Analysis Method, Device, Computer Equipment and Storage Medium for Cell Distribution State

Abstract

The present application relates to a method, apparatus, computer device, and storage medium for analyzing the cell distribution state, including: obtaining a grayscale image of a cell smear as the cell image to be analyzed; obtaining the image texture features corresponding to the cell image; inputting the image texture features into a pre-trained distribution state analysis model, so as to, through the distribution state analysis model, determine the probability corresponding to each of at least two preset cell distribution states based on the image texture features, and determine the distribution state result according to the probability; and determining the cell distribution state corresponding to the cell smear according to the distribution state result output by the distribution state analysis model, thereby realizing the determination of the cell distribution state in the cell smear and being able to analyze the cell distribution uniformity in advance.

Description

Technical Field

[0001] This application relates to the field of computer technology, and particularly to a method, device, computer equipment and storage medium for analyzing the cell distribution state. Background Art

[0002] With the development of medical technology, in order to obtain more reliable cell analysis results, a liquid containing a cell sample is often stirred by an instrument to make the cells as evenly distributed as possible, and then a cell smear is made, and a cell analysis result is obtained based on the cell smear.

[0003] However, in the prior art, there is a lack of means for identifying the cell distribution state in a cell smear, and it is difficult to determine the cell distribution state in the picture when analyzing the cell smear. Summary of the Invention

[0004] Based on this, it is necessary to provide a method, device, computer equipment and storage medium for analyzing the cell distribution state in view of the above technical problems.

[0005] A method for analyzing the cell distribution state, the method comprising:

[0006] Obtaining a grayscale image of a cell smear as a cell image to be analyzed;

[0007] Obtaining an image texture feature corresponding to the cell image; the image texture feature is used to characterize the relative distribution position information between cells in the cell image;

[0008] Inputting the image texture feature into a pre-trained distribution state analysis model, so as to determine, by the distribution state analysis model, the probability corresponding to each of at least two preset cell distribution states based on the image texture feature, and determining a distribution state result according to the probability;

[0009] Determining the cell distribution state corresponding to the cell smear according to the distribution state result output by the distribution state analysis model.

[0010] Optionally, the image texture feature is a texture feature matrix, and the obtaining the image texture feature corresponding to the cell image includes:

[0011] Obtaining a plurality of pixel point pairs on the cell image and the grayscale combination corresponding to each pixel point pair;

[0012] Generating a gray-level co-occurrence matrix according to the occurrence frequency of each grayscale combination, and each matrix element in the gray-level co-occurrence matrix corresponds to the occurrence frequency of each grayscale combination;

[0013] A texture feature matrix of the cell image is obtained according to multiple gray-level co-occurrence matrices of multiple pixel pairs.

[0014] Optionally, acquiring a plurality of pixel pairs on the cell image includes:

[0015] Get multiple preset offset directions;

[0016] offsetting the positions of the reference pixels on the cell image according to the multiple offset directions to obtain offset points corresponding to the respective offset directions;

[0017] A plurality of pixel pairs on the cell image are obtained according to the reference pixel points and the offset points in each offset direction.

[0018] Optionally, obtaining the grayscale combination corresponding to each pixel pair includes:

[0019] For each pixel pair, obtain the grayscale value corresponding to each pixel in the pixel pair;

[0020] Determine the grayscale corresponding to the grayscale value of each pixel according to the preset grayscale level;

[0021] According to the grayscale corresponding to each pixel in the pixel pair, the grayscale combination corresponding to the pixel pair is determined.

[0022] Optionally, the gray level co-occurrence matrix includes gray level co-occurrence matrices corresponding to a plurality of offset directions, and obtaining the texture feature matrix of the cell image according to the gray level co-occurrence matrix includes:

[0023] Splicing multiple gray-level co-occurrence matrices to obtain a spliced gray-level co-occurrence matrix;

[0024] The spliced gray-level co-occurrence matrix is subjected to data normalization to obtain a texture feature matrix of the cell image.

[0025] Optionally, performing data standardization on the spliced gray-level co-occurrence matrix includes:

[0026] Obtaining the average value corresponding to each matrix element in the spliced gray level co-occurrence matrix;

[0027] The average value is used to perform data standardization on the spliced gray-level co-occurrence matrix to obtain a texture feature matrix of the cell image.

[0028] Optionally, it also includes:

[0029] Acquire a sample image texture feature corresponding to the sample cell grayscale image and a cell distribution state label corresponding to the sample cell grayscale image;

[0030] Input the texture features of the sample image into the convolutional neural network model to be trained, and through the convolutional neural network model, determine the probabilities corresponding to each of the preset multiple cell distribution states based on the texture features of the sample image, and determine the distribution state prediction result corresponding to the grayscale image of the sample cells according to the multiple probabilities;

[0031] Determine a loss function according to the distribution state prediction result and the cell distribution state label, and adjust the model parameters of the convolutional neural network model according to the loss function. Repeat the adjustment of the model parameters of the convolutional neural network model until the training end condition is met, and determine the current convolutional neural network model as the distribution state analysis model.

[0032] An analysis device for cell distribution state, the device includes:

[0033] A cell image acquisition module, configured to acquire the grayscale image of the cell smear as the cell image to be analyzed;

[0034] An image texture feature acquisition module, configured to acquire the image texture features corresponding to the cell image; the image texture features are used to characterize the relative distribution position information between cells in the cell image;

[0035] A feature input module, configured to input the image texture features into a pre-trained distribution state analysis model, and through the distribution state analysis model, determine the probability corresponding to each cell distribution state in at least two preset cell distribution states based on the image texture features, and determine the distribution state result according to the probabilities;

[0036] A distribution state determination module, configured to determine the cell distribution state corresponding to the cell smear according to the distribution state result output by the distribution state analysis model.

[0037] A computer device includes a memory and a processor, the memory stores a computer program, and when the processor executes the computer program, the steps of the method described in any one of the above are implemented.

[0038] A computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the method described in any one of the above are implemented.

[0039] The above analysis method, device, computer device, and storage medium for a cell distribution state obtain a grayscale image of a cell smear as the cell image to be analyzed, obtain the image texture features corresponding to the cell image, and input the image texture features into a pre-trained distribution state analysis model. The distribution state analysis model determines the probability corresponding to each of at least two preset cell distribution states based on the image texture features, and determines the distribution state result according to the probability. Furthermore, the cell distribution state corresponding to the cell smear can be determined based on the distribution state result output by the distribution state analysis model, realizing the determination of the cell distribution state in the cell smear and enabling the analysis of the cell distribution uniformity in advance. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 It is a schematic flowchart of a method for analyzing a cell distribution state in an embodiment;

[0041] Figure 2 It is a schematic flowchart of the steps for obtaining a texture feature matrix in an embodiment;

[0042] Figure 3 It is a schematic flowchart of the steps for obtaining a gray-level co-occurrence matrix in an embodiment;

[0043] Figure 4 It is a schematic diagram of a method for obtaining an offset point in an embodiment;

[0044] Figure 5 It is a schematic diagram of the structure of a convolutional neural network model in an embodiment;

[0045] Figure 6 It is a block diagram of the structure of an analysis device for a cell distribution state in an embodiment;

[0046] Figure 7 It is an internal structure diagram of a computer device in an embodiment. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0047] In order to make the objectives, technical solutions, and advantages of the present application clearer, the present application will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0048] In one embodiment, as Figure 1As shown, a method for analyzing cell distribution status is provided. This embodiment uses the method applied to a terminal as an example for illustration. It is understood that the method can also be applied to a server, or to a system including a terminal and a server, and implemented through interaction between the terminal and the server. The terminal can be, but is not limited to, a personal computer, a laptop computer, a tablet computer, or a cell slide machine. After obtaining a cell smear, a grayscale image corresponding to the cell smear can be input into the terminal for processing.

[0049] Specifically, this method may include the following steps:

[0050] Step 101: Acquire a grayscale image of a cell smear as a cell image to be analyzed.

[0051] As an example, a cell smear can be prepared from a sample containing cells to be analyzed, where the cells to be analyzed can be the target of analysis, such as cells used as biological or medical experimental samples, cells in pathological tissue, etc. A grayscale image, also known as a grayscale image, is an image represented by grayscale. By grayscaling an RGB image containing red (R), green (G), and blue (B), a corresponding grayscale image can be obtained.

[0052] In practical applications, after a cell smear is prepared and obtained, a grayscale image corresponding to the cell smear can be obtained and used as a cell image to be analyzed.

[0053] Specifically, after obtaining a cell smear, the user can send an analysis request for the cell distribution state in the cell smear to the terminal. The terminal can respond to the analysis request and obtain a grayscale image as the cell image to be analyzed.

[0054] For example, after making a cell smear, the cell smear can be photographed using a microscope with a camera function to obtain a microscopic image corresponding to the cell smear. When the microscopic image is an RGB image, it can be pre-processed.

[0055] The pre-processing may include grayscale processing and image smoothing processing. The grayscale processing may be to read the original microscopic image and convert it into a three-channel RGB format, and then obtain a single-channel grayscale image after grayscale processing.

[0056] Step 102: Acquire image texture features corresponding to the cell image; the image texture features are used to characterize relative distribution position information between cells in the cell image.

[0057] As an example, a cell image may include multiple cells, such as two or more cells; the relative distribution position information may be information representing the relative positions between the two or more cells.

[0058] In practical applications, when analyzing cells, to improve the accuracy of the analysis results, the distribution state of cells in a cell smear can be determined. Based on this, after obtaining a cell image, the cell image can be analyzed to obtain the image texture features corresponding to the cell image, and the image texture features can be used to characterize the relative distribution position information of cells in the cell image.

[0059] Step 103: Input the image texture features into a pre-trained distribution state analysis model, so as to, through the distribution state analysis model, determine the probability corresponding to each of at least two preset cell distribution states based on the image texture features, and determine a distribution state result according to the probabilities.

[0060] Specifically, cells in different cell smears can have different cell distribution states. For example, a state where cells are evenly distributed or a state where cells are unevenly distributed; or, multiple distribution states can be preset according to the level of evenness. For example, a high-level distribution state, a medium-level distribution state, a low-level distribution state, and a state where cells are concentratedly distributed. The evenness of cells is positively correlated with the level. The higher the evenness, the higher the level.

[0061] After obtaining the image texture features, the image texture features can be input into a pre-trained distribution state analysis model. After the distribution state analysis model obtains the image texture features, it can determine the probability corresponding to each of at least two preset cell distribution states based on the image texture features, and determine a distribution state result according to the respective probabilities. For example, the cell distribution state with the highest probability can be determined as the distribution state result.

[0062] Step 104: Determine the cell distribution state corresponding to the cell smear according to the distribution state result output by the distribution state analysis model.

[0063] After the distribution state analysis model determines the distribution state result, the cell distribution state corresponding to the cell smear can be determined according to the distribution state result output by the model. Specifically, the output distribution state result can be determined as the cell distribution state of the cell smear, or after obtaining the distribution state result, the distribution features corresponding to the distribution state result can also be obtained. Among them, the distribution features can be used to describe the distribution mode of cells in the cell picture, such as whether it is even, the level of evenness or dispersion, the form of dispersion, etc. Further, the distribution features can be used to determine the cell distribution state.

[0064] In this embodiment, by obtaining a grayscale image of a cell smear as the cell image to be analyzed, image texture features corresponding to the cell image are obtained, and the image texture features are input into a pre-trained distribution state analysis model. The distribution state analysis model determines the probability corresponding to each of at least two preset cell distribution states based on the image texture features, and determines the distribution state result according to the probability. Then, the cell distribution state corresponding to the cell smear can be determined according to the distribution state result output by the distribution state analysis model, thereby realizing the determination of the cell distribution state in the cell smear and being able to analyze the uniformity of the cell distribution in advance.

[0065] In practical applications, although a stirring process can be added before making a cell smear to make the cells evenly distributed, the movement of cells in the sample is a microscopic process that is difficult to control. Even if a stirring step is added, it is still difficult to ensure that the cells in the cell smear are evenly distributed, which can easily affect the analysis results. In one embodiment of the present application, for multiple cell smears, after obtaining the corresponding cell distribution states, one or more cell smears with a cell distribution uniformity higher than a preset threshold can be selected from the multiple cell smears based on the cell distribution states corresponding to each cell smear as the cell smears for analysis. By screening multiple cell smears based on the cell distribution uniformity, the reliability of the cell analysis results can be improved.

[0066] In one embodiment, the image texture feature can be a texture feature matrix, such as Figure 2 As shown, the step of obtaining the image texture features corresponding to the cell image may include the following steps:

[0067] Step 201: Acquire multiple pixel pairs on the cell image and the grayscale combination corresponding to each pixel pair.

[0068] As an example, a pixel pair may refer to a pair of two pixels, for example, pixel P1 and pixel P2 may form a pixel pair (P1, P2); a grayscale combination may refer to a combination of multiple grayscale values, for example, for grayscale value G1 and grayscale value G2, there may be a grayscale combination (G1, G2).

[0069] For cell images, the cell distribution state refers to the distribution pattern of each cell. In each distribution pattern, the relative distribution positions of each cell will show a certain arrangement relationship. This arrangement relationship can be simplified to a spiderweb-like distribution of a cell radiating outward from the center point relative to the cells around it. Based on this, the cell distribution state can be described by the corresponding grayscale combination of multiple pixel pairs in the cell image.

[0070] In a specific implementation, a cell image is composed of multiple pixel points. After obtaining the cell image to be analyzed, multiple pixel point pairs can be obtained from the cell image, and each pixel point pair includes two pixel points. Since each pixel point can have a corresponding gray value. After determining the pixel point pair, for each pixel point pair, the gray values corresponding to the respective pixel points in the pixel point pair can be obtained, and a gray level combination corresponding to the pixel point pair can be determined based on the multiple gray values. For example, for the pixel point pair (P1, P2), the gray value corresponding to pixel point P1 is G1, and the gray value corresponding to pixel point P2 is G2, and the gray level combination corresponding to the pixel point pair (P1, P2) can be determined as (G1, G2).

[0071] Step 202: Generate a gray-level co-occurrence matrix according to the occurrence frequencies of the respective gray level combinations.

[0072] As an example, the gray-level co-occurrence matrix can be a matrix that describes the image texture features through the spatial characteristics of gray levels. Each matrix element in the gray-level co-occurrence matrix corresponds to the occurrence frequency of the respective gray level combination. Specifically, since the texture can be formed by the repeated occurrence of the gray level distribution in the spatial position, there can be a certain gray level relationship between two pixels separated by a certain distance in the image space, that is, the spatial characteristics of the gray levels in the image.

[0073] In practical applications, multiple gray level combinations can be obtained for multiple pixel point pairs. After obtaining multiple gray level combinations, the occurrence frequencies of the respective gray level combinations can be counted, and a gray-level co-occurrence matrix can be generated based on the occurrence frequencies corresponding to the multiple gray level combinations.

[0074] Step 203: Obtain the texture feature matrix of the cell image according to the multiple gray-level co-occurrence matrices of the multiple pixel point pairs.

[0075] After obtaining the gray-level co-occurrence matrices of the multiple pixel point pairs, the texture feature matrix of the cell image can be determined based on the multiple gray-level co-occurrence matrices. Specifically, the multiple obtained gray-level co-occurrence matrices can be used as the texture feature matrix, or the obtained gray-level co-occurrence matrices can be further processed, and the processed matrix can be used as the texture feature matrix.

[0076] In this embodiment, by obtaining multiple pixel point pairs on the cell image and the gray level combinations corresponding to the respective pixel point pairs, generating a gray-level co-occurrence matrix according to the occurrence frequencies of the respective gray level combinations, and obtaining the texture feature matrix of the cell image according to the gray-level co-occurrence matrices of the multiple pixel point pairs, the image texture features of the cell image can be characterized by the gray-level co-occurrence matrix, and the texture features can be reflected by the occurrence frequencies of the gray levels, realizing the quantitative description of the texture features.

[0077] In one embodiment, the obtaining of the multiple pixel point pairs on the cell image may include the following steps:

[0078] Acquire multiple preset offset directions; according to the multiple offset directions, offset the positions of the reference pixel points on the cell image respectively to obtain offset points corresponding to each offset direction; and obtain multiple pixel point pairs on the cell image based on the reference pixel points and offset points in each offset direction.

[0079] In a specific implementation, multiple preset offset directions can be obtained. Multiple preset offset directions can be described by direction vectors, and the direction and magnitude of the offset can be determined by the direction vectors. Those skilled in the art can set the offset direction according to actual conditions. For example, the offset directions (0, 1), (1, 0), (-1, -1) and (-1, 1) can be selected. The above four offset directions can also be called unit direction vectors. For example, different numerical combinations can be taken for the offset direction (a, b) to obtain multiple offset directions. In actual applications, for cells with a spider-web-like distribution, offsets can be made in four directions: 0°, 45°, 90° and 135°.

[0080] After determining multiple offset directions, for each offset direction, the position of a reference pixel on the cell image can be offset using that offset direction to obtain an offset point corresponding to the offset position. The reference pixel on the cell image can be any point on the image. Specifically, taking an N×N pixel cell image as an example, any point (x, y) can be selected as a reference pixel, and based on the offset direction (a, b), an offset point (x+a, y+b) can be obtained. By traversing the pixel points on the cell image, multiple reference pixels and offset points corresponding to the offset directions can be obtained.

[0081] For each offset direction, multiple pixel pairs in the direction can be obtained based on multiple reference pixels and their corresponding offset points, and then multiple pixel pairs on the cell image can be determined based on the pixel pairs corresponding to the multiple offset directions.

[0082] In this embodiment, by offsetting the positions of the reference pixels on the cell image according to multiple offset directions, the offset points corresponding to each offset direction are obtained, and multiple pixel pairs in different directions can be obtained, providing a data basis for the subsequent generation of the grayscale co-occurrence matrix.

[0083] In one embodiment, Figure 3 As shown, obtaining the grayscale combination corresponding to each pixel pair may include the following steps:

[0084] Step 301: For each pixel pair, obtain the grayscale value of each pixel in the pixel pair.

[0085] Since each pixel in a pixel pair may correspond to a grayscale value, after obtaining multiple pixel pairs, the grayscale value corresponding to each pixel in the pixel pair may be obtained for each pixel pair.

[0086] Step 302 : determining the grayscale corresponding to the grayscale value of each pixel according to a preset grayscale level.

[0087] As an example, a cell image may have a preset grayscale. Specifically, grayscale may refer to dividing the brightness variation between the brightest and darkest levels into several levels, resulting in multiple grayscale levels. For example, the grayscale level may be 16, 32, or 64 levels, and each grayscale level may correspond to a grayscale value within a preset range.

[0088] Based on this, for each pixel pair, after determining the grayscale value corresponding to each pixel in the pixel pair, the grayscale corresponding to the grayscale value of each pixel can be determined in time in combination with the preset grayscale.

[0089] Step 303 : determining a grayscale combination corresponding to the pixel pair according to the grayscale corresponding to each pixel in the pixel pair.

[0090] For each pixel pair, after determining the grayscale corresponding to each pixel in the pixel pair, the grayscale combination corresponding to the pixel can be determined based on multiple grayscales. For example, for the pixel pair (A, B), the grayscale corresponding to pixel A is G1, and the grayscale corresponding to pixel B is G2, so the grayscale combination (G1, G2) can be obtained.

[0091] In this embodiment, the grayscale corresponding to the grayscale value of each pixel is determined according to a preset grayscale level, and the grayscale combination corresponding to the pixel pair is determined according to the grayscale corresponding to each pixel in the pixel pair, providing a data basis for constructing a grayscale co-occurrence matrix.

[0092] In one embodiment, the gray level co-occurrence matrix includes gray level co-occurrence matrices corresponding to a plurality of offset directions, and obtaining the texture feature matrix of the cell image according to the gray level co-occurrence matrix may include the following steps:

[0093] Multiple gray-level co-occurrence matrices are spliced to obtain a spliced gray-level co-occurrence matrix; and data normalization is performed on the spliced gray-level co-occurrence matrix to obtain a texture feature matrix of the cell image.

[0094] In a specific implementation, after obtaining the gray-level co-occurrence matrices corresponding to multiple offset directions, the multiple gray-level co-occurrence matrices can be spliced to obtain a spliced gray-level co-occurrence matrix. After obtaining the spliced gray-level co-occurrence matrix, data normalization processing can be performed on it, and the matrix after data normalization processing is used as the texture feature matrix of the cell image.

[0095] For example, Figure 4 As shown, for the target pixel (pixel of interest) A(x, y), that is, the reference pixel in this application, the offset points corresponding to the four offset directions can be obtained according to the preset four offset directions (0, 1), (1, 0), (-1, -1), (-1, 1). By adjusting the position of the pixel center point A(x, y) and traversing each pixel on the cell image, multiple pixel pairs corresponding to each offset direction and their corresponding grayscale combinations can be obtained, and then the grayscale co-occurrence matrices (also called joint probability matrices) corresponding to the four offset directions (0, 1), (1, 0), (-1, -1), (-1, 1) can be generated. After obtaining the four grayscale co-occurrence matrices, the matrices can be spliced according to the shape of a "田" character to obtain the spliced co-occurrence matrix.

[0096] In this embodiment, by splicing multiple gray-level co-occurrence matrices to obtain a spliced gray-level co-occurrence matrix, and performing data standardization on the spliced gray-level co-occurrence matrix to obtain a texture feature matrix, a matrix reflecting the texture features in all directions can be obtained, providing a data basis for the subsequent determination of the cell distribution status.

[0097] In one embodiment, the data normalization of the spliced gray-level co-occurrence matrix may include the following steps:

[0098] Obtain the average value corresponding to each matrix element in the spliced gray-level co-occurrence matrix; use the average value to perform data standardization on the spliced gray-level co-occurrence matrix to obtain the texture feature matrix of the cell image.

[0099] In practical applications, each matrix element in the spliced gray-level co-occurrence matrix can be obtained, and the corresponding average value of each matrix element can be calculated. This average value can then be used to normalize the gray-level co-occurrence matrix data to obtain the texture feature matrix corresponding to the cell image. Each matrix element can correspond to a grayscale combination, and the matrix element can be the frequency of occurrence of the corresponding grayscale combination.

[0100] Specifically, for each occurrence frequency P, the standard deviation std(P) corresponding to each occurrence frequency can be obtained. When performing data normalization, for each occurrence frequency P in the texture feature matrix, the difference between the occurrence frequency P and the average value is calculated, and then the result is divided by the standard deviation std(P).

[0101] In this embodiment, by performing data standardization on the spliced gray-level co-occurrence matrix, the model can output more accurate results.

[0102] In another example, during the training of the distribution state analysis model, data normalization processing can also be performed on the texture features of the sample images used for training. The data normalization processing of the model training process is similar to the above steps and will not be described in detail in this application. During model training, by performing data normalization processing, the convergence speed of the model can be increased and the training efficiency can be improved.

[0103] In one embodiment, the method may further include the following steps:

[0104] Step 401 : Acquire sample image texture features corresponding to a sample cell grayscale image and cell distribution state labels corresponding to the sample cell grayscale image.

[0105] As an example, the sample cell grayscale image may be a cell grayscale image corresponding to a sample cell smear used to train the model.

[0106] In practical applications, after obtaining the sample cell grayscale image, the image texture characteristics corresponding to the image, that is, the sample image texture features, and the cell distribution state label corresponding to the sample cell grayscale image can be obtained.

[0107] In step 402, the texture features of the sample image are input into a convolutional neural network model to be trained, so that the convolutional neural network model can determine the probabilities corresponding to the preset multiple cell distribution states based on the texture features of the sample image, and determine the distribution state prediction result corresponding to the sample cell grayscale image based on the multiple probabilities.

[0108] As an example, the distribution state prediction result may be a prediction result of the uniformity of cell distribution in a sample cell grayscale image.

[0109] After obtaining the texture features of the sample image, the features can be input into the convolutional neural network model to be trained. The convolutional neural network model can determine the probabilities corresponding to the preset multiple cell distribution states through the input sample image texture features, and then determine the distribution state prediction results corresponding to the sample cell grayscale image based on multiple probabilities.

[0110] In practical applications, a convolutional neural network model may include multiple "neurons", such as convolutional layers, pooling layers, fully connected layers, and softmax layers. Figure 5 As shown in the figure, when determining the distribution state prediction result, after the convolution operation is performed on the input texture feature matrix through the convolution layer and the pooling layer, it can be input into the fully connected layer, and the result of the fully connected layer is mapped to the (0, 1) interval through the softmax layer to obtain the probability corresponding to each cell distribution state, wherein the uniformity of cell distribution can be positively correlated with the probability, that is, the closer to 0, the more uneven the distribution, and conversely, the closer to 1, the more uniform the distribution.

[0111] Step 403: Determine a loss function based on the distribution state prediction result and the cell distribution state label, and adjust the model parameters of the convolutional neural network model based on the loss function. Repeat the adjustment of the model parameters of the convolutional neural network model until the training end condition is met, and determine the current convolutional neural network model as the distribution state analysis model.

[0112] After obtaining the distribution state prediction result, the loss function can be determined based on the prediction result and the cell distribution state label, wherein the loss function L can adopt the cross entropy loss function shown below:

[0113]

[0114] Among them, y j is the jth cell distribution state, S j is the probability corresponding to the j-th cell distribution state.

[0115] After determining the corresponding cross entropy loss function, the model parameters of the convolutional neural network model can be adjusted according to the loss function. After the adjustment, the process can return to step 401 and repeat the step of adjusting the parameters of the convolutional neural network model again. The process is iterated continuously until the training end conditions are met. The current convolutional neural network model can be determined as a distribution state analysis model.

[0116] In this embodiment, by training the convolutional neural network, a trained distribution state analysis model can be obtained, which provides a model for the subsequent determination of the cell distribution state and can automatically identify the cell distribution state in various cell smears, effectively improving the analysis efficiency.

[0117] It should be understood that although Figure 1-3 The steps in the flowchart are shown in sequence as indicated by the arrows, but these steps are not necessarily executed in the order indicated by the arrows. Unless otherwise specified in this document, there is no strict order restriction for the execution of these steps, and these steps can be executed in other orders. In addition, Figure 1-3 At least part of the steps may include multiple steps or multiple stages. These steps or stages are not necessarily performed at the same time, but can be performed at different times. The order of execution of these steps or stages is not necessarily one by one, but can be performed in turn or alternately with other steps or at least part of the steps or stages in other steps.

[0118] In one embodiment, Figure 6 As shown, a device for analyzing cell distribution status is provided, and the device may include:

[0119] The cell image acquisition module 601 is configured to acquire a grayscale image of a cell smear as the cell image to be analyzed;

[0120] The image texture feature acquisition module 602 is configured to acquire the image texture features corresponding to the cell image; the image texture features are used to characterize the relative distribution position information of cells in the cell image;

[0121] The feature input module 603 is configured to input the image texture features into a pre-trained distribution state analysis model, so as to determine, by means of the distribution state analysis model, the probability corresponding to each of at least two preset cell distribution states based on the image texture features, and determine a distribution state result according to the probability;

[0122] The distribution state determination module 604 is configured to determine the cell distribution state corresponding to the cell smear according to the distribution state result output by the distribution state analysis model.

[0123] In one embodiment, the image texture features are a texture feature matrix, and the image texture feature acquisition module 602 includes:

[0124] The gray-level combination acquisition sub-module is configured to acquire a plurality of pixel point pairs on the cell image and the gray-level combinations corresponding to the respective pixel point pairs;

[0125] The gray-level co-occurrence matrix generation sub-module is configured to generate a gray-level co-occurrence matrix according to the occurrence frequencies of the respective gray-level combinations, and each matrix element in the gray-level co-occurrence matrix corresponds to the occurrence frequency of the respective gray-level combination;

[0126] The texture feature matrix acquisition sub-module is configured to obtain the texture feature matrix of the cell image according to a plurality of gray-level co-occurrence matrices of a plurality of pixel point pairs.

[0127] In one embodiment, the gray-level combination acquisition sub-module includes:

[0128] The offset direction acquisition unit is configured to acquire a plurality of preset offset directions;

[0129] The offset point acquisition unit is configured to offset the position of a reference pixel point on the cell image according to the plurality of offset directions to obtain offset points corresponding to the respective offset directions;

[0130] The pixel point pair acquisition unit is configured to obtain a plurality of pixel point pairs on the cell image according to the reference pixel point and the offset points in each offset direction.

[0131] In one embodiment, the gray-level combination acquisition sub-module includes:

[0132] a grayscale acquisition unit, configured to acquire a plurality of grayscales corresponding to the cell image, each grayscale corresponding to a grayscale value within a preset range;

[0133] The grayscale combination determining unit is used to determine the grayscale corresponding to each pixel point in each pixel point pair according to the grayscale value corresponding to each pixel point in the pixel point pair, and determine the grayscale combination corresponding to the pixel point pair according to multiple grayscales.

[0134] In one embodiment, the gray level co-occurrence matrix includes gray level co-occurrence matrices corresponding to multiple offset directions, and the texture feature matrix acquisition submodule includes:

[0135] a splicing unit, configured to splice a plurality of gray-level co-occurrence matrices to obtain a spliced gray-level co-occurrence matrix;

[0136] The normalization unit is used to perform data normalization on the spliced gray-level co-occurrence matrix to obtain a texture feature matrix of the cell image.

[0137] In one embodiment, the standardization unit includes:

[0138] An average value obtaining subunit is used to obtain the average value corresponding to each matrix element in the spliced gray level co-occurrence matrix;

[0139] The matrix normalization subunit is used to use the average value to perform data normalization on the spliced gray-level co-occurrence matrix to obtain the texture feature matrix of the cell image.

[0140] In one embodiment, the apparatus further comprises:

[0141] A sample image texture feature acquisition module is used to acquire the sample image texture feature corresponding to the sample cell grayscale image and the cell distribution state label corresponding to the sample cell grayscale image;

[0142] a sample feature input module, idiomatically inputting the sample image texture features into a convolutional neural network model to be trained, so as to determine, through the convolutional neural network model, the probabilities corresponding to each of a plurality of preset cell distribution states based on the sample image texture features, and determining a distribution state prediction result corresponding to the sample cell grayscale image based on the plurality of probabilities;

[0143] The distribution state analysis model acquisition module is used to determine the loss function based on the distribution state prediction result and the cell distribution state label, and adjust the model parameters of the convolutional neural network model according to the loss function, repeatedly adjust the model parameters of the convolutional neural network model until the training end conditions are met, and determine the current convolutional neural network model as the distribution state analysis model.

[0144] For the specific limitations of an analysis device for the cell distribution state, reference can be made to the limitations of the analysis method for the cell distribution state in the above text, which will not be elaborated here. Each module in the above analysis device for the cell distribution state can be implemented in whole or in part by software, hardware, and their combination. Each of the above modules can be embedded in the processor of the computer device in hardware form or be independent of it, or can be stored in the memory of the computer device in software form, so that the processor can call and execute the operations corresponding to each of the above modules.

[0145] In one embodiment, a computer device is provided. The computer device can be a terminal, and its internal structure diagram can be as Figure 7 shown. The computer device includes a processor, a memory, a communication interface, a display screen, and an input device connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner. The wireless manner can be achieved through WIFI, a carrier network, NFC (Near Field Communication), or other technologies. When the computer program is executed by the processor, it realizes an analysis method for the cell distribution state. The display screen of the computer device can be a liquid crystal display screen or an electronic ink display screen. The input device of the computer device can be a touch layer covering the display screen, or a button, a trackball, or a touchpad provided on the housing of the computer device, or an external keyboard, touchpad, or mouse, etc.

[0146] Those skilled in the art can understand that Figure 7 the structure shown in

[0147] is only a block diagram of some structures related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine some components, or have a different component layout.

[0148] Obtain a grayscale image of a cell smear as the cell image to be analyzed;

[0149] Obtain the image texture feature corresponding to the cell image; the image texture feature is used to characterize the relative distribution position information between cells in the cell image;

[0150] Input the image texture features into a pre-trained distribution state analysis model, so as to, through the distribution state analysis model, determine the probability corresponding to each of at least two preset cell distribution states based on the image texture features, and determine a distribution state result according to the probability;

[0151] Determine the cell distribution state corresponding to the cell smear according to the distribution state result output by the distribution state analysis model.

[0152] In one embodiment, when the processor executes the computer program, the steps in the above-mentioned other embodiments are also implemented.

[0153] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the following steps are implemented:

[0154] Obtain a grayscale image of a cell smear as a cell image to be analyzed;

[0155] Obtain the image texture features corresponding to the cell image; the image texture features are used to characterize the relative distribution position information between cells in the cell image;

[0156] Input the image texture features into a pre-trained distribution state analysis model, so as to, through the distribution state analysis model, determine the probability corresponding to each of at least two preset cell distribution states based on the image texture features, and determine a distribution state result according to the probability;

[0157] Determine the cell distribution state corresponding to the cell smear according to the distribution state result output by the distribution state analysis model.

[0158] In one embodiment, when the computer program is executed by a processor, the steps in the above-mentioned other embodiments are also implemented.

[0159] Those skilled in the art will appreciate that all or part of the processes in the above-mentioned embodiment methods can be implemented by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application may include at least one of non-volatile and volatile memory. Non-volatile memory may include read-only memory (ROM), magnetic tape, floppy disk, flash memory or optical memory, etc. Volatile memory may include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM).

[0160] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0161] The above-mentioned embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the invention patent. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the patent of the present application shall be subject to the attached claims.

Claims

1. A method for analyzing the cell distribution state, characterized in that The method includes: Obtaining a grayscale image of a cell smear as the cell image to be analyzed; Obtaining a plurality of pixel pairs on the cell image and the corresponding gray-level combinations of each pixel pair; Generating a gray-level co-occurrence matrix according to the occurrence frequencies of the respective gray-level combinations, where each matrix element in the gray-level co-occurrence matrix corresponds to the occurrence frequency of each gray-level combination; Obtaining a texture feature matrix of the cell image based on a plurality of gray-level co-occurrence matrices of a plurality of pixel pairs; Taking the texture feature matrix as the image texture feature corresponding to the cell image; the image texture feature is used to characterize the relative distribution position information between cells in the cell image; Inputting the image texture feature into a pre-trained distribution state analysis model, so as to, through the distribution state analysis model, determine the probability corresponding to each of at least two preset cell distribution states based on the image texture feature, and determine a distribution state result according to the probability; Determining the cell distribution state corresponding to the cell smear according to the distribution state result output by the distribution state analysis model; The distribution state analysis model is trained through the following steps: Obtaining a sample image texture feature corresponding to a sample cell grayscale image and a cell distribution state label corresponding to the sample cell grayscale image; Inputting the sample image texture feature into a convolutional neural network model to be trained, so as to, through the convolutional neural network model, determine the probability corresponding to each of a plurality of preset cell distribution states based on the sample image texture feature, and determine a distribution state prediction result corresponding to the sample cell grayscale image according to a plurality of probabilities; Determining a loss function according to the distribution state prediction result and the cell distribution state label, and adjusting the model parameters of the convolutional neural network model according to the loss function. Repeatedly adjusting the model parameters of the convolutional neural network model until the training end condition is met, and determining the current convolutional neural network model as the distribution state analysis model.

2. The method according to claim 1, characterized in that, The obtaining of the plurality of pixel pairs on the cell image includes: Obtaining a plurality of preset offset directions; According to the plurality of offset directions, respectively offsetting the position of a reference pixel point on the cell image to obtain offset points corresponding to each offset direction; Obtaining a plurality of pixel pairs on the cell image according to the reference pixel point and the offset points in each offset direction.

3. The method according to claim 1, characterized in that, The obtaining of the corresponding gray-level combination of each pixel pair includes: For each pixel pair, obtaining the gray-level values of the respective pixels in the pixel pair; Determining the gray level corresponding to the gray-level value of each pixel according to a preset number of gray levels; Determining the gray-level combination corresponding to the pixel pair according to the gray levels corresponding to the respective pixels in the pixel pair.

4. The method according to claim 2, wherein The gray-level co-occurrence matrix includes gray-level co-occurrence matrices corresponding to a plurality of offset directions respectively. The obtaining of the texture feature matrix of the cell image according to the gray-level co-occurrence matrix includes: Stitching a plurality of gray-level co-occurrence matrices to obtain a stitched gray-level co-occurrence matrix; Performing data normalization on the stitched gray-level co-occurrence matrix to obtain the texture feature matrix of the cell image.

5. The method according to claim 4, characterized in that, The data normalization of the stitched gray-level co-occurrence matrix includes: Obtaining the average value corresponding to each matrix element in the stitched gray-level co-occurrence matrix; Using the average value to perform data normalization on the stitched gray-level co-occurrence matrix to obtain the texture feature matrix of the cell image.

6. An analysis device for the cell distribution state, characterized in that, The device includes: A cell image acquisition module, configured to acquire a gray-scale image of a cell smear as a cell image to be analyzed; An image texture feature acquisition module, configured to acquire multiple pixel pairs on the cell image and the gray-level combinations corresponding to each pixel pair; generating a gray-level co-occurrence matrix according to the occurrence frequencies of each gray-level combination, where each matrix element in the gray-level co-occurrence matrix corresponds to the occurrence frequency of each gray-level combination; obtaining the texture feature matrix of the cell image according to multiple gray-level co-occurrence matrices of multiple pixel pairs; using the texture feature matrix as the image texture feature corresponding to the cell image; the image texture feature is used to characterize the relative distribution position information between cells in the cell image; A feature input module, configured to input the image texture feature into a pre-trained distribution state analysis model, so as to, through the distribution state analysis model, determine the probability corresponding to each of at least two preset cell distribution states based on the image texture feature, and determine a distribution state result according to the probability; A distribution state determination module, configured to determine the cell distribution state corresponding to the cell smear according to the distribution state result output by the distribution state analysis model; A sample image texture feature acquisition module, configured to acquire a sample image texture feature corresponding to a sample cell gray-scale image and a cell distribution state label corresponding to the sample cell gray-scale image; A sample feature input module, conventionally used to input the sample image texture feature into a convolutional neural network model to be trained, so as to, through the convolutional neural network model, determine the probability corresponding to each of multiple preset cell distribution states based on the sample image texture feature, and determine a distribution state prediction result corresponding to the sample cell gray-scale image according to multiple probabilities; A distribution state analysis model acquisition module, configured to determine a loss function according to the distribution state prediction result and the cell distribution state label, and adjust the model parameters of the convolutional neural network model according to the loss function, repeatedly adjust the model parameters of the convolutional neural network model until a training end condition is satisfied, and determine the current convolutional neural network model as the distribution state analysis model.

7. A computer device, comprising a memory and a processor, the memory storing a computer program, characterized in that, When the processor executes the computer program, the steps of the method according to any one of claims 1 to 5 are implemented.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, the steps of the method according to any one of claims 1 to 5 are implemented.

Images