High-resolution 3D coronary vessel wall imaging with near 100% respiratory efficiency using epicardial fat tracking: Reproducibility and comparison with standard methods

Authors

  • Andrew D. Scott MSc,

    Corresponding author
    1. Cardiovascular Magnetic Resonance Unit, National Heart and Lung Institute, Imperial College London, London, United Kingdom
    2. Cardiovascular Magnetic Resonance Unit, Royal Brompton and Harefield NHS Foundation Trust, London, United Kingdom
    • Cardiovascular Magnetic Resonance Unit, Royal Brompton Hospital, Sydney Street, London, SW3 6NP, UK
    Search for more papers by this author
  • Jennifer Keegan PhD,

    1. Cardiovascular Magnetic Resonance Unit, National Heart and Lung Institute, Imperial College London, London, United Kingdom
    2. Cardiovascular Magnetic Resonance Unit, Royal Brompton and Harefield NHS Foundation Trust, London, United Kingdom
    Search for more papers by this author
  • David N. Firmin PhD

    1. Cardiovascular Magnetic Resonance Unit, National Heart and Lung Institute, Imperial College London, London, United Kingdom
    2. Cardiovascular Magnetic Resonance Unit, Royal Brompton and Harefield NHS Foundation Trust, London, United Kingdom
    Search for more papers by this author

Abstract

Purpose

To quantitatively assess the performance and reproducibility of 3D spiral coronary artery wall imaging with beat-to-beat respiratory-motion-correction (B2B-RMC) compared to navigator gated 2D spiral and turbo-spin-echo (TSE) acquisitions.

Materials and Methods

High-resolution (0.7 × 0.7 mm) cross-sectional right coronary wall acquisitions were performed in 10 subjects using four techniques (B2B-RMC 3D spiral with alternate (2RR) and single (1RR) R-wave gating, navigator-gated 2D spiral (2RR) and navigator-gated 2D TSE (2RR)) on two occasions. Wall thickness measurements were compared with repeated measures analysis of variance (ANOVA). Reproducibility was assessed with the intraclass correlation coefficient (ICC).

Results

In all, 91% (73/80) of acquisitions were successful (failures: four TSE, two 3D spiral (1RR) and one 3D spiral (2RR)). Respiratory efficiency of the B2B-RMC was less variable and substantially higher than for navigator gating (99.6 ± 1.2% vs. 39.0 ± 7.5%, P < 0.0001). Coronary wall thicknesses (± standard deviation [SD]) were not significantly different: 1.10 ± 0.14 mm (3D spiral (2RR)), 1.20 ± 0.16 mm (3D spiral (1RR)), 1.14 ± 0.15 mm (2D spiral), and 1.21 ± 0.17 mm (TSE). Wall thickness reproducibility ranged from good (ICC = 0.65, 3D spiral (1RR)) to excellent (ICC = 0.87, 3D spiral (2RR)).

Conclusion

High-resolution 3D spiral imaging with B2B-RMC permits coronary vessel wall assessment over multiple thin contiguous slices in a clinically feasible duration. Excellent reproducibility of the technique potentially enables studies of disease progression/regression. J. Magn. Reson. Imaging 2011;33:77–86. © 2010 Wiley-Liss, Inc.

MAGNETIC RESONANCE (MR) coronary angiography has undergone considerable development over the last 20 years. Whole-heart studies are performed in the coronary rest period with navigator gating in typically around 15 minutes (1, 2) and have demonstrated sensitivities and specificities of 78% and 96%, respectively (2), for the detection of significant (≥50% reduction in lumen diameter) stenoses on a per segment basis. Like other angiographic techniques, however, MR coronary angiography is unable to detect the outward remodeling of the vessel wall which occurs in subclinical coronary artery disease prior to luminal narrowing (3). Using double inversion dark blood preparation, it is possible to selectively image the proximal and mid coronary vessel wall (4) and to detect wall thickening associated with asymptomatic coronary artery disease (5–8). This, however, is a challenging technique. Not only are the arteries very small (2–5 mm luminal diameter (9), 0.5–1.0 mm wall thickness (10–13)) and so require high spatial resolution imaging, but they also move considerably with both the cardiac (14) and respiratory (15, 16) cycles. To reduce the effects of cardiac motion, data must be acquired in a window of minimal coronary artery motion within each heartbeat, the timing and duration of which is assessed in a prescan stage (17). Respiratory motion effects may be minimized using breatholding (4, 7, 18, 19) but this limits the acquisition to a single 2D slice and compromises either the signal-to-noise ratio (SNR) or the acquired spatial resolution. Alternatively, free-breathing navigator-gated studies, which are not restricted by breathold ability, have been used for both high-resolution 2D acquisitions (20) and for 3D acquisitions (21). For vessel wall imaging, 3D imaging is preferable as it has the potential to enable detailed assessment over multiple thin contiguous slices and has higher SNR per unit acquisition time. However, the respiratory efficiency of navigator gating is inherently low and can be highly variable due to respiratory drift or erratic breathing (22). As a result, the acquisition duration cannot be predetermined and may be unfeasibly long, leading to failed acquisitions in a significant minority of patients (2).

Beat-to-beat respiratory motion correction (B2B-RMC) is a novel technique which has been implemented with a 3D spiral acquisition to allow retrospective correction of the full range of respiratory motion (ie, 100% respiratory efficiency) (23). This technique acquires a fully sampled low-resolution volume with fat-selective excitation in every cardiac cycle immediately before high-resolution spiral data. 3D displacements of the epicardial fat around the coronary artery are retrospectively obtained from the low-resolution fat images and used to correct the corresponding high-resolution data on a beat-by-beat basis. The much improved respiratory efficiency permits a substantial reduction in imaging time when compared to standard navigator gating with a 5 mm acceptance window, potentially allowing high-resolution 3D coronary vessel wall studies to be completed within a clinically feasible timescale. While conventional coronary vessel wall imaging acquires data with alternate R-wave cardiac gating (2RR gating), the B2B-RMC technique was originally demonstrated with gating every R-wave (1RR gating). The more common practice of 2RR gating increases SNR, and also reduces the sensitivity of the sequence to heart rate variations, as there is more complete longitudinal magnetization recovery between sequence repeats. However, if sufficient SNR could be routinely obtained using 1RR gating, then imaging time could be reduced by a further factor of 2, which would be of considerable benefit.

The aim of this work is to quantitatively assess cross-sectional coronary vessel wall images acquired using 3D spiral imaging with B2B-RMC at near 100% respiratory efficiency and to compare them with conventional navigator gated techniques. B2B-RMC 3D spiral datasets were acquired with both 2RR gating and 1RR gating in 10 healthy subjects, together with a navigator-gated 2D spiral acquisition and a navigator-gated 2D turbo-spin echo (TSE) acquisition, which is commonly used for coronary vessel wall imaging (4, 7, 18, 20). Measurements of vessel wall thickness were compared between techniques. In addition, all four acquisitions in all subjects were repeated on a separate occasion to assess the reproducibility of wall thickness measurements, which is crucial for studies of disease progression and regression.

MATERIALS AND METHODS

All imaging was performed on a Siemens 1.5T Avanto MR scanner (Siemens Medical Systems, Erlangen, Germany) with maximum gradient amplitude 45 mT/m and maximum slew rate 200 mT/m/msec, using an anterior phased array coil. Ten healthy subjects (mean age 38 ± 11 years old, range 28–59 years old, four female) were recruited from the staff of our research institutions according to local ethics regulations and were each scanned on two separate occasions (subsequently referred to as the initial and repeat studies).

Preparatory Imaging

Dark blood cross-sectional right coronary artery vessel wall images were planned from multiple free-breathing in plane bright blood T2-prepared 2D balanced steady state free precession (b-SSFP) images (Fig. 1). Planning acquisitions were performed in the subject-specific coronary rest period which was determined from an initial four-chamber view retrogated breathold b-SSFP cine acquisition (17) (42 msec temporal resolution). The plane selected for vessel wall imaging was located in a straight section of the proximal to mid right coronary artery (within ≈50 mm of the origin). A second retrogated breathold b-SSFP cine acquisition was performed in this plane in order to refine the estimate of the start and duration of the coronary rest period for the selected segment of artery (24). This enabled a suitable choice of gating delay and acquisition window duration for the subsequent coronary vessel wall acquisitions. The dark blood preparation for the spiral coronary vessel wall imaging incorporated a user-defined slab selective reinversion pulse (25) which was positioned to reinvert the coronary artery while avoiding reinversion of the blood flowing into the imaging slab, as shown in Fig. 1.

Figure 1.

Orthogonal in-plane right coronary artery vessel wall images (a,b). The position of the cross-sectional right coronary artery plane (c) is shown in blue. The selective reinversion pulse of the dark blood preparation is shown in red and is positioned to reinvert the coronary vessel wall while minimizing the reinversion of the chest wall and blood in the aortic root and ventricles.

Right Coronary Artery Vessel Wall Imaging

For each subject on each occasion, four acquisitions of the right coronary artery wall were obtained using the sequences described below in a randomized order (80 acquisitions in total). The sequence parameters were selected to obtain high spatial resolution and sufficient SNR within an acceptable acquisition duration. The in-plane resolution was identical for all sequences (0.7 × 0.7 mm). For comparison with the 3D B2B-RMC acquisitions, 2D acquisitions were performed with navigator gating, as is commonly used for coronary artery wall assessment. While a direct comparison with navigator gated 3D acquisitions would have been preferable, to do so with comparable spatial resolution and SNR to that used in the 3D acquisition with B2B-RMC would have resulted in a prohibitively long scan time, due to the typically low respiratory efficiency (≈40% (26)) of navigator gating. The slice thickness for the 2D acquisitions was 6.0 mm while for the 3D acquisitions, 8 × 3.0 mm slices were acquired and reconstructed to 16 × 1.5 mm slices. Although we would have ideally performed the 2D acquisitions with 3.0 mm slice thickness, this was not possible due to SNR and time considerations. To minimize through-plane partial volume effects, which may result in artificially high vessel wall thickness measurements, care was to taken to place the imaging slice/slab perpendicular to the straightest section of artery available.

B2B-RMC 3D Spiral Acquisitions

The sequence diagram for the 3D spiral acquisition with B2B-RMC is shown in Fig. 2 and was based on that described by Keegan et al (23). In every other cardiac cycle, dark blood double inversion preparation was followed by a fully sampled low-resolution 3D volume acquired with fat-selective excitation. Immediately after the low-resolution 3D acquisition, two interleaves of a high-resolution 3D stack of spirals acquisition were obtained with water-selective excitation. A standard, crossed pair diaphragmatic navigator was acquired immediately after the high-resolution spiral readouts. The dark blood inversion time was typically 625 msec (which is optimal for 2RR gating and a heart rate of 60 bpm). This, however, was reduced in subjects with an early coronary rest period so that the acquisition of the imaging data could be performed during minimal coronary motion. The low-resolution 3D volume consisted of a stack of eight single-shot spirals (15 msec readout duration, 2.5 μs sampling time, echo time (TE) 3.9 msec, repeat time (TR) 23 msec), each acquired with a 1-2-1 fat-selective binomial pulse (flip angle 15°) over a field of view (FOV) of 261 × 261 mm with an in-plane resolution of 4.8 × 4.8 mm. The slab thickness was 24 mm with 6/8ths partial Fourier encoding in kz (resulting in the acquisition of six single-shot spirals). Reverse centric kz phase encoding ensured that the center of kz in the low-resolution acquisition was acquired as close as possible to the following high-resolution data. Zero-filling in kz resulted in reconstruction of 16 × 1.5 mm slices. The total acquisition duration for the low-resolution volume was 136 msec. For the high-resolution volume, 75 spiral interleaves (10 msec readout duration, 2.5 μs sampling, TE = 3.4 msec, TR = 2RR) were required to fill kx-ky space at an FOV of 570 mm. The large FOV was used to shift the characteristic spiral artifacts away from the anatomy of interest and to improve SNR. Two spiral interleaves were acquired sequentially following each low-resolution volume (flip angles 45° and 90°), resulting in an acquisition window of 35 msec. The following navigator was used to reject and reacquire data acquired at very extreme respiratory positions. These were defined as diaphragm positions >10 mm outside the subjects normal range, as determined in an ≈30-second navigator scout acquisition. The total acquisition duration was 600 cardiac cycles (for 2RR gating and assuming 100% respiratory efficiency).

Figure 2.

The sequence diagram for the 3D spiral dark blood acquisition with beat-to-beat respiratory motion correction (B2B-RMC) using 2RR gating. Every imaging segment consists of a full low-resolution 3D spiral fat-selective (FE) acquisition immediately followed by two interleaves of a high-resolution 3D spiral acquisition with water selective excitation (WE). Finally a standard crossed pair navigator is used to monitor the position of the diaphragm. [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.]

Images were reconstructed and processed offline using in-house software written in MatLab (MathWorks, Natick, MA). 3D translational respiratory motion information was derived from the fully sampled low-resolution fat-excitation volumes. The respiratory displacements were obtained using normalized subpixel cross-correlation and retrospectively applied to the corresponding high-resolution data to compensate for respiratory motion (23).

The acquisition was repeated with 1RR gating. In this case the dark blood inversion time was reduced to ≈400 msec in order to maintain effective suppression of blood signal. All other parameters were fixed. The total acquisition duration was 300 cardiac cycles (for 1RR gating and assuming 100% respiratory efficiency).

Navigator-Gated 2D Spiral

A single-slice 2D spiral acquisition (TE = 7.0 msec) was acquired with 2RR gating (TR = 2RR). The k-space trajectories and dark blood preparation were identical to those used for the 3D spiral sequence. In every other cardiac cycle, two spiral readouts were acquired sequentially with a total cardiac acquisition window of 43 msec, using binomial water excitation (flip angles 45° and 90°). Respiratory gating was performed with a standard crossed pair navigator positioned over the right hemidiaphragm with a 5 mm end-expiratory accept/reject window and end-expiratory position tracking (24, 27). The total acquisition duration was 76 cardiac cycles (for 2RR gating and assuming 100% respiratory efficiency).

Navigator-Gated 2D Turbo-Spin Echo

A single-slice 2D TSE acquisition (TE = 27 msec, echo spacing 6.8 msec) was acquired with 2RR gating (TR = 2RR). The images were acquired with standard double inversion recovery dark blood preparation using a reinversion pulse slice thickness of 3 times the imaging slice thickness and an inversion time defined as for the 3D spiral B2B-RMC acquisition with 2RR gating. Fat suppression was performed with an asymmetric adiabatic short TI spectral inversion recovery (AASPIR) pulse (24, 28). A matrix of 576 × 576 was acquired over a 403 × 403 mm FOV. As averaging is performed on the fly, the acquisition of multiple averages was not possible in combination with navigator gating. Instead, the SNR in the 2D TSE images was increased by using phase oversampling of 77%. The echo train length was chosen to fit within the subject-specific cardiac rest period and constrained to a maximum of 15. Navigator gating was performed as for the 2D spiral technique. The total acquisition duration was 204 cardiac cycles (for 2RR gating, assuming 100% respiratory efficiency and a typical echo train length of 10).

Repeat Studies

The repeat studies for each subject were performed in an identical manner. The imaging plane for the dark blood cross-sectional right coronary acquisitions was positioned the same distance from the aortic root using screen grabs and tracings from the initial studies (see Fig. 1).

Data Analysis

Quantitative comparison of each of the four techniques was performed using the average thickness of the right coronary vessel wall. For each 3D study (both 1RR and 2RR gating), the slice with the sharpest vessel wall from the 4 × 1.5 mm slices which correspond to the 6.0 mm 2D slice was selected visually for comparison. The images were presented for analysis in a randomized order to avoid bias and the decision on which slice to analyze was made independently for each 3D acquisition (1RR and 2RR). All images were Fourier interpolated by a factor of 4 and then zoomed (using bilinear interpolation) for the subsequent analysis. Average thickness measurements were calculated from circular regions of interest manually drawn around the inner and outer vessel wall. The intra- and interobserver reproducibility of this measurement technique was determined in a randomly selected subset of 20 images as the mean ± SD of the signed differences between two measurements made by the same observer and by two independent observers, respectively.

Comparison of vessel wall thickness obtained using all four acquisitions was performed using repeated measures analysis of variance (ANOVA) on the data acquired in the initial studies. Using both the initial and repeat datasets, the interstudy reproducibility of vessel wall thickness measured with each of the techniques was assessed using the intraclass correlation coefficient (ICC) (29) and by the method of Bland and Altman (30).

All selected coronary artery wall images were scored on a 3-point scale (1 = vessel wall not visualized, 2 = adequate, 3 = good) (8). Only those coronary wall images with a score of 2 or 3 were analyzed as described above. Image quality scores were compared between techniques using the Friedman test.

RESULTS

The mean heart rate for all studies was 61 ± 8 bpm (range 46–73 bpm) and the mean duration of the diastolic right coronary rest periods was 119 ± 59 msec (range 50–226 msec). During the initial studies, a fault was found in the TSE sequence which caused the scanner to ignore the phase oversampling if the echo train length was modified. As a result, four of the TSE acquisitions in subjects with longer rest periods were acquired without the prescribed 77% phase oversampling. In three of these subjects the image quality was high regardless and they were included in the subsequent analysis. The image quality was poor in the fourth subject but this was primarily attributed to cardiac motion, as discussed below, rather than to reduced SNR. The fault was resolved prior to beginning the repeat studies and all further TSE acquisitions were performed with the phase oversampling. Example images from all techniques obtained in the initial study of one subject are shown in Fig. 3, including the four reconstructed 1.5 mm slices of each of the 3D techniques corresponding to the 6.0 mm slices obtained with the 2D acquisitions.

Figure 3.

A comparison of images obtained in the initial study of an example healthy subject (subject 10) using all four methods. The four 1.5-mm slices (a–d,e–h) of the 3D spiral techniques are shown along with the equivalent single 6.0-mm slices obtained using the 2D TSE (i) and 2D spiral (j) techniques. Both of the 3D spiral techniques with B2B-RMC were acquired with respiratory efficiencies of 100% and the range of diaphragm displacements corrected was 18 mm and 23 mm using 2RR gating and 1RR gating, respectively. The 2D images were acquired with respiratory efficiencies of 45% (2D spiral) and 39% (2D TSE). The measured vessel wall thicknesses were 1.31 mm (3D spiral with 2RR gating), 1.39 mm (3D spiral with 1RR gating), 1.29 mm (2D spiral), and 1.36 mm (2D TSE). In both 3D spiral images a degree of ghosting artifact is observed (as discussed in the text). The effect is worse in the 1RR gated images (scored 2) than in the 2RR gated images (scored 3).

The four acquisitions in each study were performed in a random order and were typically completed within ≈30 minutes. Of the 80 acquisitions performed (10 subjects with four acquisitions on two occasions), 73 (91%) produced images with image quality scores of 2 or 3 from which measurement of the right coronary vessel wall thickness could be achieved. Table 1 documents the success or failure of each acquisition in each subject, together with the heart rate and coronary rest period duration (as estimated from the cine acquisition in the same plane). Four rejected datasets were TSE acquisitions, two were 3D spiral acquisitions with 1RR gating, and one was a 3D spiral acquisition with 2RR gating. In one subject (subject 9), the failed acquisitions appeared to correspond to an increase in heart rate (discussed fully below). For the remaining subjects, however, no changes in heart rate were observed during the four acquisitions in each study and the success or failure of an acquisition was not dependent on the order in which it was performed. Of the successful datasets, 42 were scored 2 and 31 were scored 3. Of these 31 datasets that were scored 3, eight were 3D spiral acquisitions with 2RR gating, five were 3D spiral with 1RR gating, eight were 2D spiral, and 10 were 2D TSE acquisitions. There was no significant difference in image quality score between techniques (P = ns).

Table 1. Success/Failure of the High Resolution Coronary Artery Vessel Wall Acquisitions
SubjectInitialRepeat
Heart rate (BPM)Rest period (msec)3D spiral 2RR3D spiral 1RR2D spiral 2RR2D TSE 2RRHeart rate (BPM)Rest period (msec)3D spiral 2RR3D spiral 1RR2D spiral 2RR2D TSE 2RR
  • ✓, a

    Successful acquisition where vessel wall thickness could be measured (scored 2 or 3).

  • Unsuccessful acquisition where the vessel wall could not be distinguished sufficiently to enable a thickness measurement (scored 1).

  • a

    Studies acquired without the prescribed 77% phase oversampling (see text for details).

151172a6846
26175a5885
365856750
457174a56226
547223a5597
672506890
76019046182
85915366101
973507283
105414659110

Acquisition Duration and Respiratory Efficiency

For each of the techniques, Table 2 gives the respiratory efficiency, acquisition duration, acquisition duration per slice, and range of diaphragm motion in the initial studies. A one-tailed paired t-test was used to compare the respiratory efficiency of the B2B-RMC techniques with the navigator-gated techniques, indicating that the B2B-RMC technique is significantly more efficient (99.6 ± 1.2% vs. 39.0 ± 7.5%, respectively, P < 0.0001). In addition, the much lower SD of the respiratory efficiencies indicates that the B2B-RMC technique is substantially less variable than the navigator technique. While the acquisition duration of the 2D spiral technique was shorter than all other techniques, the acquisition duration per reconstructed slice of the 3D techniques is considerably less. On average the B2B-RMC technique corrected for 20 ± 6 mm of diaphragm motion in the initial study (range 15–36 mm) in contrast to the 5 mm nominal gating window used in the navigator gated acquisitions.

Table 2. Overview of Respiratory Efficiency, Duration, and Related Parameters in the Initial Study by Imaging Technique
Method3D spiral 2RR3D spiral 1RR2D spiral 2RR2D TSE 2RR
  • a

    P < 0.0001 when comparing the respiratory efficiency of the navigator gated techniques with the B2B-RMC techniques.

  • b

    Nominal navigator acceptance window.

  • c

    P < 0.001, 2D TSE vs. 2D spiral, 2D spiral vs. 3D spiral 2RR, 3D spiral 1RR vs. 3D spiral 2RR; P < 0.01, 2D TSE vs. 3D spiral 1RR; P < 0.05, 2D spiral vs. 3D spiral 1RR; P = ns, 2D TSE vs. 3D spiral 2RR.

  • d

    P < 0.001, 2D TSE vs. 2D spiral, 2D TSE vs. 3D spiral 1RR, 2D TSE vs. 3D spiral 2RR, 3D spiral 1RR vs. 3D spiral 2RR; P < 0.01, 2D spiral vs. 3D spiral 1RR, 2D spiral vs. 3D spiral 2RR.

  • e

    The acquisition duration for the 2D TSE technique was corrected for the 77% phase oversampling error. The variation between subjects therefore reflects differences in echo train length and respiratory efficiency.

Respiratory efficiency (SD)a (%)99.5 (1.6)99.8 (0.3)39.0 (9.4)39.3 (4.6)
Diaphragm motion compensated (SD) (mm)21.3 (6.9)18.4 (6.0)5.0b5.0b
Reconstructed slices (number × thickness (mm))16 × 1.516 × 1.51 × 6.01 × 6.0
Total acquisition durationc (SD) (RR intervals)603 (10)301 (1)209 (59)479e (74)
Acquisition duration per reconstructed sliced (SD)(RR intervals)37.7 (0.6)18.8 (0.1)209 (59)479e (74)
N109108

Vessel Wall Thickness

Bland–Altman plots for the intra- and interobserver reproducibility of the vessel wall thickness measurements obtained in a subset of 20 images are shown in Fig. 4. The mean intraobserver difference was 0.04 ± 0.09 mm and the mean interobserver difference was −0.05 ± 0.08 mm.

Figure 4.

Bland–Altman plots showing the intra- and interobserver reproducibility of the method used to measure the coronary vessel wall thickness. The intra- and interobserver differences were (mean ± SD) 0.04 ± 0.09 mm and −0.05 ± 0.08 mm, respectively.

Vessel wall thickness measurements were obtained in 37 datasets from the initial study (Table 3). The average vessel wall thickness for all techniques was 1.11 ± 0.19 mm with repeated measures ANOVA showing no significant differences between techniques.

Table 3. Vessel Wall Thicknesses Obtained in the Initial Study
Method3D spiral 2RR3D spiral 1RR2D spiral 2RR2D TSE 2RRP-value
Thickness (SD) (mm)1.10 (0.14)1.20 (0.16)1.14 (0.15)1.21 (0.17)ns
N109108

Reproducibility

The mean time between the initial and repeat studies was 37 ± 16 days (range 28–77 days). In one subject (subject 3), the heart rate at the time of the repeat scan was faster than during the initial scan and there was no identifiable coronary rest period. The study was consequently abandoned and repeated at a later date when the heart rate was similar to that in the initial study. Initial and repeat images (29 days apart) in an example subject are shown in Fig. 5. Vessel wall thickness measurements were made in 36 repeat datasets. The results of the reproducibility tests, comparing the initial and repeat studies for each imaging technique, are given in Table 4 and the associated Bland–Altman plots are shown in Fig. 6. The ICCs ranged from good for the 3D spiral with 1RR gating (0.65) to excellent (0.87) for the 3D spiral with 2RR gating acquisition (31). The SD of the signed differences in vessel wall thickness between studies ranged from 0.10 mm (3D 2RR) to 0.14 mm (2D TSE and 3D 1RR).

Figure 5.

The initial (a–d) and repeat (e–h) results obtained using each method in an example healthy volunteer (subject 4). The single slices chosen for quantitative vessel wall thickness comparisons from the 3D spiral techniques are shown here. There is a high degree of visual similarity between the initial and repeat images (acquired 29 days apart). Respiratory efficiencies and measured vessel wall thickness, on the top row from left to right, are: 100% 1.09 mm, 100% 1.29 mm, 40% 1.34 mm, and 38% 1.35 mm. On the bottom row, also from left to right, the respiratory efficiencies and measured vessel wall thicknesses are: 100% 1.12 mm, 100% 1.02, 55% 1.05 mm, and 40% 1.05 mm. In the initial study the B2B-RMC technique corrected for 14 mm and 20 mm of diaphragm motion when acquired with 2RR and 1RR gating, respectively. In the repeat study the B2B-RMC technique corrected for 18 mm of diaphragm motion when acquired with both 1RR and 2RR gating. The 2D TSE acquisition in the initial study was one of the four studies acquired without the 77% phase oversampling, resulting in a lower SNR. Despite this, image quality is good. 3D spiral images with 2RR gating have lower SNR than those with 1RR gating due to incomplete magnetization recovery between sequence repeats.

Figure 6.

Bland–Altman plots for the reproducibility comparison between imaging techniques. The reproducibility was highest for (a) the 3D spiral technique with 2RR gating: 0.01 ± 0.10 mm, followed by (c) the 2D spiral technique: 0.06 ± 0.12 mm, (d) the 2D TSE technique: 0.06 ± 0.14 mm, and finally (b) the 3D spiral with 1RR gating (the same SD on the vessel wall thickness differences as the TSE technique but a lower ICC): −0.01 ± 0.14 mm.

Table 4. Interstudy Reproducibility
Method3D spiral 2RR3D spiral 1RR2D spiral 2RR2D TSE 2RR
Mean difference (SD) (mm)0.01 (0.10)−0.01 (0.14)0.06 (0.12)0.06 (0.14)
Intraclass correlation0.870.650.700.72
N98106

DISCUSSION

B2B-RMC enables the acquisition of high-resolution 3D images of the coronary vessel wall with near 100% respiratory efficiency. For both 2RR and 1RR gating, the vessel wall thicknesses measured using the 3D spiral with B2B-RMC technique are not significantly different from those measured with conventional navigator techniques and the reproducibility was good (with 1RR gating) or excellent (with 2RR gating). The B2B-RMC technique compensated for 20 ± 6 mm of diaphragm motion, which results in much improved and less variable respiratory efficiency compared with conventional navigator techniques (99.6 ± 1.2% vs. 39.0 ± 7.5%, respectively, P < 0.0001). This enables 3D coverage of the vessel with high resolution in a predictable and clinically feasible imaging time. In a patient cohort the advantages of 3D imaging and B2B-RMC in coronary vessel wall imaging are expected to be even more evident and a direct comparison with navigator-gated 3D spiral imaging would be of interest. The reduction in acquisition duration using B2B-RMC should improve patient cooperation and may permit 3D vessel wall images to be obtained within a routine MR coronary angiography exam. The thinner contiguous slices obtained in this way result in reduced partial volume effects and potentially enable assessment of through-plane changes in plaque shape.

Overall, 91% of the 80 acquisitions performed were successful. Of the seven studies that failed, four were TSE acquisitions, two were B2B-RMC 3D spiral acquisitions with 1RR gating, and one was a B2B-RMC 3D spiral acquisition with 2RR gating. Although the TSE sequence resulted in the highest number of images scored 3, it also result in the largest number of failed studies (scored 1). The TSE sequence is particularly sensitive to motion and an accurate assessment of the coronary rest period is imperative, as has been demonstrated in previous studies (32, 33). In this study the mean heart rate in the failed TSE acquisitions (subjects 2 and 6 (initial studies) and subjects 1 and 9 (repeat studies)) was higher than in the successful TSE acquisitions (68 ± 5 bpm vs. 59 ± 8 bpm, P < 0.05) and the visually estimated coronary rest period was shorter (57 ± 18 msec vs. 133 ± 64 msec, P < 0.05). Retrospective quantitative analysis of the corresponding cine images in the plane of interest using local normalized cross-correlation was performed to determine the true coronary rest period. This demonstrated that in the failed acquisitions the rest period had been overestimated in the visual assessment and that the implemented echo train length was too long. One of these acquisitions (subject 2 (initial study)) was not imaged with 77% phase oversampling due to the sequence error previously discussed, but this is not thought to be a key reason for the poor image quality. The first of the rejected 3D spiral acquisitions with 1RR gating was acquired in a subject (subject 7 (initial study)) with a dominant left coronary artery system and, hence, a narrow right coronary artery (luminal diameter = 0.7 mm and outer vessel diameter = 2.7 mm, measured on the 3D spiral with 2RR gating images). A combination of the small vessel size and the reduced SNR available with 1RR gating resulted in insufficient contrast between the lumen and vessel wall to perform the thickness measurement. The equivalent repeat study was successful, possibly due to subtle changes in the slice orientation and dark blood reinversion slab location, which may have improved the definition between the vessel wall and lumen. The second failed 3D spiral acquisition with 1RR gating and the only failed 3D spiral acquisition with 2RR gating were obtained in subject 9 (repeat study) where only the 2D spiral acquisition, which was performed first, was successful. The subject reported feeling uncomfortable during the last three acquisitions, which were those that failed. The rest period was visually estimated at 83 msec from the cine acquisition performed immediately before the 2D spiral imaging, but subsequent quantitative analysis of the cine acquisition demonstrated that the actual rest period was only 41 msec. The subject's discomfort was accompanied by a heart rate increase of ≈5 bpm and a resultant reduction in the already short rest period duration is thought to be the cause of the acquisition failures.

Small errors in determining the onset and duration of the coronary rest period have greatest impact in subjects with short rest periods. While we estimated the onset and duration of the coronary rest period from a cine acquisition in the same plane as the vessel wall image, subsequent quantitative analysis of the motion of the artery showed that in some cases this was inaccurate. Automatic determination of rest period (34), improved temporal resolution of the cine acquisitions (currently 20 reconstructed frames/RR interval, actual temporal resolution 42 msec) and/or repeat cine acquisitions for determining the rest period could potentially improve image quality and reduce the number of failed acquisitions.

The B2B-RMC 3D spiral technique with 2RR gating was highly successful. Only one acquisition failed of 20 performed (subject 9, repeat study, as discussed above), 95% were scored 2 or 3, the respiratory efficiency was close to 100%, and the reproducibility measures were excellent. While the 3D spiral technique with 1RR gating allows a factor of 2 reduction in imaging time, the reproducibility was lower and problems imaging small arteries due to the reduced SNR were observed. In addition, this sequence is more sensitive to heart rate variations (due to less complete longitudinal magnetization recovery between sequence repeats) which can potentially lead to ghosting and poor dark blood suppression. The generally reduced image quality in the 3D spiral 1RR acquisitions (see Figs. 3, 6) is reflected in the poorer image quality scores and is due primarily to lower SNR. In addition, the 3D spiral images in Fig. 3 exhibit some ghosting from a section of the chest wall which is reinverted by the 180° pulse of the navigator. As the chest wall does not move rigidly with the coronary artery, the retrospective motion correction results in ghosting artifact from this reinverted section. This effect is generally worse in 1RR gated images where the navigator, which is near the end of the cardiac cycle, is temporally close to the dark blood preparation of the next sequence repeat. In the example shown in Fig. 3, the effect is further enhanced in the 1RR gated images (scored 2) when compared to the 2RR gated images (scored 3) as the mean respiratory amplitude was larger (maximum diaphragm displacement 23 mm vs. 18 mm, respectively). The section of chest wall reinverted in the subject shown in Fig. 6 was sufficiently far from the right coronary artery to avoid this artifact. An alternative approach to reducing the overall imaging time of 2RR gated acquisitions would be to increase the cardiac acquisition window (currently only 35 msec) in subjects with longer rest periods by acquiring more spiral interleaves after each low-resolution volume and this is currently under investigation.

The vessel wall thickness measured will clearly depend on the acquired spatial resolution. In this work, using an in-plane resolution of 0.7 × 0.7 mm, the average vessel wall thickness in the initial studies across all acquisitions was 1.11 ± 0.19 mm. This lies within the range obtained in healthy volunteers using MR by other groups (0.75 ± 0.17 mm (4), 1.0 ± 0.2 mm (20), 1.14 (range 0.63 – 1.74) mm (18), 0.79 ± 0.23 mm (19), 1.0 ± 0.2 mm (5) and 1.6 ± 0.2 mm (35)).

Vessel wall thickness measurements obtained using circular regions of interest drawn around the inner and outer vessel are reproducible (intra- and interobserver differences in wall thickness measurements were 0.04 ± 0.09 mm and −0.05 ± 0.08 mm, respectively). For comparison, in a previous study of reproducibility (18) using multiple cross-sectional 2D TSE images of the right coronary vessel wall a region of interest-based method was used to obtain intra- and interobserver differences in wall thickness measurements of −0.07 ± 0.10 mm and −0.02 ± 0.20 mm, respectively. The interstudy difference from the same work (18) was 0.02 ± 0.20 mm, which also compares well to the values of 0.01 ± 0.10 mm (3D spiral with 2RR gating), −0.11 ± 0.14 mm (3D spiral with 1RR gating), 0.06 ± 0.12 mm (2D spiral), and 0.06 ± 0.14 mm (2D TSE) found here. In addition, the ICCs closely agree with those obtained by Desai et al (35) (0.86) using in-plane 3D spiral imaging.

While the lack of a comparison of the 3D spiral B2B-RMC technique with a navigator-gated 3D technique could be seen as a limitation of this study; the poor respiratory efficiency of navigator gating made this unfeasible at the desired resolution. At the average navigator efficiency of 39%, for example, a similar 3D spiral study with navigator gating would have taken 1535 cardiac cycles or ≈25 minutes to acquire (at a heart rate of 60 bpm). The navigator-gated techniques used here for comparison were therefore 2D techniques, as are frequently used in assessing the coronary artery wall. All subjects in this work were healthy volunteers, but further studies will apply the 3D spiral technique with 2RR gating in a patient cohort with known coronary artery disease with the aim of showing increased vessel wall thickness and 3D imaging of localized atherosclerotic plaques. In addition, all acquisitions in this study were performed in the right coronary artery. While this is the more mobile artery—and hence the more challenging to image—its proximity to the surface coil is beneficial. Improvements in SNR at 3T and the use of improved receive coils will be important for further improvements in spatial resolution and for targeting the left coronary artery, which has a more posterior location. Key to future developments is a study into the optimal spatial resolution of the low-resolution volumes and implementing additional techniques to reduce the time required to obtain the low-resolution data, including parallel imaging and/or compressed sensing techniques (36). Following further optimization, the B2B-RMC technique should progress to performing a rapid online 3D translational prospective motion correction followed by a finer retrospective nonrigid correction. High-resolution in plane 3D spiral coronary artery wall imaging (5, 21, 35) has been demonstrated with navigator gating and the B2B-RMC technique with nonrigid capabilities would be well suited to reducing the duration of this type of acquisition. Furthermore, although we have presented the B2B-RMC technique here applied only to 3D spiral coronary artery acquisitions, it has potential benefits in many other 3D cardiac MR applications.

In conclusion, high-resolution dark blood prepared 3D spiral imaging with B2B-RMC permits highly reproducible assessment of the coronary vessel wall thickness when performed with 2RR cardiac gating. Furthermore, the significantly and substantially improved respiratory efficiency of the B2B-RMC technique over standard navigator gating with a 5-mm acceptance window permits acquisition of a 3D high-resolution volume in a reasonable time scale. This will enable full 3D assessment of coronary plaque morphology in patient cohort.

Acknowledgements

This project was undertaken at the NIHR Cardiovascular Biomedical Research Unit at the Royal Brompton and Harefield NHS Foundation Trust and Imperial College, London.

Ancillary