1991, The British Journal of Radiology, 64, 10-16

Magnetic resonance angiography of abdominal vessels: early experience using the three-dimensional phase-contrast technique By P. Vock, MD, F. Terrier, MD, H. Wegmiiller, MD, *F. Mahler, MD, tPh. Gertsch, MD, IS. P. Souza, PhD and §C. L Dumoulin, PhD Departments of Radiology, 'Internal Medicine, and tVisceral Surgery, University of Berne, Switzerland and § Research and Development Center, General Electric Company, Schenectady, New York, USA (Received April 1990) Keywords: Magnetic resonance, Angiography, Abdominal vessels, Phase contrast Abstract. Based on three-dimensional acquisition of three sequences sensitive to one flow-direction, abdominal magnetic resonance phase-contrast angiography (MRA) was performed in 13 volunteers and 20 patients. The subjects received no antiperistaltic medication and were allowed to breath normally during the three acquisition periods of 11 minutes. The frequency of demonstration of the normal aorta, superior mesenteric and right and left renal arteries was 100%/100%/91%/100%, and of the inferior vena cava, splenic, superior mesenteric and portal veins was 92%/67%/92%/100%, respectively, whereas other abdominal vessels were seen less constantly. In renal artery stenosis or occlusion, MRA detected eight out of nine pathological arteries, missed only a minimal stenosis and was never false positive. In all 10 cases of portal hypertension, MRA demonstrated the venous collaterals detected by conventional angiography and in six cases showed more collaterals, particularly paravertebral vessels. A Budd-Chiari syndrome was investigated as well. If the accuracy of MRA can be proved in larger studies, it may become an important diagnostic tool in evaluating abdominal vascular pathology, such as renal artery stenosis or portal hypertension.

Radiological angiographic methods, although they give excellent morphological information, involve ionizing radiation, the risk of side reactions to contrast media and a varying degree of invasiveness. Duplex sonography can overcome some of these disadvantages, but the access to deeper vessels may be difficult or impossible in the abdomen. Magnetic resonance angiographic methods have been introduced recently and afford functional flow information rather than morphological imaging (Wedeen et al, 1985a, 1985b; Dumoulin & Hart, 1986; Alfidi et al, 1987; Dumoulin et al, 1988, 1989a; Laub & Kaiser, 1988). Most applications have focused on the head, neck and extremities, whereas the trunk of the body with its physiological motion was initially neglected because of difficulty with breathing and peristalsis artefacts. This article reflects our 8-month experience with phase-contrast MR angiography (MRA) using threedimensional acquisition (Dumoulin et al, 1989a). The purpose of the study was to test the detectability of normal abdominal vessels and to assess the potential applications of MRA to vascular disease, especially renal artery stenosis and portal hypertension. Methods A 1.5 tesla imaging system with a shielded gradient coil subsystem (Signa, General Electric Co., Milwaukee, USA) was used. Flow-sensitive images were generated by a fast-imaging three-dimensional phase contrast technique (Dumoulin et al, 1989a). The field of view was 40 cm in the axial plane (256 left-right frequency encoding steps and 128 anteroposterior phase encoding steps) and 20 cm in the superoinferior dimension (128 10

phase encoding steps), occasionally 10 cm (for a matrix of 64). With a time to repeat (TR) of 20 ms, an echo time (TE) of 10 ms, a flip angle of 20° (occasionally 10°) and two excitations, one volume-study lasted roughly 11 minutes (5.5 minutes respectively for the matrix of 64). Three such acquisitions were obtained to encodeflowin all three directions. The three-way modulus method (square root of the sum of the squares) was used to combine all 128 (64) triplets pixel by pixel into 128 (64) axial images, each containing three-dimensional flow information. By the maximum pixel method, images of all axial anatomical planes were then both collapsed into one axial image (Fig. la) and projected into 32 coronal and oblique planes at view angle steps of 3.2° (Fig. lb). These were analysed three-dimensionally using a video loop. The same safety instructions and checks were taken as for ordinary MRI. However, using a research software, we did not control the specific absorption rate (SAR). Informed consent was given. The patients were all studied in the supine position, their abdomen being mildly compressed with a wide strap attached to the table. No medication was used to reduce gastrointestinal peristalsis, and the subjects were allowed to breath normally during the entire acquisition period. Thirteen normal volunteers (five female, eight male; aged 8^43 years) were studied. Ten had the ordinary protocol. In two we used a saturation pulse placed superiorly to the imaging volume in order to suppress the signals from the protons in the heart and the aorta; therefore, we were not able to analyse the arteries. In one volunteer the field of view was centred lower to include only the level of the renal vessels. A vessel was The British Journal of Radiology, January 1991

MR angiography of abdominal vessels

Figure 1. Normal three-dimensional phase-contrast MRA of the abdomen, (a) Axial view of 128 slices collapsed using the maximum intensity method, (b) Anteroposterior projection of the same data presented in (a). A: aorta; C: inferior vena cava; short arrows: renal arteries; arrow head: superior mesenteric artery; P: portal vein. Note the decrease of flow signal in the infrarenal aorta and, less obviously, in the suprarenal inferior vena cava.

considered to be detectable when its main portion was seen without an interruption of more than 5 mm. Twenty patients (11 female, nine male; aged 21-84 years) were studied by MRA for the presence of pathological vessels and the absence of normal vessel signals without knowledge of the results of other investigations. Nine were suspected of having renal artery stenosis or occlusion. Following the MRA study, all had conventional catheter angiography. Three had normal renal arteries, three unilateral stenosis or occlusion and three bilateral stenosis (nine pathological renal arteries). Three patients with four diseased arteries were also studied by MRA after percutaneous transluminal angioplasty (PTA). The other 11 patients had portohepatic disease; all were studied by duplex sonography, nine by catheter angiography, eight by dynamic computed tomography

(CT) and seven by magnetic resonance imaging (MRI). Ten patients had portal hypertension, three from extrahepatic and seven from intrahepatic portal venous obstruction; eight had cirrhosis, two hepatocellular carcinoma and one cholangiocarcinoma. One patient also had the Budd-Chiari syndrome caused by a malignant tumour. Results

Using the standard protocol without spatial saturation, the suprarenal aorta, superior mesenteric and left renal arteries were always and the right renal artery nearly always detected in volunteers (Fig 1; Table I). However, in the infrarenal aorta flow signal was always weaker than suprarenally and in three cases nearly absent. At the point of maximal inferior convexity of the renal arteries there was a short interruption of less

Table I. Abdominal MRA in normal volunteers: vessel detectability (« = 13") Veins

Arteries Aorta: suprarenal infrarenal

11/11 (100%) 8/11 (73%)

Common hepatic

2/10 (20%)

Splenic Superior mesenteric Renal: right left

6/10 10/10 10/11 11/11

(60%) (100%) (91%) (100%)

Vena cava Hepatic: right middle left Portal Splenic Superior mesenteric Renal: right left

12/13 7/12 7/12 0/12 10/12 8/12 11/12 6/13 8/13

(92%)* (58%) (58%) (0%) (83%) (67%) (92%) (46%) (62%)

a

In two studies, saturation of inflowing aortic flow was used, so that analysis of arteries was not possible; one study was limited to the renal vessels. *In four out of the 12 the signal intensity was much higher in the infrarenal than in the suprarenal vena cava.

Vol. 64, No. 757

11

P. Vock et al

(a) (b) Figure 2. Embolic occlusion of the right renal artery documented by conventional angiography (a) and by MRA (b); the axial view demonstrates normal flow in the left renal (arrow) and superior mesenteric artery (M), but no flow signal in the right renal artery. A: aorta; V: inferior vena cava.

than 5 mm of the renal artery signal in most of these normal volunteers. A good, nearly continuous signal from the splenic artery was observed in six of 10 volunteers, but only in two from the common hepatic artery. The inferior vena cava, left and right renal veins were well visualized in 92%, 62% and 46%, and the splenic, superior mesenteric and portal veins in 67%, 92% and 83%, respectively. Although visible, vena caval flow was often considerably weaker in the suprarenal than the infrarenal segment. The right and middle hepatic veins were shown in 58%, whereas the left hepatic vein was never more than segmentally demonstrated. In suspected renal artery pathology, MRA was truly negative in the three subjects with normal renal arteries. It correctly demonstrated reduced or absent signal over a long distance of more than 5 mm in eight out of nine renal artery stenoses or occlusions (Figs 2 and 3; Table II), and snowed normal signal intensity in one artery with minimal stenosis by conventional angiography which was not treated by PTA. In all four arteries of the three patients with a follow-up study, the signal intensity of the stenosed artery was slightly increased 1 day after PTA (Fig. 3c). Table HI compares the results of MRA {n—\G) and of catheter angiography (n = 8) in portal hypertension. Well-known collateral pathways (coronary, umbilical, short gastric and periportal veins) could be demonstrated by both methods, and MRA was at least as sensitive as catheter angiography (Fig. 4). It additionally showed curled paravertebral veins in eight out of 10 patients not seen with conventional angiography in the same patients. In one case with a history of bleeding due to portal hypertension, MRA, conventional angiography and duplex sonography were all unable to demonstrate collaterals. In one patient with neoplastic Budd-Chiari syndrome (Fig. 5), MRA demonstrated a high-grade stenosis of the right and middle hepatic veins and even more venous collaterals than detected by MRI, angiography and sonography. 12

The following were also detected: a supernumerary renal artery in one volunteer and two patients, a right hepatic artery arising from the superior mesenteric artery in a patient and a left ovarian vein in a volunteer. Discussion

Vascular disease has become one of the most common medical problems. Although there are several vascular imaging methods, each has its limitations. Conventional X-ray angiography is invasive, involves ionizing radiation and the risk of iodized contrast agents, and generally demonstrates only the injected territory. Duplex ultrasound avoids these disadvantages and offers quantitative flow information; however, its field of view is often limited and access to specific deep vessels may be difficult or impossible. Other tomographic methods, like CT and MRI, in addition to sharing some of these limitations, do not afford a topographic overview and the spatial resolution needed for small vessels. Despite its cost, a non-invasive imaging method with three-dimensional presentation capabilities like MRA might add to the spectrum of diagnostic methods. Table II. MRA of renal arteries with conventional angiographic correlation Conventional angiography

Number of patients"

MRA signal intensity Normal Reduced Absent

Normal artery 9 Embolic occlusion 1 Stenosis 8

9 0 1*

0 0 3

"Number of patients: 9; number of arteries: 18. ^Minimal degree of stenosis in fibromuscular hyperplasia at conventional angiography that did not justify percutaneous transluminal angioplasty. The British Journal of Radiology, January 1991

MR angiography of abdominal vessels

Figure 3. Right renal artery stenosis (arrow) by (a) conventional angiography and (b) MRA. The circumscribed morphological decrease in vessel diameter (a) correlates with the wide spread drop of flow signal (b). (c) After percutaneous transluminal angioplasty, improved flow signal distal to the original stenosis (long arrow) shows the decrease of obstruction. Table HI. MRA in portal hypertension: detectability of collaterals Patient

Collaterals Periportal

Coronary

R.O.

—j—

+ /H

RE.

-1-

E.B. S.T. M.U.

—/— —/—

+ /H —/-

-/-

-/-

S.A. S.C. F.R. F.U.

+/+ + /— —/—

-/-

4-/0

BO.

-10

-io

3 MRA(«=10) Conventional X-ray angiography (« = 8) 1

Paraumbilical

Short gastric

Retroperitoneal (splenorenal)

Paravertebral

Other

Epigastric —I—

Left ovarian

— /— /-

—/o 3

4-/0

-/o

-/o 2

4-/0 8

2

The first 4- (visualized), — (not visualized) or 0 (not performed) gives the result of MRA; the second gives that of conventional angiography. Vol. 64, No. 757

13

P. Vock et al

Figure 4. Portal hypertension due to liver cirrhosis, (a) Indirect portography from the superior mesenteric artery demonstrates hepato-fugal flow in the umbilical vein (arrow), (b) Axial collapsed MRA and (c) left anterior oblique projected MRA similarly demonstrate the recanalized umbilical vein (arrow). Portal vein: P.

Magnetic resonance angiography is based on the two prerequisites of sensitizing the signal to flow and of suppressing signal from static tissue. Two principal solutions have been realized, time-of-flight (Laub & Kaiser, 1988; Dumoulin et al, 1989b, Edelman et al, 1989a; Keller et al, 1989) and phase-sensitive techniques (Weeden et al, 1985a, 1985b; Dumoulin & Hart, 1986; Dumoulin et al, 1989a). Time-of-flight MRA depends on flow-related enhancement using either wash-in or wash-out phenomena, whereas phase-contrast MRA relies onflow-dependentphase shifts and a subtraction technique to detect flow in the readout, the phaseencoding or the projection direction. Volumetric acquisition (three-dimensional Fourier-transformation) (Dumoulin et al, 1989a, 1989b) affords smaller voxels with less signal loss due to magnetic susceptibility effects than sequential acquisition of two-dimensional sections. We used this with the phase-contrast technique and had to accept a longer acquisition time. Thus, in contrast to the two-dimensional time-of-flight and phase contrast methods we could not avoid respiratory motion by measuring during apnea. In a recent article, Edelman et al declared breath-holding a prerequisite for highquality angiograms of body regions subject to motion 14

artefact (Edelman et al, 1989a). Our results both in normals and in patients prove it is possible to demonstrate most major abdominal vessels during normal respiration and without antiperistaltic medication. The superior mesenteric and both renal arteries as well as the portal vein and its tributaries are reliably seen in a high percentage. Physiological abdominal aortic flow at rest above all maintains renal and visceral perfusion and, thus, is less in the infrarenal segment. On the other hand, the signal drop within the suprarenal vena cava seen in normal volunteers is most likely caused by flow pulsatility, as known from Doppler studies. Certainly, artefacts are common, such as those arising from the heart and the aorta in the phase-encoding direction that destroy the signal from the left hepatic vein. Our results still need to be improved if MRA is to contribute useful information on other abdominal vessels. However, we are confident that this will be possible with ongoing hardware and software adaptation and with new methods of saturation and motion compensation. Another new development is quantification of flow, and initial experience is very promising (Edelman et al, 1989b; Maier et al, 1989). Already, MRA is useful in difficult clinical cases, for The British Journal of Radiology, January 1991

MR angiography of abdominal vessels

Figure 5. Budd-Chiari syndrome and portal hypertension due to cholangiocarcinoma. (a) Axial MRI demonstrates the hypointense central mass lesion (m) infiltrating around the inferior vena cava (o) and splenomegaly, (b) Anteroposterior projection MRA confirms the functional stenosis of both the right and middle hepatic veins (arrows) and the inferior vena cava (o). (c) Portal hypertension is better demonstrated on the axial view by collaterals both of the umbilical/anterior abdominal veins (arrow heads) and the paravertebral veins (long arrow).

instance when iodized contrast agents or invasive angiography are contra-indicated. If our results can be confirmed in larger series, MRA might be used as a screening method for significant renal artery stenosis. We obtained a sensitivity of 89% and missed only one probably not significant stenosis. The fundamental difference between morphology-oriented conventional angiography and functional MRA must be emphasized. For instance, with significant stenosis in the order of 50% to 90%, MRA showed a complete signal void in the entire extrarenal segment of the renal artery beyond the point of obstruction. Likewise, the obviously physiological circumscribed signal drop at the point of inferior convexity of renal arteries correlated with normal morphology and might be caused either by dephasing due to intravoxel velocity gradients or by higher order motion. If this is true, loss of signal will be minimized with shorter TE. Currently, we do not know whether the morphological information of conventional angiography or, despite decreased spatial resolution, the functional information of MRA will give a better differentiation of significant stenosis. Furthermore, in residual or recurrent arterial hypertension, a non-invasive test for follow-up studies after PTA of renal artery stenosis would be extremely welcome. It is certainly too Vol. 64, No. 757

early to judge the value of MRA in detecting stenosed supernumerary renal arteries. Magnetic resonance phase-contrast angiography may soon have a significant role in portal hypertension. Although our phase-contrast method currently does not quantify flow, we add to the recent communication of four patients using the time-of-fiight technique (Edelman et al, 1989c) with our angiographically correlated data on 10 patients. In our experience MRA was at least as sensitive as and several times better than conventional angiography in demonstrating portosystemic and portohepatic collaterals. Specifically, we observed pathological paravertebral vessels not demonstrated by other methods. These probably represent portosystemic retroperitoneal collaterals and are not shown well angiographically on usual projections. Since any method using local contrast injection will reach only part of the portal territory, indirect arterial portography, splenoportography and transhepatic portography have special limitations as well as their invasiveness. However, MRA is sensitive to any laminar flow within the entire field of view without contrast injection and the potential risk of iodized contrast agents. Since the signal of MRA is based on flow, watershed collaterals with minimum flow may theoreti15

P. Vock et al

cally escape detection by this method. The non-invasive combination of spatially limited quantitative duplex sonography with the global survey of MRA may overcome such problems. Our protocol may be used to investigate veins other than the portal system, as illustrated by the case of Budd-Chiari syndrome. It will have to be adapted to any specific problem, e.g. switching the readout and the phase-encoding directions and using spatial arterial saturation might allow demonstration of all three hepatic veins. In conclusion, MRA of abdominal vessels is still in the phase of clinical development. Because its quality is not significantly degraded by respiration, the threedimensional phase-contrast method is an alternative to the two-dimensional time-of-flight technique with multiple sections during apnea. Renal artery stenosis and portal hypertension are currently the promising areas of application. Depending on confirmation of results by larger series, MRA might be used for screening or, combined with duplex sonography, occasionally replace conventional angiography.

DUMOULIN, C. L., SOUZA, S. P., WALKER, M. F. & WAGLE, W.,

1989a. Three-dimensional phase contrast angiography. Magnetic Resonance in Medicine, 9, 139-149. DUMOULIN, C. L., CLINE, H. E., SOUZA, S. P., WAGLE, W. A. &

WALKER, M. F., 1989b. Three-dimensional time-of-flight magnetic resonance angiography using spin saturation. Magnetic Resonance in Medicine, 11, 35-46. EDELMAN, R. R., WENTZ, K. U., MATTLE, H., ZHAO, B., LIU,

C , KIM, D. & LAUB, G., 1989a. Projection arteriography and venography: initial clinical results with MR. Radiology, 172, 351-357. EDELMAN, R. R., MATTLE, H., KLEEFIELD, J. & SILVER, M. S.,

1989b. Quantification of blood flow with dynamic MR imaging and presaturation bolus tracking. Radiology, 171, 551-556. EDELMAN, R. R., ZHAO, B., LIU, C , WENTZ, K. U., MATTLE, H. P., FINN, J. P. & MCARDLE, C , 1989C. Magnetic

resonance angiography and flow velocity quantification in the portal venous system. American Journal of Roentgenology, 153, 755-760. KELLER, P. J., DRAYER, B. P., FRAHM, E. K., WILLIAMS, K. D., DUMOULIN, C. L. & SOUZA, S. P., 1989. MR Angiography

with two-dimensional acquisition and three-dimensional display. Radiology, 173, 527-532. LAUB, G. A. & KAISER, W. A., 1988. MR angiography with

gradient motion refocusing. Journal of Computer Assisted Tomography, 12, 377-382. References

MAIER, S. E., MEIER, D., BOESIGER, P., MOSER, U. T. & VIELI,

ALFIDI, R. J., MASARYK, T. J., HAACKE, E. M., LENZ, G. W., Ross, J. S., MODIC, M. T., NELSON, A. D., LIPUMA, J. P. &

COHEN, A. M., 1987. MR angiography of peripheral, carotid and coronary arteries. American Journal of Roentgenology, 149, 1097-1109. DUMOULIN, C. L. & HART, H. R., 1986. Magnetic resonance

angiography. Radiology, 161, 717-720. DUMOULIN, C. L., SOUZA, S. P., WALKER, M. F. & YOSHITOME,

F., 1988. Time resolved MR angiography. Resonance in Medicine, 6, 275-286.

16

Magnetic

A., 1989. Human abdominal aorta: comparative measurements of blood flow with MR imaging and multigated Doppler US. Radiology, 171, 487^92. WEDEEN, V. J., ROSEN, B. R., CHESLER, D. & BRADY, T. J.,

1985a. MR velocity imaging by phase display. Journal of Computer Assisted Tomography, 9, 530-536. WEDEEN, V. J., MEULI, R. A., EDELMAN, R. R., GELLER, S. C , FRANK, L. R., BRADY, T. J. & ROSEN, B. R., 1985b.

Projective imaging of pulsatile flow with magnetic resonance. Science, 230, 946-948.

The British Journal of Radiology, January 1991