Data analysis apparatus and data analysis method for microscopic laminated images

Abstract

A data analysis apparatus for analyzing microscopic Z-stack images used for cell diagnosis is provided. The data analysis apparatus for microscopic Z-stack images of a specimen (S) containing a number of cells includes: a microscope (20) for obtaining Z-stack images of the specimen (S), an imaging unit (30) for capturing Z-stack images obtained by the microscope (20), a storage unit (70) for storing Z-stack images captured by the imaging unit (30), and a control unit (50). The control unit (50) detects cells from the Z-stack images stored in the storage unit (70), obtains a cell image group corresponding to each detected cell, cuts out a cell image from the images constituting the cell image group for each cell, obtains the confidence level of each of the multiple classification categories for the cell image of each cell, and visualizes the confidence level obtained for each classification category.

Description

Technical Field

[0001] This invention relates to a data analysis apparatus and a data analysis method for microscopic layered images. Background Technology

[0002] To digitize specimens with depth, such as those used in cell diagnostics, methods have been developed to obtain Z-stacked images covering the entire slide specimen. However, obtaining Z-stacked images of the entire slide specimen results in enormous data sizes. Therefore, techniques such as Extended Depth-of-Field (EDF) for generating focused synthetic images from Z-stacked images have been developed (e.g., U.S. Patent No. 9,332,190).

[0003] In addition, a technique has been developed to record data by compressing the captured Z-stacked image data, thereby reducing the data size while maintaining the information of each Z layer of the Z-stacked image (e.g., International Publication No. 2023 / 233922).

[0004] Furthermore, artificial intelligence is used to analyze large numbers of digitized cell images and to detect or classify objects. In the analysis of large numbers of cell images, there are situations where it is desirable to analyze the types, states, and proportions of cells contained within a cell population. In this case, techniques have been developed to classify the type and state of each cell using artificial intelligence, obtaining the confidence level of each classification as the output of the artificial intelligence, and depicting the confidence levels of classification for many cells using histograms, two-dimensional plots, etc. (e.g., International Publication No. 2020 / 004101). Summary of the Invention

[0005] The problem that the invention aims to solve Figure 1 is a schematic diagram of a portion of a microscopic Z-stack image containing many cells. In the example shown in Figure 1, six Z layers (L1–L6) are shown. The distance between Z layers in the Z-stack image is, for example, less than 1 μm. As shown in Figure 1, when the distance between Z layers is narrower than the depth of the cell, a single cell can be observed spanning multiple Z layers.

[0006] If such a Z-stacked image is processed using the EDF method, the images of cells C and D, which are overlapped in the depth direction, are linked, as shown in Figure 2. For example, in cell diagnosis, it is necessary to observe cell clumps with many cells. Therefore, if data processed by the EDF method is used, it is difficult to observe and resolve cell clumps with many cells.

[0007] Previously, no analytical method was established that considered the three-dimensional configuration of cell groups and treated the cell groups as a whole as an object for the many cells contained in Z-stacked images.

[0008] The purpose of this invention is to provide a data analysis device and a data analysis method for analyzing microscopic stacked images used for cell diagnosis.

[0009] Methods for solving problems The first aspect of this invention is a data analysis apparatus for a microscopic layered image of a specimen containing many cells, the microscopic layered image data analysis apparatus having: A microscope was used to obtain a Z-stacked image of the specimen. The camera unit captures the Z-stacked image obtained by the microscope; Storage unit, which stores the Z-stacked images captured by the camera unit; and Control Department, its Cells are detected from the Z-stacked image stored in the storage unit. Obtain a set of cell images corresponding to each detected cell. For each cell, a cell image is extracted from the images constituting the cell image group. For each cell image, obtain the confidence level for each of the multiple classification categories. The confidence scores obtained for each of the categories are visualized.

[0010] The second aspect of this invention is a data analysis method for microscopic layered images of a specimen containing many cells, wherein the data analysis method for the microscopic layered images, Z-stack images of the specimen were obtained using a microscope. The camera unit captures images of the Z-stack obtained by the microscope. The storage unit stores the Z-stacked images captured by the camera unit. The control unit detects cells from the Z-stacked image stored in the storage unit. The control unit obtains a set of cell images corresponding to each detected cell. The control unit extracts cell images from the images constituting the cell image group for each cell. The control unit obtains the confidence level of each category in multiple classifications for the cell images of each cell. The control unit visualizes the confidence level obtained for each category classification.

[0011] The effects of the invention According to the present invention, it is possible to analyze Z-stacked images of microscopes used for cell diagnosis. Attached Figure Description

[0012] Figure 1 is a schematic diagram of a portion of the Z-stacked image.

[0013] Figure 2 is a schematic diagram of the Z-stacked image processed by the EDF method.

[0014] Figure 3 is a schematic diagram of the data parsing device according to the embodiment.

[0015] Figure 4 is a diagram illustrating the method for obtaining a Z-stacked image.

[0016] Figure 5 is a flowchart illustrating the data parsing method of the implementation method.

[0017] Figure 6 is a diagram showing the images of the cell image group.

[0018] Figure 7 is a diagram showing cell images of each cell.

[0019] Figure 8 is a schematic diagram of cell images of each cell.

[0020] Figure 9 is a diagram showing the classification of cell images by category.

[0021] Figure 10(a) and (b) are histograms of confidence levels.

[0022] Figure 11(a), (b), and (c) are two-dimensional plots of confidence levels.

[0023] Figure 12 is the first two-dimensional plot of the multi-level gating.

[0024] Figure 13 is the second two-dimensional plot of the multi-level gating.

[0025] Symbol Explanation 10 Stage, 20 Microscope, 30 Imaging unit, 40 Focal plane moving mechanism, 42 Stage moving mechanism, 50 Control unit, 55 Image buffer, 70 Storage unit, 80 Transmission unit, S Specimen. Detailed Implementation

[0026] The following describes a data analysis apparatus and method for microscopic stacked images with reference to the accompanying drawings. The data analysis apparatus and method for microscopic stacked images of the present invention analyze microscopic Z-stack images containing many cells and obtain information about the overall cell population contained in the Z-stack image.

[0027] Figure 3 shows a schematic configuration diagram of the data analysis apparatus of this embodiment. The data analysis apparatus of this embodiment includes a microscope 20, an imaging unit 30, a focal plane moving mechanism 40, and a control unit 50. The data analysis apparatus may further include a stage 10, a stage moving mechanism 42, an image buffer 55, a storage unit 70, and a transmission unit 80. Additionally, the data analysis apparatus may also include an illumination 22.

[0028] The specimen S, which is the object of the camera, is placed on the stage 10. In the example shown in Figure 3, the stage 10 is parallel to the XY plane.

[0029] Microscope 20 acquires an image of the specimen S on the focal plane. Microscope 20 has an optical system such as lenses.

[0030] The camera unit 30 captures images obtained by the microscope 20 based on a timing signal. The camera unit 30 is, for example, a digital camera.

[0031] The focal plane movement mechanism 40 moves the focal plane of the microscope 20 along the optical axis based on the focal plane control signal. The optical axis is orthogonal to the stage 10. In the example shown in FIG3, the optical axis is set as the Z direction. The optical axis (Z direction) in this embodiment is the vertical direction, but the optical axis is not limited to the vertical direction.

[0032] The stage movement mechanism 42 moves the microscope 20 and the stage 10 relative to each other in a direction orthogonal to the optical axis (XY direction) based on the stage movement control signal.

[0033] The image buffer 55 can also be configured between the camera unit 30 and the storage unit 70. The image buffer 55 can also be configured as part of the output side of the camera unit 30. The image buffer 55 temporarily stores image data of multiple images captured by the camera unit 30.

[0034] The storage unit 70 stores the image data transferred from the image buffer 55.

[0035] The transmitting unit 80 sends the compressed image data to the outside.

[0036] The control unit 50 controls the camera unit 30, the focal plane moving mechanism 40, the stage moving mechanism 42, the storage unit 70, and the sending unit 80.

[0037] Specifically, the control unit 50 sends a focal plane control signal to the focal plane moving mechanism 40, which moves the focal plane to a predetermined focal coordinate (Z coordinate). The control unit 50 also sends a camera timing signal to the camera unit 30, indicating when to capture an image.

[0038] The control unit 50 sends a stage movement control signal to the stage movement mechanism 42, causing the microscope 20 and the stage 10 to move relative to each other in a direction orthogonal to the optical axis (XY direction). Based on the focal plane control signal and the stage movement control signal, the control unit 50 sends an imaging timing signal to the imaging unit 30 when the focal coordinates become predetermined coordinates. Thus, the imaging unit 30 is able to capture an image of the specimen S on the desired focal plane.

[0039] The control unit 50 can also control the transmission unit 80 to send image data to the outside. The control unit 50 can also add additional information such as camera conditions to the image data stored in the storage unit 70 and the image data sent to the outside from the transmission unit 80.

[0040] Here, the Z-stacked image will be explained with reference to FIG4. FIG4 shows an example of a method for obtaining a Z-stacked image. The Z-stacked image is obtained in a microscope 20 by stacking planar images of multiple Z layers (L1 to L5) with different focal planes in the depth direction. For example, using an imaging unit 30 equipped with a region sensor such as a CCD or CMOS, the image of the focal plane on the specimen S is imaged onto the optical system of the region sensor, and the distance between the objective lens of the microscope 20 and the specimen S is gradually changed while repeatedly taking pictures. In this way, planar images of multiple Z layers (L1 to L5) are obtained continuously to obtain a Z-stacked image.

[0041] Here, planar images are captured repeatedly, for example, at Z-layer spacings of 250 nm or 500 nm for each Z-layer. The Z-layer spacing is typically on the order of 200 nm to 2 μm.

[0042] It should be noted that the depth of field of the subject near the focal plane is determined by the optical system. Therefore, it is preferable to set the optical system so that the Z-layer spacing is approximately the same as the depth of field of the subject. For example, the depth of field of the subject is determined by the numerical aperture of the objective lens and the numerical aperture of the illumination light. When shooting with transmitted light, the aperture of the transmitted illumination light incident on the specimen S can be adjusted by using an aperture stop.

[0043] By using a vertical stage (Z-stage), the optical system of the microscope 20, or a portion thereof including the objective lens, can be moved along the optical axis, thereby changing the distance between the specimen S and the objective lens. Alternatively, the stage 10 carrying the specimen S can be placed on the vertical stage and moved along the optical axis, thereby changing the distance between the specimen S and the objective lens.

[0044] The vertical stage can be a single stage. Alternatively, it can be a combination of two or more stages, such as a coarse stage for retracting the optical system and a fine stage for fine-tuning the distance between the objective lens and the sample. The vertical stage allows for fine-tuning with sufficient accuracy relative to the Z-layer spacing of the stacked image.

[0045] The imaging range of the stacked images during imaging is determined based on the characteristics of the specimen S. For example, in the case of a cytological diagnostic specimen, imagine a large number of cells fixed on a slide, forming a stacked structure of cells. In this case, the imaging range is set at a sufficient height relative to the height of the stacked cells from the surface of the slide.

[0046] It is preferable to use autofocus processing in determining the imaging range of the stacked images. Specifically, for a Z-stacked image temporarily obtained for the purpose of implementing autofocus, the degree of focus of each Z-layer image is quantified, and the range of the Z-layers whose degree of focus meets certain conditions is determined as the imaging range.

[0047] As a method to quantify the degree of focus, the method described later can also be used to detect the cells or nuclei contained in the Z-stacked image, and after selecting the most suitable Z layer for observation for each cell or nucleus, the number of images of each selected cell or nucleus can be used as an indicator of the degree of focus. In this case, the range of Z layers in which a certain number of selected cells or nuclei exist can be set as the imaging range.

[0048] The control unit 50 pre-obtains parameters such as the imaging range and Z-layer spacing based on the type of specimen S. The control unit 50 switches these parameters according to the type of slide, the content of image data obtained through focusing processing, and the position of the vertical stage. This allows for the efficient acquisition of images of sufficient quality necessary for reconstructing a suitable three-dimensional image.

[0049] To integrate the acquired overlay images from multiple locations, the control unit 50 records data on the vertical stage position during image capture. For example, the control unit 50 detects the vertical stage position using an encoder mounted on the stage 10. Alternatively, the control unit 50 can accumulate and calculate the amount of movement of the stage 10 from its origin position to detect the vertical stage position. Furthermore, the control unit 50 can also detect the vertical stage position by monitoring the pulses of the stepper motor that moves the stage 10 and the control pulses input to the stepper motor's controller using digital circuitry.

[0050] In the example shown in Figure 4, an image can also be captured after the movement of the focal plane has completed and stopped. Alternatively, the focal plane can be moved continuously, and an image can be captured when the focal plane reaches a predetermined position. The latter imaging method can capture images at a higher speed compared to the former.

[0051] It should be noted that in the latter imaging method, if the distance the focal plane moves during the exposure time of the imaging unit 30 is sufficiently large relative to the depth of field of the subject in the optical system, image blurring may occur. To avoid this, it is preferable to set the exposure time and focal plane movement speed of the imaging unit 30 according to the depth of field of the subject in the optical system. Furthermore, if a higher focal plane movement speed is desired, using a flash in the illumination 22 can reduce image blurring caused by focal plane movement.

[0052] The control unit 50 sends a stage movement control signal to move the microscope 20 and the stage 10 relative to each other in a direction orthogonal to the optical axis. Additionally, the control unit 50 sends a focal plane control signal to move the focal plane to a predetermined focal coordinate. If the vertical stage is moved to the desired position for imaging, the control unit 50 sends an image capture timing signal to the imaging unit 30. The imaging unit 30 captures an image upon receiving the image capture timing signal.

[0053] The data parsing method of this embodiment will be described below with reference to FIG5. FIG5 is a flowchart showing the data parsing method of the embodiment.

[0054] First, the camera unit 30 captures the Z-stack image obtained by the microscope 20 based on the camera timing signal (step S1).

[0055] Next, the control unit 50 reads the Z-stacked image into the computer's memory (step 2). Preferably, the Z-stacked image is read into the GPU (Graphical Processing Unit) memory.

[0056] It should be noted that when the Z-stacked image captured in step S1 is compressed and saved, the compressed data is decoded. When decoding the compressed data using a GPU or a device connected to the GPU, the decoded data is output to the GPU's memory. For the compressed Z-stacked image, all Z layers can be decoded, or only a portion of the Z layers can be decoded.

[0057] Next, the control unit 50 detects cells from each Z layer of the Z-stacked image (step S3). Specifically, the control unit 50 detects cells by detecting cell nuclei as shown in FIG1. ​​For example, by binarizing the image based on the hue, density, contrast, etc. of the cell nuclei, the control unit 50 can detect cell nuclei by resolving their size, shape, etc. Alternatively, the control unit 50 can also use object detection algorithms based on deep learning such as YoLo (You Only Look Once) to detect cell nuclei.

[0058] Thus, for example, in the case shown in Figure 1, it is possible to detect that the nucleus of cell A spans three Z layers L2 to L4. Furthermore, it is possible to detect that the nucleus of cell B spans two Z layers, L4 and L5.

[0059] Next, the control unit 50 acquires a set of cell images (step S4). For example, in the example shown in FIG1, cell A is detected across five Z layers L1 to L5. Therefore, the control unit 50 acquires a set of cell images corresponding to cell A from the five Z layers L1 to L5. Similarly, the control unit 50 acquires a set of cell images corresponding to cell B from the five Z layers L2 to L6. The total number of cell image sets acquired in this way is equivalent to the total number of cells contained in the Z-stacked image of the specimen S.

[0060] Figure 6 is an example of a group of cell images. In the example shown in Figure 6, cell A spans layers Z, L1 to L9. These nine images constitute the group of cell images corresponding to cell A. Similarly, cell B spans layers Z, L1 to L5. These five images constitute the group of cell images corresponding to cell B.

[0061] It should be noted that in step S4, a method can also be used where the difference in the XY position of cell nuclei detected in adjacent Z layers is defined as the same cell image group if the width and height of the region of interest (ROI) are less than 25%. Alternatively, the area of ​​repeated regions in adjacent Z layers can be used as an indicator, or shape or color information can be considered in addition to position to obtain a cell image group.

[0062] Next, the control unit 50 extracts cell images from each cell image group (step S5). Specifically, the control unit 50 determines one or more Z layers suitable for cell observation and analysis from the images constituting the cell image group for each cell. Then, the control unit 50 extracts cell images from the determined Z layers for each cell.

[0063] More specifically, preferably, among the images constituting the cell image group, the image with the focus on the cell nucleus is selected as the cell image. In step S3, it is also possible that, in the case where cells are detected by detecting the cell nucleus, the Z-stack image of the detected cell nucleus is selected as the cell image.

[0064] Furthermore, for multiple images constituting a cell image group, statistical measures such as standard deviation and variance can be used as indicators to select cell images. In the example shown in Figure 6, the values ​​under each image represent the standard deviation of the pixel brightness values ​​of each image. The control unit 50 can also select the image with the largest standard deviation of pixel brightness values ​​as the cell image.

[0065] In addition, for multiple images that constitute a group of cell images, image processing such as edge detection can be used to select cell images.

[0066] Here, the control unit 50 preferably cuts out a cell image with the cell nucleus of the target cell positioned in the center. The size of the cell image only needs to be the size required for subsequent analysis.

[0067] Figure 7 is an example of a cell image showing each cell. Thus, as shown in Figure 7, a cell image suitable for observation and analysis is selected for each cell. Figure 8 is a schematic diagram corresponding to the cell images of each cell in Figure 1. As explained with reference to Figure 2, if the Z-stacked image as shown in Figure 1 is processed using the EDF method, the images of cells C and D, which are overlapped in the depth direction, are linked as shown in Figure 2. In contrast, according to the present invention, different images can be obtained for cells C and D, which are overlapped in the depth direction, as shown in Figure 8.

[0068] It should be noted that in step S3, the control unit 50 may not detect cell nuclei in all Z layers. Alternatively, the control unit 50 may detect cell nuclei every predetermined Z layer (e.g., every four Z layers). This reduces the time required for cell nuclei detection.

[0069] Next, the control unit 50 analyzes the cell images of each cell obtained in step S5 and performs category classification (step S6). Then, the control unit 50 obtains the confidence level for each category classification of each cell. Here, the confidence level can be obtained using, for example, the Softmax function. In this case, the sum of the confidence levels for each category is 1. Alternatively, the confidence level can also be obtained using the Sigmoid function. In this case, the sum of the confidence levels for each category is not necessarily 1. When plotting a histogram or two-dimensional graph of the confidence level, the cell populations are sometimes broken at both ends of 0 and 1 in the confidence level obtained using the Softmax function, resulting in an unnatural distribution. There are cases where it is preferable to use the Sigmoid function to achieve a more natural distribution. In the following example, the confidence level obtained using the Sigmoid function is shown.

[0070] For example, the control unit 50 performs image classification on each cell image based on the multi-Axis Vision Transformer (MaxViT) method. Figure 9 is an example of the category classification of cell images. In the example shown in Figure 9, a model that has been pre-learned is used to classify cell images into four categories: leukocytes, glandular cells (Glan), epithelial cells (Epi), and debris. It should be noted that the model has been learned to classify foreign objects such as debris and images where cells are not visible in the center of the cell image into the Debris category. In the example shown in Figure 9, the right side of the cell image represents the confidence level of each category when the cell image is classified using the model.

[0071] Then, the control unit 50 visualizes the confidence level of each category classification (step S7). Specifically, the control unit 50 outputs the confidence level of each category classification as a histogram or a two-dimensional plot. Hereinafter, an example of the control unit 50 outputting the confidence level of each category classification as a histogram will be described with reference to FIG10. FIG10 shows the results of analyzing two cytological specimens and creating a histogram of the confidence level of Debris based on the obtained results. FIG10(a) shows the histogram of cytological specimen A. FIG10(b) shows the histogram of cytological specimen B. In FIG10(a) and (b), the horizontal axis of the graph represents the confidence level of Debris. The vertical axis of the graph represents the frequency of the confidence level of Debris. The upper right corner of the graph shows a magnified view of a portion of the graph with the vertical axis enlarged.

[0072] Regarding specimen A for cell diagnosis shown in Figure 10(a), the total number of cells analyzed was 11,859. Cells with a confidence level of 0.8 or higher for Debris were identified as Debris. Therefore, 349 cells were identified as Debris. Similarly, regarding specimen B for cell diagnosis shown in Figure 10(b), the total number of cells analyzed was 133,136. Of these, 1,687 cells were identified as Debris.

[0073] Then, the control unit 50 can display cell images of cells identified as Debris, i.e., cells included in the Debris gate, to the output unit. In this way, by using a histogram, the number and proportion of cells classified into a specific category (the examples shown in Figure 10 (a) and (b) are Debris) can be visualized. In addition, by setting gates in the histogram, specific categories can be easily extracted.

[0074] It should be noted that the visualization of confidence levels for each category classification based on control unit 50 is not limited to histograms or two-dimensional plots. For example, t-distributed stochastic neighbor embedding (t-SNE) or self-organizing map (SOM), sub-image anomaly detection with deep pyramid correspondences (SPADE), uniform manifold approximation and projection (UMAP), isometric mapping (Isomap), locally linear embedding (LLE), autoencoder, variational autoencoder (VAE), principal component analysis (PCA), independent component analysis (ICA), linear discriminant analysis (LDA), factor analysis (FA), multidimensional scaling (MDS), and non-negative matrix factorization (NMF) can also be used. Dimensionality reduction and plotting are achieved using techniques such as Factorization, Kernel Principal Component Analysis (KPCA), tangent distance-based manifold learning, Maximum Variance Unfolding (MVU), Neighborhood Components Analysis (NCA), Laplacian Eigenmaps, Diffusion Maps, Random Projection (RP), Feature Agglomeration, Sparse Principal Component Analysis (SPCA), Tangent Principal Component Analysis (tPCA), and Deep Embedding Clustering (DEC).In the case of dimensionality reduction rendering, for example, by color-coding and displaying each presumed category for each cell, it is possible to determine where each type of cell is located in the drawing. Methods for classifying the presumed categories of each cell include, for example, classifying by using the category with the highest confidence level, or classifying by setting a category based on the analysis of a gate set in a histogram or two-dimensional drawing.

[0075] The following describes an example of the control unit 50 outputting the confidence level of each category in a two-dimensional plot, with reference to FIG11. FIG11 shows the results of analyzing three cellular diagnostic specimens and creating a two-dimensional plot of the confidence levels for Debris and leukocytes based on the obtained results. FIG11(a) shows the two-dimensional plot for cellular diagnostic specimen A. FIG11(b) shows the two-dimensional plot for cellular diagnostic specimen B. FIG11(c) shows the two-dimensional plot for cellular diagnostic specimen C. In FIG11(a), (b), and (c), the horizontal axis of the graph represents the confidence level of Debris. The vertical axis of the graph represents the confidence level of leukocytes. The hue represents the concentration. The square in the upper left corner of the graph represents the gate set for identifying leukocytes.

[0076] In specimen A of cell diagnosis shown in Figure 11(a), the proportion of white blood cells was 5.1%. In specimen B of cell diagnosis shown in Figure 11(b), the proportion of white blood cells was 14.0%. In specimen C of cell diagnosis shown in Figure 11(c), the proportion of white blood cells was 69.3%.

[0077] Then, the control unit 50 can display cell images of cells identified as white blood cells, i.e., cells contained in the gates (a), (b), and (c) of FIG11, to the output unit.

[0078] The following describes an example of multi-level gating with reference to Figures 12 and 13. First, as shown in Figure 12, the control unit 50 outputs a first two-dimensional graph for the cell diagnostic specimen. This first two-dimensional graph uses the horizontal axis of the graph as the confidence level for Debris and the vertical axis as the confidence level for white blood cells. In Figure 12, the square in the upper left corner of the graph represents the gate used to identify white blood cells. Additionally, the square in the lower right corner of the graph represents the gate used to identify Debris. The control unit 50 can display cell images identified as white blood cells or Debris cells on the output unit.

[0079] The control unit 50 performs the next level of gating for the cells within the gates shown in the lower left square of the graph in Figure 12. Specifically, as shown in Figure 13, the control unit 50 outputs a second two-dimensional graph for the cells within the lower left gate of the first two-dimensional graph. This second two-dimensional graph uses the horizontal axis of the graph as the confidence level for epithelial cells and the vertical axis as the confidence level for glandular cells. In Figure 13, the vertical and horizontal lines represent quadrant gates that divide the cell population into four categories. The lower right division of the graph, separated by the two lines of the quadrant gates, represents the gate used to identify epithelial cells. Additionally, the upper left division represents the gate used to identify glandular cells. Furthermore, the lower left division corresponds to the gates for other cells that do not belong to any of these categories.

[0080] The control unit 50 can display cell images of cells identified through each gate in Figures 12 and 13 to the output unit. In this way, by performing multi-level gating, detailed analysis of cell diagnostic specimens can be achieved.

[0081] It should be noted that in step S5, when the control unit 50 determines multiple Z layers suitable for cell observation and analysis from the images constituting the cell image group for each cell, the control unit 50 repeats steps S1 to S7 for each Z layer.

Claims

1. A data analysis apparatus for a microscopic layered image of a specimen containing a plurality of cells, the microscopic layered image data analysis apparatus comprising: A microscope was used to obtain a Z-stacked image of the specimen; The camera unit captures the Z-stacked image obtained by the microscope; A storage unit that stores the Z-stacked images captured by the camera unit; as well as Control Department, its Cells are detected from the Z-stacked image stored in the storage unit. Obtain a set of cell images corresponding to each detected cell. For each cell, a cell image is extracted from the images constituting the cell image group. For each cell image, the confidence level of each category in multiple classifications is obtained. The confidence scores obtained for each of the categories are visualized.

2. The data analysis apparatus for microscopic layered images according to claim 1, wherein, The control unit detects the cell by detecting the cell nucleus from the Z-stacked image stored in the storage unit.

3. The data analysis apparatus for microscopic layered images according to claim 1 or 2, wherein, When a cell is detected crossing multiple Z layers that constitute the Z-stacked image, the control unit obtains the cell image set from the multiple Z layers that the cell is detected crossing.

4. A method for data analysis of a microscopic layered image of a specimen containing many cells, wherein in the data analysis method of the microscopic layered image, Z-stack images of the specimen were obtained using a microscope. The camera unit captures images of the Z-stack obtained by the microscope. The storage unit stores the Z-stacked images captured by the camera unit. The control unit detects cells from the Z-stacked image stored in the storage unit. The control unit obtains a set of cell images corresponding to each detected cell. The control unit extracts cell images from the images constituting the cell image group for each cell. The control unit obtains the confidence level of each category in multiple classifications for the cell images of each cell. The control unit will visualize the confidence level obtained for each category classification.

5. The data analysis method for microscopic layered images according to claim 4, wherein, The control unit detects the cell by detecting the cell nucleus from the Z-stacked image stored in the storage unit.

6. The data analysis method for microscopic layered images according to claim 4 or 5, wherein, When a cell is detected crossing multiple Z layers that constitute the Z-stacked image, the control unit obtains the cell image set from the multiple Z layers that the cell is detected crossing.

Images