Author + information
- Received November 28, 1998
- Revision received March 19, 1999
- Accepted April 23, 1999
- Published online August 1, 1999.
- Matthias Stuber, PhD∗,‡,* (, )
- René M. Botnar, PhD∗,‡,
- Peter G. Danias, MD, PhD∗,
- Daniel K. Sodickson, MD, PhD∗,
- Kraig V. Kissinger, RT, MS∗,
- Marc Van Cauteren, PhD§,
- Jan De Becker, PhD§ and
- Warren J. Manning, MD∗,†
- ↵*Reprint requests and correspondence: Dr. Matthias Stuber, Beth Israel Deaconess Medical Center, Cardiovascular Division, 330 Brookline Ave., Boston, Massachusetts 02215
The goal of the present study was to develop a strategy for three-dimensional (3D) volume acquisition along the major axes of the coronary arteries.
For high-resolution 3D free-breathing coronary magnetic resonance angiography (MRA), coverage of the coronary artery tree may be limited due to excessive measurement times associated with large volume acquisitions. Planning the 3D volume along the major axis of the coronary vessels may help to overcome such limitations.
Fifteen healthy adult volunteers and seven patients with X-ray angiographically confirmed coronary artery disease underwent free-breathing navigator-gated and corrected 3D coronary MRA. For an accurate volume targeting of the high resolution scans, a three-point planscan software tool was applied.
The average length of contiguously visualized left main and left anterior descending coronary artery was 81.8 ± 13.9 mm in the healthy volunteers and 76.2 ± 16.5 mm in the patients (p = NS). For the right coronary artery, a total length of 111.7 ± 27.7 mm was found in the healthy volunteers and 79.3 ± 4.6 mm in the patients (p = NS). Comparing coronary MRA and X-ray angiography, a good agreement of anatomy and pathology was found in the patients.
Double-oblique submillimeter free-breathing coronary MRA allows depiction of extensive parts of the native coronary arteries. The results obtained in patients suggest that the method has the potential to be applied in broader prospective multicenter studies where coronary MRA is compared with X-ray angiography.
Two-dimensional (2D) breath-hold coronary magnetic resonance angiography (MRA) has been shown to be a promising and valuable method for assessment of the native coronary arteries (1–4). However, three-dimensional (3D) methods in which data are only collected in a small fraction of the cardiac cycle have inherent advantages, including minimizing bulk cardiac motion, superior signal-to-noise ratio (SNR), the acquisition of thin contiguous sections and the ability to postprocess/reformat the data set in 3D (5–7). Although free-breathing 3D coronary MRA has been applied successfully (6,8), measurement time is considerable and coverage of the full coronary artery tree may still be limited due to excessive measurement times associated with large volume acquisitions.
By expeditiously planning the 3D volume along the major axis of the coronary vessels (9,10), coronary coverage may be more optimal. Such a targeted approach allows acquisition of smaller 3D volumes which nevertheless covers the major extent of the coronary system and which depicts the arteries themselves with high in-plane spatial resolution.
In the present study, we present a novel approach for defining the plane of the relevant coronary artery anatomy using a fast, 3D free-breathing navigator sequence for scout scanning. In conjunction with a three-point planscan software tool, the double-oblique imaging plane oriented in parallel to the major coronary arteries is then defined. The technique has been applied to 15 volunteers and seven patients with X-ray angiographically defined coronary artery disease.
The study population included 22 participants, 15 healthy adult volunteers (nine male, age: 28 ± 6 years) and seven patients (six male, age: 65 ± 8 years) with X-ray angiographically confirmed coronary artery disease. No patient had undergone prior coronary revascularizations. All participants were in sinus rhythm and were without contraindications to magnetic resonance (MR) imaging. Written informed consent was obtained from all participants, and the protocol was approved by the hospital Committee on Clinical Investigation.
All participants were examined supine with electrocardiographic (ECG) leads on the anterior left hemithorax. All scans were performed during free breathing. For coronary artery localization and for navigator positioning at the right hemidiaphragm (RHD), two scout scans were utilized. The first scout was an ECG triggered, free-breathing, multislice 2D segmented gradient echo (repetition time [TR] = 11 ms, echo time [TE] = 2.4 ms, 256 × 128 matrix, 450 mm field of view) scan with nine transverse, nine coronal and nine sagittal interleaved acquisitions of the thorax. The total duration of this acquisition was 1 min. On these images, the navigator position at the RHD dome as well as the localized transverse 3D volume for the subsequent scout scan were planned.
Three-dimensional turbo field echo-echo planar imaging (TFE-EPI) scout scan
For the second scout, the MR data were acquired in diastole and at end-expiration using navigator gating and real-time motion correction (11,12)about a volume which included the coronary arteries as defined in the prior scout (Fig. 1). One TFE-EPI shot (13)included five radiofrequency excitations, each followed by 11 EPI readouts. One shot was applied per cardiac cycle, resulting in an acquisition window of <70 ms (TR = 8.8 ms, TE = 5.3 ms). A transverse 3D slab covering a volume of 30 mm thickness was scanned with an in-plane resolution of 1.3 × 1.8 mm. A total of 30 overlapping slices with a reconstructed slice thickness of 3 mm were acquired. For contrast enhancement between blood and myocardial muscle, a T2 preparation (T2-Prepulse) (14,15)preceded the navigator and the imaging portion of the sequence (Fig. 2).
Three-point planscan technique
The three-point planscan tool is part of the extended cardiac software package (CPR6) on the Philips (Best, The Netherlands) Gyroscan ACS-NT system. On the reconstructed images obtained with the 3D TFE-EPI scout scan, transverse images of three anatomic levels (Fig. 3A, B and C)are loaded into the display of the three-point planscan tool. For the right coronary artery (RCA), transverse images depicting a proximal, a mid- and a distal segment of the RCA are selected (Fig. 3). The user specifies the RCA in each image with an interactive mouse click. These three points define the geometry of the center plane of the imaged 3D volume. The geometric parameters of this plane (three angulations and three off-center values) are subsequently transferred to the scan parameters of the user interface. For the left coronary system, segments representing the bifurcation of the left main (LM), the mid-left anterior descending (LAD) and the proximal left circumflex coronary arteries (LCX) were identified. With this definition, the software prescribes the plane defined by the LM, the mid-LAD as well as the proximal/mid-LCX.
Imaging sequence for 3D coronary MRA
Two ECG gated, free-breathing navigator-gated and corrected double-oblique 3D MRA scans were performed in the patients and volunteers: one scan for the left and one scan for the right coronary system. The 3D coronary MRA sequence with real-time navigator gating and motion correction (16)was implemented on a commercial 1.5-T Gyroscan ACS-NT whole body system (Philips Medical Systems, Best, The Netherlands) equipped with a cardiac software (CPR6) and a PowerTrak 6000 gradient system (23 mT/m, 220 μs rise time). For signal acquisition, a five-element cardiac synergy coil was used.
The 3D segmented k-space TFE sequence (TE = 2.4 ms, TR = 8.8 ms) used a modular structure as displayed in Figure 2. The flow-insensitive T2-Prepulse (14,15)for contrast enhancement is followed by a localized anterior saturation prepulse (Fig. 2, REST), the navigator, the spectrally selective fat saturation pulse and finally the TFE imaging sequence (Fig. 2, HR-Scan). The utilized receiver bandwidth was 135 Hz/pixel. A double-oblique 3D volume was imaged as prescribed by the three-point planscan tool. To compensate for suboptimal slice profiles of the slice selective excitations, a 3.9-cm thick volume was imaged with subsequent oversampling (in slice selection direction) by a factor of 1.3. The 20 center slices with a slice thickness of 1.5 mm each were subsequently reconstructed rejecting the six most peripheral slices. A field of view of 360 mm and a 512 × 360 matrix yielded an in-plane resolution of 0.7 × 1.0 mm. Eight phase encoding steps were sampled during each cardiac cycle (acquisition window = 70 ms). K-space was sampled using centric ordering with priority for the low ky-profiles. One signal average was performed, and no flow-compensating gradients were used. Data acquisition was performed in middiastole.
A vertical 2D selective real-time navigator for gating and motion tracking (11,12,16)was applied at the dome of the RHD (Fig. 1). A navigator excitation angle of 60° and a diameter of 20 mm were used. The gating window for the diaphragm was 5 mm, and a correction factor of 0.6 was used to relate the superior–inferior position of the diaphragm to the superior–inferior position of the proximal coronary arteries (17). For real-time motion correction, the frequency of the slice-selective radiofrequency excitation as well as mixing phase and frequency were adapted based on the detected position (within the gating window) of the lung–diaphragm interface.
Signal-to-noise and contrast-to-noise ratios
At the level of the LM and the proximal RCA, the SNR was quantified in the ascending aorta (16). Hereby, the average signal intensity (IBlood) in a user-defined region of interest (ROI) was divided by the standard deviation (SDEVBlood) of the signal in the same ROI. The SNR of muscle, was determined in the muscle of the left ventricular free wall at the level of the proximal RCA by equation 1. IMuscleis the mean signal intensity in the user-defined ROI on the muscle, and SDEVMuscleis the standard deviation of the signal in this region. Subsequently, the contrast-to-noise ratio (CNR) between blood and muscle was assessed as equation 2: 
For 3D visualization and length measurements of the coronary arteries, the image data sets were transferred to a commercially available EasyVision workstation (Philips Medical Systems, Best, The Netherlands). The three orthogonal sections through the 3D data set are displayed, and the user navigates interactively through the entire data set. The LM, LAD and RCA were manually identified in all the three displayed planes. The 3D pathway of the coronary artery was then multiplanar reformatted, and the lengths of the individual segments of the native coronary arteries was determined manually.
During the high resolution coronary MRA scans, the scan efficiency was recorded in all individuals. Scan efficiency (%) was defined as the number of shots accepted divided by the total number of heart beats used for completing a scan. A scan efficiency of 50%, for example, refers to a scan in which on average, data acquired from every other heartbeat are accepted for reconstruction.
All values are presented as mean ± SD. Comparisons between healthy volunteers and patients were made using an unpaired Student ttest. In all cases, a two-tailed test was performed with p values ≤0.05 considered significant.
Magnetic resonance studies were completed in all participants without complications.
For the RCA (Fig. 4), a double-oblique coronal orientation in parallel to the atrioventricular groove with a left–right angulation of 25.1 ± 9.6° (range 4 to 40°) and a foot–head angulation of 32.6 ± 10.2° (range 7 to 54°) was found in the healthy volunteers using the three-point planscan tool. For patients, these two angles were 4.3 ± 11.2° (p = 0.002, range −8° to +27°) and 34.3 ± 10.3° (p = NS, range 21 to 56°), respectively. For the left coronary system, a transverse orientation with an anterioposterior angulation of 12.5 ± 10.0° (range 0 to 38°) and a left–right angulation of 2.5 ± 5.7° (range −13° to +17°) was found in the healthy volunteers. The corresponding angles for the left coronary system in patients were 5.2 ± 3.5° (p = NS, range 0 to 10°) and 5.0 ± 4.0° (p = NS, range 0 to 17°).
In Figure 5, a representative example showing the RCA and the LCX of a healthy volunteer is displayed. The images A to D refer to four adjacent (out of a total of 20) 1.5-mm thick sections. Major portions of this RCA and the LCX are oriented in-plane and displayed with a resolution of 0.7 × 1.0 mm. In Figure 5, E, the 3D planar reformatted version of this data set is presented. In Figure 6, the 3D planar reformatted right and left coronary arterial systems from a healthy volunteer are displayed. In Figure 6, A, extensive portions of the RCA including smaller branch vessels are displayed. In addition, the same figure also shows the LM and the LCX. Figure 6Bdisplays the LM, the LAD, the first diagonal and the proximal LCX. In Figure 7, four healthy volunteer cases are presented. The RCA is displayed in the upper panel (A–D) and the left coronary system of the same individuals in the lower panel (E–H). In all cases long portions of the coronaries and many first-order branches of the left coronary system are clearly visible. In Figure 8, image A shows an RCA of a patient. In image B, a zoomed region of interest of image A is displayed together with the X-ray angiogram (C) of the same patient. In agreement with the X-ray angiogram, the stenoses can clearly be localized on the coronary MRA. In Figure 9, the same situation is displayed for another patient with X-ray angiographically confirmed RCA disease. In this case, a 30% proximal RCA stenosis (dotted arrow) and a 90% mid-RCA stenosis (plain arrow) were reported from the X-ray angiogram (Fig. 9B). Figure 10shows the results for a patient with left coronary disease. Localized stenoses (arrows) were seen on the X-ray angiogram (B). The same stenoses are demonstrated on the coronary MRA (A).
The average length of the LM and the LAD was 81.8 ± 13.9 mm in the healthy volunteers and 76.2 ± 16.5 mm in the patients (p = NS). For the RCA, a total length of 111.7 ± 27.7 mm was found in the healthy volunteers and 68.0 ± 12.9 mm in the patients (p < 0.05). Three patients had total occlusion of the proximal RCA. If the three patients with a 100% proximal RCA stenosis are omitted, a total length of 79.3 ± 4.6 mm was found in the patients (p = NS vs. healthy volunteers).
Signal-to-noise and contrast-to-noise ratios
In the healthy volunteers, an SNR of 12.3 ± 2.6 was measured on the images showing the RCA, whereas 11.9 ± 2.7 was found in the patients at the same location (p = NS). For the double-oblique images showing the left coronary system, a SNR of 10.2 ± 1.0 was found in the volunteers and 9.0 ± 1.2 in the patients (p = NS). The signal of myocardium is markedly reduced when compared with blood. This could be observed in the patients (CNR left: 5.9 ± 1.4, CNR right: 8.0 ± 1.2) as well as in the volunteers (CNR left: 6.2 ± 1.6, CNR right: 9.7 ± 2.0) (p = NS).
Combined (left and right coronary system), scan efficiency was 60% in healthy volunteers and 43% in patients (p < 0.01). No significant difference between the left and right coronary system scan efficiency was found between patients or healthy volunteers (p = NS). Average scanning time per double-oblique high resolution 3D scan was less than 16 min.
Utilizing the double-oblique approach presented in this report, extensive portions of the left and right coronary arterial systems can reliably be acquired in patients and healthy volunteers using free-breathing high resolution 3D coronary MRA. The current images and the high values found for SNR and CNR demonstrate excellent contrast between muscle and blood, which allows for a good depiction of the native coronary arteries in the healthy as well as in cases with angiographically documented pathology. In comparison with earlier studies (16), we found an increase of more than 30% for SNR and CNR. These improvements may be explained by the low receiver bandwidth utilized for the present study.
Due to the application of the T2-Prepulse, the blood–muscle contrast does not primarily depend on inflow, therefore allowing for acquisition of imaging planes that are nonorthogonal to the blood flow. Major portions of the coronary arteries including branches are parallel to the imaging plane and thus displayed with the high in-plane resolution. Lower through-plane resolution minimally affects the reconstruction and thus, locally reduced resolution in the reconstructed images due to anisotropic voxel size may be avoided. No increase of the imaged volume and thus no increase in scanning time were needed for enhanced coverage of the coronary artery tree. With a purely transverse 3D volume orientation, an increase of more than 100% in scanning time would be required to cover the same extent of the right coronary system as found in the present study. For the left coronary system, a 20% increase in scanning time would be required utilizing a transverse 3D slab.
Coordinate system of the native coronary system
The advantages of the targeted double-oblique acquisition strategy result in part from the fact that the left and right coronary arteries lie predominantly in the approximate planes of the intraventricular septum and of the atrioventricular groove, respectively. Thus, these anatomically defined planes constitute a natural coordinate system for coronary MRA. Our data suggest, however, that the particular orientations of the optimal imaging planes are patient specific. As a consequence, scout-scanning techniques such as the three-point plan are required for coronary MRA to maximize the patient-specific length of the vessels oriented in-plane.
For a successful application of the three-point planscan software tool, a sufficient image quality is required for the localization of the double-oblique 3D volume. By the application of fast 3D imaging techniques such as TFE-EPI for scout scanning, the relevant portions of the coronary arteries can be identified very efficiently (2 min scan duration during free breathing) and reliably.
Three-dimensional techniques allow for the acquisition of thin and contiguous sections. Motion artifacts during the rather long scanning time may degrade the quality of the slice profiles of such thin sections. Such effects may be minimized by the application of prospective adaptive volume tracking (11). Although the slice profile of the 3D slab may be compromised with respect to 2D techniques, the excitation of a thicker slab with subsequent oversampling in slice selection direction minimizes the adverse effect on image quality if the most peripheral slices are rejected during reconstruction. This approach, however, is associated with a prolongation of measurement time. For the present study, where no dynamic information is assessed, the duty cycle is also relatively low, because the 3D data are acquired in only a small fraction of the RR interval. However, this also allows taking maximum advantage of the potential SNR improvement associated with 3D acquisition techniques.
By a visual comparison of the X-ray angiograms and the coronary MR angiograms, anatomy and localization of the stenoses could be successfully confirmed on the MR angiograms. Anatomic as well as pathologic details are confirmed in the MR images with an objective, quantitative description of the pathology remaining to be defined. The accuracy of this methodology for identification of coronary stenoses needs to be investigated in large multicenter clinical studies.
In the patients, the average visible length of the RCA was lower than in healthy volunteers. This might be explained by the total occlusion of the proximal RCA (three cases) and a 90% stenosis of the mid-RCA (two cases) which was reported from the X-ray angiograms.
Double-oblique submillimeter free-breathing 3D coronary MRA is a robust technique that has been successfully applied in healthy volunteers and in patients with X-ray angiographically documented coronary disease. The native left and right coronary systems can be displayed with a good contrast, and the 3D reconstructed vessels are seen with high in-plane resolution. Disadvantageous effects of anisotropic voxel sizes in the reconstructed images are suppressed using the double-oblique approach, and the coverage of the coronary artery tree may be enhanced without an increase in scanning time.
The orientation of the left and right coronary system is patient specific. Therefore, accurate localization strategies such as the three-point planscan technique in conjunction with the navigator-gated and motion-corrected TFE-EPI scout scanning are crucial to take full advantage of double-oblique imaging.
The good agreement between the information obtained on the X-ray angiograms and coronary MRA suggest that the currently proposed free-breathing double-oblique technique has the potential to be applied in broader multicenter prospective patient studies, in which coronary MRA is compared with X-ray angiography.
- contrast-to-noise ratio
- left anterior descending coronary artery
- left circumflex coronary artery
- left main coronary artery
- magnetic resonance
- magnetic resonance angiography
- right coronary artery
- right hemidiaphragm
- region of interest
- signal-to-noise ratio
- echo time
- turbo field echo-echo planar imaging
- repetition time
- Received November 28, 1998.
- Revision received March 19, 1999.
- Accepted April 23, 1999.
- American College of Cardiology
- Manning W.J.,
- Li W.,
- Boyle N.G.,
- Edelman R.R.
- Pennell D.J.,
- Keegan J.,
- Firmin D.N.,
- Gatehouse P.D.,
- Underwood S.R.,
- Longmore D.B.
- Botnar R.M.,
- Stuber M.,
- Danias P.G.,
- Kissinger K.V.,
- Manning W.J.
- ↵Stuber M, Botnar RM, Danias PG, Kissinger KV, Manning WJ. Submillimeter 3D coronary MRA using real-time navigator correction: comparison of navigator locations. Radiology (in press).