Reconstruction of images of the beating heart using a highly constrained backprojection

(Cross-reference of related applications) This application is filed on September 22, 2005, and is a US patent provisional application No. 60 / 719,445 with the title of the invention "HIGHLY CONSTRAINED IMAGE RECONSTRUCTION METHOD". U.S. Patent Application No. 60 / 738,444, filed on November 21, 2005, whose title is "IMAGE RECONSTRUCTION METHOD FOR CARDIAC GATED MAGNETIC RESONANCE IMAGING" based on.

(Background of invention) The field of the present invention is nuclear magnetic resonance imaging.<u style="single">About how the system works</u>.. More specifically, the present invention relates to image reconstruction from cardiac gate magnetic resonance acquisition.

A magnetic field in which substances such as human tissues are uniform (polarization magnetic field B)₀), The individual magnetic moments of the spins in the tissue try to align by this polarization magnetic field, but around them precess in random order at their inherent Larmor frequencies. This substance, that is, this structure, is in the xy plane and has a magnetic field close to the Larmor frequency (excitation magnetic field B)₁), The net alignment moment Mz rotates or "tilts" with respect to its xy plane, creating a net lateral magnetic moment Mt. Excitation signal B₁After the end, a signal is output by its excited spins, which can be received and processed to form an image.

When creating an image using these signals, the magnetic field gradient (G)_x, G_y, And G_z) Is used. Typically, the area to be imaged is scanned by a continuous measurement cycle in which these gradients fluctuate according to the particular positioning method used. Each measurement is called a "view" in the industry, and the number of views determines the resolution of the image. The resulting set of received NMR signals or views or k-space samples is digitized and processed to reconstruct the image using one of many well-known reconstruction techniques. The total scan time is, in part, determined by the number of measurement cycles or views acquired for a single image, so by reducing the number of views acquired at the expense of image resolution. Can be shortened.

The most common method for obtaining an NMR data set capable of reconstructing an image is what is called a "Fourier transform" imaging technique or a "spin-warp" technique. This technique was developed by WAEdelstein et al., "Spin-Warp NMR Imaging and Applications to Human Whole-Body Imaging" (Spin-Warp NMR Imaging and Applications to Human Whole-Body Imaging).<u style="single">Physics in Medicine and Biology</u>, Vol.25, p.751-756 (1980)). The method uses a variable amplitude phase-coded pulsed field gradient pulse to phase-code the spatial information in this gradient direction before acquiring the NMR signal. In a two-dimensional embodiment (2DFT), for example, a phase-coded gradient (G) along one direction._y) Is applied to code the spatial information in that direction, and then the read magnetic field gradient (G) in one direction orthogonal to this phase coding direction._x) Is present, the signal is acquired. This reading gradient, which is present during the acquisition of the spin-echo, encodes the spatial information in its orthogonal direction. In a typical 2DFT pulse sequence, the phase coded physics gradient pulse G_yThe magnitude of is incremented (G) in a series of views captured during the scan_y) Is increased. In a three-dimensional embodiment (3DFT), before each signal read, a third gradient (G)_z) Is applied and the coding is done along its third axis. This second phase coded gradient pulse G_zThe size of is also changed stepwise through the value being scanned. These 2DFT and 3DFT methods sample the k-space with a rectilinear pattern as shown in FIG. 2, and the k-space samples are placed in a Cartesian grid.

Magnetic resonance angiography (MRA) utilizes the phenomenon of magnetic resonance to create images of the human vascular system and heart. To enhance the diagnostic ability of MRA, a contrast agent such as gadolinium can be injected into the patient prior to MRA scanning. As described in US Pat. No. 5,417,213, this contrast amplification (CE) MRA method (trick) obtains a central k-space view when the contrast agent bolus is flowing through the vasculature of interest. That is. Collection of the central line of k-space during peak arterial amplification is important for the success of the CEMRA test. If the centerline of k-space is acquired before the contrast agent arrives, severe image artifacts can limit the diagnostic information in the image. On the other hand, the arterial image obtained after the passage of the peak arterial contrast medium is obscured by the amplification of the veins. In many anatomical sites, such as the carotid and renal arteries, the separation between arterial and venous amplification can be as short as 6 seconds.

Acquisition of MRA data is timed so that the central region of k-space is acquired when the contrast bolus reaches the artery of interest. The ability to time the arrival of contrast media varies considerably, and for dynamic studies depicting the separation and amplification of arteries and veins, it is useful to obtain a series of MRA images in many applications. A series of temporal images is also useful for observing the delayed vascular filling pattern caused by the disease. This requirement is "Time-Resolved Contrast Amplification 3D MR Angiography" by Korosec F., Frayne R, Grist T., Mistretta C. (<u style="single">Magn.Reson.Med.</u>Partially addressed by acquiring a series of time-resolved images using the 3D "Fourier" acquisition method, as described in 1996; 36: 345-351) and US Pat. No. 5,713,358. ..

More recently, projection reconstruction methods have been used to obtain time-resolved MRA data, as described in US Pat. No. 6,487,435. The projection reconstruction method, sometimes referred to as "radial" acquisition, has been known since the beginning of magnetic resonance imaging. The projection reconstruction method does not sample the k-space with a linear scan pattern as shown in FIG. 2 as is done by the Fourier imaging method, but rather the center of the k-space as shown in FIG. Get a series of views that sample radial lines that extend outward from. The number of views required to sample the k-space determines the length of the scan, and if an insufficient number of views are obtained, the reconstructed image will have streak artifacts. Created. The technique disclosed in Japanese Patent No. 6,487,435 captures continuously undersampled images in an interleaved view and shares the surrounding k-space data between the continuous images. Such streaking is reduced.

For example, as described in US Pat. No. 6,710,686, there are two methods used to reconstruct an image from the obtained set of k-space projection views. The most common method is to regrid a k-space sample from a radial sampled locus position to a Cartesian grid. The image is then reconstructed by performing a 2D or 3D Fourier transform on the gridded k-space sample. The second way to reconstruct the image is to transform the radial k-space projection view into Radon space by Fourier transforming each projection view. The image is reconstructed from those signal projections by filtering and back-projecting these signal projections into the field of view (FOV). As is well known in the art, if the acquired signal projections are insufficient in number to satisfy the Nyquist sampling theorem, the reconstructed image will have streak artifacts.

The standard back projection method is shown in Figure 4. Each acquired signal projection profile 10 is back-projected onto the FOV 12 within the profile 10 by projecting each signal sample 14 through the FOV 12 along the projection path indicated by the arrow 16. When back-projecting each signal sample 14 onto the FOV 12, it has no a priori information about the subject, the NMR signal in the FOV 12 is homogeneous, and the signal sample 14 is on each pixel through the projection path. Assume that they are evenly distributed. For example, FIG. 4 shows the projection path 18 for a single signal sample 14 in a single projection profile 10 as the projection path 18 passes through N pixels in the FOV 12. The signal value (P) of this signal sample 14 is evenly divided among these N pixels.<maths num="1"><img file="JP5113062B2_D0001.tif" /></maths>(Here, S_nIs the NMR signal value distributed to the nth pixel in the projection path having N pixels. )

Obviously, the assumption that the FOV12 NMR signal is even is incorrect. However, as is well known in the art, if a certain filtering correction is made to each signal profile 10 and a sufficient number of filtered profiles are obtained with the corresponding number of projection angles. The error caused by this false assumption is minimized and image artifacts are suppressed. A typical filter-corrected back-projection method for image reconstruction requires 400 projections for a 256 x 256 pixel 2D image and 203,000 for a 256 x 256 x 256 pixel 3D image. Projection is required. If the method described in U.S. Pat. No. 6,487,435 cited above is adopted, the number of projected views required for these same images can be reduced to 100 (2D) and 2000 (3D). it can.

When imaging an artery, such as a coronary artery, the movement of the beating heart becomes a problem. To reduce motion artifacts in MRI or MRA images, it is customary to cardiac gate the acquisition of a view using an ECG signal that displays the cardiac phase. For example, as described in US Pat. No. 5,329,925, groups or segments of views are acquired in each of one or more cardiac phases during each cardiac cycle. For example, eight different views are acquired in a particular heart phase, and after 16 heartbeats, a total of 8 × 16 = 128 different views are acquired and the image is reconstructed from those views. Since a single breath-hold is typically 16 to 20 heartbeats, it is very difficult to get all the data during the breath to avoid artifacts caused by respiratory movements. desirable.

Proper single-slice 2D images can be obtained in one or more cardiac phases during a single breath-hold using projection reconstruction and view sharing, but conventional methods have each cardiac phase. Was not fast enough to capture a single 3D image or multiple 2D slices. Such images are needed when the subject of examination does not lie in a single 2D plane (eg, coronary arteries) and multi-slice or 3D image acquisition is required.

(Summary of invention) The present invention produces a cardiac gate MR image.<u style="single">How the system works</u>And, in particular, improve the quality of highly undersampled images acquired at a particular cardiac phase.<u style="single">How the system works</u>Is. A series of undersampling image frames are acquired at the selected cardiac phase during continuous heartbeats. Views acquired during continuous heartbeats are used to sample interleaved trajectories in k-space and combine them to reconstruct a composite image that depicts an object in a selected cardiac phase. This composite image is used in a highly limited back projection of each projection view by weighting the distribution of the back projection signal samples.

The findings of the present invention use significantly fewer acquisition views if the a priori information of the NMR signal contour in FOV12 is used in the back-projection image reconstruction process instead of the assumed uniform signal contour. It is possible to generate a good quality frame image. For example, referring to FIG. 5, it can be seen that the signal contour of FOV12 includes structures such as blood vessels 18 and 20. In practice, when the back-projection path 8 penetrates these structures, the more accurate distribution of the signal sample 14 to each pixel is weighted by weighting the distribution as a function of the NMR signal contour known at that pixel position. Achieved. As a result, the majority of signal samples 14 are distributed at pixels that intersect structures 18 and 20. For the back projection path 8 with N pixels, this can be expressed as:<maths num="2"><img file="JP5113062B2_D0002.tif" /></maths>(Here, P is the NMR signal sample value, and C_nIs the signal value of the composite image at the nth pixel along the back projection path. ) The molecule of formula (2) weights each pixel using the corresponding NMR signal value in the composite image, and the denominator is the sum of the composite images, with the total back-projection signal sample reflecting the sum of projections for the image frame. The value is normalized so that it is not multiplied. This normalization can be done separately for each pixel after the back projection is performed, but in many clinical uses it is much easier to normalize the projection P before the back projection. It should be noted. In this case, the projection P is normalized by dividing by the corresponding value Pc in the projection through the composite image at the same view angle. The normalized projection P / Pc is then back-projected, and the resulting image is then multiplied by the composite image.

A 3D embodiment of the present invention is illustrated in FIG. 6 for a single 3D projected view characterized by view angles θ and φ. This projected view is Fourier transformed to form a single signal contour, which is back-projected along axis 16 and extends to the radon plane 21 at a distance r along the back-projection axis 16. Instead of filtering backprojection, where the projected signal contours are filtered and evenly distributed along axis 16 into a continuous radon plane, the projected signal contours use the information in the composite image to create the radon plane 21. Will be distributed to. The composite image in FIG. 6 includes blood vessels 18 and 20. Weighted signal contour values are placed at image positions x, y, z in the radon plane 21 based on their intensities at the corresponding positions x, y, z in the composite image. This is a simple multiplication of the voxel value of the corresponding composite image and the signal profile value. This product is then normalized by dividing this product by the profile value from the corresponding image space profile formed from the composite image. The formula for 3D reconstruction is:<maths num="3"><img file="JP5113062B2_D0003.tif" /></maths>(Here, the sum (Σ) is the total projection in the time frame, and the x, y, z values in a specific radon plane are the profile values P (r) at the appropriate r, θ, φ values for that plane. , θ, φ). Pc (r, θ, φ) is the corresponding profile value from the composite image, and C (x, y, z) r, θ, φ is (r, φ). It is a composite image value in θ, φ).)

Another finding of the present invention is that this image reconstruction method can be used advantageously in cardiac gate acquisition to allow a series of undersampled frame images to be acquired in the same cardiac phase. By interleaving the view of the continuous image frame acquisition, the views from the continuous image frame can be combined and used to reconstruct a higher quality composite image. This composite image is then used in the aforementioned back-projection reconstruction of each image frame.

Another aspect of the invention is the reconstruction of image frames acquired during cardiac gate scans using a 3D hybrid projection reconstruction pulse sequence. A projected view is taken to sample a k-space with a radial trajectory within a 2D slice, and phase coding is used to get multiple slices along the axial direction. The composite image is reconstructed for each of the multiple slice positions, and these composite images are used during the back-projection reconstruction of the 2D slice within each image frame.

(Detailed description of preferred embodiments) Particularly with reference to FIG. 1, preferred embodiments of the present invention are employed in MRI systems. The MRI system comprises a workstation 110 having a display 112 and a keyboard 114. Workstation 110 comprises processor 116, which is a commercially available programmable machine running a commercially available operating system. Workstation 110 provides an operator interface that allows scan instructions to be entered into the MRI system.

Workstation 110 is coupled to four servers: pulse sequence server 118, data acquisition server 120, data processing server 122, and data storage server 23. In a preferred embodiment, the data storage server 23 is run by the workstation processor 116 and associated disk drive interface circuitry. The remaining three servers 118, 120 and 122 are run by separate processors mounted in a single enclosure and interconnected using a 64-bit backplane bus. The pulse sequence server 118 uses a commercially available microprocessor and a commercially available 4-communication controller. Both the data acquisition server 120 and the data processing server 122 use the same commercially available microprocessor, and the data processing server 122 further includes one or more array processors based on the commercially available parallel vector processor.

The workstation 10 and the processors 118, 120 and 122 of the server 118, 120 and 122 are connected to the serial communication network. This serial network transmits data downloaded from workstation 110 to servers 118, 120 and 122, as well as tag data communicated between servers and between workstations and servers. In addition to this, a high-speed data link is provided between the data processing server 122 and the workstation 110 to transmit image data to the data storage server 23.

The pulse sequence server 118 functions to operate the gradient system 24 and the RF system 26 in response to program elements downloaded from workstation 110. The gradient waveform required to perform the specified scan is generated and given to the gradient system 24, which excites the gradient coil in the assembly 28 and uses the magnetic field gradient for position encoding of the NMR signal. G_x, G_y, And G_zTo generate. The gradient coil assembly 28 forms part of a magnet assembly 30 with a polarization magnet 32 and a full body RF coil 34.

The RF excitation waveform is applied to the RF coil 34 by the RF system 26 to perform a designated magnetic resonance pulse sequence. The responsive NMR signal detected by the RF coil 34 is received by the RF system 26 and amplified, demodulated, filtered and digitized under the command of the command generated by the pulse sequence server 118. The RF system 26 comprises an RF transmitter that produces a wide range of RF pulses used in MR pulse sequences. The RF transmitter produces RF pulses with the desired frequency, phase, and pulse amplitude waveform in response to scan instructions and instructions from the pulse sequence server 118. The generated RF pulse can be applied to the whole body RF coil 34 and can be applied to one or more local coils or coil arrays.

The RF system 26 corresponds to a plurality of corresponding local coils or in a coil array. It also includes one or more RF receiver channels that can be connected to multiple coil elements. Each RF The receiver channel is RF amplified to amplify the NMR signal received by the connected coil. Quadrature phase inspection that detects and digitizes the I and Q quadrature phase components of the instrument and the received NMR signal. Equipped with a dispenser. The magnitude of the received NMR signal is thus the sum of the squares of the I and Q components. It can be obtained at any sampling point by the root,<maths num="4"><img file="JP5113062B2_D0004.tif" /></maths>The phase of the received NMR signal can also be obtained.<maths num="5"><img file="JP5113062B2_D0005.tif" /></maths>

The pulse sequence server 118 optionally receives patient data from the physiological acquisition controller 36. The controller 36 receives signals from several different sensors connected to the patient, such as ECG signals from electrodes or breathing signals from bellows. The pulse sequence server 118 typically uses such signals to synchronize or "gate" scan performance to the patient's breathing or heartbeat.

The pulse sequence server 118 is also connected to a scan room interface circuit 38 that receives signals from various sensors related to the patient and the condition of the magnet system. It is also through the scan room interface circuit 38 that the patient alignment system 40 receives commands to move the patient to the desired position during the scan.

It should be clear that the pulse sequence server 118 provides real-time control of the MRI system elements during the scan. As a result, the hardware element needs to operate with program instructions that are executed in a timely manner by the run-time program. The command component of the scan instruction is downloaded from workstation 110 in the form of an object. The pulse sequence server 118 includes a program that receives these objects and converts them into objects used in the run-time program.

The data acquisition server 120 receives the digitized NMR signal sample generated by the RF system 26. The data acquisition server 120 operates in response to command components downloaded from workstation 110, receives real-time NMR data, and provides buffer storage so that the data is not lost due to data overruns. Depending on the scan, the data acquisition server 120 only passes the acquired NMR data to the data processor server 122. However, for scans that require information derived from the acquired NMR data to further control the performance of the scan, the data acquisition server 120 is programmed to generate such information and propagate it to the pulse sequence server 118. Will be done. For example, during a prescan, NMR data is acquired and used to calibrate the pulse sequences performed by the pulse sequence server 118. It also captures the navigator signal during scanning and uses it to adjust the operating parameters of the RF or gradient system, or view the k-space sampled. order) can be controlled. The data acquisition server 120 can also be used to process NMR signals used to detect the arrival of contrast media during an MRA scan. In all of these examples, the data acquisition server 120 acquires NMR data and processes it in real time to generate the information used to control the scan.

The data processing server 122 receives the NMR data from the data acquisition server 120 and processes it according to the command components downloaded from the workstation 110. Such processing includes, for example, Fourier transforming the unprocessed k-space NMR data to generate a two-dimensional or three-dimensional image, applying a filter to the reconstructed image, and obtaining the NMR data. It can include performing back-projection image reconstruction of, calculating a functional MR image, calculating a motion or flow image, and the like.

The image reconstructed by the data processing server 122 is transmitted to the workstation 110 again and stored. The real-time image is stored in a database memory cache (not shown) from which the image can be output to an operator display 112 or display 42 located near the magnet assembly 30 and used by the physician in charge. The batch mode image or the selected real-time image is stored in the host database on the disk storage 44. When such an image is reconstructed and transferred to storage, the data processing server 122 notifies the data storage server 23 on the workstation 110. Operators can use workstation 110 to store images, generate film, or send images to other facilities over a network.

To carry out a preferred embodiment of the invention, NMR data is acquired using projection reconstruction or radial pulse sequences as shown in FIG. This is a fast gradient cold echo pulse sequence in which selectively and asymmetrically clipped sinc rf excitation pulses 200 are generated in the presence of a slice selection gradient 202. The flip angle of the rf pulse 200 is typically 30 ° -40 ° T₁Is set near the Ernst angle for shortened blood.

As described in more detail below, this pulse sequence is used to obtain a single 2D slice by sampling in a single k-space circular slice, or this pulse sequence is 204 in FIG. , 206 and 208, can be used to sample multiple circular k-space planes. When multiple 2D slices are acquired, the axial gradient 202 is the slab selection gradient following the phase coded gradient lobe 210 and the reverse polarity rewinder gradient lobe 212. This axial phase coded gradient 210 is stepped through the values during the scan to sample from each of the 2D k-space slices 204, 206 and 208.

Two in-plane readout gradients 214 and 216 are performed during the acquisition of the NMR echo signal 218 to sample k-space along the radial trajectory within the 2D plane 204, 206 or 208. These in-plane gradients 214 and 216 are perpendicular to the axial gradient and they are orthogonal to each other. As described in more detail below, during scanning, these in-plane gradients are stepped through a series of values to rotate the view angle of the radial sampling trajectory. Each in-plane reading gradient precedes the prefading gradient lobes 220 and 222, followed by the unwinding gradient lobes 224 and 226.

Those skilled in the art will appreciate that sampling loci other than the preferred linear locus extending from one point on the peripheral boundary of k-space to the opposite point on the peripheral boundary of k-space through the center of k-space may be used. Should be understood. One variant is to obtain a partial NMR echo signal 218, which is sampled along a trajectory that does not extend over the entire range of the sampled k-space volume. Another variant equivalent to the linear projection reconstruction pulse sequence is sampling along a curved path rather than a straight line. Such pulse sequences use, for example, "Fast Three Dimensional Sodium Imaging" by FE Boada et al. (MRM, 37: 706-715, 1997) and "Spiral Projection Imaging" by KV Koladia et al. There was a high speed 3D PC-MRA (Rapid 3D PC-MRA Using Spiral Projection) Imaging) (Proc.Intl.Soc.Magn.Reson.Med.13 (2005)) and JGPipe and Koladia Spiral Projection Imaging: a new fast 3D trajectory ( It is stated in Proc.Intl.Soc.Mag.Reson.Med.13 (2005)). The present invention is available in 3D versions as well as 2D versions of these sampling methods, and the term "pixel" as used herein is intended to refer to a position in either a 2D or 3D image. It should also be understood that it is.

The pulse sequence described above is used by the MRI system of FIG. 1 to acquire a series of cardiac gate frame images. In the first preferred embodiment, only a single 2D slice is obtained within each image frame. Referring to FIGS. 1 and 9, after the subject is placed in the bore of the MRI system and the ECG signal is coupled to the physiological acquisition controller 36, the system sends the ECG trigger signal as shown in decision block 300. stand by. When the trigger signal is received, a set of image frames is acquired, as outlined in 302. This is illustrated in FIG. 10, where the cardiac cycle is initiated by the ECG trigger signal at 304 and a set of 6 image frames 306-311 is acquired at a given time or "cardiac phase" during the subsequent RR interval. In a preferred embodiment, the acquisition of each image frame in process block 312 consists of 10 projected views, which are set to a view angle that samples the 2D k-space as uniformly as possible. This is a highly undersampled dataset, and performing typical image reconstruction in the absence of the present invention will result in very poor quality images.

When the last image frame is acquired during the RR interval, as determined by decision block 314, the system loops back and waits for the next ECG trigger signal in decision block 300. A similar set of image frames is retrieved during the next RR interval, but each of the 10 views captured for each image frame 306-311 during the subsequent RR interval is interleaved with the previously captured view. The exception is that it is done. This is shown in FIG. 11, the projected view indicated by the dotted line 230 is acquired during one RR interval, the projected view indicated by the dashed line 232 is acquired during the other RR interval, and is indicated by the solid line 234. The projected view is acquired in yet another RR interval. During a typical breath-hold, 16-20 sets of such interleaved projected views may be obtained.

As pointed out above, the views within each captured image frame are arranged to sample k-space as uniformly as possible, satisfying the Nyquist sampling criteria up to radius r, as shown in FIG. The combined and interleaved projections 230, 232, and 234 further sample the k-space as uniformly as possible, but sample the k-space more densely, satisfying the Nyquist sampling criteria up to a much larger radius R. As a result, when data was acquired for the total heart rate during breath-holding, as detected in decision block 316 of FIG. 9, a significant number of interleaved and uniformly distributed views for each cardiac phase are shown. Obtained in the synthetic dataset indicated by 10 318.

Still referring to FIG. 9, for each cardiac phase, a composite image is reconstructed using composite dataset 318, as shown in process block 320. During a typical breath-hold consisting of 16-20 heartbeats, the composite dataset contains 16-20 times the data contained within a single image frame and is an artifact using traditional filtering back-projection techniques. Images that do not reasonably include can be reconstructed. As shown in process block 321, the resulting composite image may be further edited or filtered to remove unwanted structures. This can be done by manually displaying the composite image to remove unwanted structures or automatically filtering the detectable structures or tissues.

As shown in process block 324, a series of image frames may be reconstructed here for each cardiac phase. The reconstruction of one image frame will be described below. An important aspect of the present invention is that a composite image of the cardiac phase is used and the frame image is reconstructed for this cardiac phase.

In particular, with reference to FIG. 12, the first step is to transform the k-space projections of the image frame (10 in the preferred embodiment) into Radon space by Fourier transform, as shown in process block 330. The result is a set of 10 signal profiles, as shown in FIG. As shown in process block 332, each of these signal profiles is back-projected to the VOI as depicted in path 8 of FIG. This back projection is weighted by the composite image, as previously explained with reference to Eq. (2). That is, the back projection value (P) at any pixel (n) is the same pixel size (C) in the composite image._n).

Back-projection signal value (S), as shown in process block 334_n) Is added to the reconstructed image frame. The system then loops back at decision block 336 and backprojects the next signal profile 10 as shown in process blocks 338 and 332. Therefore, the signal values (S) of all back-projected signal profiles 10._n) Is added to the image frame with a weight determined by the corresponding pixel value in the higher quality composite image. Composite images are of higher quality because they are reconstructed from much more projected views and have fewer artifacts. Also, the composite image is of higher quality because the projected view used for reconstruction is acquired over a much longer period of time. In general, the SNR of an image frame is proportional to the square root of the acquisition duration. The discovery of the present invention is that this unique, highly limited reconstruction process transfers the higher quality of the composite image to the image frame.

Returning to the flowchart of FIG. 9, one cardiac phase image frame is reconstructed with the corresponding composite image, and then the next cardiac phase image frame is reconstructed as shown in process block 340. When all cardiac phase image frames have been reconstructed, as determined by decision block 342, the image frames may be displayed in many ways as shown in process block 344.

The scan described above produces a series of image frames in each of the multiple cardiac phases. This is shown in FIG. 13, where each reconstructed image frame 345 is associated with a particular heart phase and a particular heartbeat. These image frames 345 may be displayed in many different ways. First, images may be displayed that depict objects in continuous cardiac phases at selected time points during breath-holding (ie, heartbeat). For example, if the subject is the human heart, images of continuous heart phases will show the structure of the heart that changes during a single heartbeat.

Image 345 at a particular cardiac phase can also be observed over continuous heartbeats. In this case, the movement of the heart is frozen and you can see how the structure changes over time. This display mode is particularly useful when the contrast agent is used and the series of images 345 describes the contrast agent flowing into the field of view. Such a contrast amplification embodiment of the present invention will be described below.

A preferred method of using contrast media in ECG gate scanning is depicted in FIG. This particular embodiment employs two breath-holds. As shown in process block 350, the first step is to command the patient to establish the first reference breath-hold. This may be done using a monitor device, eg, a monitor device disclosed in US Pat. No. 5,363,844, entitled "Breath-hold Monitor For MR Imaging". The monitor device provides visual feedback to the patient regarding respiratory movements. During this initial breath-hold, a series of cardiac phase image frames are acquired as described above, as shown in process block 352.

The contrast agent is then administered as shown in process block 354 and a second breath hold is reestablished at the reference position as shown in process block 356. For example, using the breath-holding monitor described above, the patient inhales and exhales until the feedback light on the monitor indicates that the reference position has been reached. As shown in process block 358, when the contrast agent flows into the field of view during this second breath-hold, another set of cardiac phase image frames is acquired.

Pre-contrast mask images are generated for each cardiac phase, as shown in process block 360. This is done by combining the total projections obtained by the first breath-hold in each cardiac phase and performing conventional filtering back-projection image reconstruction on the combined projections. For example, if 20 heartbeats occur during this first breath-hold and 10 interleaved projected views are obtained during each cardiac phase, a total of 10 x 20 = 200 projected views will be used. Then, each mask image is reconstructed. Thus, as shown in FIG. 15, a mask image 362 is generated for each cardiac phase.

An unmasked composite image is then reconstructed for each cardiac phase, as shown in process block 364. This combines all the interleaved projections acquired at each cardiac phase during the second breath-hold after contrast administration, and the combined (eg, 200) interleaved projections re-image the conventional filtering back-projection. This is done by performing the configuration. Thus, an unmasked composite image 366 is generated for each cardiac phase, as shown in FIG.

Here, a final composite image is generated for each cardiac phase, as shown in process block 368. This is done by subtracting the masked image 362 of each cardiac phase from the corresponding unmasked composite image 366. Thus, a composite image 370 is generated for each cardiac phase, as shown in FIG. The masked composite image 370 displays image pixels whose intensities are changed by the arrival of the contrast medium. Such image pixels may be arteries in MRA studies or ventricles in cardiac studies. Furthermore, it is clear that the same mask composite image 370 can be generated by subtracting the projected view before contrast agent administration from the corresponding projected view after contrast agent administration and reconstructing the mask composite image 370 from the differential projection view. There will be.

Here, a set of image frames can be reconstructed for each cardiac phase, as shown in process block 372. This is done using a masked composite image of the cardiac phase 370 and projections acquired during each heartbeat for the cardiac phase, as described above and as shown in FIG. To generate the desired sparse dataset for this highly limited back-projection reconstruction, the corresponding mask projection view is subtracted from the projection view used to reconstruct this image frame. Thus, as shown in FIG. 15, the image frame 374 is reconstructed for each heartbeat and each heart phase. These image frames can be displayed in many different ways, as described above and as shown in process block 380.

Another preferred embodiment of the present invention uses the multi-slice capability of the hybrid PR pulse sequence of FIG. 7 to acquire multi-slice image frames during each heartbeat and at each cardiac phase. Multiple continuous slices provide a 3D volume, and a maximum intensity projection (MIP) image can be generated from the 3D volume. This is important when a structure that does not lie completely in one 2D plane is being imaged. This multi-slice embodiment may be adopted in the contrast amplification acquisition described above with reference to FIG. 14, but here, with reference to FIG. 16, a multi-slice embodiment without contrast amplification is described. To. This embodiment is similar in many respects to the embodiment described above with reference to FIG. 10, where essentially the same steps are identified with the same reference number in FIG.

Specifically referring to FIGS. 16 and 17, a cardiac trigger scan is performed in which a series of image frames are acquired in multiple cardiac phases during each heartbeat during breath holding. During each heartbeat, a series of image frames 306-311 is acquired in process block 301. This block is similar to the block described above. However, instead of getting a single set of 10 projected views that samples a single k-space slice, two sets of 10 projected views are taken during each cardiac phase. Each set of 10 projected views is phase-coded differently along the axial gradient (Figure 7), and two adjacent 2D slices in k-space are sampled. As will be described, an image frame composed of three 2D slices is finally formed at each cardiac phase. Referring to FIG. 8, these slices include a central k-space slice 206 referred to below as "A" and two peripheral k-space slices 204 and 208 referred to below as "B" and "C". .. However, to reduce the scan time required to capture each image frame, only two of these three slices are captured during any single cardiac phase.

As best shown in FIG. 17, during each acquisition of the frame image at a particular cardiac phase, a central slice A is acquired along with one of the peripheral slices B or C. The sampling pattern is a pattern in which the central slice A is acquired, one of the peripheral slices B or C is acquired, and the other peripheral slices can be generated from the temporal adjacency acquisition at any cardiac phase. For example, in FIG. 17, the first cardiac phase image frame 380 acquired during the nth heartbeat of breath-holding acquires A and B slices while the pulse sequence of FIG. 7 is repeated 20 times. A and C slices are obtained at 382 and 384 during the same cardiac phase of n-1 and n + 1 heartbeats. Therefore, when reconstructing the image frame for the first cardiac phase of the nth heartbeat, by interpolating between the time-adjacent C slices obtained at 382 and 384 during the n-1 and n + 1 heartbeats. Slice C data is calculated.

Revisiting FIG. 16 again, after all image frames have been acquired for the total heartbeat during breath holding, as determined by decision block 316, the composite image from the acquired data, as shown in process block 323. The set is reconstructed. First, a one-dimensional Fast Fourier Transform is performed along the axial phase-coded axis of the acquired k-space dataset. Each of the resulting hybrid spatial datasets consists of three axial slices (A, B, and C) of the projected view, as shown in FIG. For each cardiac phase, three composite images are reconstructed, one for each slice A, B, and C. More specifically, for each cardiac phase, the projections of slice A obtained during whole breath holding are combined as shown in FIG. 17 for the second cardiac phase to form the synthetic A dataset 390. , The projections of slice B are combined to form the synthetic B dataset 392, and the projections of slice C are combined to form the synthetic C dataset 394. The projections combined to form the composite datasets 390, 392, and 394 are interleaved as described with reference to FIG. 11 and sample k-space slices A, B, and C substantially uniformly. Importantly, the synthetic datasets 390, 392, and 394 sample slices A, B, and C more densely than any of the image frames acquired during the cardiac phase. As a result, composite images may be reconstructed from the composite datasets 390, 392, and 394 formed using conventional image reconstruction methods for each cardiac phase. Each of the three slices is reconstructed separately using conventional 2D image reconstruction methods. This conventional method may be a filtering back projection or a 2D projection grid change in each slice, followed by a 2D Fast Fourier Transform.

Following the reconstruction of the composite image, the image frame for each cardiac phase is reconstructed, as shown in process block 325. Similar to the embodiments described above, this image reconstruction uses highly limited backprojection and composite images to improve signal-to-noise ratio and reduce image artifacts within highly undersampled image frames.

Reconstruction of each image frame can be done in many different ways. First, a limited set of cardiac phase A, B, and C phase-coded projection views may be Fourier transformed along the axial phase-coded gradient axis to form three slices. Each of the three slices obtained is composed of a limited set of projected views, and the slice image is reconstructed from the limited set using the method described above shown in FIG. That is, each projected view is Fourier transformed into Radon space, then back-projected using a composite image of the slices, and each back-projected value is weighted. In this way, three continuous image frame slices may be reconstructed for each cardiac phase during each heartbeat and displayed as a three-dimensional image. In addition, a maximum intensity pixel projection (MIP) image may be generated from the three-dimensional image.

In the reconstruction method described above, the Fourier transform along the axial phase-coded gradient axis was first performed. This is preferred when the A, B, and C phase-coded projection views in the reconstructed image frame are not interleaved with each other. If they are interleaved, alternative methods are preferred. An alternative method is to use the highly limited back projection and composite image weighting described above to perform the A, B, and C phase-coded projection views, respectively, before performing the Fourier transform along the gradient axis. Are back-projected separately. The resulting data set is then gridded to align the samples along the axial phase coding axis. A Fourier transform along the axial phase-coded gradient axis is then performed on the resulting gridded hybrid dataset.

Instead of transforming the image reconstruction process into a process of three 2D slices, a three-dimensional reconstruction of the A, B, and C phase-encoded datasets at each cardiac phase may be performed directly. Many ways to do this are described in co-pending U.S. Patent Application No. 11 / 482,372, filed July 7, 2006, entitled "HIGHLY CONSTRAINED IMAGE RECONSTRUCTION METHOD". Explained in the book. These methods are incorporated herein by reference.

<figref num="1">It is a block diagram of the MRI system using this invention.</figref><figref num="2">It is a figure of k-space sampling using the Fourier transform technique.</figref><figref num="3">It is a figure of k-space sampling using the projection reconstruction technique.</figref><figref num="4">It is a figure of the conventional back projection reconstruction method.</figref><figref num="5">It is a figure of the back projection method which reconstructs a 2D PR image according to this invention.</figref><figref num="6">It is a figure of the back projection method which reconstructs a 3D PR image.</figref><figref num="7">FIG. 5 is a graph of a hybrid PR pulse sequence performed by the MRI system of FIG. 1 when performing a preferred embodiment of the present invention.</figref><figref num="8">It is a figure of k-space sampling using the hybrid pulse sequence of FIG.</figref><figref num="9">It is a flowchart of a step in a preferable embodiment of this invention.</figref><figref num="10">It is a figure of the heart gate acquisition of the data in one heartbeat.</figref><figref num="11">It is a figure of interleaved sampling of k-space by a radial sampling locus.</figref><figref num="12">It is a flowchart of the step of reconstructing a 2D image frame according to the present invention.</figref><figref num="13">FIG. 5 is a diagram of an image generated using the method of FIG.</figref><figref num="14">It is a flowchart of the step in the 2nd contrast amplification embodiment of this invention.</figref><figref num="15">FIG. 3 is a diagram of an image generated using the method of FIG.</figref><figref num="16">It is a flowchart of a step in another preferable embodiment of this invention.</figref><figref num="17">It is a figure of the heart gate acquisition of the data in one heartbeat when carrying out the method of FIG.</figref>