Eric

C. Wong,

#{149} Andrzej

BA

Jesmanowicz,

High-Resolution, MR Imaging with a Local The many and hand cause

small

size

F

of the wrist and, be-

of the

tissues,

their magnetic resonance (MR) imaging necessitates use of the high spatial resolution obtainable with fields of view as small as 2 cm x 2 cm x 1 mm. The authors demonstrate that the use of a local xyz gradient coil, positioned off-center in a clinical MR imager to facilitate patient positioning, permits acquisition of high-resolution images in spin-echo (SE) and

gradient-recalled-echo quences

with

as 6 msec

(GRE)

echo

(SE)

time

(TE)

or 3 msec

authors

compare

this

taming

high-resolution

seas short

(GRE).

method

The for ob-

images

with

the alternative method of using normal gradient strengths and increased pulse duration. The effects on image quality of TE, bandwidth, gradient strength, and chemical shift artifacts are presented. Images obtained with the local gradient coil of the carpal

tunnel,

carpal

bones,

interphalangeal

unteers Index

terms:

Hand,

resonance

resonance

studies,

1991;

From

the

Radiology, consin,

vol-

MR studies,

43.1214

(MR), technology (MR), surface coils

#{149} Mag#{149} Wrist,

MR

43.1214

Radiology

I

proximal

in healthy

are shown.

Magnetic

netic

and

joint

8701

c RSNA,

181:393-397

Biophysics

Section,

Department

ESR Center,

Medical

College

Watertown

Plank

reprint 1991

requests

S. Hyde,

PhD

years we have been exploring the limits of high-resolution imaging in a whole-body magnetic resonance (MR) imager. This work has primarily comprised development of pulse sequences for small field of view (FOV) (i) and achievement of advances in radio-frequency (RF) coil technology to improve the signal-to-noise ratio (SNR) (2). In this article, we report two improvements we have made to our high-resolution imaging techniques: (a) design and construction of a three-axis local gradient coil to obtain stronger and more rapidly switched gradient fields and (b) modification of pulse sequences to obtain an FOV as small as 2 cm x 2 cm X i mm with use of whole-body gradient coils. These techniques are demonstrated and discussed herein in the context of high-resolution imaging of the finger and wrist of healthy humans. Resolution in MR images can be increased in two ways: Either the gradient duration or the gradient amplitude can be increased. A basic tradeoff in SNR is presented by the choice of one of these methods. One factor in this trade-off is the effect of the bandwidth of data acquisition on SNR. For a given FOV and matrix

Rd,

of

of Wis-

Milwaukee,

to J.S.H.

OR several

is inversely proportional to the root of the data acquisition time (Tdaq) and proportional to the square root of the bandwidth. Therefore, the SNR is increased by using low bandwidth and long Tdaq. The other factor in the trade-off is the effect of echo time (TE) on signal loss from T2 decay. The MR signal is increased by using the shortest TE possible, and the minimum TE is limited by the duration of the imaging gradients. Thus, the advantage of using gradient pulses of longer duration is decreased noise as the result of a lower bandwidth, and the disadvantage is decreased signal as the result of a longer minimum TE. Conversely, the size,

WI 53226. Received March 1, 1991; revision requested April 30; revision received June 21; accepted July 1. Supported by grants CA41464 and RRO1008 from the National Institutes of Health.

Address

#{149} James

Short Echo Time ofthe Fingers and Wrist Gradient Coil’

fibrous tissues have short T2s,

of the

PhD

noise square

use of higher amplitude gradient pulses of shorter duration gives higher signal because of shorter TE but gives higher noise because of higher bandwidth. The net SNR is dependent on the minimum TE achievable with use of each of the two methods and on the T2 of the tissue being examined. Pulse sequences with longer gradient durations have the advantage that they can be implemented on a standard clinical MR imager without the need for hardware modifications. Excellent high-resolution images can be obtained by using this technique, particularly in tissues with relatively long T2. For imaging species with short T2, use of a local gradient coil allows small FOVs to be obtained with short

TE by using

larger

amplitude

and

more rapidly switched gradient pulses. We have found that many of the soft tissues in the finger and wrist are visible only with short TE Sequences. These sequences have the additional advantages of reducing chemical shift and flow artifacts.

MATERIALS Pulse

AND

METHODS

Sequences

For high-resolution imaging with whole-body gradient coils, a modified version of the pulse sequence presented by Jesmanowicz et al (1) was used. For FOVs smaller than 8 cm, the Tda5 and the duration of the readout gradient were increased from the standard 8 msec to 16 msec. The bandwidth of the digital filter was narrowed from 16 kHz to 8 kHz to maintain the same resolution (256 points in the frequency-encoding direction). For FOVs of 2-4 cm, Tda was further increased to 32 msec, and the tandwidth of the filter

Abbreviations: gradient-recalled SE = spin

Td,,,

echo,

FOV echo, SNR

data acquisition

field of view, GRE RF = radio frequency,

=

= signal-to-noise

time, TE

= echo

=

ratio, time.

393

was

to 4 kHz.

narrowed

ing

gradient

lobe

shape

of a half

mize

its duration.

of less

than

was

sine

3 mm,

and

pulse width

were increased of the RF pulse the

pulses

of 1.6-msec

ramp

time

of the

RF

gradient

crusher gradient rection after data the standard lecting full echoes,

gradient-recalledwere also modified gradient coil. RF duration,

and

200

The

was

GRE sequences used in the phase-encoding

tamable

duration

thickness.

(SE) and sequences our local

gradient

to mini-

to narrow the bandand thereby to de-

Spin-echo echo (GRE) for use with were

the

thicknesses

section-selective

section

coils

from

section

the

pulse

crease

phase-encod-

to a trapezoid For

the

The changed

p.sec.

a rewinding direction

during

transmit

cycle

For

the

ger studies, rent, bird-cage

coil. and the

SE sequences

with

TE

times this ap-

image,

All

of 8 msec

clinical Medical

sequences

using

with

TE

with

for FOVs

as small

Gradient

Coil

The using

was

local

duces

gradient

fields

and has a gradient G/cm in the x and G/cm lindrical

se-

mH

designed

along

all

cm.

an

a cyof 10.7 induc-

y gradi-

induced curare compen-

as electrical

No

or gradient waveform distortion been encountered in this configura-

tion.

The

coil

insta-

is then

con-

amplifiers.

RF Coils For was

studies of the built to fit inside

accommodate images

a wrist

of the

rangement

carpal

was

used.

wrist, the

394

#{149} Radiology

a saddle gradient

of normal tunnel,

coil coil to

size.

For

a two-coil

A saddle

used to transmit RF, and coil on the anterior surface was used in receive-only

caused

distortion

of

shield

was

obtained

on

system Milwaukee).

a i.5-T

(Signa; For

coil,

to secure

GE studies

a support the

coil

at the

approximately

remove

the clinical

the

gradient

imager

is

15 minutes.

axes

loop.

gradient

eddy

used,

coil was

a small surface of the wrist mode. The two

ar-

1.

study:

a, receive-only

mit-only d,

transmit-receive

gradient

of the

and

on

It has

support amplifiers

power

study

were

Figure

coil.

currents in this

ramps

gradient built

from

bility have

local

was

to install

of the local gradiof the feedback

to the

the local

RF coils

sated for the impedence ent coil by modification

nected

were

MR imaging Systems,

and

It is built diameter

such

images

time

three

for the x and

problems

at

RF

in an echo-planar

use of a slotted

by

The whole-body gradient coils are disconnected from the power amplifiers, and the leads to these coils are insulated so they will not The power

coils

of the

effects of eddy were observed

(3,4) and coil pro-

ents and 0.112 mH for the z gradient. The coil can support a rise time of 50 p.sec, but 200-p.sec ramps were chosen for use in this study to limit field slew rates.

that rents.

re-

Studies

structure

strength at 70 A of 8 y directions and 22

of 37.0

This RF

RF the

patient table. The gradient coil is against the wall of the patient bore when inserted into the magnet. Healthy volunteers were imaged with their hand at their side inside the gradient coil. The descent The

in the z direction. form, with inner

cm and length tance of 0.029

GRE created

side

coil was

gradient laboratory.

in our

and were

as 4 cm.

gradient

conjugate

built

of 6 msec

TE of 3 msec

of the

distortion

shield

and

coil.

factor

the

to isolate

required.

and GRE sequences with TE of 5 msec were created for FOVs as small as 2 cm. SE quences

in-

0.13-mm

between coil

of some

in the

Human

proach,

quality

study. However, in which 50-p.sec currents

a solid

placed

the gradient

No obvious this shield

gradients stronger,

four using

fin-

to accommodate

gradient

from

the expense

the

or

was the

RF fields

To obtain was shortened to either 4 or 2 msec. For a given FOV and matrix size, this required use of digital filters with bandwidths of 32 or 64 kHz and By

cm

the

constant-curwas built with

Therefore,

shield

coil(s)

fields. within

two

of 2.5

(5).

The unloaded quality factor of the RF coils decreased by as much as a factor of 3.5 when introduced unshielded into the

in the section-select diacquisition. By using of 8 msec and by colthe minimum TE at-

that were respectively.

a 12-element, coil (3,4)

side diameter fingers.

gradient

12 msec.

and

detuned

copper the

decoupled,

passively

stored

was

geometrically

coil was

gradient and a

in an SE sequence shorter TEs, T5

were

the surface

RESULTS The gradient and RF coils used in this study are shown in Figure 1. The transmit-receive finger coil was used with the whole-body gradient coils (Fig 2a) and in the local gradient coil (Fig 2b, 2c). The transmit-receive wrist coil was used in the local gradient coil (Figs 3b, 4). A similar wrist coil was used with the whole-body gradient coils (Fig 3a). The transmit-only RF coil and the receive-only RF coil were used together in the local gradient coil (Fig 3c). With the whole-body gradient coils (1 G/cm maximum), we used the pulse sequence described above to image the proximal interphalangeal joint of the third digit of a healthy volunteer at an FOV of 3 cm x 3 cm x i mm (Fig 2a). The minimum TE for this sequence was 54 msec, and the bandwidth of the filter was 4 kHz. With the local gradient coil, the minimum TE for this FOV was 12 msec, with the usual Tdaq of 8 msec and 16kHz filter. An image obtained with this TE is shown in Figure 2b. These images demonstrate that T2 decay

Gradient

RF coil. c,

and

RF coils used in this RF coil. b, trans-

surface

transmit-receive wrist

RF

finger

RF coil.

e, local

coil.

plays a dominant role in determining the SNR in these tissues. We define an overall SNR as the average signal of the smallest elliptical region of interest that completely encompasses the finger divided by the average noise. The overall SNR in Figure 2a is 4.7 and that in Figure 2b is 6.9, despite use of the higher bandwidth and half the number of signals averaged. These images were acquired in 8 mmutes (Fig 2a) and 4 minutes (Fig 2b). At this resolution, the trabecular structure of the bone and the articular cartilage are well delineated. In the image obtained with shorter TE, the volar plate and the flexor tendons are also clearly visible. For a comprehensive survey of images of the finger, see the article by Erickson et al (6). Figure 2a also demonstrates the chemical shift artifact in the misregistration of the articular cartilage with respect to the bone marrow. At the gradient strength used for this image (0.67 G/cm), the amplitude of the artifact is 0.7 mm (approximately 6 pixels), and the artifact is clearly visible. In Figure 2b, the gradient strength was mm

2.7 G/cm, giving a shift of 0.17 (1.5 pixels). Figure 2c shows an image of the same joint obtained by using a GRE sequence with a TE of 4 msec. The Tdaq was 2 msec, and the bandwidth of the digital filter was 64 kHz. The soft tissues surrounding the flexor tendons have high signal intensity with this TE, and the tendons are therefore well defined. The extensor tendons and the collateral ligaments are also well delineated. Table i shows measured T2 and T2* values from regions of interest in the finger. The T2 data were obtained from a four-echo SE image, and the T2* data were obtained from GRE images obtained at different TEs. The bandwidth for these images was 32 kHz.

November

1991

b.

a. Figure time

2. =

width,

Axial

MR

500 msec, and

four

TE

54 msec

=

averaged.

and

two signals

gradient

flexor

=

the

[500/54])

proximal

with

(b) SE image

c.

interphalangeal

joint

of the

third

digit

FOV of 3 cm x 3 cm x 1 mm was obtained (500/12)

with

FOV

of 3 cm

x 3 cm

in a healthy

with

x 1 mm

was

b.

a. Figure

=

Axial

(a) SE image

obtained

with

use

(repetition

gradient

coils,

of a local

gradient

4-kHz

band-

coil,

was obtained ac

ligaments,

16-

with

use

= articu-

C.

images of the wrist in a healthy volunteer. (a) SE image (500/28) with FOV of 6 cm x 6 cm x 2 mm was obtained with use of whole-body gradient coils and 8-kHz bandwidth. (b) GRE image (500/3) with FOV of 6 cm x 6 cm x 2 mm and a 90#{176} flip angle was obtained with use of a local gradient coil and 64-kHz bandwidth. (c) GRE image (500/5) with FOV of 4 cm x 4 cm x I mm and a flip angle of 90#{176} was obtained with use of a local gradient coil and 32-kHz bandwidth. The hyperintense streak in the carpal bones is an artifact resulting from signal from the body. Two signals were averaged in each image. ft = flexor tendons, et = extensor tendons, mn = median nerve, on = ulnar nerve,

ua

3.

volunteer.

use of whole-body

averaged. (c) GRE image (500/4) with FOV of 4 cm x 4 cm x 1 mm and a 90#{176} flip angle 64-kHz bandwidth, and one signal averaged. bm = bone marrow, vp = volar plate, ci = collateral tendons, et = extensor tendons, pda = palmar digital artery.

coil,

lar cartilage,ft

through

signals

kHz bandwidth, of a local

images

ulnar

Figure images

MR

artery.

3a and with an

3b shows axial wrist FOV of 6 cm x 6

sive survey the article

cm X 2 mm. Figure 3a is an SE image obtained by using the body gradient coils and a TE of 28 msec, the shortest

TE attainable

with

this

FOV.

Figure

3b

the

Figure carpal

the

local

of images by Middleton 3c is an tunnel

gradient

of the wrist, et al (7).

see

axial image through obtained by using

coil and

a 2.5-cm

is a GRE image obtained by using the local gradient coil, a TE of 3 msec, and a bandwidth of 64 kHz. Structures that have very short T2s and are therefore hyperintense in the image

surface coil on the anterior surface of the wrist. For this image, the FOV was 4 cm X 4 cm X i mm, the TE was 5 msec, and the bandwidth was 32 kHz. This image shows very detailed silhouettes of the flexor tendons, the

obtained

flexor

with

short

sheaths surrounding dons, the nerves and cartilage and ligaments the carpal bones. For

Volume

181

#{149} Number

TE include

the

the flexor tenskin, and the surrounding a comprehen-

2

retinaculum,

surrounded ulnar nerve tense artifact

carpal

bones

the

median

nerve

by a thin sheath, and the and artery. The hyperinthat appears within the

represents

signal

the body A small

from

exterior amount

to the

gradient

coil.

of RF is transmitted

to and received from the body through the walls of the gradient coil. Because the body experiences very little of the imaging gradients, all of the signal from the body appears to

be at or near

isocenter

of the

gradient

coil; hence, the artifact appears. Figure 4 shows coronal images

the

carpal

bones

obtained

of

by using

the local gradient coil and an FOV of 6 cm X 6 cm X 2 mm. Figure 4a is an SE image obtained with a TE of 12 msec and a bandwidth of i6 kHz, while Figure 4b is a GRE image ob-

Radiology

#{149} 395

tamed

with

bandwidth

a TE of 5 msec

and

of 32 kHz.

structure

The

the carpal bones is very well with essentially no chemical facts. An intercarpal ligament

tween

the

bones

is seen

capitate

and

clearly

a

of

defined, shift artibe-

hammate

in both

images.

DISCUSSION Herein we have described our mitial experience with high-resolution, short TE imaging of healthy humans

with

a local

compared

gradient this

lution

imaging

body

gradient

coil.

We also

technique

with coils

to high-reso-

use

of whole-

and

modified

a.

pulse sequences. Our goals in this study were to determine the advantages and disadvantages of use of a local gradient coil to drive conventional pulse sequences at very short TE and to examine the effects of SNR, bandwidth, TE, and chemical shift artifacts with use of this technique.

Our

first observation

was

that

there

are many structures with short T2 in the finger and wrist that give significant signal only at very short TE. An important trade-off in SNR exists for species with short T2 because use of a shorter TE requires use of both a shorter Td,,q and a filter with a higher bandwidth. The relevant calculation is the expected increase in SNR from a shorter TE multiplied by the expected decrease in SNR from a higher bandwidth. Table 2 shows the minimum TE and FOV available with our present software for use with differ-

ent bandwidths and In deciding between

pulse sequences. the different

TEs, in the quence SNR with width over

there is always a crossover point T2s of tissue above which a sewith longer TE gives better and below which a sequence shorter TE and a higher bandgives better SNR. This crosspoint occurs at a T2 given by = [2(TE1 TE1)]/[ln(B1)/B,)J, where B = bandwidth. For two SE sequences, one with TE of 12 msec and bandwidth of i6 kHz and another with TE of 6 msec and band-

width

of 64 kHz,

occurs

at a T2 of 8.6

Another

the

crossover

point

msec.

advantage

of using

strong

gradient fields is the reduction of the chemical shift artifact. This artifact is most apparent in the misregistration of the bone marrow with respect to the surrounding cortical bone, cartilage, and synovial fluid and can be seen easily in Figures 2a and 3a. The

amplitude

of this

proportional and

is reduced

396

#{149} Radiology

to the

artifact

is inversely

gradient

to 0.06

mm

strength for

the

b.

Figure 4. Coronal MR images of the wrist FOV of 6 cm x 6 cm x 2 mm was obtained width.

(b) GRE

local gradient triangular

image coil and fibrocartilage,

(500/5) 32-kHz

ii

with FOV bandwidth.

= intercarpal

in a healthy volunteer. (a) SE image with use of a local gradient coil and

of 6 cm x 6 cm x 2 mm was Two signals were averaged

obtained in each

(500/12) with 16-kHz bandwith use of a image. tfc =

ligament.

chemical shift between water and fat at 8.0 G/cm. This corresponds to 0.3 pixels at 4-cm FOV and 256 x 256 resolution. In some cases, particularly when the detailed structure of the bones and articular cartilage is of interest, the reduction of this artifact alone may be an appropriate reason for using strong gradient fields. An important consideration is that although improved SNR may be gained by using shorter TE, the contrast-to-noise ratio of the structure of interest may increase, decrease, or remain the same. The contrast around the tendons increases sharply at very short TE because the sheaths become hyperintense, while the tendons themselves remain hypointense. Conversely, the triangular fibrocartilage in the wrist shows low contrast at the shortest TEs because its signal intensity increases, and it becomes isointense with its surroundings. In our study, at the shortest TE of 3 msec, the only structures that remained hypointense were tendons, bone, and some ligaments. We note that in using very short TE sequences, the flow artifacts in the phase-encoding direction are somewhat reduced. This is due to the very short time allowed for the dephasing of flowing spins. Also with short TEs, more sections can be acquired in a given time because more TE intervals will fit within a given repetition time. Apart from allowing the use of very short TE sequences, the local gradient coil makes very high resolution imaging possible. Section thickness and FOV can both be reduced without requiring use of longer TEs, and any type of imaging pulse sequence can

be used. Local gradient coils are also very useful with other pulse sequences that are very demanding of gradient strength and switching times, such as echo-planar and diffusion imaging, and we have implemented these sequences with our gradient coil. The gradient coil described herein is large enough to accommodate the hand, wrist, and forearm. Larger cylindrical gradient coils can be constructed to accommodate the arm, lower extremity, and head at the cost of lower

efficiency

and/or

increased November

1991

minimum switching times. The use of surface gradient coils may allow strong gradients and fast switching times to be realized in the thorax and abdomen. The images presented herein are of healthy volunteers, and we have attempted to identify anatomic structures that can be better visualized with use of our techniques. We are currently beginning a clinical study of use of the same techniques to determine in which pathologic cases the techniques are clinically applicable. U

Acknowledgments: tance

#{149} Number

2

our

imager.

Johnson

building

study.

Milwaukee,

in incorporating for

E.C.W.

local

Robert Vavrek for assis-

gradient

We also thank the

RF coils

acknowledges

the

Medical Scientist Training cal College of Wisconsin.

Program

used

1.

D.

4.

in this

support

of the

at the Medi-

5.

A, Hyde JS, Kneeland JB. Pulse sequences for small fields of view (abstr). In: Book of abstracts: Society of Magnetic Resonance in Medicine 1988. Berkeley,

Society

of Magnetic

1988;

Kneeland

LB.

MR imaging

Resonance

JS.

(abstr).

JS,

Rilling

RJ,Jesmanowicz

AJR 7.

A.

Pas-

of surface

coils by pole 1990; 89:485-495.

in-

Reson Erickson 5, Kneeland JB, Macrandar S, Jesmanowicz A, Hyde JS. MR imaging of the finger: correlation with anatomic sections. 1989; 152:1013-1019.

Middleton WD, Kneeland JB, Kellman GM, et al. MR imaging of the carpal tunnel: normal anatomy and preliminary findings in the carpal 148:307-316.

High-resolution

with local coils. Radiology

1989; 171:1-7. Wong EC, Jesmanowicz A, HydeJS. mization of coils for MRI by conjugate

descent

in

1041.

Hyde

Hyde

sive decoupling sertion. J Magn

6.

Medicine,

of Magnetic Resonance in Medicine Berkeley, Calif: Society of Magnetic Resonance in Medicine, 1990; 517. Wong EC, Jesmanowicz A, Hyde JS. Coil optimization for MRI by conjugate gradient descent. Magn Reson Med 1991; 21:39-48. 1990.

Jesmanowicz

Calif: 2.

Society

coils

Richard

References

dient

181

Systems,

into a Signa

3.

Volume

We thank

of GE Medical

tunnel

syndrome.

AJR

1987;

Optigra-

In: Book of abstracts:

Radiology

#{149} 397