A method and system for obtaining cell homogeneity based on image analysis

Abstract

The present invention discloses a method and system for obtaining cell homogeneity based on image analysis, comprising the following steps: cell image segmentation, cell area acquisition, coefficient of variation statistics, and homogeneity scoring. According to the cell line image, the present invention automatically performs image segmentation, cell area acquisition, and coefficient of variation statistical calculation, and scores the homogeneity of the cell line based on the coefficient of variation. The entire process realizes automated analysis, and the quantitative homogeneity score of the cell line image can be obtained without manual intervention. The present invention uses the coefficient of variation of cell area and the coefficient of variation of cell roundness to perform quantitative homogeneity evaluation and analysis on the cell line image, thereby providing an objective, quantifiable, and comparable standard for the homogeneity of the cell line, which is well applicable to different cell lines with circular or non-circular shapes.

Description

Technical Field

[0001] The present invention belongs to the field of biological image processing, and more specifically, relates to a method and system for obtaining cell homogeneity based on image analysis. Background Art

[0002] In life science research, it is often necessary to observe cell populations in vivo and in vitro. One characteristic of cell populations is that their morphologies are both similar and somewhat heterogeneous. Even monoclonal cell lines grown from the same cells still do not have completely uniform morphologies. The homogeneity of cell morphology is an important feature for judging multiple cell indicators, such as stability, the proportion of cells in the cell cycle, expression levels, etc. Flow cytometry (Application of fluorescent-activated cell sorting, FACS) is often used for high-throughput cell homogeneity detection, but its judgment criteria often rely on cell staining results, and many details of cell morphological changes cannot be analyzed. Moreover, when detecting cells, they pass through the detector transiently, and single cells of interest cannot be tracked and further processed.

[0003] Existing microscope image analysis software has relatively single functions and does not have a direct function for judging cell homogeneity. It often relies on the naked eye observation of people to compare and judge the homogeneity between cell lines through experience. When multiple groups of cells need to be judged, the homogeneity of multiple groups of cells needs to be memorized, which is not easy to compare in batches and is prone to errors. For cells with relatively subtle differences in homogeneity, manual judgment is prone to misjudgment and takes a long time. Therefore, using cell image analysis to quantitatively analyze homogeneity has become an indispensable requirement. Summary of the Invention

[0004] In view of the above-mentioned defects or improvement requirements of the prior art, the present invention provides a method and system for obtaining cell homogeneity based on image analysis. The purpose is to simultaneously use the coefficient of variation of cell area and the coefficient of variation of cell roundness to perform quantitative homogeneity evaluation and analysis on cell line images, so as to provide an objective, quantifiable, and comparable standard for the homogeneity of cell lines, and provide a quantitative index for the quality evaluation of cell culture, thereby solving the technical problems of the prior art that rely on visual inspection by the naked eye and comparison by experience to judge the homogeneity between cell lines, cannot be quantitatively compared, and cannot be automatically analyzed.

[0005] To achieve the above object, according to one aspect of the present invention, a method for obtaining cell homogeneity based on image analysis is provided, which is characterized by including the following steps:

[0006] Cell image segmentation: Obtain a cell line image to be processed, perform image segmentation on the cell line image to be processed, and obtain all single-cell images in the cell line image;

[0007] Cell area acquisition: For each single cell image, obtain the edge of the cell within the single cell image through an edge detection algorithm, and determine the cell area and cell perimeter of the single cell image according to the number of pixel points within the edge;

[0008] Coefficient of variation statistics: Take Total single cell images as cell statistical samples, count the cell area and cell perimeter of each cell among them, and calculate the coefficient of variation Cv of the cell areas of the Total single cell images S And the coefficient of variation Cv of cell roundness R , where the coefficient of variation of cell area is the ratio of the standard deviation of cell area to the mean value of cell area, and the coefficient of variation of roundness is the ratio of the standard deviation of cell roundness to cell roundness;

[0009] Homogeneity scoring: According to the coefficient of variation Cv of cell area S And the coefficient of variation Cv of cell roundness R , score the homogeneity of the processed cell line images according to the principle that the smaller the coefficient of variation of cell area or the coefficient of variation of cell roundness, the greater the cell homogeneity, and obtain the homogeneity score p.

[0010] Preferably, the method for obtaining cell homogeneity based on image analysis includes the following steps:

[0011] Roundness evaluation: According to the cell area and cell perimeter of all cells in the cell image, calculate the ratio of the cell contour to the background in the mask image, judge whether the cells are close to circular, sort according to the degree of closeness to a circle, and remove irregular cells whose roundness ranking is lower than a preset threshold to obtain Total single cell images as cell statistical samples.

[0012] Preferably, in the method for obtaining cell homogeneity based on image analysis, the cell image segmentation step combines a semantic segmentation model and contour detection analysis to obtain the minimum bounding rectangle of the cell contour of a single cell as the single cell image of the cell.

[0013] Preferably, in the method for obtaining cell homogeneity based on image analysis, the cell image segmentation step is specifically:

[0014] (1-1) Model semantic segmentation: Take the cell line image to be processed as the program input of the semantic model, segment it, and obtain its mask image;

[0015] (1-2) Contour detection: After binarizing the mask image, perform contour detection, erosion, and filtering processing to obtain the cell contour;

[0016] (1-3) Image traversal: Traverse the cell line image to be processed, and cut out single cell images of the cells with the minimum bounding rectangle of the cell contour as the boundary to obtain all single cell images in the cell line image.

[0017] Preferably, for the method for obtaining cell homogeneity based on image analysis, the circularity R of the cells is calculated as follows: R = (4π * S) / L 2 , where S is the cell area and L is the cell perimeter.

[0018] Preferably, for the method for obtaining cell homogeneity based on image analysis, the homogeneity score p is calculated according to the following method:

[0019] p = kCv S α + Cv R β

[0020] Where k is the cell area coefficient, which is an empirical value; α and β are weight coefficients, and α + β = 1.

[0021] According to another aspect of the present invention, there is provided a system for obtaining cell homogeneity based on image analysis, which includes a cell image segmentation module, a circularity evaluation module, a coefficient of variation statistics module, and a homogeneity scoring module;

[0022] The cell image segmentation module is used to obtain the cell line image to be processed, perform image segmentation on the cell line image to be processed, and obtain all single cell images in the cell line image;

[0023] The cell area acquisition module is used for each single cell image, to obtain the edge of the cell in the single cell image through an edge detection algorithm and determine the cell area and cell perimeter of the single cell image according to the number of pixel points within the edge;

[0024] The coefficient of variation statistics module is used to take Total single cell images as cell statistical samples, count the cell area and cell perimeter of each cell among them, and calculate the coefficient of variation Cv of the cell areas of the Total single cell images S And the coefficient of variation Cv of the cell circularity R , where the coefficient of variation of the cell area is the ratio of the cell area standard deviation to the cell area mean value, and the coefficient of variation of the circularity is the ratio of the cell circularity standard deviation to the cell circularity ratio;

[0025] The homogeneity scoring: is used to calculate according to the coefficient of variation Cv of the cell area S And the coefficient of variation Cv of the cell circularity R, according to the principle that the smaller the coefficient of variation of cell area or the coefficient of variation of cell roundness, the greater the cell homogeneity, perform homogeneity scoring on the processed cell line images to obtain the homogeneity score p.

[0026] Preferably, the cell homogeneity acquisition system based on image analysis includes a roundness evaluation module. The roundness evaluation module is used to calculate the ratio of the cell contour to the background in the mask image according to the cell area and the perimeter of all cells in the cell image, determine whether the cells are close to circular, sort according to the degree of closeness to a circle, remove irregular cells with a roundness ranking lower than a preset threshold, and obtain Total single cell images as cell statistical samples.

[0027] According to another aspect of the present invention, there is provided an electronic device, including a memory, a processor, and a computer program stored on the memory and executable on the processor, wherein when the processor executes the program, the steps of the cell homogeneity acquisition method based on image analysis provided by the present invention are implemented.

[0028] According to another aspect of the present invention, there is provided a non-transitory computer-readable storage medium, on which a computer program is stored, wherein when the computer program is executed by a processor, the steps of the cell homogeneity acquisition method based on image analysis provided by the present invention are implemented.

[0029] Generally speaking, compared with the prior art through the above technical solutions conceived by the present invention, the following beneficial effects can be achieved:

[0030] Based on the cell line images, the present invention automatically performs image segmentation, cell area acquisition, coefficient of variation statistical calculation, and scores the homogeneity of the cell line according to the coefficient of variation. The whole process realizes automatic analysis and can obtain the quantitative homogeneity score of the cell line image without manual intervention.

[0031] The present invention uses the coefficient of variation of cell area and the coefficient of variation of cell roundness to perform quantitative homogeneity evaluation analysis on cell line images, thereby providing an objective, quantifiable, and comparable standard for the homogeneity of cell lines, which is well applicable to different cell lines with circular or non-circular shapes. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] Figure 1 is a schematic flowchart of the cell homogeneity acquisition method based on image analysis provided by an embodiment of the present invention;

[0033] Figure 2A is a micrograph of cell line A, Figure 2B is a micrograph of cell line B, Figure 2C is a micrograph of cell line C;

[0034] Figure 3A It is the masked image of cell line A, Figure 3B It is the masked image of cell line B, Figure 3C It is the masked image of cell line C;

[0035] Figure 4 It is the homogeneity score p of cell lines A, B, and C calculated in the embodiments of the present invention. Detailed implementation manners

[0036] In order to make the objectives, technical solutions and advantages of the present invention clearer, the present invention will be further described in detail below in conjunction with embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention. In addition, the technical features involved in the various embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0037] The method for obtaining cell homogeneity based on image analysis provided by the present invention includes the following steps:

[0038] (1) Cell image segmentation: Obtain the cell line image to be processed, perform image segmentation on the cell line image to be processed, and obtain all single-cell images in the cell line image; preferably, in combination with a semantic segmentation model and contour detection analysis, obtain the smallest upright rectangle of the cell contour of a single cell as the single-cell image of the cell; the specific method is as follows:

[0039] (1-1) Model semantic segmentation: Use the cell line image to be processed as the program input of the semantic model, perform segmentation on it, and obtain its masked image; the semantic model can be selected as, for example, the U-net model;

[0040] (1-2) Contour detection: After binarizing the masked image, perform contour detection, erosion and filtering processing to obtain the cell contour;

[0041] (1-3) Image traversal: Traverse the cell line image to be processed, and cut out the single-cell image of the cell with the smallest upright rectangle of the cell contour as the boundary to obtain all single-cell images in the cell line image;

[0042] (2) Cell area acquisition: For each of the single-cell images obtained in step (1), obtain the edge of the cell in the single-cell image through an edge detection algorithm, and determine the cell area and cell perimeter of the single-cell image according to the number of pixel points within the edge;

[0043] (3) Roundness evaluation: Based on the cell areas and perimeters of all cells in the cell images obtained in step (2), calculate the ratio of the cell contours to the background in the mask image to determine whether the cells are close to circular. Especially for cells with a circular shape, sort them according to their proximity to a circle, remove irregular cells with a roundness ranking lower than the preset threshold, and obtain Total individual cell images as cell statistical samples. The roundness of cells further reflects the uniformity of cells. The rounder the cells, the smaller the area difference, indicating better cell uniformity. However, cells are sensitive to roundness. Some cells with poor roundness may significantly affect the overall evaluation of the coefficient of variation of roundness. However, the overall cells in the image show a relatively uniform performance. Therefore, to exclude the influence of a small number of cells with too low roundness on the overall roundness evaluation, only cells with a roundness ranking exceeding a fixed proportion limit are used as objects for uniformity evaluation. This will not have an obvious impact even on cell lines that are not circular in shape, and the scalability between cell lines is good.

[0044] The roundness R of cells is calculated as follows: R = (4π * S) / L 2 , where S is the cell area and L is the cell perimeter.

[0045] (4) Coefficient of variation statistics: Based on the cell statistical samples composed of the Total individual cell images obtained in step (3), count the cell areas and perimeters of each cell, and calculate the coefficient of variation Cv of the cell areas of the Total individual cell images S and the coefficient of variation Cv of cell roundness R , where the coefficient of variation of cell area is the ratio of the standard deviation of cell area to the mean value of cell area, and the coefficient of variation of roundness is the ratio of the standard deviation of cell roundness to the ratio of cell roundness;

[0046] (5) Uniformity scoring: Based on the coefficient of variation Cv of cell area obtained in step (4) S and the coefficient of variation Cv of cell roundness R , score the uniformity of the processed cell line images according to the principle that the smaller the coefficient of variation of cell area or the coefficient of variation of cell roundness, the greater the cell uniformity, and obtain the uniformity score p; In the preferred scheme, the uniformity score p is calculated according to the following method:

[0047] p = kCv S α + Cv R β

[0048] where k is the cell area coefficient, which is an empirical value; α and β are weight coefficients, and α + β = 1.

[0049] Generally speaking, the most commonly used index to judge homogeneity is the standard deviation. However, cell types vary widely in size and shape, and using the standard deviation to judge will result in extremely poor generality and can only score and compare for each cell line. The present invention uses the dimensionless coefficient of variation to evaluate the degree of dispersion of data, adapts to the differences between different cell lines, and can more accurately and objectively evaluate the homogeneity of cells.

[0050] In addition to the size of the cells, we believe that the shape of the cells is also an important factor affecting cell homogeneity. The size of the cells is related to the overall development state of the cells, and the shape of the cells is closely related to the state and function of the cells. Cells assume a specific shape to maintain a specific function. The present invention uses roundness to evaluate the shape of cells and uses the coefficient of variation of roundness to characterize the stability of cell shape.

[0051] For the overall homogeneity of cells, both the size and shape of the cells are considered simultaneously. Therefore, the coefficient of variation of cell area and the coefficient of variation of cell roundness are used to jointly evaluate the homogeneity of cells according to their respective weights. According to the principle that the smaller the coefficient of variation of cell area or the coefficient of variation of cell roundness, the greater the cell homogeneity, the homogeneity score is given to the processed cell line images, and the final homogeneity score is obtained. This homogeneity score can objectively reflect the degree of homogeneity of the developmental state and functional state of cells, thus providing a quantifiable basis for the evaluation of cell morphology and culture effect.

[0052] The cell homogeneity acquisition system based on image analysis provided by the present invention includes:

[0053] A cell image segmentation module, a cell area acquisition module, a roundness evaluation module, a coefficient of variation statistics module, and a homogeneity scoring module;

[0054] The cell image segmentation module is used to obtain the cell line image to be processed, perform image segmentation on the cell line image to be processed, and obtain all single-cell images in the cell line image;

[0055] The cell area acquisition module is used for each single-cell image, to obtain the edge of the cell within the single-cell image through an edge detection algorithm and determine the cell area and cell perimeter of the single-cell image according to the number of pixel points within the edge;

[0056] The roundness evaluation module is used to calculate the ratio of the cell contour to the background in the mask image according to the cell area and cell perimeter of all cells in the cell image, judge whether the cells are close to circular, sort according to the degree of closeness to a circle, remove irregular cells whose roundness ranking is lower than a preset threshold, and obtain Total single-cell images as cell statistical samples;

[0057] The coefficient of variation statistical module is used to take Total single-cell images as cell statistical samples, count the cell area and cell perimeter of each cell among them, and calculate the coefficient of variation Cv of the cell areas of the Total single-cell images S and the coefficient of variation Cv of cell roundness R , where the coefficient of variation of cell area is the ratio of the standard deviation of cell area to the mean value of cell area, and the coefficient of variation of roundness is the ratio of the standard deviation of cell roundness to the ratio of cell roundness;

[0058] The homogeneity score: is used to perform a homogeneity score on the processed cell line images according to the coefficient of variation Cv of cell area S and the coefficient of variation Cv of cell roundness R , and obtain a homogeneity score p according to the principle that the smaller the coefficient of variation of cell area or the coefficient of variation of cell roundness, the greater the cell homogeneity.

[0059] The following are examples:

[0060] A method for obtaining cell homogeneity based on image analysis includes the following steps:

[0061] (1) Cell image segmentation: The photos of cell lines A, B, and C taken are placed in folder (A), as Figures 2A to 2C shown, and the trained cell line semantic segmentation model: U-net model B is prepared;

[0062] (1-1) Semantic segmentation: Taking the photos in folder (A) and U-net model B as inputs, running the semantic segmentation program to generate a mask image, and the mask images of cell lines A, B, and C are respectively as Figures 3A to 3C shown.

[0063] (1-3) After binarizing the obtained mask image, perform contour detection, erosion, and filtering processing to obtain cell contours and record the number of cell contours;

[0064] (2) Cell area acquisition: According to the number of cell contours obtained in step (1), then traverse each cell, obtain the edge of each cell through the edge detection algorithm, and obtain the area S and perimeter L of a single-cell image according to the number of pixel points within the edge.

[0065] (3) Roundness evaluation: Calculate the roundness R of each cell, and the calculation method is as follows: R = (4Π * S) / L 2 , where S is the cell area and L is the cell perimeter. Sort according to the degree of proximity to a circle, and remove the irregular cells ranked in the last 5% of the roundness ranking to obtain Total cell samples.

[0066] (4) Coefficient of variation statistics: Based on the area and perimeter of the single-cell images obtained in step (4), calculate the coefficient of variation of the cell area and the coefficient of variation of cell roundness of the cells. The calculation formulas are as follows:

[0067]

[0068]

[0069] Among them, Cv S is the coefficient of variation of the cell area, σ S is the standard deviation of the cell areas of Total cell samples, is the average value of the cell areas of Total cell samples, Cv R is the coefficient of variation of the cell area, σ R is the standard deviation of the cell areas of Total cell samples, is the average value of the cell areas of Total cell samples.

[0070] (5) Homogeneity score: Based on the coefficient of variation Cv S of the cell area and the coefficient of variation Cv R of cell roundness obtained in step (4), calculate the homogeneity score p of the cells. The calculation method is as follows:

[0071] p = kCv S α + Cv R β

[0072] Among them, k is the cell area coefficient, which is an empirical value; α and β are weight coefficients, and α + β = 1. In this embodiment, k ranges from 0.2 to 1, and k is selected as 0.7 in this embodiment; α ranges from 0.1 to 0.9, and α is selected as 0.7 in this embodiment; β ranges from 0.1 to 0.9, and β is selected as 0.3 in this embodiment. Calculate the homogeneity score p for cell lines A, B, and C as Figure 4 shown. It can be seen that the homogeneity score p can objectively reflect the homogeneity degree of the cells and is suitable as a homogeneity quantification standard for automated calculation to perform high-throughput cell homogeneity detection.

[0073] It is easy for those skilled in the art to understand that the above are only preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent replacements, and improvements made within the spirit and principles of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for obtaining cell homogeneity based on image analysis, characterized in that, it includes the following steps: Cell image segmentation: Obtain the cell line image to be processed, perform image segmentation on the cell line image to be processed, and obtain all single-cell images in the cell line image; Cell area acquisition: For each of the single-cell images, obtain the edge of the cell in the single-cell image through an edge detection algorithm and determine the cell area and cell perimeter of the single-cell image according to the number of pixel points within the edge; Coefficient of variation statistics: Take Total single-cell images as cell statistical samples, count the cell area and cell perimeter of each cell among them, and calculate the coefficient of variation Cv of the cell areas of the Total single-cell images S and the coefficient of variation Cv of cell roundness R , where the coefficient of variation of cell area is the ratio of the standard deviation of cell area to the mean value of cell area, and the coefficient of variation of roundness is the ratio of the standard deviation of cell roundness to the ratio of cell roundness; Homogeneity score: Based on the coefficient of variation (Cv) of cell area S and the coefficient of variation (Cv) of cell roundness R , the homogeneity of the processed cell line images is scored according to the principle that the smaller the coefficient of variation of cell area or the coefficient of variation of cell roundness, the greater the cell homogeneity, and the homogeneity score p is obtained; The homogeneity score p is calculated according to the following method: p = kCv S α + Cv R β where k is the cell area coefficient, which is an empirical value; α and β are weight coefficients, and α + β = 1.

2. The method for obtaining cell homogeneity based on image analysis according to claim 1, characterized in that, it includes the following steps: Roundness evaluation: According to the cell area and cell perimeter of all cells in the cell image, calculate the ratio of the cell contour to the background of the mask image, judge whether the cells are close to circular, sort according to the degree of closeness to a circle, and remove irregular cells whose roundness ranking is lower than a preset threshold to obtain Total single-cell images as cell statistical samples.

3. The method for obtaining cell homogeneity based on image analysis according to claim 1 or 2, characterized in that, in the cell image segmentation step, in combination with a semantic segmentation model and contour detection analysis, obtain the minimum bounding rectangle of the cell contour of a single cell as the single-cell image of the cell.

4. The method for obtaining cell homogeneity based on image analysis according to claim 1 or 2, characterized in that, the cell image segmentation step is specifically: (1-1) Model semantic segmentation: Take the cell line image to be processed as the program input of the semantic model, perform semantic segmentation on it, and obtain its mask image; (1-2) Contour detection: After binarizing the mask image, perform contour detection, erosion, and filtering processing to obtain the cell contour; (1-3) Image traversal: Traverse the cell line image to be processed, and cut with the minimum bounding rectangle of the cell contour as the boundary to obtain the single-cell image of the cell, and obtain all single-cell images in the cell line image.

5. The method for obtaining cell homogeneity based on image analysis according to claim 1 or 2, characterized in that, The circularity R of the cell is calculated as follows: R = (4π * S) / L 2 , where S is the cell area and L is the cell perimeter.

6. A system for obtaining cell homogeneity based on image analysis, characterized in that, it includes a cell image segmentation module, a cell area acquisition module, a coefficient of variation statistics module, and a homogeneity scoring module; The cell image segmentation module is used to obtain the cell line image to be processed, perform image segmentation on the cell line image to be processed, and obtain all single-cell images in the cell line image; The cell area acquisition module is used for each of the single-cell images, to obtain the edge of the cell in the single-cell image through an edge detection algorithm and determine the cell area and cell perimeter of the single-cell image according to the number of pixel points within the edge; The coefficient of variation statistical module is used to take Total single-cell images as cell statistical samples, count the cell area and perimeter of each cell among them, and calculate the coefficient of variation Cv of the cell areas of the Total single-cell images S and the coefficient of variation Cv of cell roundness R , where the coefficient of variation of cell area is the ratio of the standard deviation of cell area to the mean value of cell area, and the coefficient of variation of roundness is the ratio of the standard deviation of cell roundness to the roundness of the cell; The uniformity scoring module is used to calculate the coefficient of variation Cv of cell area S and the coefficient of variation Cv of cell roundness R , and score the uniformity of the processed cell line images according to the principle that the smaller the coefficient of variation of cell area or the coefficient of variation of cell roundness, the greater the cell uniformity, so as to obtain the uniformity score p The homogeneity score p is calculated according to the following method: p = kCv S α + Cv R β where k is the cell area coefficient, which is an empirical value; α and β are weight coefficients, and α + β = 1.

7. The cell homogeneity acquisition system based on image analysis according to claim 6, wherein, it further includes a roundness evaluation module, which is used to calculate the ratio of the cell contour to the background in the mask image according to the cell area and the perimeter of all cells in the cell image, determine whether the cells are close to a circular shape, sort them according to the degree of closeness to the circular shape, and remove irregular cells whose roundness ranking is lower than a preset threshold, so as to obtain Total single-cell images as cell statistical samples.

8. An electronic device, comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, wherein, when the processor executes the program, it implements the steps of the method for acquiring cell homogeneity based on image analysis according to any one of claims 1 to 5.

9. A non-transitory computer-readable storage medium, on which a computer program is stored, wherein, when the computer program is executed by a processor, it implements the steps of the method for acquiring cell homogeneity based on image analysis according to any one of claims 1 to 5.

Images