JP6873134B2 - Removal of image artifacts in SENSE imaging - Google Patents

Description

The present invention relates to magnetic resonance imaging, in particular to the removal of artifacts in the SENSE magnetic resonance imaging protocol.

As part of the procedure for generating images of the patient's body, a large static magnetic field is used by a magnetic resonance imaging (MRI) scanner to align the nuclear spins of atoms. This large static magnetic field is called the B0 field.

During an MRI scan, the radio frequency (RF) pulse generated by the transmitter antenna or antenna element causes a perturbation in the local magnetic field, and the RF signal emitted by the nuclear spin is the receiver antenna or antenna. Detected by an array of elements. These RF signals are used to construct an MRI image. These antennas or antenna elements are sometimes called coils. The term coil is often used synonymously to refer to both an antenna and an antenna element. In addition, the transmitter and receiver antennas can be integrated into a single transmitter / receiver antenna that performs both functions. It is understood that the use of the term transmitter / receiver antenna also refers to systems in which separate transmitter and receiver antennas are used. The transmitted RF field is called a B1 field. During long scans, the subject may have internal or external movement, which can corrupt the data and result in blurred or artifacted images.

SENSE is a parallel imaging technique. In parallel imaging technology, multiple antenna elements are used to acquire data at the same time. The coil sensitivity map (CSM) includes the spatial sensitivity of all antenna elements. In this case, "coil" refers to the antenna element. The coil sensitivity map is used to combine the data acquired using each of the individual antenna elements into a single composite image. SENSE significantly speeds up the acquisition of magnetic resonance imaging. The magnetic resonance parallel imaging reconstruction technique is briefly outlined in Section 13.3 of "the handbook of MRI Pulse Sequences" by Bernstein et al. (Published by Elsevier Academic Press, 2004) (hereafter Bernstein et al.).

Publication article by Winkelmann et al., "Ghost artifact removal using a parallell imaging approach", Magn. Reson. Med. 2005 Oct; 54 (4): 1002-9 (hereafter Winkelmann et al.) Describes a ghost artifact removal algorithm using parallel imaging. Use the extended SENSE formulation to remove ghosting artifacts. Enhanced SENSE reconstruction is determined by computationally trying different sources for ghosting artifacts and ranking them using a consistency scale. The disadvantage of this method is that it requires a large amount of calculation.

The present invention provides, in an independent claim, a magnetic resonance imaging system, a computer program product, and a method of operating a magnetic resonance imaging system. An embodiment is given in the dependent claim.

As will be appreciated by those skilled in the art, aspects of the invention may be embodied as devices, methods or computer program products. Accordingly, aspects of the present invention are all generally hardware embodiments, entirely software embodiments (including firmware, resident software, microcode, etc.) or "circuits", "modules" herein. Alternatively, it may take the form of an embodiment that combines software and hardware aspects that may be referred to as a "system." Further, aspects of the invention may take the form of a computer program product embodied in one or more computer-readable media having computer-executable code embodied on a computer-readable medium.

Any combination of one or more computer-readable media may be utilized. The computer-readable medium may be a computer-readable signal medium or a computer-readable storage medium. As used herein, "computer-readable storage medium" includes any tangible storage medium that can store instructions that can be executed by the processor of a computing device. Computer-readable storage media may also be referred to as computer-readable non-temporary storage media. Computer-readable storage media may also be referred to as tangible computer-readable media. In some embodiments, the computer-readable storage medium may also be capable of storing data that can be accessed by the processor of the computing device. Examples of computer-readable storage media include floppy (registered trademark) disks, magnetic hard disk drives, semiconductor hard disks, flash memory, USB thumb drives, random access memory (RAM), read-only memory (ROM), optical disks, magnetic optical disks, and Includes, but is not limited to, processor register files. Examples of optical discs include compact discs (CDs) such as CD-ROMs, CD-RWs, CD-Rs, DVD-ROMs, DVD-RWs, or DVD-R discs and digital versatile discs (DVDs). The term computer-readable storage medium also refers to various types of recording media that can be accessed by computer devices via networks or communication links. For example, the data may be read by a modem, by the Internet, or by a local area network. The computer-executable code embodied on a computer-readable medium is, but is not limited to, any suitable medium, including, but not limited to, wireless, wired, fiber optic cable, RF, etc., or any suitable combination of the above. It may be transmitted using.

The computer-readable signal medium may include a propagated data signal with computer-executable code embodied internally, eg, in baseband or as part of a carrier wave. Such propagated signals may take any of various forms, including but not limited to electromagnetic, optical, or any suitable combination thereof. The computer-readable signal medium may be any computer-readable medium that is not a computer-readable storage medium and can communicate, propagate, or transport a program for use by or in connection with an instruction execution system, device, or device.

“Computer memory” or “memory” is an example of a computer-readable storage medium. Computer memory is any memory that has direct access to the processor. "Computer storage" or "storage" is a further example of a computer-readable storage medium. Computer storage is any non-volatile computer readable storage medium. In some embodiments, the computer storage may be computer memory or vice versa.

As used herein, "processor" includes an electronic component capable of executing a program, machine executable instruction, or computer executable code. References to computing devices, including "processors," should optionally be construed as including more than one processor or processing core. The processor is, for example, a multi-core processor. A processor also refers to a collection of processors that are distributed within a single computer system or into multiple computer systems. It should be understood that the term computer device may refer to a collection or network of computer devices, each having one or more processors. Computer-executable code is executed by multiple processors within the same computer device or distributed among multiple computer devices.

The computer executable code may include machine executable instructions or programs that cause the processor to perform aspects of the invention. Computer executable code for performing operations according to aspects of the invention is an object-oriented programming language such as Java®, Smalltalk, or C ++ and conventional procedural programming such as a "C" programming language or similar programming language. It may be written in any combination of one or more programming languages, including languages, and compiled into machine executable instructions. In some cases, computer-executable code may be in high-level language or ahead-of-time form and may be used with an interpreter that flexibly generates machine-executable instructions.

The computer executable code is entirely on the user's computer, partially on the user's computer, as a stand-alone software package, partially on the user's computer and partially on the remote computer, or completely remote. It can be run on a computer or server. In the latter case, the remote computer may be connected to the user's computer through any type of network, including a local area network (LAN) or wide area network (WAN), or this connection to an external computer (eg, for example). It may be done (through the internet using an internet service provider).

Aspects of the present invention will be described with reference to flowcharts, diagrams and / or block diagrams of methods, devices (systems) and computer program products according to embodiments of the present invention. It will be appreciated that each block or portion of a block of a flowchart, diagram, and / or block diagram, where applicable, can be implemented by computer program instructions in the form of computer executable code. It is further understood that combinations of blocks in different flowcharts, diagrams, and / or block diagrams may be combined as long as they are not mutually exclusive. These computer program instructions are for instructions executed through the processor of a computer or other programmable data processor to perform a function / action specified in one or more blocks of a flowchart and / or block diagram. It may be provided to a general purpose computer, a special purpose computer, or the processor of another programmable data processing device to make the machine to give rise to the means.

These computer program instructions are also such that the instructions stored on a computer readable medium make a product containing instructions that perform a specified function / action in one or more blocks of a flowchart and / or block diagram. , Other programmable data processing equipment, or may be stored on a computer-readable medium that can instruct other devices to function in a particular way.

Computer program instructions also provide a process for instructions executed on a computer or other programmable device to perform a specified function / action in one or more blocks of a flowchart and / or block diagram. , A computer, other programmable data processor, or other device to give rise to a computer-implemented process by allowing a series of operating steps to take place on a computer, other programmable device, or other device. May be loaded on.

As used herein, a "user interface" is an interface that allows a user or operator to interact with a computer or computer system. The "user interface" is sometimes referred to as a "human interface device." The user interface can provide information or data to the operator and / or receive information or data from the operator. The user interface may allow input from the operator to be received by the computer and may provide output from the computer to the user. That is, the user interface may allow the operator to control or operate the computer, and the interface may allow the computer to indicate the result of the operator's control or operation. Displaying data or information on a display or graphical user interface is an example of providing information to an operator. Data via keyboard, mouse, trackball, touchpad, indicator stick, graphic tablet, joystick, gamepad, webcom, headset, gearstick, steering wheel, pedal, wired glove, dancepad, remote control, and accelerometer Receiving is an example of all user interface elements that allow the reception of information or data from the operator.

As used herein, "hardware interface" includes an interface that allows a processor in a computer system to interact with and / or control an external computing device and / or device. The hardware interface may allow the processor to send control signals or instructions to external computing devices and / or devices. The hardware interface may also allow the processor to exchange data with external computing devices and / or devices. Examples of hardware interfaces are universal serial bus, IEEE1394 port, parallel port, IEEE1284 port, serial port, RS-232 port, IEEE488 port, Bluetooth® connection, wireless LAN connection, TCP / IP connection, Ethernet (registration). Trademarks) Includes, but is not limited to, connectivity, control voltage interfaces, MIDI interfaces, analog input interfaces, and digital input interfaces.

As used herein, a "display" or "display device" includes an output device or user interface configured to display an image or data. The display may output visual, audio, and / or tactile data. Examples of displays are computer monitors, television screens, touch screens, tactile electronic displays, braille screens, cathode line tubes (CRTs), storage tubes, bistable displays, electronic paper, vector displays, flat panel displays, vacuum fluorescent displays (VFs). , Light emitting diode (LED) displays, electroluminescent displays (ELDs), plasma display panels (PDPs), liquid crystal displays (LCDs), organic light emitting diode displays (OLEDs), projectors, and head mount displays. Not limited.

Magnetic resonance (MR) data is defined herein as the recorded measurement result of a radio frequency signal emitted by an atomic spin by the antenna of a magnetic resonance apparatus during a magnetic resonance imaging scan. Magnetic resonance data is an example of medical image data. Magnetic resonance imaging (MRI) images are defined herein as a restored two-dimensional or three-dimensional visualization of the anatomical data contained within the magnetic resonance imaging data. This visualization can be done using a computer.

In one aspect, the present invention provides a magnetic resonance imaging system comprising a radio frequency system for acquiring magnetic resonance data from an imaging zone. Radio frequency systems include multiple antenna elements. The magnetic resonance imaging system further includes memory containing machine executable and pulse sequence commands. The pulse sequence command causes the processor to acquire magnetic resonance data from multiple antenna elements according to the SENSE protocol. The SENSE or sensitivity coding method is a well-known parallel imaging reconstruction method used in magnetic resonance imaging. The SENSE protocol is outlined, for example, in Chapter 13 Section 3 of the above-mentioned textbook by Bernstein et al., Handbook of MRI Pulse Sequences.

The magnetic resonance imaging system further comprises a processor. By executing machine executable instructions, the processor controls the magnetic resonance imaging system with pulse sequence commands to acquire magnetic resonance data. By executing the machine executable instruction, the processor further reconstructs the preliminary image using the magnetic resonance imaging data. In various examples, the preliminary image can take different forms. For example, the pulse sequence command may include a command that causes the processor to control the magnetic resonance imaging system to acquire a survey image or a scout image before the SENSE image data is acquired. In another example, the preliminary image is a SENSE image reconstructed using the SENSE imaging protocol.

By executing machine executable instructions, the processor further calculates the fit or alignment between the anatomical model and the preliminary image. The anatomical model may include detailed anatomical data that is aligned or fitted to the preliminary image. In other cases, the anatomical model is a set of procedures or actions used to identify a region of the preliminary image. For example, an anatomical model can be used to identify a region at a predetermined distance from the anterior edge of the subject. The identification of this area can be considered as fitting or alignment.

The anatomical model fits or aligns with the preliminary image. Various types of anatomical models include models that identify anatomical landmarks, models that use anatomical maps aggregated from many different subjects, and models that are modified to fit specific preliminary images. A deformable shape model can be mentioned. The anatomical model includes a motion likelihood map. The motion likelihood map has data representing the likelihood that a particular region identified by the anatomical model is moving during the acquisition of magnetic resonance data. For example, areas such as the diaphragm, lungs, or heart are very likely to be moving. In another example, it marks a particular blood vessel that can be a potential source of flow artifacts, for example due to a new influx of blood. These can be given higher or relative values than other areas in the anatomical model. In some examples, the motion likelihood map is a probability distribution, but the motion likelihood map may contain simple weighting that is neither normalized nor scaled.

By executing machine executable instructions, the processor also uses motion likelihood maps and fits or alignments, at least in part, to identify at least one source of image artifacts. Fit or alignment correlates position in the anatomical model with the preliminary image. This makes it possible to apply the motion likelihood map to the preliminary image. The combination of the motion likelihood map and fit or alignment makes it possible to identify areas in the preliminary image that are likely to cause or cause motion artifacts. In some examples, the motion likelihood map is applied directly to the preliminary image to identify at least one source of image artifacts. In another example, the motion likelihood map is used as a weight in the search algorithm used to identify at least one image artifact source.

By executing machine executable instructions, the processor further uses at least one image artifact source, at least in part, to determine the extended SENSE expression. During the reconstruction of the SENSE image, the images acquired from each of the plurality of antenna elements are reconstructed, then combined and developed into a single image. The extended SENSE equations are expressed as simultaneous linear equations, where the sensitivity matrix is ranked up with additional entries to suppress or eliminate ghosting and motion artifacts. For example, one or more additional columns can be added to the sensitivity matrix. Upon execution of machine executable instructions, the processor further constructs a corrected SENSE image using an extended set of SENSE expressions. This embodiment has the benefit of providing a means of effectively reducing the effects of image artifacts within the SENSE reconstructed image. This embodiment has the advantage of reduced motion due to prior knowledge incorporated into the motion likelihood map of the anatomical model.

At least one source of image artifacts is, in some cases, considered to be a physical location, or a location that corresponds to a physical location relative to the reconstruction of the preliminary or later SENSE image. For example, if the preliminary image is a survey or scout image, the location, and in detail the voxels, may differ, but the spatial location identified by the anatomical model is still used in the reconstruction of the SENSE image. be able to.

In another embodiment, by executing a machine executable instruction, the processor uses magnetic resonance data to reconstruct a measurement coil image for each of the plurality of antenna elements. Upon execution of the machine executable instruction, the processor further constructs a preliminary SENSE image by combining measurement coil images for each of the plurality of antenna elements according to the SENSE protocol, using a set of coil sensitivities. This is useful because the preliminary SENSE image can be used to identify at least one source of image artifacts.

In another embodiment, the preliminary SENSE image is made using an overdetermined reconstruction. In this case, the number of plurality of antenna elements is larger than that of the SENSE factor with some additional constraints. For example, if there are 16 antenna elements, the SENSE factor is 3, and there are two locations identified as image artifact sources for each x value, then 16 is greater than 3 + 2, so SENSE reconstruction is excessive. It is determined.

In another embodiment, the preliminary image comprises a preliminary SENSE image. In other words, the preliminary image may be a preliminary SENSE image, or the preliminary SENSE image may be part of the image data used to create the preliminary image.

In another embodiment, by executing a machine executable instruction, the processor further uses a spare SENSE image and coil sensitivity to form a back-projection image for each of the plurality of antenna elements. During SENSE reconstruction, individual images from each of the antenna elements are combined using coil sensitivity. In back projection, a preliminary SENSE image or a reconstructed SENSE image is used with coil sensitivity to calculate an image for each of the plurality of antenna elements. If the acquired magnetic resonance data and the coil sensitivity are perfect, the back projection image and the measurement coil image will be exactly the same. However, this is usually not the case. There is an error in the coil sensitivity determinant, which can cause a difference between the back projection image and the measurement coil image. Due to these errors, the preliminary SENSE image or a part of the preliminary SENSE image may be damaged. Therefore, by comparing the back-projection image for each of the plurality of antenna elements with the measurement coil image thereof, it is possible to obtain a means for evaluating how successful the SENSE reconstruction is.

Upon execution of the machine executable instruction, the processor further compares the back-projected image with the measurement coil image for each of the antenna elements to identify the set of voxels affected for each of the antenna elements. The back-projection image is compared with the measurement coil image on a voxel basis. If the voxels in the back-projected image fluctuate beyond a predetermined amount from the measurement coil image or according to a statistical scale, the voxels can be added to the affected set of voxels. Specific voxels in the affected set of voxels differ between the back projection image and the measurement coil image sufficiently to indicate that there may be an error in that voxel or in the corresponding voxel in the preliminary SENSE image. Voxel.

Identification of at least one image artifact source is done in image space. Identification of at least one image artifact source is made using at least a partially affected set of voxels and, at least in part, a motion likelihood map and fit or alignment. The set of affected voxels is identified, for example, by comparing the voxel values between the back-projection image and the corresponding measurement coil image. If the value of a particular voxel fluctuates beyond a given threshold, that voxel is added to the set of affected voxels. This method has the advantage that the set of affected voxels, motion likelihood map, and fit or alignment are useful in identifying the source of the artifacts caused in the preliminary SENSE image in space. .. For example, the affected set of voxels is a motion-induced ghost image or ghosting. The combination of these three is used to identify at least one source of image artifacts.

In another embodiment, the at least one image artifact source calculates the maximum value of the consistency scale from each of the at least one image artifact source within a predetermined neighborhood before constructing the corrected SENSE image. It is corrected by searching for. The use of the consistency scale is described in detail by Winkelmann et al. The consistency scale corresponds to the consistency test described in FIG. 3 by Winkelmann et al., An example of a consistency scale is Equation 7 of Winkelmann et al. The difficulty in using ghost artifact removal as described by Winkelmann et al. Is that it is computationally extremely demanding. The examples described herein have the advantage that the computational process can be dramatically accelerated using pre-existing identification of areas in the motion likelihood map where there is likely to be motion. Have.

The consistency measure depends on the difference between the set of affected voxels in the preliminary SENSE image and the back-projected trial SENSE image for each of the plurality of antenna elements. The back-projected trial SENSE image is composed of the trial SENSE image. The trial SENSE image is constructed using the trial SENSE equation. An example of the trial SENSE equation is Winkelmann's equation 3. There is an additional column added to the coil sensitivity determinant. However, it is unknown where the cause of the ghost is derived. For this reason, we construct several different trial SENSE equations and try them to find out which one provides the best consistency scale. The present embodiment has the advantage of significantly speeding up the search for enhanced SENSE equations that give the best results, using prior knowledge of where at least one image artifact source is located. Prior knowledge is expressed in the form of an anatomical model.

In another embodiment, the trial SENSE equation that maximizes the consistency scale is an extended SENSE equation.

In another embodiment, by executing a machine executable instruction, the processor further modifies at least one image artifact source by aligning the affected set of voxels with the preliminary image. For example, if a set of affected voxels identifies ghosting artifacts in an image, it is possible to simply align that set of affected voxels directly to the preliminary image. This helps identify the source of at least one image artifact. In some cases, motion likelihood maps can also be used to obtain trial locations for aligning a set of affected voxels with a preliminary image. This helps speed up the computational method even further.

In another embodiment, the extended SENSE equation comprises an extended coil sensitivity determinant. The extended coil sensitivity determinant is chosen to minimize the contribution from at least a portion of at least one image artifact source.

This can be done in several different ways. In one example, only the motion likelihood map is used to identify the regions included in the extended SENSE reconstruction. In one embodiment, the preliminary image is reconstructed and then the anatomical model is aligned with the preliminary image. Therefore, the motion likelihood map is also then aligned with the preliminary image. Regions above a particular value or threshold in the motion likelihood map are then identified as regions or voxels in the preliminary image that are likely to cause motion artifacts. The contribution of these regions to the artifacts is then suppressed by adding the sensitivity of these locations to the extended coil sensitivity determinant.

In another embodiment, the preliminary image comprises a survey scan image. The use of survey scan images may be beneficial in some cases. For example, instead of performing a SENSE reconstruction, a body coil or body antenna is used to obtain a survey scan image. Survey scan images are less likely to have image artifacts in them. This is useful in aligning the anatomical model.

In another embodiment, the at least one image artifact is two-dimensional or three-dimensional. This is useful, for example, because the method described in detail by Winkelmann et al. Is very computationally intensive and can be difficult to correct for the 2D or 3D region causing motion artifacts.

In another aspect, the invention provides a computer program product with machine executable instructions executed by a processor controlling a magnetic resonance imaging system. The magnetic resonance imaging system comprises a radio frequency system for acquiring magnetic resonance data from the imaging zone. Radio frequency systems include multiple antenna elements. By executing machine executable instructions, the processor controls the magnetic resonance imaging system with pulse sequence commands to acquire magnetic resonance data. The pulse sequence command causes the processor to acquire magnetic resonance data from multiple antenna elements according to the SENSE protocol. By executing the machine executable instruction, the processor further reconstructs the preliminary image using the magnetic resonance imaging data. By executing machine executable instructions, the processor further calculates the fit or alignment between the anatomical model and the preliminary image.

The anatomical model includes a motion likelihood map. By executing machine executable instructions, the processor also uses motion likelihood maps and fits or alignments, at least in part, to identify at least one source of image artifacts. By executing machine executable instructions, the processor further uses at least one image artifact source, at least in part, to determine the extended SENSE expression. Upon execution of the machine executable instruction, the processor further constructs a corrected SENSE image using the extended SENSE expression. The advantages of this have been described above.

In another aspect, the invention provides a method of operating a magnetic resonance imaging system. The magnetic resonance imaging system comprises a radio frequency system for acquiring magnetic resonance data from the imaging zone. Radio frequency systems include multiple antenna elements. The method comprises controlling the magnetic resonance imaging system with pulse sequence commands to acquire magnetic resonance data. The pulse sequence command causes the processor to acquire magnetic resonance data from multiple antenna elements according to the SENSE protocol. The method further comprises reconstructing a preliminary image using magnetic resonance imaging data.

The method further comprises the step of calculating the fit or alignment between the anatomical model and the preliminary image. The anatomical model includes a motion likelihood map. The method further comprises the step of identifying at least one image artifact source using a motion likelihood map and fit or alignment, at least in part. The method further comprises the step of determining an extended SENSE equation using at least one image artifact source, at least in part. The method further comprises, at least in part, using the extended SENSE equation to construct a corrected SENSE image.

It should be understood that one or more of the aforementioned embodiments of the present invention may be combined unless they are mutually exclusive.

Hereinafter, preferred embodiments of the present invention will be described by way of example only with reference to the following drawings.

FIG. 1 shows an example of a magnetic resonance imaging system. It is a flowchart explaining the method of operating the magnetic resonance imaging system of FIG. It is a figure of still another example of a magnetic resonance imaging system. It is a flowchart explaining the method of operating the magnetic resonance imaging system of FIG. It is a figure explaining the technique of extended SENSE reconstruction guided by an annotated body model. It is a figure which schematically explains the influence of the movement in a magnetic resonance image. It is still another figure which schematically explains the influence of the movement in a magnetic resonance image. FIG. 5 schematically illustrates the selection of regions that are minimized during SENSE reconstruction extended to reduce motion artifacts. It is a figure which shows the magnetic resonance image modified to simulate the movement in a fat layer. It is a figure which shows the simulation which removes a motion artifact from the image shown in FIG.

Elements with similar reference numbers in the figure are either equivalent elements or perform the same function. The elements discussed earlier are not necessarily considered in later figures if their functions are equivalent.

FIG. 1 shows an example of a magnetic resonance imaging system 100 including a magnet 104. The magnet 104 is a superconducting cylindrical magnet 104 having a bore 106 penetrating it. Different types of magnets can be used, for example both split cylindrical magnets and so-called open magnets. Split cylindrical magnets are similar to standard cylindrical magnets, except that the cryostat is split into two parts to allow access to the equi-plane of the magnet, such magnets, for example charged particle beam therapy. Used with. The open magnet has two magnet parts, one of which is above the other with a space large enough to receive the subject, and the arrangement of the two parts of the region is a Helmholtz coil. Similar to that of. Open magnets are widespread because the degree to which the subject is trapped is low. Inside the cryostat of a cylindrical magnet, there is a group of superconducting coils. Inside the bore 106 of the cylindrical magnet 104, there is an imaging zone 108 in which the magnetic field is strong and uniform enough to perform magnetic resonance imaging.

Inside the bore 106 of the magnet is also a set of magnetic gradient coils 110 that are used to acquire magnetic resonance data and spatially encode magnetic spins within the imaging zone 108 of the magnet 104. The magnetic field gradient coil 110 is connected to the magnetic field gradient coil power supply 112. The magnetic field gradient coil 110 is intended to be representative. In general, the magnetic field gradient coil 110 includes three separate sets of coils for spatially encoding in three orthogonal spatial directions. The magnetic field gradient power supply supplies current to the magnetic field gradient coil. The current supplied to the magnetic field gradient coil 110 is controlled as a function of time and is gradiented or pulsed out.

Inside the bore 106 of the magnet 104 is an optional body coil 114. The body coil 114 is sometimes called a body antenna. The body coil 114 is connected to the transmitter / receiver 116 in the figure. In some embodiments, the body coil 114 is also connected to a whole body coil radio frequency amplifier and / or receiver, which is not shown in this example. When both the transmitter and the receiver are connected to the whole body coil 114, a means for switching between the transmission mode and the reception mode may be provided. For example, a circuit with a pin diode can be used to select the transmit mode or the receive mode. The subject support 120 supports the subject 118 within the imaging zone.

In the figure, the transmitter / receiver 122 is connected to the magnetic resonance imaging antenna 124. In this example, the magnetic resonance imaging coil 124 is a surface coil comprising a plurality of antenna elements 126, 126 ″, 126 ″, 126 ″ ″. The transmitter / receiver 122 can operate to transmit and receive individual RF signals to the individual coil elements 126, 126 ″, 126 ″, 126 ″ ″. In this example, the transmitter / receiver 116 and the transmitter / receiver 122 are shown as separate devices. However, in another example, device 116 and device 122 are combined.

The magnetic resonance imaging system includes a computer system 130. The transmitter / receiver 116, the transceiver 122, and the magnetic field gradient coil power supply are connected to the hardware interface 132 of the computer 130 in the figure. The computer 130 further includes a processor 134 that can operate to execute machine-readable instructions in the figure. Computer 130 further includes a user interface 136, computer storage 138, and computer memory 140 in the figure, all of which are accessible from processor 134 and are connected to processor 134.

Computer memory 138 includes the pulse sequence command 150 in the figure. The pulse sequence command includes instructions that allow the processor 134 to control the magnetic resonance imaging system 100 to acquire magnetic resonance data 152 according to the SENSE protocol. The pulse sequence command may include a command that allows the processor 134 to acquire magnetic resonance data according to two or more imaging protocols. The pulse sequence command 150 allows the processor 134 to retrieve data according to the SENSE protocol, but allows the processor 134 to use other protocols, such as capturing a survey scan or a scout scan before performing the SENSE protocol. May be good. Computer storage 138 includes, in the figure, magnetic resonance data 152 acquired using the pulse sequence command 150 to control acquisition.

Computer storage 138 further includes a preliminary image 154 reconstructed from magnetic resonance data 152 in the figure. Computer storage 138 further includes an anatomical model 156 that can be fitted or aligned with the preliminary image 154 in the figure. Computer storage 138 further includes a motion likelihood map 158 that is aligned with or is part of the anatomical model 156 in the figure. Computer storage 138 further includes a fit or alignment 159 calculated between the preliminary image 154 and the anatomical model 156 in the figure. Computer storage 138 further includes location 160 of the image artifact source identified in the preliminary image 154 in the figure. The location 160 of the image artifact source is relative to one location or set of locations in the preliminary image 154. If the magnetic resonance imaging system 100 is calibrated correctly, the location 160 of the image artifact source can also be stored in the form of coordinates with respect to the subject 118.

Computer storage 138 further includes an extended SENSE equation 162 in the figure. The extended SENSE equation includes in its sensitivity matrix one or more additional columns used to minimize the sensitivity of the reconstruction result to motion at the location of the image artifact source. Computer storage 138 further includes a corrected SENSE image 164 in the figure. The computer memory 140 has machine executable instructions 170 in the figure. Machine executable instructions 170 include instructions that allow the processor 134 to control the operation of the magnetic resonance imaging system 100, as well as reconstruct the image and modify the magnetic resonance data and the magnetic resonance image. The machine executable instruction 170 allows the processor 134 to perform the computer implementation method described in any one of the following method flow charts.

FIG. 2 shows a flowchart illustrating a method of operating the magnetic resonance imaging system 100 of FIG. The following method steps can be performed by the machine executable instruction 170 shown in FIG.

First, in step 200, the magnetic resonance imaging system is controlled by the pulse sequence command 150 to acquire magnetic resonance data 152. Next, in step 202, the preliminary image 154 is reconstructed from the magnetic resonance data 152. Then in step 204, the fit or alignment between the anatomical model 156 and the preliminary image 154 is calculated. The anatomical model 156 includes a motion likelihood map 158. At step 206, at least one image artifact source 160 is identified using the motion likelihood map 158 and the fit or alignment 159, at least in part. Then, in step 208, the extended equation 162 is constructed using knowledge about at least one image artifact source 160, at least in part. Finally, in step 210, the SENSE image 164 corrected according to the extended SENSE equation 162 is configured.

In one variant, the extended SENSE equation is constructed by adding one or more additional columns to the coil sensitivity determinant to form an extended coil sensitivity determinant. The extended coil sensitivity determinant is chosen to minimize the contribution from at least a portion of at least one image artifact source.

FIG. 3 shows yet another example of the magnetic resonance imaging system 300. The system 300 is similar to the system 100 shown in FIG. The computer storage 138 further includes a measurement coil image 302 reconstructed from the magnetic resonance data 152 in the figure. The computer storage 138 further shows a preliminary SENSE image configured using the measurement coil image 302 and the coil sensitivity set 306 stored in the computer memory 138 in the figure. The coil sensitivity set 306 is known in advance or is measured during the SENSE calibration step. For example, a body coil can be used to obtain a survey image, which can be used to calibrate each of the antenna elements 126, 126 ″, 126 ″, 126 ″ ″.

Computer storage 138 further includes a preliminary SENSE image 304 and a back-projection image 308 calculated from the coil sensitivity set 306 in the figure. The back-projection image 308 is used to create an image showing what the measurement coil image 302 would look like if the set of coil sensitivities 306 were fully known and there were no image artifacts. .. However, the back projection image 308 is different from the measurement coil image 302. By comparing the measurement coil image with the back-projection image for each of the coil elements 126, 126 ″, 126 ″, 126 ″ ″, the affected voxel set 310 can be identified for each coil. These are used to identify image artifacts in the measurement coil image. These may lead to the identification of image artifacts in the preliminary SENSE image 304. Computer memory 140 again indicates machine executable instructions 170. The machine executable instruction causes, for example, the processor 134 to perform the computer implementation method described in FIG. 2 or also described in FIG. 4 below.

FIG. 4 shows a flow chart illustrating a method similar to the method described in FIG. In FIG. 4, some additional steps are added to the method. FIG. 4 begins at steps 200 and 202 as described in FIG. Next, in step 400, the magnetic resonance data 152 is used to reconstruct the measurement coil image for each of the plurality of antenna elements 126, 126 ″, 126 ″, 126 ″ ″. Then in step 402, using the coil sensitivity set 306, by combining the measurement coil images 302 of each of the plurality of antenna elements 126, 126', 126'', 126''' according to the SENSE magnetic resonance imaging protocol. , Preliminary SENSE image 304 is configured. In some examples, the preliminary SENSE image 304 is the preliminary image 154. In this case, step 204 is exactly the same as steps 400 and 402, so step 204 is performed here.

However, in other examples, the preliminary image is different or distinct from the preliminary SENSE image 304. For example, the pulse sequence command 150 causes the magnetic resonance imaging system 100 to acquire a survey scan or a scout scan, which is then fitted to the anatomical model. In yet another example, both a survey scan or scout scan image and a preliminary SENSE image are used to fit the anatomical model while performing this computer practice.

As mentioned above, the method then optionally performs step 204. In the method shown in FIG. 4, steps 404, 406, 408 are more detailed instructions as to how step 206 of FIG. 2 is performed. First, in step 404, a preliminary SENSE image 304 and a set of coil sensitivities 306 are used to construct back-projection images for each of the plurality of antenna elements 126, 126', 126'', 126'''. .. Then, in step 406, the back-projection image 308 is compared with the measurement coil images 302 of each of the plurality of antenna elements 126, 126', 126'', 126''' to identify the affected set 310 of voxels. To do. The affected voxel set 310, motion likelihood map 158, and fit or alignment 159 are then used, at least in part, to identify at least one image artifact source. Then, in step 408, at least one by computationally searching for the maximum value of the consistency scale within a predetermined neighborhood from each of the at least one image artifact source before constructing the corrected SENSE image 164. Correct the image artifact source 160.

For example, the computational method described by Winkelmann et al. Can be applied, in which case the search area is limited to a predetermined neighborhood of a particular voxel. In this case, the algorithm of Winkelmann et al. Is applied, but the search area is significantly reduced by using the processing in the image space. This can dramatically increase the computational efficiency in finding at least one source of image artifacts. Steps 404, 406, and 408 may be computationally repeated during the step of searching for the maximum value of the consistency scale. Steps 208 and 210 are then performed as described in FIG.

SENSE is one of the technologies for accelerating the acquisition of magnetic resonance (MR) data. To facilitate deployment, coil sensitivity information and measured data must be available as essential inputs to the SENSE algorithm. Even with perfect coil sensitivity, SENSE image artifacts can still appear due to errors in the measured data. The most prominent of this type of artifact caused by the distortion of the Fourier coding process is ghosting, which alternates the appearance of parts of the actual signal. However, although the Fourier coding process is compromised, the actual signal is correctly sensitive coded, which gives an opportunity to identify and eliminate ghosting structures. See the paper by Winkelmann et al.

The key to this idea is to identify dangerous signals by consistency testing and solve the extended SENSE problem that includes the replaced signal components (ghosts), and for the replaced signal components, the phase. The correct source must be found in the search procedure along the coding direction. The computational load of this search can be feasible with 1D undersampling, but it presents problems in two dimensions and can lead to erroneous minimums in some bad cases. In some examples, the idea is to incorporate a suitable, patient-fitting body model into an extended SENSE reconstruction, speeding up and stabilizing the search for ghost sources by incorporating specific prior knowledge. That is. Based on the 3D Scout MRI data (or other images) obtained at the start of the examination, a suitable body model (or anatomical model) can be tailored to the patient. This model is for annotated organs (ie, liver, lungs, etc.) and blood vessels and fluids that are likely to be sources of MR ghosts (eg, due to influx, flow, movement effects, etc.). It can include structures such as a filled chamber (heart, etc.), chest wall, etc. Fitted to the superimposed SENSE / consistency data, this model can guide the search procedure for ghost sources, reducing computational effort and increasing reliability.

SENSE is a method of parallel imaging. Even if the coil sensitivity information is perfect, the quality of the SENSE image still depends on the measured data. The measured data may contain some data inconsistencies caused by movements such as volumetric movements, flow, and new magnetization inflows, which actually impair the Fourier coding process. there is a possibility. Ghosting, apart from blurring, is the most prominent image artifact, which cannot be eliminated by standard SENSE reconstruction.

However, ghosts have been found to be correctly sensitively coded, albeit shifted in the phase coding direction in the final SENSE image (due to the impaired Fourier coding process). However, the actual location, that is, the location where it occurred, is unknown (1). Therefore, voxels damaged by the ghost generation signal are included in the reconstructed SENSE image by analyzing the consistency between the reconstructed SENSE image and the underlying, superposed data. You can find it at (1). Consistency tests show ghosts because they were exposed to different receive coil sensitivities during signal detection compared to those assumed during SENSE reconstruction.

In order to eliminate the ghosts corresponding to these damaged voxels, we have to formulate an extended SENSE problem that seeks additional signal contributions, that is, ghosts that originate elsewhere and overlap the damaged voxels. It doesn't become. The resulting signal model differs from the normal SENSE model only _{by the additional terms Si, g} δ in Equation [1] by Winkelmann et al.
C _i = ΣS _{i, j} ρ _j + S _{i, g} δ [1]

Here, the vector C includes the superimposed coil signal measured with respect to the coil i, that is, the measurement data. S represents the sensitivity matrix, ρ is the vector containing the actual voxel signal to be acquired, δ is the contribution of the ghost signal originating from the unknown location g, and the sum spans all coils. This extended SENSE problem can only be established if the SENSE problem is overdetermined. When there is one phase coding direction (2D imaging), the rank of the extended SENSE matrix is increased by 1 (only one ghost is expected), and when there are two directions like 3D imaging, There may be two sources of ghosting, which further increase the rank of the extended SENSE matrix, resulting in a reduced number of pseudo-inverse element conditions and making the solution more susceptible to noise. The extended SENSE problem is one of the key elements of optimization to get the best data integrity (fixed as a penalty term) with respect to the potential location g of the ghost. While this search can be feasible with 1D undersampling (see Figure 3a by Winkelmann et al.), It presents problems in two dimensions and can lead to false minimums in some bad cases. ..

In FIG. 3a by Winkelmann et al., The one-dimensional search along the phase coding direction for ghosting artifacts is 2D imaging. The integrity logarithm (P) of the extended SENSE reconstruction is plotted as a function of the potential source of ghost δ along the phase coding direction. Two different ghosting voxels in the final SENSE image are shown (I, II). At the appearance of ghosts (LAA-location of artifacts in the final image), the problem becomes peculiar, and at ghost sources (LAO-location of artifact sources), consistency is maximal. For ghosts (I), the major maximums are generally well defined, and for ghosts (II) the problem can be more difficult.

Furthermore, the inverse condition of equation [1] becomes defective when the source g of the artifact is close to the location I of the voxel under consideration, making it more difficult to resolve the extended SENSE.

It is proposed to incorporate prior knowledge in order to speed up and stabilize the process of solving the extended SENSE problem equation [1]. This is achieved by using a suitable, patient-fitting body model during the extended SENSE reconstruction. This model shows potential areas at risk of causing ghosting artifacts, guides exploration, and helps eliminate potential false positives.

Based on the 3D Scout MRI data obtained at the beginning of the examination, the body model can be specifically tailored to the patient and adequately reflect the key features of the patient's anatomy and shape dimensions. .. The model was filled with annotated organs (liver, lungs, etc.) and structures (blood vessels, fluids, etc.) that are likely to be sources of MR ghosts (eg, due to influx, flow, movement effects, etc.). Includes chamber (heart, etc.), chest wall, etc.).

The model can be adapted to the expanded SENSE / consistency data to guide the search procedure for ghost sources, reducing computational effort and increasing reliability.

In addition, the fitted body model can help guide the SENSE signal model in other directions as well. Based on the fitted body model, areas in the coil sensitivity map that are at risk of being incorrect can be identified. These areas can be added as columns to the sensitivity matrix to further extend the sensitivity matrix.

In one example, a liver test is performed on the subject or patient. The liver is surrounded by the corresponding multi-element receiving coil and is located in the isocenter of the magnet. For planning purposes, multi-slice or 3D low resolution scout scans or other scans are measured. The data is further sent to an image processing algorithm, which uses elastic alignment to fit the body model of the corresponding predefined body region to the data. Rough body area identification is either automatic or driven in context. After the alignment process, the patient's shape dimensions are matched with information about the risk of motion artifacts and their nature. This information will be available for each voxel inside the patient and all subsequent algorithms will be available.

Coil sensitivity information is obtained using a SENSE reference scan. A receiver array is used to perform a 3D diagnostic liver scan performed by holding breath (15-20 seconds), which covers the entire abdominal region. Due to the movement of the heart, fresh blood is pumped into the 3D volume and scanned, resulting in a natural influx effect on the intensity modulation of the MR signal, resulting in a Fourier ghost in the subsampled SENSE imaging. Causes developmental artifacts. This artifact also propagates throughout the FOV during SENSE reconstruction.

This artifact is identified by the consistency test proposed by Winkelmann et al. This test measures how the measured, sub-sampled reduced FOV image fits into the image finally reconstructed with SENSE.

To facilitate this comparison, the final SENSE image is back-projected into individual reduced FOV coil images for each channel. Each individual's difference to the measured data must match the noise level of the receiver and is evaluated using the appropriate probability (see Winkelmann et al.). Mismatches can identify damaged voxels that require another solution to the SENSE problem using the extended SENSE scheme (see Winkelmann et al.). Here, a geometrically matched model is involved (shown briefly in Figure 5 below), which is a search along the corresponding phase coding direction performed to find the source of the ghost. Support / guide. Therefore, the extended SENSE scheme is enhanced by an annotated body model that tells the algorithm which pixels in each full field (FOV) or subsampled FOV are potential sources of artifacts. Will be done. Such pixels and their adjacent pixels are tested for achievable integrity, assuming they are the cause. In addition, the information obtained from the model can be used to additionally regularize other potential solutions (locations) to avoid erroneous solutions.

This method can be used to speed up the algorithm and avoid noise breakthroughs or false positives.

FIG. 5 shows an idealized representation of the preliminary SENSE image 304, an anatomical model 156, and a corrected SENSE image 164. Preliminary SENSE image 304 shows a cross section of the subject, with some ghost artifacts 500 of the aorta 502. The movement of blood in the aorta 502 produces the ghost artifact 500. The anatomical model 156 may be assigned a motion likelihood map 158. In this map, the aorta 502 is identified as a region with a high likelihood of causing ghosting artifacts. The model 156 is then fitted to the preliminary SENSE image 304 and the corrected SENSE image 164 is calculated by applying the methods described in FIGS. 3 and 4. FIG. 5 illustrates an enhanced SENSE reconstruction technique guided by an annotated body model. The SENSE-reconstructed MR image shows the ghosting artifacts identified by the integrity test (phase coding direction, ie, replication of the vertically beating aorta). A geometrically matched, fitted body model (middle) shows a potential source of this ghosting artifact (the aorta, or high-risk area, highlighted in red). This information is used to guide the search for an enhanced SENSE reconstruction and remove the generally shown artifacts (right).

Such models also have many other purposes, such as, for example.
-Organ-specific planning-Automatic navigator positioning-Automatic shim volume positioning-Can help with automatic external volume suppression (REST slab) positioning.

In MRI, areas of non-clinical interest often have scattered artifacts in areas of clinical interest. One typical example is the aorta, which disperses fluid artifacts in the liver, and another example is the breast or heart, which causes movement to produce artifacts on the spine. See FIG. 6 below as an example. The figure schematically illustrates the sagittal plane through the human body, including the spine, which is of clinical interest, and the heart, which is not of interest in the study.

FIG. 6 shows a sagittal plane view 600 showing the heart 602. The heart produces a ghosting image of image artifact 604 or heart 602 as it moves.

In some cases, this problem can be overcome by positioning the so-called REST (short for "Regional Saturation Technology") on the problematic (usually moving) body area. However, this is often constrained, for example, areas other than straight slabs are usually impractical. There can also be significant drawbacks in terms of scan time, achievable number of iterations, and so on.

A large number of receivers can be used to solve this problem in a completely different way. The coil combination algorithm can be made to minimize sensitivity to areas known to cause problems when the results are combined.

This can be done by the user pointing to an area known to cause motion artifacts. This has some similarities to the placement of REST slabs, also planned by the user. The difference is that the indication of the presence of movement for the optimal combination of synergies can be given before or after the actual measurement (assuming that the raw data of the measurement is still held in memory). For this reason, this technique is called "post-scan REST".

However, the need for user input hinders the workflow. Therefore, an object of the present invention is to apply the same principle without an interface.

Especially in abdominal imaging, it is assumed that motion artifacts are mainly caused by the patient's anterior subcutaneous fat. The idea is to detect the region and create a coil combination algorithm that minimizes the sensitivity to that region in the reconstruction process.

FIG. 7 shows a diagram showing an abdominal cross-sectional MRI. The region of reference numeral 702 represents fat. The region of reference numeral 704 represents an organ of interest such as the liver. The various lines 706 represent image or ghosting artifacts caused by the movement of the fat layer 702.

In some examples, this method applies to sagittal or axial abdominal imaging performed with multiple receiving antennas or antenna elements. In principle, the present invention is equivalent to any type of acquisition sequence and does not require modification thereof. The present invention comprises modified reconstructed "regular" data.

One element of some examples is the detection of the anterior margin of the anterior subcutaneous fat region. This can be done by analyzing the first reconstruction of the current scan, the image of the same area obtained in the previous scan, the coil reference scan, and the like.

Optionally, low resolution chemical shift imaging scans (water-fat separation scans, silicon-fat separation scans, etc.) can provide estimates of fat layer thickness. Alternatively, the "typical" fat layer thickness may be pre-programmed. Fully automated image segmentation, such as intensity-based or map-based segmentation, can also help generate initial estimates of objects for which signal strength should be suppressed.

Given an initial estimate of the object, the foremost part of the object is considered the "problematic" fat layer. Combining the leading edge with thickness gives a region (2D region in a slice, or 3D region in a large number of slices).

Instead, the position of the curve (or surface in multislice or 3D) representing the center of the anterior fat region can be determined.

Given that region or central plane, reconstructions can be created that minimize sensitivity to that region. This is done by selecting the most appropriate weighting of the coil elements that will be unresponsive to the area of the anterior fat area for each pixel being reconstructed.

FIG. 8 is similar to that shown in FIG. The schematic view of this MRI also shows an abdominal cross section of 700. In this case, there is a region 800 in which the combination of coil element weights is unresponsive. This eliminates the image artifact 706 or ghosting artifact shown in FIG.

In some examples, the reconstruction algorithm involves _{adding an extra row s a} to the coil sensitivity matrix S (or a large number of rows s _a1 , ... s _ak , ... s _aK . Here. k spans all identified locations that can be sources of motion artifacts). As a result, an "extended coil sensitivity matrix" S _E = [S s _a ] is obtained (or instead, S _E = [S s _a1 ... s _ak ... s _aK ]). Using this matrix, the canonical solution of the SENSE problem can be applied, i.e.

Except for the subscript "E", this is similar to the very well known SENSE equation, where p represents the resulting set of pixels, Ψ is the noise covariance between the acquisition channels, and R Is a regularization matrix and m is the measured coil array data. Nevertheless, here both p and R contain one (or set) extra element. In the present invention, the SENSE equation does not generate an equidistant set of expanded pixels, but rather a "regular" equidistant set of pixels plus pixels at the location of the artifact cause (ie, the anterior fat edge). It produces an estimate of intensity and, within the scope of the present invention, the extra results are not considered of interest.

Similarly, the difference between the "regular" regularization matrix R and " _RE " is that the latter must also indicate the expected signal level at the location of the anterior fat, which is the other diagonal of the matrix R. As well known as the element.

9 and 10 show magnetic resonance images 900 and 1000. FIG. 9 shows a magnetic resonance imaging 900 with simulated motion artifacts added to the image.

FIG. 10 shows magnetic resonance imaging 1000 in which one region of the coil weighting element is unresponsive to remove artificial motion artifacts in image 900.

In the following description, "Possup" is used as an abbreviation for "suppression of post-scan sensitivity to the anterior fat layer". Basically, the idea is to slightly modify the SENSE reconstruction and the resulting combination of antenna elements is relative to the (automatically detected) area of the patient, eg, the anterior subcutaneous fat area in abdominal imaging. The sensitivity should be minimized.

An image of the abdomen that stopped breathing was taken as the starting point. This image showed relatively few "natural" movement artifacts. After that, we devised four coil sensitivity maps that attenuate as 1 / x from the front edge, right edge, trailing edge, and right edge of the image.

Motion was simulated by creating six different "distorted" images. The rear half of the object was left unaffected when distorting each, but the front half so that the top edge of the object is moved forward by 1, 2, 3, 4, 5, or 6 pixels. It was stretched. (The object is about 200 pixels in size, so the front half is about 100 pixels, which means that in the most distorted one, the front half is stretched by about 6%). A distorted image is shown in FIG.

All of these sets were Fourier transformed into k-space, and each ky line in the simulation was randomly selected from one of these six sets.

Reconstructing the above simulation using CLEAR yields the image on the right.

This image was used as the first iteration to detect the leading edge of the object. We then defined two regions parallel to its leading edge. One was at a depth of 5 pixels (front) and one was at a depth of 12 mm. Knowledge of fat layer thickness was entered manually. This results in two curves F1 (x) and F2 (x).

The Possup reconstruction calculates the weighting of the coil element that maximizes the sensitivity to (x, y) and minimizes the sensitivity to the regions F1 (x) and F2 (x) for each location (x, y). After that, the coil element data is combined.

Unexpectedly, the anterior fat region is suppressed at depths between 5 and 12 pixels. Mathematically, this consists of extending the SENSE coil sensitivity matrix with two extra columns. This is relevant here as it allows for easy combination with SENSE. By applying Possup, this artifact is also largely removed, as shown in FIG.

Although the present invention has been illustrated and described in detail in the drawings and the aforementioned description, such illustration and description should be considered to be descriptive or exemplary and not limiting. That is, the present invention is not limited to the disclosed embodiments.

Other variations of the disclosed embodiments may be understood and realized by those skilled in the art who practice the claimed invention from the drawings, the present disclosure and the examination of the accompanying claims. In the claims, the word "comprising" does not exclude other elements or steps, and the indefinite articles "a" or "an" do not exclude more than one. A single processor or other unit performs the function of some of the claimed items. The mere fact that certain means are listed in different dependent claims does not indicate that the combination of these means is not used in an advantageous manner. Computer programs may be stored / distributed on any suitable medium, such as an optical storage medium or solid state medium, supplied with or as part of other hardware, but on the Internet or other wired or wireless. It may be distributed in other forms, such as via the telecommunications system of. Any reference code in the claims should not be construed as limiting the scope of the invention.

100 Magnetic resonance imaging system 104 Magnet 106 Magnet bore 108 Imaging zone 110 Magnetic gradient coil 112 Magnetic gradient coil power supply 114 Body coil or body antenna 116 Transmitter / receiver 118 Subject 120 Subject support 122 Transmitter / receiver 124 Magnetic resonance image antenna 126 Antenna Element 126'Antenna element 126'' Antenna element 126''' Antenna element 130 Computer 132 Hardware interface 134 Processor 136 User interface 138 Computer storage 140 Computer memory 150 Pulse sequence command 152 Magnetic resonance data 154 Preliminary image 156 Anatomical model 158 Motion likelihood map 159 Fit or alignment 160 Location of image artifact source 162 Enhanced SENSE expression 164 Corrected SENSE image 170 Machine-executable instructions 200 Magnetic resonance imaging system controlled by pulse sequence commands to control magnetic resonance data 202 Steps to reconstruct the preliminary image using magnetic resonance imaging data 204 Steps to calculate the fit or alignment between the anatomical model and the preliminary image, where the anatomical model is in motion Step 206, including the likelihood map Step 206 Identify at least one image artifact source using at least a motion likelihood map and fit or alignment Step 208 At least partially at least one image artifact Step 210 to determine the extended SENSE equation using the source Step 210 At least in part, to construct the SENSE image corrected according to the enhanced SENSE reconstruction using the enhanced SENSE equation 300. Magnetic Resonance Imaging System 302 Measurement Coil Image 304 Preliminary SENSE Image 306 Coil Sensitivity Set 308 Back Projection Image 310 Affected Boxel Set 400 Magnetic resonance data is used to reconstruct the measurement coil image for each of multiple antenna elements. Step 402 Preliminary by combining measurement coil images for each of multiple antenna elements according to the SENSE protocol using a set of 402 coil sensitivities. Step 404 to construct a SENSE image Step 406 to construct a back-projection image for each of a plurality of antenna elements using a preliminary SENSE image and coil sensitivity. Step 408 Identifying the Affected Set of Boxels for Each of Multiple Antenna Elements Compared to an Image Before constructing a corrected SENSE image, within a predetermined neighborhood from each of at least one image artifact source. Steps to correct at least one source of image artifacts by computationally searching for the maximum value of the integrity scale 500 Ghost artifacts 502 Aoria 600 Sagittal view 602 Heart 604 Image artifacts 700 Abdominal cross section 702 Fat 704 Of interest Organ 706 Image artifact 800 Region 900 Magnetic resonance image 1000 Magnetic resonance image

Claims (14)

A radio frequency system with multiple antenna elements for acquiring magnetic resonance data from the imaging zone,
With the processor
A memory comprising a machine executable instruction and a pulse sequence command, wherein the pulse sequence command is a magnetic resonance imaging system comprising a memory that causes the processor to acquire magnetic resonance data from the plurality of antenna elements according to the SENSE protocol. There,
By executing the machine executable instruction, the processor
To acquire the magnetic resonance data by controlling the magnetic resonance imaging system with the pulse sequence command, and
Reconstructing the preliminary image using the magnetic resonance data and
To calculate the fit between the anatomical model and the preliminary image, that the anatomical model includes a motion likelihood map.
Using the motion likelihood map and the fit, at least in part, to identify at least one source of image artifacts.
At least in part, using the at least one image artifact source to determine the extended SENSE equation,
A magnetic resonance imaging system that constructs and performs a corrected SENSE image using the extended SENSE equation. By executing the machine executable instruction, the processor
Using the magnetic resonance data to reconstruct the measurement coil image for each of the plurality of antenna elements,
The magnetic resonance according to claim 1, wherein a preliminary SENSE image is constructed by combining the measurement coil images of each of the plurality of antenna elements according to the SENSE protocol using a set of coil sensitivities. Imaging system. The magnetic resonance imaging system according to claim 2, wherein the preliminary image includes the preliminary SENSE image. Execution of the machine executable instruction further causes the processor to
Using the preliminary SENSE image and the coil sensitivity to construct a back projection image for each of the plurality of antenna elements.
The back-projection image is compared with the measurement coil image for each of the plurality of antenna elements to identify the set of voxels affected for each of the plurality of antenna elements.
The identification of the at least one image artifact source is made in image space, and the identification of the at least one image artifact source uses, at least in part, the set of affected voxels and at least partially. The magnetic resonance imaging system according to claim 2 or 3, wherein the motion likelihood map and the fit are used in the above. The at least one image artifact source computationally searches for the maximum value of the consistency scale within a predetermined neighborhood from each of the at least one image artifact source before constructing the corrected SENSE image. Corrected by this, the consistency measure depends on the difference between the set of affected voxels in the preliminary SENSE image and the back-projected trial SENSE image for each of the plurality of antenna elements. The magnetic resonance imaging system according to claim 4, wherein the back-projected trial SENSE image is composed of a trial SENSE image, and the trial SENSE image is constructed using a trial SENSE formula. The magnetic resonance imaging system according to claim 5, wherein the trial SENSE equation that maximizes the consistency scale is the extended SENSE equation. The execution of the machine executable instruction further modifies the at least one image artifact source by aligning the affected set of voxels with the preliminary image, claims 4, 5, and 5. Or the magnetic resonance imaging system according to 6. The magnetic resonance imaging system according to any one of claims 1 to 3, wherein the extended SENSE equation is selected to minimize the contribution from at least a portion of the at least one image artifact source. The magnetic resonance imaging system according to any one of claims 1 to 8, wherein the preliminary image includes a survey scan image. The magnetic resonance imaging system according to any one of claims 1 to 9, wherein the at least one image artifact source is two-dimensional or three-dimensional. A computer program with machine-executable instructions executed by a processor that controls a magnetic resonance imaging system, said magnetic resonance imaging system comprising a radio frequency system for acquiring magnetic resonance data from an imaging zone. The radio frequency system comprises a plurality of antenna elements, and by executing the machine executable instruction, the processor can be subjected to.
Controlling the magnetic resonance imaging system with a pulse sequence command to acquire the magnetic resonance data, the pulse sequence command causes the processor to acquire magnetic resonance data from the plurality of antenna elements according to the SENSE protocol. To let and
Reconstructing the preliminary image using the magnetic resonance data and
To calculate the fit between the anatomical model and the preliminary image, the anatomical model includes a motion likelihood map.
Using the motion likelihood map and the fit, at least in part, to identify at least one source of image artifacts.
Using at least one image artifact source, at least in part, to determine the extended SENSE equation,
A computer program that uses the extended SENSE expression to construct a corrected SENSE image. A method of operating a magnetic resonance imaging system, wherein the magnetic resonance imaging system includes a radio frequency system for acquiring magnetic resonance data from an imaging zone, the radio frequency system includes a plurality of antenna elements, and the method. Is
The step of controlling the magnetic resonance imaging system with a pulse sequence command to acquire the magnetic resonance data, wherein the pulse sequence command causes a processor to acquire magnetic resonance data from the plurality of antenna elements according to the SENSE protocol. , Steps and
A step of reconstructing a preliminary image using the magnetic resonance data,
A step of calculating the fit between the anatomical model and the preliminary image, wherein the anatomical model includes a motion likelihood map.
A step of identifying at least one image artifact source using the motion likelihood map and the fit, at least in part.
The step of determining the extended SENSE equation using at least one image artifact source, at least in part,
A method comprising the steps of constructing a corrected SENSE image using the extended SENSE equation. A step of reconstructing a measurement coil image for each of the plurality of antenna elements using the magnetic resonance data.
A step of constructing a preliminary SENSE image by combining the measurement coil images for each of the plurality of antenna elements according to the SENSE protocol using a set of coil sensitivities.
A step of constructing a back projection image for each of the plurality of antenna elements using the preliminary SENSE image and the coil sensitivity.
It further comprises a step of comparing the back-projected image with the measurement coil image for each of the plurality of antenna elements to identify the set of voxels affected for each of the plurality of antenna elements.
The identification of the at least one image artifact source is made in image space, and the identification of the at least one image artifact source uses, at least in part, the set of affected boxels and at least partially. Is performed using the motion likelihood map and the fit or alignment, and the at least one image artifact source generates the at least one image artifact before constructing the corrected SENSE image. Corrected by computationally searching for the maximum value of the consistency scale from each of the sources within a predetermined neighborhood, the consistency scale is the effect of the preliminary SENSE image on each of the plurality of antenna elements. Depending on the difference between the set of boxels to receive and the back-projected trial SENSE image, the back-projected trial SENSE image is composed of a trial SENSE image, and the trial SENSE image uses a trial SENSE equation. 12. The method of claim 12. The extended SENSE determinant includes an extended coil sensitivity determinant, and the extended coil sensitivity determinant is selected to minimize contributions from at least a portion of the at least one image artifact source. , The method according to claim 12 or 13.

Images