SEARCH

SEARCH BY CITATION

Keywords:

  • non-contrast-enhanced angiography;
  • inversion recovery;
  • magnetic resonance angiography;
  • parallel imaging;
  • spin-labeling

Abstract

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. CONCLUSION
  7. Acknowledgements
  8. REFERENCES

Inversion-prepared pulse sequences can be used for noncontrast MR angiography (MRA) but suffer from long scan times when acquired using conventional nonaccelerated techniques. This work proposes a subtraction-based spin-labeling, three-dimensional fast inversion recovery MRA (FIR-MRA) method for imaging the intracranial arteries. FIR-MRA uses alternating cycles of nonselective and slab-selective inversions, leading to dark-blood and bright-blood images, respectively. The signal difference between these images eliminates static background tissue and generates the angiogram. To reduce scan time, segmented fast gradient recalled echo readout and parallel imaging are applied. The inversion recovery with embedded self-calibration method used allows for parallel acceleration at factors of 2 and above. An off-resonance selective inversion provides effective venous suppression, with no detriment to the depiction of arteries. FIR-MRA was compared against conventional three-dimensional time-of-flight angiography at 3 T in eight normal subjects. Results showed that FIR-MRA had superior vessel conspicuity in the distal vessels (P < 0.05), and equal or better vessel continuity and venous suppression. However, FIR-MRA had inferior vessel sharpness (P < 0.05) in four of nine vessel groups. The clinical utility of FIR-MRA was demonstrated in three MRA patients. Magn Reson Med, 2010. © 2010 Wiley-Liss, Inc.

As has been recently reviewed (1), non-contrast-enhanced MR angiography (MRA) methods have had a long history of development and have been applied in virtually every vascular system in the body. The discovery in the last several years of an association between gadolinium-based contrast agents and nephrogenic systemic fibrosis in patients with renal impairment (2, 3) has prompted renewed interest in non-contrast-enhanced MRA (CE-MRA) methods. One class of non-CE-MRA utilizes flow-related enhancement of blood into a region previously saturated by selective excitation. This class includes spin-labeling or tagging methods (4, 5) and time-of-flight (TOF) methods (6, 7). These methods are suited for imaging the relatively fast flow of the intracranial arteries, specifically for detection and diagnostic evaluation of aneurysms, arteriovenous malformations (AVM), and vascular stenosis. The pulse sequence most frequently utilized for intracranial MRA at our institution is three-dimensional (3D) TOF (7) acquired with flow compensation (8) and the multiple overlapping thin-slab acquisition technique (9). Because the TOF sequence is subject to signal saturation, multiple overlapping thin-slab acquisition can provide improved vessel conspicuity over a single-slab acquisition of equal thickness. However, the spatial saturation pulses that are applied for venous suppression do not necessarily discriminate between veins and arteries, which can result in the appearance of discontinuous vessels at slab interfaces.

The goal of this work was to develop a 3D non-CE-MRA method for imaging the intracranial arteries, with high spatial resolution and image quality. Techniques for generating vascular contrast using inversion recovery have been described in the literature (4, 5, 10, 11). One permutation of such techniques (4) uses nonselective and selective inversion cycles to generate respectively control (dark-blood) and labeled (bright-blood) images, similar to that described in perfusion imaging with spin labeling (12–15). Rather than acquiring two separate scans (4), both datasets can be acquired in an alternating fashion as a single scan (13–15) to reduce misregistration between both datasets. The complex difference between both datasets results in the elimination of static tissue and the generation of vascular contrast from unsaturated blood flowing into the selective inversion slab. Such spin-labeling MRA can provide high vessel-to-background contrast and high vessel conspicuity, as seen in 2DFT carotid and intracranial imaging with cine (16, 17). Sampling at high spatial resolution and 3DFT imaging would also provide superior depiction of vessel morphology (18), similar to that seen in 3D TOF. However, 3D spin-labeling MRA can potentially have very long scan times (>15 min) due to the long inversion intervals and the need for two inversion cycles for data collection.

In this work, we propose the fast inversion recovery (FIR-MRA) technique for high-resolution, 3DFT, non-CE-MRA. FIR-MRA uses nonselective and selective inversion pulses, along with an appropriate inversion time (TI) for interleaved acquisition of dark-blood and bright-blood T1-weighted images in a single scan. Subtraction of the two images yields an MR angiogram. Use of multiple repetitions of a fast gradient recalled echo (GRE) sequence for readout (19) provides an initial speedup. Parallel imaging (20–22) at factors of 2 and above is applied. To exploit the intrinsic delay intervals within the inversion-prepared sequence for calibration acquisition, the inversion recovery with embedded self-calibration (IRES) method (23) is applied, which provides effectively increased acceleration over that of standard self-calibration. Additionally, an off-resonance selective inversion provides effective venous suppression. The principal parameters of FIR-MRA were investigated using simulations and in vivo imaging. The image quality of 3D FIR-MRA was compared to 3D TOF in eight normal subjects, using qualitative evaluation criteria. FIR-MRA was further assessed in three patient studies.

MATERIALS AND METHODS

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. CONCLUSION
  7. Acknowledgements
  8. REFERENCES

Basic Principles of FIR-MRA Technique

Similar to the flow-sensitive alternating inversion recovery (14, 15) perfusion imaging method applied in angiography (16, 17), 3D FIR-MRA is a noncontrast technique that employs two different cycles of inversion preparation. As shown schematically in Fig. 1a, during the first inversion cycle a nonselective inversion is applied, whereby the TI is chosen to null the signal of the inverted blood. In the second inversion cycle, a selective inversion is applied to the targeted axial imaging slab. Blood proximal to the inversion slab remains fully or highly magnetized and flows into the slab during a prescribed in-flow interval, set to equal TI of the first cycle. Every cycle pair acquires the same points in k-space. A complex difference between signals acquired from the two inversion cycles eliminates static tissue in the imaging slab and generates an angiogram due to the inflowing, magnetized blood.

thumbnail image

Figure 1. The FIR-MRA technique. a: Illustration of angiogram generated by taking the difference between signals from the nonselective and slab-selective inversions. The directions of blood flow are indicated by the arrows. The selective inversion slab is extended by a fraction ε in the superior (cephalad) direction for suppression of venous signal. b: Illustration of the FIR-MRA pulse sequence, whereby consecutive inversion pulses have an interval of TC. Simulated magnetizations of fully magnetized arterial blood (Arteries1), arterial blood subjected to previous nonselective inversion pulses (Arteries2), and intraslab venous blood and white matter tissue are shown. Acquisition takes place within the shaded region of the plots.

Download figure to PowerPoint

As shown in the pulse sequence diagram of Fig. 1b, in each cycle, TI defines the time interval from the inversion pulse to initiation of the readout, as performed with a train of nR spoiled GRE pulses of flip angle α selective for the targeted imaging slab. The delay time from the end of the readout train to the next inversion pulse allows for magnetization recovery. The cycle time (TC) defines the overall interval between consecutive inversion pulses. Sampling of ky-kz-space by the readout train is done using an interleaved elliptical-centric phase-encoding order (24). The simulated longitudinal magnetizations of blood (arteries and veins) and tissue are also shown in Fig. 1b. For the arteries, two possible scenarios are depicted. In the first scenario, the arterial blood prior to reaching the imaging slab is assumed to be uninverted prior to reaching the imaging slab and thus has full magnetization at the beginning of each TC interval. In the second scenario, arterial blood is assumed to be slower and has experienced one or more previous nonselective inversion pulses, resulting in high but not full magnetization at the beginning of each TC interval. TI is chosen to simultaneously null arterial blood during the nonselective cycle and to provide sufficient time for inflow of fully or highly magnetized blood during the selective cycle.

In generating the plots of Fig. 1b, it was assumed that venous blood and brain tissue remained within the imaging slab, resulting in identical magnetization behavior for both cycles. On the other hand, inflowing arterial blood is fully or highly magnetized in the selective case and is inverted for the nonselective case. During the subsequent readout, the multiple α radiofrequency pulses cause a transient behavior of the measured arterial signal, particularly for the selective case.

Characterization of Principal Parameters

A characteristic of the magnetization-prepared FIR-MRA method presented here is that the measured signal levels within a cycle vary from one repetition of the segmented readout to the next. This causes signal modulation across the phase-encode directions of k-space. There are two primary effects that lead to the signal modulation of the arterial blood. The first effect is T1 recovery from inverted blood, seen mainly during the first (black-blood) cycle, which results only in mild signal modulation. The second effect is signal saturation due to the application of multiple α pulses within a readout cycle. To study these effects, the concept of an arrival time similar to that described in Warmuth et al. (16) is introduced here. The arrival time, t, is defined as the time taken for blood just proximal to the selective inversion slab to arrive at an imaged pixel. The t of a pixel increases with the intraslab distance blood must travel from the proximal slab boundary to that pixel. All imaged pixels experience the first signal modulation effect. In addition, pixels at increasingly positive t values experience progressively greater signal modulation (second effect) because the blood has been subjected to an increasing number of α pulses. The pixels with t > TI will have negligible difference signal because blood will not have arrived at such pixels prior to the start of readout.

Based on early feasibility studies, four parameters of importance to arterial signal modulation and vessel conspicuity were (I) TI; (II) GRE flip angle, α; (III) the number of repetitions per inversion cycle, nR; and (IV) the parallel acceleration factor, R. These parameters were evaluated with simulations and in vivo experiments. Simulations were performed using standard Bloch equations, which assumed T1 relaxation times at 3.0 T of 1600 msec and 850 msec for blood and brain tissue, respectively. TI values of 600 msec, 750 msec, and 900 msec were deemed suitable for visualizing the intracranial vasculature. α Values of 10°, 15°, and 20° were chosen to determine the effects of signal saturation due to the α pulses. nR Values between 30 and 60 were tested to understand the effects of increased signal modulation with increased nR. Since increasing nR also reduces scan time, a constant scan time analysis of R was performed to understand the tradeoff between parallel acceleration (R) and fast GRE (nR). R was varied between 1 and 3 such that nR ·R was constant.

Venous Suppression With Selective Slab Extension

For venous suppression, the slab-selective inversion was extended in the superior direction for in vivo imaging, as depicted in Fig. 1a. The objective was to extend minimally beyond the top of the brain. This required a shift in the center frequency of the selective inversion pulse and an appropriate decrease in magnitude of the slab-select gradient. A fifth parameter, the (V) fractional superior extension of the inversion slab for venous suppression, ε, was studied with in vivo experiments. When expressed as a fraction of the section thickness of the imaging slab, ε was varied from 0 (no extension) to 1.0 (doubled thickness). In multislab acquisitions stacked in the superior-inferior direction, ε was defined with respect to the most superior imaging slab. The most superior extent of that slab was then used as the most superior extent for all other imaging slabs.

Comparing Standard Self-Calibrated Parallel Imaging With IRES

For parallel acquisition, self-calibrated methods (22, 26) may provide certain advantages over separate calibration scans that acquire k-space center samples in an acquisition separate from the accelerated acquisition (21). However, the net acceleration of self-calibration is reduced from nominal because calibration samples at the k-space center are incorporated into the acquisition. In addition, this can cause increased signal modulation in FIR-MRA because the k-space center is more densely sampled compared to the k-space periphery. The use of separate calibration avoids added signal modulation but requires a separate calibration scan. We used the IRES technique (23), which performs calibration acquisitions during the delay time interval used for magnetization recovery. Because the same degree of undersampling is used for the entirety of k-space, IRES preserves the nominal acceleration of the undersampling factor and avoids the extra modulation caused by self-calibration.

FIR-MRA acquired with IRES was compared to that acquired with standard self-calibration using both signal simulations and in vivo imaging. Acceleration factors of R = 2 to 4 were tested. Like the α pulses, the IRES calibration acquisitions also employed GRE but with a flip angle β that was typically smaller than α. The number of β pulses per cycle in this work was typically 14, which was deemed sufficient for accurate calibration while still being allowable within the delay time intervals used.

In Vivo Imaging of Normal Subjects

Eight normal subjects (ages 25 to 46, five male) were recruited for comparing the new FIR-MRA to a TOF MRA sequence optimized for routine intracranial imaging at our institution, using a protocol approved by our institutional research review board. Studies were performed on a 3.0-T whole-body scanner (GE Healthcare; Signa 14× software) using an eight-element coil of 75-cm circumference (In vivo Corp, Orlando FL, USA). Parameters are shown in Table 1. For TOF, spatial saturation bands 4 cm thick were also applied superior to each slab for venous suppression. For FIR-MRA, an adiabatic inversion pulse (pulse width of 16 msec, bandwidth = 2 kHz) (27) was used.

Table 1. Pulse Sequence Parameters of 3D TOF and 3D FIR-MRA
 3D TOF3D FIR-MRA
TC/TI/TR/TE (msec)−/−/38/3.91600/750/9.5/3.2
α/β (°)25/−15/5
Acquisition bandwidth (kHz)±15.63±15.63
Sampling matrix (X × Y × Z)384 × 210 × 32288 × 240 × 32
Image display matrix512 × 512 × 144512 × 512 × 144
Sampling resolution (mm3)0.5 × 0.8 × 1.40.6 × 0.7 × 1.4
Image display resolution (mm3)0.4 × 0.4 × 0.70.4 × 0.4 × 0.7
Partial echo0.750.75
Number of slab overlap slices66
Slab thickness (cm)4.54.5
Number of slabs33
Total slab thickness (cm)1010
Axial field of view (cm)1818
Parallel acceleration factor2, in Y (R/L)2, in Y (R/L)
Phase-encoding orderLinearElliptical centric
Repetitions per cycle, nR-60
Use of flow compensation gradientsYes, in X and ZNo
Total scan time (minutes)6.47.7

For both examinations, axial excitation slabs were used. The frequency encoding was anterior-posterior and phase-encoding was right-left. An R = 2, one-dimensional acceleration was applied along the phase-encoding direction. For TOF, reconstruction for parallel imaging was performed using a commercial, image space-based method (ASSET; GE Healthcare; Signa v14); for 3D FIR-MRA, the IRES (23) method was implemented and images were reconstructed offline using MatLab (R2008a; The MathWorks Inc., Natick, MA) with k-space-based GRAPPA (22) in the x-kY-kz hybrid space.

The first two normal subjects were also imaged more extensively for characterizing the parameters of the FIR-MRA sequence. The parameters were varied one at a time, keeping the remaining parameters constant as reflected in Table 1.

Patient Studies

In addition to the eight normal subjects, the clinical feasibility of FIR-MRA was evaluated in studies of three patients who were referred for MRA. In the first patient study (67-year-old female), a saccular aneurysm at the right middle cerebral artery trifurcation was detected with x-ray CT angiography. The patient underwent MRA with 3D TOF, 3D FIR-MRA, and a higher-resolution 3D FIR-MRA (0.7 mm slice thickness). In the second study, a patient (76-year-old female) with an aneurysm of the right internal carotid artery was followed up with MRA 1 year after the endovascular coil procedure to evaluate any remnants of the aneurysm. The patient underwent 3D TOF, 3D FIR-MRA, and CE-MRA acquired after a timing bolus. In the third and final study (29-year-old male), the filling and draining patterns of a left temporal lobe AVM were evaluated with 3D TOF, 3D CE-MRA, and 3D FIR-MRA. In the second and third patients, x-ray digital subtraction angiography images acquired prior to the MRA were available.

Evaluations

The FIR-MRA versus TOF comparison was evaluated as follows. For each subject, two neuroradiologists well experienced in MRA were presented with full-volume, axial maximum intensity projections (MIPs) and source images from both examinations suitable for interactive multiplanar review using a workstation (Advantage Windows; GE Healthcare; version 4.4). FIR-MRA results were compared with TOF for each subject, using three criteria (vessel conspicuity, continuity, and sharpness) and evaluated using a five-point scale in each of nine vessel groups (Table 2,a). These vessel groups were the internal carotid arteries, the middle cerebral arteries M1 to M3 segments, middle cerebral artery M4 segments and beyond, the anterior cerebral arteries A1 to A3 segments, the anterior frontal branches, the vertebral-basilar arteries, the posterior cerebral arteries P1 to P3 segments, posterior cerebral artery P4 segments and beyond, and the cerebellar arteries. Additionally, FIR-MRA and TOF examinations were evaluated individually using four criteria (venous signal, motion artifacts, aliasing artifacts, and boundary artifacts), defined on a four-point scale (Table 2, b). The nonparametric Wilcoxon signed rank method (28) was used to test for a significant difference in image quality between TOF and FIR-MRA for each criterion, where P < 0.05 was taken to be statistically significant.

Table 2. •••
a. Scheme of Relative Evaluation Criteria for Comparison Between TOF and FIR-MRA
Evaluation criteria
  • a

    Vessel groups: 1. internal carotid arteries; 2. middle cerebral arteries (MCA) M1 to M3 segments; 3. MCA M4 segments and beyond; 4. anterior cerebral arteries A1 to A3 segments; 5. anterior frontal branches; 6. vertebral-basilar arteries; 7. posterior cerebral arteries (PCA) P1 to P3 segments; 8. PCA P4 segments and beyond; 9. cerebellar arteries.

 CriterionDescription
A.Vessel conspicuityPresence and extent of presence of vessels
B.Vessel continuityThe degree of continuous vessel presence through slabs
C.Vessel sharpnessThe crispness of edges of vessels
Evaluation scale and vessel groups
 ScoreInterpretationVessel groups
 −2TOF is markedly better than FIR-MRANine vessel groupsa
 −1TOF is slightly better than FIR-MRA
 0Negligible differences
 +1FIR-MRA is slightly better than TOF
 +2FIR-MRA is markedly better than TOF
b. Scheme of Absolute Evaluation Criteria for Comparison Between TOF and FIR-MRA
Evaluation criteria
 CriterionDescription 
D.Venous signalPresence of venous signal 
E.Motion artifactsPresence of motion artifacts 
F.Aliasing artifactsAliasing due to parallel imaging 
G.Boundary artifactsArtifacts observed at slab boundaries 
Evaluation scale
 ScoreInterpretation 
 0Negligible presence 
 1Mild presence does not degrade diagnostic quality 
 2Mild presence that degrades diagnostic quality 
 3Presence that severely degrades diagnostic quality 
Table 3. •••
Results of Relative Criteria Evaluation, Showing Mean (Range) Scores*
Vessel groupa/criterionA. Vessel conspicuityB. Vessel continuityC. Vessel sharpness
1.Internal carotid arteries0 (0)0 (0)−0.1 (−1 to 0)
2.MCA M1-M3 segments0 (0)0 (0)−0.4 (−1 to 0)
3.MCA M4 and beyond+1.3 (+1 to +2)+0.4 (0 to +1)−0.4 (−1 to 0)
4.ACA A1-A3 segments0 (0)0 (0)−0.9 (−1 to 0)
5.Anterior frontal branches+0.6 (0 to +2)+0.3 (0 to +1)−0.8 (−1 to 0)
6.Vertebral-basilar arteries0 (0)0 (0)0 (0)
7.PCA P1-P3 segments0 (0)0 (0)−0.6 (−1 to 0)
8.PCA P4 and beyond+1.5 (+1 to +2)+0.8 (0 to +2)−0.8 (−1 to 0)
9.Cerebellar arteries0 (0)+0.5 (0 to +1)−0.9 (−1 to 0)
Results of Absolute Criteria Evaluation, Showing Mean (Range) Scores
Acquisition/criterionD. Venous signalaE. Motion artifactsaF. Aliasing artifactsaG. Boundary artifactsb
  • *

    Statistical significance as tested using a Wilcoxon signed rank test is indicated in bold font (P < 0.05).

  • a

    Vessel groups as listed in Table 1.

  • a, b

    Test for significance is based on score difference between TOF and FIR-MRA.

TOF0.6 (0 to +2)0 (0)0.5 (0 to +1)0 (0)
FIR-MRA0.1 (0 to +1)0 (0)0 (0)0 (0)

RESULTS

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. CONCLUSION
  7. Acknowledgements
  8. REFERENCES

Characterization of Principal Parameters

Figure 2a illustrates various arrival time positions. These positions are indicated in Fig. 2b-e, which shows the variation of the principal parameters (I to IV) in plots of vessel contrast versus kr, the radial distance from the origin in kY-kz-space. Figure 2b shows that the lowest TI of 600 msec produces the largest peak signal at t = 0 msec but has no signal in k-space center at t = 750 msec. Figure 2c shows that an increase in α results in both an increase in peak signal at kr = 0 and an undesirable increase in signal modulation. Figure 2d shows that an increase in nR results in increased signal modulation. Figure 2e shows the constant scan time analysis for different nR ·R pairs, whereby the trend for a decrease in R is similar to that of an increase in α (Fig. 2c).

thumbnail image

Figure 2. An (a) illustration of arrival time t, and the (b-e) simulated vascular signals obtained at various arrival time positions indicated as 1 (t = 0), 2 (t = 750 msec when t ≤ TI), and 3 (t = 750 msec when t > TI). The plots show vascular signal (obtained as a difference between signal acquired between the selective and nonselective cycles, similar to the Arteries2 plots of Fig. 1b) versus kr, the radial distance in ky-kz-space. The simulations are used to evaluate (b) Parameter I (TI), (c) Parameter II (α), (d) Parameter III (nR), and (e) Parameter IV (R) under a constant scan time analysis, which takes into account noise increase due to R. The control parameters are reflected in Table 1.

Download figure to PowerPoint

Figure 3 shows experimental demonstrations of varying the principal parameters. Increasing TI results in decreased vessel signal but increased distal vessel conspicuity (Parameter I, Fig. 3a-c, arrows). Increasing α reduces background noise but at the cost of reduced vessel sharpness (Parameter II, Fig. 3d-f). Increasing nR from 30 to 60 reduces vessel sharpness (Parameter III, Fig. 3g-j). Decreasing R under the constant scan time analysis (Parameter IV, Fig. 3k-m) provides similar observations to that seen in Fig. 3g-j (Parameter III). Venous signal was diminished when ε was increased from 0.0 to 0.8 (Parameter V, Fig. 3n-p) and was eliminated at ε = 0.8. Over this range, no discernible changes in arterial signal were observed.

thumbnail image

Figure 3. In vivo results acquired for a normal subject (30-year-old male) with axial MIPs (8 cm thick) shown in inverse video. Parameter I, TI at (a) 600 msec, (b) 750 msec, and (c) 900 msec. Arrows in (b) and (c) depict enhancement of the distal branches of the middle cerebral circulation not seen in (a). The dashed box in (a) defines the subvolume shown for (d-m). Parameter II, α at (d) 10°, (e) 15°, and (f) 20°, all at R = 3. Parameter III, nR at (g) 30, (h) 40, and (j) 60, and the constant scan time analysis of Parameter IV with (k)nR = 40, R = 3, (l)nR = 60, R = 2, and (m)nR = 120, R = 1. Parameter V, ε at (n) 0.0, (o) 0.4, and (p) 0.8. Undesirable venous signal in the superior sagittal sinus was seen when ε < 0.8 (arrows, (n), (o)). The control parameters are reflected in Table 1.

Download figure to PowerPoint

Parallel Imaging With IRES

Figure 4a shows FIR-MRA signal simulations of acquisitions with standard self-calibration and IRES calibration. The signal modulation of IRES is independent of R, but the standard self-calibrated acquisition experiences increased signal modulation with increased R. Figure 4b,d shows the FIR-MRA acquisitions with IRES, whereby net accelerations (Rnet) were equivalent to the applied acceleration (R). Figure 4c,e shows that acquisitions with standard self-calibration experience both reduced Rnet and increased signal modulation, especially at the higher acceleration of R = 3 (Fig. 4e).

thumbnail image

Figure 4. Comparisons of IRES versus standard self-calibration (SC) in 3D FIR-MRA acquisitions of a normal subject (30-year-old male). a: Signal modulation of vessel contrast against k-space; and (b-e) targeted, coronal reformat MIPs (4 cm thick) of FIR-MRA acquisitions of the proximal slab using parameters indicated in Table 1. The acquisitions are (b) IRES R = Rnet = 2, (c) SC R = 2 (Rnet = 1.8), (d) IRES R = Rnet = 3, and (e) SC R = 3 (Rnet = 2.4). SC had increased signal modulation versus IRES calibration at the same R, resulting in increased signal modulation (arrows, (c), (e)).

Download figure to PowerPoint

In Vivo Study: Comparing FIR-MRA to Conventional TOF

Table 3a,b summarizes the results from the evaluation. For vessel conspicuity, FIR-MRA was shown to be significantly superior to TOF in two of the nine vessel groups, superior but not statistically so in one, and similar to TOF in the remaining six groups. For vessel continuity, FIR-MRA was deemed superior to TOF for four distal vessel groups and similar to TOF for the remaining vessel groups. However, none of the vessel groups had statistically significant improved vessel continuity. For vessel sharpness, TOF was superior in eight of nine groups but statistically so in only four of them. In terms of venous signal and aliasing artifacts (Table 3b), the overall image quality of both examinations was good. Statistical significance was not shown in any of these groups, but FIR-MRA was scored equal or better than TOF in every subject. Neither substantial motion nor boundary artifacts were noted for either examination.

Figure 5 shows TOF and FIR-MRA images from one of the normal subjects. In addition to illustrating some of the findings in the qualitative comparison, the images also show superior background suppression of FIR-MRA (Fig. 5e) compared to TOF (Fig. 5a). An improved tissue and CSF contrast was seen in the bright-blood source images of FIR-MRA (Fig. 5f) as compared to TOF (Fig. 5b).

thumbnail image

Figure 5. 3D TOF (top row, (a-d)) and 3D FIR-MRA (bottom row, (e-h)) images from one normal subject (37-year-old male). Full volume axial MIPs shown in inverse video (a,e) demonstrate the superior distal vessel conspicuity and superior background suppression of FIR-MRA. Improved tissue contrast and CSF signal suppression were seen in the FIR-MRA bright-blood source image (f) compared to the corresponding axial slice from TOF (b). The targeted, sagittal MIPs of the left carotid circulation (c,g) and that of the vertebral circulation (d,h) show that FIR-MRA had both superior vessel conspicuity and inferior vessel sharpness in the distal vessels. Venous signal was seen in TOF ((a,d), arrows) but was not discernible in FIR-MRA. The loss of vessel continuity due to spatial saturation bands ((d), arrowhead) was apparent in TOF.

Download figure to PowerPoint

Patient Studies

Figure 6 shows results from the first patient study. The right middle cerebral artery aneurysm could be seen in the CT angiography examination (Fig. 6a), TOF (Fig. 6b), and FIR-MRA (Fig. 6c). Both MRA methods revealed the presence of another aneurysm of diameter 4 mm in the cavernous region of the left internal carotid artery. A higher-resolution, single-slab FIR-MRA (Fig. 6d) provided an improvement in vessel sharpness over the standard FIR-MRA (Fig. 6c).

thumbnail image

Figure 6. Images from the first patient study (67-year-old female). A 4 mm × 7 mm saccular aneurysm at the right middle cerebral artery trifurcation (arrowhead in all figure parts) was detected with (a) CT angiography shown in coronal reformat. The same aneurysm was well depicted with MRI, shown in targeted, coronal MIPs of (b) 3D TOF (1.4 mm slice thickness superior-inferior), (c) 3D FIR-MRA (1.4 mm slice thickness, acquisition time of 7.7 min), and (d) high-resolution single-slab 3D FIR-MRA (0.7 mm slice thickness, acquisition time of 15.4 min). A second, 4 mm-diameter cavernous aneurysm in the left internal carotid artery was detected with MRA (arrow, (b-d)) and was retrospectively seen in CT angiography (arrow, (a)). Improved vessel sharpness was seen in the (d) higher-resolution FIR-MRA as compared to the standard FIR-MRA (c).

Download figure to PowerPoint

In the second patient study (Fig. 7), the x-ray digital subtraction angiography (Fig. 7a,b) of the right carotid is shown prior to and after the endovascular coil procedure. The aneurysm remnants were best seen in the CE-MRA (Fig. 7d), followed by FIR-MRA (Fig. 7e) and then TOF (Fig. 7c). Superior vessel conspicuity was observed in FIR-MRA relative to TOF in this elderly patient (76 years old).

thumbnail image

Figure 7. Images from the second patient study (76-year-old female) of a coiled aneurysm of the right internal carotid artery. Right carotid, sagittal x-ray digital subtraction angiography shows the aneurysm (a) prior to and (b) after the endovascular coil procedure. Sagittal projections (4 cm thick) of (c) TOF, (d) CE-MRA acquired after a timing bolus, and (e) FIR-MRA show that the appearance of aneurysm remnants (arrows) of FIR-MRA was more similar to the CE-MRA than the TOF was. Superior vessel conspicuity was also observed in FIR-MRA relative to TOF (arrowheads).

Download figure to PowerPoint

In the final patient study (Fig. 8), x-ray digital subtraction angiography (Fig. 8a,b) of the left carotid circulation is shown. The nidus of the AVM was best visualized on the FIR-MRA angiogram (Fig. 8e) compared to TOF (Fig. 8c) and CE-MRA (Fig. 8d). In addition, the feeding arteries, the nidus, and the large draining vein were distinguishable from each other and from the surrounding brain tissue in the FIR-MRA bright-blood image (Fig. 8f), unlike the other two MRA techniques (Fig. 8c,d).

thumbnail image

Figure 8. Images from the third patient study (29-year-old male), showing x-ray digital subtraction angiography images of the left temporal lobe AVM (3-cm diameter) in the (a) coronal and (b) sagittal orientations. The dashed line indicates the position of the cropped axial projections (15 mm thick) as depicted by (c) TOF, (d) CE-MRA, (e) FIR-MRA angiogram, and (f) FIR-MRA bright-blood data. The feeding arteries (arrows) were well visualized in all images. The nidus (arrowhead) was best seen with the FIR-MRA angiogram. The vein (chevron) was best seen on the CE-MRA but was indistinguishable from the nidus and feeding arteries. On the FIR-MRA bright-blood image (f), all three vessel components (feeding arteries, nidus, vein) identified previously were distinguishable from each other and from surrounding brain tissue.

Download figure to PowerPoint

DISCUSSION

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. CONCLUSION
  7. Acknowledgements
  8. REFERENCES

We have developed a high-resolution, 3D, non-contrast-enhanced, FIR-MRA technique utilizing spin-labeling for intracranial angiography. The principal parameters of the sequence were identified and optimized. The technique was both tested in normal human subjects and used in three patients referred for clinical MRA. The choice of employing a single-scan spin-labeling technique provided superior conspicuity of distal vessels over 3D TOF. Simultaneously, dark-blood and bright-blood T1-weighted images are produced, which may have added diagnostic benefits in discriminating arteries from veins and brain tissue. The off-center selective inversion provides an effective venous suppression scheme that discriminates venous blood from arterial blood. The combination of fast GRE and parallel imaging allowed for a high-resolution, 3D, intracranial, MRA examination of 7.7 min. In particular, the IRES method for parallel imaging resulted in neither loss of net acceleration nor further signal modulation. The principal limitation of FIR-MRA versus TOF was some loss of vessel sharpness. This is due to the signal modulation during the readout. FIR-MRA also did have a somewhat longer scan time than TOF (6.4 min with calibration scan). Conceivably, improved receiver coils could allow higher acceleration, thereby reducing the degree of signal modulation and reducing the loss of sharpness.

To evaluate vessel continuity apart from vessel conspicuity, vessels that were conspicuous in FIR-MRA but not in TOF were given a zero score for vessel continuity. This was noted most prominently in the anterior frontal branches (vessel group 5). The aliasing artifacts in TOF were typically seen in the pontine region of the brain and were attributed to the image-space-based reconstruction used, rather than to the TOF technique. Ghosting artifacts were observed in TOF but did not factor in the evaluations as these were attributed mostly to the linear phase-encoding order of TOF. While the presence of veins in TOF did not result in nondiagnostic images, the application of saturation bands for venous suppression also suppressed arteries that reenter the imaging slab, contributing to the appearance of vessel discontinuity. In comparison, the venous suppression method in FIR-MRA was more effective and was not detrimental to the depiction of arteries.

There are some other limitations of FIR-MRA. In spite of FIR-MRA being acquired in a single-scan manner, motion that occurs at time scales smaller than TC may result in image artifacts. Motion between the acquisitions of consecutive slabs may also occur in both FIR-MRA and TOF. However, no motion artifacts were observed (Table 3b). A limitation of spin-labeling-based techniques mentioned in the literature (26) is the signal dropout in the proximal vessels at large TI (∼1500 msec). This arises from the scanner's maximum field of view that limits the extent of the labeling region proximal to the imaging slab. In this work, these signal dropout effects were not observed, given the smaller TI (≤900 msec) used, and the maximum field of view was 48 cm. Finally, because the FIR-MRA technique relies on the long T1 of blood, FIR-MRA must be performed prior to any intravenous injection of gadolinium-based contrast agents.

There are several parameters that were not demonstrated in this work. TC could be increased to increase vascular contrast but would also result in scan time increase. The number of slabs could be reduced by increasing the slab thickness of each slab, but this could potentially reduce vessel conspicuity. The readout bandwidth could be reduced to increase SNR, but doing so would increase signal dephasing due to a longer echo time. The use of balanced steady-state free precession as seen in other non-CE-MRA methods (29–31) could provide increased efficiency due to shorter echo time and increased received signal. In this work, however, the choice of spoiled GRE avoided the issues of banding artifacts and specific absorption rate intrinsic to balanced steady-state free precession. All acquisitions in this study were obtained at 3 T, which from theory is advantageous versus 1.5 T because a longer T1 of blood at 3 T increases vessel signal.

There are several potential benefits of FIR-MRA. The improved conspicuity of distal vessels, aneurysm remnants (Fig. 7), and the nidus of the AVM (Fig. 8) all suggest that FIR-MRA may be better than TOF in depicting vascular territories with slow flow. The ability to distinguish vessel components from tissue with draining veins appearing dark in bright-blood FIR-MRA may provide a new avenue for evaluating AVMs. The complete background suppression in FIR-MRA allowed visualizations with volume rendering and MIPs to be performed without the need for image segmentation, facilitating, for example, automatic aneurysm detection algorithms (32, 33). Branches of the external carotid arteries were unevaluated but were observed to also be better depicted in FIR-MRA due to obscuration by bright, subcutaneous fat in TOF. Hence, FIR-MRA may be useful for evaluation of giant cell arteritis, which characteristically involves the superficial temporal artery. Finally, for multicontrast carotid plaque imaging (34), FIR-MRA can simultaneously provide high luminal signal and blood-nulled T1 contrast, hence avoiding the need for image registration.

CONCLUSION

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. CONCLUSION
  7. Acknowledgements
  8. REFERENCES

The noncontrast angiographic technique of 3D FIR-MRA has demonstrated improved distal vessel conspicuity relative to standard 3D TOF in intracranial imaging and has demonstrated clinical utility in imaging aneurysms and AVMs.

Acknowledgements

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. CONCLUSION
  7. Acknowledgements
  8. REFERENCES

The authors acknowledge the valuable technical help of Matt A. Bernstein, Ph.D., Eric A. Borisch, Roger C. Grimm, Clifton R. Haider, Tom C. Hulshizer, Joseph M. Kreidermacher, Gary M. Miller, M.D., Phillip J. Rossman, and Diane M. Sauter.

REFERENCES

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. CONCLUSION
  7. Acknowledgements
  8. REFERENCES