JP2002209867A - Method for separating leuco substance, ectocinerea and cerebrospinal fluid from brain magnetic resonance image and calculating volume - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a method for separating leuco substance, ectocinerea and cerebrospinal fluid from a brain image which is provided from a magnetic resonance imaging instrument and for calculating each volume. SOLUTION: In the method, the leuco substance, the ectocinerea and the cerebrospinal fluid are separated from a proton density image photographed concerning the same transection of a brain part and from a T2 emphasis image among slices generated by the magnetic resonance imaging instrument and the each volume is calculated. The method is characterized by comprising a first process for calculating the number of pixels which constitute respective components from a whole part in consideration of the rate of the two mutually different components, a second process for deciding a judgement value by which two parts are separated in an area where the two components are superimposed and blurred based on the number of pixels calculated in the first process, and a third process for obtaining the volume of each component from the image separated by the judgement value of the second process.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

The present invention relates to a brain image processing method, and more particularly, to a magnetic resonance imaging apparatus.
The present invention relates to a method for separating white matter, gray matter, and cerebrospinal fluid from a brain image provided by (MR) and calculating respective volumes.

[0002]

2. Description of the Related Art Medical research on the structure and role of the brain, which supervises all biological functions of humans, has already been carried out to a considerable extent, and medical technology for the discovery and cure of various brain diseases has been developed. Many progress has been made. Imaging diagnosis using computerized tomography equipment (CT) and magnetic resonance imaging equipment (MR) has played a very important role in the development of such medical technology. We expect that the diagnostic technology will continue to evolve. But,
In the case of degenerative brain disease with cerebral contraction, such as Alzheimer's disease or cerebral palsy, the presence or absence of the disease only depends on the subjective judgment of the diagnostic radiologist using magnetic resonance imaging with the naked eye. Atrophy of white matter and gray matter in the cerebral spinal fluid
A method for quantitatively calculating the degree of increase in (CSF) and for early diagnosis of the occurrence of disease based on objective criteria such as examining the degree of atrophy of each tissue has not been developed.

[0003] Atrophy of white matter and gray matter and the resulting increase in cerebrospinal fluid are the most common phenomena exhibited by patients with degenerative brain disease. Tracking the changes in the disease and examining the degree of atrophy in comparison with the volume of a normal person can be utilized as the most basic and objective diagnostic method for diagnosing the occurrence of disease. However, due to the complex structure of the brain and the physical properties of magnetic resonance imaging, a certain thickness (usually 3-5 mm)
When a slice is taken and expressed as a single brightness value at each position, there is a problem of partial volume (artifact) depending on the size of the pixel (voxel), which causes the brightness phenomenon to be blurred To date, no method has yet been developed that can separate white matter, gray matter, and cerebrospinal fluid from magnetic resonance images in consideration of partial volumes. Therefore, a method capable of quantitatively calculating the volumes of white matter, gray matter, and cerebrospinal fluid constituting the human cerebrum has not been developed until now. However, some research teams use elementary research such as separating the whole brain region from the magnetic resonance image of the brain and semi-automated methods by the user's manual operation on T1 weighted images (T1 weighted image). Attempts have been made to separate white matter and gray matter areas.

[0004] However, since such a method also targets a T1-weighted image in which white matter and gray matter are not well compared, accurate separation of each tissue is not achieved, and manual operation by a user is required. It has the problem of inefficiency in processing time and the problem that cerebrospinal fluid cannot be processed separately.

[0005]

SUMMARY OF THE INVENTION It is an object of the present invention to provide a method for separating white matter, gray matter and cerebrospinal fluid from a brain image provided by a magnetic resonance imaging apparatus and calculating the volume thereof. .

It is still another object of the present invention to provide a method for determining light and dark values of white matter, gray matter and cerebrospinal fluid from a given brain image without absolute criteria, and determining a discriminant value that can separate each of them. Is to do.

Still another object of the present invention is to calculate the volume of white matter, gray matter, and cerebrospinal fluid in a patient who has a possibility of degenerative brain disease, compare the volume with that of a normal person, and follow the change in volume. And to provide a method for early prediction and diagnosis of brain disease.

[0008]

In order to achieve the above object, an image processing method according to the present invention uses the same axial section of the cerebrum in a slice made by a magnetic resonance imaging apparatus. Proton density taken for
(proton density) A method of separating white matter, gray matter and cerebrospinal fluid from an image and a T2-weighted image and calculating the volume of the matter; considering the ratio of two different components to each other, the number of pixels constituting each component from the whole is determined. A first step of calculating, a second step of determining a discriminant value that can separate two parts in a blurred area where two components overlap based on the number of pixels calculated in the first step, and the second step And a third step of obtaining the volume of each component from the image separated by the discrimination value according to (1).

A detailed feature of the method for separating and calculating the volume of white matter, gray matter and cerebrospinal fluid according to the present invention is the proton density (P
D) removing the brain-excluding portion from the brain image of the video, removing the same region from the corresponding T2 video, and removing the cerebrospinal cord from the T2 video after the first process is completed. A second step of separating the cerebrospinal cord of the same region from the proton density (PD) image, and a partial volume of each component from the proton density (PD) image consisting of only white matter and gray matter by separating the cerebrospinal cord by the second step A third step of calculating the volumes of the white matter and the gray matter using the information on the partial volumes of the white matter and the gray matter calculated by the third step and the slice information provided from the tomography apparatus. It includes four steps.

[0010]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS The configuration and operation of the present invention will be described below in detail. FIG. 1 is a control block diagram showing a control and processing relationship with an apparatus and the like related to magnetic resonance imaging according to the present invention.

An image file provided from a magnetic resonance (MR) image device 1 is transmitted to an image processing unit 3 via an interface device 2. At this time, the image processing unit 3 stores the transmitted image information in the main storage device 4 and loads the program for white matter and gray matter separation prepared for the present invention stored in the auxiliary storage device 5. I do. After the pre-processing process, the separated white matter and gray matter images are output through the image output device 6 through the determination of the discrimination value for the separation of white matter and gray matter, and the volumes of the simultaneously calculated white matter and gray matter are calculated. Is stored in the video database. The image database maintains information on the calculated white matter and gray matter images and the calculated volume, and information on the gender, age, name, etc. of the person (normal person or patient) corresponding to the handled image. I have.

[0012] Examples of the brain imaging apparatus include a computer tomography apparatus and a magnetic resonance imaging apparatus. The image file processed by the present invention means an image file obtained from the magnetic resonance apparatus. The extracted video output device has a display function so that a user can recognize the processed video information, such as a printer or a monitor.

The size of one image slice for the brain provided by the magnetic resonance imaging apparatus is 265 × 256 size, and includes an image in which each pixel can be expressed by a light-dark value from 0 to 4095. One pixel consists of 2 bytes and one of the magnetic resonance imaging slices of the brain is 256 × 256
× 2 = 128 Kbytes. In addition, the header of this file has various types of information ranging from information on the structure and format of the image to information on the patient's body.

The present invention further converts such image slice information into an 8-bit PGM (Portable Gray Map) format file, and then uses it for separating white matter and gray matter. In this process, each pixel of the image file is expressed as an image having 8-bit brightness values from 0 to 255.

The image slices used in the present invention for the separation and volume calculation of white matter and gray matter and cerebrospinal fluid are based on the same cross-section of the cerebral part in the whole slice created by the magnetic resonance imaging apparatus. A proton density image and a T2-weighted image. Each of these two types of video slices is formed inside and outside 15 slices.

The cross section of the cerebrum is photographed with a thickness of 5 mm each while rising upward from the lower end (eye level) in a cross section.
Video slices generated with the compressed light and dark values will look different depending on the position. In the lower slices, the cerebellum, part of the eyes, the brainstem, etc. are represented along with the cerebrum with a relatively small area, and at the middle position, the image size is the largest and it has an elliptical shape. The image has a butterfly-shaped ventricle in the center of the image. Then, the size of the image becomes gradually smaller and elliptical toward the upper end.

FIG. 2 is a proton density image slice, and FIG.
Is a T2-weighted video slice. The T2-weighted image and the proton density image taken for the same cross section have different characteristics in the distribution of light and dark values. The T2-weighted image is generally darker than the proton density image, but the cerebrospinal part is relatively brighter than the proton density image, so it is easy to extract the cerebrospinal cord. used. However, the white matter and the gray matter are relatively darker than the proton density image, and it is extremely difficult to distinguish the two components.
On the other hand, in the proton density image, it is not easy to distinguish between the cerebrospinal part and the white matter part and the gray matter part are easier than the T2-weighted image. Will be

In the present invention, the remaining part which does not correspond to the brain region before the separation of white matter and gray matter is first removed.
In this process, first, after removing the black background of the image in the proton density image, the inner region of the brain is separated from the outer skin and the fat layer surrounding the inner region of the brain. The removal of the black background of the image is performed as follows. The initial proton density image has a black background with a brightness value close to 0 together with the elliptical brain image.

FIG. 4 is a histogram for the eighth image. Pixels forming a high peak between light and dark values 0 to 10 are pixels forming a black background of the brain image and a fat layer surrounding the inside of the brain. In addition, pixels having a light-dark value of 90 or more are pixels constituting the outer skin and the inside of the brain. In the present invention, a black background of an image is removed by using such a difference between light and dark values. That is, starting from an arbitrary pixel (0,0) constituting the black background of the image, pixels corresponding to a brightness value of 90 or less are defined as 4 pixels.
-After tracking by 4-connectivity, the tracked pixels were removed. On the other hand, a fat layer having a brightness value of 90 or less that surrounds the inside of the brain in the outer skin of the brain is not removed because it is not connected to the pixels constituting the black background.

FIG. 5 shows a proton density image with the background removed. Next, after the black background of the image is removed as shown in FIG. 5, the inside of the brain is separated from the outer skin and fat layer surrounding the inside of the brain. However, in the present invention, the distribution of the light and dark values of the proton density image was adjusted before separation into the inner region of the brain.

In a magnetic resonance image of the brain, the distribution of light and dark values of pixels constituting the image appears differently for each individual, and even for the same person, it may appear differently for each slice.
Therefore, the light and dark values of a portion constituting white matter and gray matter in the brain may be expressed differently for each slice. As can be seen from FIG. 5, in the case of a partial slice photographed for the upper end of the brain, the brightness of the region inside the brain appears brighter than the other slices. In the case of such a partial slice for the upper end, since the contrast between the inside of the brain and the fat layer surrounding it is not clear, it is not possible to accurately extract the brain region, and it is also difficult to separate the white matter and gray matter regions. In some cases, the determination value used for the determination is not accurately determined. Therefore, the distribution of the light and dark values was adjusted so that the distribution of the light and dark values inside the brain became similar to each other for all slices.

According to the present invention, when the average brightness in the inner region of the brain where white matter and gray matter are mixed is adjusted to 120 through observation of various magnetic resonance images, the inside of the brain and the fat layer surrounding the same are adjusted. It was found that the contrast between the light and dark values became clear, the brain was easily extracted, and finally, the discrimination value for separating white matter and gray matter could be stably determined.

FIG. 6 shows a proton density image in which the brightness value is adjusted, and it can be seen that a part of an image at the lower end is slightly darker than before, so that all the proton density images are similar. It will have a light-dark distribution.

After adjusting the light and dark values of the proton density (PD) image, the inside of the brain is separated and extracted from the outer skin and the fat layer. After the adjustment of the light and dark values, the inside of the brain will have an overall light and dark value of 90 or more, and is surrounded by an elliptical fat layer having a light and dark value of 90 or less. Therefore, from some locations that fall inside the brain
If pixels with light and dark values of 90 or more are tracked by 4-connectivity, only the fat layer and the outer skin of the brain will remain. At this time, some pixels existing within the brain, which correspond to a brightness value of 90 or less, are excluded. However, even if such a portion has a brightness value of 90 or less, it is a portion existing in the inner region of the brain and must be separated as being regarded as the inner region of the brain. Therefore, if one pixel constituting the outer skin of the brain is further tracked by adjacent pixels by 4-connectivity, the outer skin excluding the inner region of the brain and the fat layer can be tracked to a connected area. When the extracted region is separated from the background-removed image as shown in FIG. 4 in which the background is removed, a region corresponding to the inside of the brain can be separated from the image of the entire brain.

FIG. 7 shows a state in which pixels having light and dark values of 90 or more are tracked and removed after such a process. FIG. The region is separated from the image of the whole brain.

On the other hand, a portion corresponding to the inside of the brain can be separated from the corresponding T2-weighted image using the proton density image in which the internal region is separated as shown in FIG. FIG. 9 shows a T2-weighted image in which the inner region is separated.

The extraction of white matter, gray matter, and cerebrospinal fluid and the calculation of the volume of cerebrospinal fluid which are not performed in the present invention are performed on the image of the inner region of the brain separated from the proton density image as shown in FIG. The image of the inner region of the brain separated from the T2-weighted image is used for cerebrospinal fluid extraction and analysis.

In the present invention, each volume calculation through the separation of white matter, gray matter, and cerebrospinal fluid, which is finally performed, is performed on the inner region of the brain shown in the image of FIG. The most important point in the separation of white matter and gray matter and cerebrospinal fluid is to separate white matter and gray matter by taking into account the partial volume of white matter and gray matter through interpretation of blurred light and dark values.

The problem of blurring of light and dark values due to the problem of the partial volume generally occurs depending on the size of a voxel of an image. Normally, a magnetic resonance imaging device (MR) scans a human body to a certain thickness (3 mm to 10 mm) to generate an image slice, and when scanning at a universal thickness of 5 mm, 13 About 16 video slices are generated. At this time, when the thickness of the scan is small, the size of the pixel is relatively small, and when the thickness of the scan is large, the size of the pixel is also increased. If the smaller pixel contains only fat and / or moisture, the larger pixel will contain all of the two components, and therefore the amount of fat and water contained in the larger pixel will be reduced. It will have the same signal strength as the average weight. As a result, the brightness value is determined by the arithmetic mean of the signal intensities generated from the pixels constituting a more constant thickness, and then determined as the brightness value of one pixel in a two-dimensional slice.

In FIG. 10, reference numeral 11 denotes a section having a constant thickness which passes through for photographing a human body, and reference numeral 12 denotes a pixel constituting a thickness scanned from any one axis. . After the arithmetic average value of the signal intensities emitted from each pixel is expressed as one light-dark value, it is determined as the light-dark value of one pixel forming a two-dimensional cross section as shown in FIG. .

Most of the light / dark blurring phenomena shown in the magnetic resonance images taken on the upper surface of the cerebral axis have different amounts of hydrogen atoms within a certain thickness through which the magnetic field passes. It is manifested by the presence of white matter and gray matter or cerebrospinal fluid together. That is, different tissues present in a constant thickness will each emit different amounts of energy during the relaxation time of the nucleus,
Depending on the amount of energy released, it is determined as a different light and dark value. At this time, the magnetic resonance imaging apparatus takes an arithmetic average of the light and dark values based on the signal intensity emitted from each pixel determined within a certain thickness, and generates a single light and dark value. As a result, a two-dimensional magnetic resonance image shows a blur phenomenon of light and dark values. FIG. 11 (a) is a proton density image of the upper surface of the brain taken at a thickness of 3 mm, and FIG. 11 (b) is a proton density image taken of a 10 mm thickness. As already mentioned, in case (b),
As the size of the pixel increases, the problem of the partial volume becomes more pronounced, so that the phenomenon of light and dark values becomes more blurred as compared with (a), and the resolution of the image decreases.

In the present invention, based on such a principle of generating a magnetic resonance image, when an attempt is made to separate two components from a proton density image showing a blur phenomenon of light and dark values, it appears in a region showing the blur phenomenon. A method is used to determine the appropriate light / dark value as the discriminant value from the light / dark values that are present and separate the two components.After separation, the part showing the blur phenomenon is treated as a specific component depending on each light / dark. Was.

FIG. 12 (a) shows that two components having different light and dark values exist in a constant thickness, and FIG. 12 (b) shows such a cross section obtained by a magnetic resonance imaging apparatus. After the image is captured, the image is shown in a blurred state depending on the ratio of each component included. (C) shows a state in which the video of (b) is separated by an appropriate discrimination value. At this time, the state of the original cross section such as the picture (a) is not known from the separated video such as (c), but the amount of each component is calculated in reverse from the brightness value expressed at each position. Will be.

In the present invention, a component having a light-dark value of y1 within a certain thickness is z%, and a component having a light-dark value of y2 is (100-z) based on the above-mentioned facts. % Is calculated by the following formula based on the principle that it is generated by an arithmetic average based on the ratio of two light and dark values, and finally, from the light and dark value G that appears due to blur development. Conversely, the content ratio of each component was calculated.

[0035]

(Equation 7) Therefore, on the contrary, the partial volume of each component included in the original thickness is calculated from the brightness value G indicating the blur phenomenon.

That is, the partial volume of the specific component can be inferred from the above equation as follows. If the brightness of the A component is y
If the light and dark value of the B component is y2, and if such a component is mixed in a certain thickness to show the light and dark value of G, then G
A component is z% and B component is
It could be calculated that (1−z)%.

By such a method, it is possible to find out the number of pixels constituting each component from the whole in consideration of the ratio of the A or B component in the image compressed as shown in FIG. Based on the number of such pixels, a light-dark value that can separate two parts in a blurred area where two components overlap is determined. Such a light-dark value is called a discrimination value, and FIG. 12C shows an image separated by the determined discrimination value.

If (a) in FIG. 12 is 5 mm in the actual brain,
Referring to FIG. 12, (b) is an image generated by a magnetic resonance imaging apparatus. Here, the actual state of the brain as shown in FIG. 12 (a) cannot be known from the video as shown in FIG. 12 (b).
However, after calculating the content ratio of each component from the light and dark values expressed at each position through the analysis of the light and dark values proposed in the present invention, the discriminant value is determined using the calculated ratio, and the determined discriminant value is used. An image as shown in FIG. 12 (b) is separated like an image as shown in FIG. 12 (c). Similarly, the original state as shown in FIG. 12 (a) cannot be known from the image as shown in FIG. 12 (c), but the pixels constituting each component shown in FIG. The number is expressed in consideration of the partial volume of each component in FIG. 12A which is an actual cross section, and the volume of each component can be accurately calculated based on this.

Hereinafter, the algorithm up to now will be described in detail with respect to the image generated by the magnetic resonance apparatus for the human brain used in the present invention. On the other hand, the video processed by applying the algorithm described below is shown only for the third, eighth, and twelfth video in the video of FIG. 8, and the remaining video slices are processed by the same process. You.

In the present invention, cerebrospinal fluid was first extracted using T2-weighted images. The T2-weighted image is generally darker than the proton density image, but the cerebrospinal fluid portion is brightly enhanced. Therefore, after cerebrospinal fluid is first extracted from the T2-weighted image, the pixel at the same position as the cerebrospinal fluid extracted from the T2-weighted image for the corresponding proton density image is extracted as the cerebrospinal fluid of the proton-density image Will be.

In a T2-weighted image, the distribution of light and dark values of pixels corresponding to cerebrospinal fluid may differ from slice to slice. Therefore, when simply calculating pixels having a brightness value equal to or higher than the light-dark value specified by the general observation value as cerebrospinal fluid, accurate extraction is not performed because the calculation of cerebrospinal fluid has different characteristics for each slice. Sometimes. Thus, in the present invention, the Gaussian distribution curve is used to determine the lower limit of the light and dark value of the cerebrospinal fluid pixel regardless of the light and dark value distribution of each slice for each slice. That is, after obtaining a histogram for the T2-weighted image of the brain internal region, the largest light-dark value among the light-dark values overlapping the Gaussian distribution curve was determined as the lower limit of the light-dark value of the pixel corresponding to the cerebrospinal fluid. FIG. 13 shows a state in which the histogram generated for the eighth image overlaps the Gaussian distribution curve. In this case, the light and dark value 112 was determined as the lower limit of the cerebrospinal fluid.

FIG. 14 shows the third, eighth, and twelfth images among the T2-weighted images from which the cerebrospinal fluid was extracted. FIG. 15 shows the T2-weighted images shown in FIG. 14 from the corresponding proton density images. 5 is an image in which pixels at the same position as the cerebrospinal fluid portion are extracted and removed as cerebrospinal fluid. On the other hand, cerebrospinal fluid extracted from the proton density image of FIG. 15 is shown in FIG.

The cerebrospinal fluid extracted from the proton density image is T2
It has a distribution of light and dark values different from that of the emphasized image. When only pure cerebrospinal fluid is present in a certain thickness, it is expressed relatively bright, but when mixed with white or gray matter, it is expressed in proportion to these amounts and darkened. ing. Figure
The histogram of 17 shows the relationship between the cerebrospinal fluid extracted from the T2-weighted image shown in FIG. 14 and the cerebrospinal fluid extracted from the proton density image as shown in FIG.

The pixels extracted as cerebrospinal fluid from the T2-weighted image are widely distributed from A to C in the proton density image.
Among them, the pixels corresponding to the light and dark values of B or more correspond to pure cerebrospinal fluid, and the pixels corresponding to the light and dark values of A to B are mixed with white matter or gray matter. From these pixels, the partial volume of the cerebrospinal fluid must be calculated. FIG. 18 shows a proton density image from which cerebrospinal fluid has been removed and a histogram for the extracted cerebrospinal fluid. In FIG. 18, a region A is a histogram for a proton density image from which cerebrospinal fluid has been removed. In this part, the partial volumes of white matter and gray matter are also calculated. Area B
And C are histograms of cerebrospinal fluid extracted from the proton density image, where C corresponds to pure cerebrospinal fluid and B is the portion of cerebrospinal fluid mixed with white matter and gray matter. Here, y1 is regarded as the lower limit of the light and dark value that white matter can have, and y3 is regarded as the upper limit of the light and dark value that gray matter can have and at the same time as the lower limit of the light and dark value that pure cerebrospinal fluid can have.

The distribution function of the histogram for the area A is H
H (G) is the number of pixels corresponding to the light and dark value G.
Here, the component having the brightness value of y1 in a certain thickness is z%,
When a component having a light-dark value of y3 is included (100-z)%,
The light-dark value G expressed by the blurring effect is as follows.

[0046]

(Equation 8) If the light-dark value G indicates a light-dark value corresponding to an intermediate point between y1 and y3, 50% of the pixels corresponding to such light-dark values correspond to white matter, and the remaining 50% corresponds to gray matter. It can be seen as the corresponding component. Therefore, the number of pixels corresponding to the partial volume of white matter in the area A is calculated as follows, and z is applied to a different ratio according to each light-dark value. That is,
It applies to 100% for y1 values and 0% for y3 values. Therefore, the number of pixels corresponding to the partial volume of white matter in the area A is as follows.

[0047]

(Equation 9) The partial volume of white matter must also be calculated from region B. However, such a portion is mixed with the cerebrospinal fluid, and pixels corresponding to a partial volume of the cerebrospinal fluid must be removed first. When W is a distribution function for the regions B and C, W (G) is the number of pixels corresponding to the brightness value G. Here, first, pixels corresponding to a light-dark value of y3 or more are extracted as pure cerebrospinal fluid, and the number of pixels corresponding to the extracted pure cerebrospinal fluid is as follows.

[0048]

(Equation 10) Next, pixels corresponding to the partial volume of the cerebrospinal fluid are extracted and removed from each of the light and dark values of the region B. Youth, y3 in a certain thickness
The light and dark values G appearing when the cerebrospinal fluid having a light and dark value of z% and the white matter and gray matter components having a light and dark value of y2 (100-z)% are contained are as follows.

[0049]

[Equation 11] At this time, the number of pixels corresponding to the partial volume of the cerebrospinal fluid in each of the light and dark values y2 to y3 of the region B is as follows.

[0050]

(Equation 12) Therefore, the sum PCSF of PCSF1 and PCSF2 is the number of pixels corresponding to the entire cerebrospinal fluid in a given proton density image. Now, the number of pixels corresponding to the partial volumes of white matter and gray matter can be calculated from the area B. When W ′ is the distribution function of the histogram for the remaining portion after removing pixels corresponding to the partial volume of cerebrospinal fluid from region B, W ′
(G) is the number of pixels corresponding to the brightness value G in the area B. At this time, the number of pixels corresponding to the white matter partial volume is as follows.

[0051]

(Equation 13) The sum PV of PV1 and PV2 is the number of pixels corresponding to white matter in a given PD image. At this time, the number of pixels corresponding to gray matter can be calculated as follows.

[0052]

[Equation 14] FIG. 19 is a histogram showing the process of calculating the white matter partial volume described above.

After removing pure and cerebrospinal fluid corresponding to a partial volume from a given proton density image, a light / dark value for separating white matter and gray matter will be determined.

When T is a histogram for a proton density image from which the influence of the cerebrospinal cord has been removed, the number of pixels at each light-dark value is accumulated from y1, which is the lower limit of the light-dark value for white matter components, and the number of accumulated pixels is calculated. Determined the light-dark value t that includes the sum of PV1 and PV2 for the first time as the discrimination value.

[0055]

(Equation 15) FIG. 20 is a histogram showing the determination of such a discrimination value, and FIG. 21 is a proton density image in which each component is finally separated based on the determined discrimination value. The brightest part of the brain is the cerebrospinal fluid, the darkest part is white matter and the rest is gray matter.

After white matter, gray matter and cerebrospinal fluid are separated from each slice, the volume of white matter and gray matter can be calculated using the number of pixels constituting each separated component. At this time, among the information stored in the header of the magnetic resonance image, information on the size of each pixel and the interval between slices is used, and in the present invention, the volume of each white matter and gray matter and cerebrospinal fluid is calculated by the following equation. Was calculated.

[0057]

(Equation 16) N: number of processed magnetic resonance imaging slices Si: slice number Wp: number of pixels constituting the extracted component X, Y: horizontal and vertical length of one pixel

The partial volumes of white matter, gray matter, and cerebrospinal fluid contained in a certain thickness are calculated from the images generated by the magnetic resonance imaging apparatus of the brain through light / dark value analysis. A method for determining discriminant values that can separate white matter, gray matter, and cerebrospinal fluid from each image slice based on the above is described. In addition, a method for calculating the volume of white matter and gray matter using the number of pixels constituting the separated white matter and gray matter and the cerebrospinal fluid and the information in the header of the image was also presented. Therefore, using a series of methods proposed in the present invention, first calculate the volume of white matter and gray matter and cerebrospinal fluid for normal humans for each age and then obtain statistics, then the possibility of degenerative brain disease Calculate the volume of white matter, gray matter, and cerebrospinal fluid for certain patients and compare them to normal subjects, and track changes in volume to help with early-stage prediction and diagnosis of such brain diseases. It is determined that it can be used.

[Brief description of the drawings]

FIG. 1 is a control block diagram showing a configuration of a brain image processing device according to the present invention.

FIG. 2 is a diagram showing a PD slice used for white matter and gray matter extraction.

FIG. 3 is a diagram showing a T2-weighted image slice used for cerebrospinal fluid extraction.

FIG. 4 is a diagram illustrating a histogram of an initial proton density image.

FIG. 5 is a diagram illustrating a proton density image slice from which black characters in the background have been removed;

FIG. 6 is a diagram illustrating a proton density image slice in which the brightness value is adjusted.

FIG. 7 is a diagram showing a proton density image slice in which the inside of the brain is tracked.

FIG. 8 is a diagram showing a proton density image slice in which the inside of the brain is finally separated.

FIG. 9 is a diagram showing a T2-weighted image slice in which the inside of the brain is separated.

FIG. 10 is an exemplary view showing a process of calculating an average light-dark value of each pixel at a position having a predetermined thickness and different signal intensities.

FIG. 11a shows a magnetic resonance imaging slice of the top of the axis taken at a thickness of 3 mm.

FIG. 11b shows a magnetic resonance imaging slice of the top of the axis taken at a thickness of 10 mm.

FIG. 12 is a view showing an example of image separation based on discrimination values in a blur region of two components different from each other.

FIG. 13 is a diagram illustrating a determination example of a lower limit value of cerebrospinal fluid brightness of a T2-weighted image.

FIG. 14 is a diagram showing a T2-weighted image slice from which cerebrospinal fluid has been removed.

FIG. 15 shows a proton density image slice from which cerebrospinal fluid has been removed.

FIG. 16 is a diagram showing an image slice of cerebrospinal fluid extracted from a proton density image.

FIG. 17 is an exemplary view showing the correspondence between the cerebrospinal fluid of the T2-weighted image and the cerebrospinal fluid of the proton density image.

FIG. 18 is a diagram illustrating a histogram of a proton density image.

FIG. 19 is a diagram illustrating a histogram used for calculating a partial volume of white matter.

FIG. 20 is a diagram showing a histogram for determining a discrimination value.

FIG. 21 shows proton density image slices separated into white matter, gray matter, and cerebrospinal fluid.

[Explanation of symbols]

 REFERENCE SIGNS LIST 1 MR image processing device 2 interface device 3 image processing unit 4 main storage device 5 auxiliary storage device 6 image output device

────────────────────────────────────────────────── ───

[Procedure amendment]

[Submission date] July 17, 2001 (2001.7.1)
7)

[Procedure amendment 1]

[Document name to be amended] Statement

[Correction target item name] Claims

[Correction method] Change

[Correction contents]

[Claims]

Wherein said third process, first removes the black from the brain images of human proton density image
And 3-1 stage, the proton density image to have a distribution profile of the internal and brightness values as comparison of brightness values of the fat layer becomes clear surrounded this is similar to each other in the brain for all slices and the 3-2 step to adjust the brightness value distribution, the interior of the brain was separated from the skin and fat layer regulating brightness value surrounds the interior of the brain from the performed protons density image by the first 3-2 step Stage 3-3, which is extracted by using the proton density image from which the inner region is separated, and the corresponding T2
Part 3 to separate the part corresponding to the inside of the brain from the emphasized image
The human of claim 1 , comprising four steps.
Of white matter and gray matter and cerebrospinal fluid from magnetic resonance images of human brain and calculation of volume.

3. The method according to claim 1, wherein the fourth step is performed according to a histogram for a T2-weighted image of an inner region of the brain.
Of the light and dark values that overlap the curved Gaussian distribution curve
2. The separation and volume calculation of white matter, gray matter and cerebrospinal fluid from a magnetic resonance image of the brain according to claim 1 , wherein a value is obtained and determined as a lower limit of a light and dark value of a pixel corresponding to the cerebrospinal fluid. Method.

Wherein said fifth process, first the white matter and the proton density, which contains only gray matter (PD) video
A second 5-1 generating a histogram (H), a first histogram generated by the first 5-1 stage
Generating a second 5-2 calculating a partial volume of the white matter (PV1) from (H), of the definitive cerebrospinal image to the fourth step, the second histogram from the remaining image by removing the pure cerebrospinal to a second 5-3 phase, the and the 5-4 calculating a partial volume of pure cerebrospinal second histogram generated by the 5-3 stage, a region where the pure cerebrospinal portion is removed and the 5-5 calculating a partial volume (PV2) occupied by white matter from the from the entire pixel value of the first histogram second 5-
By subtracting two stages and the sum of the parts by volume of the white matter, which is calculated by the 5-5 stage (PV1 + PV2), includes a first 5-6 calculating a partial volume of the gray matter, the first histogram and the second histogram a second 5-7 generating a third histogram, while increasing the brightness value in the third histogram, among which calculates an integrated value of the distribution function, than the sum of the partial volumes of the value the white matter (PV1 + PV2) Step 5-8 in which the first light and dark value that becomes larger is set as the discrimination value, and an image in which white matter and gray matter are separated for each slice with reference to the discrimination value set in step 5-8. white matter and gray matter from the magnetic resonance imaging of the brain according to claim 1, characterized in that it comprises a second 5-9 step of generating
And volume calculation method of cerebrospinal fluid .

Wherein said first 5-2 step, the whole cerebrospinal are removed, a histogram of PD image containing only white matter and gray matter to produce H in slice of a certain thickness, pure extracting the brightness value y 2 of pure gray matter and brightness values y1 white matter, z% is white matter having a brightness value of y1 in a predetermined thickness, the gray matter component having a brightness value of y3 (100- z)%
The light and dark value G that appears when included

(Equation 1) Using the method of calculating z% from the number of pixels corresponding to the white matter partial volume from such a light and dark value, the sum of the number of pixels corresponding to the white matter partial volume from the light and dark values y1 to y3 is calculated from the formula

(Equation 2) 5. The method according to claim 4 , wherein the calculation is performed by using a formula, wherein white matter, gray matter, and cerebrospinal fluid are calculated from the magnetic resonance image of the brain.

Wherein said first 5-4 stage, the fifth proton density produced by a process (PD) pure cerebrospinal from the image is removed, the histogram W for cerebrospinal images are mixed with white matter and gray matter in a dark value y3 of cerebrospinal, z% cerebrospinal having brightness values of y 3 extracts the brightness value y2 of the other components, ingredients not cerebrospinal having brightness values of y2 (100-z) G value that appears when% is included
To

(Equation 3) The sum of the number of pixels corresponding to the partial volume of the cerebrospinal cord from the light and dark values y2 to y3 is calculated by the formula

(Equation 4) 5. The method according to claim 4 , wherein the calculation is performed by using a magnetic resonance image of the brain, wherein white matter, gray matter, and cerebrospinal fluid are calculated from the magnetic resonance image.

Wherein said first 5-out, after removal of the pixels corresponding to partial volume of pure No脊marrow of W ', when the distribution function of the histogram for the remaining portion, the partial volume of the white matter

(Equation 5) 5. The method according to claim 4 , wherein the calculation is performed by using a formula, wherein white matter, gray matter, and cerebrospinal fluid are calculated from the magnetic resonance image of the brain.

Total volume of 8. white matter by overall image slice calculated by the sixth step (or gray matter) the extracted white matter number of slices containing the (or gray matter) is N, an interval of the slice D And the number of pixels constituting the white matter (or gray matter) extracted from the i-th slice is Lp
When the horizontal length of one pixel is X, the vertical length of one pixel is Y, and the slice number is Si,

(Equation 6) 2. The method for separating and calculating white matter, gray matter and cerebrospinal fluid from a magnetic resonance image of the brain according to claim 1 , wherein the volume is calculated from the following equation.

 ──────────────────────────────────────────────────続 き Continued on the front page F term (reference) 4C096 AA04 AA05 AA20 AB01 AB08 AB50 AC01 AD14 DC06 DC18 DC19 DC21 DC24 DC35 5B057 AA09 BA07 CA08 CA12 CA16 CB08 CB12 CC01 CE03 CE09 CH08 DA08 DB02 DB09 DC16 DC23 DC36 5L0913 AA06 BA DA01 EA35 FA14 FA35 FA43 FA54 GA09

Claims (9)

[Claims] 1. A proton density image taken for the same axial section of a cerebral part of a slice made by a magnetic resonance imaging apparatus and a T2 weighted image.
mage) a method of separating white matter, gray matter and cerebrospinal fluid from an image and calculating a volume; a first step of calculating the number of pixels constituting each component from the whole in consideration of a ratio of two components different from each other; A second step of determining a discrimination value that can separate two parts in a blurred area where two components overlap based on the number of pixels calculated in the first step; and a video separated by the discrimination value in the second step. A method for separating white matter, gray matter, and cerebrospinal fluid from a magnetic resonance image of a brain and calculating a volume, the method comprising a third step of determining the volume of each component. 2. A first step of removing a portion excluding the brain from a brain image of a proton density (PD) image and removing the same region from a corresponding T2 image, and a step of removing a brain from the T2 image after the first step is completed. Removing the spinal cord and separating the cerebrospinal cord of the same region from the corresponding proton density (PD) image; and separating the cerebral spinal cord number by the second step, the proton density consisting of only white matter and gray matter ( PD) a third step of calculating the partial volume of each component from the image, white matter and white matter using the slice information provided from the tomography apparatus and information on the partial volume of white matter and gray matter calculated in the third step A method of separating white matter and gray matter from a brain magnetic resonance image and calculating a volume, comprising a fourth step of calculating each volume of gray matter. 3. The first process comprises: a first step of removing a black background from a brain image of a proton density image; and a contrast between a brightness value of a fat layer surrounding the brain and a brightness value of a fat layer surrounding the brain for all slices. A second step of adjusting the distribution of light and dark values of the proton density image so that the distribution of light and dark values is similar to each other; The third step of extracting and extracting the inside of the brain from the outer skin and fat layer surrounding the inside of the, and the corresponding T2 using the proton density image from which the inner region has been separated
3. The method according to claim 2, further comprising a fourth step of separating a portion corresponding to the inside of the brain from the emphasized image. 4. The method according to claim 2, wherein the white matter, the gray matter, and the cerebrospinal fluid are calculated from the magnetic resonance image of the brain. . 4. The second process comprises: obtaining a histogram for a T2-weighted image of an inner region of the brain; and determining a largest light-dark value among light-dark values overlapping with a Gaussian distribution curve for a pixel corresponding to cerebrospinal fluid. 3. The method for separating and calculating white matter, gray matter and cerebrospinal fluid from a brain magnetic resonance image according to claim 2, wherein the lower limit of the light and dark values is determined. 5. The method according to claim 1, wherein the third step is a first step of a proton density (PD) image including only white matter and gray matter.
A first step of generating a histogram (H), and a first histogram (H) generated by the first step
A second step of calculating a white matter partial volume (PV1) from the cerebrospinal image separated by the second process,
A third step of generating a second histogram from the image remaining after removing the pure cerebrospinal cord, a fourth step of calculating a partial volume of the pure cerebrospinal cord from the second histogram generated by the third step, A fifth step of calculating a partial volume (PV2) occupied by white matter from the region from which the pure cerebrospinal part has been removed; and a white matter calculated by the second and fifth steps from the whole pixel values of the first histogram. Sum of partial volumes of
Subtracting (PV1 + PV2) to calculate a partial volume of gray matter; generating a third histogram including the first histogram and the second histogram; and increasing the brightness value in the third histogram. While calculating the integral value of the distribution function, the first light-dark value whose value is larger than the sum of the partial volumes of the white matter (PV1 + PV2) is set as the discrimination value, The method according to claim 2, further comprising a ninth step of generating an image in which white matter and gray matter are separated for each slice based on the set discrimination value. Method of white matter and gray matter separation and volume calculation. 6. The second step is to generate a histogram H of a PD image including only white matter and gray matter in a certain thickness by removing the entire cerebral spinal cord, to generate pure white matter light and dark. The value y1 and the brightness value y3 of pure gray matter are extracted, and the white matter having the brightness value y1 in a certain thickness is z.
%, A light-dark value G appearing when (100−z)% of a gray matter component having a light-dark value of y3 is expressed as follows: The sum of the pixels corresponding to the white matter partial volume from the light and dark values y1 to y3 is calculated using the method of calculating z% from the light and dark values using the pixels corresponding to the white matter partial volume. The method for separating white matter and gray matter from a magnetic resonance image of a brain and calculating a volume according to claim 3, wherein the calculation is performed by an equation. 7. The fourth step comprises: removing a pure cerebrospinal cord from a proton density (PD) image generated by the third step, and obtaining a histogram W for a cerebrospinal image mixed with white matter and gray matter, The light and dark value y3 of the cerebral spinal cord and the light and dark value y2 of the other component are extracted, and the cerebrospinal cord having the light and dark value of y3 in a certain thickness is z%, y2
The light-dark value G that appears when a component other than the cerebral spinal cord having a light-dark value of (100−z)% is contained is expressed as The sum of the pixels corresponding to the partial volume of the cerebrospinal cord from the light and dark values y2 to y3 is calculated by the following equation. 4. The method for separating white matter and gray matter from a brain magnetic resonance image and calculating the volume according to claim 3, wherein the calculation is performed by an equation. 8. In the fifth step, when W ′ is a distribution function of a histogram for a remaining portion after removing pixels corresponding to a partial volume of pure cerebrospinal fluid, the white matter partial volume is expressed as ] The method according to claim 3, wherein the white matter and the gray matter are separated from the brain magnetic resonance image and the volume is calculated by an equation. 9. The total volume of white matter (or gray matter) of the entire image slice calculated in the fourth step is defined as follows: N is the number of slices including the extracted white matter (or gray matter), and D is the interval between slices. And the number of pixels constituting the white matter (or gray matter) extracted from the i-th slice is Lp
Where X is the horizontal length of one pixel, Y is the vertical length of one pixel, and Si is the slice number. 3. The method for separating white matter and gray matter from a brain magnetic resonance image and calculating the volume according to claim 2, wherein the calculation is performed using the following equation.