JPH05154131A - Magnetic resonance diagnostic device - Google Patents

Abstract

PURPOSE:To facilitate the selection of a processing boxel by obtaining a three- dimensional <1>H image and dividing it into boxels of prescribed size, deriving a pixel value in each boxel, and selecting a boxel as a spectrum processing boxel when the number of contained pixels having a larger pixel value than a prescribed value is more than a prescribed number. CONSTITUTION:From a desired area of a body to be examined, three- dimensional data of a <1>H image is obtained through a data collecting part 9. Subsequently, other processing is executed by a computer system 10. That is, first of all, the obtained <1>H image data is divided by a lattice corresponding to a boxel obtained by chemical shift imaging. Next, a threshold of a pixel existing in a each boxel is derived, and whether or not the number of pixels above the threshold is a prescribed number or above in each boxel is decided. As a result, the boxel deciced to have the prescribed number of pixels or above is set as a spectrum processing boxel and its boxel position is stored. Further, a spectrum processing is executed only to the selected boxel.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a magnetic resonance diagnostic apparatus which obtains anatomical information and biochemical information of a subject by utilizing a magnetic resonance phenomenon.

[0002]

2. Description of the Related Art In recent years, with the development of a magnetic resonance diagnostic apparatus, non-invasive observation and observation of minute metabolites in a living body are possible.
MRSI (Magnetic Resonance) that can be imaged
Spectroscopic Imaging) is in practical use.

MRSI is for imaging the abundance of in vivo metabolites by chemical shift imaging, and is effective for early diagnosis of diseases, determination of therapeutic effect on cancer, and the like. The procedure up to imaging is as shown in the flowchart in FIG. 17. First, after collecting NMR signals by chemical shift imaging (step ST1), processing such as FFT (fast Fourier transform) is performed to obtain ¹ H images. Reconfigure (step ST2).

Thereafter, the operator refers to the ¹ H image, selects a voxel in which a tissue exists, and designates a voxel for performing spectrum processing (step ST3). Next, spectrum processing is performed for each designated voxel (step ST4), and an in-vivo metabolite is imaged based on the estimated area of the spectrum of each compound (step S).
T5).

In such an operation, spectrum processing is not performed for voxels in which no tissue exists, so that
The processing time can be shortened. However, in order to carry out this operation, it is necessary for the operator to draw a grid of a size corresponding to the voxel size on the ¹ H image, judge the presence or absence of tissue in each grid, and specify the processing voxel. Therefore, there is a drawback that it takes a lot of trouble.

Therefore, in order to solve this problem, a voxel having a maximum power spectrum of each voxel equal to or greater than a predetermined threshold value is selected, and it is determined that tissue exists in the selected voxel. It is easy to think of a method of performing spectrum processing only on voxels. However, in chemical shift imaging, there is a problem in terms of S / N ratio, and data acquisition is performed with a coarser matrix compared to the sequence for ¹ H image acquisition. Therefore, when the above method is used, the effect of Gibbs ringing occurs. May be strongly received, and a signal may be generated in a voxel in which no tissue actually exists.

Therefore, it is difficult to judge the presence or absence of the tissue even by using the method in which the voxels whose maximum value of the power spectrum is equal to or more than the predetermined threshold value are treated voxels.

On the other hand, in recent years, images of proton compounds using MRSI have been actively used. For example, in some tumor tissues, the concentration of NAA (N-acetylalbaric acid) is decreased compared to that in normal tissues, and Ch
Since the concentration of oline tends to be higher than that in normal tissue, imaging these distributions allows us to understand the site of metabolic abnormality in the body before it exhibits morphological changes.

In order to calculate image information reflecting the concentrations of various metabolites contained in the living body, it is necessary to calculate the area of each chemical shift spectrum (real part of spectrum). However, in the collected spectrum, phase rotation based on the system, sequence parameter, etc., or various causes other than those is observed,
This phase must be corrected when calculating the area.

Generally, the phase is approximately 1 with respect to frequency.
Since it can be approximated by the following functions, conventionally, the method of calculating the phase characteristic at the peak of the observed spectrum using the least square method and correcting the phase has been widely adopted.

However, since the concentrations of these compounds in vivo are generally low, the signal-to-noise ratio (S / N ratio) of the spectrum is not always good. Therefore, since a large error is included in the peak detection and the phase calculation in the phase correction processing, the spectrum after the phase correction does not draw an accurate absorption curve (real part spectrum) or dispersion curve (imaginary part spectrum), In some cases, only mixed spectra of parts and imaginary parts can be obtained. In such a case, it becomes difficult to obtain an accurate metabolic distribution image, and finally it is necessary to artificially perform phase correction on each of the many voxels acquired in MRSI. For example,
In the case of 3D-proton MRSI, often 32 ×
Since 32 matrix images are used, there is a drawback that it takes a lot of time to process the spectrum data when the phase correction is artificially performed.

In the case of MRI images, a method has been reported in which the magnitude of phase rotation is measured in advance and the phase is corrected using this data. That is, this is a method of measuring the phase distribution such that the imaginary part of each pixel of the image obtained for the phantom becomes 0, and performing the phase correction of the image measured thereafter. However, when such a method is applied to MRSI, the phase characteristics greatly change depending on the set position of the surface coil, for example, when the surface coil is used, and thus it is difficult to apply in practice.

The Journal of Magnetic Resonance, 69, 151-155 (1986), or Magne
tic Resonance in Medeicine 14,26-30 (19
As described in 90), a method of correcting the phase of each voxel by utilizing the fact that the phase of the spectrum can be corrected from the phase characteristic calculated in the time domain is conceivable. In this method, in principle, the primary phase characteristic Has a drawback that it cannot be corrected in each voxel, and that the phase calculation error is large due to the influence of the signal-to-noise ratio to obtain the phase characteristic in the time domain.

Further, in order to observe the proton metabolites, it is necessary to suppress a water signal having a magnitude of about 10 ^{4 with} respect to the concentration of the metabolite, and at this time it is difficult to completely remove the water signal. In this case, a water signal remains and a phenomenon in which the water spectrum and the metabolite spectrum overlap with each other is observed as shown in FIG. In such a case, there is a problem that the phase at the peak 31 of the metabolite spectrum is distorted by the residual water signal, so that the phase cannot be accurately corrected.

[0015]

As described above, in the conventional magnetic resonance diagnostic apparatus, since the operator designates the voxels for spectrum processing while referring to the ¹ H image, a lot of labor is required. There is. Further, even in the method in which the voxel having the maximum value of the power spectrum of each voxel is the set threshold value or more, it is difficult to determine the presence or absence of the tissue due to the influence of Gibbs ringing.

On the other hand, MR for collecting distribution images of metabolites
In SI, in order to accurately obtain a distribution image that reflects the concentration of each metabolite, spectrum phase correction must be performed, and conventionally, phase correction is artificially performed for each voxel. Therefore, there is a problem that it requires a lot of labor.

The present invention has been made to solve such a conventional problem, and a first object thereof is to provide a magnetic resonance diagnostic apparatus capable of easily designating a voxel for performing spectrum processing. A second object is to provide a magnetic resonance diagnostic apparatus that can easily carry out phase processing correction.

[0018]

To achieve the above object, the present invention applies a high frequency magnetic field and a gradient magnetic field to a subject placed in a uniform static magnetic field according to a predetermined pulse sequence, In a magnetic resonance diagnostic apparatus for detecting and imaging the magnetic resonance signal of, the means for imaging a predetermined site of the subject based on a metabolite having a high resolution and dividing the image into voxels of a predetermined size including a plurality of pixels. And, imaging the predetermined site based on a low-resolution metabolite, the size of the voxel, means for dividing into the same voxel as the absolute position, and the voxel in each voxel of the high-resolution metabolite image. A means for determining whether or not there are a certain number or more of pixels whose power spectrum value is equal to or greater than a predetermined threshold value; To further comprising a means for performing a spectrum processing of the low resolution metabolites image.
Further, in the invention of claim 1, means for counting the number of pixels whose power spectrum value of the voxel existing in each voxel determined to be the voxel to be subjected to the spectral processing is counted, and the pixel. Means for determining the tissue volume in each voxel by multiplying the number by the pixel volume.

Further, a high-frequency pulse is applied to the subject placed in a static magnetic field to collect spectral information of the metabolites in the subject, the phase shift of the spectrum is corrected, and then a distribution image of the metabolites is obtained. In the magnetic resonance diagnostic apparatus for obtaining a means for observing the water spectrum from each voxel divided when imaging the metabolite, and detecting the maximum value of the observed spectrum for each voxel, this maximum Means for setting the phase value of the spectrum in the value as a phase reference value, means for storing the phase characteristic of the spectrum depending on the system and the pulse sequence, and the metabolism based on the phase reference value and the phase characteristic of the spectrum. And means for correcting the phase of the spectrum forming the product image.

[0020]

With the configuration described above, when selecting a voxel to be spectrally processed from among the voxels acquired by chemical shift imaging, first, the 3D data of the ¹ H image signal is acquired, and this 3D data is acquired. It is divided into lattices corresponding to voxels acquired by chemical shift imaging.
Then, the pixel value of the pixel forming each voxel is detected, and a voxel having a predetermined number (usually “1”) of pixels having a pixel value larger than a predetermined threshold value is selected as a voxel for spectrum processing. .. This makes selection of voxels for spectrum processing easy and accurate.

When determining the volume of tissue contained in each voxel, the pixel value is determined for each pixel in each voxel forming the ¹ H image, and whether each pixel value is larger than the threshold value of Determine whether or not. Then, the number of pixels larger than the threshold value is obtained, and this number is multiplied by the volume per pixel to obtain the volume of the tissue in the voxel. This facilitates the calculation of the tissue volume and is extremely useful for calculating the concentration.

When correcting the phase shift of the spectrum, first, the water spectrum is observed, and the phase of the chemical shift spectrum of the metabolite is corrected based on this observation result. Therefore, it is possible to perform accurate phase correction without requiring a lot of manual labor.

[0023]

Embodiments of the present invention will be described below with reference to the drawings. FIG. 1 is a block diagram showing the configuration of a magnetic resonance diagnostic apparatus according to the present invention.

In the figure, a static magnetic field magnet 1 and a gradient coil 2 and a shim coil 4 provided inside the static magnetic field magnet 1 make a uniform static magnetic field on an object (not shown) and x, y, z which are orthogonal to each other in the same direction. A gradient magnetic field having a linear gradient magnetic field distribution in the direction is applied. The gradient coil 2 is driven by a gradient coil power supply 5, and the shim coil 4 is driven by a shim coil power supply 6. The probe 3 provided inside the gradient coil 2 applies a high-frequency magnetic field to the subject by being supplied with the high-frequency signal from the transmitter 7, and receives a magnetic resonance signal from the subject. The probe 3 may be used for both transmission and reception, or may be separately provided for transmission and reception. The magnetic resonance signal received by the probe 3 is detected by the receiving unit 8 and then transferred to the data collecting unit 9, where it is A / D converted and then the computer system 10
And processed for data processing.

The above gradient coil power supply 5, shim coil power supply 6, receiving unit 8 and data collecting unit 9 are all controlled by the sequence control unit 12, and the sequence control unit 1 is also used.
2 is controlled by the computer system 10. The computer system 10 is controlled by a command from the console 11. Then, the magnetic resonance signal input from the data collection unit 9 to the computer system 10 is subjected to Fourier transform or the like, and image data of the density distribution of desired nuclei in the subject is reconstructed based on the Fourier transform. Then, this image data is sent to the image display 13 and displayed as an image.

Next, the operation of designating a voxel for spectrum processing will be described with reference to the flowchart shown in FIG.

First, three-dimensional data of a ¹ H image signal is acquired from a desired region of a subject (step ST11).
This is performed by, for example, the pulse sequence for spatial 3D ¹ H image acquisition shown in FIG. 3 or the multi-slice sequence shown in FIG.

Next, the acquired ¹ H image data is divided by the grid corresponding to the voxels acquired by chemical shift imaging (step ST12). For example, 3D ¹
When the H image is acquired, the grid 22 is determined for the three-dimensional image 21 as shown in FIG. When a multi-slice ¹ H image is acquired, the grid 22 is selected for the multi-slice image 23 as shown in FIG.

After that, the threshold value of the pixels existing in each voxel is determined, and it is determined whether or not there are a predetermined number Np or more of pixels having the threshold value or more in each voxel (step ST13). Normally, the number Np is set to "1". Further, the voxels to be processed at this time are signal acquisition voxels, and for example, when chemical shift imaging by the local excitation method is used, as shown in FIG. 7, only the voxels in the local excitation region 24 are processed. To do.

Then, for a voxel determined to have Np or more pixels equal to or more than the threshold value (YES in step ST13), the voxel position is stored as a voxel for spectrum processing. That is, when even one pixel in the voxel is equal to or larger than the threshold value (when Np = 1), it is determined that the tissue exists in this voxel, and the spectrum processing is performed.

On the other hand, for a voxel determined not to have Np or more pixels equal to or more than the threshold value (NO in step ST13), this voxel is set as a voxel that does not require spectrum processing (step ST15). Then, it is determined whether or not all voxels are to be processed voxels (step ST16).

As a result, a voxel to be subjected to spectrum processing is selected from all voxels, and if the spectrum processing is performed only on the selected voxel, the abundance of the desired in-vivo metabolite can be imaged. You can do it.

In this way, in this embodiment, the voxels to be subjected to the spectral processing are selected based on the ¹ H image. Therefore, the spectrum processing for unnecessary voxels can be omitted, and the voxels to be processed are automatically determined, so that the manual labor can be reduced.

Further, the ¹ H image has a good S / N ratio,
Since the data can be collected in a fine matrix, it is not affected by Gibbs ringing. Therefore, the voxel to be processed is not erroneously recognized.

When performing the spectrum processing, the voxel position previously stored in the computer memory may be called and the spectrum processing may be performed only on the voxels judged to be for processing. Store the voxel position that you decide not to do in the computer memory,
A method of performing spectrum processing on voxels other than those stored when performing spectrum processing may be adopted.

Next, a second embodiment of the present invention will be described. In the first embodiment, the processing voxels are determined based on the ¹ H image, but if the user's region of interest is limited, it is not necessary to perform the spectral processing on all the selected voxels. Therefore, in this embodiment, a method of designating voxels in the region of interest will be described. FIG. 8 is a flowchart showing the operation procedure of the second embodiment. First, the region where the tissue exists is extracted from the acquired ¹ H image according to the procedure of the flowchart shown in FIG. 2 (step ST21).

Next, the position in the slice axis direction on the chemical shift imaging is calculated based on the position information in the slice axis direction of the ¹ H image (step ST22). Here, the slice center of the ¹ H image is used for calculating the slice position.

After that, the region of interest is surrounded on the ¹ H image, and this position information is input to the computer system 10 (step ST23). At this time, a pen, a mouse, or the like that can input the locus by tracing on the CRT with a pen is used. As a result, a region of interest 25 is set as shown in FIG. 9, for example. Then, it is determined which voxel position the line surrounding the region of interest corresponds to (step ST24). For example, as shown in FIG. 10, positions 27 at the four corners of each voxel
Is stored, and a voxel including the line 26 surrounding the region of interest may be searched for. This provides information on the outer edges of voxels within the region of interest.

After that, the voxels inside the obtained outer edge voxels are set as the voxels for spectrum processing, and the spectrum processing is performed only on these voxels (step ST25).

In this way, in the second embodiment,
A region of interest is further set to the voxels to be spectrally processed selected by the operation of the first embodiment, and spectral processing is performed only on the voxels in this region. Therefore, unnecessary spectrum processing of voxels can be omitted, and the spectrum processing can be shortened.

Next, a third embodiment of the present invention will be described. In this embodiment, a method for obtaining the volume of tissue existing in each voxel is proposed. In the above-described first embodiment, each pixel value in the voxel is obtained, and a process of determining whether the pixel value is larger than a predetermined threshold value is performed. In this embodiment, in each voxel, the number N (x) of pixels having a pixel value larger than a predetermined threshold value and the number N (y) of pixels having a smaller pixel value are obtained, and the number of pixels N (x), N (x) is obtained. The volume of the tissue is calculated based on (y).

Now, the number N of pixels in a voxel
When (y) is larger than the predetermined value M (usually “1”), there is a part other than the tissue in this voxel.
The tissue volume V (x) can be calculated by the following equation (1).

V (x) = N (x) · Vp (1) Here, Vp is the volume of the pixel.

When the number N (y) of pixels in a voxel is smaller than the predetermined value M, the volume V (x) of the tissue is equal to the volume of the voxel because all the voxels are tissues. That is, it is expressed by the following equation (2).

V (x) = N · Vp (2) where N is the number of pixels in one voxel In this way, in the third embodiment, the tissue volume contained in the voxel can be easily calculated. Therefore, it is extremely useful when obtaining the density of an image.

Next, a fourth embodiment of the present invention will be described. In this embodiment, a method for automatically performing phase processing correction in MRSI will be shown.

FIG. 11 is a Lough chart showing the operation procedure of this embodiment. First, the reference data for performing the phase correction of the target MRSI data is imaged (step ST31). The measurement of the phase data is performed using the pulse sequence shown in FIG. 12, which excludes the water signal suppression process, that is, the pulse sequence shown in FIG. 13, from the example of the pulse sequence for observing the proton metabolites. .. Here, a localized MRSI pulse sequence for imaging a spatial two-dimensional metabolite distribution is shown. At this time, the repetition time of the pulse sequence is set shorter than when observing the proton metabolites, and the measurement time can be shortened. If possible, the observation time can be shortened by roughening the frequency resolution. Next, the maximum value of the spectrum (absolute value) by the water signal of each voxel is detected (step ST32).
At this time, it is preferable not to perform subsequent processing on voxels whose spectrum size is smaller than a preset threshold value.

Thereafter, the phase φ ₁ (x, y, z) at the peak position of the spectrum of each voxel is calculated and stored (step ST33). Here, the phase in the vicinity of the peak of the spectrum usually changes abruptly as shown in 14 (a) and 14 (b), so a large error may be included in the calculation of the phase. Therefore, when the spectrum line width is narrow, such error can be reduced by performing spectrum line loading processing by multiplying by an exponential function or the like in the time domain as is well known.

Next, the frequency of the transmitter / receiver is changed by Δf, the processes from step 31 to step 33 are executed, and the phase φ ₂ (x, y, z) at the peak position of the spectrum of each voxel at this time is calculated. Ask and memorize this.

Then, the obtained phase φ ₁ (x, y,
Based on z) and φ ₂ (x, y, z), the phase characteristic θ (x, y, z, f) in each voxel is calculated and stored (step ST35). Here, the phase characteristic is usually
Phase characteristics up to the 0th and 1st order are calculated, but if necessary, higher order characteristics can be obtained by changing the frequency Δf by several points.

Assuming that the frequency of the transceiver at the time of measuring the phase φ ₁ is f ₀ , the phase characteristic θ (x, y, z,
f) and the phases φ ₁ (x, y, z), φ ₂ (x, y,
z) is the phase correction coefficient a (x, y,
z) and b (x, y, z) are shown by the following equations (3) to (5).

Θ (x, y, z, f) = a (x, y, z) f + b (x, y, z) (3) φ ₁ = a (x, y, z) · f ₀ + b ( x, y, z) (4) φ ₂ = a (x, y, z) · (f ₀ + Δf) + b (x, y, z) (5) Therefore, the first-order phase correction coefficient a (x , Y, z), and 0
The next phase correction coefficient b (x, y, z) is calculated by the following (6), (7)
It is shown by the formula.

A (x, y, z) = (φ ₂ −φ ₁ ) / Δf (6) b (x, y, z) = φ ₁ − (φ ₂ −φ ₁ ) · f ₀ / Δf (7) After that, MRSI data is collected by the pulse sequence for metabolite observation shown in FIG. 12 (step ST3
6). The spectrum Sm of each voxel observed at this time
(X, y, z) is the phase characteristic θ (x, y, z, f)
Therefore, the spectrum distortion is exhibited for the spectrum shape S (x, y, z) in which the real part and the imaginary part are separated.

Therefore, the spectrum Sm (x, y, z) is expressed by the following equation (8).

Sm (x, y, z) = S (x, y, z) .exp (-j.theta. (X, y, z, f)) (8) Then, the above equations (6) and (7) Using the phase characteristics shown in (8)
If the phase term of the equation is removed, the observed spectrum Sm
A spectrum S (x, y, z) in which the real part and the imaginary part are separated can be obtained from (x, y, z) (step ST3).
7).

Thereafter, the area of the obtained real part spectrum is calculated, and in some cases, the peak height of the spectrum is calculated (step ST38), and the metabolite distribution image is displayed based on this (step ST39). ..

In this way, a phase-corrected metabolite distribution image can be obtained. In this way, in the fourth embodiment, the phase characteristic is calculated in advance by using the water spectrum, and the phase around the spectrum of the proton metabolite is corrected using this, so that manual labor is required. Instead, the phase can be easily corrected. As a result, a physically meaningful image reflecting the concentration distribution of metabolites can be obtained in a short time.

In the fourth embodiment, when there is a second-order or higher-order phase characteristic depending on the system or the like, it is possible to measure the higher-order phase characteristic in advance and correct it as described above. However, if the filter characteristic that is the cause of the high-order phase characteristic is existing as shown in FIG. 15, such a characteristic is stored and the phase characteristic obtained by the method is superimposed to correct the characteristic. You can also do it.

Further, if the procedure shown in the fourth embodiment is used, the phase generated in the spectrum collected by the pulse sequence having a so-called dead time due to the encoding time for adding the position information as shown in FIG. It is also possible to correct the first-order distortion of.

[0060]

As described above, according to the present invention, when imaging the abundance of in vivo metabolites by chemical shift imaging, the voxels to be spectrally processed can be easily and accurately selected. Therefore, the manual labor can be reduced, and the processing time can be shortened.

Further, the pixel value of the ¹ H image is obtained, the number of pixels having a pixel value of a predetermined value or more existing in the voxel is counted, and the volume of the tissue contained in the voxel is obtained based on this counted value. There is. Therefore, the tissue volume can be easily calculated, which is extremely useful for calculating the concentration.

When obtaining spectral information using MRSI, a water spectrum is used to obtain a phase characteristic in advance, and the phase rotation of the spectrum of the metabolite is corrected using this phase characteristic. Therefore, it is possible to easily perform the phase correction without requiring manual labor, and as a result, it is possible to obtain an accurate image that reflects the concentration distribution of the metabolite in a short time.

[Brief description of drawings]

FIG. 1 is a block diagram showing a configuration of a magnetic resonance diagnostic apparatus according to the present invention.

FIG. 2 is a flowchart showing the operation of the first embodiment of the present invention.

FIG. 3 is a pulse sequence diagram for acquiring a spatial three-dimensional ¹ H image.

FIG. 4 is a pulse sequence diagram when a ¹ H image is acquired by multi-slice.

FIG. 5 is an explanatory diagram showing a grid set on a three-dimensional image.

FIG. 6 is an explanatory diagram showing grids set on a multi-slice image.

FIG. 7 is an explanatory diagram showing a local excitation region.

FIG. 8 is a flowchart showing the operation of the second embodiment of the present invention.

FIG. 9 is an explanatory diagram showing a region of interest to be set.

FIG. 10 is an explanatory diagram showing outer edge information of voxels.

FIG. 11 is a flowchart showing the operation of the fourth embodiment of the present invention.

FIG. 12 is a pulse sequence diagram for observing proton metabolites.

FIG. 13 is a pulse sequence diagram for observing proton metabolites without water suppression processing.

FIG. 14 is a phase characteristic diagram of a spectrum.

FIG. 15 is an explanatory diagram showing an example of filter characteristics.

FIG. 16 is a pulse sequence diagram of MRSI having a dead time.

FIG. 17 is a flowchart showing a conventional chemical shift imaging procedure.

FIG. 18 is an explanatory diagram showing overlapping of spectra.

[Explanation of symbols]

1 static magnetic field magnet 2 gradient coil 3 probe 4 shim coil 7 transmitter 8 receiver 9 data processor 10 computer system 11 console 21 ¹ H three-dimensional image 22 lattice 23 ¹ H multi-slice image 24 local excitation region 25 region of interest

Claims (4)

[Claims] 1. A magnetic resonance diagnostic apparatus for applying a high-frequency magnetic field and a gradient magnetic field to a subject placed in a uniform static magnetic field according to a predetermined pulse sequence, and detecting a magnetic resonance signal from the subject to visualize it. In the above, a predetermined site of the subject is imaged based on a metabolite having a high resolution, a means for dividing the image into voxels of a predetermined size composed of a plurality of pixels, and the predetermined site is imaged based on the metabolite having a low resolution. And a means for dividing this into voxels of the same size and absolute position as the voxels, and in each of the voxels of the high-resolution metabolite image, the value of the power spectrum of the voxels is equal to or more than a predetermined threshold pixel. Means for determining whether or not there are more than a certain number, and spectral processing of the low-resolution metabolite image is performed only for the voxels determined to have the certain number or more. Magnetic resonance imaging apparatus characterized by comprising cormorants and means. 2. A means for counting the number of pixels having a power spectrum value of the voxel existing in each voxel determined to be the voxel to be subjected to the spectrum processing, the pixel volume being equal to or larger than a predetermined threshold value. The magnetic resonance diagnostic apparatus according to claim 1, further comprising: means for multiplying by to obtain a tissue volume in each voxel. 3. A metabolite distribution image after applying a high-frequency pulse to a subject placed in a static magnetic field to collect spectrum information of the metabolite in the subject and correcting the phase shift of the spectrum. In the magnetic resonance diagnostic apparatus to obtain, means for observing the water spectrum from each voxel divided when imaging the metabolite, and detecting the maximum value of the observed spectrum for each voxel, this maximum Means for setting the phase value of the spectrum in the value as the phase reference value, means for storing the phase characteristic of the spectrum depending on the system and the pulse sequence, and the metabolism based on the phase reference value and the phase characteristic of the spectrum. Means for correcting the phase of the spectrum forming the product image, and a magnetic resonance diagnostic apparatus comprising: 4. A metabolite distribution image obtained by applying a high-frequency pulse to a subject placed in a static magnetic field to collect spectral information of the metabolite in the subject and correcting the phase shift of the spectrum. In the magnetic resonance diagnostic apparatus to obtain, the water spectrum from each voxel divided when imaging the metabolite, a means for observing multiple times at different resonance frequencies, and the maximum value of the spectrum at each frequency for each voxel And a means for obtaining a phase characteristic with respect to frequency from the phase value of the spectrum at each maximum value, and a means for correcting the phase of the spectrum forming the metabolite image based on the phase characteristic. A characteristic magnetic resonance diagnostic apparatus.