Introduction	Choose Top of pageABSTRACTIntroduction <<Materials and MethodsResultsDiscussionReferencesCITING ARTICLES

Gadolinium-enhanced three-dimensional (3D) MR angiography has become a preferred means of accurate noninvasive rapid evaluation of vascular disease [1,2,3]. Advantages over conventional contrast angiography and CT angiography include avoidance of the potential risks of arterial catheterization or iodinated contrast administration [4, 5]. Vascular conspicuity in gadolinium-enhanced 3D MR angiography is primarily based on the hyperintensity of the lumen on the rapidly acquired T1-weighted image caused by T1 shortening of the blood itself. This T1 shortening makes the technique unaffected by in-plane saturation and less subjected to signal loss caused by dephasing, both of which are common limitations in other MR angiography techniques [3, 6].

However, we have encountered cases of pseudostenosis at or near vessel origins on gadolinium-enhanced 3D MR angiography, which has not been previously described in the radiology literature. The pseudostenosis was also reproduced in a pulsatile flow phantom experiment. Examples of this observation are presented and possible causes discussed.

Materials and Methods	Choose Top of pageABSTRACTIntroductionMaterials and Methods <<ResultsDiscussionReferencesCITING ARTICLES

We performed a database search of 883 gadolinium-enhanced chest and abdominal MR imaging examinations from June 1997 to March 1998. Gadolinium-enhanced 3D MR angiographic images were obtained from 314 examinations.

All examinations were performed on Signa 1.5-T MR scanners (General Electric Medical Systems, Milwaukee, WI), software version 5.5 or 5.6, at two gradient strengths. One scanner had a high-performance gradient system (25 mT/m) and two had a standard gradient system (10 mT/m). A torso phased array coil was used in all cases except a few for which the built-in body coil was used. On the high-performance gradient system, the gadolinium-enhanced 3D MR angiography sequence parameters were enhanced fast gradient-echo 3D; TR range/TE range, 7-9/2-3 msec; flip angle, 50°; matrix, 256 × 160-192 (frequency × phase); half-Fourier acquisition (number of excitations, 0.5); receiver bandwidth, ±31.2 kHz; and fat suppression. The standard gradient systems had the following parameters: 3D fast spoiled gradient-echo; TR/TE, 13/2.3; flip angle, 60°; matrix, 256 × 160-192 (frequency × phase) or 512 × 160-192; number of excitations, one; bandwidth, ±32 kHz; and no fat or spatial saturation pulses. The field of view and the number of partitions were adjusted to accommodate each patient, yielding effective slice thickness of 1.5-3 mm and in-plane resolution of 0.94-2.37 mm. Rectangular field of view (fractional phase-encoding) of 75% was used for axial sequences, 50% for sagittal sequences, and 100% for coronal sequences. Frequency-encoding directions were superior-inferior for sagittal and coronal sequences and right-left for the axial sequence.

All examinations included an unenhanced 3D sequence and at least two successive sequences during a dynamic injection. The acquisition time for each 3D sequence was 18-26 sec (with breath-holding) for the high-performance gradient and 84-107 sec (without breath-holding) for the standard gradient systems. Injection of gadopentetate dimeglumine (Magnevist; Berlex Laboratories, Wayne, NJ) at doses of 0.1-0.2 mmol/kg and a rate of 2 mL/sec for the high-performance gradient, and 0.3 mmol/kg and rate of 0.8-2.0 mL/sec for the standard gradient system, was performed manually through a 20- to 22-gauge peripheral IV catheter. A spoiled gradient-echo sagittal test bolus timing sequence (6.2/1.5; flip angle, 90°; matrix, 256 × 128; field of view, 30 cm; number of excitations, one; slice thickness, 20 mm) was used to set the injection time so that the peak enhancement occurred during the central point of K-space. The 3D sequence on the standard gradient system used sequential ordering of K-space whereby the weakest phase-encoding steps (center of K-space) were acquired in the middle of the sequence. The 3D sequence on the high-performance gradient system also used sequential ordering, but with half-Fourier acquisition (number of excitations, 0.5). The weak phase-encoding gradients were applied near the beginning of the acquisition, rendering it closer to centric ordering. Additional two-dimensional time-of-flight MR angiography (25/1.5; flip angle, 60°; slice thickness, 2-2.5 mm) and 3D phase-contrast MR angiography (TR/TE range, 34/7-9; flip angle, 20°; encoding velocity, 30 cm/sec) sequences were obtained in several patients.

Postprocessing of 3D images was performed at an Advantage Windows station, software version 1.2 (General Electric Medical Systems) by an experienced radiologist using maximum intensity projection and multiprojection volume reconstruction. These images and the source images were reviewed on a digital workstation (Canon Medical, Los Angeles, CA). The major aortic branches were reviewed for the presence of stenosis. The degree of stenosis was categorized subjectively as normal, mild, moderate, severe, or occluded.

In vitro studies used a pulsatile flow phantom with a 2-cm main tube diameter and an 8-mm branch diameter with an orifice angulation of 110°. The total volume of circulating fluid in the phantom was 5 L. The phantom and surrounding reservoir contained a non-Newtonian polymeric solution of carboxymethylcellulose with T1 of 1415, T2 of 201 msec, and viscosity of 0.036 poise (poise = 0.1 kg·m^-1·sec^-1), values similar to those of blood. Pump settings included stroke volume of 60 mL; peak velocity, 100 cm/sec; rate, 70 beats per min; and systolic/diastolic flow ratio, 2:1.

Administration of gadolinium was performed using two different rates of manual injection: a rapid bolus injection and a slow injection. The experiment was performed once for each rate of injection using 20 mL of 10% gadopentetate dimeglumine solution. Gadolinium was administered into the phantom approximately 1 m from the magnet bore of the high-performance gradient system after a sagittal test bolus timing sequence identical to that previously described for patients. The enhanced fast gradient 3D sequences were obtained in a coronal plane, including the main tube and branch of the phantom, in parameters similar to those of the clinical cases, using 7/2.3; flip angle, 50°; matrix, 256 × 256 (frequency × phase); half-Fourier acquisition (number of excitations, 0.5); bandwidth, ±31.2 kHz; and fat suppression. Frequency encoding was superior-inferior. Acquisition time was 15 sec. For the rapid injection experiment, the bolus was injected at the rate of 2 mL/sec so that the gadolinium concentration varied during acquisition. The 3D sequence was first obtained before injection followed by three successive 3D sequences beginning 4 sec after the initiation of injection. Then the fluid was flushed from the system and replaced by unenhanced fluid. This was followed by the slow-injection experiment using a slow rate of injection (<1 mL/sec) to produce an approximately steady concentration of gadolinium throughout image acquisition. Three successive 3D sequences were acquired beginning approximately 10 sec after the onset of slow injection. Images were reviewed and reconstructed in the same manner as in patients.

Results	Choose Top of pageABSTRACTIntroductionMaterials and MethodsResults <<DiscussionReferencesCITING ARTICLES

Of all cases that had an apparent stenosis on the first-pass 3D gadolinium-enhanced MR angiography, a pseudostenosis was noted in six patients (approximately 2% of 314 contrast-enhanced 3D MR angiography examinations). Findings of no stenosis, or a lesser degree of stenosis, on the second-pass gadolinium-enhanced 3D MR angiography were seen in all cases. In two cases, the 3D phase-contrast MR angiography showed normal vessel caliber. One patient underwent catheter angiography to confirm the lack of stenosis (Fig. 1A,1B,1C,1D). The pseudostenoses were noted at the origin of a major branch of the thoracic or abdominal aorta and were seen on both source and reconstructed images (Fig. 2A,2B,2C,2D). The observation was made in three cases on the high-performance gradient and three cases on the standard gradient systems.

View larger version (115K)

Fig. 1A. —56-year-old woman examined for possible renovascular hypertension. Coronal source MR image of first-pass three-dimensional (3D) gadolinium-enhanced MR angiogram (fast spoiled gradient-echo; TR/TE, 13/2.3 msec; flip angle, 60°; slice thickness, 1.5 mm; acquisition time, 107 sec) shows moderate stenosis of proximal left renal artery (arrow).

View larger version (134K)

Fig. 1B. —56-year-old woman examined for possible renovascular hypertension. Coronal source MR image of second-pass acquisition shows only minimal narrowing of left renal artery (arrowhead).

View larger version (81K)

Fig. 1C. —56-year-old woman examined for possible renovascular hypertension. Coronal maximum-intensity-projection image of 3D phase-contrast MR angiography sequence (34/8.6; encoding velocity, 30 cm/sec) shows minimal narrowing of left renal artery (straight arrow). Accessory left renal artery is shown (curved arrow).

View larger version (145K)

Fig. 1D. —56-year-old woman examined for possible renovascular hypertension. Slightly oblique anterior projection of iodinated contrast angiography reveals minimal narrowing of proximal left renal artery (arrowhead) and accessory left renal artery (arrow).

View larger version (114K)

Fig. 2A. —58-year-old man with history of coronary artery disease and chest pain. Sagittal source image (A) and maximum-intensity-projection image (B) of first-pass three-dimensional (3D) gadolinium-enhanced MR angiography (fast spoiled gradient-echo; TR/TE, 13/2.3 msec; flip angle, 60°; slice thickness, 1.5 mm; acquisition time, 107 sec) show moderate stenosis of origin of left common carotid artery (arrow).

View larger version (117K)

Fig. 2B. —58-year-old man with history of coronary artery disease and chest pain. Sagittal source image (A) and maximum-intensity-projection image (B) of first-pass three-dimensional (3D) gadolinium-enhanced MR angiography (fast spoiled gradient-echo; TR/TE, 13/2.3 msec; flip angle, 60°; slice thickness, 1.5 mm; acquisition time, 107 sec) show moderate stenosis of origin of left common carotid artery (arrow).

View larger version (119K)

Fig. 2C. —58-year-old man with history of coronary artery disease and chest pain. Sagittal source image (C) and maximum-intensity-projection image (D) of second-pass 3D gadolinium-enhanced MR angiogram show normal caliber of origin of left common carotid artery (arrowhead).

View larger version (111K)

Fig. 2D. —58-year-old man with history of coronary artery disease and chest pain. Sagittal source image (C) and maximum-intensity-projection image (D) of second-pass 3D gadolinium-enhanced MR angiogram show normal caliber of origin of left common carotid artery (arrowhead).

This artifact was replicated in vitro (Fig. 3A,3B,3C,3D). A pseudostenosis was prominent during the first pass of rapid-injection experiment (Fig. 3A), was decreased during the second pass (Fig. 3B), and was absent during both the third pass (Fig. 3C) and a subsequent slow injection (Fig. 3D).

View larger version (188K)

Fig. 3A. —Pulsatile phantom on coronal gadolinium-enhanced fast gradient-echo three-dimensional MR angiography (TR/TE, 7/2.3 msec; flip angle, 50°; matrix, 256 × 256; number of excitations, 0.5; fat suppression; acquisition time, 15 sec). MR image acquired 4 sec after rapid gadolinium bolus initiation, causing gadolinium concentration to vary during acquisition, shows pseudostenosis (arrow).

View larger version (192K)

Fig. 3B. —Pulsatile phantom on coronal gadolinium-enhanced fast gradient-echo three-dimensional MR angiography (TR/TE, 7/2.3 msec; flip angle, 50°; matrix, 256 × 256; number of excitations, 0.5; fat suppression; acquisition time, 15 sec). MR image acquired approximately 25 sec after rapid bolus initiation shows pseudostenosis (arrow) to be less severe than in A.

View larger version (201K)

Fig. 3C. —Pulsatile phantom on coronal gadolinium-enhanced fast gradient-echo three-dimensional MR angiography (TR/TE, 7/2.3 msec; flip angle, 50°; matrix, 256 × 256; number of excitations, 0.5; fat suppression; acquisition time, 15 sec). MR image acquired approximately 45 sec after rapid bolus initiation shows no pseudostenosis (arrow).

View larger version (183K)

Fig. 3D. —Pulsatile phantom on coronal gadolinium-enhanced fast gradient-echo three-dimensional MR angiography (TR/TE, 7/2.3 msec; flip angle, 50°; matrix, 256 × 256; number of excitations, 0.5; fat suppression; acquisition time, 15 sec). MR image obtained during slow injection, causing steady concentration of inflowing gadolinium, shows no pseudostenosis (arrow).

Discussion	Choose Top of pageABSTRACTIntroductionMaterials and MethodsResultsDiscussion <<ReferencesCITING ARTICLES

We have observed apparent stenoses of normal aortic branches on first-pass gadolinium-enhanced 3D MR angiography. The second pass and other MR angiography techniques showed normal vessel caliber. This pseudostenosis phenomenon was seen in only a small proportion (approximately 2%) of our patients and was reproducible in a phantom. The frequency of this pseudostenosis in the in vitro setting could not be determined because we performed the experiment only once for each rate of injection. However, the pseudostenosis was most severe with the rapid-injection technique, particularly during the early enhanced sequences, and was not seen during the slow infusion.

The explanation of this first-pass pseudostenosis phenomenon has not been established. We speculate that the apparent reduction in the lumen size seen only on the first contrast-enhanced sequence is a result of the rapidly changing T1 of blood during sampling of K-space soon after injection. The second-pass sequence has less variability of gadolinium concentration during its acquisition. The phantom experiment replicates the pseudostenosis seen during the early gadoliniun-enhanced period and during the rapid bolus injection, which had rapidly changing gadolinium concentration. However, the experiment does not explain why this phenomenon is most conspicuous just beyond the takeoff of the aortic branches, particularly at the branches that are approximately at a right angle to the main lumen. Variable enhancement during acquisition of the periphery of K-space, which determines edge resolution, might be necessary for this artifact to occur.

Alternative explanations include incomplete mixing at the branch origin as a result of complex flow. The degree of spin dephasing from turbulent flow is not likely to produce a loss of signal as a result of the nature of the short TE of the 3D gradient-echo technique. Additionally, this artifact was not seen on the second-pass images, which should be equally susceptible to spin dephasing. Furthermore, the pseudostenosis was absent on 3D phase-contrast MR angiography, which was far more susceptible to spin dephasing [7].

The edge-ringing or vessel-broadening artifacts described in the in vitro study by Maki et al. [8] are attributed to the filling of the center of the K-space during the rapid rising of contrast concentration. These artifacts may lead to over-estimation of vessel caliber, in contrast to the first-pass pseudostenosis artifact described in our study. The effect of phase- and frequency-encoding directions relative to the vessel orientation does not account for the fact that we observed this pseudostenosis in all situations.

A few limitations of this study should be noted. Because of the retrospective nature of our study, we were unable to determine the exact rate and the timing of injection for each case, which may have contributed to the variability of rate of change and onset of T1 shortening effect relative to the central point of K-space. The power injectors were not available during the study period. Despite the routine use of the sagittal test bolus timing sequence, individual differences in injection methods may have occurred.

Although only one patient underwent catheter angiography as a gold standard, we believe that the routinely performed second-pass sequences are valuable. Both first-pass and second-pass 3D sequences should be obtained. Although during the first pass the vessel-to-background contrast is superior (avoidance of venous signal and, consequently, superior reconstructed MR angiogram), vessel origins are usually visible on the second pass. Therefore, we suggest that second-pass images should be reviewed when the diagnosis of a significant stenosis is suspected because potential inconsistency in the timing of bolus injection does occur in routine clinical practice. It is possible that using other methods of K-space ordering or an improved consistency and precision in the bolus injection technique may reduce the frequency of this first-pass pseudostenosis artifact [9].

Our observation of a pseudostenosis on gadolinium-enhanced 3D MR angiography should alert the radiologist to a potential error in diagnosing significant stenoses. Analysis of other secondary findings, such as renal morphology and enhancement pattern, and of other MR angiographic techniques, particularly 3D phase-contrast MR angiography, is also helpful [7]. To increase the specificity and accuracy of this technique, interpretation should include a careful review of other images, particularly the second-pass images from the same injection.