Arterial flow characteristics in the presence of vascular disease and implications for FSE-based noncontrast MR angiography

Abstract

Purpose

To investigate arterial flow characteristics in the setting of vascular disease, and examine their effect on the performance of fast spin-echo (FSE)-based noncontrast MR angiography (NC-MRA).

Materials and Methods

Seventeen patients were recruited from among those scheduled for routine contrast-enhanced MR angiography (CE-MRA) of the lower extremities at 1.5T. The research portion of the exam was performed prior to the clinically-indicated protocol and included phase-contrast imaging at multiple levels in the legs and FSE-based NC-MRA in the calf and thigh, using a 3D ECG-gated technique that exploits differences in arterial flow velocity between diastole and systole.

Results

Vascular occlusions were associated with reduced systolic velocity, a delayed systolic peak, and, in two middle-aged patients, an increase in diastolic velocity. Elevated systolic and diastolic velocities were observed in a subject with a non-healing ulcer. NC-MRA allowed visualization of arteries with systolic velocities as low as 3 cm/s, and exhibited comparable depiction to CE-MRA for diastolic velocities as high as 6 cm/s. At the highest diastolic velocities observed (15 cm/s) arterial depiction was severely degraded.

Conclusion

FSE-based NC-MRA as presently implemented performs successfully over a wide range of flow patterns, but does not accommodate extremely low systolic velocities or very high diastolic velocities.

INTRODUCTION

Gadolinium-enhanced MR angiography (MRA) has become the technique of choice for noninvasive diagnosis and evaluation of peripheral arterial disease (1). However, concerns about the risk of nephrogenic systemic fibrosis in patients with renal insufficiency (2) have stimulated renewed interest in imaging methods that do not require exogenous contrast agents. One such method, pioneered by Miyazaki et al. (3), exploits the pulsatility of arterial flow to distinguish arteries from veins and background tissue. Two sets of images are acquired using a 3D fast spin-echo (FSE) sequence, one during fast systolic flow, when the arterial signal is suppressed due to flow-induced dephasing, and the other during diastole, when the arterial signal is higher. A bright-blood angiogram can then be obtained by subtracting the two image sets (Figure 1).

Figure 1.

Illustration of the noncontrast MRA technique in an 84-year old man with claudication. Images were acquired using an ECG-gated 3D fast spin-echo (FSE)sequence at peak systole (left) and end-diastole (center), with trigger delays of 190 ms and 0 ms, labeled TD_s and TD_d, respectively. Other parameters included a refocusing pulse FA of 90° and TE of 19 ms. Note that the arteries appear dark during systole and bright during diastole (arrows). Subtracting the systolic images from the diastolic images produces a bright-blood angiogram, of which a maximum intensity projection (MIP) is shown on the right. Plotted below is the flow velocity time course obtained from phase-contrast imaging through the superficial femoral arteries at the level indicated by the dashed line on the MIP, together with the acquisition windows for systolic and diastolic imaging. Note that the systolic velocity is lower in the right femoral artery than the left, probably reflecting the greater severity of disease on the right.

Exquisite arterial depiction has been demonstrated using this approach (4–7). However, it is not yet known how well the technique will perform in situations where the flow pattern is altered due to pathology. It has previously been shown that the flow sensitivity of FSE sequences can be adjusted by varying the flip angle (FA) of the refocusing pulses (8,9); low FA provides greater flow sensitivity, while higher FA produces less flow sensitivity. This offers some means to tailor the technique to individual flow patterns. However, it remains to be shown whether it provides sufficient control to achieve accurate arterial depiction over a wide range of disease states.

The goal of this exploratory study was to investigate arterial flow characteristics in patients with vascular disease and examine their impact on the performance of FSE-based noncontrast MRA (NC-MRA). Phase-contrast imaging was used to document flow velocity in the major arteries of the legs, in particular at points proximal and distal to an occlusion or stenosis. NC-MRA was performed in the calf and thigh and arterial depiction on the resulting images was compared to that obtained with contrast-enhanced MR angiography (CE-MRA).

MATERIALS AND METHODS

Subject cohort

Patients were recruited from among those scheduled for routine CE-MRA of the lower extremities, with or without angiography of the abdomen and pelvis. Seventeen consecutive patients who consented to participate in the study were prospectively included. The subject cohort comprised 9 men and 8 women, aged 22 – 88 years (mean 67.4 years). Reasons for the clinical exam included claudication (N=14), non-healing ulcer (N=1), deep vein thrombosis (N=1) and Leriche syndrome (N=1).

Image Acquisition

Imaging was performed on a 1.5T Siemens Avanto system using a peripheral vascular coil in combination with the spine array in the patient table and one or more body matrix coils as needed over the abdomen and pelvis.

Phase-contrast imaging was used to document flow characteristics and to determine optimal parameters for the NC-MRA acquisition. Images were obtained in an axial plane with retrospective electrocardiographic (ECG) gating, velocity-encoding factor = 50 – 100 cm/s, in-plane voxel size = (1.4 – 1.6) × (1.4 – 1.6) mm², slice thickness = 5 mm, 40 phases, 3 lines per segment, receiver bandwidth (BW) = 391 Hz/pixel, FA= 30°, echo time (TE) = 3.53 – 3.68 ms and repetition time (TR) = 40.70 – 41.60 ms, where TR includes all lines in a segment and represents the true temporal resolution of the acquisition. Measurements were performed at three levels in each station prior to NC-MRA. Following acquisition of the NC-MRA images, the levels were compared to the location of the pathology and further measurements were performed if necessary. Commercial software (Argus, Siemens Medical Solutions) was used to measure flow velocities.

NC-MRA was conducted in the calf and thigh using an ECG-gated 3D FSE sequence with nonselective excitation and refocusing pulses. Frequency-encoding was chosen in the head-foot direction to provide sensitivity to flow along the principal orientation of the vessels. Two datasets were acquired within the same sequence, one during fast systolic flow, with a trigger delay (TD) determined by phase-contrast imaging, and the other at end-diastole, with a TD of 0 ms (Figure 1). The sequence was run in ‘constant FA’ mode, with the FA of the refocusing pulses chosen in the range 60° – 120°. Sixty degrees represents the minimum FA for this sequence and 120° was chosen as the upper limit since higher values have been found to cause ghosting and loss of branch vessels (9). A high FA was chosen if the patient exhibited fast flow, while a low FA was used in cases of slow flow, and the acquisition was repeated with a different FA if time permitted. The TE was chosen following experience in healthy volunteers(9), which suggested that a long TE could help improve suppression of background tissue if the subject’s R-R interval varied with time, due, for example, to respiration. These results from the fact that the effective TR is determined by the R-R interval, so any variation in heart rate will affect the amplitude of the signal, and may compromise the cancellation of tissue signal between the systolic and diastolic acquisitions. In the present study, a long TE (81 – 104 ms) was used in 8 patients, and the minimum TE (19 – 20 ms) in the others. These choices corresponded to full Fourier sampling and 55% Fourier sampling respectively, since the sequence used linear phase encoding. The acquisition window was also determined by the partial Fourier factor and fell within the ranges 160 – 206 ms and 101 – 121 ms respectively. Other parameters included: field of view (FOV)= 400 – 450 mm, base resolution = 320 and phase resolution = 83 – 88%, giving an in-plane voxel size of (1.3 – 1.4) × (1.5 – 1.7) mm², generalized partially parallel acquisition (GRAPPA) with nominal acceleration factor 2 and 24 reference lines, 80 – 112 partitions with nominal slice thickness 1.5 mm (calf) or 2.0 mm (thigh), slice resolution 75%, slice partial Fourier factor 6/8, 2 echo trains per partition, BW =977 Hz/pixel, echo spacing 2.76 – 2.84 ms, restore pulse on, and data acquisition every cardiac cycle. The total acquisition time depended on the subject’s heart rate and the number of partitions acquired, but was approximately 3 – 4 minutes per station.

CE-MRA was performed after the NC-MRA and included time-resolved imaging of the calves followed by a multi-station bolus chase using a stepping-table approach. A total of 0.15 mmol/kg gadopentetate dimeglumine (Magnevist, Bayer Healthcare) was injected, using 0.04 mmol/kg for time-resolved imaging, and 0.11 mmol/kg for the bolus chase.

Dynamic imaging of the calves was performed using Time-resolved angiography With Interleaved Stochastic Trajectories (TWIST) (10). The area of the fully-sampled central region was set at 10% and the sampling density in the periphery was chosen to be 25%. Other parameters included: TE = 1.0 ms, TR = 3.1 ms, FA = 25°, FOV = 450 – 500 mm, base resolution = 512 and phase resolution = 75%, giving an in-plane voxel size of (0.9 – 1.0) × (1.2 – 1.3) mm², phase partial Fourier factor 6/8, nominal GRAPPA acceleration factor 3 with 24 reference lines, 88 partitions with nominal slice thickness 1.3 mm, slice resolution 93%, slice partial Fourier factor 6/8, BW = 440 Hz/pixel and temporal resolution 5.73 – 6.88s.

Bolus chase imaging (11) was performed with a 3D Fast Low-Angle SHot (FLASH) sequence. Acquisition parameters for the thigh station included: TE = 0.95 ms, TR = 2.94 ms, FA = 25°, FOV = 500 mm, base resolution = 512 and phase resolution = 60%, giving an in-plane voxel size of 1.0 X 1.6 mm², phase partial Fourier factor 6/8, nominal GRAPPA acceleration factor 3 with 24 reference lines, 72 – 88 partitions with nominal slice thickness 1.4 – 1.65 mm, slice resolution 60%, slice partial Fourier factor 6/8, BW = 440 Hz/pixel, and acquisition time = 10 – 13s. Parameters for the calf station of the bolus chase were identical to those listed above for dynamic imaging except for a slice resolution of 100% and an acquisition time of 20 – 23s.

Image Assessment

Two board-certified radiologists, each with four years of experience in MRA, independently rated arterial depiction by NC-MRA in comparison to CE-MRA on a 5-point scale: better/same/slightly worse/much worse/not visualized. Separate grades were assigned to each of several predefined arterial segments provided those segments were patent and within the FOV. In the calves they comprised the distal popliteal artery, tibioperoneal trunk, and proximal and distal halves of the anterior tibial, peroneal, and posterior tibial arteries. In the thighs they included the common femoral artery, profunda femoris, proximal and distal halves of the superficial femoral artery, and the proximal popliteal artery. A single average rating was assigned to collaterals for each leg and station in which they were present. Interobserver agreement was evaluated in terms of the linear-weighted Cohen’s Kappa statistic. Source images from the diastolic and systolic acquisitions were not used by the radiologists in their assessment of arterial depiction, but were analyzed retrospectively in cases of poor depiction to determine the cause.

RESULTS

Systolic velocity

Over the femoral and popliteal arteries, 8 of the 17 patients exhibited differences in peak systolic velocity of greater than 30% between left and right legs. In 7 of these patients, all of whom had claudication, the leg with lower peak velocity corresponded to the side with greater disease (see for example Figure 1). Reduced velocity was found both proximal and distal to the pathology in all cases for which such data were available. In the eighth patient, who had a non-healing ulcer, the ipsilateral leg had higher flow velocity than the contralateral leg in all arterial segments evaluated.

No evidence was found in any of the patients that depiction of the femoral or popliteal arteries was compromised due to insufficient systolic flow speed. However, in two patients, depiction of at least one of the more distal arteries in the calf appeared to have been affected by low systolic velocity. One case arose in a patient with complete occlusion of the proximal popliteal artery, while the other (shown in Figure 2) occurred in a patient with complete occlusion of the right common iliac artery. As demonstrated by the graphs, peak flow velocity in this patient is lower in the arteries of the right calf than those of the left. In particular, the systolic velocity is 6.0 ± 0.4 cm/s and 3.4 ± 0.3 cm/s in the right anterior tibial and right peroneal arteries respectively. Although these arteries appear on the contrast-enhanced images to have similar, if not larger, caliber than the corresponding arteries of the left leg, their intensity on the noncontrast angiogram is diminished. In particular, the right peroneal artery (arrow) is only faintly visible.

Figure 2.

Comparison between noncontrast and contrast-enhanced angiography (NC-MRA and CE-MRA respectively) in a 56-year old woman with complete occlusion of the right common iliac artery. Sequence parameters for NC-MRA included: refocusing pulse FA = 90° (calf) and 100° (thigh), systolic TD = 250 ms (calf) and 226 ms (thigh), and TE = 19 ms. The graphs show flow velocity in the posterior tibial, peroneal and anterior tibial arteries at the level indicated by the dashed line on the noncontrast angiogram. Note the much lower peak velocities in the arteries of the right calf compared to those of the left. The right peroneal artery, which has the lowest systolic velocity (3.4 ± 0.3 cm/s), is only faintly visible by NC-MRA (arrow). The lower signal in the right common femoral artery as compared to the left is probably due to B₁⁺ inhomogeneity.

Diastolic velocity

Three patients showed evidence of signal loss on the noncontrast angiograms due to dephasing in diastole. Figure 3 presents an example in a 54-year old man with complete occlusion of the superficial femoral arteries. The left popliteal artery distal to the occlusion (arrow) exhibited a diastolic velocity of 15.2 ± 0.5 cm/s and was visible only as a faint outline on the noncontrast angiogram. Inspection of the source images (inset) revealed signal loss in diastole over most of the lumen. This occurred despite the use of a high FA (120°), which was chosen to minimize flow sensitivity.

Figure 3.

Comparison between NC-MRA and CE-MRA in a 54-year old man with claudication, together with graphs of flow velocity at the levels indicated by the dashed lines. Sequence parameters for NC-MRA included: refocusing pulse FA = 120°, systolic TD = 240 ms (thigh) and 316 ms (calf), and TE = 104 ms. Note the poor depiction of the left popliteal artery by NC-MRA as compared to CE-MRA. Inspection of the noncontrast source images (bottom left) reveals low signal in the left popliteal artery even in diastole, resulting in poor visualization on the difference images (arrows). The signal loss is presumably due to high diastolic velocity in the left popliteal artery, as demonstrated by phase-contrast imaging (see graph above). NC-MRA also provides poor depiction of the profunda femoris in both legs, probably reflecting high diastolic velocity on the left and a combination of moderately fast diastolic velocity and locally reduced B₁⁺ on the right.

Another patient (22-year old man with a non-healing ankle ulcer) exhibited elevated diastolic and systolic flow proximal to the wound. Measurements in the tibioperoneal trunk revealed velocities of 15.6 ± 1.0 cm/s in diastole and 81.6 cm/s at peak systole. Depiction of the popliteal artery and tibioperoneal trunk was degraded even at high FA (120°) due to dephasing in the lumen on the diastolic images.

In a third subject (56-year old man with claudication), the popliteal artery exhibited a diastolic velocity of 6.1 ± 0.4 cm/s distal to an occlusion. It was well depicted by NC-MRA at a FA of 120°, but showed reduced signal for FA = 90°, and was only faintly visible at FA = 70°. By contrast, the popliteal artery on the contralateral side, which had a lower diastolic velocity of 2.2 ± 0.8 cm/s, was well depicted for all FA tested. To quantify these observations, the ratio of arterial intensities on the ipsilateral and contralateral sides was measured, yielding values of 82%, 56% and 36% for the three FA respectively.

Timing of systole

Measured values for the time to peak systole ranged from 94 – 336 ms in the femoral artery and 143 – 368 ms in the distal popliteal artery. The minimum values occurred in a 68 year-old woman with claudication, while the maximum values were recorded respectively in a 72 year-old man with Leriche syndrome and in a 54-year old man distal to an occlusion (see Figure 3). Three patients, all with claudication, exhibited differences of greater than 70 ms in the time of peak systole between left and right legs. In two cases the extent of disease was markedly asymmetric, and the systolic peak was delayed distal to the region most severely affected. In the third case (Figure 4), the disease was bilateral, and a substantial delay (>90 ms) in peak systole was observed between arterial segments proximal and distal to the occlusions. Despite the large timing differences between left and right sides and between proximal and distal segments in this patient, the arterial depiction is adequate throughout. In particular, the right popliteal artery is well visualized even though the TD used for systolic imaging (200 ms) falls well before peak flow in that segment.

Figure 4.

Comparison between NC-MRA and CE-MRA in an 85-year old woman with claudication, together with graphs of flow velocity at the levels indicated by the dashed lines. Sequence parameters for NC-MRA included: refocusing pulse FA = 90°, systolic TD = 200 ms (thigh) and 243 ms (calf), and TE = 19 ms. Note that despite the difference in the timing of peak systole between the common femoral and popliteal arteries and between the two legs, arterial depiction by NC-MRA is adequate throughout. In particular, the right popliteal artery is well visualized even though its flow velocity peaks well after the trigger delay used for systolic imaging (indicated by the vertical line and labeled TD_s). The lower signal in the right common femoral artery as compared to the left is probably due to B₁⁺ inhomogeneity.

Given that the diastolic acquisition, which requires fairly low flow, was performed with a TD of 0 ms, it is also interesting to observe in Figure 4 that the early systolic upslope in the common femoral arteries does not appear to have substantially impacted their depiction. It should be noted that this patient was imaged using a TE of 19 ms, resulting in a relatively short acquisition window (120 ms). In two other patients with early systolic upslope, a TE of 94 – 95 ms was used, corresponding to a longer acquisition window (186 – 188 ms), and arterial depiction was degraded.

Vessel orientation

In four patients, loss of signal was observed on the noncontrast angiograms in the proximal anterior tibial artery where it ran perpendicular to the frequency-encoding direction (see example in Figure 5). This artifact, which may mimic a stenosis or occlusion (6), could be traced in each case to insufficient dephasing on the systolic acquisition.

Figure 5.

Comparison between NC-MRA and CE-MRA in an 88-year old man with claudication, together with a graph of flow velocity at the level indicated by the dashed line. Sequence parameters for NC-MRA included: refocusing pulse FA = 90°, systolic TD = 229 ms, and TE = 19 ms. Note the localized signal loss in the proximal left anterior tibial artery where it runs perpendicular to the frequency-encoding direction. Inspection of the source images reveals incomplete dephasing in this region on the systolic acquisition (inset).

Other factors

Besides flow-related factors, other issues affecting arterial depiction by NC-MRA included subject motion and B₁⁺ inhomogeneity (12). Motion obscured the vessels in 2 patients, while B₁⁺ inhomogeneity caused varying degrees of signal attenuation in the arteries of the upper right thigh in 6 patients (see for example Figures 2 and 4). In addition, one patient had a total hip replacement, which caused greater signal loss on the noncontrast angiogram (TE = 20 ms) than the contrast-enhanced images (TE = 0.95 ms). In two patients, depiction of the proximal popliteal artery in the thigh station was better by NC-MRA than CE-MRA due to late arrival of the contrast bolus.

Assessment of arterial depiction

In the calf station, assessments were made of 243 arterial segments and collaterals in 9 legs. Over those vessels, the pooled ratings by both readers for depiction by NC-MRA in comparison to CE-MRA on a 5-point scale of better/same/slightly worse/much worse/not visualized were 10% / 71% / 12% / 6% / 1% respectively, with κ = 0.47 (moderate inter-observer agreement). In the thigh, ratings were given for 138 arterial segments and collaterals in 9 legs, and the results were 4%/57%/20%/10%/9% respectively, with κ = 0.65 (substantial inter-observer agreement).

DISCUSSION

FSE-based NC-MRA performed successfully over a remarkably wide range of flow patterns; both arteries with high flow and those with low flow could be accurately depicted provided their signals during the systolic and diastolic acquisitions were adequately different. However, alterations in both systolic and diastolic velocities were observed in our patient cohort, and were sufficient, in some cases, to pose challenges for the technique. Stenoses and occlusions were associated with reduced systolic velocity, presumably because of increased resistance to flow. In two middle-aged patients, the diminished pulsatility was accompanied by elevated diastolic velocity, a phenomenon that may depend on the compliance of the proximal vessels. In another young patient, hyperemia due to a non-healing wound increased both diastolic and systolic flow velocities.

A FA of 120° was found to provide good depiction of arteries with a diastolic velocity as high as 6 cm/s, while a FA of 90° enabled visualization of arteries with a systolic velocity as low as 3 cm/s, consistent with observations in phantoms(9,13). Although the FA provides some control over flow sensitivity, our results suggest that it may not be sufficient to accommodate the full range of flow patterns encountered in patients with vascular disease. For example, arteries with a diastolic velocity of 15 cm/s were poorly depicted even at a FA of 120°. This was the maximum FA used in this study, since higher values have been found to cause ghosting and loss of branch vessels and may exceed specific absorption rate (SAR) limits(9).

In the case of large arteries, the source images may provide a fallback if depiction on the subtracted images is compromised due to vessel orientation or high diastolic velocity. However, small arteries, whose depiction is more likely to be affected by low systolic flow, can be difficult to identify on the source images.

A delayed systolic peak, characteristic of the so-called ‘tardus-parvus’ waveform (14), was observed distal to occlusions in some patients. This gave rise to substantial timing differences between proximal and distal arterial segments and between left and right legs. The technique was found to be remarkably robust to such variations; provided a TD could be chosen that produced signal suppression during systole in all arterial segments, good depiction could be achieved throughout. It was found in cases of early systolic upslope that a short TE (i.e. short acquisition window) was required to avoid dephasing during the diastolic acquisition, which was performed with a TD of 0 ms. The time of the systolic upslope differed among patients, presumably due to variations in vessel compliance, and occurred earlier in the thigh than the calf. An alternative approach might be to use a long TD, so that the diastolic acquisition occurs after the systolic peak (7). However, this can cause mistriggering if the heart rate varies.

Depiction by NC-MRA was rated as good as or better than CE-MRA in 81% of arterial segments in the calf, and 61% of vessels in the thigh. The poorer performance in the thigh is due in part to the earlier systolic upslope, and could be mitigated by minimizing the TE. Besides flow-related issues, which were the focus of this work, another problem encountered with the present implementation of the technique was sensitivity to subject motion (7). Since the arterial signal during diastole is not substantially higher than the intensity of background tissue, any motion between the systolic and diastolic acquisitions can obscure the vessels.

Alternative noncontrast approaches to peripheral MRA overcome some of these problems. The flow-sensitive dephasing (FSD)-prepared balanced steady-state free precession (SSFP) technique of Fan and Li (15) avoids the issue of high diastolic flow by applying the flow-sensitive preparation only during the systolic acquisition (16). One potential drawback of that technique, however, is that the SSFP readout is sensitive to static field (B₀) inhomogeneity, which limits the usable FOV. A completely different approach is the 2D inflow-based quiescent interval steady state (QISS) technique recently proposed by Edelman et al. (17,18). That method is independent of diastolic flow, robust against motion, and easy to set up since it does not require individual adjustment of sequence parameters. However, its performance may be compromised in arteries with retrograde flow.

The present work was intended as an exploratory study, rather than a validation of the technique, and included only a small number of patients. Other limitations included the fact that we were not able to compare different FA in each patient due to time constraints in the clinical setting.

In conclusion, FSE-based NC-MRA as presently implemented is remarkably robust to variations in flow pattern, but can fail in situations of extremely low systolic flow or very high diastolic flow.