A method and device for identifying cervical cancer cells

Abstract

The application discloses a cervical cancer cell recognition method and device, and the method comprises the following steps: acquiring a cell microscopic image of a pathological sample; inputting the cell microscopic image into a contour recognition model to output a label containing a contour of a target cell; cutting the cell microscopic image based on the label to obtain an image of each target cell; extracting a cell nucleus feature and a cytoplasm feature in the image of each target cell to obtain a feature vector corresponding to each target cell; and inputting the feature vectors into a cervical cancer cell recognition model in sequence to obtain a confidence degree of the target cell corresponding to each group of feature vectors, wherein the confidence degree represents a probability that the target cell is a cervical cancer cell. Through two-stage processing of the cell microscopic image by the contour recognition model and the cervical cancer cell recognition model, the contour of a single suspected cervical cancer cell is obtained first, and then the single cell is subjected to cervical cancer cell recognition, so that the balance between the accuracy and the efficiency of cervical cancer cell recognition is achieved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of cell image processing, and particularly relates to a cervical cancer cell recognition method and device. BACKGROUND

[0002] Traditional cervical cancer examination methods mainly use artificial interpretation screening methods, which have problems such as high cost, large workload, and reliability and accuracy affected by the professional skills and subjective emotions of doctors. The purpose of cervical cell image recognition technology research is to identify whether there are diseased epithelial cells in the cervical cell image, reduce the workload of doctors, and reduce false positives and false negatives in cervical cell recognition.

[0003] In the prior art, for cervical cancer cell screening, features are usually extracted from the entire cell group image, and then recognition is performed. However, this recognition method is often affected by various non-cancer cells, resulting in insufficient feature matching degree and inaccurate recognition of cervical cancer cells. In addition, due to the large amount of cells in the image, the feature data level is large, resulting in low recognition efficiency. SUMMARY

[0004] The present application solves the technical problems of low accuracy and low efficiency in the process of recognizing cervical cancer cells in the prior art, and provides a cervical cancer cell recognition method and device.

[0005] In one aspect of the embodiment of the present application, a cervical cancer cell recognition method is provided, including the following steps: obtaining a cell microscopic image of a pathological sample; inputting the cell microscopic image into a contour recognition model to output a label containing a contour of a target cell, wherein the contour recognition model is a model pre-trained according to multiple contour shapes of cervical cancer cells, and is used for recognizing the contour of the cervical cancer cell in the image, and the target cell is a cell having a similar contour shape as the cervical cancer cell; cropping the cell microscopic image based on the label to obtain an image of each target cell; extracting a cell nucleus feature and a cytoplasm feature in each target cell image to obtain a feature vector corresponding to each target cell; and inputting the feature vectors into a cervical cancer cell recognition model in sequence to obtain a confidence of the target cell corresponding to each feature vector, wherein the confidence represents the probability that the target cell is a cervical cancer cell.

[0006] Optionally, while the cell microscopic image is cropped based on the mark to obtain the image of each target cell, the method further comprises: determining cell information within a preset range around each target cell, the cell information comprising information about whether the cells within the preset range are target cells and a ratio of target cells within the preset range to the total number of cells; and wherein the feature vector is input into the cervical cancer cell recognition model in sequence, and the cell information is also input into the cervical cancer cell recognition model.

[0007] Optionally, before the cell microscopic image is input into the contour recognition model, the method further comprises: performing enhancement processing on the cell microscopic image to determine the contour of each cell.

[0008] Optionally, the contour recognition model comprises at least two convolution layers, and each convolution layer comprises at least N convolution kernels, wherein N is the number of pre-classifications of the contours of cervical cancer cells, and each convolution kernel corresponds to one contour.

[0009] Optionally, the nucleus features comprise color features, shape features and texture features of the nucleus, and the cytoplasm features comprise color features of the cytoplasm.

[0010] Another aspect of the embodiment of the application further provides a device for recognizing cervical cancer cells, comprising: an acquisition module configured to acquire a cell microscopic image of a pathological sample; a marking module configured to input the cell microscopic image into a contour recognition model to output a mark comprising a contour of a target cell, wherein the contour recognition model is a model pre-trained according to various contour shapes of cervical cancer cells and is used to recognize the contour of a cervical cancer cell in an image, and the target cell is a cell having a similar contour shape to a cervical cancer cell; a cropping module configured to crop the cell microscopic image based on the mark to obtain an image of each target cell; an extraction module configured to extract nucleus features and cytoplasm features in the image of each target cell to obtain a feature vector corresponding to each target cell; and a recognition module configured to input the feature vectors into a cervical cancer cell recognition model in sequence to obtain a confidence degree of each target cell corresponding to each group of feature vectors, wherein the confidence degree represents a probability that the target cell is a cervical cancer cell.

[0011] Optionally, the device further comprises a determination module configured to determine cell information within a preset range around each target cell while the cell microscopic image is cropped based on the mark to obtain the image of each target cell, the cell information comprising information about whether the cells within the preset range are target cells and a ratio of target cells within the preset range to the total number of cells; and the recognition module is further configured to input the cell information into the cervical cancer cell recognition model while the feature vectors are input into the cervical cancer cell recognition model in sequence.

[0012] Another aspect of the embodiment of the present application also provides a computer device, characterized in that comprising: at least one processor; and a memory connected with the at least one processor in communication; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to perform the method as described above.

[0013] Another aspect of the embodiment of the present application also provides a computer readable storage medium, characterized in that the computer readable storage medium stores computer instructions, and the computer instructions are used to make a computer execute the method as described above.

[0014] The technical effects achieved by the present application are as follows:

[0015] According to the embodiment of the present application, the cell microscopic image is processed in two levels through the contour recognition model and the cervical cancer cell recognition model, the contour of a single suspected cervical cancer cell is obtained first, and then the single cell is recognized as a cervical cancer cell, so that the balance between the accuracy and the efficiency of the cervical cancer cell recognition is achieved.

[0016] Since the single cell contains a small amount of characteristic information, the recognition efficiency of the cervical cancer cell is high, and since the single cervical cancer cell is recognized, there are fewer interference factors compared to the recognition of the whole pathological sample image, so that the recognition accuracy of the cervical cancer cell is greatly improved.

[0017] In the process of recognizing the single cervical cancer cell, the population change reaction of the cancer cell is fully considered, the cell information in the preset range around the target cell is taken as the recognition input feature of the cervical cancer cell, so that the recognition accuracy of the cervical cancer cell is further improved, and the misjudgment caused by the change of the single cell is avoided. BRIEF DESCRIPTION OF DRAWINGS

[0018] In order to more clearly illustrate the specific embodiments of the present application or the technical solutions in the prior art, the drawings needed in the specific embodiments or the prior art description will be briefly introduced as follows. Obviously, the drawings in the following description are some embodiments of the present application, and those skilled in the art can also obtain other drawings according to these drawings without creative labor.

[0019] Figure 1 A flow chart of a specific example of the cervical cancer cell recognition method in the embodiment 1 of the present application;

[0020] Figure 2 A principle block diagram of a specific example of the cervical cancer cell recognition device in the embodiment 2 of the present application;

[0021] Figure 3 Fig. 1 is a structural schematic diagram of a computer device of the present application. DETAILED DESCRIPTION

[0022] The technical solutions of the present application will be described clearly and completely below in conjunction with the drawings. Obviously, the described embodiments are part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative efforts fall within the scope of protection of the present application.

[0023] In the description of the present application, it should be noted that the terms "center", "upper", "lower", "left", "right", "vertical", "horizontal", "inner", "outer", and the like indicate the orientation or positional relationship shown in the drawings, and are only for the convenience of describing the present application and simplifying the description, and do not indicate or imply that the device or element referred to must have a particular orientation, be constructed and operated in a particular orientation, and therefore cannot be understood as a limitation on the present application. In addition, the terms "first", "second", "third" are only for descriptive purposes and cannot be understood as indicating or implying relative importance.

[0024] In the description of the present application, it should be noted that unless otherwise explicitly specified and limited, the terms "mounting", "connecting", "connecting" should be understood broadly, for example, it can be fixedly connected, or it can be detachably connected, or integrally connected; it can be mechanically connected, or it can be electrically connected; it can be directly connected, or it can be indirectly connected through an intermediate medium; it can be the internal communication of two elements; it can be wireless connection, or it can be wired connection. For those skilled in the art, the specific meaning of the above terms in the present application can be understood according to the specific circumstances.

[0025] In addition, the technical features involved in the different embodiments of the present application described below can be combined with each other as long as they do not conflict with each other.

[0026] Embodiment 1

[0027] The embodiment of the present application provides a method for identifying cervical cancer cells, which can be used for identifying cervical cancer cells, so as to facilitate the capture of cervical cancer cells, such as Figure 1 As shown, the method comprises the following steps:

[0028] Step S101, obtaining a cell microscopic image of a pathological sample.

[0029] The pathological sample can be a biological sample of the object collected by cervical scraping, and then a microscopic image capable of observing cells is obtained by microscopic imaging. The cell microscopic image contains a large number of cervical cancer cells and also contains some normal cells.

[0030] Step S102, input the cell microscopic image into a contour recognition model, and output a label containing the contour of the target cell, wherein the contour recognition model is a model pre-trained according to multiple contour shapes of cervical cancer cells, and is used to recognize the contour of the cervical cancer cell in the image, and the target cell is a cell having a similar contour shape as the cervical cancer cell.

[0031] After the image is acquired, the cell microscopic image can be pre-processed to obtain the contour of each cell. Specifically, the cell microscopic image can be enhanced before being input into the contour recognition model to determine the contour of each cell. In this way, clear cell contour lines can be obtained, facilitating the labeling of the target cell.

[0032] The contour recognition model can be a model pre-trained for contour labeling of suspected cervical cancer cells (i.e., target cells). Due to pathological changes, the shape of cervical cancer cells changes and the cells increase in size. The shape of cervical cancer cells includes fiber shape, tadpole shape, rhombus shape, and other irregular shapes. During training of the contour recognition model, the shape of each type of cervical cancer cell can be used as a training set to train the contour recognition model. The model can be trained using machine learning to recognize different types of contours and perform corresponding labeling.

[0033] Specifically, in the embodiment of the present application, the contour recognition model includes at least two convolution layers, and each convolution layer includes at least N convolution kernels, wherein N is the number of pre-classification of the contour of the cervical cancer cell, and each convolution kernel corresponds to a contour.

[0034] The contour of the cervical cancer cell is recognized using a convolutional neural network, and multiple convolution layers are used for calculation. Each convolution kernel recognizes the contour of the corresponding shape type, so as to accurately recognize different types of cervical cancer cells.

[0035] In the embodiment of the present application, the label of the target cell can be a contour line data consistent with the shape of the target cell, so as to facilitate reading and recognition by the computer, thereby quickly obtaining the coordinate data of the shape of the target cell.

[0036] Step S103, cropping the cell microscopic image based on the label to obtain an image of each target cell.

[0037] After the contour of the target cell is determined, the target cell on the cell microscopic image is cropped based on the label marked by the contour recognition model, so as to obtain an image of a single target cell, i.e., a single cell image.

[0038] The embodiment of the present application cuts out suspected cervical cancer cells, i.e., target cells, by using contour marks, thereby eliminating normal cells from the original cell microscopic image, including only the target cells of mononuclear cells, reducing the data amount of cervical cancer cell recognition, and improving the recognition efficiency.

[0039] In step S104, the nucleus features and cytoplasm features in the image of each target cell are extracted to obtain a feature vector corresponding to each target cell.

[0040] For each target cell, its nucleus features and cytoplasm features are extracted respectively, and these features are combined to form a feature vector corresponding to the target cell, which prepares for subsequent cervical cancer cell recognition. Specifically, the nucleus features include color features, shape features, and texture features of the nucleus; and the cytoplasm features include color features of the cytoplasm.

[0041] Since cancer cells are cells after pathological changes, their nuclei will change, specifically, the nuclei will be large, and the nucleus-cytoplasm ratio will increase; the nuclei will be deformed, elongated, and the nuclear edge will be jagged, and the nuclei will have concave, long bud, lobed, mulberry, or crescent shapes, etc. Correspondingly, the cytoplasm color will also change after processing. By these features possessed by cervical cancer, it is comprehensively judged whether the target cell is a cervical cancer cell.

[0042] In step S105, the feature vectors are sequentially input into a cervical cancer cell recognition model to obtain a confidence of the target cell corresponding to each group of feature vectors, wherein the confidence represents the probability that the target cell is a cervical cancer cell.

[0043] After the features are extracted and the feature vectors are generated, they are input into a cervical cancer cell recognition model, wherein the cervical cancer cell recognition model is a neural network model obtained by pre-training, and is used to identify whether a mononuclear cell is a cervical cancer cell. Specifically, a large number of mononuclear cervical cancer cell images can be collected, and then the feature vectors of the cervical cancer cells are extracted, and a machine learning technique is used for model training to obtain the cervical cancer cell recognition model. After obtaining the confidence of the target cell, whether the target cell is a cervical cancer cell can be determined based on the confidence. Specifically, a threshold can be set to determine whether the confidence reaches the threshold. When the threshold is reached, it is determined to be a cervical cancer cell; otherwise, it is not a cervical cancer cell.

[0044] Since the mononuclear cells contain a small amount of feature information, the recognition efficiency of the cervical cancer cells is high. At the same time, since it is the recognition of mononuclear cervical cancer cells, there are fewer interference factors compared to the recognition of the entire pathological sample image, which greatly improves the recognition accuracy of the cervical cancer cells.

[0045] According to the embodiment of the present application, the cell microscopic image is processed in two levels through the contour recognition model and the cervical cancer cell recognition model, the contour of a single suspected cervical cancer cell is obtained first, and then the single cell is recognized as a cervical cancer cell, so that the balance between the accuracy and efficiency of the cervical cancer cell recognition is achieved.

[0046] As an optional implementation of the embodiment of the present application, when the cell microscopic image is cropped based on the mark to obtain the image of each target cell, the cell information in the preset range around each target cell is determined, the cell information includes the information whether the cells in the preset range are target cells and the ratio of the target cells in the preset range to the total number of cells, and the feature vector is input into the cervical cancer cell recognition model while the cell information is input into the cervical cancer cell recognition model.

[0047] In the embodiment of the present application, during the process of recognizing the single cervical cancer cell, the population change of the cancer cells is fully considered, the cell information in the preset range around the target cell is used as the input feature of the cervical cancer cell recognition, so that the accuracy of the cervical cancer cell recognition is further improved, and the misjudgment caused by the change of the single cell is avoided.

[0048] Embodiment 2

[0049] The embodiment provides a cervical cancer cell recognition device, which can be used to execute the recognition method of the above-mentioned embodiments, as shown in the following Figure 2 The device comprises:

[0050] The acquisition module 201 is configured to acquire a cell microscopic image of a pathological sample.

[0051] The marking module 202 is configured to input the cell microscopic image into a contour recognition model, and output a mark containing the contour of a target cell, wherein the contour recognition model is a model pre-trained according to multiple contour shapes of cervical cancer cells, and is used to recognize the contour of a cervical cancer cell in an image, and the target cell is a cell having a similar contour shape as the cervical cancer cell.

[0052] The cropping module 203 is configured to crop the cell microscopic image based on the mark to obtain an image of each target cell.

[0053] The extraction module 204 is configured to extract the nucleus feature and the cytoplasm feature in the image of each target cell to obtain a feature vector corresponding to each target cell.

[0054] The identification module 205 is used to input the feature vectors sequentially into the cervical cancer cell identification model to obtain the confidence level of the target cell corresponding to each set of feature vectors, wherein the confidence level represents the probability that the target cell is a cervical cancer cell.

[0055] According to embodiments of the present invention, a two-stage processing method is used to process cell microscopic images through a contour recognition model and a cervical cancer cell recognition model. First, the contour of a single suspected cervical cancer cell is obtained, and then the single cell is identified as a cervical cancer cell, thus achieving a balance between accuracy and efficiency in cervical cancer cell recognition.

[0056] Optionally, the identification device further includes: a determination module, configured to determine cell information within a preset range around each target cell while cropping the cell microscopic image based on the marker to obtain an image of each target cell, wherein the cell information includes whether the cells within the preset range are target cells and the ratio of target cells within the preset range to the total number of cells; the identification module is further configured to input the cell information into the cervical cancer cell identification model while sequentially inputting the feature vector into the cervical cancer cell identification model.

[0057] In this embodiment of the invention, during the identification of individual cervical cancer cells, the population change response of cancer cells is fully considered. Cell information within a preset range around the target cell is used as the input feature for identifying cervical cancer cells, thereby further improving the accuracy of cervical cancer cell identification and avoiding misjudgment caused by changes in individual cells.

[0058] For a detailed description of the device, please refer to the method embodiments, which will not be repeated here.

[0059] Example 3

[0060] In one embodiment of the present invention, a computer device is also provided, the internal structure of which can be shown in the figure below. Figure 3 As shown. The computer device includes a processor and memory connected via a system bus, and may also include a display screen and input devices. The processor provides computing and control capabilities. The memory includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores an operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. When the computer program is executed by the processor, it implements the identification method of the above embodiment. The computer device may also include a display screen and input devices. The display screen may be a liquid crystal display screen or an electronic ink display screen. The input devices may be a touch layer covering the display screen, buttons, a trackball, or a touchpad mounted on the computer device casing, or an external keyboard, touchpad, etc.

[0061] Those skilled in the art can understand that, Figure 3 The structure shown in the figure is only a block diagram of part of the structure related to the scheme of the present application, and does not constitute a limitation on the computer device to which the scheme of the present application is applied. The specific computer device can include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0062] Those of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiments can be completed by instructing relevant hardware through a computer program. The computer program can be stored in a non-volatile computer readable storage medium, and when executed, can include the processes of the above-mentioned embodiments. Any reference to memory, storage, database or other medium used in the embodiments provided by the present application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. As an illustration but not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDR SDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0063] The above-described embodiments only express several implementation manners of the present application, and the description is more specific and detailed, but it should not be understood as a limitation on the scope of the patent of the present application. It should be pointed out that for those of ordinary skill in the art, without departing from the concept of the present application, several modifications and improvements can be made, which all belong to the protection scope of the present application. Therefore, the protection scope of the patent of the present application should be subject to the appended claims.

[0064] The preferred embodiments of the present application are described in detail above. It should be understood that those of ordinary skill in the art can make many modifications and changes without creative labor based on the concept of the present application. Therefore, any technical solution obtained by logical analysis, reasoning or limited experiment based on the prior art according to the concept of the present application shall be within the protection scope determined by the claims.

Claims

1. A method for identifying cervical cancer cells, characterized in that, Includes the following steps: Obtain cellular microscopic images of pathological samples; The cell microscopic image is input into the contour recognition model, and the output is a label containing the contour of the target cell. The contour recognition model is a model pre-trained based on various contour shapes of cervical cancer cells, used to identify the contour of cervical cancer cells in the image. The target cell is a cell with a similar contour shape to cervical cancer cells. The cell microscopic image is cropped based on the markers to obtain an image of each target cell; Determine cell information within a preset range around each target cell, the cell information including whether the cells within the preset range are target cells and the ratio of target cells within the preset range to the total number of cells; Extract the nuclear and cytoplasmic features from the image of each target cell to obtain the feature vector corresponding to each target cell; The feature vector and the cell information are sequentially input into the cervical cancer cell identification model to obtain the confidence level of the target cell corresponding to each set of feature vectors, wherein the confidence level represents the probability that the target cell is a cervical cancer cell.

2. The identification method according to claim 1, characterized in that, Before inputting the cell microscopic image into the contour recognition model, the method further includes: The cell microscopic images are enhanced to determine the outline of each cell.

3. The identification method according to claim 1, characterized in that, The contour recognition model includes at least two convolutional layers, each of which includes at least N convolutional kernels, where N is the number of pre-classified cervical cancer cell contours, and each convolutional kernel corresponds to one contour.

4. The identification method according to claim 1, characterized in that, The nuclear features include: color features, shape features, and texture features of the nuclear nucleus; the cytoplasmic features include the color features of the cytoplasm.

5. A device for identifying cervical cancer cells, characterized in that, include: The acquisition module is used to acquire cellular microscopic images of pathological samples; The labeling module is used to input the cell microscopic image into the contour recognition model and output a label containing the contour of the target cell. The contour recognition model is a model pre-trained based on various contour shapes of cervical cancer cells and is used to identify the contour of cervical cancer cells in the image. The target cell is a cell with a similar contour shape to cervical cancer cells. The cropping module is used to crop the cell microscopic image based on the markers to obtain an image of each target cell; An extraction module is used to extract the nuclear and cytoplasmic features from the image of each target cell to obtain a feature vector corresponding to each target cell. The identification module is used to sequentially input the feature vectors into the cervical cancer cell identification model to obtain the confidence level of the target cell corresponding to each set of feature vectors, wherein the confidence level represents the probability that the target cell is a cervical cancer cell; The determination module is used to determine the cell information within a preset range around each target cell while cropping the cell microscopic image based on the marker to obtain an image of each target cell. The cell information includes whether the cells within the preset range are target cells and the ratio of the target cells within the preset range to the total number of cells. The identification module is also used to input the cell information into the cervical cancer cell identification model while sequentially inputting the feature vector into the model.

6. A computer device, characterized in that, include: At least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to perform the method as described in any one of claims 1-4.

7. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions for causing a computer to perform the method as described in any one of claims 1-4.

Images