WO2013109095A1

WO2013109095A1 - Magnetic resonance image processing method

Info

Publication number: WO2013109095A1
Application number: PCT/KR2013/000420
Authority: WO
Inventors: 박재석; 최상천
Original assignee: 고려대학교 산학협력단
Priority date: 2012-01-19
Filing date: 2013-01-18
Publication date: 2013-07-25
Also published as: KR101358427B1; KR20130085386A

Abstract

The magnetic resonance image processing method according to the present invention comprises the steps of: receiving image data generated following the application of a turbo spin echo sequence; removing noise in the slice direction and noise in the coil region from the image data; and reconstructing a distortion profile on the basis of the image data from which the noise has been removed. In the noise removing step, the noise in the slice direction is removed via the steps of: converting the image data into square matrix format; removing 0 value in 0 value and the signal value comprised in the square matrix, and generating a substitution matrix in which the remaining signal value is allocated to the central or a side region of the square matrix; and removing noise by using low-coefficient approximation via eigenvalue decomposition with respect to the substitution matrix.

Description

Magnetic Resonance Image Processing Method

The present invention relates to a magnetic resonance image processing method.

Magnetic resonance image processing apparatus (MRI) is a device for obtaining a tomographic image of a specific part of a patient by using a resonance phenomenon according to the supply of electromagnetic energy. Such magnetic resonance image processing apparatuses are widely used because they do not cause radiation exposure problems caused by X-ray imaging apparatuses or computed tomography (CT) devices, and tomographic images are relatively easily obtained. However, since the magnetic field is used, there is a problem in that artifacts are generated by metal materials (for example, metal for implants, iron cores for fracture treatment, etc.) attached to the patient's body.

As a method for removing such artifacts, an image processing method called slice encoding for Metal Artifact Correction (SEMAC) has recently been introduced.

1 is a view for explaining a conventional SEMAC image processing method.

The SEMAC image processing method is based on a view angle tilting (VAT) spin echo sequence, and a Z-phase encoding step is applied to each excited slice to correct metal artifacts. At this time, the Z-phase encoding step decomposes the distorted profile. That is, after acquiring the added three-dimensional data by extending the distorted profile due to the metal in the slice direction, the same method is applied to each slice, and the images are corrected by combining the images. For example, the main configuration is to correct metal artifacts through a sequence as shown in FIG. 1. In general, the distortion caused by the metal can be classified into inplane distortion and through distortion, and the inplane distortion can be suppressed by the VAT spin echo sequence. However, through distortion must be corrected in a state in which a plurality of data are combined, and the image reconstructed using the SEMAC sequence has a problem in that the signal-to-noise ratio (SNR) is not good.

On the other hand, related to the technology of the present invention, Japanese Patent Laid-Open Publication No. 2009-519086 (electronic tracking method and apparatus for metal artifact compensation using a modular array of reference sensors) has a metal artifact using a modular array of reference sensors. Disclosed are a method and apparatus of electronic tracking for compensation.

SUMMARY OF THE INVENTION The present invention has been made to solve the above-mentioned problems of the prior art, and an object of some embodiments of the present invention is to provide a magnetic resonance image processing method of improving the metal artifact compensation performance.

As a technical means for achieving the above-described technical problem, the magnetic resonance image processing method according to the present invention, receiving the image data generated by the application of the turbo spin echo sequence; Removing noise in a slice direction and noise in a coil region with respect to the image data; And restoring a distortion profile based on the image data from which the noise has been removed, wherein removing the noise comprises: converting the image data into a square matrix form; Removing a zero value among the signal values and zero values included in the square matrix, and placing the remaining signal values in the center or one region of the square matrix to generate a substitution matrix and eigenvalue decomposition for the substitution matrix. Noise is removed in the slice direction by applying a low coefficient approximation method.

According to the above-described problem solving means of the present invention, by effectively modeling the noise in the slice direction and the noise in the coil region, by obtaining the optimal level of eigenvalues through decomposition of the eigenvalues in each slice direction, the noise that appears during image reconstruction The effect of can be reduced as much as possible. Also, in the coil region, noise of each coil may be removed by using a coil information map and a noise correlation matrix. Through this modeling, it is possible to reconstruct an image having an improved signal-to-noise ratio compared to a conventional method using a SEMAC sequence.

1 is a view for explaining a conventional SEMAC image processing method.

2 is a flowchart illustrating a magnetic resonance image processing method according to an embodiment of the present invention.

FIG. 3 is a diagram for describing a TSEMAC sequence used in a magnetic resonance image processing method according to an embodiment of the present invention.

4 is a view for explaining a magnetic resonance image processing method according to an embodiment of the present invention.

5 is a view for explaining a low coefficient approximation method of a magnetic resonance image processing method according to an embodiment of the present invention.

6 is a view for explaining a low coefficient approximation method of a magnetic resonance image processing method according to another embodiment of the present invention.

7 is a flowchart illustrating a magnetic resonance image processing method according to another embodiment of the present invention.

DETAILED DESCRIPTION Hereinafter, exemplary embodiments of the present invention will be described in detail with reference to the accompanying drawings so that those skilled in the art may easily implement the present invention. As those skilled in the art would realize, the described embodiments may be modified in various different ways, all without departing from the spirit or scope of the present invention. In the drawings, parts irrelevant to the description are omitted in order to clearly describe the present invention, and like reference numerals designate like parts throughout the specification.

Throughout the specification, when a part is "connected" to another part, this includes not only "directly connected" but also "electrically connected" with another element in between. . In addition, when a part is said to "include" a certain component, which means that it may further include other components, except to exclude other components unless otherwise stated.

First, image data is received from a magnetic resonance image processing apparatus using a combination of a turbo spin echo sequence and a SEMAC sequence (S210).

The turbo spin echo sequence refers to a sequence in which a plurality of refocus pulses are continuously applied after the application of the RF excitation pulse. With reference to the drawings will be described in detail.

TSEMAC sequence is to improve the time efficiency, and to solve the problem of the conventional image processing method using the SEMAC sequence. That is, since the SEMAC sequence performs an additional Z-phase encoding step, there is a problem that the scan time increases, and it is necessary to improve time efficiency.

After the application of the RF excitation pulse 310, the

refocus pulses

312 and 314 are sequentially applied a plurality of times to form a plurality of spin echo intervals. Since each spin echo period is spatially encoded, a plurality of K spatial lines may be sampled after the excitation pulse is applied. Accordingly, the time required for the entire image processing can be reduced.

Meanwhile, a Z-phase encoding step is performed during the spin echo period, thereby distorting the distorted profile. K-space signal measured during read out and z-phase encoding

Is defined according to the following equation.

[Equation 1]

At this time,

,

Will have the same relationship as

Denotes the self-rotation ratio, Gx denotes the size of the read interval, Gs denotes the size of the VAT compensation gradient, tx denotes the time spent in the read interval, and ts denotes the time spent in the VAT compensation interval.

Referring back to FIG. 2, an operation of removing noise in a slice direction with respect to the received image data is performed (S220). At this time, the matrix structure is transformed to reflect the characteristics of the temporal and spatial data well, and then low rank approximation is used through eigen decomposition.

With reference to the drawings will be described in more detail.

4 is a view for explaining a magnetic resonance image processing method according to an embodiment of the present invention, Figure 5 is a view for explaining a low coefficient approximation method of a magnetic resonance image processing method according to an embodiment of the present invention, 6 is a view for explaining a low coefficient approximation method of a magnetic resonance image processing method according to another embodiment of the present invention.

First, as shown in the left side of FIG. 4, it is assumed that distortion of an image occurs in the ROI. Since the shape of the data is irregular due to the distortion of the image, the window size of the resolved data is set to include a plurality of slices. Thereafter, each of the decomposition data is superimposed to restore an image signal, and each decomposition data includes a noise signal in addition to the image signal. As such decomposition data overlaps, a problem may occur in which a noise signal is also overlapped. Accordingly, an operation of processing the noise signal in advance on each piece of decomposition data is performed.

To this end, as shown in FIG. 5, noise is removed using a low coefficient approximation method.

The signal yi measured by TSEMAC corresponds to Cartesian sampling of k-t space and includes noise Wi, and can be defined in the form of Equation 2 below.

[Equation 2]

Next, the temporal profile of the same voxel position in the decomposition data may be extended to an n-dimensional vector as shown in Equation 3 below, and has a strong correlation in the plurality of decomposition data.

[Equation 3]

In this case, l (x, t) represents a signal, x represents a spatial position, and t represents time. In addition, Xj (j = 0,1,2, ..., m-1) represents the j-th voxel at the same position.

In order to use the correlation in the decomposition data, the signal coordinates l (x, t) are rearranged in a matrix form as shown in Equation 4 below.

[Equation 4]

The number of rows of the L matrix corresponds to the number of slices of a TSEMAC 3D image, and the number of rows of the L matrix also represents the same time profile as the number of slices. Since the L matrix includes zero values, the L matrix may have a diagonal matrix form as shown in Equation 4. For example, as illustrated in (a) of FIG. 5, a diagonal matrix having an n * n form may be provided. As described above, a process of converting image data into a square matrix form is performed.

Next, a zero matrix value is removed from the signal value and the zero value included in the L matrix, and a substitution matrix is generated in which the remaining signal values are arranged at the center or one side of the square matrix.

For example, as shown in (c) of FIG. 5, the L matrix is permutated such that the

values

510 and 514 located at the side edges of the L matrix are located at the center portion 512 of the square matrix. . According to the generation of such a substitution matrix, the number of rows of the substitution matrix is equal to the number of slices n, but the number of columns of the substitution matrix is equal to the number e of performing the Z-phase encoding step.

Meanwhile, according to another embodiment, as shown in FIG. 6, zero values of signal values and zero values included in the L matrix are removed, and the remaining signal values 600 are moved to the left side of the square matrix. You can create a replacement matrix.

Therefore, the substitution matrix Lp has a stronger correlation than the matrix L before substitution.

Next, Fourier transform the substitution matrix along t. In this case, according to the Fourier transform result shown in (c) of FIG. 5, it can be seen that the substitution matrix is very strongly related.

Next, as shown in (d) of FIG. 5, since there are not many singular values, it can be seen that the decomposition data can be approximated by a low rank matrix. Therefore, the rank of the substitution matrix Lp may be reduced as shown in the following equation.

[Equation 5]

r <n * e

Meanwhile, in performing singular value decomposition (SVD) on the substitution matrix Lp in each coil image, the substitution matrix Lp having the low coefficient r may be decomposed as follows.

[Equation 6]

Where U and V are n * r, e * r matrices, and singular values (

) Is positive.

Next, a method of restoring the low coefficient matrix L will be described.

The low coefficient matrix L can be obtained by solving the equation (7).

[Equation 7]

On the other hand, since the rank penalty is nonconvex, it must be transformed into a convex function such as nucleus norm, which is expressed by Equation 8.

[Equation 8]

On the other hand, the nucleus norm of the L matrix can be expressed by Equation (9).

[Equation 9]

By solving the cost function defined by Equation 8, the optimal coefficient approximation matrix can be expressed as the following equation.

[Equation 10]

In this way, noise can be removed by substituting the matrix structure and using a low coefficient approximation method through eigenvalue decomposition for the substitution matrix.

Referring back to Figure 2, after removing the noise in the slice direction, the noise in the coil region is removed (S230). To this end, a noise correlation matrix between coils is obtained and a best linear unbiased estimator (BLUE) is used.

Since noise obtained through different phase-array channels is partially related, it is appropriate to use the best linear unbiased estimation method. If the highest linear discomfort estimate includes a noise covariance and a coil information map, the signal-to-noise ratio of the reconstructed image using the same can be improved.

The equation representing the highest linear discomfort estimate is shown in Equation 11.

[Equation 11]

In this case, C represents the sensitivity matrix of the K-th coil,

Matrix is

Represents a noise covariance matrix. In Equation 11,

The matrix indicates the level and correlation of noise from the receive channel to optimize the signal to noise ratio. Therefore, it is possible to remove the noise in the coil region through this.

remind

The matrix represents an image with improved signal-to-noise ratio and can suppress not only noise in the slice direction, but also noise in different phase-array channels.

Next, the distorted profile is restored for the data of the image space from which the noise is removed (S240). As the noise in the slice direction and the noise in the phase-array channel direction in the previous step are removed, the remaining data has a sparse state or sparsity. For the data in this state, the L1 minimization norm is obtained by solving the following equation (12).

[Equation 12]

There may be various algorithms for solving the equation (12). For example, a Basic Pursuit (BP) method or an Orthogonal Matching Pursuit (OPM) method may be used.

For example, the orthogonal matching seeking method reconstructs the distorted profile in the image-frequency (x-f) region, and has an advantage of high speed and relatively easy implementation. The algorithm creates a signal model starting from an empty model, with the most important new atom added at every step, which is performed repeatedly. This repetition continues until the maximum residual in the x-f region reaches the noise level. The final image can be obtained through inverse Fourier transform of the predicted x-f image.

In the meantime, the magnetic resonance image processing method described above is described based on the step of removing the noise in the coil region after removing the noise in the slice direction. However, the order of the steps may be changed. That is, according to another embodiment of the present invention, unlike the embodiment of FIG. 2, the magnetic resonance image may be processed by performing the noise removing step in the slice direction after performing the noise removing step in the coil region.

In detail, FIG. 7 is a flowchart illustrating a method of processing a magnetic resonance image, according to another exemplary embodiment.

As shown in FIG. 7, first, image data is received from a magnetic resonance image processing apparatus using a combination of a turbo spin echo sequence and a SEMAC sequence (S710).

In operation S720, noise in the coil region is removed from the received image data.

Then, the noise in the slice direction is removed (S730).

Next, the distorted profile is restored with respect to the data of the image space from which the noise is removed (S740).

As described above, each step of FIG. 7 may apply the noise canceling method in the coil region and the noise canceling method in the slice direction described above with reference to FIGS. 2 to 6.

In this way, the order of the noise removal in the slice direction and the removal in the coil area can be changed by the operator. In addition, in some cases, the order of noise cancellation in the slice direction and noise removal in the coil region may be simultaneously performed in parallel.

One embodiment of the present invention can also be implemented in the form of a recording medium containing instructions executable by a computer, such as a program module executed by the computer. Computer readable media can be any available media that can be accessed by a computer and includes both volatile and nonvolatile media, removable and non-removable media. In addition, computer readable media may include both computer storage media and communication media. Computer storage media includes both volatile and nonvolatile, removable and non-removable media implemented in any method or technology for storage of information such as computer readable instructions, data structures, program modules or other data. Communication media typically includes computer readable instructions, data structures, program modules, or other data in a modulated data signal such as a carrier wave, or other transmission mechanism, and includes any information delivery media.

Although the methods and systems of the present invention have been described in connection with specific embodiments, some or all of their components or operations may be implemented using a computer system having a general purpose hardware architecture.

The foregoing description of the present invention is intended for illustration, and it will be understood by those skilled in the art that the present invention may be easily modified in other specific forms without changing the technical spirit or essential features of the present invention. will be. Therefore, it should be understood that the embodiments described above are exemplary in all respects and not restrictive. For example, each component described as a single type may be implemented in a distributed manner, and similarly, components described as distributed may be implemented in a combined form.

The scope of the present invention is shown by the following claims rather than the above description, and all changes or modifications derived from the meaning and scope of the claims and their equivalents should be construed as being included in the scope of the present invention. do.

Claims

In the magnetic resonance image processing method,

Receiving image data generated according to application of a turbo spin echo sequence;

Removing noise in a slice direction and noise in a coil region with respect to the image data; And

Restoring a distortion profile based on the noise-free image data;

Removing the noise,

Converting the image data into a square matrix;

Generating a substitution matrix by removing a zero value among signal values and zero values included in the square matrix and disposing the remaining signal values into a center or one region of the square matrix; And

And removing noise by applying a low coefficient approximation method through eigenvalue decomposition to the permutation matrix.
The method of claim 1,

Receiving the image data,

After applying the RF excitation pulse, continuously applying a refocus pulse a plurality of times; and

And performing a Z-phase encoding and reading operation during a spin echo period after the application of the refocus pulse.
The method of claim 1,

And the number of rows and columns of the square matrix generated according to the step of converting the image data into a square matrix form is equal to the number of slices of the image data.
The method of claim 1,

The number of rows of the substitution matrix generated by generating the substitution matrix is equal to the number of slices of the image data, and the number of columns of the substitution matrix is the same as the number of times Z-phase encoding is performed. Treatment method.
The method of claim 1,

The generating of the substitution matrix is a magnetic resonance image processing method for arranging a signal value disposed in the side edge region of the square matrix to the center of the square matrix.
The method of claim 1,

The generating of the substitution matrix may include arranging the remaining signal values to the left side of the square matrix.
The method of claim 1,

Removing the noise,

A magnetic resonance image processing method for removing noise in the coil region by applying a best linear unbiased estimator to the noise correlation matrix calculated for the image data.
The method of claim 1,

Restoring the distortion profile,

Magnetic resonance image processing method for restoring the distortion profile by removing the distortion signal through an orthogonal matching pursuit.
A computer-readable recording medium having recorded thereon a program for performing the method according to any one of claims 1 to 8.