JP2023069917A - Magnetic resonance imaging apparatus and image processing method - Google Patents

Abstract

To obtain a synthetic image in which an artifact is effectively reduced with simple processing for an imaging range including a region (spatial region) with different influences of movement in imaging using a multi-channel reception coil.SOLUTION: A magnetic resonance imaging apparatus estimates data of a channel with an artifact using data of a channel without an artifact by applying GRAPPA arithmetic of parallel imaging to synthesize a channel, sequentially applies arithmetic processing using a difference between an image after data estimation and an original image to suppress and remove noise increase following estimation data creation.SELECTED DRAWING: Figure 10

Description

The present invention relates to a technique for acquiring an image with reduced artifacts in magnetic resonance imaging (hereinafter referred to as MRI).

MRI processes nuclear magnetic resonance signals generated from a subject placed in a static magnetic field space by irradiating the subject with a high-frequency magnetic field to generate an image of the subject. Two-dimensional or three-dimensional images can be obtained by encoding the nuclear magnetic resonance signals by applying a gradient magnetic field that gives a gradient to the static magnetic field when acquiring the nuclear magnetic resonance signals. Encoding is typically applied in a phase-encoding direction orthogonal to the slice direction that selects the imaging plane. In addition, the MRI system uses a receiving coil composed of multiple small coils (channels) as a receiving coil for receiving nuclear magnetic resonance signals. A common method is to reconstruct one image by synthesizing the sets).

The target object of MRI is usually a living body such as a person, and moving objects such as respiratory motion, pulsation, and blood flow during imaging affect imaging, and artifacts occur in the phase encoding direction. . In order to reduce such artifacts, a large number of crusher pulses are generally applied. A method of reducing artifacts using information of multi-channel coils has also been proposed (Non-Patent Document 1).

In the technique described in Non-Patent Document 1, based on the degree of matching between measurement data obtained by synthesizing measurement data from a plurality of channels and actually acquired measurement data, a k-space region in which movement is occurring is specified, By appropriately combining the synthesized measurement data and the actually measured measurement data according to the magnitude of movement in this region, the influence of movement is reduced.

``Data Convolution and Combination Operation (COCOA) for Motion Ghost Artifacts Reduction'', Magnetic Resonance in Medicine 64:157-166 (2010)

In the technique described in Non-Patent Document 1, a process is performed to identify a region on the k-space that is affected by motion and to optimize the data in this region. Identification of a region on the k-space is identification of data along the time axis, which is excellent for detecting motion along the time axis, but identifies a spatial region affected by motion within the imaging range. However, it does not address the effects of motion on the motion of a portion of the imaged region, eg, continuous motion such as blood flow.

Furthermore, in the technique described in Non-Patent Document 1, based on the magnitude and standard deviation of the movement, it is determined whether it is a temporary movement or a periodic movement, and processing is performed when combining the synthetic measurement data and the actual measurement data. have to be different.

An object of the present invention is to provide a technique capable of effectively and uniformly reducing artifacts in an imaging range including areas (spatial areas) differently affected by motion.

In order to solve the above problems, the present invention estimates the data of channels with artifacts with the data of channels without artifacts, and synthesizes the channels. Preferably, the difference between the image after data estimation and the original image is used to apply sequential arithmetic processing to suppress an increase in noise that accompanies the generation of estimated data.

That is, the MRI apparatus of the present invention has a receiving coil having a plurality of channels, a measurement unit that receives nuclear magnetic resonance signals generated from an examination object on each channel, and a nuclear magnetic resonance signal received on each channel. an image processing unit that generates an image of an inspection object, the image processing unit includes a reconstruction unit that generates a channel image for each channel, a synthesis unit that generates a synthesized image by synthesizing the channel images, and a channel image. and an artifact remover for removing included artifacts.
The artifact removing unit has a data estimating unit that estimates k-space data of a channel image containing artifacts using k-space data of a channel image that does not contain artifacts, and the reconstructing unit includes: A channel image is created using the k-space data estimated by the data estimation unit for the image containing the artifact.

Further, the image processing method of the present invention is an image processing method for processing nuclear magnetic resonance signals respectively received by receiving coils having a plurality of channels to generate a magnetic resonance image. estimating the k-space data of the artifact-free channel image from the k-space data of the artifact-free channel image to generate an artifact-removed image from the artifact-free channel image; and compositing the channel image without the artifacts and the channel image with the artifacts removed.

An image obtained from a plurality of channels (an image for each channel, hereinafter referred to as a channel image) has a signal value distribution corresponding to its reception sensitivity. If there is an area affected by motion (an area where artifacts are generated) in part of the imaging range, there are channel images with high signal values and channel images with low signal values for that area. In the present invention, artifacts are removed or reduced by estimating the k-space data of a channel image (an image with artifacts) in which an area where artifacts are generated has a high signal from the k-space data of a channel image without artifacts. image can be obtained.

1 is an overall configuration diagram of an MRI apparatus to which the present invention is applied; Functional block diagram of the image processing unit A diagram showing the processing flow of the image processing unit FIG. 4 is a diagram schematically showing processing of the first embodiment; Figures showing examples of channel images, where (A) is a channel image with artifacts, (B) is a channel image without artifacts, and (C) is an image of channel (A) estimated from the channel image of (B). , (D) is a difference image between the image of (A) and the image of (C). FIG. 4 is a diagram for explaining data estimation processing according to the first embodiment; The figure which shows the effect of embodiment FIG. 11 is a diagram for explaining data estimation processing according to Modification 3 of Embodiment 1; Functional block diagram of the image processing unit of the second embodiment FIG. 5 is a diagram for explaining the processing of the second embodiment;

First, the overall configuration of an MRI apparatus to which this embodiment is applied will be described.

As shown in FIG. 1, the MRI apparatus includes a measurement unit 10 that generates nuclear magnetic resonance signals in a subject and measures the generated nuclear magnetic resonance signals, and an operation such as image reconstruction using the nuclear magnetic resonance signals. and a control unit 30 for controlling the operations of the measurement unit 10 and the image processing unit 20 .

The image processing unit 20 and the control unit 30 can be built on a computer 50 or a work station equipped with a CPU and memory, although not limited thereto, and the computer can upload programs that implement their respective functions. , the functions are executed. However, it is also possible to realize part of the calculations performed by the image processing unit 20 by hardware such as ASIC and FPGA. The image processing unit and the control unit (computer 50), like a normal computer, include a UI unit 40 having an output device such as an input device for user input and a display device, calculation results, data necessary for calculation, etc. An external storage device 60 or the like can be provided for storing the .

The configuration of the measurement unit 10 is the same as that of a general MRI apparatus, and as shown in FIG. A transmission unit 12 that irradiates a high-frequency magnetic field that causes nuclear magnetic resonance to the atomic nuclei (mainly protons) of contained elements, a reception unit 13 that receives nuclear magnetic resonance signals generated from the subject, and a magnetic field gradient in the static magnetic field. and a sequencer 15 that operates the transmitter 12, receiver 13, and gradient magnetic field generator 14 according to a predetermined pulse sequence.

The transmission unit 12 includes a high-frequency oscillator 121 that generates a high-frequency signal, a modulator 122, a high-frequency amplifier, and a transmission RF coil 123 that generates a high-frequency magnetic field by being supplied with the amplified high-frequency signal.

The receiving unit 13 includes a receiving coil 131 arranged near the subject, a quadrature detector 132, an A/D converter 133, and the like, and converts the nuclear magnetic resonance signal received by the receiving coil 131 into two systems of digital signals. It is converted and transferred to the image processing unit 20 . The receiving coil 131 is a multi-channel coil that combines a plurality of small coils. The receiving unit 13 receives a nuclear magnetic resonance signal for each small coil and uses it as respective measurement data.

The gradient magnetic field generator 14 includes three sets of gradient magnetic field coils 141 that generate gradient magnetic fields in mutually orthogonal three-axis (X, Y, Z) directions, and a gradient magnetic field power supply 142 that drives them. When a nuclear magnetic resonance signal is generated, a section (slice) in an arbitrary direction for generating a nuclear magnetic resonance signal is selected by appropriately combining and applying gradient magnetic field pulses generated by three sets of gradient magnetic field coils. and can provide (ie encode) position information in any direction. Generally, the direction of the gradient magnetic field for frequency encoding one nuclear magnetic resonance signal is called the readout direction, and the direction of the gradient magnetic field for encoding with different gradient magnetic field strengths for each nuclear magnetic resonance signal is called the phase encoding direction.

The sequencer 15 calculates a pulse sequence used for imaging using the set pulse sequence and imaging parameters under the control of the control unit 30, and according to it, the transmitting unit 12, the receiving unit 13, and the gradient magnetic field generating unit described above. 14 is operated repeatedly. There are various pulse sequences that differ according to imaging methods and imaging purposes, and are stored in advance in a storage device within the MRI apparatus or an external storage device. When the user selects a desired pulse sequence or selects a pulse sequence preset as an inspection protocol, it is set in the sequencer 15 via the control section 30 .

Nuclear magnetic resonance signals (digital signals) collected by repeated operations according to the pulse sequence are passed to the image processing unit 20 as measurement data for each channel. The measurement data is, for example, two-dimensional or three-dimensional space data (k-space data) with one axis as the readout direction and the other one or two axes as the phase encoding direction.

The image processing unit 20 reconstructs the k-space data for each channel acquired by the measurement unit 10 and creates an image of the subject. For this reason, as shown in FIG. 2, the image processing unit 20 includes a reconstruction unit 21 that reconstructs an image that is real space data by performing computation such as inverse Fourier on the k-space data, and synthesizes the images of each channel. A synthesizing unit 23 is provided. The MRI apparatus 1 of this embodiment further includes an artifact removal unit 25 that removes artifacts that may occur in the reconstructed image after synthesis. The artifact removing unit 25 performs estimation using the measurement data of the image without the artifact for the channel image with the artifact among the images of each channel, creates the estimated measurement data, and performs processing to make the image without the artifact. .

FIG. 3 shows an overview of the processing of the image processing unit 20. As shown in FIG. The reconstruction unit 21 converts the measurement data of each channel into image data (channel image) (S1). The artifact removal unit 25 divides channel images into images with artifacts and images without artifacts (S2), and performs estimation processing using the measurement data of images without artifacts for the measurement data of the channel images with artifacts. (S3). The presence or absence of artifacts in each channel image may be determined by the user by presenting the channel image generated by the reconstruction unit 21 to the user and may be accepted by the user, or may be determined within the image processing unit 20. It is also possible to provide means. Details will be described later.

For the estimation of the measurement data, calculation using the GRAPPA calculation used in parallel imaging (PI), in which k-space data is thinned out at a predetermined rate of magnification (R-factor) and measured, can be used. GRAPPA calculation calculates the weight (kernel) of the measurement points around one measurement point from the k-space center data (measurement data of a predetermined area measured for auto-calibration: ACS), and uses this weight to calculate a plurality of This is a method of estimating the data of unmeasured points by performing a convolution operation on the measured data of the peripheral measurement points of the channel. In this embodiment, the object of estimation is not unmeasured data but measurement data of an image with artifacts, and the GRAPPA method is applied to this to perform estimation from the measurement data of a plurality of channels.

The reconstruction unit 21 uses the estimated measurement data to generate a channel image from which artifacts have been removed (S4). Since the artifact-removed channel image was reconstructed from measurement data estimated using multiple measurement data, it may have increased noise compared to the original image with artifacts. In this case, processing for removing the increased noise may be performed (S5). However, this step S5 is not essential. Finally, the synthesizing unit 23 synthesizes the artifact-removed channel image and another channel image (originally no artifact-free channel image) to generate a synthesized image (S6).

A specific embodiment of the processing of the image processing unit 20 will be described below.

<Embodiment 1>
In this embodiment, the image processing unit 20 receives the result of the user's determination of the presence or absence of artifacts in the images reconstructed from the measurement data of each channel, and receives the result of the user determining the presence or absence of artifacts. ) are estimated from the measurement data of channel images without artifacts (abbreviated as AF-free Ch images).

As shown in FIG. 2, the configuration of the image processing unit 20 of the present embodiment includes a reconstruction unit 21, a synthesis unit 23, and an artifact removal unit 25. The artifact removal unit 25 includes a user and a data estimating unit 253 for estimating the measurement data of the AF-enabled Ch image.

Details of the processing will be described below with reference to FIG. The flow of processing is the same as the flow shown in FIG. 3, and reference is made to FIG. 3 as appropriate.

First, the channel image (S1) generated by the reconstruction unit 21 is displayed on the display device of the UI unit 40, and the reception unit 251 receives the user's determination of the presence or absence of artifacts. Examples of images presented to the user are shown in FIGS. As shown, the channel images have different areas of high signal strength due to the difference in reception sensitivity of the channels. For example, the image on the left (A) shows artifacts due to blood flow in the lower region of the image of the subject, but the channel images (three images on the right) with low sensitivity for this region (B). At , no artifacts appear due to the low signal strength. In this example, there is one image with artifacts, but there may be multiple images. That is, the receiving unit 251 may receive a plurality of channel images as images with artifacts.

When the user designates a Ch image with AF (S2), the data estimation unit 253 performs estimation processing using measurement data of other channels for the designated channel image (S3). Specifically, as with the GRAPPA calculation, first, for the central region of the k-space data, each of the data that minimizes the difference between the data S ^est (ch1) estimated from the peripheral data and the original measurement data S ^measured (ch1). A weighting factor (kernel) W of the peripheral measurement points is determined by the least squares method from the following equation (1).

(Number 1)
S ^est (ch1)=W・S ^mesured (ch) (1)
The weight W is a value that uses the measurement line, channel, and measurement line to be estimated as coefficients (indexes).

FIG. 6 schematically shows the relationship between the measurement point to be estimated and the surrounding measurement points. In the figure, the area surrounded by a square indicates the measurement point and the range of the surrounding measurement points. In the GRAPPA calculation used in the PI method, the kernel size representing the range of the peripheral measurement points is usually 2×3 or 2×5 because the adjacent measurement points of two measurement lines sandwiching the unmeasured line are used. In the embodiment, 3×3 is used because there are no unmeasured points. However, the kernel size is not limited to that shown in the figure, and can be set arbitrarily.

6 shows a case where the number of channels is 3, the image of channel 1 (Ch1) is a Ch image with AF, and the images of channels Ch2 and Ch3 are Ch images without AF, in order to simplify the explanation. However, the number of channels may be greater than three.

Next, using this weighting factor W and the artifact-free measurement data of channels Ch2 and Ch3, the convolution operation shown in the following equation (2) is performed to estimate the measurement point data of channel Ch1.

式中、Ｓｊ（ｋｙ）はチャンネルｊにおけるｋ空間上の計測ライン、ｂは計測ラインのインデックスで図６の例では-1,0,+1のいずれか、ｌはチャンネルのインデックスを表す。

In the formula, Sj(ky) is a measurement line on the k space in channel j, b is the index of the measurement line, which is -1, 0 or +1 in the example of FIG. 6, and l is the index of the channel.

By performing the above data estimation processing (S4) for all the measurement points of the measurement data of the channel with artifacts, the estimation data of channel Ch1 is generated as shown in FIG.

After that, the synthesizing unit 23 synthesizes the channel image reconstructed from the estimated data and the artifact-free channel image to generate one image (composite image) (S6). Synthesis can be performed in the same manner as for general multi-channel image synthesis (MAC synthesis). As shown in Fig. 5, the image (C) reconstructed from the estimated Ch measurement data has fewer artifacts than the original image (A), and the artifacts are removed from the difference image (D). .
Further, as shown in FIG. 7, it can be seen that artifacts are also suppressed in the synthesized image by performing the processing of the present embodiment.

According to the present embodiment, the GRAPPA operation is applied to estimate the measurement data for the AF-affected Ch image, and after the artifact-free Ch image is synthesized with another AF-free Ch image, artifacts are eliminated. A suppressed image can be obtained. In addition, since it is premised on the determination of the presence or absence of artifacts in the image, there is no need to temporally identify the movement that causes b artifacts as in the technique of Non-Patent Document 1, and artifacts that occur locally in space can be effectively eliminated. can be suppressed to

<Modification 1 of Embodiment 1>
In the first embodiment, the user determines the presence or absence of artifacts in the channel image, and the reception unit 251 receives the determination result. can also be provided. In that case, the reception unit 251 in FIG. 2 is replaced with an artifact determination unit (not shown).

As a method for the artifact determination unit to determine the presence or absence of artifacts, for example, a method of determining from an image profile (method 1), a method of using a deep learning model such as CNN, etc. can be considered, and any of them may be adopted. .

In the case of Method 1, flow artifacts appear not only inside the image where the object exists, but also around it. Separately, in an image with artifacts, there is a signal due to the artifacts around the region of interest. Signals caused by artifacts can be detected by using a mask or the like for the subject region, and the presence or absence of artifacts can be determined based on this.

According to this modification, the process from measurement to image synthesis can be performed completely automatically. Even when this modification is adopted, in order to confirm the validity of the results, intermediate processing such as artifact determination results and final processing results are presented to the user, and user changes are accepted as necessary. can be

<Modification 2 of Embodiment 1>
In the first embodiment, as the data estimation process, the case of estimating the measurement data of Ch images with artifacts using the measurement data of Ch images without artifacts has been described. As the data, estimated Ch measurement data may also be used. In this case, for example, when calculating the kernel in the range indicated by the square in the measurement data of Ch1 in FIG. The convolution operation shown in Equation 2 is performed by applying the weights to the measurement data themselves.

According to this modification, the estimated Ch data can include the information of the measured data of the Ch image with the artifact, so that the estimated Ch data can be prevented from greatly deviating from the measured data of the channel to be estimated. can be done.

<Modification 3 of Embodiment 1>
In the first embodiment, the case where the measurement data of each channel is full-sampling data has been described as an example, but the same processing can be performed on the measurement data under-sampled at a predetermined speed ratio.

For example, as shown in FIG. 8, in parallel imaging, there are 1 to (R-1) unmeasured lines between the measurement lines (R is the speed ratio). FIG. 8 shows an example of double-speed measurement in which every other line is measured, and unmeasured lines are indicated by white circles. In this case, ignoring unmeasured lines and focusing only on measured lines, data estimation is performed using data of other AF-free Ch images in the same manner as in FIG. are partially omitted, but estimation is performed using data within the kernel size in the same manner as in FIG. 6). This yields undersampled estimated Ch data for the Ch image with AF. As for the data estimation (S3) in this case as well, the measurement data of the measurement data itself to be estimated may be used for the estimation as in the second modification.

After that, the synthesizing unit 23 creates a synthesized image using PI calculation. In the PI calculation, in addition to the calculation (such as the SENSE method) that unfolds aliasing that occurs in the image using the sensitivity distribution of each receiving coil in the image space, and the calculation that performs reconstruction by estimating the data in the measurement space (such as GRAPPA), Various methods such as the SPIRiT method and the CAIPIRINHA method have been proposed, and any of them may be adopted.

As a further modification, instead of estimating data using the undersampled k-space data as described above, ACS is first used to estimate measurement data using the normal PI method GRAPPA, etc. After making full sampling measurement data, the presence or absence of artifacts in each Ch image is determined, and data estimation (S3), estimated Ch image reconstruction (S4), and image synthesis (S5) are performed in the same manner as in the first embodiment. may

According to this modified example, by combining the PI method, it is possible to increase the measurement speed and obtain an image with reduced artifacts.

<Embodiment 2>
This embodiment is characterized by performing noise removal processing after data estimation in addition to the processing of the first embodiment described above. FIG. 8 shows a configuration example of the image processing unit 20 of this embodiment. In this example, a sequential calculation unit 27 is added as means for removing noise. The functions and processing contents of other elements are the same as those of the first embodiment and its modifications, and the points different from the first embodiment will be mainly described below.

FIG. 10 shows an overview of the processing of this embodiment. As shown in the figure, in the present embodiment as well, measurement data is estimated for channel images with artifacts using measurement data of channel images without artifacts (measurement data of channel images with artifacts may be included). , to generate an estimated Ch image.

Since the Ch image estimated in this way is obtained by convolution of a plurality of measurement data, there is a possibility that the noise is amplified compared to the original Ch image with AF. Therefore, first, the difference between the estimated Ch image and the original AF-existing Ch image is calculated to create a difference image composed of artifacts and noise. Sequential calculations such as compression sensing are performed so that the difference image becomes an image of only artifacts (a noise-removed difference image). At this time, an area without artifacts is restricted so that it does not largely deviate from the original image with AF.

式（３）において、I_denoize、I_wArtifact、I_estは、それぞれノイズ除去後のCh画像、元のAF有Ch画像、データ推定後の推定Ch画像を表し、右辺の第一項は上述した制約を表している。λは項の重みを制御するための係数である。 This sequential calculation can be represented, for example, by the following equation (3).

In Equation (3), I _denoize , I _wArtifact , and I _est represent the Ch image after noise removal, the original Ch image with AF, and the estimated Ch image after data estimation, respectively, and the first term on the right side is the constraint described above. represents. λ is a coefficient for controlling the weight of the term.

By performing such sequential calculations, the noise level of the estimated Ch image is finally reduced to the noise level of the original AF-affected Ch image, and information other than artifacts included in the original AF-affected Ch image A Ch image (I _denoize ) in which is sufficiently reflected can be obtained.

After that, using the obtained Ch image and Ch images of other channels to generate a composite image is the same as in the first embodiment.

According to this embodiment, it is possible to suppress an increase in noise associated with data estimation and obtain an image with better image quality.

1: MRI apparatus, 10: measurement unit, 20: image processing unit, 21: reconstruction unit, 23: synthesis unit, 25: artifact removal unit, 251: reception unit (or artifact determination unit), 253: data estimation unit, 27: sequential calculation unit, 30: control unit, 40: UI unit, 50: computer, 60: storage device.

Claims (13)

A measuring unit that has a receiving coil with a plurality of channels and receives nuclear magnetic resonance signals generated from an object to be inspected in each channel, and generates an image of the object to be inspected using the nuclear magnetic resonance signals received in each channel. and an image processing unit,
The image processing unit includes a reconstruction unit that generates a channel image for each channel, a synthesizing unit that generates a synthesized image by synthesizing the channel images, and an artifact removal unit that removes artifacts included in the channel images. ,
The artifact removing unit includes a data estimating unit that estimates k-space data of a channel image containing artifacts using k-space data of a channel image that does not contain artifacts, and the reconstructing unit. A magnetic resonance imaging apparatus according to claim 1, wherein a channel image is created using the k-space data estimated by the data estimation unit for the image containing the artifact. The magnetic resonance imaging apparatus according to claim 1,
A magnetic resonance imaging apparatus according to claim 1, wherein the data estimator performs data estimation by convoluting data points corresponding to data points to be estimated in k-space data used for estimation using a predetermined kernel. The magnetic resonance imaging apparatus according to claim 1,
A magnetic resonance imaging apparatus, wherein the data estimating unit uses the k-space data of the channel image containing the artifact in addition to the k-space data of the channel image not containing the artifact as k-space data used for estimation. . The magnetic resonance imaging apparatus according to claim 1,
The magnetic resonance imaging apparatus, wherein the image processing unit further includes a receiving unit that receives designation of an image including artifacts or an image not including artifact images among the channel images. The magnetic resonance imaging apparatus according to claim 1,
A magnetic resonance imaging apparatus, wherein the image processing unit further includes an artifact determination unit that determines images containing artifacts and images not containing artifact images. The magnetic resonance imaging apparatus according to claim 1,
A magnetic resonance imaging apparatus, wherein the k-space data of each channel is undersampled data, and the synthesizing unit generates a synthesized image of each channel by parallel imaging calculation. The magnetic resonance imaging apparatus according to claim 1,
A magnetic resonance imaging apparatus, wherein the image processing unit further comprises a sequential operation unit that performs noise reduction processing using the k-space data estimated by the data estimation unit. The magnetic resonance imaging apparatus according to claim 7,
A magnetic resonance imaging apparatus, wherein the processing performed by the sequential calculation unit includes a constraint condition that an image reconstructed from k-space data after data estimation does not greatly deviate from an original image before data estimation. The magnetic resonance imaging apparatus according to claim 7,
Magnetic resonance imaging, wherein the processing performed by the sequential calculation unit includes processing for making the noise level of an image reconstructed from k-space data after data estimation equal to the noise level of the original image before data estimation. Device. The magnetic resonance imaging apparatus according to claim 1,
A magnetic resonance imaging apparatus, wherein the artifact is a flow artifact. An image processing method for generating a magnetic resonance image by processing nuclear magnetic resonance signals respectively received by receiving coils having a plurality of channels,
Among channel images obtained from nuclear magnetic resonance signals of each channel, k-space data of channel images with artifacts are estimated from k-space data of channel images without artifacts, and artifacts are removed from the channel images with artifacts. generating an image of the
and combining the artifact-free channel image and the artifact-removed channel image. The image processing method according to claim 11,
The step of generating the artifact-removed image includes using a kernel determined by a relationship between k-space central region data of the artifact-free channel image and k-space central region data of the artifact-free channel image to remove the artifact. estimating the channel image with the artifact by convolving the k-space data of the channel image without the artifact. The image processing method according to claim 11,
1. An image processing method comprising, after the step of removing artifacts, performing sequential processing for making the noise level of a channel image after removing the artifacts equal to the noise level of the original image of the channel image.

Images