Cervical cell identification method and system

Abstract

The invention provides a cervical cell identification method and system. The method comprises the following steps: identifying a qualified visual field area in a pathological image to be detected; extracting contour coordinate information of each to-be-detected cervical cell in the qualified view area, and segmenting and extracting a feature vector of each to-be-detected cervical cell; and querying in a query set based on the feature vector of each to-be-detected cervical cell to obtain the category of each to-be-detected cervical cell. According to the method, the image quality of each view area in the pathological image to be detected is analyzed, so that the output reliability of the final recognition result is ensured, and prediction deviation and errors caused by low-quality data are avoided; each cervical cell in the qualified visual field area is identified, so that main targets to be detected are not missed, and the workload of labeling is reduced by learning the common morphological characteristics of the bottom layers of various cervical cells; through a query classification mode, the satisfactory identification accuracy can be achieved only through fewer abnormal cell annotations compared with a traditional method.

Description

Technical Field

[0001] The present invention relates to the field of pathological cytology detection, and more particularly, to a cervical cell recognition method and system. Background Art

[0002] At present, the main methods for cervical cancer screening mainly include human papillomavirus (HPV) nucleic acid detection and cytological examination, and those with positive HPV detection usually need further cytological examination for triage. The current mainstream cytological examination method is liquid-based thin-layer cytological examination (TCT), which has high specificity, but its sensitivity largely depends on the technical level and experience of cytologists, and the lack of experienced pathological cytologists is a common social problem. On the other hand, the inspection method that completely relies on manual film reading for judgment also has the shortcomings of low efficiency and uncontrollable quality.

[0003] In recent years, the rapidly developing artificial intelligence (AI) technology has the potential to overcome the above obstacles and can significantly improve the efficiency of film reading and screening. Based on digital microscopic images, the algorithm model can quickly locate and detect cervical abnormal cells according to their special morphological, color, karyotype, boundary and other characteristics, so as to make further interpretations of the samples (such as normal / abnormal detection or abnormal cell category judgment, etc.). As a low-cost auxiliary diagnosis tool, this technology can overcome the time limit of manual film reading and the shortage of technical personnel, and avoid biases caused by subjective factors, which is helpful for implementing cervical cancer screening at the social level in resource-poor areas, and also has a broad market prospect.

[0004] At present, many domestic and foreign enterprises are deploying and developing TCT AI-assisted diagnosis products, but the technical routes are highly similar, and there is generally a bottleneck of insufficient generalization performance in the industry. Specifically, a relatively common solution is to collect multi-center TCT pathological image files for large-scale annotation to form an abnormal cell data set, and then train a supervised recognition or segmentation model for this type of abnormal cells. Given the complexity of cervical cytological diagnosis criteria, the diversity of abnormal cell categories and the non-standardization of pathological slides, these models often have more or less performance degradation on samples produced by other medical institutions, thus affecting the quality of clinical diagnosis (such as the missed detection of important abnormal cells leading to false negatives, etc.). In addition, large-scale clinical cell image annotation also has significant time and economic costs (acquisition of high-quality annotation data sources and long-term investment of senior cytologists). Summary of the Invention

[0005] In view of the technical problems existing in the prior art, the present invention provides a cervical cell recognition method and system, which can overcome the problem of low efficiency of model training that requires a large amount of annotation.

[0006] According to a first aspect of the present invention, there is provided a method for cervical cell recognition, including: Identifying a qualified field of view area in a pathological image to be detected; Extracting contour coordinate information of each cervical cell to be detected in the qualified field of view area; Based on the contour coordinate information of each cervical cell to be detected, segmenting each cervical cell image to be detected from the qualified field of view area; Extracting a feature vector of each cervical cell to be detected from each cervical cell image to be detected; Based on the feature vector of each cervical cell to be detected, querying in a query set to obtain the category of each cervical cell to be detected, where the query set includes feature vectors of cells of multiple categories; Determining the screening and recognition classification of the pathological image to be detected according to the categories of all cervical cells to be detected in the pathological image to be detected.

[0007] According to a second aspect of the present invention, there is provided a cervical cell recognition system, including: An identification module for identifying a qualified field of view area in a pathological image to be detected; A first extraction module for extracting contour coordinate information of each cervical cell to be detected in the qualified field of view area; A segmentation module for segmenting each cervical cell image to be detected from the qualified field of view area based on the contour coordinate information of each cervical cell to be detected; A second extraction module for extracting a feature vector of each cervical cell to be detected from each cervical cell image to be detected; A category query module for querying in a query set based on the feature vector of each cervical cell to be detected to obtain the category of each cervical cell to be detected, where the query set includes feature vectors of cells of multiple categories; determining the screening and recognition classification of the pathological image to be detected according to the categories of all cervical cells to be detected in the pathological image to be detected.

[0008] The cervical cell recognition method and system provided by the present invention ensure the reliability of the output of the final recognition result by analyzing the image quality of each field of view area in the pathological image to be detected, avoiding prediction deviation and errors caused by low-quality data; identifying each cell in the qualified field of view area ensures that the main targets to be detected are not missed, and reduces the workload of annotation by learning the underlying common morphological features of various cervical cells; by means of query classification, only relatively fewer abnormal cell annotations are required compared to traditional methods to achieve a satisfactory diagnostic accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] Figure 1Flowchart of a cervical cell recognition method provided by an embodiment of the present invention; Figure 2 Schematic structural diagram of a cervical cell recognition system provided by an embodiment of the present invention; Figure 3 Schematic hardware structure diagram of a possible electronic device provided by the present invention; Figure 4 Schematic hardware structure diagram of a possible computer-readable storage medium provided by the present invention. Detailed implementation manners

[0010] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Apparently, the described embodiments are some, but not all, of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention. In addition, the technical features in each embodiment or a single embodiment provided by the present invention can be combined with each other arbitrarily to form a feasible technical solution. Such combination is not restricted by the order of steps and / or the pattern of structural composition, but must be based on the fact that those of ordinary skill in the art can implement it. When the combination of technical solutions is contradictory or cannot be implemented, it should be considered that such combination of technical solutions does not exist and is not within the protection scope required by the present invention.

[0011] Figure 1 Flowchart of a cervical cell recognition method provided by an embodiment of the present invention, as Figure 1 shown, the method includes: Step 1, identify a qualified field-of-view area in the pathological image to be detected.

[0012] In an embodiment of the present invention, the identifying a qualified field-of-view area in the pathological image to be detected includes: dividing the pathological image to be detected into multiple field-of-view areas according to a fixed-size sliding window; analyzing the image quality of each of the field-of-view areas based on an image quality control model to obtain a quality control result for each of the field-of-view areas, where the quality control result includes that the image quality of the field-of-view area is qualified and the image quality of the field-of-view area is unqualified.

[0013] Specifically, the pathological image to be detected is divided into multiple image blocks according to a fixed sliding window, each image block is a field-of-view area, and the image quality of each field-of-view area is analyzed based on a trained image quality control model to identify the field-of-view areas with qualified image quality and the field-of-view areas with unqualified image quality in the pathological image to be detected. Subsequently, only the field-of-view areas with qualified image quality are processed.

[0014] Among them, the process of constructing the image quality control model is as follows: collect high-resolution pyramid pathological image data of TCT samples (such as TMAP, SVS or TIFF formats, ensuring a certain richness in specimen imaging, covering as many real usage scenarios as possible without specific requirements for sample identification and classification). Horizontally and vertically traverse the imaging layer at a specified magnification (20X in this embodiment) with a fixed-size sliding window patch (1280*1280 pixels in this embodiment) to obtain a patch data set D1, and the patch data set D1 includes a large number of image patches. Manually screen out unqualified samples with significant image quality problems in the patches (such as severely blurred scanning due to defocus, large-area stains, etc. that seriously affect diagnosis), and then randomly sample positive samples in the remaining clear-quality patch samples with a total quantity close to that of the unqualified samples to jointly construct the training set D_qc of the quality control classification model (300 positive and negative patch samples each in this embodiment).

[0015] Based on the obtained data set D_qc, train the image quality control model M_qc (using the ResNet34 network model in this embodiment), and retain the model weight file after completion.

[0016] Input each field area after dividing the pathological image to be detected into the trained image quality control model, output the quality control results of each field area, and obtain the qualified field areas and unqualified field areas among these field areas.

[0017] Step 2, extract the contour coordinate information of each cervical cell to be detected in the qualified field areas.

[0018] In an embodiment of the present invention, extracting the contour coordinate information of each cervical cell to be detected in the qualified field areas includes: Based on the pan-cell recognition model, extract the contour coordinate information of each cervical cell to be detected in the qualified field areas.

[0019] Among them, training the pan-cell recognition model includes: Collect multiple qualified field areas, and label the position contours of each target cell in each qualified field area based on an image annotation tool, where the position contours are polygon contours; Obtain a training data set, where the training data set includes multiple training samples, and each training sample includes a qualified field area and the position contours of each target cell therein; Train the pan-cell recognition model based on the training data set.

[0020] It can be understood that a pan-cell recognition model is constructed and trained, and the contour coordinate information of each cervical cell to be detected in each qualified field area is extracted using the pan-cell recognition model.

[0021] Among them, the process of constructing the pan-cellular recognition model includes: Constructing a training data set: Reusing the D1 data set of the above steps, selecting a part from the samples with qualified clear image quality for cell-level annotation (100 cases are taken in this embodiment, covering the main cell morphology categories). Specifically, using an image annotation tool (labelme is used in this embodiment), the contour of each target cell in the image sample is accurately annotated with a polygon to obtain the corresponding mask of the image, that is, the coordinate contour information of each target cell. The annotated image patch and the mask mask are used to form the training data set D_det of the pan-cellular recognition model.

[0022] Training the pan-cellular recognition model M_det (Mask-RCNN network model is used in this embodiment) based on the obtained training data set D_det, and retaining the model weight file after completion.

[0023] Based on the trained pan-cellular recognition model, the position coordinate contour information of each cervical cell to be detected in each qualified visual field area of the pathological image to be detected.

[0024] Step 3, based on the contour coordinate information of each cervical cell to be detected, segmenting each cervical cell image to be detected from the qualified visual field area.

[0025] In an embodiment of the present invention, the segmenting each cervical cell image to be detected from the qualified visual field area based on the contour coordinate information of each cervical cell to be detected further includes: based on the contour coordinate information of each cervical cell to be detected, segmenting the mask mask of each cervical cell to be detected from the qualified visual field area, and placing the mask mask of each cervical cell to be detected in the middle of a square white background canvas with a set size according to the morphological ratio to obtain a single cell image.

[0026] Specifically, each cervical cell image to be detected is segmented and intercepted from the original sample image through the contour coordinate information of each cervical cell to be detected, and is placed in the middle of a square white background canvas with a suitable size according to the morphological ratio to obtain a single cell picture data set D_cell. In the embodiment of the present invention, the cell image is accurately segmented by contour interception instead of position rectangle frame interception, because it is to avoid information interference caused by other adjacent cells or overlapping cells within the position frame.

[0027] Step 4, extracting the feature vector of each cervical cell to be detected from each cervical cell image to be detected.

[0028] In an embodiment of the present invention, the extracting the feature vector of each cervical cell to be detected from each cervical cell image to be detected includes: Input each cervical cell image to be tested into the self-supervised basic model to obtain the feature vector of each cervical cell to be tested.

[0029] The training process of the self-supervised basic model includes: Obtain normal cervical cell sample images and abnormal cervical cell sample images respectively, and form multiple sample images; Extract qualified visual field regions from each sample image, and segment each target cell image in each qualified visual field region as a sample target cell image; Train the self-supervised basic model to extract the feature vector of the sample target cell image based on the sample target cell image.

[0030] Specifically, after each cervical cell image to be tested is segmented from the original image, input each segmented cervical cell image to be tested into the self-supervised basic model to obtain the feature vector of each cervical cell to be tested.

[0031] It should be noted that the training of the self-supervised basic model does not require annotation and is an unsupervised learning process, self-learning the feature vectors of target cells. Since the dataset does not require annotation, a large amount of annotation time will be saved.

[0032] The training dataset of the self-supervised basic model is the target cell images segmented according to step 3. The dataset D_cell is used to train the self-supervised model M_cell (in this embodiment, DINOv2 is used). After completion, the model weight file is retained (the inference output of this model is a vector with a shape of [1, 1024]).

[0033] Step 5, based on the feature vector of each cervical cell to be tested, query in the query set to obtain the category of each cervical cell to be tested. The query set includes the feature vectors of target cells of multiple categories.

[0034] In an embodiment of the present invention, the querying in the query set based on the feature vector of each cervical cell to be tested to obtain the category of each cervical cell to be tested, where the query set includes the feature vectors of target cells of multiple categories, includes: Calculate the similarity between the feature vector of any cervical cell image to be tested and the feature vectors of multiple target cells of each category in the query set, and use the category with the highest similarity as the category of any cervical cell to be tested.

[0035] Among them, first, a query set is constructed. The specific process of constructing the query set is as follows: Select typical positive cells of each type from each target cell, and construct the query set Q_cell together with a part of randomly sampled normal cells. In the embodiment of the present invention, assume there are N types of cells (normal cervix is one type, abnormal cervix includes N - 1 types, a total of N types), then Q_cell contains N subsets, each type includes m cells and the feature vectors of m cells. For example, the feature vectors of 100 target cells are taken for each query category.

[0036] Calculate the similarity between the feature vector of each cervical cell to be detected and the m target cells of each category in the query set, obtaining m similarities between the cervical cell to be detected and the m cells of each category. Screen out m' similarities greater than the preset similarity threshold from the m similarities, calculate the average of the m' similarities, obtaining the average similarity between the cervical cell to be detected and the cells of each category in the query set, and take the category with the highest average similarity as the category of the cervical cell to be detected.

[0037] Step 6: Determine the screening and recognition classification of the to-be-detected pathological image according to the categories of all cervical cells to be detected in the to-be-detected pathological image.

[0038] Specifically, the to-be-detected pathological image includes multiple cervical cells to be detected. According to Step 5, obtain the category of each cervical cell to be detected, and take the category with the highest degree of abnormality among the categories of all cervical cells to be detected as the abnormal category of the cervical cells in the to-be-detected pathological image.

[0039] It should be noted that for the unqualified regions in the to-be-detected pathological image, the categories also need to be identified according to the above model. The specific steps include: For each unqualified visual field region in the to-be-detected pathological image, extract the contour coordinate information of each cervical cell to be detected in the unqualified visual field region; Based on the contour coordinate information of each cervical cell to be detected, segment each cervical cell image from the unqualified visual field region; Extract the feature vector of each cervical cell to be detected from each cervical cell image; Based on the feature vector of each cervical cell to be detected, query in the query set to obtain the category of each cervical cell to be detected. The query set includes the feature vectors of cervical cells of multiple categories; Determine the category of the unqualified visual field region in the to-be-detected pathological image according to the categories of all cervical cells to be detected in all the unqualified visual field regions in the to-be-detected pathological image.

[0040] So far, the categories of qualified visual field regions and unqualified visual field regions in the pathological image to be detected are respectively identified. Finally, the quality control results of each visual field region of the pathological image to be detected, as well as the position contours and abnormal categories of each cervical cell to be detected, are marked on the pathological image to be detected for visual display.

[0041] See Figure 2 , which provides a cervical cell recognition system according to an embodiment of the present invention. The system includes: An identification module 201 for identifying qualified visual field regions in a pathological image to be detected; A first extraction module 202 for extracting the contour coordinate information of each cervical cell to be detected in the qualified visual field region; A segmentation module 203 for segmenting each cervical cell image to be detected from the qualified visual field region based on the contour coordinate information of each cervical cell to be detected; A second extraction module 204 for extracting the feature vector of each cervical cell to be detected from each cervical cell image to be detected; A category query module 205 for querying in a query set based on the feature vector of each cervical cell to be detected to obtain the category of each cervical cell to be detected, where the query set includes the feature vectors of cells of multiple categories; and determining the screening and identification classification of the pathological image to be detected according to the categories of all cervical cells to be detected in the pathological image to be detected.

[0042] It can be understood that a cervical recognition system provided by the present invention corresponds to the cervical cell recognition methods provided in the foregoing embodiments. The relevant technical features of the cervical cell recognition system can refer to the relevant technical features of the cervical cell recognition method, which will not be elaborated here.

[0043] Please refer to Figure 3 , Figure 3 which is a schematic diagram of an embodiment of an electronic device provided by an embodiment of the present invention. As Figure 3 shown, an embodiment of the present invention provides an electronic device 300, including a memory 310, a processor 320, and a computer program 311 stored on the memory 310 and executable on the processor 320. When the processor 320 executes the computer program 311, the steps of the cervical cell recognition method are implemented.

[0044] Please refer to Figure 4 , Figure 4 which is a schematic diagram of an embodiment of a computer-readable storage medium provided by the present invention. As Figure 4 shown, this embodiment provides a computer-readable storage medium 400, on which a computer program 411 is stored. When the computer program 411 is executed by a processor, the steps of the cervical cell recognition method are implemented.

[0045] A cervical cell recognition method and system provided by an embodiment of the present invention ensure the reliability of the output of the final recognition result by analyzing the image quality of each field of view area in the pathological image to be detected, and avoid prediction deviation and errors caused by low-quality data. Each cell in the qualified field of view area is recognized, ensuring that the main targets to be detected are not missed, and the self-supervised basic model can learn the underlying common morphological features of various cervical cells without additional annotation, reducing the workload of annotation; through the method of query classification, only relatively fewer abnormal cell annotations are required compared to traditional methods to achieve a satisfactory diagnostic accuracy rate.

[0046] It should be noted that in the above embodiments, the descriptions of each embodiment have their own emphases. For the parts not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0047] Those skilled in the art should understand that the embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can be implemented in the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present invention can be implemented in the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0048] The present invention is described with reference to the flowcharts and / or block diagrams of methods, apparatuses (systems), and computer program products according to the embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, and the combination of processes and / or blocks in the flowchart and / or block diagram, can be realized by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded computer, or other programmable data processing devices to generate a machine, such that the instructions executed by the processor of the computer or other programmable data processing devices generate a device for realizing the functions specified in Figure 1 one process or multiple processes and / or blocks Figure 1 one block or multiple blocks.

[0049] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a specific manner, such that the instructions stored in the computer-readable memory generate a manufactured article including an instruction device, and the instruction device realizes the functions specified in Figure 1 one process or multiple processes and / or blocks Figure 1 one block or multiple blocks.

[0050] These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus, causing a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process, so that the instructions executed on the computer or other programmable apparatus provide steps for implementing the functions specified in one process or a plurality of processes and / or blocks Figure 1 one process or a plurality of processes and / or blocks Figure 1 and / or steps for implementing the functions specified in one block or a plurality of blocks.

[0051] Although the preferred embodiments of the present invention have been described, additional changes and modifications can be made by those skilled in the art once they learn of the basic inventive concept. Therefore, the appended claims are intended to be construed to cover the preferred embodiments as well as all changes and modifications falling within the scope of the present invention.

[0052] 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 equivalent technologies, the present invention is also intended to include these modifications and variations.

Claims

1. A method for identifying cervical cells, characterized in that: include: Identify qualified visual field areas in the pathological image to be detected; Extracting the contour coordinate information of each cervical cell to be tested in the qualified visual field area; Based on the contour coordinate information of each cervical cell to be tested, segmenting each cervical cell image to be tested from the qualified visual field area; Extracting a feature vector of each cervical cell to be tested from each cervical cell image to be tested; Based on the feature vector of each cervical cell to be tested, querying in a query set to obtain the category of each cervical cell to be tested, wherein the query set includes feature vectors of multiple categories of cervical cells; According to the categories of all the cervical cells to be detected in the pathological image to be detected, the screening and identification classification of the pathological image to be detected is determined.

2. The cervical cell identification method according to claim 1, characterized in that: The step of identifying a qualified visual field area in a pathological image to be detected includes: The pathological image to be detected is divided into multiple visual field areas according to a fixed-size sliding window; The image quality of each of the field of view areas is analyzed based on the image quality control model to obtain a quality control result for each of the field of view areas, wherein the quality control result includes whether the image quality of the field of view area is qualified or unqualified.

3. The cervical cell identification method according to claim 1, characterized in that: The step of extracting the contour coordinate information of each cervical cell to be detected in the qualified visual field area includes: Extracting the contour coordinate information of each cervical cell to be detected in the qualified visual field area based on the pan-cell recognition model; Wherein, training the pan-cellular recognition model comprises: Collecting a plurality of qualified visual field areas, and annotating the position outline of each target cell in each of the qualified visual field areas based on an image annotation tool, wherein the position outline is a polygonal outline; Acquire a training data set, wherein the training data set includes a plurality of training samples, each of which includes a qualified visual field area and a position profile of each target cell therein; The pan-cell recognition model is trained based on the training data set.

4. The cervical cell identification method according to claim 1, characterized in that: The method of segmenting each image of the cervical cells to be tested from the qualified visual field area based on the contour coordinate information of each cervical cell to be tested also includes: Based on the contour coordinate information of each cervical cell to be tested, the mask of each cervical cell to be tested is segmented from the qualified field of view, and the mask of each cervical cell to be tested is placed in the middle of a square white background canvas of a set size according to the morphological proportion to obtain a single cell image.

5. The cervical cell identification method according to claim 1, characterized in that: The step of extracting a feature vector of each cervical cell to be tested from each cervical cell image to be tested comprises: Input each cervical cell image to be tested into the self-supervised basic model to obtain the feature vector of each cervical cell to be tested; The training process of the self-supervised basic model includes: Acquire normal cervical cell sample images and abnormal cervical cell sample images respectively to form a plurality of sample images; Extracting a qualified visual field area from each sample image, and segmenting each target cell image in each qualified visual field area as a sample target cell image; Based on the sample target cell image, the self-supervised basic model extracts a feature vector of the sample target cell image for training.

6. The method for identifying cervical cells according to claim 1, characterized in that: The feature vector based on each cervical cell to be tested is used to query the query set to obtain the category of each cervical cell to be tested, wherein the query set includes feature vectors of multiple categories of cells, including: The similarity between the feature vector of any cervical cell image to be tested and the feature vectors of multiple target cells of each category in the query set is calculated, and the category with the highest similarity is used as the category of any cervical cell to be tested.

7. The method for identifying cervical cells according to claim 1 or 6, characterized in that: Determining the screening and identification classification of the pathological image to be detected according to the categories of all the cervical cells to be detected in the pathological image to be detected includes: The category with the highest abnormality among all the categories of cervical cells to be detected in the pathological image to be detected is used as the screening and identification category of the pathological image to be detected.

8. The method for identifying cervical cells according to claim 1, characterized in that: Also includes: For each unqualified visual field area in the pathological image to be detected, extracting the contour coordinate information of each cervical cell to be detected in the unqualified visual field area; Based on the contour coordinate information of each cervical cell to be tested, segmenting each cervical cell image to be tested from the unqualified visual field area; Extracting a feature vector of each cervical cell to be tested from each cervical cell image to be tested; Based on the feature vector of each cervical cell to be tested, querying in a query set to obtain the category of each cervical cell to be tested, wherein the query set includes feature vectors of cells of multiple categories; According to the categories of all the cervical cells to be detected in all the unqualified visual field areas in the pathological image to be detected, the screening and identification classification of the unqualified visual field areas in the pathological image to be detected is determined.

9. The method for identifying cervical cells according to claim 1 or 8, characterized in that: The categories of all the cervical cells to be detected in the pathological image to be detected are determined to screen and identify the pathological image to be detected, and then the following steps are further included: The quality control results of each visual field area of ​​the pathological image to be detected, as well as the position outline of each cervical cell to be detected and the corresponding cervical cell abnormality category are marked on the pathological image to be detected for visual display.

10. A cervical cell identification system, characterized in that: include: A recognition module, used for identifying a qualified visual field area in a pathological image to be detected; A first extraction module is used to extract the contour coordinate information of each cervical cell to be detected in the qualified visual field area; A segmentation module, used for segmenting each image of the cervical cells to be tested from the qualified visual field area based on the contour coordinate information of each cervical cell to be tested; The second extraction module is used to extract the feature vector of each cervical cell to be tested from each cervical cell image to be tested; A category query module is used to query a query set based on the feature vector of each cervical cell to be tested to obtain the category of each cervical cell to be tested, wherein the query set includes feature vectors of multiple categories of cells; and determine the screening and identification category of the pathological image to be tested according to the categories of all cervical cells to be tested in the pathological image to be tested.

Images