Guy

H. Sebag,

MD

#{149} Sheila

G. Moore,

MD

Effect of Trabecular Bone on the Appearance of Marrow in Gradient-Echo Imaging of the Appendicular Skeleton’

‘

This prospective study evaluated the effect of trabecular bone on the appearance of marrow in gradientecho (GRE) images of the appendicular skeleton in vivo at high magnetic field strength. Magnetic resonance (MR) imaging of 10 normal extremities in five patients was performed with spin-echo (SE) and GRE sequences. The latter were obtamed with gradient recalled acquisition in a steady state. SE and GRE sequences had identical spacing and planes of imaging. Cortical bone appeared as a signal void regardless of the pulse sequences and parameters. Marrow in contact with trabecular bone exhibited a shortened effective transverse relaxation time (T2*) and resultant signal loss because of local field inhomogeneities where mineralized matrix interfaced with it. This T2* effect was increased in regions with more trabecular bone (epiphysis) than regions with little trabecular bone (diaphysis). A low signal intensity on GRE images may represent fatty marrow with a high content of trabecular bone and should not be interpreted only as hematopoietic marrow. Index

terms: Bones, MR studies, 40.1214 #{149} Extremities, MR studies, 40.1214 #{149} Magnetic resonance (MR), tissue characterization #{149} Magnetic (MR), pulse

resonance

Radiology

1

ogy, 052,

From the Stanford Stanford,

Department of Diagnostic Radio!University Medical Center, S. CA 94305-5105 (G.H.S., 5GM.),

ceived

August

Enfants

of Pediatric Malades,

7, 1989;

ber 1 1; revision received November 15. Supported from the Georges Lurcy York. 0

Address reprint RSNA, 1990

revision

Radiology, Paris

(G.H.S.).

requested

Re-

Octo-

October 31; accepted in part by a grant Foundation, New requests to 5GM.

(GRE) sequences used for mus-

RADIENT-ECHO

are

increasingly

culoske!etal

low fast imaging

imaging

ation

magnetic with the

because

they

al-

(MR) of threedimensional acquisition. GRE sequences have several contrast characteristics that are distinct from those of spin-echo (SE) sequences (1). The signal intensity of tissue is dependent on effective transverse relaxtime

(T2*)

resonance advantage

as opposed

to the

T2

dependence of SE sequences. To our knowledge, the consequence of this T2* dependence on bone marrow contrast has not been previously studied in vivo in the appendicular skeleton. A 1986 study (2) has shown that fat and water in contact with trabecular bone in vitro exhibits shortened T2* relaxation times. To determine if this T2* effect extends to the appendicular skeleton in vivo at high field strength, a prospective study using GRE was undertaken. PATIENTS MR imaging

AND

METHODS

of 10 normal

extremities

in five patients (aged 5-40 years; three females and two males) was performed on a 1.5-T superconducting MR system with both SE and GRE pulse sequences. SE sequences were used with a repetition time (TR) of 600-800 msec and an echo time (TE) of 20 msec for Ti-weighted images and a TR of 2,000 msec and TE of 80 msec for T2-weighted images. GRE images

were

1990; 174:855-859

and the Department H#{244}pital des

sequences

G

obtained

with

gradient

recalled

ac-

quisition in a steady state (GRASS) (3). A TR of 60 msec, flip angle of 30#{176}, and TE of 10, 12, 14, 16, 20, 24, 30, and 40 msec were used. An additional set of images was obtamed with a TR of 250 msec, flip angle of 300, and TE of 10, 12, 14, 16, 20, 24, 30, and 40 msec. Spacing and plane of imaging were the same for GRE and SE sequences. Five-millimeter-thick corona! and transverse sections of the femur, tibia, or both, including proximal and distal epiphyses, were obtained with a 24-48cm field of view, six to eight repetitions, and a 256 X 256 matrix. At the intermediate flip angle (300) and short TR (60 msec) used for this study, the

signal intensity of the tissue is dependent on the ratio of T2* and Ti relaxation times. Tissues with a long T2*/T1 give a high signal intensity. T2* dependence on both short TR (60 msec) and long TR (250

msec) GRASS sequences is increased by lengthening the TE (3). Furthermore, because of its different spectral components, fat signal exhibits two chemical shift-induced modulations on GRASS images at 1.5 T. The

first

modulation has a period 4 msec and the secof approximately 20 msec

(p) of approximately ond

a period

(4). For each

modulation,

the spectra!

components are in phase when the TE is equal to integer multiples of the period - p. They are out of phase when the TE is equal to [(2n + 1)12] . p (ie, for both modulations, the spectral components are in phase at approximately 20 and 40 msec

n

and out of phase at approximately 10 and 30 msec). We therefore used variable echo delays to take sideration.

this

phenomenon

into

con-

RESULTS Cortical bone appeared as a signal void regardless of the pulse sequences and parameters. On SE images, fatty (yellow) marrow was seen as increased signal intensity on Tiweighted images (Fig 1) and decreased signal intensity on T2weighted images commensurate with the decrease in signal intensity of subcutaneous fat. Signal intensity of hematopoietic (red) marrow was intermediate (similar to that of muscle) on SE sequences. On the GRASS images the signal intensity of marrow was the same regardless

msec

of which

TR

or 250 msec).

was

used

(60

intensity of diaphyseal fatty marrow was similar to that of subcutaneous fat regardless of TE. Signal intensity of the epiphyseal fatty marrow, however, was moderately less than that of subcuta-

Abbreviations:

Signal

GRASS

in a steady state, echo, SE = spin echo, TE repetition time.

gradient

quisition

GRE =

=

recalled

ac-

gradient

echo time, TR

=

855

Figure

1.

Transverse Ti-weighted (b) the metaphysis, and

epiphysis,

neous

fat

on

the

(TR

800 msec,

=

(c) the

lO-msec-TE

diaphysis.

TE

20 msec)

Fatty

marrow

SE images

of the distal

is isointense

to subcutaneous

of transverse

GRASS

left femur fat

in a 40-year-old in the

three

man

segments

show of the

(a) the femur.

image

(Fig 2a). On the images obtained with increasing TE and constant TR and flip angle, the signal intensity of the epiphyseal fatty marrow decreased as compared with the signal intensity

of subcutaneous

fat

(Fig

2c). Epiphyseal signal remained creased regardless of the phase subcutaneous thus displaying phase/out-of-phase

fat

relative

2b,

deof

to muscle,

no significant effect

in(Fig

3).

The effect on metaphysea! marrow was intermediate to the effect on diaphyseal or epiphyseal marrow. The signal

intensity

marrow

of the

was

metaphyseal

moderately

less

than

that of subcutaneous msec-TE image.

fat However,

on

the iOthe signal

intensity did not

metaphyseal to the same

marrow degree

of the decrease

as that

of the

epiphysea!

marrow

with increasing TE. The presence of hematopoietic versus fatty marrow in the metaphysis did not affect the decrease in metaphyseal marrow signal

2a.

Figure illustrates

2.

Matrix the

interaction

between

umn distal metaphysis, right and 14 msec (bottom row) (Fig

images

marrow

column

site

of the same (left

patient

column

middiaphysis)

=

and

as in Figure

1. This series middle colTE. (a) TE of 12 msec (top row) distal

epiphysis,

2 continues).

intensity.

DISCUSSION GRE based

gle than

is a fast on

the

excitation

imaging technique use of a partial flip

pulse

90#{176}) followed

(typically by

an-

less

a gradient

re-

versa! to refocus the echo and generate the signal (1). Factors that influence GRE imaging contrast are the acquisition technique and tissue characteristics. It is reported that images acquired with an appropriate

TR, TE, and

flip

contrast similar (i,3). However,

appearance

angle

can

GRE

provide

a

to that of SE imaging we describe marrow

on GRE

images

that

dif-

fers from marrow appearance on SE images in that fatty marrow signal is not always isointense to subcutane-

856

Radiology

#{149}

ous fat but varies according to the anatomic location of the marrow and the TE selected on the GRE sequence. The amount of trabecular bone in addition to marrow composition (red versus yellow) is reflected on GRE images. On SE images, hydrogen nuclei associated with aliphatic carbon chains (fatty marrow) and water protons (hematopoietic marrow) are the major contributors to the signal pattern (5), as opposed to trabecular bone, which has a minor or absent contribution. Regardless of the SE sequence, bone marrow signal is homogeneous taneous skeleton

and isointense to the subcufat in those regions of the containing exclusively yel-

low marrow (ie, epiphyses and adult distal appendicular skeleton). Several contrast characteristics of GRE images are distinct from those of SE images and have a significant influence on signal intensity. Most notably, signal intensity on GRE images is dependent on the T2*, as opposed to the T2 dependence on SE sequences. On GRE images, there is no rephasing of local magnetic field inhomogeneities by a i800 pulse, and thus T2* dephasing is not recovered. Therefore GRE imaging is very sensitive to field inhomogeneities arising either extrinsically (nonuniformity in the applied main static field) or intrinsically

(heterogeneity

of the

March

body

1990

quence. The hydrogen nuclei in the fatty acid chains exhibited a signal decrease in regions containing more trabecular metaphysis)

bone (the epiphysis than regions with

and little

or no trabecular bone (the diaphysis and subcutaneous fat) (Fig 4). The progressive increase in signal loss with increasing TE indicated that fatty marrow

bone

in contact

exhibits

T2*

and

with

trabecular

a markedly

shortened

dephasing

secondary

a spin

to local field inhomogeneities. The effect on marrow signal seen in our study is not due to dephasing by proton diffusion across local field gradients during the interecho time of a

iso0

SE pulse.

This

phenomenon,

af-

fecting only water, cannot be recovered by a i80#{176} pulse and results in T2 shortening. Therefore the subsequent

signal

seen

decrease

should

be also

on SE sequences.

Adipose

tissue

spectral finic

contains

components, protons

several

including

ole-

of unsaturated

fatty

acid and central long-chain fatty of approximately 64 MHz resonance

methy!ene of the acid. Chemical shift 4.i ppm (260 Hz at frequency) be-

tween

and

ene

the

vinyl

protons

quency

central

results

methyl-

in a high-fre-

modulation

with

a period

of

approximately 4 msec. In addition, there is a second, low-frequency modulation with a period of approximately 20 msec related to the chemical shift difference between the central CH2 and the -CH2-Oprotons

of the

Hence,

fatty

acid

fat signal

shift-induced

chains

other,

a). Epiphyseal signal seal signal intensity dle

column,

loss is increased is intermediate

with increasing TE (right column, b and c). Metaphybetween the diaphyseal and epiphyseal intensities (mid-

a-c).

itself) (6). On SE images TR shortening increases the degree of Ti weighting, but with TR shortening (