Magnetic resonance image denoising

Abstract

A method for magnetic resonance image reconstruction is disclosed herein. The method comprises: receiving sub-sampled k-space data that has been acquired for imaging an object; reconstructing at least one aliasing image from the sub-sampled k-space data using a first magnetic resonance image reconstruction method; denoising is carried out on the at least one aliasing image; deriving k-space data from the at least one denoised aliasing image using a Fourier transform, the derived k-space data referred to as denoised k-space data; a magnetic resonance image of the subject is reconstructed from the denoised k-space data using a second magnetic resonance image reconstruction method, the second magnetic resonance image reconstruction method comprising the first magnetic resonance image reconstruction method and additional processing operations.

Description

Technical Field

[0001] This invention relates to magnetic resonance imaging, and more particularly to denoising magnetic resonance (MR) images. Background Technology

[0002] The conventional magnetic resonance imaging (MRI) image reconstruction process is a pipeline that includes processing steps such as Fourier transform, noise decorrelation, filtering, and image interpolation. The resulting reconstructed image can exhibit spatially varying noise levels, depending on the sensitivity of the receiving coil elements and the sampling mode of the acquisition. The reconstructed image is also characterized by local and global noise correlations, especially when acquisition is accelerated through subsampling. Both of these phenomena complicate denoising because they are either accurately modeled during training, in which case denoising is tailored to a specific acquisition and loses broader applicability; or they are broadly covered during training, in which case denoising suffers suboptimal performance due to overgeneralization.

[0003] US Patent Application US2020 / 0278410 discloses a method for acquiring subsampled k-space data and reconstructing it, including a denoising step. This denoising step operates in wavelet space (i.e., on subsampled k-space data with sparse transform). Summary of the Invention

[0004] The present invention provides medical systems, computer programs, and methods in the independent claims. Embodiments are given in the dependent claims.

[0005] The embodiment can perform denoising trained to remove white noise from aliased images, each of which is directly reconstructed using Fourier transform based on subsampled k-space data acquired using a single receiving coil element. Because the noise levels in these aliased images are uniform and the noise is uncorrelated, the loss in terms of wider applicability or optimal performance experienced with dedicated or general denoising methods applied to conventional reconstructed images can be avoided.

[0006] In one aspect, the present invention provides a medical system for magnetic resonance image reconstruction. The medical system includes: a memory storing machine-executable instructions; and a computing system, wherein execution of the machine-executable instructions causes the computing system to: receive subsampled k-space data acquired for imaging an object; reconstruct at least one aliased image based on the subsampled k-space data using a first magnetic resonance image reconstruction method; denoise the at least one aliased image; derive k-space data from the at least one denoised aliased image using a Fourier transform, the derived k-space data being referred to as denoised k-space data; and reconstruct a magnetic resonance image of the object based on the denoised k-space data using a second magnetic resonance image reconstruction method, the second magnetic resonance image reconstruction method including the first magnetic resonance image reconstruction method and additional processing operations. The terms "processing step" and "processing operation" are used interchangeably herein.

[0007] This invention relates to the problem of denoising in magnetic resonance images reconstructed from k-space data acquired based on subsampling. Since subsampling in k-space aliasing occurs in image space, the associated unwrapping of aliasing in image space also affects the noise characteristics of the unwrapped reconstructed magnetic resonance image. The insight of this invention is to apply denoising at an early stage of the image reconstruction process, particularly (e.g., by performing a Fourier transform from the subsampled k-space data) applying denoising to the (directly) reconstructed aliased image. In practice, multiple subsampled k-space datasets may exist, acquired by corresponding RF receiver antennas (coils (elements)) with corresponding spatial sensitivity distributions. These multiple subsampled k-space datasets can be reconstructed into multiple aliased images (aliased coil images) for the corresponding RF antennas. According to the invention, denoising is applied to (one or more) aliased images to primarily remove irrelevant noise while largely preserving the aliased signal. For example, this can be implemented as follows: denoising primarily removes white noise from (one or more) aliased images. Subsequently, the denoised aliased image is inversely transformed into denoised aliased (subsampled) k-space data. Thus, the final (unwound) magnetic resonance image was reconstructed using a parallel imaging reconstruction method and spatial sensitivity distribution. Because denoising is performed before the dealiasing / unwound process, the noise component of the denoising input is unaffected by the dealiasing / unwound process. Therefore, denoising does not need to be modeled according to the k-space sampling mode and can be directed towards the removal of, for example, white noise.

[0008] This subject matter can provide magnetic resonance images of the volume of an object. The volume can be, for example, the shoulder, heart, brain, or any other part of the object that can be imaged by a magnetic resonance imaging system. In one example, the magnetic resonance imaging system can be located remotely from the medical system. This allows for flexible implementation of this subject matter. For example, the medical system can be configured to perform this method using k-space data acquired by multiple remote magnetic resonance imaging systems (e.g., in a hospital). Alternatively, the medical system may include a magnetic resonance imaging system. This allows for, for example, local processing of the acquired k-space data during a scan of the object.

[0009] Volumetric magnetic resonance images can be obtained by applying specific acquisition techniques and image reconstruction schemes. Acquisition techniques can be performed based on spatial resolution. Spatial resolution can be defined by the field of view (FOV) and the matrix size. The matrix size can be the number of frequency encoding steps and the number of phase encoding steps. The field of view can define the size of the area covered by the matrix. Therefore, the acquisition technique can define the size of the k-space, the initial number of samples in the k-space, and the sampling pattern of the samples in the k-space.

[0010] Acquisition techniques can use pulse sequences to fill samples in k-space. K-space can be partially filled by acquiring fewer samples than the initial number. In this case, the acquired k-space data can be referred to as undersampled k-space data or subsampled k-space data. This saves scan time while maintaining spatial resolution. K-space subsampling can be performed in any coding direction (e.g., partition coding direction). In one example, k-space subsampling can be performed in the phase coding direction by reducing the number of k-space lines. However, the missing information from the uncollected k-space samples can lead to image aliasing due to the reduced desired field of view. In fact, if the Nyquist criterion is not satisfied, the field of view becomes smaller than the volume, resulting in aliased image outcomes. For example, if the volume is larger than the reduced field of view in the phase direction, folding artifacts may occur.

[0011] In one example, the received subsampled k-space data can be k-space data acquired from one or more slices of a volume. The subsampled k-space data can represent the number of slices S, where S is greater than or equal to 1. The received subsampled k-space data can be acquired by one or more receiving coils of a magnetic resonance imaging system. The received subsampled k-space data can be acquired by R receiving coils, where R is greater than or equal to 1. Therefore, the structure of the received subsampled k-space data can be provided based on the acquisition technique and the number of receiving coils. The receiving coil can also be referred to as a receiving coil element, receiving coil, or radio frequency receiving coil. In practice, radio frequency radiation can be emitted into an object, where the radio frequency radiation excites the tissue to magnetize. Radio frequency signals are then emitted from the tissue. Radio frequency receiving coils can be used to receive the superposition of these radio frequency signals.

[0012] The image reconstruction scheme can process k-space data to produce reconstructed magnetic resonance images. Specifically, this image reconstruction scheme can provide volumetric magnetic resonance images with optimal noise reduction. To this end, the reconstruction scheme is defined by two magnetic resonance image reconstruction methods, referred to herein as a first magnetic resonance image reconstruction method and a second magnetic resonance image reconstruction method, respectively. The second magnetic resonance image reconstruction method can, for example, be a conventional magnetic resonance imaging process including a fully pipelined approach for reconstructing magnetic resonance images. The second magnetic resonance image reconstruction method can include the first magnetic resonance image reconstruction method and additional processing steps. Each of the additional processing steps can process k-space data and / or image spatial data obtained from the reconstruction of the k-space data.

[0013] A first magnetic resonance image reconstruction method may, for example, include a minimum number of steps that enable a pipeline to obtain a magnetic resonance image. The first magnetic resonance image reconstruction method may, for example, be derived from a second magnetic resonance image reconstruction method. The first magnetic resonance image reconstruction method may be a second magnetic resonance image reconstruction method that excludes at least a portion of the additional processing steps. Therefore, the spatial distribution and / or spatial correlation of noise in an image generated by the first magnetic resonance image reconstruction method differs from (e.g., is more uniform) in the noise distribution in an image reconstructed using only the second magnetic resonance image reconstruction method.

[0014] The first magnetic resonance image reconstruction method can be used to reconstruct at least one aliased image based on received subsampled k-space data. For example, N aliased images can be reconstructed based on received subsampled k-space data, where N is an integer greater than or equal to 1 (N>=1).

[0015] For example, subsampled k-space data can include subsets of k-space data. Each subset of k-space data can be received by a specific receiving coil among multiple receiving coils and / or represent one or more specific slices of a volume. Each subset of k-space data can have the same field of view. Additionally, each subset of k-space data can have the same sampling pattern. This enables uniform subsampling. Uniform subsampling can lead to significant aliasing because only some shifted copies of the original image are superimposed. Therefore, uniform subsampling may be easier to handle than non-uniform subsampling for typical denoising methods. In this example, a first magnetic resonance image reconstruction method can be used to reconstruct multiple aliased images based on the received subsampled k-space data, where each aliased image is obtained from a corresponding subset of k-space data. A second magnetic resonance image reconstruction method can, for example, be configured to perform parallel imaging (PI) reconstruction techniques.

[0016] Alternatively, the subsampled k-space data can be received by a single receiving coil and represent a slice (S=1, R=1). In this case, a aliased image (N=1) can be reconstructed from the received subsampled k-space data using a first magnetic resonance imaging (MRI) method, or the first MRI method can be performed multiple times on the received subsampled k-space data to reconstruct multiple aliased images (N>1). Multiple images can be combined (e.g., averaged) to obtain a single image.

[0017] N aliased images can be denoised individually to obtain N denoised images. For example, denoising can be performed by inputting N aliased images into a noise filtering module. In response to receiving N aliased images as input, the noise filtering module can output N denoised images. The noise filtering module can denoise the N aliased images sequentially or in parallel. The noise filtering module can be a software module. The noise filtering module can, for example, be configured to remove additive noise from the input magnetic resonance images. For example, additive noise can be white noise. Therefore, denoising can remove white noise from the N aliased images.

[0018] Fourier transform can be applied to N denoised aliased images to derive k-space data, which is referred to as denoised k-space data in this paper. The Fourier transform can be a two-dimensional discrete Fourier transform. The sampling pattern of the denoised k-space data can be the same as the sampling pattern of the received subsampled k-space data. A second image reconstruction method can be used to reconstruct the volumetric magnetic resonance image from the denoised k-space data.

[0019] This topic offers the following advantages. Altered images can be directly reconstructed from subsampled k-space data using Fourier transform. Since the noise levels in these aliased images can be uniform and uncorrelated, the loss of broader applicability or optimal performance in denoising methods can be avoided. Furthermore, different types of denoising methods can be used. These different types of denoising methods can include those that do not have the ability to generalize to other image types from the image types they were trained on.

[0020] In one example, subsampled k-space data can comprise multiple subsets of k-space data. Subsampling can be performed in the phase-encoding direction by reducing the number of k-space lines. Subsampling of the k-space data subsets is done in the same way. That is, each subset of the k-space data subsets can cover the same area with the same number of samples following the same sampling pattern. Such uniform subsampling can lead to significant aliasing because only some shifted copies of the original image are superimposed. This may be much easier than non-uniform subsampling for typical denoising methods, but it can lead to more noise-like aliasing. This example can be applied both to standard sensitivity coding (SENSE) and to individual leaves in periodically rotating overlapping parallel lines imaged using enhancement reconstruction (PROPELLER).

[0021] As k-space data passes through the reconstruction pipeline, both signal and noise undergo the same processing steps, and ultimately both are mapped to each individual pixel in the image. However, some or all of the additional processing steps can introduce spatial correlation and / or spatial variation noise into the magnetic resonance image. Accordingly, this subject provides a first magnetic resonance image reconstruction method.

[0022] According to one example, the first magnetic resonance image reconstruction method is a Fourier transform operation. That is, the first magnetic resonance image reconstruction is a second magnetic resonance image reconstruction that excludes all additional processing steps. Excluding all additional processing steps avoids the intentional or unintentional introduction of spatial correlations or spatial variations in noise. This example also has the advantage of accelerating the method while still improving the denoising of the magnetic resonance image. The method can be accelerated because the aliased image can be directly reconstructed using the Fourier transform.

[0023] According to one example, a subset of additional processing steps can be selected, such that the first magnetic resonance image reconstruction method includes a Fourier transform step and a selected subset of processing steps. The selected subset of processing steps can be referred to as a maintenance step. The remaining unselected processing steps can be referred to as exclusion steps. Exclusion steps can be processing steps that are non-uniformly applied to subsampled k-space data and / or corresponding image data. Exclusion steps can be processing steps that cause spatial correlation and / or spatial variation noise. In practice, the algorithms(one or more) used in the exclusion steps can be very complex and non-linear, causing the effect on the noise distribution to become complex or even unpredictable. For example, any algorithm that processes a pixel based on other pixels (e.g., neighboring pixels) will cause spatially correlated noise or spatially varied noise. In contrast to the previous example, the first magnetic resonance image reconstruction method is not limited to Fourier transform because it only excludes steps that cause spatially correlated and / or spatially varied noise. Selecting a maintenance step can, for example, include: receiving input from a user, wherein the input indicates the processing steps to be excluded; and performing the selection based on the input. Alternatively, the selection of maintenance steps can be performed automatically, for example, using predefined rules.

[0024] Excluding certain parts of the additional processing steps can improve denoising of magnetic resonance images and the resulting aliased images for the following reasons: Maintaining the processing steps can achieve improved aliased images compared to applying only Fourier transform. Improved aliased images can lead to improved denoising.

[0025] According to one example, multiple receiving coils of a magnetic resonance imaging (MRI) system are used to acquire subsampled k-space data. The subsampled k-space data may include a subset of k-space data from each receiving coil; for example, each subset may be received by a corresponding receiving coil. In this example, the second MRI reconstruction method may be a PI reconstruction method (e.g., SENSE). According to this example, the exclusion step may include a weighting step for applying the sensitivity of the receiving coils to the corresponding k-space data subset and / or a weighting step for applying compensation for variations in sampling density in k-space to the corresponding k-space data.

[0026] Excluding these two steps can improve noise characteristics for denoising purposes for the following reasons: Reconstructed images can exhibit spatially varying noise levels, largely dependent on the sensitivity of the receiving coil elements and the sampling mode of the acquisition. Reconstructed images are also characterized by local and global noise correlations, especially when acquisition is accelerated through subsampling. Both of these phenomena complicate denoising because they are either accurately modeled, in which case denoising is tailored to a specific acquisition and loses broader applicability; or they are broadly overridden, in which case denoising suffers suboptimal performance due to overgeneralization.

[0027] According to one example, the exclusion step may include a sampling density compensation step and a correlation weighting step. Correlation weighting can be used with the PROPELLER acquisition technique. Correlation weighting can be a downweighting of data from blades with poor correlation to the reference blade as a means of reducing artifacts from through-plane motion. The exclusion step may also include a contrast weighting step, for example, prioritizing data acquired closer to the nominal echo time over earlier or later data as a means of improving contrast. The exclusion step may also include a signal attenuation correction step, for example, for scaling the data to compensate for expected signal attenuation during acquisition as a means of avoiding generated artifacts. This example can be advantageous because it excludes the maximum number of processing steps that could lead to inefficient denoising from the first magnetic resonance image reconstruction method.

[0028] According to one example, subsampled k-space data comprises multiple subsets of k-space data acquired using multiple receiving coils according to the PROPELLER acquisition technique. Each subset of k-space data represents: a PROPELLER blade, or a PROPELLER blade acquired by a specific receiving coil in the receiving coils. In this case, N aliased images comprise one aliased image of each of the k-space data subsets. The number of k-space data subsets can be equal to the number of aliased images N produced. Denoising k-space data comprises N denoised k-space data subsets, with each denoised aliased image corresponding to one denoised k-space data subset. The sampling mode in each denoised k-space data subset is the sampling mode used in the corresponding received noisy k-space data subset.

[0029] The PROPELLER acquisition technique can define the number of blades D, where the number of blades D can be greater than or equal to the number of slices S (e.g., S = D in the case where each blade provides one slice), where each blade is subsampled in the phase encoding direction by a factor F. The width of the blade becomes (L / F). L is the initial number of lines along the phase direction in the blade, which enables full sampling of the k-space of the blade. The field of view of each blade in k-space can be the same. In addition, the sampling pattern in each blade in k-space can be the same. This enables uniform subsampling. One or more (R>=1) receiving coils can be used to acquire subsampled k-space data. If each subset of k-space data represents a corresponding PROPELLER blade, then the entire received subsampled k-space data can include N subsets of k-space data, where N is equal to the number of blades, i.e., N = D. If a subset of k-space data represents a PROPPLER blade that has been sampled by a specific receiving coil in the receiving coil, then the entire received subsampled k-space data can include N subsets of k-space data, where N is the product of the number of blades and the number of receiving coils, i.e., N = D × R.

[0030] According to one example, denoising an aliased image includes grouping the aliased image according to a corresponding blade. Alternatively, denoising an aliased image includes grouping the aliased image according to a corresponding receiving coil. Alternatively, denoising an aliased image includes grouping the aliased image according to both a corresponding blade and a receiving coil. Grouping can result in multiple groups of aliased images. Denoising can be performed jointly on each of the multiple groups of aliased images. This example can implement joint denoising of aliased images using either the SENSE technique or the PROPELLER technique. For example, joint denoising can be performed using the technique described by Jelle Veraart et al. in NeuroImage (142, 2016, pp. 394-406).

[0031] As an example, one or more sets of aliased images include: These represent multiple sets of aliased images of the PROPELLER blades, or multiple sets of aliased images whose k-space data has been acquired by the receiving coils, or multiple sets of aliased images of the PROPELLER blades and the receiving coils. This enables joint denoising to be performed on aliased images from all blades, aliased images from all receiving coil elements, or aliased images from all blades and all receiving coil elements.

[0032] According to one example, the subsampled k-space data includes multiple subsets of k-space data that have been acquired using multiple receiving coils, wherein the at least one aliased image includes one aliased image of each subset of k-space data in the subsets of k-space data, and wherein the denoised k-space data includes one denoised k-space data subset of each denoised aliased image.

[0033] As an example, denoising of N aliased images can be partial. Only partial denoising of the aliased images can be performed to reduce the risk of structural removal.

[0034] As an example, a trained machine learning model can be used to perform denoising on N aliased images. Machine learning models include convolutional neural networks (CNNs). Examples of neural networks that can be used for image denoising include DnCNN, recurrent CNNs, UNet, and a simplified version of DenseNet. In particular, Mohan et al. describe a deep convolutional neural network in “Robust and interpretable blind image denoising via bias-free convolutional neural networks” (arXiv:1906.05478) that can be adapted to implement a noise filtering module.

[0035] According to one example, the first magnetic resonance image reconstruction method may be Fourier transform, and the second magnetic resonance image reconstruction method may be a combination of compressed sensing (CS) and parallel imaging (PI). According to this example, the received subsampled k-space data may include multiple subsets acquired using the PROPELLER technique, wherein each k-space data subset represents: a PROPELLER blade, or a PROPELLER blade that has been acquired by a specific receiving coil in the receiving coil.

[0036] As an example, subsampled k-space data has been acquired to reduce the field of view by a percentage value less than a defined threshold. For example, the subsampled k-space data could be 1.5 times the subsampled k-space data.

[0037] In another aspect, the present invention relates to a computer program comprising machine-executable instructions for execution by a computing system; wherein execution of the machine-executable instructions causes the computing system to: receive subsampled k-space data acquired for imaging an object; reconstruct at least one aliased image based on the subsampled k-space data using a first magnetic resonance image reconstruction method; denoise the at least one aliased image; derive k-space data based on the at least one denoised aliased image using a Fourier transform, the derived k-space data being referred to as denoised k-space data; and reconstruct a magnetic resonance image of the object based on the denoised k-space data using a second magnetic resonance image reconstruction method, the second magnetic resonance image reconstruction method comprising the first magnetic resonance image reconstruction method and additional processing operations.

[0038] In another aspect, the present invention relates to a method for magnetic resonance image reconstruction. The method includes: receiving subsampled k-space data acquired for imaging an object; reconstructing at least one aliased image based on the subsampled k-space data using a first magnetic resonance image reconstruction method; denoising the at least one aliased image; deriving k-space data from the at least one denoised aliased image using a Fourier transform, the derived k-space data being referred to as denoised k-space data; and reconstructing a magnetic resonance image of the object based on the denoised k-space data using a second magnetic resonance image reconstruction method, the second magnetic resonance image reconstruction method including the first magnetic resonance image reconstruction method and additional processing operations.

[0039] 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.

[0040] 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.

[0041] 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.

[0042] 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.

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

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

[0045] 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 a programmable gate array.

[0046] 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).

[0047] 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, which run via the computing system of 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.

[0048] These machine-executable instructions or computer program instructions may also be stored in a computer-readable medium that can instruct a computer, other programmable data processing apparatus or other device to 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.

[0049] 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.

[0050] As used herein, a "user interface" is an interface that allows a user or operator to interact with a computer or computer system. A "user interface" can also be referred to as a "human-machine interface device." A user interface can provide information or data to and / or receive information or data from an operator. A user interface enables input from an operator to be received by the computer and 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.

[0051] As used herein, "hardware interface" encompasses the interfaces that enable a computer system to interact with and / or control external computing devices and / or devices. A hardware interface allows 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.

[0052] 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.

[0053] 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.

[0054] Magnetic resonance imaging, or MR imaging, is defined in this paper as a two-dimensional or three-dimensional visualization reconstructed from anatomical data contained within k-space data. Such visualization can be performed using a computer. Attached Figure Description

[0055] Preferred embodiments of the invention will now be described by way of example only with reference to the accompanying drawings, in which: Figure 1 An example of a medical system is illustrated; Figure 2 A flowchart illustrating a method for reconstructing magnetic resonance images is shown; Figure 3 An example of a medical system is illustrated; Figure 4 A flowchart of a method for denoising aliased images is shown; Figure 5 The original folded blades and the noise-reduced folded blades are shown; Figure 6 The magnetic resonance images obtained without and with this method are shown respectively. Figure 7 The magnetic resonance image obtained by denoising the image jointly reconstructed from all the overlapping blades is shown. Detailed Implementation

[0056] 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.

[0057] Figure 1 An example of a medical system 100 is illustrated. In this example, the medical system 100 includes a computer 102 having a computing system 104. Computer 102 is intended to represent one or more computers or computer systems. Computing system 104 is intended to represent one or more computing systems or computing cores. Computer 102 is also shown to include an optional hardware interface 106 connected to computing system 104. If other components of the medical system 100 are present or included (e.g., a magnetic resonance imaging system), hardware interface 106 can be used to exchange data and commands with these other components. The medical system 100 is also shown to include an optional user interface 108, which can provide various devices for relaying data and receiving data and commands from an operator.

[0058] The medical system 100 is also shown to include a memory 110 connected to a computing system 104. The memory 110 is intended to represent various types of memory accessible by the computing system 104. The memory 110 is shown to contain machine-executable instructions 120. The machine-executable instructions 120 are instructions that enable the computing system 104 to perform various control, data processing, and image processing tasks. The memory 110 is also shown to contain subsampled k-space data 122 that has been acquired for imaging the volume of an object. The subsampled k-space data 122 may, for example, be 1.5 times the subsampled k-space data. According to this subject matter, the subsampled k-space data 122 may, for example, be used to reconstruct a volumetric magnetic resonance image 124.

[0059] Figure 2 This is a flowchart illustrating a method for magnetic resonance image reconstruction based on examples from this topic. For illustrative purposes, in Figure 2 The method described in [the document] can be used in [the following context] Figure 1 The method is implemented in the system shown, but is not limited to this implementation. For example, it can be performed by a medical system 100.

[0060] In step 201, subsampled k-space data (e.g., 122) acquired for imaging the object can be received. In step 203, at least one aliased image can be reconstructed from the subsampled k-space data using a first magnetic resonance image reconstruction method. In step 205, at least one aliased image can be denoised. Denoising can be employed to improve the signal-to-noise ratio. Different methods can be used for denoising. For example, methods relying on convolutional neural networks may be particularly advantageous. They can be applied to complex or amplitude images.

[0061] In step 207, Fourier transform can be used to derive denoised k-space data from at least one denoised aliased image.

[0062] In step 209, a second magnetic resonance image reconstruction method can be used to reconstruct the magnetic resonance image of the object based on the denoised k-space data. The second magnetic resonance image reconstruction method includes the first magnetic resonance image reconstruction method and additional processing operations.

[0063] The first magnetic resonance image reconstruction method can be derived, for example, from the second magnetic resonance image reconstruction method. The first magnetic resonance image reconstruction method can be the second magnetic resonance image reconstruction method, excluding at least part of the additional processing steps. The first magnetic resonance image reconstruction method can be used to reconstruct an aliased image based on subsampled k-space data. The noise level in the resulting aliased image may differ from the noise level obtained when reconstructing the aliased image using the second magnetic resonance image reconstruction method. This is because any additional processing steps can affect the noise, making denoising inefficient. This topic addresses improving denoising of magnetic resonance images by using the first magnetic resonance image reconstruction method followed by the second magnetic resonance image reconstruction method.

[0064] Figure 3 Another example of a medical system 300 is illustrated. Figure 3 The medical system in China is similar to Figure 1 The difference lies in the medical system in China. Figure 3 The medical system also includes an additional magnetic resonance imaging system 302 controlled by the computing system 104.

[0065] The magnetic resonance imaging system 302 includes a magnet 304. The magnet 304 is a superconducting cylindrical magnet with a hole 306 passing through it. Different types of magnets are also possible; 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 sections 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 a sufficiently large space between them to receive the object: the arrangement of the two sections is similar to that of Helmholtz coils. Open magnets are popular because the object is less restricted. An assembly of superconducting coils is located inside the cryostat of the cylindrical magnet.

[0066] An imaging region 308 exists within the aperture 306 of the cylindrical magnet 304, in which the magnetic field is sufficiently strong and uniform to perform magnetic resonance imaging. A field of view 309 is shown within the imaging region 308. Typically, k-space data is acquired with respect to the field of view 309. A region of interest may be identical to the field of view 309, or it may be a sub-volume of the field of view 309. An object 318 is shown supported by an object support 320 such that at least a portion of the object 318 is within both the imaging region 308 and the field of view 309.

[0067] A set of magnetic field gradient coils 310 is also present within the aperture 306 of the magnet. These coils are used to acquire preliminary k-space 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 comprise three separate sets of coils for spatial encoding in three orthogonal spatial directions. The magnetic field gradient power supply powers 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 ramp-varying or pulse-varying.

[0068] Adjacent to the imaging region 308 is an RF coil 314, which is used to manipulate the orientation of the magnetic spin within the imaging region 308 and to receive radio transmissions from the spin also located within the imaging region 308. The RF antenna may comprise multiple coil elements. The RF antenna may also be referred to as a channel or antenna. The RF coil 314 is connected to an RF transceiver 316. The RF coil 314 and the RF transceiver 316 can be replaced by separate transmit and receive coils, as well as separate transmitters and receivers. It should be understood that the RF coil 314 and the RF transceiver 316 are representative. The RF coil 314 is also intended to represent a dedicated transmit antenna and a dedicated receive antenna. Similarly, the transceiver 316 can also represent separate transmitters and receivers. The RF coil 314 may also have multiple receive / transmit elements, and the RF transceiver 316 may have multiple receive / transmit channels.

[0069] Transceiver 316 and gradient controller 312 are shown as hardware interface 106 connected to computer system 102.

[0070] Memory 110 is also shown to contain pulse sequence commands 330. Pulse sequence commands 330 are commands or data that can be converted into commands that enable computing system 104 to control the magnetic resonance imaging system to acquire subsampled k-space data 332. Memory 110 shows subsampled k-space data 332 already acquired by the magnetic resonance imaging system using pulse sequence commands 330.

[0071] Figure 4 This is a flowchart illustrating a method for denoising aliased images, based on examples from this topic. For illustrative purposes, in Figure 4 The method described in [the document] can be used in [the following context] Figure 1 The method is implemented in the system shown, but is not limited to this implementation. It can be performed, for example, by a medical system 100, where the subsampled k-space data 122 is acquired using the PROPELLER acquisition technique.

[0072] In step 401, the aliased images can be grouped according to the corresponding blades and receiving coils, thereby generating multiple sets of aliased images. In step 403, joint denoising can be performed on each set of aliased images in the multiple sets of aliased images.

[0073] Figure 5-7 An example result is shown obtained using PROPELLER in shoulder imaging, where the number of blades D = 46 and the number of receiving coils R = 15. Figure 5 The original folded blade 501 and the noise-reduced folded blade 503 are shown. Figure 5In this paper, for eight of the 15 receiving coil elements, an aliased image reconstructed using Fourier transform of 1.5 times subsampled k-space data from one of the 46 blades is shown. These reconstructed aliased images are not weighted by the sensitivity of the receiving coil elements, as is typically done in coil combinations. Therefore, the noise level remains spatially invariant. However, these reconstructed aliased images have been scaled to equalize the noise level across the receiving coil elements. The k-space data are also not weighted according to the sampling density, as is typically done in joint reconstruction of k-space data based on all blades. Therefore, the noise remains uncorrelated. Denoising using a CNN trained to remove only white noise is applied to these aliased images. The aliased images are again transformed into 1.5 times subsampled k-space data from individual blades and receiving coil elements using an inverse Fourier transform. These k-space data are then jointly subjected to image reconstruction, in this case, a combination of compressed sensing (CS) and parallel imaging (PI) methods.

[0074] Figure 6 The original image 601 and the processed image 603 obtained by denoising the individual overlapping blades 501 in the image space are shown, as follows: Figure 5 As shown.

[0075] exist Figure 7 In, it provides access via from Figure 6 Image 701 is obtained by directly performing the same denoising process trained to remove only white noise on the original image 601. Figure 6 Unlike the processed image 603, the noise is suppressed very unevenly, and the noise level appears even more uneven.

Claims

1. A medical system for magnetic resonance image reconstruction, the medical system comprising: a memory storing machine executable instructions; and a computing system, wherein execution of the machine executable instructions cause the computing system to: receive sub-sampled k-space data that has been acquired for imaging a subject; reconstruct at least one aliased image from the sub-sampled k-space data using a first magnetic resonance image reconstruction method; de-noise the at least one aliased image; derive k-space data from the at least one de-noised aliased image using a Fourier transform, the derived k-space data referred to as de-noised k-space data; reconstruct a magnetic resonance image of the subject from the de-noised k-space data using a second magnetic resonance image reconstruction method, the second magnetic resonance image reconstruction method comprising the first magnetic resonance image reconstruction method and an additional processing operation.

2. The medical system of claim 1, wherein, the de-noising is done by de-noising trained to remove white noise from the aliased image that is directly reconstructed from sub-sampled k-space data using a Fourier transform.

3. The medical system of claim 1, the first magnetic resonance image reconstruction method comprising a Fourier transform operation and a subset of the additional processing operations selected, wherein, the unselected processing operation is an operation that is applied non-uniformly to the sub-sampled k-space data.

4. The medical system of claim 1 or 3, wherein, the sub-sampled k-space data has been acquired using one receive coil or multiple receive coils, the first magnetic resonance image reconstruction method comprising a Fourier transform operation and a subset of the additional processing operations selected, wherein the unselected processing operation comprises at least one of: a weighting operation for applying a sensitivity of the receive coil to the respective k-space data, and a weighting operation for applying a compensation for a variation of a sampling density in k-space to the respective k-space data.

5. The medical system of any one of the preceding claims, the subsampled k-space data comprising a plurality of subsets of k-space data that have been acquired using a plurality of receive coils, respectively; wherein, the at least one aliased image comprises one aliased image per k-space data subset of the k-space data subsets, wherein the de-noised k-space data comprises one de-noised k-space data subset per de-noised aliased image.

6. The medical system of any of the preceding claims 1 to 4, the subsampled k-space data comprising a plurality of subsets of k-space data that have been acquired according to a periodically rotating overlapping parallel lines with enhanced reconstruction (PROPELLER) acquisition technique; wherein, the at least one aliased image comprises one aliased image per k-space data subset of the k-space data subsets, wherein the de-noised k-space data comprises one de-noised k-space data subset per de-noised aliased image, wherein each k-space data subset represents: one PROPELLER blade, or one PROPELLER blade that has been acquired by a particular receive coil.

7. The medical system of claim 5 or 6, wherein, each k-space data subset of the k-space data subsets is sub-sampled k-space data, wherein the k-space data subsets are uniformly sub-sampled in the phase encoding direction.

8. The medical system of any of the preceding claims 5 to 7, wherein, execution of the machine executable instructions cause the computing system to perform the de-noising of the aliased images by: grouping the aliased images according to respective blades and / or receive coils, resulting in multiple groups of aliased images; performing the de-noising jointly for each group of aliased images of the multiple groups of aliased images.

9. The medical system of claim 8, one or more groups of aliased images comprising: multiple groups of aliased images representing the PROPELLER blades, respectively; multiple groups of aliased images that have had their k-space data acquired by the receive coils, respectively; or multiple groups of aliased images representing the PROPELLER blades and the receive coils, respectively.

10. The medical system of any one of the preceding claims, wherein, The medical system further comprises a magnetic resonance imaging system, wherein the memory further contains pulse sequence commands configured to control the magnetic resonance imaging system to acquire k-space data according to a magnetic resonance imaging protocol, wherein execution of the machine executable instructions further causes the computing system to acquire the subsampled k-space data by controlling the magnetic resonance imaging system.

11. The medical system of claim 10, the magnetic resonance imaging system comprising a plurality of receive coils for receiving the subsampled k-space data.

12. The medical system of any one of the preceding claims, wherein, Execution of the machine executable instructions causes the computing system to perform the denoising of the at least one aliased image partially.

13. The medical system of any one of the preceding claims, wherein, Execution of the machine executable instructions causes the computing system to perform the denoising of the at least one aliased image using a trained machine learning model, the machine learning model comprising a convolutional neural network, CNN.

14. A computer program comprising machine executable instructions, wherein, Execution of the machine executable instructions causes the computing system to: receive subsampled k-space data that has been acquired for imaging a subject; reconstruct at least one aliased image from the subsampled k-space data using a first magnetic resonance image reconstruction method; denoise the at least one aliased image; derive k-space data from the at least one denoised aliased image using a Fourier transform, the derived k-space data being referred to as denoised k-space data; reconstruct a magnetic resonance image of the subject from the denoised k-space data using a second magnetic resonance image reconstruction method, the second magnetic resonance image reconstruction method comprising the first magnetic resonance image reconstruction method and an additional processing operation.

15. A method for magnetic resonance image reconstruction, the method comprising: receiving subsampled k-space data that has been acquired for imaging a subject; reconstructing at least one aliased image from the subsampled k-space data using a first magnetic resonance image reconstruction method; denoising the at least one aliased image; deriving k-space data from the at least one denoised aliased image using a Fourier transform, the derived k-space data being referred to as denoised k-space data; reconstructing a magnetic resonance image of the subject from the denoised k-space data using a second magnetic resonance image reconstruction method, the second magnetic resonance image reconstruction method comprising the first magnetic resonance image reconstruction method and an additional processing operation.

Images