Guy Paul

Marchal, MD #{149} Hilde Bosmans, MSc #{149} Luc Van Van Hecke, PhD #{149} Christian Plets, MD #{149} Albert

Intracranial and Clinical Time-of-Flight

fraeyenhoven, L

MD

#{149} Guy

Wilms,

T

arteries of the brain are relatively small and tortuous, with rapid flow during both systole and diastole. The veins are large and have slow but continuous flow. In brain HE

tissue, motion is restricted pulsations. Since normal vessels are anatomically

to vascular

intracranial restricted

to

a relatively small volume, three-dimensional techniques can be applied within an acceptable acquisition time. The magnetic resonance (MR) angiognaphy method used is based on the detection of inflowing spins into a saturated three-dimensional volume as described by Laub and Kaiser (1). In this method, inflow of unsaturated spins of blood into the volume of interest produces higher signal intensity than the spins of stationary tissues, which are continuously subject

to the

short

repetition

time

(TR)

gradient-echo sequence and are thus quickly saturated. With the use of refocusing gradients for first-order molion, phase dispersion can be prevented in laminar flow. However, the method fails to properly compensate

for

higher-order

bulence, Regions this

type

are

middle

the

cerebral

of the

osities

(2).

motion

Bifurcations,

stenoses herent

and

tur-

resulting in signal dropouts. particularly sensitive to of artifact

imaging

in brain

anterior artery

internal

vessel

confluences,

of signal

teries is partially sive saturation

in the

and

to inco-

peripheral

into the three-dimensional slab (2). This effect is even more dramatic in slow venous flow, with important consequences for the visualization of venous anatomy. The aim of the present study was to optimize the parameters of a threedimensional

an-

caused by pnogresof inflowing spins

Radiology

(MR),

1990: 175:443-448

pulse

sequences

of Radiology (G.M., and Neurology (C.P.), University Hospitals KU Leuven, Herestraat 49, 5-3000 Leuven, Belgium. Received October 27, 1989; revision requested DcI

From

H.B.,

the Departments

L.V.f.,

cember

accepted to G.M. C RSNA,

G.W.,

P.V.H.,

AND patients

Twenty-six

intracra-

METHODS with

proved

vascu-

lax anomalies of the brain, including 12 with arteniovenous malformation (AVM)

and 14 with

venous

spectively respectively)

angioma,

studied, with

MR angiography.

Gd-DTPA

proand

11,

(diethyleneBen-

acid) (Scheming, (tetraazacyclododecane-

tniaminepentaacetic

lin)

were

mainly (seven gadolinium-enhanced

or Cd-DOTA

tetraacetic acid) (Guenbert, Aulnay-sousBois, France) was administered by slow intravenous mmol/kg

injection immediately

at a dose of 0.1 before image ac-

quisition.

For each patient the MR angiogram was compared with the available conventional brain angiogram (n 17) and/or the previous MR study (n 23). The AVMS were assessed for the size of the nidus, origin

of the

feeding

arteries,

and

the

pattern of venous drainage (4). One patient with an AVM had already been treated with partial embolization, so a strict comparison with the pretherapeutic angiogram

case.

with

was

Venous regard

no

longer

possible

angiomas were to their location

in

that

evaluated and pattern

of drainage. Three-dimensional

time-of-flight MR angiography was performed on a 1.5-T superconducting magnet (Magnetom; Sicmens, Erlangen, Federal Republic of GemIn the

initial

phase,

brain

images

were acquired with a linearly polarized head coil. Later, they were obtained with a circularly polarized head coil that is

A.L.B.)

12; revision

received

January

22. Address

1990

sequence

of the

MATERIALS

many). Index terms: Angioma, central nervous systern, 10.365, 10.75 #{149} Arteniovenous malformations, cerebral, 17.75 #{149} Cerebral angiography, 17.1214 #{149} Cerebral blood vessels, MR studies, 17.1214 #{149} Magnetic resonance (MR), cine study. Magnetic resonance (MR), comparative studies. Magnetic resonance (MR), image display.

time-of-flight

for MR angiognaphy nial vessels.

the

knees of the and the tortucarotid artery

are also susceptible flow (2,3).

Loss

resonance

#{149}

Vascular Lesions: Optimization Evaluation of Three-dimensional MR Anglography’

The clinical value of three-dimensional time-of-flight magnetic resonance (MR) angiography was prospectively evaluated in 26 patients with congenital intracranial vascular lesions; 12 had arteriovenous malformations (AVMs), and 14 had venous angiomas. In the initial phase of the study the entire region of interest was imaged with one large acquisition volume (60-120mm-thick slab). Later, the angiograms were obtained with adjacent but slightly overlapping, 30-mmthick slabs, which clearly improved vascular detail. Gadolinium enhancement slightly improved depiction of veins but not of arteries. MR angiograms were compared with available conventional angiograms and MR studies. The topography of the AVM nidus was equally well appreciated on the MR as on the conventional angiograms. However, in six of 12 patients the hyperdynamic afferent arteries were incompletely shown on MR angiograms because of incomplete rephasing. In three patients, venous drainage was also incompletely visualized. Compared with conventional MR studies, MR angiography offered the same detection rate but better anatomic insight. Thirteen of the 14 yenous angiomas were also identified on MR angiograms. Detailed imaging, however, necessitated gadoliniurn enhancement and thin-slab acquisition.

Magnetic

MD

MD

Baert,

January

reprint

16, 1990; requests

Abbreviations:

AVM

arteniovenous

malfor-

mation, FISP fast imaging with steady pnecession, FLASH fast low-angle shot, TE echo time, TR = repetition time.

443

“

‘.

.‘,.

a.

b.

Figure

(a)

C.

Three-dimensional MR angiognams of healthy angiognam obtained in single large volume (120 above the sinuses is due to susceptibility artifacts. 1.

MR

gions

because

of saturation.

only minimal multiple-volume a and b have

Figure

effect

almost

Images

2.

(b) MR angiogram

on the acquisition.

obtained

visualization Dramatic

completely

of patient

with

of peripheral improvement

volunteer mm thick) Only the

a circularly

(TRITE in a linearly larger arteries

polarized

arterial branches. in vascular detail,

40/10, 20#{176} flip angle, FISP, maximum-intensity projection). polarized head coil. The higher signal intensity in the meare well shown. Most of the venous signal (arrow) was lost

head

There was both arterial

coil.

There

no effect on and venous,

is slight

improvement

in vascular

the veins (arrow). (c) MR can be seen. Susceptibility

detail

angiogram artifacts

but with seen

on

disappeared.

with

AVM

in

the night frontal region. (a, b) Selective angiograms of the right carotid artery. Huge frontal AVM drains into the sagittal sinus (arrow) and into the deep venous system (arrowheads). The feeding arteries are not distinct. angiognams tion (40/

(c, d, e) Gadolinium-enhanced with multiple-volume 10, 20#{176} flip angle, FISP).

MR acquisiIn c, the

complete maximum-intensity projection shows the nidus, an afferent artery (long anrow) branching from the middle cerebral artery, and the frontal (top arrow) and deep venous (bottom arrow) draining systems. In

d, a corresponding of the affenent e, a corresponding frontal

subvolume

shows

arterial branches subvolume

draining

veins

some

(arrow). In shows the

(arrow).

a.

b.

commercially available. (Siemens, Erlangen, Federal Republic of Germany). The different methods were compared on the basis of the following subjective criteria: the extent and detail to which the various cerebral arteries and veins were depicted and the presence of signal intensity

variations

on signal

dropouts,

which

define the continuity of vessels. The parameter settings of the thnee-dimensional time-of-flight MR angiography sequence were experimentally defined in healthy volunteers. To keep the echo time (TE) below 10 msec, only firstorder flow refocusing was used in the section-selection and frequency-encoding directions.

The

three-dimensional

ume

was

tion. ticed

No significant between fast

(FLASH)

on fast

cession

zoom

444

partitioned

(FISP)

factor

sequences.

Radiology

#{149}

axial

difference low-angle

imaging

and

vol-

in the

thus

the

was shot

with

steady

The

maximal

field

was

direc-

of view

limited

by the TE. The best resolution plane was obtained with the TE = 10 msec and field of view 20 cm, rather than with a shorten TE of 8 msec and a field of view of 25 cm.

duced versus

tion

direction,

the

smallest

Although

tion

thickness

was

used.

in the axial combination

no-

pre-

dephasing

is theoretically

me-

by shortening TE, a TE of 8 msec a TE of 10 msec did not significant-

ly affect

the

the

resolution

best

dephasing

artifacts.

in the

To

obtain

section-selecpossible This

parti-

resulted

May

in

1990

/ / b

I

;

/ b. Figure

3. Images of patient the posterior middle cerebral hypertrophic both the

nidus

feeding of the

C.

with AVM in the panietooccipital region. (a) Selective angiogram artery (solid arrow) and the draining cortical vein (open arrow).

arteries lesion

(arrow). and the

(c) Three-dimensional draining veins (arrow).

MR angiognam The hypertrophic

(40/10, feeding

20#{176} flip artery,

shows large (b) Spin-echo angle) with however,

feeding arteries branching from MR image (3,000/15) shows the

multiple-volume is not seen.

Figure

4.

lamic

Images

AVM.

giogram trophic

branches

echo

shows

thalamic

choroidal

and

arrow)

shows

with

right

tha-

right

vertebral

AVM

fed by hyper-

an-

thalamus-perforating

of the posterior

MR image

(solid

of patient

(a) Selective

(thin arrow) and venous drainage

‘l

acquisition

cerebral

strongly (thick

artery

hypentrophic arrow). (b) Spin-

(520/15)

shows

deep

the nidus

and the draining Three-dimensional

veins (open MR angio-

arrow). (c) gram (40/ 10, 20#{176} flip angle) with multiplevolume acquisition (maximum-intensity projection) shows the nidus and veins (arrow). The small feeding

however,

are not seen.

(d) Subvolume

angiogram at the level ison of vascular detail

the incomplete with

the

draining arteries,

MR

of the AVM. Companin c and d illustrates

rendering

of a large

maximum-intensity

volume

projection.

0

an anisotnopic

b.

in

__________________________________

. .‘

-a

J

.4

.

.2..

l._.

A)

.4,-.

-

.4

J

I d. Volume

175

Number

#{149}

2

plane

resolution

voxel

size and

of 0.8 X 0.8 mm2 a i-mm

section

thickness. The adverse influence of partial-volume effects was thereby also minimized. TR and flip angle have an immediate effect on vessel contrast. Suppression of stationary tissues obtained with increased saturation is TR dependent. However, decreasing TR not only suppresses signal from stationary tissues but also from yessels with slow blood flow. A TR of 40 msec gave the best overall results. The flip angle also influences contrast between the different components of stationary tissues. An angle between 15#{176} and 20#{176} provided the best results. To minimize the adverse effect of saturation on spins flowing slowly through the three-dimensional acquisition volume, the total volume of interest was acquired in successive and overlapping 30-mmthick, 32-partition slabs in the axial dimection (Fig 1). With the parameters used (TR

Radiology

.

445

msec/TE

msec

=

acquisition,

256

acquisition

time

Finally,

the

40/

10,

X 256 was

32 partitions, image

5.5

minutes

systematic

of gadolinium

the

saturation, of veins

(Fig 2). MR angiograms cine replay projections,

Table 1 Three-dimensional Time-of-Flight MR Angiography Conventional Angiography and MR Imaging

the

per

slab.

administration

lowered

and reduced sualization

one

matrix),

Ti

were

evaluated

viPatient

Region Interest

1

A

+++

+

0

N

+++

+++

+++t

on a

of the maximum-intensity calculated for a rotation

on

a

minutes.

This between

three-dimensional ing time.

was an clinically

display

acceptable useful

and

++

++

++

+++

++

+++

+++

+++

++

V

0

++

0

3

A

+++

. . .

+++ +++ +++ +++ +++

...

+++

. . .

+++

4

N V A N V

5

A

+++

7

ly

polarized

able in

and small

the quality of the MR improved dramatically study, when the circularhead coil became availthe angiogram was acquired

successive

intensity

projection

8

9

10

three-dimension-

a! subvolumes (Fig 1). However, all AVMs were successfully detected with MR angiography independently of the acquisition technique used (Figs 2-4). Gadolinium injection somewhat improved the visualization of the veins but had no effect on the visualization of the arteries. In particular, signal dropout due to dephasing in the hyperdynamic afferent arteries was not altered by gadolinium enhancement (Fig 3). The final two-dimensional angiograms, obtained with the maximumalgorithm,

often

12

Note-MR with

were, however, beyond the scope of this study). Comparison of the threedimensional MR angiognams with two-dimensional MR images or conventional angiograms is shown in Table 1 for the AVM patients. In six of 12 patients, the large feeding yessels were incompletely rendered on the MR angiograms because of low signal intensity, isointensity, or signal void in these vessels (Fig 3). However, the size and topography of the nidus were equally well assessed

Radiology

1 2 3 4

5 6 7

++

+++

+++ +++

+++

++

0

+++ +++

+++ +++

++ 0

RPCA

A

+++

0

+

Lfrontal

N

+++

+++

+++t

LACA

V

++

++

++

A N

+++ +++

+++ +++

+++

V

+++

0

+++

A

+++

+

N

+++

+++

+++

+

+++

. . .

+++

+++

. . .

+++

+++t

+

+++

. . .

+++

. . .

+++

N

. . .

+++ +++

++ +++

V

. . .

+++

+++

in patients

acquisition.

A

1-6 was with feeding

artery,

N

nidus.

MR Imaging

+++

region,

RACAandRMCA

Loccipital region, L?vfCAarnILPCA? R

thalamic region, R+LMCAandPCA

R

thalamic

region, and R MCA

R ACA

acquisition V

venous

and

parietal

+t

R

++

+

Posterior

. . .

+++t

++

L

++t

Posterior

+t

Posterior

fossa

0

Posterior

fossa

+

R

+++t

13

+++

+++

+++f

14

+++

+

++t

+++t ++t

+++t +1

exact

on the projected MR angiograms. The pattern of venous drainage was well demonstrated (Fig 4) in nine patients. Thirteen of the 14 venous angiobe identified

on

the

MR

single-volume

anatomy,

++

region region

Posterior fossa Posterior fossa R panietal region Lfrontal region R occipital region

+++t

1-9 was with +++

frontal

Rpanietal

++

+++ +++t

in patients

region

+t

+++

angiography

L panietal

+ + + t

region fossa

++

. . .

. . .

panietal

region fossa

++t

. . .

. . .

anterior

Location of Venous Angioma

MR Angiognaphy

. . .

exact

Angiomas

+++

. . .

7-12

+++

ACA

. . .

+++ +++ +++

in patients

drainage,

0 no visualization. middle cerebral artery,

Time-of-Flight MR Angiography of Venous Conventional Angiography and MR Imaging

Angiography

L MCA

Rfrontal

single-volume

9 10 11 12

could

region,

+

. . .

region,

L parietooccipital

+++t

N

region,

R occipital

+++

A

. . .

RMCA

0

V

with multiple-volume acquisition. visualization. 0 no visualization. * R = right. L = left. f Gadolinium enhancement.

mas

Rfrontal region, RACAand RMCA R panietal region,

+++

. . .

region,

+1

8

Note-MR

panietal RMCA

++ +++ ++

anatomy. ++ = incomplete anatomy, + partial visualization, * R = right, L = left, PCA posterior cerebral artery, MCA cerebral artery. I Gadolinium enhancement.

Patient

R

N V

angiography

with

region,RMCA

A

multiple-volume

Table 2 Three-dimensional

Rpanietooccipital

++

+++ +++ +++

+++ +++ +++

region,

RPCA

+++

A N V A

Compared

#{149}

+

R occipital

N V

V

11

provided only part of the vascular information from the acquired threedimensional data sets (Figs 2, 4). In particular, after gadolinium enhancement, small vessels tended to fade into the increased signal intensity of subcutaneous soft tissues. In such cases, vascular detail can be improved with three-dimensional cine display, reconstruction of a thinner slab out of the three-dimensional data set (Fig 5), or application of postprocessing filter procedures (which

446

Angiography

with

Location of AVM, Origin of Feeding Antery*

MR Angiography

N

6

In general, angiograms later in the

MR Imaging

A

comput-

RESULTS

of

V

2

Kontron-Mipron system (Kontron, Munich, Federal Republic of Germany) in steps of 1 #{176}around an axis in the sagittal plane (eg, -90#{176} -* 0#{176}: coronal -#{247}axial view). Total computing time varied from 40 to 80 compromise

Compared

of blood

improving the not of arteries

but

of AVMs

acquisition incomplete

and

in patients

anatomy.

+

10-14 partial

angiograms (Table 2). Three patients were studied with plain MR angiography and 1 1 with gadolinium-enhanced MR angiography. In venous angiomas, gadolinium enhancement

May

1990

Figure

Images

5.

venous giogram

angioma. shows

of patient

with

rolandic

(a) Digital subtraction typical venous angioma

an(an-

row) with umbrella-shaped distribution of medullany veins and large draining veins to the superior sagittal sinus. (b) Gadoliniumenhanced spin-echo MR image (600/15) shows strong enhancement of the lesion, with demonstration of both the medullany veins and draining veins. (c) Gadoliniumenhanced MR angiogram (40/10, 20#{176} flip angle, FLASH) with single-volume (90 mm thick)

acquisition.

Vascular

contrast

of the

angioma (arrow) is largely lost when the entire acquisition volume is projected. (d) Selective projection of the volume of interest (30 mm thick). The angioma (arrow) is seen with improved contrast.

:

\\‘

b.

a.

creased creases

4-6).

/

ond

:

I /Z’”\L ‘ ..

..

4 d.

c.

clearly

improved vascular detail. In the dilated medullary veins, converging to the central draining vein, were better visualized. Compared with contrast material-enhanced Ti-weighted images and, a fortioni, to conventional angiograms, MR angiograms acquired in one large three-dimensional volume showed poor detail, even after gadolinium enhancement (Fig 5). However, in four of five patients in whom a multislab acquisition was performed the overall quality of the angiograms was excellent (Fig 6). particular,

DISCUSSION MR angiography based on threedimensional gradient-echo sequences is an interesting approach for brain angiography (1 -3,5). Unfortunately, this technique, based on the refocusing of unsaturated inflowing spins, necessarily suffers from progressive saturation and incomplete refocus-

Volume

.

,,

.

175

Number

#{149}

2

most

technique dependent

,.,

..

However,

cant effect Incomplete

I,

.:“

with gadolinium, the Ti relaxation

ing. Progressive saturation is flow dependent and occurs more rapidly in slow-flowing venous blood. However, rapidly flowing arterial blood is also sensitive to this phenomenon (6,7). The degree of saturation depends on flow velocity (lower in elderly patients), radio-frequency deposition, slab thickness (acquisition volume), and the course of the vessel (perpendicular, oblique, on parallel to the acquisition volume) (1,8). Of these factors, only radio-frequency deposition and slab thickness are operator dependent. Preliminary studies in volunteers (Fig 1) showed that a TR of 40 msec and a flip angle of 20#{176} resulted in the best compromise for maximum vessel-to-background contrast, signal-to-noise ratio, and radio-frequency deposition. In addition, the use of the minimal available slab thickness (30 mm with 32 partitions) acquired in the axial plane further decreased saturation effects (68). Saturation in veins was also de-

this

which inrate (Figs 2,

had

in arteries. refocusing important

no signifiis the

drawback

(Fig 3). This and occurs

again because

secof this

is flow it is

difficult to compensate for high-onden motion with extra gradients. Theoretically, the best solution is to de-

crease

TE and

thereby

reduce

phase

shifts. However, this is limited by gradient strength, gradient switching, the decay of the induced eddy currents, and the field of view. In our clinical setting, we found that a field of view of 20 cm, at the expense of a somewhat larger TE (10 msec), was the best compromise for optimal spatial resolution and acceptable dephasing artifacts. Even though MR angiography easily demonstrated all AVMs and all except one venous angioma, there are still important shortcomings that prevent its application as a routine imaging procedure in the evaluation of symptomatic patients. Indeed, the goal of any treatment of a patient with AVM is to eliminate the risk of intracranial hemorrhage. To do so, more is required than just detection. Treatment success depends on total resection or embolization. Therefore, the diagnostic workup should indude a detailed assessment of the number and location of afferent feeding arteries, the size and topographic location of the nidus, and the anatomy of the draining veins (4,9). In this series, the major shortcoming of MR angiognaphy was the incomplete depiction of the hyperdynamic afferent arteries in about 40% of the AVMs (Fig 3). A similar experience was reported by Edelman et al (10). On the other hand, the nidus of these lesions was as easily and precisely assessed on the three-dimen-

Radiology

447

#{149}

a.

Figure nor

6. sagittal

hanced veins

Images sinus.

of patient with venous (b) Gadolinium-enhanced

MR angiogram (arrowheads).

angioma. (a) Digital subtraction spin-echo MR image (520/15)

(40/ 10, 20#{176} flip angle, The

components

of the

sional projections as with conventional angiography (Figs 2-4). Demonstration of the venous anatomy was also not a major problem in AVMs (Figs 2-4); however, this was cleanly more problematic in venous angiomas. During the early phase of the study, when the angiograms were acquired in one large volume 70-120 mm thick, only the larger transcenebral draining veins were successfully visualized (Fig 5). Despite the use of gadolinium to compensate for saturation, the smaller medullary veins were only occasionally seen. However, with the application of our multiple small-volume technique, MR angiography offered almost identical information as conventional angiography in venous angiomas (Fig 6). In MR angiognaphy based on gradient-echo sequences and time-offlight effects, it is possible that ernoneous information is provided by the paramagnetic effect of blood breakdown products, which shorten Ti and ties

lead to local signal hyperintensior, on the other hand, shorten

FISP) angioma

demonstrates are

Radiology

#{149}

peripheral seen

shows draining

branches

a large and

of anterior

angioma medullary

and

4.

GA,

with

gradient

put

Kaiser

Assist

WA. motion

Tomogr

1988;

TJ, Modic

MT.

tracranial

circulation: with

5.

Ross tailed 1988;

6.

JS, et al.

preliminary

In-

clinical (volume)

Radiology Modic

MT.

Ruggieni

1989; PM,

HJ,

Strother

CM,

in the

cortical

Kikuchi

Y, et al.

management

intracranial

JS.

of supra-

AVMs.

MR angiography

vascular 4:32-39.

Marchal

images.

C,

Wilms

AJNR

1988;

furnishes

de-

mations

of the

Hecke

brain de

Society 1989.

Intracranial

quence

(abstr).

(volume)

brain.

1-

JP, Engels based MR In:

Book

of

Resonance Calif: Society

Masaryk

1989; TJ, Modic

pulse-se-

in three-dimenMR

angiography.

ogy 1989; 171:785-791. Stein BM, Mohr JP. of the

July

in Medicine,

circulation:

considerations

sional

Congr#{232}s

Paris,

of Magnetic Berkeley,

of Magnetic Resonance 1:99. Ruggieni PM, Laub GA, MT.

P. et al.

17e

Radiologie,

methods

Abstracts: in Medicine

Int

malfor-

(abstr).

7, 1989. de Graaf RG, van Dijk P, Groen JLM. A comparison of inflow angiography

8.

Imaging

of arteriovenous

International

7.

Diagn

G, Van

MR angiography

Radiol-

Vascular malformaJ Med 1988;

N EngI

Edelman RR, Wentz KU, O’Reilly GV, et al. Evaluation of intracerebral arteniovenous malformations using selective magnetic resonance raphy (abstr). ety of Magnetic 1989.

TJ,

and

supe-

9:225-235.

10.

12:378-382.

Ross

three-dimensional

MR angiography. 171:793-799. Masaryk

arteries

the

319:368-370.

MR angiography J Corn-

refocusing.

Masaryk results

3.

Smith

tions

Laub

draining into (c) Gadolinium-en-

cerebral

tentonial

9.

2.

middle

MR imaging

References 1.

(arrow) veins.

(arrow).

T2, destroying vascular signal locally. Both effects would mimic vascular lesions. In conclusion, important progress has been made in the quality of MR angiograms of the brain. The data thus obtained and interpreted together with conventional two-dimensional images provide accurate information about the anatomy and topography of AVMs and venous angiomas in the brain. Although this technique is sufficient for diagnosis, it does not yet provide the detailed vascular and hemodynamic information obtained with conventional angiography, which is needed to plan either surgical on endovasculan treatment. U

et al.

Berkeley,

Resonance

In:

arteniography and venogBook of Abstracts: SociResonance in Medicine Calif:

in Medicine,

Society

1989;

of Magnetic

1:161.

Three-dimensional (volume) gradientecho imaging of the carotid bifurcation: preliminary clinical experience. Radiology

448

also

angiogram shows the

1989;

171:801-806.

May

1990