Intracranial arterial wall imaging using three-dimensional high isotropic resolution black blood MRI at 3.0 Tesla

Authors

  • Ye Qiao PhD,

    1. The Russell H. Morgan Department of Radiology and Radiological Sciences, The Johns Hopkins Hospital, Baltimore, Maryland, USA
    Search for more papers by this author
  • David A. Steinman PhD,

    1. Biomedical Simulation Laboratory, Department of Mechanical and Industrial Engineering, University of Toronto, Toronto, Ontario, Canada
    Search for more papers by this author
  • Qin Qin PhD,

    1. The Russell H. Morgan Department of Radiology and Radiological Sciences, The Johns Hopkins Hospital, Baltimore, Maryland, USA
    2. F.M. Kirby Research Center for Functional Brain Imaging, Kennedy Krieger Institute, Baltimore, Maryland, USA
    Search for more papers by this author
  • Maryam Etesami MD,

    1. The Russell H. Morgan Department of Radiology and Radiological Sciences, The Johns Hopkins Hospital, Baltimore, Maryland, USA
    Search for more papers by this author
  • Michael Schär PhD,

    1. The Russell H. Morgan Department of Radiology and Radiological Sciences, The Johns Hopkins Hospital, Baltimore, Maryland, USA
    2. Philips Healthcare, Cleveland, Ohio, USA
    Search for more papers by this author
  • Brad C. Astor PhD,

    1. Department of Epidemiology, The Johns Hopkins Bloomberg School of Public Health, Baltimore, Maryland, USA
    2. Department of Medicine, The Johns Hopkins University School of Medicine, Baltimore, Maryland, USA
    Search for more papers by this author
  • Bruce A. Wasserman MD

    Corresponding author
    1. The Russell H. Morgan Department of Radiology and Radiological Sciences, The Johns Hopkins Hospital, Baltimore, Maryland, USA
    • Johns Hopkins Hospital, 367 East Park Building, 600 North Wolfe Street, Baltimore, MD 21287
    Search for more papers by this author

Abstract

Purpose:

To develop a high isotropic-resolution sequence to evaluate intracranial vessels at 3.0 Tesla (T).

Materials and Methods:

Thirteen healthy volunteers and 4 patients with intracranial stenosis were imaged at 3.0T using 0.5-mm isotropic-resolution three-dimensional (3D) Volumetric ISotropic TSE Acquisition (VISTA; TSE, turbo spin echo), with conventional 2D-TSE for comparison. VISTA was repeated for 6 volunteers and 4 patients at 0.4-mm isotropic-resolution to explore the trade-off between SNR and voxel volume. Wall signal-to-noise-ratio (SNRwall), wall-lumen contrast-to-noise-ratio (CNRwall-lumen), lumen area (LA), wall area (WA), mean wall thickness (MWT), and maximum wall thickness (maxWT) were compared between 3D-VISTA and 2D-TSE sequences, as well as 3D images acquired at both resolutions. Reliability was assessed by intraclass correlations (ICC).

Results:

Compared with 2D-TSE measurements, 3D-VISTA provided 58% and 74% improvement in SNRwall and CNRwall-lumen, respectively. LA, WA, MWT and maxWT from 3D and 2D techniques highly correlated (ICCs of 0.96, 0.95, 0.96, and 0.91, respectively). CNRwall-lumen using 0.4-mm resolution VISTA decreased by 27%, compared with 0.5-mm VISTA but with reduced partial-volume-based overestimation of wall thickness. Reliability for 3D measurements was good to excellent.

Conclusion:

The 3D-VISTA provides SNR-efficient, highly reliable measurements of intracranial vessels at high isotropic-resolution, enabling broad coverage in a clinically acceptable time. J. Magn. Reson. Imaging 2011;. © 2011 Wiley-Liss, Inc.

THE PRESENCE OF intracranial vascular disease is highly predictive of stroke (1). However, disease prevalence may be underestimated due to the lack of an appropriate diagnostic tool to depict the intracranial vessel wall (2). Black blood MR imaging (BBMRI) has emerged as an effective method to measure wall thickness and identify pathological features of extracranial vessels (3–5). Recently, its application has been extended to evaluate intracranial vessels, specifically to detect atherosclerosis (6–9) and vasculitis (7, 10). Measuring intracranial vessel wall thickness remains a technical challenge given the small size of these vessels. Furthermore, the techniques introduced thus far have been standard two-dimensional (2D) black blood sequences, which are prone to partial volume artifacts amplified by the inherent curving course of intracranial vessels (11). This adds to the challenge of covering the numerous intracranial sites that are prone to atherosclerosis formation (e.g., basilar artery [BA], middle cerebral artery [MCA], and petrous internal carotid artery [ICA]) (12) by 2D imaging.

Three-dimensional acquisitions enable high isotropic resolution that can minimize the overestimation of wall thickness as a consequence of the tortuosity of these small vessels; however, 3D techniques suffer from long scan times and suboptimal flow suppression (13). For example, double inversion recovery techniques (14, 15) typically used in 2D acquisitions generally provide inadequate flow suppression in 3D acquisitions because of the relatively thick re-inversion pulse required. Furthermore, the long echo train length (ETL) used to suppress flow by dephasing effects in 2D turbo spin echo (TSE) techniques (16) are not possible at 3D without impractically long scan times. A recently proposed 3D technique, Volumetric ISotropic TSE Acquisition (VISTA, Philips), uses variable-flip-angle refocusing pulses to achieve a longer ETL for more effective flow suppression without compromising signal and at relatively short scan times (17). In fact, this technique has been shown to have higher signal-to-noise ratio (SNR) efficiency and stronger black-blood effects compared with conventional 3D TSE sequences (17–19).

A 3D variable flip-angle refocusing pulse sequence has been used to image carotid (19) and peripheral (20) arterial walls. However, one cannot intuit the successful application of this technique to intracranial wall imaging because these vessels are structurally unique. For example, they are surrounded by cerebrospinal fluid (CSF) rather than soft tissue (e.g., fat). Thus, our aim was to develop and optimize a high, isotropic resolution 3D BBMRI (i.e., VISTA) protocol to measure intracranial arterial wall size in a clinically acceptable scan time, using a conventional 2D BBMRI sequence (i.e., double inversion TSE) as a reference.

METHODS

Three-dimensional VISTA - Technical Considerations

The applied 3D isotropic resolution VISTA sequence is a variant of TSE with variable-flip-angle (FA) nonselective refocusing RF pulses and radial view ordering. The variable refocusing FA modulation is designed to achieve a target signal level by a precipitous drop in the initial FAs, and then maintain a pseudo-steady-state signal level over the remainder of the echo train by gradually increasing FAs. This minimizes signal blurring from T2 decay while reducing RF power (21).

The primary mechanisms for the intrinsic black-blood effects of 3D VISTA include the following: (i) Intravoxel dephasing of moving blood spins. Blood with a spectrum of velocities and accelerations flowing across a magnetic field gradient leads to widespread phase dispersion that results in signal loss. In particular, the complex state of motion such as turbulence or pulsation contributes to the spread of velocities and accelerations, and serendipitously induces additional signal attenuation (16). Furthermore, the flow suppression is more effective for vessels with small diameters, such as cerebral vessels (22). (ii) The use of low FA refocusing pulses causes the formation of simulated echoes, which store magnetization along the longitudinal axis and exhibits a complicated phase evolution between the longitudinal and transverse planes that results in signal loss (17). Furthermore, the FA impacts flow-related signal loss, and a smaller FA leads to greater flow suppression (17, 23).

The signal of the vessel wall achieved using the VISTA sequence can be optimized by enabling radial-ordering modulation in which the center of K space is sampled at the beginning of the echo train (17). This has the added benefit of minimizing the T2-weighting of the image, thereby darkening the signal of the surrounding CSF.

The VISTA pulse sequence implemented herein was based on the 3D proton density (PD)-weighted TSE technique described by Busse et al (17), that uses a variable FA refocusing control, autocalibrating 2D-accelerated parallel imaging, and radial view ordering to produce isotropic high-resolution images. Parameters (e.g., TE, ETL, and resolution) were modified to facilitate intracranial wall imaging.

Study Population

Thirteen healthy volunteers (8 males; ages, 22–82 years; mean, 44 years) with no history of intracranial vascular disease were recruited. Four patients (1 male; ages, 38, 42, 44, and 61 years) with intracranial stenosis based on a preceding MR angiogram (MRA) or CTA were recruited (one BA stenosis, three MCA stenoses). Institutional review board approval was obtained and participants provided informed consent.

MRI Examination

All exams were performed on a 3T MRI scanner (Achieva; Philips Healthcare, The Netherlands) using the body coil for transmission and an eight-channel head coil for reception. A 3D time-of-flight (TOF) MRA was first acquired to localize the intracranial arteries. 3D VISTA images were then acquired in a coronal plane (45-mm-thick slab) to cover the major intracranial vessels as identified on the TOF MRA. Imaging parameters were as follows: repetition time/echo time (TR/TE), 2000 ms/38 ms; TSE factor, 60 including 4 startup echoes; echo spacing, 6.1 ms; sense factor, 2 (right–left direction); oversampling factor, 1.8; and number of averages, 1. The FOV was 200 × 166 × 45 mm3 at a matrix of 400 × 332 × 90 for an acquired voxel volume of 0.5 × 0.5 × 0.5 mm3 (scan time, 7.9 min). To explore the trade-off between SNR and voxel volume, a VISTA sequence was repeated with an acquired resolution of 0.4 × 0.4 × 0.4 mm3 for 6 volunteers and 4 patients using a half scan factor (partial Fourier) of 0.6 to approximate the same coverage and scan time (scan time, 7.6 min). The variable-flip-angle scheme for the VISTA acquisitions is illustrated in Figure 1. Radial k-space ordering was used in the phase-encoding and partition-encoding directions, and no fat suppression or electrocardiography (ECG) trigger was applied.

Figure 1.

Refocusing variable-flip-angle scheme used for VISTA acquisitions. Parameters: αmin = 50°, αmax = 120°, ETL = 56 + 4 startup echoes.

The 2D BBMRI images were acquired for all volunteers using an ECG-gated double inversion recovery TSE sequence with the following parameters: TR/Turbo factor/TE: 2 RR/10/9 ms; FOV, 120 × 90 mm2; 1 excitation; slice thickness, 2 mm with 0 gap; number of averages, 2. Two sets of 2D BBMRI images were acquired with resolutions of 0.25 × 0.25 × 2 mm3 and 0.5 × 0.5 × 2 mm3, and scan times of 74 s /slice and 37 s /slice, respectively. The MRI slices were oriented perpendicular to the vessel axis at three standard locations that represent common sites for intracranial atherosclerosis (12) (Fig. 2): (a) basilar trunk, 5–6 mm proximal to its terminal bifurcation; (b) M1 segment of MCA, 5–6 mm beyond the origin of M1; (c) horizontal petrous segment of the ICA, 4–5 mm proximal to the cavernous segment. The side of the MCA and ICA used for imaging was randomly chosen before imaging for each segment. Two MRI slices were obtained at each location. For patients, three to five 2D BBMRI slices were acquired centered at the most stenotic regions.

Figure 2.

The 3D TOF MRA collapsed Maximum Intensity Projection (MIP) image (AP view) demonstrates normal intracranial vessels in a 38-year-old male (a). 2D black blood MRI (BBMRI) images were acquired at two in-plane resolutions (0.25 × 0.25 mm2 versus 0.5 × 0.5 mm2) and oriented orthogonal to the M1 segment of the MCA (b,c), distal BA (d,e) and horizontal petrous ICA (f,g) as prescribed on the TOF MRA (a, lines). Cross-sectional view of vessel of interest is identified by arrow.

In addition, a 3D fluid-attenuated inversion recovery (FLAIR) VISTA sequence (acquired resolution, 0.5 × 0.5 × 0.5 mm3), was also acquired in 2 volunteers to study the effect of CSF suppression on SNR and the visual conspicuity of the vessel wall.

Image Analysis

Comparison Between 3D VISTA and 2D TSE

MRI images were processed using customized software (VesselMass, Leiden University Medical Center, the Netherlands). The 3D VISTA images (acquired resolution, 0.5-mm isotropic) were reconstructed to 0.5-mm and 2-mm slice thicknesses at orientations identical to the 2D TSE slices using the Multi-Planar Reformations (MPR) tool (Fig. 3). For signal comparison, the reconstructed 0.5-mm-thick VISTA images (0.5 × 0.5 × 0.5 mm3) were matched with the native 2D TSE images (0.2 5 × 0.25 × 2 mm3), having identical voxel volumes. For morphologic comparison, the reconstructed 2-mm-thick VISTA images (0.5 × 0.5 × 2 mm3) were matched with the native 2D images (0.5 × 0.5 × 2 mm3) for the same in-plane resolution and slice thickness to test whether they provided comparable wall thickness and lumen and wall area measurements. To minimize recall bias, the 2D TSE and reconstructed VISTA images were analyzed in separate sessions by at least 2 weeks.

Figure 3.

The 3D VISTA images (0.5-mm isotropic resolution) of a BA in a 38-year-old healthy volunteer. A 3D VISTA image is reconstructed to visualize the long axis of the BA (arrows, a). Short axis VISTA images are reconstructed to 0.5-mm-thickness (b) and 2-mm-thickness (c) at the same position as the 2D TSE image (Fig. 2d,e) using the Multi-Planar Reformations (MPR) tool (dashed line, a). Cross-sectional view of BA is identified by arrow (b,c).

Images were analyzed by two readers using a semi-automated contouring feature of VesselMass software. Contours were generated using a gradient image that displays the spatial derivatives in image intensity (i.e., edges) extracted from the original gray-scale image. These edges provide an objective definition for soft tissue boundaries, which eliminates the influence of subjective window/level settings for vessel contour detection (24). Lumen and outer wall contours were drawn using the gradient image by bisecting the band of high intensity that represents the lumen and wall interface, as well as the band representing the interface between the wall and surrounding tissue (Fig. 4d). Lumen area (LA), wall area (WA), mean wall thickness (MWT), and maximum wall thickness (maxWT) values were generated (Fig. 4). For regions without a clear boundary (e.g., 2–4 clock in Fig. 4d), the contour was traced to maintain the continuity of the vessel's curvature based on the magnitude image.

Figure 4.

The 3D VISTA images (0.5-mm isotropic resolution) of a basilar artery in an 82-year-old healthy volunteer. Long-axis view of the basilar artery (a) to orient short axis view (b, reconstructed at line shown in a). A gradient image (c) is generated using Sobel operator (24) to guide contour placement (d) using VesselMass software (Leiden University, the Netherlands). Contours are transferred to the magnitude image (e), and then used to divide the vessel wall into 12 radial segments with thickness and area measurements generated by VesselMass software (f). Basilar artery, arrow.

The SNR of lumen (SNRlumen) and wall (SNRwall) measurements were calculated: SNR = S/SDnoise, where S is the averaged signal intensity of the region of interest, and SDnoise is the standard deviation of noise. Because of the inhomogeneous noise distribution encountered in parallel imaging, we measured noise from an ROI of 25 mm2 manually placed in the adjacent white matter (20, 25) instead of using the air. The contrast-to-noise ratio (CNR) of wall versus lumen (CNRwall-lumen) was calculated as CNRwall-lumen= SNRwall-SNRlumen. The CNR efficiency (CNReff) was determined to account for differences in scan times between 2D and 3D techniques to enable a fair comparison. CNReff was calculated as: CNReff = CNR/(VOXEL(TAslice)1/2), where VOXEL is the voxel volume (in mm3) and TAslice is the scan time per slice (in minute) (20).

Comparison Between VISTA Acquired at 0.4 and 0.5 Isotropic Resolution

The 3D VISTA dataset acquired at 0.4-isotropic resolution was reconstructed at 0.4-mm-thick slices at three standard locations as prescribed for the 0.5-isotropic 3D VISTA images (i.e., based on the positioning of the 2D slices). In addition, the native coronal view image that best displayed the supraclinoid ICA segment in cross section was selected from the two 3D datasets, and the slice locations were matched. Therefore, four vessel segments were analyzed from the participants who underwent both 0.4-isotropic and 0.5-isotropic VISTA imaging. Signal and morphologic measurements were assessed in the same manner as described in the previous section.

Statistical Analysis

Data were analyzed using SPSS 18.0 (SPSS Inc, Chicago, IL). All signal-based measurements (SNRlumen, SNRwall, CNRwall-lumen and CNReff(wall-lumen)) were determined for each slice (n ≤ 6) and a single value was used for each participant based on the average of all slices. Morphological variables (LA, WA, MWT, and maxWT) were reported as the average of both slices for each vessel segment (i.e., MCA, BA, petrous ICA, and supraclinoid ICA), as wall thickness may vary by location. All signal-based and morphological measurements were compared between 3D VISTA and 2D TSE sequences using two-tailed paired t-tests. The same test was conducted to compare the VISTA images acquired at 0.4-isotropic versus 0.5-isotropic resolution. Agreement between MRI measurements obtained from 2D and 3D techniques were assessed using Bland-Altman plots (26) and intraclass correlation coefficients (ICC) (27). Inter- and intra-reader variability was assessed using ICC, and reliabilities below 0.4 were characterized as poor, 0.4 to 0.75 as fair to good, and above 0.75 as excellent (28). Repeated measures analysis of variance was used to calculate between-subject variance and between-reader variance for MWT of each vessel segment and each spatial resolution. Data are presented as means ± standard deviations.

RESULTS

The 3D VISTA images were successfully acquired from 17 participants (13 volunteers and 4 patients). The major intracranial vessel walls were clearly visualized in all participants, and no atherosclerotic plaques were noted in healthy volunteers. Minimal flow artifact was identified in two cases as a wisp of faint signal projecting into the lumen from the inferior vessel wall at the junction between the horizontal and vertical segments of the petrous ICA, but not elsewhere including in the MCA, BA, and ICA beyond the petrous segments for all participants.

Comparison Between VISTA and 2D TSE

The 3D VISTA 0.5-mm isotropic-resolution images were reconstructed and matched with corresponding 2D MRI images in 12 volunteers. One volunteer moved between the VISTA and 2D TSE sequences, prohibiting comparison. Only those cases with adequate image quality on the 2D TSE sequence were compared. A total of 54 pairs of 2D and 3D images at the MCA, BA and petrous ICA locations were used for comparison.

Compared with 2D TSE (0.25 × 0.25 × 2.0 mm3) image measurements, 3D VISTA images acquired at the same voxel volume (0.5 × 0.5 × 0.5 mm3) and reconstructed to the same location provided 58% improvement in SNRwall (6.34 ± 1.84% versus 10.01 ± 2.45; P < 0.01), 74% improvement in CNRwall-lumen (3.70 ± 1.20 versus 6.45 ± 1.84; P < 0.01), and 484% improvement in CNReff (wall-lumen) (45.69 ± 13.26 versus 266.93 ± 65.33; P < 0.01). A difference in SNRlumen could not be detected between the 3D and 2D acquisitions (3D VISTA, 2.89 ± 1.40 versus 2D TSE, 2.68 ± 0.82). For a comparison of morphology, the 3D VISTA images reconstructed to the same voxel dimension as the 2D TSE images (0.5 × 0.5 × 2.0 mm3) revealed excellent agreement between measurements of LA, WA, MWT and maxWT for each vessel segment (ICCs of 0.96, 0.95, 0.96, and 0.91, respectively; Table 1). There was no difference in LA, WA, MWT and maxWT measurements for BA, petrous ICA and MCA segments compared between 2D and 3D acquisitions (P value not significant). Bland-Altman analysis showed good agreement without a bias between techniques (mean wall thickness shown in Fig. 5).

Figure 5.

Bland-Altman plot of the percent difference versus mean for the 2D and 3D paired MWT measurements. 3D images (0.5-mm isotropic resolution) were reconstructed to match the voxel dimensions of the 2D sequence (0.5 × 0.5 × 2 mm3). Measurement pairs show good agreement.

Table 1. Morphologic Measurements for Intracranial Vessel Segments From 3D VISTA and 2D TSE Images*
 2D TSE (0.5x0.5x2 mm3)3D VISTA (0.5x0.5x2 mm3)
MCABAPetrous ICAMCABAPetrous ICA
  • *

    P not significant for paired (2D TSE-3D VISTA) measurements.

Lumen area (mm2)7.61±2.397.33±2.4118.88±0.377.96±2.737.70±2.4218.17±0.38
Wall area (mm2)6.46±1.427.15±1.8816.09±0.076.39±1.597.15±1.8816.60±0.06
Mean wall thickness (mm)0.57±0.070.65±0.080.97±0.160.58±0.060.65±0.070.96±0.14
Maximum wall thickness (mm)0.77±0.100.85±0.171.42±0.320.77±0.100.84±0.121.35±0.28

Comparison Between 0.4-mm and 0.5-mm Isotropic Resolution VISTA Sequences

The 3D VISTA images acquired at 0.4-mm resolution in six volunteers were reconstructed at four locations (BA, MCA, petrous ICA, and supraclinoid ICA) and matched with corresponding images reconstructed at 0.5-mm resolution (37 image pairs). Compared with 0.5-mm resolution image measurements, 0.4-mm resolution images showed a 27% decrease in CNRwall-lumen (6.43 ± 2.16 versus 4.67 ± 1.25; P < 0.01). For morphologic measurements, the MWT combined per arterial location for all participants obtained from 0.4-mm images decreased by an average of 10.2% compared with corresponding 0.5-mm images (paired differences were significant, P < 0.05). We observed a qualitative improvement in plaque delineation for the patient exams due to diminished partial volume effects related to the improved resolution at 0.4 mm (Figs. 6 and 7).

Figure 6.

TOF MRA MIP (acquired resolution, 0.52 × 0.70 × 1.4 mm3) of the basilar artery demonstrates a high-grade stenosis of its mid segment (arrow, a) in a 42-year-old patient. A multiplanar reconstruction of the MRA dataset oriented through the long axis of the basilar artery depicts the narrowing at its mid segment (arrow, b). Reconstructed 3D VISTA image at 0.5-mm isotropic resolution (c) at the same position as the MRA reconstruction (b) demonstrates wall thickening responsible for the high-grade stenosis (arrow, c). Reconstructed 3D VISTA image at 0.4-mm isotropic resolution (d) at the same position shows improved delineation of the lumen at the point of narrowing (arrow, d) due to reduced partial volume effects. A 2D BBMRI slice (0.25 × 0.25 × 2.0 mm3) acquired coronally at this location (not for the purpose of this study) is also shown (e) for comparison, but suffers from insufficient SNR. Magnified long-axis 3D VISTA 0.4-mm-resolution reconstruction image shows the slice orientation (line, f) used to position a short axis reconstruction of the VISTA dataset through the mid-basilar plaque (g, arrowheads delineate outer wall, arrow points to lumen).

Figure 7.

The 3D VISTA image of the basilar artery (arrow) shown in Figure 4 reconstructed in the same long-axis plane but acquired at a higher resolution (0.4-mm isotropic). Note the sharper delineation of the wall compared with Figure 4a. Also noteworthy is the slight asymmetry in wall thickness of this curved vessel, with the thicker side along its concave bend where there is compensatory thickening likely due to the greater tensile stresses as described for bending vessels (40).

Comparison Between FLAIR-VISTA and VISTA at 0.5-mm Isotropic Resolution

CSF suppression was applied to the VISTA sequence and resulted in poor conspicuity of the vessel wall with a 76% reduction in SNRwall (15.88 ± 1.69 versus 3.81 ± 0.09; P < 0.05) and an 83% reduction in CNRwall-lumen (11.87 ± 0.13 versus 2.03 ± 0.14; P < 0.05). We simulated the wall signal using the formulas for steady-state excitation and inversion recovery (29) and calculated a 72% signal reduction when using CSF suppression (assuming a TR of 2000 ms, a T1CSF of 4300 ms (30), and a T1wall of 1198 ms) (31, 32).

MRI Measurement Reproducibility

Intra- and inter-reader reliability (ICC) for MRI measurements (e.g., MWT, LA, and WA) of petrous ICA, supraclinoid ICA, and BA segments ranged from 0.84 to 0.98 (Table 2). Reliability estimates were lower for MCA measurements, which seemed due to its confluence with adjacent brain parenchyma with little surrounding CSF reducing conspicuity of its outer wall.

Table 2. MRI Measurement Reliability Based on VISTA Acquired at 0.5-mm Isotropic Resolution*
  Petrous ICASupraclinoid ICABAMCA
Mean wall thickness (mm)Inter-reader0.910.940.840.64
Intra-reader0.980.920.900.57
Lumen area (mm2)Inter-reader0.980.920.940.82
Intra-reader0.980.930.970.94
Wall area (mm2)Inter-reader0.980.920.940.82
Intra-reader0.990.990.930.95

Between-subject variance of MWT was 0.125 mm, 0.042 mm, 0.068 mm, and 0.064 mm based on 0.5-isotropic VISTA images, and was 0.122 mm, 0.022 mm, 0.048 mm, and 0.033 mm based on 0.4-isotropic VISTA images for petrous ICA, supraclinoid ICA, BA, and MCA segments, respectively. Between-reader variance was approximately 0.04 mm for all segments at both resolutions: 0.035 mm, 0.041 mm, 0.037 mm, and 0.040 mm based on 0.5-isotropic VISTA images, and 0.0.041 mm, 0.044 mm, 0.037 mm, and 0.041 mm based on 0.4-isotropic VISTA images for petrous ICA, supraclinoid ICA, BA, and MCA segments, respectively.

DISCUSSION

We introduce a new MRI method for high-isotropic resolution imaging of intracranial arterial walls at 3T without the anticipated difficulties of suboptimal flow suppression. This acquisition can cover a large volume of intracranial vessels, inclusive of the typical sites of atherosclerosis formation, in a clinically acceptable scan time of approximately 7 min to provide highly reliable measurements of vessel wall size. In particular, the superior SNR efficiency afforded by the variable-flip-angle refocusing pulses, along with the inherent ability to reconstruct this isotropic imaging volume in any plane, enable better vessel wall visualization compared with 2D TSE black blood sequences used for intracranial arterial imaging.

Once thought to be uncommon, intracranial atherosclerotic disease is now known to be as prevalent as extracranial atherosclerosis (33, 34). Despite a growing recognition of the importance of identifying intracranial atherosclerosis (34), only a few studies have attempted to image intracranial atherosclerosis using MRI (7–9, 35, 36). Until now a 2D BBMRI technique has been the only approach used, but its application is limited by (i) low spatial resolution in the slice-select direction (in general, 2 or 3 mm), thus making 2D images more prone to obliqueness artifact from partial volume effects, which is particularly troublesome for the inherently tortuous intracranial vessels; (ii) long acquisition times needed to achieve high resolution with sufficient SNR to measure the wall and depict fine intracranial lesions; (iii) difficulty positioning 2D slices in one scan to capture multiple intracranial vessels with varying orientations (basilar, MCA, or ICA segments). In comparison, our 3D VISTA sequence has demonstrated high intrinsic SNR/CNR efficiency, allowing for volume acquisitions with 0.4- to 0.5-mm resolution along the slice direction and with broad coverage (45-mm) in one acquisition. Our test against the 2D technique was particularly rigorous considering there were some 2D–3D paired cases not analyzed because of inadequate 2D image quality. Of note, with the aid of MPR, 3D acquisitions enable retrospective visualization of the vessel wall and lumen in flexible planes, therefore allowing for accurate monitoring of disease progression and regression.

The 3D BBMRI techniques have been developed for extracranial arterial wall imaging and are commonly steady-state free precession (SSFP) sequences combined with a black blood preparation pulse (37, 38). However, SSFP for intracranial vessel wall imaging is hampered by strong susceptibility effects from air (e.g., sinuses) and adjacent bony structures (e.g., skull base), particularly at high fields such as 3.0T. In contrast, a 3D TSE technique with a dedicated refocusing sweep and a long echo train (e.g., VISTA) is less sensitive to these field inhomogeneities (19, 20). Our study is the first application of the VISTA sequence for intracranial vessel wall imaging.

It is known that 3D TSE has intrinsic black-blood effects from the dephasing of moving blood spins (16, 22). With VISTA, the intrinsic black-blood effect is further enhanced by the long echo train. Additionally, the low-flip-angle refocusing pulses induce stimulated echoes that increase the phase dispersion (17). The tortuous nature of intracranial vessels promotes appreciable secondary flows that lead to increased dephasing of spins, particularly when the vessels are small (22). In our study, blood signal was effectively suppressed through the 4.5-cm slab with flow suppression comparable to that of the 2D TSE sequence and even provided superior contrast between wall and lumen compared with that achieved by 2D TSE.

Although CSF suppression could theoretically improve the conspicuity of the intracranial vessel wall, our results demonstrated a deleterious effect because of the SNR penalty. To gain contrast between the wall and CSF, we chose radial instead of linear view ordering to obtain T1/PD-weighted images where CSF appeared dark (17). Surrounding CSF seemed to improve wall conspicuity, which was the reason we found MCA thickness measurement reliability to be less than for other vascular segments. We would expect this to improve for the MCA in older individuals with more CSF surrounding the vessel due to age-related brain involution, especially compared with the relatively young volunteers in our study.

Our results show the 3D VISTA sequence can detect lesions and measure the intracranial vessel wall in normal human arteries. It may provide a reference standard of the normal vessel wall to discern pathological changes. Furthermore, we observed that reducing the resolution from 0.5-mm to 0.4-mm isotropic should allow for a reduction in partial-volume-based overestimation of wall thickness and a sharper depiction of wall features (Figs. 6 and 7). As illustrated in Figure 8, for a typical normal cerebral artery wall thickness of 0.5–0.7 mm, the measured thickness decreases by 15–20% in going from 0.5- to 0.4-mm resolution. This is broadly consistent with our measured 10–20% reductions. More importantly, as can be seen by the leveling off of the curves in Figure 8, inadequate spatial resolution serves to “compress” differences in true wall thickness, consequently decreasing the ability to resolve actual differences in wall thickness. Consider, for example, the task of discriminating between a 0.5- and 0.6- or 0.6- and 0.7-mm wall. Referring to Figure 8, at 0.5-mm spatial resolution, the apparent difference would be on the order of 0.02 mm, well below the precision of the measurements. At 0.4-mm spatial resolution, however, the apparent difference is on the order of 0.4 mm, close to the inter-reader variability. In other words, for discriminating differences in cerebrovascular wall thickness, 0.4-mm spatial resolution appears to offer a two-fold increase in apparent resolving power compared with 0.5-mm spatial resolution.

Figure 8.

Effect of spatial resolution on wall thickness measurement. Curves obtained using the techniques outlined in Antiga et al (11) for the case of vessel segmented without using Gaussian smoothing. Symbols are placed at true thicknesses of 0.5, 0.6, and 0.7 mm, with horizontal dashed lines identifying the thickness that would be measured at that true wall thickness.

Limitations to our study include the following: (i) The inability to effectively resolve small intracranial vessels (distal branches of the Circle of Willis). This limitation is more theoretical than clinically relevant because atherosclerosis is a disease of large arteries and our targets (BA, M1 segment MCA, and intracranial portions of the ICA) are typical sites of plaque formation; (ii) Small sample size. We only included a small number of healthy volunteers varied in age and race, though the thin vessels encountered in our healthy and relatively young group of volunteers poses a greater technical challenge than might be encountered in a population more susceptible to atherosclerosis. A large population study will establish the association of risk factors (i.e., age, race) with vessel wall thickness; (iii) We report excellent observer reliability but did not test scan reliability. However, based on the largest investigation reported on extracranial carotid wall MRI measurement reliability (5), overall scan reliability was found to be primarily related to reader variability. Furthermore, the 3D nature of our sequence obviates the need for prospective slice placement by the MRI technologist, which is the most important reason for scan variability (39); (iv) Lack of a standard reference for intracranial vessels. Future studies with autopsy specimen correlation are necessary to test the agreement between MRI measurements and histology. In conclusion, the 3D VISTA sequence offers high isotropic spatial resolution with excellent flow suppression to reliably measure intracranial vessel wall thickness and depict lesions with broad coverage in approximately 7 min at 3.0T. This technique may provide important insight into stroke risk by enabling the assessment of plaque burden not otherwise achievable by conventional angiographic techniques.

Acknowledgements

D.A.S. acknowledges salary support from his Heart and Stroke Foundation Career Investigator Award. The authors would like to thank Dr. YiuCho Chung for his input that helped lead to this study.

Ancillary