a
central line algorithm is applied on a distance image,
which is created by applying a 3D distance transformation
to the binary image of the entire vessel structure to obtain
“central” paths inside the vessels (Fig 2). All image elements on these paths will serve as “seed” image elements
in the separation. The central line algorithm chooses the
local maximum as a cost function for creating best paths.
Fuzzy objects (vessel, artery, and vein) are represented
by a data structure called a “shell,” and shell rendering is
used for 3D visualization of vascular information (26).
For distinguishing arteries and veins, the shell renditions
can be colored differently for the two objects in a composite display, or the objects can be turned on and off
selectively. Shell rendering retains object heterogeneity,
and, therefore, object composition can be more accurately
depicted. Two-dimensional object images (artery and vein)
in the slice can also be colored differently and superimposed
in one display by using the 3DVIEWNIX library.
RESULTS
This approach has been applied to EPIX Medical’s
CE-MRA data (MS-325 [AngioMARK; EPIX Medical,
Cambridge, Mass]). A total of 109 original CE-MRA data
sets (from 28 patients) from three university hospitals
have been processed. In our study, CE-MRA data were
acquired during the precontrast, dynamic (arterial phase),
and early and late postcontrast steady state, with various
resolutions and orientations, stored in DICOM (Digital
Imaging and Communications in Medicine) format. In all
of our case studies, unified parameter settings were
adopted without requiring per-study adjustments. 3D images and movies of objects (vessels, arteries, and veins)
have been created for each study. Shell renditions are
colored differently for the two objects in a composite dis-
S129
LEI ET AL
Academic Radiology, Vol 9, Suppl 1, 2002
Figures 3– 6. (3) Image of vessel of example 1. (4) Image of arteries of example 1.
(5) Image of veins of example 1. (6) Images of arteries and veins of example 1.
play, or the objects can be turned on and off selectively.
All of our case studies were performed on a personal
computer (PC)(Gateway 2000 [300-MHz/256-MB memory]; Pentium) under the Linux operating system. The
whole procedure, which includes vessel segmentation,
artery-vein separation, and shell rendering, is completed
in a few minutes per study.
Three examples for vessel segmentation and arteryvein separation are reported here to demonstrate the performance of this approach. In these examples, the early
postcontrast steady-state images were used. The segmented vessel images for each example (from the belly to
the feet) are shown in Figures 3, 7, and 11; and the separated artery and vein images for each example are shown
in Figures 4, 8, and 12 and Figures 5, 9, and 13, respectively. The composite displays of artery (in red) and vein
(in blue) for each example are given in Figures 6, 10, and
14. Images shown are selected from 90-frame movies.
S130
DISCUSSION
CE-MRA has fundamentally changed angiography and
is established as an important complementary technique to
conventional angiography. However, this technique requires advanced vascular visualization as part of clinical
protocols, and its utility can be significantly enhanced for
evaluation and treatment of diseases when artery and vein
structures are further separated. Our technique, by combining the strengths of fuzzy connected object definition,
object separation, and shell rendering, provides a highquality 3D display of vascular information in a single
CE-MRA image. This technique seems to be the only
image-processing approach available for artery and vein
separation in a general scope (3–10,27–29).
We have presented an image-processing approach for
artery-vein separation and for their 3D visualization. The
results show that this approach is quite automated, unsu-
Academic Radiology, Vol 9, Suppl 1, 2002
3D MRA VISUALIZATION AND ARTERY-VEIN SEPARATION
Figure 7–10. (7) Image of vessel of Example 2. (8) Image of arteries of Example 2.
(9) Image of veins of Example 2. (10) Images of arteries and veins of Example 2.
pervised, and robust. We are in the process of refining
this development. The following issues will be addressed
in the near future.
Toward a fully automated and entirely unsupervised
approach, the “seed” selection will be improved. Because
the intensity of image elements in the vessels of MS-325
CE-MRA data is far above the 95th percentile of the volume histogram, it is possible to directly apply a higher
threshold on the original CE-MRA images to obtain some
vessel regions (instead of painting a few “seed” image
elements) as the bank of “seed” image elements for vessel
segmentation, thus eliminating one user interaction step.
Both vessel segmentation and artery-vein segmentation
require the computation of connectedness. We have recently developed, independent of this research, a “queue”
method in connectedness computation. After the affinity
is computed, at the stage of tracking, the queue method
arranges the strengths of connectedness in a descending
order; then it removes the image elements that have the
highest strength. This operation is repeated until the
tracking is over. It has been shown that this method can
shorten connectedness computation time by a factor of
about 20 because it cuts off the process of revisiting the
image elements with the highest strength of connectedness. For example, the segmentation of the brain parenchyma in a 256 ⫻ 256 ⫻ 60 volume can be accomplished
in about 5–10 seconds on a 300-MHz Pentium PC. The
queue method will be included in the processes of computing connectedness in both vessel segmentation and
artery-vein separation, making all processes interactive.
S131
LEI ET AL
Academic Radiology, Vol 9, Suppl 1, 2002
Figure 11–14. (11) Image of vessel of Example 3. (12) Image of arteries of Example 3.
(13) Image of veins of Example 3. (14) Images of arteries and veins of Example 3.
The accuracy of separation and the level of the vessel
tree up to which meaningful separation can be achieved
will be determined through manual segmentation of a subset
of MS-325 CE-MRA data and comparison with the computer-produced results. A more general subjective accuracy
evaluation procedure will be developed and tested for both
vessel segmentation and artery-vein separation.
REFERENCES
1. Grist TM, Korosec FR, Peters DC, et al. Steady-state and dynamic MR
angiography with MS-325: initial experience in humans. Radiology
1998; 207:539 –544.
2. Lauffer R, Parmelee D, Dunham S, et al. MS-325: albumin-targeted
contrast agent for MR angiography. Radiology 1998; 207:529 –538.
3. Davis C, Ladd M, Romanowski B, Wildermuth S, Knoplioch J, Debatin
J. Human aorta: preliminary results with virtual endoscopy based on
three-dimensional MR imaging data sets. Radiology 1996; 199:37– 40.
S132
4. A.T.F. on Appropriateness Criteria. American College of Radiology appropriateness criteria. Reston, Va: American College of Radiology,
1995.
5. Bluemke DA, Darrow RD, Gupta R, Tadikonda SK, Dormoulin CL. 3D
contrast enhanced phase contrast angiography: utility for arterial/venous segmentation (abstr).ISMRM 1999; Proceedings 2:1237.
6. Mazaheri Y, Carroll TJ, Mistretta CA, Korosec FR, Grist TM. Vessel
segmentation in 3D MR angiography using time resolved acquisition
curves (abstr). ISMRM 1999; Proceedings 3:2181.
7. Foo TK, Ho V, Hood MN, et al. A novel method for MR arterial and
venous discrimination using gated phase contrast and vein selection
(abstr). ISMRM 1999; Proceedings 3:2182.
8. Bock M, Schoenberg SO, Flomer F, Gran A, Strecker R, Schad LR.
Artery-vein separation in 3D contrast enhanced pulmonary MRA using
correlation (abstr). ISMRM 1999; Proceedings 1:486.
9. Sonka M, Stetacik R, Tadikonda S. Feasibility of automated separation
of arteries and veins using a graph searching (abstr). ISMRM 1999;
Proceedings 3:2183.
10. Niessen W, van Swijndregt AM, Elsman B, Wink O, Viergever M, Mali
W. Enhanced artery visualization in blood pool MRA: results in the peripheral vasculature. IPMI Proc 1999; 1613:340 –345.
Academic Radiology, Vol 9, Suppl 1, 2002
11. Lei T, Udupa J, Saha P, Odhner D. 3D MR angiographic visualization
and artery-vein separation. SPIE Proc 1999; 3658:52–59.
12. Lei T, Udupa J, Saha P, Odhner D, Baum R. Artery-vein separation
using MR angiographic data: in 25 patients (abstr). ISMRM 1999; Proceedings 3:1235.
13. Udupa JK, Samarasekera S. Fuzzy connectedness and object
delineation: theory, algorithm, and applications in image segmentation.
Graphical Models Image Processing 1996; 58:246 –261.
14. Udupa JK, Odhner J, Tian J, Holland G, Axel L. Automatic clutter-free
volume rendering for MR angiography using fuzzy connectedness.
SPIE Proc 1997; 3034:114 –119.
15. Samarasekera S, Udupa JK, Miki Y, Grossman RI. A new computerassisted method for enhancing lesion quantification in multiple sclerosis. J Comput Assist Tomogr 1997; 21:145–151.
16. Miki Y, Grossman RI, Udupa JK, et al. Differences between relapsing
remitting and chronic progressive multiple sclerosis as determined
with quantitative MR imaging. Radiology 1999; 210:769 –774.
17. Udupa JK, Wei L, Samarasekera S, Miki Y, van Buchem MA, Grossman RI. Multiple sclerosis lesion quantification using fuzzy connected
principles. IEEE Trans Med Imaging 1997; 16:598 – 609.
18. Miki Y, Grossman RI, Samarasekera S, et al. Clinical correlation of
computer assisted enhancing lesion quantification in multiple sclerosis. Am J Neuroradiol 1997; 18:705–710.
19. Phillips M, Grossman RI, Miki Y, et al. Comparison of t2 lesion volume
and magnetization transfer ratio histogram analysis and atrophy and
measures of lesion burden in patients with multiple sclerosis. AJNR
Am J Neuroradiol 1998; 19:1055–1060.
3D MRA VISUALIZATION AND ARTERY-VEIN SEPARATION
20. Miki Y, Grossman RI, Udupa JK, et al. Relapsing-remitting multiple
sclerosis: longitudinal analysis of MR images—lack of correlation between changes in T2 lesion volume and clinical findings. Radiology
1999; 213:395–399.
21. Udupa JK, Tian J, Hemmy D, Tessier P. A Pentium-based craniofacial
3D imaging and analysis system. J Craniofac Surg 1997; 8:333–339.
22. Saha PK, Udupa JK. Scale-based fuzzy connectivity: a novel image
segmentation methodology and its validation. SPIE Proc 1999; 3661:
246 –257.
23. Saha PK, Udupa JK, Odhner D. Scale-based fuzzy connected image
segmentation: theory, algorithm, and validation. Comput Vision Image
Understanding 2000; 77:145–174.
24. Udupa JK, Saha PK, Lotufo A. Fuzzy-connected object definition in
images with respect to co-object. SPIE Proc 1999; 3661:236 –245.
25. Udupa JK, Odhner D, Samarasekera S, et al. 3DVIEWNIX: an open,
transportable, multidimensional, multimodality, multiparametric imaging software system. SPIE Proc 1994; 2164:58 –73.
26. Udupa JK, Odhner D. Shell rendering. IEEE Comput Graphics Applications 1993; 13:58 – 67.
27. Gatti C, Accomazzi V, Chann R, Stapleton S, Done E. Vascular visualization using IAP. SPIE Proc 1999; 3658:117–124.
28. Kobashi S, Hata Y, Tokimoto Y, Ishikawa M. Automatic segmentation
of blood vessels from MR angiography volume data. SPIE Proc 1999;
3661:968 –976.
29. Smedby O, Svensson S, Lofstrand T. Greyscale connectivity concept
for visualizing MRA and CTA volumes. SPIE Proc 1999; 3658:212–220.
S133
Improvement of Vascular Signal Intensity in
Contrast-enhanced MRA with Gd-BOPTA:
Comparison with Gd-DTPA1
Paolo Pavone, MD, Carlo Catalano, MD, Filippo Cademartiri, MD, Giacomo Luccichenti, MD, Roberto Passariello, MD
RATIONALE AND OBJECTIVES
The need to reduce the acquisition time in contrast
material– enhanced magnetic resonance angiography
(MRA) requires a reduction of the signal-to-noise ratio
(SNR). The SNR can be raised by administering double
doses of gadopentetate dimeglumine (Gd-DTPA). Our
purpose was to evaluate the possibility of performing contrast-enhanced MRA with a high-relaxivity paramagnetic
contrast agent, that is, gadobenate dimeglumine (GdBOPTA).
MATERIALS AND METHODS
Twelve patients with suspected carotid artery stenosis
were examined on a 1.5-T magnet (Vision Plus; Siemens,
Erlangen, Germany) using a time-resolved MRA seAcad Radiol 2002; 9(suppl 1):S134
1From Department of Radiology, University of Parma, Via Gramsci 14,
43100 Parma, Italy (P.P., F.C., G.L.), and Department of Radiology, University of Rome (C.C., R.P.). Address correspondence to P.P.
©
AUR, 2002
S134
quence. In all patients, MRA was performed twice within
24 hours by using the same sequence parameters, injecting (at the same flow rate of 2 mL/s) 15 mL of Gd-DTPA
(Magnevist; Berlex Laboratories, Wayne, NJ, USA) in the
first study and 10 mL of Gd-BOPTA (MultiHance; Bracco,
Milano, Italy) in the second study.
RESULTS
With Gd-BOPTA, signal intensity showed a mean increase of 1.5 times the intensity with Gd-DTPA, and the
SNR increased 1.4 times compared with Gd-DTPA. The
differences between the two groups were significant (P ⬍
.05) regarding both signal intensity and SNR.
CONCLUSION
Gd-BOPTA– enhanced MRA can provide superior vascular signal intensity and SNR, as compared with GdDTPA, due to its higher relaxivity, even at lower doses.
Therefore, Gd-BOPTA can be used to reduce costs and
improve vascular enhancement.
Comparative Studies on the Efficacy of MRI
Contrast Agents in MRA1
Hanns-Joachim Weinmann, PhD, Hans Bauer, PhD, Wolfgang Ebert, PhD, Thomas Frenzel, PhD
Bernd Radüchel, PhD, Johannes Platzek, PhD, Heribert Schmitt-Willich, PhD
RATIONALE AND OBJECTIVES
The objective was to analyze the potential of various
para- and superparamagnetic agents as contrast agents for
magnetic resonance angiography (MRA).
MATERIALS AND METHODS
Values for the signal intensity (SI) of plasma samples
spiked with different concentrations (0.005–25 mmol/L)
of various paramagnetic agents (gadopentetate dimeglumine [Magnevist; Schering AG, Berlin, Germany]; gadoxetic acid [Eovist; Schering AG]; Gadomer-17 [Schering
AG]; gadobenate dimeglumine [MultiHance; Bracco, Milan, Italy]; MS-325 [EPIX Medical, Cambridge, Mass);
and the ultrasmall superparamagnetic iron oxide [USPIO],
SH U 555C [Schering AG]) were measured at 1.5 T (Allegra; Siemens AG, Erlangen, Germany). A three-dimensional (3D) MRA sequence (4.7/1.89 [repetition time
msec/echo time msec], 25° flip angle) was used. The SI
values of the maximum intensity projections (MIP) were
measured, and the apparent T1 and T2* relaxivities were
calculated from the SI-versus-concentration curves using
the known dependence of the SI from sequence parameters and relaxation times (1).
MRA studies using a superconducting MR imager (Allegra; Siemens AG) were performed in rabbits after intravenous injection of two dose regimens. The contrast agents
were injected after starting the imaging sequence to achieve
optimum bolus concentrations during the k space of the sequence. Additional images were taken at 3, 10, and 30 minAcad Radiol 2002; 9(suppl 1):S135–S136
1 From the Research Laboratories, Schering AG, Müllerstrasse 178,
D-13342 Berlin, Germany. Address correspondence to H.J.W.
©
AUR, 2002
utes after dosing. A dose of 0.2 mmol/kg was used for all
agents to allow a comparison of the various agents with the
currently performed practice with extracellular gadolinium
chelates (eg, Magnevist). A second dose was adjusted to the
calculated T1 relaxivity of Magnevist in plasma.
RESULTS
All para- and superparamagnetic agents elicited similar
SI-versus-concentration profiles. After reaching maximum
SI, the values dropped with increasing concentrations (Figure). Maximum SI was reached at concentrations between 3
and 10 mmol/L, depending on the contrast agent used (Table). The gadolinium compounds produced comparable maximum SI values. The USPIO displayed a somewhat different
profile. Maximum SI was about 60% of the values measured
for gadolinium compounds. The highest SI of this USPIO
was obtained at about 3 mmol/L.
The low-molecular-weight gadolinium compounds elicited a T1 relaxivity of 3.5–5.2 L mmol⫺1 sec⫺1; the compounds with a larger molecular weight or strong plasma
protein binding exhibited T1 relaxivities of about 12 L
mmol⫺1 sec⫺1. The T2* relaxivities of all gadolinium
compounds were similar and significantly larger than the
T1 relaxivities. The T2* relaxivity of the iron agent displayed a much larger value of about 70 L mmol⫺1 sec⫺1.
The high dose of 0.2 mmol/kg yielded excellent angiograms. The arterial phase occurred during the first 12 seconds of imaging; later images displayed both arterial and
venous enhancement. The bolus MRA images immediately after the injection of the gadolinium agents were
superior to the USPIO images. The strong susceptibility
effect of the USPIO diminished the image quality during
the very first imaging after bolus injection.
MultiHance and Eovist exhibit similar enhancement characteristics to Magnevist. The blood enhancement declined
S135
WEINMANN ET AL
Academic Radiology, Vol 9, Suppl 1, 2002
Parameters of SI versus Concentration Curves Obtained by MRI Measurements of Plasma Tubes Spiked
with Various Concentrations of Contrast Agents
Agent
Relative SImax (au)
Concentration at SImax
(mmol L⫺1)
T1 (L mmol⫺1 sec⫺1)
T2* (L mmol⫺1 sec⫺1)
Magnevist
Eovist
MultiHance
MS-325
Gadomer-17
SH U 555 C
700
750
700
675
675
400
10
7.5
7.5
5
5
3.4
3.5
5.2
4.4
11
11
6.4
22
26
28
28
28
72
Note.—3D MRA (4.7, 1.89, 25° flip angle). au ⫽ arbitrary units.
rapidly due to extravasation and excretion. A longer lasting
venous and arterial enhancement was seen with Gadomer-17
and MS-325, due to the particular pharmacokinetics of these
agents.
The USPIO demonstrated good enhancement even at
30 minutes after injection. However, the arterial MRA
images do not display significant differences if the dose is
adjusted to the T1 relaxivity of Magnevist.
CONCLUSION
Gadolinium complexes with plasma protein binding or
increased molecular weight elicit higher T1 relaxivities than
Magnevist. The highest values are obtained with Gadomer-17 and MS-325, the agent with the strongest protein
binding. The T1 relaxivities of the latter agents are lower
than previously reported for 0.47 T (2). The suboptimal correlation times of these gadolinium complexes at 1.5 T
weaken their T1 relaxivities (3) in such a way that differences seen at 0.47 T between Gadomer-17 and MS-325 disappear. Complexes with weak protein binding, such as MultiHance and Eovist, exhibit a somewhat stronger T1 relaxivity than Magnevist. Strong T2* relaxivities influence the SI
even at that extremely short echo time. These effects cause a
strong SI reduction at high concentrations, most prominent
with USPIO. The T2* relaxivities of the gadolinium complexes do not differ significantly because of the identical
magnetic moments of the compounds.
The T2*/T1 ratio determines the maximum achievable
SI to some extent. Practically no maximum SI (SImax)
differences are observed among the various gadolinium
compounds. In other words, similar maximal contrast enhancement is achievable independent of the gadolinium
chelate used. Because of the disadvantageous large
T2*/T1 ratio, USPIO compounds are less suitable for arterial bolus contrast-enhanced MRA than gadolinium
S136
SI (in arbitrary units on x axis) versus concentration of Gadomer-17 and the USPIO, SH U 555 C, in plasma samples spiked
with various concentrations of the contrast agents. Data were obtained by MRI measurements (3D MRA) (4.7, 1.89, 25° flip angle).
complexes. To avoid dominating T2* effects, lower concentrations of such agents are beneficial. Consequently,
delayed images with USPIO do not display large susceptibility effects because the concentrations are about five
times lower than in the bolus peak phase.
The imaging experiments demonstrate the favorable
characteristics of gadolinium agents in the early bolus
phase, whereas the advantages of large complexes (MS325 and Gadomer-17) and USPIO compounds are most
evident in the distribution phase.
REFERENCES
1. Haase A. Snapshot FLASH MRI: applications to T1, T2, and chemicalshift imaging. Magn Reson Med 1990; 13:77– 89.
2. Lauffer RB, Parmelee DJ, Dunham SU, et al. MS325: albumin-targeted
contrast agent for MR angiography. Radiology 1998; 207:529 –538.
3. Lauffer RB, Brady TJ. Preparation and water relaxation properties of
proteins labeled with paramagnetic metal chelates. Magn Reson Imaging 1985; 3:11–16.
Arterial Concentration Profiles of Two Blood Pool
Agents and Gd-DOTA after Intravenous
Injection in Rabbits1
Claire A. Corot, PhD, Xavier Violas, Philippe Robert, MS, Marc Port, PhD
RATIONALE AND OBJECTIVES
Gadolinium-enhanced magnetic resonance angiography
(MRA) is a technique influenced by the injection parameters of the contrast agent and the acquisition parameters.
Different techniques have been proposed to standardize
MRA protocols (1).
Because there is no linear relationship between the
contrast and the gadolinium blood concentrations (2), it is
difficult with imaging modalities to analyze the contribution of each injection parameter and also to compare contrast agents. An experimental model in rabbits was developed to mimic an MRA protocol to determine the gadolinium blood concentrations during the time immediately
following injection. The influences of the injection parameters (injection rate, duration of injection, volume of injection, dose) were tested, and two blood pool agents
(P760, P717) were compared to the nonspecific agent gadoterate meglumine (Gd-DOTA) in similar conditions.
P717, a Gd-DOTA– dextran derivative, is a slow-clearance blood pool agent (SC-BPA); and P760, a macromolecular Gd-DOTA derivative, is a low-diffusion agent
(LDA), as defined previously (3).
MATERIALS AND METHODS
Products
Gd-DOTA (0.5 M Dotarem; Guerbet, Roissy, France)
was used as an example of a nonspecific agent. To test
Acad Radiol 2002; 9(suppl 1):S137–S139
1 From Guerbet Research, BP 50400, 95943 Roissy CDG, France. Address
correspondence to C.A.C..
©
AUR, 2002
the influence of the injection rate, Gd-DOTA at a dose of
300 ␮mol/kg was injected in the same volume using injection rates of 0.08, 0.13, 0.25, 0.5, and 1 mL/sec. Under
these conditions, the total duration of the injection was
25, 14, 8, 4, and 1.25 seconds, respectively.
To allow the same injection parameters as with P760,
Gd-DOTA was diluted to 0.033 M and injected at a dose
of 30 ␮mol/kg (duration of injection, 4.2 seconds; injection rate, 0.5 mL/sec).
P760 (0.033 M, LDA) and P717 (0.1 M, SC-BPA,
CMD-A2-Gd-DOTA), both research products from Guerbet, were tested at the dose of 30 ␮mol/kg at an injection
rate of 0.5 mL/sec. The duration of injection was 4.2 seconds for P760 (4) and 1.6 seconds for P717 (5).
Pharmacokinetic Experiment
All animal experiments were performed in compliance
with the European Economic Community directive (86/
609/EEC) on animal welfare. Male New Zealand rabbits
(2.5–3.5 kg) were anesthetized by intramuscular injection
of ketamine at 50 mg/kg and xylazine at 0.5 mL/kg. The
femoral artery was catheterized with a 1–1.5-mm catheter
(Vygon, East Rutherford, NJ). The ear marginal vein was
catheterized (Insyte Vialon 22G; Becton Dickinson, Pont
de Clairx, France) and maintained with sticking plaster.
The products were injected through an automatic injector
(Perfuseur Sage M362; Orion Europe, Cambridge, England) using 5-mL syringes (Becton Dickinson, Madrid,
Spain) connected to the ear marginal vein catheter.
Starting at the beginning of the injection, femoral artery blood samples were collected continuously over 30
seconds. Blood was directly transferred from the catheter
into a series of tubes. Samples of approximately 300 ␮L
each were taken every second. After this period, samples
were obtained every 5 seconds for l minute and then at 2
S137
COROT ET AL
Academic Radiology, Vol 9, Suppl 1, 2002
and 5 minutes. All experiments were performed in triplicate, with the exception of the injection rate experiment
(n ⫽ 1, injection rate condition).
Gadolinium Measurements
Blood gadolinium concentrations were measured using
spectroscopic emission atomic inductively coupled plasma
(SEA-ICP) equipment (Spectraspan VII; Perkin Elmer,
Boston, Mass) after water dilution (1/20).
Parameters
The maximum concentration (Cmax) and the time to
Cmax (Tmax) were estimated from the graph of the blood
gadolinium concentrations versus time. These two parameters describe the bolus phase. The postbolus phase was
described by the parameter Ci/C0, which corresponds to
the blood gadolinium concentration measured at time i
after the injection, divided by the theoretical initial concentration C0 at time zero (C0 ⫽ theoretical concentration
based on the injected dose in micromoles per kilogram,
which is instantaneously homogenized in the blood volume of 60 mL/kg).
Figure 1. Arterial gadolinium blood concentrations after the injection of Gd-DOTA at a dose of 0.1 mmol/kg at different injection
rates (0.08, 0.13, 0.25, 0.5, and 1 mL/sec).
RESULTS
For the same injected dose of Gd-DOTA, 300 ␮mol/
kg, the injection rate (and, consequently, the total injected
volume) modifies the bolus peak profile (Fig 1). Increasing the injection rate produces a sharpening of the peak
(Cmax increase, Tmax decrease, peak length decrease). At a
low injection rate, the first pass presents a plateau form.
The Cmax is 10.1, 10.3, 20.1, 22.8, and 25.3 mmol/L, and
Tmax is 31, 20, 12, 11, and 6 seconds at respective injection rates of 0.08, 0.13, 0.25, 0.5, and 1 mL/sec. Concentrations at the postbolus phase are identical for all of the
injection rates at 1 and 5 minutes after injection.
To evaluate the pharmacokinetic contribution of the
blood pool properties in this model, P760, P717, and GdDOTA were injected at the same dose (0.03 mmol/kg).
Under these conditions, the three products showed the
same bolus phase (Fig 2). The earlier Tmax and slightly
higher Cmax for P717 are attributed to the higher concentration of the injected solution (0.1 M) compared to GdDOTA and P760 (0.033 M solution for both in this experiment). At the postbolus phase, there is a clear differentiation in the gadolinium concentration as a function of
the blood pool properties of each agent (Fig 3). Forty-five
seconds after injection, 84% of the P717 molecules are
still present in the blood compartment, whereas 65% of
S138
Figure 2. Arterial gadolinium blood concentrations after the injection of a dose of 0.03 mmol/kg (0.03 mmol gadolinium/kg) of
Gd-DOTA, P760, and P717 at 0.5 mL/sec: bolus phase.
Figure 3. Arterial gadolinium blood concentrations after the
injection of a dose of 0.03 mmol/kg (0.03 mmol gadolinium/kg)
of Gd-DOTA, P760, and P717 at 0.5 mL/sec: postbolus phase.
the P760 molecules and only 51% of the Gd-DOTA molecules remain.
Academic Radiology, Vol 9, Suppl 1, 2002
ARTERIAL GADOLINIUM CONCENTRATIONS IN MRA-LIKE MODEL
CONCLUSION
The maximum concentration is very sensitive to the injection rate and, consequently, to the total duration of the
injection. A decrease in the injection rate leads to a decrease
in the Cmax, the bolus having a plateau form. Thus, the duration of the plateau depends on the total duration of the injection. Conversely, an increase in the injection rate sharpens
the bolus peak and increases the Cmax. The changes in the
bolus profile according to the injection parameters have to
be taken into account for each MRA protocol (6). Substantial changes in the gadolinium concentrations during signal
acquisition induce artifacts (7). Furthermore, the hemodynamic parameters (cardiac output, blood pressure) influence
the bolus profile (Tmax, Cmax, length).
Injected under similar conditions, blood pool agents
P760 and P717 have a comparable profile to Gd-DOTA
during the bolus phase. The site of injection is in the ear
marginal vein, and arterial blood samples are collected
from the femoral artery. This means that no significant
extravasation is observed, although the products traverse
once through the lung, the heart, and large arteries. The
short arrival time of the bolus peak (ie, Tmax ⬍ 7 seconds)
indicates that the product measured from this arterial site
remains mainly in the macrocirculation. On the basis of
these data, blood pool agent properties in this protocol
have no significant advantage over nonspecific agents in
the bolus phase, with the bolus phase being mainly used
for breath-hold techniques (8).
Conversely, at the postbolus phase, the concentrations
of the blood pool agents are higher than with Gd-DOTA.
This is due to the absence of extravasation or to delayed
extravasation and renal excretion compared to Gd-DOTA.
Forty-five seconds after injection, approximately 50% of
the Gd-DOTA has left the blood compartment, compared
to only 35% of P760 and 16% of P717. It can be assumed for P717 that due to the polydispersity of the dextran derivative (I ⫽ 1.66), a portion of small molecules
may extravasate, which could explain why the compound
is not fully intravascular at this time (5). P760 is not truly
a blood pool agent because even at early times after injection, a small portion of the compound has already left
the blood compartment. Due to its higher molecular volume, P760 presents a lower rate of interstitial diffusion
than nonspecific agents (4).
The increase in the arterial blood concentration with
blood pool agents in the postbolus phase leads to an increase in the temporal imaging window. This is of particular interest for territories like the coronary arteries,
where a longer time interval is required (9,10). These data
corroborate imaging results obtained with P717 and P760,
where a significant enhancement was observed at later
times after the contrast agent injection (11–13).
This model was developed to mimic an MRA protocol
in the rabbit to analyze the influence of different parameters on blood gadolinium concentrations, which will determine signal efficacy. Two phases are identified: the
bolus phase and the postbolus phase.
The injection parameters (injection rate, total duration
of injection, volume, concentration of the solution, dose)
strongly influence the pattern of the bolus phase. The
postbolus phase is sensitive only to the injected dose.
In this model, blood pool agents present a similar bolus profile to Gd-DOTA, whereas in the postbolus phase,
the intravascular restriction of these blood pool agents
leads to higher blood concentrations than Gd-DOTA.
REFERENCES
1. Marchal G, Bosmans H. Contrast-enhanced magnetic resonance angiography. In: Thomsen HS, Muller RN, Mattrey RF, eds. Trends in
contrast media. New York, NY: Springer-Verlag, 1999; pt 23:267–283.
2. Westenberg J, Wasser M, Van der Geest RJ. Scan optimization of Gd
contrast enhanced three dimensional MRA of peripheral arteries with
multiple bolus injections and in vitro validation of stenosis quantification. Magn Reson Imaging 1999; 17:47–57.
3. Port M, Meyer D, Bonnemain B, et al. P760 and P775: MRI contrast
agents characterized by new pharmacokinetic properties. MAGMA
1999; 8:172–176.
4. Corot C, Port M, Raynal I, et al. Physical, chemical and biological
evaluations of P760: a new gadolinium complex characterized by a
low rate of interstitial diffusion. J Mag Res Imaging 2000; 11:182–196.
5. Corot C, Schaefer M, Meyer D. Physical, chemical and biological evaluations of CMD-A2-Gd-DOTA. Acta Radiol 1997; 5412:91–99.
6. Maki J, Prince M, Chenevert T. Optimizing three-dimensional gadolinium-enhanced magnetic resonance angiography. Invest Radiol 1998;
33:528 –537.
7. Boos M, Scheffler K, Haselhorst R, Reese E, Frohlich J, Bongartz G.
Evaluation of arterial first pass contrast medium dynamics as a function of intravenous injection parameters (abstr). In: Proceedings of the
Seventh Meeting of the International Society for Magnetic Resonance
in Medicine. Berkeley, Calif: International Society for Magnetic Resonance in Medicine, 1999; 1175.
8. Kouwenhoven M. Contrast-enhanced MR angiography. Acta Radiol
1997; 38(suppl 412):57– 67.
9. Grist T, Korosec F, Peters D. Steady-state and dynamic MR angiography with MS325: initial experience in humans. Radiology 1998; 207:
539 –544.
10. Woodard P, Li D, Haacke E, Gropler R. Coronary MR angiography.
MRI Clin North Am 1999; 7:365–378.
11. Marchand B, Douek P, Canet E. Noninvasive evaluation of the pharmacokinetic and relaxivity of new MR contrast agents for MR angiography. MAGMA 1998; 6:84.
12. Kroft LJ, Doornbos J, Van der Geest RJ, Benderbous S, De Roos A. Infarcted myocardium in pigs: MR imaging enhanced with slow-interstitialdiffusion gadolinium compound P760. Radiology 1999; 212:467– 473.
13. Kroft LJ, Doornbos J, Van der Geest RJ, Benderbous S, De Roos A.
Equilibrium phase MR angiography of the aortic arch and abdominal
vasculature with the blood pool contrast agent CMD-A2-Gd-DOTA in
pigs. J Magn Res Imaging 1999; 9:777–785.
S139
Visualization Techniques for Blood Pool Contrastenhanced Magnetic Resonance Angiography1
E. Kent Yucel, MD
The conventional technique for processing time-of-flight
and dynamic arterial phase contrast material– enhanced
magnetic resonance (MR) angiographic images has been
the maximum intensity pixel (MIP) projection algorithm.
This is a powerful technique for creating projectional images from source data that are not very anatomically complex. However, in the case of images obtained using MR
imaging blood pool contrast agents, the MIP technique
does not provide images that are appropriate for most
diagnostic purposes. This is due to the complex overlap
of arterial and venous structures in the projectional images, which makes it nearly impossible to disentangle the
vessels when they lie in close proximity. Therefore, alternative visualization techniques will be required to derive
the full diagnostic benefit of these contrast agents.
The simplest technique to use includes viewing of
source and multiplanar reformatted (MPR) images. Modern workstations permit rapid viewing of these large multislice data sets in a convenient and rapid fashion. MPR
in planes orthogonal to the vessel axis can be especially
useful in assessing cross-sectional vessel measurements,
such as area stenosis, and in clearly separating adjacent
arteries and veins. Currently available software packages
permit the rapid placement of points along the course of a
vascular bundle in two planes with rapid generation of a
set of vessel-orthogonal MPR slices. Despite the power of
these MPR techniques, they are still rather user-intensive
in that the viewer must physically look at a large number
of individual slices, either individually or in a cine format. It would be much more convenient for users to have
access to panoramic viewing of these image sets so the
Acad Radiol 2002; 9(suppl 1):S140 –S142
1 From the Department of Radiology, Brigham & Women’s Hospital, 75
Francis St, Boston, MA 02115. Address correspondence to the author.
©
AUR, 2002
S140
viewer can come to a rapid diagnosis as to the general
state of the vasculature, with dedicated viewing of source
and MPR images reserved for selected areas where higher
precision is needed.
The two basic approaches to this problem are (a) tagging of each image voxel on the basis of some physiologic feature during the acquisition of the image, and
(b) pure postprocessing of high-resolution image data,
relying on spatial relationships of vessels to separate arteries and veins.
In acquisition tagging, a physiologic feature measurable by MR imaging is associated with each voxel. This
feature is then used to identify anatomic structures, such
as arteries and veins, which can be differentiated by this
feature. Images can be displayed by colonization with the
magnitude of the feature displayed as color hue and the
sign (if applicable) associated with different colors (eg,
red or blue). Features that can be used for this purpose
include the time of appearance of contrast, oxygenation,
and flow direction.
With current MR sequences, full three-dimensional
(3D) volumes can be acquired with temporal resolution of
several (⬍10) seconds. This permits clear distinction of
arterial and venous voxels. In one implementation, initial
estimates of the arterial and venous volumes are made
using a two-dimensional (2D) correlation plot. Initially,
the sum of squared deviations between a voxel’s time
values and those of the reference artery and vein curves
are plotted on a 2D plot. Initial estimates of pure arterial
and venous voxels are made from the peaks in the 2D
plot. Temporally ambiguous voxels are assigned based on
spatial connectivity. However, most of the segmentation
is provided by the 2D correlation (1). In the case where
higher resolution steady-state images are acquired with
lower resolution dynamic images for temporal data, inter-
VISUALIZATION TECHNIQUES FOR MRA
Academic Radiology, Vol 9, Suppl 1, 2002
polation must be used to map the temporal data onto the
steady-state image.
The main limitation of temporal labeling is that only a
single field of view can be imaged throughout the dynamic and steady-state phases required. It would be helpful to have techniques that are applicable to the steadystate alone so multiple vascular regions could be imaged,
taking full advantage of the longer temporal duration of
the steady-state phase of contrast agent distribution.
Phase-contrast agent imaging is one approach that meets
this requirement. The physiologic feature that is most
readily encodable using phase data is blood velocity, although differences in blood oxygenation may also be exploited to provide phase-contrast images (2). Phase-contrast imaging can be used to differentiate arteries and
veins based on flow velocity (3) or flow direction (4).
The use of flow direction is appealing, especially in regions where arteries and veins are closely entwined in
vascular bundles, such as the lower extremity circulation.
In such vascular territories, arterial and venous flows are
strictly distinguishable based on flow direction relative to
the axial plane. The upper extremity and carotid circulations are examples of other vascular territories that exhibit
this feature. The fact that arterial and venous flows are in
opposite directions also has the advantage that phase encoding only needs to be in a single direction, parallel to
the vascular bundle.
Alternatively, pure postprocessing strategies may be
applied to a high-resolution image. Volume rendering is a
technique that incorporates data from all pixels in an image to provide a 3D representation. By incorporating cut
planes to eliminate unwanted structures and by varying
the degree of opacity of structures, image viewing may be
optimized. This technique can be effective at depicting
arteries and veins even when these structures are adjacent.
By providing depth cues and with the ability to rotate the
vessels in three dimensions, even complex anatomy of,
for example, the carotid bifurcation can be analyzed
quickly and accurately. This technique is most effective
when used on vessels that, although adjacent, are not
completely intertwined, such as the carotid artery and
jugular vein. Another feature of the carotid artery that
makes the volume-rendering approach especially effective
is the fact that a focal area of the carotid artery, the carotid bifurcation, is of greatest interest, so the volume can
be rotated with the aim of visualizing this specific region.
In the peripheral circulation, this approach is less successful because the complex intertwining of artery and vein
makes separating the vessels along their entire course a
very tedious process, and it is necessary to fully visualize
the entire length of the artery because disease can occur
at any location along it.
In the periphery, another goal of postprocessing is to
completely segment the arterial from the venous circulations. The simplest method of doing this is a connectivity
approach, whereby a seed is placed within an arterial or
venous voxel, and the program is asked to generate a
structure based on pixels above a certain threshold that
are connected to the seed point. Although attractive in
principle, this simple approach has several limitations
based on focal zones of diminished signal due to artifact
as well as bridging between structures, which can occur
with loss of just a single pixel separating them.
More advanced techniques for artery-vein segmentation
are under investigation. Two recently described approaches are forward-looking graph search (5) and
“fuzzy” connectedness (6). As an example, in the former
technique, a graph search is performed on the 3D image
data starting at a set of user-defined points. Node costs
are determined by gray level, 3D edge strength, and a
model of preferred branch direction. The graph search
program then identifies the lowest-cost path through the
3D volume. Fuzzy connectedness aims to overcome the
limitations of pure intensity thresholding by establishing
affinity relationships between regions. These affinities can
be composed of a variety of characteristics but, in practice, usually consist primarily of a signal homogeneitybased component and an object feature-based component.
The strength of these postprocessing techniques is that
they do not rely on special acquisitions but can be performed post hoc on any high-resolution data set. However, it will need to be seen how robust these techniques
are in a range of pathologic conditions, including long
occlusions.
In conclusion, intravascular contrast agents provide a
unique opportunity for high-resolution MR imaging of
vascular structures. However, to maximize the utility of
these agents, it will be essential to have techniques to
separate overlapping arteries and veins. Both acquisitionbased and pure postprocessing approaches have shown
promise; further work will be required to demonstrate
which approaches will prove most robust in clinical use.
REFERENCES
1. Mazaheri Y, Carroll TJ, Mistretta CA, et al. Vessel segmentation in 3D
MR angiography using time resolved acquisition curves (abstr). In: Proceedings of the Seventh Annual Meeting of the International Society for
S141
YUCEL
Magnetic Resonance in Medicine. Berkeley, Calif: International Society
for Magnetic Resonance in Medicine, 1999; 2181.
2. Wang Y, Haacke EM, Yu Y, Li D, Bae KT. Differentiating arteries and
veins using susceptibility induced venous phase (abstr). In: Proceedings
of the Seventh Annual Meeting of the International Society for Magnetic
Resonance in Medicine. Berkeley, Calif: International Society for Magnetic Resonance in Medicine, 1999; 17.
3. Foo TKF, Ho VB, Hood MN, et al. A novel method for MR arterial and
venous discrimination using gated phase contrast and VENC selection
(abstr). In: Proceedings of the Seventh Annual Meeting of the International Society for Magnetic Resonance in Medicine. Berkeley, Calif: International Society for Magnetic Resonance in Medicine, 1999; 2182.
4. Bluemke DA, Darrow RD, Gupta R, Tadikonda SK, Dumoulin CL. 3D
contrast enhanced phase contrast angiography: utility for arterial/ve-
S142
Academic Radiology, Vol 9, Suppl 1, 2002
nous segmentation (abstr). In: Proceedings of the Seventh Annual
Meeting of the International Society for Magnetic Resonance in Medicine. Berkeley, Calif: International Society for Magnetic Resonance in
Medicine, 1999; 1237.
5. Sonka M, Stefancik R, Tadikonda S. Feasibility of automated separation
of arteries and veins using a graph searching technique (abstr). In: Proceedings of the Seventh Annual Meeting of the International Society for
Magnetic Resonance in Medicine. Berkeley, Calif: International Society
for Magnetic Resonance in Medicine, 1999; 2183.
6. Lei T, Udupa JK, Saha PK, et al. Artery-vein separation using MR angiographic data in 25 patients (abstr). In: Proceedings of the Seventh
Annual Meeting of the International Society for Magnetic Resonance in
Medicine. Berkeley, Calif: International Society for Magnetic Resonance
in Medicine, 1999; 1235.
Ultrasmall Superparamagnetic Iron
Oxide– enhanced MR Imaging of Atherosclerotic
Plaque in Hyperlipidemic Rabbits1
Stefan G. Ruehm, MD, Claire Corot, PhD, Peter Vogt, MD, Heidi Cristina, DVM, Jörg F. Debatin, MD
RATIONALE AND OBJECTIVES
Chronic inflammatory changes of the arterial wall are
inherent in the genesis of atherosclerosis (1). Despite the
inhomogeneous makeup of atherosclerotic plaque, it is
assumed that atherosclerotic changes originate from fatty
streaks that occur early in life and tend to progress to
advanced lesions (2,3). Particular plaque configurations
yield high risks for thrombosis independent of the degree
of associated luminal narrowing (4). Whereas conventional angiographic techniques display luminal narrowing,
the presented strategy aims at the detection of atherosclerosis-associated inflammatory changes in the vessel wall,
based on the observation that ultrasmall superparamagnetic iron oxide (USPIO) particles undergo phagocytosis
by cells of the mononuclear phagocytic system (5).
The purpose of this study was to evaluate the performance of USPIO (Sinerem; Laboratoire Guerbet, Aulnay
Sous Bois, France) as a marker of macrophage activity in
atherosclerotic plaque.
MATERIALS AND METHODS
Experiments were conducted on five heritable hyperlipidemic rabbits aged 6 –10 months. Three white New Zealand rabbits served as a control group. The studies were
performed in accordance with all regulations governing
animal experiments.
Acad Radiol 2002; 9(suppl 1):S143–S144
1 From the Institutes of Diagnostic Radiology (S.G.R., J.F.D) and Pathology
(P.V.), Central Biological Laboratory (H.C.), University Hospital Zurich,
Ramistr 100, CH-8091 Zurich, Switzerland; and Laboratoire Guerbet,
Aulnay Sous Bois, France (C.C.). Address correspondence to S.G.R. at
Universitätsklinikum Essen, Zentralinstitut für Röntgendiagnostik, Operatives Zertrum II, Hufelandstrasse 55, D-45122 Essen, Germany.
©
AUR, 2002
Magnetic resonance (MR) imaging was performed with
the rabbits fully anesthetized. To maximize the signal-tonoise ratio, a quadrature transmit/receive head coil was
used. The protocol was as follows:
1. All rabbits underwent three-dimensional (3D) contrast material– enhanced MR angiography of the thoracic
aorta with 2 mL of gadoterate meglumine (Gd-DOTA)
(Dotarem; Laboratoire Guerbet, Aulnay Sous Bois,
France) diluted with 10 mL of saline. Contrast material
injection was performed with an automated injector at a
flow rate of 0.1 mL/sec.
2. At least 1 day (maximum, 10 days) after the 3D MR
angiography, USPIO (Sinerem) was injected intravenously at
a dose of 1 mmol Fe/kg. MR imaging with a 3D ejectionfraction gradient-recalled echo (GRE) sequence (6.7/1.6 [repetition time msec/echo time msec], 30° flip angle, field of
view of 28 ⫻ 19.6 cm, and two signals acquired) was performed daily up to 5 days after USPIO application. Maximum intensity projections of the 3D data sets were reconstructed. Subsequently, the rabbits were killed for ex vivo
imaging of the aorta and histopathologic evaluation.
3. For ex vivo imaging, the aortic specimens were
closed with surgical thread at both ends, filled with 1:50
diluted Gd-DOTA, and placed in a small plastic container
filled with saline. 3D MR data sets were obtained with
the parameters described previously.
4. Histopathologic evaluation of the aorta was performed,
including histochemical staining (Prussian blue staining).
RESULTS
3D MR angiography with Gd-DOTA (Dotarem) revealed
no abnormal findings in either the five hyperlipidemic or
three control rabbits (Fig 1). After application of the USPIO
(Sinerem), the signal intensity in the vessel lumen steadily
increased over the 5-day imaging period, paralleling the deS143
RUEHM ET AL
Academic Radiology, Vol 9, Suppl 1, 2002
Figure 1. Coronal source image of 3D contrast-enhanced MR angiogram (3D GRE) (6.7/
1/6, 30° flip angle) with Gd-DOTA shows the
aorta of a hyperlipidemic rabbit. No abnormalities are depicted.
Figure 2. Coronal source image of 3D
contrast-enhanced MR angiogram (3D GRE)
(6.7/1/6, 30° flip angle) of the same hyperlipidemic rabbit as in Figure 1 at 5 days after
injection of USPIO (Sinerem). Note the susceptibility artifacts in the vessel wall representing Fe uptake in macrophages embedded in plaque.
crease in T2*— effects due to the decreasing concentration
of USPIO (Sinerem). Luminal signal intensity reached its
maximum (best angiographic effect) after 4–5 days. On the
3D GRE images, the aorta of the hyperlipidemic rabbits
showed marked susceptibility artifacts in the vessel wall in
vivo (Fig 2), as well as ex vivo.
In the atherosclerotic rabbits, histopathologic examination proved marked uptake of Fe particles in macrophages
embedded in atherosclerotic plaque found in the aortic
wall. No such changes were seen in the control rabbits.
Similarly, no such changes were seen in one hyperlipidemic rabbit killed before receiving the USPIO (Sinerem).
CONCLUSION
The results of the study indicate that USPIOs, which
undergo phagocytosis by macrophages in atherosclerotic
plaques of the aortic wall, lead to susceptibility artifacts
S144
that can be readily detected with 3D GRE sequences.
USPIO uptake may thus serve as a marker of early atherosclerotic changes before luminal narrowing is present.
The presented imaging protocol is suited to combine MR
angiographic imaging with the detection of early atherosclerotic wall changes.
REFERENCES
1. Ross R. Atherosclerosis: an inflammatory disease. N Engl J Med 1999;
340:115–126.
2. Berliner JA, Navab M, Fogelman AM, et al. Atherosclerosis: basic
mechanisms— oxidation, inflammation, and genetics. Circulation 1995;
91:2488 –2496.
3. Ross R. The pathogenesis of atherosclerosis: a perspective for the
1990s. Nature 1993; 362:801– 809.
4. Fuster V, Badimon L, Badimon JJ, Chesebro JH. The pathogenesis of
coronary artery disease and the acute coronary syndromes (1). N Engl
J Med 1992; 326:242–250.
5. Saini S, Stark DD, Hahn PF, Wittenberg J, Brady TJ, Ferrucci JT Jr.
Ferrite particles: a superparamagnetic MR contrast agent for the reticuloendothelial system. Radiology 1987; 162:211–216.
Noninvasive Assessment of Wound-healing
Angiogenesis with Contrast-enhanced MRI1
Thomas H. Helbich,2 Timothy P. L. Roberts, Mark D. Rollins, David M. Shames, Karl Turetschek, Harriet W. Hopf
Matthias Mühler, Thomas K. Hunt, Robert C. Brasch, MD
RATIONALE AND OBJECTIVES
Angiogenesis is a process by which new vessels grow
toward and into the tissue (1,2) and is a critical component of the wound-healing process. During this process,
angiogenesis leads to increased endothelial permeability
of the microvasculature, which is triggered by plasma
proteins such as the vascular endothelial growth factor
(1,2). Based on physiologic data, there are similarities
between wounds and tumors (1). Previous studies demonstrated that macromolecular contrast material– enhanced
magnetic resonance imaging (MRI) is able to quantify
microvascular permeability and plasma volume, representatives of angiogenesis, in tumors (3,4).
The purpose of this study was twofold: (a) to quantitatively assess the time course of microvascular permeability and fractional plasma volume (fPV), representatives of
angiogenesis, in an established wound-healing model induced in rodents; and (b) to correlate the MRI parameters
with histologic and physiologic data of the wound-healing
process.
MATERIALS AND METHODS
Twelve healthy young adult female Sprague-Dawley
rats weighing 225–250 g were used for all wound model
implantations. Polypropylene woven mesh cylinders were
Acad Radiol 2002; 9(suppl 1):S145–S147
1 From the Center for Pharmaceutical and Molecular Imaging, Department
of Radiology (T.H.H., T.P.L.R., D.M.S., K.T., M.M., R.C.B.), and the Wound
Healing Laboratory, Department of Surgery (M.D.R., H.W.H., T.K.H.), University of California, San Francisco, Box 0628, 515 Parnassus Ave, San
Francisco, CA 94143-0628. Address correspondence to R.C.B.
2
©
Current address: Department of Radiology, University of Vienna, Austria.
AUR, 2002
constructed from 27 ⫻ 27-mm square pieces of mesh
with a mesh opening of 1,000 ␮m and a thickness of
1,020 ␮m (Spectrum, Houston, Tex), using a method of
construction similar to stainless steel wire mesh HuntSchilling wound chambers (5). Final dimensions of the
cylinder were an internal diameter of 8 mm and length of
28 mm. Rats were anesthetized with pentobarbital (35
mg/kg, intraperitoneal), buprenorphine hydrochloride
(0.05 mg/kg, intraperitoneal), and atropine (0.8 mg/kg,
intraperitoneal). Four cylinders were inserted through two
1-cm dorsal midline incisions and implanted into two
pairs of bilateral subcutaneous pockets. Wound fluid was
aspirated on days 1, 4, 9, and 22 after implantation using
brief halothane anesthesia. For histologic purposes, three
rats were killed on each of days 1, 4, 9, and 22. When
aspirated, wound fluid was taken from a cylinder in
which fluid had not been previously drawn. Vascular endothelial growth factor (VEGF) levels were assayed by
commercial enzyme-linked immunosorbent assay (ELISA)
(R&D Systems, Minneapolis, Minn) from the wound
fluid.
MRI was performed on an imaging system (Omega
CSI-II System; Bruker Instruments, Fremont, Calif) operating at 2 T. This system is equipped with self-shielded
gradient coils (⫾20 G/cm, 15-cm inner diameter)
(Acustar S-150; Bruker Instruments). Before MRI, pentobarbital (Abbott Laboratories, North Chicago, Ill) was
injected intraperitoneally at a dosage of 50 mg/kg of body
weight. For contrast medium injection during MRI, a 23gauge butterfly cannula (Abbott Laboratories) was inserted into a tail vein of each animal. Animals were
placed supine in a “birdcage” radio-frequency coil. A dilute gadopentetate– containing phantom was included in
the field of view, close to each animal. Wound chambers
were examined with a T1-weighted three-dimensional
S145
HELBICH ET AL
spoiled gradient-refocused (3D-SPGR) acquisition in a
steady state. 3D-SPGR images were obtained with the
following parameter settings: echo time (TE), 1.4 msec;
repetition time (TR), 50 msec; one acquisition; flip angle,
90°; matrix, 128 ⫻ 128 ⫻ 16; section thickness, 3 mm;
field of view, 50 ⫻ 50 ⫻ 48 mm; and acquisition time, 1
minute 42 seconds per volume of 16 sections. Precontrast
longitudinal relaxation rates (R1 ⫽ 1/T1) were calculated
based on a set of seven similar SPGR sequences with
varying TR values between 50 and 3,200 msec. To monitor albumin–(Gd-DTPA)30 enhancement, three initial precontrast images and 30 dynamic postcontrast images (with
fixed TR at 50 msec) were acquired over a 1-hour period.
Albumin-(Gd-DTPA)30, a prototype macromolecular contrast medium for MRI, was administered in a dose of 0.03
mmol of Gd per kilogram of body weight (6).
Animals were imaged on days 1, 4, 9, and 22 after
wound chamber implantation. For histologic purposes,
three rats were killed on each of days 1, 4, 9, and 22 immediately after MRI and wound fluid aspiration.
All MRI data were transferred to a workstation (Sun
Sparc 10; Sun Microsystems, Mountain View, Calif). Signal intensity (SI) values for each time point were obtained
from regions of interest (ROIs) defined in the most enhancing areas close to the wound chamber. Three to five
ROIs over several different anatomic image sections were
summed and evaluated using an image analysis program
(MRVision, Menlo Park, Calif). In addition, signal in the
inferior vena cava (IVC) and signal in a gadolinium phantom in the field of view were measured.
The mean SI values from the tumor and the IVC blood
were corrected for spectrometer variation over time by
dividing the SI at each time point by the SI of the phantom. Precontrast Rl (1/T1) estimates for tumors were obtained by curve fittings based on seven unenhanced SPGR
sets with TRs varying from 50 to 3,200 msec. The precontrast Rl value for IVC blood was assumed to be 0.752
sec⫺1 (1/1.33) (7). Postcontrast R1 values can be calculated based on SI and knowledge of precontrast Rl values
(7,8). The difference between the postcontrast R1 value at
each time point and the precontrast Rl, ⌬R1(t), is taken to
be directly proportional to the local gadolinium concentration in tissue at time t (9). Our techniques for determining
the fractional plasma volume of the tumor tissue (fPV, in
mmol cc⫺1 of tissue) and the permeability of the tumor
tissue to a particular contrast agent, as estimated by the
endothelial transfer coefficient (Kps, in mL min⫺1 100
cc⫺1 of tissue), have been reported in detail elsewhere
(10). Briefly, a two-compartment model including micro-
S146
Academic Radiology, Vol 9, Suppl 1, 2002
vascular permeability was fitted to the ⌬R1(t) data from
IVC and tumor after the former was corrected for hematocrit assuming a value of 0.42 (10). The ⌬R1(t) from the
IVC, fitted by a monoexponential function for macromolecular contrast medium, served as forcing functions for
the tumor tissue model, which had a two-compartment
structure corresponding to plasma and interstitial water
spaces. The model was fitted to the data using the SAAM
II software (SAAM Institute, Seattle, Wash).
Immediately after MRI, animals were killed by an
overdose of pentobarbital. Wound chambers were resected, fixed in methylmethacrylate (Technovit 9100;
Heraeus Kulzer, Wehrheim, Germany), sectioned (100
␮m) in the same plane as the MRI with a low-speed saw
(Buehler, Lakebluff, Ill), and stained with epoxy tissue
staining. For each specimen, 5–10 areas with the highest
microvascular density (MVD) were chosen for vessel
counting at low power (⫻40 and ⫻100).
RESULTS
The results of the permeability parameters (Kps, fPV),
VEGF levels, and the MVD obtained from the wound
models on days 1, 4, 9, and 22 in all 12 animals are summarized in the Table. Microvascular permeability of the
wound, characterized by Kps, increased significantly from
day 1 (Kps ⫽ 0 mL min⫺1 100 cc⫺1 of tissue) to day 9
(Kps ⫽ 0.046 mL min⫺1 100 cc⫺1 of tissue; P ⬍ .01) and
remained constant through day 22 (Kps ⫽ 0.053 mL
min⫺1 100 cc⫺1 of tissue). The fPV increased progressively with time (day 1, fPV ⫽ 0.03 mL cc⫺1 of tissue;
day 22, fPV ⫽ 0.07 mL cc⫺1 of tissue). The VEGF levels
increased significantly from day 1 (489.5 pg/dL) to day 9
(1,215.9 pg/dL) and decreased on day 22 (849.6 pg/dL).
The correlation between MRI data and VEGF levels was
weak (r ⫽ 0.3; Spearman correlation). However, protease,
which is known to interact with the VEGF, seems to be a
reasonable explanation for these results. The MVD increased significantly with time (day 1 ⫽ 0; day 22 ⫽
6.4). Both Kps and fPV correlated closely with vessel
counts of the wounds.
CONCLUSION
Our results demonstrate that contrast-enhanced MRI is
able to quantitatively assess the time-dependent changes
of angiogenesis in the wound-healing process. Representatives of microvasculature, such as Kps, correspond with
the histologic and physiologic parameters of the wound-
Academic Radiology, Vol 9, Suppl 1, 2002
MR ASSESSMENT OF WOUND-HEALING ANGIOGENESIS
Results for Kps, fPV, VEGF, and MVD of the Wound Studies Obtained on Different Days
Day
Kps (mL min⫺1 100 cc⫺1 tissue)*
fPV (mL cc⫺1 tissue)†
VEGF (pg/dL)‡
MVD§
1
4
9
22
0
0.014 (0.006)
0.046 (0.017)
0.053 (0.006)
0.03
0.03
0.04
0.07
489.5 (95.9)
894.6 (254.4)
1,215.9 (401.7)
849.6 (522.9)
0
1.1
5.9
6.4
Note.—Values in parentheses are SD; medians are reported for fPV and MVD.
*Significant difference between days 1 vs 4, 9, and 22 and days 4 vs 9 and 22 (P ⬍ .001; Scheffe test).
†Significant difference between days 1 and 4 vs 22 (P ⬍ .05; Kruskal-Wallis test).
‡Significant difference between days 1 vs 4, 9, and 22 (P ⬍ .05; Scheffe test).
§Significant difference between days 1 and 4 vs 9 and 22 (P ⬍ .05; Kruskal-Wallis test).
healing angiogenesis. Thus, MRI may offer a noninvasive
quantitative assay for the wound-healing process and may
be used to monitor treatment response.
REFERENCES
1. Dvorak H. Tumors: wounds that do not heal. N Engl J Med 1986; 315:
1650 –1659.
2. Suh D, Hunt T. Time line of wound healing. Clin Podiatr Med Surg
1998; 15:1–9.
3. Brasch R, Pham C, Shames D, et al. Assessing angiogenesis using
macromolecular MRI contrast media. J Magn Reson Imaging 1997;
7:68 –74.
4. van Dijke C, Brasch R, Roberts T, et al. Mammary carcinoma model:
correlation of macromolecular contrast enhanced MR imaging characterizations of tumor microvasculature and histologic capillary density.
Radiology 1996; 198:813– 818.
5. Schilling J, Favata B, Radakovick M. Studies of fibroplasia in wound
healing. Surg Gynecol Obstet 1953; 96:143–149.
6. Ogan MD, Schmiedl U, Moseley ME, Grodd W, Paajanen H, Brasch
RC. Albumin labeled with Gd-DTPA: an intravascular contrast-enhancing agent for magnetic resonance blood pool imaging—preparation
and characterization. Invest Radiol 1987; 22:665– 671. [Erratum: Invest
Radiol 1988; 23:961.]
7. Schwickert HC, Roberts TP, Shames DM, et al. Quantification of liver
blood volume: comparison of ultra short TI inversion recovery echo
planar imaging (ULSTIR-EPI), with dynamic 3D-gradient recalled echo
imaging. Magn Reson Med 1995; 34:845– 852.
8. Roberts TPL, Brasch RC, Schwickert HC, et al. Quantification of tissue
gadolinium concentration using magnetic resonance imaging: comparison of ultrashort inversion time inversion recovery echoplanar and
three-dimensional spoiled gradient-recalled approaches with in vitro
measurements. Acad Radiol 1996; 3:282–285.
9. Koenig SH, Spiller M, Brown RD, Wolf GL. Relaxation of water protons
in the intra- and extracellular regions of blood containing Gd(DTPA).
Magn Reson Med 1986; 3:791–795.
10. Shames D, Kuwatsuru R, Vexler V, Muehler A, Brasch R. Measurement of capillary permeability to macromolecules by dynamic magnetic resonance imaging: a quantitative noninvasive technique. Magn
Reson Med 1993; 29:616 – 622.
S147