Soap-Bubble Visualization and Quantitative Analysis of 3D Coronary Magnetic Resonance Angiograms

Transcription

1 Magnetic Resonance in Medicine 48: (2002) Soap-Bubble Visualization and Quantitative Analysis of 3D Coronary Magnetic Resonance Angiograms Alex Etienne, 1,3 René M. Botnar, 1,2 Arianne M.C. van Muiswinkel, 2 Peter Boesiger, 3 Warren J. Manning, 1 and Matthias Stuber 1,2 * In order to compare coronary magnetic resonance angiography (MRA) data obtained with different scanning methodologies, adequate visualization and presentation of the coronary MRA data need to be ensured. Furthermore, an objective quantitative comparison between images acquired with different scanning methods is desirable. To address this need, a software tool ( Soap-Bubble ) that facilitates visualization and quantitative comparison of 3D volume targeted coronary MRA data was developed. In the present implementation, the user interactively specifies a curved subvolume (enclosed in the 3D coronary MRA data set) that closely encompasses the coronary arterial segments. With a 3D Delaunay triangulation and a parallel projection, this enables the simultaneous display of multiple coronary segments in one 2D representation. For objective quantitative analysis, frequently explored quantitative parameters such as signal-to-noise ratio (SNR); contrast-to-noise ratio (CNR); and vessel length, sharpness, and diameter can be assessed. The present tool supports visualization and objective, quantitative comparisons of coronary MRA data obtained with different scanning methods. The first results obtained in healthy adults and in patients with coronary artery disease are presented. Magn Reson Med 48: , Wiley-Liss, Inc. Key words: coronary MRA; postprocessing; reformatting; quantitative coronary analysis; analysis package; coronary angiography; quantitative angiography Recent studies suggest that coronary magnetic resonance angiography (MRA) is emerging as a valuable tool for the assessment of proximal to mid coronary artery integrity (1). However, MR scanner hardware, software, and methodology are under constant development. As a consequence, a growing number of coronary MRA imaging methods, including steady-state free precession (SSFP) (2), echo-planar imaging (EPI) (3 5), fast spin echo (6), spiral (7,8), and more conventional segmented k-space gradientecho sequences with (9 11) and without (12) the application of exogenous contrast agents, have been and continue to be explored. Furthermore, imaging at different field strengths, parallel imaging (13,14), and motion-compensation strategies such as breath-holding (5,15), prospective (16) or retrospective (17) navigators, and mid- (18) and late-diastolic (19,20) image acquisition with predetermined or patient-specific trigger delays (20) are currently under evaluation. In order to compare data obtained with different approaches, adequate visualization and presentation of the coronary MRA data need to be ensured (21). Furthermore, an objective quantitative comparison between images acquired with different methods is desirable (12). To address this need, we developed a quantitative software tool ( Soap-Bubble ) that facilitates multiplanar reformatting (MPR) of 3D volume-targeted coronary MRA data sets, while also providing frequently explored quantitative measures such as signal-to-noise (SNR) and contrast-tonoise (CNR) ratios, vessel length and diameter, and local vessel sharpness. The first results obtained with this software implementation are presented. METHODS Implementation The Soap-Bubble tool for coronary reformatting and quantitative coronary analysis was implemented under IDL 5.2 (Interactive Data Language; Research Systems Inc., Boulder, CO), on a commercial Microsoft Windows2000 (Microsoft Corp., Redmond, WA) 1.4 MHz Pentium 4 platform (OptiPlex GX400, Dell Computers, Austin, TX) equipped with a three-button mouse (Logitech, Fremont, CA). Reformatting Contemporary coronary MRA techniques include thin-slab 3D volume-targeted acquisitions along the major segments of the left and right coronary arterial systems (5,22). These relatively thin slabs encompass contiguous sections of the major proximal to mid coronary arteries. However, the geometry of the coronary arterial tree together with the tortuous nature of the individual coronary segments generally prevent visualization of the entire coronary anatomy in one single image (Fig. 1). Therefore, multiplanar reformatting (MPR) and maximum intensity projections (MIPs) have been used for improved visualization and quantitative coronary analysis. In the present report, we propose an alternative technique for visualization and subsequent quantitative analysis of 3D coronary MRA data sets. The technique includes 2 Philips Medical Systems, Best, The Netherlands. a user-assisted preselection of a curved subvolume in- 3 Institute of Biomedical Engineering, ETH, Zurich, Switzerland. cluded in the 3D coronary MRA data set. Ideally, this 1 Department of Medicine (Cardiovascular Division), Beth Israel Deaconess Medical Center and Harvard Medical School, Boston, Massachusetts. *Correspondence to: Matthias Stuber, Ph.D., Beth Israel Deaconess Medical curved subvolume includes all the coronary segments and Center, Cardiovascular Division, 330 Brookline Ave., Boston, MA branches enclosed in the acquired volume. All the voxels mstuber@caregroup.harvard.edu that are not included in this user-prescribed subvolume Received 21 January 2002; revised 9 May 2002; accepted 9 May DOI /mrm are discarded and are not used for subsequent visualization and analysis. Published online in Wiley InterScience ( Wiley-Liss, Inc. 658

2 Coronary Reformatting and Analysis Tool 659 FIG. 1. Four individual slices (out of a 10-slice 3D coronary MRA data set) in which adjacent segments of the RCA are seen (arrows). The present implementation is based on the underlying assumption that the coronary anatomy within the acquired volume V (Fig. 2) fits to a relatively smooth 3D surface D. Conceptually, D can be regarded as the surface of a soap bubble. However, the surface D has to be manipulated for a more adequate fit to the local coronary geometry. The definition of this manipulated surface D (Fig. 2) is performed by the user, who identifies a series of points P i (P i x, P i y, P i z) on the coronary arterial tree (included in V) with interactive mouse clicks (i - 1..n). Employing a previously described 3D Delaunay triangulation algorithm (23), the points P 1...P n prescribe the manipulated, curved, convex hull D (Fig. 2). With a parallel projection in the direction of the normal vector N (0, 0, 1), each pixel value in D is subsequently projected on a plane normal to N. A volume (with the skin thickness ds) encompassing D can be defined by the user. With the image data included in this targeted subvolume, an MIP parallel to the normal vector N is subsequently performed (the thickness ds is constant in the direction of N, but not perpendicular to the individual triangles in D ). The resulting 2D image displays a planar reconstruction of the user-selected 3D anatomy, in which true distances are not maintained. User Interface The user interface of the present integrated Soap-Bubble coronary visualization and quantitative analysis tool consists of five graphical viewports, as shown in Fig. 3. After loading the MR data set from the MR scanner ( Load File button), manual coronary reformatting is performed in viewports 1 3, in which three orthogonal sections of the 3D coronary MRA data are displayed. Using an interactive mouse click of the left mouse button in one of viewports 1 3, the user can selectively navigate in this 3D data set. The crosshair moves to the user-specified location and indicates the intersections with the two other orthogonal planes, which are updated and displayed in the two other viewports. This enables the identification of any arbitrary point in the 3D data set. Alternatively, using the Move Up, Move Down, etc., buttons below viewports 1 3, the displayed plane can be moved incrementally for finer adjustment. To facilitate the localization of smaller-diameter structures, the 3D data set can be selectively zoomed ( Zoom In, Zoom Out buttons) and panned (using the Center button, the image is centered at the location of the crosshair) and the image data in viewports 1 3 are updated accordingly. Zooming of the image data is performed using a bilinear interpolation. Using a mouse click of the right button, a coronary point P i can be defined in the 3D image data set (Fig. 3). Previously defined points can be selectively deleted using the middle mouse button. After both definition and removal of a point P i, the reformatted image is updated in viewport 4. Complementary features include adaptations of the window and level of the grayscale images, as well as the option to rotate and mirror the displayed image data. To facilitate viewing of contiguous segments displayed in adjacent slices, the individual slices can be played in a movie mode. Upon completion of the definition of P 1..P n, these data points can be stored on the hard drive of the computer ( Store Points button) or previously defined geometries can be reloaded ( Load Points button) for further processing or analysis. The reformatted image data obtained in viewport 4 can be zoomed, panned, and selectively stored in an electronic GIF, JPEG, or TIFF format. The skin thickness ds (Fig. 2) in which the MIP is performed can be adjusted using the and buttons adjacent to viewport 4. FIG. 2. Soap-Bubble coronary reformatting. 3D coronary MRA data are acquired in a Cartesian coordinate system (x,y,z) and in the volume V. The user-identified points P i define the manipulated surface D, which is shown in a close-up view as a curved (3D Delaunay triangulation), convex hull D. A parallel MIP of this manipulated surface (parallel to the slice-selection direction N) leads to the planar coronary reformat. MIP can selectively be performed in a volume that closely encompasses the coronary arterial tree and is specified by the skin thickness ds.

3 660 Etienne et al. FIG. 3. User interface of the Soap-Bubble tool. The user manually identifies points on the coronary arterial tree (Pi) as visualized in the three orthogonal sections displayed in the viewports 1 3. The reformatted image of an RCA MRA data set (see Fig. 1) is shown in viewport 4, and simultaneously displays the LM and a proximal segment of both the LAD and the LCX with more distal branching vessels (dotted arrow). Quantitative Coronary Analysis Vessel Length Using the Length Measurement mode (Fig. 4, right side of viewport 4), measurements of coronary segment lengths can be performed. A 2D projection Pⴕ1..Pⴕn (Pⴕi (Pix, Piy, 0)) of the user-specified coronary points P1..Pn is graphically overlaid to the reformatted image in viewport 4 (Fig. 4, yellow points). On this reformatted representation of the coronary anatomy, the user identifies a pathway along individual coronary segments, for which length quantification is requested (Fig. 4, blue points overlaid to the right coronary artery (RCA)). A subset Pⴕi..Pⴕk of Pⴕ1..Pⴕn (along the coronary segment on which length measurement is performed) is identified by interactive mouse clicks (1 ⱕ i, k ⱕ n). Since the corresponding 3D representation of Pⴕi..Pⴕk is easily obtained (Pi..Pk), a true 3D length measurement of the individual coronary segments can be calculated and the numerical value of the length of the underlying 3D pathway is subsequently updated on the display (Fig. 4, arrow). Vessel Border, Sharpness, and Diameter Changing the Mode from Length Measurement to Vessel Navigator initiates the computation of the edge image (viewport 5, Fig. 5) and vessel tracking is performed. As earlier described by Botnar et al. (12), the local vessel sharpness can be obtained utilizing a Deriche algorithm (24). In brief, this algorithm calculates an edge image using a first-order derivative of the input image. The local value in a Deriche image represents the magnitude of local change in signal intensity ( vessel sharpness, vessel edge value). In the present implementation of the software, the above-described reformatted coronary is used for vessel sharpness assessment. Vessel sharpness is calculated on the Deriche image normal to the coronary segment Pⴕi..Pⴕk. A vessel sharpness of 100% refers to a maximum signal intensity change at the vessel border. A lower edge value is consistent with inferior vessel sharpness and vice versa. For the identification of the vessel edges along the path Pⴕi..Pⴕk, a semi-automatic vessel tracking algorithm was implemented. In a first step, equidistant points Qj are defined as Qj Pⴕi j * (Pⴕi 1 Pⴕi); ( j j * 0.1; j 0..10). At each position Qj, local image intensity profiles (normal to the vessel) on both the anatomical image IA(l) and the Deriche image ID(l) are determined as shown in Fig. 6. Hereby, l denotes the coordinate along both profiles and G refers to the maximum of the profile IA(l). Full width at half maximum (FWHM) refers to the width of IA(G).

4 Coronary Reformatting and Analysis Tool 661 FIG. 4. Reformatted RCA MRA data (viewport 4) overlaid with the user-specified data points (Pⴕ1..Pⴕn, yellow). Length measurement is performed along the highlighted (blue points) coronary segment, and the numerical value of the highlighted RCA segment length is updated (arrow). Within the search-window defined by FWHM, the two maxima, S1 and S2 on ID(l), are identified at the positions l1 and l2. Hereby, S1 and S2 are the local edge or vessel sharpness values and the distance between l1 and l2 defines the local vessel diameter d. To ensure local measurements of the smallest diameter, a series of intensity profiles with an angular sweep of 30, in steps of 1, are determined at each position Qj. The minimal value found for d as a function of together with its related values S1, S2, l1, and l2 are then used for further analysis. This procedure is repeated until 1, and until the analysis for the entire segment Pⴕi..Pⴕk is completed. Using this algorithm, the location of the vessel border, and the vessel diameter and sharpness of the user-specified coronary segment is defined semi-automatically. The quantitative values along this coronary segment are stored electronically in an ASCII file for further numerical processing and comparisons. SNR and CNR SNR and CNR measurements are initiated using the SNR/ CNR button above viewport 5. SNR analysis can be selectively performed on a user-specified slice of the original 3D data or on the reformatted image. Two regions of interest (ROIs) can be manually defined by the user. Typically, ROIA is positioned in the blood-pool of the ascending aorta or in a coronary segment. ROIB can be localized anterior to the chest-wall in the air. SNR is subsequently determined as shown in Eq. [1], in which the mean signal intensity in ROIA is IBlood and the standard deviation of ROIB is SDEVAir. Both the numerical SNR values and the ROI position/size can be stored and reloaded for future analysis. SNR (I Blood)/SDEV Air [1] In a method similar to that used for the SNR measurements, CNR can be determined using three different ROIs. Typically, one ROI is localized in the ascending aorta or in a coronary segment (ROIA), another ROI is localized in the air anterior to the chest wall (ROIB), and the third ROI is positioned on the myocardium (ROIC). CNR is subsequently defined as shown in Eq. [2], in which the average signal intensities obtained in ROIA and ROIC are denoted as IBlood and IMuscle, respectively. As for SNR measurements, the ROI positions/sizes can be stored and reloaded for future analysis.

5 662 Etienne et al. FIG. 5. Graphical overlays on the user-specified path define the local vessel diameter (viewport 5) together with the location of maximum vessel sharpness along the path (viewport 4). CNR (I Blood I Muscle )/SDEV Air [2] RESULTS Reformatting and quantitative analysis of the RCA takes 3 min, while the same analysis of the left coronary arterial system (left main (LM), left anterior descending (LAD), and left coronary circumflex (LCX)) may take up to 10 min, depending on the number of branching vessels seen on the image (typically, points are needed for the left coronary system, and points for the right coronary system). As shown in Figs. 3, 7, and 8, a simultaneous display of multiple coronary branches is enabled. Original images obtained using a previously described 3D coronary MRA acquisition technique (22) are shown in Fig. 1, in which four slices of a double oblique 3D RCA coronary MRA data set (including 10 slices) are displayed. Despite the volume-targeted acquisition, the geometry of the RCA prevents a contiguous full-length display of the RCA in one slice. However, using the Soap-Bubble reformatting methodology, the full extent of this RCA segment is visualized as documented in viewport 4 of Fig. 3. On this reformat, the RCA, LM, and LCX are displayed simultaneously with more distal branching vessels (dotted arrow). Using the integrated length measurement, a total length of 133 mm was found for this reformatted contiguous RCA segment. The average vessel diameter was 2.8 mm with a vessel sharpness of 63%. SNR was 26.3 and CNR 18.5, respectively. In Fig. 7a, example reformatted coronary MRA data acquired in a healthy adult post-intravenous administration of the blood-pool agent (B22956; Bracco SpA, Milan, Italy) (25) are displayed adjacent to the more conventional T2Prep (Fig. 7b) data obtained in the same subject. Numerical data obtained for the LAD (proximal 4 cm) in this subject include a 2.6-mm vessel diameter with T2Prep (vs. 2.7 mm with contrast agent), a 11.5-cm contiguous length of the LAD (vs with contrast agent), a vessel sharpness of 50% (vs. 60% with contrast agent), and an SNR/CNR of 32/19 (vs. 37/44 with contrast agent). Figure 8 displays Soap-Bubble reformats of 3D coronary MRA data acquired with different acquisition techniques. For all the acquisitions, free-breathing and real-time navigator technology were used for respiratory motion suppression. The video-inverted left coronary arterial system acquired with a dual-inversion fast spin-echo sequence (26) in Fig. 8a shows the LM and 14 cm of the contiguous LAD simultaneously with a septal branch vessel (S1). In Fig. 8b, a left coronary arterial system acquired with an arterial spin labeling technique is shown (27). A simulta-

6 Coronary Reformatting and Analysis Tool 663 FIG. 6. (a) Reformatted RCA data and (b) the corresponding Deriche first-order derivative image on which vessel sharpness is defined. Local cross-sectional intensity profiles of the anatomic data I A (l) and the Deriche image I D (l) are displayed simultaneously. FIG. 7. a: Reformatted coronary MRA of a left coronary arterial system acquired with an inversion technique after intravenous administration of the blood-pool agent B b: A reformatted T2Prep endogenous contrast-enhanced coronary MRA acquired in the same subject.

7 664 Etienne et al. FIG. 8. Reformatted images of 3D coronary MRA data acquired with real-time navigator technology during free breathing. (a) A videoinverted black-blood coronary MRA is displayed adjacent to a spin-tagged acquisition of (b) the left coronary arterial system. c: An RCA together with a left coronary arterial system, including the LM, LAD, LCX, and some smaller-caliber branching segments (using a T2Prep segmented k-space gradient-echo acquisition). d: An anomalous RCA (dashed arrow) from the left coronary cusp (L) acquired with a T2Prep SSFP sequence is shown (R right coronary cusp; RV right ventricle; LA left atrium). e and f: Right and left coronary arterial systems acquired with an interleaved spiral imaging sequence. neous visualization of multiple coronary segments of left and right coronary arterial systems acquired with a double-oblique 3D T2Prep technique (22) is seen in Fig. 8c, where the proximal RCA is displayed along with the LM, LAD, LCX, and multiple branching vessels. An anomalous (dashed arrow) RCA acquired with an SSFP sequence in conjunction with a T2Prep for contrast enhancement (2) is displayed in Fig. 8d, on which smaller-diameter RCA branching vessels can be appreciated (dotted arrows). Both acquisitions in Fig. 8e and f show reformats of a right (e) and left (f) coronary arterial system acquired with a previously reported dual-interleave spiral imaging technique (8). The RCA coronary MRA data (3D T2Prep (12) technique (Fig. 9a)) of a patient with x-ray-defined coronary artery disease is shown adjacent to the x-ray angiogram (Fig. 9b). The x-ray-defined lesion in the proximal RCA (solid arrow) is visually confirmed on the coronary MRA reformat, and corresponding anatomical details can be appreciated in the regions of the dotted arrows. DISCUSSION The present implementation of the Soap-Bubble tool facilitates visualization and quantitative assessment of volumetric coronary MRA data sets that are acquired in parallel to the major axes of both the left and the right coronary arterial system. In the past, frequently explored quantitative measures of coronary MRA images included SNR and CNR (2,8,28,29), and vessel length, sharpness, and diameter (12,30). Using the present integrated implementation of the Soap-Bubble tool, these values can readily be obtained during the same analysis procedure with an ASCII output for further statistical analysis. Together with the implementation on a commercially available software and computer platform, this may facilitate a more widespread use of the present software. Contiguous coronary segments that are included in multiple adjacent slices of the 3D data set can be combined and displayed in one single image. When compared to more conventional planar MIPs, the Soap-Bubble implementation offers the advantage of localized MIPs in parallel to the local coronary anatomy. This may help to minimize adverse influences of partial volume effects associated with thicker slab MIPs for enhanced volumetric coverage. As a consequence, multiple segments of the coronary arterial tree can be displayed simultaneously and with maximized contrast. While true distances and ves-

8 Coronary Reformatting and Analysis Tool 665 FIG. 9. Reformatted T2Prep coronary MRA (a) together with the x-ray angiogram (b) acquired in a patient with x-ray-defined RCA disease. The lesion is shown in the region of the solid arrows together with corresponding smaller-diameter branching vessels (dotted arrows). sel segment lengths can be quantified, true distances are not maintained on the reformatted image, consistent with conventional parallel MIP algorithms. For larger 3D data sets or non-volume-targeted acquisitions (in which a more distinctive out-of-plane orientation of the coronary segments is expected), the present implementation may not be sufficient, and more sophisticated visualization and manipulation procedures (21,31) may have to be added. Alternative extensions to the present parallel projection method also include surface mapping techniques adopted from cortex mapping, as used in functional MRI (fmri) (32,33). However, given the 3D definition of the userspecified data points P 1..P n, such an adaptation of the visualization part of the software could be made at minimum cost. While a more automated identification of the major coronary segments would be desirable (34), the present approach primarily depends on a priori knowledge and the ability of the operator to identify the coronary anatomy in the 3D coronary MRA data set. Since the definition of the data points P 1..P n is user-specific, slightly different surface maps may be obtained for independent readers. However, very similar results are expected for users familiar with the coronary anatomy. In clinical cases with known coronary artery disease, the present tool has demonstrated its usefulness in displaying and visualizing the diseased or anomalous coronary segments, while the lesions could be visually identified. However, further research and software adaptations may be required for the quantitative assessment and grading of coronary disease. This may include an extension from a planar to a volumetric definition of coronary vessel sharpness and local coronary diameter. Such analysis may be limited by the anisotropic resolution obtained using contemporary coronary MRA data acquisition schemes (35). However, at higher field strengths, high-resolution isotropic coronary MRA data may be more easily obtained. An extension to a volumetric assessment of vessel sharpness and diameter remains to be examined. This will also support the quantitative assessment of eccentric coronary stenoses. Consistent with the intense research in the field of coronary MRA, a growing number of imaging sequences (2 12) are presently being examined. Furthermore, comparisons of these imaging sequences at different field strengths and with varying motion-compensation strategies (5,15 20,36 40) are being explored. The added value of such novel implementations often needs to be defined in comparison with an already established methodology. Therefore, the availability of a software tool that enables objective, quantitative analysis supports comparisons among different scanning techniques. Furthermore, the ability to load previously stored coronary artery geometries and ROIs enables consistent visualization and analysis of coronary MRA data obtained with different protocols in the same subjects. CONCLUSIONS The present coronary analysis tool, Soap-Bubble, enables the visualization of volumetric coronary MRA data, and the simultaneous display of multiple contiguous coronary vessels is feasible. It further facilitates quantitative coronary analysis and appears to be well suited for objective, quantitative comparisons among different scanning techniques. The use of this tool in quantitative coronary MRA analysis and grading of stenosis is an ultimate goal, and it remains to be investigated in patients with x-ray-defined coronary artery disease.

9 666 Etienne et al. REFERENCES 1. Kim WY, Danias PG, Stuber M, Flamm SD, Plein S, Nagel E, Langerak SE, Weber OM, Pedersen EM, Schmidt M, Botnar RM, Manning WJ. Coronary magnetic resonance angiography for the detection of coronary stenoses. N Engl J Med 2001;345: Deshpande VS, Shea SM, Laub G, Simonetti OP, Finn JP, Li D. 3D magnetization-prepared true-fisp: a new technique for imaging coronary arteries. Magn Reson Med 2001;46: Bornert P, Jensen D. Coronary artery imaging at 0.5 T using segmented 3D echo planar imaging. Magn Reson Med 1995;34: Goldfarb JW, Edelman RR. Coronary arteries: breath-hold, gadoliniumenhanced, three-dimensional MR angiography. Radiology 1998;206: Wielopolski PA, van Geuns RJ, de Feyter PJ, Oudkerk M. Breath-hold coronary MR angiography with volume targeted imaging. Radiology 1998;209: Stuber M, Botnar RM, Kissinger KV, Manning WJ. Free-breathing blackblood coronary MR angiography: initial results. Radiology 2001;219: Meyer CH, Hu BS, Nishimura DG, Macovski A. Fast spiral coronary artery imaging. Magn Reson Med 1992;28: Bornert P, Stuber M, Botnar RM, et al. Direct comparison of 3D spiral vs. Cartesian gradient-echo coronary magnetic resonance angiography. Magn Reson Med 2001;46: Li D, Dolan RP, Walovitch RC, Lauffer RB. Three-dimensional MRI of coronary arteries using an intravascular contrast agent. Magn Reson Med 1998;39: Hofman MBM, Henson RE, Kovacs SJ, et al. Blood pool agent strongly improves 3D magnetic resonance coronary angiography using an inversion pre-pulse. Magn Reson Med 1999;41: Stuber M, Botnar RM, Danias PG, et al. Contrast agent-enhanced, freebreathing, three-dimensional coronary magnetic resonance angiography. J Magn Reson Imaging 1999;10: Botnar RM, Stuber M, Danias PG, Kissinger KV, Manning WJ. Improved coronary artery definition with T2-weighted, free-breathing, three-dimensional coronary MRA. Circulation 1999;99: Sodickson DK, Manning WJ. Simultaneous acquisition of spatial harmonics (SMASH): fast imaging with radiofrequency coil arrays. Magn Reson Med 1997;38: Pruessmann KP, Weiger M, Scheidegger MB, Boesiger P. SENSE: sensitivity encoding for fast MRI. Magn Reson Med 1999;42: Edelman RR, Manning WJ, Burstein D, Paulin S. Coronary arteries: breath-hold MR angiography. Radiology 1991;181: McConnell MV, Khasgiwala VC, Savord BJ, et al. Prospective adaptive navigator correction for breath-hold MR coronary angiography. Magn Reson Med 1997;37: Li D, Kaushikkar S, Haacke EM, et al. Coronary arteries: three-dimensional MR imaging with retrospective respiratory gating. Radiology 1996;201: Stuber M, Botnar RM, Danias PG, Kissinger KV, Manning WJ. Submillimeter three-dimensional coronary MR angiography with real-time navigator correction: comparison of navigator locations. Radiology 1999;212: Wang Y, Watts R, Mitchell I, et al. Coronary MR angiography: selection of acquisition window of minimal cardiac motion with electrocardiography-triggered navigator cardiac motion prescanning initial results. Radiology 2001;218: Kim WY, Stuber M, Kissinger KV, Andersen NT, Manning WJ, Botnar RM. Impact of bulk cardiac motion on right coronary MR angiography and vessel wall imaging. J Magn Reson Imaging 2001;14: Cline HE, Thedens DR, Irarrazaval P, et al. 3D MR coronary artery segmentation. Magn Reson Med 1998;40: Stuber M, Botnar RM, Danias PG, et al. Double-oblique free-breathing high resolution three-dimensional coronary magnetic resonance angiography. J Am Coll Cardiol 1999;34: Wrazidlo W, Brambs HJ, Lederer W, Schneider S, Geiger B, Fischer C. An alternative method of three-dimensional reconstruction from twodimensional CT and MR data sets. Eur J Radiol 1991;12: Deriche R. Fast algorithms for low-level vision. IEEE Trans Pattern Anal Machine Intel 1990;PAMI-12: Cavagna FM, Anelli PL, Lorusso V, et al. B-22956, a new intravascular contrast agent for MR coronary angiography. In: Proceedings of the 9th Annual Meeting of ISMRM, Glasgow, Scotland, p Stuber M, Botnar RM, Spuentrup E, Kissinger KV, Manning WJ. Threedimensional high-resolution fast spin-echo coronary magnetic resonance angiography. Magn Reson Med 2001;45: Stuber M, Boernert P, Spuentrup E, Botnar RM, Manning WJ. Selective three-dimensional visualization of the coronary arterial lumen using arterial spin tagging. Magn Reson Med 2002;47: Zheng J, Li D, Cavagna FM, et al. Contrast-enhanced coronary MR angiography: relationship between coronary artery delineation and blood T1. J Magn Reson Imaging 2001;14: Botnar RM, Stuber M, Danias PG, Kissinger KV, Manning WJ. A fast 3D approach for coronary MRA. J Magn Reson Imaging 1999;10: Shea SM, Kroeker RM, Deshpande V, et al. Coronary artery imaging: 3D segmented k-space data acquisition with multiple breath-holds and real-time slab following. J Magn Reson Imaging 2001;13: Rensing BJ, Bongaerts AH, van Geuns RJ, van Ooijen PM, Oudkerk M, de Feyter PJ. Intravenous coronary angiography using electron beam computed tomography. Prog Cardiovasc Dis 1999;42: Dale AM, Fischl B, Sereno MI. Cortical surface-based analysis. I. Segmentation and surface reconstruction. Neuroimage 1999;9: Fischl B, Sereno MI, Dale AM. Cortical surface-based analysis. II. Inflation, flattening, and a surface-based coordinate system. Neuroimage 1999;9: Cline HE, Thedens DR, Meyer CH, Nishimura DG, Foo TK, Ludke S. Combined connectivity and a gray-level morphological filter in magnetic resonance coronary angiography. Magn Reson Med 2000;43: Botnar RM, Stuber M, Kissinger KV, Manning WJ. Free-breathing 3D coronary MRA: the impact of isotropic image resolution. J Magn Reson Imaging 2000;11: Hardy CJ, Saranathan M, Zhu Y, Darrow RD. Coronary angiography by real-time MRI with adaptive averaging. Magn Reson Med 2000;44: Nehrke K, Bornert P, Manke D, Bock JC. Free-breathing cardiac MR imaging: study of implications of respiratory motion initial results. Radiology 2001;220: Jhooti P, Gatehouse PD, Keegan J, Bunce NH, Taylor AM, Firmin DN. Phase ordering with automatic window selection (PAWS): a novel motion-resistant technique for 3D coronary imaging. Magn Reson Med 2000;43: Weiger M, Bornert P, Proksa R, Schaffter T, Haase A. Motion-adapted gating based on k-space weighting for reduction of respiratory motion artifacts. Magn Reson Med 1997;38: Huber ME, Hengesbach D, Botnar RM, et al. Motion artifact reduction and vessel enhancement for free-breathing navigator-gated coronary MRA using 3D k-space reordering. Magn Reson Med 2001;45: