Cell growth state detection method and system based on image processing

Abstract

The present application relates to the technical field of cell image processing, and particularly relates to a cell growth state detection method and system based on image processing. The method comprises: obtaining a suspected cell region degree of each pixel point in a gray image of a cell image; clustering the pixel points to obtain an enlarged boundary pixel point of a class cluster, and then obtaining a confidence degree of each class cluster pixel point; multiplying the suspected cell region degree and the confidence degree and normalizing to obtain an importance degree; analyzing the gray degree of the pixel points in the eight-neighborhood of a pixel point to obtain a gray degree variation degree of the pixel point; obtaining neighboring pixel points of the pixel point, and analyzing to obtain a continuity degree of the pixel point; respectively optimizing an enhancement coefficient of each pixel point according to the importance degree, the gray degree variation degree and the continuity degree of the pixel point, and then performing image enhancement; and using the enhanced cell image to detect the cell growth state. The present application can improve the accuracy of cell growth state detection.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of cell image processing, in particular to a cell growth state detection method and system based on image processing. BACKGROUND

[0002] In the process of cell culture, accurate detection of the growth state of cells is crucial for scientific research and clinical applications. The traditional method mainly relies on manual microscope observation and manual sampling combined with flow cytometry. Manual microscope observation relies on the experience and judgment of the observer, which is highly subjective and prone to errors. Long-time observation can lead to fatigue, further affecting the accuracy of the results. Manual sampling for counting may interfere with the cell culture environment, increase the risk of contamination, and is tedious and inefficient. With the development of computer vision technology, image processing technology can be used to detect the growth state of cells. The acquired cell images are enhanced to retain the details of the cell area. The morphology, texture and other features of a large number of cells in different growth states are obtained and trained. Machine learning is used to detect the growth state of cells. During cell culture, as the cells grow, cell stacking may occur in the culture dish due to the gradual increase in cells. Due to the depth of field of the high-power microscope, when the cells are not on the same focal plane, some structures will be blurred due to defocusing. Therefore, image enhancement processing is needed when analyzing and detecting the growth state of cell growth images.

[0003] Traditional image enhancement methods (such as linear enhancement and histogram equalization) often adjust the gray scale range of the entire image, which may result in over-enhancement of low-frequency information (such as background) and submersion of high-frequency details (such as cell edges). The Laplace operator mainly acts on high-frequency components and has little effect on low-frequency components. Therefore, it can enhance cell edges and details while preserving background information, avoiding interference from background noise. Therefore, the present application uses laplace transformation to enhance the image. After obtaining the calculation results of all pixel points in the image using the laplace operator, the laplace image of the original image is obtained. The Laplace image and the original image are superimposed with a set enhancement coefficient to generate a sharpened image. Traditional methods generally use a fixed enhancement coefficient to superimpose with the original image, which may result in enhancement of unnecessary interference parts (such as impurities in the culture dish), thereby interfering with the subsequent detection of cell growth conditions and affecting the judgment of cell growth state. SUMMARY

[0004] To solve the above technical problems, the technical solution adopted by the present application is as follows:

[0005] In a first aspect, one embodiment of the present application provides a cell growth state detection method based on image processing, the method comprising:

[0006] The acquired cell image is preprocessed to obtain a gray image; a window of different scales is established with a pixel point as the center, and the suspected cell region degree of the pixel point is obtained according to the gray value change in each window;

[0007] The pixel points are clustered to obtain class clusters; the boundaries of the class clusters are expanded to obtain expanded boundary pixel points; the confidence of the pixel points of a class cluster is obtained according to the pixel points at the boundaries of the adjacent class clusters and the expanded boundary pixel points; and the importance degree is obtained by multiplying and normalizing the suspected cell region degree and the confidence;

[0008] The absolute value of the difference between the average gray value of the pixel points in the eight-neighborhood of a pixel point and the gray value of the pixel point is normalized to obtain the gray value change degree of the pixel point;

[0009] The pixel point with the smallest gray difference in the eight-neighborhood of a pixel point is taken as the adjacent pixel point, and the continuity degree of the pixel point is obtained based on the difference between the gray value change degrees of the pixel point and the adjacent pixel point;

[0010] The enhancement coefficients of the pixel points are optimized according to the importance degree, the gray value change degree and the continuity degree of each pixel point, respectively, to obtain the final enhancement coefficients and enhance the cell image; and the cell growth state is detected by using the enhanced cell image.

[0011] Preferably, the suspected cell region degree of a pixel point is obtained according to the gray value change in each window, comprising:

[0012] The product of the standard deviation and the entropy value of the gray value of the pixel points in each window corresponding to a pixel point is obtained and averaged to obtain the suspected cell region degree of the pixel point.

[0013] Preferably, the pixel points are clustered to obtain class clusters, comprising:

[0014] The suspected cell region degree of a pixel point and the coordinates of the pixel point are combined to form a feature vector of the pixel point, and the pixel points are clustered according to the difference between the feature vectors of each two pixel points to obtain class clusters.

[0015] Preferably, the boundaries of the class clusters are expanded to obtain expanded boundary pixel points, comprising:

[0016] The original boundaries of the class clusters are expanded outward by a pixel point as the expansion standard step, the expanded boundaries of the class clusters are obtained, and the pixel points in the expanded boundaries are the expanded boundary pixel points.

[0017] Preferably, the confidence of the pixel points of a cluster is obtained according to the pixel points of adjacent clusters, the pixel points at the boundary of the cluster and the pixel points of the enlarged boundary, comprising:

[0018] The distance between the center of a cluster and the center of other clusters is obtained, denoted as a first distance; the preset number of clusters with the minimum first distance in the other clusters are adjacent clusters of the cluster;

[0019] The absolute value of the difference between the number of pixel points between a cluster and its corresponding adjacent cluster and the reciprocal of the product of the first distance is taken as the inter-cluster feature corresponding to the adjacent cluster; the absolute value of the difference between the average gray value of the pixel points at the original boundary of a cluster and the average gray value of the pixel points of the enlarged boundary corresponding to the cluster is taken, and the reciprocal of the standard deviation of the pixel points of the enlarged boundary is multiplied to obtain the boundary feature of the cluster; the inter-cluster features corresponding to each adjacent cluster of a cluster are summed and multiplied by the boundary feature to obtain the confidence of the pixel points of the cluster.

[0020] Preferably, the continuity of a pixel point is obtained based on the difference in the degree of gray change of the pixel point and its adjacent pixel points, comprising:

[0021] The continuity of a pixel point is obtained by normalizing the absolute value of the difference in the degree of gray change of the pixel point and its adjacent pixel points.

[0022] Preferably, the enhancement coefficient of each pixel point is optimized according to the importance, the degree of gray change and the continuity of each pixel point, comprising:

[0023] A gray change degree threshold is set, and when the degree of gray change of a pixel point is greater than the gray change degree threshold, the pixel point is marked as a high gray change pixel point, and vice versa;

[0024] For a high gray change pixel point, the reduction step of the high gray change pixel point is obtained by multiplying the enhancement coefficient optimization sensitivity coefficient and the degree of gray change of the high gray change pixel point, and the difference between the initial enhancement coefficient and the reduction step is the final enhancement coefficient of the high gray change pixel point;

[0025] For a low gray change pixel point, the multiplication of the enhancement coefficient optimization sensitivity coefficient and the importance, the continuity and the degree of gray change of the low gray change pixel point is the increase step, and the sum of the initial enhancement coefficient and the increase step is the final enhancement coefficient of the low gray change pixel point.

[0026] In a second aspect, the present application also provides a cell growth state detection system based on image processing, comprising a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the computer program implements the steps of the cell growth state detection method based on image processing when executed by the processor.

[0027] The embodiment of the present application has at least the following beneficial effects: the present application obtains a gray image of a cell image, establishes a window of different scales with a pixel point as the center, analyzes the change of the gray value in different windows to obtain the suspected cell region degree of the pixel point, and uses the window of different scales to analyze, which can avoid the problem that a too small window cannot completely capture the uneven gray feature and a too large window contains too many features to cover the features of the target pixel point itself, so that the analysis is more accurate; further, the pixel points are clustered to obtain an enlarged boundary pixel point of a cluster, the boundary feature and the feature between the local adjacent clusters are analyzed, the confidence of the cluster pixel point is obtained, and then the suspected cell region degree and the confidence are combined to improve the accuracy of the importance analysis; then, the gray change degree and the continuity degree of the pixel points are obtained, the enhancement coefficients of the pixel points are optimized in combination with the importance of the pixel points, the final enhancement coefficients of the pixel points are obtained, the laplace operator is used for image enhancement, and when the laplace image is superimposed on the original image, the cell region edge and the details are properly enhanced (the weak edge is further enhanced, the details are highlighted; the strong edge is properly enhanced, and the artifacts are eliminated), the cell region edge and the texture details in the cell image are more prominent, and thus the cell growth state can be more intuitively analyzed and detected. BRIEF DESCRIPTION OF DRAWINGS

[0028] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, and the advantages thereof, the drawings needed to be used in the embodiments or the prior art description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and for those skilled in the art, other drawings can be obtained from these drawings without creative labor.

[0029] Figure 1 A method flow chart of a cell growth state detection method based on image processing provided by the embodiment of the present application is shown in the figure.

[0030] Figure 2 An edge expansion schematic diagram of a cell growth state detection method based on image processing provided by the embodiment of the present application is shown in the figure. DETAILED DESCRIPTION

[0031] In order to further illustrate the technical means and effects taken by the present application to achieve the predetermined object of the application, the specific implementation, structure, features and effects of the image processing-based cell growth state detection method and system according to the present application are described in detail as follows in combination with the accompanying drawings and preferred embodiments. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. In addition, the specific features, structures or characteristics in one or more embodiments can be combined in any suitable form.

[0032] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which the present application belongs.

[0033] The specific scheme of the image processing-based cell growth state detection method provided by the present application is described in detail below in combination with the accompanying drawings.

[0034] Embodiment 1

[0035] The main application scenario of the present application is: for cell images, the laplace operator is used to enhance them, improve the quality of the images, and make the cell regions, edges and texture details in the cell images more prominent, so that the cell growth state can be more intuitively analyzed and detected.

[0036] Please refer to Figure 1 which shows a method flowchart of the image processing-based cell growth state detection method provided by one embodiment of the present application, and the method comprises the following steps:

[0037] Step S1: pre-processing the obtained cell images to obtain gray-scale images; establishing windows of different scales with a pixel point as the center, and obtaining the suspected cell region degree of the pixel point according to the gray value change in each window.

[0038] A high-definition camera is installed at a suitable position of the cell preparation device to capture cell growth state images in the microscope at each stage, that is, according to the cell growth cycle, the cells in the culture device are photographed every certain period of time to obtain cell images.

[0039] For the obtained cell images, first, the gray-scale processing is performed, and then in order to make the subsequent laplace transformation enhancement effect better, the high-frequency noise is eliminated by Gaussian filtering to obtain the gray-scale image of the cell image, and the next step of processing is performed.

[0040] Because different organelles such as mitochondria, endoplasmic reticulum, Golgi body, lysosome are distributed in different regions of the cell, the density and composition of these organelles are different from the cytoplasm matrix, which leads to the difference of electron density or optical density in local regions, and thus the uneven gray scale in microscopic imaging. Therefore, the degree of the suspected cell region of each pixel point is obtained according to the local gray scale distribution uniformity.

[0041] After obtaining the gray scale image, because the sizes of the organelles in the cell are different, a too small window cannot completely capture the gray scale unevenness feature, and a too large window contains too many features which can cover the features of the target pixel point itself. Therefore, a window of different scales is established with any pixel point in the image as the center, three scales of the first scale window, the second scale window and the third scale window are established in the present application, the scale of the window is n*n, n=3, 5, 7, and the implementer can determine the scale of the window according to the actual situation. The gray scale values of the pixel points in the window are obtained, and then the corresponding gray scale distribution uniformity is calculated as the degree of the suspected cell region of each pixel point.

[0042] Further, the degree of the suspected cell region of the pixel point is obtained according to the gray scale value change in each window. Specifically, the product of the standard deviation and the entropy value of the gray scale value of the pixel points in each window corresponding to one pixel point is obtained and the average value is calculated to obtain the degree of the suspected cell region of the pixel point.

[0043] The specific calculation model is as follows:

[0044]

[0045] wherein Yi represents the degree of the suspected cell region of the i th pixel point, is the standard deviation of the gray scale value in the k th window of the i th pixel point, that is, the fluctuation degree. The greater the fluctuation degree of the gray scale value, the greater the gray scale fluctuation in the window, that is, the poorer the local uniformity, which is more consistent with the gray scale distribution characteristics in the cell region. is the entropy value of the gray scale value of the pixel points in the k th window of the i th pixel point. The greater the gray scale entropy, the more complex the gray scale information contained in the window, that is, the poorer the local gray scale consistency, which is more consistent with the poor gray scale distribution uniformity characteristics in the cell region. represents the number of windows of different scales corresponding to one pixel point, which is 3 in the present embodiment, represents that the average value of the gray scale distribution uniformity of the windows of different scales is taken as the degree of the suspected cell region Yi of the i th pixel point.

[0046] In step S2, the pixel points are clustered to obtain class clusters, the boundaries of the class clusters are expanded to obtain expanded boundary pixel points, the confidence of the pixel points of one class cluster is obtained according to the pixel points at the boundaries of the adjacent class clusters and the expanded boundary pixel points, and the importance degree is obtained by multiplying the degree of the suspected cell region and the confidence and normalizing.​

[0047] Step S1 obtains the degree of suspected cell region of each pixel point. In the growth process of cells, due to various reasons, there will be precipitates in the nutrient solution, and the gray scale of the precipitate region will also be uneven. If it is excessively enhanced, it will interfere with the detection of cell texture, so the importance of each pixel point belonging to the cell region needs to be obtained, and the enhancement coefficient of image enhancement using the laplace operator needs to be adjusted.

[0048] Because cells may naturally fall off debris during growth, or drugs (such as chemotherapy drugs) or fixatives (such as paraformaldehyde) added in experiments may form precipitates, so there may be impurities in the gray scale image of the cell image in addition to cells; the characteristics of the impurity region are similar to those in the cell, both of which show local gray scale unevenness, so the confidence of each pixel point belonging to the cell region needs to be obtained by combining the local distribution characteristics of each pixel point, and the degree of suspected cell region of each pixel point and the coordinates are used as feature vectors. According to the difference between the feature vectors of each two pixel points, the pixel points are clustered to obtain a plurality of clusters; due to the existence of cell membranes or cell walls, each cell has obvious boundaries, that is, clustered distribution, and the size of the cluster is relatively uniform; and the impurity region does not exist due to constraints, so the size of the cluster where it is located is not uniform, and the shape is different, and then the importance of each pixel point is obtained by combining the corresponding degree of suspected cell region.

[0049] For the obtained clusters, because the characteristics of the cell region are similar to those of some interference regions, the confidence of the degree of suspected cell region of each pixel point in the cluster needs to be obtained by combining the local distribution characteristics of each pixel point in the cluster.

[0050] First, because the cell boundaries separate each cell, the gray scale difference between the cell boundary part and the pixel points in the cell region is large. Taking a pixel point as an expansion standard step, the original boundary of each cluster is expanded outward to obtain the expanded boundary of each cluster, which is used as the cell membrane (wall), and the pixel points in the expanded boundary are expanded boundary pixel points; as shown in Figure 2 The green region in the figure is the expanded boundary, and the dark yellow part and the light yellow part are a cluster, and the light yellow part is the original boundary of the cluster.

[0051] Further, the cells are bound by cell membranes, and the same type of cells are generally cultured in the culture dish, so the cells are generally distributed in clusters (the distance between each cell is similar), and the area is similar (the number of pixel points in the cluster is similar), so the adjacent clusters of each cluster need to be obtained. Specifically, the distance between the center of a cluster and the center of other clusters is obtained, denoted as the first distance; the preset number of clusters with the smallest first distance in other clusters are the adjacent clusters of the cluster, wherein the preset number is 5, and the implementer adjusts according to the actual situation. The difference between the area and the distance between each adjacent cluster and the corresponding cluster is smaller, and the cluster is more consistent with the cell region morphology and distribution characteristics.

[0052] Finally, the confidence of the pixel points of a cluster is obtained according to the adjacent clusters of the cluster, the pixel points at the boundary, and the expanded boundary pixel points. Specifically, the reciprocal of the product of the absolute value of the difference between the number of pixel points of a cluster and the corresponding adjacent cluster and the first distance is obtained as the inter-cluster feature corresponding to the adjacent cluster; the absolute value of the difference between the mean value of the gray values of the pixel points at the original boundary of a cluster and the mean value of the gray values of the expanded boundary pixel points of the cluster is obtained, and the reciprocal of the standard deviation of the expanded boundary pixel points is multiplied to obtain the boundary feature of the cluster; the inter-cluster features corresponding to each adjacent cluster of a cluster are summed and multiplied by the boundary feature to obtain the confidence of the pixel points of the cluster.

[0053] The specific calculation model is: ,

[0054] Wherein, Zi is the confidence of the ith pixel point, that is, the confidence of the pixel points of the cluster in which the ith pixel point is located is taken as the confidence of the ith pixel point; si is the number of pixel points of the cluster in which the ith pixel point is located, and sj represents the number of pixel points of the jth adjacent cluster of the cluster in which the ith pixel point is located. represents the first distance between the cluster in which the ith pixel point is located and the jth adjacent cluster of the cluster in which the ith pixel point is located (the distance between the centers of the two clusters). represents the inter-cluster feature corresponding to the jth adjacent cluster of the cluster in which the ith pixel point is located, the more similar the area (pixel point data) is, the smaller the first distance is, that is, the more densely distributed, the more consistent with the cell distribution characteristics of the cell region, and the more likely the cluster in which the ith pixel point is located is a cell region.

[0055] and respectively represent the mean value of the gray values of the expanded boundary pixel points corresponding to the cluster in which the ith pixel point is located and the mean value of the gray values of the pixel points at the original boundary of the cluster in which the ith pixel point is located, the standard deviation of the gray values of the expanded boundary pixel points, which reflects the consistency of the gray values of the region, The boundary feature is represented, and the product of the gray scale difference value of the original boundary and the enlarged boundary pixel points and the gray scale consistency of the enlarged boundary is represented. The greater the gray scale difference value and the higher the gray scale consistency, the more the boundary part feature of the cell region is consistent, and the more likely the class cluster of the i-th pixel point is the cell region.

[0056] After obtaining the confidence of each pixel point, the importance degree corresponding to each pixel point can be obtained by combining the suspected cell region degree. The importance degree is obtained by multiplying the suspected cell region degree and the confidence and normalizing. The specific calculation model is: Ci is the importance degree of the i-th pixel point; Yi is the suspected cell region degree of the i-th pixel point; and Zi is the confidence of the i-th pixel point. Thus, the importance degree of each pixel point can be obtained.

[0057] In step S3, the gray scale change degree of the pixel point is obtained by normalizing the absolute value of the difference between the gray scale mean value of the pixel points in the eight-neighborhood of the pixel point and the gray scale value of the pixel point.

[0058] After obtaining the importance degree of each pixel point, since the edge degrees of each pixel point are different, the corresponding gray scale change degrees are also different, and the laplace operator responds more strongly to the regions with obvious gray scale change in the image. Therefore, when optimizing the enhancement coefficient of each pixel point, the local gray scale change, continuity degree and importance degree of the pixel point need to be combined; the weak edge is further enhanced to highlight the details; and the strong edge is appropriately enhanced to eliminate the artifacts.

[0059] Since the laplace operator acts on the eight-neighborhood of each pixel point, when analyzing the local gray scale change of each pixel point, the pixel points in the eight-neighborhood of the pixel point are analyzed. The gray scale change degree of the pixel point is obtained by normalizing the absolute value of the difference between the gray scale mean value of the pixel points in the eight-neighborhood of the pixel point and the gray scale value of the pixel point. The specific calculation model is:

[0060] ,

[0061] Wherein, Bi is the gray scale change degree of the i-th pixel point, Hj is the gray scale value of the j-th neighborhood pixel point in the eight-neighborhood of the i-th pixel point, and norm represents normalization. The difference between the i-th pixel point and the gray scale mean value of the neighborhood pixel points is represented. When the local gray scale change degree of the i-th pixel point is greater, the gray scale difference value between the i-th pixel point and the neighborhood pixel points is greater, that is, the gray scale change degree is higher, the local laplace operator response degree is stronger, and it is more likely to be over-enhanced. Thus, the continuity degree of each pixel point can be obtained.

[0062] Step S4, a pixel point in the eight-neighborhood of the pixel point with the smallest gray level difference is taken as the adjacent pixel point, and the continuity of the pixel point is obtained based on the difference in the gray level variation degree of the pixel point and the adjacent pixel point.

[0063] After the gray level variation degree of each pixel point is obtained in step S3, the continuity of the pixel point needs to be analyzed. Specifically, the adjacent pixel point of a pixel point is obtained, and the pixel point in the eight-neighborhood of the pixel point with the smallest gray level difference is taken as the adjacent pixel point of the pixel point. When the variation degree of the gray level in the eight-neighborhood of the pixel point is closer to the variation degree of the gray level in the eight-neighborhood of the adjacent pixel point, the continuity of the pixel point and the adjacent pixel point is better, and it is less likely to be an isolated noise or an interference pixel point.

[0064] The continuity of the pixel point is obtained based on the difference in the gray level variation degree of the pixel point and the adjacent pixel point. Specifically, the absolute value of the difference between the gray level variation degree of the pixel point and the adjacent pixel point is normalized to obtain the continuity of the pixel point. The specific calculation model is as follows: ,

[0065] wherein Li is the continuity of the i-th pixel point; Bi is the gray level variation degree of the i-th pixel point; is the gray level variation degree of the adjacent pixel point of the i-th pixel point, and the calculation method is the same as that of the gray level variation degree of the i-th pixel point. When the variation degree of the i-th pixel point is closer to the variation degree of the adjacent pixel point of the i-th pixel point, the continuity of the i-th pixel point and the adjacent pixel point is better, and it is less likely to be an isolated noise or an interference pixel point. When the local laplace response is weak, the enhancement coefficient should be increased.

[0066] Step S5, the enhancement coefficient of each pixel point is optimized according to the importance, the gray level variation degree and the continuity of each pixel point, the final enhancement coefficient is obtained, and the cell image is enhanced. The cell growth state is detected by using the enhanced cell image.

[0067] After the importance, the gray level variation degree and the continuity of each pixel point are obtained, for the pixel point with a high gray level variation degree, in order to ensure the overall clarity of the image, only the gray level variation degree needs to be adjusted. For the pixel point with a low gray level variation degree, only the importance and the continuity need to be optimized.

[0068] The enhancement coefficients of each pixel point are optimized according to the importance, the gray scale change degree and the continuity of each pixel point, the final enhancement coefficient is obtained, and the cell image is enhanced, specifically, a gray scale change degree threshold is set, when the gray scale change degree of a pixel point is greater than the gray scale change degree threshold, the pixel point is marked as a high gray scale change pixel point, otherwise, the pixel point is marked as a low gray scale change pixel point; preferably, the gray scale change degree threshold is 0.6, which can be set by the user according to the user's demand, and the setting range is generally [0.5, 0.9], the larger the value, the higher the overall sharpening intensity of the image, and vice versa.

[0069] For a high gray scale change pixel point, the multiplication of the enhancement coefficient optimization sensitivity coefficient and the gray scale change degree of the high gray scale change pixel point is the reduction step of the high gray scale change pixel point, and the difference between the initial enhancement coefficient and the reduction step is the final enhancement coefficient of the high gray scale change pixel point.

[0070] For a low gray scale change pixel point, the multiplication of the enhancement coefficient optimization sensitivity coefficient and the importance, the continuity of the low gray scale change pixel point is the increase step, and the sum of the initial enhancement coefficient and the increase step is the final enhancement coefficient of the low gray scale change pixel point.

[0071] The specific calculation model is:

[0072] ,

[0073] Wherein, represents the final enhancement coefficient of the pixel point, Xi represents the initial enhancement coefficient, that is, the fixed enhancement coefficient of the laplace operator in the image enhancement in the present application, Bi represents the change degree of the i-th pixel point, 0.6 is the gray scale change degree threshold, 0.5 is the enhancement coefficient optimization sensitivity coefficient, the larger the coefficient, the greater the enhancement coefficient optimization intensity, and the greater the image sharpening intensity, and a suitable sensitivity coefficient needs to be set to prevent the image from being distorted by excessive sharpening; -0.5Bi represents the reduction step, when the gray scale change degree of the i-th pixel point is high, the enhancement coefficient needs to be reduced to ensure the clarity of the whole image, and the pixel point does not need to be considered whether it is located in the cell area; Li is the continuity of the i-th pixel point; Ci is the importance of the i-th pixel point; 0.5Li*Ci represents the increase step, when the gray scale change degree of the i-th pixel point is low, the i-th pixel point needs to be enhanced, that is, the corresponding enhancement coefficient is increased, and the basis for enhancement is the degree of belonging to the cell area and the continuity, which excludes the enhancement of interference pixels to interfere with the detection.

[0074] After obtaining the optimized enhancement coefficient of each pixel point, the whole cell image is subjected to laplace transformation and enhancement, and an enhanced image is obtained, and then the next step is processed.

[0075] The growth state of the cells includes six states of cell cycle phases (G1, S, G2, M), apoptosis, necrosis, and the like. Further, the historical enhanced cell image is segmented (segmentation of foreground and background) by using a threshold segmentation method to obtain a segmented image. The growth state of the cells in the segmented image is labeled by a person according to the morphological features and texture features of the cells in different growth states. Then, the labeled image is used to train an image recognition model (for example, a CNN) to obtain a trained model. Finally, the current enhanced cell image is segmented and input into the trained image recognition model to obtain the number of cells in different growth states. The detection result is output on a display screen of the cell growth device to display the growth state of the cells in the device in real time, which helps the staff to understand the growth state of the cells and adjust the device to better assist the cell growth.

[0076] Embodiment 2:

[0077] The embodiment provides a cell growth state detection system based on image processing, which comprises a memory, a processor, and a computer program stored in the memory and executable on the processor. The computer program implements the steps of a cell growth state detection method based on image processing when executed by the processor. Since Embodiment 1 has described a cell growth state detection method based on image processing in detail, it will not be described here.

[0078] It should be noted that the above-mentioned sequence of the embodiments of the present application is only for description, and does not represent the advantages and disadvantages of the embodiments. The above describes a specific embodiment of the present application. In addition, the processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multi-task processing and parallel processing are possible or can be advantageous.

[0079] Each embodiment in the present specification is described in a progressive manner, and the same or similar parts between each embodiment can be referred to each other. Each embodiment mainly describes the differences from other embodiments.

[0080] The above only describes the preferred embodiments of the present application and does not limit the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principles of the present application shall be included in the protection scope of the present application.

Claims

1. An image processing-based cell growth state detection method, characterized by, The method comprises: The acquired cell image is preprocessed to obtain a gray image; a window of different scales is established with a pixel point as the center, and a suspected cell region degree of the pixel point is obtained according to the gray value change in each window; The pixel points are clustered to obtain class clusters; the boundaries of the class clusters are expanded to obtain expanded boundary pixel points; the confidence of the pixel points of a class cluster is obtained according to the adjacent class clusters, the pixel points at the boundaries and the expanded boundary pixel points of the class cluster; the importance degree and the confidence are multiplied and normalized to obtain the importance degree; The absolute value of the difference between the average gray value of the pixel points in the eight-neighborhood of a pixel point and the gray value of the pixel point is normalized to obtain the gray change degree of the pixel point; The pixel point with the smallest gray difference in the eight-neighborhood of a pixel point is taken as the adjacent pixel point, and the continuity degree of the pixel point is obtained based on the difference between the gray change degrees of the pixel point and the adjacent pixel point; The importance degree, the gray change degree and the continuity degree of each pixel point are used to optimize the enhancement coefficient of each pixel point, respectively, to obtain the final enhancement coefficient and enhance the cell image; and the cell growth state is detected by using the enhanced cell image; The optimization of the enhancement coefficient of each pixel point according to the importance degree, the gray change degree and the continuity degree of each pixel point comprises: A gray change degree threshold is set, and when the gray change degree of a pixel point is greater than the gray change degree threshold, the pixel point is marked as a high gray change pixel point, otherwise, the pixel point is marked as a low gray change pixel point; For a high gray change pixel point, the multiplication of the enhancement coefficient optimization sensitivity coefficient and the gray change degree of the high gray change pixel point is taken as the reduction step of the high gray change pixel point, and the difference between the initial enhancement coefficient and the reduction step is the final enhancement coefficient of the high gray change pixel point; For a low gray change pixel point, the multiplication of the enhancement coefficient optimization sensitivity coefficient and the importance degree, the continuity degree and the like of the low gray change pixel point is taken as the increase step, and the sum of the initial enhancement coefficient and the increase step is the final enhancement coefficient of the low gray change pixel point.

2. The method of claim 1, wherein the method comprises: The suspected cell region degree of the pixel point is obtained according to the gray value change in each window, and the method comprises: The product of the standard deviation and the entropy value of the gray value of the pixel points in each window corresponding to a pixel point is obtained, and the average value is taken to obtain the suspected cell region degree of the pixel point.

3. The method of claim 1, wherein the method comprises: The pixel points are clustered to obtain class clusters, and the method comprises: The suspected cell region degree and the coordinates of a pixel point are taken as the feature vector of the pixel point, and the pixel points are clustered according to the difference between the feature vectors of each two pixel points to obtain class clusters.

4. The method of claim 1, wherein the method comprises: The boundaries of the class clusters are expanded to obtain expanded boundary pixel points, and the method comprises: The original boundaries of the class clusters are expanded outward by taking a pixel point as the expansion standard step to obtain the expanded boundaries of the class clusters, and the pixel points in the expanded boundaries are the expanded boundary pixel points.

5. The method of claim 1, wherein the method comprises: The confidence of the pixel points of a class cluster is obtained according to the adjacent class clusters, the pixel points at the boundaries and the expanded boundary pixel points of the class cluster, and the method comprises: Obtaining a distance between a center of a cluster and centers of other clusters, denoted as a first distance; a preset number of clusters with the smallest first distance among the other clusters are neighboring clusters of the cluster; Obtaining a product of an absolute value of a difference between a number of pixel points between a cluster and a corresponding neighboring cluster of the cluster and an inverse of the first distance as an inter-cluster feature corresponding to the neighboring cluster; obtaining a product of an absolute value of a difference between a mean value of gray scale values of pixel points of an original boundary of a cluster and a mean value of gray scale values of pixel points of an expanded boundary of the cluster and an inverse of a standard deviation of the pixel points of the expanded boundary as a boundary feature of the cluster; and obtaining a confidence of a pixel point of the cluster by summing the inter-cluster features corresponding to the neighboring clusters of the cluster and multiplying the boundary feature.

6. The method of claim 1, wherein the method comprises: The continuity of a pixel point is obtained based on a difference in a degree of change in gray scale of the pixel point and neighboring pixel points of the pixel point, and the continuity of the pixel point is obtained by normalizing an absolute value of the difference. The computer program, when executed by a processor, implements steps of a cell growth state detection method based on image processing according to any one of claims 1-6.

7. An image processing-based cell growth state detection system comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, wherein The computer program, when executed by a processor, implements steps of a cell growth state detection method based on image processing according to any one of claims 1-6.

Images