Real-time design of radio frequency and gradient pulses in magnetic resonance imaging

Abstract

Disclosed herein is a medical system (100, 300) comprising a memory (110) storing machine executable instructions (120) and a convolutional neural network (122). The convolutional neural network is configured to receive as input a complex array (128) encoding a selection (124) of at least one excitation field of view (324, 900) and, in response, output a radio frequency waveform (130) and a plurality of spatially selective gradient pulse waveforms (132). The convolutional neural network is a multi-task convolutional neural network. Execution of the machine executable instructions causes a computing system (104) to: receive (200) the selection (124) of the at least one excitation field of view; receive (202) an initial pulse sequence command (126); encode (204) the complex array using the at least one excitation field of view; receive (206) the radio frequency waveform and the plurality of spatially selective gradient pulse waveforms in response to inputting the complex array into the convolutional neural network; and construct (208) a modified pulse sequence command (134) by modifying the initial pulse sequence command with the radio frequency waveform and the plurality of spatially selective gradient pulse waveforms.

Description

Technical Field

[0001] This invention relates to magnetic resonance imaging, and more particularly to the design of pulse sequences for controlling magnetic resonance imaging systems. Background Technology

[0002] As part of the process of generating images of a patient's body, a magnetic resonance imaging (MRI) scanner uses a large static magnetic field to align the nuclear spins of atoms. This large static magnetic field is called the B0 field or main magnetic field. MRI can be used to spatially measure various quantities or properties of an object. Spatial coding in MRI is performed using a combination of radio frequency (RF) waveforms (or RF pulses) used to control the MRI scanner's transmit coils and multiple spatially selective gradient pulse waveforms (gradient pulses).

[0003] Vinding et al. published an article titled "Ultrafast (milliseconds), multidimensional RF pulse design with deep learning" (Magn. Reson. Med. 2019; 82: 586-599) which discloses a neural network that uses an input map that considers the desired region of interest for excitation and outputs a single-channel multidimensional RF pulse.

[0004] The paper “Joint Design of RF and gradient waveforms via auto-differentiation for 30 tailored excitation in MRI” by Tianrui Luo et al. (retrieved from ARXIV.org, Cornell University Library, August 2012, 24, 2020) discloses a method for jointly designing RF and gradient waveforms based on an automatically differentiable Bloch simulator. Summary of the Invention

[0005] The present invention provides a medical system, a computer program, and a method in the independent claims. Embodiments are given in the dependent claims.

[0006] The embodiments can provide improved selective RF excitation. This can be achieved by using a convolutional neural network trained to output both a radio frequency waveform for generating the B1 field and an accompanying spatially selective gradient pulse waveform. Various types of neural network architectures have been investigated, and multi-task neural networks provide the best results. Multi-task neural networks provide multiple outputs in response to receiving one input. In the example described herein, the convolutional neural network employs a complete complex array already encoded with the desired RF excitation field of view and outputs a complete radio frequency waveform and a complete spatially selective gradient pulse waveform. In the case of acquiring a two-dimensional slice, there are two spatially selective gradient pulse waveforms, and if the acquisition is three-dimensional, there are three spatially selective gradient pulse waveforms output.

[0007] In one aspect, the invention provides a medical system including a memory storing machine-executable instructions and a convolutional neural network. The convolutional neural network is configured to receive a complex array as input, encoding a selection of at least one excitation field of view. The complex array may, for example, represent a two-dimensional slice to be acquired by magnetic resonance imaging (MRI) or a three-dimensional volume to be acquired by MRI.

[0008] The convolutional neural network is configured to output a radio frequency (RF) waveform and a plurality of spatially selective gradient pulse waveforms in response to receiving the complex array. The RF waveform may be a waveform used to control the transmitter in a magnetic resonance imaging system to generate a B1 magnetic field. The plurality of spatially selective gradient pulse waveforms are each of a gradient pulse waveform matched with the RF waveform to acquire magnetic resonance data for the at least one excitation field of view.

[0009] The convolutional neural network is a multi-task convolutional neural network, which has a first output for the radio frequency waveform and a separate output for each of the plurality of spatially selective gradient pulse waveforms. For a two-dimensional magnetic resonance imaging protocol, there will be two spatially selective gradient pulse waveforms, and for a three-dimensional magnetic resonance imaging protocol, there will be three spatially selective gradient pulse waveforms.

[0010] The medical system also includes a computing system. The execution of the machine-executable instructions causes the computing system to receive the selection of the at least one excitation field of view. This could, for example, be the field of view of interest to the operator or physician for imaging. The execution of the machine-executable instructions also causes the computing system to receive an initial pulse sequence command configured to control the magnetic resonance imaging system to acquire k-space data describing the object. The execution of the machine-executable instructions further causes the computing system to receive the radio frequency waveform and the plurality of spatially selective gradient pulse waveforms in response to inputting the complex array into the convolutional neural network.

[0011] The execution of the machine-executable instructions also enables the computing system to construct a modified pulse sequence command by modifying the initial pulse sequence command using the radio frequency waveform and the plurality of spatially selective gradient pulse waveforms, such that the pulse sequence command is configured to control the magnetic resonance imaging system to acquire the k-space data from the at least one excitation field of view. This embodiment can be advantageous because it allows the operator of the magnetic resonance imaging system to easily specify the excitation field of view. It also allows the specification of excitation fields of view of arbitrary size or shape. In other cases, it also allows the specification of multiple excitation fields of view for a single acquisition.

[0012] The initial pulse sequence command can be, for example, a preliminary or templated pulse sequence command, which includes RF pulses and gradient pulse regions for which slice selection needs to be specified. Therefore, embodiments can provide customization of the pulse sequence command for a specific acquisition by utilizing RF waveforms and multiple spatially selective gradient pulse waveforms to modify the pulse sequence command.

[0013] This invention relates to the automatic generation of radio frequency (RF) waveforms and multiple (two or more) spatially selective gradient waveforms based on an input excitation field of view, and to the automatic generation of RF waveforms and multiple (two or more) spatially selective gradient waveforms by modifying an initial pulse sequence command, with respect to the RF waveforms and gradient waveforms involved. In response to the input excitation field of view, the automatically generated RF waveforms and gradient waveforms are returned via a multi-task convolutional neural network. This multi-task convolutional neural network has several parallel outputs (for the corresponding waveforms) and is not fully connected. This invention achieves very good accuracy in the obtained waveforms, i.e., with only a small deviation from standard data information, while employing a relatively limited number of nodes and weights, compared to using a fully connected convolutional network.

[0014] In another embodiment, the convolutional neural network is trained by the computational system that repeatedly performs the following steps: a first repeated step is to generate a training radio frequency waveform and a plurality of training spatially selective gradient pulse waveforms using a selective excitation pulse design algorithm. Another repeated step is to calculate the model excitation field of view by inputting the training radio frequency waveform and the plurality of training spatially selective gradient pulse waveforms into a magnetic resonance imaging signal model. A further repeated step is to receive the forward-propagating radio frequency waveform and the plurality of forward-propagating spatially selective training gradient pulse waveforms by inputting the model excitation field of view into the convolutional neural network.

[0015] Another repetitive step is to update the parameters in the convolutional neural network by performing backpropagation using the training radio frequency waveform and the forward propagation radio frequency waveform, and the backpropagation is also performed using matching pairs of the plurality of training spatially selective gradient pulse waveforms and the plurality of forward propagation spatially selective training gradient pulse waveforms. This embodiment may be advantageous because it provides a convolutional neural network capable of specifying patient field of view with arbitrary shapes and sizes of intervals. It should be noted that these steps may be repeated cyclically, but they are typically performed in parallel, and the backpropagation used to train the convolutional neural network is performed as a vector computation process.

[0016] In another embodiment, the selected excitation pulse design algorithm is the Shinnar-Le Roux algorithm.

[0017] In another embodiment, the selected excitation pulse design algorithm is a small flip angle approximation algorithm.

[0018] In another embodiment, the selected excitation pulse design algorithm is a numerical optimal control algorithm.

[0019] In another embodiment, the magnetic resonance imaging signal model is a numerical solution to the Bloch equation.

[0020] In another embodiment, the pulse sequence command is configured to acquire the k-space data according to a parallel imaging magnetic resonance imaging protocol. This embodiment can be particularly advantageous because, when combined with a custom excitation field of view, very accurate and rapid magnetic resonance images can be acquired.

[0021] In another embodiment, the execution of the machine-executable instructions further causes the computing system to receive a survey magnetic resonance image. The execution of the machine-executable instructions also causes the computing system to plot the survey magnetic resonance image on a display. The execution of the machine-executable instructions further causes the computing system to receive a selection of at least one excitation field of view in response to displaying the survey magnetic resonance image. The selection of the at least one excitation field of view is within the survey magnetic resonance image. This embodiment is advantageous because it allows for the customization of the excitation field of view for a specific object and magnetic resonance imaging protocol.

[0022] In another embodiment, the selection of the at least one field of view is received from an automatic image segmentation algorithm. For example, if a radiologist orders a cardiac magnetic resonance imaging protocol, automatic segmentation can be used to very precisely limit and specify the imaging of the heart.

[0023] In another embodiment, the selection of the at least one field of view is received from a user interface. For example, the system operator can precisely depict the desired field of view. This can provide a more flexible means of performing magnetic resonance imaging.

[0024] In another embodiment, the selection of the at least one field of view is received from an automatic image segmentation algorithm and from a user interface. For example, a portion of the selected excitation field of view can be received from the automatic image segmentation algorithm, and other portions can be received from the user interface. The operator can also use the user interface to correct the segmentation from the algorithm.

[0025] In another embodiment, the plurality of spatially selective gradient pulse waveforms are two spatially selective gradient pulse waveforms. Each of the at least one excitation field of view is a two-dimensional excitation field of view. For example, two gradient pulse waveforms are specified, and these are used to acquire two-dimensional slices. However, it can be configured to acquire multiple slices.

[0026] In another embodiment, the plurality of spatially selective gradient pulse waveforms are three spatially selective gradient pulse waveforms. Each of the at least one excitation field of view is a three-dimensional excitation field of view. This can be advantageous because any three-dimensional excitation field of view can be specified.

[0027] In another embodiment, each of the at least one excitation field of view has an arbitrary shape. The excitation field of view is specified by placing complex values ​​into a complex array used to encode the selection of at least one excitation field of view. This enables the specification of excitation fields of view with arbitrary shapes.

[0028] In another embodiment, each of the at least one excitation field of view is spatially disconnected. For example, since the excitation field of view can be arbitrarily specified for three-dimensional acquisition, multiple excitation fields of view can be specified. An example would be a magnetic resonance imaging scan in which magnetic resonance images of breast tissue from two different breasts are acquired simultaneously, but for different radio frequency excitation fields of view.

[0029] In another embodiment, the neural network is a multi-scale convolutional neural network.

[0030] In another embodiment, the convolutional neural network is a U-Net convolutional neural network.

[0031] In another embodiment, the medical system includes the magnetic resonance imaging system. The execution of the machine-executable instructions further enables the computing system to acquire the k-space data by controlling the magnetic resonance imaging system using modified pulse sequence commands. The execution of the machine-executable instructions also enables the computing system to reconstruct the magnetic resonance imaging data based on the k-space data. This embodiment may be advantageous because the excitation field of view has been customized for a specific magnetic resonance imaging acquisition. This can, for example, provide superior image quality compared to cases where a pre-calculated excitation field of view has already been used.

[0032] In another aspect, the present invention provides a computer program including machine-executable instructions for execution by a computing system controlling a medical system. The computer program further includes a convolutional neural network configured to receive a complex array encoding a selection of at least one excitation field of view as input. The convolutional neural network is configured to output a radio frequency waveform and a plurality of spatially selective gradient pulse waveforms in response to receiving the complex array. The convolutional neural network is a multi-task convolutional neural network having a first output for the radio frequency waveform and a separate output for each of the plurality of spatially selective gradient pulse waveforms.

[0033] The execution of the machine-executable instructions causes the computing system to receive the selection of the at least one excitation field of view. The execution of the machine-executable instructions also causes the computing system to receive an initial pulse sequence command configured to control the magnetic resonance imaging system to acquire k-space data describing the object. The execution of the machine-executable instructions further causes the computing system to encode the complex array using the at least one excitation field of view.

[0034] The execution of the machine-executable instructions further causes the computing system to receive the radio frequency waveform and the plurality of spatially selective gradient pulse waveforms in response to inputting the complex array into the convolutional neural network. The execution of the machine-executable instructions also causes the computing system to construct a modified pulse sequence command by modifying the initial pulse sequence command using the radio frequency waveform and the plurality of spatially selective gradient pulse waveforms, such that the pulse sequence command is configured to control the magnetic resonance imaging system to acquire the k-space data from the at least one excitation field of view.

[0035] In another aspect, the present invention provides a medical imaging method using a convolutional neural network. The convolutional neural network is configured to receive a complex array encoding the selection of at least one excitation field of view as input. The convolutional neural network is configured to output a radio frequency waveform and a plurality of spatially selective gradient pulse waveforms in response to receiving the complex array. The convolutional neural network is a multi-task convolutional neural network having a first output for the radio frequency waveform and a separate output for each of the plurality of spatially selective gradient pulse waveforms.

[0036] The method includes receiving the selection of the at least one excitation field of view. The method further includes receiving an initial pulse sequence command configured to control a magnetic resonance imaging system to acquire k-space data describing an object. The method further includes encoding the complex array using the at least one excitation field of view. The method further includes receiving the radio frequency waveform and the plurality of spatially selective gradient pulse waveforms in response to inputting the complex array into the convolutional neural network. Finally, the method further includes constructing a modified pulse sequence command by modifying the initial pulse sequence command using the radio frequency waveform and the plurality of spatially selective gradient pulse waveforms, such that the pulse sequence command is configured to control the magnetic resonance imaging system to acquire the k-space data from the at least one excitation field of view.

[0037] In another aspect, the present invention provides a method for training a convolutional neural network. The method includes repeatedly performing the following steps. It should be understood that these steps may be repeated sequentially multiple times, or they may be performed in parallel multiple times. The method uses a selected excitation pulse design algorithm to generate training radio frequency waveforms and multiple training spatially selective gradient pulse waveforms. The method further includes calculating the model excitation field of view by inputting the training radio frequency waveforms and the multiple training spatially selective gradient pulse waveforms into a magnetic resonance imaging signal model.

[0038] The method further includes receiving a forward-propagating radio frequency waveform and a plurality of forward-propagating spatially selective training gradient pulse waveforms by inputting the model excitation field of view into the convolutional neural network. The method also includes updating the parameters of the convolutional neural network by performing backpropagation using the training radio frequency waveforms and the forward-propagating radio frequency waveforms. Backpropagation is then performed using matching pairs of the plurality of training spatially selective gradient pulse waveforms and the plurality of forward-propagating spatially selective training gradient pulse waveforms.

[0039] It should be understood that one or more of the foregoing embodiments of the present invention can be combined as long as the combined embodiments are not mutually exclusive.

[0040] As those skilled in the art will recognize, various aspects of the present invention can be implemented as apparatus, method, or computer program product. Accordingly, various aspects of the present invention can take the form of a completely hardware embodiment, a completely software embodiment (including firmware, resident software, microcode, etc.), or an embodiment combining software and hardware aspects (all of which may be referred to herein as "circuit," "module," or "system" in general). Furthermore, various aspects of the present invention can take the form of a computer program product implemented in one or more computer-readable media having computer-executable code implemented thereon.

[0041] Any combination of one or more computer-readable media can be used. The computer-readable media can be a computer-readable signal medium or a computer-readable storage medium. As used herein, "computer-readable storage medium" encompasses any tangible storage medium capable of storing instructions executable by a processor or computing system of a computing device. A computer-readable storage medium may be referred to as a computer-readable non-transitory storage medium. A computer-readable storage medium may also be referred to as a tangible computer-readable medium. In some embodiments, a computer-readable storage medium may also be capable of storing data accessible by a computing system of a computing device. Examples of computer-readable storage media include, but are not limited to: floppy disks, magnetic hard disk drives, solid-state drives, flash memory, USB thumb drives, random access memory (RAM), read-only memory (ROM), optical discs, magneto-optical discs, and register files of computing systems. Examples of optical discs include compact discs (CDs) and digital universal discs (DVDs), such as CD-ROMs, CD-RWs, CD-Rs, DVD-ROMs, DVD-RWs, or DVD-R discs. The term computer-readable storage medium also refers to various types of recording media accessible by a computer device via a network or communication link. For example, data can be retrieved on a modem, the Internet, or a local area network. Computer-executable code implemented on a computer-readable medium may be transmitted using any suitable medium, including but not limited to wireless, wired, fiber optic cable, RF, or any suitable combination thereof.

[0042] Computer-readable signal media may include propagated data signals having computer-executable code implemented therein, for example, in baseband or as a carrier wave. Such propagated signals may take any variety of forms, including but not limited to electromagnetic, optical, or any suitable combination thereof. A computer-readable signal medium may be any computer-readable medium that is not a computer-readable storage medium and is capable of conveying, propagating, or transmitting a program used by or in conjunction with an instruction execution system, apparatus, or device.

[0043] "Computer memory" or "memory" is an example of a computer-readable storage medium. Computer memory is any memory that can be directly accessed by a computing system. "Computer storage device" or "storage device" is another example of a computer-readable storage medium. A computer storage device is any non-volatile computer-readable storage medium. In some embodiments, a computer storage device may also be computer memory, or vice versa.

[0044] As used herein, "computing system" encompasses electronic components capable of running programs or machine-executable instructions or computer-executable code. References to computing systems, including examples of "computing systems," should be interpreted as potentially encompassing more than one computing system or processing core. A computing system may, for example, be a multi-core computing system. A computing system can also refer to a collection of computing systems within a single computer system or distributed across multiple computer systems. The term "computing system" should also be interpreted as potentially referring to a collection or network of computing devices, each including a processor or computing system. The machine-executable code or instructions may be run by multiple computing systems or processors, which may reside within the same computing system or even be distributed across multiple computing systems.

[0045] Machine-executable instructions or computer-executable code may include instructions or programs that cause a processor or other computing system to perform aspects of the present invention. Computer-executable code for performing operations related to aspects of the present invention may be written in any combination of one or more programming languages ​​and compiled into machine-executable instructions, including object-oriented programming languages ​​such as Java, Smalltalk, C++, etc., and conventional procedural programming languages ​​such as the "C" programming language or similar programming languages. In some instances, the computer-executable code may be in the form of a high-level language or in a pre-compiled form and used in conjunction with an interpreter that generates machine-executable instructions at runtime. In other instances, the machine-executable instructions or computer-executable code may be in the form of programming for programmable gate arrays.

[0046] The computer-executable code may be executed entirely on the user's computer, partially on the user's computer (as a standalone software package), partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In the latter case, the remote computer may be connected to the user's computer via any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., via the Internet provided by an Internet service provider).

[0047] Aspects of the invention are described with reference to flowchart illustrations, diagrams, and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It should be understood that, when applicable, each block or portion of a flowchart illustration, diagram, and / or block diagram can be implemented by computer program instructions in the form of computer-executable code. It should also be understood that combinations of blocks from different flowchart illustrations, diagrams, and / or block diagrams can be combined when not mutually exclusive. These computer program instructions can be provided to a computing system of a general-purpose computer, a special-purpose computer, or other programmable data processing apparatus that produces the machine, such that the instructions, executable via the computer or other programmable data processing apparatus, create units for implementing the functions / actions specified in the flowchart illustrations and / or one or more block diagram blocks.

[0048] These machine-executable instructions or computer program instructions may also be stored in a computer-readable medium that can instruct a computer, other programmable data processing apparatus or other device to operate in a particular manner, such that the instructions stored in the computer-readable medium produce an article of writing including instructions that implement the functions / actions specified in flowcharts and / or one or more block diagrams.

[0049] The machine-executable instructions or computer program instructions may also be loaded onto a computer, other programmable data processing apparatus or other device to cause a series of operational steps to be performed on the computer, other programmable apparatus or other device, thereby producing a computer-implemented process, such that the instructions, which execute on the computer or other programmable apparatus, provide a process for the function / action specified in the flowchart and / or one or more block diagram boxes.

[0050] As used herein, a "user interface" is an interface that allows a user or operator to interact with a computer or computer system. A "user interface" can also be referred to as a "human-machine interface device." A user interface can provide information or data to and / or receive information or data from an operator. A user interface enables input from an operator to be received by the computer and output from the computer to the user. In other words, the user interface allows an operator to control or manipulate the computer, and the interface allows the computer to indicate the effects of the operator's control or manipulation. The display of data or information on a monitor or graphical user interface is an example of providing information to an operator. The reception of data via a keyboard, mouse, trackball, touchpad, pointing stick, graphics tablet, joystick, game controller, webcam, headset, pedal, wired gloves, remote control, and accelerometer are all examples of user interface components that implement the reception of information or data from an operator.

[0051] As used herein, "hardware interface" encompasses the interfaces that enable a computer system to interact with and / or control external computing devices and / or devices. A hardware interface allows the computing system to send control signals or instructions to external computing devices and / or devices. It also enables the computing system to exchange data with external computing devices and / or devices. Examples of hardware interfaces include, but are not limited to: Universal Serial Bus (USB), IEEE 1394 port, parallel port, IEEE 1284 port, serial port, RS-232 port, IEEE-488 port, Bluetooth connectivity, wireless LAN connectivity, TCP / IP connectivity, Ethernet connectivity, control voltage interfaces, MIDI interfaces, analog input interfaces, and digital input interfaces.

[0052] As used herein, “display” or “display device” encompasses an output device or user interface suitable for displaying images or data. Displays can output visual, audio, and / or tactile data. Examples of displays include, but are not limited to: computer monitors, television screens, touchscreens, tactile electronic displays, Braille screens, cathode ray tubes (CRTs), memory 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 (OLED) displays, projectors, and head-mounted displays.

[0053] K-space data are defined in this paper as measurements of radio frequency signals emitted by atomic spins, recorded by the antenna of a magnetic resonance imaging (MRI) device during a magnetic resonance imaging (MRI) scan. Magnetic resonance data is an example of tomographic medical image data.

[0054] Magnetic resonance imaging (MRI) data, or magnetic resonance (MR) images, are defined in this paper as reconstructed two-dimensional or three-dimensional visualizations of anatomical data contained within k-space data. Such visualizations can be performed using a computer. Attached Figure Description

[0055] In the following description, preferred embodiments of the invention will be illustrated by way of example only and with reference to the accompanying drawings, in which:

[0056] Figure 1 An example of a medical system is illustrated;

[0057] Figure 2 The illustration shows the use of Figure 1 A flowchart of the methods used in medical systems;

[0058] Figure 3 This illustration shows another example of a medical system;

[0059] Figure 4 The illustration shows the use of Figure 3 A flowchart of the methods used in medical systems;

[0060] Figure 5 A flowchart illustrating the method for training a convolutional neural network is shown.

[0061] Figure 6 An example of a convolutional neural network is illustrated.

[0062] Figure 7 The illustration shows the effectiveness of convolutional neural networks in providing the desired excitation field of view;

[0063] Figure 8 The desired field of view is illustrated; and

[0064] Figure 9 The illustration shows the field of view generated using a convolutional neural network and the field of view generated using the SLR algorithm.

[0065] List of reference numerals

[0066] 100 Medical Systems

[0067] 102 Computer

[0068] 104 Computing System

[0069] 106 Optional hardware interfaces

[0070] 108 optional user interfaces

[0071] 110 Memory

[0072] 120 Machine-executable instructions

[0073] 122 Convolutional Neural Networks

[0074] 124 Selection of at least one excitation field of view

[0075] 126 Initial Pulse Sequence Command

[0076] 128 complex array

[0077] 130 RF waveform

[0078] More than 132 spatially selective gradient pulse waveforms

[0079] 134 Modified Pulse Sequence Command

[0080] 200 Selecting to receive at least one excitation field of view

[0081] 202 Receives an initial pulse sequence command configured to control the magnetic resonance imaging system to acquire k-space data describing the object.

[0082] 204 Encoding the complex array using at least one excitation field of view

[0083] 206 Receives radio frequency waveforms and multiple spatially selective gradient pulse waveforms in response to inputting a complex array into a convolutional neural network.

[0084] 208 A modified pulse sequence command is constructed by modifying the initial pulse sequence command using radio frequency waveforms and multiple spatially selective gradient pulse waveforms, such that the pulse sequence command is configured to control the magnetic resonance imaging system to acquire k-space data from at least one excitation field of view.

[0085] 300 Medical System

[0086] 302 Magnetic Resonance Imaging System

[0087] 304 magnet

[0088] 306 Magnet Chamber

[0089] 308 Imaging Area

[0090] 310 Magnetic Gradient Coil

[0091] 312 Magnetic Gradient Coil Power Supply

[0092] 314 RF coil

[0093] 316 transceiver

[0094] 318 Objects

[0095] 320 Object support

[0096] 322 Anatomical Structure

[0097] 324 Excitation Field

[0098] 330 k spatial data

[0099] 332 Magnetic Resonance Imaging Data

[0100] 400 Acquiring k-space data by controlling the magnetic resonance imaging system using modified pulse sequence commands.

[0101] 402 Reconstructing Magnetic Resonance Imaging Data from k-space Data

[0102] 500 uses a selective excitation pulse design algorithm to generate training RF waveforms and multiple training space selective gradient pulse waveforms.

[0103] 502 The modeled excitation field of view is calculated by inputting training radio frequency waveforms and multiple training spatially selective gradient pulse waveforms into the magnetic resonance imaging signal model.

[0104] 504 The modeled excitation field of view is input into a convolutional neural network to receive forward-propagating radio frequency waveforms and multiple forward-propagating spatially selective training gradient pulse waveforms.

[0105] 506 Updates the parameters of a convolutional neural network by utilizing training radio frequency waveforms and forward propagation radio frequency waveforms, and by performing backpropagation using matched pairs of multiple training spatially selective gradient pulse waveforms and multiple forward propagation spatially selective training gradient pulse waveforms.

[0106] 600 convolutional kernels with ReLU and max-pooling layers

[0107] 602 Convolutional kernels with ReLU and max-pooling layers

[0108] 604 convolutional kernels with ReLU and max-pooling layers

[0109] 606 Planarization layer

[0110] 608 Fully Connected Layer

[0111] 608' Fully Connected Layer

[0112] 608'' Fully Connected Layer

[0113] 700 Generated Excitation Field of View

[0114] 702 Standard Data Excitation Field of View

[0115] 800 Expected Motivational Field

[0116] 900 Excitation field of view generated by CNN

[0117] 902 Excitation field of view generated by the use of RF pulses in SLR design Detailed Implementation

[0118] Elements with similar numbers in these figures are either equivalent or perform the same function. If the functions are equivalent, elements already discussed will not need to be discussed in later figures.

[0119] Figure 1 A diagram illustrating an example of a medical system 100 is shown. Figure 1The medical system 100 is depicted as including a computer 102, which includes a computing system 104. The computing system 104 is intended to represent one or more processing cores or computing systems that may be located in the same location or distributed. The computer 102 is shown as including an optional hardware interface 106 and an optional user interface connected to the computing system 104. The hardware interface 106 may, for example, enable the computing system 104 to exchange data with and control other components of the medical system 100. The user interface 108 enables the operator of the medical system 100 to control and interact with it. The user interface 108 may, for example, include a graphical user interface or other graphical control device.

[0120] The medical system 100 is also shown as including a memory 110 that communicates with the computing system 104. The memory 110 is intended to represent any combination of memories that can be accessed by the computing system 104.

[0121] Memory 110 is shown as containing machine-executable instructions 120. Machine-executable instructions 120 are instructions that enable the computing system to perform tasks such as controlling other components and performing various digital and image processing tasks. Memory 110 is also shown as including a convolutional neural network 122. This convolutional neural network is a multi-task convolutional neural network having a first output for outputting a radio frequency waveform and separate outputs for each of a plurality of spatially selective gradient pulse waveforms. The convolutional neural network outputs these in response to receiving a complex array encoding the selection of at least one excitation field of view for magnetic resonance imaging acquisition.

[0122] Memory 110 is also shown as containing the selection of at least one excitation field of view 124. Memory is also shown as containing initial pulse sequence commands 126. These may be, for example, a set of pulse sequence commands retrieved from a database or selected when a particular protocol is desired to be executed.

[0123] The memory 110 is also shown to contain a complex array for encoding the selection of at least one excitation field of view 124. The memory is also shown to contain a radio frequency waveform 130 and a plurality of spatially selective gradient pulse waveforms 132 received from the convolutional neural network 122 in response to the input complex array 128. The memory 110 is also shown to contain a modified pulse sequence command 134. The modified pulse sequence command 134 is constructed based on an initial pulse sequence command 126 and is modified to include the radio frequency waveform 130 and the plurality of spatially selective gradient pulse waveforms 132.

[0124] Figure 2 The illustrated operation is shown. Figure 1The flowchart describes a method for a medical system 100. First, in step 200, the selection of at least one excitation field of view 124 is received. Next, in step 202, an initial pulse sequence command 126 is received. Next, in step 204, a complex array 128 is encoded using at least one excitation field of view 124. In step 206, a radio frequency waveform 130 and multiple spatially selective gradient pulse waveforms 132 are received in response to inputting the complex array 128 into a convolutional neural network 122. Finally, in step 208, a modified pulse sequence command is constructed by integrating the radio frequency waveform 130 and the multiple spatially selective gradient pulse waveforms 132 into the initial pulse sequence command 126.

[0125] Figure 3 The illustration shows another example of a medical system 300. Medical system 300 is similar to... Figure 1 The medical system 100 includes, in addition to, a magnetic resonance imaging system 302 controlled by a computing system 104.

[0126] The magnetic resonance imaging system 302 includes a magnet 304. Magnet 304 is a superconducting cylindrical magnet with a bore 306 passing through it. The use of different types of magnets is also possible; for example, both split cylindrical magnets and so-called open magnets are possible. A split cylindrical magnet is similar to a standard cylindrical magnet, except that the cryostat has been divided into two parts to allow access to the isoplanar surface of the magnet; such a magnet can be used, for example, in conjunction with charged particle beam therapy. An open magnet has two magnet sections, one above the other, with a sufficiently large space between them to receive the object: the arrangement of the two sections is similar to that of a Helmholtz coil. Open magnets are preferred because the object is subject to less constraint. An assembly of superconducting coils is located inside the cryostat of the cylindrical magnet.

[0127] Within the bore 306 of the cylindrical magnet 304, there exists an imaging region 308 with a sufficiently strong and uniform magnetic field to perform magnetic resonance imaging. Magnetic resonance data is typically acquired for a region of interest. The object 318 is shown supported by an object support 320 such that at least a portion of the object 318 is located within the imaging region 308.

[0128] The magnet chamber 306 also contains an assembly of magnetic field gradient coils 310, which are used to acquire preliminary magnetic resonance data for spatial encoding of magnetic spins within the imaging region 308 of the magnet 304. The magnetic field gradient coils 310 are connected to a magnetic field gradient coil power supply 312. The magnetic field gradient coils 310 are intended to be representative. Typically, the magnetic field gradient coils 310 contain three separate sets of coils for spatial encoding in three orthogonal spatial directions. The magnetic field gradient power supply supplies current to the magnetic field gradient coils. The current supplied to the magnetic field gradient coils 310 is time-controlled and can be either slanted or pulsed.

[0129] Adjacent to the imaging region 308 is a radio frequency (RF) coil 314, which is used to manipulate the orientation of magnetic spins within the imaging region 308 and to receive radio transmissions from spins also located within the imaging region 308. The RF antenna may comprise multiple coil elements. The RF antenna may also be referred to as a channel or antenna. The RF coil 314 is connected to an RF transceiver 316. The RF coil 314 and the RF transceiver 316 may be replaced by separate transmit and receive coils, as well as separate transmitters and receivers. It should be understood that the RF coil 314 and the RF transceiver 316 are representative. The RF coil 314 is intended to also represent a dedicated transmit antenna and a dedicated receive antenna. Similarly, the transceiver 316 may also represent separate transmitters and receivers. The RF coil 314 may also have multiple receive / transmit elements, and the RF transceiver 316 may have multiple receive / transmit channels. For example, if a parallel imaging technique (such as SENSE) is performed, the RF coil 314 will have multiple coil elements.

[0130] Transceiver 316 and gradient controller 312 are shown as hardware interface 106 connected to computer system 102.

[0131] Within imaging region 308, there exists an anatomical structure 322 of object 318. It can be seen that there exists an excitation field of view 324 that has been modified to fit precisely the anatomical structure 322. The excitation field of view 324 is specified by selecting at least one excitation field of view 124 and has been encoded into the complex array 128.

[0132] The memory 110 is also shown to contain k-space data 330 that has been acquired by the magnetic resonance imaging system 302 by controlling it using a modified pulse sequence command 134. The k-space data 330 describes anatomical structures 322. The memory 110 is also shown to contain magnetic resonance imaging data 332 that has been reconstructed based on the k-space data 330.

[0133] Figure 4 The illustrated operation is shown. Figure 3 A flowchart of a method for a medical system 300. This method uses, as shown in... Figure 2 The illustrated steps 200-208 begin. After step 208 has been executed, step 400 is executed, and then step 402 is executed. In step 400, k-space data 330 is acquired by controlling the magnetic resonance imaging system 302 using a modified pulse sequence command 134. Finally, in step 402, magnetic resonance imaging data 332 is reconstructed from the k-space data 330.

[0134] Figure 5A flowchart illustrating a method for training a convolutional neural network 122 is shown. Steps 500, 502, 504, and 506 are shown as being executed cyclically or repeatedly. However, these steps can be executed in parallel. That is, all training data can be processed by executing step 504 multiple times, then step 506 multiple times, then step 500 multiple times, and then step 502 multiple times. First, in step 500, a training radio frequency waveform and multiple training spatially selective gradient pulse waveforms are generated using a selected excitation pulse design algorithm. Next, in step 502, the model excitation field of view is calculated by inputting the training radio frequency waveforms and the training spatially selective gradient pulses into a magnetic resonance imaging signal model.

[0135] Subsequently, in step 504, the forward propagation radio frequency waveform and multiple forward propagation spatially selective training gradient pulse waveforms are received by inputting the model excitation field of view into the convolutional neural network. Then, in step 506, the convolutional neural network is trained by performing backpropagation to update the parameters of the convolutional neural network. Backpropagation is performed using the training radio frequency waveform and the forward propagation radio frequency waveform. Backpropagation is also performed using matching pairs of multiple training spatially selective gradient pulse waveforms and multiple forward propagation spatially selective training gradient pulse waveforms.

[0136] Modern MRI systems typically load pre-defined radiofrequency (RF) pulses and accompanying gradients during clinical scans, with minimal adaptation to the specific requirements of each scan. An example can be provided using a multi-task, multi-scale CNN method for real-time design of the excitation RF pulses and accompanying gradient waveforms to achieve spatial two-dimensional selectivity. The CNN-designed RF and gradients closely approximate their SLR counterparts, achieving an NRMSE of 0.0075 ± 0.0038 on 400 test data points. Phantom imaging using the predicted RF also closely approximates the excitation of the SLR design. The algorithm runs on a commercially available workstation within 500 ms.

[0137] While occasional requirements such as self-refocusing or adiabatic properties are necessary, the most common requirement for two-dimensional excitation pulses is an excitation distribution in two dimensions. Therefore, RF pulse design presents an inverse problem given the desired spatial and frequency distributions, where small tip angle approximations apply to small excitation angles, and the Shinnar-Le Roux (SLR) algorithm has been shown to handle any flip angle from 0° to 180°. For modern MRI instruments, conventional RF pulses and accompanying gradients are typically designed at system deployment using either of these methods, based on hardware constraints including the maximum permissible B1 field, gradient field strength, and gradient field transition rate, and the resulting waveforms are stored on the system. For each specific scan, some of these waveforms are loaded into the spectrometer, where they are scaled in amplitude or stretched in time.

[0138] While the SLR method is practical and allows for the acquisition of routine imaging sequences, it limits its applications, preventing any acquisition where highly specific signal excitations would be beneficial. For example, spatial two-dimensional excitations have been used clinically in diffusion imaging of the spinal cord, breast, pancreas, prostate, and cervix to limit phase-encoded FOV in order to improve resolution and reduce geometric distortion. Beyond using the SLR algorithm to generate simple two-dimensional excitations, it is possible to generate RF and gradient waveforms with completely arbitrary shapes, thereby optimizing the balance between RF performance and duration. However, the design of these arbitrary RF and gradient waveforms requires iterative algorithms, which are time-consuming and preclude real-time waveform generation.

[0139] This example discloses a neural network-based two-dimensional selective RF design, where both the RF waveform and gradient waveform are generated via a multi-task convolutional neural network in less than one second. The method can be executed at each specific scan time to optimize RF performance under given imaging conditions, including hardware constraints (RF and gradient constraints and the number of transmission channels), system defects (eddy currents and B0 inhomogeneities), imaging object-related challenges (local magnetic field inhomogeneities in both B0 and B1), and specification-related requirements (FOV of interest). Variable rate selective excitation (VERSE) methods have sometimes been used clinically, and their iterative design algorithms have recently been simplified; however, they only address a subset of the problems listed above. Here, we present pilot results from this comprehensive real-time RF design project. This demonstrates the utility of DL as a path for real-time optimization of arbitrary excitation pulses.

[0140] Figure 6 The diagram illustrates an example of the topology or structure of a convolutional neural network 122. The neural network 122 receives an input array 128. A first layer 600 exists, comprising a convolutional kernel plus ReLU plus max pooling. This is then fed into a second layer, comprising another convolutional kernel plus ReLU plus max pooling. This is then fed into a third convolutional kernel or layer 604, which also has ReLU and max pooling layers. This is then fed into a flattening layer 606, which is then connected to three separate fully connected layers 608, 608', and 608''. Layer 608 outputs a radio frequency waveform 130. Only the connected layers 608' and 608'' output two spatially selective gradient pulse waveforms 132. Figure 6 The illustrated neural network 122 can be used for two-dimensional slicing. The basic structure can be reused and modified for three-dimensional excitation fields. For example, additional connection layers can be added for third-space selective gradient pulse waveforms, and the input can also be modified to accept complex three-dimensional arrays. The exact structure of the various convolutional layers is not required. However, the use of individual connection layers connected to the flattening layer 606 has shown superior results compared to other types of neural network topologies.

[0141] Example multi-task CNNs can be built based on convolutional frameworks (hereinafter referred to as CNNs). As shown above... Figure 6 As shown, it includes an input layer; three convolutional blocks 600, 602, and 604; a flattening layer 606; fully connected layers 608, 608', and 608''; and an output layer. The input layer takes the real and imaginary parts of the desired activation distribution and is therefore a three-dimensional matrix with a third dimension of size 2 (e.g., 101×101×2). The first convolutional block 600 contains 16 convolutional kernels and rectified linear units (ReLU), followed by a max-pooling step. The second convolutional block 602 and the third convolutional block 604 contain 32 and 64 convolutional kernels, respectively, both followed by the same ReLU activation and max-pooling. The flattening layer 606 is then configured to transform the data into a 9216 vector, which is concatenated by the fully connected layer to produce high-level inference with an output of matching dimensions. The convergence of the network is evaluated by minimizing the loss, which is measured as the root mean square error (RMSE). The network parameters are tuned until the optimal network settings are achieved.

[0142] RF, GRS, and GRF waveforms are considered as three channels of the RF pulse to generate the excitation distribution, and a multi-task learning technique is employed, in which a flat layer carrying shared features is connected in parallel with three fully connected layers. Furthermore, the convolutional kernels are adjusted to multiple sizes of 1×1, 3×3, 5×5, and 7×7, similar to the strategy in the initial module. Compared to regular convolutional operations, convolutions with multiple kernel sizes capture feature information at different scales, and this avoids potential overfitting problems if deeper networks are used.

[0143] An example may contain one or more of the following characteristics:

[0144] 1) Generate a sufficiently large set of 2DRF impulses using the standard SLR algorithm used to train CNNs.

[0145] 2) Generate the excitation distribution for each pulse using B1och simulation, which is used to train the CNN.

[0146] 3) Custom convolutional neural network architecture with optimized layers and connections.

[0147] 4) Use the generated excitation distribution as input and the corresponding RF pulse as output to train the network.

[0148] 5) Input the desired stimulus distribution into the trained network to obtain the RF of the CNN design.

[0149] Figure 7The illustration shows the effectiveness of the radio frequency waveform 130 and multiple spatially selective gradient pulse waveforms 132 provided by the convolutional neural network 122. In this figure, the exemplary radio frequency waveform 130 and multiple spatially selective gradient pulse waveforms 132 are illustrated at the top. These waveforms 130, 132 are used to provide an excitation distribution 700 in the middle. This closely approximates the standard data excitation field view depicted in box 702 below.

[0150] Furthermore, the predicted RF can actually provide an excitation distribution that is very close to the desired FOV excitation distribution. This was tested using phantom imaging on a clinical MR 3.0T scanner.

[0151] Also used Figure 8 and 9 The diagram illustrates the effectiveness of convolutional neural networks. Figure 8 The desired excitation field of view is shown at 800°. Figure 9 In Figure 900, the excitation field of view generated by the use of a convolutional neural network is shown. On the right, box 902 shows the excitation field of view generated by an SLR-designed RF pulse. It can be seen that the convolutional neural network matches the SLR-designed RF pulse very well.

[0152] An example may contain one or more of the following characteristics:

[0153] 1) CNN-based RF design for both RF and gradient waveforms.

[0154] 2) Multitasking techniques are used in conjunction with a shared layer to connect the RF waveform and the two gradients.

[0155] 3) Multi-scale kernels in CNNs are used to capture details at different scales in the input FOV image domain.

[0156] 4) Select echo plane trajectories for RF generation of the training dataset.

[0157] This technology may be tested through experiments on scanning and acquiring images.

[0158] Spatial two-dimensional or three-dimensional excitation can be applied in imaging of the brain, spinal cord, head and neck tumors, breast, pancreas, and prostate. This allows for improved spatial resolution and reduced geometric distortion for diffusion-weighted imaging.

[0159] Although the invention has been described and illustrated in detail in the accompanying drawings and the foregoing description, such description and illustration are to be regarded as illustrative or exemplary rather than restrictive; the invention is not limited to the disclosed embodiments.

[0160] Those skilled in the art, through studying the accompanying drawings, description, and claims, will be able to understand and implement other variations of the disclosed embodiments in practicing the claimed invention. In the claims, the word "comprising" does not exclude other elements or steps, and the words "a" or "an" do not exclude a plurality. A single processor or other unit may fulfill the functions of several items recited in the claims. Although specific elements are recited in dissimilar dependent claims, this does not indicate that combinations of these elements cannot be advantageously used. Computer programs may be stored and / or distributed on suitable media, such as optical storage media or solid-state media provided with or as part of other hardware, but computer programs may also be distributed in other forms, such as via the Internet or other wired or wireless telecommunications systems. No reference numerals in the claims shall be construed as limiting the scope.

Claims

1. A medical system (100, 300), comprising: A memory (110) storing machine-executable instructions (120) and a convolutional neural network (122), wherein the convolutional neural network is configured to receive a complex array (128) encoding a selection (124) of at least one excitation field of view (324, 900) as input, wherein the convolutional neural network is configured to output a radio frequency waveform (130) and a plurality of spatially selective gradient pulse waveforms (132), wherein the convolutional neural network is a multi-task convolutional neural network having a first output for the radio frequency waveform and a separate output for each of the plurality of spatially selective gradient pulse waveforms; and - Computing system (104), wherein the execution of machine-executable instructions enables the computing system to: - Receive (200) the selection of the at least one excitation field of view; - Receive (202) an initial pulse sequence command (126), the initial pulse sequence command being configured to control the magnetic resonance imaging system (302) to acquire k-space data (330) of the descriptive object (318); - Encode the complex array using the at least one excitation field of view (204); - In response to inputting the complex array into the convolutional neural network, the radio frequency waveform and the plurality of spatially selective gradient pulse waveforms are received (206); and - By modifying the initial pulse sequence command using the radio frequency waveform and the plurality of spatially selective gradient pulse waveforms, a modified pulse sequence command (208) is constructed such that the pulse sequence command is configured to control the magnetic resonance imaging system to acquire the k-space data from the at least one excitation field of view.

2. The medical system according to claim 1, wherein, The convolutional neural network is trained by the computational system that repeatedly performs the following steps: - Use a selective excitation pulse design algorithm to generate (500) training RF waveforms and multiple training space selective gradient pulse waveforms; - The excitation field of view modeled by (502) is calculated by inputting the training radio frequency waveform and the plurality of training spatially selective gradient pulse waveforms into the magnetic resonance imaging signal model; - By inputting the modeled excitation field of view into the convolutional neural network to receive (504) forward propagation radio frequency waveforms and multiple forward propagation spatially selective training gradient pulse waveforms; - The parameters of the convolutional neural network are updated (506) by performing backpropagation using the training radio frequency waveform and the forward propagation radio frequency waveform and by using matching pairs of the plurality of training spatially selective gradient pulse waveforms and the plurality of forward propagation spatially selective training gradient pulse waveforms.

3. The medical system according to claim 2, wherein, The selective excitation pulse design algorithm is any one of the following: the Shinnar–Le Roux algorithm, the small flip angle approximation algorithm, and the numerical optimal control algorithm.

4. The medical system according to claim 2 or 3, wherein, The magnetic resonance imaging signal model is a numerical solution to the Bloch equation.

5. The medical system according to any one of the preceding claims, wherein, The pulse sequence command is configured to acquire the k-space data according to a parallel imaging magnetic resonance imaging protocol.

6. The medical system according to any one of the preceding claims, wherein, The execution of the machine-executable instructions also enables the computing system to: -Receive and investigate magnetic resonance images; - Plot the investigated magnetic resonance images on the display; and - In response to displaying the surveyed magnetic resonance image, the selection of the at least one excitation field of view is received, wherein the selection of the at least one excitation field of view is within the surveyed magnetic resonance image.

7. The medical system according to any one of the preceding claims, wherein, The selection of the at least one field of view is received from any of the following: an automatic image segmentation algorithm, a user interface, and a combination thereof.

8. The medical system according to any one of the preceding claims, wherein, The plurality of spatially selective gradient pulse waveforms are two spatially selective gradient pulse waveforms, wherein each of the at least one excitation field of view is a two-dimensional excitation field of view.

9. The medical system according to any one of claims 1 to 7, wherein, The plurality of spatially selective gradient pulse waveforms are three spatially selective gradient pulse waveforms, wherein each of the at least one excitation field of view is a three-dimensional excitation field of view.

10. The medical system according to any one of the preceding claims, wherein, Each of the at least one excitation field of view has an arbitrary shape and / or is spatially disconnected.

11. The medical system according to any one of the preceding claims, wherein, The convolutional neural network is any one of the following: a multi-scale convolutional neural network and a U-Net convolutional neural network.

12. The medical system according to any one of the preceding claims, wherein, The medical system includes the magnetic resonance imaging system, and the execution of the machine-executable instructions further enables the computing system to: - The magnetic resonance imaging system is controlled using the modified pulse sequence commands to acquire (400) the k-space data (330); and - Reconstruct (402) magnetic resonance imaging data (332) based on the k-space data.

13. A computer program comprising machine-executable instructions (120) for execution by a computing system (104) controlling a medical system, wherein, The computer program further includes a convolutional neural network (122) configured to receive a complex array (128) encoding a selection (124) of at least one excitation field of view (324, 900) as input, wherein the convolutional neural network is configured to output a radio frequency waveform (130) and a plurality of spatially selective gradient pulse waveforms (132), wherein the convolutional neural network is a multi-task convolutional neural network having a first output for the radio frequency waveform and a separate output for each of the plurality of spatially selective gradient pulse waveforms; The execution of the machine-executable instructions enables the computing system to: - Receive (200) the selection of the at least one excitation field of view; - Receive (202) an initial pulse sequence command (126), the initial pulse sequence command being configured to control the magnetic resonance imaging system (302) to acquire k-space data (330) of the descriptive object (318); - Encode the complex array using the at least one excitation field of view (204); - In response to inputting the complex array into the convolutional neural network, the radio frequency waveform and the plurality of spatially selective gradient pulse waveforms are received (206); and - By modifying the initial pulse sequence command using the radio frequency waveform and the plurality of spatially selective gradient pulse waveforms, a modified pulse sequence command (208) is constructed such that the pulse sequence command is configured to control the magnetic resonance imaging system to acquire the k-space data from the at least one excitation field of view.

14. A method for medical imaging using a convolutional neural network (122), wherein, The convolutional neural network is configured to receive a complex array (128) encoding a selection (200) of at least one excitation field of view (324, 900) as input, wherein the convolutional neural network is configured to output a radio frequency waveform (130) and a plurality of spatially selective gradient pulse waveforms (132), wherein the convolutional neural network is a multi-task convolutional neural network having a first output for the radio frequency waveform and a separate output for each of the plurality of spatially selective gradient pulse waveforms. The method includes: - Receive (200) the selection of the at least one excitation field of view; - Receive (202) an initial pulse sequence command, which is configured to control the magnetic resonance imaging system (302) to acquire k-space data (330) of the descriptive object (318); - Encode the complex array using the at least one excitation field of view (204); - In response to inputting the complex array into the convolutional neural network, the radio frequency waveform and the plurality of spatially selective gradient pulse waveforms are received (206); and - By modifying the initial pulse sequence command using the radio frequency waveform and the plurality of spatially selective gradient pulse waveforms, a modified pulse sequence command (208) is constructed such that the pulse sequence command is configured to control the magnetic resonance imaging system to acquire the k-space data from the at least one excitation field of view.

15. A method for training a convolutional neural network (122), wherein, The method includes repeatedly performing the following steps: - Use a selective excitation pulse design algorithm to generate (500) training RF waveforms and multiple training space selective gradient pulse waveforms; - The excitation field of view modeled by (502) is calculated by inputting the training radio frequency waveform and the plurality of training spatially selective gradient pulse waveforms into the magnetic resonance imaging signal model; - By inputting the modeled excitation field of view into the convolutional neural network to receive (504) forward propagation radio frequency waveforms and multiple forward propagation spatially selective training gradient pulse waveforms; - The parameters of the convolutional neural network are updated (506) by performing backpropagation using the training radio frequency waveform and the forward propagation radio frequency waveform and by using matching pairs of the plurality of training spatially selective gradient pulse waveforms and the plurality of forward propagation spatially selective training gradient pulse waveforms.

Images