Robert Dennis

R. Edelman, J. Atkinson,

MD MS

Segmented Method for of the Liver

#{149} Bernd #{149} Sanjay

#{149} Andrew

Wallner, MD Saini, MD

TurboFLASH: Breath-Hold with Flexible M

Singer,

MR Imaging Contrast’

A method called segmented turboFLASH imaging allows high-resolution, multisection, short-inversiontime (TI) inversion-recovery (STIR), Ti- or T2-weighted magnetic resonance (MR) studies of the liver to be completed within a breath-hold interval. The method was applied in a phantom and in 19 patients with hepatic lesions. Sequence comparisons were performed among segmented turboFLASH, single-shot turboFLASH, Ti-weighted gradient-echo with ultrashort echo time, and T2weighted spin-echo (SE) techniques. Signal from fat and liver could be nulled with the segmented turboFLASH method, with TIs of 10 and 300 msec, respectively; signal from these tissues could not be eliminated with the single-shot approach. Signal-difference-to-noise ratios and contrast for the best segmented sequences were comparable with those of the best T2-weighted SE and Ti-weighted gradient-echo techniques. It is concluded that it is feasible to obtain breath-hold images with arbitrary tissue contrast by means of segmented turboFLASH imaging. The method may prove helpful for the detection and characterization of hepatic lesions and will likely have applications to other anatomic regions such as the chest and pelvis.

scribed

Index

imaging times of 1 second or less. Echo-planar imaging is the fastest

terms: Liver, MR studies, 761.1214. Magnetic resonance (MR), experimental #{149} Magnetic resonance (MR), pulse sequences #{149} Magnetic resonance (MR), rapid imaging #{149} Phantoms

Radiology

I From Brookline

Ulm,

Ulm,

1990;

Federal

been

the Department of Radiology, 1990; revision requested April print requests to R.R.E. c RSNA, 1990

of Germany

(B.W.);

imag-

with

FLASH,

vari-

for studying disease inupper abdomen. There

particular

interest

izing hepatic ous hemangiomas.

masses

in using

such Because

as cavernrespira-

tory motion can degrade the quality of MR images of the upper abdomen, a number of techniques have been suggested to reduce motion artifacts. These strategies include signal averaging (1), fat suppression by means of short inversion time (TI) inversion recovery (STIR) sequences (2), dy-

consists

of a modified

gradi-

ent-echo technique and can be used on standard imaging systems (10-12). It uses gradient-echo sequences with very short (eg, 7 msec) repetition times (TRs). Tissue contrast can be altered by applying radio-frequency (RF) prepulses. For instance, a 180#{176} prepulse allows the creation of Tiweighted inversion recovery (IR)type images, whereas a driven-equilibrium analog prepulse sequence (90#{176}-180#{176}-90#{176}) permits more T2 weighting. However, given the gradient limitations of existing commercia! whole-body imaging systems, the minimum acquisition time for a

namic reordering of the phase-encoding steps (3), gradient motion rephasing (4), and breath holding in conjunction with fast imaging sequences such as FLASH (fast low-an-

high-resolution (eg, 256 X 128) image is nearly 1 second. The long acquisi-

gle shot) (5-8). All these strategies involve compromises. For instance, breath holding and extensive signal

a single section can be imaged, which renders the technique sensitive to misregistration of sequentially acquired images. We have developed a variant of these sequences that we call a segmented k-space turboFLASH acquisition. It permits arbitrary tissue contrast to be generated in high-resolution MR images. Moreover, these images can be acquired in a multisection mode within a single breath

averaging

are

not

to obtaining STIR reduces

readily

applicable

T2-weighted respiratory

images; artifacts

from

subcutaneous

fat

those tion.

due to splenic Gradient-echo

and bowel sequences

ultrashort high-quality

echo Ti-

times but

imaging ing times

However, specialized

have

permit

technique, of less

Siemens

mowith

MR

than

been

de-

imaging

permitting 100 msec

available. Another called snapshot FLASH

Systems,

Iselin,

time

tissue trast

contrast, and is not obtained.

or turbo-

Boston (5.5.). Received March 19; accepted June 20. Address

of and

28,

re-

method

MATERIALS

tech-

(D.J.A.);

The

entails

a substantial

loss

of

true IR or T2 conMoreover, only

was

used

to image

the liver in a series of patients, its performance was compared that of standard pulse sequences.

imag(9).

School, 330 University NJ

tion

hold.

with

this approach necessitates hardware that is not yet

Medical

Massachusetts General Hospital, 20; final revision received June

increases

(TEs) allow not T2-weighted

images to be obtained. Recently, methods that

but

of Radiology, Beth Israel Hospital and Harvard Medical MA 02215 (R.R.E., B.W., AS.); the Department of Radiology,

Republic

(MR)

resonance been used

MR imaging as a screening technique for liver metastasis and for character-

generally nique,

177:515-521

the Department Ave. Boston,

AGNETIC ing has

able success volving the has

#{149}

MD

AND

and with

METHODS

The single-shot turboFLASH sequence spans k space (ie, acquires all lines of data) after a single RF prepulse. For instance, to obtain IR-type contrast (IR turboFLASH), a single 180#{176} RF prepulse is

Abbreviations:

FLASH fast low-angle shot, inversion recovery, RF radio frequency, SD = signal difference, SE spin echo, SI signal intensity, S/N signal-to-noise ratio, STIR = short TI inversion recovery, TE = echo time, TI = inversion time, TR repetition time. IR

=

5i5

applied. With use of a TRITE of 7 msec/4 msec, 128 lines of data are acquired in 0.9 second. In contrast, the segmented turbo-

10-mm section thickness. Region-of-interest measurements compassing at least 25 pixels were

enused to

FLASH

measure

the

technique

segments

spans

of 32 steps

256

X 128-pixel

fourth

line

k space

each

image of data

in four

to generate

(Fig

a

1). Every

is acquired

sequential-

ly for each segment. With use of a TR/TE of 7/4, each segment is acquired in 224 msec. A repetition delay of approximately

3 seconds

is interposed

quisitions

for sequential

low Ti relaxation

between

data ac-

segments

to occur.

to al-

The RF pre-

pulse is applied at the beginning of each segmental acquisition. In addition, three preparation sequences (total duration, 21 msec) are applied to create a steady state

before each segmental data acquisition. The total imaging time (ie, the breath-

signal

intensities

holding interval) is 12-14 seconds. The contrast obtained with the singleshot and segmented IR turboFLASH se-

quences

was assessed

in a phantom

and TIs of 10, 150, 300, 600, i,200, 2,400

msec

A total

were

and

of 19 patients

with

hepatic

masses were studied. All patients had undergone computed tomographic (CT), ultrasound (US), or radionuclide scanning that documented the presence of one or more hepatic lesions. Findings included three cysts, three cavernous hemangio-

mas, one hepatoma, noma, lung,

one cholangiocarci-

and 1 1 metastases or colon primary

from tumors.

breast, Diagnoses

were based on biopsy proof or typical appearance at CT, at US, or on a technetium99m-tagged Studies whole-body

red blood cell study. were performed on a 1.5-T imaging system (Siemens

Medical Systems, Iselin, NJ) with standard hardware. The following sequences were used in all subjects: (a) T2-weighted spin echo (SE) (2,500/40, 90, 140; two cxcitations; first-order gradient motion mephasing for all three echoes; and imaging time of 10.7 minutes); (b) Ti-weighted breath-hold FLASH (110/5, one excitation, flip angle of 80#{176}, six sections per breath hold, and imaging time of 14 seconds); and (c) single-shot IR turboFLASH sequence (7/4; one excitation; flip angle

of 10#{176}; 180#{176} prepulse 500, time

700, and 1,500 of 0.9 second,

In addition,

three

with

TIs of 7, 300,

msec; and imaging not including the

types

of segmented

TIs).

ac-

quisitions were used: (a) IR turboFLASH (180#{176} prepulse with TIs of 10, 300, and 800 msec); (b) SE turboFLASH (90#{176}-TE/2180#{176}-TE/2-90#{176}prepulses with TEa of 40 and 80 msec); and (c) IR-SE turboFLASH (180#{176}-TI-90#{176}-TE/2-180#{176}-TE/2-90#{176} pre-

pulses with a TI of 10 msec and a TE of 40 msec). A magnitude reconstruction of the segmented turboFLASH data was performed on the host VAX II microcomputer (Digital Equipment, Maynard, Mass) with standard two-dimensional Fourier transformation; reconstruction time per image was 2 seconds. All images were acquired with a 256 X 128 matrix and

516 #{149} Radiology

liv-

Mz

RF

prepulses

RESULTS The the

phantom

differences

study

(Fig

in the

contrast

ior of single-shot boFLASH images. msec,

used.

in

a.

made

from tubes filled with various concentrations of copper sulfate. Ti ranged from 85 to 2,175 msec, and T2 from 54 to 529 msec. For both sequences, a flip angle of 10#{176}

(SIs)

em and lesions. Background SI and the standard deviation were measured with a large region of interest encompassing the entire region anterior to the abdomen. The ratio of lesion-liver signal difference (SD) to noise was computed as follows: SD/noise = Sliiver)/(standard deviation of background). Lesion-liver contrast was computed as follows: Contmast (Sliesion Slliver)/(Sllesion + Slliver). The signal-to-noise ratio (S/N) of liver was also noted. SD/noise and contrast for the various pulse sequences were cornpared by means of a t test for paired samples.

the

quence phantom markedly

reduced a Ti

the

turboFLASH

able

to null

the of 153

segmented

shot

turboFLASH

se-

the signal from the a Ti of 306 msec and

tom

with

behav-

and segmented turWith a TI of 10

segmented

nulled with

2) shows

SI of the msec.

version,

the

sequence

the

signals

phantoms with short test Ti of any phantom signal could be nulled

phanPhase-encoding

Unlike

singlewas

from

un-

rIhIhIhthIhIhIh1.h1hIhIhI

the

umes

3-5.

Images

obtained

the shorin which the with this se-

of good

with

quality

all pulse

quences in the i9 patients. hold images were uniformly respiratory artifacts. Signal

SeBreathfree from

of fat

and liver could be nulled with the segmented IR turboFLASH method with TIs of 10 and 300 msec, respectively; not be

signal from eliminated

these with

tissues could the single-

shot approach. Maximum lesion-liver SD/noise was achieved with the IRSE sequence (21.0 ± 22.7), although the differences were slight (-17.5 ± 10.0 with the short TE FLASH sequence and 16.0 ± iO.9 with the conventional T2-weighted with a TE of 90 msec).

ences SD/noise significant.

in the

absolute

ratios

With nulling liver, excellent was obtained;

were

!_L._1

Tis;

quence was 1,075 msec. Transverse banding artifacts, which presumably relate to buildup of unspoiled transverse magnetization, were worse in the single-shot than in the segmented sequence images. The results of the clinical study are summarized in the Table and in Figwere

SE sequence The differ-

values not

lines

of these statistically

of the signal from lesion-liver contrast hepatic lesions ap-

the

SEGMENT

SEGMENT

(j-

Hz

2

1

_VT ri>3ec

f

prepulse5

prepulses

b.

Figure 1. Comparison of single-shot kspace coverage (a) and segmented k-space coverage (b). In a, a single set of RF prepulses is applied initially; in b, the prepulses are applied at the start of each segment, with a 3-second delay between segments (only two of four segments are illustrated, and the preparation imaging is not shown). The duration of the acquisition in b is one-fourth as long as in a. In a, each

line

of data

every

fourth

is acquired line

sequentially;

of k space

is acquired

in b, se-

quentially. This minimizes signal discontinuities between the last line of one segment and the first line of the next.

peared bright against a dark background. Both the IR turboFLASH sequence and the T2-weighted SE Sequence with a TE of i40 msec produced the best lesion-liver contrast (0.63 ± 0.iO and 0.63 ± 0.18, mespectively). The next best contrast was achieved with a segmented IR-SE sequence

with

a TE of 40 msec

November

(0.58

i990

. . .

‘4 S

S

S S

..

#{149}.

.

S

S.

S

. .

S

S

S

S S

.

S

S

b.

a.

Figure 2. Images of phantom consisting of tubes containing various concentrations of copper sulfate. Ti /T2 values (in milliseconds), from bottom up, were as follows: Left column-I 85/54, 2 153/100, 3 306/i75. Right column-4 583/234, 5 = 1,075/406, 6 = 1,930/430, 7 = 2,175/529. (a) Single-shot turboFLASH sequence with TI values (in milliseconds) of 10 (top left), 150 (top right), 300 (middle left), 600 (middle right), 1,200 (bottom left), and 2,400 (bottom right). (b) Segmented IR turboFLASH sequence with same TI values. Note that the SI of the phantom with a Ti of 306 msec was nulled and the SI of the phantom with a Ti of 153 msec was markedly decreased with use of the segmented acquisition and a TI of 10 msec; on the other hand, the single-shot acquisition nulled the SI of the phantom with the much longer Ti of 1,075 msec at the same TI. The single-shot sequence did not null the SI of any of the phantoms with shorter Tls. Horizontal banding appears worse in the single-shot images, presumably because the acquisition of each phase-encoding step in sequence (single-shot method) allows a greater buildup of transverse magnetization than the acquisition of every fourth step (segmented method).

± 0.20). quence

The T2-weighted SE Segave a contrast value of 0.47 ± O.i7, and the gradient-echo sequence with a TE of 5 msec a value of -0.24 ± 0.09. Single-shot turboFLASH sequences gave maximum contrast

with a TI of 700 msec (-0.48 Comparison of the singleIR turboFLASH sequence that

0.21).

±

shot

produced msec) that

the with

best

the

produced

contrast

segmented the

best

(TI

700 (TI

=

300 msec) showed that the segmented sequence was significantly better (P < .02). The T2-weighted SE sequence with a TE of 140 msec and the segmented IR-SE sequence also produced better shot method

contrast than (P < .02).

the

The

single-shot

Volume

177

tumboFLASH

#{149} Number

2

and

et al were

a TR of 3

trast

able

positive.

to acquire

a

is negative The

and

T2 contrast

combination

opposing

time

results in a net loss of contrast. In addition, single-shot tumboFLASH is intrinsically a single-section imaging technique. The advantage is that the short imaging time (