Method and apparatus for measuring and correcting motion effects using navigator echoes

[0002] This patent application claims the benefit of U.S. Provisional Patent Applications No. 60/329,811, filed Oct. 16, 2001 and No. 60/373,820, filed Apr. 19, 2002.

[0004] The present invention relates generally to medical diagnosis, and more particularly relates to magnetic resonance image data acquisition.

[0005] Magnetic resonance imaging (MRI) is the method of choice for noninvasive diagnosis of soft tissue disease in humans, and has wide application in cardiovascular diseases. Fast gradient technology has made high-resolution 3D imaging possible, including magnetic resonance angiography (MRA) of coronary and pulmonary arteries. High resolution MRI is primarily impeded by physiologic motion such as respiration and gross patient movement.

[0006] Industry responded to the physiologic motion impediment by developing methods to suppress the artifacts caused by physiological motion. Motion correction requires precise motion information. One suppression approach is the navigator method, which detects motion prior to data acquisition and modifies data acquisition accordingly. Navigator techniques use image space navigator echoes for detecting motion during image data acquisition. Physiological motion causes global displacement in the navigator echo and results in a shift of the image space navigator echo compared to an image space reference navigator echo. The accuracy in extracting motion information from navigator echoes is crucial to the effectiveness of the navigator technique. An image space least squares method has been demonstrated to be an accurate method for extracting displacement. The image space least squares algorithm computes the variance for a shifted navigator echo with respect to a reference echo and determines the displacement by locating the shift position that gives the minimum variance. Other methods used include correlation and edge detection methods. However, these methods are time consuming on most host computers of MRI scanners and do not allow flexible real-time control of data acquisition due to the long processing times required.

[0007] For effective real-time control of data acquisition, the processing time for processing the navigator echo must be short enough so that there is little motion in the processing duration. When this occurs, the detected motion in the navigator echo is approximately the same as that in the acquired k-space data, thereby allowing effective real-time modification of data acquisition to suppress motion effects.

[0008] One approach taken to decrease processing time is to extract linear motion information from raw k-space data. If motion is linear in 3D, only the k-space phase data carries motion information. According to the Fourier shift theorem, a global displacement of d pixels of a reference navigator echo corresponds to a linear phase shift in the k-space data. Equation 1 shows this correspondence. $\begin{array}{cc}{s}_{\mathrm{dis}}\left[x\right]=s_{\mathrm{ref}}\left[x+d\right]\stackrel{\mathrm{DFT}}{⟷}{S}_{\mathrm{dis}}\left[k\right]=S_{\mathrm{ref}}\left[k\right]\exp \left(i\frac{2\pi }{N}\mathrm{kd}\right)& \left[1\right]\end{array}$

[0009] In equation 1, s_ref[x] is the reference navigator profile in image space, s_dis[x] is the displaced profile in image space, S_ref[k] is the reference navigator echo in k-space, S_dis[k] is the displaced echo in k-space, and d is the displacement. Equation 1 illustrates that displacement is proportional to the slope of the phase change in k-space data.

[0010] One method to extract displacement information based upon the Fourier shift theorem was proposed by Ahn and Cho. The Ahn and Cho method is fast, requiring only 4N multiplications and 4N additions per displacement estimate. The Ahn and Cho method derives the relation in equation 2 for the displacement d from equation 1: $\begin{array}{cc}d=\frac{N}{2\pi }\left(\arg \left\{S_{\mathrm{dis}}^{*}\left[k\right]S_{\mathrm{dis}}\left[k+1\right]\right\}-\arg \left\{S_{\mathrm{ref}}^{*}\left[k\right]S_{\mathrm{ref}}\left[k+1\right]\right\}\right)& \left(2\right)\end{array}$

[0011] In equation 2, arg{·} is the phase operator, and S*_ref[k]S_ref[k+1] may be regarded as a kernel of the nearest neighbor correlation.

[0012] According to equation 2, only a pair of adjacent points is theoretically sufficient to derive the displacement information. In practice, all data points are contaminated with noise and the displacement derived from two points contains substantial error. To account for noise, all other points in the navigator echo can be averaged to reduce the noise error. Ahn and Cho proposed an average by taking the nearest neighbor correlation prior to the phase operator as shown in equation 3: $\begin{array}{cc}\frac{N}{2\pi }\arg E\left\{S^{*}\left[k\right]S\left[k+1\right]\right\}=\frac{N}{2\pi }\arg \sum _{k=-N/2}^{N/2-2}{S}^{*}\left[k\right]S\left[k+1\right]& \left(3\right)\end{array}$

[0013] E{·} is the expectation operator and the divider is dropped as it has no effect on the phase. Accordingly, the displacement is: $\begin{array}{cc}{\hat{d}}_{\mathrm{AHN}}=\frac{N}{2\pi }\left(\arg \sum _{k=-N/2}^{N/2-2}{S}_{\mathrm{dis}}^{*}\left[k\right]S_{\mathrm{dis}}\left[k+1\right]-\arg \sum _{k=-N/2}^{N/2-2}{S}_{\mathrm{ref}}^{*}\left[k\right]S_{\mathrm{ref}}\left[k+1\right]\right)& \left(4\right)\end{array}$

[0014] It can be seen from equation 4 that points in k-space contribute equally to the final displacement estimate. Consequently, this method is susceptible to noise, particularly contributions from points at the edges of k-space where the noise dominates the signal. Additionally, the Ahn and Cho method is geared towards detecting translation (i.e., displacement). As such, it is suitable for detecting one dimensional motion, but is not effective when detecting general motion.

[0015] General motion involves rotation, dilation, and contraction and existing techniques do not work. The rotation, dilation, and contraction effects must be known to correct their motion effects. For example, the widely used pencil beam navigator echo techniques provides a good indication of cardiac motion. However, the pencil beam navigator echo technique cannot detect multiple motion components simultaneously. For example, it cannot effectively monitor the motion of coronary arteries because the motion of chamber blood interferes with the detection of heart motion.

[0016] In view of the foregoing, it is a general aim of the present invention to enhance the speed and accuracy of suppressing motion effects in MRI systems.

[0017] In that regard, it is also an object of the present invention to minimize noise effects in navigator echoes.

[0018] A feature of the present invention is to detect a general motion due to rotation, dilation, and displacement.

[0019] A further feature is to detect general motion of coronary arteries due to rotation, dilation, and displacement from data acquired with volume selective excitation.

[0020] A further feature is to utilize algorithms which employ weighting factors based on the signal to noise ratio of each point.

[0021] In accordance with an embodiment of the instant invention, a method of detecting displacement comprises by fitting a function to the motion-induced phase shift using a k-space weighted least squares minimization to find the displacement where the weighting factor takes into account that the noise in k-space phase data is inversely proportional to the signal to noise ratio.

[0022] In accordance with a further embodiment of the instant invention, a method of detecting displacement comprises performing a linear regression on k-space data by fitting a straight line to the motion-induced phase shift using a k-space weighted least squares minimization to find the displacement, and using a weighting factor which takes into account that the noise in k-space phase data is inversely proportional to the signal to noise ratio.

[0023] In accordance with a further alternate embodiment of the instant invention, a method of detecting a general motion due to rotation, dilation, and displacement comprises determining a rotation angle and dilation scaling factors from the magnitude k-space data by a least squares minimization. Once the rotation angle and dilation scaling factors are determined, a displacement vector is obtained from the phase data in k-space using a least squares minimization. For example, in one embodiment, general motion of coronary arteries is detected by volume-selective excitation of the arteries. This is accomplished by transmitting a spatial spectral selective pulse to eliminate signals from the chest wall and to excite epicardial fat to produce an excited signal. The coronary motion is detected as described above.

[0024] Other objectives and advantages of the invention will become more apparent from the following detailed description when taken in conjunction with the accompanying drawing.

[0025] The accompanying drawings incorporated in and forming a part of the specification illustrate several aspects of the present invention, and together with the description serve to explain the principles of the invention. In the drawings:

[0026] FIG. 1 is a block diagram generally illustrating an exemplary computer system on which the present invention resides;

[0027] FIG. 2 is a flow chart of a process for acquiring linear motion displacement from raw k-space phase data in accordance with the present invention;

[0028] FIG. 3 illustrates standard deviations of estimation errors at different noise levels using a simulated navigator profile;

[0029] FIGS. 4aand 4billustrate an example of actual and estimated displacements of a motion phantom;

[0030] FIG. 5aillustrates a coronal image of a resolution motion phantom obtained without motion;

[0031] FIG. 5billustrates a coronal image of the resolution motion phantom of FIG. 5aobtained without motion correction;

[0032] FIG. 5cillustrates a coronal image of the resolution motion phantom of FIG. 5bobtained with prior art motion correction;

[0033] FIG. 5dillustrates a coronal image of the resolution motion phantom of FIG. 5bobtained with motion correction in accordance with the teachings of the instant invention;

[0034] FIG. 6aillustrates estimated displacements of a diaphragm in a human volunteer using a prior art k-space motion correction and a standard image space least squares motion correction;

[0035] FIG. 6billustrates estimated displacements of a diaphragm in a human volunteer using k-space motion correction of the instant invention and a standard image space least squares motion correction;

[0036] FIGS. 7a-fillustrate exemplary k-space trajectories for detecting motion of an object;

[0037] FIG. 8 is a flow chart of a process for acquiring rotation, dilation, and displacement from raw k-space magnitude and phase data in accordance with the present invention;

[0038] FIG. 9 illustrates an exemplary k-space trajectory for detecting coronary artery motion;

[0039] FIG. 10ais an axial image of heart volume without water saturation;

[0040] FIG. 10bis an axial image of heart volume with water saturation;

[0041] FIG. 10cis an axial image of heart volume with water saturation and with spatial fat saturation;

[0042] FIG. 11 illustrates the navigator echoes corresponding to the k-space trajectory of FIG. 9;

[0043] FIG. 12aillustrates RL displacement of human coronary arteries obtained with the teachings of the present invention;

[0044] FIG. 12billustrates AP displacement of human coronary arteries obtained with the teachings of the present invention;

[0045] FIG. 12cillustrates SI displacement of human coronary arteries obtained with the teachings of the present invention;

[0046] FIG. 13aillustrates diaphragm motion in the SI direction without breath-holding obtained with the teachings of the present invention;

[0047] FIG. 13billustrates coronary motion in the SI direction without breath-holding obtained with the teachings of the present invention;

[0048] FIG. 13cillustrates coronary motion in the AP direction without breath-holding obtained with the teachings of the present invention;

[0049] FIG. 13dillustrates coronary motion in the RL direction without breath-holding obtained with the teachings of the present invention;

[0050] FIG. 14aillustrates diaphragm motion in the SI direction with breath-holding obtained with the teachings of the present invention;

[0051] FIG. 14billustrates coronary motion in the SI direction with breath-holding obtained with the teachings of the present invention;

[0052] FIG. 14cillustrates coronary motion in the AP direction with breath-holding obtained with the teachings of the present invention;

[0053] FIG. 14dillustrates coronary motion in the RL direction with breath-holding obtained with the teachings of the present invention; and

[0054] FIG. 15 illustrates one cycle of cardiac motion in the RL direction with ECG gating.

[0055] While the invention will be described in connection with certain preferred embodiments, there is no intent to limit it to those embodiments. On the contrary, the intent is to cover all alternatives, modifications and equivalents as included within the spirit and scope of the invention as defined by the appended claims.

[0056] Turning to the drawings, wherein like reference numerals refer to like elements, the invention is illustrated as being implemented in a suitable MRI data acquisition environment. Although not required, the invention will be described in the general context of computer-executable instructions, such as program modules, being executed by a personal computer. The invention may also be practiced in distributed computing environments where tasks are performed by remote processing devices that are linked through a communications network.

[0057] With reference to FIG. 1, a block diagram of an exemplary system 100 for implementing the invention is shown. The exemplary system includes a computer 102, a magnet system 104, a data acquisition system 106, and display 108. The magnet system 104, data acquisition system 106 and display 108 are well known in the art and need not be described in detail herein. In general terms, the computer 102 controls the gradient and RF magnets or coils (not shown) in the magnet system 104 via amplifiers (not shown). The computer 102 also controls the data acquisition system 106, processes the data acquired, and outputs an image to display 108.

[0058] Computer 102 typically includes at least some form of computer readable media. By way of example, and not limitation, computer readable media may comprise computer storage media and communication media. Computer storage media includes volatile and non-volatile, removable and non-removable media implemented in any method or technology for storage of information such as computer readable instructions, data structures, program modules or other data. Computer storage media includes, but is not limited to, RAM, ROM, EEPROM, flash memory, or other memory technology, CD-ROM, digital versatile disks (DVD) or other optical storage, magnetic cassettes, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other medium which can be used to store the desired information. Communication media typically embodies computer readable instructions, data structures, program modules, or other data in a modulated data signal such as a carrier wave or other transport mechanism and includes any information delivery media.

[0059] The present invention fits a function to the variance from the k-space data of an object at a displaced position to the k-space data of the object at a reference position and weights the function to account for noise in the k-space data. In fitting a weighted function to the variance, the weighting can be done by weighting the variance, weighting the overall function, or weighting the variance and the overall function. The displacement is found by using a least squares minimization on the weighted function.

[0060] One embodiment of the present invention is based on linear regression of the k-space navigator phase data. The displacement is found by fitting a straight line to the motion induced phase shift using a k-space weighted least squares approach according to the equation: $\begin{array}{cc}\hat{d}\text{:}\min \left\{\sum _{k=-N/2}^{N/2-1}\frac{|S\left[k\right]{|}^2}{2{\sigma }^2}{\left(\Delta \varphi \left[k\right]-\frac{2\pi }{N}\mathrm{kd}\right)}^2\right\}& \left(5\right)\end{array}$

[0061] where Δφ[k]=arg S_dis[k]−arg S_ref[k].

[0062] The noise in the k-space phase data varies according to the inverse of the k-space signal to noise ratio (SNR), |S[k]|/σ, when the SNR is sufficiently high because the noise in the signal is approximately Gaussian. The present invention minimizes noise effects by selecting the weighting factor to be $\frac{|S\left[k\right]{|}^2}{2{\sigma }^2}.$

[0063] The weighting factor used results in the linear fit of motion-induced phase shift to be weighted more strongly to the high SNR points near the k-space center and only weakly to the poor SNR points at the edges of k-space. Additionally, the weightings are such that the average power of noise in the estimate is minimized. This makes the present method robust against noise, in contrast to Ahn and Cho's method where equal weightings are used for all data points.

[0064] The weighted least-squares method illustrated in equation 5 has the following closed-form solution: $\begin{array}{cc}{\hat{d}}_{\mathrm{kLS}}=\frac{N}{2\pi }\frac{\sum _{k=-N/2}^{N/2-1}k|S\left[k\right]{|}^2\Delta \varphi \left[k\right]}{\sum _{k=-N/2}^{N/2-1}{k}^2|S\left[k\right]{|}^2}& \left(6\right)\end{array}$

[0065] Equation 6 shows that the method of the present invention requires only N multiplications and N additions per displacement estimate, if k|S[k]|²and $\sum _{k=-N/2}^{N/2-1}{k}^2|S\left[k\right]{|}^2$

[0066] are pre-computed.

[0067] The Fourier shift theorem shown in equation 1 suggests that the phase change caused by motion is a linear function of spatial frequency k. In practice, the phase of a complex number is defined in the range of −π to π. When the observed phase change is calculated as Δφ[k]=arg S_dis[k]−arg S_ref[k], it may contain discontinuities (e.g., jumps) that must be removed or unwrapped. In one embodiment, a simple phase unwrapping consisting of replacing jumps greater than π with their 2π complements is used. This unwrapping works well when a SNR threshold is used to cut off noisy data points to ensure that the phase variation between adjacent k-space data points is sufficiently smooth. In one embodiment, the SNR threshold is approximately ten.

[0068] FIG. 2 shows a flow diagram of the steps taken to determine displacement from the navigator echo phase when there is no rotation or dilation. Those skilled in the art will recognize that some of the steps shown being performed sequentially can be performed in parallel. A reference navigator echo is acquired (step 200). Displaced navigator echoes are acquired (step 202). The k|S[k]|²and $\sum _{k=-N/2}^{N/2-1}{k}^2|S\left[k\right]{|}^2$

[0069] are calculated (step 204) and the phase change Δφ[k] is computed (step 206). If there is a jump in Δφ[k] that is greater than a threshold (step 208), a phase unwrapping is performed (step 210). The threshold in one embodiment is π. If jumps greater than π occur, the jumps are replaced with their 2π complements. The estimated displacement is then calculated (step 212). If there are more data points (step 214), steps 206 to 214 are repeated.

[0070] Computer simulation and motion extraction experiments were performed to evaluate the accuracy and computational efficiency of the present invention and compared to the Ahn and Cho method. The effectiveness of these methods was then demonstrated by correcting for motion effects on a resolution motion phantom with obtained displacement information.

[0071] Turning now to FIG. 3, Monte-Carlo simulation was used to evaluate the accuracy of the present invention against Ahn and Cho's method for an idealized navigator profile at different noise levels. In each run, white Gaussian noise was independently added to the reference profile and a shifted profile of known displacement. These navigator profiles were processed by both methods, and the obtained displacement estimates were subtracted from the true displacement to determine the estimation errors. The standard deviations of these errors were used to compare the accuracy of the algorithms. FIG. 3 shows the standard deviations of estimation errors averaged over 5000 Monte-Carlo runs at various contrast-to-noise ratios (CNR's). A CNR is defined as the ratio of navigator profile peak and noise standard deviation. These error curves match well with theoretical curves. In FIG. 3, curve 300 is the theoretical error curve and curve 302 is the simulated estimation error curve for the method of the present invention Curve 304 is the theoretical curve and curve 306 is the simulated estimation error curve for the Ahn and Cho method. It can be seen that the present method outperforms the Ahn and Cho method.

[0072] Turning now to FIGS. 4aand 4b,a motion phantom was used to produce a known motion waveform for comparison with navigator measurements. The phantom consisted of a computer-controlled stepper motor, which drove a stagecoach in the direction of the magnet bore. A bottle of water was fixed on the stagecoach as the imaging object. To simulate the SI motion of human diaphragm, the motion waveform was set to a sinusoid with peak-to-peak amplitude of 4 cm and period of 4 sec. A pencil-beam navigator pulse sequence was used to collect navigator data. The sequence was run on a 1.5T MR imaging system (GE Medical Systems, Milwaukee, Wis.) and had the following parameters: TR=150 ms, TE=5 ms, flip angle=15°, bandwidth=15.63 KHz, FOV=30 cm, 256-point navigator resolution. Navigator echoes were obtained and post-processed on a Sun Ultra2 workstation (Sun Microsystems, Palo Alto, Calif.).

[0073] FIG. 4aillustrates the displacement obtained with Ahn and Cho's method on the motion phantom and FIG. 4billustrates the displacement obtained using the method of the present invention. Curves 402 and 406 illustrate the true motion. Curve 404 illustrates the estimated displacement obtained using the Ahn and Cho method and curve 408 illustrates the estimated displacement obtained using the method of the present invention. The average CPU times to process a 256-point phantom navigator echo on a Sun Ultra2 computer were 0.53 msec for the Ahn and Cho method and 0.57 msec for the present invention. The absolute errors averaged over 256 echoes were 2.0 mm for the Ahn and Cho method and 0.5 mm for the method of the present invention. The method of the present invention has four times the error reduction. The errors in peak-to-peak motion amplitudes averaged over 10 cycles of the sinusoid were 5.5 mm for Ahn and Cho's method and 1.3 mm for the method of the present invention, which results in four times the error reduction.

[0074] Turning now to FIGS. 5a-5d,displacement information obtained with the method of the present invention and Ahn and Cho's method were used to retrospectively correct motion effects on a resolution motion phantom. The phantom was programmed to undergo a sinusoidal motion with peak-to-peak amplitude of 4 cm and period of 4 sec. Motion correction was done in k-space by offsetting the motion phase shift according to equation 1. The effectiveness of motion correction using displacement estimates obtained is illustrated in FIGS. 5a-5d.

[0075] FIG. 5ais a coronal image of the stationary phantom. FIG. 5bshows the image of the same phantom in motion obtained without motion correction, and a substantial amount of ghosting and blurring can be observed. These artifacts are suppressed considerably by applying motion correction to the same raw k-space data. FIG. 5cis the corrected image using motion derived from Ahn and Cho's method, and FIG. 5dis the corrected image using motion derived from the method of the present invention. FIG. 5dshows a better motion suppression compared to FIG. 5c,indicating a more accurate motion extraction using the present invention.

[0076] Turning now to FIGS. 6aand 6b,the Ahn and Cho method and the method of the present invention were also applied to data from the diaphragm of a human volunteer to detect respiratory motion. In this case the true motion is unknown, so an accurate but less efficient image-space least-squares fitting algorithm was used to provide the standard for comparison. The corresponding displacements obtained are illustrated in FIGS. 6aand 6b.FIG. 6aillustrates the estimated displacement using the Ahn and Cho method. Curve 600 is the image-space least squares fitting method and curve 602 is the estimated displacement using the Ahn and Cho method. FIG. 6billustrates the estimated displacement using method in accordance with the present invention. Curve 604 is the image-space least squares fitting method and curve 606 is the estimated displacement using the method of the present invention.

[0077] To detect multiple motion components such as rotation, dilation and displacement using a single navigator excitation, more sophisticated navigator echoes sensitized to motion such as defined by a k-space circle or spiral are needed. The trajectories are chosen to make processing easier. Examples of k-space trajectories include trajectories defined by orthogonal functions spanned over the duration of the navigator echo acquisition, such as sinusoidal functions such as (sin 2π(t/T−½), cos 2n(t/T−½), 1) (see FIG. 7a), Legendre functions (2t/T−1, ½(3(2t/T−1)²−1), 1) (see FIG. 7c), or higher order functions (see FIGS. 7band 7d), where T is the sampling duration. Since image energy is concentrated in the center of k-space, functions that provide trajectories through or close to the center are more desirable. For example, the Legendre set which passes near the center and higher order functions permitted by gradient switching power may be more desirable (see FIGS. 7b,7c,and 7d). Another approach to selecting a k-space trajectory is to use non-overlapping functions (see FIG. 7e), or some hybrid such as a k-space spiral trajectory (see FIG. 7f). In such cases, motion components are encoded in phase and/or magnitude of navigator data.

[0078] In an alternate embodiment of the present invention, the displacement is extracted from the navigator signal phase when motion is composed of translation, rotation, and dilation. General motion (i.e., rotation, dilation and displacement) is detected by first exciting the object whose motion is to be measured using selective volumetric excitation. Selective volumetric excitation of an object is done by transmitting a slab-selective pulse centered on the excitation frequency of the object to produce an excited signal that is sampled. If water is present, a spectral saturation pulse centered on water frequency is transmitted prior to the slab-selective saturation pulse to eliminate water from the excited signal. Spatial saturation RF pulses may also be transmitted to eliminate signals that may interfere with the excited signal.

[0079] Once the object has been excited, the excited signal is sampled in k-space using one or more k-space trajectories that is sensitized to the motion of interest of the object to obtain k-space data. For example, the k-space trajectory shown in FIG. 7fis used to detect rotation and dilation of an object. The motion parameters are then extracted from the k-space data using the weighted k-space least squares method described hereinbelow. General motion experienced by a heart (e.g., coronary arteries) shall be used to illustrate this embodiment. Those skilled in the art recognize that the present invention can be used in other situations where there is general motion. For example, motion of the brain may be detected using the present invention.

[0080] The heart is made of chamber blood, myocardium, and epicardial fat. The motion of coronary arteries correlates directly with the motion of the myocardium and epicardial fat, but not the motion of chamber blood. The myocardium and blood have the same resonance frequency, which makes it difficult to excite the myocardium to determine motion without getting interference that is caused by exciting the chamber blood. However, by spatially selective excitation of epicardial fat, coronary artery motion can be directly monitored because the epicardial fat can be excited without exciting the chamber blood. Turning now to FIG. 8, the overall steps to extract displacement are shown. A reference navigator echo is acquired (step 802). A displaced navigator echo is acquired (step 804). Volume-selective excitation is used to acquire the navigator echoes. For coronary motion, volume-selective excitation is accomplished by transmitting a spatial spectral selective pulse to eliminate signals from the chest wall and to excite epicardial fat to produce an excited signal.

[0081] The spatial spectral selective pulse may be implemented several ways. For example, it can be implemented by transmitting a spatial saturation pulse to suppress signals outside the six walls of the heart (i.e., the interior, superior, right, left, anterior, and posterior walls) and transmitting a one dimensional selective pulse centered on epicardial fat frequency to excite the epicardial fat. Alternatively, the implementation may be transmitting a spatial saturation pulse to suppress signals outside four walls of the heart and transmitting a two dimensional selective train in the direction of the heart that was not saturated and centered on the epicardial fat frequency to excite the epicardial fat. For example, if the superior and interior walls are not saturated, the direction of the heart chosen is along the S-I direction. In an alternate embodiment, the implementation is transmitting a spatial saturation pulse to suppress signals outside two walls of the heart and transmitting a three dimensional selective pulse in the two directions of the heart not saturated and centered on the epicardial fat frequency to excite the epicardial fat. For example, if the two walls that are saturated are the right and left walls, the selective pulse is transmitted in the S-I direction and the A-P direction. If the two walls that are saturated are the anterior and posterior walls, the selective pulse is transmitted in the S-I direction and the R-L direction. If the two walls that are saturated are the interior and superior walls, the selective pulse is transmitted in the A-P direction and the R-L direction. In a further alternate embodiment, the implementation is transmitting a four dimensional selective pulse in the S-I, A-P and R-L directions of the heart and centered on the epicardial fat frequency to excite the epicardial fat. The pulses can be transmitted in a single pulse train or as separate pulses. The delay between pulses should be as small as possible. The excited epicardial fat signal is then sampled in k-space. A piece-wise linear k-space trajectory along the k_x, k_y, and k_zaxes that may be used is illustrated in FIG. 9. It should be noted that other trajectories such as a k-space circle or spiral may also be used.

[0082] FIG. 10ais an axial image of heart volume without water saturation and FIG. 10bis an axial image of heart volume with water saturation. FIG. 10cis an axial image of heart volume with water saturation and spatial fat saturation (i.e., the figure only contains epicardial fat signal from the heart). The data was gathered using a 1.5T MR system (General Electric, Milwaukee, Wis.). The following parameters were used: FOV=30 cm, slice thickness=90 mm, TR=58ms, flip angle=30°, BW=15.63 kHz, 1.17 mm resolution. FIG. 11 shows a sample of the acquired navigator echo in one TR. The three echo peaks correspond to the three line segments along the k-space axes. The phase changes are proportional to the corresponding displacement of the epicardial fat.

[0083] Returning now to FIG. 8, the rotation angle and scaling factors due to contraction and respiration are determined from the k-space magnitude data as described below (step 806). The displacement vector from the k-space phase data is then determined (step 808). The rotation angle, scaling factors, and displacement vector of general motion of a heart is determined as follows. Let x be the position vector (x, y, Z)^T, D_rthe displacement vector associated with respiration, and the dilation/contraction along the xyz axis as a_x, a_y, a_z. The respiratory motion component can be expressed in a matrix form: $\begin{array}{cc}{x}^{\prime }=\left(\begin{array}{ccc}{a}_x& 0& 0\\ 0& a_y& 0\\ 0& 0& a_z\end{array}\right)\left(x-D_r\right)\equiv C_r\left(x-D_r\right)& \left(7\right)\end{array}$

[0084] Cardiac motion can be expressed conveniently in a coordinate system XYZ, with Z the long axis of the heart and XZ in the septum plane. The transformation from the xyz coordinate system to the XYZ coordinate system consists of 1) rotating about the x-axis until the z-axis becomes the Z-axis and 2) rotating about the Z-axis until the x-axis gets to the septum (XZ) plane. Both rotating angles can be easily measured from scout scans of the heart. The Jacobian associated with this transformation (J) is the product of these two rotations and can be obtained from a scout scan. The rotation and contraction associated with cardiac motion, adding the respiratory component, is composed of transforming from xyz to XYZ, contraction/dilation, rotation, and transforming from XYZ back to xyz:

[0085] x″=J⁻¹(R_cC_cJ−D_cJ)C_r(x−D)≡A(x−D)  [8]

[0086] In equation 8, R_cis the rotation matrix pertaining to cardiac motion, C_cis the contraction matrix pertaining to contraction, C_ris the respiration matrix pertaining to respiration, D_cis the displacement due to cardiac motion, and D is the displacement. The k-space expression of the motion effects f_m(x)=f(A(x−D)) is:

[0087] F_m(k)=e^{ik D}F((A⁻¹)^Tk)  [9]

[0088] The six motion parameters in the matrix A (i.e., the rotation angle in R_c, the two scaling factors in C_c, and the three scaling factors in C_r) are determined from the magnitude k-space data by using a least squares minimization:

[0089] min{Σ_k|S(k)/σ|²(|F_m(A^Tk)|−|F(k)|)²}  [10]

[0090] where the factor $\frac{2\pi }{N}$

[0091] is factored into the unit for k.

[0092] The motion parameters in matrix A can be obtained by sampling a k-space volume where A^Tk might be located around k, and a resampling search as known in the art. Once the matrix A is determined, the displacement vector D can be obtained from the phase difference using the least squared approach:

[0093] min{Σ_k|S(k)/σ|²(ΔΦ(k)−kD)²}.  [11]

[0094] FIGS. 12a-cshows the detected waveforms of a volunteer in the right-left (RL), anterior-posterior (AP), and cranial-caudal (superior-inferior, SI) directions without the effects of rotation and dilation taken into account. FIG. 12ais the RL direction waveform, FIG. 12bis the AP direction waveform, and FIG. 12cis the SI direction waveform. The waveforms of FIG. 12 were extracted from signal phases around the peaks of the navigator echo of FIG. 11.

[0095] Respiratory and coronary motion detection obtained with the method of the present invention on healthy human volunteers was performed using a 1.5T MR system (General Electric, Milwaukee, Wis.). The navigator pulse sequence used for spatially selective excitation of epicardial fat (i.e., volumetric excitation) was: FOV=180 mm, slice thickness=90 mm, 1-4-6-4-1 binomial spatial-spectral pulse centered on water frequency, flip angle=30°, BW=15.63 kHz, TR=40 ms. Data was acquired successively with and without breath-holding to correlate the respiratory motion of the diaphragm with coronary motion. Coronary imaging is often performed by gating data acquisition to the ECG signal. The coronary motion was sampled immediately after arrival of the ECG signal.

[0096] Bulk motion of the diaphragm and coronary arteries without breath-holding obtained using the method of the present invention are illustrated in FIGS. 13a-d.The sampling frame rate in the figures is 60 ms/frame. FIG. 13ashows the diaphragm motion in the SI direction. FIG. 13bshows cardiac motion in the SI direction. FIG. 13cshows cardiac motion in the AP direction and FIG. 13dshows cardiac motion in the RL direction. The figures show that coronary motion is affected by respiratory motion primarily in the SI direction as expected, but also in the RL and AP direction. This effect is absent with breath-holding, as shown in FIGS. 14a-d.The sampling frame rate in the figures is 60 ms/frame. FIG. 14ashows the diaphragm motion in the SI direction. FIG. 14bshows cardiac motion in the SI direction, FIG. 14cshows cardiac motion in the AP direction, and FIG. 13dshows cardiac motion in the RL direction. The small drift of the diaphragm position in FIG. 14ais due to imperfect breath-holding. The effects of the imperfect breath-holding observed coronary motion is seen in FIGS. 14b-d.This demonstrates the high motion sensitivity of the volumetric navigator echo method of the present invention.

[0097] FIG. 15 shows a full cycle of cardiac motion in the RL direction with ECG gating and breath-holding obtained using the method of present invention. The sampling frame rate in the figure is 40 ms/frame. The rest period of approximately 200 milliseconds and the optimal delay time of approximately 800 milliseconds can be seen in FIG. 15. The heart rate of the volunteer was 50 beats per minute. The pattern of coronary motion shows a gradual slowing of motion before the rest period and substantial motion after the rest period. The rest period observed in volunteers with tachycardia is shorter (not shown). These observations are consistent with known cardiac physiology.

[0098] The volumetric navigator method of the present invention provides accurate bulk translation information of the coronary arteries. It can be used to detect the respiratory motion of the heart directly and may replace pencil-beam navigator echo for detecting respiratory motion. The methods of the present invention is also sensitive to cardiac motion and can identify the optimal delay time and the rest period within the cardiac cycle. It has the potential of replacing the ECG signal for triggering data acquisition and can provide the optimal data sampling window to minimize both respiratory and cardiac motion artifacts.

[0099] The foregoing description of various embodiments of the invention has been presented for purposes of illustration and description. It is not intended to be exhaustive or to limit the invention to the precise embodiments disclosed. Numerous modifications or variations are possible in light of the above teachings. The embodiments discussed were chosen and described to provide the best illustration of the principles of the invention and its practical application to thereby enable one of ordinary skill in the art to utilize the invention in various embodiments and with various modifications as are suited to the particular use contemplated. All such modifications and variations are within the scope of the invention as determined by the appended claims when interpreted in accordance with the breadth to which they are fairly, legally, and equitably entitled.