Pathological image cell contour display system based on deep gray level variation technology

Abstract

The application discloses a pathological image cell contour display system based on deep gray level change technology, which comprises a first acquisition module, a preprocessing module, a deep gray level change processing module, an identification module, a post-processing module and a visualization module.

Description

Technical Field

[0001] This invention relates to the field of image processing technology, and in particular to a system for displaying cell contours in pathological images based on deep grayscale variation technology. Background Technology

[0002] Digital pathology images (whole slide image, WSI) are high-resolution digital images acquired through scanning with a fully automated microscope or optical magnification system. These images are then processed and stitched together seamlessly across multiple fields of view using a computer, resulting in multi-level visualization. Currently, accurate cell contour extraction is crucial for disease diagnosis and treatment in digital pathology image analysis. However, due to the complexity and noise interference of digital pathology images, accurate cell contour extraction is a challenging task. Traditional image processing methods are often affected by factors such as image noise, lighting variations, and cell morphological diversity, leading to inaccurate or unstable extraction results.

[0003] Existing methods for contour recognition in digital pathology images include: 1. Contour extraction based on traditional image processing methods: Traditional image processing methods such as thresholding and edge detection can be used to extract cell contours. However, due to factors such as image noise and illumination variations, the extraction results may be inaccurate or unstable. 2. Cell contour recognition based on machine learning: Using machine learning algorithms, such as Support Vector Machines (SVM) and Random Forests, models can be trained to automatically recognize and extract cell contours. This method can improve extraction accuracy by learning from a large amount of labeled image data, but it may have certain limitations for complex image scenes and situations with significant cell morphological variations. 3. Image segmentation based on deep learning: Deep learning technologies, such as Convolutional Neural Networks (CNN) and U-Net, have achieved significant results in the field of image segmentation. By using deep learning models, cell contour features in images can be automatically learned and accurately segmented. This method is complex to operate, expensive, and not widely used, generally suitable for scientific research. Summary of the Invention

[0004] This invention aims to at least partially solve one of the technical problems in the aforementioned technologies. Therefore, the purpose of this invention is to propose a system for displaying cell contours in pathological images based on deep grayscale variation technology. This system facilitates improved accuracy and efficiency in contour recognition, supports complex image scenes, possesses strong analytical capabilities, can adapt to complex image scenes and cell morphology changes, and enhances the processing capability for diverse pathological images.

[0005] To achieve the above objectives, embodiments of the present invention propose a system for displaying cell contours in pathological images based on deep grayscale variation technology, comprising: The first acquisition module is used to acquire digital pathology images; The preprocessing module is used to preprocess digital pathology images to obtain preprocessed images; The deep grayscale transformation processing module is used to process the preprocessed image based on the deep grayscale transformation model to obtain the transformed image; The recognition module is used to perform cell contour recognition on the transformed image and determine the contour of each cell. The post-processing module is used to post-process the cell contours and obtain the post-processing results; The visualization module is used to visualize the post-processing results, obtain the target image, and display it.

[0006] According to some embodiments of the present invention, the preprocessing module includes: The noise reduction module is used to process digital pathological images to obtain denoised images; The brightness adjustment module is used to adjust the brightness of the denoised image to obtain the adjusted image. The enhancement module is used to perform contrast enhancement processing on the adjusted image to obtain a preprocessed image.

[0007] According to some embodiments of the present invention, the identification module includes: The first determining module is used to convert the transformed image into an n*m pixel matrix and perform rasterization processing, with each pixel corresponding to a raster, to obtain an n*m raster matrix, and to establish a coordinate system to determine the coordinate information of each raster. The calculation module is used to calculate the Gaussian coefficient of each grid cell and the angle of its direction. The judgment module is used to determine, in the coordinate system, whether the gray value of the pixel corresponding to each current grid is greater than the gray value of the pixel corresponding to the adjacent grid in the direction of the current grid; if so, the gray value of the pixel corresponding to the current grid remains unchanged; otherwise, the gray value of the pixel corresponding to the current grid is determined to be 0. The second determining module is used to determine the maximum gray value at the angle of the direction of each grid, set the gray value of the pixels of all other grids to 0, and connect the pixels corresponding to the maximum gray value based on the double threshold algorithm to determine the outline of each cell.

[0008] According to some embodiments of the present invention, the post-processing module includes: The second acquisition module is used to acquire parameter information of the contour of each cell; the parameter information includes contour number and contour geometric parameters; The first elimination module is used to match the geometric parameters of the contour with the preset geometric parameters according to the contour number, eliminate the contours with a matching degree lower than the preset threshold, and re-number them to obtain the correction parameter information of the contours of the remaining cells. The second elimination module is used to construct an outer processing box for each cell based on the correction parameter information, thereby obtaining an outer processing box parameter set; construct a corresponding kd-tree index based on the outer processing box parameter set, and use the outer processing box filtering conditions to traverse the kd-tree index to filter out outer processing boxes without adjacency relationships and eliminate them, thereby obtaining a set of neighboring cells; the set of neighboring cells is a set of cells with adjacency relationships; the adjacency relationships include external tangency, intersection, internal tangency, and inclusion relationships; The third determining module is used for: Cells with intersecting, tangent, or inclusive relationships are selected from adjacent cell sets and merged to obtain the first treatment result; Cells with ectocervical relationships were selected from adjacent cell sets and subjected to tangential enhancement treatment to obtain the second treatment result; Based on the first and second processing results, the post-processing result is determined.

[0009] According to some embodiments of the present invention, the second rejection module includes: The fourth determination module is used to determine the feature points of the outline of each cell based on the correction parameter information; The building module is used to construct an external processing box based on feature points.

[0010] According to some embodiments of the present invention, the visualization module includes: The segmentation module is used to obtain the contour information of cells in the processed image corresponding to the post-processing result, perform mask processing, and perform multi-scale segmentation to obtain the segmentation result; The scene recognition module is used to perform scene recognition based on the scene recognition model according to the segmentation results, and obtain cell detection results at each segmentation scale. The labeling module is used to fuse cell detection results at each segmentation scale, determine the cell category of the corresponding cells, and label them in the processed image. The visualization processing module is used to visualize the processed images labeled with cell categories, obtain the target image, and display it.

[0011] According to some embodiments of the present invention, the marking module includes: The third acquisition module is used for: The cell detection results at each segmentation scale were obtained and numerically processed to obtain the first value; The feature information of corresponding pixels in the cell detection results at each segmentation scale during the fusion process is obtained and numerically processed to obtain the second value. The fifth determining module is used for: Each first value is compared with the second value to determine the segmentation scale with the greatest change and the segmentation scale with the smallest change. Based on the segmentation scale with the greatest and least change, a preset data table is queried to determine the cell category of the corresponding cell and mark it in the processed image.

[0012] According to some embodiments of the present invention, the visualization processing module includes: The sixth determination module is used to parse the processed image with labeled cell categories and determine the parsing information; The allocation module is used to allocate and process the target image based on the display elements set in the visualization model according to the parsed information, so as to obtain the target image and display it.

[0013] According to some embodiments of the present invention, the display elements include at least one of bar charts, pie charts, and line charts.

[0014] Beneficial effects 1. Improve the accuracy of cell outlines: By inferring the color values ​​of the image, grayscale background noise in the image can be accurately removed, thereby improving the accuracy of cell outlines.

[0015] 2. Improve extraction efficiency: By adopting deep grayscale transformation technology, digital pathological images can be processed automatically, greatly reducing the need for manual intervention and improving processing efficiency.

[0016] 3. Supports complex image scenes: It has strong resolution capabilities and can adapt to complex image scenes and changes in cell morphology, improving the ability to process diverse pathological images.

[0017] 4. Provides visualization: By displaying and visualizing cell outlines, it can help doctors or pathologists observe and analyze digital pathology images more clearly, assisting in disease diagnosis and treatment decisions.

[0018] Other features and advantages of the invention will be set forth in the following description, and will be apparent in part from the description, or may be learned by practicing the invention. The objects and other advantages of the invention may be realized and obtained by means of the structures particularly pointed out in the written description and the accompanying drawings.

[0019] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0020] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings: Figure 1 This is a block diagram of a pathological image cell contour display system based on deep grayscale change technology according to an embodiment of the present invention; Figure 2 This is a block diagram of a preprocessing module according to an embodiment of the present invention; Figure 3 This is a block diagram of an identification module according to an embodiment of the present invention. Detailed Implementation

[0021] The preferred embodiments of the present invention will be described below with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are for illustration and explanation only and are not intended to limit the present invention.

[0022] like Figure 1 As shown, this embodiment of the invention proposes a system for displaying cell contours in pathological images based on deep grayscale variation technology, comprising: The first acquisition module is used to acquire digital pathology images; The preprocessing module is used to preprocess digital pathology images to obtain preprocessed images; The deep grayscale transformation processing module is used to process the preprocessed image based on the deep grayscale transformation model to obtain the transformed image; The recognition module is used to perform cell contour recognition on the transformed image and determine the contour of each cell. The post-processing module is used to post-process the cell contours and obtain the post-processing results; The visualization module is used to visualize the post-processing results, obtain the target image, and display it.

[0023] The working principle of the above technical solution is as follows: In this embodiment, the digital pathological image is preprocessed, such as denoising, adjusting the image brightness and contrast, to obtain a preprocessed image, which facilitates the improvement of the effect of subsequent processing.

[0024] In this embodiment, the deep grayscale transformation processing module employs a deep learning model or algorithm to process the preprocessed image and remove grayscale background noise. The deep grayscale transformation model can learn image features by training on a large amount of digital pathology image data, thereby accurately removing background noise, facilitating the elimination of its influence, and simultaneously highlighting cell contours, thus improving the accuracy and efficiency of cell contour extraction. Deep grayscale transformation technology, in this context, refers to a processing technique for digital pathology images used to adjust grayscale background noise in an image to highlight cell contours. It employs a deep learning model or algorithm, utilizing neural networks and other technologies to learn and infer features from the image, thereby achieving the adjustment of the grayscale background.

[0025] In this embodiment, the recognition module is used to perform cell contour recognition on the transformed image, determine the contours of each cell, and display the cell contours after removing the grayscale background. Edge detection algorithms or other image segmentation techniques may be used to extract the cell contours.

[0026] In this embodiment, post-processing includes removing irregular contours and merging contacting cells, which facilitates the organization and analysis of cell contour data, thereby enabling visualization processing.

[0027] In this embodiment, the target image is the image after visualization processing.

[0028] The beneficial effects of the above technical solution are as follows: Improved accuracy of cell contours: By inferring image color values, grayscale background noise in the image can be accurately removed, thereby improving the accuracy of cell contours. Increased extraction efficiency: Utilizing deep grayscale transformation technology, digital pathological images can be processed automatically, greatly reducing the need for manual intervention and improving processing efficiency. Support for complex image scenes: It has strong analytical capabilities, can adapt to complex image scenes and changes in cell morphology, and improves the ability to process diverse pathological images. Provided visualization: By displaying and visualizing cell contours, doctors or pathologists can more clearly observe and analyze digital pathological images, assisting in disease diagnosis and treatment decisions.

[0029] like Figure 2 As shown, according to some embodiments of the present invention, the preprocessing module includes: The noise reduction module is used to process digital pathological images to obtain denoised images; The brightness adjustment module is used to adjust the brightness of the denoised image to obtain the adjusted image. The enhancement module is used to perform contrast enhancement processing on the adjusted image to obtain a preprocessed image.

[0030] The beneficial effects of the above technical solution are: noise reduction, brightness adjustment and contrast enhancement of digital pathological images, elimination of noisy pixels in digital pathological images, achieving brightness balance and contrast enhancement, improving the accuracy of digital pathological images, and thus facilitating accurate recognition of cell contours.

[0031] like Figure 3 As shown, according to some embodiments of the present invention, the identification module includes: The first determining module is used to convert the transformed image into an n*m pixel matrix and perform rasterization processing, with each pixel corresponding to a raster, to obtain an n*m raster matrix, and to establish a coordinate system to determine the coordinate information of each raster. The calculation module is used to calculate the Gaussian coefficient of each grid cell and the angle of its direction. The judgment module is used to determine, in the coordinate system, whether the gray value of the pixel corresponding to each current grid is greater than the gray value of the pixel corresponding to the adjacent grid in the direction of the current grid; if so, the gray value of the pixel corresponding to the current grid remains unchanged; otherwise, the gray value of the pixel corresponding to the current grid is determined to be 0. The second determining module is used to determine the maximum gray value at the angle of the direction of each grid, set the gray value of the pixels of all other grids to 0, and connect the pixels corresponding to the maximum gray value based on the double threshold algorithm to determine the outline of each cell.

[0032] The working principle of the above technical solution is as follows: In this embodiment, the first determining module is used to convert the transformed image into an n*m pixel matrix and perform rasterization processing. Each pixel corresponds to a raster, resulting in an n*m raster matrix. A coordinate system is established to determine the coordinate information of each raster. The main characteristics of the raster are: explicit attributes and implicit positioning. That is, the data directly records the attributes themselves, while the location is converted into corresponding coordinates according to the row and column numbers. In other words, the positioning is obtained based on the position of the data in the dataset. In the raster structure, a point is represented by a raster cell; linear features are represented by a group of adjacent raster cells along the line direction. Each raster cell has at most two adjacent cells on the line. For example, the position of a pixel is the coordinate value of the coordinate system, that is, the position of the pixel in the second row and second column in the coordinate system is (2,2), which is used as the coordinate information of the raster.

[0033] In this embodiment, the calculation of the Gaussian coefficients for each grid cell includes: ; in, For position is The Gaussian coefficients of the raster, x=1,2…n; y=1,2,…m; It is a natural constant; Each grid cell has a direction when its edge changes. The direction of the ray extending from the grid cell to this direction is the direction of the grid cell. The angle formed by this ray and the positive direction of the grid cell's horizontal axis is the angle of the grid cell's direction.

[0034] ; in, The angle of the direction in which the grid is located. for right Do the partial derivative, for right Do the partial derivative.

[0035] In this embodiment, the judgment module is used to determine, in the coordinate system, whether the gray value of the pixel corresponding to each current grid is greater than the gray value of the pixel corresponding to the adjacent grid in the direction of the current grid; if so, it determines that the gray value of the pixel corresponding to the current grid remains unchanged; otherwise, it determines that the gray value of the pixel corresponding to the current grid is 0. The second determination module is used to determine the maximum gray value in the direction of each grid, set the gray values ​​of the pixels of all other grids to 0, and connect the pixels corresponding to the maximum gray value based on a double threshold algorithm to determine the contour of each cell. Grids that facilitate the determination of the maximum gray value in all directions are used as feature grids, i.e., feature edge points. The pixels corresponding to the maximum gray value are connected based on the double threshold algorithm to determine the contour of each cell. This improves the accuracy of determining the contour of each cell.

[0036] The beneficial effects of the above technical solution are as follows: Converting pixels in the transformed image into a raster format facilitates pixel vectorization, determines the Gaussian coefficients and directional angles of each raster, and thus facilitates the determination of the raster with the maximum grayscale value across all directions, serving as the feature raster, i.e., feature edge points. Based on a dual-threshold algorithm, the pixels corresponding to the maximum grayscale values ​​are connected to determine the contours of each cell. This improves the accuracy of determining the contours of each cell.

[0037] According to some embodiments of the present invention, the post-processing module includes: The second acquisition module is used to acquire parameter information of the contour of each cell; the parameter information includes contour number and contour geometric parameters; The first elimination module is used to match the geometric parameters of the contour with the preset geometric parameters according to the contour number, eliminate the contours with a matching degree lower than the preset threshold, and re-number them to obtain the correction parameter information of the contours of the remaining cells. The second elimination module is used to construct an outer processing box for each cell based on the correction parameter information, thereby obtaining an outer processing box parameter set; construct a corresponding kd-tree index based on the outer processing box parameter set, and use the outer processing box filtering conditions to traverse the kd-tree index to filter out outer processing boxes without adjacency relationships and eliminate them, thereby obtaining a set of neighboring cells; the set of neighboring cells is a set of cells with adjacency relationships; the adjacency relationships include external tangency, intersection, internal tangency, and inclusion relationships; The third determining module is used for: Cells with intersecting, tangent, or inclusive relationships are selected from adjacent cell sets and merged to obtain the first treatment result; Cells with ectocervical relationships were selected from adjacent cell sets and subjected to tangential enhancement treatment to obtain the second treatment result; Based on the first and second processing results, the post-processing result is determined.

[0038] The working principle of the above technical solution is as follows: In this embodiment, the contours are numbered 1, 2, 3… The contour geometric parameters represent shape parameters, size parameters, etc.

[0039] In this embodiment, the preset geometric parameters are shape and size standards. The first rejection module is used to match the geometric parameters of the contours with the preset geometric parameters according to the contour number, reject contours with a matching degree lower than a preset threshold, and re-number them to obtain the corrected parameter information of the contours of the remaining cells; this facilitates the rejection of contours that do not meet the size requirements or have irregular shapes, which are regarded as noise contours, thus avoiding accurate extraction of cells in the subsequent process.

[0040] In this embodiment, the second elimination module is used to construct an outer processing box for each cell based on the correction parameter information, obtaining an outer processing box parameter set, that is, to wrap the cell based on the outer processing box. A kd-tree (short for k-dimensional tree) is a data structure for segmenting k-dimensional data space, mainly used for searching key data in multi-dimensional space, with filtering conditions including range search and nearest neighbor search. Using the filtering conditions of the outer processing boxes, the kd-tree index is traversed to filter out outer processing boxes without adjacency relationships and eliminate them, obtaining a set of adjacent cells; the set of adjacent cells is a set of cells with adjacency relationships; the adjacency relationships include external tangency, intersection, internal tangency, and containment relationships.

[0041] In this embodiment, cells with intersecting, tangent, or inclusive relationships are screened from the adjacent cell set and merged to obtain the first processing result; to facilitate the merging of contacting cells, two or more contacting cells are merged and fused into independent cells.

[0042] In this embodiment, cells with tangential relationships are selected from the adjacent cell set and subjected to tangential enhancement processing to obtain a second processing result. This facilitates a clearer representation of cells with tangential relationships, avoiding errors in the number of cells identified during the recognition process. Based on the first and second processing results, a post-processing result is determined. This facilitates the merging and sharpening of cell contours in the image, making it easier to clearly determine the contour information of each individual cell.

[0043] The beneficial effects of the above technical solution are: to remove irregular contours and merge contacting cells, to perform tangential enhancement processing on cells with external tangential relationships, to facilitate the merging and clearing of cell contours in the image, to facilitate the clear determination of the contour information of each independent cell, and to facilitate clear visualization.

[0044] According to some embodiments of the present invention, the second rejection module includes: The fourth determination module is used to determine the feature points of the outline of each cell based on the correction parameter information; The building module is used to construct an external processing box based on feature points.

[0045] The working principle of the above technical solution is as follows: In this embodiment, the feature points are the maximum and minimum values ​​of the contour in a preset direction, determined based on the correction parameter information. For example, the pixel points corresponding to the maximum and minimum ordinates, maximum and minimum abscissas. The fourth determining module is used to determine the feature points of the contour of each cell based on the correction parameter information; the construction module is used to construct an external processing box based on the feature points.

[0046] The beneficial effects of the above technical solution are: it facilitates the rapid and accurate construction of external processing boxes based on feature points.

[0047] According to some embodiments of the present invention, the visualization module includes: The segmentation module is used to obtain the contour information of cells in the processed image corresponding to the post-processing result, perform mask processing, and perform multi-scale segmentation to obtain the segmentation result; The scene recognition module is used to perform scene recognition based on the scene recognition model according to the segmentation results, and obtain cell detection results at each segmentation scale. The labeling module is used to fuse cell detection results at each segmentation scale, determine the cell category of the corresponding cells, and label them in the processed image. The visualization processing module is used to visualize the processed images labeled with cell categories, obtain the target image, and display it.

[0048] The working principle and beneficial effects of the above technical solution are as follows: In this embodiment, the segmentation module is used to obtain the contour information of cells in the processed image corresponding to the post-processing result, perform mask processing, and perform multi-scale segmentation to obtain the segmentation result; the scene recognition module is used to perform scene recognition based on the segmentation result and a scene recognition model to obtain the cell detection result at each segmentation scale; the cell detection result is the feature information of pixels at each segmentation scale. The labeling module is used to perform fusion processing on the cell detection results at each segmentation scale, determine the cell category of the corresponding cell, and label it in the processed image; from the local segmentation scales to the overall comprehensive size, it is convenient to accurately determine the cell category of the corresponding cell. The visualization processing module is used to perform visualization processing on the processed image with labeled cell categories to obtain the target image and display it. This facilitates the determination of the included cell categories, makes it convenient for users to view the process, and improves the user experience.

[0049] According to some embodiments of the present invention, the marking module includes: The third acquisition module is used for: The cell detection results at each segmentation scale were obtained and numerically processed to obtain the first value; The feature information of corresponding pixels in the cell detection results at each segmentation scale during the fusion process is obtained and numerically processed to obtain the second value. The fifth determining module is used for: Each first value is compared with the second value to determine the segmentation scale with the greatest change and the segmentation scale with the smallest change. Based on the segmentation scale with the greatest and least change, a preset data table is queried to determine the cell category of the corresponding cell and mark it in the processed image.

[0050] The working principle and beneficial effects of the above technical solution are as follows: Cell detection results at each segmentation scale are acquired and numerically processed to obtain a first value (i.e., the numerical result corresponding to the cell detection result at each segmentation scale before fusion); feature information of corresponding pixels in the cell detection results at each segmentation scale during the fusion process is acquired and numerically processed to obtain a second value (i.e., the numerical result corresponding to the cell detection result at each segmentation scale during the fusion process); a fifth determination module is used to compare each first value with the second value to determine the segmentation scale with the greatest change and the segmentation scale with the least change; the feature information of different cells at different segmentation scales will change differently during the fusion process. A preset data table is a preset data table of the segmentation scale with the greatest change - the segmentation scale with the least change - cell category. Based on the segmentation scale with the greatest change and the segmentation scale with the least change, the preset data table is queried to determine the cell category of the corresponding cell and mark it in the processed image. This facilitates accurate determination of cell categories and their marking in the processed image.

[0051] According to some embodiments of the present invention, the visualization processing module includes: The sixth determination module is used to parse the processed image with labeled cell categories and determine the parsing information; The allocation module is used to allocate and process the target image based on the display elements set in the visualization model according to the parsed information, so as to obtain the target image and display it.

[0052] According to some embodiments of the present invention, the display elements include at least one of bar charts, pie charts, and line charts.

[0053] The working principle and beneficial effects of the above technical solution are as follows: The sixth determination module is used to analyze the processed image of the labeled cell category and determine the analysis information; the allocation module is used to allocate the processing based on the analysis information and the display elements set in the visualization model to obtain the target image and display it. The analysis information is displayed through different display elements, and the processed image is displayed together with it, that is, the target image is displayed, so that users can easily view the processed image and the automatically analyzed data. This data is displayed through at least one of bar charts, pie charts, and line charts.

[0054] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A system for displaying cell contours in pathological images based on deep grayscale variation technology, characterized in that, include: The first acquisition module is used to acquire digital pathology images; The preprocessing module is used to preprocess digital pathology images to obtain preprocessed images; The deep grayscale transformation processing module is used to process the preprocessed image based on the deep grayscale transformation model to obtain the transformed image; The recognition module is used to perform cell contour recognition on the transformed image and determine the contour of each cell. The post-processing module is used to post-process the cell contours and obtain the post-processing results; The visualization module is used to visualize the post-processing results, obtain the target image, and display it.

2. The system for displaying cell contours in pathological images based on deep grayscale change technology as described in claim 1, characterized in that, The preprocessing module includes: The noise reduction module is used to process digital pathological images to obtain denoised images; The brightness adjustment module is used to adjust the brightness of the denoised image to obtain the adjusted image. The enhancement module is used to perform contrast enhancement processing on the adjusted image to obtain a preprocessed image.

3. The system for displaying cell contours in pathological images based on deep grayscale change technology as described in claim 1, characterized in that, The identification module includes: The first determining module is used to convert the transformed image into an n*m pixel matrix and perform rasterization processing, with each pixel corresponding to a raster, to obtain an n*m raster matrix, and to establish a coordinate system to determine the coordinate information of each raster. The calculation module is used to calculate the Gaussian coefficient of each grid cell and the angle of its direction. The judgment module is used to determine, in the coordinate system, whether the gray value of the pixel corresponding to each current grid is greater than the gray value of the pixel corresponding to the adjacent grid in the direction of the current grid; if so, the gray value of the pixel corresponding to the current grid remains unchanged; otherwise, the gray value of the pixel corresponding to the current grid is determined to be 0. The second determining module is used to determine the maximum gray value at the angle of the direction of each grid, set the gray value of the pixels of all other grids to 0, and connect the pixels corresponding to the maximum gray value based on the double threshold algorithm to determine the outline of each cell.

4. The system for displaying cell contours in pathological images based on deep grayscale change technology as described in claim 1, characterized in that, The post-processing module includes: The second acquisition module is used to acquire parameter information of the contour of each cell; the parameter information includes contour number and contour geometric parameters; The first elimination module is used to match the geometric parameters of the contour with the preset geometric parameters according to the contour number, eliminate the contours with a matching degree lower than the preset threshold, and re-number them to obtain the correction parameter information of the contours of the remaining cells. The second elimination module is used to construct an outer processing box for each cell based on the correction parameter information, thereby obtaining an outer processing box parameter set; construct a corresponding kd-tree index based on the outer processing box parameter set, and use the outer processing box filtering conditions to traverse the kd-tree index to filter out outer processing boxes without adjacency relationships and eliminate them, thereby obtaining a set of neighboring cells; the set of neighboring cells is a set of cells with adjacency relationships; the adjacency relationships include external tangency, intersection, internal tangency, and inclusion relationships; The third determining module is used for: Cells with intersecting, tangent, or inclusive relationships are selected from adjacent cell sets and merged to obtain the first treatment result; Cells with ectocervical relationships were selected from adjacent cell sets and subjected to tangential enhancement treatment to obtain the second treatment result; Based on the first and second processing results, the post-processing result is determined.

5. The system for displaying cell contours in pathological images based on deep grayscale change technology as described in claim 4, characterized in that, The second rejection module includes: The fourth determination module is used to determine the feature points of the outline of each cell based on the correction parameter information; The building module is used to construct an external processing box based on feature points.

6. The system for displaying cell contours in pathological images based on deep grayscale change technology as described in claim 1, characterized in that, The visualization module includes: The segmentation module is used to obtain the contour information of cells in the processed image corresponding to the post-processing result, perform mask processing, and perform multi-scale segmentation to obtain the segmentation result; The scene recognition module is used to perform scene recognition based on the scene recognition model according to the segmentation results, and obtain cell detection results at each segmentation scale. The labeling module is used to fuse cell detection results at each segmentation scale, determine the cell category of the corresponding cells, and label them in the processed image. The visualization processing module is used to visualize the processed images labeled with cell categories, obtain the target image, and display it.

7. The system for displaying cell contours in pathological images based on deep grayscale change technology as described in claim 6, characterized in that, The marking module includes: The third acquisition module is used for: The cell detection results at each segmentation scale were obtained and numerically processed to obtain the first value; The feature information of corresponding pixels in the cell detection results at each segmentation scale during the fusion process is obtained and numerically processed to obtain the second value. The fifth determining module is used for: Each first value is compared with the second value to determine the segmentation scale with the greatest change and the segmentation scale with the smallest change. Based on the segmentation scale with the greatest and least change, a preset data table is queried to determine the cell category of the corresponding cell and mark it in the processed image.

8. The system for displaying cell contours in pathological images based on deep grayscale change technology as described in claim 7, characterized in that, The visualization processing module includes: The sixth determination module is used to parse the processed image with labeled cell categories and determine the parsing information; The allocation module is used to allocate and process the target image based on the display elements set in the visualization model according to the parsed information, so as to obtain the target image and display it.

9. The system for displaying cell contours in pathological images based on deep grayscale change technology as described in claim 8, characterized in that, The display elements include at least one of bar charts, pie charts, and line charts.

Images