TOF-MRA Using Multi-Oblique-Stack Acquisition (MOSA)

Transcription

1 JOURNAL OF MAGNETIC RESONANCE IMAGING 26: (2007) Technical Note TOF-MRA Using Multi-Oblique-Stack Acquisition (MOSA) Ed X. Wu, PhD, 1,2 * Edward S. Hui, BEng, 1,2 and Jerry S. Cheung, BEng 1,2 Purpose: To alleviate the intrinsic limitation of current time-of-flight magnetic resonance angiography (TOF-MRA) techniques, which are insensitive to in-plane blood flow due to the flow saturation effect. Materials and Methods: A multi-oblique-stack acquisition (MOSA) technique was proposed in this study to tackle this TOF problem by acquiring two or more TOF image stacks bearing different orientations. The MOSA approach was evaluated in human brain MR angiograms by integrating it with the widely used multiple overlapping thin-slab acquisition (MOTSA) technique. Two TOF image sets in different orientations were acquired using MOTSA. They were combined pixel-by-pixel by taking the maximum intensity value. Maximum intensity projection (MIP) was then performed, and the resulting MRA quality was assessed. Results: The results demonstrated that MR angiograms obtained by MOSA-MOTSA were clearly improved compared to those acquired by conventional MOTSA in the same scan time. Conclusion: MOSA clearly demonstrates its advantage over conventional TOF acquisition in visualizing in-plane blood flows. Key Words: time-of-flight; MRA; multiple angles; multiple acquisitions; vessel enhancement J. Magn. Reson. Imaging 2007;26: Wiley-Liss, Inc. IN TIME-OF-FLIGHT MAGNETIC RESONANCE AN- GIOGRAPHY (TOF-MRA), the signal from blood is maximized when the blood vessel is orthogonal to the imaging plane (i.e., when the inflow effect is maximal) (1). However, when the blood vessel is parallel to the imaging plane, the signal from blood will be saturated, resulting in hypointense signal or discontinuity in the MR 1 Laboratory of Biomedical Imaging and Signal Processing, University of Hong Kong, Hong Kong. 2 Department of Electrical and Electronic Engineering, University of Hong Kong, Hong Kong. Contract grant sponsors: University of Hong Kong Committee on Research and Conference Grants; Hong Kong Jockey Club Charities Trust. *Address reprint requests to: E.X.W., Department of Electrical and Electronic Engineering, University of Hong Kong, Pokfulam, Hong Kong. ewu@eee.hku.hk Received August 28, 2006; Accepted January 25, DOI /jmri Published online in Wiley InterScience ( angiogram (1). Such spin saturation can be reduced by aligning the slice selection to the direction of flow, and hence maximizing the inflow effect within the imaging slab (2). However, given a single imaging stack, there will always exist some vessels that are parallel to the stack or are inclined at an angle from the slice-selection direction because of the general complexity of threedimensional (3D) vascular morphology. In addition, with traditional 3D TOF-MRA techniques, the slab is generally thick, resulting in more spin saturation (3). This problem can be partly alleviated by using multiple overlapping thin slab acquisition (MOTSA) (4) and applying a tilted optimized nonsaturating excitation (TONE) pulse whose flip angle linearly varies across the imaging slab (5). To address the intrinsic in-plane flow saturation problem in TOF-MRA, a multi-angle TOF- MRA approach is proposed in this study. MATERIALS AND METHODS Theory Signal from blood is maximized in TOF-MRA images when blood flows perpendicularly to the imaging plane (1). The blood signal in a TOF slice is (1,6) 1 qn S sin M ze M 0 M ze (1) n 1 q, where q e TR/T1 cos, n integer of TH/ vt R, and M ze M 0 1 e TR/T1 / 1 q. In these equations represents the flip angle, v is the blood flow speed, TH is the slice thickness, TR is the repetition time, T 1 is the spin-lattice relaxation time for blood, n is the number of blood velocity segments in the blood vessel within an excited slice that experience different number of RF excitations (1), M 0 is the longitudinal equilibrium magnetization, and M ze is the spoiled equilibrium longitudinal magnetization value. When the vessel traverses the imaging plane not perpendicularly but with an oblique angle (see Fig. 1), the number of blood velocity segments, a subslice with thickness vt R, within an excited slice, n, is larger than that for a perpendicular vessel: n n cos. (2) 2007 Wiley-Liss, Inc. 432

2 TOF-MRA Using Multi-Oblique-Stack Acquisition 433 maximum intensity projection (MIP) of the new TOF image matrix. Figure 1. Illustrative depiction of a tilted blood vessel within an excited slice. It shows a blood vessel that is inclined at an angle from the slice-selection direction within an excited slice of thickness, TH. Within the slice there are n number of velocity segments, with thickness vt R. The segments experience an increasing number of RF pulses as they move from the bottom to the top of the slice. The blood signal within the tilted vessel, S, for the same volume of blood, becomes 1 qn S sin M ze M 0 M ze (3) n 1 q. Equations [2] and [3] show that there exists a complex relationship between the blood signal and the angle. Because there are more velocity segments within a tilted blood vessel, more blood spins will experience more RF excitations and saturation. Therefore, S is always smaller than S, except for the case of very fastflowing blood and a thin slice, where n n 1. During TOF-MRA acquisition, rotating the normal axis of the imaging plane toward the blood vessel direction by a finite angle can reduce the n in Eq. [2] to n n cos, (4) thus increasing the blood signal S in Eq. [3]. However, due to the general vascular complexity, simply rotating a single imaging stack by angle will lead not only to an increase of inflow effect for some vessels, but also to a decrease of inflow effect for some other vessels as well. It is therefore beneficial to acquire multiple TOF imaging stacks with various orientations and a broad range of blood inflow effects, and then combine them for improved MR angiograms. This multi-oblique-stack acquisition (MOSA) technique is illustrated in Fig. 2, in which, for example, two stacks are applied for brain TOF-MRA. With each stack of TOF images bearing a particular orientation, the blood inflow effect and vessel coverage in each stack will be different. To obtain the best vasculature coverage from all TOF image sets, the sets can be combined by taking the maximum pixel value, pixel by pixel, to form a single new TOF image matrix. MR angiograms can then be reconstructed by Experiments All MRI experiments were carried out on six healthy volunteers and performed on a 3T whole-body scanner (Achieva, Philips Medical Systems, The Netherlands). Written informed consent was obtained from all subjects. An eight-channel quadrature head coil was used. In the current study, two oblique stacks were acquired using MOTSA, each of which was inclined at 15 or 15 with respect to the anterior posterior (AP) axis for the MOSA-MOTSA TOF images. For comparison between MOSA and conventional TOF-MOTSA, a single stack of TOF-MOTSA images along the AP axis without tilting was also acquired for each subject using the same parameters, except for the number of signal averages (NSA 2). Isotropic TOF images were obtained with a TOF-MOTSA T 1 fast-field echo (FFE) sequence using five slabs and a TONE pulse. The experimental parameters were as follows: TR/TE/ 25/3.5 msec/ 18, image resolution mm 3, FOV mm 3, acquisition matrix size with zero-filling to , SENSE factor of 4 (factor of 4 for phase), phaseencoding direction right left (RL), vein saturation band of 20-mm thickness and 12-mm gap, and NSA 1 for each acquisition of the two tilted stacks. MOTSA image volumes were constructed from MOTSA slabs using the chunk acquisition and reconstruction method (CHARM; Philips Medical Systems, The Netherlands) (4). The scan time was 10.5 minutes for both the two-oblique MOSA and the conventional MOTSA protocols. Image Processing Prior to MR angiogram reconstruction, the MOSA- MOTSA TOF images were coregistered and resliced with respect to the conventional TOF-MOTSA images by using SPM2 (Wellcome Department of Cognitive Neurology, London, UK). The two coregistered MOSA-MOTSA TOF image sets were then combined by taking the larger intensity value, pixel by pixel, between the two to form a new TOF image matrix using MATLAB (Math- Works, Natick, MA, USA). An identical mask was man- Figure 2. Illustration of MOSA, with a blood vessel perpendicular to the FH axis covered by two imaging stacks tilted at (stack 1) and (stack 2) from the AP axis.

3 434 Wu et al. Figure 3. Typical brain MR angiograms on the axial plane of a normal volunteer. a: Transverse MIP views of TOF-MRA obtained by conventional 3D MOTSA. b: Transverse MIP views of TOF-MRA obtained by 3D MOSA-MOTSA with two oblique stacks. With MOSA, the signal intensity from the medium-sized blood vessel is clearly higher, especially the vessels indicated by the open arrow (vessel I), curly bracket (vessel II), double arrows (vessel III), and solid arrow (vessel IV). Note that blood flow within vessels I IV was mostly in-plane for conventional MOTSA. ually drawn and applied to cut away the skull, eyes, and tissues within the pharynx in the combined MOSA- MOTSA TOF images and the conventional TOF-MOTSA images for better vessel visualization. Afterwards, MR angiograms were MIP-reconstructed from these two data sets using etdips (NIH, USA). The MIP MR angiogram computed from the combined TOF images was analyzed using ImageJ (NIH, USA). To compare the MR angiogram quality for both methods, regions of interest (ROIs) were manually drawn in the MIP images with exactly the same area of ROI for both MOSA and MOTSA images. ROIs were placed in the background brain tissue and a portion of blood vessels. The ROIs of blood vessels included areas of obvious improvement in vasculature visualization in the MR angiogram from MOSA-MOTSA. Because of the variation in noise level across the FOV in parallel imaging, measurement of noise from ROIs in the background air is not appropriate (7). Instead, vessel contrast (VC) was used and defined as VC 100 S vessel S tissue S tissue (5) where S vessel represents the signal intensity in the ROI of blood vessels, and S tissue represents the signal intensity in the ROI in background brain tissue. RESULTS The MOSA approach yielded a general improvement of MR angiogram quality in all six subjects studied. Figures 3 and 4 illustrate a typical comparison in one subject, showing the axial and sagittal views of TOF- MRA obtained by conventional MOTSA (Figs. 3a and 4a) and those obtained by the proposed MOSA-MOTSA with two oblique stacks (Figs. 3b and 4b). TOF images from conventional MOTSA and MOSA-MOTSA are also shown in Fig. 5 to show the appreciable difference in vessel appearance between the two methods. With MOSA, the signal intensity from the medium-sized blood vessel in all views is clearly higher, particularly the vessels indicated by the open arrow (vessel I), curly bracket (vessel II), double arrows (vessel III), and solid arrow (vessel IV). Enlarged views of these vessels are shown in their corresponding figures. Blood flow within vessels I IV was mostly in-plane for the conventional MOTSA. Note that spatial coverage by overlap of two

4 TOF-MRA Using Multi-Oblique-Stack Acquisition 435 Figure 4. Typical brain MR angiograms on the sagittal plane of a normal volunteer. a: Sagittal MIP views of TOF-MRA obtained by conventional 3D MOTSA. b: Sagittal MIP views of TOF-MRA obtained by 3D MOSA-MOTSA. MOSA sets was slightly smaller than that of the MOTSA set. Also note that, without cutout, MIP vessel visualization using MOSA-MOTSA is still superior to that achieved using conventional MOTSA. Given the same total scan time, one potential drawback of MOSA is the increase in background noise level due to the application of fewer signal averages for each acquisition of a particular imaging stack. Tables 1 and 2 show the background tissue signal, vessel signal, and VC analysis results for the datasets illustrated in Fig. 3. One background brain-tissue ROI and four vessel ROIs used in the analysis are depicted in Fig. 3b as a dashed circle and rectangles, respectively. The ROIs of the vessels were drawn inside the dotted box indicated in Fig. 3b. The ROI areas were 3.9 mm 2 for vessel I, 2.7 mm 2 for vessel II, 5.0 mm 2 for vessel III, and 5.6 mm 2 for vessel IV. As shown in Table 2, the signal intensity of blood vessels S vessel was significantly increased. Therefore, the overall performance of MOSA-MOTSA is superior to that of conventional MOTSA. For example, for vessel II in Figs. 3 and 4, which exhibited significant improvement, the signal intensity rose approximately by 30.0%. More importantly, the contrast between the vessels and the background brain tissue in the MIP images were higher in MOSA-MOTSA. DISCUSSION The results demonstrate that the MOSA-MOTSA approach produces better MR angiogram quality as com- Table 1 Background Brain Tissue Signal Measurements in MIP Images Image type Method S tissue SD a Figure 5. TOF images on the same axial plane acquired using conventional 3D MOTSA (a), MOSA-MOTSA (b), MOTSA at 15 (c), and MOTSA at 15 (d). The images are from the same data sets shown in Figs. 3 and 4. Magnified views of one vessel are shown in the insets. MIP MOTSA (NSA 2) MOSA-MOTSA images (NSA 1) images (NSA 1) a Values are mean SD. S tissue the signal intensity in the ROI in background brain tissue.

5 436 Wu et al. Table 2 Signal and Contrast Measurements of the Four Vessels Marked in Fig. 3b in MIP Images Image type Method Vessel I Vessel II Vessel III Vessel IV S vessel VC S vessel VC S vessel VC S vessel VC MIP MOTSA MOSA-MOTSA S vessel signal intensity in the ROI in the blood vessel, VC vessel contrast between the blood vessel and background brain tissue. pared to conventional MOTSA. For the same scan time, MOSA approach reduces the overall influence of the in-plane blood flow saturation effect in complex 3D vasculature. Despite the fact that fewer signal averages have to be applied for each tilting angle, which results in a higher background noise level in MOSA TOF images, the signal from blood is generally higher in MOSA- MOTSA MIP images compared to MOTSA images, thus enabling stronger contrast between the blood vessels and background brain tissue. The optimum implementation of MOSA depends on the vascular anatomy and the TOF acquisition parameters used. In the brain, the major arteries bend themselves away from the foot head (FH) axis. The choice of the anterior-posterior (AP) axis being perpendicular to the FH axis as the reference axis for tilting seems to produce more improvement for the two oblique angle acquisitions, based on the six volunteers studied. In addition, it is necessary to find a set of optimal angulations for the imaging planes based on how the vessels in an organ are oriented. For the brain, the optimum angulations with respect to the AP axis seem to be in the range of With large angulations, the vein suppression based on the saturation at one side of the slab can become problematic. In conventional brain TOF- MRA, the venous blood saturation band is put on top of the axial imaging stack because venous flow is assumed to be along the head foot direction. However, when the stack is tilted at a large angle, there can be some intracranial arterial blood going through the stack from top to bottom with respect to the stack, leading to undesired arterial suppression instead. Blood-flow artifacts are known to distort vessel appearance in TOF-MRA, depending on the flow direction with regard to the in-plane orientation. Thus MOSA may lead to varying distortions for different oblique angles, and make the vessels appear smeared or larger after the maximum-picking operation of all TOF-MOSA image sets. Enlargement of certain vessels was observed in a few volunteer scans in this study, likely caused by this issue. In addition, patient motion during MOSA could lead to a similar problem. Note that successful implementation of TOF-MOSA approach also requires appropriate interpolation and correct coregistration of multiple TOF-MOSA data sets in order to accurately perform the pixelwise combination into a single data set. In practice, multiple TOF-MOSA image sets cannot be simply regridded to a single 3D image matrix correctly if any gradient nonlinearity causing non-rigid image distortions exists that depends on the orientation of the imaging. However, we did not observe any misregistration due to this problem, likely because the gradient nonlinearity is negligible in the relatively small central region used for brain imaging. Nevertheless, this problem can be fully resolved if nonrigid image distortions are corrected in the TOF image sets based on the known gradient nonlinearity prior to their pixelwise combination. In the current study, only two oblique stacks of brain TOF images were employed for the MOSA demonstration. In fact, more than two stacks can be used for even better vessel visualization. MOSA is applicable to the study of blood vessels in the brain and other organs, and it can be applied to both 2D and 3D TOF-MRA protocols in general, although the collective optimization of MOSA and TOF parameters remains to be explored. In practice, the acquisitions of these oblique stacks can be interleaved to reduce patient-motion artifacts during scan. The MOSA approach may be particularly useful for low- and mid-field MRI scanners, where multiple averaging is often employed. REFERENCES 1. Haacke EM, Brown RW, Thompson MR, Venkatesan R. Magnetic resonance imaging physical principles and sequence design. New York: Wiley-Liss; p. 2. Ozsarlak O, Goethem JWV, Maes M, Parizel PM. MR angiography of the intracranial vessels: technical aspects and clinical applications. Neuroradiology 2004;46: Bosmans H, Wilms G, Dymarkowski S, Marchal G. Basic principles of MRA. Eur J Radiol 2001;38: Blatter DD, Parker DL, Robison RO. Cerebral MR angiography with multiple overlapping thin slab acquisition. Part I. Quantitative analysis of vessel visibility. Radiology 1991;179: Atkinson D, Brant-Zawadzki M, Purdy D, Laub G. Improved MR angiography: magnetization transfer suppression with variable flip angle excitation and increased resolution. Radiology 1994;190: Haacke EM, Masaryk TJ, Wielopolski PA, Zypman FR. Optimizing blood vessel contrast in fast three-dimensional MRI. Magn Reson Med 1990;14: Kellman P, McVeigh ER. Image reconstruction in SNR units: a general method for SNR measurement. Magn Reson Med 2005;54: