Stuart

A. Bobman,

MD

Postoperative Chemical

a

Scott

Shift

W. Atlas,

Index

terms:

Magnetic Magnetic

shift

comparative (MR),

contrast

tebral

disks,

#{149}

studies

resonance (MR), resonance (MR), #{149} Magnetic resonance

enhancement 336.783

#{149} Spine,

336.1214

Radiology

1991;

179:557-562

#{149} Spine.

MR

studies,

interver-

a

John

Listerud,

MD,

Lumbar Spine: MR Imaging’

A modified fat-suppression pulse sequence (consisting of combined frequency-selective fat presaturation followed by a spin-echo acquisition when fat and water magnetization vectors have opposite phase) was used to optimize the conspicuity of intravenous enhancement by gadopentetate dimeglumine on magnetic resonance images in 10 patients previously operated on for lumbar discogenic disease as well as in two patients with herniated disks who had not previously undergone surgery. This technique produced the greatest degree of fat suppression in the phantom study. In six of the patients who had previously undergone surgery, epidural enhancement was more obvious on the fatsuppressed images than on conventional spin-echo images, while in four patients, enhancement was equivalent. The herniated disks in two patients not previously operated on were not enhanced with cither technique. Contrast enhancement was universally distinguishable from fat signal and from nonenhancing water-containing tissue on the fat-suppressed images obtained after contrast material administration. This technique may reduce the need for precontrast imaging. Furthermore, postoperative enhancement of nerve roots was more obvious on fat-suppressed images in seven of eight patients. This finding might represent previously undiagnosed degrees of arachnoidal inflammation, which may be a factor in the failed back syndrome.

chemical

MD

T

HE potential

value

PhD

fat

on

of spin-echo

saturation

spin-echo

images

with short echo times. This necessitates MR imaging before administration of contrast material, which consequently doubles the time of the MR imaging examination and requires comparison of precontrast and postcontrast images, rather than simply allowing the definition of areas of enhancement and nonenhancement on one image. One proposed solution to this problem is chemical shift imaging to suppress the fat signal (91 i). This study was designed to optimize fat-suppressed MR imaging of the lumbar spine and to determine whether such imaging can be used to increase the conspicuity of contrast enhancement in patients with postoperative changes. MATERIALS

AND

I. Grossman,

METHODS

pulse

a spin

echo.

each

excitation

utilizes

and opposed, water protons

From the Department of Radiology, Hospital of the University of Pennsylvania, 3400 Spruce St. Philadelphia, PA 19104. From the 1989 RSNA scientific assembly. Received August 16, 1990; revision requested September 24; final revision received December 4; accepted December 10. Address reprint requests to S.W.A. C RSNA, 1991

of

of a 90#{176} radio-fre-

respectively, an image of can be acquired without

postprocessing of each chopper

by addition of the signals pair together to cancel

the fat proton component. In previous vestigations involving such a sequence, the 180#{176} pulse of one of the excitations shifted spins

temporally, are

out

so that

of phase

fat and

inis

water

(9-il).

In our study, we attempted to optimize the fat suppression in the sequence in a phantom study before studying patients. Our initial pulse was a frequency-selective 90#{176} 1:3:3:1 pulse of 8-msec duration (12).

The

bandwidth

of maximal

excita-

tion was 50 Hz centered 220 Hz (3.5 ppm) from the resonant frequency of water. This presaturation pulse was followed by

a “spoiler” gradient to dephase the excited fat spins. The unsaturated water resonances were then excited in the conventional manner. The echo was collected after a 180#{176} pulse with the fat and water spins in phase, opposed, or alternately in phase and opposed as selected by the operator. Each chopper signal pair was then

added. Thus, the pulse sequence produced chemical shift images with fat and water magnetization vectors in phase or or with

opposed,

the

cancelling

cy-selective

shift have fat

by collection

of the echoes

quency pulse that is phase-alternated with the previous 90#{176} pulse) with the fat and water magnetization vectors in phase

tations-either

Study

Previous techniques of chemical imaging used to suppress fat signal involved use of a frequency-selective

followed

By collection

from a sequence involving two excitations alternately as halves of “chopper excitation pairs” (“chopper excitation” refers to a two-excitation scheme in which

vector Phantom

MD

Contrast-enhanced

magnetic resonance (MR) imaging of the lumbar spine in the management of failed-back-surgery syndrome has been suggested by several investigators (1-4). It has been shown that enhancement with paramagnetic contrast material allows, in most cases, differentiation between recurrent avascular disk herniation and vascular epidural fibrosis (5-8). Poor conspicuity of contrast enhancement results from the background of hyperintense

#{149} Robert

on the A fat ployed quence.

with fat

fat magnetization

itself

on alternate

or without

exci-

frequen-

presaturation-depending

operator’s specification. and water phantom was emto optimize the fat-suppression SeIt consisted of two bottles, one

containing 100% pure soybean oil and the other a dilute aqueous copper sulfate solution. The phantom was imaged in a

quadrature copy

head

was

coil. First,

performed

with

MR spectrosa pulse

se-

quence of 1,000/25 (repetition time msec/ echo time msec) to validate the fat and water phantom (13). Then, at constant transmit section

and receive axial images

attenuations, (1,000/25)

were

multiob-

557

a.

a.

b.

c.

b.

d.

e.

f.

Figure 1. Six spin-echo images (1,000/23) of the phantom, with the water bottle on the left and the fat bottle on the right in each pain. The images in the top row were acquired without fat presaturation (a-c); those in the bottom row were acquired with the fat pnesaturation pulse and its spoiler (d-f). The images in the left column (a, d) were acquired with the fat and water magnetization vectors in phase; those in the middle column (b, e), with the fat and water magnetization vectors out of phase (opposed); those in the right column (c, f), with the fat magnetization vector cancelling itself on alternate excitations (chopper Dixon). The upper left image (a) is, therefore, a conventional spin-echo image without any suppression of

fat. The

number

displayed

immediately

below

the fat bottle

in each

image

is the absolute

C.

signal intensity within the fat phantom obtained with its corresponding combination of fat suppression options. The percentage displayed below this number represents the signal intensity relative to the conventional spin-echo standard. The greatest degree of fat signal suppression is seen in e, where there is 96% suppression (notated as 4% relative residual signal intensity), compared with the conventional spin-echo technique.

tamed

with

of the

operator-selectable

each

of the six combinations

on comparison

ages)

fat-suppression

parameters as described above to determine the options affording the optimal reduction in

of fat signal

the

clinical

Patient

portion

for subsequent of the

use

study.

Study

A total all with discogenic ulopathy).

of 12 patients

were

examined,

symptoms suggestive of lumbar disease (eg, back pain or radicTen of the 12 had previously

undergone ages ranged

lumbar laminectomies. from 24 to 79 years

43.5

seven

years);

The laminectomies weeks to 12 years

of the

10 were

Their (mean, men.

had been performed 4 before their MR exami-

nation (mean, 2.2 years). Three of the 10 had undergone two separate surgical procedures. Three of the 10 had undergone operations at multiple levels. Two patients (one man aged 36 years and one

woman

aged

42 years)

with

disk hernia-

tion who had not undergone surgery were also studied. The conventional fat-suppressed images were compared side by side in a nonblinded fashion

and by

two neuroradiologists (S.A.B., S.W.A.) who subjectively judged the conspicuity of enhancement. graded as either

558

a

Radiology

Enhancement was first present or absent (based

and

with

then

nonenhanced

quaiitatively

im-

compared,

by

means of visual inspection, between fatsuppressed and conventional images obtamed after administration of contrast material as either equivalent, less evident, or more prominent. In each patient, the specific regions of enhancement were noted (eg, epidural tissue, intrathecal nerve roots, nerve root ganglia, disk ma-

tenial). In the clinical

part of the study, the lumbar spine was imaged with a 5 X 11inch (12.5 X 27.5-cm) posteriorly placed

surface waukee)

coil (GE Medical Systems, Miloperating in receive mode. Sag-

ittal spin-echo for localizing

had been nonangled tamed

images (600/20) the levels of the

were disks

used that

operated on. Five-millimeter axial images were then ob-

(1,000/20-23,

two

to four

signals

averaged, 256 X 128 matrix, 1-mm intersection gap, 20-cm field of view) through the

area

ventional

of prior

spin

surgery

echo,

with

as well

posed spin echo with discussed previously.

both

con-

as the op-

fat presaturation Immediately

after

intravenous administration of 0. 1 mmol/ kg gadopentetate dimeglumine (Magnevist; Berlex Imaging, Wayne, NJ), both spin-echo and fat-suppressed spin-echo axial

same

Figure 2. (a) Axial conventional spin-echo image (600/25) of water and fat phantom consisting of bottle of dilute aqueous CuSO4 (left) and pure soybean oil (night). (b) Line volume through phantom in a stimulated for spectroscopy. (c) Proton MR spectrum (1,000/25) of the entire stimuated line volume (selected by the intersection of the section-selective 90#{176} radio-frequency pulse and the orthogonal section selected by the 180#{176} pulse) is displayed as bands (intensity band map), with the horizontal position representing the voxel position within the line volume, the vertical position representing the chemical shift, and the brightness representing the signal intensity. Also, the spectrum for a single voxel within the fat phantom is superimposed (lower right), with the location of the voxel marked by the line Iabeled profile. The soybean oil spectrum possesses not only a large methylene peak, but also a much smaller olefinic peak. The peaks representing 60-Hz artifact are also labeled.

imaging levels.

was

repeated

through

the

RESULTS Phantom

Study

The absolute and normalized signals from the six combinations of the fat-suppression pulse sequence options were imaged (Fig 1). The greatest degree of fat suppression (96% reduction in fat signal relative to the conventional spin echo) was achieved with the opposed spin echo in combi-

May

1991

Figure 35)

3.

(a) Axial

through

the

MR spectrum

spin-echo

left

leg.

(1,500/35)

b.

b. 4.

On

non-fat-suppressed,

the

silatenal

Recurrent epidural

tionable

intense

of a line

minimal

sue

on

the

The

nature

enhancement

in left

foramen).

al to a larger

herniation

with

of

this

epidunal

nonenhancing of

the

tissue

central enhanced

and

179

a

Number

2

foramen

fat-suppressed

area

(solid

fat-suppressed

(a),

clear.

laterally

space

nation with fat presaturation (Fig le). This technique was subsequently empboyed for the clinical trial. Proton MR spectra were imaged through a single line of the fat and water phantoms (Fig 2) (13). The MR spectra demonstrate that the soybean oil phantom (Fig 2c) does model the signal behavior of biologic fat, such as that in the tibial marrow and subcutaneous region (Fig 2b). As expected, the stimulated voxels in the region corresponding to the CuSO4 bottle possess a single strong peak corresponding to the water resonance. The soybean oil bottle has a strong methylene peak, as well as a much smaller olefinic resonance downfield, shifted nearly as far as the water peak (Fig 2c). This spec-

Volume

scar in axial image

is not

is suggested

lateral

Contrast-enhanced

basis

enhancing spin-echo

(300/ Proton

volume

(OL).

d.

C.

conventional

fat.

tissue

opposite

disk

(b)

through the tibia in a is displayed in the same manner as in Figure 2c. Both the intensity band map (upper left) and conventional spectrum (lower right) obtained through the tibia as marked by the vertical profile line show a strong methylene peak (ME) and a weak olefinic peak This spectrum and that of the fat phantom in Figure 2c are similar.

(PR)

Figure

image

lower

arrow). image,

On

along

images

there the

the

(arrow)

(1,000/23)

is left intravenously

left

facet

image

(d)

This

was

although

clearly

shows

believed the

small

to represent focal

trum is similar to that leg obtained through which also demonstrates

tissue

of marked

areas

of enhancement

a recurrent

enhancement

of the lower the tibia (Fig the strong

could

spin-echo

Study

Fat-suppressed images obtained in six of the 10 postoperative patients demonstrated more conspicuous gadolinium enhancement than the conventionab spin-echo images (Figs 4, 5). The conspicuity of enhancement in four postoperative patients was equivalent on both the fat-suppressed and spin-echo images. The

image

(c) shows

with

epidural

roots

image

arrows)

herniation

indicate

before

within

imaging.

obliterates

the

(b),

hypen-

is present

posterior

surrounding

ip-

ques-

slightly

of fat (as

(open

nerve 3),

which

hypointensity

disk also

on 3 years

(arrow),

fat-suppressed

region

water and methylene peaks, with the weaken olefinic peak resonating at approximately the same frequency as water.

Patient

was operated

soft

non-fat-suppressed

Nonenhanced

expected

which

epidunal

enhanced, joint.

replacing

at L5-S1,

ventrolatenal

in

and

medi-

scar

tis-

veins.

the

thecal

sac

of

eight of these 10 patients demonstrated definite enhancement, which was more evident on the fat-suppressed images in seven of these cases (Fig 6). Six of the eight patients with root enhancement had previously undergone myebography; the other two underwent surgery, and one was complicated by dural laceration. The nerve roots were enhanced within the thecal sac, within the nerve root sheaths, and distally (including the dorsal root ganglia). The two patients in the series who were not operated on had acutely herniated disks that were not enhanced with either fat-suppressed or conventional sequences (Fig 7). One of these two studies demonstrated

Radiology

a

559

a.

b.

d.

C.

Figure

5. Distinguishing between enhancing scan and epidural fat in axial images (1,000/23) through the inferior aspect of L-4 obtained 5 months after laminectomy. Following contrast material administration, there is some enhancement (open arrow) ventrolateral to the left of the thecal sac on the enhanced conventional spin-echo image (b) as compared with the nonenhanced spin-echo image (a). There is high intensity in the right ventrolateral epidural fat (solid arrow, a and b). On fat-suppressed images obtained before (c) and after (d) administnation

of

guished clearly

contrast

material,

from the differentiated

the

right

ventrolateral

slightly hypenintense from enhancement

epidural

left ventral scar on a fat-suppressed

fat

(open

(solid

arrow,

c and

arrow, c), which postcontrast image

d)

is markedly

hypointense

is enhanced alone (d).

dramatically

and

(open

therefore

arrow,

easily

d).

The

distin-

ventral

fat

is

b.

a. Figure

6.

Enhancing

intrathecal

nerve

roots

in postoperative,

postmyelography

patient.

(a)

Four

axial

the level of the superior end plate of L-5 (preand postcontnast [GD] and with and without fat suppression, previously undergone two laminectomies and two myelography examinations. Conspicuity of nerve root fat-suppression technique (arrows, lower night image) versus the spin-echo technique (lower left image). pressed axial images (i,000/23), obtained in the same patient, extending from the level of the L-4 pedicle right, top to bottom) show diffuse and extensive nerve root enhancement.

penidiscal enhancement that was more evident with the fat-suppressed technique (Fig 7b). The images in both patients (neither of whom had previously undergone myebography) did not show nerve root enhancement with either the fat-suppressed or non-fat-suppressed methods.

DISCUSSION The was

560

a

purpose to evaluate

Radiology

of this investigation the utility of opti-

mized-fat-suppression MR imaging (i4,i5) in the lumbar spine of postoperative patients. This technique has the potential of increasing the conspicuity of abnormal contrast enhancement. A phantom study was initially employed as a model for optimization of the fat-suppression pulse sequence prior to imaging in patients. The phantom study made evident that fat suppression is supenior when the fat presaturation pulse and spoiler are applied, regardless of

images

(1,000/23)

obtained

through

as labeled) in a patient who had enhancement is increased with the (b) Contrast-enhanced fat-supto the top of the sacrum (left to

whether the spin echo is collected in phase, out of phase, or with chopper cancellation of fat signal (Fig 3). Intuitively, one might expect that the chopper-Dixon technique with the fat magnetization vectors subracted out on alternate chopper signals should yield a greaten degree of fat suppression than the opposed spinecho technique. This is clearly the case when the fat presaturation pulse is not

phantom

applied

(Fig

experiment

1).

However,

showed

the

that

May

1991

b.

C.

Figure 7. Penidiscal ated disk not operated echo image (1,000/20) Si disk in this patient tomatic

for

6 weeks

enhancement in hennion. (a) Sagittal spinshows a herniated L5who had been sympand

had

not

undergone

23)

(b) Axial spin-echo image (1,000/ was obtained at a position 6 mm below

the

center

surgery.

of the

herniated

fragment. (1,000/23) was obtained at the same level as b. (d) Axial contrast-enhanced spin-echo image (1,000/ 23) was obtained at the same level as b.

(c) Axial

fat-suppressed

(e) Axial

contrast-enhanced

disk

image

fat-suppressed

image (1,000/23) was obtained at the same level as b. Herniated disk material (solid anrow, b-e) is seen on all images (b-d). Note penidiscal enhancement (open arrows, d and e) compared with the corresponding nonen-

e.

when the fat is presatunated, the opposed spin-echo technique (Fig le) provides superior fat suppression than does the true chopper-Dixon method (Fig if). This behavior is also observed in biologic fat. A likely explanation would be that the fat presaturation pulse, which is frequency-selective only for the methylene component of fat, markedly diminishes the methylene signab vector. The small remaining methylene magnetization vector is opposed by the small unsaturated olefinic magnetization vector, which resonates at approximately the same frequency as water, thus more complete intravoxel tion

allowing cancella-

of fat signal. The clinical portion of the study confirmed that the opposed spinecho technique with fat saturation increased the conspicuity of contrast enhancement against a background of epidunal fat when compared with conventional spin-echo images. It is generally recognized that the most important use of MR imaging in patients with recurrent or persistent symptoms after back surgery is to di-

Volume

179

a

Number

2

agnose recurrent disk herniation and to distinguish it from scar tissue. Therefore, the radiologist desires the technique that is both most sensitive and most specific for abnormal enhancement to make this assessment. Frequently, enhancement can be detected only on spin-echo images when compared with unenhanced images. In every case in this small series, enhancement was sufficiently conspicuous and distinct from both fat and nonenhancing water-containing tissues that unenhanced images would not have been needed to document the enhancement. If confident identification of enhancement and separation of it from nonenhancing disk material are consistently possible with this technique, the need for precontrast imaging would be diminished, thereby increasing clinical efficiency. Furthermore, subtle areas of minimal enhancement on spin-echo images were obvious on the fat-suppressed images. These features support the use of this technique in postoperative lumbar spines. It should be noted that the overall signal is some-

hanced

image

obvious on hancement granulation veins (9), or

(b and

c). which

is much

fat-suppressed image presumably represents tissue (6) or distended both.

more

(e). The either epidural

en-

what diminished on fat-suppressed images as compared with conventional spin-echo images. The fat-suppression studies showed nerve root enhancement on images obtained in a high percentage of postoperative and postmyebographic patients in this series. The nerve roots were enhanced both within the thecal sac, within the nerve root sheaths, and peripheral to and including the dorsal root ganglia. While enhancement distal to the portion of the roots possessing bloodbrain barrier is normal (5), that within the thecab sac presumably is not and may represent arachnoiditis in these patients. All patients in this study were symptomatic, in that all had either recurrent bow back pain or radicubopathy. It is not certain to what degree arachnoiditis may have contributed to their clinical symptoms. It is not clear why root enhancement should

Radiology

a

561

be more conspicuous on fat-suppressed images against a background of non-fat-containing cerebrospinal fluid than on spin-echo images. This might be related to better use of the dynamic range of the anabog-to-digitab converter with respect to the signals of the tissues of interest when only water protons are imaged. We conclude that contrast enhancement is more evident on fatsuppressed MR images than on conventional spin-echo images in the evaluation of the postoperative bumbar spine. This is potentially of great significance, since the treatment of recurrent back pain in these patients often hinges on the distinction of recurrent disk herniation from scar formation. We also note a surprisingly high incidence of abnormal nerve root enhancement in this same set of patients, clearly seen on the fat-suppressed images. This finding may help explain the high incidence of failed back syndrome, since it may indicate that more of these patients have arachnoiditis. Furthermore, with this technique, the need for obtaming precontrast images for com-

562

a

Radiology

panison may be diminished because it is both extremely sensitive and specific for abnormal enhancement. U

8.

References 1.

2.

3.

Bundschuh CV, Modic Ml, Ross JS, Masaryk TJ, Bohlman H. Epidural fibrosis and recurrent disk herniation in the lumbar spine: MR imaging assessment. AJR 1988; 150:923-932. Hochhauser L, Kieffer SA, Cacayorin ED, Petro GR, Teller WF. Recurrent postdiskectomy low back pain: MR-surgical correlation. AJR 1988; 151:755-760. Masanyk TJ, Ross JS. Modic Ml, Boumphrey F, Bohlman H, Wilber G. High-resolution

4.

5.

6.

7.

MR

imaging

of sequestered

9.

10:1083-1088. Simon JH, nial

with

for

Radiology

Szumowski PW, Plewes

and

chopper 12.

imaging.

Hone

water

MR

averaging

Radiology

1987; PJ.

transform

improved

1989;

1989; 9:379-388. Szumowski J, Plewes lipid

shift

contrast

mate-

lesion

de-

171:539-543.

J, Eisen JK, Vinitski S. Haake DB. Hybrid methods of

chemical-shift

11.

Chemical

paramagnetic

enhancement

piction.

10.

J.

Szumowski

imaging

lum-

bar intervertebral disks. AJR 1988; 150:1155-1162. Sotiropoulos 5, Chafetz NI. Lang P. et al. Differentiation between postoperative scar and recurrent disk herniation: prospective comparison of MR. CT, and contrast-enhanced CT. AJNR 1989; 10:639643. Breger RK, Williams AL, Daniels DL, et al. Contrast enhancement in spinal MR imaging. AJR 1989; 153:387-391. Hueftle MG, Modic Ml, Ross JS, et al. Lumbar spine: postoperative MR imaging with Gd-DTPA. Radiology 1988; 167:817824. Ross JS, Delamarter R, Hueftle MG, et al. Gadolinium-DTPA-enhanced MR imaging

of the postoperative lumbar spine: time course and mechanism of enhancement. AJR 1989; 152:825-834. Ross JS, Blaser S. Masaryk IJ, et al. GdDIPA enhancement of posterior epidural scar: an experimental model. AJNR 1989;

Magn

DB.

Reson

Separation

imaging

in the

Med

of

signals

time

by

domain.

165:247-250.

Solvent

suppression

in

nuclear

magnetic

resonance.

Fourier

Magn

13.

14. 15.

Reson 1983; 55:283-300. J, Lenkinski RE, Axel L, Roberts M. Hydrogen ultrathin phase-encoded spectroscopy (HUPSPEC). Magn Reson Med 1990; 14:507-521. Brateman L. Chemical shift imaging: a review. AJR 1986; 146:971-980. Listerud

Dixon imaging.

WI.

Simple Radiology

proton 1984;

spectroscopic 153:189-194.

May

1991