WO2015170394A1 - Image capturing device, image processing device, and image processing method - Google Patents

Abstract

The present invention suppresses artifacts which occur in a high-speed image capturing method by thinning out data. A low-dimensional image is created by converting measurement data and an image into a low-dimensional space, and a new type of constraint in image estimation processing is created using the low-dimensional image. By providing the new type of constraint that is different in nature from the constraint of sparseness in compressed sensing and a constraint based on coil sensitivity in parallel imaging, the range of the solution of an estimated image is narrowed to thereby bring the estimated image close to a correct solution.

Description

Imaging apparatus, image processing apparatus, and image processing method

The present invention relates to an imaging apparatus such as a magnetic resonance imaging apparatus or a CT apparatus, and more particularly to a technique for reconstructing a high-quality image using thinned data acquired by these imaging apparatuses.

In an imaging apparatus such as a magnetic resonance imaging (hereinafter referred to as MRI) apparatus or a CT apparatus, in order to shorten the imaging time, data that is less than the number of data originally necessary for image reconstruction or the like (deletion data) A method has been developed in which thinned data is collected and images are reconstructed by making up for deficiencies from other information.

For example, in an MRI apparatus, there is a technique (parallel imaging) in which a true image is reconstructed by using a sensitivity distribution for thinned data obtained by a plurality of small coils constituting a receiving coil (Patent Document 1). ). Patent Document 1 describes a technique for generating a coil sensitivity distribution used in parallel imaging from data obtained by combining thinned data.

Recently, even when there is not enough data, a technique (compressed sensing) that reconstructs an image using that property when the data has the property that it can be converted into sparse data by some kind of conversion has been proposed. . Examples of applying this technique to an MRI apparatus are disclosed in, for example, Patent Document 2 and Patent Document 3. Patent Document 2 proposes a method for increasing the quality of an estimated image by repeatedly calculating a correction amount corresponding to an artifact for each of a plurality of data sets having different thinning methods and correcting the original frame data. ing. Patent Document 3 discloses a technique for improving the quality of an estimated image by converting a difference from an image created by a technique (“Sliding window”) that captures a part of k-space data while shifting it for each frame. Has been proposed. Moreover, although it is not compression sensing, the example which uses an average image for optimization in the image reconstruction which compensates a defect | deletion from other information is disclosed by patent document 4. FIG.

Non-Patent Documents 1 to 5 propose a technique that further evolves parallel imaging, a peripheral technique thereof, and a technique related to compressed sensing. For example, Non-Patent Document 1 and Non-Patent Document 2 disclose a coil sensitivity distribution. An operator (referred to as SPIRIT operator or GRAPPA operator) that supplements the thinned data based on the above has been proposed. Non-Patent Document 3 describes a technique for determining an optimum weight for adding data of each coil. Non-Patent Document 4 proposes a technique that uses a plurality of sensitivity maps, and Non-Patent Document 5 proposes a technique that changes a threshold value used in repeated reconstruction of compressed sensing during repetition.

As for the CT apparatus, reconfiguration by repetitive calculation has become realistic due to the speeding up of the computer compared to the back projection method having a short calculation time which has been the mainstream until now. In order to improve temporal resolution in imaging moving objects such as the heart and to reduce the exposure dose, it is desired to perform reconstruction from the thinned data, and application of the above-described compressed sensing is also attempted.

JP 2003-325473 A US Patent Publication No. 2013/310678 US Patent Publication No. 2009/262996 US Patent Publication No. 2005/189942

Conventional image estimation methods using compressed sensing may have limitations on the target frame data or limitations on the improvement of the estimated image (matching with the true image) depending on each method. . For example, consideration is given to artifacts (Patent Document 2), and conversion to sparse data is devised (Patent Document 3). Is performed within the framework of estimating images whose data is sparse in the transformed space, and there is a wide range of images that satisfy the constraints of consistency, and the estimated image is a true image There is a problem that it may not be close enough. Further, Patent Document 2 discloses a technique that can effectively suppress artifacts, but the application conditions are severe such that the artifacts can be obtained analytically and a plurality of data sets can be fused.

It is an object of the present invention to provide an imaging apparatus and an image processing technique that can be applied not only to an MRI apparatus but also to a CT apparatus or any other imaging apparatus and make an estimated image closer to a true image. Moreover, this invention makes it a subject to provide the technique which can suppress an artifact effectively.

In order to solve the above-mentioned problem, the first aspect of the present invention converts measurement data and images, which are thinned data, into a low-dimensional space when an image is reconstructed from the thinned data by estimation processing. Then, a new type of constraint that matches in the space is created, the range of the solution of the estimated image in the estimation process is narrowed down, and the content is to bring the estimated image closer to the correct solution. This new type of constraint is different in nature from the constraint of sparseness in compressed sensing and the constraint based on the coil sensitivity of parallel imaging disclosed in the above prior art document.

The second aspect of the present invention includes artifact extraction and artifact suppression using the extraction result. The artifact extraction includes target measurement data image, target measurement data, and other measurement data. And a composite image.

The first aspect and the second aspect of the present invention can be implemented independently or in combination.

That is, the imaging apparatus of the present invention includes an imaging unit that collects a data set necessary for image reconstruction, and an image processing unit that reconstructs an image using the data set, and the data set collected by the imaging unit is A data set including missing data, wherein the image processing unit uses the data set to estimate an image under a constraint condition of coincidence between the data set and the image, and the data set; Update the estimated image estimated by the image estimation unit under the constraint condition of the low-dimensional image and the low-dimensional data creation unit that lowers the image, and the low-dimensional image set created by the low-dimensional data creation unit An image updating unit (first mode).

The imaging apparatus of the present invention includes an imaging unit that collects a data set necessary for image reconstruction, and an image processing unit that reconstructs an image using the data set, and the data set collected by the imaging unit is A data set including missing data, wherein the image processing unit uses a plurality of data sets to estimate an image under a constraint condition of coincidence between the data set and the image, and the missing data An artifact extraction unit that extracts an artifact included in an image reconstructed from a data set including: a data correction unit that corrects the data set using the artifact information extracted by the artifact extraction unit, and the artifact. The extraction unit includes one data set and a combined data set of the one data set and other data sets. With, which calculates the artifacts included in the one data set, the image estimation unit performs image estimation processing for the data set that is corrected by the data correcting section (second embodiment).

The image processing apparatus of the present invention has the configuration of the image processing unit of the imaging apparatus described above, and may be part of the imaging apparatus or independent of the imaging apparatus.

The imaging apparatus and image processing apparatus of the present invention are applied to various imaging apparatuses such as an MRI apparatus, a CT apparatus, an ultrasonic imaging apparatus, a PET apparatus, and the like.

According to the present invention, in high-speed imaging that measures thinned data, by adding a new type of constraint to compressed sensing, artifacts can be suppressed and an image close to a true value can be acquired. Further, if the artifact is the same level as before, it is possible to increase the data thinning rate and speed up imaging.

1 is a block diagram showing an overall outline of an MRI apparatus to which the present invention is applied. Block diagram showing the image processing unit of the first embodiment The flowchart which shows operation | movement of the image process part of 1st embodiment. The figure explaining the data which the image processing part of 1st embodiment handles, and its processing The figure which shows the effect of 1st embodiment The figure which shows the example of a change of the data handled by 1st embodiment The figure which shows another example of a change of the data which the image process part of 1st embodiment handles The figure which shows the example of a change of the averaging process of the image process part of 1st embodiment. The figure which shows another example of a change of the averaging process of the image process part of 1st embodiment. The figure which shows another example of a change of the averaging process of the image process part of 1st embodiment. The figure which shows another example of a change of the averaging process of the image process part of 1st embodiment. The figure which shows another example of a change of the averaging process of the image process part of 1st embodiment. The figure which shows another example of a change of the averaging process of the image process part of 1st embodiment. The figure which shows an example of the added UI function of 1st embodiment. The figure which shows another example of the added UI function of 1st embodiment. The block diagram which shows the image processing part of 2nd embodiment. The flowchart which shows operation | movement of the image process part of 2nd embodiment. The figure explaining the data which the image processing part of 2nd embodiment handles, and its processing The flowchart which shows the example of a change of operation | movement of the image process part of 2nd embodiment. The flowchart which shows another example of change of operation of the image processing part of a second embodiment. The flowchart which shows another example of a change of the operation | movement of the image processing part of 2nd embodiment. The figure which shows the effect of 2nd embodiment The block diagram which shows the whole outline | summary of CT apparatus with which this invention is applied. The flowchart which shows operation | movement of the image processing part of 3rd embodiment. The figure explaining the data which the image processing part of 3rd embodiment handles, and its processing Flowchart showing the operation of the image processing unit of the fourth embodiment The figure explaining the data which the image processing part of 4th embodiment handles, and its processing

The imaging apparatus according to the present embodiment includes an imaging unit that collects a data set necessary for image reconstruction, and an image processing unit that reconstructs an image using the data set. The image processing unit processes the data set including the missing data collected by the imaging unit, and uses the data set to estimate an image under a constraint condition of coincidence between the data set and the image. A low-dimensional data creation unit for reducing the order of the data set and the image, and the image estimation unit under a constraint condition for matching between the low-dimensional data set created by the low-dimensional data creation unit and the low-dimensional image. And an image update unit that updates the estimated image.

The imaging apparatus is, for example, an MRI apparatus. In this case, the imaging section receives a magnetic field generation unit that generates a static magnetic field and a gradient magnetic field, a transmission unit that generates a high-frequency magnetic field, and a nuclear magnetic resonance signal generated by the high-frequency magnetic field. The data set is k-space data.

The imaging apparatus is, for example, a CT apparatus. In this case, the imaging unit includes an X-ray source, an X-ray detector that detects X-rays emitted from the X-ray source, the X-ray source, and the X-ray detection. The data set is made up of projection data collected at different rotation angles of the rotation unit.

Hereinafter, an embodiment of an imaging apparatus of the present invention will be described with reference to the drawings. In all the drawings for explaining the embodiments of the present invention, those having the same function are denoted by the same reference numerals, and repeated explanation thereof is omitted.

First, an embodiment in which the imaging apparatus of the present invention is applied to an MRI apparatus will be described. FIG. 1 shows an outline of the MRI apparatus. The MRI apparatus 100 obtains a tomographic image of the subject 101 using the NMR phenomenon. As shown in FIG. 1, the MRI apparatus 100 generates a static magnetic field, and a subject placed in the static magnetic field. A gradient magnetic field generator 130 that applies a gradient magnetic field to 101, a transmitter 150 that transmits a high-frequency magnetic field pulse that excites magnetization of the subject 101 with a predetermined flip angle, and an echo signal generated by the subject 101 are received. A receiving unit 160, an image reconstructed from echo signals received by the receiving unit 160, and a control unit 170 that controls operations of the gradient magnetic field generating unit 130, the transmitting unit 150, and the receiving unit 160 according to an imaging sequence, and a sequencer 140. The magnetic field generation unit (the static magnetic field generation unit 120 and the gradient magnetic field generation unit 130), the transmission unit 150, and the reception unit 160 correspond to the imaging unit of the present invention.

The static magnetic field generator 120 generates a uniform static magnetic field in the space around the subject 101. A static magnet type, normal conduction type, or superconducting type static magnetic field generation source arranged around the subject 101 is used. Prepare. There are a vertical magnetic field method and a horizontal magnetic field method depending on the direction of the generated static magnetic field. In the vertical magnetic field method, a static magnetic field is generated in a direction perpendicular to the body axis of the subject 101, and in the horizontal magnetic field method, a static magnetic field is generated in the body axis direction.

The gradient magnetic field generation unit 130 is a gradient magnetic field coil 131 wound in the three-axis directions of X, Y, and Z, which is a coordinate system (apparatus coordinate system) of the MRI apparatus 100, and a gradient magnetic field power source that drives each gradient magnetic field coil. 132. The gradient coil 131 applies gradient magnetic field pulses Gx, Gy, and Gz in the three axis directions of X, Y, and Z by driving the gradient magnetic field power supply 132 of each gradient coil 131 in accordance with a command from the sequencer 140. . Each of the gradient magnetic field pulses Gx, Gy, and Gz is applied in a direction orthogonal to the slice plane (imaging cross section) at the time of imaging, and sets the slice plane for the subject 101 and is orthogonal to the set slice plane. In addition, there is a role of applying the information in the remaining two directions orthogonal to each other and encoding position information in two directions into the NMR signal (echo signal).

The transmitter 150 irradiates the subject 101 with a high-frequency magnetic field pulse (hereinafter referred to as “RF pulse”) in order to cause nuclear magnetic resonance to occur in the nuclear spins of the atoms constituting the biological tissue of the subject 101. And a high-frequency oscillator (synthesizer) 152, a modulator 153, a high-frequency amplifier 154, and a high-frequency coil (transmission coil) 151 on the transmission side. The high frequency oscillator 152 generates an RF pulse and outputs it at a timing according to a command from the sequencer 140. The modulator 153 amplitude-modulates the output RF pulse. The high-frequency amplifier 154 amplifies the amplitude-modulated RF pulse and supplies the amplified RF pulse to the transmission coil 151 disposed close to the subject 101. The transmission coil 151 irradiates the subject 101 with the supplied RF pulse.

The receiving unit 160 detects a nuclear magnetic resonance signal (echo signal, NMR signal) emitted by nuclear magnetic resonance of the nuclear spin constituting the living tissue of the subject 101, and receives a high-frequency coil (receiving coil) on the receiving side. 161, a signal amplifier 162, a quadrature detector 163, and an A / D converter 164. The reception coil 161 is disposed in the vicinity of the subject 101 and detects an NMR signal of the response of the subject 101 induced by the electromagnetic wave irradiated from the transmission coil 151. The detected NMR signal is amplified by the signal amplifier 162 and then divided into two orthogonal signals by the quadrature phase detector 163 at the timing according to the command from the sequencer 140, and each is digitally converted by the A / D converter 164. The amount is converted and sent to the controller 170. Although the receiving coil 161 may be composed of a single coil, a multiple array coil (phased array coil) in which a plurality of small coils are combined is often used.

The sequencer 140 repeatedly applies RF pulses and gradient magnetic field pulses according to a predetermined pulse sequence. The pulse sequence describes the high-frequency magnetic field, the gradient magnetic field, the timing and intensity of signal reception, and is stored in the control unit 170 in advance. The sequencer 140 operates according to an instruction from the control unit 170 and transmits various commands necessary for collecting tomographic image data of the subject 101 to the transmission unit 150, the gradient magnetic field generation unit 130, and the reception unit 160.

The control unit 170 controls the entire MRI apparatus 100, performs operations such as various data processing, displays and stores processing results, and includes a CPU 171, a storage device 172, a display device 173, and an input device 174. The storage device 172 includes an internal storage device such as a hard disk and an external storage device such as an external hard disk, an optical disk, and a magnetic disk. The display device 173 is a display device such as a CRT or a liquid crystal. The input device 174 is an interface for inputting various control information of the MRI apparatus 100 and control information of processing performed by the control unit 170, and includes, for example, a trackball or a mouse and a keyboard. The input device 174 is disposed in the vicinity of the display device 173. The operator interactively inputs instructions and data necessary for various processes of the MRI apparatus 100 through the input device 174 while looking at the display device 173.

The imaging sequence from which the CPU 171 of the control unit 170 gives a control signal to the sequencer 140 includes a pulse sequence in which application timing of RF pulses and gradient magnetic field pulses is determined, application intensity of RF pulses and gradient magnetic field pulses, application timing, and the like. It is determined by the designated imaging parameter. The pulse sequence is preset and held in the storage device 172.

The CPU 171 implements each process of the control unit 170 such as control of the operation of each unit of the MRI apparatus 100 and various data processing by executing a program stored in advance in the storage device 172 according to an instruction input by the operator. . For example, when data from the receiving unit 160 is input to the control unit 170, the CPU 171 executes processing such as signal processing and image reconstruction, and displays the tomographic image of the subject 101 as a result on the display device 173. At the same time, it is stored in the storage device 172.

The data from the receiving unit 160 is normally data arranged in a data space called k-space. The coordinates (position) of the k space are determined by the application amount of the gradient magnetic field pulse given to the echo signal, and the points (k space matrix) to be acquired are determined according to the image size and the visual field. Conventional image reconstruction is performed using a data set that fills the entire k-space matrix. In this embodiment, the CPU 171 uses a data set obtained by thinning out k-space data and performs compressed sensing. It has a function to perform image processing and create images. The thinning measurement is performed by thinning out the encoding step of the gradient magnetic field to be encoded. That is, in the MRI apparatus of this embodiment, the control unit 170 controls the pulse sequence so as to perform thinning measurement, obtains thinned data, and performs predetermined image processing on the data set obtained by thinning measurement. To get an image.

Details of the image processing function of the control unit 170 are shown in the block diagram of FIG. As shown in the figure, the image processing unit 180 performs an operation such as Fourier transform and inverse Fourier transform on measurement data, an estimation operation based on compressed sensing, and the like, and an image reconstruction unit 186 that reconstructs an image, and the measurement data and image An average image creation unit 181 that creates data with reduced dimensions and an image update unit 182 that updates the estimated image using the reduced-order data and the reduced-order image created by the average image creation unit 181. Note that the reduced-order data created by the average image creation unit 181 is not limited to averaged data and average images, but may be data created by a median instead of the average, for example. This part is named an average image creation part 181.

Although not shown, the image reconstruction unit 186 uses an operator creation unit that creates an operator necessary for image reconstruction, such as a coil sensitivity map, an operator used by the operator creation unit, a sparse transformation matrix based on compressed sensing, and the like. An image estimation unit that estimates and updates an image, a Fourier transform unit that converts measurement data into image data, and the like. The function of updating the image using an operator such as coil sensitivity is distinguished from the image updating unit 182 that updates the estimated image using the reduced-dimensional image, and is used as the second image updating unit.

These functions of the image reconstruction unit 186, the averaged image creation unit 181 and the image update unit 182 are executed by a program incorporated in the CPU 171. That is, each of these units corresponds to a program including a predetermined algorithm installed in the CPU 171. However, a part of this function may be replaced by a known ASIC (Application Specific Integrated Circuit).

The MRI apparatus of the present embodiment is characterized in that the image processing unit 180 includes an averaged image creation unit 181 and an image update unit 182. Hereinafter, each embodiment of the operation of the image processing unit 180 will be described.

<First embodiment>
In the first embodiment, a case will be described in which the receiving coil is a phased array coil composed of a plurality of small coils, and the data set to be subjected to image processing is k-space data of 2D imaging moving images. The k-space data constituting the moving image is a plurality of time-series k-space data, and one k-space data is referred to as frame data.

FIG. 3 is a flow showing the operation of the image processing unit. This operation is based on the conventional iterative reconstruction method disclosed in Non-Patent Document 1 and the like, but an averaged image creating step S103 and an estimated image updating step S109 based on the averaged image are added. Is a feature of this embodiment. Note that the iterative reconstruction method is not limited to that described below, and the contents of steps other than steps S103 and S109 can be changed accordingly.

<< Step S101 >>
First, the image processing unit 180 receives the data 201 of each frame. FIG. 4 schematically shows frame data received by the image processing unit 180. In the figure, a portion where no data exists is displayed in white (205), and a portion where measurement data exists is displayed in black (204). In the example shown in the figure, the kx direction is the encoding direction in which data is thinned out, and each frame data is measured with the low frequency 202 in the kx direction not thinned and the high frequency 203 sparsely measured.

<< Step S103 >>
Next, the average image creation unit 181 creates an average image from a plurality of frame data. As shown in FIG. 4, the average process is a process for creating the average image data in the k-space by the average of the frames. However, the average process is different from a general average in that the average is performed only for the frames in which the data exists. . That is, the data having no data in only one frame is used as it is as indicated by the arrow 207, but the data having the data in a plurality of frames is averaged as indicated by the arrow 208. The average image 209 is obtained by performing inverse Fourier transform on the data thus averaged, that is, the data 206 in the k space. This average image 209 is used for a process of updating an image with a new type of constraint condition in an update process (step S109) described later. Note that, as described above, the average image creation unit 181 may replace the average with a median, thereby creating reduced-order data. In the following description, an average case will be described as a representative.

<< Step S104 >>
In step S104, a coil sensitivity map may be created and used for the coil sensitivity in S105, or the operator creation unit may use an operator (operator related to coil sensitivity) using the data of the low frequency 202 that is densely acquired. ) May be created. As an operator, a SPIRIT operator (Non-Patent Document 1) or a GRAPPA operator (Non-Patent Document 2) may be used depending on the type of reconstruction method using coil sensitivity. In the flowchart of FIG. 3, a sensitivity map or the like is shown as a representative. Further, the data acquisition method used when obtaining the sensitivity map or the like is not limited to this. For example, the measurement data that is the target of image processing does not have to be data that is densely measured in the low frequency range, and data in which the low frequency range 202 is acquired in advance before the main measurement can also be used. Also, a method of creating a sensitivity map by combining frames (for example, Patent Document 1) can be adopted.

<< Steps S105 to S109 >>
Thereafter, the processes shown in steps S105 to S109 are repeated. These processes are for solving the following equations (1) to (4) by a POCS (Projection on Convex Sets) algorithm. Except Expression (3), the expression (constraint condition) is the same as that employed by the prior art (Non-Patent Document 1). Expression (1) is Step S108, Expression (2) is Step S105, and Expression (3) is Step S109. Equation (4) is a constraint condition for each processing in step S110. If a single coil is used instead of a multiple coil, the process of step 105 is omitted. In the flow of FIG. 4, the repetitive processing is performed in the order of steps S105, S108, S109, and S110. However, the order is not limited to this and is arbitrary.

In the equation, x is a vector representing an image corresponding to each coil of acquired data, y is a vector representing acquired data, Ψ is a matrix representing conversion to a sparse space in the compressed sensing technique, and is assumed to be a wavelet transform as an example. JointL1 is expressed by equation (5),

(Where c is the coil index and r is the position index)
The sparse space coefficient w _rc is the square root of the sum of squares with respect to c and the sum with respect to r. As described above, the specific procedure of the iterative reconstruction method is not limited to this. For example, in Equation (5), other norms such as the L1 norm may be minimized without using JointL1.

G in the formula (2) is, for example, a SPIRIT operator,

The matrix notation x = Gx in Equation (6). In the equation (6), i and j are coil indexes (corresponding to the coil index c and the position index r), and a convolution of g _ij and x _i is taken with respect to the position. In step S104, G is obtained by optimizing so that Gx = x is satisfied in the low frequency range 202 in which data is densely acquired.

Indicates the norm of x. The L2 norm is used as an example, but not limited to this. For example, the norm weighted according to the coil

But it ’s okay. In the above equation, kc represents the weight of each coil.

In the equations (2) to (4), ε ₁ , ε ₂ , and ε ₃ are used to allow the data to not completely match due to various errors.

M and M ′ in Equation (3) are matrices representing conversion to an average. M is a simple frame average. M ′ is a frame average in an area where data exists. F is a matrix representing the Fourier transform, and F ⁻¹ is a matrix representing the inverse Fourier transform. Note that the same symbol is used for the same conversion, but the specific matrix shape is changed according to the vector to be converted.

The restriction on the average image is not limited to the expression (3). For example, the coil may be synthesized as shown in Expression (3) ′. Alternatively, the order of coil synthesis and averaging may be changed as in equation (3) ".

D in Equation (4) is a matrix representing a transformation that extracts only elements corresponding to the acquired data y.

The processing of each step will be specifically described based on the outline of steps S105 to S110 described above.

Step S105 is processing corresponding to equation (2) (processing by the second image updating unit), and updates the estimated image using a sensitivity map (here, SPIRIT operator). That is, in the k-th iteration, the (k−1) th estimated image x _k−1 is received, and the estimated image is updated to m _k as shown in Expression (7). Initial value x of the estimated image ₀ average image generated in step S103 for example, the image such that the entire frame.

Step S108 is processing corresponding to Expression (1), and as shown in Expression (8), the estimated image m _k in Step S105 is updated to w _k based on the sparse conversion of compressed sensing.

Here, S _λ (x) is a Joint-Soft-thresholding function, and the soft-thresholding is performed after collecting the coils corresponding to the equation (5) (hereinafter the same). The function S ′ _λ () for the coil summary is defined as follows for the vector w _r = (w _r1 , w _r2 ,..., W _rNc ) for the wavelet coefficient coil at position r. Nc is the number of coils. S _λ () applies S ′ _λ () to all positions r.

Step S109 corresponds to Expression (3), and the estimated image update unit 182 updates the estimated image to a _k as represented by Expression (10). Note that λ represents the strength of the constraint, and may be automatically determined as a noise level, or a method of adjusting during repetition (Non-Patent Document 5) may be employed.

This process gives new data matching restrictions, and unlike the conventional data matching restrictions expressed by the equation (4), the estimated image x and the acquired data y are converted to different irreversible transformations M. And M ′ are converted into a low-dimensional space, and data consistency is considered. The average image transformation M considered in this example is a simple frame average for x. Here, if simple frame averaging is performed on the acquired data y, it only has the meaning of weighting the data in the equation (4), but F ⁻¹ M′y is converted, and artifacts due to thinning are reduced. Since the suppressed frame averaging is performed, it is possible to add a coincidence restriction that suppresses artifacts. From the definition, F ⁻¹ M′y better represents the average image than the simple frame average. Further, since the thinning method is not largely biased, F ⁻¹ M′y is considered to be an approximation of the measured value of the average image, and can certainly be used as a restriction on the consistency of the estimated image with the average image Mx.

Unlike the method in which the difference between each frame and the average of each frame is used for optimization described in Patent Document 4, this coincidence restriction is close to an average image that is not a true image. 4 does not include possible problems.

Note that λ2 that determines the strength of this constraint may be automatically determined in the same manner as λ1 in the case of equations (8) and (9), or may be adjustable in repetition. Also, the user may be able to set.

Next, step S110 is a process corresponding to equation (4), and the estimated image a _k in step S109 is updated to x _{k as} in equation (11).

Here, D is the same conversion as defined in Equation (4), in which only the elements corresponding to the acquired data are extracted. That is, the acquired data is replaced with the data by the processing of Expression (11). It should be noted, may be replaced by exactly the same data as in Equation (11), may be Motashi the error using the S _λ.

<< Steps S111 and S112 >>
The above steps S105 to S110 are repeated until a predetermined end condition is satisfied (S111), and if the end condition determined in step S111 is satisfied, the process proceeds to creation of an output image in step S112. The termination condition is, for example, a condition that whether or not the repetition has been made a predetermined number of times. Various known conditions can be used as the end condition, such as determining the update amount of the estimated image.

In step S112, the estimated image created up to step S111 is created in a format desirable for the user. In this example, the estimated image is an image received by each coil, and the output image is created as shown in Expression (12).

In Expression (12), N _c is the number of coils. The creation of the output image is not limited to this. For example, as in the technique described in Non-Patent Document 3, a method of adaptively combining coils may be used. Further, as in the technique described in Non-Patent Document 4, in step S104 in FIG. 3, an estimation image is created for a plurality of coil sensitivity maps, and in the repetition of steps S105 to S110, estimation for each coil sensitivity map is performed. A method of creating an image and synthesizing an estimated image for each coil sensitivity map in step S112 can also be employed.

In the flow shown in FIG. 3, after receiving all the frame data in step S101, the process may proceed to the subsequent steps, but it is not necessary to receive all the frame data. For example, data may be acquired sequentially while performing image reconstruction processing.

According to the present embodiment, in addition to the restriction of sparse compression sensing and the restriction based on the coil sensitivity of parallel imaging, a new type of restriction different from these, that is, matching the converted image into a low-dimensional space. By adding the restriction on the nature, the range of the solution of the estimated image can be narrowed down, and the estimated image can be brought closer to the correct solution than the conventional compressed sensing technique.

The effect by this embodiment is shown in FIG. Data without thinning was prepared as a reference. Reconstruction was performed on data obtained by thinning out reference data (thinning rate: 50%), and a difference image from the reconstructed image of the reference data was defined as d. Images were evaluated using RMS (root mean square) of S _λ (d) in an appropriate ROI (region of interest) as an index of difference. Since the coil is already organized, S _λ (d) is the same as the Soft-Threshold function. As shown in FIG. 5, when reconstruction is performed with an average image constraint (image estimation including step S109) based on a difference index when reconstruction is performed without average image constraint (image estimation without step S109). It can be seen that the difference index is smaller and an image close to the true image is obtained.

As mentioned above, although embodiment which applies the 1st aspect of this invention to the frame data of the moving image imaged with the MRI apparatus was demonstrated, it is possible to add a various change on the basis of this embodiment. Hereinafter, some examples of various modifications will be described.

<Modification Example 1 of First Embodiment>
In the embodiment shown in FIG. 4, the data set thinned out in the kx direction in the k space is used. However, the first modification example is an example using data thinned out in both the kx and ky axes in the k space. is there.

FIG. 6 shows a data set 201 of each frame and averaged data 206 created from them. Also in this case, the method of creating averaged data by the average image creating unit 181 is the same as in the first embodiment described above, and the data is used as it is for an area in which data exists only in one frame. For areas where data is present in the frames, the data are averaged. In this way, averaged data 206 is created. As a method of obtaining the average image 209 (FIG. 4) from the created average data 206, if there is little missing data in the average data 206, the image may be reconstructed by inverse Fourier transform as in the first embodiment. The average image may be reconstructed by parallel imaging using a sensitivity map.

The processes (steps S105 to S110) shown in FIG. 3 are repeatedly performed using the average image thus created and the subsequent process S112 is the same as in the first embodiment.

<Modification Example 2 of First Embodiment>
In the first embodiment, the average image creation unit 181 creates data obtained by averaging the acquired data in the frame direction, and creates an average image from the averaged data. First, after reconstructing an image from the acquired data, The images may be averaged. This is shown in FIG. In FIG. 7, 210 is each frame image, 211 is a subject, and 212 is an average image. In this case, the order of F ⁻¹ M′y in Equation (3) is changed to M′F ⁻¹ y. Moreover, a simple average or a weighted average can be used as a conversion matrix at the time of averaging.

<Modification 3 of the first embodiment>
The average image creation unit 181 performs the average on the partial frames 201P as shown in FIG. 8 instead of the average over all frames, and sets the constraint condition shown in Expression (3) for the average image, You may perform the process of Formula (10).

Further, as shown in FIG. 9, a plurality of partial average images may be created. In that case, a plurality of average images 209A, 209B,... Are passed to the constraint condition, and the constraint of Expression (3) corresponding to each is given. That is, a plurality of expressions (3) are prepared. The specific update process (10) is also performed a plurality of times corresponding to each average image.

Furthermore, the data weight of each frame (230A, 230B, 230C,...) Of the estimated image included in the partial average image may be different. In that case, the average of each frame of the estimated image is also averaged by applying a corresponding weight. For example, in FIG. 9, the average image of the estimated image corresponding to the average image 209A is pixel (230A) × (3/4) + pixel (230B) × (1/4), and the estimated image corresponding to the average image of 209B The average image is assumed to be pixel (230B) × (2/4) + pixel (230C) × (2/4).

<Modification 4 of the first embodiment>
Although the first embodiment is intended for moving image frame data, the present embodiment can be applied to a plurality of data sets having different conditions such as imaging parameters instead of moving image frame data.

FIG. 10 shows an example in which the present invention is applied to an imaging method in which TE (echo time) is changed to obtain an image 220 corresponding to each. In this example, averaging processing is performed on a plurality of images 220 having different TEs, and the average image obtained by this processing is used as a constraint condition for the estimation processing of each image 220. Further, FIG. 11 is an example applied to an imaging method in which TR (repetition time) is further changed in addition to TE to acquire an image 221 corresponding to each. In this imaging method, averaging may be performed in a set 220 of images having different TEs, or may be averaged in a set 221 of images having different TR changes. Moreover, you may average with respect to both TE and TR, respectively.

<Modification 5 of the first embodiment>
Although the first embodiment is intended for a plurality of 2D data sets such as a plurality of frame data, the present embodiment can also be applied to one data or a 3D data set.

FIG. 12 is a diagram illustrating a case where image estimation processing is performed on one 2D data set. In this example, sparsely measured high frequency data is averaged in the kx-axis direction to create averaged data parallel to the ky axis. The average image obtained by reconstructing such averaged data corresponds to pseudo projection data depending on the nature of the image. The use of this for the constraint of image estimation is the same as in the first embodiment, thereby obtaining one image. Note that FIG. 12 illustrates the case where the averaging process in the kx-axis direction is performed, but the axis is arbitrary, and the averaging process may be performed in the ky-axis direction.

Further, when the imaging is 3D imaging and the measurement data is a 3D data set, the averaging process may be performed in the z direction as shown in FIG. Also in this case, similarly to the fourth modification, it is possible to perform partial averaging processing in the z direction or weighting. In addition, when averaged data is created from the measurement data and the averaged data is reconstructed to obtain an average image, the averaging process may be performed in the kz direction.

<Extended example of the first embodiment>
The MRI apparatus of the first embodiment can add a function for supporting the image estimation processing according to the first embodiment based on the apparatus configuration shown in FIGS. 1 and 2. As an additional example, a function addition example using a user interface (UI) will be described.

As described above, when the image processing unit of the first embodiment performs image estimation processing by adding new constraints using the thinned data, the creation of an average image for determining new constraints Various averaging can be performed, such as averaging of directions, averaging of k-space in each axial direction, averaging of different TEs and TRs, and the like. The axis to be averaged may be automatically determined according to other imaging parameters, for example, in the time direction when capturing a moving image of the heart. However, a user can prepare a UI as shown in FIG. You may let me decide.

Also, regarding the strength λ2 of the constraint condition used in step S109, a UI as shown in FIG. 15 may be prepared and determined by the user.

These user interfaces can be constructed in the display device 173 and the input device 174 and the CPU 171 shown in FIG. The image processing unit 180 (average image creation unit 181) performs an averaging process according to the axis set via the UI, and the estimated image update unit 182 uses the equation (10) with the strength constraint set via the UI. ) To update the image.

As mentioned above, some modified examples and extended examples based on the first embodiment have been described, but these modified examples can be arbitrarily combined as long as there is no technical contradiction, and these arbitrary combinations are also implemented in this embodiment. Included in the form.

<Second embodiment>
This embodiment is characterized by having an artifact suppression function. That is, the imaging apparatus according to the present embodiment includes an imaging unit that collects a data set necessary for image reconstruction, and an image processing unit that reconstructs an image using the data set, and the data collected by the imaging unit. The set is a data set including missing data, and the image processing unit uses the data set to estimate an image under a constraint condition of coincidence between the data set and the image, and the missing data An artifact extraction unit that extracts an artifact included in an image reconstructed from a data set that includes data; and a data correction unit that corrects the data set using the artifact information extracted by the artifact extraction unit, and The artifact extraction unit is a data set composed of one data set and the one data set and other data sets. With bets, which calculates the artifacts included in the one data set, the image estimation unit performs image estimation processing for the data set that is corrected by the data correcting section.

In the imaging apparatus of the present embodiment, the image processing unit may or may not include an average image creation unit and an estimated image update unit based on the average image, as in the first embodiment. There may be. Hereinafter, the present embodiment will be described using an example in which the imaging apparatus is an MRI apparatus.

The overall configuration of the apparatus is the same as the block diagram shown in FIG. The configuration of the image processing unit, which is a feature of this embodiment, is shown in the block diagram of the control unit in FIG. In FIG. 16, the same elements as those described in FIG. 2 are denoted by the same reference numerals, and description thereof is omitted.

As shown in the figure, in addition to the image reconstruction unit 187, the image processing unit 180 includes a frame synthesis data creation unit 183, an artifact extraction unit 184, and an artifact suppression unit 185. Similar to the image reconstruction unit 186 of the first embodiment, the image reconstruction unit 187 includes an operator creation unit that creates an operator necessary for image reconstruction such as a coil sensitivity map, an operator used by the operator creation unit, and compressed sensing. An image estimation unit that estimates an image using a sparse transformation matrix based on the data, a Fourier transform unit that converts measurement data into image data, and the like. The image reconstruction unit 187 may further include an average image creation unit 181 and an image update unit 182 included in the image processing unit 180 of the first embodiment.

The frame composition data creation unit 183 composes a plurality of data sets and creates a composite image. The artifact extraction unit 184 extracts artifacts using an image created from a data set before synthesis and a synthesized image. The artifact suppression unit 185 suppresses artifacts in the estimated image using the artifact information extracted by the artifact extraction unit 184. That is, the artifact suppression unit 185 functions as a data correction unit.

Next, the operation of the image processing unit having the above configuration will be described with reference to the flow of FIG. This flow shows the operation when the image processing unit 180 includes the average image creation unit 181 and the image update unit 182 as in the first embodiment. Compared with the flow of FIG. Data creation), step S106 (artifact extraction), and step S107 (artifact suppression) are added. Detailed description of the same processing as in FIG. 3 is omitted. In the flowchart of FIG. 17, the data flow is indicated by “no synthesis”, “frame synthesis A”, and “frame synthesis B” in order to indicate whether each step handles data without synthesis or data with frame synthesis. Yes. Since the data can be applied to data in k-space or image data, the flowchart does not distinguish whether the data is in k-space or an image.

<< Step S101 >>
First, the image processing unit 180 receives data of each frame. Data received by the image processing unit 180 is, for example, k-space data at least partially measured sparsely as shown in FIGS. FIG. 18 shows, as an example, frame data thinned out in the kx direction with data in the k-space of a 2D moving image similar to FIG. The image processing unit 180 may receive sequentially, instead of receiving the frame data of all the frames.

<< Step S102 >>
The frame synthesis data creation unit 183 creates two frame synthesis data for one frame data by using the frame data before and after the frame data. This is shown in FIG.

When creating the nth frame composite data, the frame synthesis A data 251 is created using the (n-1) th frame data 301 and the nth frame data 302, and the nth frame data 302 and the (n + 1) th frame data. The frame composition B data 252 is created using the information 303. At this time, the low-frequency data of the frame composition A data 251 and the frame composition B data 252 is the low-frequency data measured closely of the n-th frame data 302. The high-frequency data is data randomly selected from the data without synthesis. Specifically, the frame composition A data 251 is created so that the low frequency is data of the nth frame of the non-synthesis data, and the high frequency is data randomly selected from the nth and (n-1) th frames. Is done. The frame synthesis B data 252 is created so that the low frequency is data of the nth frame of the non-combined data, and the high frequency is data randomly selected from the nth and n + 1th frames.

By doing so, the arrangement of data becomes random, and in the frame synthesis A and B, artifacts due to thinning are less likely to interfere. The method of creating the data for frame synthesis A and B is not limited to the method described above. For example, the low frequency may be selected at random from two frames. In addition, when selecting at random in frame synthesis, when all the frames are considered, the data may be allocated so as not to overlap too much. That is, data that was not used to create the nth composite frame is used to create the (n−1) th or (n + 1) th composite frame. In this way, the frame average images can be made completely equal for both no synthesis and frame synthesis.

<< Steps S103 and S104 >>
Similar to the first embodiment, an average image and a sensitivity map used for image estimation processing are created.

<< Step S105 >>
The image reconstruction unit (image estimation unit) receives “no synthesis”, “frame synthesis A”, and “frame synthesis B” data from the frame synthesis data creation unit 183, and for each, the sensitivity created in step S104. The estimated image based on the coil sensitivity is updated by a map, for example, by a SPIRIT operator. Here, for each data, the estimated image is updated according to the above-described equation (2).

Thereafter, the data of no synthesis and frame synthesis A and B are passed to the artifact extraction unit 183 (step S106), and the process proceeds to step S108.

<< Step S106 >>
In step S106, an artifact is extracted. Assume that an image without synthesis is x, a frame synthesis A image is a, a frame synthesis B image is b, and an artifact image to be extracted is c. Let c _{i be obtained} as follows, where x _i , a _i , b _i , and c _i are the pixel values of each complex number in pixel _i .

In formula (13), max (m, n) means the larger of m and n.

The intention of this processing will be described with reference to FIG. As described above, the frame composite images a and b are composed of data randomly selected from the two non-compositing data, and the artifact interference is suppressed. The artifact remains strong. However, since the images a and b are mixed with the adjacent frame, some difference from the adjacent frame remains. In the two difference images (the difference image between the uncombined image x and the image a and the difference image between the uncombined image x and the image b), it is expected that the artifact is generated in the same manner. Can be emphasized. In Expression (13), the process of multiplying and extracting only positive signs and taking the square root to return the intensity corresponds to the process of taking an AND. Also, a sign is assigned to sign to leave information on whether the artifact was generated so as to increase or decrease the absolute value.

In this way, by taking an AND, it is possible to extract an artifact without leaving an error that occurs when a composite image is created, compared to a case where an artifact is simply extracted from a difference from the composite image. Further, in the conventional method (for example, Patent Document 3), it is premised on the existence of data having different artifacts for the frames to be superimposed. For this reason, the data that can be applied is limited in the conventional method, but in this embodiment, the data in which the artifact is suppressed is newly created by mixing the frames, and the data with different artifact is not essential. Wide range. Also, artifacts can be extracted from images of different frames by taking AND between frames without relying on fusion.

<< Step S107 >>
In step S107, artifacts are suppressed as follows.
Here, arg (x) is a declination of x. Since the artifact has no phase information and only positive and negative information, in this process, the absolute value | xi | of the xi is first changed so as to cancel the artifact (ci) with the absolute value, and then the phase information is returned. . S _λ3 is a Joint-Soft-Thresholding function defined by Equation (9). Alternatively, only Soft-Thresholding that does not summarize coils may be performed.

By the above processing, data without synthesis with artifacts suppressed can be obtained. This unsynthesized data is passed to the image estimation unit (step S108).

<< Steps S108 to S111 >>
In steps S108, S109, S110, and S111, the same processing as that of the first embodiment is performed on each of the non-combined data processed in step S107 and the frame combined data A and B passed from step S105. That is, update of the estimated image based on the sparse conversion of compressed sensing (for example, Expression (1)), update of the estimated image using the constraint based on the average image created in Step S103 (for example, Expression (3)), and the acquired data The estimated image is updated based on the coincidence (for example, Expression (4)). In step S109, only the average image of the non-compositing data is received, and all the data are updated based on the same average image.

Step S111 and subsequent steps are the same as in the first embodiment.

As mentioned above, although the imaging device of the second embodiment has been described with reference to the flowchart of FIG. 17, the present invention is not limited to this, and various changes can be made.

For example, the artifact suppression processing shown in steps S102, S106, and S107 of FIG. 17 can be performed independently of the processing (S103, S109) that uses the average image of the first embodiment. In this case, as shown in FIG. 19, only the processing for suppressing artifacts is added to the conventional image estimation processing (repetitive reconstruction).

FIG. 17 and FIG. 19 are examples of performing artifact suppression during repetitive reconstruction. However, artifact suppression may be performed like filter processing on an image. In that case, as shown in FIG. 20, only the processing of steps S102, S106, and S107 is performed.

Further, in the flowcharts of FIGS. 17 and 20, the non-compositing data is passed to steps S103 and S104. However, as shown in FIG. 21, the data of frame composition A and B is created immediately after step S101 in step S102. The data of frame synthesis A and B may also be passed to steps S103 and S104, and an average image and a coil sensitivity map may be created for each and used for subsequent processing.

In addition, when performing the same process as the process of 1st embodiment, it is possible to employ | adopt the various modification examples demonstrated in 1st embodiment independently or in combination. Also, a UI function can be added as in the first embodiment.

According to the present embodiment, a plurality of synthesized data is created, and the artifacts are extracted by taking the AND of the difference between the synthesized data and the data before synthesis, thereby extracting the artifacts with less error and less errors. Can be suppressed. Thereby, the accuracy of the image estimation process can be further improved.

The effect by this embodiment is shown in FIG. Data without thinning was prepared as a reference. Reconstruction is performed on data obtained by thinning out reference data (thinning rate: 50%), and a difference image from the reconstructed image of the reference data is defined as d. Images were evaluated using the RMS of Sλ (d) at an appropriate ROI as an index of difference. Since the coil is already organized, S _λ (d) is the same as the Soft-Threshold function. It can be seen that the difference index when the reconstruction is performed with the average image constraint is smaller than the difference index when the reconstruction is performed without the average image restriction, and an image close to the true image is obtained.

In the above, each embodiment which applied the imaging device of the present invention to the MRI apparatus and its modification example were explained. Next, an embodiment in which the imaging apparatus of the present invention is applied to a CT apparatus will be described. Before describing each embodiment, the overall configuration of the CT apparatus will be described with reference to the drawings.

FIG. 23 is a block diagram showing an overall outline of the CT apparatus. In this CT apparatus, a rotating plate 193 in which an X-ray source 191 and an X-ray detector 192 are arranged to face each other, and an opening provided in the center of the rotating plate 193 are arranged in a direction crossing the rotating direction of the rotating plate 193. An X-ray source 191 emits X-rays to a subject (not shown) laid on the bed 194, and the X-ray detector 192 detects X-rays passing through the subject. To do. The X-ray detector 192 includes a large number of X-ray detection elements arranged in an arc shape, and a one-dimensional array detector along the arc or a two-dimensional array detector arranged in a direction orthogonal to the rotation direction. Is used.

The timing of irradiating the X-ray from the X-ray source 191 is controlled by the X control unit 197. The rotation of the rotating plate 193 and the movement of the bed 194 are controlled by the rotating plate control unit 199 and the bed control unit 198, respectively. Under these controls, projection data for different rotation angles of the rotating plate 193 is obtained by repeating the above-described X-ray irradiation and X-ray detection while rotating the rotating plate 193 at a predetermined rotation speed. It is done.

Projection data is collected by the data collection unit 195 and used for reconstruction of a tomographic image of the subject. The data collected by the data collection unit 195 is data in which one axis is a channel direction (X-ray detection element arrangement direction) and the other axis is a view direction (rotation angle direction of the rotating plate).

The image processing unit 196 reconstructs a tomographic image of the subject using the data collected by the data collecting unit 195. For the reconstruction of tomographic images, in addition to the back projection method, a successive approximation reconstruction method for reconstructing an image by iterative calculation is adopted.

Based on the above configuration, an embodiment of a CT apparatus to which the present invention is applied will be described below.

<Third embodiment>
This embodiment will also be described by taking as an example a case where time-series frame data such as heart moving image capturing is obtained, but the frame data is not limited to time-series data, and can be replaced with a plurality of data. .

The configuration of the image processing unit 196 in the present embodiment is similar to the configuration of the image processing unit 180 of the MRI apparatus shown in FIG. 2, and the image reconstruction unit 186 of the image processing unit 180 in FIG. Since it has the same function (average image creation unit 181 and estimated image update unit 182) except that it is an image reconstruction unit having a function of performing projection and successive approximation reconstruction, illustration is omitted, Description will be made with the aid of FIG. 2 as necessary.

Hereinafter, the operation of the image processing unit 196 of the CT apparatus of this embodiment will be described with reference to the flowchart of FIG. In this flow, steps S103, S108, and S109 are added to the conventional successive approximation reconstruction. Note that S108 is a step of updating an estimated image based on a known sparse conversion of compressed sensing.

First, data is received from the data collection unit 195 (S101). An example of data used for reconstruction is shown in FIG. As described above, data constituting one frame data is projection data in which one axial direction is a view direction and the other axial direction is a channel direction. The data x handled by MRI is a complex number (vector), but here it is a real number. Data with different view angles is acquired for each frame 201. This shortens the acquisition time of one frame data and captures an image with high time resolution in moving image capturing.

The frame data is transferred to the average image creation unit 181 (step S103), and image estimation processing based on compressed sensing is performed (S108 to S110). As shown in FIG. 25, the average image is created by using the data for data in only one frame, and for the data in a plurality of frames by using the average of the data of these frames as the averaged data 206. . This averaged data 206 is used as an average image 209 using back projection or a successive approximation method. The average image 209 is used for updating the estimated image in the image estimation process (step S109).

The processes in steps S108 to S110 are the same as those in the first embodiment except that there is no update of the estimated image based on the coil sensitivity in the first embodiment, and are the same as those in the first embodiment, and equations (1), (3), (4 ) Can be expressed by the following formulas (1A), (3A), and (4A).

In Formulas (3A) and (4A), P ⁻¹ and P are conversion processes between data images. Specifically, filter-corrected backprojection-projection, image reconstruction processing by a successive approximation method, and its inverse conversion are performed. Represents.

That is, in step S108, the estimated image is updated based on the sparse transformation according to the equation (1A), and the estimated image is further updated according to the equation (3A) with respect to the estimated image updated in step S108. Since there are no multiple coils, the JointL1 process is the same as the normal L1 norm. In step S110, the estimated image is updated based on the consistency of the acquired data according to the equation (4A). Specifically, the estimated image of the frame is updated as shown in Expression (11A).

An end condition is determined in step S111, and an output image of CT is created in step S112.

The embodiment in which the present invention is applied to the image processing unit of the CT apparatus has been described above. However, the modified example of the first embodiment described in the first embodiment can be similarly applied to the third embodiment. . For example, instead of using data thinned out in the view direction, the present invention can be applied to data thinned out in the channel direction or data thinned out in both directions. Further, as shown in FIG. 25, the data to be handled may be not only data that is sparse evenly, but also data that is biased between sparse and dense portions. For example, data obtained by alternately collecting half-scan data from 0 ° to 180 ° and half-scan data from 180 ° to 0 ° for each frame may be used.

Also, it is possible to use a partial average instead of an average for all frames as a constraint. Furthermore, it is possible to use an average image obtained by averaging not only time-series frame data but also 3D data in the z direction as a constraint. In addition, various modifications can be applied. Furthermore, a UI function such as a UI for setting the strength of restrictions may be added as appropriate.

According to the present embodiment, it is possible to obtain a high-quality image in which artifacts are suppressed using measurement data with a reduced number of data even in a CT image. For example, in the case of performing successive approximation reconstruction with moving image capturing of the heart, if data with different projection angles is acquired for each cardiac phase, data with an increased number of types of projection angles can be obtained on the same average in the time direction, An average image with reduced artifacts can be created. By adding a restriction based on such an average image, a more accurate image, that is, an image close to a true image can be obtained.

<Fourth embodiment>
In this embodiment, the configuration of the image processing unit of the second embodiment is applied to a CT apparatus.
The configuration of the image processing unit 196 in this embodiment is similar to the configuration of the image processing unit 180 of the MRI apparatus shown in FIG. 16, and the image reconstruction unit 186 of the image processing unit 180 in FIG. It has the same functions (frame synthesis data creation unit 183, artifact extraction unit 184, artifact suppression unit 185) except that it is an image reconstruction unit having a CT image reconstruction function for performing successive approximation reconstruction. Therefore, illustration is abbreviate | omitted and FIG. 16 is used and demonstrated as needed. The image processing unit 196 of the present embodiment may or may not include the average image creation unit 181 and the estimated image update unit 182 included in the image processing unit of the third embodiment.

Hereinafter, the operation of the image processing unit of this embodiment will be described with reference to the flowchart shown in FIG. Also in this figure, in order to make it easy to understand the data handled in each step, the data flow is indicated by “original frame” (no synthesis), “synthesis frame 1”, and “synthesis frame 2”.

The processing shown in FIG. 26 is based on the assumption that the image processing unit 196 includes an average image creation unit 181 and an estimated image update unit 182. Compared with FIG. 24, the processing of step S102, steps S106, and S107 is included. Is different. In other words, the processing of the image processing unit according to the present embodiment has steps S102, S103, S106, S107, S108, and S109 added to the conventional successive approximation reconstruction.

The processes in steps S103, S108, and S109 are the same as those described in the third embodiment. In the third embodiment, these processes are performed on data that has not been combined. The point of processing is different.

Steps S102, S106, and S107 are the same as in the second embodiment except that the data to be handled is CT data (projection data or image data). FIG. 27 shows a synthesis example using half scan data. In this example, the data for one rotation is divided in half, and one half and the other half are obtained alternately. In the frame synthesis 1, in the n-th frame synthesis, the data on one side of the (n−1) th frame and the data on the n-th side are used as frame synthesis 1 data. In frame composition 2, data on one side of the nth frame and data on the n + 1th one side are used as frame composition 2 data. In this way, the synthesis is completed simply by combining the data of the two frames.

Note that FIG. 27 shows a synthesis example using data obtained by half scan, but the present invention can also be applied to frame data as shown in FIG. 25 and data thinned out in an arbitrary axial direction, for example.

Thereafter, the artifact is extracted by taking the AND of the difference between the image of the frame composition 1 and the image of the frame without the composition (original frame) and the difference between the image of the frame composition 2 and the image of the original frame, and the original frame Suppresses image artifacts. After that, image estimation processing (S108 to S110) is performed as in the third embodiment, and a CT image is output in step S112.

As described above, the present invention (second aspect) can also be applied to CT apparatus or data measured by a CT apparatus, and the same effect as that of an MRI apparatus, that is, an image with reduced artifacts is obtained. be able to.

Note that this embodiment can be modified in the same way as described in the second embodiment. For example, when the image processing unit 196 does not perform the restriction using the average image, the same as the flowchart of FIG. In addition, steps S103 and S109 can be omitted. The present embodiment can also be applied to a case where compressed sensing is not performed. In this case, the artifact suppression process can be regarded as a filter process as in the flowchart of FIG.

As described above, the embodiment in which the present invention is applied to the CT apparatus which is an imaging apparatus other than the MRI apparatus has been described according to the third embodiment and the fourth embodiment. However, the first aspect of the present invention is a repetition of acquiring missing data. If an imaging apparatus that performs reconstruction is added, the same image processing unit as described in this embodiment is added to various imaging apparatuses other than the MRI apparatus and the CT apparatus, such as an ultrasonic imaging apparatus and a PET apparatus. Can be applied. The second aspect of the present invention can also be applied to an image processing unit that does not repeatedly perform reconstruction, and an artifact suppression effect can be obtained.

Further, in the first to fourth embodiments described above, the embodiment in which the present invention is applied to the image processing unit of the imaging apparatus has been described. However, the present invention is an image processing apparatus independent from the imaging apparatus, that is, data acquired by the imaging apparatus. Any device having a function of receiving and reconstructing an image can be applied. As the apparatus configuration as an independent image processing apparatus, for example, the configuration of the image processing unit shown in FIGS. 2 and 16 can be used as it is.

According to the present invention, a new image processing technique capable of reconstructing a high-quality image in which artifacts are suppressed is provided.

DESCRIPTION OF SYMBOLS 100 ... MRI apparatus, 101 ... Subject, 120 ... Static magnetic field generation part (imaging part), 130 ... Gradient magnetic field generation part (imaging part), 131 ... Gradient magnetic field coil, 132. ..Gradient magnetic field power supply, 140... Sequencer, 150... Transmission unit (imaging unit), 151... Transmission coil, 152 .synthesizer, 153. DESCRIPTION OF SYMBOLS 160 ... Reception part (imaging part), 161 ... Reception coil, 162 ... Signal amplifier, 163 ... Quadrature phase detector, 164 ... A / D converter, 170 ... Control part 171 ... CPU, 172 ... storage device, 173 ... display device, 174 ... input device, 180, 196 ... image processing unit, 181 ... average image creation unit (low-dimensional data) Creation unit), 182... Estimated image update 183: Frame composition data creation unit, 184: Artifact extraction unit, 185 ... Artifact suppression unit, 186, 187 ... Image reconstruction unit, 191 ... X-ray source (imaging unit), 192: X-ray detector (imaging unit), 193 ... rotating plate, 194 ... bed, 195 ... data collection unit, 201 ... data of each frame, 201P ... partial Frame data 202 ... Low frequency, 203 ... High frequency, 204 ... Data present portion, 205 ... Data missing portion, 206 ... Averaged data, 207, 208 ... Average processing, 209, 209A, 209B ... Average image, 220 ... Image corresponding to TE, 221 ... Image corresponding to TR, 230A, 230B, 230C ... Estimated image Frame, 251 ... Frame Composite A, 252 ... Frame Composite B, 301 ... n-1 th frame data, 302 ... n-th frame data, 303 ... n + 1 th frame data.

Claims (18)

1. An imaging apparatus comprising: an imaging unit that collects a data set necessary for image reconstruction; and an image processing unit that reconstructs an image using the data set,
   The data set collected by the imaging unit is a data set including missing data,
   The image processing unit, using the data set, an image estimation unit that estimates an image under a constraint condition of coincidence between a data set and an image;
   A low-dimensional data creation unit for reducing the dimensions of the data set and the image;
   An image update unit that updates the estimated image estimated by the image estimation unit under a constraint condition of matching between the low-dimensional data set and the low-dimensional image created by the low-dimensional data creation unit;
   An imaging apparatus comprising:
2. The imaging apparatus according to claim 1,
   The imaging apparatus, wherein the low-dimensional data creation unit creates a low-dimensional data set by performing an averaging process on an area where data of the data set exists.
3. The imaging apparatus according to claim 1,
   The low-dimensional data creation unit creates a low-dimensional data set by filtering the plurality of data sets.
4. The imaging apparatus according to claim 1,
   The imaging unit includes a magnetic field generation unit that generates a static magnetic field and a gradient magnetic field, a transmission unit that generates a high frequency magnetic field, and a reception unit that receives a nuclear magnetic resonance signal generated by the high frequency magnetic field, and the data set is k-space data. There is an imaging apparatus.
5. The imaging apparatus according to claim 4, wherein the receiving unit includes a plurality of receiving coils, and the plurality of data sets are data sets generated from nuclear magnetic resonance signals respectively received by the plurality of receiving coils. ,
   The imaging apparatus, comprising: a second image updating unit that updates an estimated image based on sensitivities of the plurality of receiving coils.
6. The imaging apparatus according to claim 4,
   The data set collected by the imaging unit is frame data obtained in time series,
   The low-dimensional data creation unit creates a low-dimensional data set by averaging a plurality of frame data in the frame direction.
7. The imaging apparatus according to claim 4,
   The low-dimensional data creating unit creates a low-dimensional data set by averaging the k-space data in a desired axial direction of k-space.
8. The imaging apparatus according to claim 1,
   The imaging unit includes an X-ray source, an X-ray detector that detects X-rays emitted from the X-ray source, and a rotating unit that rotates the X-ray source and the X-ray detector facing each other. ,
   The imaging apparatus, wherein the data set includes projection data collected at different rotation angles of the rotation unit.
9. The imaging apparatus according to claim 8, wherein
   The data set is a plurality of frame data obtained in time series, and the plurality of frame data includes projection data having different rotation angles from each other.
10. The imaging apparatus according to claim 1, further comprising:
    An artifact extraction unit that extracts an artifact included in an image reconstructed from the data set including the missing data;
    A data correction unit that corrects the data set using the artifact information extracted by the artifact extraction unit;
    The artifact extraction unit calculates an artifact included in the one data set by using one data set and a composite data set of the one data set and the other data set.
    The imaging apparatus, wherein the image estimation unit performs an image estimation process on the data set corrected by the data correction unit.
11. An imaging apparatus comprising: an imaging unit that collects a data set necessary for image reconstruction; and an image processing unit that reconstructs an image using the data set,
    The data set collected by the imaging unit is a data set including missing data,
    The image processing unit uses a plurality of data sets, an image estimation unit that estimates an image under a constraint condition of coincidence between the data set and the image,
    An artifact extraction unit that extracts an artifact included in an image reconstructed from the data set including the missing data;
    A data correction unit that corrects the data set using the artifact information extracted by the artifact extraction unit;
    The artifact extraction unit calculates an artifact included in the one data set by using one data set and a composite data set of the one data set and the other data set.
    The imaging apparatus, wherein the image estimation unit performs an image estimation process on the data set corrected by the data correction unit.
12. The imaging device according to claim 10 or 11,
    The artifact extraction unit creates a plurality of composite sets having different data sets other than the one data set as the composite data set, and uses the one data set and a plurality of composite data sets to generate the one data set. An imaging apparatus characterized by calculating an artifact included in the image.
13. The imaging device according to claim 11 or 12,
    The imaging unit includes a magnetic field generation unit that generates a static magnetic field and a gradient magnetic field, a transmission unit that generates a high-frequency magnetic field, and a reception unit that receives a nuclear magnetic resonance signal generated by the high-frequency magnetic field. An imaging device that is characterized.
14. The imaging device according to claim 11 or 12,
    The imaging unit includes an X-ray source, an X-ray detector that detects X-rays emitted from the X-ray source, and a rotating unit that rotates the X-ray source and the X-ray detector so as to face each other. ,
    An imaging apparatus, which is an X-ray CT apparatus.
15. An image processing apparatus that estimates an image under a constraint condition of coincidence between a data set and an image using a data set including missing data,
    A low-dimensional data creation unit for reducing the dimensions of the data set and the image;
    An image update unit that updates an estimated image estimated under a constraint condition of coincidence between a low-dimensional data set created by the low-dimensional data creation unit and a low-dimensional image;
    An image processing apparatus comprising:
16. An image processing method for estimating an image using a plurality of data sets including missing data under a constraint condition of coincidence between the data set and the image,
    Obtaining a reduced-order data set using the data set;
    An image estimation step of sparsely transforming the data set and estimating an image;
    An image update step of updating the estimated image under a constraint condition of coincidence between the reduced-dimension data set and a low-dimensional image;
    An image processing method comprising: a re-updating step of further updating the updated image updated in the image updating step under a constraint condition of coincidence between the data set and the updated image.
17. The image processing method according to claim 16, comprising:
    An image processing method comprising repeating the image estimation step, the image update step, and the re-update step.
18. An imaging apparatus comprising: an imaging unit that collects a data set necessary for image reconstruction; and an image processing unit that reconstructs an image using the data set,
    The data set collected by the imaging unit is a data set including missing data,
    The image processing unit is an artifact extraction unit that extracts an artifact included in an image reconstructed from a data set including the missing data;
    A data correction unit that corrects the data set using the artifact information extracted by the artifact extraction unit;
    The artifact extraction unit calculates an artifact included in the one data set by using one data set and a composite data set of the one data set and the other data set. apparatus.

Images