Remote-assisted intelligent blood cell recognition method and apparatus, and storage medium

Abstract

The present invention relates to the technical field of cell morphology detection. Disclosed are a remote-assisted intelligent blood cell recognition method and apparatus, and a storage medium. The method comprises the following steps: making a peripheral blood cell smear, and scanning a cell image; performing noise processing and isolated-point removal on the cell image on the basis of a first algorithm and a second algorithm; segmenting the cell image into a background area and a target area on the basis of a grayscale value of the cell image; obtaining color features of the cell image on the basis of a color page model, and processing the color features by means of both boundary transformation and morphological transformation, so as to obtain boundary features and morphological features of cells in the cell image; and performing recognition on the boundary features and the morphological features on the basis of a constructed recognition space, and outputting matched cell types. In the present application, an optimization process is added in a cell recognition process, and a coverage problem is taken into consideration from a local area first, and is then extended from the local to the entity, thereby ensuring the comprehensiveness of recognition, and improving the recognition precision.

Description

A remotely assisted blood cell intelligent identification method, device and storage medium Technical Field

[0001] The present invention relates to the technical field of cell morphology detection, and in particular to a remotely assisted intelligent blood cell identification method, device and storage medium. Background Art

[0002] Peripheral blood samples contain three types of visible blood cells: red blood cells, white blood cells, and platelets. White blood cells are further divided into granulocytes, monocytes, and lymphocytes based on the differences in the granules contained in their nuclei and protoplasm. Granulocytes are further divided into eosinophils, basophils, and neutrophils based on the properties of their granules. The percentage of each type of white blood cell relative to the total white blood cell count is a key component of peripheral blood tests. After Wright's staining, the types of white blood cells are distinguished based on the microscopic appearance of mature white blood cells, including nuclear staining, cytoplasmic staining, and nuclear shape.

[0003] Qualitative analysis of cell images using microscopy and visual methods to diagnose health conditions has long played a crucial role in clinical pathology. With the advancement of computer technology, image processing and analysis techniques are playing an increasingly important role in clinical diagnosis and treatment. The development of new image analysis systems that automatically process immunohistochemical cell images and perform quantitative analysis, thus assisting physicians in making rapid and accurate judgments, holds significant promise for the diagnosis of medical diseases. However, current image analysis systems are not sufficiently accurate for identifying and segmenting overlapping cells, often requiring manual intervention by physicians.

[0004] Patent application number CN202011399853.9 discloses a method for morphological analysis and counting of peripheral blood cells, comprising the following steps: S1: Scanning a smear and preprocessing; S2: Dividing the scanned image into several regions; S3: Setting different color value zones for different cells and collecting a color value distribution profile of the RGB three-color space of the counting region image; S4: Retaining the a1 color value zone in the counting region and setting the remaining pixels to black to obtain and store a separate A1 cell image b1; S5: Running the C-means clustering algorithm on each image to segment the image; S6: Counting cells in the segmented image; S7: Calculating data for each type of cell, comparing the data with normal reference values, establishing associations with existing diseases, and performing associative analysis. The accuracy of this image processing method depends on the quality of the scanned image, resulting in low cell recognition accuracy.

[0005] Summary of the Invention

[0006] In view of this, an object of the present invention is to provide a remote-assisted intelligent blood cell identification method, device and storage medium to solve the problem of low cell identification accuracy in existing image analysis systems.

[0007] The present invention solves the above technical problems through the following technical means:

[0008] In a first aspect, the present application provides a remotely assisted blood cell intelligent identification method, comprising the following steps:

[0009] S110: Prepare peripheral blood cell smears and scan cell images;

[0010] S120: performing noise processing and isolated point removal on the cell image based on the first algorithm and the second algorithm;

[0011] S130: Segmenting the cell image into a background area and a target area based on the grayscale value of the cell image;

[0012] S140: obtaining color features of the cell image based on a color page model, and processing the color features through boundary transformation and morphological transformation respectively to obtain boundary features and morphological features of cells in the cell image;

[0013] S150: Identify the boundary features and the morphological features based on the constructed recognition space, and output matching cell types.

[0014] In conjunction with the first aspect, in some optional embodiments, the pixels of the cell image are set to A times B, the center point is represented as (x, y), and the first algorithm includes:

[0015] Where S represents the coordinate set of all points in the cell image, M represents all coordinate points, f(x, y) describes the noise points in the cell image, and g(x, y) represents the noise points after smoothing. The second algorithm can remove the isolated points in the cell image, including:

[0016] Where E is the cell image and X is the calculation element.

[0017] In conjunction with the first aspect, in some optional implementations, the color page model includes:

[0018] Among them, x', y' and z' are the color features of any point in the cell image, and R, G, and B are the color channel intensities of the corresponding points.

[0019] In combination with the first aspect, in some optional embodiments, the boundary transformation includes: connecting the global features of the cell image and the boundary pixels of the target part features to form a closed area, establishing a histogram of the size and direction of the boundary part of the closed area, and obtaining the boundary features based on the histogram.

[0020] In combination with the first aspect, in some optional embodiments, the morphological transformation includes extracting the morphological features based on the grayscale co-occurrence matrix, that is, predicting the probability that a pixel point with a first grayscale value will generate a second grayscale value in the target length direction and target direction.

[0021] In combination with the first aspect, in some optional embodiments, the feature set and center function of the cell image in the recognition space are transformed into a spatial plane to map the cell image in the space without covering the center, construct a covering area, complete the recognition of the boundary features and the morphological features, and output the cell type.

[0022] In a second aspect, the present application provides a remotely assisted intelligent blood cell identification device, comprising:

[0023] Input unit: making peripheral blood cell smears and scanning cell images;

[0024] A first processing unit is configured to perform noise processing and isolated point removal on the cell image based on a first algorithm and a second algorithm;

[0025] A second processing unit is configured to segment the cell image into a background area and a target area based on a grayscale value of the cell image;

[0026] A third processing unit is configured to obtain color features of the cell image based on a color page model, and process the color features through boundary transformation and morphological transformation to obtain boundary features and morphological features of cells in the cell image;

[0027] Recognition unit: recognizes the boundary features and the morphological features based on the constructed recognition space, and outputs the matching cell types.

[0028] In a third aspect, the present application provides a computer storage medium, wherein the computer storage medium stores a computer program, and when the computer program is run on a computer, it can execute the remote-assisted intelligent blood cell identification method as described above.

[0029] Beneficial effects of the present invention:

[0030] 1. This application adds an optimization process in the cell recognition process, and first considers the coverage problem from the local area, and then extends from the local to the whole, ensuring the comprehensiveness of recognition and improving the recognition accuracy.

[0031] 2. This application combines variables such as matrix contrast and correlation to complete the extraction of cell morphological features, which can maximize the reflection of the morphological characteristics of pixels on the cell image, and reduce the phenomenon of changes in resolution and feature extraction accuracy caused by the susceptibility of cell morphology to the influence of conditions such as lighting when identifying single pixels.

[0032] 3. This application removes image noise from cell images, achieves background separation of cell images, and further improves the image quality of cell images. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] FIG1 is a schematic diagram of a flow chart of a remotely assisted intelligent blood cell identification method according to the present invention;

[0034] FIG2 is a schematic diagram of a remote-assisted intelligent blood cell identification device according to the present invention;

[0035] in,

[0036] 10: input unit; 20: first processing unit; 30: second processing unit; 40: third processing unit; 50: recognition unit. DETAILED DESCRIPTION

[0037] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0038] As shown in FIG1 , a remote-assisted intelligent blood cell identification method of the present invention includes the following steps:

[0039] S110: Prepare peripheral blood cell smears and scan cell images;

[0040] S120: performing noise processing and isolated point removal on the cell image based on the first algorithm and the second algorithm;

[0041] S130: Segmenting the cell image into a background area and a target area based on the grayscale value of the cell image;

[0042] S140: obtaining color features of the cell image based on a color page model, and processing the color features through boundary transformation and morphological transformation respectively to obtain boundary features and morphological features of cells in the cell image;

[0043] S150: Identify the boundary features and the morphological features based on the constructed recognition space, and output matching cell types.

[0044] In step S110, a peripheral blood cell smear is prepared and a cell image is scanned. The peripheral blood cell smear is prepared by placing a drop of collected peripheral blood 1 cm from one end of a glass slide. The left hand holds the glass slide flat, and the right hand holds the smear, approaching the blood drop from the back or front. The blood is spread out along the edge of the smear to an appropriate width. The smear is then immediately angled at a 30-45° angle to the glass slide, and the edge of the smear is gently pressed to form a blood smear of appropriate thickness. After the blood film dries, it is stained with a dye, typically using Wright's stain, and rinsed. This is how the Wright-stained image of peripheral blood cells observed under a microscope is produced. After the peripheral blood cell smear is magnified by a microscope, data is transmitted using a digitizing device, such as a color CCD camera or image acquisition card. The digitized image is then stored in a computer. The computer can then process the peripheral blood cell image, thereby enabling the application of digital image processing technology to peripheral blood cell images.

[0045] In step S120, noise processing and isolated point removal are performed on the cell image based on the first algorithm and the second algorithm. The pixel size of the cell image is set to A times B, the center point is represented as (x, y), and the points around the center point are called the perimeter. The average grayscale value of all pixels in the perimeter of the center point (x, y) is calculated and the initial pixel value of the center point (x, y) is replaced. The expression of the first algorithm is:

[0046] Where x = 0, 1, ...A; y = 0, 1, ...B-1, S represents the coordinate set of all points in the cell image, M represents all coordinate points, f(x, y) describes the noise points in the cell image, and g(x, y) represents the noise points after smoothing.

[0047] The second algorithm is developed from the corrosion and expansion operations. Its main function is to remove isolated points, reduce protrusions, and blur the boundaries. It is used to adjust the boundaries of cell colonies in cell images. The second algorithm includes:

[0048] The two expressions above are in a dual relationship. The first expression represents the calculation element X processing the cell image E. The elements in the structure set obtained after the calculation must be greater than the elements before the calculation. The second expression is the opposite.

[0049] In step S130, the cell image is segmented into a background area and a target area based on the grayscale value of the cell image. In this application, the image is segmented into two parts, the background area and the target area, in combination with the grayscale value of the cell image, which is more conducive to image recognition. The inter-class variance of the grayscale value between the background area and the target area determines the error rate of the two. The larger the variance, the more significant the difference between the two. The segmentation process is as follows: Set the grayscale value of the target image, usually 0-256, assuming that the coordinates of a certain point in the image are (x, y), and the calculation formula of its perimeter average grayscale value D(x, y) is as follows:

[0050] Where C represents the perimeter width and F(x, y) is the probability of the image appearing. If the image pixels are represented as A times B, the probability density of the center point (x, y) is expressed as:

[0051] In this application, a two-dimensional histogram is used to describe the relationship between the target and the background. Assume that (U', T') is the threshold in the histogram, G' and K' represent the background and noise respectively, V and W represent the edge and the target respectively, and the probability of the background area and the target area appearing is recorded as E0 and F1 respectively. Then

[0052] Without considering the influence of noise, the sum of E0 and F1 is equal to 1, then the overall mean expression of the target area and the background area is:

[0053] Therefore, the threshold expression between the target area and the background area is obtained:

[0054] The above algorithm can be used to calculate all the pixels in the cell image, thereby achieving separation of the target area and the background area.

[0055] In step S140, the color features of the cell image are obtained based on the color page model. The color features are processed through boundary transformation and morphological transformation respectively to obtain the boundary features and morphological features of the cells in the cell image. The number of different tobacco pixels relative to the total number of pixels in the entire image is described in the form of a histogram, where the horizontal axis of the histogram represents the color level and the vertical axis represents the color ratio. The color page model using RGB (red, green, and blue color channels) to extract color can be expressed as:

[0056] Among them, x', y' and z' are the color features of any point in the cell image, and R, G, and B are the color channel intensities of the corresponding points.

[0057] The current trend in peripheral blood cell testing is to leverage relevant experience, combine cell boundaries and morphological features to distinguish cell types, and then transform existing classification methods into computer-readable language, using artificial intelligence to provide cell judgment results. When performing boundary characterization, the obtained results are used as shape feature parameters. Using a Hough transform, the boundary pixels of the global feature and the target part feature are connected to form a closed region. A histogram of the size and direction of the boundary section is constructed, and the cell boundary characteristics can be obtained based on this histogram.

[0058] Morphological changes are extracted based on the grayscale co-occurrence matrix. This predicts the probability that a pixel with a first grayscale value L will generate a second grayscale value L' along the target length direction d and the target direction θ. This can also be considered as a prediction of the probability density of grayscale values ​​in a cell image. The co-occurrence matrix can best reflect the morphological characteristics of cells. This application combines matrix contrast, correlation, energy, and other variables to achieve texture feature extraction.

[0059] It is known that the pixels of a cell image E are A×B and the gray level is set to H' level, then the co-occurrence matrix of the cell image is expressed as

[0060] P(x,y,d,θ)=P(I(l,w))

[0061] =(l+d cos(θ), w+dsin(θ)=y)

[0062] Where I(l, w) represents the grayscale of the pixel (l, w), d is the distance between two points (x, y), and θ is the angle between them. If P is a matrix with an H'×H' structure, once d and θ are determined, the co-occurrence matrix at different angles can be calculated by combining the distance and angle between the two points. Finally, the different grayscale co-occurrence matrices are normalized to extract morphological features.

[0063] In step S150, the boundary features and the morphological features are identified based on the constructed recognition space, and the matching cell types are output. In the recognition space, the feature set and center function of the cell image are transformed into the space plane, and the cell image is mapped to the top coverage center in the space to construct the coverage area, complete the recognition of the boundary features and the morphological features, and output the cell type. If the input set of a straight image is represented by K = [(x1, y1), (x2, y2), ... (x p ,y p )], where k represents the set of cell image pixels in n-dimensional space, and the number of samples in the set is p. Divide k into ξ subsets and construct a three-layer network N * , so that it meets the following requirements: After passing through the network, it belongs to K i’ The output value of each point is y i’, and y i’ =(0,1,0,...0), that is, the i-th ’ The first component is equal to 1, and the other components are equal to 0. The process of using the center function coverage algorithm to identify cell images is as follows:

[0064] T ′ :O→Z, x∈O, [T′(x),T′(y)]=K(x,y)

[0065] The above conversion process means: in the n-dimensional hyperplane O, the transformation T' maps the points on the O plane into the P recognition space, then the recognition space input set is represented as P T’ .

[0066] In the recognition space, the first a values ​​of the output set Y are set to be different. After the above changes, the coverage C in the recognition space can be calculated. i’ The acquisition process is as follows: select any x in the sample set i’ , combined with x i’ Create coverage C i’ , the center of this cover is x i’ . Delete P after creating the coverage T’ The points that have been covered in P T’ A coverage center is selected within the image. After multiple operations, until all points are deleted, the entire coverage neighborhood is established, thus completing the identification of cell types in all cell images. This method adds an optimization process and first considers the coverage problem from a local area, then expands from the local area to the whole area, ensuring comprehensive recognition and improving recognition accuracy.

[0067] As shown in FIG2 , the present application further provides a remote-assisted intelligent blood cell identification device, which includes at least one software function module stored in a storage module in the form of software or firmware or in an operating system (OS) embedded in a control device. Processing is used to execute executable modules stored in the storage module, such as software function modules and computer program modules included in the remote-assisted intelligent blood cell identification device.

[0068] The apparatus includes an input unit 10 , a first processing unit 20 , a second processing unit 30 , a third processing unit 40 and a recognition unit 50 .

[0069] Input unit 10: making peripheral blood cell smears and scanning cell images;

[0070] A first processing unit 20 is configured to perform noise processing and isolated point removal on the cell image based on a first algorithm and a second algorithm;

[0071] The second processing unit 30 is configured to segment the cell image into a background area and a target area based on the grayscale value of the cell image;

[0072] The third processing unit 40 obtains color features of the cell image based on the color page model, processes the color features through boundary transformation and morphological transformation, and obtains boundary features and morphological features of cells in the cell image;

[0073] Recognition unit 50: recognizes the boundary features and the morphological features based on the constructed recognition space, and outputs the matching cell types.

[0074] When the device is actually used, it is set on the client side, and the recognition results can be sent to the remote auxiliary terminal through the server side in the form of sampling review. The remote auxiliary terminal controlled by the expert verifies the recognition results. For example, in the cell images recognized in the same batch, 5% of the recognition results are randomly selected for verification. When the expert verification results are all accurate or close to accurate, the cell image recognition results of this batch can be output to the next step. When the expert finds an error in the recognition result during verification, the error can be marked and sent to the client side for calibration.

[0075] In this embodiment, the storage module may be, but is not limited to, a random access memory, a read-only memory, a programmable read-only memory, an erasable programmable read-only memory, an electrically erasable programmable read-only memory, or the like. In this embodiment, the storage module may be used to store the identification space in the identification unit 50, for example. Of course, the storage module may also be used to store programs, which the processing module executes upon receiving execution instructions. The acquisition, processing, and identification of cell images can be referenced to the aforementioned remotely assisted intelligent blood cell identification method and will not be further elaborated upon here.

[0076] An embodiment of the present application also provides a computer storage medium, in which a computer program is stored. When the computer program runs on a computer, the computer executes the remote-assisted intelligent blood cell identification method as described in the above embodiment.

[0077] Through the description of the above implementation methods, those skilled in the art can clearly understand that the present application can be implemented through hardware or by means of software plus a necessary general hardware platform. Based on this understanding, the technical solution of the present application can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (which can be a CD-ROM, a USB flash drive, a mobile hard disk, etc.), including a number of instructions for enabling a computer device (which can be a personal computer, an electronic device, or a network device, etc.) to execute the methods described in each implementation scenario of the present application.

[0078] In summary, the present application provides a remote-assisted intelligent blood cell identification method, device, and storage medium, the method comprising the following steps: making a peripheral blood cell smear and scanning a cell image; performing noise processing and isolated point removal on the cell image based on a first algorithm and a second algorithm; segmenting the cell image into a background area and a target area based on the grayscale value of the cell image; obtaining the color features of the cell image based on a color page model, processing the color features through boundary transformation and morphological transformation respectively, and obtaining the boundary features and morphological features of the cells in the cell image; identifying the boundary features and morphological features based on a constructed recognition space, and outputting the matching cell types. The present application adds an optimization process to the cell recognition process, and first considers the coverage problem from the local area, and then extends from the local to the whole, thereby ensuring the comprehensiveness of the recognition and improving the recognition accuracy.

[0079] The above embodiments are intended only to illustrate the technical solutions of the present invention and are not intended to limit the scope of the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art will appreciate that the technical solutions of the present invention may be modified or replaced with equivalents without departing from the spirit and scope of the technical solutions of the present invention, and such modifications or equivalents shall be encompassed by the claims of the present invention. Any techniques, shapes, and structures not described in detail herein are well known.

Claims

1. A remote-assisted intelligent blood cell identification method, characterized in that: The following steps are involved: S110: Prepare peripheral blood cell smears and scan cell images; S120: performing noise processing and isolated point removal on the cell image based on the first algorithm and the second algorithm; S130: Segmenting the cell image into a background area and a target area based on the grayscale value of the cell image; S140: obtaining color features of the cell image based on a color page model, and processing the color features through boundary transformation and morphological transformation respectively to obtain boundary features and morphological features of cells in the cell image; S150: Identify the boundary features and the morphological features based on the constructed recognition space, and output matching cell types.

2. The remote-assisted intelligent blood cell identification method according to claim 1, characterized in that: The pixels of the cell image are set to A times B, and the center point is represented as (x, y). The first algorithm includes: Wherein S represents the coordinate set of all points in the cell image, M represents all coordinate points, f(x, y) describes the noise points in the cell image, g(x, y) represents the noise points after smoothing, and the second algorithm can remove the isolated points in the cell image, including: Among them, E is the cell image and X is the calculation element.

3. The remote-assisted intelligent blood cell identification method according to claim 2, characterized in that: The color page model includes: Among them, x', y' and z' are the color features of any point in the cell image, and R, G, B are the color channel intensities of the corresponding points.

4. The remote-assisted intelligent blood cell identification method according to claim 3, characterized in that: The boundary transformation includes: connecting the boundary pixels of the global feature of the cell image and the target part feature to form a closed area, establishing a histogram of the size and direction of the boundary part of the closed area, and obtaining the boundary feature based on the histogram.

5. The remote-assisted intelligent blood cell identification method according to claim 4, characterized in that: The morphological transformation includes extracting the morphological features based on a grayscale co-occurrence matrix, that is, predicting the probability of a pixel point with a first grayscale value generating a second grayscale value in a target length direction and a target direction.

6. The remote-assisted intelligent blood cell identification method according to claim 5, characterized in that: In the recognition space, the feature set and center function of the cell image are transformed in the space plane to map the cell image to the top coverage center in the space, construct a coverage area, complete the recognition of the boundary features and the morphological features, and output the cell type.

7. A remote-assisted intelligent blood cell identification device, characterized in that: include: Input unit (10): making peripheral blood cell smears and scanning cell images; A first processing unit (20) performs noise processing and isolated point removal on the cell image based on a first algorithm and a second algorithm; A second processing unit (30): segmenting the cell image into a background area and a target area based on the grayscale value of the cell image; A third processing unit (40) is configured to obtain color features of the cell image based on a color page model, and process the color features through boundary transformation and morphological transformation to obtain boundary features and morphological features of cells in the cell image; Recognition unit (50): recognizes the boundary features and the morphological features based on the constructed recognition space, and outputs the matching cell types.

8. A computer storage medium, characterized in that The computer storage medium stores a computer program, and when the computer program is run on a computer, the remote-assisted intelligent blood cell identification method according to any one of claims 1 to 6 can be executed.

Images