Cell Image Analysis Method

Abstract

This cell image analysis method involves: a step for acquiring a cell image (10) displaying cells (90); a step for acquiring index values (21) obtained by analyzing the cells using a learned model having learned cell analysis; a step for acquiring abnormal cell regions (91) in each of which, among the index values, a first index value indicating that a cell is not a normal cell and is an abnormal cell is greater than a predetermined determination reference value (20); and a step for representing the abnormal cell regions in an identifiable manner.

Description

Technical Field

[0001] The present invention relates to a method for analyzing cell images, and more particularly to a cell analysis method for analyzing cells using a pre-trained model.

Background Art

[0002] Conventionally, a cell analysis method for analyzing cells using a pre-trained model has been known. Such a cell analysis method is disclosed, for example, in International Publication No. 2019 / 171546.

[0003] International Publication No. 2019 / 171546 discloses a cell image analysis method for analyzing an image of cells captured by an imaging device. The cell image analysis method disclosed in International Publication No. 2019 / 171546 classifies a normal cell region and an abnormal cell region using the analysis result of a pre-trained model. Specifically, International Publication No. 2019 / 171546 discloses a configuration for classifying a normal cell region and an abnormal cell region by a segmentation process for determining to which category each pixel of a cell image belongs.

Prior Art Documents

Patent Documents

[0004]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0005] Here, although not disclosed in International Publication No. 2019 / 171546, when classifying a normal cell region and an abnormal cell region using the analysis result (index value) of a learned model, among the index values for each pixel, the highest index value determines whether the pixel belongs to the normal cell region or the abnormal cell region. That is, for each pixel, if the index value indicating that it is a normal cell is greater than the index values indicating abnormal cells and the background, it is classified as the normal cell region. Also, for each pixel, if the index value indicating that it is an abnormal cell is greater than the index values indicating normal cells and the background, it is classified as the abnormal cell region. Therefore, even in a region classified as the normal cell region, there may be abnormal cells included. Thus, with the method of classifying the normal cell region and the abnormal cell region by segmentation processing as in International Publication No. 2019 / 171546, it is difficult to determine whether there is suspicion of abnormal cells in the region classified as the normal cell region. Therefore, a cell image analysis method capable of identifying a region suspected of containing abnormal cells in a cell image is desired.

[0006] This invention is made to solve the above problems, and one object of this invention is to provide a cell image analysis method capable of identifying a region suspected of containing abnormal cells in a cell image.

Means for Solving the Problems

[0007] To achieve the above object, a cell image analysis method according to one aspect of this invention includes a step of acquiring a cell image in which cells are captured, a step of obtaining index values obtained by analyzing the cells captured in the cell image using a learned model trained to analyze cells, a step of obtaining an abnormal cell region which is a region where a first index value indicating that the cell is an abnormal cell rather than a normal cell among the index values is greater than a predetermined determination reference value, and a step of displaying the abnormal cell region in an identifiable manner. , the predetermined determination reference value includes a first determination reference value and a second determination reference value lower than the first determination reference value. In the step of obtaining the index value, as the first index value, for each pixel of the cell image, a first probability value indicating the probability that the cell is an abnormal cell and a second probability value indicating the probability that the cell is a normal cell are obtained. The step of obtaining the abnormal cell region includes a step of obtaining, as the first abnormal cell region, the abnormal cell region where the first index value is greater than the first determination reference value, and a step of obtaining, as the second abnormal cell region, the abnormal cell region where the first index value is greater than the second determination reference value. In the step of obtaining the first abnormal cell region, even if the first probability value is smaller than the second probability value, when the first probability value is greater than the first determination reference value, the pixel of the cell image is obtained as the first abnormal cell region. In the step of obtaining the second abnormal cell region, even if the first probability value is smaller than the second probability value, when the first probability value is greater than the second determination reference value, the pixel of the cell image is obtained as the second abnormal cell region. In the step of displaying the abnormal cell region in a distinguishable manner, the first abnormal cell region and the second abnormal cell region are displayed in a distinguishable manner 。

Effects of the Invention

[0008] In the cell image analysis method in the above-described one aspect, as described above, among the index values, a step of obtaining an abnormal cell region, which is a region where a first index value indicating that a cell is an abnormal cell rather than a normal cell is larger than a predetermined determination reference value, is provided. As a result, since a region where the first index value is larger than the predetermined determination reference value is obtained as the abnormal cell region, even when the first index value is smaller than other index values, it can be obtained as the abnormal cell region. Further, as described above, a step of discriminatively displaying the abnormal cell region is provided. As a result, for example, when the determination reference value is decreased, the obtained abnormal cell region is discriminatively displayed as a region suspected of being an abnormal cell, so that the abnormal cell region suspected of being an abnormal cell can be visually displayed. As a result of these, it is possible to provide a cell image analysis method capable of grasping a region suspected of being an abnormal cell in a cell image.

Brief Description of the Drawings

[0009]

Figure 1

Figure 2

Figure 3

Figure 4

Figure 5

Figure 6

Figure 7

Figure 8

Figure 9

Figure 10

Figure 11

Figure 12

Figure 13

Figure 14

Figure 15

Embodiments for Carrying Out the Invention

[0010] Hereinafter, embodiments embodying the present invention will be described with reference to the drawings.

[0011] With reference to FIG. 1, the configuration of a cell image analysis apparatus 100 according to an embodiment will be described.

[0012] (Configuration of Cell Image Analysis Apparatus) As shown in FIG. 1, the cell image analysis apparatus 100 includes an image acquisition unit 1, a processor 2, a storage unit 3, a display unit 4, and an input reception unit 5.

[0013] The image acquisition unit 1 is configured to acquire the cell image 10. The cell image 10 is an image in which cells 90 (see FIG. 2) are captured. Specifically, the cell image 10 is an image of cultured cells cultured in a container. In the present embodiment, the image acquisition unit 1 is configured to acquire the cell image 10 from a device that captures the cell image 10, such as a microscope 7 to which an imaging device is attached. The image acquisition unit 1 includes, for example, an input / output interface.

[0014] The processor 2 is configured to analyze the acquired cell image 10. The processor 2 includes a CPU (Central Processing Unit), a ROM (Read Only Memory), a RAM (Random Access Memory), a GPU (Graphics Processing Unit), or an FPGA (Field-Programmable Gate Array) configured for image processing. Further, the processor 2 composed of a CPU or the like as hardware includes a control unit 2a, an image analysis unit 2b, an image processing unit 2c, and a superimposed cell image generation unit 2d as functional blocks of software (program). The processor 2 functions as the control unit 2a, the image analysis unit 2b, the image processing unit 2c, and the superimposed cell image generation unit 2d by executing a program stored in the storage unit 3. The control unit 2a, the image analysis unit 2b, the image processing unit 2c, and the superimposed cell image generation unit 2d may be individually configured by hardware by providing dedicated processors (processing circuits).

[0015] The control unit 2a is configured to control the cell image analysis device 100. Further, the control unit 2a is configured to perform control to display the superimposed cell image 50 on the display unit 4. Details of the superimposed cell image 50 will be described later.

[0016] In this embodiment, the image analysis unit 2b analyzes whether the cells 90 shown in the cell image 10 are normal cells or abnormal cells. Specifically, the image analysis unit 2b is configured to obtain an index value 21 (see FIG. 4) obtained by analyzing the cells 90 shown in the cell image 10 using a learned model 6 that has been trained to analyze the cells 90 (see FIG. 2). Details of the index value 21, normal cells, and abnormal cells will be described later.

[0017] The image processing unit 2c is configured to obtain a background label image 11 (see FIG. 4), a normal cell label image 12 (see FIG. 4), and an abnormal cell label image 13 (see FIG. 4) based on the index value 21. Details of the configuration in which the image processing unit 2c obtains each label image will be described later. The image processing unit 2c obtains an abnormal cell region 91 (see FIG. 6A), which is a region where a first index value indicating that the cell 90 is an abnormal cell rather than a normal cell among the index values 21 is larger than a predetermined determination reference value 20.

[0018] The superimposed cell image generation unit 2d is configured to generate a superimposed cell image 50 in which the abnormal cell region 91 can be identified. Details of the configuration in which the superimposed cell image generation unit 2d generates the superimposed cell image 50 will be described later.

[0019] The storage unit 3 is configured to store the cell image 10, the learned model 6, and the determination reference value 20. The predetermined determination reference value 20 includes a first determination reference value 20a and a second determination reference value 20b that is lower than the first determination reference value 20a. Details of the first determination reference value 20a and the second determination reference value 20b will be described later. The storage unit 3 is also configured to store various programs executed by the processor 2. The storage unit 3 includes a storage device such as an HDD (Hard Disk Drive) or an SSD (Solid State Drive), for example.

[0020] The display unit 4 is configured to display the superimposed cell image 50 generated by the superimposed cell image generation unit 2d and the like. The display unit 4 includes a display device such as a liquid crystal monitor, for example.

[0021] The input reception unit 5 is configured to receive operation inputs by an operator. The input reception unit 5 includes input devices such as a mouse and a keyboard.

[0022] (Cell image) Referring to FIG. 2, the cell image 10 will be described. The cell image 10 is an image in which cultured cells are captured. In the present embodiment, the cell image 10 is an image of cells 90 having differentiation ability as cultured cells. For example, the cells 90 include iPS cells (induced Pluripotent Stem Cells), ES cells (Embryonic Stem Cells), and the like. The image analysis unit 2b (see FIG. 1) is configured to analyze whether the cells 90 shown in the cell image 10 are undifferentiated cells or undifferentiated deviated cells. Note that undifferentiated cells are cells having differentiation ability. Also, undifferentiated deviated cells are cells in which differentiation into specific cells has started and which do not have differentiation ability. Cells having differentiation ability such as iPS cells are cultured while remaining in an undifferentiated state. Therefore, in the present embodiment, undifferentiated cells are regarded as normal cells. Also, undifferentiated deviated cells are regarded as abnormal cells. Also, during culturing, the culture environment is controlled so as to maintain the undifferentiated state, so the appearance rate of abnormal cells is sufficiently small. That is, abnormal cells are rare cells with respect to the entire cultured cells.

[0023] Also, as shown in FIG. 2, the cell image 10 may include a region where noise is captured (noise region 80). The noise includes scratches on the culture vessel when culturing the cells 90.

[0024] (Image analysis method) Next, with reference to FIG. 3, a method for analyzing the cell image 10 by the cell image analysis method according to the present embodiment will be described. In the present embodiment, a configuration will be described in which the cell image analysis apparatus 100 analyzes the cell image 10 to determine whether the cell 90 shown in the cell image 10 is a normal cell or an abnormal cell. In the present embodiment, the cell image analysis apparatus 100 determines whether the cell 90 shown in the cell image 10 is a normal cell or an abnormal cell by analyzing the cell image 10 using the learned model 6. When the cell image 10 is input, the learned model 6 outputs an index value 21. The index value 21 includes a first probability value 21a indicating the probability that the cell 90 is an abnormal cell, a second probability value 21b indicating the probability that the cell 90 is a normal cell, a third probability value 21c indicating the probability of being a background, and a fourth probability value 21d indicating the probability of being noise. In the present embodiment, the learned model 6 is learned to output the index value 21 for each pixel of the cell image 10. Note that the first probability value 21a is an example of the "first index value" in the claims.

[0025] FIG. 3 is a block diagram showing the flow of image processing according to the present embodiment. As shown in FIG. 3, in the present embodiment, the cell image analysis method broadly includes an image analysis method 101 and a method 102 for generating the learned model 6.

[0026] (Learning model generation) The method 102 for generating the learned model 6 according to this embodiment generates the learned model 6 by causing the learning model 6a to learn to classify each pixel of the cell image 10 into a normal cell, an abnormal cell, and a background. Specifically, the method 102 for generating the learned model 6 generates the learned model 6 by causing the learning model 6a to learn using the teacher cell image 30 and the teacher label image 31. That is, the method 102 for generating the learned model 6 uses the cell image 10 as the teacher cell image 30 as input data, and the images labeled for normal cells, the images labeled for abnormal cells, the images labeled for the background, and the images labeled for noise as the teacher label image 31 as output data. From this, the method 102 for generating the learned model 6 causes the learning model 6a to learn which of normal cells, abnormal cells, the background, and noise each pixel of the input image is.

[0027] Specifically, the method 102 for generating the learned model 6 includes a step 102a of inputting the teacher cell image 30 to the learning model 6a, and a step 102b of causing the learning model 6a to learn to output the teacher label image 31. The learned model 6 is, for example, a convolutional neural network (CNN) shown in FIG. 3, or includes a convolutional neural network in part. The learned model 6 generated by causing the learning model 6a to learn is stored in the storage unit 3 (FIG. 1) of the cell image analysis apparatus 100. Note that the first probability value 21a is the probability value that the cell 90 shown in the teacher cell image 30 is an abnormal cell. Also, the second probability value 21b is the probability value that the cell 90 shown in the teacher cell image 30 is a normal cell. Also, the third probability value 21c is the probability value that the pixel of the teacher cell image 30 is the background.

[0028] In addition, in this embodiment, in step 102b of training the learning model 6a to output the teacher label image 31, when training the analysis of the state of the cell 90, it is also trained that the region that is similar to the abnormal cell but recognized as the background part is the noise region 80, thereby creating the learned model 6. That is, in this embodiment, the learning model 6a is trained with four classification classes: normal cells, abnormal cells, background, and noise. As a result, the learned model 6 outputs, as the index value 21, a first probability value 21a indicating the probability of being a normal cell, a second probability value 21b indicating the probability of being an abnormal cell, a third probability value 21c indicating the probability of being the background, and a fourth probability value 21d indicating the probability of being noise for the input cell image 10. When training that it is the noise region 80, the training is performed by using a teacher image in which an operator has previously labeled the noise region 80. As a result, compared with the configuration of training three classification classes: normal cells, abnormal cells, and background, the number of classification classes to be discriminated can be increased. Therefore, among the three classification classes, abnormal cells that were classified as the background may slightly increase in the appearance of abnormal cells due to the increase in the noise classification class. As a result, by suppressing the classification of abnormal cells as the background, the classification accuracy of the learned model 6 can be improved.

[0029] (Image Analysis Method) The image analysis method 101 according to this embodiment is an image analysis method for classifying the cells 90 shown in the cell image 10 acquired by the image acquisition unit 1 from a microscope 7 or the like into normal cells and abnormal cells. The image analysis method 101 according to this embodiment includes a step of acquiring the cell image 10, a step of acquiring the index value 21 of each pixel of the cell image 10, a step of acquiring the abnormal cell region 91, and a step of displaying the abnormal cell region 91 in an identifiable manner. Details of the processing of each step of the image analysis method 101 will be described later.

[0030] In this embodiment, as shown in FIG. 3, the step of acquiring the cell image 10 is performed by the image acquisition unit 1. The image acquisition unit 1 acquires the cell image 10 from an image capturing device such as a microscope 7. Further, the image acquisition unit 1 outputs the acquired cell image 10 to the image analysis unit 2b.

[0031] Also, in this embodiment, as shown in FIG. 3, the step of analyzing the cell image 10 is performed by the image analysis unit 2b. The image analysis unit 2b inputs the cell image 10 into the learned model 6 to obtain the index value 21 for each pixel of the input cell image 10. Further, the image analysis unit 2b outputs the acquired index value 21 to the image processing unit 2c.

[0032] The image processing unit 2c obtains the abnormal cell region 91 from the cell image 10 based on the index value 21. The image processing unit 2c outputs the acquired abnormal cell region 91 to the superimposed cell image generation unit 2d.

[0033] The superimposed cell image generation unit 2d generates a superimposed cell image 50 based on the abnormal cell region 91 and the cell image 10, and causes the display unit 4 to display it. In this embodiment, the cell image analysis apparatus 100 displays the abnormal cell region 91 in an identifiable manner by displaying the superimposed cell image 50. Details of the superimposed cell image 50 will be described later.

[0034] (Label Image Acquisition Process) In this embodiment, the image processing unit 2c obtains the abnormal cell region 91 based on the index value 21 output by the learning model 6a. Specifically, as shown in FIG. 4, the image processing unit 2c obtains the abnormal cell region 91 by obtaining a label image based on the index value 21.

[0035] As shown in FIG. 4, the image analysis unit 2b obtains the index value 21 by inputting the cell image 10 into the learned model 6. The image analysis unit 2b obtains the index value 21 for each pixel of the cell image 10. In this embodiment, the image analysis unit 2b outputs the first probability value 21a, the second probability value 21b, the third probability value 21c, and the fourth probability value 21d as the index value 21.

[0036] The image processing unit 2c (see FIG. 1) acquires a background label image 11, a normal cell label image 12, an abnormal cell label image 13, and a noise region label image (not shown) based on the index value 21 output by the learned model 6.

[0037] The background label image 11 is an image with the value of the third probability value 21c indicating the probability of being a background as a pixel value. Specifically, in the background label image 11, the larger the value of the third probability value 21c, the larger (blacker) the pixel value, and the smaller the value of the third probability value 21c, the smaller (whiter) the pixel value. Also, in the hatched portion, the pixel value is larger as the hatching becomes darker. In the background label image 11, as shown in the legend 8, by applying one hatching, it is illustrated that a probability value within a predetermined range is included. For example, in the black region of the background label image 11, the third probability value 21c indicating the probability of being a background is greater than 80% and less than or equal to 100%.

[0038] The normal cell label image 12 is an image with the value of the second probability value 21b indicating the probability of being a normal cell as a pixel value. Specifically, in the normal cell label image 12, the larger the value of the second probability value 21b, the larger (blacker) the pixel value, and the smaller the value of the second probability value 21b, the smaller (whiter) the pixel value. Also, in the hatched portion, the pixel value is larger as the hatching becomes darker. In the normal cell label image 12 as well, as shown in the legend 8, by applying one hatching, it is illustrated that a probability value within a predetermined range is included. For example, in the black region of the normal cell label image 12, the first probability value 21a indicating the probability of being a normal cell is greater than 80% and less than or equal to 100%.

[0039] The abnormal cell label image 13 is an image with the value of the first probability value 21a indicating the probability of being an abnormal cell as the pixel value. Specifically, in the abnormal cell label image 13, the larger the value of the first probability value 21a, the larger (blacker) the pixel value, and the smaller the value of the first probability value 21a, the smaller (whiter) the pixel value. Also, in the hatched portion, the darker the hatching, the larger the pixel value. Note that in the abnormal cell label image 13 as well, as shown in the legend 8, by applying one hatching, it is illustrated that a probability value within a predetermined range is included. For example, in the black region in the abnormal cell label image 13, the second probability value 21b indicating the probability of being an abnormal cell is larger than 80% and means a probability value of 100% or less.

[0040] (Pixel value of abnormal cell label image) Here, with reference to FIG. 5, the distribution of the pixel values of the abnormal cell label image 13 will be described.

[0041] FIG. 5(A) is a schematic diagram showing the abnormal cell label image 13. Also, FIG. 5(B) is an enlarged image 15 obtained by enlarging the region 14 of the abnormal cell label image 13 in FIG. 5(A). Further, FIG. 5(C) is a graph 17 in which pixel values are plotted along the straight line 16 shown in the enlarged image 15. The graph 17 is a graph in which the vertical axis represents the pixel value and the horizontal axis represents the position of the pixel.

[0042] The enlarged image 15 shown in FIG. 5(B) is an image obtained by enlarging the region 14, which is a region with a small pixel value, in the abnormal cell label image 13. As shown in the graph 17 of FIG. 5(C), although the pixel value is small, there is a possibility of being an abnormal cell, although with a low probability.

[0043] Here, when determining whether each pixel is a normal cell, an abnormal cell, or a background based on the order of the magnitudes of the index values 21 output by the learned model 6, the region 14 of the abnormal cell label image 13 shown in FIG. 5(A) may not be determined as an abnormal cell. Specifically, among the index values 21 of the same pixel as the region 14, if the first probability value 21a is smaller than the second probability value 21b and the third probability value 21c, it is not determined as an abnormal cell. However, if the first probability value 21a is not 0, it has a slight possibility of being an abnormal cell. Therefore, in the case of determination based on the magnitude of the index value 21, the region of the cell 90 shown in the enlarged image 15 may not be recognized as the abnormal cell region 91 by the operator.

[0044] Therefore, in the present embodiment, the image processing unit 2c is configured to acquire the abnormal cell region 91 based on the determination reference value 20 and the index value 21. Specifically, the image processing unit 2c performs so-called threshold processing on the index value 21 using the determination reference value 20 to acquire the abnormal cell region 91. In the present embodiment, the image processing unit 2c is configured to acquire the abnormal cell region 91, which is a region where the first index value (first probability value 21a) indicating that the cell 90 is an abnormal cell rather than a normal cell among the index values 21 is larger than a predetermined determination reference value 20.

[0045] (Difference in Abnormal Cell Label Image Due to Difference in Determination Criterion Value) Next, with reference to FIG. 6, the difference in the abnormal cell label image 13 due to the difference in the determination criterion will be described.

[0046] The schematic diagrams shown in FIGS. 6(A) to 6(E) are abnormal cell label images 13 when the determination reference value 20 is changed. Specifically, the schematic diagrams shown in FIGS. 6(A) to 6(E) are abnormal cell label images 13 when the determination reference value 20 is changed to a value greater than 0%, 10% or more, 20% or more, 30% or more, and 40% or more. Further, the abnormal cell label images 13a to 13e shown in FIGS. 6(A) to 6(E) illustrate the abnormal cell region 91 as a black region. Also, the abnormal cell label images 13a to 13e shown in FIGS. 6(A) to 6(E) illustrate the regions other than the abnormal cell region 91 as white regions.

[0047] FIG. 6(A) is a schematic diagram showing the abnormal cell label image 13a when the determination reference value 20 is set to a value greater than 0. That is, the abnormal cell label image 13a is a label image obtained by imaging the abnormal cell region 91 in which the first probability value 21a is greater than 0.

[0048] Further, FIG. 6(B) is a schematic diagram showing the abnormal cell label image 13b when the determination reference value 20 is set to a value of 10% or more. That is, the abnormal cell label image 13b is a label image obtained by imaging the abnormal cell region 91 in which the first probability value 21a is 10% or more. In FIG. 6(B), the abnormal cell region 91 when the determination reference value 20 is greater than 0 is illustrated by a dashed line 70. As shown in FIG. 6(B), when the determination reference value 20 is set to a value of 10% or more, the size of the abnormal cell region 91 becomes smaller.

[0049] Further, FIG. 6(C) is a schematic diagram showing an abnormal cell label image 13c when the determination reference value 20 is set to a value of 20% or more. That is, the abnormal cell label image 13c is a label image obtained by imaging an abnormal cell region 91 where the first probability value 21a is 20% or more. In FIG. 6(C) as well, the abnormal cell region 91 when the determination reference value 20 is greater than 0 is illustrated by a dashed line 70. As shown in FIG. 6(C), when the determination reference value 20 is set to a value of 20% or more, the size of the abnormal cell region 91 becomes smaller. Also, when comparing the abnormal cell label image 13b shown in FIG. 6(B) and the abnormal cell label image 13c shown in FIG. 6(C), the abnormal cell region 91 shown in the abnormal cell label image 13c in FIG. 6(C) is smaller than the abnormal cell region 91 shown in the abnormal cell label image 13b in FIG. 6(B).

[0050] Further, FIG. 6(D) is an abnormal cell label image 13d when the determination reference value 20 is set to a value of 30% or more. That is, the abnormal cell label image 13d is a label image obtained by imaging an abnormal cell region 91 where the first probability value 21a is 30% or more. In FIG. 6(D) as well, the abnormal cell region 91 when the determination reference value 20 is greater than 0 is illustrated by a dashed line 70. As shown in FIG. 6(D), when the determination reference value 20 is set to a value of 30% or more, the size of the abnormal cell region 91 becomes smaller. Also, as shown in FIGS. 6(B) to 6(D), as the value of the determination reference value 20 increases, the abnormal cell region 91 shown in the abnormal cell label image 13 becomes smaller.

[0051] FIG. 6(E) is a schematic diagram showing an abnormal cell label image 13e when the determination reference value 20 is set to a value of 40% or more. That is, the abnormal cell label image 13e is a label image obtained by imaging an abnormal cell region 91 where the first probability value 21a is 40% or more. In FIG. 6(E) as well, the abnormal cell region 91 when the determination reference value 20 is greater than 0 is illustrated by a dashed line 70. As shown in FIG. 6(E), when the determination reference value 20 is set to a value of 40% or more, the size of the abnormal cell region 91 becomes smaller. Also, as shown in FIGS. 6(B) to 6(E), as the value of the determination reference value 20 increases, the abnormal cell region 91 shown in the abnormal cell label image 13 becomes smaller.

[0052] That is, as shown in FIGS. 6(A) to 6(E), as the value of the determination reference value 20 increases, the size of the abnormal cell region 91 becomes smaller. In other words, the smaller the value of the determination reference value 20, the larger the number and area of the regions acquired as the abnormal cell region 91. However, in reality, the possibility of acquiring regions that are not actually abnormal cells increases. That is, there are cases where regions that do not need to be discriminably displayed as the abnormal cell region 91 are acquired as the abnormal cell region 91. Also, the larger the value of the determination reference value 20, the higher the possibility that the cells 90 included in the acquired abnormal cell region 91 are abnormal cells. However, the number and area of the acquired abnormal cell region 91 become smaller. That is, there may be cases where an acquisition omission of the abnormal cell region 91 occurs.

[0053] Therefore, in the present embodiment, as shown in FIG. 7, the image processing unit 2c is configured to acquire the abnormal cell region 91 using a plurality of determination reference values 20. Specifically, the image processing unit 2c is configured to acquire the abnormal cell region 91 using two determination reference values 20. In the present embodiment, the image processing unit 2c is configured to acquire the abnormal cell region 91 using a first determination reference value 20a and a second determination reference value 20b that is lower than the first determination reference value 20a. Hereinafter, in the present embodiment, the first determination reference value 20a is set to 10% or more, and the second determination reference value 20b is set to a value greater than 0.

[0054] In this embodiment, the image processing unit 2c acquires, as a first abnormal cell region 91a, an abnormal cell region 91 in which the first index value is larger than the first determination reference value 20a, such as the abnormal cell region image 13f shown in FIG. 7. Specifically, the image processing unit 2c acquires, as the first abnormal cell region 91a, an abnormal cell region 91 in which the first probability value 21a is larger than the first determination reference value 20a. Further, the image processing unit 2c acquires, as a second abnormal cell region 91b, an abnormal cell region 91 in which the first index value is larger than the second determination reference value 20b, such as the abnormal cell region image 13f shown in FIG. 7. Specifically, the image processing unit 2c acquires, as the second abnormal cell region 91b, an abnormal cell region 91 in which the first probability value 21a is larger than the second determination reference value 20b.

[0055] (Acquisition of cell region) Here, when determining whether the index value 21 of all the pixels of the cell image 10 is larger than the first determination reference value 20a and the second determination reference value 20b, the processing load of the image processing unit 2c increases. Therefore, as shown in FIG. 8, in this embodiment, the image processing unit 2c determines whether the index value 21 of each pixel of the cell region 93, which is the cell 90 shown in the cell image 10, is larger than the first determination reference value 20a and the second determination reference value 20b. In this embodiment, the image processing unit 2c is configured to acquire a cell region 93, which is the region of the cell 90, and a region 94 other than the cell 90 based on the normal cell label image 12 and the abnormal cell label image 13. Specifically, the image processing unit 2c acquires a cell region image 18 by adding the pixel values at the same positions of the normal cell label image 12 and the abnormal cell label image 13. Note that the cell region image 18 shown in FIG. 8 is a binary image in which the pixel value of the cell region 93 is 1 and the pixel value of the region 94 other than the cell 90 is 0. That is, the region illustrated in black in the cell region image 18 is the cell region 93. Further, the region illustrated in white in the cell region image 18 is the region 94 other than the cell 90.

[0056] (Overlapped cell image) Next, with reference to FIG. 9, the configuration in which the overlapping cell image generation unit 2d generates the overlapping cell image 50 will be described. FIG. 9 shows an example in which the overlapping cell image generation unit 2d generates the overlapping cell image 50 shown in FIG. 9(B) from the cell image 10 shown in FIG. 9(A).

[0057] The overlapping cell image generation unit 2d generates the overlapping cell image 50 by superimposing the abnormal cell region 91 on the cell image 10. Specifically, the overlapping cell image generation unit 2d generates the overlapping cell image 50 by superimposing the first abnormal cell region 91a and the second abnormal cell region 91b on the cell image 10.

[0058] Also, in the present embodiment, the overlapping cell image generation unit 2d generates an image in which the abnormal cell region 91 can be identified as the overlapping cell image 50. Specifically, the overlapping cell image generation unit 2d generates, as the overlapping cell image 50, an image in which the entire inside of the first abnormal cell region 91a is displayed in a predetermined color. The overlapping cell image generation unit 2d, for example, displays the entire inside of the first abnormal cell region 91a in a colored state of red. In the example shown in FIG. 9(B), it is illustrated that the entire inside of the first abnormal cell region 91a is displayed in a colored state by hatching the first abnormal cell region 91a with a predetermined color.

[0059] Also, the overlapping cell image generation unit 2d generates, as the overlapping cell image 50, an image in which the frame line 91c surrounding the second abnormal cell region 91b is emphasized. In the example shown in FIG. 9(B), the frame line 91c is emphasized by illustrating the frame line 91c with a thick line. Note that the frame line 91c is the contour line of the second abnormal cell region 91b. In the present embodiment, the overlapping cell image generation unit 2d superimposes a thick line (frame line 91c) indicating the contour line of the cell region 93 of the cell image 10, and in the abnormal cell region 91, divides the second probability value 21b into two stages, and generates a colored image in which each region corresponding to the second probability value 21b is stained with two colors.

[0060] In addition, the superimposed cell image generation unit 2d generates, as a superimposed cell image 50, an image in which the normal cell region 92 is displayed in a color different from that of the abnormal cell region 91. That is, the superimposed cell image generation unit 2d displays the normal cell region 92 and the abnormal cell region 91 in different colors, and generates, as the superimposed cell image 50, an image in which the entire area within the first abnormal cell region 91a is displayed in a predetermined color and the frame line 91c surrounding the second abnormal cell region 91b is highlighted. For the abnormal cell region 91, for example, in the present embodiment, the superimposed cell image generation unit 2d displays the normal cell region 92 in a state of being colored blue.

[0061] (Superimposed cell image according to the comparative example) FIG. 10 shows a comparative example in which a superimposed cell image 50 shown in FIG. 10(B) is generated from a cell image 10 shown in FIG. 10(A) by obtaining the normal cell region 92 and the abnormal cell region 91 according to the magnitude of the index value 21.

[0062] The superimposed cell image 40 according to the comparative example is an image in which the normal cell region 96 and the abnormal cell region 95 are obtained based on the largest index value 21 for each pixel and superimposed on the cell image 10. Specifically, at each pixel, the probability value with the largest value among the first probability value 21a, the second probability value 21b, and the third probability value 21c is obtained, and the normal cell region 96 and the abnormal cell region 95 are obtained. Therefore, when the first probability value 21a indicating the probability of being an abnormal cell is smaller than the second probability value 21b indicating the probability of being a normal cell and the third probability value 21c indicating the probability of being a background, it is not displayed as the abnormal cell region 91. Therefore, the area of the abnormal cell region 95 shown in the superimposed cell image 40 according to the comparative example is smaller than the area of the abnormal cell region 91 shown in the superimposed cell image 50 according to the present embodiment. That is, in the superimposed cell image 40 according to the comparative example, the abnormal cell region 95 is difficult to grasp at a glance.

[0063] When comparing the superimposed cell image 50 according to this embodiment with the superimposed cell image 40 according to the comparative example, more and wider abnormal cell regions 91 are displayed in the superimposed cell image 50 according to this embodiment. Therefore, the operator can more accurately grasp the abnormal cell regions 91 that may be abnormal cells.

[0064] (Number of abnormal cells) Here, there may be cases where the operator wants to confirm the number of abnormal cells among the cells 90 shown in the cell image 10. Therefore, in this embodiment, the image processing unit 2c is configured to acquire the number of abnormal cells shown in the cell image 10. When counting the number of abnormal cells, if adjacent abnormal cells are counted individually, the ratio of abnormal cells to normal cells may become too high. Therefore, in this embodiment, in some cases, it is preferable that the image processing unit 2c measures the abnormal cell region 91 as one abnormal cell instead of counting the abnormal cells individually. Therefore, in this embodiment, the image processing unit 2c is configured to acquire the number 60 of the second abnormal cell regions 91b shown in the cell image 10 as the number of abnormal cells.

[0065] (Ratio of the area of the first abnormal cell region to the area of the second abnormal cell region) The second abnormal cell region 91b is a region where the possibility of being an abnormal cell is low, and the first abnormal cell region 91a is a region where the possibility of being an abnormal cell is high. The higher the ratio of the area of the first abnormal cell region 91a included in the second abnormal cell region 91b, the higher the possibility of being an abnormal cell. However, it is difficult to easily grasp the ratio of the area of the first abnormal cell region 91a included in the second abnormal cell region 91b just by a cursory look at the superimposed cell image 50.

[0066] Therefore, in this embodiment, the image processing unit 2c is configured to acquire the ratio 61 of the area of the first abnormal cell region 91a to the area of the second abnormal cell region 91b. Further, the image processing unit 2c is configured to cause the display unit 4 to display the acquired ratio 61 of the area of the first abnormal cell region 91a to the area of the second abnormal cell region 91b.

[0067] (Selection of Abnormal Cell Regions to be Identifiably Displayed) Next, with reference to FIG. 11, a configuration in which the image processing unit 2c selects abnormal cell regions 91 to be identifiably displayed will be described. In the present embodiment, the image processing unit 2c acquires abnormal cell regions 91 based on a first determination reference value 20a and a second determination reference value 20b. Here, if something with an extremely low possibility of being an abnormal cell region 91 is displayed, it will interfere with the operator when checking the abnormal cell region 91. Therefore, in the present embodiment, the image processing unit 2c is configured to select the abnormal cell regions 91 to be identifiably displayed.

[0068] In the abnormal cell label image 13g shown in FIG. 11, as shown in region 19a, a second abnormal cell region 91b including the first abnormal cell region 91a inside is shown. Also, in the abnormal cell label image 13g, as shown in region 19b, a second abnormal cell region 91b with an extremely small area is shown. Further, in the abnormal cell label image 13g, as shown in region 19c, a second abnormal cell region 91b that does not include the first abnormal cell region 91a inside is shown.

[0069] In the present embodiment, the image processing unit 2c excludes, from the objects to be identifiably displayed from the first abnormal cell region 91a, the second abnormal cell regions 91b having an area equal to or less than a predetermined area among the second abnormal cell regions 91b.

[0070] Also, in the present embodiment, the image processing unit 2c excludes, from the objects to be identifiably displayed from the first abnormal cell region 91a, the second abnormal cell regions 91b that do not include the first abnormal cell region 91a inside. That is, as in the abnormal cell label image 13h shown in FIG. 11, the image processing unit 2c targets only the second abnormal cell regions 91b that are larger than a predetermined area and include the first abnormal cell region 91a inside as the objects to be identifiably displayed from the first abnormal cell region 91a.

[0071] FIG. 12 is a schematic diagram showing an example in which the display unit 4 displays the overlapping cell image 50, the number 60 of the second abnormal cell regions 91b, and the ratio 61 of the area of the first abnormal cell region 91a to the area of the second abnormal cell region 91b.

[0072] As shown in FIG. 12, the image processing unit 2c causes the display unit 4 to display the superimposed cell image 50. Further, the image processing unit 2c causes the display unit 4 to display the number 60 of the second abnormal cell regions 91b. Further, the image processing unit 2c causes the display unit 4 to display the ratio 61 of the area of the first abnormal cell region 91a to the area of the second abnormal cell region 91b. In the present embodiment, the image processing unit 2c displays the superimposed cell image 50, the number 60 of the second abnormal cell regions 91b, and the ratio 61 of the area of the first abnormal cell region 91a to the area of the second abnormal cell region 91b in a side-by-side state.

[0073] (Superimposed Cell Image Display Process) Next, with reference to FIG. 13, the process by which the cell image analysis apparatus 100 displays the superimposed cell image 50 will be described.

[0074] In step 200, the image acquisition unit 1 acquires the cell image 10 in which the cells 90 are imaged.

[0075] In step 201, the image analysis unit 2b acquires the index value 21 obtained by analyzing the cells 90 imaged in the cell image 10 using the learned model 6 in which the analysis of the cells 90 has been learned. Specifically, in step 201, the image analysis unit 2b acquires, as the first index value, the first probability value 21a indicating the probability that the cell 90 is an abnormal cell for each pixel of the cell image 10. More specifically, in step 201, the image analysis unit 2b acquires, as the index value 21, the second probability value 21b indicating the probability that the cultured cell is a normal cell together with the first probability value 21a. In the present embodiment, the image analysis unit 2b acquires the index value 21 using the learned model 6 in which the state of the cells 90 and whether or not it is the noise region 80 are determined.

[0076] In step 202, the image processing unit 2c acquires the cell region 93 that is the region of the cells 90 and the region 94 other than the cells 90.

[0077] In step 203, the image processing unit 2c acquires an abnormal cell region 91 in which the first index value is greater than the first determination reference value 20a as a first abnormal cell region 91a. Specifically, the image processing unit 2c acquires an abnormal cell region 91 in which the first probability value 21a is greater than the first determination reference value 20a as the first abnormal cell region 91a. Note that in step 203, the image processing unit 2c acquires, as the first abnormal cell region 91a, a region in which the first probability value 21a for each pixel of the cell region 93 is greater than the first determination reference value 20a. Also, in step 203, even when the first probability value 21a is smaller than the second probability value 21b, if the first probability value 21a is greater than the first determination reference value 20a, the image processing unit 2c acquires it as the first abnormal cell region 91a.

[0078] In step 204, the image processing unit 2c acquires an abnormal cell region 91 that is a region where the first index value indicating that the cell 90 is an abnormal cell rather than a normal cell among the index values 21 is greater than a predetermined determination reference value 20. Specifically, the image processing unit 2c acquires an abnormal cell region 91 in which the first index value is greater than the second determination reference value 20b as a second abnormal cell region 91b. More specifically, the image processing unit 2c acquires an abnormal cell region 91 in which the first probability value 21a is greater than the second determination reference value 20b as the second abnormal cell region 91b. In the present embodiment, the image processing unit 2c acquires, as the second abnormal cell region 91b, a region in which the first probability value 21a is greater than 0 as the second determination reference value 20b. Also, even when the first probability value 21a is smaller than the second probability value 21b, if the first probability value 21a is greater than the second determination reference value 20b, the image processing unit 2c acquires it as the second abnormal cell region 91b. Further, the image processing unit 2c acquires, as the second abnormal cell region 91b, a region in which the first probability value 21a for each pixel of the cell region 93 is greater than the second determination reference value 20b.

[0079] In step 205, a normal cell region 92 is acquired.

[0080] In step 206, the image processing unit 2c excludes, from the targets to be displayed in a distinguishable manner from the first abnormal cell region 91a, the second abnormal cell regions 91b having an area equal to or smaller than a predetermined area among the second abnormal cell regions 91b.

[0081] In step 207, the image processing unit 2c excludes, from the targets to be displayed in a distinguishable manner from the first abnormal cell region 91a, the second abnormal cell regions 91b that do not contain the first abnormal cell region 91a inside.

[0082] In step 208, the superimposed cell image generation unit 2d generates a superimposed cell image 50 based on the cell image 10, the first abnormal cell region 91a, the second abnormal cell region 91b, and the normal cell region 92. Specifically, the superimposed cell image generation unit 2d generates the superimposed cell image 50 by superimposing the first abnormal cell region 91a, the second abnormal cell region 91b, and the normal cell region 92 on the cell image 10.

[0083] In step 209, the image processing unit 2c displays the superimposed cell image 50 on the display unit 4. That is, the image processing unit 2c displays the superimposed cell image 50 on the display unit 4, thereby displaying the abnormal cell region 91 in a distinguishable manner. In addition, the image processing unit 2c displays the superimposed cell image 50 on the display unit 4, thereby displaying the first abnormal cell region 91a and the second abnormal cell region 91b in a distinguishable manner. Further, the image processing unit 2c displays the entire inside of the first abnormal cell region 91a in a predetermined color and highlights the frame line 91c surrounding the second abnormal cell region 91b by displaying the superimposed cell image 50 on the display unit 4. In addition, the image processing unit 2c superimposes and displays the first abnormal cell region 91a and the second abnormal cell region 91b on the cell image 10 by displaying the superimposed cell image 50 on the display unit 4. Further, the image processing unit 2c superimposes and displays the first abnormal cell region 91a, the second abnormal cell region 91b, and the region of normal cells on the cell image 10 by displaying the superimposed cell image 50 on the display unit 4. Thereafter, the process ends.

[0084] Note that the processes from step 203 to step 205 may start from any step. Also, the processes of step 206 and step 207 may start from either process.

[0085] (Abnormal cell count display process) Next, referring to FIG. 14, the process in which the image processing unit 2c acquires and displays the count 60 of the second abnormal cell region 91b will be described. Note that the process in which the image processing unit 2c acquires the count of abnormal cells is performed after step 207 of generating the superimposed cell image 50 is completed.

[0086] In step 300, the image processing unit 2c acquires the count 60 of the second abnormal cell region 91b shown in the cell image 10.

[0087] In step 301, the image processing unit 2c displays the acquired count 60 of the second abnormal cell region 91b. Then, the process ends.

[0088] (Display process of ratio of area of first abnormal cell region to area of second abnormal cell region) Next, referring to FIG. 15, the process in which the image processing unit 2c acquires and displays the ratio 61 of the area of the first abnormal cell region 91a to the area of the second abnormal cell region 91b will be described. Note that the process in which the image processing unit 2c acquires and displays the ratio 61 of the area of the first abnormal cell region 91a to the area of the second abnormal cell region 91b is started when there is an input in which the operator selects the abnormal cell region 91 in the superimposed cell image 50.

[0089] In step 400, the area of the first abnormal cell region 91a is acquired.

[0090] In step 401, the area of the second abnormal cell region 91b is acquired.

[0091] In step 402, the image processing unit 2c acquires the ratio 61 of the area of the first abnormal cell region 91a to the area of the second abnormal cell region 91b.

[0092] In step 403, the image processing unit 2c displays the ratio 61 of the area of the first abnormal cell region 91a to the area of the second abnormal cell region 91b. Thereafter, the process ends.

[0093] (Effect of this embodiment) In this embodiment, the following effects can be obtained.

[0094] In this embodiment, as described above, the cell image analysis method includes a step of acquiring a cell image 10 in which a cell 90 appears, a step of acquiring an index value 21 obtained by analyzing the cell 90 appearing in the cell image 10 using a learned model 6 in which the analysis of the cell 90 is learned, a step of acquiring an abnormal cell region 91 which is a region where a first index value indicating that the cell 90 is an abnormal cell rather than a normal cell among the index values 21 is larger than a predetermined determination reference value 20, and a step of displaying the abnormal cell region 91 in an identifiable manner.

[0095] As described above, by providing the step of acquiring the abnormal cell region 91 which is a region where the first index value indicating that the cell 90 is an abnormal cell rather than a normal cell among the index values 21 is larger than the predetermined determination reference value 20, a region where the first index value is larger than the predetermined determination reference value 20 is acquired as the abnormal cell region 91. Therefore, even when the first index value is smaller than other index values 21, it can be acquired as the abnormal cell region 91. Further, as described above, by providing the step of displaying the abnormal cell region 91 in an identifiable manner, for example, when the determination reference value 20 is decreased, the acquired abnormal cell region 91 is displayed in an identifiable manner as a region suspected of being an abnormal cell. Therefore, the abnormal cell region 91 having even a slight possibility of being an abnormal cell can be visually displayed. As a result, it is possible to provide a cell image analysis method capable of grasping a region (abnormal cell region 91) suspected of being an abnormal cell in the cell image 10.

[0096] Also, in the above embodiment, by configuring as follows, the following further effects can be obtained.

[0097] That is, in the present embodiment, as described above, the predetermined determination reference value 20 includes a first determination reference value 20a and a second determination reference value 20b that is lower than the first determination reference value 20a. The step of acquiring the abnormal cell region 91 includes a step of acquiring, as a first abnormal cell region 91a, an abnormal cell region 91 in which the first index value is greater than the first determination reference value 20a, and a step of acquiring, as a second abnormal cell region 91b, an abnormal cell region 91 in which the first index value is greater than the second determination reference value 20b. In the step of displaying the abnormal cell region 91 in a distinguishable manner, the first abnormal cell region 91a and the second abnormal cell region 91b are displayed in a distinguishable manner. As a result, the operator can grasp the first abnormal cell region 91a determined to be an abnormal cell by the first determination reference value 20a and the second abnormal cell region 91b determined to be an abnormal cell by the second determination reference value 20b that is smaller than the first determination reference value 20a. As a result, the operator can distinguishablely grasp the second abnormal cell region 91b that has a possibility of being an abnormal cell even slightly and the first abnormal cell region 91a that has a higher possibility of being an abnormal cell than the second abnormal cell region 91b, so that the abnormal cell region 91 can be grasped more accurately.

[0098] Also, in the present embodiment, as described above, in the step of acquiring the index value 21, as the first index value, a first probability value 21a indicating the probability that the cell 90 is an abnormal cell is acquired for each pixel of the cell image 10. In the step of acquiring the first abnormal cell region 91a, an abnormal cell region 91 in which the first probability value 21a is greater than the first determination reference value 20a is acquired as the first abnormal cell region 91a. In the step of acquiring the second abnormal cell region 91b, an abnormal cell region 91 in which the first probability value 21a is greater than the second determination reference value 20b is acquired as the second abnormal cell region 91b. Thereby, by comparing the first probability value 21a with the first determination reference value 20a and the second determination reference value 20b, it is possible to easily determine whether the cell 90 is in the first abnormal cell region 91a or the second abnormal cell region 91b.

[0099] Also, in the present embodiment, as described above, the cell image 10 is an image in which cultured cells are imaged. In the step of obtaining the index value 21, as the index value 21, together with the first probability value 21a, a second probability value 21b indicating the probability that the cultured cells are normal cells is obtained. In the step of obtaining the first abnormal cell region 91a, even when the first probability value 21a is smaller than the second probability value 21b, if the first probability value 21a is larger than the first determination reference value 20a, it is obtained as the first abnormal cell region 91a. In the step of obtaining the second abnormal cell region 91b, even when the first probability value 21a is smaller than the second probability value 21b, if the first probability value 21a is larger than the second determination reference value 20b, it is obtained as the second abnormal cell region 91b. As a result, even when the first probability value 21a is smaller than the second probability value 21b, if it is larger than the first determination reference value 20a or the second determination reference value 20b, it is determined that it is the abnormal cell region 91. Therefore, the abnormal cell region 91 can be obtained regardless of the magnitude of the value of the first probability value 21a. As a result, since it is possible to suppress the acquisition as the normal cell region 92 due to the first probability value 21a being smaller than the second probability value 21b, the operator can accurately grasp the abnormal cell region 91.

[0100] Also, in the present embodiment, as described above, in the step of obtaining the second abnormal cell region 91b, as the second determination reference value 20b, a region where the first probability value 21a is larger than 0 is obtained as the second abnormal cell region 91b. As a result, if the first probability value 21a is larger than 0, it can be obtained as the second abnormal cell region 91b. As a result, it is possible to suppress the omission of the acquisition of the second abnormal cell region 91b, so that the operator can more accurately grasp the abnormal cell region 91.

[0101] Further, in the present embodiment, as described above, the method further includes a step of obtaining a cell region 93, which is a region of the cell 90, and a region 94 other than the cell 90 from among the cell images 10. In the step of obtaining the first abnormal cell region 91a, a region where the first probability value 21a for each pixel of the cell region 93 is greater than the first determination reference value 20a is obtained as the first abnormal cell region 91a. In the step of obtaining the second abnormal cell region 91b, a region where the first probability value 21a for each pixel of the cell region 93 is greater than the second determination reference value 20b is obtained as the second abnormal cell region 91b. Thereby, it becomes possible to determine whether a cell is an abnormal cell only for the cell region 93. As a result, for example, compared with a configuration for determining whether a cell is an abnormal cell with respect to the index value 21 of all the pixels of the cell image 10, the processing load can be reduced.

[0102] Further, in the present embodiment, as described above, in the step of displaying the abnormal cell region 91 in a distinguishable manner, the entire inside of the first abnormal cell region 91a is displayed in a predetermined color, and a frame line 91c surrounding the second abnormal cell region 91b is highlighted. Thereby, the first abnormal cell region 91a and the second abnormal cell region 91b can be easily distinguished at a glance.

[0103] Further, in the present embodiment, as described above, in the step of displaying the abnormal cell region 91 in a distinguishable manner, the first abnormal cell region 91a and the second abnormal cell region 91b are superimposed and displayed on the cell image 10. Thereby, in the cell image 10, the first abnormal cell region 91a and the second abnormal cell region 91b can be easily distinguished. As a result, in the cell image 10, the operator can easily grasp the first abnormal cell region 91a with a strong suspicion of being an abnormal cell and the second abnormal cell region 91b with a weak suspicion of being an abnormal cell.

[0104] Also, in the present embodiment, as described above, in the step of visibly displaying the abnormal cell region 91, the first abnormal cell region 91a, the second abnormal cell region 91b, and the region 92 of normal cells are superimposed and displayed on the cell image 10. Thereby, in the cell image 10, normal cells and abnormal cells can be easily distinguished by the operator.

[0105] Also, in the present embodiment, as described above, in the second abnormal cell region 91b, a step of excluding the second abnormal cell region 91b having an area equal to or less than a predetermined area from the object to be visibly displayed separately from the first abnormal cell region 91a is further provided. Thereby, since the second determination reference value 20b is low, a region determined as the second abnormal cell region 91b but not an abnormal cell can be excluded from the object to be visibly displayed. As a result, it is possible to prevent the operator from recognizing a region other than abnormal cells as the abnormal cell region 91.

[0106] Also, in the present embodiment, as described above, a step of excluding the second abnormal cell region 91b that does not include the first abnormal cell region 91a inside from the object to be visibly displayed separately from the first abnormal cell region 91a is further provided. Thereby, among the second abnormal cell regions 91b, a region that does not contain abnormal cells can be excluded from the object to be visibly displayed. As a result, it is possible to prevent the operator from recognizing the second abnormal cell region 91b that does not contain abnormal cells as the abnormal cell region 91.

[0107] Also, in the present embodiment, as described above, a step of acquiring the number 60 of the second abnormal cell regions 91b shown in the cell image 10 and a step of displaying the acquired number 60 of the second abnormal cell regions 91b are further provided. Thereby, a region having even a slight possibility of being an abnormal cell can be counted as one abnormal cell. As a result, compared with the case of counting abnormal cells individually, it is possible to suppress an increase in the number of abnormal cells, so that it is possible to suppress an increase in the ratio of abnormal cells to normal cells.

[0108] In addition, in the present embodiment, as described above, there are further provided a step of obtaining a ratio 61 of the area of the first abnormal cell region 91a to the area of the second abnormal cell region 91b, and a step of displaying the ratio 61 of the area of the first abnormal cell region 91a to the area of the second abnormal cell region 91b. Thereby, when the ratio 61 of the area of the first abnormal cell region 91a to the area of the second abnormal cell region 91b is large, it can be determined that there is a strong suspicion that the cell is an abnormal cell. Also, when the ratio 61 of the area of the first abnormal cell region 91a to the area of the second abnormal cell region 91b is low, it can be determined that there is a weak suspicion that the cell is an abnormal cell. As a result, the operator can grasp the accuracy (possibility) of the cell being an abnormal cell as a numerical value, so that the accuracy of the cell being an abnormal cell can be easily grasped.

[0109] In addition, in the present embodiment, as described above, when learning to analyze the state of the cell 90, a step of creating a learned model 6 is further provided by also learning that a region that is similar to an abnormal cell but recognized as a background portion is a noise region 80. In the step of obtaining the index value 21, the learned model 6 that has been learned to determine whether the state of the cell 90 is a noise region 80 or not is used to obtain the index value 21. Thereby, it is possible to suppress the determination that the abnormal cell region 91 is also a background region because the learned model 6 estimates that the noise region 80 similar to the abnormal cell is a background region. As a result, the estimation accuracy of the abnormal cell region 91 can be improved.

[0110] [Modification Example] The embodiments disclosed this time should be considered as illustrative in all respects and not restrictive. The scope of the present invention is shown not by the description of the above embodiments but by the claims, and further includes all changes (modification examples) within the meaning and scope equivalent to the claims.

[0111] For example, in the above-described embodiment, an example of the configuration in which the image processing unit 2c acquires the first abnormal cell region 91a and the second abnormal cell region 91b using the first determination reference value 20a and the second determination reference value 20b has been shown. However, the present invention is not limited to this. For example, the image processing unit 2c may be configured to acquire the abnormal cell region 91 using one determination reference value 20.

[0112] Further, in the above-described embodiment, an example of the configuration in which the image processing unit 2c determines whether or not it is the abnormal cell region 91 based on the first probability value 21a and the second probability value 21b has been shown. However, the present invention is not limited to this. For example, the image processing unit 2c may be configured to determine whether or not it is the abnormal cell region 91 based only on the first probability value 21a.

[0113] Further, in the above-described embodiment, an example of the configuration in which the first determination reference value 20a is greater than 0 and the second determination reference value 20b is 10% or more has been shown. However, the present invention is not limited to this. If the first determination reference value 20a is a value smaller than the second determination reference value 20b, the first determination reference value 20a and the second determination reference value 20b can be set to arbitrary values.

[0114] Further, in the above-described embodiment, an example of the configuration in which the image processing unit 2c determines whether or not it is the abnormal cell region 91 for each pixel of the cell region 93 has been shown. However, the present invention is not limited to this. For example, the image processing unit 2c may be configured to determine whether or not it is the abnormal cell region 91 for all the pixels of the cell image 10.

[0115] Further, in the above-described embodiment, an example of the configuration in which the superimposed cell image generation unit 2d generates the superimposed cell image 50 that displays the entire inside of the first abnormal cell region 91a in a predetermined color and highlights the frame line 91c surrounding the second abnormal cell region 91b has been shown. However, the present invention is not limited to this. As long as the first abnormal cell region 91a and the second abnormal cell region 91b are distinguishable, the superimposed cell image generation unit 2d may generate the superimposed cell image 50 in any manner.

[0116] In the above embodiment, an example of the configuration in which the superimposed cell image generation unit 2d superimposes and displays the first abnormal cell region 91a and the second abnormal cell region 91b on the cell image 10 has been shown. However, the present invention is not limited to this. For example, the superimposed cell image generation unit 2d may be configured to generate an image in which the first abnormal cell region 91a is superimposed on the cell image 10 and an image in which the second abnormal cell region 91b is superimposed on the cell image 10, and display those images side by side.

[0117] In the above embodiment, an example of the configuration in which the superimposed cell image generation unit 2d generates a superimposed cell image 50 that superimposes and displays the first abnormal cell region 91a, the second abnormal cell region 91b, and the region 92 of normal cells on the cell image 10 has been shown. However, the present invention is not limited to this. For example, the superimposed cell image generation unit 2d may generate a superimposed cell image 50 that superimposes and displays the first abnormal cell region 91a and the second abnormal cell region 91b on the cell image 10, and it is not always necessary to superimpose the region 92 of normal cells.

[0118] In the above embodiment, an example of the configuration in which the image processing unit 2c excludes the second abnormal cell region 91b having an area equal to or smaller than a predetermined area from the targets to be displayed so as to be distinguishable from the first abnormal cell region 91a has been shown. However, the present invention is not limited to this. For example, the image processing unit 2c may not exclude the second abnormal cell region 91b having an area equal to or smaller than a predetermined area from the targets to be displayed so as to be distinguishable from the first abnormal cell region 91a.

[0119] In the above embodiment, an example of the configuration in which the image processing unit 2c excludes the second abnormal cell region 91b that does not include the first abnormal cell region 91a inside from the targets to be displayed so as to be distinguishable from the first abnormal cell region 91a has been shown. However, the present invention is not limited to this. For example, the image processing unit 2c may not exclude the second abnormal cell region 91b that does not include the first abnormal cell region 91a inside from the targets to be displayed so as to be distinguishable from the first abnormal cell region 91a.

[0120] In the above-described embodiment, an example of the configuration in which the image processing unit 2c acquires the number 60 of the second abnormal cell regions 91b shown in the cell image 10 and displays it on the display unit 4 has been shown. However, the present invention is not limited to this. For example, the image processing unit 2c may be configured to acquire the number of the first abnormal cell regions 91a shown in the cell image 10 and display it on the display unit 4.

[0121] In the above-described embodiment, an example of the configuration in which the image processing unit 2c acquires the number 60 of the second abnormal cell regions 91b shown in the cell image 10 and displays it on the display unit 4 has been shown. However, the present invention is not limited to this. For example, the image processing unit 2c does not necessarily have to acquire the number 60 of the second abnormal cell regions 91b shown in the cell image 10.

[0122] In the above-described embodiment, an example of the configuration in which the image processing unit 2c acquires the ratio 61 of the area of the first abnormal cell region 91a to the area of the second abnormal cell region 91b and displays the acquired ratio 61 on the display unit 4 has been shown. However, the present invention is not limited to this. For example, the image processing unit 2c does not necessarily have to acquire the ratio 61 of the area of the first abnormal cell region 91a to the area of the second abnormal cell region 91b.

[0123] (Item 1) A step of acquiring a cell image in which cells are shown, and the cells is a normal cell or an abnormal cell analysis Method A step of acquiring an index value obtained by analyzing the cells shown in the cell image using a learned model obtained by learning the analysis of the cells, and Among the index values, a step of acquiring an abnormal cell region that is a region where a first index value indicating that the cells are abnormal cells is larger than a predetermined determination reference value, and The A step of acquiring an abnormal cell region that is a region where a first index value indicating that the cells are abnormal cells is larger than a predetermined determination reference value, and A step of displaying the abnormal cell region in an identifiable manner 、 The predetermined determination reference value includes a first determination reference value and a second determination reference value lower than the first determination reference value. In the step of obtaining the index value, as the first index value, for each pixel of the cell image, a first probability value indicating the probability that the cell is the abnormal cell and a second probability value indicating the probability that the cell is the normal cell are obtained. The step of obtaining the abnormal cell region includes a step of obtaining, as the first abnormal cell region, the abnormal cell region where the first index value is greater than the first determination reference value, and a step of obtaining, as the second abnormal cell region, the abnormal cell region where the first index value is greater than the second determination reference value. In the step of obtaining the first abnormal cell region, even if the first probability value is smaller than the second probability value, when the first probability value is greater than the first determination reference value, the pixel of the cell image is obtained as the first abnormal cell region. In the step of obtaining the second abnormal cell region, even when the first probability value is smaller than the second probability value, if the first probability value is larger than the second determination reference value, pixels of the cell image are obtained as the second abnormal cell region. In the step of discriminatively displaying the abnormal cell region, the first abnormal cell region and the second abnormal cell region are discriminatively displayed. , Cell image analysis method.

[0127] (Item 2 ) In the step of obtaining the second abnormal cell region, as the second determination reference value, a region where the first probability value is greater than 0 is obtained as the second abnormal cell region, Item 1 The cell image analysis method according to.

[0128] (Item 3 ) The method further includes a step of obtaining a cell region that is a region of the cell and a region other than the cell in the cell image, In the step of obtaining the first abnormal cell region, a region where the first probability value for each pixel of the cell region is greater than the first determination reference value is obtained as the first abnormal cell region, In the step of obtaining the second abnormal cell region, a region where the first probability value for each pixel of the cell region is greater than the second determination reference value is obtained as the second abnormal cell region, Item 1 or 2 The cell image analysis method according to.

[0129] (Item 4 ) In the step of discriminatively displaying the abnormal cell region, the entire inside of the first abnormal cell region is displayed in a predetermined color, and a frame line surrounding the second abnormal cell region is highlighted, Item 1~3 The cell image analysis method according to any one of.

[0130] (Item 5 ) In the step of discriminatively displaying the abnormal cell region, the first abnormal cell region and the second abnormal cell region are superimposed and displayed on the cell image, Item 4 The cell image analysis method according to.

[0131] (Item 6 ) In the step of visibly displaying the abnormal cell region, the first abnormal cell region, the second abnormal cell region, and the region of the normal cells are superimposed and displayed on the cell image. 5 The cell image analysis method according to

[0132] (Item 7 ) The method further includes a step of excluding, from the objects to be visibly displayed separately from the first abnormal cell region, the second abnormal cell regions having an area equal to or smaller than a predetermined area among the second abnormal cell regions. 4~6 The cell image analysis method according to any one of

[0133] (Item 8 ) The method further includes a step of excluding, from the objects to be visibly displayed separately from the first abnormal cell region, the second abnormal cell regions that do not include the first abnormal cell region inside. 4~7 The cell image analysis method according to any one of

[0134] (Item 9 ) A step of obtaining the number of the second abnormal cell regions shown in the cell image; A step of displaying the obtained number of the second abnormal cell regions. 1~8 The cell image analysis method according to any one of

[0135] (Item 10 ) A step of obtaining the ratio of the area of the first abnormal cell region to the area of the second abnormal cell region; A step of displaying the ratio of the area of the first abnormal cell region to the area of the second abnormal cell region. 1~9 The cell image analysis method according to any one of

[0136] (Item 11 ) When training the analysis of the state of the cells, the method further includes the step of creating the learned model by also training that a region similar to the abnormal cells but recognized as a background part is a noise region. In the step of obtaining the index value, the index value is obtained using the learned model that has been trained to determine the state of the cells and whether it is the noise region. 1~10 The cell image analysis method according to any one of items 1 to 3.

[0137] (Item 12) The step of obtaining the number of the second abnormal cell regions shown in the cell image, The cell image analysis method according to any one of items 2 to 11, further including the step of displaying the obtained number of the second abnormal cell regions.

[0138] (Item 13) The step of obtaining the ratio of the area of the first abnormal cell region to the area of the second abnormal cell region, The cell image analysis method according to any one of items 2 to 12, further including the step of displaying the ratio of the area of the first abnormal cell region to the area of the second abnormal cell region.

[0139] (Item 14) When training the analysis of the state of the cells, the method further includes the step of creating the learned model by also training that a region similar to the abnormal cells but recognized as a background part is a noise region. In the step of obtaining the index value, the index value is obtained using the learned model that has been trained to determine the state of the cells and whether it is the noise region. The cell image analysis method according to any one of items 2 to 13.

Explanation of Signs

[0140] 6 Learned model 10 Cell image 20 Judgment reference value 20a First determination reference value 20b Second determination reference value 21 Index value 21a First probability value (first index value) 21b Second probability value 50 Overlay image 60 Number of second abnormal cell regions 61 Ratio of the area of the first abnormal cell region to the area of the second abnormal cell region 80 Noise region 90 Cells (cultured cells) 91 Abnormal cell region 91a First abnormal cell region 91b Second abnormal cell region 91c Frame line (frame line surrounding the second abnormal cell region) 92 Normal cell region 93 Cell region 94 Region other than cells

Claims

1. A step of obtaining a cell image in which cells are depicted; A step of obtaining an index value obtained by analyzing the cells depicted in the cell image using a learned model that has learned an analysis method for determining whether the cells are normal cells or abnormal cells; A step of obtaining an abnormal cell region which is a region where a first index value indicating that the cells are the abnormal cells among the index values is larger than a predetermined determination reference value; A step of visibly displaying the abnormal cell region; and The predetermined determination reference value includes a first determination reference value and a second determination reference value lower than the first determination reference value; In the step of obtaining the index value, as the first index value, for each pixel of the cell image, a first probability value indicating the probability that the cell is the abnormal cell and a second probability value indicating the probability that the cell is the normal cell are obtained; The step of obtaining the abnormal cell region includes a step of obtaining, as a first abnormal cell region, the abnormal cell region where the first index value is larger than the first determination reference value, and a step of obtaining, as a second abnormal cell region, the abnormal cell region where the first index value is larger than the second determination reference value; In the step of obtaining the first abnormal cell region, even when the first probability value is smaller than the second probability value, if the first probability value is larger than the first determination reference value, the pixel of the cell image is obtained as the first abnormal cell region; In the step of obtaining the second abnormal cell region, even when the first probability value is smaller than the second probability value, if the first probability value is larger than the second determination reference value, the pixel of the cell image is obtained as the second abnormal cell region; A cell image analysis method for visibly displaying the first abnormal cell region and the second abnormal cell region in the step of visibly displaying the abnormal cell region.

2. The cell image analysis method according to claim 1, wherein in the step of obtaining the second abnormal cell region, as the second determination reference value, a region where the first probability value is larger than 0 is obtained as the second abnormal cell region.

3. Further comprising a step of obtaining a cell region which is a region of the cells and a region other than the cells in the cell image; In the step of obtaining the first abnormal cell region, a region where the first probability value for each pixel of the cell region is larger than the first determination reference value is obtained as the first abnormal cell region; In the step of obtaining the second abnormal cell region, a region where the first probability value for each pixel of the cell region is greater than the second determination reference value is obtained as the second abnormal cell region, according to the cell image analysis method of claim 1 or 2.

4. In the step of displaying the abnormal cell region in a distinguishable manner, the entire inside of the first abnormal cell region is displayed in a predetermined color, and a frame line surrounding the second abnormal cell region is highlighted, according to the cell image analysis method of any one of claims 1 to 3.

5. In the step of displaying the abnormal cell region in a distinguishable manner, the first abnormal cell region and the second abnormal cell region are superimposed and displayed on the cell image, according to the cell image analysis method of claim 4.

6. In the step of displaying the abnormal cell region in a distinguishable manner, the first abnormal cell region, the second abnormal cell region, and the region of the normal cells are superimposed and displayed on the cell image, according to the cell image analysis method of claim 5.

7. The method further includes a step of excluding, from objects to be displayed in a distinguishable manner from the first abnormal cell region, the second abnormal cell regions having an area equal to or less than a predetermined area among the second abnormal cell regions, according to the cell image analysis method of any one of claims 4 to 6.

8. The method further includes a step of excluding, from objects to be displayed in a distinguishable manner from the first abnormal cell region, the second abnormal cell regions that do not include the first abnormal cell region inside, according to the cell image analysis method of any one of claims 4 to 7.

9. A step of obtaining the number of the second abnormal cell regions shown in the cell image; A step of displaying the obtained number of the second abnormal cell regions, according to the cell image analysis method of any one of claims 1 to 8.

10. A step of obtaining a ratio of the area of the first abnormal cell region to the area of the second abnormal cell region; A step of displaying the ratio of the area of the first abnormal cell region to the area of the second abnormal cell region, according to the cell image analysis method of any one of claims 1 to 9.

11. When learning the analysis of the state of the cells, the method further includes a step of creating the learned model by also learning that a region that is similar to the abnormal cells but recognized as a background part is a noise region. In the step of obtaining the index value, the index value is obtained by using the learned model that has been trained to determine the state of the cell and whether it is in the noise region, according to any one of claims 1 to 10.

Images