Joseph
W. Carlson,
PhD
#{149} David
MR Relaxometry Work
relaxation
rates
in extremities
heads
of healthy volunteers. The of the measurement is sufficient to obtain a distinctive relaxation rate dispersion behavior for different tissues. sensitivity
Index terms: MR. 452.1214
technology
Brain, MR, 10.1214 #{149} Magnetic resonance #{149} Magnetic resonance
Extremities, (MR), (MR), tissue
#{149}
characterization Radiology
Antonio
Brito,
#{149}
BS
1992;
184:635-639
resonance (MR) imaging is a useful diagnostic tool in clinical imaging because it provides a wide range of image contrasts. It is well known that image contrast in MR imaging is determined mainly on the basis of the longitudinal and transverse relaxation times Ti and T2, the density of hydrogen nuclei haying a substantial but comparably smaller variability in the body. It is also well known that Ti of tissue varies considerably with different magnetic field strengths. These differences are ultimately due to interactions of ‘
lvi
AAGNETIC
the hydrogen cal environment. Most of the
strength rates
nucleus observable
dependence in the
with
range
its chemifield
of relaxation of field
the Radiologic of California, Dr, South
Imaging Laboratory, San Francisco, 400 San Francisco, CA 94080
(J.W.C., A.B., L.K.); and Tosiba America MRI Inc. South San Francisco, Calif (D.M.G.). Received 13, 1992; revision requested March 16; revision received March 30; accepted April 8. Address reprint requests to J.W.C. C RSNA, 1992
January
PhD
trometers
(2-4)
by
MR
imaging
sequence.
MATERIALS
high
field
strengths.
In
the intermediate regime-a range of field strengths currently being used with clinical imagers-we should see the transition behavior between the limiting cases of very high and very low field strengths. Techniques for measuring the field strength dependence of Ti have been devised for use with laboratory spec-
cy-
article,
Apparatus
Measurement performed with
of relaxation an existing
system for
A 0.064-T
system
ica
South
MRI,
modified
backs. well
magnet,
for
this
of large can
be
variations
a pulsed aperture human resting patients
(20-cm gap). investigation on their sides on their
with
is centered
A pulsed netization value; no reception
magnet
the
relatively
strength.
small
In addition,
magnetic
on
gion of the relaxation
is
because
in relaxation
will see that the baseline strength
was
electro-
imager
changes
explored
Amer-
Calif)
application
of field
field
strength.
whole-body
The low-field-strength suited
elec-
Toshiba
Francisco,
by including
regimen rates
field
(ACCESS; San
was MR
a pulsed of a cycle
magnetic
permanent
imaging
rates clinical
by adding application
the baseline
of a clini-
METHODS
Experimental
magnet with a small This small gap limits to heads for patients or to extremities for
at very
In this
AND
corresponding
of 1
of field
we describe the implementation field-cycling relaxometry with cal, whole-body MR imager.
onto
to a frequency
means
cling. During part of the pulse sequence, an electromagnet is turned on and magnetization is allowed to approach equilibrium in that field. The electromagnet is then turned off and the magnetization measured quickly by means of a conventional
relevant to imaging systems is a result of hydrogen interactions with biologic macromolecules, particularly proteins (1). The typical rotational correlation time for a protein molecule is 1 psec,
stant
From University Grandview
Kaufman,
imaging tromagnet
strengths
MHz. Longitudinal relaxation of a hydrogen spin occurs more rapidly when the Larmor frequency is much lower than the characteristic frequency for protein reorientation than when it is much higher. In general, Ti will approach distinct limiting values at very high and very low field strengths in proteinaceous tissues. At Larmor frequencies much less than i MI-Iz (corresponding to a field strength of 200 G), Ti will decrease to a value independent of field strength. At frequencies much higher than 1 MHz, Ti will approach a higher value and again become con-
I
Leon
#{149}
in Progress’
of relaxation rate disperfrom magnetic resonance images was demonstrated on a clinical, whole-body imaging system. Study of the behavior of relaxation rates over a range of field strengths probes the structural environment of imaged hydrogen protons and reveals information about the composition of tissue. The authors deterand
PhD
Imaging
Acquisition sion curves
mined
M. Goldhaber,
an
we
field
interesting
re-
rate spectrum. is used
to approach radio-frequency is performed
to allow
mag-
an
equilibrium transmission while the pulsed
electromagnet
is on, nor
magnetization
undergoing
is any
or
transverse
free preces-
sion. Therefore, the homogeneity of the electromagnet is relatively unimportant. For example, although the homogeneity of the static magnetic ppm, the homogeneity
tromagnet typical
field is typically of the pulsed
is on the order field
Abbreviations: repetition time.
of 30%
50 elec-
across
a
of view.
ROl
=
region
of interest,
TR
=
635
A small electromagnet duce power requirements ing.
The
of two
was used to refor fast switch-
electromagnet
circular
we
bundles
cross-sectional
consisted
(ROIs) in particular tissues. A numerical best fit of the data to a monoexponential approach to an equilibrium value was
with
formed
used
of wires
dimension
7 x 7
of
a
cm
an average radius of 10 cm. A current 60 A provided a maximum variation 450
G with
these
tromagnet
was
parable fiers.
measurements.
driven
to those Voltage
supplies
as gradient
limitations
resulted
The
com-
ampli-
on the current
in a rise
time
to full
cycle
field strength of 30 msec. The fall time of the electromagnet was reduced substantially to approximately 8 msec with the use of a flying capacitor dump circuit. The rate of change of the magnetic field strength in these measurements was 6 T/sec, which is sufficiently high to induce retinal magnetophosphenes (5) but below levels for motor nerve stimulation reported in echo-planar imaging (6). In the imaging pulse sequence, we applied an initial 90 pulse to flip magnetization
from
the
static
field
direction
into
the
transverse plane. The pulsed electromagnet was then turned on for a time r, ranging from 50 to 350 msec. Transverse magnetization is quickly and irretrievably dephased in the highly nonuniform field. The cycle field was then turned off, and
the field was allowed
to settle.
shielding of the pole piece counterwound “pancake”
nated
eddy
dard
imaging
formed
currents
without
contamination
baseline field strength. To obtain an accurate essential
that
a stan-
the
of the
through
it is the
elec-
tromagnet be turned off completely. This requires an introduction of additional delay time for the current to fully settle. The time between turning off the amplifier pulse and start of a pulse sequence was 25 msec.
After
the current
two-dimensional
echo sequence
settled, Fourier
a standard transform
was performed
spin-
nal from an asymmetric echo was collected, with an echo time of 30 msec, a section thickness of 10 mm, a 1.7 x 1.4-mm in-plane resolution, and a 128 x 256 ma-
trix. Field
cycle time was the same
for all
phase-encoded acquisitions, then it was increased and the imaging sequence repeated. The time of cycle field application ranged from 50 to 350 msec in steps of 75 msec, for a total of five different values. With the 25-msec delay time for current settling and 150 msec for performing ra-
dio-frequency
pulses,
data
acquisition,
and unused sequence time, the repetition time (TR) was 225-525 msec, for a total imaging time of 4 minutes per field strength value. Current investigations with this apparatus have included extremities and heads of healthy volunteers and a rat model.
Data
Analysis
Mean signal from manually
636
Radiology
#{149}
intensities located
were regions
echo
recorded of interest
be
Two
in-
into
ac-
taken
sequence
results
in some
ph
zation,
magnetization
90#{176}-I80#{176} measurement following form: m
+ {[m(f)
-
exp_b/TI(fo)J}
‘1’ ‘
The signal
intensity
an
of the collected magnetization
exponential
attenuation
data is
multiplied from
T2
decay. The first term in braces in Equation (1) contains the two instrumental details described earlier; this term is a constant independent of the cycle field strength. The second term in braces is linearly dependent on cycle field strength by means of the equilibrium magnetization value mO(f). The derivation of Equation (1) is
explained The
in the Appendix. most
reliable
estimates
of TI are fit to all data from each of the selected ROIs for all cycle field strength values. Thirty different cycle field strength values were used to obtain a typical set of images in these preliminary studies. As a result, data analysis required a 33-parameter fit to the data: one parameter for the constant term in Equation (I), two for the linear function in Equation (1), and 30 for the Tls at the different field strengths. In analyzing the data, we did not assume any relationship between the Tls at different field strengths. The computational effort required for this analysis is modest. Before performing a global fit of the data, only the data from a single field strength value are fit to a three-parameter model. These results are then used as a starting point in a simplex search for a global best fit. The entire data analysis for each ROI at all field strength values required approximately 2 minutes with use of a minicomputer. Errors were estimated for the individual data points by evaluating the variance of the data points from the best-fit curve, that is, the reduced 2 was set equal to 1. Upper and lower confidence limits of Tl
by performing
a global
the
inhomogeneous
cycle
field.
The
expT/I(f)J.
-
by
The regrowth occurs for a time r, at an increased or decreased field strength. The electromagnet is turned off, and, after allowing the current to completely shut off, the magnetization is measured with a conventional spin-echo two-dimensional Fourier transform sequence. rf = radio frequency.
mz]exp_b/TI(fo)}
-
-----
dephased
(m) before the sequence has the
+ m,(f0)[l
mexpbhhl(fo)
=
times.
-
Figure 1. Diagram of pulses for the relaxation rate measurements. An initial 90#{176} pulse flips longitudinal magnetization into the transverse plane, where it is quickly
relaxation
and the cycle and delay
longitudinal
d
r
to the baseline field strength magnetization at the TI of that field. The assumed model of signal intensity depends on the TI at the cycle field strength, Tl(f), the TI at the baseline field strength, Tl(f,), the equilibrium magnetization values at the cycle and baseline field strengths, m0(f) and m,(f0), respectively, the residual longitudinal magneti-
achieved
(Fig 1). Sig-
density. must
per-
relaxation
count for this data analysis to be successful. First, transmitter coil nonuniformity or misadjustment will result in nonzero residual magnetization, m,, along the static field after application of the 90#{176} prepulse. Second, the delay time 6 between turning off the current and the start of the spin-
by
measurement,
current
details
the longitudinal
to be per-
the tissue
magnetization
x [1
with use of a winding elimi-
sequence
and
Active
and allowed
pulse
rate
strumental
elec-
by amplifiers
used
and
of of
to calculate
f,Ld
I
were established by performing a numeric search for the values of TI to find where x2 increased by 1 (7). The confidence limits describe only random errors in the data and do not help estimate effects from systematic errors in the instrumentation. Cycle field strength over the entire field of view varies by approximately 30%. Therefore, we must be careful to calibrate the local field strength when presenting data of relaxation rates. The regions studied herein were of the electromagnet,
always
ations were much been calibrated.
close so that
less
than
to the center the field van-
30%
and
had
RESULTS The relaxation process for hydrogen nuclei is the result of a combination of many effects at the molecular level. Individual effects, if they act independently, contribute additively to the overall relaxation rate (the inverse
of Ti).
pendent come
Contributions
relaxation more
plotted
from
processes
apparent
when
as relaxation
rates.
Phantom
inde-
will data
beare
Studies
Archetypal behavior of the relaxation rates as determined with phantom studies are shown in Figure 2. Relatively constant relaxation rate behavior over the field strengths measured
was
seen
in phantoms
contain-
ing simple solutions (Fig 2a), such as a solution of a paramagnetic relaxation agent, nickel chloride, and water or an emulsion of mineral oil and water (approximately i :1 mixture of mineral oil and
water
oleate and monooleate Phantoms romolecules,
with
sorbitan
mono-
polyoxyethylene-sorbitan added as emulsifiers). containing biologic macsuch as hen egg albuSeptember
1992
6
T
300 400
2 1000
400
Field
600
Strength
800
000
ues were returned
400
decreased
:
600
200
300 :
6
4
S
)
2OO
are probably artifacts of the order of data acquisition. Data were initially acquired at the baseline field strength of the magnet; the field strength val-
150
150
1200
0
200
400
(guss)
F1d
600
Strength
800
1000
1.00
( g.ss)
b. Figure
2. rates
ation emulsion strength phantom strengths associated phantom
Relaxation rate dispersion curves for in vitro samples. (a) Curves show the relaxfor phantoms containing a solution of nickel chloride and water (solid line) and an of mineral oil and water (dotted line). Relaxation rates were insensitive to field over the ranges investigated and showed no apparent structure. (b) Curves from a containing hen egg albumen show a comparably larger change over the field studied. The denatured egg phantom (dotted line) exhibited peaks at 500 and 650 G with nitrogen cross-relaxation that were not present in the native egg albumen (solid line).
higher than the native value. Also, structure appeared in the relaxation rate curve, with the appearance of two peaks at approximately 500 and 650 C. These peaks were ascribed to cross-relaxation effects of proton interactions with nitrogen nuclei in cross-linked protein molecules (2). These peaks represent not only the chemical composition of the phantom but also the structure of the composition. These peaks in relaxation rates
are a result
of the
short-range
the tissue caused by heating phantom (8). The field strength values
peaks
Figure
3.
Representative
MR image
ob-
tamed in the knee of a healthy volunteer (TR = 525 msec, echo time = 30 msec [525/ 30]). This is one of the five images used to determine relaxation rates. The cycle field used in this image was set to give a total field strength of 1,200 G.
men (Fig 2b), exhibited substantial changes in the relaxation rates for these same field strengths. The lower curve in Figure 2b exhibits a decrease
in the
relaxation
rate
from
3 sec-1
(Ti,
333 msec) at 200 G to 2 sec (Ti, 500 msec) at 1,100 C. Substantial changes occur in egg albumen with heat denaturation. After a phantom containing albumen was heated at approximately 60#{176}C for 10 minutes, the overall relaxation rate increased, as did the dispersion in the rates. At 200 C, the rate had nearly doubled, to roughly 5 sec (Ti, 200 msec), and decreased to 2.5 sect (Ti, 400 msec) at 1,100 C, a rate 25%
Volume
184
Number
#{149}
3
are
a result
of the
order
of these
quadrupole
transition frequencies of the nitrogen nuclei and are independent of the chemical composition of the material. Winter and Kimmich (2) calculated three field strength values for these peaks: i95, 490, and 685 C, with the peak at lowest field seen only in highly ordered biologic tissues. Koenig and Brown (i) reported slightly different values for the higher two peaks, of 500 and 655 C.
Human
Volunteer
marrow,
and
Study
fat; data
The
were
macromolecular
tissue
with
substan-
tially less structure. Data obtained the region of the pons exhibited stantial structure in the relaxation behavior (Fig 5c).
in subrate
Study
The measurements obtained in human volunteers are intended only to help assess the sensitivity of the technique in the differentiation of normal tissue. To determine both the feasibility and potential of the technique in the evaluation of pathologic tissue, we obtained measurements in a rat with an implanted mammary adenocarcinoma (9). A representative transaxial image of two rats is shown in Figure 6. The tumor had a discernable necrotic center surrounded by a viable region. Relaxation rates for the necrotic center, viable region, and normal rat muscle tissue are shown in Figure 7. The overall
relaxation
rate
of the
necrotic
center
of the slightly
tumor (Fig compared
7b) decreased with that
viable
region
7a). A small,
statistically seen in the Conversely,
field tially This
(Fig
of the
but
significant, structure was necrotic center at 650 C. the normal muscle (Fig
strengths, as well larger cross-relaxation enables a qualitative
tative differentiation and muscle in this
higher at lower
re-
as a substanstructure. and quanti-
between particular
tumor model.
DISCUSSION
pro-
cessed as described earlier. The basic behavior of the three regions is shown in Figure 4. The fat (ROI volume, 1.2 cm3) and bone marrow regions (ROI volume, 2.5 cm3) showed relatively flat relaxation rate behavior, with little structure apparent. The small structures seen at 640 C, the baseline field strength of the magnet,
then muscle
region, however (ROI volume, 1.5 cm3), exhibited an overall behavior similar to that seen in the cross-linked protein phantom of the previous experiments. This same sequence was repeated with transaxial sections of the brain of the healthy volunteer. Relaxation rate dispersion in white (Fig 5a) or gray (Fig 5b) matter is typical of that of
7c) showed a substantially laxation rate, particularly
A set of measurements was obtamed from sagittal images acquired in the knee of a healthy 34-year-old man. A representative image of the knee is shown in Figure 3. ROIs were located in different regions of muscle,
bone
incrementally, value, and
incrementally.
Rat Model
in
the
increased to baseline
These
intended
preliminary
only
of relaxation rate a clinical whole-body
imaging
time
experiments
to show
the
are
feasibility
determination imager.
of 4 minutes
with With an
per
field
point, the total data collection time for the 32 field points used in the human studies was approximately 2 hours.
Patient
acceptance
with
the
imaging
Radiology
637
#{149}
I00
100
100
8
r ISO
ISO
S
6
‘, 3)
20(1
400
600
Field
800
StI’ngth
a. Figure
1000
200
200
300
300
400
400
1300
600
1000
1000
200
I
200
400
(glJ)
FIeld
600
800
1000
200
300 400
2 600
200
rates for fat (a), bone marrow indicative of the short-range
400
(g.us)
Strength
600
800
Strength
Field
b.
4. Relaxation cross-relaxation peaks
1000
1200
000
12
0
(gauss)
C.
(I,), and muscle (c) in a healthy order in this tissue.
volunteer.
The long
muscles
of the extremity
showed
large
200
200
300
300
400
400
600
600
i S
10011
CO
200
400
600
800
1000
1000
120(1
I Fiald
200
F,td
a.
600
400
(gauss)
Strangth
600
1000
1200
0
sion
5. Relaxation components with
ROIs
were
times
3.0 cm3, I .9 cm3, and
used
would
rate dispersion little apparent
in these
be poor,
of field
Our
data
determine structure
show
matter,
in the of the
brain pons
of a healthy volunteer. (c) shows a much larger
gray
matter,
and
pons,
shorten reducing
by
will
points that
than were We can
allow
can
used reduce
us to acquire
in this the
Figure
images
positioning). is needed
Furto deter-
minutes.
The clinical applicability of this procedure may lie in observations of changes in nitrogen peaks associated with structural changes in the underPathologic
such
structural
as demyelination
conditions
changes
in
of brain
6.
Considerations
of total
imaging
aging.
In both
ence
is that
dimensions
hence, tional
spectroscopy
spectroscopy of spatial
total imaging to the product
ings.
However,
relax-
from along differ-
requires
MR images
(525/30)
disperof the
of two rats with
im-
phase
total
encoding;
imaging
time
is
independent of the number of frequency points of spectroscopy information. In relaxation rate dispersion measurements, the total imaging time is dependent on the product of only one
of the
spatial
dimensions
and
strength
of the
magnet,
we
the
field
were
to probe the cross-relaxation peak structure and very low field relaxation rate dispersion with relatively small changes in field strength. Attempts to perform this procedure with a mediumor high-field-strength
MR imager
two
time is proporof spatial encod-
peaks.
Radiology
remiim-
and
number of field points sampled. Because of the low baseline
#{149}
Transaxial
resolution are in spectroscopic
or scar tissue in muscle, may be mdicated by changes in cross-relaxation
638
show large The volumes
planted mammary adenocarcinomas. A small necrotic region is visible in the center of a large tumor in the leg of the rat on the right side of the figure. As in Figure 3, only one of five images obtained at 1,200 G is shown.
of
mine if we can reliably calculate Ti with less than five different time delays; if so, we could reduce examination time further, to approximately 15
tissue,
(b) matter peak effect.
clearly
ometry, additional information each image voxel is acquired with the image. The primary
with
gray
peak fewer
including patient ther investigation
tissue.
White (a) and cross-relaxation
measured. we
time and spatial niscent of those
associated
1200
respectively.
the head in 64 phase-encoded steps. These reductions will result in a total examination time of 30 minutes (not
lying
1000
the the
number of field points to 15 and still have sufficient data to characterize both cross-relaxation peaks and the overall trend in relaxation rates. A sacrifice in image resolution to 1.4 x 3 mm
800
( aa.s)
we anticipate
cross-relaxation with substantially
data points investigation.
600 Strngth
C.
determined A region
2.0 cm3, for white
400 F,1d
experiments
but
that we can substantially overall imaging time
number
curves structure.
200
S trngth
b.
Figure
1000
able
tions
This culties
larger field technical
with
changes,
experience Future limited
require baseline more
associated
strength
field
will
from the will cause
and
larger the
variastrength. diffi-
field patient
will
larger rates of flux change. work with use of uniform
cycle
coils
would
to extremity
probably
and
head
be
imag-
ing because of the drawbacks associated with larger-aperture coils. However, a nonuniform “surface” field coil could be used to investigate relax-
ation rates nonuniformity
of the heart muscle. The of the cycle field September
1992
.
ISO 6
150
ISO
6 200_
200
4
200
I
4 300
300
400
400
400
600
600
2
_
1000
200
400 Fild
600
800
1000
1000
1200
200
400
(gauss)
St,angth
800 (gauss)
600
Strsngth
Field
a.
1000
Figure
7.
region
(a) and
would niques
Relaxation
rate
necrotic
dispersion
region
curves
determined
in the center
of a tumor
require data processing techto calibrate the field of the coil,
_O
this
is not
a formidable
problem.
for
a rat
(b) decreased,
of this
next
[m2exp_ThhI(f)
x expStTI(fo)
APPENDIX understand
the derivation
+ m,(f)[l
m2exp’T1(I)
-
with
an
with
implanted
little
mammary
variation
magnetization period
+ m,(f)(l
adenocarcinoma.
compared
at the
The cycle field is then turned off and declines to within I % of the baseline field strength within 8 msec. An additional 17 time
delay
is necessary
for
the
re-
maining cycle field to completely decay. We can approximate that the field strength for the entire 25-msec period is the baseline field strength. The evolution
Volume
184
Number
#{149}
3
with
2.
-
+ m0(f0)(I
expT1f)1 -
3.
exp8/Tl(1o)).
dent on the TI at the baseline field strength and the TR. However, because this term is small and can usually be neglected in comparison to the regrowth contribution, this effect is not important. U
Winter
600
800
Strength
1000
1200
(gauss)
Koenig SH, Brown RD III. Relaxometry tissues. In: Gupta RK, ed. NMR spectroscopy of cells and organisms. Vol 2. Boca ton, Fla:CRC, 1987; 75-114.
4.
Relaxation
muscle
F, Kimmich
R.
6.
8.
9.
viable
Y, Muller
RN.
Field-cycling
relaxometry: medical applications. Radiology 1988; 168:843-849. Marg E. Magnetostimulation of vision: direct noninvasive stimulation of the retina and the visual brain. Optom Vision Sci 1991; 68:427-440. Cohen MS. Weisskopff RM, Rzedzian RR,
Kantor varying 7.
for
(c).
NMR field-cycling of bovine serum
spectroscopy
Haverbeke
5.
rates
of normal
albumin, muscle tissue, micrococcus lufeus and yeast: ‘4N’H-quadrupole dips. Biochim Biophys Acta 1982; 719:292-298. Koenig SH, Brown RD III, Adams D, Emerson D, Harrison CG. Magnetic field dependence of 1/T1 in tissue. Invest Radiol 1984; 19:76-81. Rinck PA, Fisher HW, Vander Elst L, Van
KL. Sensory stimulation by timemagnetic fields. Magn Reson Med
1989; 14:409-414. Bevington PR. Data reduction and error analysis for the physical sciences. New York: McGraw-Hill, 1969; 242-243. Beaulieu CF. ClarkJl, Brown RD. Spiller M,
Koenig
References 1.
that
relaxation
is
of
exp’T1(f)J.
400
Field
After a rearrangement of terms, this expression is Equation (I). Equation (1), start with the assumption Of the two contributions to the first that before the application of the initial 90#{176} term in braces in Equation (1), the residual pulse, magnetization is purely longitudilongitudinal magnetization factor is much nal and aligned with the static baseline smaller than the term from regrowth durmagnetic field. Application of the initial, ing the period 8. For example, if the typinominal 90#{176} pulse flips magnetization into cal Ti at the baseline field strength is 300 the transverse plane, with some possible msec, the regrowth in 25 msec is 8% of m,; residual component in the z direction, m2. if the ffip angle is misset by 5#{176}, the residual The cycle field is applied for a time ‘r. The magnetization is only 0.4%. magnetization will exponentially apIf the magnetization has not fully come proach the equilibrium value m0(f) with to equilibrium at the baseline field the characteristic TI constant of the cycle strength by the end of the measurement field. The longitudinal magnetization after phase of the pulse sequence, the residual the time i is magnetization would be slightly depenTo better
200
C.
of the longitudinal end
1000
1200
b.
msec
300
2 600
but
0
SH.
Relaxometry
of calf lens ho-
mogenates, including cross-relaxation by crystallin NH groups. Magn Reson Med 1988; 8:45-57. Davis PL, Sheldon P. Kaufman L, et al. Nuclear magnetic resonance imaging of mammary adenocarcinomas in the rat. Cancer 1983; 51:433-439.
of Ra-
Radiology
639
#{149}
W. Carlson,
PhD
#{149} David
MR Relaxometry Work
relaxation
rates
in extremities
heads
of healthy volunteers. The of the measurement is sufficient to obtain a distinctive relaxation rate dispersion behavior for different tissues. sensitivity
Index terms: MR. 452.1214
technology
Brain, MR, 10.1214 #{149} Magnetic resonance #{149} Magnetic resonance
Extremities, (MR), (MR), tissue
#{149}
characterization Radiology
Antonio
Brito,
#{149}
BS
1992;
184:635-639
resonance (MR) imaging is a useful diagnostic tool in clinical imaging because it provides a wide range of image contrasts. It is well known that image contrast in MR imaging is determined mainly on the basis of the longitudinal and transverse relaxation times Ti and T2, the density of hydrogen nuclei haying a substantial but comparably smaller variability in the body. It is also well known that Ti of tissue varies considerably with different magnetic field strengths. These differences are ultimately due to interactions of ‘
lvi
AAGNETIC
the hydrogen cal environment. Most of the
strength rates
nucleus observable
dependence in the
with
range
its chemifield
of relaxation of field
the Radiologic of California, Dr, South
Imaging Laboratory, San Francisco, 400 San Francisco, CA 94080
(J.W.C., A.B., L.K.); and Tosiba America MRI Inc. South San Francisco, Calif (D.M.G.). Received 13, 1992; revision requested March 16; revision received March 30; accepted April 8. Address reprint requests to J.W.C. C RSNA, 1992
January
PhD
trometers
(2-4)
by
MR
imaging
sequence.
MATERIALS
high
field
strengths.
In
the intermediate regime-a range of field strengths currently being used with clinical imagers-we should see the transition behavior between the limiting cases of very high and very low field strengths. Techniques for measuring the field strength dependence of Ti have been devised for use with laboratory spec-
cy-
article,
Apparatus
Measurement performed with
of relaxation an existing
system for
A 0.064-T
system
ica
South
MRI,
modified
backs. well
magnet,
for
this
of large can
be
variations
a pulsed aperture human resting patients
(20-cm gap). investigation on their sides on their
with
is centered
A pulsed netization value; no reception
magnet
the
relatively
strength.
small
In addition,
magnetic
on
gion of the relaxation
is
because
in relaxation
will see that the baseline strength
was
electro-
imager
changes
explored
Amer-
Calif)
application
of field
field
strength.
whole-body
The low-field-strength suited
elec-
Toshiba
Francisco,
by including
regimen rates
field
(ACCESS; San
was MR
a pulsed of a cycle
magnetic
permanent
imaging
rates clinical
by adding application
the baseline
of a clini-
METHODS
Experimental
magnet with a small This small gap limits to heads for patients or to extremities for
at very
In this
AND
corresponding
of 1
of field
we describe the implementation field-cycling relaxometry with cal, whole-body MR imager.
onto
to a frequency
means
cling. During part of the pulse sequence, an electromagnet is turned on and magnetization is allowed to approach equilibrium in that field. The electromagnet is then turned off and the magnetization measured quickly by means of a conventional
relevant to imaging systems is a result of hydrogen interactions with biologic macromolecules, particularly proteins (1). The typical rotational correlation time for a protein molecule is 1 psec,
stant
From University Grandview
Kaufman,
imaging tromagnet
strengths
MHz. Longitudinal relaxation of a hydrogen spin occurs more rapidly when the Larmor frequency is much lower than the characteristic frequency for protein reorientation than when it is much higher. In general, Ti will approach distinct limiting values at very high and very low field strengths in proteinaceous tissues. At Larmor frequencies much less than i MI-Iz (corresponding to a field strength of 200 G), Ti will decrease to a value independent of field strength. At frequencies much higher than 1 MHz, Ti will approach a higher value and again become con-
I
Leon
#{149}
in Progress’
of relaxation rate disperfrom magnetic resonance images was demonstrated on a clinical, whole-body imaging system. Study of the behavior of relaxation rates over a range of field strengths probes the structural environment of imaged hydrogen protons and reveals information about the composition of tissue. The authors deterand
PhD
Imaging
Acquisition sion curves
mined
M. Goldhaber,
an
we
field
interesting
re-
rate spectrum. is used
to approach radio-frequency is performed
to allow
mag-
an
equilibrium transmission while the pulsed
electromagnet
is on, nor
magnetization
undergoing
is any
or
transverse
free preces-
sion. Therefore, the homogeneity of the electromagnet is relatively unimportant. For example, although the homogeneity of the static magnetic ppm, the homogeneity
tromagnet typical
field is typically of the pulsed
is on the order field
Abbreviations: repetition time.
of 30%
50 elec-
across
a
of view.
ROl
=
region
of interest,
TR
=
635
A small electromagnet duce power requirements ing.
The
of two
was used to refor fast switch-
electromagnet
circular
we
bundles
cross-sectional
consisted
(ROIs) in particular tissues. A numerical best fit of the data to a monoexponential approach to an equilibrium value was
with
formed
used
of wires
dimension
7 x 7
of
a
cm
an average radius of 10 cm. A current 60 A provided a maximum variation 450
G with
these
tromagnet
was
parable fiers.
measurements.
driven
to those Voltage
supplies
as gradient
limitations
resulted
The
com-
ampli-
on the current
in a rise
time
to full
cycle
field strength of 30 msec. The fall time of the electromagnet was reduced substantially to approximately 8 msec with the use of a flying capacitor dump circuit. The rate of change of the magnetic field strength in these measurements was 6 T/sec, which is sufficiently high to induce retinal magnetophosphenes (5) but below levels for motor nerve stimulation reported in echo-planar imaging (6). In the imaging pulse sequence, we applied an initial 90 pulse to flip magnetization
from
the
static
field
direction
into
the
transverse plane. The pulsed electromagnet was then turned on for a time r, ranging from 50 to 350 msec. Transverse magnetization is quickly and irretrievably dephased in the highly nonuniform field. The cycle field was then turned off, and
the field was allowed
to settle.
shielding of the pole piece counterwound “pancake”
nated
eddy
dard
imaging
formed
currents
without
contamination
baseline field strength. To obtain an accurate essential
that
a stan-
the
of the
through
it is the
elec-
tromagnet be turned off completely. This requires an introduction of additional delay time for the current to fully settle. The time between turning off the amplifier pulse and start of a pulse sequence was 25 msec.
After
the current
two-dimensional
echo sequence
settled, Fourier
a standard transform
was performed
spin-
nal from an asymmetric echo was collected, with an echo time of 30 msec, a section thickness of 10 mm, a 1.7 x 1.4-mm in-plane resolution, and a 128 x 256 ma-
trix. Field
cycle time was the same
for all
phase-encoded acquisitions, then it was increased and the imaging sequence repeated. The time of cycle field application ranged from 50 to 350 msec in steps of 75 msec, for a total of five different values. With the 25-msec delay time for current settling and 150 msec for performing ra-
dio-frequency
pulses,
data
acquisition,
and unused sequence time, the repetition time (TR) was 225-525 msec, for a total imaging time of 4 minutes per field strength value. Current investigations with this apparatus have included extremities and heads of healthy volunteers and a rat model.
Data
Analysis
Mean signal from manually
636
Radiology
#{149}
intensities located
were regions
echo
recorded of interest
be
Two
in-
into
ac-
taken
sequence
results
in some
ph
zation,
magnetization
90#{176}-I80#{176} measurement following form: m
+ {[m(f)
-
exp_b/TI(fo)J}
‘1’ ‘
The signal
intensity
an
of the collected magnetization
exponential
attenuation
data is
multiplied from
T2
decay. The first term in braces in Equation (1) contains the two instrumental details described earlier; this term is a constant independent of the cycle field strength. The second term in braces is linearly dependent on cycle field strength by means of the equilibrium magnetization value mO(f). The derivation of Equation (1) is
explained The
in the Appendix. most
reliable
estimates
of TI are fit to all data from each of the selected ROIs for all cycle field strength values. Thirty different cycle field strength values were used to obtain a typical set of images in these preliminary studies. As a result, data analysis required a 33-parameter fit to the data: one parameter for the constant term in Equation (I), two for the linear function in Equation (1), and 30 for the Tls at the different field strengths. In analyzing the data, we did not assume any relationship between the Tls at different field strengths. The computational effort required for this analysis is modest. Before performing a global fit of the data, only the data from a single field strength value are fit to a three-parameter model. These results are then used as a starting point in a simplex search for a global best fit. The entire data analysis for each ROI at all field strength values required approximately 2 minutes with use of a minicomputer. Errors were estimated for the individual data points by evaluating the variance of the data points from the best-fit curve, that is, the reduced 2 was set equal to 1. Upper and lower confidence limits of Tl
by performing
a global
the
inhomogeneous
cycle
field.
The
expT/I(f)J.
-
by
The regrowth occurs for a time r, at an increased or decreased field strength. The electromagnet is turned off, and, after allowing the current to completely shut off, the magnetization is measured with a conventional spin-echo two-dimensional Fourier transform sequence. rf = radio frequency.
mz]exp_b/TI(fo)}
-
-----
dephased
(m) before the sequence has the
+ m,(f0)[l
mexpbhhl(fo)
=
times.
-
Figure 1. Diagram of pulses for the relaxation rate measurements. An initial 90#{176} pulse flips longitudinal magnetization into the transverse plane, where it is quickly
relaxation
and the cycle and delay
longitudinal
d
r
to the baseline field strength magnetization at the TI of that field. The assumed model of signal intensity depends on the TI at the cycle field strength, Tl(f), the TI at the baseline field strength, Tl(f,), the equilibrium magnetization values at the cycle and baseline field strengths, m0(f) and m,(f0), respectively, the residual longitudinal magneti-
achieved
(Fig 1). Sig-
density. must
per-
relaxation
count for this data analysis to be successful. First, transmitter coil nonuniformity or misadjustment will result in nonzero residual magnetization, m,, along the static field after application of the 90#{176} prepulse. Second, the delay time 6 between turning off the current and the start of the spin-
by
measurement,
current
details
the longitudinal
to be per-
the tissue
magnetization
x [1
with use of a winding elimi-
sequence
and
Active
and allowed
pulse
rate
strumental
elec-
by amplifiers
used
and
of of
to calculate
f,Ld
I
were established by performing a numeric search for the values of TI to find where x2 increased by 1 (7). The confidence limits describe only random errors in the data and do not help estimate effects from systematic errors in the instrumentation. Cycle field strength over the entire field of view varies by approximately 30%. Therefore, we must be careful to calibrate the local field strength when presenting data of relaxation rates. The regions studied herein were of the electromagnet,
always
ations were much been calibrated.
close so that
less
than
to the center the field van-
30%
and
had
RESULTS The relaxation process for hydrogen nuclei is the result of a combination of many effects at the molecular level. Individual effects, if they act independently, contribute additively to the overall relaxation rate (the inverse
of Ti).
pendent come
Contributions
relaxation more
plotted
from
processes
apparent
when
as relaxation
rates.
Phantom
inde-
will data
beare
Studies
Archetypal behavior of the relaxation rates as determined with phantom studies are shown in Figure 2. Relatively constant relaxation rate behavior over the field strengths measured
was
seen
in phantoms
contain-
ing simple solutions (Fig 2a), such as a solution of a paramagnetic relaxation agent, nickel chloride, and water or an emulsion of mineral oil and water (approximately i :1 mixture of mineral oil and
water
oleate and monooleate Phantoms romolecules,
with
sorbitan
mono-
polyoxyethylene-sorbitan added as emulsifiers). containing biologic macsuch as hen egg albuSeptember
1992
6
T
300 400
2 1000
400
Field
600
Strength
800
000
ues were returned
400
decreased
:
600
200
300 :
6
4
S
)
2OO
are probably artifacts of the order of data acquisition. Data were initially acquired at the baseline field strength of the magnet; the field strength val-
150
150
1200
0
200
400
(guss)
F1d
600
Strength
800
1000
1.00
( g.ss)
b. Figure
2. rates
ation emulsion strength phantom strengths associated phantom
Relaxation rate dispersion curves for in vitro samples. (a) Curves show the relaxfor phantoms containing a solution of nickel chloride and water (solid line) and an of mineral oil and water (dotted line). Relaxation rates were insensitive to field over the ranges investigated and showed no apparent structure. (b) Curves from a containing hen egg albumen show a comparably larger change over the field studied. The denatured egg phantom (dotted line) exhibited peaks at 500 and 650 G with nitrogen cross-relaxation that were not present in the native egg albumen (solid line).
higher than the native value. Also, structure appeared in the relaxation rate curve, with the appearance of two peaks at approximately 500 and 650 C. These peaks were ascribed to cross-relaxation effects of proton interactions with nitrogen nuclei in cross-linked protein molecules (2). These peaks represent not only the chemical composition of the phantom but also the structure of the composition. These peaks in relaxation rates
are a result
of the
short-range
the tissue caused by heating phantom (8). The field strength values
peaks
Figure
3.
Representative
MR image
ob-
tamed in the knee of a healthy volunteer (TR = 525 msec, echo time = 30 msec [525/ 30]). This is one of the five images used to determine relaxation rates. The cycle field used in this image was set to give a total field strength of 1,200 G.
men (Fig 2b), exhibited substantial changes in the relaxation rates for these same field strengths. The lower curve in Figure 2b exhibits a decrease
in the
relaxation
rate
from
3 sec-1
(Ti,
333 msec) at 200 G to 2 sec (Ti, 500 msec) at 1,100 C. Substantial changes occur in egg albumen with heat denaturation. After a phantom containing albumen was heated at approximately 60#{176}C for 10 minutes, the overall relaxation rate increased, as did the dispersion in the rates. At 200 C, the rate had nearly doubled, to roughly 5 sec (Ti, 200 msec), and decreased to 2.5 sect (Ti, 400 msec) at 1,100 C, a rate 25%
Volume
184
Number
#{149}
3
are
a result
of the
order
of these
quadrupole
transition frequencies of the nitrogen nuclei and are independent of the chemical composition of the material. Winter and Kimmich (2) calculated three field strength values for these peaks: i95, 490, and 685 C, with the peak at lowest field seen only in highly ordered biologic tissues. Koenig and Brown (i) reported slightly different values for the higher two peaks, of 500 and 655 C.
Human
Volunteer
marrow,
and
Study
fat; data
The
were
macromolecular
tissue
with
substan-
tially less structure. Data obtained the region of the pons exhibited stantial structure in the relaxation behavior (Fig 5c).
in subrate
Study
The measurements obtained in human volunteers are intended only to help assess the sensitivity of the technique in the differentiation of normal tissue. To determine both the feasibility and potential of the technique in the evaluation of pathologic tissue, we obtained measurements in a rat with an implanted mammary adenocarcinoma (9). A representative transaxial image of two rats is shown in Figure 6. The tumor had a discernable necrotic center surrounded by a viable region. Relaxation rates for the necrotic center, viable region, and normal rat muscle tissue are shown in Figure 7. The overall
relaxation
rate
of the
necrotic
center
of the slightly
tumor (Fig compared
7b) decreased with that
viable
region
7a). A small,
statistically seen in the Conversely,
field tially This
(Fig
of the
but
significant, structure was necrotic center at 650 C. the normal muscle (Fig
strengths, as well larger cross-relaxation enables a qualitative
tative differentiation and muscle in this
higher at lower
re-
as a substanstructure. and quanti-
between particular
tumor model.
DISCUSSION
pro-
cessed as described earlier. The basic behavior of the three regions is shown in Figure 4. The fat (ROI volume, 1.2 cm3) and bone marrow regions (ROI volume, 2.5 cm3) showed relatively flat relaxation rate behavior, with little structure apparent. The small structures seen at 640 C, the baseline field strength of the magnet,
then muscle
region, however (ROI volume, 1.5 cm3), exhibited an overall behavior similar to that seen in the cross-linked protein phantom of the previous experiments. This same sequence was repeated with transaxial sections of the brain of the healthy volunteer. Relaxation rate dispersion in white (Fig 5a) or gray (Fig 5b) matter is typical of that of
7c) showed a substantially laxation rate, particularly
A set of measurements was obtamed from sagittal images acquired in the knee of a healthy 34-year-old man. A representative image of the knee is shown in Figure 3. ROIs were located in different regions of muscle,
bone
incrementally, value, and
incrementally.
Rat Model
in
the
increased to baseline
These
intended
preliminary
only
of relaxation rate a clinical whole-body
imaging
time
experiments
to show
the
are
feasibility
determination imager.
of 4 minutes
with With an
per
field
point, the total data collection time for the 32 field points used in the human studies was approximately 2 hours.
Patient
acceptance
with
the
imaging
Radiology
637
#{149}
I00
100
100
8
r ISO
ISO
S
6
‘, 3)
20(1
400
600
Field
800
StI’ngth
a. Figure
1000
200
200
300
300
400
400
1300
600
1000
1000
200
I
200
400
(glJ)
FIeld
600
800
1000
200
300 400
2 600
200
rates for fat (a), bone marrow indicative of the short-range
400
(g.us)
Strength
600
800
Strength
Field
b.
4. Relaxation cross-relaxation peaks
1000
1200
000
12
0
(gauss)
C.
(I,), and muscle (c) in a healthy order in this tissue.
volunteer.
The long
muscles
of the extremity
showed
large
200
200
300
300
400
400
600
600
i S
10011
CO
200
400
600
800
1000
1000
120(1
I Fiald
200
F,td
a.
600
400
(gauss)
Strangth
600
1000
1200
0
sion
5. Relaxation components with
ROIs
were
times
3.0 cm3, I .9 cm3, and
used
would
rate dispersion little apparent
in these
be poor,
of field
Our
data
determine structure
show
matter,
in the of the
brain pons
of a healthy volunteer. (c) shows a much larger
gray
matter,
and
pons,
shorten reducing
by
will
points that
than were We can
allow
can
used reduce
us to acquire
in this the
Figure
images
positioning). is needed
Furto deter-
minutes.
The clinical applicability of this procedure may lie in observations of changes in nitrogen peaks associated with structural changes in the underPathologic
such
structural
as demyelination
conditions
changes
in
of brain
6.
Considerations
of total
imaging
aging.
In both
ence
is that
dimensions
hence, tional
spectroscopy
spectroscopy of spatial
total imaging to the product
ings.
However,
relax-
from along differ-
requires
MR images
(525/30)
disperof the
of two rats with
im-
phase
total
encoding;
imaging
time
is
independent of the number of frequency points of spectroscopy information. In relaxation rate dispersion measurements, the total imaging time is dependent on the product of only one
of the
spatial
dimensions
and
strength
of the
magnet,
we
the
field
were
to probe the cross-relaxation peak structure and very low field relaxation rate dispersion with relatively small changes in field strength. Attempts to perform this procedure with a mediumor high-field-strength
MR imager
two
time is proporof spatial encod-
peaks.
Radiology
remiim-
and
number of field points sampled. Because of the low baseline
#{149}
Transaxial
resolution are in spectroscopic
or scar tissue in muscle, may be mdicated by changes in cross-relaxation
638
show large The volumes
planted mammary adenocarcinomas. A small necrotic region is visible in the center of a large tumor in the leg of the rat on the right side of the figure. As in Figure 3, only one of five images obtained at 1,200 G is shown.
of
mine if we can reliably calculate Ti with less than five different time delays; if so, we could reduce examination time further, to approximately 15
tissue,
(b) matter peak effect.
clearly
ometry, additional information each image voxel is acquired with the image. The primary
with
gray
peak fewer
including patient ther investigation
tissue.
White (a) and cross-relaxation
measured. we
time and spatial niscent of those
associated
1200
respectively.
the head in 64 phase-encoded steps. These reductions will result in a total examination time of 30 minutes (not
lying
1000
the the
number of field points to 15 and still have sufficient data to characterize both cross-relaxation peaks and the overall trend in relaxation rates. A sacrifice in image resolution to 1.4 x 3 mm
800
( aa.s)
we anticipate
cross-relaxation with substantially
data points investigation.
600 Strngth
C.
determined A region
2.0 cm3, for white
400 F,1d
experiments
but
that we can substantially overall imaging time
number
curves structure.
200
S trngth
b.
Figure
1000
able
tions
This culties
larger field technical
with
changes,
experience Future limited
require baseline more
associated
strength
field
will
from the will cause
and
larger the
variastrength. diffi-
field patient
will
larger rates of flux change. work with use of uniform
cycle
coils
would
to extremity
probably
and
head
be
imag-
ing because of the drawbacks associated with larger-aperture coils. However, a nonuniform “surface” field coil could be used to investigate relax-
ation rates nonuniformity
of the heart muscle. The of the cycle field September
1992
.
ISO 6
150
ISO
6 200_
200
4
200
I
4 300
300
400
400
400
600
600
2
_
1000
200
400 Fild
600
800
1000
1000
1200
200
400
(gauss)
St,angth
800 (gauss)
600
Strsngth
Field
a.
1000
Figure
7.
region
(a) and
would niques
Relaxation
rate
necrotic
dispersion
region
curves
determined
in the center
of a tumor
require data processing techto calibrate the field of the coil,
_O
this
is not
a formidable
problem.
for
a rat
(b) decreased,
of this
next
[m2exp_ThhI(f)
x expStTI(fo)
APPENDIX understand
the derivation
+ m,(f)[l
m2exp’T1(I)
-
with
an
with
implanted
little
mammary
variation
magnetization period
+ m,(f)(l
adenocarcinoma.
compared
at the
The cycle field is then turned off and declines to within I % of the baseline field strength within 8 msec. An additional 17 time
delay
is necessary
for
the
re-
maining cycle field to completely decay. We can approximate that the field strength for the entire 25-msec period is the baseline field strength. The evolution
Volume
184
Number
#{149}
3
with
2.
-
+ m0(f0)(I
expT1f)1 -
3.
exp8/Tl(1o)).
dent on the TI at the baseline field strength and the TR. However, because this term is small and can usually be neglected in comparison to the regrowth contribution, this effect is not important. U
Winter
600
800
Strength
1000
1200
(gauss)
Koenig SH, Brown RD III. Relaxometry tissues. In: Gupta RK, ed. NMR spectroscopy of cells and organisms. Vol 2. Boca ton, Fla:CRC, 1987; 75-114.
4.
Relaxation
muscle
F, Kimmich
R.
6.
8.
9.
viable
Y, Muller
RN.
Field-cycling
relaxometry: medical applications. Radiology 1988; 168:843-849. Marg E. Magnetostimulation of vision: direct noninvasive stimulation of the retina and the visual brain. Optom Vision Sci 1991; 68:427-440. Cohen MS. Weisskopff RM, Rzedzian RR,
Kantor varying 7.
for
(c).
NMR field-cycling of bovine serum
spectroscopy
Haverbeke
5.
rates
of normal
albumin, muscle tissue, micrococcus lufeus and yeast: ‘4N’H-quadrupole dips. Biochim Biophys Acta 1982; 719:292-298. Koenig SH, Brown RD III, Adams D, Emerson D, Harrison CG. Magnetic field dependence of 1/T1 in tissue. Invest Radiol 1984; 19:76-81. Rinck PA, Fisher HW, Vander Elst L, Van
KL. Sensory stimulation by timemagnetic fields. Magn Reson Med
1989; 14:409-414. Bevington PR. Data reduction and error analysis for the physical sciences. New York: McGraw-Hill, 1969; 242-243. Beaulieu CF. ClarkJl, Brown RD. Spiller M,
Koenig
References 1.
that
relaxation
is
of
exp’T1(f)J.
400
Field
After a rearrangement of terms, this expression is Equation (I). Equation (1), start with the assumption Of the two contributions to the first that before the application of the initial 90#{176} term in braces in Equation (1), the residual pulse, magnetization is purely longitudilongitudinal magnetization factor is much nal and aligned with the static baseline smaller than the term from regrowth durmagnetic field. Application of the initial, ing the period 8. For example, if the typinominal 90#{176} pulse flips magnetization into cal Ti at the baseline field strength is 300 the transverse plane, with some possible msec, the regrowth in 25 msec is 8% of m,; residual component in the z direction, m2. if the ffip angle is misset by 5#{176}, the residual The cycle field is applied for a time ‘r. The magnetization is only 0.4%. magnetization will exponentially apIf the magnetization has not fully come proach the equilibrium value m0(f) with to equilibrium at the baseline field the characteristic TI constant of the cycle strength by the end of the measurement field. The longitudinal magnetization after phase of the pulse sequence, the residual the time i is magnetization would be slightly depenTo better
200
C.
of the longitudinal end
1000
1200
b.
msec
300
2 600
but
0
SH.
Relaxometry
of calf lens ho-
mogenates, including cross-relaxation by crystallin NH groups. Magn Reson Med 1988; 8:45-57. Davis PL, Sheldon P. Kaufman L, et al. Nuclear magnetic resonance imaging of mammary adenocarcinomas in the rat. Cancer 1983; 51:433-439.
of Ra-
Radiology
639
#{149}
