JP2017150923A - Cell image analysis method and cell image analyzer - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a cell image analysis method and a cell image analyzer in which binarization of a cell image is adequately carried out and identification of cells in the cell image is performed, even when the cells are proximate and adjacent to each other and congested, and even when large variation in pixel intensity exists in the same cell group.SOLUTION: A cell image analysis method includes: a cell image acquisition step for acquiring a cell image in which specific cells are specifically dyed; and a cell region specification step for specifying cell regions on the basis of the cell image. The cell region specification step includes: setting peripheral regions of multiple different shapes for each binarization object pixel in the cell image; calculating a threshold for the peripheral regions on the basis of luminance of each pixel in the peripheral regions; subjecting the cell image to dynamic binarization processing on the basis of whether any pixel value of the object pixels in each peripheral region exceeds the threshold or not; and specifying the cell regions on the basis of the cell image subjected to the dynamic binarization processing.SELECTED DRAWING: Figure 1

Description

  The present invention relates to a cell image analysis method and a cell image analysis apparatus for analyzing a cell image obtained by photographing a sample containing a plurality of cell tumors and extracting a target cell, and in particular, from cell images of a plurality of densely and aggregated cell groups. The present invention relates to a cell image analysis method and a cell image analysis apparatus for specifying a region of a specific cell group.

  Circulating tumor cells (CTC) are found in the blood of patients with breast cancer, lung cancer, prostate cancer, pancreatic cancer, etc., and the number reflects the metastatic nature of cancer. It is attracting attention as it becomes important information above.

However, the density of CTC in blood is extremely low (in the case of low, about 1 to 10 per 10 mL of whole blood), and its detection and counting are not easy.
Conventionally, in order to distinguish such rare cells from a sample mixed with multiple cell tumors, a bright field image obtained by imaging a sample containing multiple cell tumors with specific staining of the target rare cells Image processing, such as increasing the contrast of specific cell tumors in cell images such as fluorescent images and fluorescent images, detects rare cells that have undergone specific staining.

  Specifically, when extracting a cell region from such a cell image, for example, in a fluorescence image of a cell, a binary image is generated with a certain threshold for pixel intensity, The area is specified.

  However, when the background density of the cell image is uneven, or when the variation in pixel intensity of the target cell group is large, binarization cannot be appropriately performed with one threshold value for one image.

For this reason, the cell image is also binarized using dynamic binarization in which a threshold value is set for each pixel of the cell image.
In dynamic binarization, for example, the influence of background intensity can be reduced by setting a threshold value based on the average intensity of pixels adjacent to or close to the target pixel.

  However, when cells are closely packed and close to each other or adjacent, or even when the same cell group has a large variation in pixel intensity, the dynamic based on the average intensity as described above can be used for the following reasons. Even if binarization is performed, the cell image cannot be binarized appropriately.

  In the case described above, in the situation where the low-intensity cells are surrounded by high-intensity cells, the pixel intensity of the low-intensity cells is smaller than the average pixel intensity of the surrounding cells, and the binarization is performed. Are treated as having no low intensity, that is, low-intensity cells, and low-intensity cells cannot be identified.

  In the present invention, in view of such a current situation, binarization of a cell image is appropriately performed even when cells are densely adjacent or adjacent to each other or when the variation in pixel intensity is large even in the same cell group. It is an object of the present invention to provide a cell image analysis method and a cell image analysis apparatus that can perform identification of cells in a cell image.

The present invention was invented in order to solve the problems in the prior art as described above, and in order to achieve at least one of the objects described above, a cell image analysis reflecting one aspect of the present invention. The method is
A cell image analysis method for identifying a cell region in which a cell exists in a cell image of a sample containing cells,
A cell image acquisition step of acquiring a cell image that specifically stains a specific cell;
A cell region identification step for identifying a cell region based on the cell image;
With
The cell region specifying step includes
For each binarization target pixel of the cell image, set a peripheral region having a plurality of different shapes,
Based on the luminance of each pixel in the peripheral area, calculate a threshold value in the peripheral area,
Based on whether any of the pixel values of the target pixel in each peripheral region exceeds the threshold, the cell image is subjected to dynamic binarization processing,
A cell region is specified based on the cell image subjected to the dynamic binarization process.

Further, a cell image analysis method reflecting one aspect of the present invention,
A cell image analysis method for identifying a cell region in which a cell exists in a cell image of a sample containing cells,
A cell image acquisition step of acquiring a cell image that specifically stains a specific cell;
A cell region identification step for identifying a cell region based on the cell image;
With
The cell region specifying step includes
For each binarization target pixel of the cell image, set a peripheral region,
Based on the luminance of each pixel in the peripheral area, calculate a threshold value in the peripheral area,
Based on whether or not the pixel value of the target pixel in the peripheral region exceeds the threshold value, the cell image is dynamically binarized,
Identifying a cell region based on the dynamic binarized cell image;
The pixel image is repeated a predetermined number of times while changing the shape of the peripheral region for pixels that are not identified as a cell region.

In addition, a cell image analysis apparatus reflecting one aspect of the present invention,
A cell image analysis device for identifying a cell region in which a cell exists in a cell image of a sample containing cells,
A cell image obtaining means for obtaining a cell image obtained by specifically staining a specific cell;
Cell region specifying means for specifying a cell region based on the cell image;
With
The cell region specifying means includes
For each binarization target pixel of the cell image, set a peripheral region having a plurality of different shapes,
Based on the luminance of each pixel in the peripheral area, calculate a threshold value in the peripheral area,
Based on whether any of the pixel values of the target pixel in each peripheral region exceeds the threshold, the cell image is subjected to dynamic binarization processing,
A cell region is specified based on the cell image subjected to the dynamic binarization process.

In addition, a cell image analysis apparatus reflecting one aspect of the present invention,
A cell image analysis device for identifying a cell region in which a cell exists in a cell image of a sample containing cells,
A cell image obtaining means for obtaining a cell image obtained by specifically staining a specific cell;
Cell region specifying means for specifying a cell region based on the cell image;
With
The cell region specifying means includes
For each binarization target pixel of the cell image, set a peripheral region,
Based on the luminance of each pixel in the peripheral area, calculate a threshold value in the peripheral area,
Based on whether or not the pixel value of the target pixel in the peripheral region exceeds the threshold value, the cell image is dynamically binarized,
Identifying a cell region based on the dynamic binarized cell image;
The pixel image is repeated a predetermined number of times while changing the shape of the peripheral region for pixels that are not identified as a cell region.

  According to the present invention, since a plurality of peripheral regions are set for the binarization target pixels of the cell image and the binarization processing is dynamically performed based on the threshold value in each peripheral region, the cells are densely packed. Appropriate cell image binarization and identification of cells in cell images, even when they are close to each other or adjacent to each other, or when there is a large variation in pixel intensity even in the same cell group Can do.

FIG. 1 is a block diagram illustrating a schematic configuration of a cell image analysis apparatus according to an aspect of the present invention. FIG. 2 is a schematic diagram illustrating an example of the shape of the peripheral region. FIG. 3 is a schematic diagram illustrating another example of the shape of the peripheral region. FIG. 4 shows, as an example, cell images (A) obtained by photographing a sample containing only white blood cells as test samples, and images (B) to (F) obtained by binarizing the cell images. It is the image shown. FIG. 5 is an image obtained by performing the same processing on a cell image (A) obtained by photographing a test sample different from that in FIG. FIG. 6 shows a comparison of binarization results based on the size of the peripheral region in the binarization processing by the cell image analysis method of this example. FIG. 7 is a process diagram showing a flow of detecting a target cell. FIG. 8 is a schematic diagram for explaining a method of determining whether or not the stained region of the intermediate image and the stained region of the first cell image are close to each other in the target cell detection step. FIG. 9 is a schematic diagram for explaining another method for determining whether or not the stained region of the intermediate image and the stained region of the first cell image are close to each other in the target cell detection step.

Hereinafter, embodiments (examples) of the present invention will be described in more detail based on the drawings.
FIG. 1 is a block diagram illustrating a schematic configuration of a cell image analysis apparatus according to an aspect of the present invention.

  As shown in FIG. 1, the cell image analysis apparatus 10 of the present embodiment includes a cell image acquisition unit 12 that acquires a cell image of a sample including a plurality of types of cells (hereinafter referred to as “test sample”). The cell area specifying means 14 for specifying the cell area from the cell image acquired by the cell image acquiring means 12 is provided as will be described later.

  The cell image analysis device 10 can be configured by a computer or the like, and the cell image acquisition unit 12 and the cell region specifying unit 14 may be software executed by a computer.

  In addition, the cell image acquisition means 12 is configured to acquire a cell image obtained by specifically staining an arbitrary cell. Specifically, the cell image analysis device converts a cell image captured in advance into data. 10 may be configured only by software for capturing the image, or may further include an image capturing device 13 such as a digital camera for capturing a cell image as shown in FIG.

In the present example, the cells included in the test sample will be described as including “target cells”, “non-target cells”, and “residual cells”.
In the present specification, the “target cell” can be any cell to be detected from a test sample, for example, a rare cell such as a blood circulating cancer cell (CTC).

  The “non-target cell” is a cell other than the “target cell” among the cells contained in the test sample and has a relatively high content in the test sample, for example, non-white blood cells and the like. It can be a rare cell. The “non-target cell” may include a plurality of types of cells.

  Further, “residual cells” mean cells that are neither “target cells” nor “non-target cells” contained in a test sample. That is, normally, cells that are not stained by antibodies that specifically stain target cells or antibodies that specifically stain non-target cells. In addition, although the present Example demonstrates as what a residual cell exists, there may be no residual cell.

  Such a test sample containing target cells, non-target cells, and residual cells is not particularly limited, and examples thereof include humans or other animals such as blood, urine, lymph, tissue fluid, and body cavity fluid. Samples collected from

The cell image can be taken as follows.
1. Staining Step First, specific staining is performed on specific cells of the test sample.
In addition, as a method of specifically staining the target cell, it can be performed by contacting a test sample with a solution containing a labeled antibody that binds to an antigen of a cell to be detected.
The antibody that shows a specific reaction with the antigen of the cell to be detected is not particularly limited. For example, anti-CK antibody, anti-β-actin antibody, anti-α-tubulin antibody, anti-β- Tubulin antibodies, anti-vimentin antibodies and the like can be used.

In addition, a method for specifically staining non-target cells can be performed by bringing a solution containing a labeled antibody that stains the entire cell membrane into contact with a test sample.
The antibody that stains the entire cell membrane is not particularly limited, and examples thereof include an anti-CD45 antibody, an anti-CD2 antibody, an anti-CD13 antibody, an anti-CD14 antibody, an anti-CD15 antibody, an anti-CD19 antibody, and an anti-CD203c antibody. Can be used.

  In addition, when specifically staining non-target cells, it is preferable to stain the whole non-target cells so that the outline of the non-target cells can be identified in the cell image by staining with a plurality of stains. .

  As such a labeled antibody, an antibody labeled with a known fluorescent dye, an antibody labeled with an antibody labeled with a dye, or the like can be used. When using fluorescent dyes, different excitation wavelengths and fluorescent wavelengths are used so that the fluorescent dye of the labeled antibody that stains the target cell and the fluorescent dye of the labeled antibody that stains the non-target cell can be measured separately. A fluorescent dye having the following may be used.

  As a method for staining all cells, staining can be performed by staining the cell nucleus. For example, DAPI (4 ′, 6-diamidino-2-phenylindole), Hoechst dye (Hoechst 33342, Hoechst 33258) Etc.), Acridine Orange, DAMO, Ethidium Bromide (Ethidium bromide), Ethidium Homodimer, Propidium Iodide and the like.

Since staining is performed in this manner, when all cells are stained, each cell is locally stained.
In addition, it is preferable to perform an appropriate pretreatment before staining the test sample in this way. Hereinafter, a pretreatment method when blood, which is a representative sample, is used as a test sample will be described.

-Anticoagulation treatment Blood (whole blood) collected and taken out of the body will coagulate over time if exposed to air as it is, and the cells contained therein cannot be recovered and observed. Therefore, it is preferable that the collected blood is immediately subjected to anticoagulation treatment.

  Various anticoagulants for whole blood are known and can be used in accordance with conditions such as general concentration and treatment time. For example, anticoagulant of the type that inhibits coagulation by binding to calcium ion by chelating action and removing it from the reaction system, represented by ethylenediaminetetraacetic acid (EDTA) and citric acid (including salts such as sodium salt) And anticoagulant of the type that inhibits coagulation by forming a complex with antithrombin III in plasma, represented by heparin, and suppressing the production of thrombin. A blood collection tube in which such an anticoagulant is previously stored may be used.

-Density gradient centrifugation The density gradient centrifugation can fractionate the components in the blood containing cells according to the specific gravity. In particular, it removes erythrocytes contained in a large amount in the blood and removes target cells such as CTC. Only white blood cells can be used for cell observation.

  The separation liquid used for the density gradient centrifugation may be any material that has a specific gravity suitable for the fractionation of cells in blood and can be welded so as to have an osmotic pressure and pH that do not destroy the cells. For example, commercially available sucrose solutions such as Ficoll (registered trademark) and Percoll (registered trademark) can be used.

  When density gradient centrifugation is performed after adjusting the specific gravity of this separated liquid to be smaller than the specific gravity of red blood cells and larger than the specific gravity of white blood cells, the blood sample is divided into `` a fraction rich in red blood cells '' and `` other than red blood cells '' Can be separated into at least two layers.

For example, when the specific gravity of the separation liquid is preferably 1.113 or less, more preferably 1.085 or less, the mixing rate of erythrocytes in the “fraction containing many cells other than erythrocytes” is 2 to 6% or less. Can be suppressed. Cell observation using “a fraction containing a large amount of cells other than erythrocytes” is preferable because the risk of failure to detect target cells such as CTC due to erythrocytes is reduced and the accuracy of diagnosis can be increased. .

  In addition, immobilization treatment, permeabilization treatment, blocking treatment, and other pretreatments can be performed on the cells contained in the test sample as necessary. These pretreatments can be performed at any time and in any order as long as the desired effect is exhibited.

  For example, after obtaining the cell fraction by density gradient centrifugation, immobilize cells before immobilizing on the observation substrate, immobilize on the observation substrate, and then permeabilize and block. May be performed simultaneously (ie, the cells are treated with a solution containing both a penetrant and a blocking agent).

  When cells are immobilized before immobilization on the observation substrate, the cells do not lose their ability to adsorb to the observation substrate even after immobilization. (Physical (intermolecular) interaction) between the substrate and the substrate.

When cells are immobilized after being immobilized on the observation substrate, if the cells have already been adsorbed on the observation substrate, the immobilization by adsorption can be maintained even after immobilization.
Further, in some cases, pretreatment for cells is performed before separating a cell fraction from a sample such as blood (for example, immediately after collecting a blood sample, cell anti-coagulation treatment and cell immobilization treatment are performed, Thereafter, density gradient centrifugation is performed to obtain a cell fraction).

-Immobilization treatment Immobilization treatment is a treatment performed to delay cell autolysis and spoilage and maintain its morphology and antigenicity. In the present invention, the detection of target cells such as CTC is enhanced. In addition, it is possible to suppress damage to cells or antigens by a dissociating agent such as an acid used in the antigen-antibody dissociation step.

  In particular, when an experiment is performed by repeating the binding and dissociation multiple times, the effect of the dissociation agent on the cells is greatly increased, so cell deterioration may increase and the experiment may not continue normally. .

  However, the immobilization treatment as in this step increases the retention of cells in form and antigenicity, and the effect of making immunostaining less likely to deteriorate even when the experiment is repeated multiple times. Bring.

  Various fixing agents are known, and examples thereof include aldehydes such as formaldehyde, glutaraldehyde and glyoxal; ketones such as acetone and methyl ethyl ketone; and alcohols such as ethanol and methanol. Among these, aldehydes such as formaldehyde and ketones such as acetone are immobilizing agents that act as a cross-linking agent by reacting an aldehyde group and a ketone group with a specific amino acid residue to form a covalent bond. It is a preferred immobilizing agent in the present invention because it can gel the cell protoplasm and suppress enzyme activity.

  In addition, although it does not act directly as a fixing agent, a donor that liberates the fixing agent upon hydrolysis or the like, such as a formaldehyde donor, can also be used as one form of the fixing agent. An aqueous solution obtained by dissolving these immobilizing agents in an appropriate solvent such as PBS (or diluting a commercially available solution such as formalin from the beginning) can be used as an immobilization treatment solution.

  The immobilization treatment can be performed by bringing the immobilization treatment solution having an appropriate concentration into contact with the cells for an appropriate time. The concentration of the immobilizing agent in the immobilizing treatment liquid can be adjusted as appropriate. For example, in the case of aldehydes, it is about 0.1 to 10 w / w%, and in the case of alcohols, it is about 70 to 100 v / v%. . The contact time between the immobilization treatment solution and the cells (immobilization treatment time) can be adjusted as appropriate, and is, for example, about 10 minutes to 1 hour at room temperature.

  The immobilization treatment may be performed before or during the staining step, and may be performed before or during the first staining step, or after the Nth dissociation step and (N + 1) th. It may be repeated before or during the dyeing process.

-Permeabilization treatment The permeabilization treatment is a treatment for improving the permeability of the cell membrane to facilitate the penetration of substances from the outside of the cell into the inside of the cell. In the present invention, the antigen in the cell (protoplasm) The effect of improving the immunostaining by facilitating the binding of a fluorescent-labeled antibody to the antibody.

  Various penetrants are known. For example, Triton X-100 (polyoxyethylene alkylphenyl ether), Tween 20 (polyoxyethylene sorbitan monolaurate), saponin, digitonin, Leukoperm, NP-40, etc. Surfactants (commercial products). An aqueous solution in which these permeabilizing agents are dissolved in an appropriate solvent such as PBS can be used as the permeation treatment liquid.

  The permeabilization treatment can be performed by bringing the permeabilization treatment solution having an appropriate concentration into contact with the cells for an appropriate time. The concentration of the permeabilizing agent in the permeabilizing liquid can be adjusted as appropriate. For example, in the case of the surfactant as described above, it is about 0.01 to 1.0 v / v%. The contact time between the permeabilization solution and the cells (permeabilization time) can be adjusted as appropriate, and is, for example, about 10 minutes to 1 hour at room temperature.

-Blocking treatment The blocking agent is used to prevent non-specific staining of the cells other than the target antigen. The cell blocking treatment may be performed before or during the staining step, and may be performed either or both.

  In the case of immobilizing a cell containing a blocking agent, it is necessary to form a microchamber or the like on the observation substrate as described above (using a structural method relating to the method of immobilizing a cell described later).

  In addition, by covering the part that is likely to cause adsorption (interaction) with the observation substrate on the cell surface, it is possible to prevent the cells from adsorbing to parts other than the part related to the solid phase of the cell such as a micro chamber. It is also used for accommodating cells in a predetermined site.

  Various blocking agents are known, and examples include hydrophilic polymers such as casein, skim milk, bovine serum albumin (BSA), and polyethylene glycol, and low molecular compounds such as phospholipid, ethylenediamine, and acetonitrile. An aqueous solution obtained by diluting these blocking agents with an appropriate solvent such as PBS can be used as a blocking treatment solution.

  The blocking treatment can be performed by bringing a blocking treatment solution having an appropriate concentration into contact with cells or an observation substrate for an appropriate time. Although the density | concentration of the blocking agent in a blocking process liquid can be adjusted suitably, it is about 0.01-20 w / v%, for example. The contact time between the blocking treatment solution and the cells (blocking treatment time) can be adjusted as appropriate, and is, for example, about 5 to 120 minutes at room temperature.

2. Imaging Step In the staining step described above, a cell image is captured using the image capturing device 13 for the test sample subjected to the three types of staining.

  When labeling with a fluorescent dye, excitation that excites the fluorescent dye of the labeled antibody that stains the target cells, the fluorescent dye of the labeled antibody that stains non-target cells, and the fluorescent dye that labels all cells. What is necessary is just to image | photograph the cell image based on the fluorescence which light is irradiated sequentially and light-emitted from each fluorescent dye.

  The image capturing device 13 that captures such a cell image is not particularly limited, but is preferably a device that captures a cell image as image data that can be handled by software executed by a computer. For example, a digital camera using a CCD image sensor or a CMOS image sensor can be used.

  The image data of the cell image photographed by the image photographing device 13 is read into the cell image analyzing device 10 by the cell image acquiring means 12 and the cell region specifying means 14 performs the image analysis of the cell image.

The cell area specifying means 14 specifies an area of an arbitrary cell from the cell image as follows.
3. Cell area specifying step The cell area is specified by dynamically binarizing each pixel of the cell image as follows. In the binarization processing in this embodiment, the value of a pixel having a luminance smaller than the threshold is set to 0, and the value of a pixel having a luminance higher than the threshold is set to 1.

First, a peripheral region having a plurality of different shapes is set for an arbitrary target pixel (binarization target pixel) of a cell image.
In this specification, the “peripheral region” means a region around the target pixel used when determining a threshold value for an arbitrary target pixel. That is, the conventional dynamic binarization means a region of a pixel group adjacent to or close to the target pixel for determining the threshold value.

The shape of the peripheral region is not particularly limited as long as the plurality of peripheral regions have different shapes, but a plurality of peripheral regions can be set as follows.
First peripheral region: four-fold symmetry with respect to an arbitrary target pixel and a line-symmetric region with a straight line including the target pixel as a symmetry axis Second peripheral region: with respect to an arbitrary target pixel, An area that is two-fold symmetric and that is not four-fold symmetric and that has a line of symmetry that includes a straight line including the target pixel as an axis of symmetry In this specification, n-fold symmetry is defined as (360) around the central axis. / N) A shape that overlaps itself when rotated.

  Specifically, as the first peripheral region 17, as shown in FIG. 2A, a square whose sides are parallel to the arrangement of the pixels 15, and as shown in FIG. Examples include a substantially square in which each diagonal line is parallel to the 15 arrays, a substantially circular shape shown in FIG. Reference numeral 16 denotes an arbitrary target pixel on which binarization processing is performed.

  Further, as the second peripheral area 18, as shown in FIG. 3A, a rectangle whose sides are parallel to the arrangement of the pixels 15, and as shown in FIG. Are substantially rhombuses whose diagonals are parallel to each other, a rectangle shown in FIG. 3C, a substantially ellipse shown in FIG.

  Although not shown, for example, an area having a shape obtained by rotating the second peripheral area by ± 90 ° can be set as the third peripheral area, and more than four peripheral areas can be set. Is also possible. Further, the peripheral region may have a symmetry of 6 times or more about an arbitrary target pixel. That is, the peripheral region can be an object of n times (n is an even number) with an arbitrary target pixel as the center, and can have a line-symmetric shape with a straight line including the target pixel as a symmetry axis.

  Moreover, it is preferable that the size of each peripheral region is substantially the same. For example, when the first peripheral region 17 is a region of m × m pixels, the second peripheral region 18 is n × m pixels ( The region of n <m) and the third peripheral region are preferably set like a region of m × n pixels (n <m).

  In order to determine each threshold value from the plurality of peripheral areas set in this way, a representative value of the luminance of each peripheral area can be used. Here, as the representative value of luminance, for example, an average value, a median value, or a value obtained by adding a predetermined constant to these can be used, and more generally a value represented by the following formula (1): Can be used.

  Here, (x, y) is the coordinate of the pixel in the peripheral region, h (x, y) is the weighting factor at the coordinate (x, y), and I (x, y) is at the coordinate (x, y). Luminance and α are predetermined constants.

  However, Σh (x, y) = 1, and h = 0 outside the peripheral region. For example, when the peripheral area is a square of 11 × 11 pixels, the average value of the luminance of the peripheral area can be obtained as a threshold if the weighting coefficient h (x, y) of each pixel in the peripheral area is 1/121.

  Binarization of the target pixel is performed in each peripheral region using the threshold value of each peripheral region obtained in this manner, and when the result in all the peripheral regions is 0, the pixel value of the target pixel is set to 0. When the result in any peripheral region is 1, the pixel value of the target pixel is set to 1.

In the present embodiment, as described above, the threshold value of the target pixel is determined from the peripheral region. However, the present invention is not limited to this, and various known methods can be used as appropriate.
As a result of performing such binarization processing on all the pixels of the cell image, a pixel value is 1, that is, a portion expressed in white in the image becomes a cell region.

  When simple binarization or conventional dynamic binarization is performed by using a cell image analysis method such as this example, cells that are difficult to properly identify cell areas are densely packed and close to each other. -Even in the case of cell images that are adjacent or have a large variation in pixel intensity even in the same cell group, a cell region can be appropriately specified.

In the cell region specifying step, not only the above-described binarization processing but also a plurality of binarization processing and other image processing can be performed in combination as will be described later.
In FIG. 4, as an example, a cell image (A) obtained by photographing a sample containing only white blood cells as a test sample, and images (B) to (F) obtained by binarizing the cell image are shown. Show.

  The cell image (A) is obtained by staining the white blood cells arranged on the substrate with a fluorescent dye labeled with an anti-CD45 antibody, and trimming a part of the image photographed using a 5 × objective lens. The maximum luminance in this cell image (A) was 0.45. Further, regarding the cell image (A), when the number of cells was visually counted, 66 cells could be identified.

  Images (B) to (D) are images obtained by subjecting the cell image (A) to conventional binarization processing, and the image (B) sets a threshold value using the Otsu binarization method. Then, the image subjected to simple binarization processing, the image (C) is set to a threshold smaller than the threshold value in the image (B), and the image subjected to simple binarization processing, the image (D) is a periphery of 21 × 21 pixels. It is an image that has been subjected to dynamic binarization processing using the average luminance value of the region +0.003 as a threshold value.

  As shown in the image (B), when a threshold is set using the binarization method of Otsu and simple binarization is performed, low intensity cells may not be detected. Regarding the image (B), when the number of cells was visually counted, 44 cells could be identified.

  On the other hand, in order to detect such low-intensity cells, when the threshold value is set small (in this embodiment, the threshold value is set to 0.035), the gap between the cells as in the image (C). That is, the cell region is specified up to the place where the cell does not exist. Note that it is difficult to visually count the number of cells in the image (C).

  In addition, when the conventional dynamic binarization process is performed as in the image (D), low-intensity cells can be detected as compared with the image (B), but the cell image (A) As shown in the figure, when low-intensity cells and high-intensity cells coexist, high-intensity cells often exist around low-intensity cells. There are low-strength cells that cannot. As for the image (D), when the number of cells was visually counted, 46 cells could be identified.

  The image (E) is an image obtained by subjecting the cell image (A) to dynamic binarization processing by the cell image analysis method of the present embodiment, and is 21 × 21 pixels, 21 × 7 pixels, 7 × 21. As described above, binarization processing is performed using three peripheral regions of pixels. In each peripheral area, the average luminance value of the peripheral area +0.003 is set as a threshold value.

  Thus, by using the cell image analysis method of the present embodiment, low-intensity cells that could not be detected by conventional simple binarization and dynamic binarization can also be detected. In addition, regarding the image (E), when the number of cells was visually counted, 53 cells could be identified.

  The image (F) includes a result obtained by performing dynamic binarization processing by the cell image analysis method of the present embodiment after substituting a predetermined value or more for a pixel value on the cell image (A), and a threshold value Is an image generated on the basis of the result of simple binarization processing with 0.02.

  Specifically, in the cell image (A), after replacing a pixel having a pixel value of 0.025 or more with 0.025, a dynamic binarized image obtained by binarization in the same manner as the image (E) is obtained. . Next, a simple binarized image that has been subjected to simple binarization processing with a higher threshold value is obtained.

  Then, the dynamic binarized image and the simple binarized image are compared for each pixel. If both have a pixel value of 0, the pixel value of the target pixel is set to 0. For example, an image generated with the pixel value of the target pixel as 1 is the image (F).

  In this way, by performing a combination of a plurality of binarization processes and other image processes, it is possible to detect many low-intensity cells as in the image (F). As for the image (F), when the number of cells was visually counted, 63 cells could be identified.

  As a combination of such image processing, as described above, dynamic binarization processing by the cell image analysis method of this embodiment and known image processing can be appropriately combined. Compare the result of the dynamic binarization processing by the cell image analysis method of the example and the result of the simple binarization processing by setting a lower threshold value for each pixel. If there is, the pixel value of the target pixel is set to 1, and if any one of the values is 0, the pixel value of the target pixel can be set to 0 to specify the cell region.

  In addition, binarization processing is performed on each pixel of the cell image by the first peripheral region, and as a result, only for each pixel that is not identified as the cell region, that is, the pixel value is 0. The binarization process may be performed using two peripheral areas.

  In this way, the cell region is identified using a computer by sequentially repeating the binarization process a predetermined number of times while changing the shape of the peripheral region for pixels that have not been identified as a cell region in the cell image. The time required to do so can be shortened. Note that the number of repetitions can be set as appropriate.

FIG. 5 is an image obtained by performing the same processing on a cell image (A) obtained by photographing a test sample different from that in FIG.
As in the case of FIG. 4, it is possible to detect more cells in the image subjected to the binarization processing by the cell image analysis method of the present embodiment than the image subjected to the conventional binarization processing. did it.

FIG. 6 shows a comparison of binarization results based on the size of the peripheral region in the binarization processing by the cell image analysis method of this example.
The cell image (A) is obtained by staining white blood cells arranged on a substrate with a fluorescent dye labeled with an anti-CD45 antibody and trimming a part of an image taken with a 5 × objective lens. For photographing, a digital camera using a CCD image sensor is used, and the pixel size of the CCD is 6.45 μm × 6.45 μm. Further, the pixel pitch in the cell image (A) is 1.3 μm, and the white blood cell diameter is about 9 pixels.

  For this cell image (A), the sizes of the surrounding areas are (B) 7 × 7 pixels, (C) 9 × 9 pixels, (D) 19 × 19 pixels, (E) 29 × 29 pixels, ( F) Dynamic binarization processing was performed as 36 × 36 pixels and (G) 61 × 61 pixels.

  When the size of the peripheral region is smaller than the diameter of the cell (white blood cell) as in (B), the central portion of the cell region is black, that is, it is not a cell region. Cannot be identified.

This is a result of reflecting uneven brightness in the same cell because the size of the peripheral region is smaller than the cell diameter.
In addition, as shown in (C), by specifying the size of the peripheral region to be approximately the same as the cell diameter, the cell region can be specified almost accurately.

  On the other hand, as shown in (G), when the size of the peripheral region is significantly larger than the diameter of the cell, the portion where the cell does not exist is white, that is, the cell region, The cell area cannot be specified.

  This is because the size of the surrounding area is too large for the cell diameter, and as a result, the proportion of cells in the surrounding area is reduced. As a result, the luminance average value in the surrounding area is reduced, and the threshold value is low. Since binarization processing is performed, even a region of a gap between cells having higher brightness than the background is specified as a cell region.

In addition, as shown in (F), when the size of the peripheral region is about four times the cell diameter, the cell region can be specified almost accurately.
From the above, as the size of the peripheral region, the long axis is preferably 1 to 4 times the cell diameter.

In this specification, “long axis” means, for example, one side of a square, a long side of a rectangle, a circular diameter, an elliptical long axis, a diagonal line of a rhombus, and the like.
Further, as the size of the peripheral region, the short axis is preferably 0.5 to 1 times the cell diameter.

In the present specification, the “short axis” means, for example, a rectangular short side, an elliptical short axis, and the like.
In this embodiment, the pixel pitch of the cell image (A) is set to 1.3 μm, which is about 1/9 of the cell diameter. However, a resolution that allows the cell to recognize its shape on the image can be obtained. The pixel pitch is preferably 1/5 or less of the cell diameter.

  Hereinafter, the flow of detecting circulating cancer cells (CTC) in blood from a cell image of blood cells using the cell image analysis method of the present embodiment will be described. The blood cells contain CTC as target cells and leukocytes as non-target cells.

As described above, in order to detect a target cell from a test sample in which target cells and non-target cells are mixed, generally, cell images subjected to the following three types of staining are used.
(1) Staining using an antibody that specifically reacts with a target cell (CTC) 1
(2) Staining with an antibody that reacts specifically with non-target cells (white blood cells) 2
(3) Staining 3 for all cells (target cells, non-target cells, residual cells)
Based on a first cell image 22 that is a cell image that has been stained 1, a second cell image 24 that is a cell image that has been stained 2, and a third cell image 26 that is a cell image that has been stained 3. The target cell (CTC) in the test sample is detected.

FIG. 7 is a process diagram showing a flow of detecting a target cell.
First, the intermediate image 28 is generated by removing the overlapping portion of the second cell image 24 from the third cell image 26.

  Specifically, the second cell image 24 is binarized by the cell image analysis method of the present embodiment, and the third cell image 26 is binarized by the cell image analysis method of the present embodiment or the conventional simple binary. The binarization process may be performed by binarization or dynamic binarization, and the second cell image 24 may be subtracted from the binarized third cell image 26.

  In this specification, “subtraction processing” means subtraction in image processing, and subtracting the value of the pixel at the same position in the other cell image from the value of the pixel in one cell image It can be obtained by performing it in the pixel. In general, the pixel value is preferably set to 0 when the minimum value is 0, and when the pixel value becomes negative as a result of the subtraction process.

Next, the target cell is detected by comparing the intermediate image 28 with the first cell image 22.
Specifically, after binarizing the first cell image 22 by binarization or conventional simple binarization or dynamic binarization by the cell image analysis method of the present embodiment, as follows, Stained region 29 in intermediate image 28 (region in which target cells and a part of residual cells are stained) and stained region 23 in first cell image 22 (target cells, non-target cells, and part of non-cellular foreign matter are stained. And a region where both stained regions overlap or are close to each other is detected as a target cell. A cell including a region expressed in white in the detection image can be determined as a region of the target cell (CTC).

The determination whether or not both staining regions are close to each other can be performed as follows, for example.
In a state where the intermediate image 28 and the first cell image 22 are overlapped, as shown in FIG. 8, the minimum value t of the distance between the pixels of both stained regions is calculated, and this minimum value t is calculated from the intermediate image 28. The shortest distance between the stained region 29 and the stained region 23 of the first cell image 22 is set. If the shortest distance is equal to or smaller than a predetermined proximity determination value j, it can be determined that the shortest distance is close.

  Note that the proximity determination value j can be an arbitrary value depending on the cell to be detected. For example, when the average size of the target cell is considered to be equal to that of the white blood cell, the radius of the white blood cell is When it is assumed that it is about 7 μm, it is desirable that the proximity determination value is 7 μm. In addition, when it is assumed that the radius of white blood cells is about 15 μm, it is desirable that the proximity determination value be 15 μm. This is because if the cells are located within a radius range such as leukocytes, it can be determined that the cells are one cell unless the cells overlap in the vertical direction.

  In addition, with the intermediate image 28 and the first cell image 22 superimposed, as shown in FIG. 9, both staining regions are processed with known image processing algorithms (for example, nearest neighbor method, bilinear interpolation method, bicubic interpolation method). , Lanczos interpolation method or the like), and it is possible to determine that they are close to each other when the inflated dyeing regions 29a and 23a are superposed.

  The expansion value d can be an arbitrary value depending on the cell to be detected, and can be a value slightly larger than half of the proximity determination value j. For example, when the size of the target cell is assumed to be the same as that of the white blood cell, it is assumed that the diameter of the white blood cell is about 15 μm because there is a possibility that the objects to be stained exist within the range of the white blood cell diameter. In this case, the expansion value can be 8 μm.

  Further, if the objects to be stained are within the radius range of leukocytes, it can be determined that the possibility of the same is higher than the above, and if the radius of leukocytes is assumed to be about 7 μm, the expansion value d Can be 4 μm.

Further, by performing a combination of the above-described proximity determinations, it is possible to determine more accurately whether or not both staining regions are close to each other.
Further, when the distance between the center (centroid) position of one staining region and the center (centroid) position of the other staining region is smaller than a predetermined proximity determination value, it is determined that both staining regions are close to each other. You may make it do.

  The preferred embodiment of the present invention has been described above. However, the present invention is not limited to this, and in the above embodiment, a rare cell such as CTC is detected as the target cell. The cells are not limited to rare cells, and various modifications can be made without departing from the object of the present invention, such as use for detecting any cell.

  In addition, a program for causing the above-described cell image analysis method to be executed by a computer and the program recorded therein, for example, a magnetic tape (digital data storage (DSS), etc.), a magnetic disk (hard disk drive (HDD), a flexible disk (FD) )), Optical disc (compact disc (CD), digital versatile disc (DVD), Blu-ray disc (BD), etc.), magneto-optical disc (MO), flash memory (SSD (Solid State Drive), memory card, USB memory, etc.) And the like are also included as one embodiment of the present invention.

  The present invention is particularly effective when detecting a very small number of target cells from a large number of non-target cells. For example, when circulating blood cancer cells (CTC) are detected from blood cells, it is necessary to separate CTCs, which are target cells, from many white blood cells, which are non-target cells.

  As a method for detecting CTC, a suspension of leukocytes and CTC is poured on a cell development substrate, cells are arranged on the substrate, and a fluorescent image is obtained by appropriate fluorescent staining, and the number of CTCs in the fluorescent image is determined. Count.

  At this time, two types of staining, CTC-specific staining 1 and leukocyte-specific staining 2, were performed, and a cell that was stained with staining 1 and not stained with staining 2 was determined to be CTC. To do. For example, anti-cytokeratin (CK) antibody is used for staining 1 and anti-CD45 antibody is used for staining 2. Since both staining 1 and staining 2 are nonspecifically stained, CTC is specified by combining two types of specific staining.

  At this time, the CD45 intensity of white blood cells varies even in the same specimen, and white blood cells that shine brightly and white blood cells that shine darkly are mixed. Further, in order to develop a large number of cells on the cell development substrate with high density, the cells are in close proximity.

  Here, when white blood cells are identified from a fluorescent image of blood cells stained with an anti-CD45 antibody, if white blood cells with low CD45 intensity sandwiched between white blood cells with high CD45 intensity are detected and missed, CTC is overdetected.

  On the other hand, when there is a CTC adjacent to a white blood cell having a high CD45 intensity, the fluorescence of the white blood cell may be detected in a broad manner in the CTC region, leading to a CTC detection loss. Therefore, it is important to accurately identify white blood cells without excess or deficiency from fluorescent images of blood cells stained with anti-CD45 antibodies.

DESCRIPTION OF SYMBOLS 10 Cell image analyzer 12 Cell image acquisition means 13 Image capturing device 14 Cell area specifying means 15 Pixel 16 Binarization target pixel 17 First peripheral area 18 Second peripheral area 22 First cell image 23 Stained area 23a After expansion Stained region 24 Second cell image 26 Third cell image 28 Intermediate image 29 Stained region 29a Stained region after expansion

Claims (24)

A cell image analysis method for identifying a cell region in which a cell exists in a cell image of a sample containing cells,
A cell image acquisition step of acquiring a cell image that specifically stains a specific cell;
A cell region identification step for identifying a cell region based on the cell image;
With
The cell region specifying step includes
For each binarization target pixel of the cell image, set a peripheral region having a plurality of different shapes,
Based on the luminance of each pixel in the peripheral area, calculate a threshold value in the peripheral area,
Based on whether any of the pixel values of the target pixel in each peripheral region exceeds the threshold, the cell image is subjected to dynamic binarization processing,
A cell image analysis method for identifying a cell region based on the cell image subjected to the dynamic binarization process. A cell image analysis method for identifying a cell region in which a cell exists in a cell image of a sample containing cells,
A cell image acquisition step of acquiring a cell image that specifically stains a specific cell;
A cell region identification step for identifying a cell region based on the cell image;
With
The cell region specifying step includes
For each binarization target pixel of the cell image, set a peripheral region,
Based on the luminance of each pixel in the peripheral area, calculate a threshold value in the peripheral area,
Based on whether or not the pixel value of the target pixel in the peripheral region exceeds the threshold value, the cell image is dynamically binarized,
Identifying a cell region based on the dynamic binarized cell image;
A cell image analysis method that repeats a predetermined number of times while changing the shape of the peripheral region for pixels that are not identified as cell regions in the cell image.   The peripheral region has n-fold symmetry (n is an even number) around the binarization target pixel, and has a line-symmetric shape with a straight line including the binarization target pixel as a symmetry axis. 3. The cell image analysis method according to 1 or 2.   The cell image analysis method according to any one of claims 1 to 3, wherein a major axis of the peripheral region is 1 to 4 times a diameter of the cell.   The cell image analysis method according to claim 4, wherein the short axis of the peripheral region is 0.5 to 1 times the diameter of the cell.   The cell image analysis method according to any one of claims 1 to 5, wherein the threshold value is an average value, a median value, or a predetermined constant added to a luminance value of each pixel in the peripheral region. The cell image analysis method according to any one of claims 1 to 5, wherein the threshold value is represented by the following formula (1).
Here, (x, y) is the coordinate of the pixel in the peripheral region, h (x, y) is the weighting factor at the coordinate (x, y), and I (x, y) is at the coordinate (x, y). The luminance and α are predetermined constants, and Σh (x, y) = 1.   The cell region specifying step specifies the cell region based on the cell image subjected to the dynamic binarization process and the cell image subjected to the simple binarization process using a predetermined threshold value. The cell image analysis method described.   The cell image analysis method according to claim 8, wherein an area specified as the cell area is specified as a cell area at least one of the cell image subjected to the dynamic binarization process and the cell image subjected to the simple binarization process.   The cell image analysis method according to claim 8, wherein a region identified as the cell region in both the dynamic binarized cell image and the simple binarized cell image is identified as a cell region.   The cell image analysis method according to any one of claims 1 to 10, wherein the cell region specifying step further includes, for each pixel of the cell image, replacing a pixel value greater than or equal to a predetermined value with a predetermined value. A cell image analysis device for identifying a cell region in which a cell exists in a cell image of a sample containing cells,
A cell image obtaining means for obtaining a cell image obtained by specifically staining a specific cell;
Cell region specifying means for specifying a cell region based on the cell image;
With
The cell region specifying means includes
For each binarization target pixel of the cell image, set a peripheral region having a plurality of different shapes,
Based on the luminance of each pixel in the peripheral area, calculate a threshold value in the peripheral area,
Based on whether any of the pixel values of the target pixel in each peripheral region exceeds the threshold, the cell image is subjected to dynamic binarization processing,
A cell image analysis apparatus for identifying a cell region based on the cell image subjected to the dynamic binarization process. A cell image analysis device for identifying a cell region in which a cell exists in a cell image of a sample containing cells,
A cell image obtaining means for obtaining a cell image obtained by specifically staining a specific cell;
Cell region specifying means for specifying a cell region based on the cell image;
With
The cell region specifying means includes
For each binarization target pixel of the cell image, set a peripheral region,
Based on the luminance of each pixel in the peripheral area, calculate a threshold value in the peripheral area,
Based on whether or not the pixel value of the target pixel in the peripheral region exceeds the threshold value, the cell image is dynamically binarized,
Identifying a cell region based on the dynamic binarized cell image;
A cell image analysis apparatus that repeats a predetermined number of times while changing a shape of the peripheral region for a pixel that is not specified as a cell region in the cell image.   The peripheral region has n-fold symmetry (n is an even number) around the binarization target pixel, and has a line-symmetric shape with a straight line including the binarization target pixel as a symmetry axis. The cell image analysis apparatus according to 12 or 13.   The cell image analyzer according to any one of claims 12 to 14, wherein a major axis of the peripheral region is 1 to 4 times a diameter of the cell.   The cell image analysis apparatus according to claim 15, wherein a short axis of the peripheral region is 0.5 to 1 times the diameter of the cell.   The cell image analysis device according to any one of claims 12 to 16, wherein the threshold value is an average value, a median value, or a predetermined constant added to the luminance value of each pixel in the peripheral region. The cell image analyzer according to any one of claims 12 to 16, wherein the threshold value is represented by the following formula (1).
Here, (x, y) is the coordinate of the pixel in the peripheral region, h (x, y) is the weighting factor at the coordinate (x, y), and I (x, y) is at the coordinate (x, y). The luminance and α are predetermined constants, and Σh (x, y) = 1.   The cell region specifying means specifies the cell region based on the cell image subjected to the dynamic binarization process and the cell image subjected to the simple binarization process using a predetermined threshold. The cell image analysis apparatus described.   The cell image analysis apparatus according to claim 19, wherein an area specified as the cell area is specified as a cell area at least one of the cell image subjected to the dynamic binarization process and the cell image subjected to the simple binarization process.   The cell image analysis apparatus according to claim 19, wherein an area identified as the cell area in both the cell image subjected to the dynamic binarization process and the cell image subjected to the simple binarization process is identified as a cell area.   The cell image analysis device according to any one of claims 12 to 21, wherein the cell region specifying step further includes, for each pixel of the cell image, replacing a pixel value greater than or equal to a predetermined value with a predetermined value.   A program for causing a computer to execute the cell image analysis method according to any one of claims 1 to 11.   A computer-readable recording medium on which the program according to claim 23 is recorded.