A method and system for image recognition of hollowed-out cells in cytological smears

Abstract

The present invention proposes an image recognition method for hollow cells in a cytological smear, comprising: obtaining a cell image corresponding to a prepared cervical liquid-based cell smear; graying the obtained cell image, and preliminarily screening the hollow cells according to the first grayscale feature of the grayed cell image and a pre-established feature database; wherein the database corresponding to the grayscale feature of the cell image and the cell morphological feature at least includes the correspondence between the first grayscale feature of the cell image and the first morphological feature of the cell, and the correspondence between the second grayscale feature of the cell image and the second morphological feature of the cell; the recognition and confirmation of the hollow cells are completed according to the second grayscale feature of the hollow cell image after the preliminary screening and the feature database. The present invention also proposes an image recognition system for hollow cells in cytological smears, which effectively improves the accuracy and efficiency of image recognition of hollow cells in cytological smears.

Description

Technical Field

[0001] The present invention relates to the field of medical informationization, and in particular to an image recognition method and system for hollowed-out cells in cytological smears. Background Art

[0002] Cervical cancer is the most common malignant tumor of the female reproductive tract, and its incidence rate ranks second among female malignant tumors, and even ranks first in some underdeveloped countries and regions. Therefore, cervical cancer seriously threatens women's health. Among all the cancer lesions known to humans, cervical cancer is the only malignant disease with a clear cause. Timely detection and treatment of cervical precancerous lesions can significantly reduce its morbidity and mortality. Studies have found that the occurrence of cervical cancer is related to long-term and persistent infection with human papillomavirus (HPV), and human papillomavirus causes cervical cell lesions to form hollow cells that can be observed by the human eye. Therefore, hollow cells can be used as one of the important indicators for the early detection of cervical cancer and have significant clinical significance. If hollow cells can be observed, it will have a good prompting effect on the occurrence of cervical cancer.

[0003] At present, there are defects in the detection methods of cervical cancer and the team of pathological cytologists in China. The specific problems are as follows: (1) In terms of detection methods, the most commonly used cervical cancer screening detection methods are HPV detection and cytology detection; (2) HPV detection. Persistent infection with human papillomavirus (HPV) has been confirmed to be an important factor in the development of cervical cancer. High-risk HPV detection has also become an internationally recognized important means of cervical cancer screening; however, HPV detection has many shortcomings, including: excessive specificity, which can easily lead to false positive judgments; inability to make accurate judgments on the pathological type and course of cervical cancer; long detection time, high requirements for detection equipment, and correspondingly high costs; (3) Cytology. Cervical cytology includes traditional cervical cell smear (Pap smear) method and liquid-based thin layer cytology (TCT). The Pap smear method is to take a small amount of cell samples from the cervix, place them on a glass slide, and then observe whether there are lesions under a microscope. The lesions are divided into 5 levels according to the severity of the lesions. However, there is a false negative rate of 15%-40%. Liquid-based cytology uses a special cervical brush and liquid-based preservation technology to collect and fix cell components, store the collected cells in a fixative, and then examine the filtered cells. The overall effect is better than that of the conventional Pap smear method, with a low false negative rate of about 12.8%, which greatly improves the quality of cytological preparation and the detection rate and accuracy of epithelial cell lesions. However, it depends on the experience and technical level of the examiner, is highly subjective, and is prone to missed diagnosis.

[0004] At present, cervical cytology smear examination is the most common, economical, effective and least side-effect early cervical cancer screening method, and is particularly suitable for large-scale physical examinations for risk groups. The advent of liquid-based cytology technology is a major innovation in film-making technology. However, both HPV testing and liquid-based cytology testing are inseparable from the pathologist's film reading diagnosis (cytology slide reading diagnosis). However, there is a serious shortage of pathologists in my country, and film reading still depends on pathological cytologists to use the naked eye to observe hundreds of thousands or even hundreds of thousands of cells under a microscope to search for diseased cells. Pathologists not only have a high workload, but long-term microscopic observation can easily make people tired, and require doctors to have comprehensive medical knowledge and rich clinical experience. In addition, the accuracy and reliability of manual interpretation results are easily affected by the doctor's subjective factors, analysis conditions and other factors, making the interpretation results show a high false negative and false positive rate. Summary of the invention

[0005] In order to solve the problems existing in the prior art, the present invention innovatively proposes an image recognition method and system for hollow cells in cytological smears, which effectively solves the problems of low accuracy and efficiency of recognition results caused by manual film reading and diagnosis by military physicians in the prior art, and effectively improves the accuracy and efficiency of image recognition of hollow cells in cytological smears.

[0006] The first aspect of the present invention provides an image recognition method for hollowed-out cells in a cytological smear, comprising:

[0007] Obtaining a cell image corresponding to the prepared cervical liquid-based cell smear;

[0008] The acquired cell image is grayed, and the hollowed-out cells are preliminarily screened according to the first grayscale feature of the grayed-out cell image and a pre-established feature database; wherein the database corresponding to the grayscale feature of the cell image and the cell morphological feature at least includes the corresponding relationship between the first grayscale feature of the cell image and the first morphological feature of the cell, and the corresponding relationship between the second grayscale feature of the cell image and the second morphological feature of the cell; the first morphological feature of the cell and the second morphological feature of the cell are both used to characterize the hollowed-out cells;

[0009] The identification and confirmation of the hollowed-out cells are completed based on the second grayscale features of the hollowed-out cell images after preliminary screening and the feature database.

[0010] Optionally, the first morphological feature of the cell is the presence of a cavity in the hollowed-out cell, and the first grayscale feature of the corresponding cell image is that the grayscale values ​​of the cell image include, from the outside to the inside, the cytoplasm grayscale value, the cavity grayscale value, and the cell nucleus grayscale value.

[0011] Furthermore, the gray value of the cavity is greater than the gray value of the cytoplasm, and the gray value of the cytoplasm is greater than the gray value of the nucleus.

[0012] Optionally, the second morphological feature of the cell is the dispersed distribution of grayscale values ​​in the hollowed-out cell, and the second grayscale feature of the corresponding cell image is that the mean square error of the cell image grayscale is greater than a preset first value, and the grayscale edge change rate of the cell image grayscale is greater than a preset second value.

[0013] Furthermore, the mean square error of the grayscale of the cell image is obtained as follows:

[0014] x k represents the grayscale value of the kth pixel in the cytoplasm, N represents the number of pixels in the cytoplasm, μ represents the average grayscale value of the cytoplasm, and σ is the mean square error of the grayscale of the cell image.

[0015] Optionally, the grayscale edge change rate of the cell image grayscale is obtained as follows:

[0016] Where f(i, j) represents the grayscale value of the pixel with coordinates (i, j), f(in, j) represents the grayscale value of the pixel with coordinates (in, j), f(i, jn) represents the grayscale value of the pixel with coordinates (i, jn), n is 1, Δf i (i, j) is the grayscale edge change rate of the pixel with coordinates (i, j) in the i direction, Δf j (i, j) is the grayscale edge change rate of the pixel with coordinates (i, j) in the i direction.

[0017] Furthermore, the grayscale edge change rate of the cell image grayscale is greater than a preset second value specifically: the grayscale edge change rate of the grayscale of the pixel point with coordinates (i, j) in the cell image grayscale in the i direction is greater than the preset second value, and the grayscale edge change rate of the grayscale of the pixel point with coordinates (i, j) in the cell image grayscale in the j direction is greater than the preset second value.

[0018] Optionally, the feature database further includes: a correspondence between a third grayscale feature of the cell image and a third morphological feature of the cell; the third morphological feature of the cell is used to characterize the hollowed-out cell.

[0019] Furthermore, the third morphological feature of the cell is that the number of cell nuclei is greater than 1, and the corresponding third grayscale feature of the cell image is that the grayscale value of the cell nuclei in the cell image is greater than 1.

[0020] A second aspect of the present invention provides an image recognition system for hollowed-out cells in cytological smears, comprising:

[0021] An image acquisition module, for acquiring a cell image corresponding to the prepared cervical liquid-based cell smear;

[0022] A processing and screening module performs grayscale processing on the acquired cell image, and preliminarily screens the hollowed-out cells according to the first grayscale feature of the grayscale processed cell image and a pre-established feature database; wherein the database corresponding to the grayscale feature of the cell image and the cell morphological feature includes the corresponding relationship between the first grayscale feature of the cell image and the first morphological feature of the cell, and the corresponding relationship between the second grayscale feature of the cell image and the second morphological feature of the cell; the first morphological feature of the cell and the second morphological feature of the cell are both used to characterize the hollowed-out cells;

[0023] The identification and confirmation module completes the identification and confirmation of the hollowed-out cells according to the second grayscale features of the hollowed-out cell images after preliminary screening and the feature database.

[0024] The technical solution adopted by the present invention includes the following technical effects:

[0025] 1. The present invention reduces the workload of pathologists and reduces missed diagnosis of hollow cells by performing preliminary screening and identification confirmation on the grayscale features of the cell images corresponding to the prepared cervical liquid-based cell smears. It effectively solves the problem of low accuracy and efficiency of recognition results caused by manual film reading and diagnosis by military physicians in the prior art, and effectively improves the accuracy and efficiency of image recognition of hollow cells in cytological smears.

[0026] 2. The technical solution of the present invention not only includes preliminary screening, but also re-identifies and confirms the images that are initially screened as hollow cells, further reducing the false negative rate of diagnosis, improving the work efficiency and diagnostic level of cytopathologists, and further improving the accuracy and reliability of image recognition of hollow cells in cytological smears.

[0027] 3. In the technical solution of the present invention, the grayscale edge change rate of the grayscale of the pixel point with coordinates (i, j) in the grayscale of the cell image in the i direction is not only greater than the preset second value, but also the grayscale edge change rate of the grayscale of the pixel point with coordinates (i, j) in the grayscale of the cell image in the j direction is also greater than the preset second value, thereby further improving the accuracy and reliability of image recognition of hollow cells in cytological smears.

[0028] 4. In the technical solution of the present invention, the feature database also includes: the correspondence between the third grayscale feature of the cell image and the third morphological feature of the cell, which improves the flexibility of hollow cell image recognition and enables doctors to choose different recognition schemes according to different situations.

[0029] It is to be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0031] Figure 1 This is a schematic diagram of a process of a method in Example 1 of the present invention;

[0032] Figure 2 It is a schematic diagram of the structure of the system of Example 2 in the solution of the present invention. DETAILED DESCRIPTION

[0033] In order to clearly illustrate the technical features of the present solution, the present invention is described in detail below through specific implementation methods and in conjunction with the accompanying drawings. The disclosure below provides many different embodiments or examples for realizing different structures of the present invention. In order to simplify the disclosure of the present invention, the components and settings of specific examples are described below. In addition, the present invention may repeat reference numbers and / or letters in different examples. This repetition is for the purpose of simplification and clarity, and does not itself indicate the relationship between the various embodiments and / or settings discussed. It should be noted that the components illustrated in the accompanying drawings are not necessarily drawn to scale. The present invention omits the description of known components and processing techniques and processes to avoid unnecessary limitations on the present invention.

[0034] Embodiment 1

[0035] like Figure 1 As shown, the present invention provides an image recognition method for hollowed-out cells in a cytological smear, comprising:

[0036] S1, obtaining a cell image corresponding to a prepared cervical liquid-based cell smear;

[0037] S2, graying the acquired cell image, and preliminarily screening the hollowed-out cells according to the first grayscale feature of the grayed-out cell image and a pre-established feature database; wherein the database corresponding to the grayscale feature of the cell image and the cell morphological feature at least includes the corresponding relationship between the first grayscale feature of the cell image and the first morphological feature of the cell, and the corresponding relationship between the second grayscale feature of the cell image and the second morphological feature of the cell; the first morphological feature of the cell and the second morphological feature of the cell are both used to characterize the hollowed-out cells;

[0038] S3, completing the identification and confirmation of the koilocytes according to the second grayscale features of the koilocyte images after preliminary screening and the feature database.

[0039] Wherein, in step S1, the prepared cervical liquid-based cell smear can be scanned by a scanner or other equipment to obtain a corresponding cell image. The preparation process of the cervical liquid-based cell smear can be:

[0040] Sampling - Use a small brush or a special sampler to collect cervical epithelial cells, place them in a preservation solution bottle, and then vortex to allow the cells to enter the preservation bottle;

[0041] Smear - pipette a drop of preservation solution onto the slide, distributing it as evenly as possible without air bubbles;

[0042] Staining - using dyes to better distinguish the nucleus and cytoplasm and highlight the internal structure of the cell; the commonly used Papanicolaou staining method makes the chromatin and ribosomes in the cytoplasm blue-purple or purple-red, which has good transparency and high nuclear staining saturation, and can more clearly highlight the cell structure;

[0043] Reading - Place the cervical liquid-based cell slide prepared through the above steps under a microscope for observation. When reading the slide, make sure that the entire slide is covered. First, check it from left to right and from top to bottom with a low-power microscope. Correspondingly, when scanning the prepared cervical liquid-based cell slide with a scanner or other scanning equipment, make sure that the entire slide is scanned. You can first check it from left to right and from top to bottom with a low-power microscope, or you can first check it from top to bottom and from left to right with a low-power microscope.

[0044] In step S2, the acquired cell image is grayed, and the hollowed-out cells are preliminarily screened according to the first grayscale features of the grayed cell image and a pre-established feature database.

[0045] The establishment of a feature database can be the correspondence between the grayscale features of the image and the corresponding cell morphological features. The grayscale features of the image can be based on the perception of the specific features of the image (cell morphological features), and then the perceived image features are conceptually described. Then, based on the grayscale features of the image, the hollow cells are screened and identified. The feature database can be updated regularly or irregularly (added, modified, deleted, etc.) based on the pathologist's judgment and cognition of the hollow cells.

[0046] Specifically, the morphological characteristics of the koilocytes are: from the inside out, the nucleus is enlarged, deformed, darkly stained, sometimes binucleated or multinucleated, there is a hollow halo around the nucleus, it is pool-like, corroded, or burned, necrosis is sudden, the boundaries are clear, the lines are sharp, the cytoplasm edge is thickened, accompanied by poor keratinization, and there is cell atypia. The morphological characteristics of the koilocytes are summarized as the first morphological characteristics and the second morphological characteristics of the cells in the feature database.

[0047] Specifically, the first morphological feature of the cell is the presence of a cavity in the hollowed-out cell, and the first grayscale feature of the corresponding cell image is that the grayscale values ​​of the cell image include, from the outside to the inside, the cytoplasm grayscale value, the cavity grayscale value, and the nucleus grayscale value, wherein the cavity grayscale value is greater than the cytoplasm grayscale value, and the cytoplasm grayscale value is greater than the nucleus grayscale value.

[0048] In terms of image features, the hollowed-out cells are composed of three parts: the mutated nucleus, the cavity around the nucleus, and the cytoplasm outside the cavity. In terms of color depth, the cavity part is the lightest, followed by the cytoplasm, and the nucleus is the darkest. In terms of grayscale, the grayscale value of the cavity area is the highest, followed by the cytoplasm, and the nucleus has the smallest grayscale value.

[0049] A digital pathology scanner is used to obtain a cell image (generally a single cervical cell image) of a prepared liquid-based cell slide. The cell image can be a full-slice digital image with high pixel resolution and high quality. First, the area range of the cell cavity is determined by image segmentation, and the cells are preliminarily screened. The grayscale value difference between the cytoplasm and the cavity area is analyzed. The grayscale value of the cavity in the hollow cell is higher than that of the cytoplasm. An appropriate segmentation threshold is selected, and the contact area between the outside of the cavity and the cytoplasm is used as the segmentation point to segment the image, and determine whether the cell has a cavity for preliminary screening of the hollow cell. Specifically, during image segmentation, if the cell image can be divided into three numerical intervals according to the grayscale value, the largest numerical interval corresponds to the grayscale value of the cavity, the smallest numerical interval corresponds to the grayscale value of the cell nucleus, and the remaining numerical interval corresponds to the grayscale value of the cytoplasm, then the cervical cell corresponding to the cell image is a hollow cell; if the cell image can be divided into two numerical intervals according to the grayscale value, the largest numerical interval corresponds to the grayscale value of the cytoplasm, and the smallest numerical interval corresponds to the grayscale value of the cell nucleus, then the cervical cell corresponding to the cell image is a normal cell. The number of grayscale value intervals can be divided in a sequential direction from the inside to the outside (or from the outside to the inside), the number of times the grayscale value corresponding to the unit pixel (i.e., the grayscale difference between two pixels separated by m pixels, where m is an integer greater than 0) is greater than a preset change value. If the number of times the grayscale value corresponding to the unit pixel is greater than the preset change value is twice, that is, the number of grayscale mutations is twice, then the cell image can be divided into three value intervals according to the grayscale value, with the grayscale value at each grayscale mutation as the endpoint value of different value intervals; if the number of times the grayscale value corresponding to the unit pixel is greater than the preset change value is once, that is, the number of grayscale mutations is 1, then the cell image can be divided into two value intervals according to the grayscale value, with the grayscale value at the grayscale mutation as the endpoint value of different value intervals.

[0050] Specifically, the second morphological feature of the cell is that the grayscale value distribution in the hollowed-out cell is dispersed, and the second grayscale feature of the corresponding cell image is that the mean square error of the cell image grayscale is greater than a preset first value, and the grayscale edge change rate of the cell image grayscale is greater than a preset second value.

[0051] Furthermore, the mean square error of the grayscale of the cell image is obtained as follows:

[0052] x k represents the grayscale value of the kth pixel in the cytoplasm, N represents the number of pixels in the cytoplasm, μ represents the average grayscale value of the cytoplasm, and σ is the mean square error of the grayscale of the cell image.

[0053] Furthermore, the grayscale edge change rate of the cell image grayscale is obtained as follows:

[0054] Where f(i, j) represents the grayscale value of the pixel with coordinates (i, j), f(in, j) represents the grayscale value of the pixel with coordinates (in, j), f(i, jn) represents the grayscale value of the pixel with coordinates (i, jn), n is 1, Δf i (i, j) is the grayscale edge change rate of the pixel with coordinates (i, j) in the i direction, Δf j (i, j) is the grayscale edge change rate of the pixel with coordinates (i, j) in the i direction.

[0055] In order to reflect the grayscale changes of the edge, the grayscale gradient method is used for calculation. Digital images are discrete in nature, so the difference method is used to reflect the derivative calculation.

[0056] Specifically, the grayscale edge change rate of the cell image grayscale is greater than the preset second value. Specifically, the grayscale edge change rate of the grayscale of the pixel point with coordinates (i, j) in the cell image grayscale in the i direction is greater than the preset second value, and the grayscale edge change rate of the grayscale of the pixel point with coordinates (i, j) in the cell image grayscale in the j direction is greater than the preset second value.

[0057] In a normal grayscale cell image, the grayscale value distribution is relatively uniform; in addition to the nucleus, the hollow cell mainly includes two parts, namely the cavity and the cytoplasm. The nucleus is surrounded by the cavity part with a larger grayscale value, and the periphery of the cavity is the cytoplasm part with a relatively small grayscale value. There is a certain difference in the grayscale of the two parts, so the grayscale value distribution of the hollow cell is relatively dispersed. Based on this feature, the mean square error (or coefficient of variation, i.e., standard deviation) of the grayscale and the grayscale edge change rate can be used to characterize the discrete degree of the grayscale value for final identification and confirmation. The preset first value can be 0.08, the preset second value can be 1.0, or other values ​​(which can be adjusted according to actual conditions). The present invention is not limited here. When the two conditions of the second grayscale feature are met at the same time, it is a hollow cell.

[0058] Furthermore, the feature database also includes: a correspondence between the third grayscale feature of the cell image and the third morphological feature of the cell; the third morphological feature of the cell is used to characterize the hollowed-out cell. If the cell image has a third grayscale feature corresponding to the third morphological feature of the cell, the cervical cell corresponding to the cell image is a hollowed-out cell; otherwise, the cervical cell corresponding to the cell image is not a hollowed-out cell.

[0059] The third morphological feature of the cell is that the number of cell nuclei is greater than 1, and the corresponding third grayscale feature of the cell image is that the grayscale value of the cell nuclei in the cell image is greater than 1. Specifically, there are greater than 1 grayscale minimum values ​​in the cell image (different cell nuclei have the same grayscale value), or the number of grayscale minimum values ​​in the cell nucleus region (different cell nuclei have different grayscale values, that is, different grayscale minimum values, but the difference between different grayscale minimum values ​​is less than the preset third value) is greater than 1.

[0060] The third morphological feature and the third grayscale feature of the cell can be directly applied based on the first morphological feature and the first grayscale feature of the cell, or can be applied based on the second morphological feature and the second grayscale feature of the cell, that is, the application order can be after the first grayscale feature and the second grayscale feature, or it can be other application orders, and the present invention does not limit this.

[0061] The present invention reduces the workload of pathologists and reduces missed diagnosis of hollow cells by performing preliminary screening and identification confirmation on the grayscale features of cell images corresponding to the prepared cervical liquid-based cell smears, effectively solving the problems of low accuracy and efficiency of recognition results caused by manual film reading and diagnosis by military physicians in the prior art, and effectively improving the accuracy and efficiency of image recognition of hollow cells in cytological smears.

[0062] The technical solution of the present invention not only includes preliminary screening, but also re-identifies and confirms the images that are initially screened as hollow cells, further reducing the false negative rate of diagnosis, improving the work efficiency and diagnostic level of cytopathologists, and further improving the accuracy and reliability of image recognition of hollow cells in cytological smears.

[0063] In the technical solution of the present invention, the grayscale edge change rate of the grayscale of the pixel point with coordinates (i, j) in the grayscale of the cell image in the i direction is not only greater than the preset second value, but also the grayscale edge change rate of the grayscale of the pixel point with coordinates (i, j) in the grayscale of the cell image in the j direction is also greater than the preset second value, thereby further improving the accuracy and reliability of image recognition of hollow cells in cytological smears.

[0064] In the technical solution of the present invention, the feature database also includes: the correspondence between the third grayscale feature of the cell image and the third morphological feature of the cell, which improves the flexibility of hollow cell image recognition and enables doctors to choose different recognition schemes according to different situations.

[0065] Embodiment 2

[0066] like Figure 2 As shown, the technical solution of the present invention also provides an image recognition system for hollowed-out cells in cytological smears, comprising:

[0067] An image acquisition module 101 acquires a cell image corresponding to a prepared cervical liquid-based cell smear;

[0068] The processing and screening module 102 performs grayscale processing on the acquired cell image, and preliminarily screens the hollowed-out cells according to the first grayscale feature of the grayscale processed cell image and a pre-established feature database; wherein the database corresponding to the grayscale feature of the cell image and the cell morphological feature includes the corresponding relationship between the first grayscale feature of the cell image and the first morphological feature of the cell, and the corresponding relationship between the second grayscale feature of the cell image and the second morphological feature of the cell; the first morphological feature of the cell and the second morphological feature of the cell are both used to characterize the hollowed-out cells;

[0069] The identification and confirmation module 103 completes the identification and confirmation of the hollowed-out cells according to the second grayscale features of the hollowed-out cell images after preliminary screening and the feature database.

[0070] The present invention reduces the workload of pathologists and reduces missed diagnosis of hollow cells by performing preliminary screening and identification confirmation on the grayscale features of cell images corresponding to the prepared cervical liquid-based cell smears, effectively solving the problems of low accuracy and efficiency of recognition results caused by manual film reading and diagnosis by military physicians in the prior art, and effectively improving the accuracy and efficiency of image recognition of hollow cells in cytological smears.

[0071] The technical solution of the present invention not only includes preliminary screening, but also re-identifies and confirms the images that are initially screened as hollow cells, further reducing the false negative rate of diagnosis, improving the work efficiency and diagnostic level of cytopathologists, and further improving the accuracy and reliability of image recognition of hollow cells in cytological smears.

[0072] In the technical solution of the present invention, the grayscale edge change rate of the grayscale of the pixel point with coordinates (i, j) in the grayscale of the cell image in the i direction is not only greater than the preset second value, but also the grayscale edge change rate of the grayscale of the pixel point with coordinates (i, j) in the grayscale of the cell image in the j direction is also greater than the preset second value, thereby further improving the accuracy and reliability of image recognition of hollow cells in cytological smears.

[0073] In the technical solution of the present invention, the feature database also includes: the correspondence between the third grayscale feature of the cell image and the third morphological feature of the cell, which improves the flexibility of hollow cell image recognition and enables doctors to choose different recognition schemes according to different situations.

[0074] Although the above describes the specific implementation mode of the present invention in conjunction with the accompanying drawings, it is not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art on the basis of the technical solution of the present invention without creative work are still within the scope of protection of the present invention.

Claims

1. A method for image recognition of hollowed-out cells in cytological smears, characterized in that: include: Obtaining a cell image corresponding to the prepared cervical liquid-based cell smear; The acquired cell image is grayed, and the hollowed-out cells are preliminarily screened according to the first grayscale feature of the grayed-out cell image and a pre-established feature database; wherein the cell image grayscale feature and cell morphological feature correspondence database at least includes the correspondence between the first grayscale feature of the cell image and the first morphological feature of the cell, and the correspondence between the second grayscale feature of the cell image and the second morphological feature of the cell; the first morphological feature of the cell and the second morphological feature of the cell are both used to characterize the hollowed-out cells; wherein the first morphological feature of the cell is the presence of a cavity in the hollowed-out cell, and the first grayscale feature of the corresponding cell image is that the grayscale values ​​of the cell image include, from the outside to the inside, the cytoplasm grayscale value, the cavity grayscale value, and the cell nucleus grayscale value; the cavity grayscale value is greater than Cytoplasm gray value, and the cytoplasm gray value is greater than the nucleus gray value; during image segmentation, if the cell image can be divided into 3 value intervals according to the gray value, the largest value interval corresponds to the cavity gray value, the smallest value interval corresponds to the nucleus gray value, and the remaining value interval corresponds to the cytoplasm gray value, then the cervical cell corresponding to the cell image is a hollow cell; the number of gray value intervals is divided according to the order from inside to outside, or from outside to inside, the number of times the gray value corresponding to the unit pixel is greater than the preset change value, if the number of times the gray value corresponding to the unit pixel is greater than the preset change value is twice, then the cell image can be divided into 3 value intervals according to the gray value, and the gray value at each gray value mutation is used as the endpoint value of different value intervals; The second morphological feature of the cell is that the gray value distribution in the hollowed-out cell is dispersed, and the second gray feature of the corresponding cell image is that the mean square error of the grayscale of the cell image is greater than a preset first value, and the grayscale edge change rate of the grayscale of the cell image is greater than a preset second value; The identification and confirmation of the hollowed-out cells are completed based on the second grayscale features of the hollowed-out cell images after preliminary screening and the feature database.

2. The method for image recognition of hollowed-out cells in a cytological smear according to claim 1, characterized in that: The mean square error of the cell image grayscale is obtained as follows: x k represents the grayscale value of the kth pixel in the cytoplasm, N represents the number of pixels in the cytoplasm, μ represents the average grayscale value of the cytoplasm, and σ is the mean square error of the grayscale of the cell image.

3. The image recognition method of hollowed-out cells in a cytological smear according to claim 1, characterized in that: The grayscale edge change rate of the cell image grayscale is obtained as follows: Where f(i, j) represents the grayscale value of the pixel with coordinates (i, j), f(in, j) represents the grayscale value of the pixel with coordinates (in, j), f(i, jn) represents the grayscale value of the pixel with coordinates (i, jn), n is 1, Δf i (i, j) is the grayscale edge change rate of the pixel with coordinates (i, j) in the i direction, Δf j (i, j) is the grayscale edge change rate of the pixel with coordinates (i, j) in the i direction.

4. The method for image recognition of hollowed-out cells in a cytological smear according to claim 1, characterized in that: The grayscale edge change rate of the cell image grayscale is greater than the preset second value specifically: the grayscale edge change rate of the grayscale of the pixel point with coordinates (i, j) in the cell image grayscale in the i direction is greater than the preset second value, and the grayscale edge change rate of the grayscale of the pixel point with coordinates (i, j) in the cell image grayscale in the j direction is greater than the preset second value.

5. The method for image recognition of hollow cells in cytological smears according to any one of claims 1 to 4, characterized in that the feature database It also includes: a corresponding relationship between the third grayscale feature of the cell image and the third morphological feature of the cell; the third morphological feature of the cell is used to characterize the hollowed-out cell.

6. The method for image recognition of hollowed-out cells in a cytological smear according to claim 5, characterized in that: The third morphological feature of the cell is that the number of cell nuclei is greater than 1, and the corresponding third grayscale feature of the cell image is that the grayscale value of the cell nuclei in the cell image is greater than 1.

7. An image recognition system for hollowed-out cells in cytological smears, characterized in that: include: An image acquisition module, for acquiring a cell image corresponding to the prepared cervical liquid-based cell smear; The processing and screening module performs grayscale processing on the acquired cell image, and preliminarily screens the hollowed-out cells according to the first grayscale feature of the grayscale processed cell image and a pre-established feature database; wherein the database corresponding to the grayscale feature of the cell image and the cell morphological feature includes the correspondence between the first grayscale feature of the cell image and the first morphological feature of the cell, and the correspondence between the second grayscale feature of the cell image and the second morphological feature of the cell; the first morphological feature of the cell and the second morphological feature of the cell are both used to characterize the hollowed-out cells; wherein the first morphological feature of the cell is the existence of a cavity in the hollowed-out cell, and the first grayscale feature of the corresponding cell image is that the grayscale values ​​of the cell image include, from the outside to the inside, the grayscale value of the cytoplasm, the grayscale value of the cavity, and the grayscale value of the nucleus; the grayscale value of the cavity The value is greater than the gray value of the cytoplasm, and the gray value of the cytoplasm is greater than the gray value of the nucleus; during image segmentation, if the cell image can be divided into three value intervals according to the gray value, the largest value interval corresponds to the gray value of the cavity, the smallest value interval corresponds to the gray value of the nucleus, and the remaining value intervals correspond to the gray value of the cytoplasm, then the cervical cells corresponding to the cell image are hollow cells; the number of gray value intervals is divided according to the order from inside to outside, or from outside to inside, the number of times the gray value corresponding to the unit pixel is greater than the preset change value, if the number of times the gray value corresponding to the unit pixel is greater than the preset change value is twice, then the cell image can be divided into three value intervals according to the gray value, and the gray value at each gray value mutation is used as the endpoint value of different value intervals; The second morphological feature of the cell is that the gray value distribution in the hollowed-out cell is dispersed, and the second gray feature of the corresponding cell image is that the mean square error of the grayscale of the cell image is greater than a preset first value, and the grayscale edge change rate of the grayscale of the cell image is greater than a preset second value; The identification and confirmation module completes the identification and confirmation of the hollowed-out cells according to the second grayscale features of the hollowed-out cell images after preliminary screening and the feature database.

Images