Automatic adjustment of under-sampling factor

Abstract

A method of operating a medical system (100, 300) is disclosed herein. The method includes receiving (200) a pulse sequence command (124) configured to control a magnetic resonance imaging system (302) to acquire k-space data (330) according to a compressed sensing magnetic resonance imaging protocol. The method also includes receiving (202) magnetic resonance scan parameters describing a configuration of the pulse sequence command and a configuration of the magnetic resonance imaging system. The method also includes receiving (204) a predicted undersampling factor (128) in response to inputting the magnetic resonance scan parameters into a neural network, wherein the neural network is configured to output the predicted undersampling factor in response to receiving magnetic resonance scan parameters. The method also includes adjusting (206) the pulse sequence command (130) to select or modify sampling of the k-space data based on the predicted undersampling factor.

Description

Technical Field

[0001] This invention relates to magnetic resonance imaging, and more particularly to compression sensing magnetic resonance imaging. Background Technology

[0002] Magnetic resonance imaging (MRI) scanners use a large, static magnetic field to align the nuclear spins of atoms as part of the process of generating images of a patient's body. This large, static magnetic field is called the B0 field or main magnetic field. MRI systems sample data in k-space and then reconstruct MRI images from that k-space data.

[0003] MRI can be used to measure various quantities or properties of an object in space. A drawback of MRI is the time required to acquire k-space data. During k-space data acquisition, it can be difficult to keep the object stationary. Compressed sensing MRI reduces acquisition time by reconstructing MRI images using undersampled k-space data. Currently, the operator selects an undersampling factor (and is also known to select an acceleration factor), which affects how the k-space data is sampled. If too little k-space data is acquired, acquisitions need to be repeated with fewer undersampled samples.

[0004] U.S. Patent Application Publication US 2018 / 0203081 discloses a system and method for estimating quantitative parameters of an object using a dictionary-based magnetic resonance (“MR”) system. The dictionary may include multiple signal templates that sparsely sample the acquisition parameters used during data acquisition. A neural network is used to compare the acquired data with the dictionary. Therefore, a more computationally efficient system and method are provided compared to conventional MRF reconstruction systems and methods, and data storage requirements are reduced. U.S. Patent Application US 2015 / 108978 relates to strategies for sparse sampling in magnetic resonance imaging. More specifically, such known strategies involve selecting a basic variable density sampling pattern. This basic variable density sampling pattern is selected based on criteria provided by the user. Subsequently, the scan time for the basic variable density sampling pattern is determined by simulation analysis or according to a lookup table. To address unacceptable scan times, the variable density sampling pattern is modified to maximize the sampled k-space area without increasing the scan time. Summary of the Invention

[0005] The present invention provides, in one aspect, medical systems, computer programs, and methods. In another aspect, embodiments are provided.

[0006] The embodiments provide an improved means of selecting a predicted undersampling factor. A neural network is configured or trained to output a predicted undersampling factor in response to received magnetic resonance scan parameters. The magnetic resonance scan parameters describe the configuration of the magnetic resonance imaging system, including the configuration of pulse sequence commands for controlling the system. The predicted undersampling factor represents a prediction of the appropriate undersampling value before acquiring the magnetic resonance signal, i.e., predicting or estimating the undersampling before scanning the k-space. Therefore, by scanning the k-space according to a sampling pattern and sampling density function consistent with the prior predicted undersampling factor, the predicted undersampling can be made available at the start of sampling the magnetic resonance signal. Before MR data acquisition via sampling the k-space, the trained neural network makes the predicted undersampling factor available by returning it to the input scan parameters. The neural network can be trained based on historical data of successful image acquisitions associated with combinations of appropriate undersampling factors and scan parameters (a set).

[0007] In one aspect, the present invention provides a medical system including a memory storing machine-executable instructions. The memory also stores a neural network. The neural network is configured to output a predicted undersampling factor in response to receiving magnetic resonance imaging (MRI) scan parameters. The MRI scan parameters used herein include the configuration of an MRI system and / or the configuration of pulse sequence commands for controlling the MRI system.

[0008] Individual settings or adjustments in the pulse sequence commands, as well as the configuration of the MRI system, can affect the predicted undersampling factor. The undersampling factor is the undersampling factor when executing a compressed-sensing MRI protocol. MRI scan parameters describe the configuration of the MRI system. This configuration of the MRI system also includes the configuration of the pulse sequence commands.

[0009] The medical system also includes a computing system configured to control the magnetic resonance imaging (MRI) system. Execution of the machine-executable instructions causes the computing system to receive pulse sequence commands configured to control the MRI system to acquire k-space data according to a compressed-sensing MRI protocol. Execution of the machine-executable instructions also causes the computing system to receive MRI scan parameters. The pulse sequence commands and MRI scan parameters can be received in various ways. In some instances, MRI scan parameters and pulse sequence commands with specific configurations for the pulse sequence commands can be received from a user interface. In other cases, they can be received by retrieving the pulse sequence commands and MRI scan parameters from memory.

[0010] The execution of the machine-executable instructions also causes the computing system to receive the predicted undersampling factor in response to inputting the magnetic resonance scan parameters into the neural network. The execution of the machine-executable instructions also causes the computing system to adjust the pulse sequence command based on the undersampling factor to modify the sampling of the k-space data. For example, when a magnetic resonance imaging system acquires k-space data, it acquires it in the form of k-space data sets, which are single lines or commonly referred to as excitations. The adjustment of the pulse sequence command modifies how the k-space data is sampled to match the predicted undersampling factor.

[0011] The undersampling factor is a factor pre-undersampled relative to the Nyquist theorem. This implementation can be advantageous because it provides an improved means of setting the undersampling factor. If the undersampling factor is not reduced sufficiently, it will not have a detrimental effect on the magnetic resonance image. However, it will take longer to acquire k-space data compared to the optimal undersampling factor setting. If the undersampling factor is too low, the resulting magnetic resonance image may be corrupted. Using neural networks allows for setting the undersampling factor using a wider variety of factors, including those that a human operator would not consider. Typically, the operator will manually adjust the undersampling factor. Humans can examine various factors and then adjust the undersampling factor. This is usually constructed based on the operator's experience and is often a stochastic process.

[0012] In another embodiment, the magnetic resonance scanning parameters include the radio frequency coil configuration. This may include the number and arrangement of the radio frequency coils.

[0013] In another embodiment, the magnetic resonance scanning parameters include a scanning mode specifying whether a two-dimensional or three-dimensional scan is performed. This essentially identifies how k-space data is acquired—whether it is acquired in a three-dimensional manner or as two-dimensional slices.

[0014] In another embodiment, the magnetic resonance scanning parameters include a sequence type that specifies the contrast of the pulse sequence command. Various parameters within the pulse sequence command can be used to alter the image contrast.

[0015] In another embodiment, the magnetic resonance scanning parameters include echo time. This is a basic value that can be set in the pulse sequence command.

[0016] In another embodiment, the magnetic resonance scanning parameters include the pulse repetition time.

[0017] In another embodiment, the magnetic resonance scanning parameters include voxel size or three-dimensional spatial resolution.

[0018] In another embodiment, the magnetic resonance scanning parameters include a three-dimensional field of view.

[0019] Voxel size, or 3D spatial resolution, and the 3D field of view and voxel size, which together provide information about the field of view and matrix size, or voxel size and matrix size, are somewhat redundant. When configuring a magnetic resonance imaging system, many parameters do indeed have some overlap and redundancy.

[0020] In another embodiment, the magnetic resonance scanning parameters include the radio frequency bandwidth during k-space sampling.

[0021] The magnetic resonance scanning parameters mentioned above can include the core of the scanning parameters, which, when used to train a neural network, cause the generation of an accurate undersampling factor.

[0022] In another embodiment, the magnetic resonance scan parameters include the number of times the signal averaging is performed.

[0023] The magnetic resonance scanning parameters described below are those that can have an additional effect on estimating undersampling factors.

[0024] In another embodiment, the magnetic resonance scanning parameters also include the type of fat suppression protocol being used.

[0025] In another embodiment, the magnetic resonance scanning parameters also include the flip angle specified in the pulse sequence command.

[0026] In another embodiment, the magnetic resonance scanning parameters also include scanning time.

[0027] In another embodiment, the magnetic resonance scanning parameters also include the orientation of the field of view.

[0028] In another embodiment, the magnetic resonance scanning parameters also include the folding direction.

[0029] In another embodiment, the magnetic resonance scanning parameters also include the number of dynamic scans.

[0030] In another embodiment, the magnetic resonance imaging (MRI) scan parameters also include the type of contrast agent used. The type of contrast agent used can certainly be an important scan parameter when it is used in a particular MRI protocol. However, not all MRI protocols use contrast agents.

[0031] In another embodiment, the magnetic resonance scanning parameters also include the reconstruction voxel size or reconstruction matrix size.

[0032] In another embodiment, the magnetic resonance scanning parameters also include the type or selection of the prepulse used in the pulse sequence command.

[0033] In another embodiment, the magnetic resonance scanning parameters further include an implementation of a partial Fourier half-scan protocol or a selection of an implementation of a partial Fourier half-scan protocol.

[0034] In another embodiment, the magnetic resonance scan parameters also include the anatomical portion being examined. This could be, for example, a specific view and / or region of the body being examined.

[0035] In another embodiment, the magnetic resonance scanning parameters also include the type of excitation used. This excitation is a set of k-space data points acquired as a single acquisition.

[0036] In another embodiment, the magnetic resonance scanning parameters further include a k-space contour order.

[0037] In another embodiment, the magnetic resonance scanning parameters also include a k-space trajectory.

[0038] In another embodiment, the magnetic resonance scanning parameters also include the type of physiological synchronization. This could be, for example, synchronization with the cardiac phase or respiratory phase.

[0039] In another embodiment, the magnetic resonance scanning parameters also include a diffusion coding technique type.

[0040] In another embodiment, the magnetic resonance scanning parameters further include a k-space segmentation factor.

[0041] In another embodiment, the magnetic resonance scanning parameters also include the number of echoes used to acquire the same k-space lines.

[0042] In another embodiment, execution of the machine-executable instructions also causes the computing system to retrieve archived scan parameter data from a magnetic resonance scan parameter database. These parameters may include, for example, various parameters for pulse sequences used for various types of scans. This also includes undersampling factors. The method further includes a process of constructing archived training data based on the archived scan parameter data. This may, for example, involve extracting the values ​​of the undersampling factors and the magnetic resonance scan parameters used. The training data can then include the magnetic resonance scan parameters as input to a neural network, and the actual sampling factor used can then be compared with the output of the neural network. Execution of the machine-executable instructions also causes the computing system to train the neural network using the archived training data. This can be done, for example, using a backpropagation algorithm.

[0043] In another embodiment, the archived training data is received remotely.

[0044] In another embodiment, the archived training data is received remotely via a network connection. This allows, for example, the use of data from various locations and sites to train a neural network.

[0045] In another embodiment, the medical system further includes the magnetic resonance imaging system. The execution of the machine-executable instructions also causes the computing system to control the magnetic resonance imaging system to acquire the k-space data by utilizing the pulse sequence commands. The execution of the machine-executable instructions also causes the computing system to reconstruct magnetic resonance image data based on the k-space data. The magnetic resonance image data is data that can be plotted in two-dimensional or three-dimensional form to create a magnetic resonance image.

[0046] In another embodiment, the medical system further includes a user interface. Execution of the machine-executable instructions also causes the computing system to display the undersampling factor and at least a portion of the magnetic resonance scan parameters on the user interface before adjusting the pulse sequence command. Execution of the machine-executable instructions also causes the computing system to receive a predicted undersampling factor from the user interface in response to displaying the undersampling factor. The pulse sequence command is adjusted using the predicted undersampling factor. In this embodiment, the neural network still provides the undersampling factor, but the operator has the opportunity to use the user interface to correct or change the undersampling factor.

[0047] In another embodiment, execution of the machine-executable instructions further causes the computing system to construct user-specific training data based on the magnetic resonance scan parameters and the predicted undersampling factor. Execution of the machine-executable instructions also causes the computing system to train the neural network using the user-specific training data. The user-specific training data may, for example, include extracting the magnetic resonance scan parameters and the predicted undersampling factor, and then creating data that can be used for backpropagation to train the neural network. This can be advantageous, for example, because it can be used to train the neural network for local preferences and / or for the local pulse sequence commands or protocols used.

[0048] In another embodiment, the neural network is a multi-layer neural network. Experiments show that multi-layer neural networks can predict undersampling factors well during training.

[0049] In another embodiment, the multilayer neural network comprises at least six layers. Each of the at least six layers is fully connected to its adjacent layers. In the example described later, a seven-layer multilayer neural network is used to perform the prediction of the undersampling factor. A six-layer multilayer neural network will work correctly. A seven-layer multilayer neural network will perform even better.

[0050] In another aspect, the present invention provides a method for training a neural network. The method includes retrieving archived scan parameter data from a magnetic resonance imaging (MRI) scan parameter database. The method further includes: constructing archived training data based on the archived scan parameter data, and then using the archived training data to train the neural network. The training may be performed using a backpropagation algorithm. This method can be used to pre-train the neural network of the aforementioned medical system.

[0051] In another aspect, the present invention provides a method for operating a medical system. The method includes receiving a pulse sequence command configured to control a magnetic resonance imaging system to acquire k-space data according to a compressed induction magnetic resonance imaging protocol. The method further includes receiving magnetic resonance scanning parameters describing the configuration of the pulse sequence command and the configuration of the magnetic resonance imaging system.

[0052] The method further includes receiving a predicted undersampling factor in response to inputting the magnetic resonance scan parameters into a neural network. The neural network is configured to output the predicted undersampling factor in response to receiving the magnetic resonance scan parameters. The method also includes adjusting the pulse sequence command based on the predicted undersampling factor to modify the sampling mode of the k-space data.

[0053] In another aspect, the present invention provides a computer program including machine-executable instructions for execution by a computing system configured to control a medical system. The computer program may further include a neural network. Execution of the machine-executable instructions causes the computing system to receive pulse sequence commands configured to control a magnetic resonance imaging (MRI) system to acquire k-space data according to a compressed inductive magnetic resonance imaging (CIMI) protocol. Execution of the machine-executable instructions also causes the computing system to receive magnetic resonance scanning parameters describing the configuration of the pulse sequence commands and the configuration of the MRI system.

[0054] The execution of the machine-executable instructions also causes the computing system to receive a predicted undersampling factor in response to inputting the magnetic resonance scan parameters into a neural network. The neural network is configured to output the predicted undersampling factor in response to receiving the magnetic resonance scan parameters. The execution of the machine-executable instructions also causes the computing system to adjust the pulse sequence command based on the undersampling factor to modify the sampling or sampling mode of the k-space data.

[0055] It should be understood that one or more embodiments of the foregoing embodiments of the present invention may be combined, as long as the combined embodiments are not mutually exclusive.

[0056] Those skilled in the art will recognize that aspects of the present invention can be implemented as apparatus, method, or computer program product. Therefore, 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 collectively referred to herein as "circuit," "module," or "system." Furthermore, aspects of the present invention can take the form of a computer program product implemented on one or more computer-readable media having computer-executable code implemented thereon.

[0057] Any combination of one or more computer-readable media can be used. A computer-readable medium 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-transient 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 multi-purpose 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 computer devices via a network or communication link. For example, data can be retrieved on a modem, on the Internet, or on a local area network. Any suitable medium may be used to transmit computer-executable code implemented on a computer-readable medium, including but not limited to: wireless, wired, fiber optic cable, RF, etc., or any suitable combination thereof.

[0058] Computer-readable signal media may include, for example, propagated data signals in baseband or as a portion of a carrier wave, in which computer-executable code is implemented. Such propagated signals 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 and is capable of delivering, propagating, or transmitting a program for use by or in conjunction with an instruction execution system, apparatus, or device.

[0059] "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.

[0060] As used herein, the term "computing system" encompasses electronic components capable of running programs or machine-executable instructions or computer-executable code. References to computing systems that include the term "computing system" should be interpreted as potentially containing more than one computing system or processing core. A computing system may, for example, be a multi-core processor. A computing system may also refer 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 potentially referring to a collection or network of multiple computing devices, each of which includes a processor or computing system. Machine-executable code or instructions can be executed by multiple computing systems or processors that may be within the same computing device or even distributed across multiple computing devices.

[0061] Machine-executable instructions or computer-executable code may include instructions or programs that instruct a processor or other computing system to perform an aspect of the invention. Computer-executable code for performing operations toward the aspects of the invention may be written in any combination of one or more programming languages, including object-oriented programming languages ​​(e.g., Java, Smalltalk, C++, etc.) and conventional programming languages ​​(e.g., the "C" programming language or similar programming languages), and compiled into machine-executable instructions. In some instances, the computer-executable code may be in the form of a high-level language or in a pre-compiled form, and may be used in conjunction with an interpreter that generates the machine-executable instructions at runtime. In other instances, the machine-executable instructions or computer-executable code may be in the form of programming against a programmable gate array.

[0062] Computer executable code can 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 scenario, the remote computer can 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 it can be connected to an external computer (e.g., via the Internet provided by an Internet service provider).

[0063] Aspects of the invention have been described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It should be understood that each block or portion of the flowchart, illustration, and / or block diagram can be implemented by computer program instructions in the form of computer-executable code, where appropriate. It should also be understood that blocks in different flowcharts, illustrations, and / or block diagrams can be combined without mutual exclusion. These computer program instructions can be provided to a computing system of a general-purpose computer, special-purpose computer, or other programmable data processing apparatus to produce a machine, such that the instructions, executable via the computer or other programmable data processing apparatus, create units for implementing the functions / actions specified in one or more blocks of the flowchart and / or block diagram.

[0064] 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 function in a particular manner, such that the instructions stored in the computer-readable medium produce an article of manufacture, the article of manufacture including instructions that implement functions / actions specified in flowcharts and / or one or more block diagrams.

[0065] 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 creating a computer-implemented process, such that the instructions running on the computer or other programmable apparatus provide for performing the functions / actions specified in the flowchart and / or one or more block diagram boxes.

[0066] 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" may 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 can provide output from the computer to the user. In other words, a user interface allows an operator to control or manipulate a computer, and the interface allows the computer to indicate the effects of the operator's control or manipulation. Displaying data or information on a monitor or graphical user interface is an example of providing information to an operator. Receiving data via a keyboard, mouse, trackball, touchpad, pointing stick, graphics tablet, joystick, game controller, webcam, head-mounted device, foot pedal, wired gloves, remote control, and accelerometer are all examples of user interface components that enable the reception of information or data from an operator.

[0067] 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 a computing system to send control signals or commands to external computing devices and / or devices. A hardware interface also enables a 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.

[0068] As used herein, the term "display" or "display device" encompasses an output device or user interface suitable for displaying images or data. A display can output visual, auditory, and / or tactile data. Examples of displays include, but are not limited to: computer monitors, television screens, touchscreens, haptic 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.

[0069] k-space data are defined in this paper as measurements of radio frequency signals emitted via atomic spins, recorded by the antenna of a magnetic resonance imaging (MRI) device during a magnetic resonance imaging (MRI) scan. MRI data is an example of tomographic medical image data.

[0070] Magnetic resonance imaging (MRI) images or MR images or magnetic resonance imaging data are defined in this paper as two-dimensional or three-dimensional visualizations reconstructed from anatomical data contained within magnetic resonance imaging data. Such visualizations can be performed using a computer. Attached Figure Description

[0071] Preferred embodiments of the invention will now be described by way of example only with reference to the accompanying drawings, in which:

[0072] Figure 1 An example of a medical system is illustrated;

[0073] Figure 2 The illustrated operation is shown. Figure 1 A flowchart illustrating an example of a method for a medical system;

[0074] Figure 3 The illustration shows another example of a medical system;

[0075] Figure 4 The illustrated operation is shown. Figure 3 A flowchart illustrating an example of a method for a medical system;

[0076] Figure 5 The diagram illustrates the training of a neural network;

[0077] Figure 6 The diagram shows... Figure 5 The use of neural networks;

[0078] Figure 7 The illustration shows the integration of a neural network into a magnetic resonance imaging system;

[0079] Figure 8 A plot showing the test results for the neural network is displayed;

[0080] Figure 9 A pie chart showing additional test results for the neural network is displayed; and

[0081] Figure 10 A pie chart showing additional test results for the neural network is shown.

[0082] List of reference numerals in the attached diagram:

[0083] 100 Medical Systems

[0084] 102 Computer

[0085] 104 Hardware Interfaces

[0086] 106 Computing System

[0087] 108 User Interface

[0088] 110 Memory

[0089] 120 Machine-executable instructions

[0090] 122 Neural Networks

[0091] 124 Pulse Sequence Command

[0092] 126 Magnetic Resonance Scan Parameters

[0093] 128 Predicted undersampling factor

[0094] 130 Adjusted Pulse Sequence Command

[0095] 200 Receives a pulse sequence command, the pulse sequence command being configured to control the magnetic resonance imaging system to acquire k-space data according to a compressed sensing magnetic resonance imaging protocol.

[0096] 202 Receive Magnetic Resonance Scan Parameters

[0097] 204 Receives the predicted undersampling factor in response to inputting magnetic resonance scan parameters into the neural network.

[0098] 206 Adjusting the pulse sequence command based on the predicted undersampling factor to modify the sampling of k-space data

[0099] 300 Medical System

[0100] 302 Magnetic Resonance Imaging System

[0101] 304 magnet

[0102] 306 Magnet Chamber

[0103] 308 Imaging Area

[0104] 309 Areas of Interest

[0105] 310 Magnetic Gradient Coil

[0106] 312 Magnetic Gradient Coil Power Supply

[0107] 314 RF coil

[0108] 316 transceiver

[0109] 318 Objects

[0110] 320 Object support

[0111] 330 k spatial data

[0112] 332 Magnetic Resonance Imaging Data

[0113] 400 Acquiring k-space data by controlling the magnetic resonance imaging system using pulse sequence commands.

[0114] 402 Reconstructing Magnetic Resonance Imaging Data Based on k-space Data

[0115] 500 Input Layer

[0116] 502 Fully Connected Layer

[0117] 504 output Detailed Implementation

[0118] In these figures, elements with the same number are either equivalent elements or perform the same function. If the functions are equivalent, elements that have already been discussed need not be discussed again in the following figures.

[0119] Figure 1An example of a medical system 100 is illustrated. In this example, the medical system 100 includes a computer 102. The medical system 100 also includes a hardware interface 104 connected to a computing system 106. The computing system 106 is intended to represent one or more processors or other computing systems that may be located in one or more locations. The hardware interface 104, if present, can be used to control other components of the medical system 100 (e.g., in the case where the medical system 100 includes a magnetic resonance imaging system). The computing system 106 is also shown connected to a user interface 108 and a memory 110. The memory 110 is intended to represent any type of memory that can be connected to or accessed by the computing system 106.

[0120] Memory 110 is shown to contain machine-executable instructions 120. The machine-executable instructions 120 enable the computing system 106 to control other components of the medical system 100 via hardware interface 104. The machine-executable instructions 120 also enable the computing system 106 to perform various data processing and image processing tasks. Memory 110 is also shown to contain a neural network. The neural network has been trained to output a predicted undersampling factor for a compressed-sensing magnetic resonance imaging protocol in response to received magnetic resonance scan parameters. The magnetic resonance scan parameters describe the configuration of the magnetic resonance imaging system and the configuration of the pulse sequence commands.

[0121] Memory 110 is also shown to contain pulse sequence commands 124. Memory 110 is also shown to contain magnetic resonance scan parameters 126. Memory 110 is also shown to contain a predicted undersampling factor 128, which the neural network 122 has received in response to the input of the magnetic resonance scan parameters 126. The undersampling factor 128 can be used, for example, to adjust the k-space mode or sampling mode. Memory 110 is also shown to contain adjusted pulse sequence commands 130. These commands are pulse sequence commands 124 that, after being adjusted, match the predicted undersampling factor 128.

[0122] Figure 2 The illustrated operation is shown. Figure 1 The flowchart describes a method for a medical system. First, in step 200, a pulse sequence command 124 is received. The pulse sequence command 124 is configured to control the magnetic resonance imaging system to acquire k-space data according to a compressed sensing magnetic resonance imaging protocol. Next, in step 202, magnetic resonance scan parameters are received. Then, in step 204, a predicted undersampling factor 128 is received by inputting the magnetic resonance scan parameters 126 into a neural network 122. Finally, in step 206, the pulse sequence command is adjusted using the predicted undersampling factor 128. This may include adjusting the sampling mode in k-space.

[0123] Figure 3 The illustration shows another example of a medical system 300, which is related to... Figure 1 Similar to the medical system 100, except that the medical system 300 additionally includes a magnetic resonance imaging system 302.

[0124] The magnetic resonance imaging system 302 includes a magnet 304. Magnet 304 is a superconducting cylindrical magnet with a bore 306 passing through it. Different types of magnets can also be used; for example, split cylindrical magnets and so-called open magnets can also be used. 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 equiplanar plane 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 sufficient space between them to receive the object: the territorial arrangement of these two sections is similar to that of a Helmholtz coil. Open magnets are popular because the object is less constrained. An assembly of superconducting coils is located inside the cryostat of the cylindrical magnet.

[0125] An imaging region 308 is located within the bore 306 of a cylindrical magnet 304, wherein the magnetic field is sufficiently strong and uniform to perform magnetic resonance imaging. A region of interest 309 is shown within the imaging region 308. Typically, magnetic resonance data is acquired for the region of interest. An object 318 is shown supported by an object support 320 such that at least a portion of the object 318 is within the imaging region 308 and the region of interest 309.

[0126] Within the bore 306 of the magnet, there is also a set of magnetic field gradient coils 310. These coils are used to acquire preliminary magnetic resonance data for spatial encoding of the 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 independent sets of coils used 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 controlled as a function of time and can be either ramped or pulsed.

[0127] 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 RF transceiver 316 can be replaced by separate transmit and receive coils, and separate transmitters and receivers. It should be understood that the RF coil 314 and RF transceiver 316 are representative examples.

[0128] The RF coil 314 is also intended to represent a dedicated transmitting antenna and a dedicated receiving antenna. Similarly, the transceiver 316 can also represent a separate transmitter and receiver. The RF coil 314 can also have multiple receiving / transmitting elements, and the RF transceiver 316 can have multiple receiving / transmitting channels. For example, if parallel imaging techniques such as SENSE or acceleration techniques such as compressed sensing are being performed, the RF coil 314 will have multiple coil elements.

[0129] Transceiver 316 and gradient controller 312 are shown as hardware interface 106 connected to computer system 102.

[0130] The memory 110 is also shown to contain k-space data 330, which is acquired by controlling the magnetic resonance imaging system 302 using a tuned pulse sequence command 130. The memory 110 is also shown to contain magnetic resonance imaging data 332 reconstructed from the k-space data 330.

[0131] Figure 4 The illustrated operation is shown. Figure 3 The flowchart of the method of medical system 300. Figure 4 The method in [the text] is similar to Figure 2 The method shown. Figure 4 The method in the text begins with step 200, and steps 200, 202, 204, and 206 are as follows: Figure 2 The steps are illustrated in the diagram. After performing step 206, the method proceeds to step 400. In step 400, the magnetic resonance imaging system 302 is controlled using the adjusted pulse sequence command 130 to acquire k-space data 330. Finally, in step 402, magnetic resonance image data 332 is reconstructed based on the k-space data 330.

[0132] MRI is a widely used diagnostic tool, offering a high degree of imaging contrast and functionality. MR image acquisition is controlled by a large number of parameters that are routinely accessible in clinical settings. Optimization of these imaging parameters is performed at each individual location. To date, protocol optimization has not been standardized, and results and image quality depend heavily on operator experience.

[0133] Applications can use artificial intelligence (AI)-based methods to automatically predict the optimal compressed sensing acceleration factor (128) as the predicted undersampling factor for each protocol, thereby reducing the variability of protocol variations between clients and reducing the dependence of results on the experience of application experts.

[0134] Examples can provide a very effective way to leverage the correlations among a large number of parameters and allow parameters to be directly linked to results (e.g., image quality) by using training data.

[0135] For well-defined applications (e.g., using compressed sensing to accelerate image acquisition), neural network techniques (e.g., deep learning) can be used to predict the optimal compressed sensing factor for any given parameter settings by using compressed sensing successfully implemented by experienced application experts as training data. These predictions can then be used as a starting point (an educated guess) for each application expert or can be directly provided as guidance for optimizing their own parameters during or after application training.

[0136] The example can address one or more of the following problems and drawbacks:

[0137] 1. The results of sequence parameter optimization depend on the experience of application experts:

[0138] a. Better comparability between protocols in different locations.

[0139] 2. Increased workload for application specialists (especially during the introduction of new products and series), leading to a shortage of application specialists:

[0140] a. Reduced workload for application experts through automated guidance.

[0141] 3. Utilize protocol optimization to address individual and continuous overload issues for customers.

[0142] a. Provide automated guidance to customers

[0143] Examples can use neural networks, such as those trained using deep learning-based computer algorithms trained with well-controlled MRI protocol parameters derived from protocol optimization using compressive sensing. Then, depending on other scan parameter settings, a computer algorithm is used to predict the optimal compressive sensing factor.

[0144] Examples can provide algorithms based on multilayer artificial neural networks (neural network 122).

[0145] Training: In the initial training phase, a set of sequence parameter settings that are well-received in terms of image quality and maximum compressible sensing acceleration is collected. These sequence parameter settings (Magnetic Resonance Scanning Parameters 126) can be derived from trained application experts or currently used sequence parameter settings. These collected sequence parameter settings are referred to below as the initial training data. The following... Figure 5 A schematic diagram of the training phase of an artificial neural network is shown.

[0146] During the initial training phase, a subset of sequence parameters from the initial training data is defined as input parameters, and the compressed sensing acceleration factor is defined as the output parameter of the artificial neural network and used for network training.

[0147] Figure 5 A schematic diagram of the artificial neural network (neural network 122) during the initial training phase is shown. In this phase, the selected sequence parameters or magnetic resonance scanning parameters 126, along with the corresponding compressed sensing acceleration factor for the predicted undersampling factor 128 from an evaluated or previous dataset, are fed into neural network 122 to train the network. Arrow 126 indicates the known magnetic resonance scanning parameters 126. These items are input into input layer 500. The input layer is then connected to fully connected layer 502. The final fully connected layer 502 is connected to output 504, which provides the value of the predicted undersampling factor 128 or compressed sensing acceleration factor.

[0148] During the evaluation phase, the sequence parameters are then fed into the trained artificial neural network as input parameters, and the network calculates the compression sensing factor as the output parameter, as follows: Figure 6 As shown. Figure 6 The diagram illustrates the neural network 122 during the evaluation or usage phase. In this state, the neural network 122 has been trained. During usage, the magnetic resonance scanning parameters 126 are input into the input layer 500. Then, the fully connected layer 502 acquires this output and provides the predicted undersampling factor 128 at the output 504 as a response. At this stage, the trained artificial neural network is used to calculate the optimal compressed sensing acceleration factor based on multiple input parameters.

[0149] Examples include integrating neural networks directly into scanning software to allow for “automatic” settings for the selection of the compressed sensing acceleration factor (or “CS-SENSE”). Figure 7This scenario is illustrated in summary. If "Auto" is selected for "Compressed Sensing," numerous scanning parameters are fed directly into the trained neural network, and the calculated compressed sensing acceleration factor is then displayed in the software and used for measurement. If additional optimizations are made to the compressed sensing acceleration factor beyond what the algorithm calculates, these optimizations can be used as additional training data through feedback or reinforcement learning.

[0150] Figure 7 This illustrates how neural network 122 can be integrated into medical system 300. The user interface 108 of the magnetic resonance imaging system 302 has a page for inputting scan parameters. The user interface can provide scan parameters 126, which are then input into the artificial neural network 122. In response, a predicted undersampling factor 128 can be provided. It should be noted that... Figure 7 In this example, the magnetic resonance scanning parameter 126 may not be the actual parameter input into the neural network.

[0151] exist Figure 7 In this process, if compressed sensing reduction is set to "automatic," the optimal compressed sensing acceleration factor (CS-SENSE factor) is predicted using a pre-trained artificial neural network. The predicted CS-SENSE factor is displayed and used for verification.

[0152] Proof of Principle: The proof-of-principle implementation was tested using approximately 3000 datasets. Each dataset in these datasets uses MR sequence parameter settings for compressed sensing optimized by application experts. For the initial training of the artificial neural network, the data was split into 2934 training datasets (training data) and 227 test datasets (test data). The artificial neural network was trained using the training data. The optimal compressed sensing acceleration factor was predicted based on a set of input parameters using the test data. Subsequently, the predicted optimal compressed sensing acceleration factor was compared with the compressed sensing acceleration factor optimized by application experts (see below). Figure 8 ).

[0153] Figure 8 The diagram illustrates the test of neural network 122. Figure 8 The plot in the image illustrates the relationship between the true factor 800 and the predicted factor 802. A dataset of approximately 3000 MRI sequences (optimized by application experts) was divided into 2934 training datasets for initial training of the artificial neural network. The trained artificial neural network was tested using 227 datasets by predicting the optimal compressible sensing acceleration factor based on 17 predefined sequence parameters. The plot demonstrates a close agreement between the application expert-optimized compressible sensing acceleration factor and the compressible sensing acceleration factor predicted by the artificial neural network.

[0154] Field Testing: Field testing was conducted in collaboration with application experts. A database of archived scan parameter data was used prior to implementing compressed sensing for this field test. A trained artificial neural network was used to calculate a set of predicted compressed sensing acceleration factors. Figure 5 The difference between the compression sensing factor predicted by the artificial neural network and the compression sensing factor estimated by application experts is shown. In approximately 72% of the scans, the difference between the predicted and actual compression sensing factors is less than 1, and for 98% of the scans, this difference is less than 1.5, indicating that the proposed solution has very promising performance.

[0155] The neural network is constructed by comparing its output with an undersampling factor of 128 (denoted as "δ" in this paper) of the predictions from actual use of clinical data. Figure 9 and Figure 10 This is how the output of the neural network is compared to the accuracy of predicting the undersampling factor 128 used in a clinical setting. The lower the δ value, the more accurately the neural network corresponds to the predicted undersampling factor 128 actually used in clinical practice.

[0156] Figure 9 The δ values ​​for 194 comparisons are shown in pie chart form. The pie chart is divided into different δ levels.

[0157] Figure 10 The same data is shown in the following format, where 72% of the values ​​are δ less than 1. Figure 9 and Figure 10 The diagram shows that the neural network provides a predicted undersampling factor of 128 that is similar to the predicted undersampling factor of 128 used clinically.

[0158] Figure 9 and Figure 10 Performance in field testing. In 98% of scans, the difference between the predicted compression sensing factor and the actual compression sensing factor used was less than 1.5, and in 72% of scans, this difference was less than 1.

[0159] MRI parameters affecting image acceleration

[0160] The MRI scan parameters listed below affect the optimal image acceleration (undersampling factor 128). However, there can often be strong correlations between different parameters. This means that it is impossible to determine which acceleration factor is optimal based on a single parameter or a very limited set of parameters. This makes the selection of the optimal acceleration factor a complex multidimensional optimization problem. The parameters listed below are more or less general and independent of the MRI system manufacturer; however, naming conventions vary significantly between different manufacturers. Furthermore, the implementation of the parameters may differ from manufacturer to manufacturer, and not all parameters are accessible to MRI users.

[0161] The following section will discuss some of the magnetic resonance imaging (MRI) parameters in more detail. These MRI parameters may include one or more of the following:

[0162] 1. Coil (RF coil configuration)

[0163] The connected coils provide various information.

[0164] a. The number of coil elements affects the performance of image acceleration.

[0165] b. Coil geometry affects image acceleration performance.

[0166] c. The body parts being examined can be partially assumed to be: knee coil – most likely the knee; head coil – most likely a head / brain examination.

[0167] 2. Scanning modes (3D and 2D)

[0168] a. 3D allows for a higher speedup factor because scanning can be accelerated in two spatial dimensions.

[0169] 3. Sequence types (spin echo, gradient echo, balanced SSFP, inversion recovery, fast spin echo, FLASH, EPI)

[0170] a. Scanning techniques include information about image contrast (T1, T2, T2*, T1 / T2-bSSFP).

[0171] b. Whether gradient balancing, gradient destruction, or RF destruction sequences are used.

[0172] c.-> The two parameters are used to describe this.

[0173] d. Fast imaging mode includes information about image contrast and how the k-space is acquired (one k-space line per excitation vs. several k-space lines per RF excitation).

[0174] 4. Echo Time (TE) and Repeat Time (TR)

[0175] a. TE is the time distance between the signal excitation and the acquisition at the k-space center.

[0176] b. TR is the time between two successive radio frequency excitations of the same imaging volume.

[0177] 5. Flip angle

[0178] a. The flip angle is the excitation power of the radio frequency pulse used to excite the spin during the imaging sequence.

[0179] 6. ACQ voxel size / spatial resolution (including slice thickness) in all three dimensions.

[0180] a. Voxel sizes collected in all three spatial dimensions

[0181] b. In the tests presented in this paper, two parameters were used to describe this.

[0182] 7. Three-dimensional field of view (FOV)

[0183] a. FOV is the coverage area scanned in all three spatial dimensions.

[0184] b.-> We use two parameters to describe this.

[0185] 8. Matrix size

[0186] a. The matrix size is the number of voxels or pixels along the three spatial dimensions.

[0187] 9. Scanning time

[0188] a. Time required for scanning with / without acceleration

[0189] 10. Fat suppression (this may not be relevant to all MR protocols)

[0190] a. Depending on the imaging sequence, it may be necessary to suppress signals from fat.

[0191] b. There are different fat-inhibiting technologies: mDixon, STIR, SPIR, SPAIR, and PROSET. Each technology has a different impact on acceleration performance.

[0192] c. Use three parameters to describe this.

[0193] 11. Water-fat transfer (WFS): (This may not be relevant to all MR protocols)

[0194] a. Shift of water and fat signals in acquired images represented by voxels

[0195] 12. Bandwidth (BW)

[0196] a. Bandwidth of data sampling during acquisition.

[0197] 13. Number of signal averaging steps (NSA)

[0198] a. The number of acquisitions per scan, averaged to provide a decent image.

[0199] 14. Dynamic scan count

[0200] a. Dynamic numbers in dynamic scanning

[0201] In addition to the magnetic resonance scanning parameters mentioned above, one or more of the following parameters may also be advantageously included:

[0202] 1. Reconstructing voxel size / reconstruction matrix

[0203] a. MR images are typically interpolated during image reconstruction.

[0204] b. Reconstructing voxel sizes provides interpolated voxel sizes.

[0205] c. The reconstruction matrix provides the number of voxels in each of the three spatial dimensions.

[0206] 2. Pre-pulse type

[0207] a. Use different types of RF prepulses before RF signal excitation.

[0208] b. Different types of pre-pulses: T2 preparation, inversion, saturation, MDME, MTC, etc.

[0209] 3. Partial Fourier Transform (Half-Scan)

[0210] a. Partial Fourier or half-scan is a technique that acquires only a portion of the k-space and uses k-space symmetry to reconstruct the complete image.

[0211] 4. Orientation of imaging volume / slice orientation

[0212] a. Image acquisition direction: axial, coronal, or sagittal

[0213] 5. Folding direction

[0214] a. In which direction is phase-coded (folded), and in which direction is frequency-coded (unfolded)?

[0215] 6. Use of contrast agents

[0216] a. Was contrast agent used in the scan? If so, more signal is available and the speed can be increased.

[0217] 7. Excitation Mode (Single Excitation vs. Multiple Excitation)

[0218] Is the AK space collected in one step or in multiple steps?

[0219] 8. k-space contour order

[0220] The order in which the ak-space lines are collected: linearly from one side to the other, starting from the center of the k-space, starting from the edge of the k-space, asymmetrically, or randomly.

[0221] 9. k-space trajectory

[0222] How AK space is collected: Cartesian, spiral, radial, etc.

[0223] 10. Physiological synchronization

[0224] a. Is the sequence synchronized with cardiac motion (e.g., via ECG)?

[0225] b. Is the sequence synchronized with respiratory movements (e.g., camera or breathing belt)?

[0226] 11. Diffusion Coding

[0227] a. Is diffusion coding used in techniques such as DTI or DWI?

[0228] 12. k-space partitioning factor

[0229] a. The Fast Field Echo (TFE) and Fast Spin Echo (TSE) factors describe how many k-space lines were acquired in a set of excitations (TFE) or during a sequence of echoes (TSE).

[0230] 13. Echo count

[0231] a. Describe the number of echoes collected from the same k-spaceline.

[0232] Although the invention has been illustrated and described in detail in the accompanying drawings and the foregoing description, such illustrations and descriptions should be considered illustrative or exemplary, and not restrictive; the invention is not limited to the disclosed embodiments.

[0233] Those skilled in the art, through studying the accompanying drawings, disclosure, and claims, will be able to understand and implement other variations of the disclosed embodiments when 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 multiple. A single processor or other unit may implement the functions of several items recited in the claims. Although certain measures are recited in dissimilar dependent claims, this does not imply that combinations of these measures cannot be advantageously used. Computer programs may be stored / distributed on suitable media, such as optical storage media or solid-state media supplied together with or as part of other hardware, but 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 should be construed as limiting the scope.

Claims

1. A medical system (100, 300), comprising: A memory (110) configured to store machine-executable instructions (110), wherein the memory is also configured to store a trained neural network (122). A computing system (106) configured to control a magnetic resonance imaging system, wherein the execution of machine-executable instructions causes the computing system to: Receive pulse sequence commands, which are configured to control the magnetic resonance imaging system (302) to acquire k-space data according to the compressed sensing magnetic resonance imaging protocol; Receive magnetic resonance scanning parameters describing the configuration of the magnetic resonance imaging system; In response to inputting the magnetic resonance scanning parameters into the trained neural network, an undersampling factor is predicted for the compressed sensing magnetic resonance imaging protocol, wherein the trained neural network has been trained based on historical data of successful image acquisitions associated with a combination of appropriate undersampling factors and magnetic resonance scanning parameters; and The pulse sequence command (206) is adjusted based on the predicted undersampling factor to select or modify the sampling of the k-space data.

2. The medical system according to claim 1, wherein, The magnetic resonance scanning parameters include: radio frequency coil configuration, scanning mode for specifying two-dimensional or three-dimensional scanning, sequence type for specifying the contrast of the pulse sequence command, echo time, pulse repetition time, voxel size or three-dimensional spatial resolution, three-dimensional field of view, and radio frequency bandwidth during k-space sampling.

3. The medical system according to claim 2, wherein, The magnetic resonance imaging (MRI) scan parameters also include any of the following: the type of fat suppression protocol being used, the flip angle, the scan time, the orientation of the field of view, the folding direction, the number of dynamic scans, the type of contrast agent used, the number of signal averaging operations, and combinations thereof.

4. The medical system according to claim 3, wherein, The magnetic resonance scanning parameters also include any of the following: reconstructed voxel size or reconstruction matrix size, type of prepulse used, implementation of a partial Fourier half-scan protocol, anatomical part being examined, type of excitation used, k-space contour order, k-space trajectory, physiological synchronization, type of diffusion coding technique, k-space segmentation factor, number of echoes used to acquire the same k-space lines, and combinations thereof.

5. The medical system according to any one of claims 1-4, wherein, The execution of the machine-executable instructions also enables the computing system to: Retrieve archived scan parameter data from the magnetic resonance imaging (MRI) scan parameter database; The archived training data is constructed based on the archived scan parameter data; and The neural network is trained using the archived training data.

6. The medical system according to claim 5, wherein, The archived training data was retrieved remotely.

7. The medical system according to claim 6, wherein, The archived training data was retrieved remotely via a network connection.

8. The medical system according to any one of claims 1-4, wherein, The medical system also includes the magnetic resonance imaging system (302), wherein the execution of the machine-executable instructions further enables the computing system to: The magnetic resonance imaging system is controlled using the pulse sequence commands to acquire (400) the k-space data; and (402) magnetic resonance image data are reconstructed based on the k-space data.

9. The medical system according to claim 8, wherein, The medical system includes a user interface, wherein the execution of machine-executable instructions further enables the computing system to: Before adjusting the pulse sequence command, the predicted undersampling factor and at least a portion of the magnetic resonance scan parameters are displayed on the user interface; and The predicted undersampling factor is received from the user interface in response to displaying the undersampling factor, wherein the pulse sequence command is adjusted using the predicted undersampling factor.

10. The medical system according to claim 9, wherein, The execution of the machine-executable instructions also enables the computing system to: User-specific training data is constructed based on the magnetic resonance scanning parameters and the predicted undersampling factor; and The neural network is trained using the user-specific training data.

11. The medical system according to any one of claims 1-4, wherein, The magnetic resonance imaging protocol is a parallel imaging magnetic resonance imaging protocol.

12. The medical system according to any one of claims 1-4, wherein, The neural network is a multi-layer neural network.

13. The medical system according to claim 12, wherein, The multilayer neural network comprises at least 6 layers, wherein each of the at least 6 layers is fully connected to its adjacent layers.

14. A method of operating a medical system (100, 300), comprising: Receive pulse sequence commands, which are configured to control the magnetic resonance imaging system (302) to acquire k-space data according to the compressed sensing magnetic resonance imaging protocol; Receive magnetic resonance scanning parameters, which describe the configuration of the pulse sequence command and the configuration of the magnetic resonance imaging system; In response to inputting the magnetic resonance scanning parameters into a trained neural network to predict an undersampling factor (128) for the compressed sensing magnetic resonance imaging protocol, wherein the trained neural network has been trained based on historical data of successful image acquisitions associated with a combination of appropriate undersampling factors and magnetic resonance scanning parameters; and The pulse sequence command (206) is adjusted based on the predicted undersampling factor to select or modify the sampling of the k-space data.

15. A computer program comprising machine-executable instructions (120) for execution by a computing system configured to control medical systems (100, 300), wherein, The execution of the machine-executable instructions enables the computing system to: Receive pulse sequence commands, which are configured to control the magnetic resonance imaging system (302) to acquire k-space data according to the compressed sensing magnetic resonance imaging protocol; Receive magnetic resonance scanning parameters (126), which describe the configuration of the pulse sequence command and the configuration of the magnetic resonance imaging system; In response to inputting the magnetic resonance scanning parameters into a trained neural network to predict an undersampling factor (128) for the compressed sensing magnetic resonance imaging protocol, wherein the trained neural network has been trained based on historical data of successful image acquisitions associated with a combination of appropriate undersampling factors and magnetic resonance scanning parameters; and The pulse sequence command (206) is adjusted based on the undersampling factor to select or modify the sampling of the k-space data.

Images