A cell classification and counting method and system based on computer vision

Abstract

The present invention provides a cell classification and counting method and system based on computer vision, which obtains and preprocesses the original pathological slice digital image; segments the preprocessed pathological slice digital image based on an improved watershed algorithm to obtain an image segmentation boundary; merges and segments the image segmentation boundary with the original pathological slice digital image to obtain a segmented pathological slice digital image; extracts features of the segmented pathological slice digital image, and classifies the extracted features to obtain first cell data; simultaneously, detects the segmented pathological slice digital image using a neural network training model to obtain second cell data; takes the union of the first cell data and the second cell data and removes duplicates to obtain positive and negative cell data; and realizes the application of computational imaging in the medical field to assist in diagnosis, thereby greatly reducing the workload of doctors.

Description

Technical Field

[0001] The present invention belongs to the field of cell identification, positioning, classification and counting, and in particular relates to a computer vision-based cell classification and counting method and system. Background Art

[0002] The statements in this section merely provide background information related to the present invention and do not necessarily constitute prior art.

[0003] In breast cancer immunohistochemistry, ER stands for estrogen receptor, and PR stands for progesterone receptor. Both are positive, which is called endocrine therapy sensitive, which means that the overall survival rate and disease-free survival rate of such patients who can undergo endocrine therapy are significantly higher than those of negative patients. Doctors in professional institutions can continuously observe the morphology of immunohistochemical cells under a microscope and give evaluation results. This method is simple and direct, but it requires considerable professional background knowledge and a lot of manual participation, with a high threshold and low efficiency. Therefore, many researchers have tried to use intelligent image interpretation to replace manual participation. Traditional manual identification and counting methods are inefficient and very limited by professional knowledge and energy.

[0004] At present, many computerized methods rely on color features to detect and classify cells for immunohistochemical cell evaluation technology, or use deep learning neural network methods to identify, locate and classify cells. The morphology and color of the nuclei of various cells in the stained images are extremely similar, which makes some traditional methods regard non-tumor cells as tumor cells, resulting in a large number of counting errors. The use of deep learning methods will cause inaccurate recognition rates and cannot solve problems such as cell adhesion. In supervised and unsupervised learning methods, single pixels are used as research objects, and pixels in the same category together constitute each component of the tissue. Researchers need to select representative areas of each tissue component, including all cell types, as training samples before conducting supervised classification. Its performance depends largely on the quality and comprehensiveness of the pre-defined training samples; that is, the presence or absence of supervised learning methods has a great impact on the accuracy of cell classification. Summary of the invention

[0005] In order to solve the technical problems existing in the above-mentioned background technology, the present invention provides a cell classification and counting method and system based on computer vision, which adopts the calculation method of computer vision to realize the application of computational imaging in the medical field, assists diagnosis, and greatly reduces the workload of doctors. In view of the shortcomings of inaccurate distinction using features such as color in traditional algorithms, color features, texture features and other aspects of features are used for creative identification; for adhesion cells, an improved watershed algorithm is used for segmentation; for cell classification, after completing cell segmentation, a supervised method is used for classification, thereby solving the problem of cell classification and counting.

[0006] In order to achieve the above object, the present invention adopts the following technical solution:

[0007] A first aspect of the present invention provides a computer vision-based cell classification and counting method, comprising:

[0008] Acquire original pathological slice digital images and perform preprocessing;

[0009] Based on the improved watershed algorithm, the preprocessed digital images of pathological sections are segmented to obtain the image segmentation boundaries;

[0010] Merge and segment the image segmentation boundary with the original pathological slice digital image to obtain a segmented pathological slice digital image;

[0011] Extracting features of the segmented digital image of the pathological slice, and classifying the extracted features to obtain first cell data; and detecting the segmented digital image of the pathological slice using a neural network training model to obtain second cell data;

[0012] The first cell data and the second cell data are combined and duplicates are removed to obtain positive and negative cell data.

[0013] A second aspect of the present invention provides a computer vision-based cell classification and counting system, comprising:

[0014] An image acquisition module is configured to acquire original pathological slice digital images and perform preprocessing;

[0015] The image processing module is configured to segment the preprocessed pathological slice digital image based on an improved watershed algorithm to obtain an image segmentation boundary;

[0016] An image segmentation module is configured to merge and segment the image segmentation boundary with the original pathological slice digital image to obtain a segmented pathological slice digital image;

[0017] The image cell classification module is configured to extract features of the segmented pathological slice digital image and classify the extracted features to obtain first cell data; and at the same time, the segmented pathological slice digital image is detected using a neural network training model to obtain second cell data;

[0018] The image cell counting module is configured to obtain positive and negative cell data by taking the union of the first cell data and the second cell data and removing duplicates.

[0019] A third aspect of the present invention provides a computer-readable storage medium.

[0020] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of a computer vision-based cell classification and counting method as described in the first aspect above.

[0021] A fourth aspect of the present invention provides a computer device.

[0022] A computer device comprises a memory, a processor and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the steps in a cell classification and counting method based on computer vision as described in the first aspect above are implemented.

[0023] Compared with the prior art, the present invention has the following beneficial effects:

[0024] The present invention adopts the calculation method of computer vision to realize the application of computational imaging in the medical field, assist in diagnosis, and greatly reduce the workload of doctors. In view of the shortcomings of inaccurate distinction using features such as color in traditional algorithms, multiple features such as color features and texture features are used for creative identification; for adhesion cells, an improved watershed algorithm is used for segmentation; for cell classification, after completing cell segmentation, a supervised method is used for classification, thereby solving the problem of cell classification and counting.

[0025] Advantages of additional aspects of the present invention will be given in part in the following description, and in part will become obvious from the following description, or will be learned through practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] The accompanying drawings in the specification, which constitute a part of the present invention, are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute improper limitations on the present invention.

[0027] Figure 1 is a principle flow chart of a cell classification and counting method based on computer vision in an embodiment of the present invention;

[0028] Figure 2It is a rendering of the cell classification and counting method based on computer vision in an embodiment of the present invention. DETAILED DESCRIPTION

[0029] The present invention will be further described below in conjunction with the accompanying drawings and embodiments.

[0030] It should be noted that the following detailed descriptions are all illustrative and intended to provide further explanation of the present invention. Unless otherwise specified, all technical and scientific terms used herein have the same meanings as those commonly understood by those skilled in the art to which the present invention belongs.

[0031] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit exemplary embodiments according to the present invention. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. In addition, it should be understood that when the terms "comprising" and / or "including" are used in this specification, it indicates the presence of features, steps, operations, devices, components and / or combinations thereof.

[0032] It should be noted that the flowcharts and block diagrams in the accompanying drawings illustrate the possible implementation architecture, functions and operations of the methods and systems according to various embodiments of the present disclosure. It should be noted that each box in the flowchart or block diagram can represent a module, a program segment, or a part of a code, and the module, program segment, or a part of a code may include one or more executable instructions for implementing the logical functions specified in each embodiment. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in an order different from that marked in the accompanying drawings. For example, two boxes represented in succession can actually be executed substantially in parallel, or they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the flowchart and / or block diagram, and the combination of boxes in the flowchart and / or block diagram can be implemented using a dedicated hardware-based system that performs a specified function or operation, or can be implemented using a combination of dedicated hardware and computer instructions.

[0033] Embodiment 1

[0034] like Figure 1As shown, this embodiment provides a cell classification and counting method based on computer vision. This embodiment uses the method applied to a server as an example. It can be understood that the method can also be applied to a terminal, and can also be applied to a terminal, a server, and a system, and is implemented through the interaction between the terminal and the server. The server can be an independent physical server, or a server cluster or a distributed system composed of multiple physical servers, or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network servers, cloud communications, middleware services, domain name services, security services CDN, and big data and artificial intelligence platforms. The terminal can be a smart phone, a tablet computer, a laptop computer, a desktop computer, a smart speaker, a smart watch, etc., but is not limited to this. The terminal and the server can be directly or indirectly connected via wired or wireless communication, and this application is not limited here.

[0035] In this embodiment, the method includes the following steps:

[0036] Step S1: obtaining the original pathological slice digital image and preprocessing it;

[0037] Step S2: Segment the preprocessed digital image of the pathological section based on the improved watershed algorithm to obtain the image segmentation boundary;

[0038] Step S3: merging and segmenting the image segmentation boundary and the original pathological slice digital image to obtain a segmented pathological slice digital image;

[0039] Step S4: extracting features of the segmented pathological section digital image, and classifying the extracted features to obtain first cell data; and detecting the segmented pathological section digital image using a neural network training model to obtain second cell data;

[0040] Step S5: Take the union of the first cell data and the second cell data and remove duplicates to obtain positive and negative cell data.

[0041] In step S1, the original pathological slice digital image O(x, y) is obtained and preprocessed, specifically:

[0042] Step S11: using a pathology section device to perform sampling and shooting, to obtain a pathology section digital image with high definition, aligned borders and good quality, that is, an original pathology section digital image;

[0043] Step S12: performing color channel separation on the obtained digital image of pathological slices to obtain digital images of pathological slices of different color channels;

[0044] Step S13: performing threshold image segmentation on the obtained digital images of pathological sections of different color channels to remove the influence of impurity background on recognition;

[0045] Step S14: For the pathological slice digital image obtained in step S13, impurities in the pathological slice digital image are removed by morphological processing to obtain a preprocessed pathological slice digital image V (x, y).

[0046] The morphological processing methods specifically include expansion corrosion and other means.

[0047] In step S2, based on the improved watershed algorithm, the preprocessed pathological slice digital image V (x, y) is segmented to obtain the image segmentation boundary g (x, y). The specific process is as follows:

[0048] Step S21: performing filtering and denoising processing on the preprocessed pathological slice digital image V(x, y) to obtain a filtered and denoised pathological slice digital image U(t, x);

[0049] Step S22: transforming the digital image U(t, x) of the pathological slice after filtering and noise reduction to obtain the digital image F(x, y) of the pathological slice after transformation;

[0050] Step S23: performing threshold processing on the transformed digital image of the pathological slice F(x, y) to obtain an image segmentation boundary g(x, y).

[0051] In the traditional watershed algorithm, image segmentation is achieved by selecting seeds to merge pixels in different areas, which can usually obtain good image segmentation results and is suitable for complex image segmentation. However, this method is easily affected by noise, and the segmentation results are prone to lose important edge information. When the noise content is high, over-segmentation and under-segmentation are prone to occur, and the segmentation accuracy of the algorithm is too low. To address this phenomenon, the present application improves the traditional watershed algorithm.

[0052] Specifically, in step S21, the preprocessed pathological slice digital image V (x, y) is first filtered using Gaussian curvature filtering, wherein the Gaussian curvature filtering is performed in a two-dimensional image, and the calculation method of the Gaussian curvature is:

[0053]

[0054] In the formula, U refers to V(x, y).

[0055] Then the filtered image is subjected to noise elimination to overcome the noise interference of the preprocessed pathological slice digital image and protect the edge information of the segmented target;

[0056] In the denoising algorithm, the variational model used is the Gaussian curvature variational model:

[0057]

[0058] Among them, ε(U) is the Gaussian curvature energy, ∈ is the evolution termination threshold, which is a very small number;

[0059]

[0060] It is generally based on the diffusion method, that is, the diffusion-based algorithm, which mainly adds pseudo time t until a stable state is reached or the termination condition is met:

[0061]

[0062] Initial condition U(t=0,x)=I(x):

[0063]

[0064] Thus, the filtered and denoised pathological slice digital image U (t, x) is obtained.

[0065] In step S22, a mathematical morphological transformation operation is performed on the filtered and denoised pathological slice digital image U(t, x) to enhance the contrast of the relevant image, achieve contrast separation of the background image and the foreground image, and obtain the transformed pathological slice digital image F(x, y).

[0066] For the processing of the transformed digital image F(x, y) of the pathological slice, the traditional approach is to perform a watershed transform on F(x, y) to obtain the watershed image of the image F(x, y). The boundary points between the watersheds are the watersheds. Obviously, the watershed represents the maximum point of the image F(x, y).

[0067] Therefore, in order to obtain the edge information of the image, the improved gradient image g'(x, y) is used as the input image instead of F(x, y), that is, the relationship between g'(x, y) and F(x, y) is:

[0068] g'(x, y)=grad(F(x, y)); (5)

[0069] In the formula, F(x, y) represents the image obtained by watershed transformation, and grad{.} represents the gradient operation.

[0070] In addition, although the traditional watershed algorithm has a good response to weak edges, the noise in the image and the slight grayscale changes on the surface of the object will cause over-segmentation. In order to eliminate the over-segmentation caused by the traditional watershed algorithm, two processing methods can usually be used. One is to use prior knowledge to remove irrelevant edge information. The second is to modify the gradient function so that the water basin only responds to the target you want to detect. In order to reduce the over-segmentation caused by the traditional watershed algorithm, the gradient function is usually modified. A simple method is to perform threshold processing on the gradient image to eliminate over-segmentation caused by slight changes in grayscale. That is,

[0071] g'(x, y)=max(grad(F(x, y)), gθ)), (6)

[0072] In the formula, gθ represents the threshold, and the gradient image g'(x, y) is finally obtained.

[0073] Then use the gradient image, use M1, M2, ..., Mn to represent the pixel position of each local minimum, C(Mi) is the set of pixel coordinates in the corresponding area of ​​Mi, and use n to represent the current gradient threshold;

[0074] T[n] represents a set of pixels denoted as (u, v), g(u, v) < n, that is,

[0075] T[n]={(u,v)|g(u,v)<n}

[0076] The gradient threshold increases from the lowest integer value of the image gradient range; when the gradient threshold is n, the pixel set in the plane g(x, y)<n is T[n]; the coordinate set C(Mi) that satisfies the conditions in the area where Mi is located can be regarded as a binary image: Cn(Mi)=C(Mi)∩T[n], that is, Cn(Mi)=1 in both the C(Mi) area and the T[n] area, otherwise it is 0; C[M] represents the set of pixels in the image that satisfy the gradient value less than n when the gradient threshold is n: C[n]=∪Cn(Mi)i<n, C[max+1] is the union of all areas, and max is the maximum value of the image grayscale range; initialize C[min+1]=t[min+1], and then gradually iterate.

[0077] Assume that in step n, C[n-1-1] is established. Now consider obtaining C[n] from C[n-1].

[0078] Let S represent the set of connected components in T[n]. For each connected component s∈S[n], there are three cases:

[0079] 1) s∩C[n-1] is an empty set -- C[n]=s+C[n-1]

[0080] 2) s∩C[n-1] contains a connected component in C[n-1] - C[n] = s + C[n-1]

[0081] 3) s∩C[n-1] contains more than one component in C[n-1] - it is necessary to establish a watershed in s and expand s∩C[n-1]; the final set C[n] is the segmented image, set to g′(x, y)

[0082] In step S3, the image segmentation boundary is merged and segmented with the original pathological slice digital image to obtain a segmented pathological slice digital image, specifically:

[0083] The image segmentation boundary g(x, y) is obtained by improving the watershed algorithm, and the image segmentation boundary g(x, y) is merged with the original pathological slice digital image O(x, y). The image segmentation boundary in the merged image is set to the background image color in the original pathological slice digital image O(x, y) to obtain the segmented pathological slice digital image S(X, Y).

[0084] In step S4, the features of the segmented pathological slice digital image are extracted, and the extracted features are classified to obtain the first cell data, specifically:

[0085] The segmented pathological slice digital image S (X, Y) is subjected to relevant feature extraction, including color feature c, area feature a, inertia rate feature g, and texture feature r, and the extracted relevant features are classified using a SVM classifier to obtain the first cell data Res1, wherein Res1 includes relevant first positive cell data (including the number of positive cells, position information in the image, and the segmented first positive cell image) and first negative cell data (including the number of negative cells, position information in the image, and the segmented first negative cell image).

[0086] Using classification SVM, in the case of linear separability, there will be a hyperplane that completely separates the training samples. The hyperplane can be described as:

[0087] w·x+b=0, (7)

[0088] Among them, “·” is the dot product, w is an n-dimensional vector, and b is the offset.

[0089] The optimal hyperplane is a plane that maximizes the distance between the vector closest to each type of data and the hyperplane.

[0090] The optimal hyperplane can be obtained by solving the following quadratic optimization problem:

[0091]

[0092] Satisfy the constraints: y i (w·x i +b)≥1, i=1, 2, 3,..., n, (9)

[0093] When the number of features is particularly large, this quadratic programming problem can be transformed into its dual problem:

[0094]

[0095]

[0096] b * =y i -w·x i , (12)

[0097] Satisfy the constraints:

[0098] ∑ i＝1 α i y i =0,α i ≥0, i=1, 2,...n; (13)

[0099] Here α=(α1,……,α n ) is the Lagrange multiplier, is the normal vector of the optimal hyperplane, and is the offset of the optimal hyperplane. In the solution and analysis of this type of optimization problem, the KKT condition will play an important role. In formula (13), its solution must satisfy:

[0100] α i {y i (w*x+b)-1}, i=1, 2,...n, (14)

[0101] From formula (11), we can see that those samples with =0 have no effect on classification, and only those samples with >0 have an effect on classification. These samples are called support vectors, so the final classification function is:

[0102]

[0103] Determine the ownership of x based on the sign of f(x).

[0104] In step S4, the segmented pathological slice digital image S (X, Y) is detected using the Fast-R-CNN network pre-training model, the ROI region of interest on the relevant slice is selected as a patch image block, all patches are detected and classified, and the final detection results of all patches are obtained;

[0105] The detection and classification process of each patch is:

[0106] Use the trained neural network model to detect each patch and remove duplicate detection frames of cells in the overlapped part when the patch is removed.

[0107] Traverse all detection boxes on the current patch and remove duplicate detection boxes of two different categories of the same cell. After the traversal is completed, the final detection result of the patch is obtained;

[0108] Among them, the specific method of deduplication of detection frames between different classes is:

[0109] When calculating the size of the intersection-over-union (IOU) ratio for each pair of detection frames of different categories, if the IOU is greater than the preset threshold, the one with lower confidence among the two detection frames of different categories is deleted, and the detection frame with higher confidence is retained;

[0110] The second cell data Res2 is acquired, wherein Res2 includes second positive cell data and second negative cell data.

[0111] In step S5, the first cell data and the second cell data are combined and deduplicated to obtain the third cell data, specifically:

[0112] Compare Res2 and Res1, take the union and remove duplicates to obtain the third cell data Res3, where Res3 includes the third negative cell data and the third positive cell data;

[0113] Specifically, the acquired third negative cells are counted, and the acquired number is counted to obtain final negative cell data;

[0114] The obtained third positive cells are counted, and the number obtained is counted to obtain the final positive cell data.

[0115] In addition, for ER / PR cells, the acquired third positive cell image is grayed out, and the parameter data obtained by target detection is used to obtain the position information of each cell. The third positive cell position coordinates (x, y) are selected, and the grayscale value of the position coordinates is obtained to prepare for positive cell grading (positive cell grading refers to the division of positive cells into different levels according to the different color depths).

[0116] Through experiments, we know that when testing on a common PC host, the recognition of 1080P images can be completed within 1 second, and the recognition rate is about 96%, which is much lower than the manual recognition judgment time of about 5 minutes, indicating the effectiveness of the algorithm. It can also solve the problems of over-segmentation and under-segmentation in the cell segmentation process.

[0117] Embodiment 2

[0118] This embodiment provides a computer vision-based cell classification and counting system, including:

[0119] An image acquisition module is configured to acquire original pathological slice digital images and perform preprocessing;

[0120] The image processing module is configured to segment the preprocessed pathological slice digital image based on an improved watershed algorithm to obtain an image segmentation boundary;

[0121] An image segmentation module is configured to merge and segment the image segmentation boundary with the original pathological slice digital image to obtain a segmented pathological slice digital image;

[0122] The image cell classification module is configured to extract features of the segmented pathological slice digital image and classify the extracted features to obtain first cell data; and at the same time, the segmented pathological slice digital image is detected using a neural network training model to obtain second cell data;

[0123] The image cell counting module is configured to obtain positive and negative cell data by taking the union of the first cell data and the second cell data and removing duplicates.

[0124] It should be noted that the examples and application scenarios implemented by the above modules and corresponding steps are the same, but are not limited to the contents disclosed in the above embodiment 1. It should be noted that the above modules as part of the system can be executed in a computer system such as a set of computer executable instructions.

[0125] Embodiment 3

[0126] This embodiment provides a computer-readable storage medium on which a computer program is stored. When the program is executed by a processor, the steps in the computer vision-based cell classification and counting method described in the first embodiment are implemented.

[0127] Embodiment 4

[0128] This embodiment provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, the steps in a cell classification and counting method based on computer vision as described in the first embodiment above are implemented.

[0129] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of hardware embodiments, software embodiments, or embodiments combining software and hardware. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage and optical storage, etc.) containing computer-usable program code.

[0130] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0131] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0132] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0133] A person skilled in the art can understand that all or part of the processes in the above-mentioned embodiments can be implemented by instructing the relevant hardware through a computer program, and the program can be stored in a computer-readable storage medium, and when the program is executed, it can include the processes of the embodiments of the above-mentioned methods. The storage medium can be a disk, an optical disk, a read-only memory (ROM) or a random access memory (RAM), etc.

[0134] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A cell classification and counting method based on computer vision, characterized in that: include: Acquire original pathological slice digital images and perform preprocessing; Based on the improved watershed algorithm, the preprocessed digital image of pathological sections is segmented to obtain the image segmentation boundary; Merge and segment the image segmentation boundary with the original pathological slice digital image to obtain a segmented pathological slice digital image; Extracting features of the segmented digital image of the pathological slice, and classifying the extracted features to obtain first cell data; and detecting the segmented digital image of the pathological slice using a neural network training model to obtain second cell data; Taking the union of the first cell data and the second cell data and removing duplicates to obtain positive and negative cell data; The features of the segmented pathological slice digital image are extracted, and the extracted features are classified to obtain the first cell data: Extracting the color features, area features, inertia rate features and texture features of the segmented pathological slice digital image; The features of the extracted digital images of case slices are classified using the SVM classifier; Get first cell data; At the same time, the segmented pathological slice digital image is detected using a neural network training model to obtain second cell data: The segmented digital image of the pathological slice is detected using a neural network training model, the ROI region of interest on the slice is selected as a patch image block, all patches are detected and classified, and the final detection results of all patches are obtained; The detection and classification process of each patch is: Use the trained neural network model to detect each patch and remove duplicate detection frames of cells in the overlapped part when the patch is removed. Traverse all detection boxes on the current patch and remove duplicate detection boxes of two different categories of the same cell. After the traversal is completed, the final detection result of the patch is obtained; Get the second cell data; The first cell data and the second cell data are combined and deduplicated to obtain positive and negative cell data: Compare the first cell data and the second cell data, take a union, and remove duplicates to obtain third cell data, wherein the third cell data includes third negative cell data and third positive cell data; Counting the obtained third negative cells, and calculating the number of obtained cells to obtain the final negative cell data; The obtained third positive cells are counted, and the number obtained is counted to obtain the final positive cell data.

2. A computer vision-based cell classification and counting method as claimed in claim 1, characterized in that: The original pathological slice digital image is obtained and preprocessed, specifically: Step 1: Use the pathology section equipment to take samples and obtain a pathology section digital image with high definition, aligned borders and good quality, that is, the original pathology section digital image; Step 2: performing color channel separation on the obtained digital image of the pathological section to obtain digital images of the pathological section in different color channels; Step 3: Perform threshold image segmentation on the digital images of pathological sections of different color channels; Step 4: For the data obtained in step 3, morphological processing is used to remove impurities in the digital image of the pathological section to obtain a preprocessed digital image of the pathological section.

3. A computer vision-based cell classification and counting method as claimed in claim 1, characterized in that: The improved watershed algorithm is used to segment the preprocessed pathological slice digital image to obtain the image segmentation boundary, specifically: Performing filtering and noise reduction processing on the preprocessed digital image of the pathological slice to obtain a digital image of the pathological slice after filtering and noise reduction processing; Performing transformation processing on the digital image of the pathological slice after filtering and noise reduction processing to obtain the digital image of the pathological slice after transformation processing; The transformed digital image of the pathological slice is subjected to threshold processing to obtain the image segmentation boundary.

4. A computer vision-based cell classification and counting method as claimed in claim 1, characterized in that: The image segmentation boundary is merged and segmented with the original pathological slice digital image to obtain the segmented pathological slice digital image, specifically: The image segmentation boundary is obtained by improving the watershed algorithm, and the image segmentation boundary is merged with the original pathological slice digital image. The image segmentation boundary in the merged image is set to the background image color in the original pathological slice digital image to obtain the segmented pathological slice digital image.

5. A cell classification and counting system based on computer vision, characterized in that: include: An image acquisition module is configured to acquire original pathological slice digital images and perform preprocessing; The image processing module is configured to segment the preprocessed pathological slice digital image based on an improved watershed algorithm to obtain an image segmentation boundary; An image segmentation module is configured to merge and segment the image segmentation boundary with the original pathological slice digital image to obtain a segmented pathological slice digital image; The image cell classification module is configured to extract features of the segmented pathological slice digital image and classify the extracted features to obtain first cell data; and at the same time, the segmented pathological slice digital image is detected using a neural network training model to obtain second cell data; An image cell counting module is configured to obtain positive and negative cell data by taking the union of the first cell data and the second cell data and removing duplicates; The features of the segmented pathological slice digital image are extracted, and the extracted features are classified to obtain the first cell data: Extracting the color features, area features, inertia rate features and texture features of the segmented pathological slice digital image; The features of the extracted digital images of case slices are classified using the SVM classifier; Get first cell data; At the same time, the segmented pathological slice digital image is detected using a neural network training model to obtain second cell data: The segmented digital image of the pathological slice is detected using a neural network training model, the ROI region of interest on the slice is selected as a patch image block, all patches are detected and classified, and the final detection results of all patches are obtained; The detection and classification process of each patch is: Use the trained neural network model to detect each patch and remove duplicate detection frames of cells in the overlapped part when the patch is removed. Traverse all detection boxes on the current patch and remove duplicate detection boxes of two different categories of the same cell. After the traversal is completed, the final detection result of the patch is obtained; Get the second cell data; The first cell data and the second cell data are combined and deduplicated to obtain positive and negative cell data: Compare the first cell data and the second cell data, take a union, and remove duplicates to obtain third cell data, wherein the third cell data includes third negative cell data and third positive cell data; Counting the obtained third negative cells, and calculating the number of obtained cells to obtain the final negative cell data; The obtained third positive cells are counted, and the number obtained is counted to obtain the final positive cell data.

6. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the steps in a computer vision-based cell classification and counting method as described in any one of claims 1 to 4 are implemented.

7. A computer device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the program, the steps in the computer vision-based cell classification and counting method as described in any one of claims 1 to 4 are implemented.

Images