MagflftiC Steven

D. Wolff,

MD,

PhD

#{149} John

Eng,

MD

#{149} Robert

Magnetization Transfer Method for Improving in Gradient-Recalled-Echo

A method that improves contrast in gradient-recalled-echo magnetic resonance images is described. The increase in contrast is produced by applying low-power radio-frequency irradiation 5-10 kHz from the main water proton resonance frequency to excite macromolecular hydrogen-i nuclei as part of a conventional gradient-recalled-echo sequence. The contrast so obtained is theoretically different from Ti or T2 contrast and has been termed magnetization transfer contrast. Cat head images were acquired in vivo with this method at 4.7 and 1.5 T. Compared with conventional gradient-recalled-echo images, the magnetization transfer images demonstrate increased contrast between many tissue pairs, such as between white matter and blood and between gray matter and cerebrospinal fluid. The dependence of the magnetization transfer effect on repetition time and preirradiation power were also studied. Index terms: Cerebral blood vessels, MR studies, 17.1214 #{149} Brain, MR studies, 10.1214 #{149} Cerebrospinal fluid, MR studies, 10.1214 #{149} Magnetic resonance (MR), experimental #{149} Magnetic resonance (MR), physics #{149} Magnetic resonance (MR), pulse sequences #{149} Magnetic resonance (MR), surface coils #{149} Magnetic resonance (MR), technology #{149} Muscles, MR studies, 40.1214 Radiology

1991;

179:133-137

T

Imaging

PhD

Contrast: Contrast Images’

diagnostic value of hydrogen1 magnetic resonance (MR) imaging depends on the production of images with high signal-to-noise and contrast-to-noise ratios. The contrast in these images largely results from differences in spin-lattice (Ti) and spin-spin (T2) relaxation times. Howeven, because the ability to discriminate between two tissues on the basis of Ti and T2 is limited, various at-

ponent prevents detection macromoleculan protons MR imaging procedures.

tempts

stant map provides unique information about the interaction between water and macromolecules, it mequines a long imaging time because of the need to generate a Ti map. The purpose of this study was to investigate whether a magnetization transfer image alone can be useful by providing better image contrast. Our studies were directed to imaging the cat head because of the variety of tissues present from which contrast values could be calculated (gray matter, white matter, cenebrospinal fluid [CSFJ, blood, muscle, and fat). Fastimaging gradient-recalled-echo images were collected with and without RF irradiation, and the contrast between the various tissue pairs was analyzed. The effects of repetition time (TR) and RF irradiation power on the magnetization transfer effect were studied. The results indicate that for many tissue pairs the magnetization transfer images have superior contrast without significantly increasing the time required for data

HE

have

been

made

to increase

tissue contrast on MR images. For cxample, panamagnetic agents such as gadopentetate dimeglumine have been administered intravenously to enhance differences in relaxation mates between pathologic tissue and surrounding edema in the central nervous system (1). Efforts to increase contrast in MR images without the administration of foreign agents have relied largely on pulse sequences sensitive to motion, such as that accompanying blood flow on waten diffusion (2,3). Alternatively, othem nuclei have been imaged (eg, sodium-23) in the hope of generating high-contrast images of the brain (4). Recently, a report from our labomatory described a new form of contrast, magnetization transfer contrast (MTC), which is sensitive to the interaction between water and macromolecules (5). In this technique, lowpower radio-frequency (RF) irradiation is applied off resonance to selectively saturate H-i nuclei with short T2 as part of a standard imaging sequence. The saturation of the short T2 component is then transferred to the bulk water protons via a magnetization

I From the Laboratory of Cardiac Energetics, National Institutes of Health, Bldg 1, Rm B3-07, Bethesda, MD 20892. From the 1989 RSNA scientific assembly. Received July 27, 1990; revision requested September 18; revision received October 17; accepted October 18. Address reprint requests to R.S.B. 0 RSNA, 1991

S. Balaban,

Resonance

exchange

process

(in-

volving chemical exchange and/or through space dipolan interactions [5]), which is highly selective for waten protons (5). The short T2 component of the proton spins is believed to be composed of macnomolecu.lar protons with severely restricted motion (ie, membranes, proteins, etc). The rapid T2 relaxation of this corn-

tization

transfer

image

of these in normal The magnecan

be

com-

bined with a standard image and a Ti map (collected in the presence of RF irradiation) to generate a “rate constant map” in which pixel signal intensity denotes the pseudo-first-orden nate constant for magnetization exchange

acquisition.

(6).

Although

Images

the

of the

rate

cat

con-

head

were also obtained at 1.5 T on our clinical MR imager and yielded qualitatively similar results, indicating the applicability of the magnetization transfer technique to the imaging of humans.

Abbreviations: CSF MTC = magnetization radio frequency, TE tion time.

cerebrospinal fluid, transfer contrast, RF echo time, TR repeti-

-

133

MATERIALS

AND

METHODS 70000

Animals

1.0 -#{149}-...

i60000 50000

All data were shorthair

cats

hours

prior

obtained that

from

were

not

domestic

fed

for

12

-#{149}--

,

0.8

-#{149}--

V., .5,.

0.6

-.,40000

EEEE

0.4

30000

to anesthesia.

0

1000

2000

3000

4000

0

1000

181....)

Equipment Studies

2.

at 4.7 T were

spectrometer

performed

(General

on our

Electric,

Fremont,

Calif) in a 26-cm clear bore magnet. The RF coil used in these studies was a curtom-built

low-pass

“bird-cage”

coil

10 cm

in diameter and 9.5 cm in length. It was constructed from eight foil struts placed on edge with respect to the sample to minimize dielectric losses (7). The unloaded quality factor of the coil was 850. The quality factor decreased to 45 when loaded with the cat head. The amount of tissue exposed to the RF irradiation was estimated to be about 500 g. Modifications to our spectrometer that were important to this investigation consisted of the use of an additional amplifier (model 100LM8;

Amplifier

Research)

for

provid-

ing the RF presaturation and the use of a custom-built power combiner to combine the power from the two RF channels. Anesthesia was induced by intramusculam injection of keta.mine. The cats were then intubated, ventilated (9000D; Siemens Medical Systems, Iselin, NJ), and anesthetized with a mixture of 30% oxygen, 69% nitrogen, and 1% halothane (tidat volume, All animals

1.5

L; 32 breaths

per

minute).

survived the procedure with no noticeable ill effects. The studies at 1.5 T were performed on a clinical imager (Signa; GE Medical Systems, Milwaukee) with version 3.3 software and equipped with shielded gradients. Studies were performed by addition of a custom-built RF channel for the RF irradiation without modifying existing spectrometer hardware or software (Wolff SD, Chesnick S, Balaban RS, manuscript in preparation). Anesthesia was induced and maintained on a Plexiglas cylinder 15 cm in diameter. The unloaded quality factor of the coil was 1 10. The quality factor decreased to 82 when loaded. Power values reported here were measured with a digital RF power analyzer (model 4391; Bird Electronics, Cleveland). There are many ways to report the amplitude and power of the radiation used in this

study.

These

include

the

Bi magnetic

field strength (in tesla or gauss), the precession frequency of water protons in this field (in hertz), or the actual power deposited in the tissue (in watts or watts per kilogram). We have calibrated our coils for the 4.7-T studies for all of these variables. For this coil, the Bi field (in tesla) was related to the power deposition by the following equation: Bi (X 106) /[-O.45 + (0.57 X watts)]. The precession frequency (in hertz) was related to power by the following equation: frequency J[-7,254 + (10,232 x watts)J. These equations can be used to convert power values into terms more familiar to the reader. i34

#{149} Radiology

2000

3000

4000

1800...)

3.

Figures

2, 3. (2) Plot of raw signal intensity versus TR. (3) Plot of Ms/Mo versus TR. The effectiveness of magnetization transfer was assessed by dividing the tissue signal intensity of the MTC image (Ms) by that obtained from the control image (Mo). The tissues listed are blood vessels (blood), CSF, brain gray marten (gray), brain white matter (white), and skeletal muscle (muscle). Signal intensity values are arbitrary.

Imaging

dcl

Procedure

For the 4.7-T studies, selective saturation of the shomt-T2 H-i nuclei was accomplished by applying RF irradiation (0.1-1.5 X iO T) through the decoupler channel at 10 kHz below the resonance frequency of water H-i (5,6). Previous studies from this laboratory have demonstrated that this level of irradiation is specific for the macromolecular component, with minimal direct effects on the bulk water signal (6). Figure 1 shows the pulse sequence used in these studies, in which a 50_100 flip angle was used for observation. The pulse sequence was a modified version of a simple single-section, smallflip-angle gradient-recalled-echo pulse sequence (8) in which the time period sometimes referred to as the “predelay” is replaced with a low-power BY pulse of constant

magnitude.

For

all

studies,

Analysis

For the 4.7-T studies, images constructed by two-dimensional

were reFourier

transformation

calcula-

tion

and

with

the resident

magnitude software

General

Electric

structed ry form

images were transferred to a Sun Microsystems

UNIX-based

spectrometer.

computer

analysis

gz Time

-

sequence for acquisition of magnetization transfer images. The x axis (which represents time) is not drawn to scale. dc = RF envelope of decoupler amplifier (10 kHz off resonance); rf RF envelope from standard imaging amplifier (on resonance); gx, gy, gz magnetic field gradient strengths along the x, y, and z axes, reFigure

1.

Pulse

spectively.

two tissues. For the studies at 4.7 T, hardware limitations caused an increase in noise in the magnetization transfer images due to a poor gating arrangement for the decoupler, which injected noise into the system. (No increase in noise was observed in the 1.5-T studies.) Because the increase in noise is not inherent in the technique, simple contrast values (as opposed to contrast-to-noise values) are reported. This conclusion is justified on the basis of our l.5-T studies and recent studies at 4.7 T with custom-built decoupler gating apparatus. RESULTS

recon-

in bina3/260

workstation.

of tissue

gy

in the The

Pixel

calculations and image analysis were performed with custom-designed software written in the C programming language for the Sun 3/260 workstation equipped with a Pixar image computer for image display. For the 1.5-T studies, images were reconstructed with the Signa system in the standard manner and were analyzed with the resident software. Quantitative

gx

the

power reported is for this predelay time only. No attempt was made to calculate the average power deposition. For the studies at 1.5 T, analogous experiments were performed with a custom-built second RF channel in conjunction with the GRASS (gradient-recalled acquisition in a steady state) sequence resident in the systern software (Wolff SD, Chesnick 5, Balaban RS, manuscript in preparation). Data

rt

signal

intensity was performed by examining regions of interest from each tissue. Contrast was defined as the absolute value of the difference in signal intensity between

Effects

of TR

on MTC

Because imaging time is important in MR imaging, studies were performed to determine the effectiveness of the magnetization transfer technique in pulse sequences with short TRs. MR images were collected at 4.7 T as a function of TR while preirradiation

power

was

held

con-

stant at 4 W (5 X i0 T). The signal intensities of the different tissues on the MTC images were compared with those on control images obtained unden identical conditions but without the off-resonance irradiation. Figure April

i99i

.

k

I, ‘p

:

1

I

31

b.

c.

d. Figure

4. Effect gradient-recalled-echo brain.

.

#{149}*#{188}.i 5

.

qq

“

f’

,.

Acquisition

lion

thickness,

mm; step;

four 256

parameters: 420/ 10; secfield of view, 102 per phase-encoded points acquired per

acquisitions complex

step;

and

coded steps. Irradiation Hz), (b) 0.25 W (40 Hz),

f-

ent -..#{149}--

bk..d

-.--

0000

-..#{149}-.

irradiation

V.,

the

pairs

Figure 3

4

Figure

5. Effect of irradiation power sue signal intensity. Same tissues are as in Figures 2 and 3.

on tislisted

and

intensity

versus

Figure

5

of each

tis-

irradiation

Figure which served,

that

power is increased, transfer effect

6a shows an increase while

as preimra-

for was

6b shows

2 presents

from

the

maw

different

signal

tissues

white

intensities

in the

gray

cat head

as a function of TR, with off-resonance irradiation. Figure 3 shows the ratio of tissue signal intensity (signal intensity in MTC divided by that in control) as a function of TR. Figure 3 demonstrates that almost the full

MTC effect TR of about stantial shorten

Effect Image

can be obtained 400 msec, and

effects TRs.

can

with a that sub-

be generated

images of the cat head collected at 4.7 T as a function

were of

MR

preinradiation

power

(TE)

msec

stant at 420/10. preinradiation

msec

was varied strengths

while

was

In these time was

and

TR

held

W.

angle.)

a series

of images

Volume

179

Figure

collected

#{149} Number

1

note

between

CSF.

of the cat head at 1.5 T with

were a similar

procedure to determine for improving image

the potential contrast at lower

clinical

field

MR

imaging

7 shows

strengths.

a magnetization of a cat control

100%

blood tween

observed in the these parameters,

ratio between

head image

and a ob-

4 shows

4.7-T the

studies. contrast-

enhanced

white

by about

matter

and

than 200% and CSF.

The a

purpose

of this

dient-necalled-echo sults

indicate

study

images. that

when

in this

be-

was to transfer in gra-

The MTC

required

to pro-

study

were

me-

is used

images

collected

with

tle Ti or T2* weighting small

nutation

lit-

by means

angles

and

short

of TEs

(about 10 msec). However, because the effects on signal intensity of Ti and T2 track inversely in this imaging sequence, the signal intensities of the gradient-recalled-echo images meflected some net T2 weighting even at the shortest TRs. As we previously reported (5), the signal intensities of a magnetization transfer image tend to parallel those seen on a T2-weighted image (and track inversely those seen on a Ti-weighted image). The increase in contrast might not have been realized had the gradient-mecalled-echo images been collected with larger flip angles resulting in a Ti-weighted

image.

An example of this can be illustrated by comparing the 4.7-T MTC image with a 15#{176} flip angle and the 1.5-T image in which a 30#{176} flip angle was used. As seen in the latter images, the larger flip angle resulted in a weaken signal

determine if magnetization could increase tissue contrast

gradient-ne-

acquisitions

predominantly

DISCUSSION

field the

at differ-

was

and by more gray matter

power (Bi from

fact that an on-resonance, 2.4-msec square RF pulse at 3.7 W produced 180#{176}nutation

images

to-noise

msec/

con-

preimnadiation

and

me-

T

collected

those With

studies, RF held constant

from 0 to 5.8 can be estimated

ob-

tamed without irradiation. One can see changes in contrast similar to

on

at 400

at i.5

MR

blood

and

transfer image corresponding

of Irradiation Contrast

time

MTC

also

and

matter

Figure

with

Power

echo

matter

with

duce a single image. The gradient-recalled-echo

the

maining tissue pains. Of special are increases in contrast between

in conjunction

called-echo imaging, contrast is improved between most tissue pains. This finding is even true at short TRs because a high degree of saturation is attained in the steady state due to the many

the magincreases.

tissue pairs in contrast

Figure

6 tis-

power.

4 demonstrates

diation netization

6

5

(W.It.)

P.,,,

powers, signal

sue as a function of power. Figure presents the contrasts for various sue

2

Hz),

g.

shows

I

3.7 W (196

128 phase-en-

power: (a) (c) 0.45 W

0 W (0 (60 Hz), (e) 2.0 W (131 Hz), and (g) 5.8 W (250 Hz).

(d) 0.90 W (86 Hz),

(f)

power on of the cat

3 mm;

phase-encoded

;5

of irradiation images

from

the

long

Ti

components

(ie, CSF), which resulted in a weaken MTC effect when compared with the smaller-flip-angle

data.

The

precise

mechanism by which magnetization transfer generates contrast remains be elucidated. However, since the Radiology

to

#{149} 135

1500

1500

csfblood

-0----

1000

.

.

csf-muscle

-0--S

csf-white

.

csf-gray

1000 8) 8)

C 0 C-)

muscle-blood

-0----

#{163}

graywhite gray-blood

-0---

500

500 0

musclewhite

S

0

Power

Power

(watts)

(watts)

b.

a.

Figure

6.

Effect

tissue pairs. nal intensities

er

technique

of irradiation

(Note of the

that two

power

on tissue

gray matter-blood tissues reverse.)

pamamagnetics.

Our preliminary study of the cat head at 1 .5 T suggests that this technique may be useful for clinical diagnostic studies. Because the magnetization transfer technique is noninvasive, it should be relatively easy to apply to imaging of humans. Howevem, careful consideration must be given to RF power limitations. In the present studies at 4.7 T, 4 W provided ample power for substantial magnetization transfer effect. The amount of absorbing

the

RF irradiation

was nominally 500 g. If one assumes all transmitted power was absorbed by the tissue, the irradiation power was about 8 W/kg of tissue. This exceeds the current Food and Drug Administration

sorption for the

(a) tissue

pairs

for which

contrast

contrast decreases and then increases with Same tissues are listed as in Figures 2 and 3.

magnetics chemical contrast T2 (5,6). Studies from this laboratory have shown that the magnetization transfer effect is sensitive to the intemaction between water and macnomolecules (5,9,10). Fralix et al (9) demonstrated that the correlation time (ie, motion characteristics) of the watermacmomolecule complex and the presence of hydmoxyl groups such as those found in cholesterol are impontant parameters that modulate the mate of magnetization transfer in lipids. Magnetization transfer images, as opposed to conventional images, may provide a noninvasive way to differentiate between magnetization transfer relaxation mechanisms and other relaxation processes such as

tissue

contrast:

improved increasing

with

magnetization

irradiation

power

transfer; because

the

(b) all othrelative

sig-

is not sensitive to some memechanisms such as paraand at least some forms of exchange, it will provide different from that of Ti on

laxation

136

gray-muscle

blood-white

#{163}

C0 C-)

.

spatial

peak

specific

ab-

mate guidelines of 3.2 W/kg head and is at the limit of 8

#{149} Radiology

Figure radiation;

7.

images

(200/15;

field

of view,

matrix

size,

256

four;

and

Effect of magnetization (b) irradiation power,

transfer 6 W during

9 cm; section

on the cat TR. Image

head at 1 .5 T: (a) control image, parameters were the same for

thickness,

3 mm;

number

of signal

no irboth

averages,

X 256).

W/kg for the body and extremities. However, as demonstrated in Figures 4 and 6, a considerable increase in contrast is generated with just 1 .6 W of power (corresponding to 3.2 WI kg). In addition, power requirements may be substantially decreased at the lower field strengths used for imaging humans (eg, 1.5 T). If one assumes that Bi intensity determines the extent of MTC, then for the same Bi field strength, the power deposition will decrease with the square of the resonant frequency of water protons (1 1). In other words, when compared with studies at 4.7 T, power deposition at 1.5 T should be roughly ninefold less for the same Bi assuming a constant geometry and size. This is obviously not the case when comparing a cat brain to a human brain, and great came must be taken to account for local depositions of pow-

em. Our cat head studies at 1 .5 T used 6 W of irradiation power for maximal MTC effect. This value is much highen than one would predict on the basis of the above argument, but the difference in coil design between the 1.5- and 4.7-T studies (saddle coil vs bird-cage) also complicates any direct comparisons of power requirements. In addition, it is still unclear how the magnetization transfer effect itself quantitatively changes with field strength. In our preliminary studies in humans (12), we have found significant contrast generated in the knee at less than 8 W/kg. These results suggest that power deposition in human studies

will

not

be a serious

limita-

tion of the technique. In summary, magnetization transfer when used with a conventional gradient-recalled-echo pulse seApril

1991

quence has been shown to increase image contrast without increasing imaging time. The contrast produced is unlike that generated on Ti- on T2weighted images and may also prove useful

in providing

a noninvasive

means of differentiating among relaxation mechanisms in vivo. Finally, this technique for increasing tissue contrast has the potential to be applied to imaging human beings at clinically relevant magnetic fields U

References 1.

2.

3.

and T2-weighted Magn

4.

5.

6.

7.

Carr DH, Brown J, Bydder GM, et al. Gadolinium-DTPA as a contrast agent in MRI: initial clinical experience in 20 patients. AJR 1984; 143:215-224. Nishimura DC, Macovski A, Pauly JM, et al MR angiography by selective inver-

sion recovery. Magn Reson Med 1987; 4:193-202. Mosely ME, Cohen Y, Mintorovitch J, et al. Early detection of regional cerebral ischemia in cats: comparison of diffusion-

8.

MRI and spectroscopy.

Med 1990; 14:330-346. Hilal 5K, Maudsley AA, Simon HE, et al. In vivo NMR imaging of tissue sodium in the intact cat before and after acute cerebral stroke. AJNR 1983; 4:245-249. Wolff SD, Balaban RS. Magnetization transfer contrast (MTC) and tissue water proton relaxation in vivo. Magn Reson Med 1989; 10:135-144. Eng J, Ceckler TL, Balaban RS. Quantitative ‘H magnetization transfer imaging in vivo. Magn Reson Med 1991; 17:304-314. Balaban RS, Koretsky AP, Katz LA. Loadin8 characteristics of surface coils constructed from wire and foil. J Magn Reson 1986; 68:556-560. Haase A, Frahm J, Matthaei D, et al. FLASH imaging: rapid NMR imaging using low flip-angle pulses. J Magn Reson

9.

Fralix

TA,

Ceckler TL, Wolff SD, Simon RS. Lipid bilayer and water proton magnetization transfer: effect of cholesterol. Magn Reson Med 1991; 18:214-223.

SA, Balaban

10.

Reson

Wolff RS.

SD, Fralix

TA, Simon

Magnetization

SA, Balaban

transfer a probe

spectroscopy molecuIn: Book Reso1990. Berkeley, Calif: Resonance in Mcdi-

of model systems: for the lar basis of tissue contrast in MRI. of abstracts: Society of Magnetic nance

in Medicine

Society of Magnetic 11.

12.

cine, 1990; 350. Hoult DI, Chen C-N, Sank VJ. The field dependence of NMR imaging. II. Arguments concerning an optimal field strength. Magn Reson Med 1986; 3:730-

746. Wolff

SD, Chesnick

S, Balaban

netization transfer contrast knee at 1.5 T. In: Book of

of Magnetic Berkeley,

nance in

RS.

in the

abstracts: Resonance in Medicine Calif: Society of Magnetic

Medicine,

1990;

Maghuman Society 1990.

Reso-

1186.

1986; 67:258-266.

Volume

179

#{149} Number

1

Radiology

#{149} i37