Magnetic resonance imaging apparatus and method

Abstract

To improve the quality of an image obtained by thinning-out data collection.SOLUTION: A magnetic resonance imaging device includes a collection unit, an integration unit, and a reconstruction unit. The collection unit executes thinning-out data collection while moving sample points at equal intervals in a k space, and collects time-series k space frames. The integration unit generates a plurality of time-series k space frames on a plurality of temporal resolutions on the basis of the time-series k space frames. In the plurality of time-series k space frames, sample points are positioned at equal internals with intervals mutually different from one another. The reconstruction unit generates time-series images on the basis of the plurality of time-series k space frames.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] SUMMARY OF THE INVENTION The embodiments disclosed herein and in the drawings relate to magnetic resonance imaging apparatus and methods. [Background technology]

[0002] A technique for thinning out time-series k-space data acquisition is known as kt BLAST (k-space time Broad-use Linear Acquisition Speed-up Technique) or kt SENSE (sensitivity encoding). [Prior art documents] [Patent documents]

[0003] [Patent Document 1] U.S. Patent No. 10,274,567 [Non-patent literature]

[0004] [Non-Patent Document 1] J. Tsao, P. Boesiger, KP Pruessmann, “kt BLAST and kt SENSE: Dynamic MRI With high Frame Rate exploiting spatiotemporal Correlations”, Magnetic Resonance in Medicine 50:1031-1042 (2003). Summary of the Invention [Problem to be solved by the invention]

[0005] One of the problems to be solved by the embodiments disclosed in this specification and the drawings is to improve the image quality of images obtained by thinned data collection. However, the problems to be solved by the embodiments disclosed in this specification and the drawings are not limited to the above problem. Problems corresponding to the effects of each configuration shown in the embodiments described below can also be positioned as other problems. [Means for solving the problem]

[0006] A magnetic resonance imaging apparatus according to an embodiment includes an acquisition unit, an integration unit, and a reconstruction unit. The acquisition unit acquires time-series k-space frames by thinning data acquisition while moving sample points at equal intervals in k-space. The integration unit generates multiple time-series k-space frames for multiple temporal resolutions based on the time-series k-space frames, and the multiple time-series k-space frames have sample points positioned at equal intervals but different from each other. The reconstruction unit generates time-series images based on the multiple time-series k-space frames. [Brief explanation of the drawings]

[0007] [Figure 1] FIG. 1 is a diagram showing the configuration of a magnetic resonance imaging apparatus according to this embodiment. [Figure 2] FIG. 2 is a diagram showing an example of the flow of an MR examination according to this embodiment. [Figure 3] FIG. 3 is a diagram showing an example of the arrangement of sample points in thinned data collection according to this embodiment. [Figure 4] FIG. 4 is a diagram showing a schematic diagram of integration processing of time-series k-space frames (arrangement pattern 2 is the applied reference frame). [Figure 5] FIG. 5 is a diagram showing a schematic diagram of integration processing of time-series k-space frames (arrangement pattern 3 is the applied reference frame). [Figure 6] FIG. 6 is a diagram showing the arrangement patterns of k-space frames extracted by each integration window. [Figure 7] FIG. 7 is a diagram showing an example of image conversion processing using the difference intra-frame SENSE. [Figure 8] FIG. 8 is a diagram showing an overview of SENSE reconstruction. [Figure 9] FIG. 9 is a diagram schematically illustrating an example of input and output of the machine learning model NN1. [Figure 10] FIG. 10 is a diagram schematically illustrating an example of input and output of the machine learning model NN2. [Figure 11] FIG. 11 is a diagram schematically showing the integration process of time-series k-space frames according to the first modification. DETAILED DESCRIPTION OF THE INVENTION

[0008] Hereinafter, embodiments of a magnetic resonance imaging apparatus and method will be described in detail with reference to the drawings.

[0009] Fig. 1 is a diagram showing the configuration of a magnetic resonance imaging apparatus 1 according to this embodiment. As shown in Fig. 1, the magnetic resonance imaging apparatus 1 includes a gantry 11, a bed 13, a gradient magnetic field power supply 21, a transmission circuit 23, a reception circuit 25, a bed driving device 27, a sequence control circuit 29, and a host computer 50.

[0010] The gantry 11 has a static magnetic field magnet 41 and a gradient magnetic field coil 43. The static magnetic field magnet 41 and the gradient magnetic field coil 43 are housed in a housing of the gantry 11. A hollow bore is formed in the housing of the gantry 11. A transmitting coil 45 and a receiving coil 47 are arranged in the bore of the gantry 11.

[0011] The static magnetic field magnet 41 has a hollow, approximately cylindrical shape and generates a static magnetic field inside the approximately cylinder. For example, a permanent magnet, a superconducting magnet, or a normal-conducting magnet may be used as the static magnetic field magnet 41. Here, the central axis of the static magnetic field magnet 41 is defined as the Z axis, the axis perpendicular to the Z axis is defined as the Y axis, and the axis horizontally perpendicular to the Z axis is defined as the X axis. The X axis, Y axis, and Z axis form an orthogonal three-dimensional coordinate system.

[0012] The gradient coil 43 is a hollow, approximately cylindrical coil unit attached to the inside of the static magnetic field magnet 41. The gradient coil 43 generates a gradient magnetic field by receiving a current from the gradient power supply 21. More specifically, the gradient coil 43 has three coils corresponding to the X-axis, Y-axis, and Z-axis, which are orthogonal to each other. The three coils form gradient magnetic fields whose field strength varies along each of the X-axis, Y-axis, and Z-axis. The gradient magnetic fields along the X-axis, Y-axis, and Z-axis are combined to form a slice selection gradient magnetic field Gs, a phase encoding gradient magnetic field Gp, and a frequency encoding gradient magnetic field Gr, which are orthogonal to each other, in desired directions. The slice selection gradient magnetic field Gs is used to arbitrarily determine an imaging plane (slice). The phase encoding gradient magnetic field Gp is ​​used to change the phase of a magnetic resonance signal (hereinafter referred to as an MR signal) according to a spatial position. The frequency encoding gradient magnetic field Gr is used to change the frequency of the MR signal according to a spatial position. In the following description, the gradient direction of the slice selection gradient magnetic field Gs is the Z axis, the gradient direction of the phase encoding gradient magnetic field Gp is ​​the Y axis, and the gradient direction of the frequency encoding gradient magnetic field Gr is the X axis.

[0013] The gradient magnetic field power supply 21 supplies a current to the gradient magnetic field coil 43 in accordance with a sequence control signal from the sequence control circuit 29. The gradient magnetic field power supply 21 supplies a current to the gradient magnetic field coil 43, thereby causing the gradient magnetic field coil 43 to generate gradient magnetic fields along the X-axis, Y-axis, and Z-axis. The gradient magnetic fields are superimposed on the static magnetic field formed by the static magnetic field magnet 41 and applied to the subject P.

[0014] The transmission coil 45 is disposed, for example, inside the gradient magnetic field coil 43, and receives a current from the transmission circuit 23 to generate a radio frequency pulse (hereinafter referred to as an RF pulse).

[0015] The transmission circuitry 23 supplies a current to the transmission coil 45 to apply an RF pulse to the subject P via the transmission coil 45 to excite target protons present in the subject P. The RF pulse oscillates at a resonance frequency specific to the target protons, exciting the target protons. An MR signal is generated from the excited target protons and detected by the reception coil 47. The transmission coil 45 is, for example, a whole-body coil (WB coil). The whole-body coil may be used as a transmission / reception coil.

[0016] The receive coil 47 receives MR signals emitted from target protons present in the subject P in response to the action of an RF magnetic field pulse. The receive coil 47 has multiple receive coil elements capable of receiving MR signals. The received MR signals are supplied to the receiver circuit 25 via wire or wirelessly. Although not shown in FIG. 1 , the receive coil 47 has multiple receive channels implemented in parallel. Each receive channel has receive coil elements that receive MR signals and amplifiers that amplify the MR signals. MR signals are output for each receive channel. The total number of receive channels and the total number of receive coil elements may be the same, or the total number of receive channels may be greater or less than the total number of receive coil elements.

[0017] The receiver circuitry 25 receives MR signals generated from excited target protons via the receiver coil 47. The receiver circuitry 25 processes the received MR signals to generate digital MR signals. The digital MR signals can be expressed in k-space, which is defined by spatial frequencies. Therefore, hereinafter, the digital MR signals will be referred to as k-space data. The k-space data is a type of raw data used for image reconstruction. The k-space data is supplied to a host computer 50 via a wired or wireless connection.

[0018] The above-described transmitting coil 45 and receiving coil 47 are merely examples. A transmitting / receiving coil having both transmitting and receiving functions may be used instead of the transmitting coil 45 and receiving coil 47. Furthermore, the transmitting coil 45, receiving coil 47, and transmitting / receiving coil may be combined.

[0019] A bed 13 is installed adjacent to the gantry 11. The bed 13 has a top plate 131 and a base 133. A subject P is placed on the top plate 131. The base 133 supports the top plate 131 so that it can slide along the X-axis, Y-axis, and Z-axis. A bed driving device 27 is housed in the base 133. The bed driving device 27 moves the top plate 131 under the control of a sequence control circuit 29. The bed driving device 27 may include any motor, such as a servo motor or a stepping motor.

[0020] The sequence control circuit 29 has, as hardware resources, a processor such as a CPU (Central Processing Unit) or an MPU (Micro Processing Unit) and memories such as a ROM (Read Only Memory) and a RAM (Random Access Memory). The sequence control circuit 29 synchronously controls the gradient magnetic field power supply 21, the transmission circuitry 23, and the reception circuitry 25 based on imaging conditions set by an imaging condition setting function 512 of the processing circuitry 51, performs MR imaging of the subject P in accordance with a pulse sequence corresponding to the imaging conditions, and collects k-space data regarding the subject P.

[0021] The sequence control circuit 29 according to this embodiment performs thinned data acquisition while moving sample points at equal intervals in k-space to acquire time-series k-space data. A set of k-space data forming one frame is called a k-space frame. The temporal resolution of the k-space frame acquired by thinned data acquisition is set arbitrarily. Note that the temporal resolution refers to the time width of the acquisition period for k-space data covering all sample points filled in one k-space frame. When one k-space frame is composed of k-space data covering a relatively narrow time range, this means that the temporal resolution is higher than when it is composed of k-space data covering a relatively wide time range.

[0022] As shown in FIG. 1, the host computer 50 is a computer having a processing circuit 51 , a memory 52 , a display 53 , an input interface 54 and a communication interface 55 .

[0023] The processing circuitry 51 has a processor such as a CPU as a hardware resource. The processing circuitry 51 functions as the core of the magnetic resonance imaging apparatus 1. For example, the processing circuitry 51 has an acquisition function 511, an imaging condition setting function 512, an integration function 513, a reconstruction function 514, an image processing function 515, and a display control function 516 by executing various programs. The various programs are stored in a non-transitory computer-readable recording medium such as the memory 52.

[0024] By implementing the acquisition function 511, the processing circuitry 51 acquires various data. For example, the processing circuitry 51 acquires time-series k-space frames collected by the sequence control circuitry 29. The processing circuitry 51 may acquire the time-series k-space frames directly from the sequence control circuitry 29 or the receiving circuitry 25, or may temporarily store the time-series k-space frames in a memory 52 and acquire the data from the memory 52.

[0025] The processing circuitry 51 automatically or manually sets imaging conditions related to thinned data acquisition by implementing the imaging condition setting function 512. For example, the processing circuitry 51 can set any imaging condition, such as the echo time TE, the repetition time TR, the imaging range (FOV), the k-space trajectory type, and the thinning rate (speed rate), as one of the imaging conditions.

[0026] By implementing the integration function 513, the processing circuitry 51 generates multiple time-series k-space frames for multiple temporal resolutions based on the time-series k-space frames. The multiple time-series k-space frames have sample points positioned at equal intervals (pitches) that are different from each other. The intervals correspond to the distances between sample points in k-space. The intervals differ between the multiple time-series k-space frames. In each time-series k-space frame, the sample points are positioned at equal intervals that correspond to the temporal resolution.

[0027] By implementing the reconstruction function 514, the processing circuitry 51 generates time-series images based on multiple time-series k-space frames with multiple temporal resolutions. The time-series images refer to time-series MR image data. The temporal resolution of the time-series images can be set arbitrarily. For example, the temporal resolution of the time-series images may be set to any of the temporal resolutions of the multiple time-series k-space frames that are the original data, or may be set to a temporal resolution different from the temporal resolution of the multiple time-series k-space frames that are the original data.

[0028] By implementing the image processing function 515, the processing circuitry 51 performs various types of image processing on the time-series images. For example, the processing circuitry 51 performs image processing such as volume rendering, surface rendering, pixel value projection processing, MPR (Multi-Planer Reconstruction) processing, and CPR (Curved MPR) processing. The processing circuitry 51 can also perform various types of image processing such as region extraction, image recognition, image analysis, and alignment.

[0029] By implementing the display control function 516, the processing circuitry 51 displays various information on the display 53. For example, the processing circuitry 51 displays time-series images generated by the reconstruction function 514 on the display 53.

[0030] The memory 52 is a storage device such as an HDD (Hard Disk Drive), an SSD (Solid State Drive), or an integrated circuit storage device that stores various information. The memory 52 may also be a drive device that reads and writes various information from and to a portable storage medium such as a CD-ROM drive, a DVD drive, or a flash memory. For example, the memory 52 stores imaging conditions, time-series k-space frames, time-series images, programs, and the like.

[0031] The display 53 displays various information using a display control function 516. As the display 53, for example, a CRT display, a liquid crystal display, an organic EL display, an LED display, a plasma display, or any other display known in the art can be used as appropriate.

[0032] The input interface 54 includes input devices that accept various commands from the user. Examples of input devices that can be used include a keyboard, a mouse, various switches, a touch screen, and a touch pad. Note that input devices are not limited to those equipped with physical operating components such as a mouse and a keyboard. For example, the input interface 54 also includes an electrical signal processing circuit that receives an electrical signal corresponding to an input operation from an external input device provided separately from the magnetic resonance imaging apparatus 1 and outputs the received electrical signal to various circuits. The input interface 54 may also be a voice recognition device that converts a voice signal collected by a microphone into a command signal.

[0033] The communication interface 55 is an interface that connects the magnetic resonance imaging apparatus 1 to a workstation, a PACS (Picture Archiving and Communication System), an HIS (Hospital Information System), a RIS (Radiology Information System), etc. via a LAN (Local Area Network), etc. The network IF transmits and receives various types of information to and from the connected workstation, PACS, HIS, and RIS.

[0034] The above configuration is merely an example and is not limiting. For example, the sequence control circuit 29 may be incorporated into the host computer 50. Furthermore, the sequence control circuit 29 and the processing circuit 51 may be mounted on the same board. Furthermore, the imaging condition setting function 512 does not necessarily have to be mounted in the processing circuit 51 of the magnetic resonance imaging apparatus 1. For example, the imaging condition setting function 512 may be mounted in a computer for setting imaging conditions that is separate from the magnetic resonance imaging apparatus 1. In this case, the imaging conditions generated by the computer are supplied to the magnetic resonance imaging apparatus 1 via a network, a portable recording medium, or the like. Furthermore, the storage area for the imaging conditions in the memory 52 does not have to be mounted in the magnetic resonance imaging apparatus 1, and may be mounted, for example, in a storage device connected to the magnetic resonance imaging apparatus 1 via a network.

[0035] 2 is a diagram showing an example of the flow of an MR examination according to this embodiment. When a user issues an instruction to start an MR examination via the input interface 54 or the like, the processing circuitry 51 reads out and executes a program related to the MR examination from the memory 52. ​​By executing the program, the processing circuitry 51 performs the series of processes shown in FIG. 2. The program may be configured by a single module in which the series of processes shown in FIG. 2 are described, or may be configured by multiple modules in which the series of processes are shared and described. It is assumed that, at the start of step S1, the imaging condition setting function 512 has set imaging conditions related to thinned data acquisition.

[0036] 2, the sequence control circuit 29 acquires time-series k-space data (hereinafter, k-space data corresponding to one time-series image is referred to as a k-space frame) by thinned data acquisition (step S1). In step S1, the sequence control circuit 29 synchronously controls the gradient magnetic field power supply 21, the transmission circuit 23, and the reception circuit 25 in accordance with imaging conditions, executes thinned data acquisition on the subject P, and acquires time-series k-space frames via the reception circuit 25. The k-space frames acquired by thinned data acquisition are also referred to as acquisition k-space frames, and their time resolution is also referred to as acquisition time resolution. In this embodiment, sample points at which data acquisition is performed are set so that the sample points of each time-series k-space frame after integration by the integration function 513 are thinned at equal intervals.

[0037] FIG. 3 is a diagram showing an example of the arrangement of sample points PS in thinned data acquisition according to this embodiment. FIG. 3 illustrates an example of the arrangement of sample points PS in time-series k-space. The horizontal axis represents the phase encoding direction axis ky, which is the k-space axis along which the sample points PS are thinned, and the vertical axis represents the time axis t. Data acquisition according to this embodiment uses the Cartesian method as a k-space filling trajectory, and therefore, the sample points PS are arranged in a grid pattern in k-space. This arrangement of sample points PS is called a Cartesian grid. In this embodiment, the sample points PS are thinned in the phase encoding direction, but are not thinned in the frequency encoding direction orthogonal to the phase encoding direction. Therefore, the frequency encoding direction is omitted in FIG. 3 and other figures. Note that the phase encoding direction does not have to be ky. Furthermore, the description based on FIG. 3 does not limit thinning to the two directions, ky and kz. For example, one of these directions may be treated equivalent to ky in FIG. 3, and the remaining direction may be treated as the target of a general thinned reconstruction technique.

[0038] In Figure 3, the black dots representing sample points PS1 represent sample points for which data acquisition is performed in thinned data acquisition, while the white dots representing sample points PS2 represent sample points for which data acquisition is not performed. One horizontal row corresponds to one k-space frame. The spacing between adjacent sample points PS corresponds to the spacing in a thinning rate of R=1, i.e., full sampling. In thinned data acquisition, sample points PS1 are arranged at intervals according to the thinning rate. Figure 3 illustrates the arrangement of sample points PS for a thinning rate of R=4. In this case, sample points PS1 are arranged at intervals four times the spacing in R=1. In other words, three sample points PS2 are arranged between adjacent sample points PS1. Strictly speaking, data acquisition is performed at different times for different sample points PS1 belonging to the same k-space frame, but different sample points PS1 belonging to the same k-space frame are considered to belong to the same time.

[0039] The sequence control circuit 29 performs thinning data acquisition while changing the arrangement of sample points PS1 in a time series. When the thinning rate R=M (M is a natural number greater than or equal to 2), the sample points PS1 are changed so that data acquisition is performed for all sample points PS when M k-space frames are acquired. Here, the arrangement pattern of sample points PS1 and PS2 in each k-space frame, in other words, at each time, is called an arrangement pattern. When the thinning rate R=4, four arrangement patterns are set. A set PSS of M arrangement patterns is repeated in time series. In this embodiment, the arrangement patterns or sample points PS1 are not set so that sample points PS1 move continuously in the ky direction over time. In this embodiment, sample points PS2 are set discontinuously so that the sample points in each time-series k-space frame after integration by the integration function 513 are thinned out at equal intervals. Details of the arrangement of sample points PS1 and PS2 will be described later. In the example of FIG. 3, when M k-space frames are collected, the sample points PS1 are changed so that data collection is performed for all sample points PS. However, data collection does not necessarily have to be performed for all sample points PS, and data collection may not be performed for any sample points.

[0040] When step S1 is performed, the processing circuitry 51, by implementing the integration function 513, integrates the time-series k-space frames (time-series acquired k-space frames) acquired in step S1 using multiple integration patterns to generate multiple time-series k-space frames (time-series integrated k-space frames) for multiple temporal resolutions (step S2). The processing circuitry 51 generates multiple time-series integrated k-space frames by sequentially extracting and integrating, from the time-series acquired k-space frames, a number of k-space frames corresponding to the temporal resolution or thinning rate corresponding to each of the multiple integration patterns, while shifting the time. The processing circuitry 51 integrates k-space frames that are consecutive or spaced apart in the time axis direction, a number of k-space frames corresponding to the temporal resolution or thinning rate.

[0041] In this embodiment, the processing circuitry 51 generates multiple time-series integrated k-space frames by sequentially applying integration windows corresponding to the time resolution or thinning rate corresponding to each of the multiple integration patterns to the time-series acquired k-space frames while shifting the time. When applying the integration window to the time-series acquired k-space frames, the processing circuitry 51 fixes or varies the position of the applied reference frame in the integration window over time. The integration window is a filter that extracts and integrates k-space frames included in the integration window. The applied reference frame refers to a k-space frame that is applied as a reference by the integration window. In other words, the applied reference frame is a k-space frame that serves as a reference for the integrated k-space frame. The same applied reference frame is processed by multiple integration windows with different time resolutions. Specifically, the integration may be performed by averaging the multiple time-series acquired k-space frames, such as a simple average or a weighted average.

[0042] 4 and 5 are diagrams schematically illustrating the integration process of time-series k-space frames. In FIG. 4, the k-space frames of arrangement pattern 2 are used as the application reference, and in FIG. 5, the k-space frames of arrangement pattern 3 are used as the application reference. As shown in FIGS. 4 and 5, when the thinning rate R=4, an integration window W1 corresponding to thinning rate R=1, an integration window W2 corresponding to thinning rate R=2, and an integration window W4 corresponding to thinning rate R=4 are set. The application reference k-space frame (hereinafter referred to as the application reference frame) for all integration windows W1, W2, and W4 is the same. The application reference frame is indicated by a star in FIGS. 4 and 5, etc. The integration windows W1, W2, and W4 extract and integrate the number of k-space frames corresponding to the corresponding thinning rate. In the integrated k-space frame, sample points are positioned at equal intervals based on the thinning rate. The temporal resolution improves in the order of integration windows W1, W2, and W4. Note that it is not essential to include unthinned data as temporal resolution. For example, thinning may be performed by a method that generates data of R=2 when all frames are integrated. Alternatively, for example, when thinning is applied to two directions, ky and kz, and integration processing is applied to ky, kz, to which integration processing is not applied, will be thinned out.

[0043] In this embodiment, the processing targets of the integration windows W1 and W2 are k-space frames in a time series including the applied reference frame. For example, since the integration window W1 corresponds to a thinning rate R=1, four k-space frames in a time series including the applied reference frame are extracted and integrated so that the integrated sample points are arranged at intervals corresponding to the thinning rate R=1. The position of the applied reference frame within the integration window W1 is not particularly limited. Assume that the position of the applied reference frame within the integration window W1 in FIGS. 4 and 5 is the second. That is, the integration window W1 includes the applied reference frame, a k-space frame that temporally precedes the applied reference frame, and a k-space frame that temporally follows the applied reference frame.

[0044] Since the integration window W2 corresponds to a thinning rate R=2, two k-space frames in a time series including the applicable reference frame are extracted and integrated so that the integrated sample points are arranged at an interval corresponding to the thinning rate R=2. The position of the applicable reference frame within the integration window W2 is also not particularly limited. In this case, the position of the applicable reference frame within the integration window W2 may be set to a different position for each application so that the integrated sample points are arranged at an interval corresponding to the thinning rate R=2. Specifically, when a k-space frame of arrangement pattern 2, which is the applicable reference frame, is integrated with a k-space frame of arrangement pattern 1 that precedes it in time, all sample points are arranged at equal intervals corresponding to the thinning rate R=2. However, when a k-space frame of arrangement pattern 2 is integrated with a k-space frame of arrangement pattern 3 that follows it in time, the sample points are not arranged at equal intervals. Therefore, as shown in Fig. 4, when the applied reference frame is arrangement pattern 2, of arrangement patterns 1 and 3 which are in a chronological relationship with arrangement pattern 2, arrangement pattern 1 is the one in which all sample points after integration are arranged at equal intervals corresponding to the thinning rate R = 2, and therefore the k-space frames of arrangement pattern 2 and arrangement pattern 1 are integrated. As shown in Fig. 5, when the applied reference frame is arrangement pattern 3, for the same reason, the k-space frames of arrangement pattern 3 and arrangement pattern 4 are integrated.

[0045] In this embodiment, the sequence control circuit 29 performs thinned data acquisition by discontinuously moving sample points along the time axis in a predetermined direction of k-space (e.g., the phase encoding direction) so that sample points are arranged at equal intervals in each of the multiple time-series k-space frames when the time-series k-space frames are integrated along a time series using each of the multiple integration patterns. In this embodiment, by integrating multiple acquired k-space frames in a time series, the sample points after integration are arranged at equal intervals corresponding to the desired temporal resolution. In other words, the sample point arrangement pattern for the thinned data acquisition is set so that the sample points after integration are arranged at equal intervals corresponding to the desired temporal resolution. In this case, the arrangement pattern may be set so that the sample points are arranged at equal intervals by varying the position of the applied reference frame within the integration window.

[0046] 6 is a diagram showing the arrangement patterns of k-space frames extracted by each integration window. The first frame is an applied reference frame having arrangement pattern 1, the second frame is an applied reference frame having arrangement pattern 2, the third frame is an applied reference frame having arrangement pattern 3, and the fourth frame is an applied reference frame having arrangement pattern 4. As shown in FIG. 6, the integration function 513 generates time-series k-space frames corresponding to multiple temporal resolutions (thinning rates) for each frame (each time point).

[0047] After step S2 is performed, the processing circuitry 51, by implementing the reconstruction function 514, converts the multiple time-series k-space frames generated in step S2 into multiple time-series images (step S3). In step S3, the processing circuitry 51 performs image transformation on each of the multiple time-series k-space frames, thereby converting them into multiple time-series images with multiple temporal resolutions. The image transformation method is not particularly limited as long as it converts from k-space frames to images, and possible methods include, for example, Fourier transform such as FFT (fast Fourier transform), parallel imaging reconstruction, compressed sensing reconstruction, and machine learning reconstruction. Here, a case will be described in which differential intra-frame SENSE, which is an application of parallel imaging reconstruction, is used as image transformation.

[0048] 7 is a diagram showing an example of image conversion processing using the difference intra-frame SENSE. Note that FIG. 7 shows one time-series k-space frame K1 to K N (N is a natural number of 2 or more that represents the number of k-space frames included in the time-series k-space frames) n The processing for (1≦n≦N) is illustrated.

[0049] As shown in FIG. 7, the processing circuitry 51 processes time-series k-space frames K1 to K N One k-space frame K to be processed from n The k-space frame K to be processed is extracted. n is the time series k space frames K1 to K N Next, the processing circuitry 51 extracts the time-series k-space frames K1 to K N Averaging is performed on the time series k-space frames K1 to K N Specifically, the average k-space frame Kave is generated based on the time-series k-space frames K1 to K2 (step S31). N and dividing by N. Instead of simple addition, the k-space frames K1 to K NThe average k-space frame Kave includes k-space data at all sample points, and therefore corresponds to fully sampled data.

[0050] When step S31 is performed, the processing circuitry 51 calculates the k-space frame K n and the average k-space frame Kave, and the k-space frame K n and the difference frame Ksub based on the average k-space frame Kave n (Step S32). n Specifically, k-space frame K n and the average k-space frame Kave for each sample point. n This is k-space data relating to the fluctuating portion of the

[0051] When step S32 is performed, the processing circuit 51 uses the sensitivity map to generate the difference frame Ksub n SENSE reconstruction is performed on the difference image Isub n (Step S33) SENSE reconstruction is an image reconstruction method that reconstructs an image of aliasing artifacts from k-space data obtained by undersampling (thinned-out data acquisition) using the sensitivity difference between multiple receiving coils.

[0052] Figure 8 is a diagram showing an overview of SENSE reconstruction. Figure 8 shows an example using two receive coils. Note that this can be expanded to cases where three or more receive coils are used. As shown in Figure 8, the actual scan (thinned data acquisition in step S1) results in an acquired image for coil 1 and an acquired image for coil 2. As described above, the k-space data is undersampled in the thinned data acquisition. Both images show aliasing artifacts caused by undersampling. A calibration scan is performed prior to the actual scan. The calibration scan results in a sensitivity map for coil 1 and a sensitivity map for coil 2. The sensitivity maps represent the spatial distribution of the sensitivity of each coil.

[0053] The pixel value of each pixel in an acquired image with aliasing artifacts can be expressed as a weighted sum of the pixel value (sensitivity) of the corresponding pixel in the sensitivity map and the pixel value of the corresponding pixel in an unfolded image without aliasing artifacts. Specifically, the pixel value I1 of a pixel in the acquired image of coil 1 can be obtained based on the pixel values ​​X1 and X2 of the corresponding pixel in the unfolded image and the sensitivities S11 and S12 of the corresponding pixel in the sensitivity map, as shown in equation (1). The pixel value I2 of a pixel in the acquired image of coil 2 can be obtained based on the pixel values ​​X1 and X2 of the corresponding pixel in the unfolded image and the sensitivities S21 and S22 of the corresponding pixel in the sensitivity map, as shown in equation (2). The unfolded image can be calculated by solving the simultaneous equations (1) and (2) for all pixels in the image.

[0054] I1=S11×X1+S12×X2 (1) I2=S21×X1+S22×X2 (2)

[0055] On the other hand, after step S32 is performed, the processing circuitry 51 performs FFT on the average k-space frame Kave to generate an image (hereinafter referred to as an average image) Iave (step S34). The average image Iave is generated by subtracting the average k-space frame Kave from the k-space frame K n8. In this example, since there is no thinning of sample points in the average k-space frame Kave, aliasing artifacts do not occur. Note that the method for generating the average image Iave from the average k-space frame Kave is not limited to FFT, and any method may be used. For example, if the sample points in the average k-space frame Kave are thinned data corresponding to R=2, processing similar to S33 (SENSE reconstruction) may be used in S34. Note that when there is no thinning of sample points in the average k-space frame Kave, the average k-space frame Kave may be used as self-calibration data in place of or in combination with the calibration scan in FIG. 8.

[0056] After steps S33 and S34 are performed, the processing circuit 51 generates the differential image Isub n The image Iadd is obtained by adding the average image Iave to the image Iadd (hereinafter referred to as the added image). n (Step S35). n is the k-space frame K n This corresponds to a high-quality image based on

[0057] The processing of steps S32 to S35 is performed on the time-series k-space frames K1 to K N All k-space frames K contained in n This is performed for the time-series k-space frames K1 to K N Thus, time-series images corresponding to the plurality of time resolutions are generated. The processes of steps S31 to S35 are performed on a plurality of time-series k-space frames corresponding to the plurality of time resolutions generated in step S2. As a result, a plurality of time-series images corresponding to the plurality of time resolutions are generated. As described above, according to the image conversion shown in FIG. 7, in order to generate a plurality of time-series images, the processing circuitry 51 calculates a time-series difference frame for each of the plurality of time-series k-space frames by subtracting an average k-space frame of the time-series k-space frame from the time-series k-space frame, and generates a time-series added image by adding each time-series difference image based on the time-series difference frame to an average image based on the average k-space frame.

[0058] According to this embodiment, when preparing a plurality of time-series k-space frames for a plurality of temporal resolutions, k-space frames to be used for integration are extracted from the original time-series k-space frames and integrated so that the sample points of each integrated time-series k-space frame are arranged at equal intervals according to each temporal resolution. Since the occurrence of complex aliasing components that occurs when the sample points are arranged at uneven intervals is suppressed, it is possible to improve the image quality of the time-series images.

[0059] After step S3 is performed, the processing circuitry 51 generates a single time-series image from the multiple time-series images generated in step S3 by implementing the reconstruction function 514 (step S4). In this embodiment, the processing circuitry 51 uses a machine learning model to generate a single high-resolution time-series image from the multiple time-series images in which artifacts caused by signal loss due to thinned data collection are reduced. A neural network is used as the machine learning model.

[0060] FIG. 9 is a diagram schematically illustrating an example of input and output of the machine learning model NN1. As shown in FIG. 9, the machine learning model NN1 is a neural network trained to input multiple time-series images and output high-resolution time-series images. For example, in the example shown in FIG. 9, three time-series images are input as the multiple time-series images: a time-series image with low time resolution, a time-series image with medium time resolution, and a time-series image with high time resolution. The number of input time-series images is not limited to three and may be any number equal to or greater than two. The temporal resolution of the output high-resolution time-series images depends on the temporal resolution of the output learning samples used in the machine learning, but can be set arbitrarily. For example, the temporal resolution of the high-resolution time-series images is set to the temporal resolution of the high-resolution time-series images.

[0061] The machine learning model NN1 is trained based on a large number of training samples collected in advance. The training samples include input training samples and output training samples. The input training samples include multiple time-series images with multiple time resolutions. The output training sample includes a single high-resolution time-series image. For example, the input training samples may be multiple time-series images generated by the processing of steps S1 to S3. For example, the output training samples may be time-series images based on time-series k-space frames collected by full-sampling data collection. The untrained machine learning model may be trained based on supervised learning using a combination of the input training samples and the output training samples. Specifically, the learning parameters may be updated so as to minimize the error between the output training sample and an output sample obtained by forward propagating the input training sample to the untrained machine learning model. The learning parameters are parameters of a neural network trained by machine learning, such as weight coefficients and biases. The machine learning model NN1 is generated by setting the updated learning parameters to the untrained machine learning model.

[0062] The machine learning model NN1 may be configured to input and output images in units of one frame, or in units of multiple frames, or in units of all frames.

[0063] The machine learning model NN1 outputs a single time-series image from multiple time-series images corresponding to multiple time resolutions, so it is possible to compensate for the degradation caused by thinned data collection using information from the surrounding area in time.

[0064] In step S4, a machine learning model with other inputs and outputs may also be used.

[0065] FIG. 10 is a diagram schematically illustrating an example of input and output of the machine learning model NN2. As shown in FIG. 10, the machine learning model NN2 receives multiple time-series images and an integrated pattern label and outputs a single high-resolution time-series image. The integrated pattern label is a label for identifying the integrated pattern of each image constituting the time-series image. Specifically, the integrated pattern label is a one-hot vector that identifies the position of the applicable reference frame within the integration window corresponding to the image. For example, in the case of an integration window corresponding to a thinning rate R=2, if the applicable reference frame is integrated with a k-space frame that precedes the applicable reference frame in time, the integrated pattern label may be (0,1), and if the applicable reference frame is integrated with a k-space frame that follows the applicable reference frame in time, the integrated pattern label may be (1,0). Similar integrated pattern labels can be defined for other thinning rates. By inputting the integrated pattern label, the machine learning model NN2 can recognize the position of the applicable reference frame, thereby generating high-resolution time-series images that take the position of the applicable reference frame into account.

[0066] Note that high-resolution time-series images may be generated without using a machine learning model. For example, the processing circuitry 51 may generate high-resolution time-series images by averaging multiple time-series images in units of the same frame. Alternatively, the processing circuitry 51 may generate high-resolution time-series images by performing compressed sensing or the like on multiple time-series images.

[0067] When step S4 is performed, the processing circuitry 51 displays the time-series images generated in step S4 by implementing the display control function 516 (step S5). The time-series images are dynamically displayed on the display 53 or the like. This allows the user to observe the time-series images in high resolution.

[0068] When step S5 is performed, the MR examination shown in FIG. 2 ends.

[0069] The above-described embodiment is merely an example, and the present embodiment is not limited to this. Below, several modifications of the present embodiment will be described.

[0070] (Variation 1) In the above embodiment, a plurality of temporally consecutive k-space frames are integrated. Therefore, the arrangement pattern of sample points in the thinned data acquisition is set so that the sample points are arranged at equal intervals in the integrated k-space frame. In this case, as shown in FIGS. 4 and 5, the sample points move discontinuously in the k-space over time in the thinned data acquisition. In Modification 1, thinned data acquisition is performed so that the sample points move continuously in the k-space over time. The integration process of time-series frames according to Modification 1 will be described below.

[0071] FIG. 11 is a diagram schematically illustrating integration processing of time-series k-space frames according to Modification 1. The arrangement pattern numbers correspond to the arrangement pattern numbers shown in FIGS. 4 and 5. As shown in FIG. 11, in Modification 1, the sequence control circuit 29 performs thinned data acquisition so as to continuously move sample points along the time axis in a predetermined direction of k-space. That is, arrangement pattern 1 → arrangement pattern 3 → arrangement pattern 2 → arrangement pattern 4 are cyclically repeated in time series. An integration window W5 corresponding to a thinning rate R=1 is sequentially applied to the time-series k-space frames in time series.

[0072] The processing circuitry 51 extracts a combination of k-space frames from the k-space frames included in the integration window W5 according to the desired temporal resolution or thinning rate so that the sample points are arranged at equal intervals in the integrated k-space frame. The extracted k-space frames do not need to be consecutive in time. As shown in FIG. 11 , when the applicable reference frame is arrangement pattern 2 and time-series k-space frames with a thinning rate R=2 are generated, k-space frames of arrangement pattern 2 and arrangement pattern 1 that are separated in time are extracted so that the interval between sample points in the integrated k-space frame is two.

[0073] As described above, in Modification 1, thinned data acquisition is performed so that sample points move continuously in k-space over time, and a number of k-space frames corresponding to a desired temporal resolution are extracted from the time-series acquired k-space frames so that the integrated sample points are positioned at equal intervals corresponding to the desired temporal resolution. In this case, multiple k-space frames spaced apart in time are permitted to be extracted. This eliminates the need for complex manipulation of sample point positions during thinned data acquisition, thereby reducing mechanical load.

[0074] (Variation 2) In some of the above-described embodiments, the arrangement patterns in thinned data acquisition are cyclically repeated in the order of arrangement pattern 1 → arrangement pattern 2 → arrangement pattern 3 → arrangement pattern 4 → arrangement pattern 1 → arrangement pattern 2 → arrangement pattern 3 → arrangement pattern 4, as shown in FIGS. 4 and 5 . However, this embodiment is not limited to this. For example, the arrangement patterns may be repeated in a reciprocating manner in the order of arrangement pattern 1 → arrangement pattern 2 → arrangement pattern 3 → arrangement pattern 4 → arrangement pattern 3 → arrangement pattern 2 → arrangement pattern 1, as shown in FIGS. 4 and 5 . In this case, the position of the applied reference frame within the integration window may be adjusted over time so that the sample points of the integrated k-space frame are arranged at equal intervals.

[0075] (Variation 3) In some of the above examples, it is assumed that the sample points are arranged at equal intervals throughout the entire phase encoding direction (ky direction). However, this embodiment is not limited to this. The sample points do not have to be arranged at equal intervals throughout the phase encoding direction (ky direction), and local uneven spacing may be permitted. For example, in parallel imaging using a k-space method such as GRAPPA (generalized autocalibrating partially parallel acquisitions), additional data acquisition (ACS data acquisition) may be performed using sample points that have been thinned out in the center of k-space. This corresponds to full sampling in the center of k-space where the additional data acquisition is performed, and undersampling in the peripheral k-space.

[0076] In this case, sample points are arranged at unequal intervals between the central part of k-space where additional data has been acquired and the peripheral part of k-space where additional data has not been acquired. Even in such a case, it is possible to perform integration processing of k-space frames while focusing on the peripheral part of k-space. For example, it is preferable to extract and integrate multiple k-space frames so that sample points in the peripheral part of k-space in the integrated k-space frame are arranged at equal intervals corresponding to the temporal resolution (thinning rate) of the integrated k-space frame.

[0077] That is, according to this embodiment, it is sufficient that sample points are arranged at regular intervals across the entire area or locally in the phase encoding direction of k-space in the time series acquisition k-space frames and multiple time series integrated k-space frames.

[0078] (Variation 4) The generation of multiple time-series k-space frames by the integration function 513 and / or the reconstruction of time-series images by the reconstruction function 514 may be incorporated as part of an iterative optimization using MoDL (Model-based Deep Learning) or the like.

[0079] (Additional remarks) In some of the above embodiments, the magnetic resonance imaging apparatus 1 includes a sequence control circuit 29 and a processing circuit 51. The sequence control circuit 29 performs thinned data acquisition while moving sample points at equal intervals in k-space to acquire time-series k-space frames. The processing circuit 51 generates multiple time-series k-space frames with multiple temporal resolutions based on the time-series k-space frames. The multiple time-series k-space frames have sample points positioned at equal intervals that are different from each other. The processing circuit 51 generates a single time-series image based on the multiple time-series k-space frames.

[0080] The above configuration enables generation of a single high-resolution time-series image based on multiple time-series k-space frames with multiple temporal resolutions in relation to thinned data acquisition. In this case, when preparing the multiple time-series k-space frames, the sample points of each integrated time-series k-space frame are arranged at equal intervals according to the respective temporal resolutions. In multiple time-series images based on such multiple time-series k-space frames, the occurrence of complex aliasing components that occurs when sample points are arranged at uneven intervals is suppressed. As a result, the image quality of a single time-series image based on multiple time-series images is also expected to improve.

[0081] According to at least one of the embodiments described above, it is possible to improve the quality of an image obtained by thinned data collection.

[0082] The term "processor" used in the above description refers to a circuit such as a CPU, a GPU, an application specific integrated circuit (ASIC), a programmable logic device (e.g., a simple programmable logic device (SPLD), a complex programmable logic device (CPLD), and a field programmable gate array (FPGA)). A processor realizes its functions by reading and executing a program stored in a memory circuit. Note that instead of storing a program in a memory circuit, a program may be directly embedded in the processor circuit. In this case, the processor realizes its functions by reading and executing the program embedded in the circuit. Alternatively, instead of executing a program, a function corresponding to the program may be realized by combining logic circuits. Note that each processor in this embodiment is not limited to being configured as a single circuit, but may be configured as a single processor by combining multiple independent circuits to realize its function. Furthermore, multiple components in FIG. 1 may be integrated into a single processor to realize its function.

[0083] Although several embodiments have been described, these embodiments are presented as examples and are not intended to limit the scope of the invention. These embodiments can be implemented in various other forms, and various omissions, substitutions, modifications, and combinations of embodiments can be made without departing from the spirit of the invention. These embodiments and their modifications are included within the scope and spirit of the invention, as well as within the scope of the invention and its equivalents as defined in the claims. [Explanation of symbols]

[0084] 1. Magnetic resonance imaging device 11 Mounting stand 13 berths 21 Gradient magnetic field power supply 23 Transmitting circuit 25 Receiving circuit 27 Bed drive unit 29 Sequence control circuit 41 Static magnetic field magnet 43 Gradient magnetic field coil 45 Transmitting coil 47 receiving coil 50 Host Computer 51 Processing circuit 52 memory 53 Display 54 Input Interface 55 Communication Interface 131 Top plate 133 Foundation 511 Acquisition Function 512 Imaging condition setting function 513 Integration Features 514 Reconfiguration function 515 Image Processing Function 516 Display Control Function

Claims

1. an acquisition unit that acquires first time-series k-space frames by performing thinned data acquisition while moving sample points at equal intervals in k-space; an integration unit that generates a plurality of second time-series k-space frames corresponding to a plurality of time resolutions based on the first time-series k-space frames, and the plurality of second time-series k-space frames have sample points positioned at equal intervals that are different from each other; a reconstruction unit that generates a single time-series image based on the plurality of second time-series k-space frames; Equipped with the integrating unit integrates a plurality of k-space frames included in the first time-series k-space frames with a plurality of patterns corresponding to the plurality of time resolutions, respectively, to generate the plurality of second time-series k-space frames. Magnetic resonance imaging device.

2. 2. The magnetic resonance imaging apparatus according to claim 1, wherein the integrating unit generates the plurality of second time-series k-space frames by sequentially extracting and integrating a number of k-space frames corresponding to a time resolution or a thinning rate corresponding to each of the plurality of patterns from the first time-series k-space frames while shifting the time.

3. The integration unit generating the plurality of second time-series k-space frames by sequentially applying integration windows according to time resolutions or thinning rates corresponding to the plurality of patterns to the first time-series k-space frames while shifting the time; When applying the integration window to the first time-series k-space frames, a position of an applied reference frame in the integration window is fixed or varied over time.

3. The magnetic resonance imaging apparatus according to claim 2.

4. the acquisition unit performs the thinned-out data acquisition so as to move sample points discontinuously along a time axis in a predetermined direction of k-space so that sample points are arranged at equal intervals in each of the plurality of second time-series k-space frames when the first time-series k-space frames are integrated along a time series with each of the plurality of patterns; the integrating unit integrates the number of temporally consecutive k-space frames such that sample points are arranged at equal intervals in each of the plurality of second time-series k-space frames.

3. The magnetic resonance imaging apparatus according to claim 2.

5. the acquisition unit performs the thinned-out data acquisition so as to continuously move sample points along a time axis in a predetermined direction in k-space; the integrating unit integrates the number of temporally discontinuous k-space frames such that sample points are arranged at equal intervals in each of the plurality of second time-series k-space frames.

3. The magnetic resonance imaging apparatus according to claim 2.

6. The reconstruction unit transforming the plurality of second time-series k-space frames into a plurality of time-series images; generating the single time-series image based on the plurality of time-series images; 2. The magnetic resonance imaging apparatus according to claim 1.

7. 7. The magnetic resonance imaging apparatus according to claim 6, wherein the reconstruction unit calculates, for each of the plurality of second time-series k-space frames, a time-series difference frame by subtracting an average frame of the second time-series k-space frames from the second time-series k-space frame, and generates a time-series added image by adding each time-series difference image based on the time-series difference frame to an average image based on the average frame, in order to generate the plurality of time-series images.

8. 7. The magnetic resonance imaging apparatus according to claim 6, wherein the reconstruction unit applies the plurality of time-series images to a machine learning model to generate the single time-series image in which artifacts caused by signal loss due to the thinned data acquisition are reduced.

9. the reconstruction unit applies the plurality of time-series images and a plurality of labels corresponding to the plurality of time-series images to a machine learning model to generate the single time-series image; each of the plurality of labels representing a pattern related to an integration of k-space frames used to transform a corresponding time series of images; 9. The magnetic resonance imaging apparatus according to claim 8.

10. 2. The magnetic resonance imaging apparatus according to claim 1, wherein sample points are arranged at equal intervals across the entire area or locally in the phase encoding direction of k-space in the first time-series k-space frame and the plurality of second time-series k-space frames.

11. The magnetic resonance imaging apparatus according to claim 1 , wherein the acquisition unit performs the thinned-out data acquisition while changing the arrangement of the sample points in the k-space in time series.

12. an acquisition step of acquiring first time-series k-space frames by performing thinned data acquisition while moving sample points positioned at equal intervals in k-space; an integration step of generating a plurality of second time-series k-space frames corresponding to a plurality of time resolutions based on the first time-series k-space frames, wherein the second time-series k-space frames have sample points positioned at equal intervals that are different from each other; a reconstruction step of generating a single time-series image based on the plurality of second time-series k-space frames; Equipped with the integrating step integrates a plurality of k-space frames included in the first time-series k-space frames with a plurality of patterns corresponding to the plurality of time resolutions, respectively, to generate the plurality of second time-series k-space frames; Magnetic resonance imaging methods.

Images