Generation of magnetic resonance fingerprinting pulse sequences

Abstract

A medical system (100, 300) is disclosed herein. Execution of the machine executable instructions (120) causes the computing system (104) to repeatedly: generate (200) a random input vector; receiving (202) a generated MRF pulse sequence (128) in response to inputting the random input vector into the GAN generator neural network (122); and appending (204) the generated MRF pulse sequence to an MRF pulse sequence database (130). Execution of the machine executable instructions causes the computing system to: input (208) each generated MRF pulse sequence in the MRF pulse sequence database into a trained scoring algorithm (124) to assign one or more score values to each generated MRF pulse sequence; and receiving (210) a selected MRF pulse sequence (136) from the MRF pulse sequence database by applying a predetermined criterion (134) to one or more score values of each generated MRF pulse sequence in the MRF pulse sequence database.

Description

Technical Field

[0001] The present invention relates to magnetic resonance imaging, and in particular to magnetic resonance fingerprint recognition. Background Art

[0002] Magnetic resonance fingerprinting (MRF) is a technique in which multiple RF pulses distributed in time are applied so that they cause signals from different materials or tissues to have unique contributions to the measured MR signal. The signal depends not only on the physical parameters of the tissue (T1, T2, etc.), but also on the parameters of the magnetic field, such as B0 and Bl. These are encoded in various combinations, resulting in tens or even hundreds of thousands of entries in the Magnetic Resonance Fingerprinting dictionary.

[0003] In most applications, MRF is used to determine highly resolved quantitative maps of T1, T2 or other physical tissue parameters. Optimization of the sequence becomes important because it is necessary to ensure that small changes in T1, T2 or other parameters are uniquely reflected in a measurable change in signal shape.

[0004] Journal Publication Hamilton, Jessse I, and Nicole Seiberlich. Machine Learning for Tissue Property Quantification with Rapid Magnetic Resonance Fingerprinting. Proceedings IEEE 108.1(2019):69-85 discloses that magnetic resonance fingerprinting (MRF) is a magnetic resonance imaging (MRI)-based method that can simultaneously provide quantitative maps of multiple tissue properties from a single rapid acquisition. Tissue property maps are generated by matching the complex signal evolution collected at the scanner with a dictionary of signals simulated using Bloch equations. It also discloses that in some cases, the process of dictionary generation and signal matching can be time consuming, reducing the practicality of this technique. Recently, several groups have proposed the use of machine learning to accelerate the extraction of quantitative maps from MRF data. This article provides an overview of research combining MRF and machine learning, as well as original research showing how machine learning can accelerate dictionary generation for cardiac MRF (cMRF). M. Yang et al. published a preprint paper titled "Learning Bloch Equations to Simulate fc MR Fingerprinting Games" on the Cornell University Library Arxiv.org on April 5, 2020, which involves generating MRF dictionaries from random noise inputs (rather than acquisition sequences) based on GAN. Summary of the invention

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

[0006] The above-mentioned article by Hamilton et al. points out the time-consuming process of generating a dictionary for MRF. Generating a pulse sequence for performing MRF can also be time-consuming. It may be necessary to repeatedly modify the MRF pulse sequence and then use Bloch simulation to calculate its effectiveness. The optimization of performing MRF pulse sequences can be computationally intensive. An embodiment can provide a means for generating MRF pulse sequences more quickly. A generative adversarial network (GAN) can be trained to generate suitable MRF pulse sequences. GAN is formed by a GAN generator neural network and a GAN discriminator pair. The pair is trained together with the GAN discriminator being trained to distinguish between real and fake MRF pulse sequences. In this context, the GAN discriminator is trained to identify MRF pulse sequences that provide good information encoding capabilities. The GAN generator neural network is trained to generate a generated MRF pulse sequence in response to receiving a random input vector. The dimension of the random input vector can be selected. The GAN generator neural network is trained so that the generated MRF pulse sequence can deceive the GAN discriminator. This ultimately achieves the ability to increase the GAN discriminator to improve its ability to identify suitable MRF pulse sequences.

[0007] The present invention as set out in the set of claims 1 to 15 relates to the generation of magnetic resonance fingerprinting (MRF) acquisition sequences based on GANs, combined with a trained scoring algorithm to produce score values ​​associated with the generated MRF acquisition sequences. Generative adversarial networks (GANs) have the ability to synthesize realistic new data sets from a given set, and in the present invention, additional MRF sequences are generated from a ground truth set of known good MRF acquisition sequences. GANs include generator and discriminator networks. The generator generates data based on whether the discriminator marks it as true or not. The discriminator marks the data as correct or wrong based on learning from real data. The training of these two networks is performed alternately, so that they each gradually get better at generation and discrimination, and (ideally) get better at the same speed. The present invention enables faster generation of MRF acquisition sequences, which can be generated on the fly or on demand. The present invention avoids the need for lengthy and time-consuming Bloch simulations. The generator is fed with noise to create a random sequence definition, that is, a new MRF sequence is returned in response to the noise implementation input. The discriminator is trained to distinguish such "false" MRF sequences from the "real" MRF sequences provided for training. By iteratively optimizing the weights of the two networks, the generator learns to generate MRF sequences with properties similar to the ground truth set of known MRF sequences. In order to quantify the degree of similarity between the generated MRF sequences and their ground truth sequences, a score value is assigned to each generated MRF acquisition sequence to represent the encoding ability of the MRF sequence. The generator is fed with noise to create a random sequence definition. The discriminator is trained to distinguish such "false" MRF sequences from the "real" MRF sequences provided for training. By iteratively optimizing the weights of the two networks, the generator learns to generate MRF sequences with properties similar to the known properties. The discriminator is trained to distinguish such "false or fake" MRF sequences from the "real" MRF sequences provided for training. By iteratively optimizing the weights of the two networks, the generator learns to generate MRF sequences with properties similar to the known properties. According to the present invention, the generated MRF sequence is selected based on its score value. The score value of the MRF acquisition sequence represents the encoding power of the MR acquisition sequence, that is, the level of distinction of each parameter setting in the MRF acquisition sequence. According to the present invention, a score value is returned in response to a generated MRF acquisition sequence by a trained scoring algorithm, which is trained to predict the encoding power. The score value can be implemented as an inner product between pairs of MRF sequences, optionally averaging the parameter space of imaging parameters that vary over the MRF sequence. The generator is fed with noise to create a random sequence definition. These scores provide a measure of encoding power, so that generated MRF acquisition sequences that are well distinguished from the parameter settings can be selected. The score value itself is known from the paper "Predicting the encoding power of MR fingerprint recognition sequences" by K. Sommer et al. in "Magnetic Resonance Imaging" 41 (2017) 7-14.Thus, the effect of the present invention is to generate MRF acquisition sequences based on a set of ground truth values ​​of MRF acquisition sequences, and to select the best solution among the generated acquisition sequences based on a score value representing the encoding power of the generated MRF acquisition sequences. In this way, the GAN can be further trained during its operation. As a result of the selection of the MRF acquisition sequences, only the MRF dictionary needs to be generated, for example, by a Bloch simulator for the selected MRF acquisition sequences, for which good encoding power has been established according to their score values.

[0008] Embodiments of the present invention also include a trained scoring algorithm that is trained to output one or more score values ​​in response to receiving an MRF pulse sequence. The trained scoring algorithm replaces the use of a Bloch simulator to evaluate the effectiveness of a particular MRF pulse sequence.

[0009] To generate new MRF pulse trains, random input vectors are repeatedly generated and input into the GAN generator neural network. This can be used to generate a large number of generated MRF pulse trains. These can be stored in a folder or MRF pulse train database, for example. As used herein, a database is a system for storing and retrieving data. This may include anything from a relational database to a system for locating stored files.

[0010] The trained scoring algorithm is then used to screen the MRF pulse sequence database. Each generated MRF pulse sequence can be input into the trained scoring algorithm to obtain one or more score values ​​that can be compared with a predetermined criterion. For example, the one or more score values ​​can be calculated when the generated pulse sequence is stored in the MRF pulse sequence database, or the one or more score values ​​can be calculated on the fly when searching for a selected MRF pulse sequence.

[0011] The advantage of generating selected MRF pulse sequences in this way is that it can be very fast. The GAN generator neural network can produce a large number of generated MRF pulse sequences very quickly, and can quickly use the trained scoring algorithm to screen the MRF pulse sequence database. It can, for example, be able to generate MRF pulse sequences on the fly or on demand.

[0012] A hybrid approach can also be used. For example, a trained scoring algorithm can be used to identify a subset of MRF pulse sequences that meet the screening criteria from a database of MRF pulse sequences. The numerically calculated score values ​​of the members of the subset of MRF pulse sequences can then be calculated to select the MRF pulse sequence that best matches the needs specified by the screening criteria. The numerically calculated score values ​​can be calculated using a Bloch simulator. The advantage of this solution is that the number of numerical simulations is greatly reduced.

[0013] In one aspect, the present invention provides a medical system. The medical system includes a memory storing a number of things. The memory stores machine executable instructions, a GAN generator neural network, and a trained scoring algorithm. The GAN generator neural network is configured to output a generated magnetic resonance fingerprinting (MRF) pulse sequence in response to receiving a random input vector. The trained scoring algorithm is configured to output one or more score values ​​in response to receiving the generated MRF pulse sequence as input.

[0014] The medical system also includes a computing system. Execution of the machine executable instructions causes the computing system to repeatedly generate a random input vector. Execution of the machine executable instructions also causes the computing system to repeatedly receive the generated MRF pulse sequence in response to inputting the random input vector into the GAN generator neural network.

[0015] Execution of the machine executable instructions also causes the computing system to repeatedly append the generated MRF pulse train to the MRF pulse train database. The GAN generator neural network outputs a valid or generated MRF pulse train in response to receiving a random input vector. Typically, constructing highly optimized MRF pulse trains is very computationally intensive. By training the GAN to generate them in response to receiving a random input vector, a large number of generated MRF pulse trains can be constructed in a very short period of time. This large number of generated MRF pulse trains can then be stored in a database system.

[0016] Execution of the machine executable instructions also causes the computing system to input each generated MRF pulse sequence in the MRF pulse sequence database into a trained scoring algorithm to assign one or more score values ​​to each generated MRF pulse sequence. Execution of the machine executable instructions also causes the computing system to select a selected MRF pulse sequence from the MRF pulse sequence database by applying a predetermined criterion to one or more score values ​​of each generated MRF pulse sequence in the MRF pulse sequence database. The GAN generator neural network can be used to generate a large number of MRF pulse sequences. However, for a particular application or use, it may not be obvious which is the best MRF pulse sequence to select from the MRF pulse sequence database.

[0017] In order to do this in a very fast manner, when the MRF pulse sequence is input into a trained scoring algorithm, the algorithm outputs one or more score values. These different score values ​​may have different values ​​associated with different image weights or image quality aspects. Then, the MRF pulse sequence database can be browsed and the one or more score values ​​for each pulse sequence in the MRF pulse sequence database can be viewed or compared with a predetermined criterion. For example, the predetermined criterion may emphasize a particular image weight factor, such as T1 or T2-*. In other examples, several different factors may be considered simultaneously and the weights between them may be used. In any case, using these two machine learning modules, namely the GAN generator neural network and the trained scoring algorithm, a selected MRF pulse sequence can be selected very quickly from the MRF pulse sequence database.

[0018] In another embodiment, execution of the machine executable instructions further causes the computing system to calculate a magnetic resonance fingerprinting (MRF) dictionary with predetermined dictionary entries of a selected MRF pulse sequence by performing a Bloch equation simulation controlled by the selected MRF pulse sequence for each predetermined dictionary entry in the predetermined dictionary. In this embodiment, an MRF dictionary of the selected MRF pulse sequence is prepared. This may be beneficial because an appropriate MRF dictionary may be required in order to obtain useful information from k-space data acquired using the selected MRF pulse sequence.

[0019] The MR signals in the dictionary can be calculated by modeling the response of each substance in the material as a pulse sequence using the well-known Bloch equation. These predicted MR signal values ​​can then be compared with the measured MR signal. Each substance in the dictionary of substances can potentially make a positive contribution to the measured MR signal. By comparing at all measurement times or multiple measurement times, the substances in the dictionary can be identified based on the magnetic resonance fingerprint, accurately deconvolving the composition of the region that makes up the measured magnetic resonance signal.

[0020] In another embodiment, the medical system further comprises a magnetic resonance imaging system. Execution of the machine executable instructions further causes the computing system to acquire MRF k-space data by controlling the magnetic resonance imaging system using a selected MRF pulse sequence. Execution of the machine executable instructions further causes the computing system to reconstruct one or more magnetic resonance fingerprinting images using the MRF k-space data and the MRF dictionary according to a magnetic resonance fingerprinting protocol. This embodiment may be beneficial because one or more MRF images are reconstructed from k-space data acquired using a selected MRF pulse sequence and an MRF dictionary customized therefor. This may provide magnetic resonance fingerprinting imaging with improved quality.

[0021] In some examples, the acquisition of MRF k-space data may involve adjustments to the selected MRF pulse sequence. For example, the gradient is typically adjusted to control the field of view.

[0022] In another embodiment, the GAN generator neural network is further configured to receive one or more control switches as input. The one or more control switches may be values ​​placed in the input of the GAN generator neural network. Execution of the machine executable instructions further causes the computing system to receive the one or more control switches. Execution of the machine executable instructions further causes the computing system to input the one or more control switches into the GAN generator neural network with the random input vector. Execution of the machine executable instructions further causes the computing system to store the one or more control switches in the MRF pulse sequence database together with the generated MRF pulse sequence. The selected MRF pulse sequence is selected at least in part using the selected set of control switches.

[0023] The control switches can be used, for example, to specify data about the subject, details about the acquisition, or even anatomical regions. Thus, this can enable a single GAN neural network to be used for a variety of situations. For example, if different anatomical regions are being imaged, one solution would be to have multiple GAN generator neural networks dedicated to them. Another solution is to use control switches and then also use the control switches during training, making the GAN generator more flexible and its behavior controllable and configurable.

[0024] In another embodiment, one or more of the control switches includes an anatomical region.

[0025] In another embodiment, one or more of the control switches include a clinical question. The clinical question may be, for example, the type of examination being performed.

[0026] In another embodiment, one or more control switches include timing constraints on the acquisition of k-space data. This may be useful in the presence of repetitive motion, such as heartbeat or breathing.

[0027] In another embodiment, the one or more control switches include a breath-hold duration.

[0028] In another embodiment, selecting a selected MRF pulse sequence from an MRF pulse sequence database by applying a predetermined criterion to one or more score values ​​for each generated MRF pulse sequence in the MRF pulse sequence database includes: using the one or more score values ​​to calculate a composite score, and selecting the generated MRF pulse sequence with the highest composite score. For example, the one or more score values ​​may include scores for different things, such as T1 weighting, T2 weighting, or detection of a particular tissue type. The composite score may be a way to balance all of these different factors and use them to select the selected MRF pulse sequence.

[0029] In another embodiment, execution of the machine executable instructions further causes the computing system to receive user input. The user input specifies the composition of the composite score. The selected MRF pulse sequence is selected in response to receiving the user input. For example, there may be a user interface that enables a user to select which components of one or more score values ​​are important. This can then be used to select the selected MRF pulse sequence.

[0030] In another embodiment, selecting a selected MRF pulse sequence from an MRF pulse sequence database includes selecting a subset of MRF pulse sequences that meet the screening criteria from the MRF pulse sequence database. The selection also includes calculating a set of numerically calculated score values ​​for each MRF pulse sequence in the subset of MRF pulse sequences. The selection also includes selecting a selected MRF pulse sequence from the subset of MRF pulse sequences by comparing a predetermined criterion with a set of numerically calculated scores for each MRF pulse sequence in the subset of MRF pulse sequences. A specific MRF pulse sequence can be evaluated by performing a simulation. For example, a Bloch simulation can be used to determine the effectiveness of an MRF pulse sequence in providing data that can be used for a T1-weighted image. Since the trained scoring algorithm may have limitations because it is a trained or machine-learned component, a numerical determination of one or more score values ​​can also be performed. For example, the database may have a large number of MRF pulse sequences. The trained scoring algorithm can be used for preliminary screening and reducing the number of MRF pulse sequences that need to be searched. The MRF pulse sequence identified by the trained scoring algorithm can then be refined by performing numerical calculations on the score values ​​of each MRF pulse sequence in the subset of MRF pulse sequences.

[0031] In another embodiment, the score value comprises a T1 accuracy score.

[0032] In another embodiment, the score value comprises a T2 accuracy score.

[0033] In another embodiment, the score value comprises a T2-star accuracy score.

[0034] In another embodiment, the one or more score values ​​include a proton density accuracy score.

[0035] In another embodiment, execution of the machine executable instructions further causes the computing system to receive a training MRF pulse sequence and train a GAN generator neural network and a GAN discriminator neural network using the training MRF pulse sequence according to a GAN training protocol. In a GAN, two different neural network components are trained in conjunction. There is a GAN generator neural network and a GAN discriminator neural network. As previously described, the GAN generator neural network receives a random input vector and outputs an MRF pulse sequence in response. The GAN discriminator network is trained to identify whether the MRF pulse sequence is real. The GAN generator network is trained using the GAN discriminator network, and alternately, the GAN discriminator neural network is trained using a training MRF pulse sequence generated by the GAN generator neural network and a false or forged MRF pulse sequence.

[0036] In another embodiment, execution of the machine executable instructions further causes the computing system to receive a previous MRF pulse sequence. Execution of the machine executable instructions further causes the computing system to calculate a set of numerically evaluated score values ​​for each of one or more score values ​​of the previous MRF pulse sequence. This may, for example, involve establishing a set of Bloch equations and numerically solving one or more score values. Execution of the machine executable instructions further causes the computing system to train the trained scoring algorithm using the previous MRF pulse sequence and the set of numerically evaluated score values. The exact training method will depend on the type of algorithm or machine learning module used to implement the trained scoring algorithm.

[0037] Training a GAN generator neural network and a GAN discriminator neural network (600) using a training MRF pulse sequence according to a GAN training protocol and involving training a scoring algorithm can be performed by a medical system that deploys the trained scoring algorithm for inference. The training can also be performed separately or remotely from the medical system that deploys the trained scoring algorithm. In addition, the GAN network can be remotely trained like the initial trained GAN network deployed in the medical system of the present invention, and the deployed GAN work (particularly the trained scoring algorithm) can then be continuously trained when deployed.

[0038] In another embodiment, the trained scoring algorithm is implemented as a decision tree.In another embodiment, the trained scoring algorithm is implemented as a support vector machine.

[0039] In another embodiment, the trained scoring algorithm is implemented as a multi-layer perceptron neural network. In another embodiment, the trained scoring algorithm is implemented as a ResNet neural network.

[0040] In another embodiment, the trained scoring algorithm is implemented as a convolutional neural network.

[0041] In another embodiment, the trained scoring algorithm is implemented as a U-Net neural network.

[0042] In another aspect, the present invention provides a computer program configured for execution by a computing system. Execution of the machine executable instructions causes the computing system to repeatedly generate a random input vector. Execution of the machine executable instructions also causes the computing system to repeatedly receive a generated MRF pulse sequence in response to inputting the random input vector into a GAN generator neural network. The GAN generator neural network is configured to output a generated MRF pulse sequence in response to receiving the random input vector.

[0043] Execution of the machine-executable instructions further causes the computing system to append the generated MRF pulse sequence to the MRF pulse sequence database.

[0044] Execution of the machine executable instructions also causes the computing system to input each generated MRF pulse sequence in the MRF pulse sequence database into a training scoring algorithm to assign one or more score values ​​to each generated MRF pulse sequence. The trained scoring algorithm is configured to output one or more score values ​​in response to receiving the generated MRF pulse sequence as input. Execution of the machine executable instructions also causes the computing system to receive a selected MRF pulse sequence from the MRF pulse sequence database by applying a predetermined criterion to one or more score values ​​of each generated MRF pulse sequence in the MRF pulse sequence database.

[0045] In another aspect, the present invention provides a method. The method includes repeatedly generating a random input vector. The method also includes repeatedly receiving a generated MRF pulse sequence in response to inputting the random input vector into a GAN generator neural network. The GAN generator neural network is configured to output the generated MRF pulse sequence in response to receiving the random input vector. The method also includes repeatedly appending the generated MRF pulse sequence to an MRF pulse sequence database.

[0046] The method also includes inputting each generated MRF pulse sequence in the MRF pulse sequence database into a trained scoring algorithm to assign one or more score values ​​to each generated MRF pulse sequence. The trained scoring algorithm is configured to output one or more score values ​​in response to receiving the generated MRF pulse sequence as input. The method also includes receiving a selected MRF pulse sequence from the MRF pulse sequence database by applying a predetermined criterion to one or more score values ​​of each generated MRF pulse sequence in the MRF pulse sequence database.

[0047] It should be understood that one or more of the aforementioned embodiments of the present invention may be combined as long as the combined embodiments are not mutually exclusive.

[0048] As will be appreciated by those skilled in the art, aspects of the present invention may be embodied as an apparatus, method, or computer program product. Thus, 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, which may all be collectively referred to herein 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 media having computer executable code embodied thereon.

[0049] Any combination of one or more computer-readable media can be utilized. Computer-readable media can be computer-readable signal media or computer-readable storage media. As used herein, "computer-readable storage media" covers any tangible storage media that can store instructions that can be executed by a processor or computing system of a computing device. Computer-readable storage media can be referred to as computer-readable non-temporary storage media. Computer-readable storage media can also be referred to as tangible computer-readable media. In some embodiments, computer-readable storage media can also be capable of storing data that can be accessed 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 hard disks, flash memories, USB thumb drives, random access memories (RAM), read-only memories (ROM), optical disks, magneto-optical disks, and register files of computing systems. Examples of optical disks include compact disks (CDs) and digital versatile disks (DVDs), such as CD-ROMs, CD-RWs, CD-Rs, DVD-ROMs, DVD-RWs, or DVD-R disks. The term computer-readable storage media also refers to various types of recording media that can be accessed by a computer device via a network or communication link. For example, data can be retrieved via a modem, via the Internet, or via a local area network. Computer executable code embodied on a computer readable medium may be transmitted using any appropriate medium, including but not limited to wireless, wireline, optical fiber cable, RF, etc., or any suitable combination of the foregoing.

[0050] A computer readable signal medium may include a propagated data signal in which a computer executable code is embedded, for example, in baseband or as part of a carrier wave. 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 that can convey, propagate, or transport a program for use by or in conjunction with an instruction execution system, device, or apparatus.

[0051] "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 another 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.

[0052] As used herein, "computing system" encompasses an electronic component capable of executing a program or machine executable instructions or computer executable code. References to computing systems including examples of "computing system" should be interpreted as possibly including more than one computing system or processing core. For example, a computing system may 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 possibly referring to a collection or network of computing devices that each include a processor or computing system. Machine executable code or instructions may be executed by multiple computing systems or processors located within the same computing device or even distributed across multiple computing devices.

[0053] Machine executable instructions or computer executable code may include instructions or programs that cause a processor or other computing system to perform aspects of the present invention. Computer executable code for performing operations of various 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 "C" programming language or similar programming languages ​​and compiled into machine executable instructions. In some cases, the computer executable code may be in the form of a high-level language or precompiled form, and used in conjunction with an interpreter that generates machine executable instructions on the fly. In other cases, the machine executable instructions or computer executable code may be a programming form of a programmable logic gate array.

[0054] The computer executable code may execute entirely on the user's computer, partially on the user's computer, as a stand-alone software package, partially on the user's computer and partially on 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 may be connected to an external computer (e.g., through the Internet using an Internet service provider).

[0055] Various aspects of the present invention are described with reference to the flow chart, diagram and / or block diagram of the method, device (system) and computer program product according to an embodiment of the present invention. It should be understood that each block or a part of a block in the flow chart, diagram and / or block diagram can be implemented by a computer program instruction in the form of computer executable code when applicable. It should also be understood that when not mutually exclusive, the combination of the boxes in different flow charts, diagrams and / or block diagrams can be combined. These computer program instructions can be provided to a computing system of a general-purpose computer, a special-purpose computer or other programmable data processing device to produce a machine, so that the instructions executed by the computing system of a computer or other programmable data processing device create a unit for realizing the function / action specified in the flow chart and / or block diagram block.

[0056] These machine-executable instructions or computer program instructions may also be stored in a computer-readable medium, which may direct a computer, other programmable data processing apparatus, or other device to operate in a specific manner so that the instructions stored in the computer-readable medium produce an article of manufacture including instructions for implementing the functions / actions specified in the flowchart and / or block diagram blocks.

[0057] Machine-executable instructions or computer program instructions may also be loaded onto a computer, other programmable data processing apparatus, or other device, so that a series of operational steps are performed on the computer, other programmable apparatus, or other device to produce a computer-implemented process, so that the instructions executed on the computer or other programmable apparatus provide a process for implementing the functions / actions specified in the flowchart and / or block diagram blocks.

[0058] As used herein, "user interface" is an interface that allows a user or operator to interact with a computer or computer system. "User interface" may also be referred to as a "human-machine interface device". A user interface may provide information or data to an operator and / or receive information or data from an operator. A user interface may enable the input from an operator to be received by a computer, and may provide output to a user from a computer. In other words, a user interface may allow an operator to control or manipulate a computer, and an interface may allow a computer to indicate the effect of the control or manipulation of an operator. Displaying data or information on a display or graphical user interface is an example of providing information to an operator. All examples of receiving data by a keyboard, mouse, trackball, touchpad, pointing stick, graphic input board, joystick, game controller, webcam, headset, pedal, wired gloves, remote controller, and accelerometer are examples of user interface components that enable receiving information or data from an operator.

[0059] As used herein, "hardware interface" encompasses an interface that enables a computing system of a computer system to interact with and / or control an external computing device and / or apparatus. A hardware interface may allow a computing system to send control signals or instructions to an external computing device and / or apparatus. A hardware interface may also enable a computing system to exchange data with an external computing device and / or apparatus. 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 LAN connection, a TCP / IP connection, an Ethernet connection, a control voltage interface, a MIDI interface, an analog input interface, and a digital input interface.

[0060] As used herein, a "display" or "display device" encompasses an output device or user interface suitable for displaying images or data. A display may output visual, audio, and / or tactile data. Examples of displays include, but are not limited to: a computer monitor, a television screen, a touch screen, a tactile electronic display, a Braille screen.

[0061] Cathode ray tubes (CRTs), memory tubes, bi-stable displays, electronic paper, vectorscope displays, flat panel displays, vacuum fluorescent displays (VF), 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.

[0062] K-space data is defined herein as recorded measurements of radio frequency signals emitted by atomic spins using the antenna of a magnetic resonance apparatus during a magnetic resonance imaging scan. Magnetic resonance data is an example of tomographic medical image data.

[0063] A Magnetic Resonance Imaging (MRI) image or MR image is defined herein as a reconstructed two-dimensional or three-dimensional visualization of anatomical data contained within the magnetic resonance imaging data. The visualization may be performed using a computer. BRIEF DESCRIPTION OF THE DRAWINGS

[0064] In the following preferred embodiments, preferred embodiments of the present invention will be described by way of example only and with reference to the accompanying drawings, in which:

[0065] Figure 1 An example of a medical system is shown;

[0066] Figure 2 Shows instructions for use Figure 1 A flowchart of a method for a medical system;

[0067] Figure 3 Another example of a medical system is shown;

[0068] Figure 4 Shows instructions for use Figure 3 A flowchart of a method for a medical system;

[0069] Figure 5 A flow chart illustrating the training of a trained scoring algorithm is shown;

[0070] Figure 6 shows the training of the GAN generator neural network;

[0071] Figure 7 shows one use of a GAN generator neural network;

[0072] Figure 8 shows the expansion of the trained scoring algorithm; and

[0073] Fig. 9 Shown is a modification of the GAN generator neural network. DETAILED DESCRIPTION

[0074] Elements with the same number in these figures are equivalent elements or perform the same function. Elements that have been discussed previously will not necessarily be discussed in later figures if the function is equivalent.

[0075] Figure 1 An example of a medical system 100 is shown. The medical system 100 includes a computer 102 having a computing system 104. The computer 102 can represent one or more computers at one or more locations. The computing system 104 can also represent one or more computing systems, such as one or more computing cores. The computing system 104 is shown in communication with an optional hardware interface 106 and an optional user interface 108. The hardware interface 106 can enable the computing system 104 to communicate with other components of the medical system 100 (if present) and send control and receive data from these components. The optional user interface 108 can also enable an operator to control and operate the medical system 100.

[0076] The computing system 104 is shown as further communicating with a memory 110. The memory 110 is intended to represent various types of memory that the computing system 104 can access. The memory 110 can be a non-transitory storage medium. The memory 110 is shown as containing machine executable instructions 120. The machine executable instructions 120 enable the computing system 104 to perform basic tasks, such as numerical calculations and / or image processing. The machine executable instructions 120 can also enable the computing system 104 to control other components or devices via the hardware interface 106. The memory 110 is also shown as containing a GAN generator neural network 122 and a trained scoring algorithm 124. The memory 110 is further shown as containing a random input vector 126 and a generated MRF pulse sequence 128.

[0077] The random input vector 126 can be input into the GAN generator neural network 122 to generate or output a generated MRF pulse sequence 128. By generating a large number of random input vectors 126, a large number of generated MRF pulse sequences 128 can be created. The memory 110 is further shown as containing an MRF pulse sequence database 130. For example, a large number of generated MRF pulse sequences 128 can then be stored in the MRF pulse sequence database 130. The memory 110 is further shown as containing one or more score values ​​132 generated for the MRF pulse sequence 128 by inputting the generated MRF pulse sequence 128 into the trained scoring algorithm 124. For each generated MRF pulse sequence in the generated MRF pulse sequence 128 in the MRF pulse sequence database 130, one or more score values ​​132 can be generated. This can be done at different times. For example, after the generated MRF pulse sequence 128 is generated by the GAN generator neural network 122, one or more score values ​​132 can be generated and then stored in the MRF pulse sequence database 130 at the same time. In other examples, one or more score values ​​132 may be generated on the fly when searching the MRF pulse sequence database 130. In both cases, there may be a predetermined criterion 134 for selecting a selected MRF pulse sequence 136 from the MRF pulse sequence database 130. The predetermined criterion 134 may be a means of evaluating the one or more score values ​​132 of each of the MRF pulse sequence databases 130.

[0078] Figure 2 Shows the instructions for operation Figure 1Flowchart of a method of a medical system 100 of the present invention. First, in step 200, a random input vector 126 is generated. Next, in step 202, in response to inputting the random input vector 126 into the GAN generator neural network 122, a generated MRF pulse sequence 128 is received. In step 204, the generated MRF pulse sequence 128 is appended to the MRF pulse sequence database 130. In this example, one or more score values ​​132 have not yet been generated and placed in the MRF pulse sequence database 130. However, at this time, after performing step 202, the generation of one or more score values ​​132 for a particular generated MRF pulse sequence 128 can be calculated by inputting the generated MRF pulse sequence 128 into the trained scoring algorithm 124. In this case, when the generated MRF pulse sequence 128 is appended to the MRF pulse sequence database 130, one or more score values ​​132 can be stored and associated with the record of the generated MRF pulse sequence 128 at the same time.

[0079] The method then proceeds to box 206, which is a decision box. In this case, the question is "Has the algorithm finished generating the MRF pulse sequence?". If the answer is no, the method returns to step 200, and another random input vector 126 is generated, and the process is repeated. If the answer is yes, the method proceeds to box 208. In box 208, each generated MRF pulse sequence in the MRF pulse sequence database is input into the trained scoring algorithm to assign one or more score values ​​to each generated MRF pulse sequence. Then finally, in step 210, a selected MRF pulse sequence 136 is received from the MRF pulse sequence database 130 by applying predetermined criteria to one or more score values ​​132 of each generated MRF pulse sequence 128 in the MRF pulse sequence database 130.

[0080] Figure 3 Another example of a medical system 300 is shown. In addition to including a magnetic resonance imaging system 302 controlled by the computing system 104, Figure 3 The medical system 300 depicted in Figure 1 The medical system 100 depicted in FIG.

[0081] The magnetic resonance imaging system 302 comprises a magnet 304. The magnet 304 is a superconducting cylindrical magnet having a hole 306 running through it. It is also possible to use different types of magnets; 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 is divided into two parts to allow access to the isoplane of the magnet, for example, such a magnet can be used in conjunction with charged particle beam therapy. An open magnet has two magnet parts, one above the other, with a space in between that is large enough to accommodate a subject: the arrangement of the two part areas is similar to that of a Helmholtz coil. Open magnets are popular because the subject is less constrained. Inside the cryostat of the cylindrical magnet, there is a collection of superconducting coils.

[0082] Within the bore 306 of the cylindrical magnet 304 there is an imaging zone 308 where the magnetic field is strong enough and uniform enough to perform magnetic resonance imaging. A field of view 309 is shown within the imaging zone 308. K-space data is typically acquired for the field of view 309. The region of interest may be the same as the field of view 309, or it may be a sub-volume of the field of view 309. A subject 318 is shown supported by a subject support 320 so that at least a portion of the subject 318 is within the imaging zone 308 and the field of view 309.

[0083] There is also a collection of magnetic field gradient coils 310 within the bore 306 of the magnet, which is used to acquire preliminary k-space data to spatially encode magnetic spins within the imaging region 308 of the magnet 304. The magnetic field gradient coils 310 are connected to a magnetic field gradient coil power supply 312. The magnetic field gradient coils 310 are intended to be representative. Typically, the magnetic field gradient coils 310 contain three separate coil groups 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 ramped or pulsed.

[0084] Adjacent to the imaging zone 308 is a radio frequency coil 314 for manipulating the orientation of magnetic spins within the imaging zone 308 and for receiving radio transmissions from spins also within the imaging zone 308. The radio frequency antenna may include multiple coil elements. The radio frequency antenna may also be 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 by separate transmitting coils and receiving coils and separate transmitters and receivers. It should be understood that the radio frequency coil 314 and the radio frequency transceiver 316 are representative. The radio frequency coil 314 is also intended to represent a dedicated transmitting antenna and a dedicated receiving antenna. Similarly, the transceiver 316 may also represent a separate transmitter and receiver. The radio frequency coil 314 may also have multiple receiving / transmitting elements, and the radio frequency transceiver 316 may have multiple receiving / transmitting channels.

[0085] The transceiver 316 and gradient controller 312 are shown connected to the hardware interface 106 of the computer system 102. Both of these components, as well as other components (eg, subject support providing position data), can provide sensor data 126.

[0086] The memory 110 is also shown as containing a Bloch simulator. The memory 110 is further shown as containing an MRF dictionary 332. The MRF dictionary 332 can be constructed by controlling the Bloch simulator 330 using a selected MRF pulse sequence 136. The memory 110 is further shown as containing MRF k-space data 334 that has been acquired by controlling the magnetic resonance imaging system 302 using a selected MRF pulse sequence 136. The memory 110 is also shown as containing an MRF image 336 that has been reconstructed from the MRF k-space data 334 using the MRF dictionary 332.

[0087] Figure 4 Shows the instructions for operation Figure 3 Flow chart of a method of a medical system 300 . Figure 4 The method described in is similar to Figure 2 The method described in , wherein additional steps are performed. Figure 4 The methods described in include Figure 2 . After step 210, step 400 is additionally performed. In step 400, the MRF dictionary 332 is calculated by controlling the Bloch simulator 330 using the selected MRF pulse sequence 136. Next, in step 402, the MRF k-space data 334 is acquired by controlling the magnetic resonance imaging system 302 using the selected MRF pulse sequence 136. Finally, in step 404, one or more MRF images 336 are reconstructed using the MRF k-space data 334 and the MRF dictionary 332.

[0088] MR fingerprinting (MRF) is a relatively new method for quantitative MRI. By adopting a series of RF pulses with different properties (flip angle, phase, timing, damage, etc.), multiple quantitative parameters are encoded simultaneously in the transient signal evolution. MRF sequences usually consist of hundreds of RF pulses, resulting in a huge parameter space of hundreds or even thousands of dimensions. Therefore, finding an optimized MRF sequence is a difficult task.

[0089] Generative Adversarial Networks (GANs) are relatively new AI techniques that are very good at synthesizing realistic new datasets from a given set. There are some impressive examples on the web showing examples of cartoon characters and photo generation.

[0090] GANs consist of a generator and a discriminator network. The generator generates data based on what the discriminator labels as real or not real. The discriminator labels data as correct or incorrect based on learning from real data. The training of the two networks alternates so that they each gradually get better at generation and discrimination, and (ideally) get better at the same rate.

[0091] Examples may facilitate synthesizing optimized MRF sequences from a collection of known MRF sequences by using a generative adversarial network (GAN).

[0092] Examples may contain one or more of the following characteristics:

[0093] - A collection of known useful MR fingerprinting sequence parameters.

[0094] -GAN synthesizes new sequences.

[0095] -AI algorithm scores the quality of fingerprint recognition sequences.

[0096] - Method for selecting MRF sequences synthesized by GAN based on AI prediction of quality scores

[0097] Example GANs can be used to synthesize optimized fingerprint sequences from a set of known good fingerprint sequences. It is necessary to define a measure of the quality of the MRF sequence. Typically, such a measure will consider how to distinguish different regions in the quantitative parameter space. The exact definition of the quality measure is not important for this implementation.

[0098] The sequences synthesized by the GAN will qualify as suitable fingerprinting sequences by design. The question is whether they are also good ones. In order to judge how good the generated sequences are, we propose to implement another AI network that is trained to score the sequences. This second network can be trained from already existing fingerprint sequences that have already been scored.

[0099] Using this framework, a virtually unlimited number of sequences can be generated and scored. The best scored sequences can be tried out in practice. If they prove to be good, they can again be integrated into the discriminator training dataset.

[0100] Scoring MRF sequences can be a very computationally intensive task. Typically, a large number of signals for different parameters (T1, T2, B1, B0, etc.) need to be calculated in order to determine the coding efficiency in different regions of the parameter space. If a large number of MRF sequences should be evaluated, the process can take considerable time.

[0101] Examples may use an AI algorithm (e.g., neural network, trained scoring algorithm 124) for quality scoring (providing one or more score values ​​132). In this way, scoring may be achieved more quickly to enable a fast search through a large number of MRF sequences to find the optimum.

[0102] The training process of the algorithm is as follows Figure 5 As shown. For multiple MRF sequence definitions, quality scores are calculated. The AI ​​algorithm is trained on the MRF sequence definitions with features and corresponding quality scores as labels. After training, the AI ​​algorithm will be able to predict the quality scores of unknown MRF sequence definitions.

[0103] Figure 5 A flow chart illustrating the training of the trained scoring algorithm 124, also referred to as the quality scoring AI algorithm, is shown. Using supervised learning, the MRF sequence definition 500 is input into the quality score calculation 502 and into the quality scoring algorithm 124. The results of the quality score calculation 502 are also input into the quality scoring AI algorithm 124. The MRF sequence definition 500 provides features and the quality score calculation 502 provides labels for supervised learning. A pulse sequence is essentially a timing diagram. The MRF sequence definition 500 can be used to define the parameters of a timing sequence for performing MRF acquisition.

[0104] Figure 6 The training of the GAN generator neural network 122, or also simply referred to as the GAN generator, is shown. The GAN generator 122 is paired with a GAN discriminator 600. The noise vector 126 is input into the GAN generator 122, and this generates a generated MRF pulse sequence 128. The GAN discriminator 600 is trained to learn the difference between the MRF sequence definition 500 and the generated MRF pulse sequence 128. By alternately training the GAN generator 122 and the GAN discriminator 600, they both perform better and better until the GAN generator 122 is able to generate a suitable MRF pulse sequence.

[0105] Generative Adversarial Networks (GANs) are trained to generate MRF sequences that are similar to known MRF sequences with good encoding quality. GANs consist of a discriminator network and a generator network. These two are trained in an iterative process, such as Figure 6 As shown in Figure 2. The generator is fed noise to create a random sequence definition. The discriminator is trained to distinguish such "fake" MRF sequences from the "real" MRF sequences provided for training. By iteratively optimizing the weights of the two networks, the generator learns to generate MRF sequences with properties similar to those known.

[0106] Figure 7One use of a GAN generator neural network 122 during deployment is shown. In this flow chart, a random input vector 126 is input into the GAN generator 122. This then provides a generated MRF pulse sequence 128 that is input into a trained scoring algorithm 124. The trained scoring algorithm 124 then provides one or more score values ​​132. For example, the one or more score values ​​132 can be used as a predetermined criterion 134 to select whether to use the generated MRF pulse sequence 128.

[0107] Then in Figure 7 The trained networks (GAN generator and quality scoring algorithm) are used in the indicated sequential setup. In the first step, the GAN generator creates a new MRF sequence from the noisy input. In the second step, the quality scoring AI algorithm predicts the quality score of the sequence. After repeating this process multiple times, the MRF sequence with the best quality score is selected as the optimized MRF sequence.

[0108] Figure 8 Shows Figure 4 An extension of the trained scoring algorithm 124 is shown. Figure 7 In the example shown, additionally one or more control switches 800 are input as additional features. One or more control switches 800 are used to encode things such as the anatomical region being imaged, details about the subject, and even possible clinical questions. This enables the trained scoring algorithm 124 to be configured or tuned to specific acquisition conditions.

[0109] Fig. 9 Shows Figure 5 1. In this example, in addition to the random input vector 126, one or more control switches 700 are input into the GAN generator. Adding one or more control switches 800 to the GAN generator 122 during training enables the behavior of the GAN generator to be controlled or configured. As previously described, one or more control switches 800 can be used to encode things such as the anatomical structure being imaged, the clinical question, or details about the subject being imaged.

[0110] 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; the invention is not limited to the disclosed embodiments.

[0111] From a study of the drawings, the present disclosure, and the appended claims, those skilled in the art can 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 indefinite article "a" or "an" does not exclude a plurality. A single processor or other unit can implement the functions of several items recited in the claims. The mere fact that specific measures are described in mutually different dependent claims does not indicate that the combination of these measures cannot be fully utilized. The computer program may be stored / distributed on a suitable medium, such as an optical storage medium or solid-state medium 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. Any reference numerals in the claims should not be interpreted as limiting the scope.

[0112] Reference numerals list

[0113] 100 Medical System

[0114] 102 Computer

[0115] 104 Computing Systems

[0116] 106 Hardware Interface

[0117] 108 User Interface

[0118] 110 Memory

[0119] 120 machine executable instructions

[0120] 122 GAN generator neural network

[0121] 124 Trained scoring algorithm

[0122] 126 Random Input Vector

[0123] 128 generated MRF pulse sequences

[0124] 130 MRF pulse sequence database

[0125] 132 One or more fractional values

[0126] 134 Predetermined Standards

[0127] 136 Selected MRF pulse sequences

[0128] 200 Generate random input vector

[0129] 202 In response to inputting the random input vector into the GAN generator neural network, receiving the generated MRF pulse sequence

[0130] 204 Append the generated MRF pulse sequence to the MRF pulse sequence database

[0131] 208 Input each generated MRF pulse sequence in the MRF pulse sequence database into the trained scoring algorithm to assign one or more score values ​​to each generated MRF pulse sequence

[0132] 210 receiving a selected MRF pulse sequence from an MRF pulse sequence database by applying a predetermined criterion to one or more score values ​​of each generated MRF pulse sequence in the MRF pulse sequence database

[0133] 300 Medical System

[0134] 302 Magnetic Resonance Imaging System

[0135] 304 Magnet

[0136] 306 Magnet Hole

[0137] 308 Imaging Area

[0138] 309 Field of View

[0139] 310 Magnetic Field Gradient Coil

[0140] 312 Magnetic Field Gradient Coil Power Supply

[0141] 314 RF Coil

[0142] 316 Transceiver

[0143] 318 Main Body

[0144] 320 Subject Support

[0145] 330Bloch Emulator

[0146] 332MRF dictionary

[0147] 334MRF k-space data

[0148] 336MRF image

[0149] 400 Calculate an MRF dictionary with predetermined dictionary entries for a selected MRF pulse sequence by performing a Bloch equation simulation controlled by a selected MRF pulse sequence for each of the predetermined dictionary entries.

[0150] 402 Acquiring (402) MRF k-space data by controlling a magnetic resonance imaging system

[0151] 404 Reconstruct (404) one or more MRF images using the MRF k-space data and the MRF dictionary

[0152] 500 MRF sequence definition

[0153] 502 Quality Score Calculation

[0154] 600 GAN Discriminator

[0155] 800 One or more control switches

Claims

1. A medical system (100, 300) comprising: a memory (110) storing machine executable instructions (120), a GAN generator neural network (122), and a trained scoring algorithm (124), wherein the GAN generator neural network is configured to output a generated MRF pulse sequence (128) in response to receiving a random input vector (126), wherein the trained scoring algorithm is configured to output one or more score values ​​(132) in response to receiving the generated MRF pulse sequence as input; - A computing system (104), wherein execution of the machine executable instructions causes the computing system to repeatedly perform the following operations: - generating (200) said random input vector; - receiving (202) the generated MRF pulse sequence in response to inputting the random input vector into the GAN generator neural network; and - appending (204) the generated MRF pulse sequence to the MRF pulse sequence database (130); The execution of the machine executable instructions causes the computing system to perform the following operations: - inputting (208) each generated MRF pulse sequence in the MRF pulse sequence database into the trained scoring algorithm to assign the one or more score values ​​to each generated MRF pulse sequence; and - receiving (210) a selected MRF pulse sequence (136) from the MRF pulse sequence database by applying a predetermined criterion (134) to the one or more score values ​​of each generated MRF pulse sequence in the MRF pulse sequence database.

2. The medical system according to claim 1, wherein: Execution of the machine executable instructions also causes the computing system to compute (400) an MRF dictionary (332) having the predetermined dictionary entries for the selected MRF pulse sequence by performing a Bloch equation simulation controlled by the selected MRF pulse sequence for each of the predetermined dictionary entries.

3. The medical system according to claim 2, wherein: The medical system also includes a magnetic resonance imaging system (302), wherein execution of the machine executable instructions further causes the computing system to perform the following operations: - acquiring (402) MRF k-space data (334) by controlling the magnetic resonance imaging system using the selected MRF pulse sequence, and - reconstructing (404) one or more MRF images (336) using the MRF k-space data and the MRF dictionary.

4. A medical system according to any one of the preceding claims, wherein: The GAN generator neural network is further configured to receive one or more control switches (800) as input, wherein execution of the machine executable instructions further causes the computing system to perform the following operations: - receiving said one or more control switches; - inputting the one or more control switches into the GAN generator neural network using the random input vector; and - storing the one or more control switches and the generated MRF pulse sequence in the MRF pulse sequence database, wherein the selected MRF pulse sequence is selected at least in part using a selected set of control switches.

5. The medical system according to claim 4, wherein: The one or more control switches include any of: anatomical region, clinical question, timing limit, breath hold duration, and combinations thereof.

6. A medical system according to any one of the preceding claims, wherein: Selecting the selected MRF pulse sequence from the MRF pulse sequence database by applying a predetermined criterion to the one or more score values ​​of each generated MRF pulse sequence in the MRF pulse sequence database comprises performing the following operations: - calculating a composite score using the one or more score values; as well as - Selecting the generated MRF pulse sequence with the highest composite score as the selected MRF pulse sequence.

7. The medical system according to claim 6, wherein: Execution of the machine executable instructions further causes the computing system to perform the following operations: - receiving a user input, wherein the user input specifies a composition of the composite fraction, and wherein the selected MRF pulse sequence is selected in response to receiving the user input.

8. The medical system according to any one of claims 1 to 5, wherein: Selecting the selected MRF pulse sequence from the MRF pulse sequence database comprises performing the following operations: - selecting a subset of MRF pulse sequences that meet the screening criteria from the MRF pulse sequence database; - calculating a set of numerically calculated score values ​​for each MRF pulse sequence in the subset of MRF pulse sequences; - selecting the selected MRF pulse sequence from the subset of MRF pulse sequences by comparing the predetermined criterion with a set of numerically calculated score values ​​for each MRF pulse sequence in the subset of MRF pulse sequences.

9. A medical system according to any one of the preceding claims, wherein: The one or more score values ​​include any of: a T1 accuracy score, a T2 accuracy score, a T2-star accuracy score, a proton density accuracy score, and combinations thereof.

10. A medical system according to any one of the preceding claims, wherein: Execution of the machine executable instructions further causes the computing system to perform the following operations: - receiving a training MRF pulse sequence; and - Training the GAN generator neural network and the GAN discriminator neural network using the training MRF spike sequence according to a GAN training protocol (600).

11. A medical system according to any one of the preceding claims, wherein: Execution of the machine executable instructions further causes the computing system to perform the following operations: - receiving a previous MRF pulse sequence; - calculating a set of score values ​​for numerical evaluation of each of said one or more score values ​​of said previous MRF pulse sequence; as well as - training said trained scoring algorithm using said previous MRF pulse sequences and said set of numerically evaluated score values.

12. A medical system according to any one of the preceding claims, wherein: The trained scoring algorithm is implemented as any one of the following: a decision tree, a support vector machine, a multilayer perceptron neural network, a ResNET neural network, a convolutional neural network, and a U-Net neural network.

13. A computer program comprising machine executable instructions (120) for execution by a computing system (104), wherein: Execution of the machine executable instructions causes the computing system to repeatedly perform the following operations: - generating (200) a random input vector (126); - receiving (202) a generated MRF pulse sequence in response to inputting the random input vector into a GAN generator neural network (122), wherein the GAN generator neural network is configured to output the generated MRF pulse sequence in response to receiving the random input vector; and - appending (204) the generated MRF pulse sequence to the MRF pulse sequence database (130); The execution of the machine executable instructions causes the computing system to perform the following operations: - inputting (208) each generated MRF pulse sequence in the MRF pulse sequence database into a trained scoring algorithm (124) to assign one or more score values ​​(132) to each generated MRF pulse sequence, wherein the trained scoring algorithm is configured to output the one or more score values ​​in response to receiving the generated MRF pulse sequence as input; and - receiving (210) a selected MRF pulse sequence (136) from the MRF pulse sequence database by applying a predetermined criterion (134) to the one or more score values ​​of each generated MRF pulse sequence in the MRF pulse sequence database.

14. A method, wherein: The method comprises repeatedly performing the following operations: - generating (200) a random input vector (126); - receiving (202) a generated MRF pulse sequence (128) in response to inputting the random input vector into a GAN generator neural network (122), wherein the GAN generator neural network is configured to output the generated MRF pulse sequence in response to receiving the random input vector; and - appending (204) the generated MRF pulse sequence to a MRF pulse sequence database; The method further includes the following operations: - inputting (208) each generated MRF pulse sequence in the MRF pulse sequence database into a trained scoring algorithm (124) to assign one or more score values ​​(132) to each generated MRF pulse sequence, wherein the trained scoring algorithm is configured to output the one or more score values ​​in response to receiving the generated MRF pulse sequence as input; and - receiving (210) a selected MRF pulse sequence (136) from the MRF pulse sequence database by applying a predetermined criterion (134) to the one or more score values ​​of each generated MRF pulse sequence in the MRF pulse sequence database.

15. A method for computer-implemented training of a scoring algorithm (124) for a GAN generator neural network (122), the GAN generator neural network being configured to output a generated MRF pulse train (128) in response to receiving a random input vector (126), and wherein: The trained scoring algorithm is configured to output one or more score values ​​(132) in response to receiving the generated MRF pulse sequence as input to train the GAN generator neural network and the GAN discriminator neural network (600) using the training MRF pulse sequence according to a GAN training protocol.

Images