JP6554717B2 - Magnetic resonance apparatus and program - Google Patents

Description

  The present invention relates to a magnetic resonance apparatus that collects data including diffusion information, and a program applied to the magnetic resonance apparatus.

  In recent years, diffusion weighted imaging (DWI) has become widespread. DWI is a method of acquiring a diffusion weighted image including diffusion information of an imaging region by applying a gradient magnetic field called MPG (Motion Probing Gradient) having a large magnetic field strength. By using the diffusion weighted image, various diffusion information such as an apparent diffusion coefficient (ADC) can be acquired. However, it is known that when tissue is moved by pulsation or peristalsis during application of MPG, image intensity loss occurs. When such a loss occurs, there is a problem that an estimation error of the value of an apparent diffusion coefficient (hereinafter referred to as “ADC”) becomes large. In order to deal with this problem, a method of executing homomorphic filtering using a modulation map is known (see Non-Patent Document 1).

JP 2012-157687 A

Lim, Two-dimensional signal and image processing, Prentice Hall, 1990

However, when homomorphic filtering is performed, there is a problem that the contrast of the image changes.
Therefore, a technique that does not cause a change in contrast as much as possible is desired.

According to a first aspect of the present invention, a scanning unit that executes a first sequence for collecting data including diffusion information from a slice crossing an imaging region including a plurality of regions, a plurality of times;
First image creating means for creating a plurality of first images including diffusion information based on data collected by executing the first sequence a plurality of times;
For each of the first images, first filtering means for performing first filtering for restoring image intensity in a first portion of the first image where loss of image intensity has occurred. And a first filter means that does not perform the first filtering in a second part of the first image.

A second aspect of the present invention is a program applied to a magnetic resonance apparatus that executes a first sequence for collecting data including diffusion information from a slice crossing an imaging region including a plurality of regions a plurality of times. ,
A first image creation process for creating a plurality of first images including diffusion information based on data collected by executing the first sequence a plurality of times;
For each of the first images, a first filtering process for performing a first filtering for restoring the image intensity in a first portion of the first image where a loss of image intensity has occurred. A first filtering process that does not perform the first filtering in a second part of the first image;
Is a program for causing a computer to execute.

  Since the first filtering is not performed in the second part of the first image, a change in contrast can be prevented.

It is the schematic of the magnetic resonance apparatus of one form of this invention. It is a figure which shows the process which the processor 9 performs. It is a figure which shows the operation | movement flow of MR apparatus. It is a figure which shows the slice set when performing a scan. It is explanatory drawing of the scan SC for acquiring the diffusion emphasis image of a slice. It is explanatory drawing of scan SC1. It is a figure which shows an example of the sequence E1. Period diffusion weighted image _DW 11 obtained by sequence E1~E3 of _V1, DW _21, DW 31 a (b = b1) is a view schematically showing. It is a figure which shows roughly the diffusion weighted image obtained by sequence E1-E3 of the period V2. It is a figure which shows roughly the diffusion emphasis image acquired by performing sequence E1, E2, and E3 in the period Vn. It is explanatory drawing of scan SC2. It is a figure which shows an example of the sequence F1. It is a figure which shows roughly the diffusion emphasis image acquired in the period W1-Wn. Is a diagram illustrating an example of a diffusion weighted image DW _11. Is a diagram illustrating a diffusion weighted image _DW 11 _{~DW 1n.} It is a figure which shows schematically the production method of SD map. It is a figure which shows schematically the SD map 2 of the slice 2, and the SD map 3 of the slice 3. It is a figure which shows schematically the SD map acquired for every slice in b = b2. It is a figure which shows an example of SD map. It is explanatory drawing of the method of producing the masks MK1-MK3. It is explanatory drawing of the method of producing masks MK4-MK6. It is a figure which shows a specific example of mask MK1. It is explanatory drawing of the creation method of a modulation map. It is explanatory drawing of the example which performs a homomorphic filtering using Formula (1), without limiting the area | region where a homomorphic filtering is performed. It is explanatory drawing of the example which limits the area | region where homomorphic filtering is performed and performs homomorphic filtering using Formula (1). It is a figure which shows the homomorphic filter image in b = b1. It is a figure which shows the homomorphic filter image in b = b2. It is a figure which shows an example of a difference image DIF. It is explanatory drawing when mask MK1 is low-pass filtered. It is explanatory drawing when carrying out low pass filtering of the difference image DIF. It is explanatory drawing of a 2nd term. It is explanatory drawing of a 1st term. It is explanatory drawing when adding the image parts D1 and D2. It is a figure which shows the decompression | restoration image in b = b1. It is a figure which shows the decompression | restoration image in b = b2. It is a figure which shows an example of the synthetic | combination method of a restored image. It is a diagram illustrating an example of a composite image DC _1. a composite image DC ₄ slices 1 in b = b2 is a diagram schematically showing. It is explanatory drawing of simulation conditions. It is a figure which shows a simulation result. It is a figure which shows an ADC map.

  Hereinafter, although the form for inventing is demonstrated, this invention is not limited to the following forms.

FIG. 1 is a schematic view of a magnetic resonance apparatus according to one embodiment of the present invention.
A magnetic resonance apparatus (hereinafter referred to as “MR apparatus”) 100 includes a magnet 2, a table 3, a reception RF coil 4, and the like.

  The magnet 2 has a bore 21 in which the subject 13 is accommodated. The magnet 2 includes a superconducting coil, a gradient coil, an RF coil, and the like (not shown). The superconducting coil applies a static magnetic field, the gradient coil applies a gradient magnetic field, and the RF coil applies an RF pulse.

  The table 3 has a cradle 3 a for supporting the subject 13. The cradle 3a is configured to be able to move into the bore 21. The subject 13 is transported to the bore 21 by the cradle 3a.

  A reception RF coil (hereinafter simply referred to as “coil”) 4 is attached to the body of the subject 13.

  The MR apparatus 100 further includes a transmitter 5, a gradient magnetic field power supply 6, a receiver 7, a computer 8, an operation unit 11, a display unit 12, and the like.

  The transmitter 5 supplies current to the RF coil, and the gradient magnetic field power source 6 supplies current to the gradient coil. The receiver 7 performs signal processing such as detection on the signal received by the coil 4. A combination of the magnet 2, the transmitter 5, the gradient magnetic field power source 6, and the receiver 7 corresponds to the scanning means.

  The computer 8 controls the operation of each part of the MR apparatus 100. The computer 8 has a processor 9 and a memory 10.

  The memory 10 stores a program executed by the processor 9. The processor 9 reads a program stored in the memory 10 and executes processing described in the program. FIG. 2 shows processing executed by the processor 9.

The image creating unit 91 creates an image based on data collected by scanning.
The mask creating means 92 creates a mask (see FIG. 22) used when executing homomorphic filtering. Details of the mask will be described later.
The modulation map creating means 93 creates a modulation map (see FIG. 23) used when executing homomorphic filtering. Details of the modulation map will be described later.
The homomorphic filter means 94 performs homomorphic filtering on the image created by the image creating means 91 using the mask and the modulation map. The homomorphic filter means 94 corresponds to the first filter means.

The difference image creating means 95 creates a difference image (see FIG. 28) described later.
The low-pass filter unit 96 performs low-pass filtering on the mask and the difference image. The low-pass filter means 96 corresponds to the second filter means.
The restored image creating means 97 obtains a restored image based on the formula (3) described later. The restored image creation unit 97 corresponds to a second image creation unit.
The composite image creating unit 98 creates a composite image by compositing the restored image.
The ADC map creating means 99 creates an ADC map based on the composite image.

  The processor 9 is an example that constitutes the image creation means 91 to the ADC map creation means 99, and functions as these means by executing a program stored in the memory 10.

The operation unit 11 is operated by an operator and inputs various information to the computer 8. The display unit 12 displays various information.
The MR apparatus 100 is configured as described above.

  In the present embodiment, the subject is scanned using the MR apparatus 100, and an ADC map is created based on the data obtained by the scan. Hereinafter, an operation flow of the MR apparatus 100 will be described.

FIG. 3 is a diagram showing an operation flow of the MR apparatus.
In step ST1, a scan for acquiring a diffusion weighted image is executed (see FIGS. 4 and 5).

FIG. 4 is a diagram showing slices set when executing a scan, and FIG. 5 is an explanatory diagram of a scan SC for obtaining a diffusion-weighted image of the slices.
In this embodiment, a slice is set at a position that crosses the liver. FIG. 4 shows an example in which three slices 1 to 3 are set for convenience of explanation.

  The scan SC is divided into two scans SC1 and SC2. The scan SC1 is a scan for acquiring diffusion information of slices 1 to 3 by a sequence having MPG (Motion Probing Gradient) in which a b value indicating the intensity of diffusion emphasis is set to b = b1. On the other hand, the scan SC2 is a scan for acquiring the diffusion information of the slices 1 to 3 by the sequence having the MPG set to b = b2.

FIG. 6 is an explanatory diagram of the scan SC1.
FIG. 6 shows an example in which NEX (Number of Excitation) of each slice is n times (NEX = n). The scan SC1 has n periods V1 to Vn. In each period, three sequences E1, E2, and E3 are executed. Accordingly, in the scan SC1, the set of three sequences E1, E2, and E3 is executed n times. Hereinafter, the sequences E1, E2, and E3 executed in the periods V1 to Vn of the scan SC1 will be described.

  In the scan SC1, first, the sequence E1 is executed in the period V1 (see FIG. 7).

FIG. 7 is a diagram illustrating an example of the sequence E1.
The sequence E1 has the following pulses (1) to (4).
(1) RF pulse (1a) Excitation pulse α1 for exciting the slice
(1b) Refocus pulse α2 for refocusing spin
(2) Gradient pulse in the slice selection direction (2a) Slice selection gradient pulses G _s1 and G _s2 for selecting a slice
(3) Gradient pulse in frequency encoding direction (3a) Gradient pulse G _fs
(3b) MPG of b = b1
(3c) Read gradient pulse G _re for collecting echoes
(4) Gradient pulse in phase encoding direction (4a) Gradient pulse G _ps
(4b) Phase encoding gradient pulse Gp

The sequence E1 excites the slice 1 and collects k-space data necessary for the reconstruction of the diffusion weighted image (b = b1) of the slice 1. In this embodiment, the sequence E1 collects all the k-space data necessary for the reconstruction of the diffusion weighted image (b = b1) of the slice 1 in one shot (single shot method). The image creating unit 91 (see FIG. 2) creates a diffusion weighted image based on the data collected by the sequence E1.
After the sequence E1 is completed, the sequence E2 is executed.

Similarly to the sequence E1, the sequence E2 is represented by the sequence chart shown in FIG. However, in the sequence E2, the frequency of the RF pulse is set so that the slice 2 is excited. Sequence E2 excites slice 2 and collects k-space data necessary to reconstruct the slice 2 diffusion weighted image (b = b1). The image creating unit 91 creates a diffusion weighted image based on the data collected by the sequence E2.
After the sequence E2 ends, the sequence E3 is executed.

  Similarly to the sequence E1, the sequence E3 is represented by the sequence chart shown in FIG. However, in the sequence E3, the frequency of the RF pulse is set so that the slice 3 is excited. The sequence E3 excites the slice 3 and collects k-space data necessary for the reconstruction of the diffusion weighted image (b = b1) of the slice 3. The image creating unit 91 creates a diffusion weighted image based on the data collected by the sequence E3.

Therefore, by executing the sequences E1, E2, and E3 in the period V1, the diffusion weighted images of the slices 1 to 3 can be acquired. FIG. 8 schematically shows diffusion weighted images DW ₁₁ , DW ₂₁ , DW ₃₁ (b = b1) obtained by the sequences E1 to E3 in the period V1.
After the period V1, the sequence is executed in the next period V2.

In the period V2, as in the period V1, the sequences E1, E2, and E3 are executed. Therefore, the diffusion weighted images (b = b1) of the slices 1 to 3 can be acquired even in the period V2. FIG. 9 schematically shows a diffusion weighted image obtained by the sequences E1 to E3 in the period V2. In FIG. 9, in order to distinguish between the diffusion-weighted image acquired in the period V1 and the diffusion-weighted image acquired in the period V2, the diffusion-weighted images acquired in the period V2 are denoted by codes “DW ₁₂ ” and “DW ₂₂ ”. ”And“ DW ₃₂ ”.

Similarly, a set of three sequences E1, E2, and E3 is repeatedly executed. Then, in the last period Vn of the scan SC1, sequences E1, E2, and E3 are executed. FIG. 10 schematically shows a diffusion weighted image obtained by executing the sequences E1, E2, and E3 in the period Vn. In FIG. 10, the diffusion weighted images acquired in the period Vn are indicated by codes “DW _1n ”, “DW _2n ”, and “DW _3n ”.
Next, the scan SC2 will be described.

FIG. 11 is an explanatory diagram of the scan SC2.
FIG. 11 shows an example of NEX = n as in FIG. The scan SC2 has n periods W1 to Wn. In each period, three sequences F1, F2, and F3 are executed. Therefore, in the scan S2, the set of three sequences F1, F2, and F3 is executed n times. Hereinafter, the sequences F1, F2, and F3 executed in the periods W1 to Wn of the scan SC2 will be described.
In the scan SC2, first, the sequence F1 is executed in the period W1 (see FIG. 12).

FIG. 12 is a diagram illustrating an example of the sequence F1.
Compared with the sequence E1 (see FIG. 7), the sequence F1 has the MPG set to b = b2, but the other pulses are the same as the sequence E1. Therefore, the sequence F1 excites the slice 1 and collects k-space data necessary for the reconstruction of the diffusion weighted image (b = b2) of the slice 1. The image creating unit 91 creates a diffusion weighted image based on the data collected by the sequence F1.
After the sequence F1 is completed, the sequence F2 is executed.

Similarly to the sequence F1, the sequence F2 is also represented by the sequence chart shown in FIG. However, in the sequence F2, the frequency of the RF pulse is set so that the slice 2 is excited. Sequence F2 excites slice 2 and collects k-space data necessary to reconstruct the slice 2 diffusion weighted image (b = b2). The image creating unit 91 creates a diffusion weighted image based on the data collected by the sequence F2.
After the sequence F2 is completed, the sequence F3 is executed.

  The sequence F3 is also represented by the sequence chart shown in FIG. 12, similarly to the sequence F1. However, in the sequence F3, the frequency of the RF pulse is set so that the slice 3 is excited. The sequence F3 excites the slice 3 and collects k-space data necessary for the reconstruction of the diffusion weighted image (b = b2) of the slice 3. The image creating unit 91 creates a diffusion weighted image based on the data collected by the sequence F3.

  Similarly, for the periods W2 to Wn, a set of three sequences F1, F2, and F3 is repeatedly executed. Therefore, by executing the scan SC2, n diffusion-weighted images can be acquired for one slice. FIG. 13 schematically shows diffusion-weighted images acquired in the periods W1 to Wn.

  In this way, scans SC1 and SC2 are executed. The scans SC1 and SC2 may be executed while the subject is holding his / her breath, or may be executed using a respiratory synchronization method.

However, if an organ moves due to pulsation, peristalsis, etc. during application of MPG, it is known that a loss of image intensity (signal loss) appears in the diffusion weighted image due to the influence of the movement of the organ. ing. Figure 14 shows an example of a diffusion weighted image _{DW 11.} In Figure 14, a portion in the diffusion weighted image _{DW 11} (e.g., the left lobe of the liver, stomach, and spleen), the loss of image intensity LO1, LO2, and LO3 has appeared. This loss of image intensity, to create a ADC maps in step ST8 to be described later, it may cause an error of ADC values, images of the portion loss of image intensity occurs in the diffusion weighted image DW ₁₁ Strength should be restored as much as possible. Therefore, in this embodiment, after performing the scans SC1 and SC2, step ST2 is executed to restore the image intensity. Hereinafter, step ST2 will be described. Since step ST2 includes steps ST21 and ST22, steps ST21 and ST22 will be described in order.

  In step ST21, mask creating means 92 (see FIG. 2) creates a mask (see FIG. 22) used in step ST22 described later. Hereinafter, a method for creating a mask will be described.

First, the mask creating unit 92 selects the diffusion weighted images DW _{11 to} DW _1n of the slice 1 from the diffusion weighted images obtained by the scan SC1 with b = b1. FIG. 15 shows the selected diffusion weighted images DW _{11 to} DW _1n .

Next, the mask creating means 92 creates an SD map representing the standard deviation of the pixel values of the diffusion weighted images DW _{11 to} DW _1n . FIG. 16 schematically shows a method for creating an SD map.

The mask creating unit 92 selects pixel values of pixels at the same position from the diffusion weighted images DW _{11 to} DW _1n, and obtains a standard deviation of the selected pixel values. For example, if the standard deviation of the pixel value of the pixel Pa of the diffusion weighted image _DW 11 _{~DW 1n,} mask making unit 92 selects the pixel value Xa1~Xan pixel Pa from diffusion weighted image _DW 11 _{~DW 1n.} Then, the standard deviation σa of the pixel values Xa1 to Xan is calculated. The standard deviation σa calculated in this way is used as a value at the pixel Pa of the SD map 1.

In the above description, the method of obtaining the standard deviation σa in the pixel Pa of the SD map has been described, but the standard deviation in other pixels of the SD map is also calculated in the same manner. For example, the value in the pixel Pb of the SD map 1 can be obtained by calculating the standard deviation σb of the pixel values Xb1 to Xbn in the pixels Pb of the diffusion weighted images DW _{11 to} DW _1n . Similarly, by calculating the standard deviation σc of the pixel values Xc1~Xcn in the pixel Pc of the diffusion weighted image _DW 11 _{~DW 1n,} it is possible to obtain a value in the pixel Pc of SD map 1.

  Therefore, by calculating the standard deviation for each pixel, the SD map 1 of slice 1 at b = b1 can be obtained.

  15 and 16 show the procedure for creating the SD map 1 of the slice 1. However, the SD map can be obtained for the other slices 2 and 3 by the same method. FIG. 17 schematically shows an SD map 2 of slice 2 and an SD map 3 of slice 3. Thus, an SD map can be obtained for each slice.

FIGS. 15 to 17 show an example in which an SD map is created using the diffusion-weighted image acquired by the scan SC1 at b = b1, but the diffusion-weighted image acquired by the scan SC2 at b = b2. Also for an image, an SD map is created for each slice in the same manner. FIG. 18 schematically shows an SD map acquired for each slice at b = b2.
Therefore, a total of six SD maps are created. FIG. 19 shows an example of the SD map.

  Parts located near the heart, such as the left lobe, stomach, and spleen of the liver, move under the influence of pulsation, so the pixel values of parts such as the left lobe, stomach, and spleen of the liver change over time There is a tendency to change. On the other hand, since the right lobe of the liver is far from the heart, it is less affected by the heartbeat. Therefore, the temporal change in the pixel value of the right lobe of the liver tends not to be so large. For this reason, from the SD map, it is possible to roughly distinguish a part that is easily affected by the heart beat and a part that is not significantly affected by the heart beat.

  Next, the mask creating unit 92 creates a mask based on the SD map. 20 and 21 schematically show a method for creating a mask based on the SD map. The mask creating means 92 performs threshold processing TH for each of the SD maps 1 to 6. As the threshold value TH, a pixel value smaller than the threshold value among the pixel values of the SD map is set to “0”, while a pixel value equal to or higher than the threshold value is set to “1”. Accordingly, each SD map is divided into a “0” area and a “1” area. In this way, the masks MK1 to MK6 can be created by dividing each SD map into the “0” area and the “1” area. FIG. 22 representatively shows a specific example of the mask MK1 among the masks MK1 to MK6. In FIG. 22, an area with a pixel value of 1 is shown in white, and an area with a pixel value of 0 is shown in black.

As described above, in the SD map, the pixel value of a part that is easily affected by a heart beat or the like is different from the pixel value of a part that is not easily affected by a heart beat or the like. Therefore, by thresholding the SD map, a part that is easily affected by the heart beat (the left lobe of the liver, stomach, spleen, etc.) and a part that is not significantly affected by the heart beat (the liver And the right lobe). Referring to the mask MK1, a pixel value 1 is assigned to a part (left liver lobe, stomach, spleen, etc.) that is susceptible to the heart beat, and a part that is not significantly affected by the heart beat ( It can be seen that the pixel value 0 is mainly assigned to the right lobe of the liver). Also, areas that are susceptible to heart pulsation (liver left lobe, stomach, spleen, etc.) are prone to loss of image intensity, while those that are less susceptible to heart pulsation (right liver lobe). Etc.) is less likely to cause loss of image intensity. Therefore, the pixel value 1 region of the mask MK1 corresponds to a region where the loss of image intensity is likely to occur, while the region of pixel value 0 of the mask MK1 is a region where loss of image intensity is unlikely to occur. It turns out that it corresponds.
After creating the masks MK1 to MK6, the process proceeds to step ST22.

In step ST22, homomorphic filtering of the diffusion weighted image is executed. Step ST22 will be described below. The processing procedure of homomorphic filtering is the same for any diffusion-weighted image. Therefore, the procedure when the following description, among the obtained diffusion weighted images by scanning SC1 and SC2, which picks up the diffusion weighted image DW ₁₁ on behalf executes homomorphic filtering on diffusion weighted image DW ₁₁ explain.

  In step ST22, first, the modulation map creating means 93 (see FIG. 2) creates a modulation map used for homomorphic filtering (see FIG. 23).

FIG. 23 is an explanatory diagram of a method for creating a modulation map.
Modulation map creation means 93 first executes the low pass filtering of the diffusion weighted image _{DW 11.} In FIG. 23, an image obtained by low-pass filtering the diffusion weighted image DW ₁₁ (hereinafter referred to as “low-pass filter image”) is indicated by a symbol “DW _lpf ”.

Next, the modulation map creating means 93 inverts the low-pass filter image DW _lpf to create an inverted image. As a result, a modulation map D _mod1 is created.

After creating the modulation map D _mod1 , the homomorphic filter means 94 (see FIG. 2) performs homomorphic filtering on the diffusion weighted image DW ₁₁ using the modulation map D _mod1 . Specifically, homomorphic filtering is executed based on the following equation.
HM ₁₁ = DW ₁₁ * D _{mod 1} (1)

HM _{11 in} Expression (1) represents an image obtained by homomorphic filtering (hereinafter referred to as “homogeneous filter image”).

  Hereinafter, homomorphic filtering will be described. In this embodiment, when homomorphic filtering is performed, the region where homomorphic filtering is performed is limited based on the mask MK1. Hereinafter, in order to clarify the reason, an example in which homomorphic filtering is executed using Equation (1) without limiting the region in which homomorphic filtering is executed will be described. After that, the method of this embodiment in which homomorphic filtering is performed based on the mask MK1 and the homomorphic filtering is performed using the equation (1) will be described.

FIG. 24 is an explanatory diagram of an example in which homomorphic filtering is performed using Expression (1) without limiting the region in which homomorphic filtering is performed.
By multiplying the modulation map _{D mod1} the diffusion weighted image _{DW 11,} it executes the homomorphic filtering of the diffusion weighted image _{DW 11.} Thereby, a homomorphic filter image HM is obtained.

  FIG. 24 shows an example in which the image intensity is restored in LO1 and LO2 among the image intensity losses LO1, LO2, and LO3. Therefore, by homomorphic filtering, the image intensity loss LO3 in the spleen remains, but the left lobe and stomach of the liver can obtain a homomorphic filtered image HM in which the image intensity is restored.

  However, homomorphic filtering has the effect of enhancing contrast while restoring image intensity. Therefore, there is a problem that the contrast changes due to the homomorphic filtering. In particular, if the contrast changes, the ADC error calculated in step ST8, which will be described later, is likely to increase. Therefore, it is desirable to keep the contrast from changing as much as possible.

Therefore, in the present embodiment, the homomorphic filter means 94 limits the region in which the homomorphic filtering is performed based on the mask MK1 (see FIG. 22). By limiting the region where the homomorphic filtering is performed based on the mask MK1, there is an effect that it is possible to prevent a change in contrast in a portion of the diffusion weighted image DW ₁₁ where loss of image intensity is not so much generated. is there. Hereinafter, the reason why this effect is obtained will be described with reference to FIG.

The mask MK1 has a pixel value 1 region and a pixel value 0 region. The area of pixel value 1 of the mask MK1 corresponds to a part where the loss of image intensity occurs (for example, the left lobe and stomach of the liver), but the area of pixel value 0 of the mask MK1 is a loss of image intensity. This corresponds to a region where the occurrence of a little occurs (for example, the right lobe of the liver). Therefore, in the diffusion weighted image DW _11, in the portion corresponding to the region of the pixel value 1 in the mask MK1, it is necessary to perform the homomorphic filtering to restore the image intensity, but in a diffusion weighted image DW ₁₁ Thus, in the portion corresponding to the area of the pixel value 0 of the mask MK1, there is not much loss of image intensity, so it is not necessary to perform homomorphic filtering. Therefore, in this embodiment, in a diffusion weighted image DW _11, the portion in the portion corresponding to the region of the pixel value 1 in the mask MK1 executes the homomorphic filtering, which corresponds to the region of the pixel value 0 in the mask MK1 is Homomorphic filtering is not performed. Thereby, since the homomorphic filtering is not executed in the most part of the right lobe of the liver, a change in contrast can be prevented. Therefore, homomorphic filter image HM ₁₁ is the left lobe and stomach of the liver is the image intensity is restored, is prevented changes in contrast in the liver right leaves. For this reason, in the present embodiment, the region where homomorphic filtering is executed is limited based on the mask MK1 (see FIG. 22).

In the above description has described a method for obtaining a homomorphic filter image HM ₁₁ of the diffusion weighted image DW _11, for the other diffusion weighted image, it is possible to obtain a homomorphic filter images in a similar manner. Therefore, a homomorphic filter image can be obtained for each diffusion weighted image. 26 and 27 schematically show a homomorphic filter image. In FIG. 26, a homomorphic filter image at b = b1 is shown, and in FIG. 27, a homomorphic filter image at b = b2 is shown. In FIGS. 26 and 27, only the homomorphic filter image obtained from the diffusion-weighted image of slice 1 is shown due to space constraints in the drawing, but the same applies to the diffusion-weighted images of other slices. A homomorphic filter image is obtained.

As described above, by executing homomorphic filtering using the mask MK1, it is possible to restore the image intensity in the left lobe or stomach of the liver while preventing the contrast from changing in the right lobe of the liver. However, since homomorphic filtering is performed in the region of the pixel value 1 of the mask MK1, the contrast is enhanced while the image intensity is restored in the left lobe or stomach of the liver. Therefore, the homomorphic filter image HM ₁₁ shown in FIG. 25, when used as a restored image image intensity is restored, it is considered an error of ADC becomes larger in the left lobe and the stomach of the liver. Therefore, in this embodiment, after creating a homomorphic filter image in step ST22, steps ST3 to ST6 are executed. In steps ST3 to ST6, using the homomorphic filter image, a restored image having good contrast and restored image intensity is created. Hereinafter, steps ST3 to ST6 will be described. Note that the processing in steps ST3 to ST6 is the same for any homomorphic filter image. Therefore, in the following description, among the homomorphic filter image obtained in step ST22, representatively taken homomorphic filter image _{HM 11,} when executing the steps ST3~ST6 against homomorphic filter image _{HM 11} A method will be described.

At step ST3, the differential image producing means 95 (see FIG. 2), based on the following equation (2) obtains a difference image DIF between the diffusion weighted image _{DW 11} and homomorphic filter image _{HM 11.}
DIF = HM ₁₁ −DW ₁₁ (2)

FIG. 28 shows an example of the difference image DIF. Difference image DIF represents the difference between the pixel values of the diffusion weighted image _{DW 11} and homomorphic filter image _{HM 11.} FIG. 28 shows that the difference image DIF has a large pixel value in the left lobe or stomach of the liver. Therefore, it can be seen that the image intensity is restored in the left lobe and stomach of the liver by performing homomorphic filtering. After obtaining the difference image DIF, the process proceeds to step ST4.

In step ST4, the low-pass filtering means 96 (see FIG. 2) performs low-pass filtering on the mask MK1. FIG. 29 is an explanatory diagram when the mask MK1 is subjected to low-pass filtering. By performing low-pass filtering on the mask MK1, a smooth signal change can be realized at the boundary between the pixel values 1 and 0 of the mask MK1. Hereinafter, the mask MK1 after the low-pass filtering is performed is represented by a code “MK1 _lpf ”. After creating the mask MK1 _lpf , the process proceeds to step ST5.

In step ST5, the low-pass filtering means 96 performs low-pass filtering on the difference image DIF (see FIG. 28). FIG. 30 is an explanatory diagram when low-pass filtering is performed on the difference image DIF. In FIG. 30, the differential image DIF after the low-pass filtering is performed is indicated by a code “DIF _lpf ”. Since the pixel value can be changed smoothly by low-pass filtering the difference image DIF, the contrast enhanced by the homomorphic filtering can be reduced. Compared with the difference image DIF, the difference image DIF _lpf has a smoothly changed pixel value and reduced contrast. After creating the difference image DIF _lpf , the process proceeds to step ST6.

In step ST6, the restored image creating means 97 (see FIG. 2) creates a restored image having a good contrast while restoring the image intensity by using the difference image DIF _lpf . A method for creating a restored image will be described below. In this embodiment, the restored image is created according to the following equation (3).
RS ₁₁ = (1−MK1 _lpf ) * DW ₁₁ + MK1 _lpf * (DW ₁₁ + DIF _lpf )
... (3)
Here, RS ₁₁ : Restored image MK1 _lpf : Mask after low-pass filtering
DW ₁₁ : Diffusion weighted image DIF _lpf : Difference image after low-pass filtering

  Below, Formula (3) is demonstrated. Regarding the expression (3), first, the second term on the right side will be explained first, and then the first term on the right side will be explained.

FIG. 31 is an explanatory diagram of the second term.
When performing the calculation of the second term, the restored image creating unit 97 adds the difference image _{DIF lpf} the diffusion weighted image _{DW 11.} As described above, the contrast enhanced by the homomorphic filtering is reduced in the difference image DIF _lpf compared to the difference image DIF. Therefore, by adding the difference image DIF _lpf to the diffusion weighted image DW ₁₁ , it is possible to obtain an added image (DW ₁₁ + DIF _lpf ) in which the image intensity of the left lobe and stomach of the liver is restored and the contrast is reduced. .

Then, the restored image creation unit 97 multiplies the added image (DW ₁₁ + DIF _lpf ) by the mask MK1 _lpf . By multiplying the mask _{MK1 lpf,} from the addition image _(DW 11 _{+ DIF lpf),} it is possible to take out the image portion D2 corresponding to the region of the pixel value 1 in the mask _{MK1 lpf.} Note that the region of the pixel value 1 of the mask MK1 _lpf includes the left lobe of the liver and the stomach. Therefore, the image portion D2 extracted from the added image (DW ₁₁ + DIF _lpf ) includes the left lobe and stomach of the liver whose image intensity has been restored.

FIG. 32 is an explanatory diagram of the first term.
The first term (1-MK1 _lpf ) represents an inversion mask obtained by inverting the pixel value of the mask MK1 _lpf . Pixel values first mask _{MK1 lpf} is inverted mask _{(1-MK1 lpf)} in pixel value 0, the pixel value 0 in the mask _{MK1 lpf} is an inversion mask _{(1-MK1 lpf)} the pixel value 1.

When performing the calculation of the first term, the restored image creating means 97 multiplies the diffusion weighted image DW ₁₁ by an inversion mask (1-MK1 _lpf ). By multiplying the inverted mask _{(1-MK1 lpf),} can be from the diffusion weighted image _{DW 11,} it takes out the image portion D1 corresponding to the area of the pixel value 1 in the inverted mask _{(1-MK1 lpf).}

Next, the restored image creating means 97 adds the image portion D2 extracted from the added image (DW ₁₁ + DIF _lpf ) and the image portion D1 extracted from the diffusion weighted image DW ₁₁ . FIG. 33 is an explanatory diagram when the image portions D1 and D2 are added. The restored image RS ₁₁ can be obtained by addition.

Since the restored image RS ₁₁ includes the image portion D2 extracted from the added image (DW ₁₁ + DIF _lpf ), the image intensity is restored and the contrast is reduced in the left lobe and stomach of the liver.

In this embodiment, when the image portion D2 is extracted from the added image (DW ₁₁ + DIF _lpf ), a low-pass filtered mask MK1 _lpf is used. Therefore, the pixel value can be smoothly changed at the edge of the image portion D2 extracted from the added image (DW ₁₁ + DIF _lpf ). Furthermore, when taking out the image portion D1 from the diffusion weighted image _{DW 11,} low-pass filtered mask _{MK1 lpf} inversion mask _{(1-MK1 lpf)} is used. Accordingly, in the edge of the image portion D1 taken from diffusion weighted image DW _11, it is possible to smoothly change the pixel value. For this reason, even if the image portions D1 and D2 are added, the artifacts caused by the positional deviation at the boundary between the image portions D1 and D2 can be reduced.

In the above description, the method for obtaining the restored image RS ₁₁ based on the homomorphic filter image HM ₁₁ is described. However, the restored image can be obtained by the same method for other homomorphic filter images. Therefore, a restored image can be obtained for each homomorphic filter image. 34 and 35 schematically show the restored image. 34 shows a restored image at b = b1, and FIG. 35 shows a restored image at b = b2. 34 and 35, only the restored images RS _{11 to} RS _1n and RS _{41 to} RS _4n obtained from the homomorphic filter image of slice 1 are shown due to space constraints in the drawing. Similarly, a restored image can be obtained from the slice homomorphic filter image.

  In the present embodiment, a restored image (see FIG. 33) in which the image intensity of the left lobe of the liver and the stomach is restored can be obtained by using the method of homomorphic filtering (step ST22).

  However, there are portions where the image intensity cannot be sufficiently restored only by performing homomorphic filtering. Referring to FIG. 33, it can be seen that the image intensity of the left lobe and stomach of the liver has been restored, but the image intensity has not been sufficiently restored in the spleen even if homomorphic filtering is performed. Therefore, in this embodiment, step ST7 is executed to restore the image intensity of the spleen.

In step ST7, the composite image creating means 98 (see FIG. 2) synthesizes the restored image.
FIG. 36 is a diagram illustrating an example of a method for combining restored images.
First, the composite image creation means 98 selects the restored images RS _{11 to} RS _1n of slice 1 from the restored images obtained in step ST6. Then, the composite image creating unit 98 synthesizes the restored images RS _{11 to} RS _1n using the following formula.

Expression (4) weights the restored images RS _{11 to} RS _1n by an amplitude square weighted average method, and calculates the average of the weighted restored images RS _{11 to} RS _1n .

The composite image creating means 98 synthesizes the restored images RS _{11 to} RS _1n of slice 1 according to the equation (4). Therefore, it is possible to obtain a composite image DC _1. Figure 37 shows an example of the composite image DC _1.

As described above, since it is difficult to sufficiently restore the image intensity of the spleen even if the homomorphic filtering is executed, the restored images RS _{11 to} RS _1n often have a loss of image intensity in the spleen (see FIG. 33). ). However, when the restored images RS _{11 to} RS _1n are compared, in general, the position of the loss of image intensity that appears in the spleen is different due to the movement of the spleen. Therefore, by synthesizing the restored images RS _{11 to} RS _1n , the loss of image intensity that appears in the spleen of a certain restored image can be restored by another restored image. Accordingly, as shown in FIG. 37, the image intensity of the spleen it is possible to obtain a synthesized image DC ₁ reconstructed. In this embodiment, the amplitude square weighted average method is used. However, if the image intensity can be restored, the restored image may be synthesized by a method different from the amplitude square weighted average method. For example, some of the restored images RS _{11 to} RS _1n are weighted, but the rest of the restored images are not weighted, and the weighted restored image and the unweighted restored image are combined. May be. It may also be synthesized restored image _RS 11 _{to RS 1n} arithmetic average method of the restored image _RS 11 _{to RS 1n.}

In Figure 36, an example has been described for creating a composite image DC ₁ slice 1 in b = b1, for also the composite image of the slice 1 in the b = b2, it is possible to obtain a composite image by the amplitude squared weighted average method . Figure 38 schematically illustrates a combined image DC ₄ slices 1 in b = b2.

In the above description, the method for creating the composite image of slice 1 has been described. However, composite images are similarly created for the other slices. Therefore, a composite image of each slice at b = b1 and a composite image of each slice at b = b2 are created.
After creating the composite image, the process proceeds to step ST8.

In step ST8, the ADC map creating means 99 (see FIG. 2) creates an ADC map for each slice based on the composite image at b = b1 and the composite image at b = b2. For example, an ADC map of slice 1 can be created using composite images DC ₁ and DC ₄ . After creating an ADC map for each slice, the flow ends.

  In this embodiment, since the restored image is synthesized in step ST7, it is possible to restore the image intensity of the spleen that could not be restored sufficiently by homomorphic filtering.

  Further, in this embodiment, homomorphic filtering is performed in the region of the pixel value 1 of the mask MK1, but no homomorphic filtering is performed in the region of the pixel value 0 of the mask MK1. Therefore, it is possible to prevent a change in contrast in a region where homomorphic filtering is not performed.

  Further, in the sequence SC of this embodiment, two b values (b = b1 and b = b2) are used, but one of the two b values may be b = 0. When b = 0 is set, the MPG of the sequence shown in FIG. 7 (or FIG. 12) may not be applied. Also, the number of b values is not limited to two, and an example using only one b value may be executed, or an example using three or more b values may be executed. .

  Next, simulation was performed in order to verify the effect of this embodiment. Below, a simulation result is demonstrated.

FIG. 39 is an explanatory diagram of simulation conditions.
In the simulation, simulation data representing signal values at 25 locations (locations indicated by “+” in FIG. 39) randomly specified for the spin echo EPI image (reference image) was obtained.

Next, simulation data was synthesized. Specifically, the synthesis was performed by the following three methods.
(A) The simulation data was synthesized by the arithmetic mean method.
(B) The simulation data was synthesized by an amplitude square weighted average method.
(C) The homomorphic filtering was performed on the simulation data, and the homomorphic filtered data was synthesized by an amplitude square weighted average method.

FIG. 40 shows a simulation result.
FIG. 40 is a graph showing a normalized root mean squared error of the data synthesized by the above methods (a), (b), and (c). The horizontal axis represents each of the 25 locations specified in the referentless image. The vertical axis represents the normalized mean square error. Referring to FIG. 40, it can be seen that the method (c) shows the smallest error. Therefore, it can be seen that the combination of the homomorphic filtering and the amplitude square weighted averaging method is effective in restoring the image intensity.

  Also, scan A (NEX = 4) with b = 0 and scan B (NEX = 4) with b = 1000 were executed, and an ADC map was created based on the data obtained by both scans A and B. . Specifically, the ADC map was created by the following two methods.

(M1) Four images obtained by scan A were synthesized by an arithmetic mean method to obtain a synthesized image. Similarly, four images obtained by the scan B were synthesized by an arithmetic mean method to obtain a synthesized image. Then, an ADC map was created based on these two composite images.
(M2) Four images obtained by scan A were subjected to homomorphic filtering to obtain four homomorphic filter images HM _{a1 to} HM _a4 . Then, four restored images RS _{a1 to} RS _a4 were created based on the four homomorphic filter images HM _{a1 to} HM _a4 . Next, the four restored images RS _{a1 to} RS _a4 were synthesized by an amplitude square weighted average method to obtain a synthesized image DCa.

Similarly, the four images obtained by the scan B were subjected to homomorphic filtering to obtain four homomorphic filter images HM _{b1 to} HM _b4 . Then, four restored images RS _{b1 to} RS _b4 were created based on the four homomorphic filter images HM _{b1 to} HM _b4 . Next, the four restored images RS _{b1 to} RS _b4 were synthesized by an amplitude square weighted average method to obtain a synthesized image DCb. Then, an ADC map was created based on these two composite images.

FIG. 41 is a diagram showing an ADC map.
FIG. 41A shows an ADC map created using the method (m1), and FIG. 41B shows an ADC map created using the method (m2).

  In FIG. 41A, the ADC value in the left lobe or spleen of the liver is larger than the actual ADC value due to the influence of pulsation. On the other hand, in FIG. 41B, the ADC values of the left lobe and spleen of the liver are smaller than the ADC values of FIG. Therefore, it can be seen that a high-quality ADC map can be obtained by using both the homomorphic filtering and the amplitude-square weighted average method.

2 Magnet 3 Table 3a Cradle 4 Coil 5 Transmitter 6 Gradient magnetic field power supply 7 Receiver 8 Computer 9 Processor 10 Memory 11 Operation unit 12 Display unit 13 Subject 21 Bore 91 Image creation means 92 Mask creation means 93 Modulation map creation means 94 Same-type filter means 95 Difference image creation means 96 Low-pass filter means 97 Restored image creation means 98 Composite image creation means 99 ADC map creation means 100 MR apparatus

Claims (14)

Scanning means for executing a first sequence multiple times for collecting data including diffusion information from a slice crossing an imaging region including a plurality of regions;
First image creating means for creating a plurality of first images including diffusion information based on data collected by executing the first sequence a plurality of times;
First filter means for performing homomorphic filtering for restoring image intensity in a first portion of the first image where loss of image intensity has occurred for each first image, First filter means that does not perform the homomorphic filtering in a second portion of the first image;
Modulation map creating means for creating a modulation map for the homomorphic filtering based on the first image ;
The first filter means includes
A magnetic resonance apparatus that performs the homomorphic filtering using the modulation map. A mask creating means for creating a mask including a first region corresponding to the first portion of the first image and a second region corresponding to the second portion of the first image; And
The first filter means includes
The homomorphic filtering is performed in a portion of the first image corresponding to the first region of the mask, but corresponds to a second region of the mask in the first image. The magnetic resonance apparatus according to claim 1, wherein the homomorphic filtering is not performed in a portion. The mask creating means includes
The magnetic resonance apparatus according to claim 2, wherein a standard deviation map representing a standard deviation of pixel values of the plurality of first images is created, and the mask is created based on the standard deviation map. A difference image creation means for creating a difference image by subtracting the second image obtained by homomorphic filtering of the first image and the first image;
Second filter means for low-pass filtering the difference image;
A second image creation means for creating a first addition image obtained by adding the low-pass filtered difference image and the first image;
The magnetic resonance apparatus according to claim 2 , comprising: The second filter means includes
Low-pass filtering the mask so that pixel values at the boundary between the first region and the second region of the mask change smoothly;
The second image creating means includes
Using a low-pass filtered mask, extracting a first image portion from the first summed image;
Using a reversal mask of the low-pass filtered mask to extract a second image portion from the first image;
The magnetic resonance apparatus according to claim 4 , wherein a third image is created by adding the first image portion and the second image portion. The second image creating means creates a plurality of third images;
The magnetic resonance apparatus according to claim 5 , further comprising third image creation means for creating a composite image by combining the plurality of third images. The third image creating means includes:
The magnetic resonance apparatus according to claim 6 , wherein each of the plurality of third images is weighted, and the composite image is created based on the plurality of weighted third images. The third image creating means includes:
The magnetic resonance apparatus according to claim 7 , wherein each of the plurality of second images is weighted using an amplitude square weighted average method. The magnetic resonance apparatus according to claim 6 , further comprising: an ADC map creating unit that creates an ADC map based on the composite image and a composite image different from the composite image. The scanning means includes
Performing a second sequence for collecting data from the slice multiple times;
The another composite image is
The magnetic resonance apparatus according to claim 9 , wherein the magnetic resonance apparatus is created based on data obtained by executing the second sequence a plurality of times. The first sequence includes a first gradient pulse for performing diffusion emphasis, wherein a b value representing the intensity of diffusion emphasis is set to the first value,
The second sequence includes a second gradient pulse for performing diffusion emphasis, and a second gradient pulse in which a b value indicating the intensity of diffusion emphasis is set to a second value. The magnetic resonance apparatus according to 10 . The first sequence includes a first gradient pulse for performing diffusion emphasis, wherein a b value representing the intensity of diffusion emphasis is set to the first value,
The magnetic resonance apparatus according to claim 10 , wherein the second sequence does not include a gradient pulse for performing diffusion weighting. Said first portion of said first image includes a left lobe and stomach of the liver, said second portion of said first image includes a right lobe of the liver, of claim 1 to 12 The magnetic resonance apparatus as described in any one of them. A program applied to a magnetic resonance apparatus that executes a first sequence multiple times to collect data including diffusion information from a slice that crosses an imaging region including a plurality of regions,
A first image creation process for creating a plurality of first images including diffusion information based on data collected by executing the first sequence a plurality of times;
For each of the first images, a first filtering process for performing homomorphic filtering for restoring image intensity in a first part of the first image where loss of image intensity has occurred, A first filtering process that does not perform the homomorphic filtering in a second part of the first image;
A program for causing a computer to execute a modulation map creating process for creating a modulation map for performing homomorphic filtering based on the first image ,
The first filtering process is:
A program for performing the homomorphic filtering using the modulation map.