Download Key factors in the acquisition of contrast kinetic data for oncology

JOURNAL OF MAGNETIC RESONANCE IMAGING 10:254–259 (1999)
Original Research
Key Factors in the Acquisition of Contrast Kinetic
Data for Oncology
Jeffrey L. Evelhoch, PhD*
Dynamic contrast-enhanced magnetic resonance imaging
(DCE-MRI) has recently emerged as a promising method for
both diagnosis and prognosis of cancer despite considerable variation in both the methods of data acquisition and
analysis. Both to facilitate integration of results from
multiple institutions and to ensure that the data reflect
the underlying physiology as accurately as possible, several aspects of data acquisition should be taken into account when developing protocols for DCE-MRI regardless
of how the data are analyzed. Among the relevant issues are
the relationship between signal enhancement and contrast
agent concentration, intra- or inter-patient variation in
the blood contrast agent concentration as a function of
time, requirements for spatial and temporal resolution, the
impact of tumor heterogeneity, and the impact of patient
motion during the study. This review considers these
factors and, when possible, makes specific recommendations for addressing them experimentally. J. Magn. Reson. Imaging 1999;10:254–259. r 1999 Wiley-Liss, Inc.
Index terms: Gd-DTPA; tumors; data acquisition; heterogeneity; temporal resolution; spatial resolution
DYNAMIC CONTRAST-ENHANCED magnetic resonance
imaging (DCE-MRI) has recently emerged as a promising method for both diagnosis (1–6) and prognosis
(7–12) of cancer. Remarkably, these positive results
have been obtained despite considerable variation in
both the methods of data acquisition (eg, pulse sequences, acquisition parameters, spatial resolution and
coverage) and analysis (eg, visual inspection (2), parametric analysis (9), pharmacokinetic (8) or physiologic
(1) modeling). These encouraging results suggest there
are substantial physiologic/pharmacokinetic differences (ie, between benign and malignant, or between
non-responsive and responsive tumors) that are evident
independent of methods for acquisition and analysis of
the DCE-MRI data. However, integration of results from
multiple institutions and/or evaluation of the relative
merits of the various methods for data analysis are
difficult, if not impossible, when the differences in
Cancer Biology Program, Barbara Ann Karmanos Cancer Institute, and
Departments of Internal Medicine and Radiology, Wayne State University, Vaitkevicius MR Center, Detroit, Michigan 48201.
Contract grant sponsor: National Cancer Institute; Contract grant
number: U01 CA62555.
*Address reprint requests to: J.L.E., Vaitkevicius MR Center, Harper
Hospital, 3990 John R. Street, Detroit, MI 48201.
E-mail: [email protected]
Received July 12, 1999; Accepted July 13, 1999.
r 1999 Wiley-Liss, Inc.
acquisition parameters are substantial. Moreover, in
light of the introduction of a new class of anti-cancer
agents based on action against tumor angiogenesis (13)
and the potential role DCE-MRI could play in the
development of these agents (14,15), it is critical to
understand the relationship of DCE-MRI data to physiologic variables (eg, differential effects on perfusion and
permeability).
Regardless of how the data are analyzed, several
aspects of data acquisition should be taken into account when developing protocols for DCE-MRI both to
facilitate integration of results from multiple institutions and to ensure that the data reflect the underlying
physiology as accurately as possible. Among the issues
that should be considered are the relationship between
signal enhancement and contrast agent concentration,
intra- or inter-patient variation in the blood contrast
agent concentration as a function of time, requirements
for spatial and temporal resolution, the impact of tumor
heterogeneity, and the impact of patient motion during
the study. In this review, these factors are considered
both in general and, in some cases, in the context of
data analysis using the uptake integral (or initial area
under the signal-time curve [Initial AUC]) approach, a
method we have used previously for analysis of DCEMRI data (16) and have studied extensively in the
context of D2O measurements of tumor blood flow
(17,18). Given the emerging role for DCE-MRI in the
development of anti-angiogenic agents, this review will
use examples of extracting physiologic information from
the DCE-MRI data, although similar considerations
also apply to pharmacokinetic analysis.
RELATING SIGNAL INCREASE TO GADOLINIUM
CONCENTRATION
The relationship between tissue uptake and clearance
of biologically inert, externally detectable tracers and
physiologically relevant parameters such as blood flow
or permeability has been of interest since the 1940s
(19,20). Although many models have been developed
over the past half century, knowledge of the tracer
concentration is required to extract as much physiologically relevant information as possible from tracer kinetic data. This complicates analysis of DCE-MRI data
because only the change in signal amplitude (⌬S) is
generally measured, and that may not be easily related
to the concentration of contrast agent in the tissue (Ct).
254
Contrast Uptake Data for Oncology
The change in the spin-lattice relaxation rate (⌬R1,
where R1 ⫽ 1/T1) is linearly related to the change in
contrast agent concentration over the range of concentration likely to be observed in tissues (21). However, the
relationship between ⌬R1 and ⌬S for short echo time
(where T2* effects can be ignored) spoiled gradient-echo
sequences commonly used for DCE-MRI also depends
on R1 prior to contrast injection (R10), the flip angle (␣),
the repetition time (TR), and proton density (22,23).
Although methods have been introduced for rapid T1
measurements (eg, refs. 24–26), these are not widely
used. Hence, a simple means to relate ⌬S to ⌬R1 (and
hence Ct) independently of other variables is desirable.
This would not only facilitate extraction of physiologically relevant information from the DCE-MRI data, but
would also benefit integration of data from multiple
institutions regardless of how the data are analyzed.
In general, the relative change in signal amplitude
(⌬Srel, S divided by the initial signal) is used to eliminate
differences in proton density. However, as is evident in
Fig. 1a, the relationship between ⌬Srel and Ct depends
strongly on R10. In 1994, Hittmair et al (27) introduced a
straightforward method to convert ⌬Srel in a spoiled
gradient-echo image to an ‘‘enhancement factor’’ that is
linearly related to Ct independent of R10, and requires
little additional experimental time. However, for a short
TR typically used for rapid acquisition (eg, ⬍10 msec),
the flip angle for optimum contrast-to-noise ratio (28)
results in increased sensitivity to differences in R10, and
errors in the nominal flip angle can produce significant
errors in the apparent Ct (eg, for a 5 msec TR and a
nominal 16° flip angle, a –2° error results in up to -20%
error in the apparent Ct[29]). Thus, it seems prudent to
measure T1 before contrast injection to minimize the
error introduced by intra-tumor, inter-study, or interpatient variations in R10.
Another factor that should be considered to ensure
that the observed ⌬S can be related to Ct is the selection
of the flip angle. As previously noted by Pelc (28), the flip
angle that maximizes the contrast due to a difference in
T1 is not the Ernst angle, which would maximize the
signal given TR and the pre-contrast T1 (30). For example, the Ernst angle for a 5 msec TR and an initial T1
255
of 850 msec is 6°; however, as is evident in Fig. 1b, the
flip angle providing the maximum contrast is roughly
16°. This optimal flip angle also has the advantage that
the relationship between ⌬S and Ct is nearly linear up to
1 mM Gd-DTPA and small errors in the flip angle have
little impact on that relationship (eg, for a 5 msec TR
and a nominal 16° flip angle, a –2° error results in up to
3% error in the apparent Ct).
VARIATIONS IN BLOOD CONTRAST TIME COURSE
Changes in the blood contrast agent concentration as a
function of time, Ca(t) directly impact the uptake tracer
kinetics. An example of this relationship is illustrated in
Fig. 2a and b (blood time courses) and c and d (corresponding tissue time courses). For these simulations,
the Kety equation (31) was used to calculate tissue
tracer concentration as a function of time for tissue with
the same underlying physiology and three similar, but
slightly different Ca(t) corresponding to either a short
bolus injection (ie, 5–10 seconds; Fig. 2a and c) or a
longer bolus injection (ie, 25–30 seconds; Fig. 2b and d).
Such variations in Ca(t) could easily result from interstudy differences in cardiac output. Measuring Ca(t) in
every study is often difficult in oncologic studies, given
the desire to sample the entire tumor with the highest
spatial resolution possible (see Spatial and Temporal
Resolution, below). Moreover, the temporal requirements for accurately characterizing the more rapidly
changing arterial time-concentration curve are greater
than for the tissue if a rapid bolus is used (32). Nonetheless, there are clearly substantial differences in the
contrast kinetics curves due solely to the variations in
Ca(t) that should be considered. Although the impact of
such differences in Ca(t) depends on the method used
for data analysis (18), their impact on extraction of
kinetic parameters using the uptake integral approach
(33,34) will be considered as an example.
Figure 3a–c shows the relationship between the uptake integral (the AUC) and the product of extraction
fraction and flow (EF, which is directly proportional
to Ktrans; Tofts et al, (41) for the Ca(t) shown in Fig. 2a
and b. In all cases, the relationship is altered when Ca(t)
Figure 1. a: Relationship between contrast agent concentration ([Gd-DTPA]) and percent increase in signal for a spoiled
gradient-echo sequence with 5 msec TR, 16° flip angle (␣), and 500 msec ⬍ Initial T1 (T10) ⬍ 1000 msec. b: Relationship between
[Gd-DTPA] and the increase in signal for a spoiled gradient-echo sequence with 5 msec TR, 850 msec T10, and 6° ⬍ ␣ ⬍ 30°.
256
Evelhoch
Figure 2. Blood contrast time courses [Ca(t)] corresponding to the following. a: ‘‘Short’’ (5–10 second) bolus injection (standard)
altered by early or late peak. b: ‘‘Long’’ (25–30 second) bolus injection (standard, same dose as for curves represented in a) altered
by early or late peak. Contrast time course for a tissue with EF ⫽ 0.3 ml g-1 min-1, ve ⫽ 0.3, hematocrit ⫽ 0.45, and c: A ‘‘short’’
bolus injection. d: A ‘‘long’’ bolus injection.
varies. When the AUC over the first 60 seconds after
contrast arrival is used, the error introduced is slightly
greater for the short bolus (Fig. 3a, 7%–11% variation
from standard relationship) than for the longer bolus
injection (Fig. 3b, 6%–8%). However, the AUC values are
also greater so there would be less noise-related error in
the estimate. If the tracer kinetics are integrated over an
extended period (ie, 0–90 seconds) for the longer bolus
injection (Fig. 3c), the AUC values are comparable to
that for the 60 second AUC with the short bolus, and the
error introduced is still slightly less (7%–9%). In any
case, as illustrated in Fig. 3d, the impact of variation in
Ca(t) can be minimized by normalizing the AUC values to
the AUC in a ‘‘reference’’ tissue (eg, resting muscle; EF of
0.05 ml g-1 min-1 assumed for these simulations). Thus,
when using the uptake integral approach, if a normal
tissue with constant physiologic status is including in
the field of view, the impact of differences in Ca(t) can be
minimized. If a ‘‘reference’’ is not available or cannot be
used to account for the differences in Ca(t) for the
method of analysis used, Ca(t) should be sampled in
every study.
SPATIAL AND TEMPORAL RESOLUTION
The requirements for temporal and spatial resolution
for a particular oncologic application often are in direct
conflict. A good example of these conflicting demands is
provided by the desire to use DCE-MRI to differentiate
benign from malignant breast lesions (2,4). Both the
importance for high temporal resolution to characterize
contrast kinetics (32) accurately and the need for high
spatial resolution to identify distinguishing features of
lesion morphology (5) have been discussed. Although
the specific requirements for temporal and spatial resolution depend on the application and the method used
for data analysis, common factors relevant to considering these compromises include tumor heterogeneity,
patient motion during the study, and requirements for
sampling the blood contrast agent concentration as a
function of time.
Tumor Heterogeneity
Vascular heterogeneity within the tumor can affect the
temporal and spatial resolution required of DCE-MRI in
several ways. One impact is the need to sample the
entire tumor volume to take full advantage of this
non-invasive method. If the study aims to characterize
the tumor, failure to sample the entire tumor could
result in sampling errors of the type associated with
invasive assays (eg, region sampled not representative
of the entire tumor). If the application involves repeated
measurements (eg, assessing treatment response), the
impact of the sampling errors is compounded and
treatment effects could be altered if the same region of
Contrast Uptake Data for Oncology
257
Figure 3. Relationship between the uptake integral (AUC) and the product of extraction fraction and flow (EF) for (a) early,
standard, and late blood time course for ‘‘short’’ bolus injection and integration from 0 to 60 seconds after the bolus arrival; (b)
early, standard, and late blood time course for ‘‘long’’ bolus injection and integration from 0 to 60 seconds after the bolus arrival;
or (c) early, standard, and late blood time course for ‘‘long’’ bolus injection and integration from 0 to 90 seconds after the bolus
arrival. d: Relationship between the uptake integral normalized to that for ‘‘muscle’’ (EF ⫽ 0.05 ml g-1 min-1) and EF for early,
standard, and late blood time course for ‘‘short’’ bolus injection and integration from 0 to 60 seconds after bolus arrival.
the tumor is not sampled each time. Thus, multipleslice two-dimensional (2D) or 3D methods, which require longer acquisition times than might be desired for
optimal temporal sampling, are often needed in order to
sample the entire tumor.
A potential impact of tumor heterogeneity on the
requirements for spatial resolution arises because tumor vascularity is heterogeneous even at the microscopic level (35,36). As a consequence, some degree of
intra-voxel heterogeneity is likely regardless of the spatial resolution of the MRI data. It is useful to think of the
contrast kinetics in a single voxel as comprised of
contributions from a large number of microscopic volume elements, within which the vascular characteristics determining contrast kinetics are homogeneous
and/or water diffusion results in well-mixed compartments (21). The contrast kinetic curve for a given voxel
is then the sum of all the curves from the smaller
homogeneous, well-mixed volume elements contained
therein. However, the average time-signal curve from
the voxel does not correspond to that which would be
observed from a homogeneous voxel with the corresponding average contrast transfer rate constant (ie, kep
or EF/␭) due to the exponential relationship between
tissue contrast concentration and the rate constant
(37). As a consequence, in the presence of intra-voxel
heterogeneity, the rate constant determined by fitting
the kinetic data will differ from the true average rate
constant depending on the extent of heterogeneity (18).
This problem can be kept to a minimum by acquiring
data with the highest spatial resolution possible, or by
using the uptake integral approach (ie, the uptake
integral is nearly linearly related to the contrast transfer
rate constant, so the average uptake integral more
closely reflects the average rate constant [18]).
Patient Motion
Since tumor contrast kinetics are often sampled over
several minutes, the potential for patient motion is
substantial. Given the desire for the highest spatial
resolution possible to extract the underlying kinetic
parameters accurately, methods for motion correction
(eg, refs. 38 and 39) may often need to be applied. If, on
the other hand, the uptake integral approach is used to
analyze the data so that tumor heterogeneity has less
influence on the extracted kinetic parameters and partial volume effects are tolerable, ‘‘averaging’’ of the
motion effects in larger volumes, as is generally ac-
258
Figure 4. True blood contrast time course for a ‘‘rapid’’ and a
‘‘long’’ bolus injection and standard arrival (—) and the effect of
averaging over either 4 seconds (open circles) or 8 seconds
(solid circles).
Evelhoch
3. Measure contrast agent concentration in the blood
in every study to correct for inter- or intra-patient
differences if correction using a ‘‘reference’’ tissue
is not possible or inappropriate.
4. Sample entire tumor to minimize impact of tumor
heterogeneity.
5. Use highest possible spatial resolution or uptake
integral approach to minimize impact on heterogeneity on accuracy of kinetic/physiologic information.
6. Correct for inter-scan patient motion or use volume ‘‘averaging’’ with uptake integral approach if
partial volume effects are not a concern.
7. If sensitivity to regions with rapid contrast kinetics
is needed, contrast agent concentration in blood
should be measured with 1 second of temporal
resolution; techniques should be developed and
validated to combine this with spatial resolution
requirements.
cepted in positron emission tomography (PET) studies,
may be appropriate.
ACKNOWLEDGMENTS
Sampling the Blood Contrast Time Course
The author thanks Mr. Zhanquan He for invaluable
programming assistance in computer simulations.
The temporal requirements for sampling Ca(t) have been
considered in detail by Henderson et al (32). They
concluded that a rapid bolus injection reduces the error
in estimations of uptake kinetic parameters (especially
important for EF ⬎ 1 ml g-1 min-1) and that accurate
representation of Ca(t), which is changing much more
rapidly than the tissue, requires 1 second of temporal
resolution. However, a slightly prolonged bolus may
help to reduce the sensitivity to variations in the input
function (see Variations in Blood Contrast Time course),
and reduce the sampling requirements. (eg, Fig. 4
demonstrates the effects of averaging over 4 or 8 seconds for a 30 second bolus.) However, since the tissue
changes in contrast occur more slowly, it does not have
to be sampled as rapidly (32). To take advantage of this
differential, methods to sample Ca(t) with high temporal
resolution while sampling the tissue with high spatial
resolution (eg, ref. 40) need to be further developed and
validated. Alternately, Ca(t) could be measured in a large
vessel (or the heart) from a ‘‘pre-bolus’’ with high temporal resolution before beginning the tumor DCE-MRI
experiment. A ‘‘pre-bolus’’ of one-tenth the normal dose
may be sufficient given the high arterial concentrations;
however, it would have to be diluted to keep the injection
volume and duration the same. If the subject’s cardiovascular status is stable, this could be related to the Ca(t)
through either the dose ratio or measurement in the
same blood pool at the end of the DCE-MRI experiment.
SUMMARY OF RECOMMENDATIONS
1. Measure T1 prior to contrast injection (T10) in every
study to minimize the impact of its variation on the
relationship between the signal increase and contrast concentration.
2. Select the flip angle to optimize the signal increase
for the T10 and expected maximum concentration
of contrast agent.
REFERENCES
1. Hulka CA, Edmister WB, Smith BL, et al. Dynamic echo-planar
imaging of the breast: experience in diagnosing breast carcinoma
and correlation with tumor angiogenesis. Radiology 1997;205:837–
842.
2. Kuhl CK, Mielcareck P, Klaschik S, et al. Dynamic breast MR
imaging: are signal intensity time course data useful for differential
diagnosis of enhancing lesions? Radiology 1999;211:101–110.
3. Daniel BL, Yen Y-F, Glover GH, et al. Breast disease: dynamic spiral
MR imaging. Radiology 1998;209:499–509.
4. Orel SG. Differentiating benign from malignant enhancing lesions
identified at MR imaging of the breast: are time-signal intensity
curves an accurate predictor? Radiology 1999;211:5–7.
5. Nunes LW, Schnall MD, Siegelman ES, et al. Diagnostic performance characteristics of architectural features revealed by high
spatial-resolution MR imaging of the breast. AJR Am J Roentgenol
1997;169:409–415.
6. Weinstein D, Strano S, Cohen P, et al. Breast fibroadenoma:
mapping of pathophysiologic features with three- time-point, contrast-enhanced MR imaging—pilot study. Radiology 1999;210:233–
240.
7. Esserman L, Hylton N, Yassa L, et al. Utility of magnetic resonance
imaging in the management of breast cancer: evidence for improved
preoperative staging. J Clin Oncol 1999;17:110–119.
8. Hawighorst H, Weikel W, Knapstein PG, et al. Angiogenic activity of
cervical carcinoma: assessment by functional magnetic resonance
imaging-based parameters and a histomorphological approach in
correlation with disease outcome. Clin Cancer Res 1998;4:2305–
2312.
9. Mayr NA, Yuh WT, Zheng J, et al. Prediction of tumor control in
patients with cervical cancer: analysis of combined volume and
dynamic enhancement pattern by MR imaging. AJR Am J Roentgenol 1998;170:177–182.
10. Barentsz JO, Berger-Hartog O, Witjes JA, et al. Evaluation of
chemotherapy in advanced urinary bladder cancer with fast dynamic contrast-enhanced MR imaging. Radiology 1998;207:791–
797.
11. Bonnerot V, Charpentier A, Frouin F, et al. Factor analysis of
dynamic magnetic resonance imaging in predicting the response of
osteosarcoma to chemotherapy. Invest Radiol 1992;27:847–855.
12. Van Vierzen PB, Massuger LF, Ruys SH, Barentsz JO. Fast dynamic
contrast enhanced MR imaging of cervical carcinoma. Clin Radiol
1998;53:183–192.
Contrast Uptake Data for Oncology
13. Boehm T, Folkman J, Browder T. Antiangiogenic therapy of experimental cancer does not induce acquired drug resistance [see
comments]. Nature 1997;390:404–407.
14. Furman-Haran E, Grobgeld D, Margalit R, Degani H. Response of
MCF7 human breast cancer to tamoxifen: evaluation by the threetime-point, contrast-enhanced magnetic resonance imaging method.
Clin Cancer Res 1998;4:2299–2304.
15. Hawighorst H, Knapstein PG, Weikel W, et al. Angiogenesis of
uterine cervical carcinoma: characterization by pharmacokinetic
magnetic resonance parameters and histological microvessel density with correlation to lymphatic involvement. Cancer Res 1997;57:
4777–4786.
16. Evelhoch J, Cunnings A, Lucas D, et al. Early detection of therapeutic response by dynamic contrast-enhanced MRI. ISMRM Workshop
Syllabus: MR of cancer: Physiology and metabolism. Baltimore:
1996. 61–64.
17. Simpson NE, Evelhoch JL. Deuterium NMR tissue perfusion measurements using the tracer uptake approach: II. Comparison with
microspheres in tumors. Magn Reson Med 1999;420:240–247.
18. Simpson NE, He Z, Evelhoch JL. Deuterium NMR tissue perfusion
measurements using the tracer uptake approach: I. Optimization of
methods. Magn Reson Med 1999;42:42–52.
19. Smith BC, Quimby EH. Use of radioactive sodium as tracer in study
of peripheral vascular diseases. Radiology 1945;45:335.
20. Kety SS, Schmidt CF. The nitrous oxide method for quantitative
determination of cerebral blood flow in man: theory, procedure and
normal values. J Clin Invest 1948;6:476–484.
21. Donahue KM, Weisskoff RM, Burstein D. Water diffusion and
exchange as they influence contrast enhancement. J Magn Reson
Imaging 1997;7:102–110.
22. Haacke EM, Tkach JA. Fast MR imaging: techniques and clinical
applications. AJR 1990;67:258–266.
23. Haase A, Frahm J, Matthaei D, Hannicke W, Merboldt KD. FLASH
imaging. Rapid NMR imaging using low flip-angle pulses. J Magn
Reson 1986;67:258–266.
24. Kay I, Henkelman RM. Practical implementation and optimization
of one-shot T1 imaging. Magn Reson Med 1991;22:414–424.
25. Gowland PA, Leach MO. Fast and accurate measurements of T1
using a multi-readout single inversion-recovery sequence. Magn
Reson Med 1992;26:79–88.
26. Tong CY, Prato FS. A novel fast T1-mapping method. J Magn Reson
Imaging 1994;4:701–708.
27. Hittmair K, Gomiscek G, Langenberger K, et al. Method for the
quantitative assessment of contrast agent uptake in dynamic
contrast-enhanced MRI. Magn Reson Med 1994;31:567–571.
259
28. Pelc NJ. Optimization of flip angle for T1 dependent contrast in MRI.
Magn Reson Med 1993;29:695–699.
29. Krishnan S, Chenevert T, Evelhoch JL. Analysis of dynamic contrastenhanced MRI in tumors: optimization of acquisition parameters.
In Proceedings of the ISMRM 7th Scientific Meeting, Philadelphia,
1999.
30. Ernst RR, Anderson WA. Application of Fourier transform
spectroscopy to magnetic resonance. Rev Sci Instrum 1966;37:93–
102.
31. Kety SS. Peripheral blood flow measurement, In: Potter VR, editor.
Methods in medical research. Chicago: Year Book Medical Publishers; 1960. p 223–227.
32. Henderson E, Rutt BK, Lee TY. Temporal sampling requirements for
the tracer kinetics modeling of breast disease. Magn Reson Imaging
1998;16:1057–1073.
33. Herscovitch P, Markham J, Raichle ME. Brain blood flow measured
with intravenous H215O. I. Theory and error analysis. J Nucl Med
1983;24:782–789.
34. Evelhoch JL, McDouall JB, Mattiello J, Simpson NE. Measurement
of relative regional tumor blood flow in mice by deuterium NMR
imaging. Magn Reson Med 1992;24:42–52.
35. Endrich B, Reinhold HS, Gross JF, Intaglietta M. Tissue perfusion
inhomogeneity during early tumor growth in rats. J Natl Cancer
Inst 1979;62:387–395.
36. Tozer GM, Lewis S, Michalowski A, Aber V. The relationship
between regional variations in blood flow and histology in a transplanted rat fibrosarcoma. Br J Cancer 1990;61:250–257.
37. Evelhoch JL. Tracer measurements of blood flow. In: Gillies RJ,
editor. NMR in physiology and biomedicine. San Diego, CA: Academic Press; 1994. p 209–220.
38. Krishnan S, Chenevert TL, Helvie MA, Londy FL. Linear motion
correction in three dimensions applied to dynamic gadolinium
enhanced breast imaging. Med Phys 1999;26:707–714.
39. Korin HW, Felmlee JP, Ehman RL, Riederer SJ. Adaptive technique
for three-dimensional MR imaging of moving structures. Radiology
1990;177:217–221.
40. Taylor NJ, Rowland IJ, Tanner SF, Leach MO. A rapid interleaved
method for measuring signal intensity curves in both blood and
tissue during contrast agent administration. Magn Reson Med
1993;30:744–749.
41. Tofts PS, Brix G, Buckley DL, et al. Estimating kinetic parameters
from dynamic contrast-enhanced T1-weighted MRI of a diffusable
tracer: Standardized quantities and symbols. J Magn Reson Imaging 1999;10:223–232.