Mycoplasma inspection device, mycoplasma display device, mycoplasma inspection system, mycoplasma display system, learning device, computer program, mycoplasma inspection method, and learning device generation method

Abstract

To provide a mycoplasma inspection apparatus, a mycoplasma display apparatus, a mycoplasma inspection system, a mycoplasma display system, a learning arrangement, a computer program, a mycoplasma inspection method, and a method of generating a learning arrangement, which can detect a mycoplasma efficiently.SOLUTION: A mycoplasma inspection apparatus includes: an acquisition unit that acquires an image to be inspected in which a cell nucleus is photographed; a first learning device that is learned to detect a mycoplasma on the basis of the acquired image to be inspected; and a mycoplasma image generator that generates a mycoplasma image on the basis of the mycoplasma detected by the first learning device and the acquired image to be inspected.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The present invention relates to a mycoplasma inspection device, a mycoplasma display device, a mycoplasma inspection system, a mycoplasma display system, a learning device, a computer program, a mycoplasma inspection method, and a method for generating a learning device. [Background technology]

[0002] Mycoplasma is the smallest self-replicating bacterium, and is widely distributed in nature as a parasite of humans, mammals, etc. Non-Patent Document 1 describes a DNA staining method using indicator cells (so-called Mycoplasma Test Method B) as a "Mycoplasma Test for Cell Substrates Used in the Production of Biotechnology-Applied Pharmaceuticals / Biological Pharmaceuticals." [Prior art documents] [Non-patent literature]

[0003] [Non-Patent Document 1] Japanese Pharmacopoeia, 16th Edition (Ministry of Health, Labour and Welfare Notification No. 65, March 24, 2011) Summary of the Invention [Problem to be solved by the invention]

[0004] However, in the Mycoplasma Negative Test Method B, the test sample must be cultured for a specified period of time, fluorescently stained for DNA, and then visually confirmed for the presence of Mycoplasma under a fluorescent microscope, which reduces worker efficiency.

[0005] The present invention has been made in consideration of the above circumstances, and aims to provide a mycoplasma inspection device, a mycoplasma display device, a mycoplasma inspection system, a mycoplasma display system, a learning device, a computer program, a mycoplasma inspection method, and a method for generating a learning device that can efficiently detect mycoplasma. [Means for solving the problem]

[0006] A mycoplasma inspection device according to an embodiment of the present invention comprises an acquisition unit that acquires an image of an object to be inspected in which a cell nucleus is photographed, a first learning device that has been trained to detect mycoplasma based on the image of the object to be inspected acquired by the acquisition unit, and a mycoplasma image generation unit that generates a mycoplasma image based on the mycoplasma detected by the first learning device and the image of the object to be inspected acquired by the acquisition unit.

[0007] A mycoplasma display device according to an embodiment of the present invention comprises an acquisition unit that acquires an image of an object to be inspected in which a cell nucleus is photographed, and a mycoplasma image display unit that displays mycoplasma detected by a first learning device that has been trained to detect mycoplasma based on the image of the object to be inspected acquired by the acquisition unit and a mycoplasma image based on the image of the object to be inspected acquired by the acquisition unit.

[0008] A mycoplasma inspection system according to an embodiment of the present invention comprises an imaging device that images cell nuclei and the aforementioned mycoplasma inspection device, and the mycoplasma inspection device acquires an image of the inspection object captured by the imaging device.

[0009] A mycoplasma display system according to an embodiment of the present invention comprises an imaging device for imaging cell nuclei and the aforementioned mycoplasma display device, and the mycoplasma display device acquires an image of the object to be inspected captured by the imaging device.

[0010] A learning device according to an embodiment of the present invention includes a first learning device trained to detect mycoplasma based on an image of an object to be inspected in which cell nuclei are photographed; a second learning device trained to determine whether the cell nuclei are contaminated with mycoplasma based on the mycoplasma detected by the first learning device and the image of the object to be inspected; and a third learning device trained to detect the number of cell nuclei using a mycoplasma contamination result image containing at least one of contaminated and non-contaminated cell nuclei, which is generated based on the determination result of the second learning device.

[0011] A computer program according to an embodiment of the present invention causes a computer to perform a process of acquiring an image of an object to be inspected in which a cell nucleus is photographed, and a process of generating a mycoplasma image based on the mycoplasma detected by a first learning machine trained using the acquired image of the object to be inspected and the image of the object to be inspected.

[0012] A mycoplasma inspection method according to an embodiment of the present invention acquires an inspection object image in which a cell nucleus is photographed, and generates a mycoplasma image based on the mycoplasma detected by a first learning machine trained using the acquired inspection object image and the inspection object image.

[0013] A method for generating a learning module according to an embodiment of the present invention is a method for generating a learning module comprising a first learning module, a second learning module, and a third learning module, and the first learning module is generated using training data including a first image in which cell nuclei are photographed and in which mycoplasma is present and an extracted image in which mycoplasma is extracted from the first image, and training data including a second image in which cell nuclei are photographed and in which mycoplasma is not present and a specified image in which mycoplasma is not present; the second learning module is generated using training data including a training input image consisting of a set of the first image and the extracted image and a specified contaminated cell nucleus image, and training data including a training input image consisting of a set of the second image and the specified image and a specified non-contaminated cell nucleus image; and the third learning module is generated using teacher labels indicating the specified contaminated cell nucleus image, the specified non-contaminated cell nucleus image, and the number of cell nuclei. [Effects of the Invention]

[0014] According to the present invention, mycoplasma can be detected efficiently. [Brief explanation of the drawings]

[0015] [Figure 1] 1 is a block diagram showing an example of the configuration of an inspection device as a mycoplasma inspection device according to the present embodiment. FIG. [Figure 2] FIG. 2 is a schematic diagram showing the relationship between a captured image and an image to be inspected. [Figure 3]FIG. 2 is a schematic diagram illustrating an example of the configuration of each of a first learning device, a second learning device, and a third learning device. [Figure 4] FIG. 3 is a schematic diagram showing an example of a first image. [Figure 5] FIG. 1 is a schematic diagram showing an example of a mycoplasma extraction image. [Figure 6] FIG. 10 is a schematic diagram showing an example of a second image. [Figure 7] FIG. 10 is a schematic diagram showing an example of a predetermined image. [Figure 8] FIG. 2 is a schematic diagram showing an example of a learning input image formed by a set of a first image and a mycoplasma-extracted image. [Figure 9] FIG. 10 is a schematic diagram showing an example of a contaminated cell nucleus image. [Figure 10] 10 is a schematic diagram showing an example of learning input images formed by a set of a second image and a predetermined image. FIG. [Figure 11] FIG. 1 is a schematic diagram showing an example of an uncontaminated cell nucleus image. [Figure 12] FIG. 10 is a schematic diagram showing an example of a cut-out image for extracting a cell cluster from within a contaminated cell nucleus image. [Figure 13] FIG. 10 is a schematic diagram showing the relationship between learning input images and teacher labels for generating a third learning device. [Figure 14] FIG. 2 is a schematic diagram showing an example of an inspection flow of the inspection device according to the present embodiment. [Figure 15] FIG. 1 is a schematic diagram showing an example of a mycoplasma image. [Figure 16] FIG. 10 is a schematic diagram showing an example of an image showing the results of mycoplasma contamination. [Figure 17] FIG. 1 is a schematic diagram showing an example of a method for determining whether a cell nucleus is contaminated or not. [Figure 18] FIG. 10 is a schematic diagram showing an example of an identifier assigned to a cell nucleus in an image resulting from mycoplasma contamination. [Figure 19] FIG. 10 is a schematic diagram showing an example of display of cell nucleus number information. [Figure 20] FIG. 10 is a schematic diagram showing a first example of display of detailed information. [Figure 21] FIG. 10 is a schematic diagram showing a second example of display of detailed information. [Figure 22] 1 is a block diagram showing an example of the configuration of an inspection system or a display system according to an embodiment of the present invention. [Figure 23] 10 is a flowchart showing an example of a processing procedure of the inspection device according to the present embodiment. [Figure 24] 10 is a flowchart showing an example of a processing procedure of the inspection device according to the present embodiment. DETAILED DESCRIPTION OF THE INVENTION

[0016] An embodiment of the present invention will now be described with reference to the drawings. FIG. 1 is a block diagram showing an example of the configuration of an inspection device 50 serving as a mycoplasma inspection device according to this embodiment. The inspection device 50 includes a control unit 51 that controls the entire device, an input unit 52, a mycoplasma image generation unit 53, a contamination result image generation unit 54, a memory unit 55, a cell aggregate extraction unit 56, an interface unit 57, a learning unit 58, and a determination unit 59. The learning unit 58 includes a first learning unit 581, a second learning unit 582, and a third learning unit 583. The inspection device 50 can be connected to a display device 100 that serves as a mycoplasma display device. The display device 100 includes a control unit 101 that controls the entire device, an interface unit 102, a display screen 103, and an operation unit 104.

[0017] The control unit 51 can be configured with a CPU, a ROM, a RAM, and the like.

[0018] The input unit 52 functions as an acquisition unit and acquires an image of the test object in which cell nuclei are photographed. The image of the test object can be an image area obtained by scanning an image of the test specimen captured by an imaging device (e.g., a microscope) over an area of ​​a predetermined size.

[0019] The storage unit 55 can store the inspection object image acquired by the input unit 52. The storage unit 55 can record required data such as the processing results performed by the inspection device 50.

[0020] The interface unit 57 has a function of interfacing with the display device 100. The interface unit 57 also has a function of interfacing with a printing device (not shown).

[0021] FIG. 2 is a schematic diagram showing the relationship between a captured image and an inspection target image. The captured image is an image captured by an imaging device (e.g., a microscope, more specifically, a fluorescent microscope). The vertical and horizontal resolution (m×n) of the captured image varies depending on the imaging device. The resolution m or n can be, for example, approximately 1000, 1200, or 1400, but is not limited to these. The inspection target image can have, for example, a resolution of 128×128, but is not limited to this. The inspection target image can be extracted, for example, by scanning the captured image so that a region with a 128×128 resolution partially overlaps with another region. In other words, multiple inspection target images can be extracted from a single captured image. In this specification, the resolution of the inspection target image is described as 128×128, but is not limited to this.

[0022] Next, the learning device 58 will be described.

[0023] 3 is a schematic diagram showing an example of the configuration of each of the first learner 581, the second learner 582, and the third learner 583. The first learner 581, the second learner 582, and the third learner 583 can be configured, for example, by a multi-layer neural network (deep learning), and for example, a convolutional neural network can be used, but other machine learning methods may also be used. In this specification, the first learner 581, the second learner 582, and the third learner 583 are described as being configured by a convolutional neural network, but are not limited to this.

[0024] The first learning device 581 includes an input layer, an output layer, and a hidden layer between the input layer and the output layer. The hidden layer includes a multi-stage convolution layer and a multi-stage upsampling layer (also called an unpooling layer). For example, the input layer can receive 128 x 128 data, and the output layer can output 128 x 128 data, although the number of pieces of data can be changed depending on the resolution of the image to be inspected.

[0025] The second learning device 582 includes an input layer, an output layer, and a hidden layer between the input layer and the output layer. The hidden layer includes a multi-stage convolutional layer and a multi-stage upsampling layer (also called an unpooling layer). The input layer is configured with two channels, each of which can receive, for example, 128 x 128 data, and the output layer can output 96 x 96 data, although the number of data can be changed depending on the resolution of the image to be inspected.

[0026] The third learning device 583 includes an input layer, an output layer, and a hidden layer between the input layer and the output layer. The hidden layer includes a multi-stage convolutional layer and a pooling layer, and a fully connected layer at the final stage. The number of convolutional layers, pooling layers, and fully connected layers can be determined appropriately. For example, the input layer receives data according to the resolution of the cropped image described below, and the output layer can include output nodes according to the number of types of cell nuclei to be detected (e.g., 0, 1, 2, 3, 4 or more, etc.).

[0027] The first learning device 581, the second learning device 582, and the third learning device 583 can be constructed by combining hardware such as a CPU (e.g., a multi-processor having multiple processor cores), a GPU (Graphics Processing Units), a DSP (Digital Signal Processors), and an FPGA (Field-Programmable Gate Arrays).

[0028] Next, a description will be given of a method for generating (learning) learning device 58. First, a method for generating first learning device 581 will be described.

[0029] The first learning device 581 can be generated using learning data including a first image in which a cell nucleus is photographed and in which mycoplasma is present, and a mycoplasma-extracted image (extracted image) in which mycoplasma is extracted from the first image.

[0030] FIG. 4 is a schematic diagram showing an example of a first image. The resolution of the first image is the same as that of the image to be inspected, and can be, for example, 128 × 128. As shown in FIG. 4, the first image is an image of multiple cell nuclei, and is an input image for learning in which mycoplasma is present. Mycoplasma is a bacterium capable of self-replication. However, the first image is an image in which mycoplasma cannot be accurately visualized. Note that FIG. 4 illustrates the cell nuclei and mycoplasma in a schematic manner for convenience, and may differ from the actual image.

[0031] Figure 5 is a schematic diagram showing an example of a mycoplasma-extracted image. The mycoplasma-extracted image is an output image for learning in which mycoplasma has been extracted from the first image (i.e., an image in which cell nuclei have been deleted from the first image). In other words, the mycoplasma-extracted image shown in Figure 5 can be generated by extracting only mycoplasma from the first image shown in Figure 4. The resolution of the mycoplasma-extracted image is the same as that of the first image, and can be, for example, 128 x 128.

[0032] The first learning device 581 can be generated using learning data including a second image in which a cell nucleus is photographed and in which no mycoplasma is present, and a predetermined image in which no mycoplasma is present.

[0033] FIG. 6 is a schematic diagram showing an example of the second image. The resolution of the second image is the same as that of the image to be inspected, and can be, for example, 128 × 128. As shown in FIG. 6, the second image is an input image for learning in which no mycoplasma is present and multiple cell nuclei are photographed. Note that in FIG. 6, the cell nuclei are illustrated schematically for convenience and may differ from the actual image.

[0034] 7 is a schematic diagram showing an example of a predetermined image. The predetermined image is an output image for learning in which neither mycoplasma nor cell nuclei are present. The resolution of the predetermined image is the same as that of the second image, and can be, for example, 128 x 128.

[0035] As a result, when an inspection object image in which a cell nucleus is photographed is input to the first learning device 581, the first learning device 581 can detect mycoplasma.

[0036] Next, a method for generating second learning device 582 will be described.

[0037] The second learning device 582 can be generated using learning data including a learning input image consisting of a set of the first image and the mycoplasma-extracted image, and a predetermined contaminated cell nucleus image. Here, the first image and the mycoplasma-extracted image are the first image and the mycoplasma-extracted image used in generating the first learning device 581.

[0038] FIG. 8 is a schematic diagram showing an example of a training input image composed of a pair of a first image and a mycoplasma extraction image. The training input image shown in FIG. 8 is composed of a pair of a first image and a mycoplasma extraction image, and the first image can be input to one channel of the input layer, and the mycoplasma extraction image can be input to the other channel of the input layer. The resolution of the first image and the mycoplasma extraction image can be, for example, 128 x 128. Note that FIG. 8 shows the cell nuclei and mycoplasma schematically for convenience, and may differ from the actual images.

[0039] FIG. 9 is a schematic diagram showing an example of a contaminated cell nucleus image. The contaminated cell nucleus image is an image that represents contaminated cell nuclei contaminated with mycoplasma among the cell nuclei in the learning input image composed of the first image and the mycoplasma-extracted image. In other words, if there are non-contaminated cell nuclei that are not contaminated with mycoplasma, the non-contaminated cell nuclei can be excluded. In FIG. 9, for convenience, all cell nuclei are assumed to be contaminated, and the contaminated cell nuclei are illustrated with a dark pattern.

[0040] The second learning device 582 can be generated using learning data including a learning input image consisting of a set of a second image and a predetermined image, and a predetermined uncontaminated cell nucleus image. The second image and the predetermined image are the second image and the predetermined image used when generating the first learning device 581.

[0041] FIG. 10 is a schematic diagram showing an example of a learning input image composed of a pair of a second image and a predetermined image. The learning input image shown in FIG. 10 is composed of a pair of a second image and a predetermined image, and the second image can be input to one channel of the input layer, and the predetermined image can be input to the other channel of the input layer. The resolution of the second image and the predetermined image can be, for example, 128 x 128. Note that in FIG. 10, cell nuclei are illustrated schematically for convenience, and may differ from the actual images.

[0042] FIG. 11 is a schematic diagram showing an example of an uncontaminated cell nucleus image. The uncontaminated cell nucleus image is an image that represents uncontaminated cell nuclei that are not contaminated with mycoplasma among the cell nuclei in the learning input image. In other words, if there are contaminated cell nuclei contaminated with mycoplasma, the contaminated cell nuclei can be excluded. In FIG. 11, contaminated cell nuclei are shown with a light pattern.

[0043] As a result, when a mycoplasma image containing detected mycoplasma and multiple cell nuclei is input to the second learning device 582, the second learning device 582 can determine whether the cell nuclei are contaminated cell nuclei or non-contaminated cell nuclei.

[0044] Next, a method for generating third learning device 583 will be described.

[0045] The third learner 583 can be generated using a predetermined contaminated cell nucleus image, a predetermined non-contaminated cell nucleus image, and a teacher label indicating the number of cell nuclei. For example, the third learner 583 can be generated using a contaminated cell nucleus image or a non-contaminated cell nucleus image with one cell nucleus and a teacher label indicating that the number of cell nuclei is one (e.g., "1"). The same applies when the number of cell nuclei is two or three. The third learner 583 can also be generated using a contaminated cell nucleus image or a non-contaminated cell nucleus image with four or more cell nuclei and a teacher label indicating that the number of cell nuclei is four or more (e.g., "4"). In this way, when a mycoplasma contamination result image is input to the third learner 583, the third learner 583 can detect the number of cell nuclei. Note that the teacher label is independent of whether the cell nuclei are contaminated or not.

[0046] A specific example of a method for generating third learning device 583 will be described.

[0047] The cell cluster extraction unit 56 can extract cell clusters each consisting of one or more cell nuclei based on a predetermined contaminated cell nucleus image. Specifically, the cell cluster extraction unit 56 can extract cell clusters by cutting out a cut-out image containing one cell cluster from the predetermined contaminated cell nucleus image. Note that the number of cell clusters that can be extracted is the same as the number of cell clusters present in the predetermined contaminated cell nucleus image.

[0048] FIG. 12 is a schematic diagram showing an example of a cut-out image for extracting a cell cluster from a contaminated cell nucleus image. The predetermined contaminated cell nucleus image can be an image in which all or part of the cell nucleus is contaminated with mycoplasma. In the example of FIG. 12, a rectangular cut-out image surrounded by a dashed line shows a case in which the cut-out image includes a cell cluster consisting of two cell nuclei. Cut-out images can be extracted for all cell clusters from the contaminated cell nucleus image. Note that in this specification, a single cell nucleus will also be described as an example of a cell cluster.

[0049] FIG. 13 is a schematic diagram showing the relationship between the learning input images and the teacher labels for generating the third learning device 583. The learning input images are individual cropped images. The example in FIG. 13 shows five cropped images. No. 1 is a cropped image containing one cell nucleus. The cropped image No. 1 is input to the input layer of the third learning device 583, and the teacher label "1" (indicating that the number of cell nuclei is 1) is assigned to the output layer. No. 2 is a cropped image containing two cell nuclei. The cropped image No. 2 is input to the input layer of the third learning device 583, and the teacher label "2" (indicating that the number of cell nuclei is 2) is assigned to the output layer. No. 3 is a cropped image containing three cell nuclei. The cropped image No. 3 is input to the input layer of the third learning device 583, and the teacher label "3" (indicating that the number of cell nuclei is 3) is assigned to the output layer. Furthermore, No. 4 is a cut-out image containing four or more cell nuclei, and the cut-out image of No. 4 is input to the input layer of the third learning device 583, and a teacher label "4" (indicating that the number of cell nuclei is four or more) is given to the output layer. Note that the teacher label is an example and is not limited to the example in FIG. 13. Furthermore, the cell nuclei contained in the cut-out image may include both contaminated and non-contaminated cell nuclei.

[0050] As a result, when a mycoplasma contamination result image (described later) is input to the third learning device 583, the third learning device 583 can detect the number of cell nuclei. Specifically, the third learning device 583 can detect the number of cell nuclei for each extracted image, and by summing the detected numbers of cell nuclei across the entire mycoplasma contamination result image, the number of cell nuclei (the number of contaminated cell nuclei and the number of non-contaminated cell nuclei) in one mycoplasma contamination result image can be detected. Furthermore, by detecting the number of cell nuclei, it is possible to determine whether the test sample is positive or negative.

[0051] Next, the mycoplasma testing process will be described.

[0052] 14 is a schematic diagram showing an example of the inspection flow of the inspection device 50 of this embodiment. The first learning unit 581 detects mycoplasma based on the inspection object image acquired by the input unit 52.

[0053] The mycoplasma image generating unit 53 generates a mycoplasma image based on the mycoplasma detected by the first learning unit 581 and the inspection target image acquired by the input unit 52.

[0054] FIG. 15 is a schematic diagram showing an example of a mycoplasma image. The mycoplasma image is an image in which detected mycoplasma is superimposed on an image of the object of inspection. More specifically, the mycoplasma image generation unit 53 performs processing to assign the image of the object of inspection to one of the two channels and the image of the detected mycoplasma to the other channel. By generating a mycoplasma image, mycoplasma can be visualized and mycoplasma can be detected efficiently. Note that the cell nuclei and mycoplasma shown in FIG. 15 are merely a schematic representation of the essential parts for convenience and differ from the actual cell nuclei and mycoplasma.

[0055] The second learning device 582 determines whether or not the cell nucleus is contaminated with mycoplasma based on the mycoplasma image generated by the mycoplasma image generation unit 53. More specifically, the image to be inspected is input to one channel of the input layer of the second learning device 582, and the image of mycoplasma detected by the first learning device 581 is input to the other channel.

[0056] The contamination result image generating unit 54 generates a mycoplasma contamination result image including at least one of contaminated cell nuclei and non-contaminated cell nuclei based on the determination result of the second learning device 582.

[0057] FIG. 16 is a schematic diagram showing an example of an image resulting from mycoplasma contamination. Contaminated cell nuclei are cell nuclei contaminated with mycoplasma, and non-contaminated cell nuclei are cell nuclei not contaminated with mycoplasma. Images resulting from mycoplasma contamination include images of only contaminated cell nuclei, images of only non-contaminated cell nuclei, and images of a mixture of contaminated and non-contaminated cell nuclei. This allows the visualization of contaminated and non-contaminated cell nuclei, making it possible to confirm the contamination state of the cell nuclei. The example in FIG. 16 is a schematic illustration of the main parts for convenience. The boundaries of contaminated cell nuclei are shown with thick lines to distinguish them from non-contaminated cell nuclei.

[0058] The third learning unit 583 can detect the number of cell nuclei based on the mycoplasma contamination result image generated by the contamination result image generation unit 54. By detecting the number of cell nuclei, it is possible to determine whether the test sample is positive or negative.

[0059] The cell cluster extraction unit 56 can extract cell clusters each consisting of one or more cell nuclei, based on the mycoplasma contamination result image generated by the contamination result image generation unit 54. Specifically, the cell cluster extraction unit 56 can extract cell clusters by cutting out a rectangular cut-out image containing the cell cluster from the mycoplasma contamination result image. The cut-out image is the same as the cut-out image exemplified in FIG. 12.

[0060] The third learning device 583 detects the number of cell nuclei for each cell cluster extracted by the cell cluster extraction unit 56. Specifically, the third learning device 583 can detect the number of cell nuclei for each cut-out image, and by summing the detected numbers of cell nuclei across the entire image resulting from mycoplasma contamination, it is possible to detect the number of cell nuclei in one image resulting from mycoplasma contamination (the number of contaminated cell nuclei and the number of non-contaminated cell nuclei). This makes it possible to accurately detect the number of cell nuclei even when the cell nuclei exist as clusters.

[0061] The third learning unit 583 can detect a representative value based on the number of cell nuclei detected in each of the multiple mycoplasma contamination result images generated by the contamination result image generation unit 54 as the number of cell nuclei. For example, suppose the number of cell nuclei present in one mycoplasma contamination result image is less than a predetermined number (e.g., 1,000), but the total number of cell nuclei in the multiple mycoplasma contamination result images is equal to or greater than the predetermined number. In this case, the number of cell nuclei detected in all of the multiple mycoplasma contamination result images is recorded in the storage unit 55. Similar processing is performed across all captured images, and a representative value (e.g., average, maximum, etc.) based on the recorded number of cell nuclei is detected as the number of cell nuclei relative to the predetermined number of cell nuclei (the number of contaminated cell nuclei and the number of non-contaminated cell nuclei). For example, if the number of contaminated cell nuclei detected per 1,000 cell nuclei is 100, 200, or 300, the number of contaminated cell nuclei can be determined to be 200 if the average value is used, or 300 if the maximum value is used. The representative value to be used can be determined as appropriate. By using the representative value, it is possible to accurately determine whether the test sample is positive or negative.

[0062] The interface unit 57 functions as an output unit and can output cell nucleus count information that associates a predetermined number of cell nuclei with the number of contaminating cell nuclei. The cell nucleus count information may be information for display or for printing. The predetermined number may be, for example, 1,000. This makes it easy to determine whether a test sample is positive or negative.

[0063] When the cell cluster extracted by the cell cluster extraction unit 56 contains both contaminated cell nuclei and non-contaminated cell nuclei, the determination unit 59 can determine whether the cell cluster is a contaminated cell nucleus or a non-contaminated cell nucleus based on predetermined conditions.

[0064] Figure 17 is a schematic diagram showing an example of a method for determining whether a cell nucleus is contaminated or non-contaminated. As shown in Figure 17A, a cell cluster contains both contaminated and non-contaminated cell nuclei. In this case, as shown in Figure 17B, whether a cell cluster contains contaminated or non-contaminated cell nuclei can be determined by determining the larger of the contaminated and non-contaminated cell nuclei shown in Figure 17A. This can improve the accuracy of detecting the number of contaminated or non-contaminated cell nuclei.

[0065] Next, the display device 100 will be described.

[0066] The control unit 101 can be configured with a CPU, a ROM, a RAM, and the like.

[0067] The interface unit 102 has an interface function with the inspection device 50 .

[0068] The display screen 103 can be configured with a liquid crystal panel, an organic EL (Electro Luminescence) display, or the like.

[0069] The operation unit 104 is configured with, for example, a hardware keyboard, a mouse, etc., and can be used to operate icons displayed on the display panel 24, input characters, etc. The operation unit 104 may be configured with a touch panel.

[0070] The interface unit 102 functions as an acquisition unit and can acquire an inspection object image in which a cell nucleus is photographed from the inspection device 50. The interface unit 102 can acquire a mycoplasma image generated by the mycoplasma image generation unit 53. The interface unit 102 can also acquire a contamination result image generated by the contamination result image generation unit 54.

[0071] The control unit 101 has a function as a mycoplasma image display unit, and can display mycoplasma images acquired by the interface unit 102 on a display screen 103. This makes it possible to visualize mycoplasma, enabling efficient confirmation of mycoplasma.

[0072] The control unit 101 functions as a contamination result image display unit, and can display the mycoplasma contamination result image acquired by the interface unit 102 on a display screen 103. This makes it possible to visualize contaminated and non-contaminated cell nuclei, and to confirm the contamination state of the cell nuclei.

[0073] The control unit 101 can display an image of the mycoplasma contamination result by assigning an identifier for identifying the cell nucleus to at least one of the contaminated cell nuclei and the non-contaminated cell nuclei.

[0074] FIG. 18 is a schematic diagram showing an example of identifiers assigned to cell nuclei in an image showing the results of mycoplasma contamination. In the figure, cell nuclei surrounded by a thick line are contaminated cell nuclei, and the other cell nuclei are non-contaminated cell nuclei. In the example of FIG. 18, numbers are used as identifiers, but identifiers are not limited to numbers. This allows for easy identification of individual cell nuclei.

[0075] The operation unit 104 has a function as a setting unit, and can set whether to display or hide the identifiers (numbers in the example of FIG. 18) displayed on the display screen 103. For example, an icon or button can be displayed, and the setting can be made by operating the icon or button. This allows, for example, a user (examiner) to select whether or not an identifier is required depending on the purpose when checking the contamination result image, thereby improving the visibility for the user.

[0076] The control unit 101 can display the image of mycoplasma contamination results by displaying contaminated and non-contaminated cell nuclei in different display modes. The boundaries between the contaminated and non-contaminated cell nuclei and the outside world can be displayed in different colors. For example, as shown in FIG. 18, the boundaries of contaminated cell nuclei may be displayed more emphasized (e.g., as a thicker line) than the boundaries of non-contaminated cell nuclei, or the boundaries of contaminated cell nuclei may be displayed in red and the boundaries of non-contaminated cell nuclei in blue. This allows easy identification of contaminated and non-contaminated cell nuclei.

[0077] When a cell cluster containing one or more cell nuclei contains both contaminated and non-contaminated cell nuclei, the control unit 101 can display an image of the mycoplasma contamination result by classifying the cell cluster as either a contaminated cell nucleus or a non-contaminated cell nucleus based on predetermined conditions. The predetermined conditions can be, for example, the size of each of the contaminated and non-contaminated cell nuclei, as shown in FIG. 17. This allows easy identification of each of the contaminated and non-contaminated cell nuclei.

[0078] Next, a display example of the cell nucleus number information will be described.

[0079] FIG. 19 is a schematic diagram showing an example of the display of cell nucleus count information. As shown in FIG. 19, for each test specimen (No. 1, 2, 3, ...), the contamination determination result (contaminated or non-contaminated), the number of contaminated cell nuclei out of a predetermined number of 1,000 cell nuclei, and a detail icon 110 are displayed. For example, for test specimen No. 1, 6 cell nuclei out of 1,000 cell nuclei are contaminated, so the contamination determination result is contaminated (positive). Similarly, for test specimen No. 2, 0 cell nuclei out of 1,000 cell nuclei are contaminated, so the contamination determination result is non-contaminated (negative). Furthermore, for test specimen No. 3, 120 cell nuclei out of 1,000 cell nuclei are contaminated, so the contamination determination result is contaminated (positive). For each test specimen, detailed information can be displayed by operating the detail icon 110.

[0080] FIG. 20 is a schematic diagram showing a first example of the display of detailed information. FIG. 20 shows detailed information displayed by operating the detail icon 110 for test specimen No. 1 in FIG. 19. The cell nucleus count information area 121 displays cell nucleus count information that associates a predetermined number of cell nuclei with the number of contaminated cell nuclei. That is, the control unit 101 has a function as a display unit and displays cell nucleus count information that associates a predetermined number of cell nuclei with the number of contaminated cell nuclei. The predetermined number can be, for example, 1000. In the example of FIG. 20, it is displayed that the number of contaminated cell nuclei is 6 out of 1000. This makes it easy to determine whether the test specimen is positive or negative.

[0081] A contamination result image is displayed in the entire image area 122. The contamination result image can be, for example, an image such as that shown in FIG. 16 or 18. For example, the boundaries of contaminated cell nuclei may be displayed more emphasized (e.g., as a thick line) than the boundaries of non-contaminated cell nuclei, or the boundaries of contaminated cell nuclei may be displayed in red and the boundaries of non-contaminated cell nuclei in blue. This makes it easy to identify each of the contaminated and non-contaminated cell nuclei.

[0082] The contaminated cell mass image area 123 displays an image of each cell mass. That is, the control unit 101 can display an image of a cell mass containing one or more contaminated cell nuclei in association with an identifier. In the example shown in the figure, an image of a cell mass with ID=1, an image of a cell mass with ID=2, etc. are displayed. By operating the arrow 124 below the contaminated cell mass image area 123, images of cell masses with other IDs can be displayed. This makes it easy to identify individual contaminated cell masses.

[0083] Figure 21 is a schematic diagram showing a second example of the display of detailed information. Figure 21 shows detailed information displayed by operating the detail icon 110 for test specimen No. 3 in Figure 19. The cell nucleus number information area 121 displays that the number of contaminated cell nuclei is 120 out of 1000. This makes it easy to determine whether the test specimen is positive or negative.

[0084] The whole image area 122 is the same as the first example in FIG.

[0085] In the cell nucleus ID area 125, an identifier for identifying a cell nucleus and a code indicating whether the cell is a contaminated cell nucleus or a non-contaminated cell nucleus are displayed in association with each other. That is, the control unit 101 can display an identifier for identifying a cell nucleus (e.g., a number) and a code (e.g., OK, NG, etc.) indicating whether the cell nucleus is a contaminated cell nucleus or a non-contaminated cell nucleus in association with each other. For example, different identifiers may be used for contaminated cell nuclei and non-contaminated cell nuclei, and numbers, for example, may be used, but are not limited to this. This makes it easy to identify whether each cell nucleus is a contaminated cell nucleus or a non-contaminated cell nucleus.

[0086] In the cell nucleus image region 126, images of cell nuclei corresponding to individual IDs are displayed. For example, by selecting the cell nucleus ID field 4 in the cell nucleus ID region 125, the control unit 101 can display the image of the cell nucleus of the selected ID (=4). This allows the image of the cell nucleus that interests the user to easily display based on the contamination determination results for each cell nucleus. Note that in FIG. 21, cell mass IDs may be displayed instead of cell nucleus IDs.

[0087] Next, the inspection system and the display system will be described.

[0088] 22 is a block diagram showing an example of the configuration of the inspection system or display system of this embodiment. An inspection device 50 to which a display device 100 is connected is connected to a communication network 1. The communication network 1 may be wireless or may be wired communication. A PC (information processing device) 10 that controls the culture device 200 and the imaging device 300 is connected to the communication network 1.

[0089] The incubation device 200 includes incubation units 201 and 202, a controller 203 that controls the incubation process of the incubation units 201 and 202, and an operation unit 204 for setting incubation conditions, etc. Incubation vessels containing test specimens are housed in the incubation units 201 and 202, and the test specimens are incubated under predetermined conditions. The incubated test specimens are imaged by an imaging device (e.g., a microscope), and the captured image is output to the inspection device 50 via the PC 10. The inspection system of this embodiment includes at least the inspection device 50 and the imaging device 300, and may also include the incubation device 200. The display system of this embodiment includes at least the display device 100 and the imaging device 300, and may also include the incubation device 200. Note that the functions of the display device 100 may be incorporated into the inspection device 50, and the inspection device 50 and the display device 100 may be collectively referred to as the inspection device 50.

[0090] 23 and 24 are flowcharts showing an example of the processing procedure of the inspection device 50 of this embodiment. For convenience, the following description will be given with the control unit 51 as the main actor in the processing. The control unit 51 acquires an inspection target image (S11) and inputs the inspection target image to the first learning device 581 (S12). The control unit 51 detects mycoplasma (S13) and generates a mycoplasma image (S14).

[0091] The control unit 51 inputs the generated mycoplasma image to the second learning unit 582 (S15), and generates a mycoplasma contamination result image (S16) based on the determination result of the second learning unit 582. The control unit 51 inputs the generated mycoplasma contamination result image to the third learning unit 583 (S17), and determines the presence or absence of contaminated cell nuclei based on the detection result of the third learning unit 583 (S18).

[0092] If there are contaminated cell nuclei (YES in S18), the control unit 51 records the number of contaminated cell nuclei (S19) and determines whether the total number of cell nuclei involved in the determination process is equal to or greater than a predetermined number (e.g., 1,000) (S20). If the total number of cell nuclei is less than the predetermined number (NO in S20), the control unit 51 selects the next image to be inspected (S21) and repeats the processes from step S11 onwards.

[0093] If the total number of cell nuclei is equal to or greater than the predetermined number (YES in S20), the control unit 51 calculates the sum of the numbers of contaminated cell nuclei recorded in step S19 (S22) and identifies the number of contaminated cell nuclei relative to the predetermined number (the total number of contaminated cell nuclei added up in step S22) (S23). This makes it possible to identify how many of the 1,000 cell nuclei are contaminated cell nuclei.

[0094] The control unit 51 determines whether there are any other images to be inspected (S24). That is, it determines whether there are any other images to be inspected that are to be extracted by scanning the captured image. If there are any other images to be inspected (YES in S24), the control unit 51 resets the number of contaminated cell nuclei recorded in step S19 to zero (S25), and repeats the processes from step S11 onwards.

[0095] If there are no contaminated cell nuclei (NO in S18), the control unit 51 performs the process of step S24. If there are no other images to be inspected (NO in S24), the control unit 51 calculates a representative value of the identified number of contaminated cell nuclei relative to the predetermined number (S26), sets the calculated representative value as the number of contaminated cell nuclei, outputs cell nuclei number information correlating the predetermined number with the number of contaminated cell nuclei (S27), and ends the process. Note that if there are multiple captured images, the processes of Figs. 23 and 24 can be repeated.

[0096] The inspection device 50 can also be realized using a computer equipped with a CPU (processor), GPU, RAM (memory), etc. That is, the inspection device 50 can be realized on a computer by loading a computer program that defines the procedures of each process, such as those shown in Figures 23 and 24, into RAM (memory) provided in the computer and executing the computer program on the CPU (processor). The computer program may be recorded on a recording medium and distributed.

[0097] According to this embodiment, mycoplasma can be detected efficiently. Furthermore, by outputting and displaying or printing a mycoplasma image such as that shown in Fig. 15 or a mycoplasma contamination result image such as that shown in Fig. 16, the legitimacy of the detection result can be confirmed, and the reliability of the test result can be ensured, compared to conventional testing methods that simply provide the number of contaminated cell nuclei.

[0098] The mycoplasma inspection device of this embodiment comprises an acquisition unit that acquires an image of an object to be inspected in which a cell nucleus is photographed, a first learning device that has been trained to detect mycoplasma based on the image of the object to be inspected acquired by the acquisition unit, and a mycoplasma image generation unit that generates a mycoplasma image based on the mycoplasma detected by the first learning device and the image of the object to be inspected acquired by the acquisition unit.

[0099] The computer program according to this embodiment causes a computer to perform the following processes: acquiring an image of an object to be inspected in which a cell nucleus is photographed; and generating a mycoplasma image based on the mycoplasma detected by a first learning device trained using the acquired image of the object to be inspected and the image of the object to be inspected.

[0100] The mycoplasma inspection method of this embodiment acquires an inspection object image in which a cell nucleus is photographed, and generates a mycoplasma image based on the mycoplasma detected by a first learning device that has been trained using the acquired inspection object image and the inspection object image.

[0101] The acquisition unit acquires an image of the test object in which the cell nuclei are photographed. The image of the test object can be an image area obtained by scanning an image of the test specimen captured by an imaging device (e.g., a microscope) over an area of ​​a predetermined size.

[0102] The first learning device can be configured, for example, with a multi-layer neural network (deep learning), such as a convolutional neural network, but other machine learning methods may also be used. The first learning device detects mycoplasma based on the inspection target image acquired by the acquisition unit. Mycoplasma is a bacterium capable of self-replicating.

[0103] The mycoplasma image generation unit generates a mycoplasma image based on the mycoplasma detected by the first learning device and the inspection target image acquired by the acquisition unit. The mycoplasma image is an image in which the detected mycoplasma is superimposed on the inspection target image, making it possible to visualize the mycoplasma and efficiently detect the mycoplasma.

[0104] The mycoplasma testing device of this embodiment includes a second learning device that is trained to determine whether or not the cell nucleus is contaminated with mycoplasma based on the mycoplasma image generated by the mycoplasma image generation unit, and a contamination result image generation unit that generates a mycoplasma contamination result image including at least one of contaminated cell nuclei and non-contaminated cell nuclei based on the determination result of the second learning device.

[0105] The second learning device can be configured with, for example, a multi-layer neural network (deep learning), such as a convolutional neural network, or other machine learning methods. The second learning device determines whether the cell nucleus is contaminated with mycoplasma based on the mycoplasma image generated by the mycoplasma image generation unit.

[0106] The contamination result image generation unit generates a mycoplasma contamination result image including at least one of contaminated cell nuclei and non-contaminated cell nuclei based on the judgment result of the second learning device. Contaminated cell nuclei are cell nuclei contaminated with mycoplasma, and non-contaminated cell nuclei are cell nuclei not contaminated with mycoplasma. The mycoplasma contamination result image is an image of only contaminated cell nuclei, an image of only non-contaminated cell nuclei, or an image of a mixture of contaminated and non-contaminated cell nuclei. This makes it possible to visualize contaminated and non-contaminated cell nuclei and confirm the contamination state of the cell nuclei.

[0107] The mycoplasma detection device according to this embodiment includes a third learning device that is trained to detect the number of contaminated cell nuclei based on the mycoplasma contamination result image generated by the contamination result image generation unit.

[0108] The third learner can be configured, for example, with a multi-layer neural network (deep learning), such as a convolutional neural network, but other machine learning methods may also be used. The third learner detects the number of cell nuclei (the number of contaminated cell nuclei and the number of non-contaminated cell nuclei) based on the mycoplasma contamination result image generated by the contamination result image generation unit. By detecting the number of cell nuclei, it is possible to determine whether the test sample is positive or negative.

[0109] In the mycoplasma testing device of this embodiment, the third learning device detects, as the number of cell nuclei, a representative value based on the number of cell nuclei detected in each of the multiple mycoplasma contamination result images generated by the contamination result image generation unit.

[0110] The third learning machine detects a representative value based on the number of cell nuclei detected in each of the multiple mycoplasma contamination result images generated by the contamination result image generation unit as the number of cell nuclei. For example, if a predetermined number of cell nuclei (e.g., 1,000) can be extracted from all of the multiple mycoplasma contamination result images, the number of cell nuclei detected in all of the mycoplasma contamination result images is recorded. Similar processing is performed across all captured images, and a representative value (e.g., average, maximum, etc.) based on the recorded number of cell nuclei is detected as the number of cell nuclei relative to the predetermined number of cell nuclei (the number of contaminated cell nuclei and the number of non-contaminated cell nuclei). This allows for accurate determination of whether the test sample is positive or negative.

[0111] In the mycoplasma testing device according to this embodiment, the first learning machine is generated using training data including a first image in which a cell nucleus is photographed and in which mycoplasma is present and an extracted image in which mycoplasma is extracted from the first image, as well as training data including a second image in which a cell nucleus is photographed and in which mycoplasma is not present and a specified image in which mycoplasma is not present.

[0112] The first learning machine is generated using training data including a first image in which cell nuclei are photographed and mycoplasma is present, and an extracted image in which mycoplasma is extracted from the first image. The first image is an input image for training in which multiple cell nuclei are photographed and mycoplasma is present, and the extracted image is an output image for training in which mycoplasma is extracted from the first image (i.e., an image in which the multiple cell nuclei have been deleted). The first learning machine is generated using training data including a second image in which cell nuclei are photographed and mycoplasma is not present, and a specified image in which mycoplasma is not present. The second image is an input image for training in which multiple cell nuclei are photographed and mycoplasma is not present, and the specified image is an output image for training in which neither mycoplasma nor cell nuclei are present.

[0113] As a result, when an inspection target image in which a cell nucleus is photographed is input to the first learning device, the first learning device can detect mycoplasma.

[0114] In the mycoplasma testing device of this embodiment, the second learning device is generated using learning data including a learning input image consisting of a set of the first image and the extracted image and a predetermined contaminated cell nucleus image, and learning data including a learning input image consisting of a set of the second image and the predetermined image and a predetermined non-contaminated cell nucleus image.

[0115] The second learning device is generated using learning data including a learning input image consisting of a pair of the first image and the extracted image, and a predetermined contaminated cell nucleus image. The first image and the extracted image can be input to the input layer of the second learning device via separate channels. The predetermined contaminated cell nucleus image is an image representing contaminated cell nuclei contaminated with mycoplasma among the cell nuclei in the learning input image. In other words, if there are non-contaminated cell nuclei that are not contaminated with mycoplasma, the non-contaminated cell nuclei can be excluded.

[0116] The second learning device is generated using learning data including a learning input image consisting of a set of a second image and a predetermined image, and a predetermined non-contaminated cell nucleus image. The second image and the predetermined image can be input to the input layer of the second learning device via separate channels. The predetermined non-contaminated cell nucleus image is an image representing non-contaminated cell nuclei that are not contaminated with mycoplasma among the cell nuclei in the learning input image. In other words, if contaminated cell nuclei contaminated with mycoplasma are present, the contaminated cell nuclei can be excluded.

[0117] As a result, when a mycoplasma image containing detected mycoplasma and multiple cell nuclei is input to the second learning device, the second learning device can determine whether the cell nuclei are contaminated cell nuclei or non-contaminated cell nuclei.

[0118] In the mycoplasma detection device according to this embodiment, the third learning machine is generated using a predetermined contaminated cell nucleus image, a predetermined non-contaminated cell nucleus image, and a teacher label indicating the number of cell nuclei.

[0119] The third learning machine is generated using a predetermined contaminated cell nucleus image, a predetermined non-contaminated cell nucleus image, and a teacher label indicating the number of cell nuclei. For example, the third learning machine can be generated using a contaminated cell nucleus image or a non-contaminated cell nucleus image with one cell nucleus and a teacher label indicating that the number of cell nuclei is one (e.g., "1"). The same applies when the number of cell nuclei is two or three. The third learning machine can also be generated using a contaminated cell nucleus image or a non-contaminated cell nucleus image with four or more cell nuclei and a teacher label indicating that the number of cell nuclei is four or more (e.g., "4"). As a result, when a mycoplasma contamination result image is input to the third learning machine, the third learning machine can detect the number of cell nuclei.

[0120] The mycoplasma testing device according to this embodiment includes an output unit that outputs cell nuclei number information that associates a predetermined number of cell nuclei with the number of contaminating cell nuclei.

[0121] The output unit outputs cell nuclei count information that associates a predetermined number of cell nuclei with the number of contaminating cell nuclei. The predetermined number can be, for example, 1000. This makes it easy to determine whether the test sample is positive or negative.

[0122] The mycoplasma testing device according to this embodiment includes a cell cluster extraction unit that extracts cell clusters each containing one or more cell nuclei based on the mycoplasma contamination result image generated by the contamination result image generation unit, and the third learning machine detects the number of contaminating cell nuclei for each cell cluster extracted by the cell cluster extraction unit.

[0123] The cell cluster extraction unit extracts cell clusters each consisting of one or more cell nuclei based on the mycoplasma contamination result image generated by the contamination result image generation unit. Specifically, the cell cluster extraction unit can extract cell clusters by cutting out a cut-out image including the cell cluster from the mycoplasma contamination result image.

[0124] The third learning unit detects the number of cell nuclei for each cell cluster extracted by the cell cluster extraction unit. The third learning unit uses a cut-out image containing a cell cluster as a learning input image in advance, and can learn (generate) the number of cell nuclei contained in the cell cluster as a teacher label. This makes it possible to accurately detect the number of cell nuclei even when the cell nuclei exist as clusters.

[0125] The mycoplasma testing device according to this embodiment includes a determination unit that, when the cell cluster extracted by the cell cluster extraction unit contains both contaminated cell nuclei and non-contaminated cell nuclei, determines whether the cell cluster is a contaminated cell nucleus or a non-contaminated cell nucleus based on predetermined conditions.

[0126] When the cell cluster extracted by the cell cluster extraction unit contains both contaminated and non-contaminated cell nuclei, the determination unit determines the cell cluster as either a contaminated cell nucleus or a non-contaminated cell nucleus based on predetermined conditions. For example, the determination unit may determine the cell cluster as the larger of the contaminated and non-contaminated cell nuclei. This increases the accuracy of the detection of the number of contaminated or non-contaminated cell nuclei.

[0127] The mycoplasma display device of this embodiment includes an acquisition unit that acquires an image of an object to be inspected in which a cell nucleus is photographed, and a mycoplasma image display unit that displays mycoplasma detected by a first learning device that has been trained to detect mycoplasma based on the image of the object to be inspected acquired by the acquisition unit and a mycoplasma image based on the image of the object to be inspected acquired by the acquisition unit.

[0128] The acquisition unit acquires an image of the test object in which the cell nuclei are photographed. The image of the test object can be an image area obtained by scanning an image of the test specimen captured by an imaging device (e.g., a microscope) over an area of ​​a predetermined size.

[0129] The mycoplasma image display unit displays the mycoplasma detected by the first learning device, which detects mycoplasma based on the inspection object image acquired by the acquisition unit, and a mycoplasma image based on the acquired inspection object image. The mycoplasma image can be an image in which the mycoplasma detected by the first learning device is superimposed on the inspection object image. This makes it possible to visualize the mycoplasma and efficiently detect mycoplasma.

[0130] The mycoplasma display device of this embodiment includes a contamination result image display unit that displays a mycoplasma contamination result image including at least one of contaminated cell nuclei and non-contaminated cell nuclei based on the judgment result of a second learning device that has been trained to determine whether or not a cell nucleus is contaminated with mycoplasma based on the mycoplasma image.

[0131] The contamination result image display unit displays a mycoplasma contamination result image including at least one of contaminated cell nuclei and non-contaminated cell nuclei based on the determination result of the second learning device, which determines whether or not a cell nucleus is contaminated with mycoplasma based on the mycoplasma image. Contaminated cell nuclei are cell nuclei contaminated with mycoplasma, and non-contaminated cell nuclei are cell nuclei not contaminated with mycoplasma. The mycoplasma contamination result image can be an image of only contaminated cell nuclei, an image of only non-contaminated cell nuclei, or an image of a mixture of contaminated and non-contaminated cell nuclei. This makes it possible to visualize contaminated and non-contaminated cell nuclei and confirm the contamination state of the cell nuclei.

[0132] In the mycoplasma display device according to this embodiment, the contamination result image display unit displays the mycoplasma contamination result image by assigning an identifier for identifying the cell nucleus to at least one of the contaminated cell nuclei and the non-contaminated cell nuclei.

[0133] The contamination result image display unit displays the mycoplasma contamination result image by assigning an identifier to at least one of the contaminated and non-contaminated cell nuclei to identify the cell nuclei. For example, different identifiers can be used for the contaminated and non-contaminated cell nuclei, such as, but not limited to, numbers. This allows easy identification of individual cell nuclei.

[0134] The mycoplasma display device according to this embodiment includes a setting unit for setting whether the identifier is to be displayed or not.

[0135] The setting unit sets whether to display or hide the identifier. This allows, for example, a user (examiner) to select whether or not the identifier is required when checking the contamination result image, thereby improving the user's visibility.

[0136] In the mycoplasma display device according to this embodiment, the contamination result image display unit displays the mycoplasma contamination result image in a manner that differs between contaminated cell nuclei and non-contaminated cell nuclei.

[0137] The contamination result image display unit displays the mycoplasma contamination result image in a different display mode for contaminated cell nuclei and non-contaminated cell nuclei. The boundaries between the contaminated cell nuclei and the outside world can be displayed in different colors. This allows easy identification of contaminated and non-contaminated cell nuclei.

[0138] In the mycoplasma display device of this embodiment, when a cell cluster consisting of one or more cell nuclei contains both contaminated and non-contaminated cell nuclei, the contamination result image display unit displays a mycoplasma contamination result image, treating the cell cluster as either a contaminated cell nucleus or a non-contaminated cell nucleus based on predetermined conditions.

[0139] When a cell cluster containing one or more cell nuclei contains both contaminated and non-contaminated cell nuclei, the contamination result image display unit displays the mycoplasma contamination result image by classifying the cell cluster as either a contaminated cell nucleus or a non-contaminated cell nucleus based on a predetermined condition. The predetermined condition can be, for example, the size of each of the contaminated and non-contaminated cell nuclei. This allows easy identification of each of the contaminated and non-contaminated cell nuclei.

[0140] The mycoplasma display device according to this embodiment includes a display unit that displays cell nuclei number information that associates a predetermined cell nuclei number with the number of contaminating cell nuclei.

[0141] The display unit displays cell nuclei count information that associates a predetermined number of cell nuclei with the number of contaminating cell nuclei. The predetermined number can be, for example, 1000. This makes it easy to determine whether the test sample is positive or negative.

[0142] In the mycoplasma display device according to this embodiment, the display unit displays an identifier for identifying a cell nucleus and a code indicating whether the cell nucleus is a contaminated cell nucleus or a non-contaminated cell nucleus, in association with each other.

[0143] The display unit displays an identifier for identifying a cell nucleus in association with a code indicating whether the cell nucleus is a contaminated cell nucleus or a non-contaminated cell nucleus. For example, different identifiers can be used for contaminated cell nuclei and non-contaminated cell nuclei, and numbers, for example, can be used, but are not limited to this. This makes it easy to identify whether each cell nucleus is a contaminated cell nucleus or a non-contaminated cell nucleus.

[0144] In the mycoplasma display device according to this embodiment, the display unit displays an image of a cell cluster containing one or more contaminated cell nuclei in association with the identifier.

[0145] The display unit displays an image of a cell cluster containing one or more contaminating cell nuclei in association with the identifier, thereby making it possible to easily identify each individual contaminating cell nucleus.

[0146] The mycoplasma inspection system according to this embodiment includes an imaging device that images cell nuclei and the aforementioned mycoplasma inspection device, and the mycoplasma inspection device acquires an image of the inspection object captured by the imaging device.

[0147] The mycoplasma inspection device can acquire an image of the inspection object captured by an imaging device (for example, a microscope) that captures an image of a cell nucleus.

[0148] The mycoplasma display system according to this embodiment comprises an imaging device that images cell nuclei and the aforementioned mycoplasma display device, and the mycoplasma display device acquires an image of the object to be inspected that has been captured by the imaging device.

[0149] The mycoplasma display device can acquire an image of the object to be inspected that is captured by an imaging device (for example, a microscope) that captures an image of a cell nucleus.

[0150] The learning device of this embodiment includes a first learning device trained to detect mycoplasma based on an image of an object to be inspected in which cell nuclei are photographed, a second learning device trained to determine whether the cell nuclei are contaminated with mycoplasma based on the mycoplasma detected by the first learning device and the image of the object to be inspected, and a third learning device trained to detect the number of cell nuclei using a mycoplasma contamination result image containing at least one of contaminated cell nuclei and non-contaminated cell nuclei, which is generated based on the determination result of the second learning device.

[0151] The method for generating a learning module according to this embodiment is a method for generating a learning module comprising a first learning module, a second learning module, and a third learning module, and generates the first learning module using training data including a first image in which cell nuclei are photographed and in which mycoplasma is present, and an extracted image in which mycoplasma is extracted from the first image, and training data including a second image in which cell nuclei are photographed and in which mycoplasma is not present, and a specified image in which mycoplasma is not present; generates the second learning module using training data including a training input image consisting of a set of the first image and the extracted image and a specified contaminated cell nucleus image, and training data including a training input image consisting of a set of the second image and the specified image and a specified non-contaminated cell nucleus image; and generates the third learning module using teacher labels indicating the specified contaminated cell nucleus image, the specified non-contaminated cell nucleus image, and the number of cell nuclei.

[0152] The first learning machine is generated using training data including a first image in which cell nuclei are photographed and mycoplasma is present, and an extracted image in which mycoplasma is extracted from the first image. The first image is an input image for training in which multiple cell nuclei are photographed and mycoplasma is present, and the extracted image is an output image for training in which mycoplasma is extracted from the first image (i.e., an image in which the multiple cell nuclei have been deleted). The first learning machine is generated using training data including a second image in which cell nuclei are photographed and mycoplasma is not present, and a specified image in which mycoplasma is not present. The second image is an input image for training in which multiple cell nuclei are photographed and mycoplasma is not present, and the specified image is an output image for training in which neither mycoplasma nor cell nuclei are present.

[0153] As a result, when an inspection target image in which a cell nucleus is photographed is input to the first learning device, the first learning device can detect mycoplasma.

[0154] The second learning device is generated using learning data including a learning input image consisting of a pair of the first image and the extracted image, and a predetermined contaminated cell nucleus image. The first image and the extracted image can be input to the input layer of the second learning device via separate channels. The predetermined contaminated cell nucleus image is an image representing contaminated cell nuclei contaminated with mycoplasma among the cell nuclei in the learning input image. In other words, if there are non-contaminated cell nuclei that are not contaminated with mycoplasma, the non-contaminated cell nuclei can be excluded.

[0155] The second learning device is generated using learning data including a learning input image consisting of a set of a second image and a predetermined image, and a predetermined non-contaminated cell nucleus image. The second image and the predetermined image can be input to the input layer of the second learning device via separate channels. The predetermined non-contaminated cell nucleus image is an image representing non-contaminated cell nuclei that are not contaminated with mycoplasma among the cell nuclei in the learning input image. In other words, if contaminated cell nuclei contaminated with mycoplasma are present, the contaminated cell nuclei can be excluded.

[0156] As a result, when a mycoplasma image containing detected mycoplasma and multiple cell nuclei is input to the second learning device, the second learning device can determine whether the cell nuclei are contaminated cell nuclei or non-contaminated cell nuclei.

[0157] The third learning machine is generated using a predetermined contaminated cell nucleus image, a predetermined non-contaminated cell nucleus image, and a teacher label indicating the number of cell nuclei. For example, the third learning machine can be generated using a contaminated cell nucleus image or a non-contaminated cell nucleus image with one cell nucleus and a teacher label indicating that the number of cell nuclei is one (e.g., "1"). The same applies when the number of cell nuclei is two or three. The third learning machine can also be generated using a contaminated cell nucleus image or a non-contaminated cell nucleus image with four or more cell nuclei and a teacher label indicating that the number of cell nuclei is four or more (e.g., "4"). As a result, when a mycoplasma contamination result image is input to the third learning machine, the third learning machine can detect the number of cell nuclei. [Explanation of symbols]

[0158] 10 PC 50 Inspection equipment 51 Control section 52 Input section 53 Mycoplasma Image Generation Unit 54 Contamination result image generation unit 55 Storage section 56 Cell mass extraction section 57 Interface section 58 Learning Machine 581 First Learning Unit 582 Second Learning Unit 583 Third Learning Unit 59 Decision Section 100 display device 101 Control section 102 Interface section 103 Display screen 104 Operation section 200 Culture equipment 201, 202 Culture units 203 Controller 204 Operation section 300 Imaging device

Claims

1. an acquisition unit that acquires an image of an object to be inspected, the image capturing an area where cell nuclei are present and where mycoplasma may be present; A first learning device that has undergone deep learning to detect mycoplasma based on the inspection object image acquired by the acquisition unit; a mycoplasma image generation unit that generates a mycoplasma image based on the mycoplasma detected by the first learning device and the inspection target image acquired by the acquisition unit; a second learning machine that has undergone deep learning to determine whether the cell nucleus is contaminated with mycoplasma based on the mycoplasma image generated by the mycoplasma image generation unit; and Equipped with The second learning device The method is generated using learning data including a set of learning input images consisting of a first image capturing an area where cell nuclei are present and where mycoplasma may be present and a mycoplasma-extracted image in which mycoplasma is extracted from the first image, and a contaminated cell nucleus image representing a contaminated cell nucleus contaminated with mycoplasma among the cell nuclei in the learning input image, Furthermore, the image is generated using learning data including a set of learning input images in which cell nuclei are photographed and in which mycoplasma is not present, and a predetermined image in which neither mycoplasma nor cell nuclei are present, and an uncontaminated cell nucleus image representing uncontaminated cell nuclei that are not contaminated with mycoplasma among the cell nuclei in the learning input image. Mycoplasma testing device.

2. A mycoplasma inspection device as described in claim 1, which is provided with a contamination result image generation unit that generates a mycoplasma contamination result image including at least one of contaminated cell nuclei and non-contaminated cell nuclei based on the judgment result of the second learning device.

3. The mycoplasma testing device according to claim 2, further comprising a third learning device that has undergone deep learning to detect the number of cell nuclei based on the mycoplasma contamination result image generated by the contamination result image generation unit.

4. The third learning device 4. The mycoplasma testing device according to claim 3, wherein the number of cell nuclei is detected as a representative value based on the number of cell nuclei detected in each of the plurality of mycoplasma contamination result images generated by the contamination result image generation unit.

5. The first learning device 5. The mycoplasma testing device according to claim 3 or 4, which is generated using training data including a first image in which a region in which cell nuclei are present and in which mycoplasma may be present is photographed and an extracted image in which mycoplasma is extracted from the first image, and training data including a second image in which cell nuclei are photographed but in which mycoplasma is not present and a specified image in which mycoplasma is not present.

6. The third learning device 6. The mycoplasma detection device according to claim 3, wherein the mycoplasma detection device is generated using a predetermined contaminated cell nucleus image, a predetermined non-contaminated cell nucleus image, and teacher labels indicating the number of cell nuclei.

7. 7. The mycoplasma testing device according to claim 6, further comprising an output unit that outputs cell nuclei number information that associates a predetermined number of cell nuclei with the number of contaminating cell nuclei.

8. a cell cluster extraction unit that extracts cell clusters containing one or more cell nuclei based on the mycoplasma contamination result image generated by the contamination result image generation unit, The third learning device The mycoplasma testing device according to any one of claims 3 to 7, wherein the number of cell nuclei is detected for each cell cluster extracted by the cell cluster extracting unit.

9. 9. The mycoplasma testing device according to claim 8, further comprising a determination unit that, when the cell clusters extracted by the cell cluster extraction unit contain both contaminated cell nuclei and non-contaminated cell nuclei, determines whether the cell clusters are contaminated cell nuclei or non-contaminated cell nuclei based on the size of the cell nuclei.

10. an acquisition unit that acquires an image of an object to be inspected, the image capturing an area where cell nuclei are present and where mycoplasma may be present; a mycoplasma image display unit that displays mycoplasma detected by a first learning device that has undergone deep learning to detect mycoplasma based on the inspection target image acquired by the acquisition unit and a mycoplasma image based on the inspection target image acquired by the acquisition unit; a second learner that is deep-trained to determine whether the cell nucleus is contaminated with mycoplasma based on the mycoplasma image; and Equipped with The second learning device The method is generated using learning data including a set of learning input images consisting of a first image capturing an area where cell nuclei are present and where mycoplasma may be present and a mycoplasma-extracted image in which mycoplasma is extracted from the first image, and a contaminated cell nucleus image representing a contaminated cell nucleus contaminated with mycoplasma among the cell nuclei in the learning input image, Furthermore, the image is generated using learning data including a set of learning input images in which cell nuclei are photographed and in which mycoplasma is not present, and a predetermined image in which neither mycoplasma nor cell nuclei are present, and an uncontaminated cell nucleus image representing uncontaminated cell nuclei that are not contaminated with mycoplasma among the cell nuclei in the learning input image. Mycoplasma display device.

11. The mycoplasma display device according to claim 10, further comprising a contamination result image display unit that displays a mycoplasma contamination result image including at least one of contaminated cell nuclei and non-contaminated cell nuclei based on the judgment result of a second learning device that has undergone deep learning to determine whether or not a cell nucleus is contaminated with mycoplasma based on the mycoplasma image.

12. The contamination result image display unit The mycoplasma display device according to claim 11, wherein an identifier for identifying the cell nucleus is assigned to at least one of the contaminated cell nuclei and the non-contaminated cell nuclei, and an image of the mycoplasma contamination result is displayed.

13. The mycoplasma display device according to claim 12, further comprising a setting unit for setting whether the identifier is to be displayed or not.

14. The contamination result image display unit The mycoplasma display device according to any one of claims 11 to 13, wherein the mycoplasma contamination result image is displayed in different display modes for contaminated cell nuclei and non-contaminated cell nuclei.

15. The contamination result image display unit A mycoplasma display device according to any one of claims 11 to 14, wherein when a cell cluster containing one or more cell nuclei contains both contaminated and non-contaminated cell nuclei, the cell cluster is displayed as either a contaminated or non-contaminated cell nucleus based on the size of the cell nuclei as an image of mycoplasma contamination results.

16. The mycoplasma display device according to any one of claims 11 to 15, further comprising a display unit that displays cell nucleus number information that associates a predetermined cell nucleus number with the number of contaminating cell nuclei.

17. The display unit that displays the cell nucleus number information includes: The mycoplasma display device according to claim 16, which displays an identifier for identifying a cell nucleus in association with a code indicating whether the cell nucleus is a contaminated cell nucleus or a non-contaminated cell nucleus.

18. The display unit that displays the cell nucleus number information includes: The mycoplasma display device according to claim 17, wherein an image of a cell cluster containing one or more cell nuclei is displayed in association with the identifier.

19. An imaging device for imaging a cell nucleus and the mycoplasma testing device according to any one of claims 1 to 9, The mycoplasma inspection device is a mycoplasma inspection system that acquires an image of an inspection object captured by the imaging device.

20. An imaging device for imaging a cell nucleus and the mycoplasma display device according to any one of claims 10 to 18, The mycoplasma display device is a mycoplasma display system that acquires an image of an object to be inspected captured by the imaging device.

21. a first learning machine that has been deep-trained to detect mycoplasma based on an image of an object to be inspected that includes an area where a cell nucleus is present and where mycoplasma may be present; A second learning device that has undergone deep learning to determine whether the cell nucleus is contaminated with mycoplasma based on the mycoplasma detected by the first learning device and the image to be inspected; a third learner that is deep-trained to detect the number of cell nuclei using a mycoplasma contamination result image including at least one of contaminated cell nuclei and non-contaminated cell nuclei, the third learner being generated based on the determination result of the second learner; Equipped with The second learning device The method is generated using learning data including a set of learning input images consisting of a first image capturing an area where cell nuclei are present and where mycoplasma may be present and a mycoplasma-extracted image in which mycoplasma is extracted from the first image, and a contaminated cell nucleus image representing a contaminated cell nucleus contaminated with mycoplasma among the cell nuclei in the learning input image, Furthermore, the image is generated using learning data including a set of learning input images in which cell nuclei are photographed and in which mycoplasma is not present, and a predetermined image in which neither mycoplasma nor cell nuclei are present, and an uncontaminated cell nucleus image representing uncontaminated cell nuclei that are not contaminated with mycoplasma among the cell nuclei in the learning input image. Learning device.

22. On the computer, A process of acquiring an image of an object to be inspected, in which an area where cell nuclei are present and where mycoplasma may be present is photographed; When an image of an object to be inspected in which a cell nucleus is present and an area where mycoplasma may be present is captured is input, the acquired image of the object to be inspected is input to a first learning device that has been deep-trained to detect mycoplasma, and a process of generating a mycoplasma image based on the mycoplasma detected by the first learning device and the image of the object to be inspected; A process of deep learning a second learning device to determine whether the cell nucleus is contaminated with mycoplasma based on the generated mycoplasma image; Furthermore, the second learning machine is generated using learning data including a set of learning input images consisting of a first image capturing an area where cell nuclei are present and where mycoplasma may be present, and a mycoplasma-extracted image in which mycoplasma is extracted from the first image, and a contaminated cell nucleus image representing a contaminated cell nucleus contaminated with mycoplasma among the cell nuclei in the learning input image; a process of generating the second learning machine using learning data including learning input images consisting of a set of a second image in which cell nuclei are photographed and in which no mycoplasma is present and a predetermined image in which neither mycoplasma nor cell nuclei are present, and uncontaminated cell nucleus images representing uncontaminated cell nuclei that are not contaminated with mycoplasma among the cell nuclei in the learning input images; and A computer program that executes the following:

23. An image of the test object is acquired, which captures the area where cell nuclei are present and where mycoplasma may be present. When an image of an object to be inspected is input, in which a cell nucleus is present and an area where mycoplasma may be present is photographed, the acquired image of the object to be inspected is input to a first learning device that has been deep-trained to detect mycoplasma, and a mycoplasma image is generated based on the mycoplasma detected by the first learning device and the image of the object to be inspected; Deep learning a second learner to determine whether the cell nucleus is contaminated with mycoplasma based on the generated mycoplasma image; Furthermore, the second learning machine is generated using learning data including a set of learning input images consisting of a first image capturing an area where cell nuclei are present and where mycoplasma may be present, and a mycoplasma-extracted image in which mycoplasma is extracted from the first image, and a contaminated cell nucleus image representing a contaminated cell nucleus contaminated with mycoplasma among the cell nuclei in the learning input image; The second learner is generated using learning data including learning input images consisting of a set of a second image in which cell nuclei are photographed and in which no mycoplasma is present and a predetermined image in which neither mycoplasma nor cell nuclei are present, and uncontaminated cell nucleus images representing uncontaminated cell nuclei that are not contaminated with mycoplasma among the cell nuclei in the learning input images. Mycoplasma testing methods.

24. A method for generating a learning device including a first learning device, a second learning device, and a third learning device, Using training data including a first image in which a region in which cell nuclei are present and mycoplasma may be present and an extracted image in which mycoplasma is extracted from the first image, and training data including a second image in which cell nuclei are present and mycoplasma is not present, and a predetermined image in which mycoplasma is not present, a first learner is generated that has been deep-trained to detect mycoplasma when an image to be inspected in which a region in which cell nuclei are present and mycoplasma may be present is input, generating a mycoplasma image based on the mycoplasma detected by the first learning device and the inspection target image; Deep learning a second learner to determine whether the cell nucleus is contaminated with mycoplasma based on the generated mycoplasma image; Furthermore, the second learning machine is generated using learning data including a set of learning input images consisting of a first image capturing an area where cell nuclei are present and where mycoplasma may be present, and a mycoplasma-extracted image in which mycoplasma is extracted from the first image, and a contaminated cell nucleus image representing a contaminated cell nucleus contaminated with mycoplasma among the cell nuclei in the learning input image; generating the second learning machine using learning data including learning input images consisting of a set of a second image in which cell nuclei are photographed and in which no mycoplasma is present and a predetermined image in which neither mycoplasma nor cell nuclei are present, and uncontaminated cell nucleus images representing uncontaminated cell nuclei that are not contaminated with mycoplasma among the cell nuclei in the learning input images; A method for generating a learning machine that generates a third learning machine that has been deep-trained to detect the number of cell nuclei when a mycoplasma contamination result image containing at least one of contaminated cell nuclei and non-contaminated cell nuclei is input using a predetermined contaminated cell nuclei image, a predetermined non-contaminated cell nuclei image, and teacher labels indicating the number of cell nuclei.

Images