WO2019124076A1

WO2019124076A1 - Medical signal processing device

Info

Publication number: WO2019124076A1
Application number: PCT/JP2018/044729
Authority: WO
Inventors: 修平新田; 竹島　秀則
Original assignee: キヤノンメディカルシステムズ株式会社
Priority date: 2017-12-20
Filing date: 2018-12-05
Publication date: 2019-06-27

Abstract

A medical signal processing device according to an embodiment has a processing unit. The processing unit inputs a medical signal, which has a pattern of appearing at a position shifted by a known shift amount in a known direction, into a learned model having the function of outputting at least any one of a correction signal in which the medical signal is corrected to reduce the pattern, pattern-related information relating to the pattern, and disease information relating to the medical signal. Using the direction and the shift amount, the processing unit outputs any one of the correction signal, the pattern-related information, and the disease information.

Description

Medical signal processing device

Embodiments of the present invention relate to a medical signal processing apparatus.

Conventionally, in magnetic resonance imaging, artifacts due to various factors can occur in an image after Fourier transform or inverse Fourier transform. For example, when undersampling is performed to thin and collect lines in k space, aliasing may occur in the image. Also, for example, when acquisition is performed by EPI (Echo Planar Imaging), a virtual image called an N / 2 artifact may be generated in the image.

If a convolutional neural network (hereinafter referred to as CNN) or the like trained to reduce these artifacts is used for the image, the conventional CNN framework does not improve the accuracy of artifact reduction. There is an input target. That is, when a magnetic resonance image having aliasing such as N / 2 artifact is used as an input target to CNN, for example, the accuracy for reducing the aliasing may not be improved.

The problem to be solved by the present invention is to reduce the output error due to the learned model.

The medical signal processing device according to the embodiment has a processing unit. The processing unit includes: a correction signal corrected to reduce the pattern with respect to a medical signal having a pattern appearing at a position shifted by a known shift amount along a known direction; pattern related information related to the pattern; The medical signal is input to a learned model that is functioned to output any one of disease information related to the medical signal, and the correction signal is input using the direction and the shift amount And / or one of the pattern related information and the disease information.

FIG. 1 is a view showing an example of the arrangement of an image processing apparatus and a magnetic resonance imaging apparatus according to the embodiment. FIG. 2 is a diagram for explaining the forward propagation function by the processing circuit according to the present embodiment. FIG. 3 is a diagram for explaining the forward propagation function by the processing circuit according to the application example of the present embodiment. FIG. 4 is a diagram showing an example of a detailed configuration regarding the magnetic resonance imaging apparatus in the present embodiment. FIG. 5 is a flowchart showing an example of the processing procedure in the image generation processing in the present embodiment. FIG. 6 is a diagram showing an example of a temporary image, the number of channels of a first convolution result for the temporary image, and the number of channels of a second convolution result for the first convolution result in the present embodiment. FIG. 7 is a diagram showing an example of generation of a magnetic resonance image using the Nth convolution result by the Nth convolution layer in the present embodiment. FIG. 8 shows, in the application example of this embodiment, a 2-channel image obtained by dividing the temporary image into two, the number of channels resulting from the first convolution of 2-channel with respect to the 2-channel image, and 2 for the first convolution result obtained with 2-channel. It is a figure which shows an example with the channel number of the 2nd convolution result of a channel. FIG. 9 shows, in an application example of the present embodiment, generation of a two-channel magnetic resonance image using a two-channel N-th convolution result by the N-th convolutional layer and a magnetic resonance obtained by combining the two-channel magnetic resonance image It is a figure which shows an example with an image. FIG. 10 is a diagram showing an example of the configuration of a medical signal processing apparatus in an application example of the present embodiment. FIG. 11 is a flowchart illustrating an example of the procedure of the artifact reduction process in the first application example of the present embodiment. FIG. 12 is a diagram showing an example of cyclic shift processing for a magnetic resonance image having a reduction factor corresponding to 2 and having aliasing artifacts along the phase encoding direction in the first application example of the present embodiment. FIG. 13 is a diagram showing an example of cyclic shift processing for a magnetic resonance image having a reduction factor corresponding to 3 and having aliasing artifacts along the phase encoding direction in the first application example of the present embodiment. FIG. 14 is a flowchart illustrating an example of the procedure of the information generation process in the second application example of the present embodiment. FIG. 15 is a diagram showing an example of an electrocardiogram waveform as a biological signal in the third application example of the present embodiment. FIG. 16 is a diagram showing an example of a medical signal processing apparatus in a fourth application example of the present embodiment. FIG. 17 is a flowchart illustrating an example of the procedure of the information generation process in the fourth application example of the present embodiment. FIG. 18 is a diagram showing an example of aliasing preprocessing in the fourth application example of the present embodiment. FIG. 19 is a view showing an example of a medical signal processing apparatus in a fifth application example of the present embodiment. FIG. 20 is a flowchart illustrating an example of a procedure of combined image generation processing in the fifth application example of the present embodiment. FIG. 21 is a diagram showing an example of the post-aliasing process in the fifth application example of the present embodiment.

Hereinafter, embodiments of an image processing apparatus and a magnetic resonance imaging apparatus will be described in detail with reference to the drawings.

FIG. 1 is a view showing an example of the arrangement of an image processing apparatus and a magnetic resonance imaging apparatus according to the embodiment. For example, as shown in FIG. 1, the magnetic resonance imaging apparatus 100 according to this embodiment includes an image processing apparatus 150 in addition to components such as a static magnetic field magnet, a gradient magnetic field coil, and a high frequency coil, which are not shown. . In the present embodiment, the image processing device 150 generates a magnetic resonance image. The image processing apparatus 150 is, for example, a dedicated apparatus for generating a magnetic resonance image, or an apparatus used in combination with other functions. Further, in the present embodiment, the image processing apparatus 150 is described as a component of the magnetic resonance imaging apparatus 100, but the embodiment is not limited thereto. For example, the function executed in the image processing apparatus 150 is It may be another device communicably connected to the magnetic resonance imaging apparatus 100. In this case, another device as the image processing device 150 may be installed at another site outside the hospital.

The image processing apparatus 150 further includes a processing circuit 151, a memory 152, and an input / output interface 153. The processing circuit 151 is, for example, a central processing unit (CPU), a graphics processing unit (GPU), an application specific integrated circuit (ASIC), a programmable logic device (for example, a simple programmable logic device (Simple Programmable logic device). Logic Device (SPLD), Complex Programmable Logic Device (CPLD), Field Programmable Gate Array (FPGA), and the like.

The processing circuit 151 realizes a function by reading and executing a program stored in the memory 152. The program may be directly incorporated into the processing circuit 151 instead of storing the program in the memory 152. In this case, the processing circuit 151 realizes a function by reading and executing a program incorporated in the own circuit. Further, instead of reading and executing a program, a function corresponding to the program may be realized by a combination of logic circuits. Note that the processing circuit 151 of the present embodiment is not limited to the case where each processing circuit 151 is configured as a single circuit, and a plurality of independent circuits are combined to form one processing circuit 151 to realize its function. You may do it.

The processing circuit 151 according to the present embodiment generates a magnetic resonance image using deep learning, which is one of machine learning. General deep learning is a deeper layer of "neural network", which is an algorithm modeled on neurons in the brain of an organism. Further, the processing circuit according to the present embodiment generates a magnetic resonance image using a method called CNN (Convolution Neural Network) even in deep learning. In the general CNN method, local features of an image are extracted by performing image filter processing on nodes located in the vicinity of the pixel of interest in the previous layer in the intermediate layer.

On the other hand, the processing circuit 151 according to the present embodiment focuses on the occurrence of artifacts called aliasing and virtual images in the magnetic resonance image, and performs image filtering on nodes located in the vicinity of the pixel of interest. In addition to the above, the filtering process is performed on nodes that are located in a distant area away from the pixel of interest, which is different from the neighboring area.

FIG. 2 is a diagram for explaining the forward propagation function by the processing circuit 151 according to the present embodiment. In FIG. 2, when applying the learned model to the magnetic resonance image in which aliasing or a virtual image is generated, the processing circuit 151 sets nodes positioned in a plurality of regions (for example, the above-described near region and separated regions). Filter the object. The positions of the plurality of regions are determined according to the imaging conditions. The imaging conditions include, for example, a reduction factor (the number of thinning steps) indicating the degree of line thinning of k space in PI (Parallel Imaging), an imaging parameter of a pulse sequence of FOV (Field Of View), EPI method, etc. . The processing circuit 151 calculates aliasing, a virtual image, a shift due to a chemical shift, and the like in the magnetic resonance image according to the imaging condition, and derives a plurality of regions in which the same pixel may exist in the magnetic resonance image. The region includes one or more pixels. Then, the processing circuit 151 applies a filtering process to nodes located in the plurality of derived regions. That is, the processing circuit 151 is connected to one of the first nodes in the intermediate layer of the previous stage connected to the input side to each of the plurality of intermediate layers for each of the plurality of intermediate layers of the convolutional neural network corresponding to the learned model. And the output from the second node determined by the imaging condition in the intermediate layer of the previous stage are processed to be input together. The magnetic resonance image output from the learned model corresponds to an artifact reduced image with reduced artifacts.

The processing circuit 151 may be designed in the spatial direction as shown in FIG. 2 when the target of the filtering process is derived so as to include both the pixels in the vicinity region of the target pixel and the pixels in the separation region. And may be designed in the channel direction as shown in FIG. FIG. 3 is a diagram for explaining the forward propagation function by the processing circuit according to the application example of the present embodiment.

An example of the flow of processing by the magnetic resonance imaging apparatus 100 according to the present embodiment will be described. First, the magnetic resonance imaging apparatus 100 acquires magnetic resonance signals to obtain k-space data by executing pulses in accordance with predetermined imaging conditions. Further, the image processing device 150 performs Fourier transform or inverse Fourier transform on the obtained k-space data to generate a magnetic resonance image. Next, the image processing apparatus 150 reads the learned model stored in the memory 152, performs forward propagation processing on the generated magnetic resonance image, and displays the magnetic resonance image whose image quality is improved compared to the input image. Output to the output interface of the device etc. The learned model used by the image processing apparatus 150 at the time of the forward propagation process is one in which the position of the area to be subjected to the filtering process is specified according to the imaging condition at the time of collecting the input image. For example, the image processing apparatus 150 stores a plurality of learned models corresponding to the imaging condition in the memory 152, and during the forward propagation process, the learned model matching the imaging condition is selected from among the plurality of learned models. select.

Hereinafter, embodiments of a magnetic resonance imaging apparatus and an image processing apparatus will be described in detail with reference to the drawings. In the following description, components having substantially the same function and configuration are given the same reference numerals, and repeated description will be made only when necessary.

(Embodiment)
The entire configuration of the magnetic resonance imaging apparatus 100 according to the present embodiment will be described with reference to FIGS. 1 and 4. FIG. 4 is a diagram showing an example of a detailed configuration of the magnetic resonance imaging apparatus 100 in the present embodiment. As shown in FIG. 4, the magnetic resonance imaging apparatus 100 includes a static magnetic field magnet 101, a gradient magnetic field coil 103, a gradient magnetic field power supply 105, a bed 107, a bed control circuit (system control unit) 109, and a transmission circuit ( Transmission unit) 113, transmission coil 115, reception coil 117, reception circuit (reception unit) 119, imaging control circuit (collection unit) 121, system control circuit (system control unit) 123, storage device 125 , And an image processing apparatus 150. The subject P is not included in the magnetic resonance imaging apparatus 100.

The static magnetic field magnet 101 is a hollow substantially cylindrical magnet. The static magnetic field magnet 101 generates a substantially uniform static magnetic field in the internal space. For example, a superconducting magnet or the like is used as the static magnetic field magnet 101.

The gradient magnetic field coil 103 is a hollow and substantially cylindrical coil. The gradient magnetic field coil 103 is disposed inside the static magnetic field magnet 101. The gradient coil 103 is formed by combining three coils corresponding to X, Y, and Z axes orthogonal to one another. The Z-axis direction is assumed to be the same as the direction of the static magnetic field. The Y-axis direction is a vertical direction, and the X-axis direction is a direction perpendicular to the Z-axis and the Y-axis. The three coils in the gradient magnetic field coil 103 are individually supplied with current from the gradient magnetic field power supply 105 to generate gradient magnetic fields whose magnetic field strengths change along the X, Y, and Z axes.

The gradient magnetic fields of the X, Y, and Z axes generated by the gradient magnetic field coil 103 form, for example, a gradient magnetic field for slice selection, a gradient magnetic field for phase encoding, and a gradient magnetic field for frequency encoding (also referred to as readout gradient magnetic field). . The slice selection gradient magnetic field is used to arbitrarily determine the imaging cross section. The phase encoding gradient magnetic field is used to change the phase of the magnetic resonance signal according to the spatial position. The frequency encoding gradient magnetic field is used to change the frequency of the magnetic resonance signal according to the spatial position. In addition, the gradient magnetic fields of the X, Y, and Z axes generated by the gradient magnetic field coil 103 make the direction of the gradient magnetic field twice in order to refocus the phase of the spin on the XY plane in the gradient echo method. It is used as an inverted refocusing pulse. In addition, the gradient magnetic fields of the X, Y, and Z axes generated by the gradient coil 103 are used as an offset of the primary shimming of the static magnetic field.

The gradient magnetic field power supply 105 is a power supply device that supplies a current to the gradient magnetic field coil 103 under the control of the imaging control circuit 121.

The bed 107 is a device provided with a top plate 1071 on which the subject P is placed. The bed 107 inserts the top plate 1071 on which the subject P is placed into the bore 111 under the control of the bed control circuit 109. The bed 107 is installed, for example, in the examination room such that the longitudinal direction is parallel to the central axis of the static magnetic field magnet 101.

The bed control circuit 109 is a circuit that controls the bed 107. The bed control circuit 109 drives the bed 107 in accordance with an instruction from the operator via the input / output interface 153, thereby moving the top plate 1071 in the longitudinal direction, the up-down direction, and in some cases, the left-right direction.

The transmission circuit 113 supplies a high frequency pulse modulated at the Larmor frequency to the transmission coil 115 under the control of the imaging control circuit 121.

The transmission coil 115 is an RF coil disposed inside the gradient coil 103. The transmission coil 115 generates an RF (Radio Frequency) pulse corresponding to a high frequency magnetic field in response to the output from the transmission circuit 113. The transmission coil 115 is, for example, a whole-body coil (hereinafter referred to as a WB (Whole Body) coil) having a plurality of coil elements. The WB coil may be used as a transmit and receive coil. In addition, the transmission coil 115 may be a WB coil formed by one coil.

The receiving coil 117 is an RF coil disposed inside the gradient coil 103. The receiving coil 117 receives the magnetic resonance signal emitted from the subject P by the high frequency magnetic field. The receiving coil 117 outputs the received magnetic resonance signal to the receiving circuit 119. The receiving coil 117 is, for example, a coil array having one or more, typically a plurality of coil elements. Although the transmission coil 115 and the reception coil 117 are described as separate RF coils in FIG. 1, the transmission coil 115 and the reception coil 117 may be implemented as an integrated transmission / reception coil. The transmission / reception coil corresponds to the imaging region of the subject P, and is, for example, a local transmission / reception RF coil such as a head coil.

The receiving circuit 119 generates digital magnetic resonance signals (hereinafter, referred to as magnetic resonance data) based on the magnetic resonance signals output from the receiving coil 117 under the control of the imaging control circuit 121. Specifically, the receiving circuit 119 performs various signal processing on the magnetic resonance signal output from the receiving coil 117, and then performs analog / digital (A / D ( Perform Analog to Digital) conversion. The receiving circuit 119 samples (samples) A / D converted data. Thereby, the receiving circuit 119 generates magnetic resonance data. The receiving circuit 119 outputs the generated magnetic resonance data to the imaging control circuit 121.

The imaging control circuit 121 controls the gradient magnetic field power source 105, the transmitting circuit 113, the receiving circuit 119, and the like in accordance with the imaging protocol output from the processing circuit 151, and performs imaging on the subject P. The imaging protocol has various pulse sequences depending on the examination. In the imaging protocol, the magnitude of the current supplied to the gradient coil 103 by the gradient power supply 105, the timing when the current is supplied to the gradient coil 103 by the gradient power supply 105, and the transmission coil 113 are supplied by the transmission circuit 113. The timing and width of the high frequency pulse, the timing when the transmission circuit 113 supplies the high frequency pulse to the transmission coil 115, the timing when the magnetic resonance signal is received by the reception coil 117, etc. are defined.

The system control circuit 123 has a processor (not shown) as a hardware resource, a memory such as a ROM (Read-Only Memory) and a RAM (Random Access Memory), etc., and controls the magnetic resonance imaging apparatus 100 by a system control function. . Specifically, the system control circuit 123 reads out the system control program stored in the storage device 125, expands it on the memory, and controls each circuit of the magnetic resonance imaging apparatus 100 according to the expanded system control program. For example, the system control circuit 123 reads out the imaging protocol from the storage device 125 based on the imaging condition input by the operator via the input / output interface 153. The system control circuit 123 may generate an imaging protocol based on the imaging conditions. The system control circuit 123 transmits an imaging protocol to the imaging control circuit 121 and controls imaging of the subject P. When the image processing apparatus 150 is mounted on the magnetic resonance imaging apparatus 100, the system control circuit 123 may be incorporated in the processing circuit 151. At this time, the system control function is executed by the processing circuit 151, and the processing circuit 151 functions as a substitute for the system control circuit 123.

The storage device 125 stores various programs executed by the system control circuit 123, various imaging protocols, imaging conditions including a plurality of imaging parameters defining the imaging protocol, and the like. The storage device 125 is, for example, a RAM, a semiconductor memory element such as a flash memory, a hard disk drive (Hard Disk Drive), a solid state drive, an optical disk or the like. In addition, the storage device 125 may be a drive device or the like 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. When the image processing apparatus 150 is mounted on the magnetic resonance imaging apparatus 100, data stored in the storage device 125 may be stored in the memory 152. At this time, the memory 152 functions as a substitute for the storage device 125.

The image processing apparatus 150 includes a processing circuit 151, a memory 152, and an input / output interface 153. The processing circuit 151 has a reconstruction function 1511, a selection function 1513, and an image generation function 1515. Various functions performed by the reconstruction function 1511, the selection function 1513, and the image generation function 1515 are stored in the memory 152 in the form of a program that can be executed by a computer. The processing circuit 151 is a processor that reads programs corresponding to these various functions from the memory 152 and executes the read programs to realize the functions corresponding to the respective programs. In other words, the processing circuit 151 in the state where each program is read has a plurality of functions and the like shown in the processing circuit 151 of FIG. The reconstruction function 1511, the selection function 1513, and the image generation function 1515 will be described in detail later.

Although FIG. 1 has been described that these various functions are realized by a single processing circuit 151, a plurality of independent processors are combined to form the processing circuit 151, and each processor executes a program. The function may be realized by In other words, each function described above may be configured as a program, and one processing circuit may execute each program, or a specific function may be implemented in a dedicated independent program execution circuit. May be The reconstruction function 1511, the selection function 1513, and the image generation function 1515 included in the processing circuit 151 are an example of a reconstruction unit, a selection unit, and an image generation unit. The term "processor" used in the above description means, for example, a circuit such as a CPU, a GPU or an ASIC, a programmable logic device (SPLD, CPLD, and FPGA). The bed control circuit 109, the transmission circuit 113, the reception circuit 119, the imaging control circuit 121, the system control circuit 123, and the like are similarly configured by electronic circuits such as the processor.

The processing circuit 151 fills the magnetic resonance data along the readout direction of the k space by the reconstruction function 1511 according to the strength of the readout gradient magnetic field. The processing circuit 151 generates a magnetic resonance image by performing Fourier transform or inverse Fourier transform on the magnetic resonance data filled in the k space. The processing circuit 151 outputs the magnetic resonance image to the memory 152 and the input / output interface 153.

The memory 152 stores magnetic resonance data filled in the k space through the reconstruction function 1511, image data generated by the image generation function 1515, and the like. The memory 152 stores programs corresponding to various functions executed by the processing circuit 151. The memory 152 is, for example, a semiconductor memory device.

The input / output interface 153 has an input interface and an output interface. The input interface includes, for example, a pointing device such as a mouse or a circuit related to the input device such as a keyboard, an input terminal from a network, and the like. Note that the circuits included in the input interface are not limited to circuits related to physical operation parts such as a mouse and a keyboard. For example, the input interface receives an electrical signal corresponding to an input operation from an external input device provided separately from the magnetic resonance imaging apparatus 100, and outputs the received electrical signal to various circuits. And the processing circuit of The output interface is, for example, a display, an output terminal to a network, or the like. The display displays various magnetic resonance images reconstructed by the reconstruction function 1511, various magnetic resonance images generated by the image generation function 1515, and various information related to imaging and image processing under the control of the system control function. Do. The display is, for example, a display device such as a CRT display, a liquid crystal display, an organic EL display, an LED display, a plasma display, or any other display, monitor or the like known in the art.

The above is the description of the overall configuration of the magnetic resonance imaging apparatus 100 according to the present embodiment. The image generation processing implemented by the reconstruction function 1511, the selection function 1513, and the image generation function 1515 in the present embodiment will be described below. The image generation process in the present embodiment applies a learned model corresponding to the imaging condition when the magnetic resonance image input to the image processing apparatus 150 is acquired by magnetic resonance imaging to the input magnetic resonance image. The present invention is to output a magnetic resonance image with improved image quality by performing forward propagation processing to improve the image quality.

The memory 152 stores a plurality of learned models learned by a model learning device (not shown) in association with a plurality of imaging conditions as a program. The plurality of imaging conditions include, as described above, a reduction factor indicating a thinning rate in the thinning-out acquisition at equal intervals in k space, an imaging parameter of a pulse sequence of FOV, EPI method, and the like. Hereinafter, prior to the description of the image generation processing, the learned model will be described.

The learned model is generated by a model learning device (not shown). Specifically, the model learning device generates a learned model by causing the model before machine learning to perform machine learning according to the model learning program, based on the learning data stored in the learning data storage device (not shown). Do. The model learning device is a computer such as a workstation having a processor such as a CPU and a GPU. The model learning device and the learning data storage device may be communicably connected via a cable or a communication network, or the learning data storage device may be mounted on the model learning device. In this case, learning data is supplied from the learning data storage device to the model learning device via a cable or a communication network. Also, the model learning device and the learning data storage device may not be communicably connected. In this case, learning data is supplied from the learning data storage device to the model learning device via the portable storage medium in which the learning data is stored.

The learned model according to the present embodiment is a CNN that outputs a magnetic resonance image from which an artifact in the magnetic resonance image has been removed by using a magnetic resonance image whose occurrence position of the artifact is known according to imaging conditions. At this time, the total number of filters in CNN is preset. The learning data includes data of a magnetic resonance image whose occurrence position of an artifact is known according to imaging conditions, data indicating a near region and a distant region to be subjected to filter processing in the CNN, that is, convolution processing, and an artifact in the magnetic resonance image Is the data of the removed magnetic resonance image.

The near area and the distant area are preset based on the occurrence position of the artifact according to the imaging condition. For example, when the artifact is aliasing according to the reduction factor, the near region and the distant region correspond to a region centered at the position where the pixels overlap due to aliasing. Further, when the artifact is N half artifact in the EPI method, the near region and the distant region correspond to a region centered on the position at which the pixel folded back in the N half artifact overlaps. In addition, when the artifact is a chemical shift, the near region and the distant region correspond to a region centered at the position where the pixels shifted by the chemical shift overlap.

The adjacent area and the separated area correspond to a plurality of convolution positions according to the occurrence position of the artifact due to the same physical position in the CNN. For this reason, the model learning device can determine the convolution layer in the CNN so as to include both the vicinity region of the pixel of interest and the remote region including the far pixel superimposed on the pixel of interest.

The learned model generated by the model learning device is, for example, a neural network having a convolutional layer whose convolutional position is changed according to equally spaced decimation rates in decimated acquisition in Cartesian, N half artefact occurrence positions in EPI A neural network with a convolutional layer whose convolutional position is changed according to the neural network, a neural network with a convolutional layer whose convolutional position is changed according to a chemical shift, a decimation ratio in decimated collection and occurrence position of N half artifact in EPI and chemical It is a program that causes the processing circuit 151 to execute a neural network or the like having a convolutional layer whose convolutional position is changed according to the shift. These neural networks are stored in the memory 152 as a program in association with the imaging conditions.

The processing procedure of the image generation processing using the learned model will be described below. FIG. 5 is a flowchart showing an example of the processing procedure in the image generation processing. The artifact related to the explanation in this flowchart is described as being a aliasing artifact according to the reduction factor. The artifact applicable as the present embodiment is not limited to the aliasing artifact, and may be, for example, an N half artifact in the EPI method, an artifact due to a chemical shift, or the like.

(Image generation process)
(Step Sa1)
The imaging conditions are input by the instruction of the operator via the input / output interface 153. Hereinafter, in order to make the description specific, it is assumed that the imaging condition has a reduction factor of 2. The processing circuit 151 uses the selection function 1513 to select a learned model corresponding to the imaging condition using the input imaging condition. The selection of the learned model may be performed after step Sa3 described later.

(Step Sa2)
The imaging control circuit 121 collects magnetic resonance data by executing a pulse sequence in accordance with the input imaging conditions. The imaging control circuit 121 outputs the acquired magnetic resonance data to the processing circuit 151. The processing circuit 151 arranges the magnetic resonance data on the memory 152 in a data space indicating k space.

(Step Sa3)
The processing circuit 151 reconstructs a temporary image by performing Fourier transform or inverse Fourier transform on the magnetic resonance data arranged in the data space, that is, k-space data, by the reconstruction function 1511. In the temporary image, an artifact according to the imaging condition appears. For example, when a pulse sequence is executed under an imaging condition in which the reduction factor is 2, the temporary image has aliasing at a half position of the FOV in the phase encoding direction with respect to the FOV in the case where the reduction factor is 1. It becomes an image.

(Step Sa4)
The processing circuit 151 reads the program corresponding to the selected learned model from the memory 152 by the image generation function 1515. The processing circuit 151 executes a program corresponding to the read learned model. Specifically, the processing circuit 151 applies the read learned model to the temporary image to execute forward propagation processing. The processing circuit 151 generates a magnetic resonance image from which an artifact has been removed as a result of forward propagation processing. When a reduction factor is input as an imaging condition in step Sa1, the magnetic resonance image generated in this step is an image from which aliasing has been removed. The forward propagation function for executing the forward propagation process in this step will be described below with reference to FIGS. 6 and 7.

6 and 7 are diagrams showing FIG. 2 in detail. FIG. 6 is a diagram showing an example of a temporary image TempI, a channel number 1ConvR of a first convolution result for the temporary image TempI, and a channel number 2ConvR of a second convolution result for the first convolution result. In order to simplify the description, attention is focused on the target pixel NP in the temporary image TempI in FIG. The pixel value of the target pixel NP is the sum of the pixel value due to aliasing and the original pixel value not due to aliasing. For this reason, in each of the plurality of convolution layers in the learned model, the convolution position of the target pixel NP includes the vicinity area AR including the target pixel NP and the separation area SR around the position of the pixel folded back to the target pixel NP. Become.

The processing circuit 151 causes the image generation function 1515 to perform a filter operation corresponding to a convolution operation on a plurality of pixel values included in the near area AR and the distant area SR in the temporary image TempI in the first convolutional layer in the learned model. Execute the process Specifically, processing circuit 151 uses a plurality of filter coefficients in each of a plurality of filters used in the first convolutional layer in the selected learned model as weighting coefficients, and is included in neighboring region AR and remote region SR. A product-sum operation, ie, a convolution operation, is performed on the plurality of pixel values to be processed. The processing circuit 151 associates the product-sum value, which is the result of the convolution operation, in the first convolution result with the position NP1 corresponding to the target pixel NP. The processing circuit 151 executes filtering in the first convolutional layer in parallel over the total number of filters in the learned model. By these processes, the processing circuit 151 calculates a first convolution result. For example, when the total number of filters is 64, the channel number 1ConvR of the first convolution result corresponds to 64 maps.

The processing circuit 151 inputs the first convolution result to the second convolution layer in the selected learned model by the image generation function 1515. Specifically, the processing circuit 151 uses the plurality of filter coefficients in each of the plurality of filters used in the second convolutional layer as weighting coefficients to cope with the channel number 1ConvR of the first convolution result, that is, the total number of filters. A convolution operation is performed on the plurality of product-sum values included in the neighborhood area AR and the separation area SR in the plurality of maps. The convolution coefficient of the near area AR and the distant area SR may be learned using another coefficient or may be the same coefficient. The range of convolution (kernel size) need not be a square, and may be, for example, a shape according to the aspect ratio of the image, or, for example, a shape with a long readout direction may be used as the range of convolution. The processing circuit 151 associates the product-sum value, which is the result of the convolution operation by the second convolution layer, in the second convolution result with the position NP2 corresponding to the target pixel NP. The processing circuit 151 executes filtering in the second convolutional layer in parallel over the total number of filters in the learned model. By these processes, the processing circuit 151 calculates a second convolution result. Hereinafter, in the same manner, the processing circuit 151 repeats, in the forward propagation processing FFP, the operation by the filter processing over the total number N of convolutional layers in the learned model. A pooling layer, an activation layer, a contrast normalization layer, a shortcut (ResNet), a connection with previous data (DenseNet), and the like may be provided as appropriate between two adjacent convolutional layers.

FIG. 7 is a diagram showing an example of generation of a magnetic resonance image ReI using the Nth convolution result by the Nth convolution layer. As shown in FIG. 7, the processing circuit 151 generates a magnetic resonance image ReI by applying an all coupling layer to the channel numbers NconvR of the plurality of N-th convolution results by the image generation function 1515. The processing circuit 151 outputs the generated magnetic resonance image ReI to the input / output interface 153. The display at the input / output interface 153 displays the generated magnetic resonance image ReI.

According to the configuration described above, the following effects can be obtained.
According to the magnetic resonance imaging apparatus 100 in the present embodiment, a learned model corresponding to an imaging condition at the time when the input magnetic resonance image is acquired by magnetic resonance imaging is applied to the input magnetic resonance image. Thus, forward propagation processing for improving the image quality can be performed to output a magnetic resonance image. Specifically, according to the present magnetic resonance imaging apparatus 100, magnetic resonance data is collected by thinning collection at equal intervals in k space, and a temporary image is reconstructed by Fourier transform on the magnetic resonance data, and the same physical position is obtained. From a plurality of learned models each having a convolution layer learned using a plurality of convolution positions set according to an occurrence position of an artifact caused by a plurality of learning conditions using an imaging condition regarding a temporary image, A magnetic resonance image can be generated by selecting a trained model to be applied to the temporary image and applying the selected learned model to the temporary image.

Further, according to the present magnetic resonance imaging apparatus 100, a neural network having a convolutional layer whose convolutional position is changed according to a thinning rate in thinning collection as at least one of a plurality of learned models, N in echo planar imaging A known position according to imaging conditions using a neural network having a convolutional layer in which the convolutional position is changed according to the occurrence position of the half artifact and a neural network having a convolutional layer in which the convolutional position is changed according to the chemical shift. It is possible to generate a magnetic resonance image from which artifacts occurring in the

From the above, according to the present magnetic resonance imaging apparatus 100, the convolution position according to the known folding position according to the imaging condition, that is, the near area AR and the separation area SR, is effectively and efficiently necessary information. The convolution layer is designed by learning the used non-linear mapping, and a trained model having the designed convolution layer can be used to generate a magnetic resonance image from which an artifact is removed. Thus, according to the present magnetic resonance imaging apparatus 100, the image quality of the reconstructed magnetic resonance image can be improved.

(Application example)
The difference between this application example and the embodiment is that two temporary images (hereinafter referred to as a two-channel image) divided according to an occurrence position of an artifact such as a folding position are used as input of two channels to the CNN. To generate a magnetic resonance image from which an artifact has been removed. Note that the input to the CNN is not limited to the above two-channel image, and may be a multi-channel image divided in multiples according to the position where the artifact occurs. First, a learned model in the present application example will be described, and then generation of a magnetic resonance image using the learned model in the present application example will be described.

The learned model according to this application example is a CNN that outputs a magnetic resonance image from which an artifact in a temporary image is removed by using a 2-channel image obtained by dividing the temporary image in which the generation position of the artifact is known according to imaging conditions into two. It is. The learning data is data obtained by dividing a magnetic resonance image in which an occurrence position of an artifact is known according to an imaging condition into two, data indicating a near region and a distant region, and data of a magnetic resonance image from which the artifact in the magnetic resonance image is removed. is there. The image input to the CNN is doubled from one channel to two channels as compared to the present embodiment. The model learning device generates a learned model for the present application example by learning CNN using the learning data. The generated learned model is stored as a program in the memory 152 together with the corresponding imaging conditions.

The processing circuit 151 causes the image generation function 1515 to divide the temporary image into two in step Sa3. The processing circuit 151 generates a two-channel image, for example, by dividing the temporary image into two along an axis where no aliasing occurs in the temporary image. In addition, the axis which divides a temporary image into 2 is not limited to the said axis | shaft, It can set arbitrarily. In step Sa4, the processing circuit 151 applies the selected learned model to the two-channel image and performs forward propagation processing to generate a magnetic resonance image from which an artifact has been removed. Hereinafter, in order to make the description specific, an artifact related to the description in the present application example will be described as a aliasing artifact according to the reduction factor. The artifact applicable as the present application is not limited to the aliasing artifact, and may be, for example, an N half artifact in the EPI method, an artifact due to a chemical shift, or the like.

8 and 9 are diagrams showing FIG. 3 in detail. FIG. 8 shows a 2-channel image 2cTempI obtained by dividing the temporary image TempI into two, a channel number 1ConvR2ch of the first convolution result of the 2-channel for the 2-channel image 2cTempI, and a second of the 2 channels for the first convolution result of the 2-channel. Is a diagram showing an example of the number of channels 2ConvRch2 of the convolution result of. In order to simplify the description, attention is focused on the target pixel NP2c in the two-channel image 2cTempI in FIG. The pixel value of the target pixel NP2c is the sum of the pixel value due to aliasing and the original pixel value that is not due to aliasing. For this reason, the convolution position of the target pixel NP2c in the first image 2cI1 of the two-channel image 2cTempI is the second image of the two-channel image 2cTempI and the neighboring area AR including the target pixel NP2 in the first image 2cI1. This is a separation area SR centered on the position of the pixel which is folded back to the target pixel NP2 at 2cI2.

The processing circuit 151 causes the image generation function 1515 to set a plurality of pixel values included in the neighboring area AR in the first image 2cI1 and a separation area SR in the second image 2cI2 in the first convolutional layer in the learned model. A filtering process corresponding to a convolution operation is performed on a plurality of included pixel values. Specifically, processing circuit 151 uses a plurality of filter coefficients in each of a plurality of filters used in the first convolutional layer in the selected learned model as weighting coefficients, and is included in neighboring region AR and remote region SR. A product-sum operation, ie, a convolution operation, is performed on the plurality of pixel values to be processed. The processing circuit 151 associates the product-sum value, which is the result of the convolution operation, in the first convolution result of the two channels with the position NP2c1 corresponding to the target pixel NP2c. The processing circuit 151 executes filtering in parallel over the total number of filters in the learned model. Through these processes, the processing circuit 151 calculates the first convolution result of the two channels. For example, when the total number of filters is 128, the channel number 1ConvR2ch of the first convolution result of 2 channels corresponds to 128 maps.

The processing circuit 151 inputs the first convolution result by the image generation function 1515 to the second convolution layer in the selected learned model. Specifically, the processing circuit 151 uses the plurality of filter coefficients in each of the plurality of filters used in the second convolutional layer as weighting coefficients to correspond to the number of channels of the first convolution result 1ConvR2ch, that is, the total number of filters. A convolution operation is performed on the plurality of product-sum values included in the area ConvR centered on the position NP2c1 in the plurality of maps. The processing circuit 151 associates the product-sum value, which is the result of the convolution operation by the second convolution layer, in the second convolution result of the two channels with the position NP2c2 corresponding to the target pixel NP. The processing circuit 151 executes filtering in the second convolutional layer in parallel over the total number of filters in the learned model. Through these processes, the processing circuit 151 calculates the second convolution result of the two channels. Hereinafter, in the same manner, the processing circuit 151 repeats, in the forward propagation processing FFP, the operation by the filter processing over the total number N of convolutional layers in the learned model. A pooling layer, a local contrast normalization layer or the like may be provided as appropriate between two adjacent convolutional layers.

FIG. 9 shows an example of generation of a two-channel magnetic resonance image 2cReI using the two-channel N-th convolution result by the N-th convolutional layer and a magnetic resonance image ReI combining the two-channel magnetic resonance image 2cReI FIG. As shown in FIG. 9, the processing circuit 151 generates a two-channel magnetic resonance image 2cReI by applying all coupling layers to the number of channels NconvRch of the plurality of N-th convolution results by the image generation function 1515. Do. The processing circuit 151 generates a magnetic resonance image ReI from which an artifact is removed by synthesizing the generated two-channel magnetic resonance image 2cReI.

(First modification)
The difference between the present embodiment and the present modification is that a complex image is used as a temporary image and a complex operation is used as a convolution operation in the convolution layer. That is, in the present modification, the calculation of CNN is executed as a complex operation in the complex space. The learned model according to the present modification is a CNN that executes a complex operation with a complex image whose occurrence position of an artifact is known according to an imaging condition as input, and outputs a complex image from which the artifact in the complex image is removed. The learning data is data of a complex image whose occurrence position of an artifact is known according to imaging conditions, data indicating a near region and a distant region, and data of a complex image from which an artifact in the complex image is removed. The model learning device generates a learned model for the present application example by learning CNN using the learning data. The generated learned model is stored as a program in the memory 152 together with the corresponding imaging conditions.

The processing circuit 151 generates complex magnetic resonance data by performing quadrature detection on the collected magnetic resonance signals by the reconstruction function 1511. The processing circuit 151 generates a complex image by performing Fourier transform or inverse Fourier transform on complex magnetic resonance data. The processing circuit 151 applies the selected learned model to a complex image by the image generation function 1515 to perform forward propagation processing, thereby generating a complex image from which an artifact has been removed. The processing circuit 151 generates a magnetic resonance image using a complex image generated by forward propagation processing.

(Second modification)
The difference between the application and the present modification is that a real part image and an imaginary part image in a complex image are used as the two-channel image described in the application. The learned model according to the present modification receives as input the real part image and the imaginary part image whose occurrence position of the artifact is known according to the imaging condition, and removes the artefact in the real part image and the imaginary part image. It is a CNN that outputs a partial image and an imaginary part image. In the learning data, data of a real part image and an imaginary part image in which the occurrence position of an artifact is known according to imaging conditions, data indicating a near area and a distant area, and an artifact in the real part image and the imaginary part image are respectively removed Data of the real part image and the imaginary part image. The model learning device generates a learned model for the present application example by learning CNN using the learning data. The generated learned model is stored as a program in the memory 152 together with the corresponding imaging conditions.

The processing circuit 151 generates complex magnetic resonance data by performing quadrature detection on the collected magnetic resonance signals by the reconstruction function 1511. The processing circuit 151 generates a real part image by performing Fourier transform or inverse Fourier transform on real part data in complex magnetic resonance data. The processing circuit 151 generates an imaginary part image by performing Fourier transform or inverse Fourier transform on imaginary part data in complex magnetic resonance data. The processing circuit 151 applies the selected learned model to the real part image and the imaginary part image by the image generation function 1515 to execute forward propagation processing, thereby removing the real part image and the imaginary part image from which the artifact is removed. Generate The processing circuit 151 generates a magnetic resonance image using the real part image and the imaginary part image generated by the forward propagation processing.

As a modification of the present embodiment and the like, when the technical idea of the present image processing apparatus 150 is realized by a cloud or the like, a server on the Internet includes, for example, the processing circuit 151 and the memory 152 in FIGS. It becomes. At this time, the reconstruction function 1511, the selection function 1513, the image generation function 1515 and the like are realized by installing an image processing program for executing the function in the processing circuit 151 of the server and expanding them on the memory. For example, the server can execute an image generation process and the like.

According to at least one embodiment described above, the image quality of the magnetic resonance image can be improved.

(First application example)
Hereinafter, application examples of the present embodiment will be described. FIG. 10 is a diagram showing an example of the configuration of the medical signal processing apparatus 200 in the present application example. As shown in FIG. 10, the medical signal processing apparatus 200 includes an input interface 201, a memory 203, a processing circuit 205, and a display 207. The medical signal processing apparatus 200 may be mounted on the magnetic resonance imaging apparatus 100. Since the input interface 201 and the display 207 correspond to the input / output interface 153 in the embodiment, the description will be omitted. The input interface 201 may function as a communication interface for acquiring imaging conditions and medical images from, for example, a medical image diagnostic apparatus such as a magnetic resonance imaging apparatus. The hardware configuration of the memory 203 and the processing circuit 205 is the same as that of the embodiment, and thus the description thereof is omitted. The processing circuit 205 may have an acquisition function of acquiring imaging conditions and a medical image from a modality (not shown). The processing circuit 205 that realizes the acquisition function functions as an acquisition unit.

The memory 203 stores the learned model 231, the imaging condition 2031, and the medical signal 2033. The learned model 231 includes a correction signal corrected to reduce the pattern with respect to the medical signal 2033 having a pattern appearing at a position shifted by a known shift amount along a known direction, and pattern related information on the pattern. A function is provided to output any one of disease information related to the medical signal 2033. The medical signal 2033 is, for example, a magnetic resonance image generated by magnetic resonance imaging on the subject P, and corresponds to the above-described temporary image. Hereinafter, the medical signal 2033 is described as a magnetic resonance image in order to make the description specific. The pattern is, for example, an artifact generated in the magnetic resonance image in accordance with the imaging condition 2031 of the magnetic resonance imaging. The artifacts are, for example, at least one of folding artifacts, N half artifacts, chemical shift artifacts, and motion artifacts.

Hereinafter, the learned model 231 in the present application example will be described as outputting a correction signal. The case where the learned model 231 outputs pattern related information or disease information will be described in the second application example described later. Further, the case where the pattern is a non-artifact and the medical signal 2033 is a non-two-dimensional medical signal (for example, a non-image) will be described in the third application example. The processing relating to the present application example is suitable for denoising and the like with respect to medical signals, but is not limited to denoising and may be used as applications other than denoising as shown in the second and third application examples.

The known direction is a direction regarding the occurrence of an artifact, and is defined by the pulse sequence in the imaging condition 2031. For example, if the artifact is an aliasing artifact caused by parallel imaging, the known direction corresponds to the direction to be folded back in the magnetic resonance image. The direction to be folded back is not limited to the phase encode direction, and may be a direction defined by both the phase encode direction and the frequency encode direction, such as multi-slice caipirinha and two-dimensional caipirinha. Also, if the artifact is an N half artifact caused by execution of the EPI pulse sequence, the known direction corresponds to, for example, the phase encoding direction in the magnetic resonance image. If the artifact is a chemical artifact, the known direction corresponds, for example, to the frequency encoding direction in the magnetic resonance image. In addition, when the artifact is a motion artifact, the known direction corresponds to, for example, the direction of body movement or the pulsatile flow direction of the object P in the magnetic resonance image.

The known shift amount is a cyclic translation amount of the magnetic resonance image based on the occurrence position of the artifact, and is defined by the pulse sequence in the imaging condition 2031. For example, if the artifact is a aliasing artifact, the known shift amount corresponds to the Reduction factor in the pulse sequence. Also, when the artifact is N half artifact, the known shift amount corresponds to, for example, the position of the ghost appearing in the phase encoding direction in the magnetic resonance image. When the artifact is a chemical artifact, the known shift amount depends on, for example, the difference between the resonance frequency of water and fat and the strength of the static magnetic field. Also, when the artifact is a motion artifact, the known shift amount corresponds, for example, to the position of the ghost appearing in the phase encoding direction in the magnetic resonance image. The folding position with respect to the known direction and the known shift amount is defined or estimated as an occurrence place of an artifact by the imaging condition 2031.

The correction signal corresponds to a magnetic resonance image (hereinafter referred to as an artifact reduced image) in which an artifact in the magnetic resonance image is reduced. For example, if the artifact as a pattern is a aliasing artifact, the correction signal corresponds to an artifact reduced image in which aliasing artifacts are reduced. The artifact reduction image is an image corresponding to the magnetic resonance image ReI in FIGS. 7 and 9.

The learned model 231 outputs a correction signal to the input magnetic resonance image using the known direction and the known shift amount. Specifically, as shown in FIG. 10, the learned model 231 includes a circulation shift layer (Circulation shift layer) 2311 and a CNN 2313 as an example of a deep neural network (hereinafter referred to as DNN). Have. The learned model 231 is stored in the memory 203 in association with the imaging condition 2031. That is, the memory 203 stores a plurality of learned models according to the known direction and the known shift amount. The cyclic shift layer 2311 is preset according to the known direction and the known shift amount associated with the occurrence location of the artifact according to the imaging condition 2031. That is, the cyclic shift layer 2311 is a model that is not machine-learned. On the other hand, CNN 2313 is a model generated by machine learning. A learned model 231 in which a cyclic shift layer 2311 not subjected to machine learning and a CNN 2313 generated by machine learning are combined is generated by machine learning.

The cyclic shift layer 2311 generates a shift signal by shifting the magnetic resonance image cyclically by a known shift amount along a known direction. The shift signal corresponds to an image (hereinafter referred to as a shift image) in which the magnetic resonance image is shifted by a known shift amount cyclically along a known direction. The cyclical shifting corresponds to circulating the medical signal, assuming that both ends of the medical signal are connected with respect to a known direction. The processing content executed by the cyclic shift layer 2311 may be realized by deep learning. That is, the cyclic shift layer 2311 may be realized as a DNN that outputs a shift image with a magnetic resonance image having a substantially periodicity as an input. The processing content executed by the cyclic shift layer 2311 will be described later.

The CNN 2313 is a neural network that is functioned to output a correction signal using a magnetic resonance image and a shift image. A neural network having Locally Connect may be used instead of CNN 2313. For example, when the data output from the learned model 231 is an artifact reduction image, the learned model 231 includes a cyclic shift layer 2311 and a CNN 2313. Also, instead of CNN 2313, DNN as a full connect may be appropriately used according to the application of the output data output from the learned model 231. As CNN 2313 or DNN according to the present embodiment and the present application example, ResNet (Residual Network), Dense Net (Dense Convolutional Network), U-Net, etc. can be used. Also, in ResNet, DenseNet, U-Net, etc., a combination of cyclic shift layer 2311 and CNN 2313 may be repeatedly executed as appropriate.

The processing circuit 205 has a determination function 2511 and an output function 2513. The processing circuit 205 uses the determination function 2511 to determine the known direction and the known shift amount based on the imaging condition 2031 for the subject P. The processing circuit 205 determines a learned model based on the determined direction and the determined shift amount. The processing circuit 205 inputs a magnetic resonance image to the determined learned model by the output function 2513, and outputs a correction signal using a known direction and a known shift amount. The processing circuit 205 corresponds to a processing unit, and is configured by an electronic circuit such as the processor described above.

Hereinafter, a procedure for executing a process of generating an artifact reduction image as a correction signal (hereinafter, referred to as an artifact reduction process) using the learned model 231 in the application example will be described with reference to FIGS. 11 to 13. . FIG. 11 is a flowchart illustrating an example of the procedure of the artifact reduction process.

(Artifact reduction processing)
(Step Sb1)
The processing circuit 205 determines the learned model to which the magnetic resonance image is input using the imaging condition 2031 by the determination function 2511. Specifically, based on the imaging condition 2031 used for acquiring the magnetic resonance image, the processing circuit 205 determines a direction in which a pattern such as an artifact appears (corresponding to a known direction, hereinafter referred to as an artifact occurrence direction). The amount of shift by which the pattern shifts in a known direction (hereinafter referred to as a shift amount of shift) is determined. More specifically, the processing circuit 205 collates the imaging parameter associated with the magnetic resonance image with the correspondence table between the direction for the imaging parameter of the pulse sequence and the shift amount (hereinafter referred to as a direction shift amount correspondence table). . The processing circuit 205 determines the artifact occurrence direction and the shift amount by the collation using the direction shift amount correspondence table. The direction shift amount correspondence table is stored in advance in the memory 203, and is read from the memory 203 to the processing circuit 205 by the determination function 2511.

The processing circuit 205 uses the determination function 2511 to collate the artifact occurrence direction and the shift amount with the correspondence table of the learned model for the direction and the shift amount (hereinafter referred to as a model correspondence table). The processing circuit 205 determines a learned model by collation using a model correspondence table. The model correspondence table is stored in advance in the memory 203. The processing circuit 205 reads the determined learned model from the memory 203 to the processing circuit 205. The processing circuit 205 that implements the determination function 2511 corresponds to a determination unit.

(Step Sb2)
The processing circuit 205 causes the output function 2513 to generate a shift image via the cyclic shift layer 2311 in the learned model 231. Specifically, the processing circuit 205 inputs a magnetic resonance image to the cyclic shift layer 2311. The cyclic shift layer 2311 generates a shifted image, which is shifted cyclically by the shift shift amount along the artifact occurrence direction, with respect to the input magnetic resonance image.

A process (hereinafter, referred to as a cyclic shift process) performed on the magnetic resonance image by the cyclic shift layer 2311 will be described using FIGS. 12 and 13. FIG. 12 is a diagram showing an example of cyclic shift processing for the magnetic resonance image MA1 having a reduction factor corresponding to 2 and having aliasing artifacts along the phase encoding direction. In the magnetic resonance image MA1 shown in FIG. 12, the artifact generation direction is the phase encoding direction (y direction), and the shift amount is half (FOVy / 2) of FOVy in the vertical direction. At this time, the cyclic shift processing 2315 generates the shift image SI1 by cyclically translating the magnetic resonance image MA1 by FOVy / 2 along the y direction.

FIG. 13 is a diagram showing an example of cyclic shift processing on a magnetic resonance image MA2 having a reduction factor corresponding to 3 and having aliasing artifacts along the phase encoding direction. In the magnetic resonance image MA2 shown in FIG. 13, the artifact generation direction is the phase encoding direction (y direction), and the shift amount is half the FOVy in the vertical direction (FOVy / 3). At this time, the first cyclic shift processing 2317 circularly translates the magnetic resonance image MA2 by FOVy / 3 along the y direction to generate a first shifted image SI2. The second cyclic shift processing 2319 generates a second shift image SI3 by cyclically translating the magnetic resonance image MA2 by 2 × FOVy / 3 along the y direction. The known shift amount to be cyclically shifted is not limited to those shown in FIGS. 12 and 13, and the step size may be increased according to the type of artifact, the status of the artifact, and the like.

(Step Sb3)
The processing circuit 205 inputs the magnetic resonance image to the CNN 2313 together with the shift image by the output function 2513. For example, when the magnetic resonance image MA1 is as shown in FIG. 12, the processing circuit 205 inputs the generated shift image SI1 to the CNN 2313 together with the magnetic resonance image MA1. When the magnetic resonance image MA2 is as shown in FIG. 13, the processing circuit 205 inputs the first shift image SI2 and the second shift image SI3 to the CNN 2313 together with the magnetic resonance image MA1.

(Step Sb4)
The processing circuit 205 outputs the artifact reduced image ReI as a correction signal from the CNN 2313 to which the shift image and the magnetic resonance image output from the cyclic shift layer 2311 are input by the output function 2513. The processing circuit 205 outputs the artifact reduced image ReI to the memory 203 and the display 207.

(Step Sb5)
The display 207 displays the artifact reduced image ReI. The artifact reduced image ReI is a magnetic resonance image in which aliasing artifacts are reduced as shown in FIGS. 12 and 13. The processing circuit 205 may output the artifact reduced image ReI to an external device such as a medical image storage device via a network (not shown).

According to the configuration described above, the following effects can be obtained.
According to the medical signal processing apparatus 200 in this application example, the correction signal corrected to reduce the pattern with respect to the medical signal 2033 having the pattern appearing at the position shifted by the known shift amount along the known direction is used. A medical signal 2033 can be input to a learned model that has been functioned to output, and a correction signal can be output using a known direction and a known shift amount. Further, the learned model in the medical signal processing apparatus 200 includes a cyclic shift layer 2311 which generates a shift signal in which the medical signal 2033 is shifted cyclically by a known shift amount along a known direction, a medical signal 2033 And a neural network 2313 that is adapted to output a correction signal using the shift signal.

According to the medical signal processing apparatus 200 according to this application example and the present embodiment, the medical signal 2033 is a magnetic resonance image generated by magnetic resonance imaging on the subject P, and the pattern corresponds to the imaging condition 2031 of the magnetic resonance imaging Artifacts generated in the magnetic resonance image, the correction signal is an artifact reduced image with reduced artifacts, and the learned model 231 is a convolutional neural network having a plurality of intermediate layers, and a plurality of intermediate layers. For each layer, the output from any first node in the intermediate layer of the previous stage connected to the input side to each of the plurality of intermediate layers and the second node determined by the imaging condition 2031 in the intermediate layer of the previous stage Can be processed to be input together with the output of.

Further, according to the medical signal processing apparatus 200, the medical signal 2033 is a magnetic resonance image generated by magnetic resonance imaging for the subject P, and the pattern is a magnetic resonance image according to the imaging condition 2031 of the magnetic resonance imaging. The generated artifact, the correction signal is an artifact reduced artifact reduction image ReI, and the neural network is a neural network having local linear combination in each of the plurality of intermediate layers, and the known direction is A known shift amount is a translation amount based on the occurrence position of an artifact, which is a direction regarding the occurrence of an artifact. Further, according to the medical signal processing apparatus 200, the artifact is at least one artifact among the aliasing artifact, the N half artifact, the chemical shift artifact, and the motion artifact.

From these things, according to the medical signal processing apparatus 200, based on the imaging condition 2031, a magnetic resonance image in which an artifact appearing at a position shifted by a known shift amount along a known direction is reduced is generated. Therefore, the image quality of the magnetic resonance image can be improved, and the diagnostic efficiency for the subject P can be improved.

(Second application example)
The difference between the present application example and the first application example is that the learned model 231 outputs pattern related information or disease information. In the learned model 231 in the present application example, DNN may be used instead of CNN 2313. Also, in the learned model 231 in this application example, in ResNet, DenseNet, U-Net, or the like, a 1⁄2 max pool layer may be appropriately incorporated after the CNN 2313 (or DNN). At this time, by providing DNN as a full connect at the final stage in the learned model 231, the learned model 231 outputs pattern related information or disease information.

When the pattern is an artifact, the pattern related information is, for example, at least one physical parameter used to correct a magnetic resonance image, or data indicating the presence or absence (detection result) of the artifact. For example, when magnetic resonance imaging is performed on the subject P using a pulse sequence of the EPI method, the physical parameter corresponds to the estimated value of the physical quantity indicating the delay amount of the generation of the gradient magnetic field. The data indicating the presence or absence of an artifact is, for example, a binary (0 or 1) indicating the presence or absence of each of a plurality of artifacts in a magnetic resonance image.

The disease information is data indicating the recognition result of each of a plurality of diseases in the magnetic resonance image. For example, when the magnetic resonance image has an artifact appearing at a position shifted by a known shift amount along a known direction, the disease information corresponds to an index value indicating the degree of each of a plurality of diseases in the magnetic resonance image. That is, the disease information has an index value indicating the likelihood of each of a plurality of diseases. The index value may be a value (0 or 1) indicating the presence or absence of a disease in the medical signal. The index value and data indicating the presence or absence of an artifact may be output as a percentage by incorporating a sigmoid function or the like into the learned model 231.

Hereinafter, a procedure for executing a process of generating pattern related information or disease information (hereinafter, referred to as information generation process) using the learned model 231 in the present application example will be described with reference to FIG. FIG. 14 is a flowchart illustrating an example of the procedure of the information generation process in the present application example. Among the processing procedures in FIG. 14, the processing of step Sc1 to step Sc3 is the same as the processing of step Sb1 to step Sb3, and thus the description thereof is omitted.

(Information generation process)
(Step Sc4)
The processing circuit 205 outputs pattern related information or disease information from the CNN 2313 to which the shift image and the magnetic resonance image output from the cyclic shift layer 2311 are input by the output function 2513. The processing circuit 205 outputs pattern related information or disease information to the memory 203 and the display 207.

(Step Sc5)
The display 207 displays the pattern related information or the disease information output in the process of step Sc4. The processing circuit 205 may output pattern related information or disease information to an external apparatus such as a medical image storage apparatus via a network (not shown). The disease information may be stored in the memory 203, an external storage device, or the like in association with the magnetic resonance image input to the learned model 231. In addition, when the pattern related information is a physical parameter, the processing circuit 205 may correct the magnetic resonance image using the physical parameter by an image correction function (not shown) in the subsequent processing following this step.

Further, the pattern related information is data indicating the presence or absence of an artifact, and the processing circuit 205 determines that an artifact is present in the magnetic resonance image input to the learned model 231 (hereinafter referred to as artifact presence determination), The processing circuit 205 outputs, to the magnetic resonance imaging apparatus 100, an instruction to execute the magnetic resonance imaging again on the subject P (hereinafter referred to as a reimaging instruction). Specifically, the processing circuit 205 outputs a re-imaging instruction to the imaging control circuit 121 in response to the artifact presence determination. In response to the input of the re-imaging instruction, the imaging control circuit 121 performs magnetic resonance imaging on the subject P again.

According to the configuration described above, the following effects can be obtained.
According to the medical signal processing apparatus 200 in this application example, the medical signal 2033 is a magnetic resonance image generated by magnetic resonance imaging on the subject P, and the pattern is a magnetic resonance image according to the imaging condition 2031 of the magnetic resonance imaging. The pattern related information is a physical parameter used to correct the magnetic resonance image, the known direction is a direction regarding the occurrence of the artifact, and the known shift amount is based on the location of the occurrence of the artifact It is a translational amount. Further, according to the medical signal processing apparatus 200, the medical signal 2033 is a magnetic resonance image generated by magnetic resonance imaging for the subject P, and the pattern is a magnetic resonance image according to the imaging condition 2031 of the magnetic resonance imaging. The generated artifact, the pattern related information is data indicating the presence or absence of the occurrence of the artifact, the known direction is the direction regarding the occurrence of the artifact, and the known shift amount is a translation amount based on the occurrence position of the artifact is there. Further, according to the medical signal processing apparatus 200, the medical signal 2033 is a magnetic resonance image generated by magnetic resonance imaging for the subject P, and the pattern is a magnetic resonance image according to the imaging condition 2031 of the magnetic resonance imaging. The generated artifact, the disease information is information indicating the recognition result of each of a plurality of diseases in the magnetic resonance image, the known direction is the direction regarding the occurrence of the artifact, and the known shift amount is the occurrence of the artifact It is a translation amount based on position.

From these facts, according to the medical signal processing apparatus 200, a magnetic resonance image having an artifact appearing at a position shifted by a known shift amount along a known direction is used to correct the magnetic resonance image. Pattern related information such as data indicating the presence or absence of at least one physical parameter or an artifact, or disease information such as data indicating a recognition result of each of a plurality of diseases in the magnetic resonance image can be obtained. Thus, according to the medical signal processing apparatus 200, the diagnostic efficiency and the like for the subject P can be improved.

(Third application example)
The difference between the first application example and the second application example and this application example resides in using the biological signal of the subject P as the medical signal 2033. The biological signal is, for example, a one-dimensional signal such as an electrocardiogram waveform, a pulse waveform, or a respiration waveform. In this application example, a pattern appearing at a position shifted by a known shift amount along a known direction corresponds to the waveform of a biological signal. Also, the known direction is the time direction for the acquisition of the biological signal. The known shift amount is a period between predetermined time phases in the biological signal, and is, for example, a period of the biological signal.

The medical signal processing apparatus 200 may be mounted on a biological signal measuring apparatus that measures a biological signal. Hereinafter, in order to specifically explain, the biological signal is described as an electrocardiogram. At this time, the medical signal processing apparatus 200 may be mounted on, for example, an electrocardiograph. In addition, the medical signal processing device 200 may have a speaker (not shown). In addition, the learned model 231 in this application example will be described as outputting disease information.

FIG. 15 is a diagram showing an example of an electrocardiogram waveform ECGW as a biological signal in the present application example.
. As shown in FIG. 15, a portion of the electrocardiogram waveform ECGW (hereinafter referred to as the first waveform) included in R11 and a portion of the electrocardiogram waveform ECGW (hereinafter referred to as the second waveform) included in R22 are learned. Is input to the pre-processed model.

The processing circuit 205 determines the amount of shift by the determination function 2511. Specifically, the processing circuit 205 determines the amount of shift based on the interval between two adjacent R waves in the electrocardiogram waveform. In the present application, since the known direction is a time direction, it is not necessary to determine the known direction.

The cyclic shift layer 2311 cyclically shifts, for example, the first waveform and the second waveform according to the shift amount along the time direction. At this time, the shift signal is a waveform shown in the order of the second waveform and the first waveform along the time direction. The processing circuit 205 outputs disease information from the CNN 2313 to which the shift signal and the biological signal are input by the output function 2513. The k-th (k is a natural number) output y _{k in} each of the plurality of intermediate layers in CNN 2313 is the value of the i-th (i is a natural number) in the intermediate layer or input layer of the previous stage x _i Is given by, for example, the following equation.

For example, when the electrocardiogram waveform is as shown in FIG. 15, the processing circuit 205 outputs, as the disease information, an index value indicating the disease likeness of the extrasystole. The processing circuit 205 causes the display 207 to display a warning when the index value is equal to or greater than a predetermined value. At this time, the processing circuit 205 may output a warning sound from the speaker.

According to the configuration described above, the following effects can be obtained.
According to the medical signal processing apparatus 200 in this application example, the medical signal 2033 is a biological signal of the subject P, the pattern is a waveform of the biological signal, and the disease information is a recognition result of each of a plurality of diseases in the biological signal. The known direction is a time direction related to the acquisition of a biological signal, and the known shift amount is a period between predetermined time phases in the biological signal. Thereby, according to the medical signal processing apparatus 200, the recognition result of each of the plurality of diseases in the biological signal is detected for the biological signal having the pattern appearing at the position shifted by the known shift amount along the known direction. Disease information such as data shown can be obtained. Thus, according to the medical signal processing apparatus 200, by detecting an abnormality of the biological signal from the subject P, the abnormality can be notified.

Fourth application example
The difference between this application example and the first to third application examples is that a medical signal having a pattern appearing at a position shifted by a known shift amount along a known direction is based on a known direction and a known shift amount. Among the correction signal, the pattern related information, and the disease information, a plurality of partial signals are generated by dividing and generating the plurality of partial signals, and the plurality of partial signals are input to the learned model having DNN without the cyclic shift layer. One is to output one.

FIG. 16 is a diagram showing an example of a medical signal processing apparatus 300 in the present application example. In FIG. 16, the same reference numerals are assigned to components having the same functions and the like as the components of the medical signal processing apparatus 200 shown in FIG. The medical signal processing apparatus 300 includes an input interface 201, a memory 303, a partial signal generation circuit (partial signal generation unit) 304, a processing circuit 305, and a display 207. The partial signal generation circuit 304, also referred to as aliasing preprocessor AP, executes preprocessing related to aliasing (hereinafter referred to as aliasing preprocessing).

The partial signal generation circuit 304 divides the medical signal as the aliasing preprocessing based on the known direction determined by the determination function 3051 and the known shift amount as the aliasing preprocessing. Specifically, the partial signal generation circuit 304 is divided by dividing the medical signal by a division width (hereinafter referred to as a window) according to a known shift amount along a known direction. And generating a plurality of partial signals corresponding to the medical signal. The partial signal generation circuit 304 outputs a plurality of partial signals to the processing circuit 305. Specific aliasing preprocessing will be described later. The aliasing preprocessing performed by the partial signal generation circuit 304 may be performed in the processing circuit 305 as an aliasing preprocessing function. The partial signal generation circuit 304 is configured by an electronic circuit such as the processor described above.

Note that the windows may overlap for the medical signal. At this time, each of the plurality of partial signals has a region overlapping with another partial signal (hereinafter referred to as an overlapping region). Inputting a plurality of partial signals having overlapping regions to the learned model (DNN 3331) contributes to stabilization of outputting any one of the correction signal, the pattern related information, and the disease information. In addition, as described in the first application example, the partial signal generation circuit 304 may generate a plurality of partial signals by regarding that both ends of the medical signal are cyclically connected with respect to a known direction. .

The memory 303 stores the learned model 331, the imaging condition 2031, and the medical signal 2033. More specifically, the memory 303 stores a plurality of learned models associated with the total number of windows according to the imaging conditions. The learned model 331 has a DNN 331. The DNN 331 may be realized by ResNet, DenseNet, U-Net or the like. The processing relating to this application example is suitable for detection, recognition, estimation of physical parameters, and the like for medical signals, but is not limited to these and may be used for the purpose of denoising as shown in the first application example. At this time, in the learned model 331, various layers are incorporated in the subsequent stage of the DNN 331. Thereby, the learned model 331 can output the correction signal.

The processing circuit 305 determines the DNN 3331 to which the partial signal is input based on the imaging condition 2031 by the determination function 3051. Specifically, the processing circuit 305 determines the DNN 3331 corresponding to the total number of windows according to the total number of windows according to the known shift amount based on the imaging condition.

The processing circuit 305 inputs the plurality of partial signals to the plurality of different channels in the DNN 3331 determined by the determination function 3051 by the output function 3053. The processing circuit 305 inputs a plurality of partial signals to the determined learned model, and outputs any one of the correction signal, the pattern related information, and the disease information.

Hereinafter, for the sake of concrete description, the medical signal is a magnetic resonance image, the pattern is an artifact, and the partial signal is an image of the magnetic resonance image (hereinafter referred to as a partial image). The output from the learned model 331 is pattern related information or disease information. A procedure of executing information generation processing for generating pattern related information or disease information using the learned model 331 in the present application example will be described using FIGS. 17 and 18. FIG. 17 is a flowchart showing an example of the procedure of the information generation process in this application example. In addition, since the process of step Sd5 is the same as the process of step Sc5 shown in FIG. 14, description is abbreviate | omitted.

(Information generation process)
(Step Sd1)
The processing circuit 305 uses the determination function 3051 to determine the artifact occurrence direction and the shift amount using the direction shift amount correspondence table based on the imaging condition 2031. The processing circuit 305 determines the DNN 3331 corresponding to the total number of windows according to the total number of windows according to the shift amount.

(Step Sd2)
The partial signal generation circuit 304 generates a plurality of partial images by dividing the magnetic resonance image based on the known direction and the known shift amount. Specifically, the partial signal generation circuit 304 divides the magnetic resonance image by a window corresponding to the amount of shift along the artifact occurrence direction. The aliasing preprocessing for dividing the magnetic resonance image will be described with reference to FIG.

FIG. 18 is a diagram showing an example of aliasing preprocessing in this application example. As shown in FIG. 18, the magnetic resonance image MAA input to the partial signal generation circuit 304 has a reduction factor of 2 and has aliasing artifacts along the phase encoding direction. At this time, the total number of windows used for dividing the magnetic resonance image MAA is two, that is, the first window W1 and the second window W2. The partial signal generation circuit 304 divides the magnetic resonance image MAA at a position (hereinafter referred to as a division position) DP that is half the FOVy (FOVy / 2) in the phase encoding direction in the magnetic resonance image MAA. The partial signal generation circuit 304 generates the first partial image PI1 and the second partial image PI2 respectively corresponding to the first window W1 and the second window W2 by dividing the magnetic resonance image MAA at the division position DP. . The partial signal generation circuit 304 outputs the first partial image PI1 and the second partial image PI2 to the processing circuit 305.

The first window W1 and the second window W2 may be set across the division position DP with respect to the magnetic resonance image MAA. For example, when FOVy in the magnetic resonance image MAA is divided into 10 equal parts, an area of 1/10 to 8/10 is set as the first window W1 along the phase encode direction y, and an area of 6/10 to 10/10 is obtained. It may be set as the second window W2.

(Step Sd3)
The processing circuit 305 inputs a plurality of partial images to the DNN 3331 determined by the output function 3053. For example, the processing circuit 305 inputs a plurality of partial images to a plurality of different channels in the DNN 3331, respectively.

(Step Sd4)
The processing circuit 305 outputs pattern related information or disease information from the DNN 3331 to which a plurality of partial images are input by the output function 3053. The processing circuit 305 outputs pattern related information or disease information to the memory 203 and the display 207.

According to the configuration described above, the following effects can be obtained.
According to the medical signal processing apparatus 200 in this application example, a medical signal having a pattern appearing at a position shifted by a known shift amount along a known direction is divided based on the known direction and the known shift amount. A plurality of partial signals are generated, and one of a correction signal corrected to reduce a pattern to a medical signal, pattern related information on a pattern, and disease information on a medical signal is output. By inputting a plurality of partial signals to the function-added learned model, any one of the correction signal, the pattern related information, and the disease information can be output.

(Fifth application example)
The difference between this application example and the first to fourth application examples is that a plurality of partial signals generated by aliasing preprocessing are input to DNN as a learned model, and a plurality of partial correction signals output from DNN are known. A combined signal is generated by combining based on the direction of and the known shift amount, and is output.

FIG. 19 is a diagram showing an example of a medical signal processing apparatus 400 in the present application example. In FIG. 19, the same symbols are assigned to components having the same functions and the like as the components in the medical signal processing apparatus 300 shown in FIG. The medical signal processing apparatus 400 includes an input interface 201, a memory 303, a partial signal generation circuit 304, a combined signal generation circuit 306, a processing circuit 305, and a display 207. The combined signal generation circuit 306, also referred to as Aliasing Postprocessor APost, performs post-processing for aliasing (hereinafter referred to as aliasing post-processing).

According to this application example, when the image subjected to the aliasing pre-processing is applied to the DNN 3333, it is possible to perform the aliasing post-processing and output the image of the original spatial resolution. The aliasing pre-processor AP in the fourth application example generates a plurality of partial images by dividing the magnetic resonance image based on the known direction and the known shift amount. On the other hand, the post-folding post-processor APost in this application example combines the divided magnetic resonance images output from the DNN 3333 based on the known direction and the known shift amount. For example, if the aliasing pre-processor AP divides the image evenly into three in the x-axis direction, the aliasing post-processor APost combines the three output images. When using the post-folding post-processor APost, an image before combining, that is, a divided image is used for learning of the DNN 3333 as learning data.

For example, when reconstruction is performed on a small FOV, only an image in which aliasing occurs may be obtained. In an application where aliasing is removed by post-processing on such an image, for example, it may be regarded as an image having a double size in the aliasing direction, and may be configured to use only aliasing post-processing without aliasing pre-processing. .

The memory 303 stores the learned model 331, the imaging condition 2031, and the medical signal 2033. More specifically, the memory 303 stores a plurality of learned models associated with the total number of windows according to the imaging conditions. The learned model 331 has a DNN 3333. The DNN 3333 may be realized by ResNet, DenseNet, U-Net or the like. The DNN 3333 outputs a plurality of partial correction signals respectively corresponding to the plurality of partial signals input to itself. The plurality of partial correction signals correspond to, for example, the above-described correction signal in which the partial signal is subjected to denoising. The resolution of the medical signal input to the DNN 3333 and the resolution of the signal output from the DNN 3333 correspond to the resolution obtained by dividing the resolution of 2033 of the medical signal by the total number of windows. Prior to mounting on the medical signal processing apparatus 400, the DNN 3333 in this application example is learned using, as learning data, a plurality of partial signals according to the total number of windows and a partial correction signal as correct data.

The processing circuit 305 inputs a plurality of partial signals to the learned model 331 by the output function 3055, and outputs a plurality of partial correction signals respectively corresponding to the plurality of partial signals as a correction signal. Specifically, the processing circuit 305 inputs a plurality of partial signals to a plurality of different channels in the DNN 3331 determined by the determination function 3051. The processing circuit 305 inputs a plurality of partial signals to the determined learned model, and outputs a plurality of partial correction signals to the combined signal generation circuit 306.

The combined signal generation circuit 306 combines the plurality of partial correction signals output from the DNN 3333 based on the known direction determined by the determination function 3051 and the known shift amount as aliasing post-processing. Generate The partial signal generation circuit 306 outputs the combined signal to the memory 303 or the display 207. The combined signal generation circuit 306 is configured by an electronic circuit such as the processor described above. A specific post-aliasing process will be described later. The aliasing post-processing performed by the combined signal generation circuit 306 may be performed in the processing circuit 305 as an aliasing post-processing function. At this time, a program for executing the aliasing post-processing function is stored in the memory 303, an ASIC in the processing circuit 305, or the like.

Hereinafter, for the sake of concrete description, the medical signal is a magnetic resonance image, the pattern is an artifact, the partial signal is a partial image, and a plurality of partial correction signals output from the learned model 331 are The combined signal is a plurality of corrected images (hereinafter referred to as a partially corrected image) respectively corresponding to the partial images of (1), and the combined signal is an image obtained by combining a plurality of partially corrected images (hereinafter referred to as a combined image). The combined image corresponds to the above-described artifact reduced image.

In addition, a procedure for executing combined image generation processing for generating a combined image corresponding to a correction signal in this application example will be described with reference to FIGS. 20 and 21. FIG. 20 is a flowchart illustrating an example of a procedure of combined image generation processing in the present application example. The processes of steps Se1 to Se3 are the same as the processes of steps Sd1 to Sd3 shown in FIG. 17, and thus the description thereof is omitted. In addition, the process of step Se6 is the same as the process of step Sb5 in FIG. Further, as shown in FIG. 21, it is assumed that the pre-folding processor AP equally divides into two in the y-axis direction.

(Combined image generation process)
(Step Se4)
The processing circuit 305 causes the output function 3055 to output a plurality of partially corrected images from the DNN 3333 to which a plurality of partial images have been input. The plurality of partial correction images correspond to partial images with reduced artifacts (hereinafter referred to as artifact reduced partial images). That is, the processing circuit 305 outputs a plurality of artifact reduced partial images respectively corresponding to the plurality of partial images from the DNN 3333. The processing circuit 305 inputs the plurality of partial images to the DNN 3333, and outputs the plurality of artifact reduced partial images output from the DNN 3333 to the combined signal generation circuit 306.

(Step Se5)
The combined signal generation circuit 306 combines the plurality of partially corrected images based on the known direction (artifact occurrence direction) determined by the determination function 3051 and the known shift amount. By the processing, the combined signal generation circuit 306 generates a combined image. That is, the combined signal generation circuit 306 generates an artifact reduced image by combining a plurality of artifact reduced partial images based on the artifact occurrence direction and the shift amount. The combined signal generation circuit 306 outputs the artifact reduced image to the memory 303 and the display 207.

FIG. 21 is a diagram showing an example of post-aliasing processing in the present application example. In FIG. 21, the description of aliasing preprocessing and the like is the same as that of FIG. The processing circuit 305 causes the output function 3055 to input the first partial image PI1 and the second partial image PI2 to the DNN 3333 in the learned model 305, thereby reducing the first artifact reduced partial image RePI1 corresponding to the first partial image PI1. And a second artifact reduced partial image RePI2 corresponding to the second partial image PI2. The combined signal generation circuit 306 specifies the combined position of the first artifact reduction partial image RePI1 and the second artifact reduction partial image RePI2 using the artifact occurrence direction and the shift amount. The bonding position corresponds to, for example, the division position DP. Next, the combined signal generation circuit 306 generates an artifact reduced image by combining the first artifact reduced partial image RePI1 and the second artifact reduced partial image RePI2 at the combined position. A range indicated by AS in FIG. 21 indicates a process performed at half the resolution of the magnetic resonance image (hereinafter referred to as full resolution) before the aliasing preprocessing is performed. In general, when the magnetic resonance image is divided into N by aliasing preprocessing, the processing in the range indicated by AS in FIG. 21 is performed at a resolution of 1 / N of full resolution.

According to the configuration described above, the following effects can be obtained.
According to the medical signal processing apparatus 400 in this application example, a medical signal having a pattern appearing at a position shifted by a known shift amount along a known direction is divided based on the known direction and the known shift amount. A trained model operable to generate a plurality of partial signals, respectively correspond to the plurality of partial signals, and to output a plurality of partial correction signals corrected to reduce a pattern to a medical signal. In contrast, by inputting a plurality of partial signals, a plurality of partial correction signals are output as a correction signal, and by combining a plurality of partial correction signals based on a known direction and a known shift amount, A combined signal can be generated.

As a modified example of the first to third application examples and the like, when the technical idea of the medical signal processing apparatus 200 is realized by a cloud or the like, the server on the Internet may be, for example, the processing circuit 205 and the memory 203 in FIG. It will be possessed. Further, as a modification of the fourth application example and the like, when the technical idea of the medical signal processing apparatus 300 is realized by a cloud or the like, the server on the Internet is, for example, the partial signal generation circuit 304 in FIG. It becomes what has 305 and the memory 303. In addition, as a modification of the fifth application example and the like, when the technical idea of the medical signal processing apparatus 400 is realized by a cloud or the like, the server on the Internet is, for example, the partial signal generation circuit 304 in FIG. A combined signal generation circuit 306 and a memory 303 are provided. In these cases, the determination function 2511, the determination function 3051, the output function 2513, the output function 3053, the output function 3055, etc. install a program for executing the function in the processing circuit of the server and expand them on the memory. To be realized.

According to at least one embodiment described above, the output error due to the learned model can be reduced. For example, according to the medical signal processing apparatus 200, even for a medical signal having a pattern (artifact) that appears at a position shifted by a known shift amount along a known direction, the medical signal may not be a learned model 231 Can be used to output any one of a correction signal with improved noise reduction due to an artifact and pattern related information and disease information with improved recognition rate. it can. Further, according to the medical signal processing apparatus 300, the medical signal having a pattern (artifact) appearing at a position shifted by a known shift amount along a known direction is subjected to aliasing preprocessing and then processed. Input the medical signal to the main learning model 331 to output any one of a correction signal in which noise reduction due to an artifact is improved and pattern related information and disease information in which the recognition rate is improved. Can improve the diagnostic efficiency. Further, according to the medical signal processing apparatus 400, the medical signal having a pattern (artifact) appearing at a position shifted by a known shift amount along a known direction is subjected to aliasing preprocessing and then processed. By inputting the medical signal to the main learning model 331, it is possible to generate a combined signal obtained by combining a plurality of correction signals, which outputs a plurality of correction signals in which noise reduction due to an artifact is improved. Efficiency can be improved.

The image resolution is converted using resolution conversion processing such as up-sampling, down-sampling, pooling, etc., in pre-processing, post-processing, DNN, or a combination thereof, thereby outputting an image having a resolution different from that of the input. You may.

In addition, in the apparatus which implements this invention, when making an imaging method be selected, you may be shown to a user whether the learned model exists. Specifically, for example, when there is a learned model, the function of the present invention can be selected, and when it does not exist, the function of the present invention can not be selected. Alternatively, for example, when the function of the present invention is selected, only the condition in which the learned model exists may be selectable.

While certain embodiments of the present invention have been described, these embodiments have been presented by way of example only, and are not intended to limit the scope of the invention. These embodiments can be implemented in other various forms, and various omissions, replacements, and modifications can be made without departing from the scope of the invention. These embodiments and modifications thereof are included in the invention described in the claims and the equivalents thereof as well as included in the scope and the gist of the invention.

Claims

A correction signal corrected to reduce the pattern with respect to a medical signal having a pattern appearing at a position shifted by a known shift amount along a known direction, pattern related information on the pattern, and the medical signal The medical signal is input to a learned model that is functioned to output any one of disease information, and the correction signal and the pattern related information are input using the direction and the shift amount. A medical signal processing apparatus, comprising: a processing unit that outputs any one of: and the disease information.
The learned model is
A cyclic shift layer generating a shift signal in which the medical signal is shifted by the shift amount cyclically along the direction;
And a neural network configured to output any one of the correction signal, the pattern related information, and the disease information using the medical signal and the shift signal.
The medical signal processing device according to claim 1.
The medical signal is a magnetic resonance image generated by magnetic resonance imaging of a subject,
The pattern is an artifact generated in the magnetic resonance image according to an imaging condition of the magnetic resonance imaging,
The correction signal is an artifact reduced image in which the artifact is reduced,
The trained model is a convolutional neural network having a plurality of intermediate layers,
The processing unit is configured to, for each of the plurality of intermediate layers, output from any one first node in the intermediate layer of the previous stage connected to the input side to each of the plurality of intermediate layers, and the intermediate layer of the previous stage Processing so as to be input together with the output from the second node determined by the imaging condition.
The medical signal processing device according to claim 1.
The medical signal is a magnetic resonance image generated by magnetic resonance imaging of a subject,
The pattern is an artifact generated in the magnetic resonance image according to an imaging condition of the magnetic resonance imaging,
The correction signal is an artifact reduced image in which the artifact is reduced,
The neural network is a neural network having a local linear combination in each of a plurality of intermediate layers,
The direction is with respect to the occurrence of the artifact,
The shift amount is a translation amount based on the occurrence position of the artifact.
The medical signal processing device according to claim 2.
The medical signal is a magnetic resonance image generated by magnetic resonance imaging of a subject,
The pattern is an artifact generated in the magnetic resonance image according to an imaging condition of the magnetic resonance imaging,
The pattern related information is a physical parameter used to correct the magnetic resonance image, and
The direction is with respect to the occurrence of the artifact,
The shift amount is a translation amount based on the occurrence position of the artifact.
The medical signal processing device according to claim 2.
The medical signal is a magnetic resonance image generated by magnetic resonance imaging of a subject,
The pattern is an artifact generated in the magnetic resonance image according to an imaging condition of the magnetic resonance imaging,
The pattern related information is data indicating the presence or absence of the occurrence of the artifact,
The direction is with respect to the occurrence of the artifact,
The shift amount is a translation amount based on the occurrence position of the artifact.
The medical signal processing device according to claim 2.
The medical signal is a magnetic resonance image generated by magnetic resonance imaging of a subject,
The pattern is an artifact generated in the magnetic resonance image according to an imaging condition of the magnetic resonance imaging,
The disease information is data indicating a recognition result of each of a plurality of diseases in the magnetic resonance image,
The direction is with respect to the occurrence of the artifact,
The shift amount is a translation amount based on the occurrence position of the artifact.
The medical signal processing device according to claim 2.
The artifact is at least one of a folding artifact, an N half artifact, a chemical shift artifact, and a motion artifact.
The medical signal processing device according to any one of claims 3 to 7.
The medical signal is a biological signal of a subject, and
The pattern is a waveform of the biological signal, and
The disease information is data indicating a recognition result of each of a plurality of diseases in the biological signal,
The direction is a time direction relating to the acquisition of the biological signal,
The shift amount is a period between predetermined time phases in the biological signal.
The medical signal processing device according to claim 2.
A partial signal generation unit that generates a plurality of partial signals obtained by dividing a medical signal having a pattern that appears at a position shifted by a known shift amount along a known direction based on the direction and the shift amount;
A function is provided to output one of a correction signal corrected to reduce the pattern with respect to the medical signal, pattern related information on the pattern, and disease information on the medical signal. A processing unit that outputs any one of the correction signal, the pattern related information, and the disease information by inputting the plurality of partial signals to a learned model;
Medical signal processing apparatus equipped with
The processing unit outputs the plurality of partial correction signals respectively corresponding to the plurality of partial signals as the correction signal by inputting the plurality of partial signals to the learned model.
And a combined signal generation unit configured to generate a combined signal by combining the plurality of partial correction signals based on the direction and the shift amount.
The medical signal processing device according to claim 10.