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

JP2023548854A5Active Publication Date: 2024-11-08KONINKLIJKE PHILIPS NV
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
JP2023526956
Authority / Receiving Office
JP · JP
Patent Type
Applications
Current Assignee / Owner
Priority Date
2020-12-15
Filing Date
2021-11-04
Publication Date
2024-11-08
Estimated Expiration
2041-11-04

AI Technical Summary

Technical Problem

Existing MRI systems lack the ability to efficiently generate radiofrequency (RF) and gradient pulse waveforms that can accommodate arbitrarily shaped and sized excitation fields, limiting the flexibility and accuracy of magnetic resonance imaging.

Method used

A convolutional neural network (CNN) is trained to output both high-frequency RF waveforms and spatially selective gradient pulse waveforms, allowing for the customization of excitation fields in MRI systems, including two-dimensional and three-dimensional acquisitions.

Benefits of technology

The CNN-based approach enables rapid and accurate generation of RF and gradient waveforms that can specify arbitrarily shaped and sized excitation fields, improving the flexibility and quality of MRI imaging.

✦ Generated by Eureka AI based on patent content.
Patent Text Reader

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 composite array 128 encoding a selection of at least one excitation field 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 multitasking convolutional neural network. Execution of the machine-executable instructions causes the computing system 104 to receive (200) a selection 124 of at least one excitation field, receive (202) an initial pulse sequence command 126, encode (204) a composite array with the at least one excitation field, receive (206) a radio frequency waveform and a plurality of spatially selective gradient pulse waveforms in response to inputting the composite array into a 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.
Need to check novelty before this filing date? Find Prior Art

Description

[Technical Field]

[0001] The present invention relates to magnetic resonance imaging, and in particular to the design of pulse sequences for controlling magnetic resonance imaging systems. [Background technology]

[0002] A large static magnetic field is used by magnetic resonance imaging (MRI) scanners to align the nuclear spins of atoms as part of a procedure to generate images of a patient's inside the body. This large static magnetic field is called the B0 field or main magnetic field. MRI can be used to spatially measure various quantities or characteristics of a subject. Spatial encoding in magnetic resonance imaging is achieved 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., in their journal publication "Ultrafast (milliseconds), multidimensional RF pulse design with deep learning," Magn. Reson. Med. 2019, 82:586-599, discloses the use of a neural network that takes an input map of a desired excitation region of interest and outputs a single-channel multidimensional RF pulse.

[0004] The academic paper "Joint Design of RF and gradient waveforms via auto-differentiation for tailored excitation in MRI" by Tianrui Luo et al., retrieved August 24, 2020, from Cornell University Library, ARXIV.org, discloses a method for the joint design of RF and gradient waveforms based on an auto-differentiable Bloch simulator. Summary of the Invention

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

[0006] Embodiments provide improved selective RF excitation. They achieve this by using a convolutional neural network trained to output both the radio frequency waveform used to generate the B1 field and the accompanying spatially selective gradient pulse waveform. A variety of different types of neural network architectures were investigated, and a multitasking neural network provided the best results. A multitasking neural network provides multiple outputs in response to receiving one input. In the example described herein, the convolutional neural network takes the complete composite array encoded with the desired RF excitation field of view and outputs the complete radio frequency waveform and the complete spatially selective gradient pulse waveform. If two-dimensional slices are acquired, there are two spatially selective gradient pulse waveforms; if the acquisition is three-dimensional, there are three spatially selective gradient pulse waveforms output.

[0007] In one aspect, the present invention provides a medical system comprising a memory storing machine-executable instructions and a convolutional neural network configured to receive as input a composite array encoding a selection of at least one excitation field of view, the composite array representing, for example, a two-dimensional slice acquired by magnetic resonance imaging or a three-dimensional volume acquired by magnetic resonance imaging.

[0008] 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 composite array. The radio frequency waveform is a waveform used to control a transmitter in a magnetic resonance imaging system to generate a B1 magnetic field. The plurality of spatially selective gradient pulse waveforms are each gradient pulse waveforms that are matched to the radio frequency waveform to acquire magnetic resonance data for at least one excited field of view.

[0009] The convolutional neural network is a multitasking convolutional neural network having a first output for a radio frequency waveform and separate outputs for each of a plurality of spatially selective gradient pulse waveforms. For a two-dimensional magnetic resonance imaging protocol, there are two spatially selective gradient pulse waveforms, and for a three-dimensional magnetic resonance imaging protocol, there are three spatially selective gradient pulse waveforms.

[0010] The medical system further includes a computing system, wherein execution of the machine-executable instructions causes the computing system to receive a selection of at least one excitation field of view, e.g., a field of view of interest to an operator or physician for imaging. 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 subject. Execution of the machine-executable instructions also causes the computing system to receive a radio frequency waveform and a plurality of spatially selective gradient pulse waveforms in response to inputting the composite array into a convolutional neural network.

[0011] Execution of the machine-executable instructions further causes the computing system to configure a modified pulse sequence command by modifying the initial pulse sequence command with a radio frequency waveform and a plurality of 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. This embodiment is advantageous because it allows the excitation field of view to be easily specified by an operator of the magnetic resonance imaging system. This embodiment also allows for the specification of an arbitrarily sized or shaped excitation field of view. In other cases, this embodiment allows for 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 that includes regions where slice-selective RF and gradient pulses need to be specified. Thus, embodiments provide for customization of pulse sequence commands for a particular acquisition by modifying those pulse sequence commands with radio frequency waveforms and multiple spatially selective gradient pulse waveforms.

[0013] The present invention relates to the automatic generation of radio frequency (RF) waveforms and multiple (two or more) spatially selective gradient waveforms from an input excitation field and by modifying an initial pulse sequence command as far as the RF and gradient waveforms are concerned. The automatically generated RF and gradient waveforms are returned in response to the input excitation field by a multitasking convolutional neural network. This multitasking convolutional neural network has several parallel outputs (for each waveform) and is not fully connected. The present invention achieves very good accuracy of the resulting waveforms, i.e., only small deviations from ground truth information, while using a relatively reduced number of nodes and weights compared to the use of a fully connected convolutional network.

[0014] In another embodiment, the convolutional neural network is trained by a computing system that iteratively performs the following steps, where a first iteration step is generating a training radio frequency waveform and a plurality of training spatially selective gradient pulse waveforms using a selective excitation pulse design algorithm, another iteration step is calculating a model excitation field by inputting the training radio frequency waveform and the plurality of training spatially selective gradient pulse waveforms into a magnetic resonance imaging signal model, and another iteration step is receiving a forward propagated radio frequency waveform and a plurality of forward propagated training spatially selective gradient pulse waveforms by inputting the model excitation field into the convolutional neural network.

[0015] Another iterative step is to update the parameters of the convolutional neural network by backpropagating the training high-frequency waveforms and the forward-propagated high-frequency waveforms, and backpropagating is also performed with matched pairs of multiple training spatially selective gradient pulse waveforms and multiple forward-propagated training spatially selective gradient pulse waveforms. This embodiment is advantageous because it provides a convolutional neural network that can specify patient fields of view of arbitrary shape, spacing, and size. It should be noted that these steps may be repeated in a loop, but are typically performed in parallel, and that backpropagation, for training the convolutional neural network, is performed as a vector computation process.

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

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

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

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

[0020] In another embodiment, the pulse sequence commands are configured to acquire k-space data according to a parallel imaging magnetic resonance imaging protocol, which is particularly beneficial because when combined with a custom excitation field of view, it allows for very accurate and rapid magnetic resonance image acquisition.

[0021] In another embodiment, execution of the machine-executable instructions further causes the computing system to receive a survey magnetic resonance image. Execution of the machine-executable instructions further causes the computing system to render the survey magnetic resonance image on a display. 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 customization of the excitation field of view for a particular subject and magnetic resonance imaging protocol.

[0022] In another embodiment, the selection of at least one field of view is received from an automatic image segmentation algorithm. For example, when a radiologist prescribes 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 at least one field of view is received from a user interface, for example, an operator of the system may trace out the exact field of view desired, providing a more flexible means of performing magnetic resonance imaging.

[0024] In another embodiment, the selection of at least one field of view is received from an automated segmentation algorithm and a user interface. For example, some portions of the selection of the excitation field of view may be received from the automated imaging segmentation algorithm and other portions may be received from the user interface. An operator may also correct the segmentation from the algorithm using the user interface.

[0025] In another embodiment, the plurality of spatially selective gradient pulse waveforms are two spatially selective gradient pulse waveforms, each of which is used to acquire at least one excitation field, a two-dimensional excitation field. For example, two gradient pulse waveforms are specified and used to acquire two-dimensional slices. However, they may also 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 is advantageous because any three-dimensional excitation field of view may be specified.

[0027] In another embodiment, each of the at least one excitation field has an arbitrary shape, and the excitation fields are specified by filling in a composite array that is used to encode the selection of the at least one excitation field, thereby allowing the specification of excitation fields having arbitrary shapes.

[0028] In another embodiment, each of the at least one excitation field of view is spatially separated. For example, multiple excitation fields of view may be specified, as the excitation field of view may be arbitrarily specified for three-dimensional acquisition. One example is a magnetic resonance imaging scan in which magnetic resonance images of breast tissue for two different breasts are acquired simultaneously, but with 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 a magnetic resonance imaging system. Execution of the machine-executable instructions further causes a computing system to control the magnetic resonance imaging system with modified pulse sequence commands to acquire k-space data. Execution of the machine-executable instructions further causes the computing system to reconstruct the magnetic resonance imaging data from the k-space data. This embodiment is advantageous because the excitation field of view is customized for a particular magnetic resonance imaging acquisition. This provides superior imaging quality compared to, for example, situations where a pre-calculated excitation field of view is used.

[0032] In another aspect, the present invention provides a computer program comprising machine-executable instructions for execution by a computing system for controlling a medical system. The computer program further comprises a convolutional neural network configured to receive as an input a composite array encoding a selection of at least one excitation field of view. 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 composite array. The convolutional neural network is a multitasking 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] Execution of the machine-executable instructions causes the computing system to receive a selection of at least one excitation field of view, execution of the machine-executable instructions further 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 subject, and execution of the machine-executable instructions further causes the computing system to encode a composite array with the at least one excitation field of view.

[0034] Execution of the machine-executable instructions further causes the computing system to receive a radio frequency waveform and a plurality of spatially selective gradient pulse waveforms in response to inputting the composite array to the convolutional neural network. Execution of the machine-executable instructions further causes the computing system to configure a modified pulse sequence command by modifying the initial pulse sequence command with the radio frequency waveform and the plurality of spatially selective gradient pulse waveforms, wherein the pulse sequence command is configured to control the magnetic resonance imaging system to acquire k-space data from at least one excited field of view.

[0035] In another aspect, the present invention provides a method of medical imaging using a convolutional neural network. The convolutional neural network is configured to receive as an input a composite array encoding a selection of at least one excitation field of view. 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 composite array as an input. The convolutional neural network is a multitasking 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 a selection of 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 the subject. The method further includes encoding a composite array using the at least one excitation field of view. The method further includes receiving a radio frequency waveform and a plurality of spatially selective gradient pulse waveforms in response to inputting the composite array into a convolutional neural network. Finally, the method further includes constructing a modified pulse sequence command by modifying the initial pulse sequence command with 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 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 comprising repeatedly performing the following steps. It should be understood that these steps may be repeated multiple times sequentially or performed multiple times in parallel: generating a training radio frequency waveform and a plurality of training spatially selective gradient pulse waveforms using a selective excitation pulse design algorithm; and calculating a 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.

[0038] The method further includes receiving a forward-propagated high-frequency waveform and a plurality of forward-propagated training spatially-selective gradient pulse waveforms by inputting the model excitation field into a convolutional neural network. The method further includes updating parameters of the convolutional neural network by backpropagating the training high-frequency waveform and the forward-propagated high-frequency waveform. Backpropagation is then similarly performed with matched pairs of the plurality of training spatially-selective gradient pulse waveforms and the plurality of forward-propagated training spatially-selective gradient pulse waveforms.

[0039] It is understood that one or more of the above-described embodiments of the present invention may be combined, unless the combined embodiments are mutually exclusive.

[0040] As will be appreciated by one skilled in the art, aspects of the present invention may be embodied as an apparatus, a method, or a computer program product. Accordingly, aspects of the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment (including firmware, resident software, microcode, etc.), or an embodiment combining software and hardware aspects, all of which may be referred to herein inclusively as a "circuit," "module," or "system." Furthermore, aspects of the present invention may take the form of a computer program product embodied in one or more computer-readable medium(s) having computer-executable code embodied therein.

[0041] Any combination of one or more computer-readable media may be used. The computer-readable medium may be a computer-readable signal medium or a computer-readable storage medium. As used herein, "computer-readable storage medium" encompasses any tangible storage medium that stores instructions executable by a processor or computing system of a computing device. The computer-readable storage medium may also be referred to as a computer-readable non-transitory storage medium. The computer-readable storage medium may also be referred to as a tangible computer-readable medium. In some embodiments, the computer-readable storage medium may also be capable of storing data that can be accessed by the 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 hard disks, flash memory, USB thumb drives, random access memory (RAM), read-only memory (ROM), optical disks, magneto-optical disks, and register files of a computing system. Examples of optical disks include compact discs (CDs) and digital versatile discs (DVDs), such as CD-ROM, CD-RW, CD-R, DVD-ROM, DVD-RW, or DVD-R discs. The term computer-readable storage medium also refers to various types of storage media that can be accessed by a computing device over a network or communications link. For example, data may be retrieved through a modem, the Internet, or a local area network. Computer-executable code embodied on a computer-readable medium may be transmitted using any suitable medium, including but not limited to wireless, wired, fiber optic cable, RF, etc., or any suitable combination of the above.

[0042] A computer-readable signal medium may include a propagated data signal in which computer-executable code is embodied, for example, in baseband or as part of a carrier wave. Such a propagated signal may take any of a 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, but is capable of communicating, propagating, or transporting a program for use by or in connection 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 directly accessible to a computing system. "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, computer storage may also be computer memory, or vice versa.

[0044] As used herein, a "computing system" encompasses electronic components capable of executing programs, machine-executable instructions, or computer-executable code. References to a computing system, including examples of a "computing system," should be interpreted as including more than one computing system or processing core, as the case may be. A computing system is, for example, a multi-core processor. A computing system also refers to a collection of computing systems within a single computer system or distributed among multiple computer systems. The term computing system should also be interpreted as referring to a collection or network of computing devices, each comprising a processor or computing system. Machine-executable code or instructions are executed by multiple computing systems or processors within the same computing device or even distributed among multiple computing devices.

[0045] Machine-executable instructions or computer-executable code 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, 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, and may be compiled into machine-executable instructions. In some cases, the computer-executable code may be in the form of a high-level language or in pre-compiled form, and may be used with an interpreter that generates machine-executable instructions on the fly. In other cases, the machine-executable instructions or computer-executable code may be in the form of programming for a programmable logic gate array.

[0046] The computer executable code may run 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 scenario, the remote computer may be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or the connection may be to an external computer (e.g., through the Internet using an Internet Service Provider).

[0047] Aspects of the present invention will be described with reference to flowcharts, illustrations, and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowcharts, illustrations, and / or block diagrams, or a portion of the blocks, where applicable, can be implemented by computer program instructions in the form of computer-executable code. It will further be understood that combinations of blocks in various flowcharts, illustrations, and / or block diagrams may be combined, where not mutually exclusive. These computer program instructions may be provided to a general-purpose computer, a special-purpose computer, or other programmable data processing device computing system to create a machine, such that the instructions, executing via the computer or other programmable data processing device computing system, cause the machine to perform the function(s) / act(s) specified in one or more blocks of the flowcharts and / or block diagrams.

[0048] These machine-executable instructions or computer program instructions may also be stored on a computer-readable medium that can instruct a computer, other programmable data processing apparatus, or other device to function in a particular manner, such that the instructions stored on the computer-readable medium create an article of manufacture that includes instructions that perform the functions / acts specified in one or more blocks of the flowcharts and / or 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 to create a computer-implemented process, such that the instructions, executing on the computer or other programmable apparatus, provide a process for performing the functions / acts specified in one or more blocks of the flowcharts and / or block diagrams.

[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" is sometimes referred to as a "human interface device." A user interface provides information or data to an operator and / or receives information or data from an operator. A user interface may allow input from an operator to be received by a computer and may provide output from the computer to a user. In other words, a user interface may allow an operator to control or manipulate a computer or may allow a computer to show the results of the operator's control or manipulation. The display of data or information on a display or graphical user interface is an example of providing information to an operator. Receiving data via a keyboard, mouse, trackball, touchpad, wand, graphics tablet, joystick, gamepad, webcam, headset, pedals, wired gloves, remote control, and accelerometer are all examples of user interface components that allow information or data to be received from an operator.

[0051] As used herein, a "hardware interface" encompasses an interface that allows a computing system of a computer system to interact with and / or control external computing devices and / or devices. A hardware interface allows a computing system to send control signals or commands to external computing devices and / or devices. A hardware interface also allows a computing system to exchange data with external computing devices and / or devices. Examples of hardware interfaces include, but are not limited to, a universal serial bus, an IEEE 1394 port, a parallel port, an IEEE 1284 port, a serial port, an RS-232 port, an IEEE 488 port, a Bluetooth® connection, a wireless local area network connection, a TCP / IP connection, an Ethernet connection, a control voltage interface, a MIDI interface, an analog input interface, and a digital input interface.

[0052] As used herein, a "display" or "display device" encompasses an output device or user interface adapted to display images or data. A display outputs visual, audio, and / or tactile data. Examples of displays include, but are not limited to, computer monitors, television screens, touch screens, tactile electronic displays, Braille screens, cathode ray tubes (CRTs), storage tubes, bi-stable 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-mounted displays.

[0053] K-space data is defined herein as the recorded measurements of radio frequency signals emitted by atomic spins using the antenna of a magnetic resonance machine during a magnetic resonance imaging scan. Magnetic resonance data is an example of tomographic medical image data.

[0054] Magnetic resonance imaging (MRI) image data or magnetic resonance (MR) image is defined herein as a reconstructed two-dimensional or three-dimensional visualization of the anatomical data contained within the k-space data, which visualization can be performed using a computer.

[0055] Preferred embodiments of the invention will now be described, by way of example only, with reference to the drawings in which: [Brief explanation of the drawings]

[0056] [Figure 1] An example of a medical system is shown below. [Figure 2] 2 shows a flowchart illustrating a method of using the medical system of FIG. 1. [Figure 3] A further example from a healthcare system is shown. [Figure 4] 4 shows a flowchart illustrating a method of using the medical system of FIG. 3. [Figure 5] 1 shows a flowchart illustrating a method for training a convolutional neural network. [Figure 6] An example of a convolutional neural network is shown below. [Figure 7] We demonstrate the effectiveness of convolutional neural networks in providing desired excitation fields. [Figure 8] Indicate the desired field of view. [Figure 9] The view obtained from using a convolutional neural network and the view from using an LSR algorithm are shown. DETAILED DESCRIPTION OF THE INVENTION

[0057] Like numbered elements in the figures are either equivalent elements or perform the same function. An element discussed previously is not necessarily discussed in a later figure if the function is equivalent.

[0058] FIG. 1 shows a diagram illustrating an example of a medical system 100. The medical system 100 in FIG. 1 is shown as including a computer 102 that includes a computing system 104. The computing system 104 is intended to represent one or more processing cores or computing systems, which may be co-located 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, for example, allows the computing system 104 to exchange data with and control other components of the medical system 100. The user interface 108 allows an operator of the medical system 100 to control and interact with the medical system. The user interface 108 may, for example, include a graphical user interface or other graphical control device.

[0059] The medical system 100 is further shown as including a memory 110 in communication with the computing system 104. The memory 110 is intended to represent any combination of memory accessible by the computing system 104.

[0060] The memory 110 is shown as including machine-executable instructions 120, which are instructions that enable the computing system to perform tasks such as controlling other components and performing various numerical and image processing tasks. The memory 110 is further shown as including a convolutional neural network 122, which is a multitasking convolutional neural network having a first output for outputting a radio frequency waveform and a separate output for each of a plurality of spatially selective gradient pulse waveforms. The convolutional neural network outputs these in response to receiving a composite array encoding the selection of at least one excited field of view for a magnetic resonance imaging acquisition.

[0061] The memory 110 is further shown as including at least one excitation field selection 124. The memory is further shown as including initial pulse sequence commands 126. These initial pulse sequence commands may be, for example, a set of pulse sequence commands that are retrieved or selected from a database when a particular protocol is desired to be performed.

[0062] The memory 110 is further shown to include a composite array used to encode at least one excitation field selection 124. The memory is further shown to include 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 inputting the composite array 128. The memory 110 is further shown to include a modified pulse sequence command 134. The modified pulse sequence command 134 is constructed from the initial pulse sequence command 126 and modified to include the radio frequency waveform 130 and the plurality of spatially selective gradient pulse waveforms 132.

[0063] 2 shows a flowchart illustrating a method of operating the medical system 100 of FIG. 1. First, in step 200, a selection 124 of at least one excitation field is received. Next, in step 202, an initial pulse sequence command 126 is received. Next, in step 204, a composite array 128 is encoded using the at least one excitation field 124. In step 206, a radio frequency waveform 130 and a plurality of spatially selective gradient pulse waveforms 132 are received in response to inputting the composite array 128 into the convolutional neural network 122. Finally, in step 208, a modified pulse sequence command is constructed by integrating the radio frequency waveform 130 and the plurality of spatially selective gradient pulse waveforms 132 with the initial pulse sequence command 126.

[0064] 3 illustrates a further example medical system 300. The medical system 300 is similar to the medical system 100 in FIG.

[0065] The magnetic resonance imaging system 302 includes a magnet 304. The magnet 304 is a cylindrical superconducting magnet with a bore 306 extending therethrough. Various types of magnets can be used, including both segmented cylindrical magnets and so-called open magnets. A segmented cylindrical magnet is similar to a standard cylindrical magnet except that a cryostat is divided into two sections to allow access to the magnet's isosurface. Such magnets are used, for example, in conjunction with charged particle beam therapy. An open magnet has two magnet sections with a space large enough to accommodate a subject between the upper and lower magnet sections, and the arrangement of these two sections is similar to that of a Helmholtz coil. Open magnets are common because the subject is not confined. Inside the cryostat of the cylindrical magnet is a collection of superconducting coils.

[0066] Within the bore 306 of the cylindrical magnet 304 is an imaging zone 308 where the magnetic field is sufficiently strong and uniform to perform magnetic resonance imaging. The magnetic resonance data acquired is typically acquired over a region of interest. A subject 318 is shown as being supported by a subject support 320 such that at least a portion of the subject 318 is within the imaging zone 308.

[0067] Also within the magnet bore 306 are a set of magnetic field gradient coils 310, which are used for preliminary magnetic resonance data acquisition for spatially encoding magnetic spins within the imaging zone 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 exemplary. Typically, the magnetic field gradient coils 310 include three separate coil sets for spatial encoding in three orthogonal spatial directions. The magnetic field gradient power supply supplies current to the magnetic field gradient coils 310. The current supplied to the magnetic field gradient coils 310 is controlled as a function of time and may be ramped or pulsed.

[0068] Adjacent to the imaging zone 308 is a radio frequency coil 314 that manipulates the orientation of magnetic spins within the imaging zone 308 and also receives radio transmissions from the spins within the imaging zone 308. A radio frequency antenna includes multiple coil elements. A radio frequency antenna is sometimes referred to as a channel or antenna. The radio frequency coil 314 is connected to a radio frequency transceiver 316. The radio frequency coil 314 and the radio frequency transceiver 316 may be replaced with separate transmit and receive coils and separate transmitters and receivers. It is understood that the radio frequency coil 314 and the radio frequency transceiver 316 are exemplary. The radio frequency coil 314 is also intended to represent a dedicated transmit antenna and a dedicated receive antenna. Similarly, the transceiver 316 may also represent separate transmitters and receivers. The radio frequency coil 314 may also have multiple receive / transmit elements, and the radio frequency transceiver 316 may have multiple receive / transmit channels. For example, if a parallel imaging technique such as SENSE is performed, the radio frequency coil 314 has multiple coil elements.

[0069] The transceiver 316 and the gradient controller 312 are shown as being connected to the hardware interface 106 of the computer system 102 .

[0070] Within the imaging zone 308 is an anatomical structure 322 of the subject 318. It can be seen that there is an excitation field of view 324 that has been modified to closely fit the anatomical structure 322. The excitation field of view 324 is specified by at least one excitation field selection 124 and encoded into the composite array 128.

[0071] Memory 110 is further shown as including k-space data 330 acquired by magnetic resonance imaging system 302 by controlling magnetic resonance imaging system 302 with modified pulse sequence commands 134. K-space data 330 describes anatomical structure 322. Memory 110 is further shown as including magnetic resonance imaging data 332 reconstructed from k-space data 330.

[0072] 4 shows a flowchart illustrating a method for operating the medical system 300 of FIG. 3. The method begins with steps 200-208 as shown in FIG. 2. After performing step 208, step 400 is performed, followed by step 402. In step 400, k-space data 330 is acquired by controlling the magnetic resonance imaging system 302 with the modified pulse sequence command 134. Finally, in step 402, magnetic resonance imaging data 332 is reconstructed from the k-space data 330.

[0073] FIG. 5 shows a flowchart illustrating a method for training convolutional neural network 122. Steps 500, 502, 504, and 506 are shown as occurring in a loop or repeated multiple times. However, these steps may be performed in parallel. That is, all of the training data may be trained by performing step 504, then step 506, multiple times, followed by step 500, then multiple times by step 502. First, in step 500, a selective excitation pulse design algorithm is used to generate training radio frequency waveforms and multiple training spatially selective gradient pulse waveforms. Next, in step 502, a model excitation field of view is calculated by inputting the training radio frequency waveforms and training spatially selective gradient pulse waveforms into a magnetic resonance imaging signal model.

[0074] Thereafter, in step 504, the model excitation field is input to the convolutional neural network to receive a forward propagated high frequency waveform and a plurality of forward propagated training spatially selective gradient pulse waveforms. Then, in step 506, the convolutional neural network is trained by updating the parameters of the convolutional neural network by performing backpropagation. Backpropagation is performed with the training high frequency waveform and the forward propagated high frequency waveform. Backpropagation is also performed with matched pairs of a plurality of training spatially selective gradient pulse waveforms and a plurality of forward propagated training spatially selective gradient pulse waveforms.

[0075] Modern MRI systems typically apply preset RF and associated gradients during clinical scans, with minimal adaptation to each scan's specific requirements. An example is provided for a multitasking, multiscale CNN method for real-time design of excitation RF pulses and associated gradient waveforms to achieve spatial 2D selectivity. The CNN-designed RF and gradients were highly similar to their SLR-designed counterparts, with an NRMSE of 0.0075 ± 0.0038 for 400 test data sets. Phantom imaging using the predicted RF was also similar to the SLR-designed excitation. The algorithm executes within 500 ms on a commercially available workstation.

[0076] Although there are occasional requirements, such as self-refocusing and adiabaticity, the most common requirement for a two-dimensional excitation pulse is an excitation profile in both dimensions. RF pulse design therefore poses an inverse problem: given the desired spatial and frequency profile, the small-angle approximation works for small excitation angles, and the Shinnar-Le Roux (SLR) algorithm has been proven to handle any flip angle between 0° and 180°. For modern MRI machines, conventional RF pulses and accompanying gradients are typically designed at system release time using one of these two methods, subject to hardware constraints including the maximum allowable B1 field, gradient field strength, and gradient field slew rate, and the resulting waveforms are stored in the system. For each specific scan, portions of these waveforms are loaded into the spectrometer, scaled in magnitude or extended in time.

[0077] While the SLR method is practical and allows for conventional imaging sequences to be acquired, it limits its range of applications and precludes any acquisition where highly specific signal excitation is beneficial. For example, spatial two-dimensional excitation has been clinically employed for diffusion imaging of the spine, breast, pancreas, prostate, and cervix to limit the phase-encoding field of view (FOV) for improved resolution and reduced geometric distortion. Other than using the SLR algorithm to generate simple two-dimensional excitations, RF and gradient waveforms with completely arbitrary shapes are possible, optimizing the balance between RF performance and duration. However, designing these arbitrary RF and gradient waveforms requires iterative algorithms, which are time-consuming and preclude waveform generation in real time.

[0078] In the example, a neural network-based 2D selective RF design is disclosed in which both RF and gradient waveforms are generated by a multitasking convolutional neural network in less than one second. This method is performed during each specific scan for optimized RF performance under given imaging conditions, including hardware constraints (RF and gradient limitations and number of transmit channels), system imperfections (eddy currents and B0 inhomogeneity), imaging object-dependent challenges (local magnetic field inhomogeneity for both B0 and B1), and prescription-dependent requirements (FOV of interest). The variable velocity selective excitation (VERSE) method has occasionally been applied clinically, and its iterative design algorithm has recently been simplified, but it only addresses some of the issues listed above. Presented here are pilot results from this comprehensive real-time RF design project, which demonstrate the utility of DL as a pathway for real-time optimized arbitrary excitation pulses.

[0079] FIG. 6 shows an example topology or structure for a convolutional neural network 122. This neural network 122 receives an input array 128. There is a first layer 600 containing a convolution kernel, a ReLU, and max pooling. This first layer then feeds into a second layer containing another convolution kernel, ReLU, and max pooling. This second layer then feeds into a third convolution kernel or layer 604, which also has a ReLU and max pooling layer. This third layer then feeds into a smoothing layer 606, which is then coupled to three separate fully connected layers 608, 608′, and 608″. Layer 608 outputs a high-frequency waveform 130. Only the coupled layers 608, 608′ output two spatially selective gradient pulse waveforms 132. The neural network 122 shown in FIG. 6 is used for two-dimensional slices. The basic structure can be reused and modified for three-dimensional excitation fields. For example, additional connection layers can be added for a third spatially selective gradient pulse waveform, and the input can be modified to accept complex three-dimensional arrays. The exact structure of the various convolutional layers is not essential. However, the use of a separate connection layer coupled to a flattening layer 606 has shown superior results to other types of neural network topologies.

[0080] An exemplary multi-task CNN (hereinafter referred to as CNN) is constructed based on a convolutional framework. As shown in FIG. 6 above, the CNN consists of an input layer, three convolution 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 components of the desired excitation profile, and thus a three-dimensional matrix (e.g., 101×101×2) with a three-dimensional size of 2. The first convolution block 600 includes 16 convolution kernels and a rectified linear unit (ReLU), followed by a max-pooling step. The second and third convolution blocks 602 and 604 include 32 and 64 convolution kernels, respectively, followed by the same ReLU activation and max-pooling steps. A flattening layer 606 was then configured to convert the data into 9261 vectors, which were concatenated by a fully connected layer to produce a higher-level inference with an output of adapted dimensions. The network's convergence was assessed by the minimum loss, measured as the root mean square error (RMSE). The network's parameters were adjusted until an optimal network configuration was achieved.

[0081] We generated excitation profiles by treating the RF, GRS, and GRF waveforms as three channels of RF pulses, and used a multi-task learning technique in which a flattening layer that preserves shared features is connected in parallel with three fully connected layers. Furthermore, we adjusted the convolution kernel to multiply the sizes of 1x1, 3x3, 5x5, and 7x7, similar to the countermeasures in the Inception module. Convolution with multiple kernel sizes captures feature information at various scales compared to conventional convolution operations, thereby avoiding the potential overfitting problem when deeper networks are used.

[0082] Examples include one or more of the following features. 1) A sufficiently large set of 2D RF pulses for training a CNN was generated using the conventional SLR algorithm. 2) The excitation profile of each pulse was generated by Bloch simulation to train the CNN. 3) Custom convolutional neural network architecture with optimized layers and connections 4) The network was trained with the generated excitation profiles as input and the corresponding RF pulses as output. 5) The desired excitation profile was input to the trained network to obtain the CNN design RF.

[0083] 7 illustrates the effectiveness of a radio frequency waveform 130 and multiple spatially selective gradient pulse waveforms 132 provided by a convolutional neural network 122. In this figure, an exemplary radio frequency waveform 130 and multiple spatially selective gradient pulse waveforms 132 are shown at the top. These waveforms 130, 132 were used to provide an excitation profile 700 in the center, which closely resembles the ground truth excitation field 702 shown in the box below.

[0084] Furthermore, the predicted RF can indeed provide an excitation profile very similar to the desired FOV excitation profile, which was tested by phantom imaging on a clinical MR 3.0T scanner.

[0085] The effectiveness of the convolutional neural network is further demonstrated using Figures 8 and 9. Figure 8 shows the desired excitation field 800. In Figure 9, 900 shows the excitation field using the convolutional neural network. On the right, box 902 shows the excitation field with the LSR-designed RF pulse. It can be seen that the convolutional neural network compares very well with the LSR-designed RF pulse.

[0086] Examples include one or more of the following features. 1) CNN-based RF design for both RF and gradient waveforms. 2) Using multitasking techniques with a shared layer to combine the RF waveform and the two gradients. 3) Multi-scale kernels in CNN are used to capture details at different scales in the input FOV image domain. 4) Echo-planar trajectories were selected for RF generation of the training dataset. This technique may optionally be discovered by experimentation of scanning and acquiring images.

[0087] There are applications for spatial 2D or 3D excitation in imaging of the brain, spinal cord, head and neck tumors, breast, pancreas, and prostate, allowing for improved spatial resolution and reduced geometric distortion for diffusion-weighted imaging.

[0088] While the invention has been illustrated and described in detail in the drawings and foregoing description, such illustration and description are to be considered illustrative or exemplary and not restrictive, and the invention is not limited to the disclosed embodiments.

[0089] Other variations to the disclosed embodiments can be understood and effected by those skilled in the art in practicing the claimed invention, from a study of the drawings, the disclosure, and the appended claims. In the claims, the words "comprise," "include," "have," "includes," "included," "having," "comprises," "included ... [Explanation of symbols]

[0090] 100 Healthcare Systems 102 Computer 104 Computing Systems 106 Optional Hardware Interface 108 Optional User Interface 110 memory 120 machine-executable instructions 122 Convolutional Neural Networks 124 Select at least one excitation field 126 Initial Pulse Sequence Command 128 composite arrays 130 High Frequency Waveforms 132 Multiple spatially selective gradient pulse waveforms 134 Modified Pulse Sequence Commands 200 Receive the selection of at least one excitation field 202 receives an initial pulse sequence command configured to control a magnetic resonance imaging system to acquire k-space data describing the subject. 204 Encoding a composite array with at least one excitation field 206 receiving a radio frequency waveform and a plurality of spatially selective gradient pulse waveforms in response to inputting the composite array into a convolutional neural network. 208. Configuring a modified pulse sequence command by modifying the initial pulse sequence command with a radio frequency waveform and a plurality of 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 excited field of view. 300 Healthcare Systems 302 Magnetic Resonance Imaging System 304 Magnet 306 Magnet Bore 308 Imaging Zone 310 Magnetic Gradient Coil 312 Magnetic field gradient coil power supply 314 Radio Frequency Coil 316 Transceiver 318 Target 320 Subject Support Platform 322 Anatomical Structure 324 Excitation Field 330 k-space data 332 Magnetic Resonance Imaging Data 400 Acquire k-space data by controlling the magnetic resonance imaging system with modified pulse sequence commands 402 Reconstructing Magnetic Resonance Imaging Data from k-Space Data 500. Generate a training radio frequency waveform and multiple training spatially selective gradient pulse waveforms using a selective excitation pulse design algorithm. 502 Calculate the model excitation field of view by inputting the training radio frequency waveform and a plurality of training spatially selective gradient pulse waveforms into the magnetic resonance imaging signal model. 504. Inputting the model excitation field into a convolutional neural network to receive a forward propagated high frequency waveform and a plurality of forward propagated training spatially selective gradient pulse waveforms. 506. Updating parameters of a convolutional neural network by backpropagating the training high frequency waveform and the forward propagated high frequency waveform and matched pairs of a plurality of training spatially selective gradient pulse waveforms and a plurality of forward propagated training spatially selective gradient pulse waveforms. Convolution kernel with 600 ReLU layers and max pooling layers 602 Convolution kernel with ReLU and max pooling layers 604 Convolution kernel with ReLU and max pooling layers 606 Planarization layer 608 Fully connected layer 608' fully connected layer 608'' fully connected layer 700 generated excitation fields 702 Ground Truth Excitation Field of View 800 Desired excitation field of view 900 Excitation field of view by CNN 902 Excitation field using LSR designed RF pulse

Claims

1. a memory storing machine executable instructions and a convolutional neural network, the convolutional neural network receiving as an input a composite array encoding at least one selection of an excitation field of view, the convolutional neural network outputting a radio frequency waveform and a plurality of spatially selective gradient pulse waveforms, the convolutional neural network being a multitasking 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; a computing system, Execution of the machine-executable instructions causes the computing system to: receiving at least one selection of an excitation field; receiving an initial pulse sequence command for controlling a magnetic resonance imaging system to acquire k-space data describing the subject; encoding the composite array with the at least one excitation field; receiving the radio frequency waveform and the plurality of spatially selective gradient pulse waveforms in response to inputting the composite array into the convolutional neural network; and and configuring a modified pulse sequence command by modifying the initial pulse sequence command with the radio frequency waveform and the plurality of spatially selective gradient pulse waveforms such that the pulse sequence command controls the magnetic resonance imaging system to acquire the k-space data from the at least one excited field of view.

2. The convolutional neural network includes: generating a training radio frequency waveform and a plurality of training spatially selective gradient pulse waveforms using a selective excitation pulse design algorithm; calculating a 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; receiving a forward propagated high frequency waveform and a plurality of forward propagated training spatially selective gradient pulse waveforms by inputting the model excitation field into the convolutional neural network; and updating parameters of the convolutional neural network by backpropagating the training high frequency waveforms and the forward propagated high frequency waveforms and matched pairs of the plurality of training spatially selective gradient pulse waveforms and the plurality of forward propagated training spatially selective gradient pulse waveforms.

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

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

5. The medical system of claim 1 , wherein the pulse sequence commands acquire the k-space data according to a parallel imaging magnetic resonance imaging protocol.

6. Execution of the machine-executable instructions further causes the computing system to: Receive investigative magnetic resonance images; Rendering the survey magnetic resonance image on a display; The medical system of claim 1 , further comprising: a display unit configured to display a survey magnetic resonance image of a subject, the display unit configured to display a survey magnetic resonance image of the subject;

7. The medical system of claim 1 , wherein the selection of the at least one field of view is received from any one of an automatic image segmentation algorithm, a user interface, and combinations thereof.

8. 8. The medical system of claim 1, wherein the plurality of spatially selective gradient pulse waveforms are two spatially selective gradient pulse waveforms and each of the at least one excitation field is a two-dimensional excitation field.

9. 8. The medical system of claim 1, wherein the plurality of spatially selective gradient pulse waveforms are three partially selective gradient pulse waveforms and each of the at least one excitation field is a three-dimensional excitation field.

10. The medical system of claim 1 , wherein each of the at least one excitation field of view has an arbitrary shape and / or is spatially separated.

11. 11. The medical system of claim 1, wherein the convolutional neural network is one of a multi-scale convolutional neural network and a U-Net convolutional neural network.

12. The medical system includes the magnetic resonance imaging system, and execution of the machine-executable instructions further causes the computing system to: acquiring the k-space data by controlling the magnetic resonance imaging system with the modified pulse sequence commands; The medical system of claim 1 , further comprising: a first section for reconstructing magnetic resonance imaging data from the k-space data;

13. 1. A computer program comprising machine executable instructions for execution by a computing system for controlling a medical system, the computer program further comprising a convolutional neural network receiving as an input a composite array encoding at least one excitation field of view selection, the convolutional neural network outputting a radio frequency waveform and a plurality of spatially selective gradient pulse waveforms, the convolutional neural network being a multitasking 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; Execution of the machine-executable instructions causes the computing system to: receiving the selection of the at least one excitation field; receiving an initial pulse sequence command for controlling a magnetic resonance imaging system to acquire k-space data describing the subject; encoding the composite array with the at least one excitation field; receiving the radio frequency waveform and the plurality of spatially selective gradient pulse waveforms in response to inputting the composite array into the convolutional neural network; constructing a modified pulse sequence command by modifying the initial pulse sequence command with the radio frequency waveform and the plurality of spatially selective gradient pulse waveforms such that the pulse sequence command controls the magnetic resonance imaging system to acquire the k-space data from the at least one excited field of view.

14. 1. A method of medical imaging using a convolutional neural network, the convolutional neural network receiving as an input a composite array encoding at least one excitation field of view selection, the convolutional neural network outputting a radio frequency waveform and a plurality of spatially selective gradient pulse waveforms, the convolutional neural network being a multitasking 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 comprises: receiving the selection of the at least one excitation field; receiving an initial pulse sequence command to control a magnetic resonance imaging system to acquire k-space data describing the subject; encoding the composite array with the at least one excitation field; receiving the radio frequency waveform and the plurality of spatially selective gradient pulse waveforms in response to inputting the composite array into the convolutional neural network; and constructing a modified pulse sequence command by modifying the initial pulse sequence command with the radio frequency waveform and the plurality of spatially selective gradient pulse waveforms such that the pulse sequence command controls the magnetic resonance imaging system to acquire the k-space data from the at least one excitation field of view.

15. 1. A method for training a convolutional neural network, the method comprising: generating a training radio frequency waveform and a plurality of training spatially selective gradient pulse waveforms using a selective excitation pulse design algorithm; calculating a 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; receiving a forward propagated high frequency waveform and a plurality of forward propagated training spatially selective gradient pulse waveforms by inputting the model excitation field into the convolutional neural network; updating parameters of the convolutional neural network by backpropagating the training high frequency waveforms and the forward propagated high frequency waveforms and matched pairs of the plurality of the training spatially selective gradient pulse waveforms and the plurality of the forward propagated training spatially selective gradient pulse waveforms.