Determination of B0 inhomogeneity in magnetic resonance imaging

Abstract

Disclosed herein is a medical system (100, 300) comprising a memory (110) storing machine executable instructions (120) and a convolutional neural network (122), the convolutional neural network is configured to output a predetermined number of deblurred magnetic resonance images (126) in response to receiving a set of partially deblurred magnetic resonance images for each of slices of the deblurred magnetic resonance imaging dataset, the deblurred magnetic resonance images being the slices. Execution of machine executable instructions causes a computing system (104) to: receive (200) a set of partially deblurred magnetic resonance images; receiving (202) the predetermined number of deblurred magnetic resonance images in response to inputting a set of partially deblurred magnetic resonance images for each of the slices into the convolutional neural network; calculating (204) a set (128) of difference images for each of the slices by calculating a difference between the deblurred magnetic resonance image and each of the set of partially deblurred magnetic resonance images; and calculate (206) a determined B0 inhomogeneity map (130) for each of the slices by fitting the smooth manifold to B0 values determined from the set of difference images, the demodulation frequency map, and the assigned demodulation frequency for each of the set of difference images.

Description

Technical Field

[0001] The present invention relates to magnetic resonance imaging and in particular to the determination of B0 inhomogeneities. Background Art

[0002] As part of the process of producing images in a patient, a magnetic resonance imaging (MRI) scanner uses a large static magnetic field to align the nuclear spins of atoms. This large static magnetic field is called the B0 field or main magnetic field. The strength of the B0 field and any applied gradient magnetic fields determine the frequency at which spins (usually protons in hydrogen nuclei) precess. Inhomogeneities in the B0 field may cause protons to precess at frequencies different from those desired. The protons or other spins then resonate away from the frequency. The B0 field inhomogeneity map or equivalently the frequency detuning map can be measured and used to correct during the reconstruction of the magnetic resonance image. There may be several difficulties. In some cases, the B0 inhomogeneity map may not be available or may be invalid, for example, if the object shifts position or moves.

[0003] International patent application WO2021 / 197955 discloses a medical system comprising a memory storing machine executable instructions and a trained neural network. The trained neural network is configured to output corrected magnetic resonance image data in response to receiving a set of magnetic resonance images (each magnetic resonance image has a different spatial constant frequency detuning factor) as input. The medical system also includes a computing system configured to control the medical system, wherein the operation of the machine executable instructions causes the computing system to: receive k-space data acquired according to a magnetic resonance imaging protocol; reconstruct a set of magnetic resonance images according to the magnetic resonance imaging protocol, wherein each magnetic resonance image in the set of magnetic resonance images is reconstructed assuming a different spatial constant frequency detuning factor selected from a frequency detuning factor list; and receive corrected magnetic resonance image data in response to inputting the set of magnetic resonance images into the trained neural network. Summary of the invention

[0004] The invention provides a medical system, a computer program and a method in the independent claims. Embodiments are given in the dependent claims.

[0005] It is known to use neural networks to determine a B0 inhomogeneity map or to correct for B0 inhomogeneity. One difficulty is that neural networks are susceptible to out-of-distribution (OOD) errors, or may sometimes provide erroneous data, such as so-called "neural hallucinations". This can be problematic for medical imaging, as errors caused by the neural network may result in misleading or incorrect medical images. Embodiments may provide improved means of estimating a B0 inhomogeneity map, which may have a reduced likelihood of having errors. This is explained in the context of a single two-dimensional magnetic resonance image or slice. The following explanation may be extended to a three-dimensional data set comprising multiple slices.

[0006] In order to accurately estimate the B0 inhomogeneity map (for a single slice or image), a set of partially deblurred magnetic resonance images is provided, which already have different demodulation frequencies applied to the single magnetic resonance images. When the demodulation frequency is correct, these varying demodulation frequencies have the effect of deblurring the single magnetic resonance images. A convolutional neural network constructs a deblurred magnetic resonance image from the set of partially deblurred magnetic resonance images. Thereafter, a difference image is calculated by subtracting the deblurred magnetic resonance image from each of the partially deblurred magnetic resonance images, and vice versa. The difference image can then be used to algorithmically determine where and which of the partially deblurred magnetic resonance images provides the correct demodulation frequency. Instead of doing this directly, the B0 inhomogeneity value is determined by fitting a smooth manifold (or smooth surface) to the data derived from the set of difference images and the demodulation frequency map and the assigned demodulation frequency for each of the difference images. The fitting of the smooth surface or manifold has the effect of reducing or eliminating errors caused by a neural network that does not correctly provide a deblurred magnetic resonance image.

[0007] The concept of a partially deblurred image in the framework of the present invention relates to a version of a magnetic resonance image of a particular slice associated with a demodulation frequency value specified by a demodulation frequency map for the slice. The demodulation frequency is associated with the demodulation of the acquired magnetic resonance signal (k-space data) having the Larmor frequency as its radio frequency carrier frequency. The demodulation can be done before reconstructing the magnetic resonance image from the k-space data. Each partially deblurred image has one or more parts or patches that are correctly deblurred, i.e., in these parts or patches, the value of the demodulation frequency matches the actual spatial main magnetic field strength that may be spatially non-uniform (and associated with the local (Larmor) radio frequency carrier frequency). For each slice, the demodulation frequency map represents the spatially correlated values ​​of the demodulation frequency over the image area. The slice-specific demodulation frequency map can be set to a constant, i.e., a flat uniform map. The demodulation frequency map can vary from slice to slice, wherein the demodulation frequency map is offset by a slice-specific offset value per slice.

[0008] In one aspect, the present invention provides a medical system comprising a memory storing machine executable instructions and a convolutional neural network configured to output a predetermined number of deblurred magnetic resonance images in response to receiving a set of partially deblurred magnetic resonance images for each slice in a deblurred magnetic resonance imaging data set, the deblurred magnetic resonance images being the slices.

[0009] In magnetic resonance imaging, there may be so-called detuned blur. A deblurred magnetic resonance imaging data set is a single slice or a stack of slices forming a three-dimensional magnetic resonance imaging data set. Detuned blur in magnetic resonance imaging is caused by the lack of knowledge of the actual B0 magnetic field during the imaging process. A baseline B0 magnetic field can be measured for a magnetic resonance imaging system, and this can be used to compensate or deblur the magnetic resonance image. The difficulty with this is that for a specific magnetic resonance imaging protocol, the gradient magnetic field may induce eddy currents in various positions of the magnetic resonance imaging scanner or magnet. Since these eddy currents may vary with a specific magnetic resonance imaging protocol, compensating for these B0 field inhomogeneities may be very challenging. The method adopted in this example is to provide a set of partially deblurred magnetic resonance images for each slice of the measured magnetic resonance imaging data set. Each of these images can be prepared by assuming a specific B0 inhomogeneity. As a result of performing this process, the original magnetic resonance imaging slice may have some of its regions blurred or deblurred, depending on whether the assumption about the B0 magnetic field is correct or incorrect. A convolutional neural network is capable of receiving a set of partially deblurred magnetic resonance images and constructing a deblurred image therefrom. Essentially, it is a synthesis of a collection of said partially deblurred magnetic resonance images.Several variations are possible on this.

[0010] In one example, the convolutional neural network acts on only a single set at a time. This would be equivalent to deblurring only a single slice or two-dimensional data set of a measured magnetic resonance imaging data set. In this case, the predetermined number of deblurred magnetic resonance images is only one. For a three-dimensional magnetic resonance imaging data set including multiple slices, there is a deblurred magnetic resonance image provided for each of the slices. This means that there is a set of partially deblurred magnetic resonance images provided for each slice, and the neural network is able to produce a deblurred image for each of the sets. This can have several advantages over processing a single slice at a time. For example, a convolutional neural network can be trained to receive all of this data simultaneously, and then essentially use data from adjacent slices to help deblur each individual slice. If, for example, the anatomy of the brain or other anatomical structure changes from slice to slice, but there are similarities within each slice, a properly trained convolutional neural network can use this data, and it can provide excellent deblurring that is better than performing a single slice at a time.

[0011] Using a convolutional neural network that accepts a single slice can have advantages, in that it can be easier or more straightforward to train. For example, if only a single slice is being processed, it may not even need to be trained using medical imaging data. For example, a normal optical image from a camera may be acquired, and then portions of that image may be blurred. This can be used as training data, and providing that version of the convolutional neural network may be much simpler.

[0012] The medical system also includes a computing system. The computing system as used herein can take several different forms. In one case, it can be a remote or virtual computing system provided as a cloud service, for example. In other examples, the computing system can be a workstation or computer located in a radiology or other medical facility. In yet another example, the computing system can be part of a computer or control system for a magnetic resonance imaging system. The operation of the machine executable instructions causes the computing system to receive a set of partially deblurred magnetic resonance images for each of the slices. Each of the sets of partially deblurred magnetic resonance images has an assigned demodulation frequency of an offset of a specified slice-specific demodulation frequency map. Therefore, in this feature, the offset is a demodulation frequency offset, and the slice-specific demodulation frequency map is a hypothesis about a possible correct demodulation frequency. For example, a measurement of previously measured B0 inhomogeneity can give a good estimate or starting point for determining the B0 inhomogeneity of a specific pulse sequence or a specific activation sequence of a gradient coil system of a magnetic resonance imaging system. B0 inhomogeneity will of course vary spatially, so in this way, a slice-specific demodulation frequency map will be formed by slicing the existing or pre-measured B0 inhomogeneity map. The offset can then be used to shift the slice-specific demodulation frequency map by the spacing of the demodulation frequencies.

[0013] In some cases, there may be no a priori knowledge of B0 inhomogeneity, or it may be better not to assume this. In this case, the slice-specific demodulation frequency map may be set to a constant or near null value for all slices. In this case, the assigned demodulation frequency for each individual partially deblurred magnetic resonance image is then simply specified by the offset value. In this example, each of the set of partially deblurred magnetic resonance images then has an assigned modulation frequency that specifies a single offset demodulation frequency or value.

[0014] Execution of the machine executable instructions further causes the computing system to receive a predetermined number of deblurred magnetic resonance images in response to inputting the set of partially deblurred magnetic resonance images for each of the slices into the convolutional neural network. In this step, the set of partially deblurred magnetic resonance images for each slice is simultaneously input into the neural network, and in response, a predetermined number of deblurred magnetic resonance images are received as output.

[0015] Execution of the machine executable instructions further causes the computing system to calculate a set of difference images for each of the slices by calculating a difference between the deblurred magnetic resonance image and each partially deblurred magnetic resonance image in the set of partially deblurred magnetic resonance images. The difference may be calculated by performing a pixel-by-pixel subtraction. The process is repeated for each of the slices. Execution of the machine executable instructions further causes the computing system to calculate a determined B0 inhomogeneity map for each of the slices by fitting a smoothing manifold to values ​​determined from the set of difference images, the demodulation frequency map, and the assigned demodulation frequency for each of the set of difference images.

[0016] Essentially, the set of difference images provides information about which partially deblurred magnetic resonance images correctly deblur the deblurred magnetic resonance imaging data in a particular slice. Knowing this and then knowing the slice-specific demodulation frequency map and the assigned demodulation frequencies enables knowing the specific values ​​of the B0 inhomogeneity map for these regions.

[0017] The fitting of a smooth manifold provides a very effective means of accurately determining the B0 inhomogeneity map. The difficulty in using a convolutional neural network with medical imaging processing is that the convolutional neural network may provide hallucinations or erroneous data. For example, a particular voxel may provide a bad value, or a region may be partially incorrect. If a convolutional neural network is to be used, it can be trained to correctly directly output a determined B0 inhomogeneity map for each of the slices. However, there will be no good way to detect or correct errors. The techniques detailed above are extremely robust, and the subtraction of the set of difference images and the fitting of the smooth manifold automatically remove minor defects or inaccuracies in the determined B0 inhomogeneity map for each of the slices. If there is only a single slice, the smooth manifold can be a smooth surface. If there is a stack of slices, the manifold can be a function that varies smoothly in three-dimensional space.

[0018] The smooth manifold can be fitted in various ways. In one instance, the smooth manifold is fitted directly to the difference values ​​in the set of difference images so that the space spanned by the difference images is minimal. Then, the knowledge of the demodulation frequency map and the assigned demodulation frequency for each of the set of difference images allows the calculation of a determined B0 inhomogeneity map from the manifold. Alternatively, on a pixel-by-pixel basis, the demodulation frequency can be selected or interpolated for each pixel using the set of difference images. The smooth manifold can be fitted to these demodulation frequency values ​​or to the B0 inhomogeneity map calculated from the demodulation frequency values. In all of these examples, the smooth manifold can be used to smooth discontinuities and / or smooth potential errors.

[0019] In one example, the fitting of the smooth manifold is solved as an optimization problem, where the value of the set of difference images is minimized along the manifold, where the manifold is subject to constraints such as smoothness and maximum slope known from physical knowledge of the properties of B0 maps. For example, a template representing a typical or previous B0 map can be used.

[0020] The convolutional neural network can be implemented as a U-net, for example. In order to implement the U-net to receive a set of partially deblurred magnetic resonance images for each slice, the number of input or encoding branches can be increased to have an input branch or encoding input x number of slices for each image in the set of partially deblurred magnetic resonance images. Similarly, the output can be increased so that there is an output layer or branch for each individual slice. There can be jump connections between the various output or encoding branches of the U-net to share data between them, so that the conventional U-net architecture can be basically extended to provide a convolutional neural network. An alternative to using the U-net would be RESNET, which has expanded its number of inputs and outputs.

[0021] The training of the convolutional neural network can be performed in several different ways. If there are multiple slices, it is probably best to take a magnetic resonance imaging data set without any blur visible, and then artificially provide a set of partially deblurred magnetic resonance images for each of the slices. This can be performed, for example, by applying a blur kernel to parts of the image in different varying spatial patterns; it can also be achieved by acquiring the original k-space data and then artificially resampling the data using a simulated B0 inhomogeneity map. In any case, one has a set of blurred images, and one has the original images, which are either two-dimensional slices or a complete three-dimensional data set formed by a stack of two-dimensional slices that can be used as real-world data. The artificially blurred samples are input into the convolutional neural network, and then compared with the original unblurred images, and the neural network can be trained using a deep learning algorithm. In the case of a convolutional neural network that acts only on a single slice, the training data can be much more extensive, for example, a variety of photographic images can be used to train the convolutional neural network. In this example, one will take a photo, and then generate the set of partially deblurred magnetic resonance images by locally blurring different areas of the original photo. A big advantage is that a large amount of training data is available in the case of variable workloads.

[0022] In another embodiment, the execution of the machine executable instructions further causes the computing system to determine the modulation frequency for each voxel of the deblurred magnetic resonance image for each slice in the slice using the set of partially deblurred magnetic resonance images, the assigned modulation frequency for each of the sets of difference images, and the demodulation frequency map. The B0 inhomogeneity value is determined based on the demodulation frequency for each voxel. In this embodiment, the demodulation frequency for each voxel is determined individually. The deblurred magnetic resonance image can be used to select the value, or, for example, a curve can be fitted to all values ​​of a particular voxel using all available images, and the best value can be interpolated. However, it should be noted that the specific value of the demodulation frequency is not used directly; a smooth manifold is still used. As previously described, this helps to reduce the impact of errors and accuracy caused by using a convolutional neural network. This is very beneficial for generating medical images because it reduces the possibility of distribution errors and hallucinations caused by convolutional neural networks.

[0023] In another embodiment, during the fitting of the smooth manifold, voxels of the deblurred magnetic resonance image having an amplitude below a predetermined amplitude or an amplitude below a predetermined tolerance within at least a continuous predetermined volume are ignored or attenuated. If one examines a magnetic resonance image, one should notice that there are regions of the image where voxels have very low values; for example, a proton-weighted magnetic resonance image illustrates the volume of hydrogen protons or water in a spatially varying manner.

[0024] Outside the subject's body or where bones are located, the signal will be zero or a very low value. Because the signal is constantly low, it is disadvantageous to try to fit a manifold to this location. Similarly, a region of an image or magnetic resonance image may have a constant value or a value that varies at a certain noise level. If this is the case, inhomogeneities in the B0 field may not appear either. Therefore, in this case, if the voxels have a predetermined amplitude range, meaning that they vary a certain amount within a continuous predetermined volume, meaning that there is a volume or region of at least a certain space, then this region is also ignored or weakened during fitting. This also helps to provide a more accurate estimate of the determined B0 inhomogeneity map.

[0025] In another embodiment, execution of the machine executable instructions further causes the computing system to receive a single magnetic resonance image for each of the slices. Execution of the machine executable instructions further causes the computing system to compute a set of partially deblurred magnetic resonance images for each slice by applying a detuned demodulation determined by the demodulation frequency map and the assigned demodulation frequency. For each of the computed images, the demodulation frequency map and a particular value of the assigned demodulation frequency are used to determine a demodulation frequency to use on a voxel-by-voxel basis. The assigned demodulation frequency may be selected from a set of demodulation frequencies, wherein each member of the set of demodulation frequencies corresponds to one of the set of partially deblurred magnetic resonance images.

[0026] Detuning demodulation can be applied in either image space or k-space, for example using a demodulation kernel. The demodulation kernel can be a Fourier transform of the phase modulation map in k-space. Demodulation in k-space or image space is mathematically the same and can occur in either image or k-space, but what is applied is slightly different. In image space, it is convolved with a "blur kernel" (or "deblur kernel"). In k-space, it is just a point-by-point multiplication with the phasor (frequency * time each point was acquired).

[0027] Execution of the machine executable instructions further causes the computing system to receive measured k-space data. The measured k-space data has a spiral sampling pattern or a non-Cartesian sampling pattern. Execution of the machine executable instructions further causes the computing system to reconstruct a single magnetic resonance image for each of the slices of the measured k-space data. This embodiment may be beneficial because the above-described technique for calculating the determined B0 inhomogeneity map is particularly beneficial when having a spiral sampling pattern or a non-Cartesian sampling pattern.

[0028] In another embodiment, the medical system further comprises a magnetic resonance imaging system. The memory further comprises pulse sequence commands configured to control the magnetic resonance imaging system to acquire measured k-space data according to a magnetic resonance imaging protocol. For example, in some examples, the magnetic resonance imaging protocol may use a spiral sampling mode or a non-Cartesian sampling mode. The execution of the machine executable instructions further causes the computing system to acquire measured k-space data by controlling the magnetic resonance imaging system using the pulse sequence commands.

[0029] In another embodiment, a single magnetic resonance image for each slice is further reconstructed using the previous B0 inhomogeneity map. Execution of the machine executable instructions further causes the computing system to calculate a corrected B0 inhomogeneity map by modifying the previous B0 inhomogeneity map using the determined B0 inhomogeneity map. In this example, the previous B0 inhomogeneity map is initially used to correct the single magnetic resonance image for each of the slices. Therefore, in this case, some detuning ambiguity should be at least partially corrected. At the end of the process, a corrected B0 inhomogeneity map is calculated by modifying the previous B0 inhomogeneity map using the determined B0 inhomogeneity map. This may be beneficial, for example, because for a particular set of pulse sequence commands, the eddy currents may be similar from acquisition to acquisition. The corrected B0 inhomogeneity map may, for example, be used for acquisition using those particular pulse sequence commands.

[0030] In another embodiment, the execution of the machine executable instructions also causes the computing system to calculate a corrected magnetic resonance image using the measured k-space data and the corrected B0 inhomogeneity map. In this example, the corrected B0 inhomogeneity map is then used to recalculate the corrected magnetic resonance image using the corrected B0 inhomogeneity map. There are various ways to pass and calculate the corrected magnetic resonance image. In this case, the original k-space data is returned and the corrected B0 inhomogeneity map is used. This embodiment may be beneficial because it may provide a more accurate and potentially less blurry magnetic resonance image. In this example, the corrected magnetic resonance image may be, for example, a single slice, or if the data set is a three-dimensional data set, it may be a stack of slices.

[0031] In another embodiment, the execution of the machine executable instructions further causes the computing system to acquire additional k-space data by controlling the magnetic resonance imaging system using pulse sequence commands. The execution of the machine executable instructions further causes the computing system to reconstruct additional magnetic resonance images for each slice using the additional k-space data. The reconstruction of the additional magnetic resonance images is corrected using the corrected B0 inhomogeneity map. In this example, the pulse sequence command is the same pulse sequence command used during the process of determining the corrected B0 inhomogeneity map. It is very likely that any eddy currents caused by the gradient will be similar in different acquisitions. This may be very true if it is the same acquisition of the same individual at the same location, but even if the individual has changed, the eddy currents may still be very close. This can provide a means for determining the correct B0 inhomogeneity map for a particular magnetic resonance imaging pulse sequence command.

[0032] In another embodiment, execution of the machine executable instructions further causes the computing system to determine a spatially varying demodulation frequency using the determined B0 inhomogeneity map. Execution of the machine executable instructions further causes the computing system to calculate a corrected magnetic resonance image for each slice by demodulating the single magnetic resonance image using detuned demodulation using the spatially varying demodulation frequency. This can be performed, for example, in k-space or in image space. As described above, detuned demodulation can be applied in either image space or k-space, for example using a demodulation kernel.

[0033] In another embodiment the corrected magnetic resonance image of each slice is a motion corrected magnetic resonance image.

[0034] In another embodiment, the corrected magnetic resonance images are periodic cardiac magnetic resonance images.

[0035] In another embodiment, the corrected magnetic resonance image is a respiratory phase resolved magnetic resonance image.

[0036] In another embodiment the corrected magnetic resonance image is a diffusion weighted magnetic resonance image.

[0037] In another embodiment the corrected magnetic resonance image is a diffusion tensor weighted magnetic resonance image.

[0038] In another embodiment, the corrected magnetic resonance image is an arterial spin labeled magnetic resonance image.

[0039] The above-mentioned image types may benefit from B0 inhomogeneity being corrected using the determined B0 inhomogeneity map, since these techniques are particularly sensitive to errors in B0 inhomogeneity.

[0040] In another embodiment, the predetermined number of deblurred magnetic resonance images is one. In this case, there is only a single set of partially deblurred magnetic resonance images, and there is only one slice.

[0041] In another embodiment, the demodulation frequency map has a constant value. This may be, for example, a constant value, which may be, for example, zero. In this case, a set of partially deblurred magnetic resonance images is then constructed such that only a single demodulation frequency is applied to the entire image. This may be beneficial, for example, when a preliminary measure of B0 inhomogeneity is not available or when it is undesirable to make assumptions about its inhomogeneity.

[0042] In another embodiment, the set of single magnetic resonance images, deblurred magnetic resonance images and partially deblurred magnetic resonance images are three-dimensional or two-dimensional. In the case that they are three-dimensional, this would mean that there is a stack of slices, as previously described. If the set of partially deblurred magnetic resonance images is only two-dimensional, this would mean that there is only one single set, and only a single magnetic resonance image, which is two-dimensional and then deblurred.

[0043] In another aspect, the present invention provides a computer program or computer program product comprising machine executable instructions executed by a computing system controlling a medical system. The computer program product may be stored, for example, on a non-transitory storage medium. Execution of the machine executable instructions causes the computing system to receive a set of partially deblurred magnetic resonance images for each of the slices. Each of the set of partially deblurred magnetic resonance images has an assigned demodulation frequency with a specified offset and a slice-specific demodulation frequency map. The specified or assigned demodulation frequency may vary spatially within each of the partially deblurred magnetic resonance images.

[0044] Execution of the machine executable instructions further causes the computing system to receive a predetermined number of deblurred magnetic resonance images in response to inputting the set of partially deblurred magnetic resonance images for each of the slices into the convolutional neural network. The convolutional neural network is configured to output a predetermined number of deblurred magnetic resonance images in response to receiving the set of partially deblurred magnetic resonance images for each of the slices of the deblurred magnetic resonance image dataset, the deblurred magnetic resonance images being the slices. Execution of the machine executable instructions further causes the computing system to calculate a set of difference images for each of the slices by subtracting the deblurred magnetic resonance images from each partially deblurred magnetic resonance image in the set of partially deblurred magnetic resonance images, and vice versa. Execution of the machine executable instructions further causes the computing system to calculate a determined B0 inhomogeneity map for each of the slices by fitting a smoothing manifold to a value determined from the set of difference images for the modulation frequency map and an assigned demodulation frequency for each difference image in the set of difference images.

[0045] In another aspect, the present invention provides a method for medical imaging. The method includes receiving a set of partially deblurred magnetic resonance images for each of a plurality of slices. The plurality of slices are referred to herein as "slices". Each of the set of partially deblurred magnetic resonance images has an assigned demodulation frequency that specifies an offset of a slice-specific demodulation frequency map. The method also includes receiving a predetermined number of deblurred magnetic resonance images in response to inputting the set of partially deblurred magnetic resonance images for each of the slices into a convolutional neural network. The convolutional neural network is configured to output a predetermined number of deblurred magnetic resonance images in response to receiving a set of partially deblurred magnetic resonance images for each of the slices of a deblurred magnetic resonance imaging data set, the deblurred magnetic resonance images being the slices.

[0046] The method further comprises calculating a set of difference images for each of the slices by calculating a difference between the deblurred magnetic resonance image and each partially deblurred magnetic resonance image in the set of partially deblurred magnetic resonance images. The method further comprises calculating a determined B0 inhomogeneity map for each of the slices by fitting a smooth manifold to values ​​determined from the set of difference images, the demodulation frequency map, and the assigned demodulation frequency for each of the set of difference images.

[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 implemented as an apparatus, method, or computer program product. Accordingly, aspects of the present invention may take the form of a fully hardware embodiment, a fully software embodiment (including firmware, resident software, microcode, etc.), or a combination of software and hardware aspects (all of which may be generally referred to herein as "circuits," "modules," or "systems"). In addition, aspects of the present invention may take the form of a computer program product implemented in one or more computer-readable media having computer executable code implemented thereon.

[0049] Any combination of one or more computer-readable media may be utilized. The computer-readable medium may be a computer-readable signal medium or a computer-readable storage medium. As used herein, "computer-readable storage medium" encompasses any tangible storage medium that can store instructions that can be executed by a processor or a 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 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 medium 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 may be retrieved on a modem, the Internet, or 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 having computer executable code implemented therein, for example, in baseband or as part of a carrier wave. Such a propagated signal may take any of a variety of forms, including but not limited to electromagnetic, optical, or any suitable combination thereof. A computer readable signal medium may be any computer readable medium that is not a computer readable storage medium and is capable of conveying, propagating, or transmitting a program used by or in conjunction with an instruction execution system, apparatus, or device.

[0051] "Computer memory" or "memory" is an example of a computer-readable storage medium. Computer memory is any memory that is directly accessible by a computing system. "Computer storage" or "storage" is another example of a computer-readable storage medium. A computer storage is any non-volatile computer-readable storage medium. In some embodiments, a computer storage may also be computer memory, or vice versa.

[0052] As used herein, "computing system" encompasses an electronic component capable of running 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. The computing system may be, for example, a multi-core processor. A computing system may also refer to a collection of computing systems within a single computer system or distributed across multiple computer systems. The term computing system should also be interpreted as 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 run by multiple computing systems or processors that may be within the same computing device or may even be 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 for various aspects of the present invention may be written in any combination of one or more programming languages ​​and compiled into machine executable instructions, and the one or more programming languages ​​include object-oriented programming languages ​​such as Java, Smalltalk, C++, etc. and conventional procedural programming languages ​​such as "C" programming language or similar programming languages. In some instances, the computer executable code may take the form of a high-level language or take a precompiled form and be used in conjunction with a decoder that generates machine executable instructions at work. In other instances, machine executable instructions or computer executable code may take the form of programming for programmable logic gate arrays.

[0054] The computer executable code may run 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 case, 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] Aspects of the present invention are described with reference to the flowchart, 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, when applicable, each frame or part of the frame of the flowchart, diagram and / or block diagram can be implemented by a computer program instruction in the form of a computer executable code. It should also be understood that, when mutually non-exclusive, the combination of frames in different flowcharts, diagrams and / or block diagrams can be combined. These computer program instructions can be provided to the computing system of a general-purpose computer, a special-purpose computer or other programmable data processing device that produces a machine, so that the instruction running via the computing system of a computer or other programmable data processing device creates a unit for implementing the function / action specified in a flowchart and / or one or more block diagram frames.

[0056] These machine-executable instructions or computer program instructions may also be stored in a computer-readable medium, which can direct a computer, other programmable data processing apparatus, or other device to work 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 one or more block diagram blocks.

[0057] The machine-executable instructions or computer program instructions may also be loaded onto a computer, other programmable data processing apparatus, or other device to cause a series of operational steps to be performed on the computer, other programmable apparatus, or other device, thereby producing a computer-implemented process, so that the instructions running on the computer or other programmable apparatus provide a process for the functions / actions specified in the flowchart and / or one or more 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, the user interface may allow an operator to control or manipulate a computer, and the interface may allow a computer to indicate the effect of the control or manipulation of an operator. The display of data or information on a display or graphical user interface is an example of providing information to an operator. The reception of data by a keyboard, mouse, tracking ball, touch pad, pointing stick, graphic input board, joystick, game controller, webcam, earphone, pedal, wired gloves, remote controller, and accelerometer are all examples of user interface components that implement the reception of 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 a control signal or instruction 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, computer monitors, television screens, touch screens, tactile electronic displays, Braille screens, cathode ray tubes (CRTs), memory tubes, bi-stable displays, electronic paper, vectorscopes, flat panel displays, vacuum fluorescent displays (VFs), light emitting diode (LED) displays, electroluminescent displays (ELDs), plasma display panels (PDPs), liquid crystal displays (LCDs), organic light emitting diode displays (OLEDs), projectors, and head mounted displays.

[0061] K-space data is defined herein as the measurements of radio frequency signals emitted by atomic spins recorded 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.

[0062] 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 magnetic resonance imaging data. This visualization may be performed using a computer. BRIEF DESCRIPTION OF THE DRAWINGS

[0063] Preferred embodiments of the present invention will hereinafter be described, by way of example only, and with reference to the accompanying drawings, in which:

[0064] Figure 1 An example of a medical instrument is illustrated;

[0065] Figure 2 The diagram shows the use of Figure 1 A flowchart of a method of a medical instrument;

[0066] Figure 3 An example of a medical instrument is illustrated;

[0067] Figure 4The diagram shows the use of Figure 3 A flowchart of a method of a medical instrument;

[0068] Figure 5 The construction of the determined B0 inhomogeneity map is illustrated;

[0069] Figure 6 The construction of a set of difference images for a single slice is illustrated;

[0070] Figure 7 illustrates one way to determine an appropriate detuning value for a particular voxel; and

[0071] Figure 8 Cross-sections of rows or columns of the difference image are shown.

[0072] Reference numerals list

[0073] 100 Medical System

[0074] 102 Computer

[0075] 104 Computing Systems

[0076] 106 Hardware Interface

[0077] 108 User Interface

[0078] 110 Memory

[0079] 120 machine executable instructions

[0080] 122 Convolutional Neural Networks

[0081] 124 A collection of partially deblurred MRI images for each slice

[0082] 126 A predetermined number of deblurred magnetic resonance images

[0083] 128 A set of difference images for each of the slices

[0084] 130 Determined B0 inhomogeneity map for each slice

[0085] 200 Receiving a set of the partially deblurred magnetic resonance images for each of the slices

[0086] 202 Receiving a predetermined number of deblurred magnetic resonance images in response to inputting the set of partially deblurred magnetic resonance images for each of the slices into the convolutional neural network

[0087] 204 Calculate a set of difference images for each of the slices by calculating the difference between the deblurred magnetic resonance image and each partially deblurred magnetic resonance image in the set of partially deblurred magnetic resonance images

[0088] 206 Calculating a determined B0 inhomogeneity map for each of the slices by fitting a smooth manifold to values ​​determined from the set of difference images, the demodulation frequency map, and the assigned demodulation frequency for each of the set of difference images

[0089] 300 Medical System

[0090] 302 Magnetic Resonance Imaging System

[0091] 304 Magnet

[0092] 306 Magnet bore

[0093] 308 Imaging Area

[0094] 309 Field of View

[0095] 310 Magnetic Field Gradient Coil

[0096] 312 Magnetic Field Gradient Coil Power Supply

[0097] 314 RF Coil

[0098] 316 Transceiver

[0099] 318 objects

[0100] 320 Object Support

[0101] 330 Pulse sequence command

[0102] 332 Measured k-space data

[0103] 334 Single MRI image for each slice

[0104] 336 Previous B0 inhomogeneity map

[0105] 338 Corrected B0 inhomogeneity map

[0106] 340 Additional k-space data

[0107] 342 Additional MRI images for each slice in the slice

[0108] 400 Acquiring the measured k-space data by controlling the magnetic resonance imaging system using the pulse sequence command

[0109] 402 Receiving measured k-space data

[0110] 404 Reconstructing a single magnetic resonance image for each slice in the slices according to the measured k-space data

[0111] 406 Receive a single magnetic resonance image for each slice in the slices

[0112] 408 Calculating the set of partially deblurred magnetic resonance images by applying detuned demodulation having a demodulation frequency set by the demodulation frequency map and the assigned demodulation frequency

[0113] 410 Calculate a corrected B0 inhomogeneity map by modifying the previous B0 inhomogeneity map with the determined B0 inhomogeneity map

[0114] 412 Acquiring additional k-space data by controlling the magnetic resonance imaging system using the pulse sequence command

[0115] 414 Reconstructing an additional magnetic resonance image for each slice using the additional k-space data, wherein the reconstruction of the additional magnetic resonance image is corrected using the corrected B0 inhomogeneity map

[0116] 700 is equal to f i The mask point of the nearest bin DETAILED DESCRIPTION

[0117] Like numbered elements in these figures are either equivalent elements or perform the same function. Elements that have been discussed previously will not necessarily be discussed in subsequent figures if the function is equivalent.

[0118] Figure 1 An example of a medical system 100 is illustrated. The medical system 100 in this example is shown to include a computer 102 having a computing system 104. Computer 102 is intended to represent one or more computing devices that may be in the same location or may be distributed. Computing system 104 is intended to represent one or more computing systems, such as one or more processors and / or processing cores. The various computing systems 104 may also be located in the same or different locations. The medical system 100 in this example is intended to represent various configurations, for example, the medical system 100 may be a remote computer in a cloud computing or other network accessible computing system.

[0119] In other examples, the medical system 100 may be located, for example, in a hospital or radiology department. In other examples, the medical system 100 may be incorporated into a magnetic resonance imaging system. The computer 102 is shown to include a hardware interface 106. The hardware interface may, for example, enable the computing system 104 to communicate and control other components of the medical system 100, for example, if the computing system 104 additionally includes a magnetic resonance imaging system. The hardware interface 106 may also represent a network connection that may enable the computing system 104 to function at a remote location. The computer 102 is also shown to optionally include a user interface 108, which may, for example, enable a user to operate and / or control the medical system 100. The computer 102 is also shown to include a memory 110 that communicates with the computing system 104. The memory 110 is intended to represent various types of memory, such as non-transient storage media and volatile or non-volatile memory accessible to the computing system 104.

[0120] The memory 110 is shown as containing machine executable instructions 120. The machine executable instructions 120 enable the computing system 104 to perform various data processing and control tasks. For example, it can enable the computing system 104 to reconstruct a magnetic resonance image based on the k-space data and perform various other data manipulation and image processing tasks. The memory 110 is also shown as containing a convolutional neural network 122. The convolutional neural network 122 is configured to output a predetermined number of deblurred magnetic resonance images in response to receiving a set of partially deblurred magnetic resonance images for each of the slices of the deblurred magnetic resonance imaging data set, the deblurred magnetic resonance image being the slice. The deblurred magnetic resonance imaging data set may, for example, be a single slice, in which case the deblurred magnetic resonance imaging data set is a two-dimensional magnetic resonance image. In other cases, the deblurred magnetic resonance imaging data set may be a stack of slices forming a three-dimensional magnetic resonance imaging data set or image.

[0121] The memory 110 is also shown as containing a set of partially deblurred magnetic resonance images for each slice. The memory is also shown as containing a predetermined number of deblurred magnetic resonance images 126 received from the convolutional neural network 122 by inputting therein the set of partially deblurred magnetic resonance images 124 for each slice. The memory 110 is also shown as containing a set of difference images 128 for each of the slices. The memory 110 is also shown as containing a determined B0 inhomogeneity map for each slice 130.

[0122] Figure 2 The diagram shows the operation Figure 1Flowchart of a method of a medical system 100 of the invention. First, a set 124 of partially deblurred magnetic resonance images for each slice is received. Each of the set 124 of partially deblurred magnetic resonance images has an assigned demodulation frequency that specifies an offset of a slice-specific demodulation frequency map. Next, in step 202, a predetermined number of deblurred magnetic resonance images 126 are received in response to inputting the set 124 of partially deblurred magnetic resonance images for each slice into a convolutional neural network 122. Then, in step 204, a set 128 of difference images is calculated for each of the slices by subtracting the deblurred magnetic resonance image from each partially deblurred magnetic resonance image in the set of partially deblurred magnetic resonance images pixel by pixel, and vice versa. Finally, in step 206, a determined B0 inhomogeneity map 130 for each slice is calculated by fitting a smooth manifold to values ​​determined from the set of difference images, the demodulation frequency map, and the assigned demodulation frequency for each of the set of difference images.

[0123] Figure 3 Another example of a medical system 300 is illustrated. Figure 3 The medical system 300 depicted in FIG. Figure 1 The medical system 100 , except that it additionally comprises a magnetic resonance imaging system 302 controlled by the computing system 104 .

[0124] The magnetic resonance imaging system 302 includes a magnet 304. The magnet 304 is a superconducting cylindrical magnet having a bore 306 passing through it. It is also possible to use different types of magnets; for example, it is also possible to use both a split cylindrical magnet and a so-called open magnet. A split cylindrical magnet is similar to a standard cylindrical magnet, except that the cryostat has been split into two parts to allow access to the isoplane of the magnet, such a magnet can be used, for example, in conjunction with charged particle beam therapy. An open magnet has two magnet parts, one above the other, with a space between them large enough to receive an object: the arrangement of the two partial areas is similar to that of a Helmholtz coil. Open magnets are popular because the object is less confined. Inside the cryostat of the cylindrical magnet, there is a collection of superconducting coils.

[0125] 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. The acquired 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. An object 318 is shown supported by an object support 320 such that at least a portion of the object 318 is within the imaging zone 308 and the field of view 309.

[0126] There is also a set of magnetic field gradient coils 310 in the bore 306 of the magnet, which is used to acquire preliminary k-space data to spatially encode magnetic spins in the imaging zone 308 of the magnet 304. The magnetic field gradient coils 310 are connected to a magnetic field gradient coil power supply 312. The magnetic field gradient coils 310 are intended to be representative. Typically, the magnetic field gradient coils 310 include three separate sets of coils for spatial encoding in three orthogonal spatial directions. The magnetic field gradient power supply supplies current to the magnetic field gradient coils. The current supplied to the magnetic field gradient coils 310 is controlled as a function of time and can be ramped or pulsed.

[0127] Adjacent to the imaging zone 308 is a radio frequency coil 314, which is used to manipulate the orientation of magnetic spins within the imaging zone 308 and to receive 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 transmit and receive 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 transmit antenna and a dedicated receive antenna. Similarly, the transceiver 316 may also represent a separate transmitter and receiver. The radio frequency coil 314 may also have multiple receive / transmit elements, and the radio frequency transceiver 316 may have multiple receive / transmit channels.

[0128] The transceiver 316 and the gradient controller 312 are shown connected to the hardware interface 106 of the computer system 102 .

[0129] The memory 110 is also shown as containing pulse sequence commands 330. A pulse sequence command is a command or data that can be converted into a command that enables the computing system 104 to control the magnetic resonance imaging system 302 to acquire k-space data. The memory 110 is also shown as containing measured k-space data 332 acquired by the magnetic resonance imaging 302 and controlled by the pulse sequence commands 330. The memory 110 is also shown as containing a single magnetic resonance image 334 for each of the slices reconstructed from the measured k-space data 332. The memory 110 is also shown as containing a measured previous B0 inhomogeneity map 336. The memory 110 is also shown as containing a corrected B0 inhomogeneity map 338 that is calculated from the previous B0 inhomogeneity map 336 and the determined B0 inhomogeneity map for each slice 130. It should be understood that references to the B0 inhomogeneity map and / or the corrected B0 inhomogeneity map 338 may refer to an individual slice or inhomogeneity map as a whole. The memory 110 is also shown as containing additional k-space data 340, which is also acquired by controlling the magnetic resonance imaging system 302 with the pulse sequence commands 330. The memory 110 is also shown as containing an additional magnetic resonance image 342 for each of the slices reconstructed from the additional k-space data 340 and the corrected B0 inhomogeneity map 338. This may be, for example, a slice-specific corrected B0 inhomogeneity map 338.

[0130] Figure 4 The diagram shows the operation Figure 3 Flow chart of a method of a magnetic resonance imaging system 300 . Figure 4 The method shown in the figure is similar to Figure 2 , having additional steps. The method begins in step 400, where the computing system 104 controls the magnetic resonance imaging system using the pulse sequence command 330 to acquire measured k-space data 332. Next, in step 402, the measured k-space data 332 is received. This can be retrieved from the memory 110, for example. In some examples, the measured k-space data can have a spiral sampling pattern or a non-Cartesian sampling pattern. Next, in step 404, a single magnetic resonance image 334 for each of the slices is reconstructed based on the measured k-space data 332. Next, in step 406, a single magnetic resonance data 334 for each of the slices is received. This may, for example, require retrieving it from the memory 110 or receiving it via a network. Next, in step 408, a set 124 of partially deblurred magnetic resonance images for each slice is calculated by using detuned demodulation with a demodulation frequency set by the demodulation frequency map and the assigned demodulation frequency. Next, the method proceeds to Figure 2200, 202, 204 and 206 are shown. After performing step 206, the method then proceeds to step 410. In step 410, a corrected B0 inhomogeneity map 338 is calculated based on the previous B0 inhomogeneity map 336 and the determined B0 inhomogeneity map 130 for each slice. Next, in step 412, the computing system 104 controls the magnetic resonance imaging system 302 using the pulse sequence commands 330 to acquire additional k-space data 340. Then finally, in step 414, an additional magnetic resonance image for each of the slices is reconstructed based on the additional k-space data 340. During this reconstruction, the corrected B0 inhomogeneity map 338 is used to correct for inhomogeneities in the B0 field.

[0131] An example may provide an improvement to an existing prototype to perform "deblurring" of spiral images using a trained "artificial intelligence" (AI) neural network. In spiral or other non-Cartesian imaging, detuning leads to blur in the reconstructed image. It has been shown that given a collection of input images sharpened at a uniform frequency, AI can be used to directly produce a deblurred image; however, this results in the risk of introducing clinically relevant artifacts. An alternative approach is to use AI to predict a detuned field map or "B0 map", which can then be passed to a "traditional" deblurring algorithm to correct the blur. This approach is more stable, but requires the AI ​​network to learn a transformation from the image class domain to the B0 class domain, which requires an increased risk of obtaining unreliable or inaccurate predictions.

[0132] The example uses a hybrid approach that uses the advantages of both techniques while avoiding their disadvantages. As previously described, the AI ​​module (convolutional neural network 122) is first used to create a sharpened image (a predetermined number of deblurred magnetic resonance images 126). This is a simpler task for the AI ​​because it does not require a transformation to a new domain. The AI ​​output and the single frequency deblurred input are then fed into a second algorithm that calculates an "implicit" B0 map. This avoids the domain transformation required to generate the B0 map directly using the AI, while also mitigating the risk of artifacts generated by the network. Therefore, the resulting hybrid method is more stable and more accurate than either previous method.

[0133] Spiral images often contain blur due to detuning. This is typically corrected by using a detuning (or "B0") map acquired during a prescan. However, some blur will remain due to eddy currents, inaccuracies in the B0 map, and / or field drift. As part of the development of spiral imaging, we are exploring methods of using AI to sharpen the images beyond what can be achieved based on measured B0 maps. This feature is highly desired, particularly in the case of diffusion imaging, where strong eddy current effects and long readout times result in significant residual blur.

[0134] Two implementations are described below. In the first example, the AI ​​network produces a sharpened image as its output. In the second, the network produces a prediction of the detuned field, which can then be passed to a "traditional" deblurring algorithm. The advantages and disadvantages of each approach are summarized in the table below:

[0135]

[0136] The present invention exploits the advantages of image sharpening methods while avoiding the increased risks. Note that the above risks prevent implementation in products due to the expected reliability of the solution.

[0137] The present invention may include one or more of the following elements:

[0138] The AI ​​network is trained to take a collection of one or more blurry input images and produce a sharpened output image.

[0139] A method to predict a detuning map using both the input(s) and output(s) of a network. In the primary embodiment, this element will not use AI; however, a second AI network could potentially serve this purpose.

[0140] Figure 5 The construction of the determined B0 inhomogeneity map is illustrated. Figure 5 , the operation for a single slice is illustrated. First, there is a set 124 of partially deblurred magnetic resonance images for that slice. These are then input into a convolutional neural network 122, which in this case is a U-net. The U-net outputs a deblurred magnetic resonance image 126. Steps 204 and 206 are then performed, which then outputs a determined B0 inhomogeneity map 130.

[0141] By taking the frequency set f0 to f N-1 The initial (blurred) data is demodulated to generate the input for the U-Net. The demodulation is performed based on the known acquisition time of each point in k-space:

[0142]

[0143] Here, m refers to the initial data, and refers to the i-th demodulated data set. The method is based on the conjugate phase reconstruction (CPR) algorithm. Each of the N input images is locally detuned close to f i The region of will be sharp. U-Net can then be trained to produce a sharpened image from a collection of partially sharpened inputs.

[0144] The second part of the invention is to generate a detuning plot based on the network inputs and outputs. This starts by subtracting the output from each input and taking the magnitude of the result, as shown below Figure 6 shown.

[0145] Figure 6 The construction of a set 128 of difference images for a single slice is illustrated. A partially deblurred magnetic resonance image 124 for that slice has a deblurred magnetic resonance image 126 subtracted from it. In this example, absolute values ​​are taken, so in principle the order of subtraction can be reversed. The result of the subtraction is then a set 128 of difference images for that particular slice. The darkest areas in the image are where that particular deblurred magnetic resonance image best compensates for the B0 inhomogeneity.

[0146] These difference images are then fed into an algorithm to produce a smoothly varying detuning map. The algorithm exploits the fact that at detuning frequencies close to f i In the region of , the ith difference image will be close to zero. This is shown in the figure below: the top row shows the magnitude difference image, while the bottom row shows the detuning image closer to f than any other frequency (in white). i location.

[0147] Figure 7 One way to determine the appropriate detuning value for a particular voxel is illustrated. In this figure, the top row 128 shows a collection of various difference images. These difference images 128 are then used to construct a vector with a value equal to f i The mask point of the nearest bin.

[0148] Another way to see this is by looking at the rows or columns "on the edge" of the magnitude difference data, i.e., at f i The dimension is used as the in-plane axis. In doing so, we can see a band of values ​​close to zero. The curve shows the true detuning frequency for this set of pixels.

[0149] Figure 8 A cross section of a row or column of the difference image 128 is shown. The figure shows the relationship between the image pixel and f i By searching for the pixels closest to zero, the detuning values ​​of these voxels can be determined.

[0150] To create the final B0 map, we use any algorithm that meets the following requirements:

[0151] Based on the expectation of the detuning plot, some smoothing and maximum slope are imposed on the output frequency plot f(x,y).

[0152] If i min (x,y) is the one with the minimum |In(f i ,x,y)-Out(x,y)|, the output frequency map is minimized

[0153] While the main examples are based on the magnitude of the difference between the AI ​​output and input, any algorithm that uses both input and output and the expected prior knowledge of the B0 map can serve this purpose.

[0154] The advantage of this approach is that AI is used for a simple job - converting a collection of partially sharp images into a fully sharp image - while the risk of clinically relevant artifacts caused by AI is limited by using the second step to infer the B0 map.

[0155] 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.

[0156] Those skilled in the art can understand and implement other variations of the disclosed embodiments when practicing the claimed invention by studying the drawings, the disclosure and the claims. In the claims, the word "comprising" does not exclude other elements or steps, and the words "one" or "an" do not exclude multiple. A single processor or other unit can satisfy the functions of several items recorded in the claims. Although specific measures are recorded in mutually different dependent claims, this does not indicate that a combination of these measures cannot be used to advantage. The computer program can be stored and / or distributed on appropriate media, such as optical storage media or solid-state media provided with or as part of other hardware, but the computer program can also be distributed in other forms, such as via the Internet or other wired or wireless telecommunications systems. Any figure mark in the claims should not be interpreted as limiting the scope.

Claims

1. A medical system (100, 300), comprising: a memory (110) storing machine executable instructions (120) and a convolutional neural network (122) configured to output a predetermined number of deblurred magnetic resonance images (126) in response to receiving a set (124) of partially deblurred magnetic resonance images for each of slices of a deblurred magnetic resonance imaging dataset, the deblurred magnetic resonance images being the slices; and A computing system (104), wherein execution of the machine executable instructions causes the computing system to: receiving (200) a set of the partially deblurred magnetic resonance images for each of the slices, wherein each partially deblurred magnetic resonance image in the set of partially deblurred magnetic resonance images has an assigned demodulation frequency specifying an offset of a slice-specific demodulation frequency map; receiving (202) the predetermined number of deblurred magnetic resonance images in response to inputting the set of partially deblurred magnetic resonance images for each of the slices into the convolutional neural network; calculating (204) a set of difference images (128) for each of the slices by calculating a difference between the deblurred magnetic resonance image and each partially deblurred magnetic resonance image in the set of partially deblurred magnetic resonance images; and The determined B0 inhomogeneity map (130) for each of the slices is calculated (206) by fitting a smooth manifold to values ​​determined according to: the set of difference images, the demodulation frequency map, and the assigned demodulation frequency for each difference image in the set of difference images.

2. The medical system according to claim 1, wherein: Execution of the machine-executable instructions further causes the computing system to determine a demodulation frequency for each voxel of the deblurred magnetic resonance image for each of the slices using the set of partially deblurred magnetic resonance images, the assigned demodulation frequency for each difference image in the set of difference images, and the demodulation frequency map; and wherein the B0 inhomogeneity value is determined based on the demodulation frequency for each voxel.

3. The medical system according to claim 1 or 2, wherein: Voxels of the deblurred magnetic resonance image having a magnitude below a predetermined magnitude or a magnitude below a predetermined tolerance within at least a continuous predetermined volume are ignored or attenuated during the fitting of the smooth manifold.

4. The medical system according to claim 1, 2 or 3, wherein: Execution of the machine-executable instructions further causes the computing system to: receiving (406) a single magnetic resonance image (334) for each of the slices; The set of partially deblurred magnetic resonance images is calculated (408) by applying detuned demodulation using a demodulation frequency set by the demodulation frequency map and the assigned demodulation frequency.

5. The medical system according to claim 4, wherein: Execution of the machine-executable instructions further causes the computing system to: receiving (402) measured k-space data (332), wherein the measured k-space data has a spiral sampling pattern or a non-Cartesian sampling pattern; and The single magnetic resonance image (334) for each of the slices is reconstructed (404) based on the measured k-space data.

6. The medical system according to claim 5, wherein: The medical system also includes a magnetic resonance imaging system (302), wherein the memory further includes pulse sequence commands (330), the pulse sequence commands being configured to control the magnetic resonance imaging system to acquire the measured k-space data according to a magnetic resonance imaging protocol, wherein execution of the machine-executable instructions further causes the computing system to acquire (400) the measured k-space data by controlling the magnetic resonance imaging system using the pulse sequence commands.

7. The medical system according to claim 5 or 6, wherein: The single magnetic resonance image for each slice is also reconstructed using a previous B0 inhomogeneity map (336), and wherein execution of the machine-executable instructions further causes the computing system to calculate (410) a corrected B0 inhomogeneity map (338) by modifying the previous B0 inhomogeneity map using the determined B0 inhomogeneity map.

8. The medical system according to claim 7, wherein: Execution of the machine-executable instructions further causes the computing system to compute a corrected magnetic resonance image using the measured k-space data and the corrected B0 inhomogeneity map.

9. The medical system according to claim 8, wherein: Execution of the machine-executable instructions further causes the computing system to: acquiring (412) additional k-space data (340) by controlling the magnetic resonance imaging system using the pulse sequence commands; and An additional magnetic resonance image for each slice is reconstructed (414) using the additional k-space data, wherein the reconstruction of the additional magnetic resonance image (342) is corrected using the corrected B0 inhomogeneity map.

10. The medical system according to any one of claims 1 to 8, wherein: Execution of the machine-executable instructions further causes the computing system to: determining a spatially varying demodulation frequency using the determined B0 non-uniformity map; and A corrected magnetic resonance image for each slice is calculated by demodulating the single magnetic resonance image with detuned frequency demodulation using the spatially varying demodulation frequency.

11. The medical system according to claim 9 or 10, wherein: The corrected magnetic resonance image for each slice is any of the following: a motion corrected magnetic resonance image, a periodic cardiac magnetic resonance image, a respiratory phase resolved magnetic resonance image, a diffusion weighted magnetic resonance image, a diffusion tensor weighted magnetic resonance image, an arterial spin labeled magnetic resonance image.

12. The medical system according to any one of the preceding claims, wherein: The predetermined number of deblurred magnetic resonance images is one deblurred magnetic resonance image.

13. The medical system according to any one of the preceding claims, wherein: The demodulated frequency map has a constant value.

14. A computer program product comprising machine executable instructions (120) for execution by a computing system (104) for controlling a medical system (100, 300), wherein: Execution of the machine-executable instructions causes the computing system to: Receiving (200) a set (124) of partially deblurred magnetic resonance images for each slice of a plurality of slices, wherein each partially deblurred magnetic resonance image in the set of partially deblurred magnetic resonance images has an assigned demodulation frequency that specifies an offset of a slice-specific demodulation frequency map; receiving (202) a predetermined number of deblurred magnetic resonance images in response to inputting the set of partially deblurred magnetic resonance images for each of the slices into a convolutional neural network (122), wherein the convolutional neural network is configured to output the predetermined number of deblurred magnetic resonance images (126) in response to receiving the set of partially deblurred magnetic resonance images for each of the slices of a deblurred magnetic resonance imaging dataset, the deblurred magnetic resonance images being the slices; calculating (204) a set of difference images (128) for each of the slices by calculating a difference between the deblurred magnetic resonance image and each partially deblurred magnetic resonance image in the set of partially deblurred magnetic resonance images; and The determined B0 inhomogeneity map (130) for each of the slices is calculated (206) by fitting a smooth manifold to values ​​determined according to: the set of difference images, the demodulation frequency map, and the assigned demodulation frequency for each difference image in the set of difference images.

15. A method of medical imaging, wherein: The method comprises: Receiving (200) a set (124) of partially deblurred magnetic resonance images for each slice of one or more slices, wherein each partially deblurred magnetic resonance image in the set of partially deblurred magnetic resonance images has an assigned demodulation frequency that specifies an offset of a slice-specific demodulation frequency map; receiving (202) a predetermined number of deblurred magnetic resonance images (126) in response to inputting the set of partially deblurred magnetic resonance images for each of the slices into a convolutional neural network (122), wherein the convolutional neural network is configured to output the predetermined number of deblurred magnetic resonance images in response to receiving the set of partially deblurred magnetic resonance images for each of the slices of a deblurred magnetic resonance imaging dataset, the deblurred magnetic resonance images being the slices; calculating (204) a set of difference images (128) for each of the slices by calculating a difference between the deblurred magnetic resonance image and each partially deblurred magnetic resonance image in the set of partially deblurred magnetic resonance images; and The determined B0 inhomogeneity map (130) for each of the slices is calculated (206) by fitting a smooth manifold to values ​​determined according to: the set of difference images, the demodulation frequency map, and the assigned demodulation frequency for each difference image in the set of difference images.

Images