Robert R. Edelman, MD * Heinrich P. Mattle, MD * Dennis J. Atkinson, MS . Thomas Hill, MD * J. Paul Finn, MD * Chaim Mayman, MD 9 Michael Ronthal, MD * Henri M. Hoogewoud, MD . Jonathan Kleefield, MD

Cerebral Blood Flow: Assessment with Dynamic Contrast-enhanced T2*mweighted MR Imaging at 1.5 T1 The authors assessed regional cerebral blood flow dynamics with magnetic resonance (MR) imaging enhanced with gadolinium diethylenetriaminepentaacetic acid (DTPA). After bolus administration of Gd-DTPA, rapid T2*weighted gradient-echo images were acquired. Image acquisition time ranged from 2 to 3 seconds. The signal intensity (SI) of brain tissue and blood vessels markedly decreased during the first pass of contrast agent through the brain due to the local field inhomogeneity caused by the concentrated paramagnetic contrast agent. The method was used in 18 subjects with no cerebrovascular disease and 32 patients with stroke, vascular stenosis, arteriovenous malformation, and cerebral neoplasm. Comparison with intracranial angiography was performed in three patients and with single-photon emission computed tomography of blood flow in four. The change in T2* relaxation rate was approximately linearly related to the dose of contrast agent. The SI change increased as the echo time was lengthened. Regions in cerebral infarcts, metastases, and arteriovenous malformations showed different enhancement patterns than those of edema around a lesion and of normal brain tissue. Abnormal circulation times in patients with vascular stenoses were demonstrated. The method provides information about cerebral blood flow dynamics not available from conventional MR imaging and MR angiography.

diethylenetriaminepentaacetic acid (DTPA), a paramagnetic contrast agent, is routinely used to study a variety of cerebral disorders. Tl-weighted magnetic resonance (MR) images of the head that are enhanced with contrast material typically show an increase of signal intensities (SIs) in regions of bloodbrain barrier disruption and in wellvascularized tissues with incomplete or absent blood-brain barrier, such as the pituitary gland and nasal mucosa. On the other hand, conventional GdDTPA-enhanced Ti-weighted images show minimal or no change in the SIs of regions of the brain where the blood-brain barrier is intact, such as normal gray and white matter. If a compact bolus of Gd-DTPA is administered and rapid imaging is performed, a different effect may be observed. The paramagnetic agent produces local magnetic field inhomogeneity when it passes through GADOLINIUM

Index terms: Arteriovenous malformations, cerebral, 17.75 * Blood-brain barrier . Brain, infarction, 17.78 * Brain, MR studies, 10.1214 . Brain neoplasms, 10.33 . Cerebral blood vessels, flow dynamics, 17.1214 * Cerebral blood vessels, MR studies, 17.1214 . Cerebral blood vessels, stenosis or obstruction, 17.72 * Emission CT, 17.1211 * Gadolinium * Magnetic resonance (MR), contrast enhancement * Magnetic resonance (MR), rapid imaging

Radiology 1990; 176:211-220

' From the Departments of Radiology (R.R.E., H.P.M., D.J.A., H.M.H., J.K.) and Neurology (C.M., M.R.), Beth Israel Hospital, 300 Brookline Ave, Boston, MA 02215; Department of Radiology, New England Deaconess Hospital, Boston (H.P.M., T.H., J.P.F.); Siemens Medical Systems, Iselin, NJ (D.J.A.); and Department of Radiology, Hospital Cantonal, Fribourg, Switzerland (H.M.H.). Received November 15, 1989; revision requested January 12, 1990; final revision received March 29; accepted April 16. Address reprint requests to R.R.E. £ RSNA, 1990

the cerebrovascular system in high concentration. On T2*-weighted images, this results in transient SI loss in and around blood vessels. We have applied this technique in a series of patients with stroke, metastasis, and other cerebral disorders to demonstrate information about cerebral blood flow dynamics not available with conventional MR imaging methods.

PATIENTS AND METHODS MR images were acquired on a 1.5-T whole-body imaging system (Siemens Medical Systems, Iselin, NJ). Fifty subjects were recruited into the study from persons undergoing Gd-DTPA-enhanced MR imaging of the brain. The age range was 14-84 years (mean, 54.7 years). The control group consisted of 18 subjects without a history of cerebrovascular disorders and with no evidence of brain lesions on nonenhanced MR images. Most of them were referred to rule out a lesion such as acoustic neuroma or pituitary adenoma. The patient group consisted of 32 patients (Table). All studies were performed according to the guidelines of the hospital Committee on Clinical Investigations. The pulse sequence used for dynamic imaging was designed to produce a strongly T2*-weighted image that would be sensitive to the dephasing effects of the Gd-DTPA bolus during the first pass through the brain. For this purpose, a spoiled gradient-echo sequence with a long echo time (TE) was employed. To assess the dependence of signal change on TE and on the dose of the contrast agent, six patients without cerebral disease were studied with a multiecho sequence with four gradient echoes (TEs of 10, 20, 30, and 40 msec). In three of the six patients,

Abbreviations: DTPA = diethylenetriaminepentaacetic acid, ROI = region of interest, SI = signal intensity, SPECT = single-photon emission computed tomography, SRR = signal reduction ratio, TE = echo time, TR = repetition time.

211

180

1 80 -1

1 807

160-160-

160 -

160 -

160 -

140

40

140 -

0

.

-

-

180

-

U ._

.,

1oo

._

-0--- TE 10

-

---*-

100-

TE 20

100

-o---

TE 30 TE 40

* *

80 -

80 -

.,. I

F0

10

0

.

60

I

30

20

0

40

TE TE TE TE

20 30 40

160

60 -

160

' 120-

100

,

100'

*

-.-*-

60

20

30

Time (sec)

d.

80

.

1

0

40

---G

TE 20

TE 30 TE 40

0

TE 40

I

10

-

80 -

TE 30 0

120 100 -

TE 20

-

-

-

-

60

10

20

30

,

0

40

10

.

TE 10 TE 20

-

TE 30

-0

TE 40

*

.

20

*

30

.

40

Time (sec)

Time (sec)

f.

e. 1.

-

---- TE 10

-+--- TE 10

Figure

40

140 -

c 120

I

I

30

p

140 -

0

----I1

I

20

180 -

140

-

10

C.

180 -

60

I

C

10 20 30 40

Time (sec)

b. 180

O

60

0

Time (sec)

a.

TE TE TE TE

0.-

80 -

40

30

20

10

-

10

I

0

Time (sec)

80

, 120-

' 120

e 1 20

Example of a multidose, multiecho study in one patient. Gray matter signal versus time for dynamic T2*-weighted imaging fol-

lowing bolus administration of 2.5 (a), 5.0 (b), and 10.0 (c) mL of Gd-DTPA. The signal change increases with dose and TE. White matter signal versus time for dynamic T2*-weighted imaging following bolus administration of 2.5 (d), 5.0 (e), and 10 (f) mL of Gd-DTPA.

the multiecho sequence was performed three times with successively increasing doses of Gd-DTPA (2.5, 5.0, and 10.0 mL); there was a delay of approximately 5 minutes between the administration of each dose. In each of the other three patients, a single 20-mL dose was administered before the multiecho sequence was performed. For clinical studies, a gradient-echo

pulse sequence was employed; this sequence was slightly modified over the course of the study. The most currently used imaging parameters were a repetition time (TR) of 35 msec, TE of 25 msec (TR/ TE = 35/25), one excitation, a flip angle of 100, a field-of-view of 23 X 16 cm, a section thickness of 8 mm, and an acquisition matrix of 256 X 80 (in-plane spatial resolution = 0.9 X 2.0 mm). The flip angle of 100 was chosen to minimize T1 weighting. As a result of the small flip angle, precontrast differences in the SI of gray and white matter and postcontrast Tl-dependent changes in SI due to Gd-DTPA are rela212

*

Radiology

Summary of Cerebral Disorders in Study Population Disorder

No- of Patients

Control group None

Patient group Carotid irrftLnr

f rCt Verterobasilar territory intarct Transient ischemrc a ttFack Carotid stenosis or occlusion I ntracerebral hemorrha ge Trumar(men ingi olnTa, uLg cancer nmetastasis, g iubIa sltoma,

Lymphoma, pineal region tumonir) PoFrrn c1pKalIjc cysts after tumor resecti (nI

tively small. Typical imaging times were 2.8 seconds per image, with a delay of approximately 1 second between images. Axial images were obtained at a level that included the basal ganglia in subjects without evidence of cerebral disorders and at the level of the lesions in patients with abnormalities. In five subjects without cerebral disease, two patients with multiple

12 4

2 2 3 b

3

strokes in the vertebrobasilar territory and clinical suspicion of vertebrobasilar insufficiency, one patient with cerebellar stroke, and one patient with bilateral occipital strokes, coronal images were obtained through the cerebellum and parietal or occipital lobes to assess the relative blood flow dynamics through the anterior and posterior circulations.

July 1990

20

0.02

-

-

z 0

0

Z

/*-U-/--.-*

2.5 cc Gadolinum-DTPA 25cc Gadolinum-DTPA 10 cc Gadolinum-DTPA 20 cc Gadotinum-DTPA

10-

2.5 cc Gadolinium-DTPA S cc Gadolinium-DTPA 10 cc Gadolinium-DTPA 20 cc Gadolinium-DTPA

E 0.01

E

P

--

6)

ax

l,:1. To

0s

0.001

I

15

25

35

45

I

10

TE

20

30

40

50

TE

a.

b. y - .00a - 7.75C(-5.

R-squwd: 694

26.

024

0

022

Figure 2. Results of multiecho studies in six patients (three received Gd-DTPA doses of 2.5, 5.0, and 10.0 mL; three received a dose of 20 mL only). (a) Mean values for ratios of SI reduction to noise versus noise in gray matter. Little improvement is seen in TEs of more than 30 msec. (b) Mean values for maximum T2* rate change (AR20mas) versus TE. As expected, the values are essentially independent of TE. (c) Scatter plot of AR2*maX for gray matter versus dose of Gd-DTPA at a TE of 30 msec. Regression line is included. Dose is expressed in milliliters.

O

02

018 016 01 4 5

012 cq

01

006

0

006.

0

004e 002. 0

2

4

6

6

O 10

Dose

12

,............... 14 16 18 20

22

C.

Careful attention to the injection technique was necessary to ensure a compact bolus of Gd-DTPA. An 18-gauge intravenous line was placed in the antecubital fossa, usually of the right arm, with adequate extension tubing to permit injections outside the bore of the magnet. A 60-mL syringe was used to draw 40 mL of 0.9% saline solution, followed by GdDTPA (gadopentetate dimeglumine; Berlex, Wayne, NJ). If the syringe was maintained vertically and the Gd-DTPA was drawn slowly, then the contrast agent remained as a separate phase at the bottom of the syringe. This permitted faster, more uniform flushing of the contrast agent than if the flush and contrast agent were administered in separate syringes. Either 10 or 20 mL of Gd-DTPA (469 mg/ mL) was used for most studies. This corresponded to a dose in the range of 0.1-0.2 mmol per kilogram of body weight. A series of 25 images was obtained. After five images were obtained, the Gd-DTPA and saline solution was administered through a three-way stopcock as a rapid bolus. The duration of the bolus varied somewhat among patients, but the contrast agent (excluding the saline flush) could be administered in fewer than 5 seconds in most.

In addition to the dynamic study, a single flow-compensated gradient-echo image was

obtained before contrast material

injection at the same position as that used for dyamic images to determine vessel anatomy. Precontrast Ti-weighted and T2-weighted spin-echo images were also obtained after completion of the dynamic Volume 176 * Number 1

imaging. For Ti-weighted imaging, the parameters were 600/20, with one excitation; for T2-weighted imaging, 2,0002,500/40, 90, with one excitation. The section thickness was 5 mm; field of view, 23 X 23 cm; and acquisition matrix, 256 X 192. The dynamic images were analyzed by measuring pixel intensities over small, square regions of interest (ROIs), which typically included at least 25 pixels. ROIs were measured over the cerebral gray matter, cerebral white matter, basal ganglia, the diseased area, and the background. The flow-compensated gradientecho image was used as a guide to prevent placement of the ROIs near large blood vessels. The maximum signal reduction ratio (SRR) was calculated as SRR =

1

(Slpostcontrast

at peak Si

loss/Siprecontrast),

and expressed as a percentage. The T2* rate change (AR2*) was calculated as AR2* = -ln(Si/SO)/TE, where So is the precontrast SI, and S is the SI at time i. In a group of patients without significant cerebral disease in whom single-dose, single-echo studies were obtained (n = 6 at 10 mL of Gd-DTPA, n = 10 at 20 mL of Gd-DTPA), the area under the curve for the T2* rate change ( )AR20) was calculated as the sum of the AR2* values. No correction was made for recirculation of the contrast agent because the signal change from it was minimal. In addition, the maximum AR2* values (AR2*max) were noted. The values for gray and white matter were compared by means of a paired t test and simple regression analysis. Calculated maps of SI change were cre-

ated by dividing the SIs on a nonenhanced image by those on an enhanced image on a pixel-by-pixel basis, after thresholding and scaling factors were entered by the operator. The pixel intensity in the resulting image depends on the change in SI produced by the first pass of the contrast agent, with regions of greatest SI change appearing most hyperintense and regions of no SI change appearing hypointense. In addition, ratios of SI reduction to noise were computed as [(Slpostcontrast at peak SI loss SIprecontrast)/ (SD of the background SI)], where SD is the standard deviation. The standard deviation of the background SI was measured over a large region in the right upper quadrant of the image. Single-photon emission computed tomography (SPECT) of brain flow was performed following injection of 20 mCi (740 MBq) of technetium-99m hexamethylpropyleneamine oxime (Seretec; Amersham, Arlington Heights, Ill). Ten levels were scanned with a dedicated-head SPECT system (SME-810; Strichman Medical Imaging, Medfield, Mass) for 5 minutes per section (approximately 2,000,000 counts per section). Relative blood flow was determined by means of ROI measurements. Dynamic MR images were obtained on the same day of the SPECT study in two patients and within 2 and 14 days in the other two patients. The section for the MR image was selected to correspond to the level of abnormality on the SPECT scan or to the level of the basal ganglia if findings were normal on the SPECT study.

Radiology * 213

b. C. d. a. Figure 3. Dynamic T2*-weighted studies in a healthy subject. (a) Proton density-weighted spin-echo image (2,500/25). (b) Precontrast T2*weighted image (42/27). (c) T2*-weighted image obtained at peak of Gd-DTPA-induced SI reduction. The largest SI reduction is seen in and around medium-size to large vessels, followed in order by cortical gray matter, basal ganglia, and white matter. (d) Image calculated from b and c. Gray matter appears more intense than white matter, presumably due to larger blood volume. Medium-size to large vessels and nearby cerebrospinal fluid in the subarachnoid spaces appear very intense. The frontal horns of the lateral ventricles appear hypointense because there are no choroid plexus vessels. I

b. c. d. Figure 4. Subacute (3-day-old) right cerebellar infarct. (a) Axial T2-weighted spin-echo image (2,500/90). (b) Precontrast gradient-echo image (42/27). (c) T2*-weighted image obtained at peak of Gd-DTPA-induced SI reduction. Note absence of SI change in the infarct and increased SI change of adjoining brain tissue (arrows). There was no time difference in the filling of the anterior and posterior circulations. (d) Calculated image shows absence of SI change in the infarct.

a.

RESULTS Healthy Subjects The results of the multiecho, multidose studies are summarized in Figures 1-2. The SI change increased with the dose of contrast agent and TE. Correlation of gray matter AR2*max with the dose of Gd-DTPA showed an approximately linear relationship. This is shown for a TE of 30 msec in Figure 2c (r = .833, F = 22.6, P = .001). Delayed (>1 minute) T2*weighted gradient-echo images obtained after the dynamic studies showed only minimal reductions in SI within gray and white matter; these changes were not visually apparent. Artifacts due to diamagnetic susceptibility variations from the paranasal sinuses and petrous bones worsened as the TE was lengthened.

214 * Radiology

For single-echo, single-dose patient studies at a dose of 20 mL, the SRR was 27.7% + 7.7% for gray matter and 15.3% i 6.7% for white matter; the ratio of SI reduction to noise was -14.8 + 5.7. The single-echo, single-dose patient studies showed f AR2* values for gray and white matter were, respectively, .025 + .015 and .014 ± .011 for 10 mL of GdDTPA and .036 + .016 and .018 + .011 for 20 mL of Gd-DTPA. The t test showed these values were significantly different at doses of 10 mL (P = .004) and 20 mL (P = .001); linear regression analysis showed a significant correlation between the SAR2* values for gray and white matter (r = .79, F = 42.7, P = .0001). The maximum AR2* values for gray and white matter were, respectively, .008 ± .003 and .004 ± .001 for a dose of 10 mL

and .013 + .004 and .007 ± .003 for a dose of 20 mL. The t test showed the maximum AR2* values were significantly different at 10 mL (P = .007) and 20 mL (P = .0001); linear regression analysis showed significant correlation between the AR2*max values for gray and white matter (r = .89, F = 99.8, P = .0001). The ratio S AR2 gray matter! SAR2*white matter. computed for individual patients, was 2.0 ± 1.2. All dynamic studies produced good differentiation of gray and white matter (Fig 3). In subjects without cerebrovascular disease, the flow dynamics were symmetric, with appearance and disappearance of contrast agent occurring simultaneously in the two cerebral hemispheres. Coronal acquisitions showed simultaneous appearance and disappearance

July 1990

b. d. c. e. Figure 5. Large acute (less than 6 hours) stroke of the left middle cerebral artery. (a) T2-weighted spin-echo image shows an extensive area of SI abnormality and midline shift. (b-d) Three sequential calculated images from dynamic study. No SI change is shown within the stroke in b; SI change occurs along the periphery of the stroke in the late image (d). No hyperemia surrounds the stroke at this early stage. Visualization of the left middle cerebral artery (arrows) is delayed in the late image (d). (e) Delayed Tl-weighted spin-echo image shows vascular enhancement but no enhancement of the stroke.

a.

These arteries appeared artifactually enlarged, compared with their appearance in the flow-compensated gradient-echo and nonenhanced spin-echo images. The SI loss then appeared in the gray matter of the cortex and basal ganglia and in the white matter. The SI loss was later seen in the cerebral veins and sinuses. Minimal SI loss again occurred approximately 15-20 seconds after the first pass of the contrast agent, a finding consistent with recirculation of diluted contrast medium. The recirculation could not be visualized but was evident after numeric SI analysis of the sequential images. Two patients with no abnormalities on SPECT images also appeared to have no abnormalities on dynamic MR images.

a.

d.

c.

e.

f.

Figure 6. Studies in a patient with old (more than 2 months) bilateral parietooccipital strokes, occlusion of the left internal carotid artery, and collateral flow to the left middle cerebral artery from the right internal carotid artery over the anterior communicating artery. MR angiogram showed weak SI in the left middle cerebral artery, consistent with slow flow. (a) T2-weighted spin-echo image shows bilateral parietooccipital strokes. (b) Flow-compensated gradient-echo image shows the occlusion of the left internal carotid artery. (c-f) Four sequential calculated images from dynamic study show SI changes occurring in the territory of the right middle cerebral artery before those in the left, with delayed washout from the territory of the left middle cerebral artery (arrowheads in f) the late phase.

of contrast agent in the cerebellar and cerebral hemispheres, with no differences from side to side. A consistent pattern of SI changes was observed during the first pass of

Volume 176 Number 1 *

contrast agent through the brain; the

first pass effect was typically observed over three images. The first areas showing loss of SI were nearby medium-size and large arteries.

Stroke and Vascular Stenosis In 12 of 16 patients with stroke, there was minimal or no change in the SI at the core of the stroke during passage of the contrast material bolus through the brain, although the SI changes in the surrounding edema were comparable to or greater than those in white matter (Figs 4, 5). The images of three patients showed higher SRRs in the area of stroke than in normal gray matter; enhancement of the stroke borders on delayed Ti-weighted images was also seen in the same three patients. In one patient, a small subcortical defect in the SPECT study, representing a stroke, corresponded to an area of minimal SI change at dynamic MR. In a patient with stroke, an occlusion of the left internal carotid artery and collateral circulation to the left middle cerebral artery from the right internal carotid and basilar arteries, there was a one- to two-image delay in the appearance of the con-

Radiology . 215

-10~~~~~~~~~~~~~~~~-

i~m~ 0

PV

a.

c.

Figure 7. Studies in a patient with occlusion of the left internal carotid artery, severe stenosis of the right internal carotid artery, and collateral flow to the left middle cerebral artery from the right internal carotid artery and the vertebrobasilar system. (a) Four sequential calculated images display AR2* values after contrast material administration. Note slight delay in filling of the territory of the left middle cerebral artery. There was a 26% decrease of the f AR2* value in the left temporal lobe, compared with that in the right side. (b) Single image from dynamic study shows ROIs (light squares) used for measurements of SI changes. (c) Corresponding SPECT study shows an 11% decrease in the count rate in the left temporal lobe, compared with that in the right over the shown ROIs (squares).

trast agent and in the disappearance of contrast agent from the territory of the left middle cerebral artery (Fig 6). A one-image delay in the filling of the territory of the left middle cerebral artery was seen in a patient with a severe stenosis of the left internal carotid artery. Another patient with an occlusion of the left internal carotid artery and a severe stenosis of the right internal carotid artery showed collateral flow from the right internal carotid artery and from the vertebrobasilar system to the left middle cerebral artery. In this patient, there was a one-image delay in the filling of the left middle cerebral artery (Fig 7a). The SAR2* values for the right

216 . Radiology

and left temporal lobes, calculated from a large ROI corresponding to that used for the SPECT measurements (Fig 7b), were .078 and .058, respectively, values representing a 26% decrease on the left. The SPECT scan showed an 11% decrease in blood flow in the left temporal lobe relative to the right temporal lobe (Fig 7c). The frontal and occipital lobes appeared symmetric at SPECT and dynamic MR imaging. In two patients with clinical evidence of vertebrobasilar insufficiency and multiple infarcts of the brain stem, occipital lobe, and/or cerebellum seen on nonenhanced MR images, Gd-DTPAinduced SI changes of the arteries of

the posterior circulation and the cerebellum were delayed by one and two images, compared with the time course of SI changes in the cerebrum. No delay in filling of the posterior circulation was observed in coronal studies of five subjects without evidence of cerebrovascular disease or in two subjects with isolated cerebellar stroke and bilateral occipital strokes.

Neoplasms In five of six patients with brain tumors, the tumor nidus enhanced intensely during the first pass of contrast agent, with rapid but incom-

July 1990

a.

C.

Figure 8. Poorly differentiated lung carcinoma and presumed left parietal metastasis. (a) Coronal T2-weighted spin-echo image (2,500/90). (b) Sequence of calculated images (order = top left, top right, bottom left, bottom right) shows rapid enhancement of the tumor nidus, rapid washout with some retention of contrast agent within the tumor interstitium, and minimal SI change in the peritumoral edematous tissue. (c) Coronal postcontrast Ti-weighted spin-echo image (600/20) shows enhancement of tumor nidus due to breakdown of the blood-brain barrier and retention of Gd-DTPA within the tumor interstitium.

plete washout. The exception was a lymphoma, which had an SRR comparable to that of white matter. These tumors showed intense enhancement on delayed Tl-weighted images. The peritumoral edematous tissue showed less SI change than normal white matter during the first pass of contrast agent (Fig 8).

Hematoma and Arteriovenous

Malformation

In a patient with a hemorrhage of the frontal lobe due to a ruptured aneurysm of the anterior communicating artery, imaged 3 days after the acute event, the core of the hemorrhagic lesion-which was avascular at angiography-showed no SI

change. On the other hand, the SRR of hemorrhagic tissue in the periphery of the lesion was comparable to that of white matter. In a patient with an intracerebral hematoma due to an arteriovenous malformation,

the core of the hematoma showed no SI change, whereas the portion correVolume 176 * Number 1

sponding to the angiographically demonstrated arteriovenous malformation showed early SI change, compared with normal brain tissue and an SRR that was markedly greater than that of gray matter (Fig 9). The malformation could not be detected on conventional spin-echo images or on an MR angiogram.

DISCUSSION The SI changes observed in our study during the first pass of GdDTPA through the brain differ fundamentally from those seen on conventional enhanced MR images. The effect usually shown in conventional images is T1 shortening, which is due to short-range dipole-dipole interactions. Due to the existence of a blood-brain barrier and the fact that the vascular space constitutes less than 5% of the brain volume, the sensitivity for small changes in blood flow with the use of enhanced T1weighted imaging techniques is relatively low.

It has previously been shown in animal studies that a concentrated bolus of a paramagnetic agent such as Gd-DTPA can produce SI loss in the brain. The bolus of contrast agent distorts the local magnetic field homogeneity (shortens T2*). This results in dephasing and signal loss. Unlike the dipole-dipole effect seen on conventional Ti-weighted images, this T2* effect is of long range, extending beyond the vessel lumen, as shown by the large region of SI loss surrounding vessels (Fig 3c). Gradient-echo sequences performed with long TEs have been previously shown to be sensitive to T2* effects from paramagnetic substances in hemorrhage (1). At the doses of GdDTPA used in our study, the use of a long TE was also found to amplify the T2*-dependent SI change (Figs 1, 2). However, lengthening of the TE beyond 30 msec had little benefit in terms of the ratio of SI reduction to noise and also resulted in worsened diamagnetic susceptibility artifacts. The principles of using paramag-

Radiology

9

217

netic contrast agents to assess cere-

bral blood flow have been reviewed elsewhere (2). T2*-dependent signal changes likely relate primarily to flow within capillaries rather than larger vessels (eg, arterioles or venules) because of the greater fractional volume of the capillaries, although the fractional contribution from larger vessels is not known. Previous animal studies have shown an approximately linear relationship between the T2* rate change (AR2*) produced by a paramagnetic contrast agent and the concentration of the agent within the brain (2,3). The linear relationship seems to be comparable in our patient study also (Fig 2c), except at a dose of 2.5 mL, in which signal changes are noise limited. Animal studies also have shown a correlation between SAR2* and regional cerebral blood volume (4-6). When standard kinetic modeling is used, the calculated ratio of cerebral blood volume in gray and white matter was approximately 2, similar to values reported in the literature (7). Our results showed a ratio of approximately 2 at 10- and 20-mL doses of Gd-DTPA. This suggests that our f AR2* values may also correlate with cerebral blood volume, at least in healthy subjects. A few of our patients with severe stenoses or occlusions of the carotid artery showed a decrease in f AR2* in the territory supplied by the stenotic vessel. Previous studies with positron emission tomography have shown that cerebral blood volume and oxygen extraction fraction usually increase and that blood flow decreases in this situation (8-11). It may be that, at least in pathologic states, the f AR2* values represent not only blood volume but also a combination of blood volume and flow. Comparison with positron emission tomography is needed to evaluate the physiologic parameters represented by the fAR2* values. The lower limit of sensitivity of the dynamic T2* imaging method is not known but is highly dependent on the dose and magnetic moment of the contrast agent and on field strength. Because Gd-DTPA is not a purely intravascular tracer, breakdown of the blood-brain barrier can complicate image interpretation, as illustrated in Figure 8. Gd-DTPA is distributed into both the interstitial and intravascular spaces, so one might expect that the SI change of a tumor relates both to blood volume and to the volume and distribution of contrast agent that leaks into the tumor inter218 . Radiology

a.

C.

Figure 9. Studies of a 28-year-old woman with an acute intracerebral hematoma. (a) T2-weighted SE image shows left parietal hematoma. (b) Axial flow-compensated gradient-echo image (30/10, flip angle = 300) fails to show a definite abnormality other than mass effect. (c, d) Two sequential calculated images obtained after bolus administration of Gd-DTPA. Note early large SI change along the posterior margin of the hematoma (arrows in c), which occurs before enhancement of normal brain tissue. Minimal SI change is seen in the core of the hematoma. (e) Angiogram of the left internal carotid artery shows a tangle of vessels representing an arteriovenous malformation that is compressed by the hematoma.

stitium. If leakage were contributing significantly to Gd-DTPA-induced

T2*-dependent SI changes, then the AR2* of the tumor nidus should increase continuously after the first pass of contrast agent. In fact, the opposite was observed, since the AR2* within the tumor nidus rapidly de-

b.

a.

r

.*- ILi,

OR

..4 1" 4, t

%.

'..:d-

Idi.

e.

creased, rather than increased, after the first pass. This would suggest that SI changes relate primarily to the intravascular, rather than interstitial, distribution of contrast agent. The method may therefore be useful in providing a dynamic assessment of tumor vascularity.

July 1990

In 11 of 15 patients with stroke, no significant change in SI in the core of the infarct was observed during the first pass of Gd-DTPA. This would suggest a marked reduction of blood volume or flow to the region. In three patients with stroke, the SI change was equal to or greater than that of gray matter, a finding sugestive of reperfusion. However, angiographic confirmation was not available. In contrast to the edema associated with metastases, which showed only minimal SI change from GdDTPA (suggestive of reduced blood volume within the edema), the SI changes in the edema surrounding acute strokes were similar to or greater than those in the adjoining normal cerebral tissue (indicating normal or increased blood volumes). The dynamic T2*-weighted imaging method has several potential clinical applications. In evaluation of stroke, for instance, conventional MR images indicate only areas of edema, necrosis, or Wallerian degeneration. Dynamic T2*-weighted images provide additional functional information about regional blood flow dynamics. In addition, the technique provides dynamic information about circulation times, similar to previous studies with dynamic enhanced CT (12-15). For instance, in three patients with carotid stenoses and occlusions, there was a delay in enhancement of the left territory of the left middle cerebral artery. In two patients with vertebrobasilar insufficiency, delayed enhancement of the cerebellum and brain stem was noted relative to enhancement of the parietal lobes, but angiographic correlation was not available. There are several potential pitfalls to the method. First, because the sequence is T2* weighted, it is sensitive to susceptibility artifacts created by the paranasal sinuses and petrous ridges. These artifacts were generally minimal in calculated images displaying the ratios of pixel intensities in precontrast and postcontrast images. Second, there is marked enhancement of cerebrospinal fluid within the subarachnoid spaces and within the ventricles due to longrange susceptibility effects from the contrast agent as it passes through pial and choroid plexus vessels. In

patients with cerebral atrophy, care must be taken to compare the dynamic images with T2-weighted spinecho images to determine the distribution of the spaces containing cerebrospinal fluid. Finally, the method requires rapid, compact bolus admin-

Volume 176 . Number 1

istration of the contrast agent, which is feasible only if venous access is sat-

isfactory. There is disproportionate SI loss in the vicinity of middle-size and large blood vessels, which can create a false impression of vessel size or falsely suggest a hypervascular lesion. Correlation with MR angiographic images is helpful in image interpretation. A benefit of the vascular "blooming" in dynamic T2*weighted images is that flow can be detected even in vessels too small to be directly visualized with conventional MR imaging or MR angiographic techniques. The SI changes seen in gray and white matter presumably reflect effects from very small blood vessels, beyond the spatial resolution of MR angiography. This permitted the detection of an arteriovenous malformation (large SI change) that was compressed by an associated avascular hematoma (no SI change) (Fig 9). The malformation was not apparent with conventional spin-echo imaging or MR angiography. Further improvements in the temporal resolution of the dynamic T2*weighted imaging method can be achieved by eliminating the interimage delay, by using a smaller number of phase-encoding steps (eg, n = 64), and by using a shorter TR. The latter approach also necessitates a shortened TE, at the expense of sensitivity to T2* effects. Alternately, echo-planar imaging techniques (16,17) can be employed, although these require specialized hardware. The use of nonionic paramagnetic contrast agents that permit larger doses to be administered or of potent superparamagnetic agents (18) may also prove useful. In another approach for dynamic enhanced studies, an ultrafast gradient-echo sequence preceded by an inversion pulse is used. This technique generates T1 contrast similar to that in conventional inversion-recovery imaging (19-21). In our preliminary experience, the fast inversion-recovery method produces Tl-weighted images with lower spatial resolution and lower Gd-DTPA-induced ratios of SI change to noise than the T2* method but also may prove useful for assessing cerebral blood flow dynamics.

Other MR imaging approaches that do not employ contrast agents have

been proposed for assessing blood flow. For instance, it may be possible to use velocity-dependent phase effects to distinguish between very

slow molecular motion (diffusion) and faster capillary flow (perfusion) (22). Alternatively, one may assume that there is a net vector of capillary flow resulting in measurable, coherent intravoxel phase shifts (23). Further study is needed to determine the clinical utility of these methods. In conclusion, we have shown that first-pass T2*-dependent signal changes can be reliably produced in the brain after bolus administration of Gd-DTPA. Different patterns of signal change have been shown in stroke, vascular stenosis, tumor, and arteriovenous malformation. The MR imaging technique is simple to implement, gives reproducible results, adds minimally to the study time, and uses an approved contrast agent. It provides information about cerebral blood flow dynamics not available from conventional MR imaging and MR angiographic studies. U Acknowledgment: We thank Richard Buxton, PhD, for helpful discussions.

References 1. Edelman RR, Johnson K, Buxton R, et al. MR of hemorrhage: a new approach. AJNR 1986; 7:751-756. 2. Rosen BR, Belliveau JW, Chien D. Perfusion imaging by nuclear magnetic resonance. Magn Reson Q 1989; 5:263-281. 3. Fisel CR, Ackerman JL, Buxton RB, et al. MR contrast due to microscopically heterogeneous magnetic susceptibility: numerical simulations and applications to cerebral physiology. Magn Reson Med (in press). 4. Villringer A, Rosen RB, Belliveau JW, et al. Dynamic imaging with lanthanide chelates in normal brain: contrast due to magnetic susceptibility effects. Magn Reson Med 1988; 6:164-174. 5. Belliveau JW, McKinstry RM, Kennedy DN, et al. Functional imaging by gamma analysis of contrast enhanced NMR (abstr.) In: Book of abstracts: Society of Magnetic Resonance in Medicine 1989. Vol 1. Berkeley, Calif: Society of Magnetic Resonance in Medicine, 1989; 63. 6. Belliveau JW, Rosen BR, Kantor HL, et al. Functional cerebral imaging by susceptibility-contrast NMR. Magn Reson Med (in

press). 7. Greenberg JH, Alavi A, Reivich M, Kuhl D, Uzzell B. Local cerebral blood volume response to carbon dioxide in man. Circ Res 1978; 48:324-331. 8. Powers WJ. Positron emission tomography in cerebrovascular disease: clinical applications? In: Theodore WH, ed. Frontiers of clinical neuroscience. Vol 4. New York: Liss, 1989; 49-74. 9. Powers WJ, Raichle ME, Grubb RL Jr. Positron emission tomography to assess cerebral perfusion. Lancet 1985; 1:102103. 10. Raichle MR, Martin WRW, Herscovitch P. et al. Brain blood flow measured with intravenous H2150. II. Implementation and validation. J Nucl Med 1983; 24:790-798. 11. Powers WJ, Press GA, Grubb RL Jr, et al. The effect of hemodynamically significant carotid artery disease on the hemodynam-

Radiology * 219

12.

13.

14.

15.

ic status of the cerebral circulation. Ann Intern Med 1987; 106:27-35. Norman D, Axel L, Berninger WH, et al. Dynamic computed tomography of the brain: techniques, data analysis, and applications. AJR 1981; 136:759-770. Axel L. Cerebral blood flow determination by rapid-sequence computed tomography: a theoretical analysis. Radiology 1980; 137:679-686. Berninger WH, Axel L, Norman D, Napel S, Redington RW. Functional imaging of the brain using computed tomography. Radiology 1981; 138:711-716. Axel L. Tissue mean transit time from dynamic computed tomography by a simple deconvolution technique. Invest Radiol 1983; 18:94-99.

220 . Radiology

16. Pykett IL, Rzedzian RR. Instant images of the body by magnetic resonance. Magn Reson Med 1987; 5:563-571. 17. Ordidge RJ, Coxon R, Howseman A, et al. Snapshot head imaging at 0.5 T using the echo planar technique. Magn Reson Med 1988; 8:110-115. 18. Marchal G, Hecke PV, Demaerel P. et al. Detection of liver metastases with superparamagnetic iron oxide in 15 patients: results of MR imaging at 1.5 Tesla. AJR 1989; 152:771-775. 19. Haase A, Matthaei D, Bartkowski R, Duhmke E, Leibfritz D. Inversion recovery snapshot FLASH MR imaging. J Comput Assist Tomogr 1989; 13:1036-1040. 20. Atkinson Di, Burstein D, Edelman RR. First-pass cardiac perfusion: evaluation

with ultrafast MR imaging. Radiology 1990; 174:757-762. 21. Finelli DA, Kiefer B, Deimling M, et al. Dynamic contrast-enhanced perfusion studies of the brain with snapshot FLASH (abstr). Radiology 1989; 173(P):42. 22. LeBihan D, Breton E, Lallemand D, Aubin ML, Vignaud J, Laval-Jeantet M. Separation of diffusion and perfusion in intravoxel incoherent motion MR imaging. Radiology 1988; 168:497-505. 23. Young IR, Hall AS, Bryant DJ, et al. Assessment of brain perfusion with MR imaging. J Comput Assist Tomogr 1988; 12:721-727.

July 1990