System for identifying cancer area of digital pathological image

Abstract

The invention provides a system for identifying a cancer area of a digital pathological image. The system comprises an image acquisition module, a positioning module, a target tracking module and an area determination module. The image acquisition module is used for acquiring a digital pathological image; the positioning module is used for marking a cancer area in the digital pathological image; the target tracking module is used for uniformly segmenting the digital pathological image into a plurality of tiles and judging the positions of cancer cells in each tile by using a cancer cell tracking model; and the area determination module is used for calculating the cancer area of the digital pathological image according to the position of the cancer cell in each tile. The method is used for quantitatively analyzing the area of the cancer region.

Description

Technical Field

[0001] The present invention relates to the technical field of digital image processing, and in particular to a system for identifying the area of ​​a cancerous region in a digital pathological image. Background Art

[0002] In the field of digital pathology, the use of artificial intelligence to predict cancer areas in digital pathology images has become a hot area of ​​great concern and research. The traditional method of observing pathological sections has shortcomings such as subjectivity and time-consuming, and is easily affected by human factors. However, the rapid development of digital pathology has realized the digitization of pathological sections, and combined with artificial intelligence technology, it has made the automated analysis of pathological sections possible. Through the automated analysis of digital pathology images, the interference of human factors on the diagnostic results is greatly reduced, and the accuracy and efficiency of pathological diagnosis are significantly improved, especially in the diagnosis of lesion areas. Therefore, the prediction of cancer areas by artificial intelligence in digital pathology images has become a striking research hotspot in the field of digital pathology and has been widely used in clinical practice.

[0003] However, the output results of digital pathology image-assisted diagnosis systems usually only include the identification of cancer areas, and better products will include the classification of cancer areas. However, when outputting the prediction results, users often need to know the area of ​​the cancer area in the current slice and the corresponding area ratio, so as to help users make a more detailed pathological diagnosis of the slice.

[0004] For example, invention patent application CN202211423333.6-A device and method for cancer area segmentation of IHC nuclear expression pathology images, a device for cancer area segmentation of IHC nuclear expression pathology images, which includes a reading module, a preprocessing module and a cancer area segmentation module. The reading module is used to read the full-film image of the digital pathology image to be segmented; the preprocessing module is used to preprocess the full-film image, and use the sliding window cutting method to cut the full-film image into small images; the cancer area segmentation module is used to input the cut small image into a trained cancer area segmentation model for cancer area recognition and segmentation, and generate a cancer area mask of the corresponding area; the splicing module is used to stitch the generated cancer area mask according to the cutting position sequence of the corresponding small image blocks, and convert it into a full-film cancer area mask image; the contour extraction module is used to extract the contour of the cancer area, and fuse the contour of the cancer area into the full-film image to generate a cancer area segmentation effect diagram. Although the identification of cancerous areas in the above invention patent application can be performed for qualitative analysis, it is not possible to perform quantitative analysis, which means that users need to perform quantitative analysis through estimation. It is also impossible to judge the severity of the current pathological section through the data of cancer percentage, and thus it is impossible to assist users in writing a more comprehensive diagnostic report. Summary of the invention

[0005] The present invention provides a system for identifying the area of ​​cancerous regions in digital pathological images, which is used to solve the problem that although the identification and classification of cancerous regions can be qualitatively analyzed, they cannot be quantitatively analyzed, resulting in the user having to perform quantitative analysis by estimation.

[0006] A system for identifying the area of ​​cancerous areas in digital pathology images, comprising an image acquisition module, a positioning module, a target tracking module and an area determination module, wherein:

[0007] The image acquisition module is used to obtain digital pathology images;

[0008] The positioning module is used to mark the cancerous area in the digital pathology image;

[0009] The target tracking module is used to evenly divide the digital pathology image into multiple tiles and use the cancer cell tracking model to determine the location of the cancer cells in each tile;

[0010] The area determination module is used to calculate the cancer area of ​​the digital pathology image according to the position of the cancer cells in each tile.

[0011] Further, the area determination module includes a calculation unit and an accumulation unit, wherein:

[0012] The area calculation unit is used to calculate the area of ​​the pathological section and the area of ​​the cancerous area in each tile according to the position of the cancer cells in each tile;

[0013] The accumulation unit is used to accumulate the areas of the pathological sections and the cancerous areas in all tiles respectively, to obtain the areas of the pathological sections and the cancerous areas in the current digital pathological image.

[0014] Furthermore, the computing unit is configured to:

[0015] The number of slice cell and cancer cell pixels in each tile was counted using the pixel counting method;

[0016] The scaling operation is performed according to the resolution of the tile to obtain the area of ​​the pathological section and the cancer area in the tile.

[0017] Furthermore, it also includes a proportion determination module, which is used to calculate the proportion of the cancer area according to the area of ​​the digital pathology image and the area of ​​the cancer area in the pathology section.

[0018] Furthermore, the marking of cancerous areas in digital pathology images is obtained through the marking model;

[0019] The labeling model is constructed by the following steps:

[0020] Set up orthophoto references and digital surface models for cancer area marking;

[0021] Perform group detection on each pixel in the orthophoto map to obtain the detection group corresponding to each pixel;

[0022] Based on the detection group corresponding to each pixel in the orthophoto map, the connected domain of each cancer area mark in the orthophoto map is determined;

[0023] According to the connected domain, set the marking frequency points and cancer area rendering color;

[0024] Set the cancer area marking frequency parameters according to the frequency marking;

[0025] According to the cancer area rendering color and the digital surface model, the digital trigger response is set, and the digital trigger response is associated with the cancer area marking frequency point parameters.

[0026] Furthermore, the system also includes a cancer area classification module, which is used to identify cells in the current digital pathology image using a classification model and determine the change process of the cancer area cells based on the recognition result.

[0027] Furthermore, the system also includes an input response module, which is configured to:

[0028] Receiving an interactive operation on a display interface; wherein the interactive operation is displaying one or more of the cancer area, classification, area and occupied area of ​​the cancer area of ​​the digital pathology image on the display interface;

[0029] The cancer area, classification, area and / or occupied area of ​​the digital pathology image are obtained from the positioning module, classification module, area determination module and proportion determination module respectively, and the interactive operation of the display interface is responded to.

[0030] Furthermore, the steps of constructing the cancer cell tracking model specifically include:

[0031] Obtain labeled pathology atlases and randomly divide them into training and test sets;

[0032] Using the training set to train the cancer cell tracking model, to obtain a trained cancer cell tracking model;

[0033] The test set is used to verify the trained cancer cell tracking model.

[0034] Furthermore, before randomly dividing the pathology atlas into a training set and a test set, the method further comprises the step of preprocessing the pathology atlas; the preprocessing comprises:

[0035] The digital pathology image is evenly divided into a plurality of tiles; the features of each tile are extracted; the features of each tile are randomly scaled, cropped and / or arranged to obtain a pathology atlas.

[0036] Furthermore, the specific steps of extracting the features of each tile include:

[0037] Binarization is performed on the tiles to obtain tiles to be processed, and the edges of cells are extracted from the tiles to be processed with cancer area marks to obtain edge pixel point sets;

[0038] Extract color features and cell morphology features of the tiles to be processed;

[0039] Determine the cell shape distribution characteristics based on the gradient direction of each pixel in the edge pixel set;

[0040] The color features, cell morphology features and shape distribution features are fused to obtain the features of each tile.

[0041] Beneficial effects of the present invention: The present application performs quantitative analysis through the area of ​​the cancerous area, and also completes a comprehensive quantitative and qualitative analysis in combination with the typing of the cancerous area, so as to assist users in better completing the diagnosis.

[0042] Other features and advantages of the present invention will be described in the following description, and partly become apparent from the description, or understood by practicing the present invention. The purpose and other advantages of the present invention can be realized and obtained by the structures particularly pointed out in the written description and the accompanying drawings.

[0043] The technical solution of the present invention is further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:

[0045] Figure 1 This is a schematic diagram of the structure of a system for identifying the area of ​​a cancerous region in a digital pathology image in Example 1 of the present invention;

[0046] Figure 2 This is a flow chart for calculating the area of ​​pathological sections and cancerous areas in digital pathological images in Example 1 of the present invention;

[0047] Figure 3 Flow chart of constructing the labeling model in Example 1 of the present invention. DETAILED DESCRIPTION

[0048] The preferred embodiments of the present invention are described below in conjunction with the accompanying drawings. It should be understood that the preferred embodiments described herein are only used to illustrate and explain the present invention, and are not used to limit the present invention.

[0049] Embodiment 1 of the present invention provides a system for identifying the area of ​​a cancerous region in a digital pathology image, comprising an image acquisition module, a positioning module, a target tracking module and an area determination module, wherein:

[0050] The image acquisition module is used to obtain digital pathology images;

[0051] The positioning module is used to mark the cancerous area in the digital pathology image;

[0052] The target tracking module is used to evenly divide the digital pathology image into multiple tiles and use the cancer cell tracking model to determine the location of the cancer cells in each tile;

[0053] The area determination module is used to calculate the cancer area of ​​the digital pathology image according to the position of the cancer cells in each tile.

[0054] Among them, digital pathology images are first scanned and collected by a fully automatic microscope or optical magnification system to obtain high-resolution digital images, and then high-precision multi-field seamless stitching and processing are performed by a computer to obtain a multi-level visualized full-slice image.

[0055] The working principle of the above technical solution is: Figure 1 , the present application first uses the image acquisition module to obtain the digital pathology image, and then uses the positioning module to mark the cancer area in the digital pathology image. The marking of the cancer area of ​​the digital pathology image can be completed by using the cancer area prediction model, which is a deep learning model trained by the labeled data. Then, the digital pathology image that has been marked is evenly divided into multiple tiles by the target tracking module, and the cancer cell tracking model is used to determine the location of the cancer cells in each tile. Tile technology is a technology that divides a large digital image into small pieces. Since digital pathology images usually have very high resolutions, a single image file may be very large and difficult to load and process at one time. The present application uses tile technology to cut a large image into multiple smaller image blocks, so that each block usually has the same size, which solves the above problem. The size of the tile can be 256x256 pixels. The cancer cell tracking model is also a trained deep learning model, which can determine the location of the cancer cells in each tile according to the cancer area marking of the digital pathology image. Finally, the area of ​​the cancer area in the digital pathology image is calculated based on the location of the cancer cells in each tile.

[0056] The beneficial effects of the above technical solution are as follows: the present application utilizes slice segmentation and image analysis of tile technology to effectively calculate the area of ​​each region in a digital pathology image. Such a calculation method not only improves the calculation efficiency, but also makes it possible to calculate the area of ​​large-size slices, providing users with more comprehensive and accurate analysis results. The recognition system of the present application can quantitatively analyze the area of ​​the cancer area to assist users in better completing the diagnosis.

[0057] In one embodiment, the area determination module includes a calculation unit and an accumulation unit, wherein:

[0058] The area calculation unit is used to calculate the area of ​​the pathological section and the area of ​​the cancerous area in each tile according to the position of the cancer cells in each tile;

[0059] The accumulation unit is used to accumulate the areas of the pathological sections and the cancerous areas in all tiles respectively, to obtain the areas of the pathological sections and the cancerous areas in the current digital pathological image.

[0060] Among them, pathological sections are pathological sections made by pathological histological methods, that is, objects scanned and collected by fully automatic microscopes or optical magnification systems. The preparation method of pathological sections is usually to embed the diseased tissue in paraffin blocks, cut it into thin slices with a slicer, then stain it with hematoxylin-eosin (HE), and finally scan it with a fully automatic microscope or optical magnification system.

[0061] The working principle of the above technical solution is: Figure 2 ,After the digital pathology image is segmented into multiple tiles, the area of ​​the pathology ,section and the cancer area in each tile are calculated first, and the areas of the pathology ,sections and cancer areas in all tiles are accumulated respectively, and finally the area of ​​the digital pathology ,image and the area of ​​the cancer area in the pathology ,sections are obtained.

[0062] The beneficial effect of the above technical solution is: the present application calculates the area of ​​the pathological section and the area of ​​the cancer area in the current digital pathological image through the area calculation unit and the cumulative calculation unit, so as to better assist the user in diagnosis.

[0063] In one embodiment, the calculation unit is configured to: count the number of slice cells and cancer cell pixels in each tile using a pixel counting method; perform a scaling operation according to the resolution of the tile to obtain the area of ​​the pathological slice and the cancer area in the tile.

[0064] The working principle of the above technical solution is: Figure 2 ,The cancer cell tracking model can identify the location of cancer cells in each tile, and the location of slice cells is also known. Therefore, the pixel counting method is used to count the number of pixels of slice cells and cancer cells in each tile, and then scaled according to the resolution to obtain the area of ​​pathological slices and cancer areas in the tile.

[0065] The beneficial effects of the above technical solution are as follows: the present application uses the pixel counting method to calculate the number of slice cell and cancer cell pixels in each tile, and scales it according to the resolution of the tile, thereby obtaining the area of ​​the pathological slice and the cancer area in the tile, thereby improving the calculation efficiency and accuracy without the need for human intervention.

[0066] In one embodiment, a proportion determination module is further included, which is used to calculate the proportion of the cancer area according to the area of ​​the digital pathology image and the area of ​​the cancer area in the pathology section. The proportion of the cancer area is obtained by dividing the cancer area by the total area of ​​the section and multiplying it by 100%.

[0067] The working principle of the above technical solution is: the proportion determination module of the present application can calculate the proportion of the cancer area according to the area of ​​the digital pathology image and the area of ​​the cancer area in the pathology section, thereby assisting users in estimating the future development process of cancer cells.

[0068] The beneficial effects of the above technical solution are: judging the development of cancer cells through the data of the percentage of cancer area, assisting users in estimating the future development process of cancer cells, and thus better helping users to write diagnostic reports.

[0069] In one embodiment, the marking of the cancerous area in the digital pathology image is obtained by a trained marking model. The marking model includes the following steps:

[0070] Set up orthophoto references and digital surface models for cancer area marking;

[0071] Perform group detection on each pixel in the orthophoto map to obtain the detection group corresponding to each pixel;

[0072] Determine the connected domain of each cancer area mark in the orthophoto map based on the pixel group corresponding to each pixel in the orthophoto map;

[0073] According to the connected domain, set the marking frequency points and cancer area rendering color;

[0074] Set the cancer area marking frequency parameters according to the frequency marking;

[0075] According to the cancer area rendering color and the digital surface model, the digital trigger response is set, and the digital trigger response is associated with the cancer area marking frequency point parameters.

[0076] The orthophoto reference is a digital pathology image template with cancer area markings, and the digital surface model is a corresponding digital signal. The detection groups are set as needed, and the number of groups can be determined according to the malignancy of cancer cells in the data pathology image template.

[0077] The working principle of the above technical solution is: Figure 3, by performing group detection on the orthophoto reference with cancer area marking, a detection group corresponding to each pixel is obtained. Each detection group corresponds to a connected domain, and different connected domains have different marking frequencies and cancer area rendering colors. Then, the cancer area marking frequency parameters are set according to the frequency marking, and the digital trigger response is set according to the cancer area rendering color and the digital surface model, and the digital trigger response is associated with the cancer area marking frequency parameters; repeat the above steps until all pixels are traversed, and then the cancer area marking of the current digital pathology image can be obtained.

[0078] The beneficial effects of the above technical solution are: the cancer area of ​​the digital pathology image is marked by the constructed marking model, the marking model identifies each pixel, and finally obtains the cancer area mark of the current digital pathology image. The marking model can accurately identify the cancer area mark and provide a basis for the subsequent calculation of the cancer area.

[0079] In one embodiment, the system further includes a cancer area classification module, which is used to identify cells in the current digital pathology image using a classification model, and determine the change process of the cancer area cells according to the recognition result.

[0080] The beneficial effects of the above technical solution are as follows: the present application utilizes the cancer area classification module to identify cells in the current digital pathology image, thereby determining from which cells the cells in the cancer area have changed. The cancer area classification can assist the user in selecting which normal cells to compare when the user identifies the malignancy of the cancer area.

[0081] In one embodiment, the system further comprises an input response module, the input response module being configured to:

[0082] Receiving an interactive operation on a display interface; wherein the interactive operation is displaying one or more of the cancer area, classification, area and occupied area of ​​the cancer area of ​​the digital pathology image on the display interface;

[0083] The cancer area, classification, area and / or occupied area of ​​the digital pathology image are obtained from the positioning module, classification module, area determination module and proportion determination module respectively, and the interactive operation of the display interface is responded to.

[0084] The working principle of the above technical solution is: the input response module of the present application can accept and respond to the interactive operation of the display interface. The interactive operation is to display one or more of the cancer area, classification, area and occupied area of ​​the cancer area of ​​the digital pathology image in the display interface, and when responding to the interactive operation of the display interface, the cancer area, classification, area and / or occupied area of ​​the cancer area of ​​the digital pathology image are obtained from the positioning module, classification module, area determination module, and proportion determination module respectively.

[0085] The beneficial effects of the above technical solution are: responding to interactive operations, providing users with the cancer areas, types, areas and / or occupied areas required to analyze digital pathology images, and users can judge the development of cancer cells through the cancer area ratio data, assisting users in estimating the future development process of cancer cells, thereby helping users write diagnostic reports.

[0086] In one embodiment, the steps of constructing the cancer cell tracking model specifically include:

[0087] Obtain a labeled pathology atlas and randomly divide it into a training set and a test set; use the training set to train a cancer cell tracking model to obtain a trained cancer cell tracking model; and use the test set to verify the trained cancer cell tracking model.

[0088] The working principle of the above technical solution is: randomly divide the pathology atlas into a training set and a test set, use the training set to train the cancer cell tracking model, and use the test set to verify the trained cancer cell tracking model, thereby obtaining the trained cancer cell tracking model.

[0089] The beneficial effects of the above technical solution are: by constructing a cancer cell tracking model, the specific location of the cancer cells in the tile can be accurately identified, and the area of ​​the cancerous area can be obtained according to the location of the cancer cells.

[0090] In one embodiment, the method further comprises the steps of: before randomly dividing the pathology atlas into a training set and a test set, the method further comprises the steps of: preprocessing the digital pathology images;

[0091] The preprocessing step includes: evenly dividing each digital pathology image into a plurality of tiles; extracting the features of each tile; and randomly scaling, cropping and / or arranging the features of each tile to obtain a pathology atlas.

[0092] The working principle of the above technical solution is: the present application first divides the digital pathology image into multiple equal tiles, and then performs feature extraction; finally, the features of the digital pathology image are randomly scaled, cropped and / or arranged to obtain an enhanced pathology atlas, thereby increasing the sample size of the data set.

[0093] The beneficial effect of the above technical solution is: the present application increases the digital pathology images to increase the sample size and improve the accuracy of the trained recognition model.

[0094] In one embodiment, the tiles are binarized using an adaptive threshold algorithm to obtain tiles to be processed;

[0095] Extract the edge of the cell from the tile to be processed with the cancer area mark, and smooth the edge to obtain the edge pixel point set;

[0096] Extract color features and cell morphology features of the tiles to be processed; determine the cell shape distribution features based on the gradient direction of each pixel point in the edge pixel point set;

[0097] Assign weights to color features, cell morphology features, and shape distribution features; and use a fusion strategy to fuse color features, cell morphology features, and shape distribution features to obtain a comprehensive feature set.

[0098] The working principle of the above technical solution is as follows: the application first uses an adaptive threshold algorithm to binarize each tile to distinguish the foreground from the background. Then, the edge contour of the cell is extracted from the tile to be processed with the cancer area mark, and the edge is smoothed to obtain an edge pixel point set; the Canny edge detector can be used to extract the edge contour of the cell, and the Canny edge detector can accurately extract the edge. Morphological operations can be used for smoothing to improve the shape and connectivity of the features. Next, color features such as color histograms and color moments and cell morphological features are extracted from the tile to be processed; and the cell shape distribution features are determined based on the gradient direction of each pixel in the edge pixel point set; finally, the color features, cell morphological features, and shape distribution features are fused using fusion strategies such as concatenation, feature weighting, etc., to obtain a comprehensive feature vector.

[0099] The beneficial effect of the above technical solution is: through the above features, the present application can improve the performance and robustness of the cancer cell tracking model.

[0100] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.

Claims

1. A system for identifying the area of ​​cancerous areas in digital pathology images, characterized in that: It includes an image acquisition module, a positioning module, a target tracking module and an area determination module, wherein: The image acquisition module is used to obtain digital pathology images; The positioning module is used to mark the cancerous area in the digital pathology image; The target tracking module is used to evenly divide the digital pathology image into multiple tiles and use the cancer cell tracking model to determine the location of the cancer cells in each tile; The area determination module is used to calculate the cancer area of ​​the digital pathology image according to the position of the cancer cells in each tile.

2. A system for identifying the area of ​​cancerous areas in digital pathology images as claimed in claim 1, characterized in that: The area determination module includes a calculation unit and an accumulation unit, wherein: The calculation unit is used to calculate the area of ​​the pathological section and the area of ​​the cancerous area in each tile according to the position of the cancer cells in each tile; The accumulation unit is used to accumulate the areas of the pathological sections and the cancerous areas in all tiles respectively, to obtain the areas of the pathological sections and the cancerous areas in the current digital pathological image.

3. A system for identifying the area of ​​cancerous areas in digital pathology images as claimed in claim 2, characterized in that: The computing unit is configured to: The number of slice cell and cancer cell pixels in each tile was counted using the pixel counting method; The scaling operation is performed according to the resolution of the tile to obtain the area of ​​the pathological section and the cancer area in the tile.

4. A system for identifying the area of ​​cancerous areas in digital pathological images as claimed in claim 2, characterized in that: It also includes a proportion determination module, which is used to calculate the proportion of the cancer area according to the area of ​​the digital pathology image and the area of ​​the cancer area in the pathology section.

5. The system for identifying the cancer area of ​​a digital pathology image according to claim 1, characterized in that: The cancer area marking in the digital pathology image is obtained through the labeling model; The labeling model is constructed by the following steps: Set up orthophoto references and digital surface models for cancer area marking; Perform group detection on each pixel in the orthophoto map to obtain the detection group corresponding to each pixel; Based on the detection group corresponding to each pixel in the orthophoto map, the connected domain of each cancer area mark in the orthophoto map is determined; According to the connected domain, the marking frequency and the cancer area rendering color are set; according to the frequency mark, the cancer area marking frequency parameters are set; According to the cancer area rendering color and the digital surface model, the digital trigger response is set, and the digital trigger response is associated with the cancer area marking frequency point parameters.

6. A system for identifying the area of ​​cancerous regions in digital pathological images as claimed in claim 4, characterized in that: The system also includes a cancer area classification module, which is used to identify cells in the current digital pathology image using a classification model and determine the change process of the cancer area cells based on the recognition result.

7. A system for identifying the area of ​​cancerous regions in digital pathological images as claimed in claim 6, characterized in that: The system also includes an input response module, which is configured to: Receiving an interactive operation on a display interface; wherein the interactive operation is displaying one or more of the cancer area, classification, area and occupied area of ​​the cancer area of ​​the digital pathology image on the display interface; The cancer area, classification, area and / or occupied area of ​​the digital pathology image are obtained from the positioning module, classification module, area determination module and proportion determination module respectively, and the interactive operation of the display interface is responded to.

8. The system for identifying the cancer area of ​​a digital pathology image according to claim 1, characterized in that: The steps to build the cancer cell tracking model include: Obtain labeled pathology atlases and divide them into training set, test set and validation set; Using the training set to train the cancer cell tracking model, to obtain a trained cancer cell tracking model; The trained cancer cell tracking model is verified and tested using the test set and validation set.

9. A system for identifying the area of ​​cancerous regions in digital pathology images as claimed in claim 8, characterized in that: Before randomly dividing the pathology atlas into a training set and a test set, the method further includes the steps of: preprocessing the digital pathology images; The preprocessing step includes: evenly dividing each digital pathology image into a plurality of tiles; extracting the features of each tile; and randomly scaling, cropping and / or arranging the features of each tile to obtain a pathology atlas.

10. A system for identifying the area of ​​cancerous regions in digital pathological images as claimed in claim 9, characterized in that: The specific steps to extract the features of each tile include: Using the adaptive threshold algorithm, the tiles are binarized to obtain tiles to be processed; Extract the edge of the cell from the tile to be processed with the cancer area mark, and smooth the edge to obtain the edge pixel point set; Determine the color characteristics and cell morphology characteristics of the tile to be processed; and determine the cell shape distribution characteristics based on the gradient direction of each pixel point in the edge pixel point set; Based on the fusion strategy, the color features, cell morphology features and shape distribution features are fused to obtain a comprehensive feature set.

Images