Method and system for optimizing and identifying visual data of hematologic tumor cells

Abstract

The invention relates to the technical field of image processing, in particular to a visual data optimization recognition method and system for hematologic tumor cells, and the method comprises the steps: obtaining a local region of a pixel point, obtaining a plurality of clustering clusters of each pixel point, a to-be-analyzed region of each pixel point and a region outside the clustering clusters, and obtaining a self-adaptive growth threshold pixel point; obtaining a first self-adaptive adjustment parameter of each self-adaptive growth threshold pixel point according to the number and position distribution of the pixel points in the clustering cluster of the self-adaptive growth threshold pixel points; obtaining a second self-adaptive adjustment parameter of each self-adaptive growth threshold pixel point according to the shape of the cluster of each self-adaptive growth threshold pixel point and the gray value of each pixel point in the cluster; obtaining a comprehensive self-adaptive adjustment parameter of each pixel point; and obtaining a tumor area in the blood image. The invention aims to solve the problem that normal cells are mistakenly segmented into tumor cells when tumor cells in a blood image are acquired through a region growing algorithm.

Description

Technical Field

[0001] This invention relates to the field of image processing technology, specifically to a method and system for optimizing and recognizing visual data of blood tumor cells. Background Technology

[0002] Hematopoietic tumor cells refer to tumor cells in the blood. Because observing tumor cells in the blood can help doctors better understand the patient's condition, it is necessary to segment tumor cells from blood images.

[0003] When segmenting tumor cells from blood images using the region growing algorithm, the growth threshold is fixed during the growth process. However, normal blood cells and tumor cells may cluster together in blood images. Furthermore, the difference between tumor cells and normal blood cells is smaller than the difference between tumor cells and non-blood cells. This can lead to the region growing algorithm misidentifying pixels of normal blood cells as pixels of tumor cells, resulting in a significant difference between the segmented tumor cells and the actual tumor cells in the image. Consequently, the utilization rate of blood data is too low. Summary of the Invention

[0004] This invention provides a method and system for optimizing and identifying visual data of hematologic tumor cells to solve existing problems.

[0005] The present invention provides a method and system for optimizing and identifying visual data of hematological tumor cells, which adopts the following technical solution: This invention proposes a method for optimizing and identifying visual data of hematological tumor cells, which includes the following steps: A blood image is acquired, and the local region of each pixel is obtained. Based on the gray values ​​of each pixel and its surrounding pixels, as well as the local region of each pixel, several clusters of each pixel are obtained, along with the region to be analyzed and the region outside the clusters. Based on the gray values ​​of pixels in the region to be analyzed and the region outside the clusters, the positional distribution of pixels in each pixel's clusters, and the number of pixels in the region to be analyzed, the similarity between each pixel and its surrounding pixels is obtained. Based on the similarity between each pixel and its surrounding pixels, an adaptive growth threshold pixel is obtained. Based on the number and location distribution of pixels in the cluster of adaptive growth threshold pixels, the degree of cell aggregation around each adaptive growth threshold pixel is obtained, and then the first adaptive adjustment parameter of each adaptive growth threshold pixel is obtained. Based on the shape of each cluster of each adaptive growth threshold pixel and the gray value of each pixel in each cluster, the probability that each adaptive growth threshold pixel contains normal cells is obtained, and then the second adaptive adjustment parameter of each adaptive growth threshold pixel is obtained. Based on the first adaptive adjustment parameter and the second adaptive adjustment parameter of each adaptive growth threshold pixel, the comprehensive adaptive adjustment parameter of each pixel is obtained. The tumor region in the blood image is obtained by comprehensively and adaptively adjusting the parameters for each pixel.

[0006] Furthermore, the specific steps for obtaining the local region of each pixel are as follows: Let the g-th pixel in the blood image be denoted as... ,by Get a center point The local region obtained is denoted as A local area; among which, The length of the preset local area.

[0007] Furthermore, the specific steps for obtaining several clusters of each pixel and the region to be analyzed and the region outside the clusters for each pixel based on the gray values ​​of each pixel and its surrounding pixels, as well as the local region of each pixel, are as follows: Will The average of the absolute values ​​of the differences between the gray values ​​of the pixels and the gray values ​​of all pixels in the 8-neighborhood is used as The nearest neighbor threshold; by As the starting point for regional growth, with The nearest neighbor threshold is used as the growth threshold. The pixels in the 8-neighborhood direction are grown, and the first... Pixels grown in each neighborhood direction constitute The One initial cluster; by The nearest neighbor threshold is used as the growth threshold, to The Each pixel in the initial cluster is used as a growth point, and growth is performed simultaneously. The grown pixels and the resulting pixels are then compared with the first cluster. The pixels in the initial clusters constitute The Clusters; All clusters as The region to be analyzed; Will The region within a local area that is not part of the region to be analyzed is denoted as . The region outside the cluster.

[0008] Furthermore, the process of obtaining the similarity between each pixel and its surrounding pixels based on the grayscale values ​​of pixels in the region to be analyzed and the region outside the cluster, the positional distribution of pixels in the cluster of each pixel, and the number of pixels contained in the region to be analyzed includes the following specific formulas: ; In the formula, M represents the similarity between the g-th pixel and its surrounding pixels; M represents... The number of clusters; express The maximum Euclidean distance between a pixel in the m-th cluster and a pixel in the (m+1)-th cluster; express The average gray value of all pixels within the outer region of the cluster; express The average gray value of all pixels in the m-th cluster; express The standard deviation of grayscale values ​​of all pixels within the outer region of the cluster; express The standard deviation of grayscale values ​​of all pixels within a local region; express The nearest neighbor threshold; express The number of pixels contained in the region to be analyzed; Represents the absolute value function; It is an exponential function with the natural constant as its base; This represents the normalization function.

[0009] Furthermore, the specific steps for obtaining the adaptive growth threshold pixel based on the similarity between each pixel and its surrounding pixels are as follows: Set similarity threshold When the similarity between the g-th pixel and its surrounding pixels is greater than or equal to the similarity threshold When the time is right, the g-th pixel is recorded as the adaptive growth threshold pixel.

[0010] Furthermore, the specific steps for obtaining the degree of cell aggregation around each adaptive growth threshold pixel based on the number and position distribution of pixels in the cluster of adaptive growth threshold pixels, and thus obtaining the first adaptive adjustment parameter for each adaptive growth threshold pixel, are as follows: For the first The clusters of the adaptive growth threshold pixels are combined pairwise to obtain the first... A combination of multiple clusters of pixels with an adaptive growth threshold; In each cluster combination, the cluster with the fewest pixels is denoted as the smallest cluster; the cluster with the most pixels is denoted as the largest cluster. Calculate the Euclidean distance between each pixel in the smallest cluster and each pixel in the largest cluster in any cluster combination. The pixel in the smallest cluster corresponding to the smallest Euclidean distance is recorded as the principal analysis point of the cluster combination. The Euclidean distance between the principal analysis point of the cluster combination and the boundary points of each of the largest clusters in the cluster combination is used as several adjacency distances in the cluster combination to obtain the number of pixels in the smallest cluster of the cluster combination. The product of the products, rounded up, is taken as the number of neighbors for the cluster combination. The distance of the smallest number of neighbors among all the neighbor distances of the cluster combination is taken as a number of neighbor distances in the cluster combination; where R is equal to a constant 2. Get the The number of pixels contained in the minimum bounding rectangle of the region to be analyzed for each adaptive growth threshold pixel; According to the The nearest neighbor distance in the cluster combination of the adaptive growth threshold pixels, the first The number of pixels contained in the minimum bounding rectangle of the region to be analyzed for the adaptive growth threshold pixel, and the number of pixels in the region to be analyzed for the adaptive growth threshold pixel. The number of pixels contained in each cluster of the adaptive growth threshold pixel is obtained. The degree of cell aggregation around the adaptive growth threshold pixel, thus obtaining the degree of cell aggregation around the first pixel. The first adaptive adjustment parameter for each adaptive growth threshold pixel.

[0011] Furthermore, the obtained first The degree of cell aggregation around the adaptive growth threshold pixel, thus obtaining the degree of cell aggregation around the first pixel. The first adaptive adjustment parameter for each adaptive growth threshold pixel includes the following specific formula: ; In the formula, Indicates the first The degree of cell aggregation around the i-th adaptive growth threshold pixel; denoted as the i-th adaptive growth threshold pixel. ; express The number of clusters; express The average number of pixels contained in each cluster; express The number of cluster combinations; B represents The maximum number of neighbor distances in each cluster combination; express The (b+1)th nearest neighbor distance in the nth cluster combination; express The b-th nearest neighbor distance in the nth cluster combination; express The number of pixels in the region to be analyzed; express The number of pixels contained in the smallest bounding rectangle of the region to be analyzed; It is an exponential function with the natural constant as its base; ; In the formula, Indicates the first The first adaptive adjustment parameter for each adaptive growth threshold pixel. This is the normalization function.

[0012] Furthermore, the specific steps for obtaining the probability that each adaptive growth threshold pixel contains normal cells based on the shape of each cluster and the gray value of each pixel in each cluster, and thus obtaining the second adaptive adjustment parameter for each adaptive growth threshold pixel, are as follows: The gradient value of each pixel in the blood image is obtained by using the Sobel operator. The chain code is obtained for each cluster of each adaptive growth threshold pixel, and the edge chain code of each cluster is obtained. Based on the gradient, grayscale, and edge chain code of each pixel within the analysis region of each adaptive growth threshold pixel, the probability of normal cells surrounding each adaptive growth threshold pixel and the second adaptive adjustment parameter of each adaptive growth threshold pixel are obtained. The specific calculation formula is as follows: ; ; In the formula, Indicates the first The probability that a pixel with an adaptive growth threshold contains normal cells. express The number of cluster combinations; K represents The number of chain code directions in the edge chain codes of each cluster; express In the nth cluster combination, the edge chain code of the two clusters is the nth... The absolute value of the difference between the number of chain code values ​​in each direction; express The standard deviation of the gradient values ​​of pixels in each cluster; It is an exponential function with the natural constant as its base; Indicates the first The second adaptive adjustment parameter for each adaptive growth threshold pixel.

[0013] Furthermore, the specific steps for obtaining the comprehensive adaptive adjustment parameters for each pixel based on the first adaptive adjustment parameters and the second adaptive adjustment parameters for each adaptive growth threshold pixel are as follows: ; In the formula, Indicates the first A comprehensive adaptive adjustment parameter for each adaptive growth threshold pixel; Indicates the first The first adaptive adjustment parameter for each adaptive growth threshold pixel; Indicates the first The second adaptive adjustment parameter for each adaptive growth threshold pixel; Set the overall adaptive adjustment parameter of pixels that are not adaptive growth threshold pixels to 1.

[0014] The present invention also proposes a visualization data optimization and identification system for hematologic tumor cells, including a memory, a processor, and a computer program stored in the memory and running on the processor. The processor executes the computer program stored in the memory to implement the steps of the aforementioned visualization data optimization and identification method for hematologic tumor cells.

[0015] The beneficial effects of the technical solution of the present invention are as follows: When obtaining tumor cells from blood images through the region growing algorithm, an adaptive growth threshold pixel is first obtained based on the difference in gray value between each pixel and its surrounding pixels. This eliminates the need to calculate the comprehensive adaptive growth threshold of all pixels, reducing the computational load when calculating the comprehensive adaptive growth threshold of pixels. Based on the relatively regular shape of normal cells and the relatively similar gray values ​​of pixels within normal cells, when calculating the probability that each adaptive growth threshold pixel contains normal cells, the probability that the adaptive growth threshold pixel near normal cells contains normal cells is greater than the probability that the adaptive growth threshold pixel near tumor cells contains normal cells. This results in a more accurate adaptive growth threshold for the adaptive growth threshold pixel, leading to more accurate segmentation of tumor cells. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0017] Figure 1 This is a flowchart illustrating the steps of a visual data optimization and identification method for blood tumor cells according to the present invention. Figure 2 Image of a blood sample. Detailed Implementation

[0018] To further illustrate the technical means and effects adopted by the present invention to achieve its intended purpose, the following, in conjunction with the accompanying drawings and preferred embodiments, details the specific implementation, structure, features, and effects of a visualization data optimization and identification method and system for blood tumor cells proposed according to the present invention. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.

[0019] 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 this invention pertains.

[0020] The following description, in conjunction with the accompanying drawings, details the specific scheme of the visualization data optimization and identification method and system for blood tumor cells provided by this invention.

[0021] Please see Figure 1 The diagram illustrates a flowchart of a method for optimizing and identifying visual data of hematological tumor cells according to an embodiment of the present invention. The method includes the following steps: Step S001: Acquire a blood image to obtain the local region of each pixel; based on the gray values ​​of each pixel and its surrounding pixels, and the local region of each pixel, obtain several clusters of each pixel, as well as the region to be analyzed and the region outside the clusters for each pixel; based on the gray values ​​of pixels in the region to be analyzed and the region outside the clusters for each pixel, the positional distribution of pixels in the clusters of each pixel, and the number of pixels contained in the region to be analyzed, obtain the similarity between each pixel and its surrounding pixels; based on the similarity between each pixel and its surrounding pixels, obtain the adaptive growth threshold pixels.

[0022] The main purpose of this embodiment is to extract tumor cells from the blood, thereby obtaining a blood image.

[0023] Specifically, blood samples are obtained by drawing blood from the patient through venous or finger-prick sampling; a small drop of blood is smeared onto a glass slide and flattened with another glass slide to create a blood slide; the blood slide is stained using the Giemsa staining method, and then placed under a microscope to obtain an image of the stained blood slide, which is recorded as a blood image. Figure 2 As shown. The Giemsa staining method is a well-known technique and will not be described in detail in this embodiment.

[0024] It should be noted that in blood images, normal cells and tumor cells may cluster together, and the difference between normal cells and tumor cells is smaller than the difference between tumor cells and non-cellular regions. This means that when segmenting tumor cells from blood images using region growing algorithms, some pixels of normal cells may be mistakenly segmented as tumor cells, making accurate tumor cell segmentation impossible. Therefore, this embodiment improves the region growing algorithm with a fixed growth threshold to obtain an adaptive growth threshold region growing algorithm. This adaptive growth threshold algorithm can then obtain more accurate tumor cell regions in blood images.

[0025] It should be further noted that the grayscale value difference between each pixel and its surrounding pixels in a blood image may vary, resulting in different probabilities of non-tumor cells growing at each growth point during region growth. Therefore, in the region growth algorithm for obtaining the adaptive growth threshold, the similarity between each pixel and its surrounding pixels is first calculated to obtain the adaptive growth threshold pixel. Since the number of cells clustered around each pixel may differ, the probability of non-tumor cells growing at each growth point during region growth also varies. In this embodiment, the first adaptive adjustment parameter for each adaptive growth threshold pixel is obtained by calculating the degree of cell clustering around it. Because cell clustering in blood images can present three scenarios—normal cells clustering together, normal cells clustering together with tumor cells, and tumor cells clustering together—and the probability of non-tumor cells growing at the growth point differs in these three scenarios, this embodiment calculates the probability of normal cells surrounding the adaptive growth threshold pixel after calculating the degree of cell clustering around each pixel. Based on this probability, the second adaptive adjustment parameter for the pixel is obtained. Finally, based on the first and second adaptive adjustment parameters of each adaptive growth threshold pixel, the adaptive growth threshold for each pixel is obtained, followed by a region growing algorithm for the adaptive growth threshold. Finally, the tumor cell region in the blood image is obtained using this region growing algorithm.

[0026] It should be further explained that when obtaining the similarity between a pixel and its surrounding pixels, it is based on the grayscale difference between each pixel and its surrounding pixels. Therefore, it is necessary to obtain the region to be analyzed for each pixel. When growing the growth point using the region growing algorithm, the pixels within the 8-neighborhood of the growth point are considered as the growth point. The region to be analyzed for each pixel is obtained based on the absolute value of the grayscale difference between each pixel and all pixels within its 8-neighborhood. The similarity between each pixel and its surrounding pixels is obtained based on the distribution of grayscale values ​​within the region to be analyzed for each pixel.

[0027] Specifically, taking the first blood image as an example... Taking a pixel as an example, let the g-th pixel in the blood image be denoted as... .by Get a center point The local region obtained is denoted as The local area. The length of the local area preset in this embodiment. This example is used for illustration; other values ​​can be set in other implementations. It should be noted that if... The surrounding pixels are insufficient to generate a The window displayed is subject to change and the actual situation shall prevail.

[0028] Furthermore, The average of the absolute values ​​of the grayscale differences of all pixels in the 8-neighborhood is used as The nearest threshold, to As the starting point for regional growth, The pixels in the 8-neighborhood direction are grown, and the first... Pixels grown in each neighborhood direction constitute The From one initial cluster, multiple initial clusters are obtained. The nearest neighbor threshold is used as the growth threshold, to The Each pixel in the initial cluster is used as a growth point, and growth is performed simultaneously. The grown pixels and the resulting pixels are then compared with the first cluster. The pixels in the initial clusters constitute The There are several clusters. All clusters as The region to be analyzed. Among them, Each pixel within the region to be analyzed can belong to only one cluster. The region within a local area that is not part of the region to be analyzed is called... The region outside the cluster.

[0029] Furthermore, the similarity between the g-th pixel and its surrounding pixels is... The calculation formula is as follows: ; Formula explanation; M represents the similarity between the g-th pixel and its surrounding pixels; M represents... The number of clusters; This represents the g-th pixel in the blood image; express The maximum Euclidean distance between a pixel in the m-th cluster and a pixel in the (m+1)-th cluster; express The average gray value of all pixels within the outer region of the cluster; express The average gray value of all pixels in the m-th cluster; express The standard deviation of grayscale values ​​of all pixels within the outer region of the cluster; express The standard deviation of grayscale values ​​of pixels within a local region; express The nearest neighbor threshold; express The number of pixels contained in the region to be analyzed; Represents the absolute value function; It is an exponential function with the natural constant as its base; this embodiment uses... The model is used to represent the inverse proportional relationship, with x as the input to the model. The implementer can set the inverse proportional function and the normalization function according to the actual situation. This represents a linear normalization function, where the normalization object is each pixel. .

[0030] It should be noted that, express The average Euclidean distance between each cluster and its surrounding clusters within a local region is used to reflect the proximity between the clusters. The smaller the average Euclidean distance, the closer the connection between the clusters. After normalization by exp(), the greater the proximity, indicating that the g-th pixel is more similar to its surrounding pixels. express Clustered pixels within a local region The difference between pixels in the region outside the cluster is determined by the average gray level difference between the cluster and the region outside the cluster, as well as the gray level standard deviation. The greater the gray level difference between the two, the greater the difference between them, and the less similar the g-th pixel is to its surrounding pixels.

[0031] The above steps are used to obtain the similarity between each pixel and its surrounding pixels.

[0032] Furthermore, set a similarity threshold. When the similarity between the g-th pixel and its surrounding pixels is greater than or equal to the similarity threshold When the time is right, the g-th pixel is the adaptive growth threshold pixel. The preset similarity threshold in this embodiment... Let's take this as an example.

[0033] Thus, the adaptive growth threshold pixels and the region to be analyzed for each pixel are obtained.

[0034] Step S002: Based on the number and position distribution of pixels in the cluster of adaptive growth threshold pixels, obtain the degree of cell aggregation around each adaptive growth threshold pixel, and then obtain the first adaptive adjustment parameter for each adaptive growth threshold pixel.

[0035] It should be noted that, since region growth is performed by growing each growth point towards the pixels in its 8-neighborhood, this embodiment obtains the degree of cell aggregation around each adaptive growth threshold pixel based on the distribution of pixels in the clusters within the region to be analyzed for each pixel.

[0036] Specifically, for each adaptive growth threshold pixel, all clusters are paired to obtain multiple cluster combinations. In each cluster combination, the cluster with the fewest pixels is designated as the minimum cluster; the cluster with the most pixels is designated as the maximum cluster. If two clusters in a cluster combination have the same number of pixels, one is designated as the minimum cluster and the other as the maximum cluster. The Euclidean distance between each pixel in the minimum cluster and each pixel in the maximum cluster is calculated, and the pixel in the minimum cluster corresponding to the smallest Euclidean distance is designated as the principal analysis point of that cluster combination. The Euclidean distance between the principal analysis point of that cluster combination and each boundary point in the maximum cluster of that cluster combination is used as several adjacency distances in that cluster combination, thus obtaining the minimum cluster containing the most pixels and the maximum cluster containing the most pixels. The product of the products, rounded up, is taken as the number of neighbors for that cluster combination. The neighborhood distance with the smallest number of neighbors among all the neighborhood distances of that cluster combination is taken as a subset of the neighborhood distances for that cluster combination. This embodiment uses a preset ratio threshold. This example is used for illustration; other values ​​can be set in other implementations. Simultaneously, the number of pixels contained in the minimum bounding rectangle of the region to be analyzed for each pixel is obtained, yielding the degree of cell aggregation around each adaptive growth threshold pixel. The specific calculation formula is as follows: ; In the formula, Indicates the first The degree of cell aggregation around the i-th adaptive growth threshold pixel; denoted as the i-th adaptive growth threshold pixel. ; express The number of clusters; express The average number of pixels contained in each cluster; express The number of cluster combinations; B represents The maximum number of neighbor distances in each cluster combination; express The (b+1)th nearest neighbor distance in the nth cluster combination; express The b-th nearest neighbor distance in the nth cluster combination; express The number of pixels in the region to be analyzed; express The number of pixels contained in the smallest bounding rectangle of the region to be analyzed; As an exponential function with the natural constant as the base, this embodiment uses the exp(-x) model to represent the inverse proportional relationship, where x is the input of the model. The implementer can set the inverse proportional function according to the actual situation.

[0037] It should be noted that, This indicates that the cluster distance of the current region to be analyzed is used to analyze the first... The degree of cell aggregation around an adaptive growth threshold pixel. express The number of clusters present in the region to be analyzed; the larger the value, the stronger the clustering. The greater the degree of aggregation of surrounding cells; express The average number of pixels contained in each cluster in the region to be analyzed; the larger the value, the more significant the difference. The greater the degree of aggregation of surrounding cells; This involves determining the distance between two clusters under different cluster combinations. The principal analysis point of the smaller cluster is used to analyze the Euclidean distance difference between corresponding points in the larger clusters. A smaller proximity distance between adjacent cluster combinations indicates... The greater the degree of aggregation of surrounding cells; This represents the ratio of the number of pixels in the region to be analyzed to the number of pixels in its bounding rectangle. A larger ratio indicates... The greater the degree of aggregation of surrounding cells.

[0038] Furthermore, the degree of cell aggregation around each adaptive growth threshold pixel is obtained using the above method. Then, based on the degree of cell aggregation around each adaptive growth threshold pixel, the first adaptive adjustment parameter for each adaptive growth threshold pixel is obtained, and its calculation formula is as follows: ; In the formula, Indicates the first The first adaptive adjustment parameter for each adaptive growth threshold pixel. Indicates the first The degree of cell aggregation around an adaptive growth threshold pixel. It is a linear normalization function that normalizes the degree of cell clustering around each adaptive growth threshold pixel.

[0039] It should be noted that the first The greater the degree of cell aggregation around the first adaptive growth threshold pixel, the better. The smaller the first adaptive adjustment parameter of each adaptive growth threshold pixel, the better.

[0040] Step S003: Based on the shape of each cluster of each adaptive growth threshold pixel and the gray value of each pixel in each cluster, the probability that each adaptive growth threshold pixel contains normal cells is obtained, and then the second adaptive adjustment parameter of each adaptive growth threshold pixel is obtained.

[0041] It should be noted that in blood images, cell aggregation can occur in three ways: normal cells clustering together, normal cells clustering together with tumor cells, and tumor cells clustering together. When the cell aggregation around a growth point differs, the adaptive growth threshold for that growth point should also differ. Therefore, this embodiment determines the probability that each adaptive growth threshold pixel is surrounded by normal cells based on the region to be analyzed; and based on this probability, a second adaptive adjustment parameter is obtained for each adaptive growth threshold pixel.

[0042] It should be further noted that in blood images, the chromatin staining state of normal cells is relatively similar, meaning that the gradient value of normal cell pixels is small, and the shape of normal cells is relatively regular, meaning that the edges of normal cells are relatively similar. Therefore, this embodiment determines the probability that each adaptive growth threshold pixel contains normal cells by using the gradient of each pixel within the analysis area of ​​each adaptive growth threshold pixel and the shape of the edge of each cluster.

[0043] Specifically, the gradient value of each pixel in the blood image is obtained using the Sobel operator. Chain codes are then acquired for each cluster of each adaptive growth threshold pixel, yielding the edge chain codes for each cluster. The Sobel operator and chain code acquisition are well-known techniques, and their specific methods are not described here. Based on the gradient, grayscale value of each pixel within the analysis region of each adaptive growth threshold pixel, and the edge chain codes of each cluster, the probability of normal cells surrounding each adaptive growth threshold pixel is calculated. The specific calculation formula is as follows: ; In the formula, Indicates the first The probability that the i-th adaptive growth threshold pixel contains normal cells; let the i-th adaptive growth threshold pixel be represented as... ; express The number of cluster combinations; K represents The number of chain code directions in the edge chain codes of each cluster, that is, the number of chain code values ​​contained in the chain code, which is the number of chain code value types here; express In the nth cluster combination, the edge chain code of the two clusters is the nth... The absolute value of the difference between the number of chain code values ​​in each direction; express The standard deviation of the gradient values ​​of pixels in each cluster. As an exponential function with the natural constant as the base, this embodiment uses the exp(-x) model to present the inverse proportional relationship and normalization processing. x is the input of the model, and the implementer can set the inverse proportional function and normalization function according to the actual situation.

[0044] It should be noted that, Reflecting the regularity of normal cells, the smaller the difference in the number of chain code types in each direction within a cluster combination, the more similar the edge shapes between the cluster combinations, i.e., the... The greater the adaptive growth threshold pixel, the more likely it is to contain normal cells around it. Normal cells exhibit a similar chromatin staining pattern. Because normal cells stain uniformly, the internal point gradient values ​​are relatively consistent. Therefore, the smaller the standard deviation, the more uniform the staining. The greater the likelihood that a pixel with an adaptive growth threshold will be surrounded by normal cells.

[0045] According to the The probability that a pixel with an adaptive growth threshold contains normal cells is obtained by calculating the probability of the first pixel being surrounded by normal cells. The second adaptive adjustment parameter for each adaptive growth threshold pixel is calculated as follows: ; In the formula: Indicates the first The second adaptive adjustment parameter for each adaptive growth threshold pixel. Indicates the first The probability that a pixel with an adaptive growth threshold contains normal cells.

[0046] Step S004: Based on the first adaptive adjustment parameter and the second adaptive adjustment parameter of each adaptive growth threshold pixel, obtain the comprehensive adaptive adjustment parameter of each pixel.

[0047] The first adaptive adjustment parameter and the second adaptive adjustment parameter of each adaptive growth threshold pixel are obtained through the above method. Then, the comprehensive adaptive adjustment parameter of each adaptive growth threshold pixel is obtained, and the calculation formula is as follows: ; In the formula, Indicates the first A comprehensive adaptive adjustment parameter for each adaptive growth threshold pixel; Indicates the first The first adaptive adjustment parameter for each adaptive growth threshold pixel; Indicates the first The second adaptive adjustment parameter for each adaptive growth threshold pixel.

[0048] Set the overall adaptive adjustment parameter of pixels that are not required to adaptively grow the threshold to 1.

[0049] Therefore, we can obtain the comprehensive adaptive adjustment parameters for each pixel in the image.

[0050] Step S005: Obtain the tumor region in the blood image by comprehensively and adaptively adjusting the parameters for each pixel.

[0051] Projecting in the image There are 10 seed points, and a fixed growth threshold is preset. The fixed growth threshold is multiplied by the comprehensive adaptive adjustment parameter of each pixel, and the product is used as the adaptive growth threshold of each pixel. The grown area is recorded as the tumor region in the blood image; wherein the fixed growth threshold preset in this embodiment is... The preset number of seed points Let's take this as an example.

[0052] This concludes the embodiment.

[0053] The present invention also proposes a visualization data optimization and identification system for hematologic tumor cells, including a memory, a processor, and a computer program stored in the memory and running on the processor. The processor executes the computer program stored in the memory to implement the steps of the aforementioned visualization data optimization and identification method for hematologic tumor cells.

[0054] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for optimizing the identification of blood tumor cells by visual data, characterized in that: The method comprises the following steps: Acquire a blood image and obtain a local area of ​​each pixel; obtain a number of clusters of each pixel and an area to be analyzed and an area outside the cluster of each pixel according to the grayscale values ​​of each pixel and surrounding pixels and the local area of ​​each pixel; obtain the similarity between each pixel and surrounding pixels according to the grayscale values ​​of pixels in the area to be analyzed and the area outside the cluster of each pixel, the position distribution of pixels in the cluster of each pixel and the number of pixels contained in the area to be analyzed; obtain an adaptive growth threshold pixel according to the similarity between each pixel and surrounding pixels; According to the number and position distribution of pixels in the cluster of the adaptive growth threshold pixel point, the degree of cell aggregation around each adaptive growth threshold pixel point is obtained, and then the first adaptive adjustment parameter of each adaptive growth threshold pixel point is obtained; According to the shape of each cluster of each adaptive growth threshold pixel and the gray value of each pixel in each cluster, the possibility of normal cells being contained around each adaptive growth threshold pixel is obtained, and then the second adaptive adjustment parameter of each adaptive growth threshold pixel is obtained; According to the first adaptive adjustment parameter and the second adaptive adjustment parameter of each adaptive growth threshold pixel point, a comprehensive adaptive adjustment parameter of each pixel point is obtained; According to the comprehensive adaptive adjustment parameters of each pixel, the tumor area in the blood image is obtained.

2. The method for optimizing the visual data recognition of blood tumor cells according to claim 1, characterized in that: The specific steps of obtaining the local area of ​​each pixel are as follows: The g-th pixel in the blood image is recorded as ,by As the center point, get a The local area is recorded as of the local area; The length of the preset local area.

3. The method for optimizing the visual data recognition of blood tumor cells according to claim 2, characterized in that: The method of obtaining a plurality of clusters of each pixel and an area to be analyzed and an area outside the cluster of each pixel according to the grayscale values ​​of each pixel and surrounding pixels and the local area of ​​each pixel includes the following specific steps: Will The average of the absolute values ​​of the differences between the grayscale values ​​of all pixels in the 8-neighborhood is taken as The neighboring threshold of by As the starting point of regional growth, The neighboring threshold is used as the growth threshold pair The pixel points in the 8 neighborhood directions are grown, and the The pixels that grow in the neighborhood direction constitute No. Initial clusters; by The neighboring threshold of is used as the growth threshold. No. Each pixel in the initial cluster is used as a growth point, and the growth is carried out at the same time. The grown pixel points and the first The pixels in the initial clusters are composed of No. clusters; All clusters of The area to be analyzed; Will The area in the local area that is not the area to be analyzed is recorded as out-of-cluster region.

4. The method for optimizing the visual data recognition of blood tumor cells according to claim 3, characterized in that: The similarity between each pixel and the surrounding pixels is obtained based on the grayscale values ​​of the pixels in the area to be analyzed and the area outside the cluster, the position distribution of the pixels in the cluster of each pixel, and the number of pixels contained in the area to be analyzed. The specific formula is as follows: ； In the formula, Indicates the similarity between the g-th pixel and the surrounding pixels; M indicates The number of clusters; express The maximum Euclidean distance between the pixel in the mth cluster and the pixel in the m+1th cluster; express The average gray value of all pixels in the outer area of ​​the cluster; express The average gray value of all pixels in the mth cluster; express The standard deviation of the gray values ​​of all pixels in the outer region of the cluster; express The standard deviation of the gray values ​​of all pixels in the local area; express The neighboring threshold of express The number of pixels contained in the area to be analyzed; represents the absolute value function; is an exponential function with a natural constant as base; Represents the normalization function.

5. The method for optimizing the visual data recognition of blood tumor cells according to claim 1, characterized in that: The specific steps of obtaining the adaptive growth threshold pixel point according to the similarity between each pixel point and the surrounding pixels are as follows: Setting the similarity threshold , when the similarity between the g-th pixel and the surrounding pixels is greater than or equal to the similarity threshold When , the g-th pixel is recorded as the adaptive growth threshold pixel.

6. The method for optimizing the visual data recognition of blood tumor cells according to claim 1, characterized in that: The method of obtaining the degree of cell aggregation around each adaptive growth threshold pixel point according to the number and position distribution of the pixel points in the cluster of the adaptive growth threshold pixel point, and then obtaining the first adaptive adjustment parameter of each adaptive growth threshold pixel point, comprises the following specific steps: For The clusters of pixels with adaptive growth thresholds are combined in pairs to obtain the Multiple cluster combinations of adaptive growth threshold pixels; In each cluster combination, the cluster with the least number of pixels is recorded as the smallest cluster; the cluster with the largest number of pixels is recorded as the largest cluster; Calculate the Euclidean distance between each pixel in the smallest cluster and each pixel in the largest cluster in any cluster combination, and record the pixel in the smallest cluster corresponding to the smallest Euclidean distance as the main analysis point of the cluster combination; The Euclidean distance between the main analysis point of the cluster combination and each boundary point in the largest cluster of the cluster combination is used as a number of adjacent distances in the cluster combination to obtain the number of pixels and the minimum number of pixels in the cluster combination. The product of is rounded up as the number of neighbors of the cluster combination, and the neighbor distance with the smallest number of neighbors among all the neighbor distances of the cluster combination is taken as the number of neighbor distances in the cluster combination; where R is equal to a constant 2; Get the The number of pixels contained in the minimum bounding rectangle of the area to be analyzed with the adaptive growth threshold pixel points; According to The neighboring distance in the cluster combination of the adaptive growing threshold pixel points, The number of pixels contained in the minimum bounding rectangle of the area to be analyzed for the adaptive growth threshold pixel point, and the The number of pixels contained in each cluster of the adaptive growth threshold pixel points is obtained. The degree of cell aggregation around the adaptive growth threshold pixel point is then obtained. The first adaptive adjustment parameter of the adaptive growth threshold pixel point.

7. The method for optimizing the visual data recognition of blood tumor cells according to claim 6, characterized in that: The obtained The degree of cell aggregation around the adaptive growth threshold pixel point is then obtained. The first adaptive adjustment parameter of the adaptive growth threshold pixel point includes the following specific formula: ； In the formula, Indicates The degree of cell aggregation around the adaptive growth threshold pixel point; the i-th adaptive growth threshold pixel point is represented as ; express The number of clusters; express The average number of pixels contained in each cluster; express The number of cluster combinations; B represents The maximum value of the number of neighboring distances in each cluster combination; express The b+1th neighbor distance in the nth cluster combination; express The bth neighbor distance in the nth cluster combination; express The number of pixels in the area to be analyzed; express The number of pixels contained in the minimum bounding rectangle of the area to be analyzed; is an exponential function with a natural constant as base; ； In the formula, Indicates The first adaptive adjustment parameter of the adaptive growth threshold pixel point, is the normalization function.

8. The method for optimizing the visual data recognition of blood tumor cells according to claim 7, characterized in that: The method of obtaining the possibility of normal cells around each adaptive growth threshold pixel point according to the shape of each cluster of each adaptive growth threshold pixel point and the gray value of each pixel point in each cluster cluster, and then obtaining the second adaptive adjustment parameter of each adaptive growth threshold pixel point, comprises the following specific steps: The gradient value of each pixel in the blood image is obtained by the Sobel operator, and the chain code is obtained for each cluster of each adaptive growth threshold pixel, and the edge chain code of each cluster is obtained; According to the gradient and grayscale of each pixel in the to-be-analyzed area of ​​each adaptive growth threshold pixel and the edge chain code of each cluster, the possibility of normal cells around each adaptive growth threshold pixel and the second adaptive adjustment parameter of each adaptive growth threshold pixel are obtained. The specific calculation formula is as follows: ； ； In the formula, Indicates The probability that there are normal cells around the adaptive growth threshold pixel; express The number of cluster combinations; K represents The number of chain code directions in the edge chain codes of each cluster; express The edge chain codes of the two clusters in the nth cluster combination The absolute value of the difference between the number of chain code values ​​in each direction; express The standard deviation of the gradient values ​​of pixels in each cluster; is an exponential function with a natural constant as base; Indicates The second adaptive adjustment parameter of the adaptive growth threshold pixel point.

9. The method for optimizing the visual data recognition of blood tumor cells according to claim 1, characterized in that: The step of obtaining the comprehensive adaptive adjustment parameter of each pixel point according to the first adaptive adjustment parameter and the second adaptive adjustment parameter of each adaptive growth threshold pixel point includes the following specific steps: ； In the formula, Indicates Comprehensive adaptive adjustment parameters of adaptive growth threshold pixels; Indicates A first adaptive adjustment parameter of an adaptive growth threshold pixel point; Indicates A second adaptive adjustment parameter of an adaptive growth threshold pixel point; The comprehensive adaptive adjustment parameter of the pixel points that are not the adaptive growth threshold pixels is set to 1.

10. A visual data optimization identification system for blood tumor cells, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that: When the computer program is executed by a processor, the steps of the method for optimizing the identification of visual data of blood tumor cells as described in any one of claims 1 to 9 are implemented.

Images