WO2021217851A1 - 异常细胞自动标注方法、装置、电子设备及存储介质 - Google Patents

异常细胞自动标注方法、装置、电子设备及存储介质 Download PDF

Info

Publication number
WO2021217851A1
WO2021217851A1 PCT/CN2020/098969 CN2020098969W WO2021217851A1 WO 2021217851 A1 WO2021217851 A1 WO 2021217851A1 CN 2020098969 W CN2020098969 W CN 2020098969W WO 2021217851 A1 WO2021217851 A1 WO 2021217851A1
Authority
WO
WIPO (PCT)
Prior art keywords
image
interest
cytopathological
region
group
Prior art date
Application number
PCT/CN2020/098969
Other languages
English (en)
French (fr)
Inventor
郭冰雪
王季勇
初晓
王坚
平波
喻林
Original Assignee
平安科技(深圳)有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 平安科技(深圳)有限公司 filed Critical 平安科技(深圳)有限公司
Publication of WO2021217851A1 publication Critical patent/WO2021217851A1/zh

Links

Images

Classifications

    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06TIMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00Image analysis
    • G06T7/0002Inspection of images, e.g. flaw detection
    • G06T7/0012Biomedical image inspection
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06TIMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00Image analysis
    • G06T7/10Segmentation; Edge detection
    • G06T7/12Edge-based segmentation
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06TIMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00Image analysis
    • G06T7/10Segmentation; Edge detection
    • G06T7/136Segmentation; Edge detection involving thresholding
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06TIMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00Image analysis
    • G06T7/10Segmentation; Edge detection
    • G06T7/194Segmentation; Edge detection involving foreground-background segmentation
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06TIMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00Indexing scheme for image analysis or image enhancement
    • G06T2207/20Special algorithmic details
    • G06T2207/20016Hierarchical, coarse-to-fine, multiscale or multiresolution image processing; Pyramid transform
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06TIMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00Indexing scheme for image analysis or image enhancement
    • G06T2207/20Special algorithmic details
    • G06T2207/20081Training; Learning
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06TIMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00Indexing scheme for image analysis or image enhancement
    • G06T2207/20Special algorithmic details
    • G06T2207/20084Artificial neural networks [ANN]
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06TIMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00Indexing scheme for image analysis or image enhancement
    • G06T2207/30Subject of image; Context of image processing
    • G06T2207/30004Biomedical image processing
    • G06T2207/30024Cell structures in vitro; Tissue sections in vitro

Definitions

  • This application relates to image processing technology, which is applied in the field of smart medical treatment, and particularly relates to methods, devices, electronic equipment, and readable storage media for automatic labeling of abnormal cells.
  • the present application provides a method, device, electronic equipment, and computer-readable storage medium for automatic labeling of abnormal cells, the main purpose of which is to improve the accuracy of labeling abnormal cells and reduce the pressure of calculation and storage.
  • an automatic labeling method for abnormal cells includes:
  • the abnormal cell annotation set is mapped to the cytopathological picture to obtain an abnormal cell annotation map.
  • the present application also provides an electronic device, wherein the electronic device includes:
  • At least one processor and,
  • a memory communicatively connected with the at least one processor; wherein,
  • the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor, so that the at least one processor can execute the method for automatically labeling abnormal cells as described below:
  • the abnormal cell annotation set is mapped to the cytopathological picture to obtain an abnormal cell annotation map.
  • the present application also provides a computer-readable storage medium, wherein the computer-readable storage medium stores an abnormal cell automatic labeling program, and the abnormal cell automatic labeling program can be used by one or more processors. Execute to realize the steps of the method for automatic labeling of abnormal cells as described below:
  • the image pyramid generating step is used to obtain a cytopathological picture, perform Gaussian convolution smoothing for a preset number of times on the cytopathological picture, generate multiple cytopathological pictures, and obtain an image pyramid according to the cytopathological picture;
  • the detection and fitting step is used to detect and fit the image pyramid to obtain a low-power fitting region of interest
  • the mapping and segmentation step is used to map the low-power fitted region of interest to the acquired cytopathological picture and perform image magnification operations to generate a fitted high-power image, and obtain the region of interest in the fitted high-power image And then use the image coordinates to segment the region of interest in the fitted high-magnification image to generate a segmented high-magnification image;
  • An annotation cutting step is used for annotating and cutting the segmented high-power image through an adaptive threshold segmentation algorithm to obtain an abnormal cell annotation set;
  • the abnormal generation step is used to map the abnormal cell annotation set to the cytopathological picture to obtain an abnormal cell annotation map.
  • the present application also provides an automatic labeling device for abnormal cells, wherein the device includes:
  • the image pyramid generating module is used to obtain a cytopathological picture, perform Gaussian convolution smoothing for a preset number of times on the cytopathological picture, generate multiple cytopathological pictures, and obtain an image pyramid according to the cytopathological picture;
  • the detection and fitting module is used to detect and fit the image pyramid to obtain a low-power fitting region of interest
  • the mapping and segmentation module is used to map the low-power fitted region of interest to the acquired cytopathological picture and perform image magnification operations to generate a fitted high-power image, and obtain the region of interest in the fitted high-power image And then use the image coordinates to segment the region of interest in the fitted high-magnification image to generate a segmented high-magnification image;
  • An annotation cutting module which is used to annotate and cut the segmented high-magnification image through an adaptive threshold segmentation algorithm to obtain an abnormal cell annotation set;
  • the abnormal generation module is used to map the abnormal cell annotation set to the cytopathological picture to obtain an abnormal cell annotation map.
  • the original cell pathological pictures are smoothed by Gaussian convolution to obtain an image pyramid composed of multiple cell case pictures with higher and higher resolution from top to bottom.
  • the pathological data features in these cell case pictures are more obvious , Improve the accuracy of abnormal cell annotation; further, because the resolution of the image pixels in the image pyramid is very large, the general image processing method cannot directly use the computer for image analysis processing, so the embodiment of the application fits the pixels to The low-magnification area reduces the pixel and resolution of the image, and further releases the storage pressure of the computer while not affecting the accuracy of subsequent abnormal cell detection; in addition, when the low-magnification fitting area of interest is obtained, the embodiments of the present application further By zooming in, it becomes a fitting high-magnification image.
  • the high-magnification image has a high resolution, which can effectively improve the accuracy of the algorithm for labeling and cutting. Therefore, the method, device, electronic device, and computer-readable storage medium for automatic labeling of abnormal cells proposed in this application can solve the problems of low accuracy of abnormal cell labeling and high pressure of calculation and storage.
  • FIG. 1 is a schematic flowchart of an automatic labeling method for abnormal cells provided by an embodiment of the application
  • FIG. 2 is a schematic diagram of modules of an automatic labeling device for abnormal cells provided by an embodiment of the application
  • FIG. 3 is a schematic diagram of the internal structure of an electronic device for executing the method for automatically labeling abnormal cells according to an embodiment of the application;
  • This solution can be applied in the field of smart medical care to promote the construction of smart cities.
  • the purpose of the embodiments of this application is to label abnormal cells through adaptive threshold segmentation, which is used to improve the accuracy of abnormal cell labeling and reduce the pressure of calculation and storage.
  • the cytopathological pictures are stratified to make the data characteristics after stratification more obvious.
  • the abnormal cells in the stratified cytopathological pictures are marked, and the abnormal cells are mapped through coordinate mapping.
  • the method maps the abnormal cells in the low-pixel image in the cervical cytopathological picture to the high-pixel image for subsequent abnormal cell labeling and segmentation.
  • FIG. 1 it is a schematic flowchart of a method for automatically labeling abnormal cells provided by an embodiment of this application.
  • the method can be executed by a device, and the device can be implemented by software and/or hardware.
  • the method for automatically labeling abnormal cells includes:
  • the cytopathological pictures described in the embodiments of the present application include cervical cytopathological pictures.
  • performing Gaussian convolution smoothing processing on the cytopathological picture for a preset number of times to generate a plurality of cytopathological pictures, and obtaining an image pyramid according to the cytopathological picture includes:
  • Step A Perform an enlargement operation on the acquired cytopathological pictures to obtain a group a and b layer images, where the initial values of a and b are both 1, and the a group and b layer images are Gaussian rolls
  • the product function performs the smoothing operation of Gaussian convolution to obtain the b+1th layer image of the ath group, where the Gaussian convolution function is:
  • represents the smoothing factor
  • G (x, y, ⁇ ) represents the convolution of x and y
  • x and y represent the image coordinates
  • Step B Multiply the smoothing factor ⁇ by the scale factor k to obtain a new smoothing factor k ⁇ , and use the Gaussian convolution function to Gaussian the a-th group and the b+1th layer image through the new smoothing factor k ⁇ Convolutional smoothing operation to obtain the b+2th layer image of the ath group, and repeat this step until the Lth layer image of the ath group is obtained, where L is a predefined value;
  • Step C Perform a down-sampling operation on the a-th group of L-th layer images to obtain the a+1-th group of b-th layer images, and perform Gaussian convolution on the a+1-th group and b-th layer images using the Gaussian convolution function described above Product smoothing operation to obtain the a+1th group of b+1th layer images, and repeat the step B until the a+1th group of Lth layer images are obtained;
  • Step D Perform the loop operation of Step C and Step B on the result obtained in Step C, until the O-th group of L-th layer images is obtained, where O is a predefined value;
  • Step E Combine images from the a-th group and b-th layer images to the O-th group and L-th layer images to generate the image pyramid.
  • each image from the a-th group of b-th layer images to the O-th group of L-th layer images is a picture layer of the image pyramid.
  • the cytopathological picture can be split into multiple cytopathological pictures with higher and higher resolution from top to bottom, which makes the characteristics of the pathological data more obvious and improves the accuracy of abnormal cell labeling.
  • the S2 includes: performing gray-scale conversion of the image pyramid to generate a gray-scale image; performing binarization processing on the gray-scale image to obtain a contour image; and adopting a random selection method from the contour image , Obtain the candidate region; use the Hough circle transform detection method to identify the region of interest in the candidate region, and fit the region of interest to obtain the low-power fitting region of interest.
  • the using the Hough circle transform detection method to identify the region of interest in the candidate region, and fitting the region of interest to obtain the low-power fitting region of interest includes:
  • Step a Detect the edge of the candidate area in the image space of the candidate area by using an edge detection algorithm to obtain n edge pixel point sets;
  • Step b Map the n edge pixel point sets to the parameter space with a predefined value r as a radius:
  • ⁇ [0,2 ⁇ ) x and y represent the corresponding coordinates (x, y) of the n edge pixel point sets, and (a, b) represent the reference point in the parameter space coordinate;
  • Step c Count all the coordinate points in the parameter space, and traverse ⁇ , so that when the edge pixels in the image space are mapped to a circle in the parameter space, the circle is regarded as the Hough circle, and the Hough circle is passed through the Hough circle.
  • the circle constitutes the region of interest
  • Step d Fitting the region of interest to generate a low-power fitting region of interest.
  • the region of interest in the cytopathological picture can be arbitrarily marked by the doctor, and then identified and fitted by the Hough circle transformation detection method.
  • the high-magnification image size of the cytopathological picture is very large, and the general image processing method is difficult to load for direct processing, and it is impossible to directly use a computer for image analysis and processing. Therefore, the embodiment of this application adopts the low-magnification
  • the fitted region of interest is mapped to the acquired high-magnification image, a fitted high-magnification image is generated, and the image coordinate information of the region of interest in the fitted high-magnification image is obtained.
  • the image is segmented.
  • the mapping is to divide the low-resolution image into multiple sub-regions, and add each sub-region to the high-resolution image by multiplying the coordinates of the sub-regions and a multiple, and the multiple is the high-resolution image.
  • This application implements all the image coordinates of the region of interest acquired by fitting, and the cell region of interest can be segmented to generate a segmented high-magnification image with a size of, for example, 3000x3000.
  • S5. Perform labeling and cutting on the segmented high-power image through an adaptive threshold segmentation algorithm to obtain an abnormal cell label set.
  • the S5 includes:
  • Step I Pass each pixel i of the segmented high-magnification image through the formula:
  • Step II Perform a pre-compensation calculation on the normalized value h by using a pre-defined gamma correction compensation value to obtain a pre-compensation constant value f;
  • Step III Perform denormalization calculation on the pre-compensation constant value f through the formula f*256-0.5 to obtain the corrected image of the segmented high-magnification image;
  • Step IV The image gray level of the corrected image and the pixel point gray level of the coordinate in the corrected image are formed into a two-tuple, the mean value and variance of all the two-tuples are calculated, and the two-dimensionality is established by the mean value and the variance A maximum between-cluster variance model, calculating the two-dimensional maximum between-cluster variance model through an adaptive particle clustering algorithm to generate the optimal threshold of the corrected image;
  • Step V segment the corrected image by using the optimal threshold to generate a background and foreground segmented image
  • Step VI Perform opening and closing operations on the background and foreground segmented images to generate an abnormal cell annotation set.
  • the background and foreground segmented images are sequentially opened and calculated through a morphological algorithm, and isolated small points in the segmented images are removed to generate a first image set, and the first image set is
  • the morphological algorithm is used to sequentially perform closed calculations to fill in the small cracks between the image cells in the first image set, and clear the holes in the images in the first image set, so that the first image sets the foreground cells and all the cells in the first image set.
  • the background area in the first image set is separated to generate an abnormal cell annotation set.
  • region-based segmentation methods Traditional cell image segmentation methods are roughly divided into two categories: region-based segmentation methods and edge-based segmentation methods.
  • the basic principle of region-based segmentation methods is to achieve segmentation by classifying adjacent regions with similar characteristics into one category.
  • the edge-based segmentation method generally uses the gray level or the structure with a sudden change as the edge to perform segmentation. This application adopts the edge-based segmentation method to perform image segmentation through an adaptive threshold segmentation algorithm to obtain an abnormal cell annotation set.
  • the abnormal cell annotation set is used to locate and identify target cells through a deep learning method to generate an abnormal cell annotation map. If the edge of the abnormal cell annotation map is jagged, the edge smoothing method is used to perform Edge adjustment.
  • the above-mentioned abnormal cell annotation set may also be stored in a node of a blockchain.
  • FIG. 2 it is a functional block diagram of the device for automatic labeling of abnormal cells of the present application.
  • the apparatus 100 for automatically labeling abnormal cells described in the embodiment of the present application may be installed in an electronic device.
  • the device for automatically labeling abnormal cells may include a detection and fitting module 101, a mapping segmentation module 102, a labeling and cutting module 103, and an abnormality generation module 104.
  • the module described in the present invention can also be called a unit, which refers to a series of computer program segments that can be executed by the processor of an electronic device and can complete fixed functions, and are stored in the memory of the electronic device.
  • each module/unit is as follows:
  • the image pyramid generating module 101 is configured to obtain a cytopathological picture, perform Gaussian convolution smoothing processing for a preset number of times on the cytopathological picture, generate multiple cytopathological pictures, and obtain an image pyramid according to the cytopathological picture;
  • the detection and fitting module 102 is configured to detect and fit the image pyramid by the Hough transform circle detection method to obtain a low-power fitting region of interest;
  • the mapping and segmentation module 103 is configured to map the low-power fitting region of interest to the acquired cytopathological picture and perform an image magnification operation to generate a fitting high-power image, and acquire the interest in the fitted high-power image All the image coordinates of the region, and then use the image coordinates to segment the region of interest in the fitted high-magnification image to generate a segmented high-magnification image;
  • the annotation cutting module 104 is configured to perform annotation cutting on the segmented high-power image through an adaptive threshold segmentation algorithm to obtain an abnormal cell annotation set;
  • the abnormal generation module 105 is configured to map the abnormal cell annotation set to the cytopathological picture to obtain an abnormal cell annotation map.
  • each module of the abnormal cell automatic labeling device is as follows:
  • the image pyramid generating module 101 obtains a cytopathological picture, performs a preset number of Gaussian convolution smoothing processing on the cytopathological picture, generates a plurality of cytopathological pictures, and obtains an image pyramid according to the cytopathological picture.
  • the cytopathological pictures described in the embodiments of the present application include cervical cytopathological pictures.
  • performing Gaussian convolution smoothing processing on the cytopathological picture for a preset number of times to generate a plurality of cytopathological pictures, and obtaining an image pyramid according to the cytopathological picture includes:
  • Step A Perform an enlargement operation on the acquired cytopathological pictures to obtain a group a and b layer images, wherein the initial values of a and b are both 1, and the a group and b layer images are Gaussian rolls
  • the product function performs the smoothing operation of Gaussian convolution to obtain the b+1th layer image of the ath group, where the Gaussian convolution function is:
  • represents the smoothing factor
  • G (x, y, ⁇ ) represents the convolution of x and y
  • x and y represent the image coordinates
  • Step B Multiply the smoothing factor ⁇ by the scale factor k to obtain a new smoothing factor k ⁇ , and use the Gaussian convolution function to Gaussian the a-th group and the b+1th layer image through the new smoothing factor k ⁇ Convolutional smoothing operation to obtain the b+2th layer image of the ath group, and repeat this step until the Lth layer image of the ath group is obtained, where L is a predefined value;
  • Step C Perform a down-sampling operation on the a-th group of L-th layer images to obtain the a+1-th group of b-th layer images, and perform Gaussian convolution on the a+1-th group and b-th layer images using the Gaussian convolution function described above Product smoothing operation to obtain the a+1th group of b+1th layer images, and repeat the step B until the a+1th group of Lth layer images are obtained;
  • Step D Perform the loop operation of Step C and Step B on the result obtained in Step C, until the O-th group of L-th layer images is obtained, where O is a predefined value;
  • Step E Combine images from the a-th group and b-th layer images to the O-th group and L-th layer images to generate the image pyramid.
  • each image from the a-th group of b-th layer images to the O-th group of L-th layer images is a picture layer of the image pyramid.
  • the cytopathological picture can be split into multiple cytopathological pictures with higher and higher resolution from top to bottom, which makes the characteristics of the pathological data more obvious and improves the accuracy of abnormal cell labeling.
  • the detection and fitting module 102 detects and fits the image pyramid by a Hough transform circle detection method to obtain a low-power fitting region of interest.
  • the detection and fitting module 102 converts the image pyramid to grayscale to generate a grayscale image; performs binarization processing on the grayscale image to obtain a contour image; adopts random from the contour image The selected method is used to obtain the candidate region; the Hough circle transform detection method is used to identify the region of interest in the candidate region, and the region of interest is fitted to obtain the low-power fitting region of interest.
  • the using the Hough circle transform detection method to identify the region of interest in the candidate region, and fitting the region of interest to obtain the low-power fitting region of interest includes:
  • Step a Detect the edge of the candidate area in the image space of the candidate area by using an edge detection algorithm to obtain n edge pixel point sets;
  • Step b Map the n edge pixel point sets to the parameter space with a predefined value r as a radius:
  • ⁇ [0,2 ⁇ ) x and y represent the corresponding coordinates (x, y) of the n edge pixel point sets, and (a, b) represent the reference point in the parameter space coordinate;
  • Step c Count all the coordinate points in the parameter space, and traverse ⁇ , so that when the edge pixels in the image space are mapped to the parameter space as a circle, the circle is regarded as the Hough circle, and the Hough circle is passed through the Hough circle.
  • the circle constitutes the region of interest
  • Step d Fitting the region of interest to generate a low-power fitting region of interest.
  • the region of interest in the cytopathological picture can be arbitrarily marked by the doctor, and then identified and fitted by the Hough circle transformation detection method.
  • the mapping and segmentation module 103 maps the low-power fitting region of interest to the acquired cytopathological picture and performs an image magnification operation to generate a fitting high-power image, and acquire the interest in the fitted high-power image All image coordinates of the area.
  • the high-magnification image size of the cytopathological picture is very large, and the general image processing method is difficult to load for direct processing, and it is impossible to directly use the computer for image analysis and processing. Therefore, the embodiment of the application adopts the low-magnification
  • the fitted region of interest is mapped to the acquired high-magnification image, a fitted high-magnification image is generated, and the image coordinate information of the region of interest in the fitted high-magnification image is obtained.
  • the image is segmented.
  • the mapping is to divide the low-resolution image into multiple sub-regions, and add each sub-region to the high-resolution image by multiplying the coordinates of the sub-regions and a multiple, and the multiple is the high-resolution image.
  • the mapping segmentation module 103 uses the image coordinates to segment the region of interest in the fitted high-magnification image to generate a segmented high-magnification image.
  • This application implements all the image coordinates of the region of interest acquired by fitting, and the cell region of interest can be segmented to generate a segmented high-magnification image with a size of, for example, 3000x3000.
  • the labeling and cutting module 104 performs labeling and cutting on the segmented high-magnification image through an adaptive threshold segmentation algorithm to obtain an abnormal cell labeling set.
  • the labeling and cutting of the segmented high-magnification image to obtain an abnormal cell labeling set includes:
  • Step I Pass each pixel i of the segmented high-magnification image through the formula:
  • Step II Perform a pre-compensation calculation on the normalized value h by using a pre-defined gamma correction compensation value to obtain a pre-compensation constant value f;
  • Step III Perform denormalization calculation on the pre-compensation constant value f through the formula f*256-0.5 to obtain the corrected image of the segmented high-magnification image;
  • Step IV The image gray level of the corrected image and the pixel point gray level of the coordinate in the corrected image are formed into a two-tuple, the mean value and variance of all the two-tuples are calculated, and the two-dimensionality is established by the mean value and the variance A maximum between-cluster variance model, calculating the two-dimensional maximum between-cluster variance model through an adaptive particle clustering algorithm to generate the optimal threshold of the corrected image;
  • Step V segment the corrected image by using the optimal threshold to generate a background and foreground segmented image
  • Step VI Perform opening and closing operations on the background and foreground segmented images to generate an abnormal cell annotation set.
  • the background and foreground segmented images are sequentially opened and calculated through a morphological algorithm, and isolated small points in the segmented images are removed to generate a first image set, and the first image set is
  • the morphological algorithm is used to perform closed calculations in sequence to fill in the small cracks between the image cells in the first image set, and clear the holes in the images in the first image set, so that the first image sets the foreground cells and all the cells in the first image set.
  • the background area in the first image set is separated to generate an abnormal cell annotation set.
  • region-based segmentation methods Traditional cell image segmentation methods are roughly divided into two categories: region-based segmentation methods and edge-based segmentation methods.
  • the basic principle of region-based segmentation methods is to achieve segmentation by classifying adjacent regions with similar characteristics into one category.
  • the edge-based segmentation method generally uses the gray level or the structure with a sudden change as the edge to perform segmentation. This application adopts the edge-based segmentation method to perform image segmentation through an adaptive threshold segmentation algorithm to obtain an abnormal cell annotation set.
  • the abnormal generation module 105 maps the abnormal cell annotation set to the cytopathological picture to obtain an abnormal cell annotation map.
  • the abnormal cell annotation set is used to locate and identify target cells through a deep learning method to generate an abnormal cell annotation map. If the edge of the abnormal cell annotation map is jagged, the edge smoothing method is used to perform Edge adjustment.
  • the above-mentioned abnormal cell annotation set may also be stored in a node of a blockchain.
  • FIG. 3 it is a schematic structural diagram of an electronic device that implements the method for automatically labeling abnormal cells of the present application.
  • the electronic device 1 may include a processor 10, a memory 11, and a bus, and may also include a computer program stored in the memory 11 and running on the processor 10, such as an abnormal cell automatic labeling program 12.
  • the memory 11 includes at least one type of readable storage medium, and the readable storage medium includes flash memory, mobile hard disk, multimedia card, card-type memory (such as SD or DX memory, etc.), magnetic memory, magnetic disk, CD etc.
  • the memory 11 may be an internal storage unit of the electronic device 1 in some embodiments, for example, a mobile hard disk of the electronic device 1.
  • the memory 11 may also be an external storage device of the electronic device 1, such as a plug-in mobile hard disk, a smart media card (SMC), and a secure digital (Secure Digital) equipped on the electronic device 1. , SD) card, flash card (Flash Card), etc.
  • the memory 11 may also include both an internal storage unit of the electronic device 1 and an external storage device.
  • the memory 11 can be used not only to store application software and various types of data installed in the electronic device 1, such as codes for an automatic labeling program for abnormal cells, etc., but also to temporarily store data that has been output or will be output.
  • the processor 10 may be composed of integrated circuits in some embodiments, for example, may be composed of a single packaged integrated circuit, or may be composed of multiple integrated circuits with the same function or different functions, including one or more Combinations of central processing unit (CPU), microprocessor, digital processing chip, graphics processor, and various control chips, etc.
  • the processor 10 is the control unit of the electronic device, which uses various interfaces and lines to connect the various components of the entire electronic device, and runs or executes programs or modules stored in the memory 11 (such as executing Automatic labeling program for abnormal cells, etc.), and call the data stored in the memory 11 to execute various functions of the electronic device 1 and process data.
  • the bus may be a peripheral component interconnect standard (PCI) bus or an extended industry standard architecture (EISA) bus, etc.
  • PCI peripheral component interconnect standard
  • EISA extended industry standard architecture
  • the bus can be divided into address bus, data bus, control bus and so on.
  • the bus is configured to implement connection and communication between the memory 11 and at least one processor 10 and the like.
  • FIG. 3 only shows an electronic device with components. Those skilled in the art can understand that the structure shown in FIG. 3 does not constitute a limitation on the electronic device 1, and may include fewer or more components than shown in the figure. Components, or a combination of certain components, or different component arrangements.
  • the electronic device 1 may also include a power source (such as a battery) for supplying power to various components.
  • the power source may be logically connected to the at least one processor 10 through a power management device, thereby controlling power
  • the device implements functions such as charge management, discharge management, and power consumption management.
  • the power supply may also include any components such as one or more DC or AC power supplies, recharging devices, power failure detection circuits, power converters or inverters, and power status indicators.
  • the electronic device 1 may also include various sensors, Bluetooth modules, Wi-Fi modules, etc., which will not be repeated here.
  • the electronic device 1 may also include a network interface.
  • the network interface may include a wired interface and/or a wireless interface (such as a WI-FI interface, a Bluetooth interface, etc.), which is usually used in the electronic device 1 Establish a communication connection with other electronic devices.
  • the electronic device 1 may also include a user interface.
  • the user interface may be a display (Display) and an input unit (such as a keyboard (Keyboard)).
  • the user interface may also be a standard wired interface or a wireless interface.
  • the display may be an LED display, a liquid crystal display, a touch-sensitive liquid crystal display, an OLED (Organic Light-Emitting Diode, organic light-emitting diode) touch device, etc.
  • the display can also be appropriately called a display screen or a display unit, which is used to display the information processed in the electronic device 1 and to display a visualized user interface.
  • the abnormal cell automatic labeling program 12 stored in the memory 11 in the electronic device 1 is a combination of multiple instructions. When running in the processor 10, it can realize:
  • the abnormal cell annotation set is mapped to the cytopathological picture to obtain an abnormal cell annotation map.
  • the integrated module/unit of the electronic device 1 is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium.
  • the computer-readable storage medium may be non-volatile or volatile.
  • the computer-readable storage medium may include: any entity or device capable of carrying the computer program code, recording medium, U disk, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (ROM, Read-Only Memory) ).
  • the computer-readable storage medium may mainly include a storage program area and a storage data area, where the storage program area may store an operating system, an application program required by at least one function, etc.; the storage data area may store Data created by the use of nodes, etc.
  • the blockchain referred to in this application is a new application mode of computer technology such as distributed data storage, point-to-point transmission, consensus mechanism, and encryption algorithm.
  • Blockchain essentially a decentralized database, is a series of data blocks associated with cryptographic methods. Each data block contains a batch of network transaction information for verification. The validity of the information (anti-counterfeiting) and the generation of the next block.
  • the blockchain can include the underlying platform of the blockchain, the platform product service layer, and the application service layer.
  • modules described as separate components may or may not be physically separated, and the components displayed as modules may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules can be selected according to actual needs to achieve the objectives of the solutions of the embodiments.
  • the functional modules in the various embodiments of the present application may be integrated into one processing unit, or each unit may exist alone physically, or two or more units may be integrated into one unit.
  • the above-mentioned integrated unit may be implemented in the form of hardware, or may be implemented in the form of hardware plus software functional modules.

Landscapes

  • Engineering & Computer Science (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Health & Medical Sciences (AREA)
  • General Health & Medical Sciences (AREA)
  • Medical Informatics (AREA)
  • Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
  • Radiology & Medical Imaging (AREA)
  • Quality & Reliability (AREA)
  • Image Analysis (AREA)

Abstract

涉及图像处理技术,应用于智慧医疗领域中,揭露了一种异常细胞自动标注方法,包括:对细胞病理图片进行高斯卷积平滑处理,得到图像金字塔;对所述图像金字塔通过霍夫变换圆检测方法检测拟合得到低倍拟合感兴趣区域;将所述低倍拟合感兴趣区域映射到获取的所述细胞病理图片上,生成拟合高倍图像;将所述拟合高倍图像中的感兴趣区域分割出来,生成分割高倍图像;对所述分割高倍图像通过自适应阈值分割算法进行标注切割,得到异常细胞标注集;将所述异常细胞标注集映射到所述细胞病理图片上,得到异常细胞标注图。提高异常细胞标注的准确性以及减轻计算和存储压力。此外,还涉及区块链技术,所述异常细胞标注集可存储于区块链中。

Description

异常细胞自动标注方法、装置、电子设备及存储介质
本申请要求于2020年04月27日提交中国专利局、申请号为202010348583.2、发明名称为“异常细胞自动标注方法、装置、电子设备及存储介质”的中国专利申请的优先权,其全部内容通过引用结合在申请中。
技术领域
本申请涉及图像处理技术,应用于智慧医疗领域中,尤其涉及异常细胞自动标注方法、装置、电子设备及可读存储介质。
背景技术
目前,由于病理医生和细胞学检测设备的缺乏,使得各种人工智能辅助筛查的设备***渐渐出现,市场上也有大量利用深度学习神经网络进行特征提取和训练检测异常细胞的方法,但发明人意识到现有的方法还是需要大量的病理医生对病理数据进行标注,增加了病理医生的劳动强度,同时由于神经网络异常细胞的检测方法需要大量的病理数据,因此对于计算机的计算压力和存储性能都具有挑战性,因此需要一种异常细胞自动标注方法,用来提高异常细胞标注的准确性以及减轻计算和存储压力。
发明内容
本申请提供一种异常细胞自动标注方法、装置、电子设备及计算机可读存储介质,其主要目的在于提高异常细胞标注的准确性以及减轻计算和存储压力。
为实现上述目的,本申请提供的一种异常细胞自动标注方法,包括:
获取细胞病理图片,对所述细胞病理图片进行预设次数的高斯卷积平滑处理,生成多个细胞病理图片,根据所述细胞病理图片得到图像金字塔;
对所述图像金字塔进行检测及拟合,得到低倍拟合感兴趣区域;
将所述低倍拟合感兴趣区域映射到获取的所述细胞病理图片上并进行图像放大操作,生成拟合高倍图像,并获取所述拟合高倍图像中感兴趣区域的所有图像坐标;
利用所述图像坐标将所述拟合高倍图像中的感兴趣区域分割出来,生成分割高倍图像;
对所述分割高倍图像通过自适应阈值分割算法进行标注切割,得到异常细胞标注集;
将所述异常细胞标注集映射到所述细胞病理图片上,得到异常细胞标注图。
为了解决上述问题,本申请还提供一种电子设备,其中,所述电子设备包括:
至少一个处理器;以及,
与所述至少一个处理器通信连接的存储器;其中,
所述存储器存储有可被所述至少一个处理器执行的指令,所述指令被所述至少一个处理器执行,以使所述至少一个处理器能够执行如下所述的异常细胞自动标注方法:
获取细胞病理图片,对所述细胞病理图片进行预设次数的高斯卷积平滑处理,生成多个细胞病理图片,根据所述细胞病理图片得到图像金字塔;
对所述图像金字塔进行检测及拟合,得到低倍拟合感兴趣区域;
将所述低倍拟合感兴趣区域映射到获取的所述细胞病理图片上并进行图像放大操作,生成拟合高倍图像,并获取所述拟合高倍图像中感兴趣区域的所有图像坐标;
利用所述图像坐标将所述拟合高倍图像中的感兴趣区域分割出来,生成分割高倍图像;
对所述分割高倍图像通过自适应阈值分割算法进行标注切割,得到异常细胞标注集;
将所述异常细胞标注集映射到所述细胞病理图片上,得到异常细胞标注图。
为了解决上述问题,本申请还提供一种计算机可读存储介质,其中,所述计算机可读存储介质上存储有异常细胞自动标注程序,所述异常细胞自动标注程序可被一个或者多个处理器执行,以实现如下所述的异常细胞自动标注方法的步骤:
图像金字塔生成步骤,用于获取细胞病理图片,对所述细胞病理图片进行预设次数的高斯卷积平滑处理,生成多个细胞病理图片,根据所述细胞病理图片得到图像金字塔;
检测拟合步骤,用于对所述图像金字塔进行检测及拟合,得到低倍拟合感兴趣区域;
映射分割步骤,用于将所述低倍拟合感兴趣区域映射到获取的所述细胞病理图片上并进行图像放大操作,生成拟合高倍图像,并获取所述拟合高倍图像中感兴趣区域的所有图像坐标,再利用所述图像坐标将所述拟合高倍图像中的感兴趣区域分割出来,生成分割高倍图像;
标注切割步骤,用于对所述分割高倍图像通过自适应阈值分割算法进行标注切割,得到异常细胞标注集;
异常生成步骤,用于将所述异常细胞标注集映射到所述细胞病理图片上,得到异常细胞标注图。
为了解决上述问题,本申请还提供一种异常细胞自动标注装置,其中,所述装置包括:
图像金字塔生成模块,用于获取细胞病理图片,对所述细胞病理图片进行预设次数的高斯卷积平滑处理,生成多个细胞病理图片,根据所述细胞病理图片得到图像金字塔;
检测拟合模块,用于对所述图像金字塔进行检测及拟合,得到低倍拟合感兴趣区域;
映射分割模块,用于将所述低倍拟合感兴趣区域映射到获取的所述细胞病理图片上并进行图像放大操作,生成拟合高倍图像,并获取所述拟合高倍图像中感兴趣区域的所有图像坐标,再利用所述图像坐标将所述拟合高倍图像中的感兴趣区域分割出来,生成分割高倍图像;
标注切割模块,用于对所述分割高倍图像通过自适应阈值分割算法进行标注切割,得到异常细胞标注集;
异常生成模块,用于将所述异常细胞标注集映射到所述细胞病理图片上,得到异常细胞标注图。
本申请实施例对原始细胞病理图片通过高斯卷积平滑处理,得到由自上而下分辨率越来越高的多个细胞病例图片组成的图像金字塔,这些细胞病例图片中的病理数据特征更明显,提高了异常细胞标注的准确性;进一步地,由于图像金字塔中的图像像素的分辨率很大,一般的图像处理方法无法直接用计算机做图像分析处理,于是本申请实施例将像素拟合到低倍区域,缩小了图像的像素和分辨率,在不影响后续异常细胞检测准确率的同时,进一步释放了计算机的存储压力;另外当得到低倍拟合感兴趣区域后,本申请实施例进一步通过放大操作变为拟合高倍图像,高倍图像的分辨率高,可有效提高使用算法进行标注切割的准确率。因此本申请提出的异常细胞自动标注方法、装置、电子设备及计算机可读存储介质,可以解决异常细胞标注的准确性低以及计算和存储压力高的问题。
附图说明
图1为本申请一实施例提供的异常细胞自动标注方法的流程示意图;
图2为本申请一实施例提供的异常细胞自动标注装置的模块示意图;
图3为本申请一实施例提供的执行所述异常细胞自动标注方法的电子设备的内部结构示意图;
本申请目的的实现、功能特点及优点将结合实施例,参照附图做进一步说明。
具体实施方式
为使本申请实施例的目的、技术方案和优点更加清楚,下面将结合附图对本申请的各实施方式进行详细的阐述。然而,本领域的普通技术人员可以理解,在本申请各实施方式 中,为了使读者更好地理解本申请而提出了许多技术细节。但是,即使没有这些技术细节和基于以下各实施方式的种种变化和修改,也可以实现本申请所要求保护的技术方案。
本方案可应用于智慧医疗领域中,从而推动智慧城市的建设,本申请实施例的目的是通过自适应阈值分割对异常细胞进行标注,用于提高异常细胞标注的准确性以及减轻计算和存储压力,例如,在癌细胞筛查过程中,通过对细胞病理图片进行分层,使得分层后的数据特征更明显,对分层后的所述细胞病理图片中异常细胞进行标注,通过坐标映射的方法将所述宫颈细胞病理图片中的低像素图像中异常细胞映射到高像素图像中,用于后续的异常细胞标注分割。
下面对本实施方式的异常细胞自动标注方法实现细节进行具体的说明,以下内容仅为方便理解提供的实现细节,并非实施本方案的必须。
参照图1所示,为本申请一实施例提供的异常细胞自动标注方法的流程示意图。该方法可以由一个装置执行,该装置可以由软件和/或硬件实现。
在本实施例中,异常细胞自动标注方法包括:
S1、获取细胞病理图片,对所述细胞病理图片进行预设次数的高斯卷积平滑处理,生成多个细胞病理图片,根据所述细胞病理图片得到图像金字塔。
例如,在***筛查领域内,本申请实施例所述细胞病理图片包括宫颈细胞病理图片。
详细地,所述对所述细胞病理图片进行预设次数的高斯卷积平滑处理,生成多个细胞病理图片,根据所述细胞病理图片得到图像金字塔,包括:
步骤A:将所述获取的细胞病理图片执行扩大操作,得到第a组第b层图像,其中,a、b的初始值均为1,并将所述第a组第b层图像利用高斯卷积函数进行高斯卷积的平滑操作,得到第a组第b+1层图像,其中,所述高斯卷积函数为:
Figure PCTCN2020098969-appb-000001
其中σ表示平滑因子,G (x,y,σ)表示x与y的卷积,x、y代表图像坐标;
步骤B:将所述平滑因子σ乘以比例系数k,得到新的平滑因子kσ,通过所述新的平滑因子kσ对所述第a组第b+1层图像利用上述高斯卷积函数进行高斯卷积的平滑操作,得到第a组第b+2层图像,并重复该步骤,直到直到得到第a组第L层图像,其中L为预定义值;
步骤C:对所述第a组第L层图像进行降采样操作,得到第a+1组第b层图像,对所述第a+1组第b层图像利用上述高斯卷积函数进行高斯卷积的平滑操作,得到第a+1组第b+1层图像,重复所述步骤B,直至得到第a+1组第L层图像;
步骤D:对所述步骤C得到的结果进行所述步骤C和步骤B的循环操作,直至得到第O组第L层图像,其中O为预定义值;
步骤E:将从所述第a组第b层图像到所述第O组第L层图像组合起来,生成所述图像金字塔。其中,所述第a组第b层图像到所述第O组第L层图像中的每一个图像就是所述图像金字塔的一个图片层。
通过高斯卷积平滑处理后,可以将细胞病理图片拆分为自上而下分辨率越来越高的多个细胞病理图片,使得病理数据特征更明显,提高了异常细胞标注的准确性。
S2、对所述图像金字塔通过霍夫变换圆检测方法进行检测及拟合,得到低倍拟合感兴趣区域。
详细地,所述S2包括:将所述图像金字塔进行灰度转化,生成灰度图像;对所述灰度图像执行二值化处理,得到轮廓图像;从所述轮廓图像中采用随机选取的方法,得到候选区域;利用霍夫圆变换检测方法识别出所述候选区域中的感兴趣区域,对所述感兴趣区 域进行拟合,得到所述低倍拟合感兴趣区域。
进一步的,所述利用霍夫圆变换检测方法识别出所述候选区域中的感兴趣区域,对所述感兴趣区域进行拟合,得到所述低倍拟合感兴趣区域,包括:
步骤a、在所述候选区域的图像空间中通过边缘检测算法对所述候选区域的边缘进行检测,得到n个边缘像素点集;
步骤b、将所述n个边缘像素点集,以预定义值r为半径映射到参数空间:
Figure PCTCN2020098969-appb-000002
其中,r为预定义值,θ∈[0,2π),x与y表示所述n个边缘像素点集相应的坐标(x,y),(a,b)表示所述参数空间中参考点坐标;
步骤c、将所述参数空间中所有坐标点进行统计,对θ进行遍历,使得图像空间上的边缘像素点映射到参数空间为一个圆时,将这个圆作为霍夫圆,通过所述霍夫圆构成所述感兴趣区域;
步骤d、将所述感兴趣区域进行拟合,生成低倍拟合感兴趣区域。
通过本申请实施例,细胞病理图片的感兴趣区域可以通过医生任意标注区域,经过霍夫圆变换检测方法识别并拟合出来的。
S3、将所述低倍拟合感兴趣区域映射到获取的所述细胞病理图片上并进行图像放大操作,生成拟合高倍图像,并获取所述拟合高倍图像中感兴趣区域的所有图像坐标。
本申请实施例中,细胞病理图片的高倍图像尺寸由于很大,一般的图像处理方法难以载入进行直接处理,无法直接用计算机做图像分析处理,因此,本申请实施例通过将所述低倍拟合感兴趣区域映射到获取的高倍图像上,生成拟合高倍图像,获取所述拟合高倍图像中的感兴趣区域图像坐标信息,所述感兴趣区域图像坐标信息用于后续对所述高倍图像进行分割。
本申请实施例中,所述映射是将低分辨图像划分为多个子区域,通过子区域坐标与倍数相乘,将各个所述子区域添加至高分辨率图像中,所述倍数为所述高分辨率图像相对于低分辨率图像的放大倍数。
S4、利用所述图像坐标将所述拟合高倍图像中的感兴趣区域分割出来,生成分割高倍图像。
本申请实施通过拟合获取的感兴趣区域所有的图像坐标,可以将细胞感兴趣区域进行分割出来,生成尺寸,例如为3000x3000的分割高倍图像。
S5、对所述分割高倍图像通过自适应阈值分割算法进行标注切割,得到异常细胞标注集。
详细的,所述S5包括:
步骤Ⅰ、将所述分割高倍图像的每一个像素i通过公式:
Figure PCTCN2020098969-appb-000003
转换为0-1之间的实数,得到归一化值h;
步骤Ⅱ、利用预定义的伽马校正补偿值将所述归一化值h进行预补偿计算,得到预补常值f;
步骤Ⅲ、将所述预补常值f通过公式f*256-0.5进行反归一化计算,得到所述分割高倍图像的校正图像;
步骤Ⅳ、将所述校正图像的图像灰度级和所述校正图像中坐标的像素点灰度级构成二元组,计算所有二元组的均值和方差,通过所述均值和方差建立二维最大类间方差模型,对所述二维最大类间方差模型通过自适应粒子群集算法进行计算,生成所述校正图像的最佳阈值;
步骤Ⅴ、利用所述最佳阈值对所述校正图像进行分割,生成背景前景分割图像;
步骤Ⅵ、将所述背景前景分割图像进行开运算与闭运算处理,生成异常细胞标注集。
本申请实施例中,对所述背景前景分割图像通过形态学算法依次进行开运算计算,将分割的所述图像中孤立的小点去除,生成第一图像集,将所述第一图像集,通过形态学算法依次进行闭运算计算,将所述第一图像集中图像细胞之间的小裂缝进行填充,清除所述第一图像集中图像内部的空洞,使得所述第一图像集中前景细胞与所述第一图像集中背景区进行分离,生成异常细胞标注集。
传统的细胞图像分割方法大致分为两类:基于区域的分割方法以及基于边缘的分割方法,基于区域的分割方法的基本原理是通过把具有相似特征的相邻区域归为一类来实现分割,基于边缘的分割方法一般通过把灰度级或者结构具有突变的地方作为边缘来进行分割,本申请采取基于边缘的分割方法通过自适应阈值分割算法进行图像分割,得到异常细胞标注集。
S6、将所述异常细胞标注集映射到所述细胞病理图片上,得到异常细胞标注图。
本申请实施例中,所述异常细胞标注集用于对目标细胞通过深度学***滑的方法进行边缘调整。
需要强调的是,为进一步保证上述异常细胞标注集的私密和安全性,上述异常细胞标注集还可以存储于一区块链的节点中。
如图2所示,是本申请异常细胞自动标注装置的功能模块图。
本申请实施例所述异常细胞自动标注装置100可以安装于电子设备中。根据实现的功能,所述异常细胞自动标注装置可以包括检测拟合模块101、映射分割模块102、标注切割模块103和异常生成模块104。本发所述模块也可以称之为单元,是指一种能够被电子设备处理器所执行,并且能够完成固定功能的一系列计算机程序段,其存储在电子设备的存储器中。
在本实施例中,关于各模块/单元的功能如下:
图像金字塔生成模块101,用于获取细胞病理图片,对所述细胞病理图片进行预设次数的高斯卷积平滑处理,生成多个细胞病理图片,根据所述细胞病理图片得到图像金字塔;
检测拟合模块102,用于对所述图像金字塔通过霍夫变换圆检测方法进行检测及拟合,得到低倍拟合感兴趣区域;
映射分割模块103,用于将所述低倍拟合感兴趣区域映射到获取的所述细胞病理图片上并进行图像放大操作,生成拟合高倍图像,并获取所述拟合高倍图像中感兴趣区域的所有图像坐标,再利用所述图像坐标将所述拟合高倍图像中的感兴趣区域分割出来,生成分割高倍图像;
标注切割模块104,用于对所述分割高倍图像通过自适应阈值分割算法进行标注切割,得到异常细胞标注集;
异常生成模块105,用于将所述异常细胞标注集映射到所述细胞病理图片上,得到异常细胞标注图。
详细地,所述异常细胞自动标注装置各模块的具体实施步骤如下:
所述图像金字塔生成模块101、获取细胞病理图片,对所述细胞病理图片进行预设次数的高斯卷积平滑处理,生成多个细胞病理图片,根据所述细胞病理图片得到图像金字塔。
例如,在***筛查领域内,本申请实施例所述细胞病理图片包括宫颈细胞病理图片。
详细地,所述对所述细胞病理图片进行预设次数的高斯卷积平滑处理,生成多个细胞病理图片,根据所述细胞病理图片得到图像金字塔,包括:
步骤A:将所述获取的细胞病理图片执行扩大操作,得到第a组第b层图像,其中,a、b的初始值均为1,并将所述第a组第b层图像利用高斯卷积函数进行高斯卷积的平滑操作,得到第a组第b+1层图像,其中,所述高斯卷积函数为:
Figure PCTCN2020098969-appb-000004
其中σ表示平滑因子,G (x,y,σ)表示x与y的卷积,x、y代表图像坐标;
步骤B:将所述平滑因子σ乘以比例系数k,得到新的平滑因子kσ,通过所述新的平滑因子kσ对所述第a组第b+1层图像利用上述高斯卷积函数进行高斯卷积的平滑操作,得到第a组第b+2层图像,并重复该步骤,直到直到得到第a组第L层图像,其中L为预定义值;
步骤C:对所述第a组第L层图像进行降采样操作,得到第a+1组第b层图像,对所述第a+1组第b层图像利用上述高斯卷积函数进行高斯卷积的平滑操作,得到第a+1组第b+1层图像,重复所述步骤B,直至得到第a+1组第L层图像;
步骤D:对所述步骤C得到的结果进行所述步骤C和步骤B的循环操作,直至得到第O组第L层图像,其中O为预定义值;
步骤E:将从所述第a组第b层图像到所述第O组第L层图像组合起来,生成所述图像金字塔。其中,所述第a组第b层图像到所述第O组第L层图像中的每一个图像就是所述图像金字塔的一个图片层。
通过高斯卷积平滑处理后,可以将细胞病理图片拆分为自上而下分辨率越来越高的多个细胞病理图片,使得病理数据特征更明显,提高了异常细胞标注的准确性。
所述检测拟合模块102、对所述图像金字塔通过霍夫变换圆检测方法进行检测及拟合,得到低倍拟合感兴趣区域。
详细地,所述检测拟合模块102:将所述图像金字塔进行灰度转化,生成灰度图像;对所述灰度图像执行二值化处理,得到轮廓图像;从所述轮廓图像中采用随机选取的方法,得到候选区域;利用霍夫圆变换检测方法识别出所述候选区域中的感兴趣区域,对所述感兴趣区域进行拟合,得到所述低倍拟合感兴趣区域。
进一步的,所述利用霍夫圆变换检测方法识别出所述候选区域中的感兴趣区域,对所述感兴趣区域进行拟合,得到所述低倍拟合感兴趣区域,包括:
步骤a、在所述候选区域的图像空间中通过边缘检测算法对所述候选区域的边缘进行检测,得到n个边缘像素点集;
步骤b、将所述n个边缘像素点集,以预定义值r为半径映射到参数空间:
Figure PCTCN2020098969-appb-000005
其中,r为预定义值,θ∈[0,2π),x与y表示所述n个边缘像素点集相应的坐标(x,y),(a,b)表示所述参数空间中参考点坐标;
步骤c、将所述参数空间中所有坐标点进行统计,对θ进行遍历,使得图像空间上的边缘像素点映射到参数空间为一个圆时,将这个圆作为霍夫圆,通过所述霍夫圆构成所述感兴趣区域;
步骤d、将所述感兴趣区域进行拟合,生成低倍拟合感兴趣区域。
通过本申请实施例,细胞病理图片的感兴趣区域可以通过医生任意标注区域,经过霍夫圆变换检测方法识别并拟合出来的。
所述映射分割模块103、将所述低倍拟合感兴趣区域映射到获取的所述细胞病理图片上并进行图像放大操作,生成拟合高倍图像,并获取所述拟合高倍图像中感兴趣区域的所有图像坐标。
本申请实施例中,细胞病理图片的高倍图像尺寸由于很大,一般的图像处理方法难以载入进行直接处理,无法直接用计算机做图像分析处理,因此,本申请实施例通过将所述低倍拟合感兴趣区域映射到获取的高倍图像上,生成拟合高倍图像,获取所述拟合高倍图 像中的感兴趣区域图像坐标信息,所述感兴趣区域图像坐标信息用于后续对所述高倍图像进行分割。
本申请实施例中,所述映射是将低分辨图像划分为多个子区域,通过子区域坐标与倍数相乘,将各个所述子区域添加至高分辨率图像中,所述倍数为所述高分辨率图像相对于低分辨率图像的放大倍数。进一步地,所述映射分割模块103利用所述图像坐标将所述拟合高倍图像中的感兴趣区域分割出来,生成分割高倍图像。
本申请实施通过拟合获取的感兴趣区域所有的图像坐标,可以将细胞感兴趣区域进行分割出来,生成尺寸,例如为3000x3000的分割高倍图像。
所述标注切割模块104、对所述分割高倍图像通过自适应阈值分割算法进行标注切割,得到异常细胞标注集。
详细地,所述对所述分割高倍图像进行标注切割,得到异常细胞标注集包括:
步骤Ⅰ、将所述分割高倍图像的每一个像素i通过公式:
Figure PCTCN2020098969-appb-000006
转换为0-1之间的实数,得到归一化值h;
步骤Ⅱ、利用预定义的伽马校正补偿值将所述归一化值h进行预补偿计算,得到预补常值f;
步骤Ⅲ、将所述预补常值f通过公式f*256-0.5进行反归一化计算,得到所述分割高倍图像的校正图像;
步骤Ⅳ、将所述校正图像的图像灰度级和所述校正图像中坐标的像素点灰度级构成二元组,计算所有二元组的均值和方差,通过所述均值和方差建立二维最大类间方差模型,对所述二维最大类间方差模型通过自适应粒子群集算法进行计算,生成所述校正图像的最佳阈值;
步骤Ⅴ、利用所述最佳阈值对所述校正图像进行分割,生成背景前景分割图像;
步骤Ⅵ、将所述背景前景分割图像进行开运算与闭运算处理,生成异常细胞标注集。
本申请实施例中,对所述背景前景分割图像通过形态学算法依次进行开运算计算,将分割的所述图像中孤立的小点去除,生成第一图像集,将所述第一图像集,通过形态学算法依次进行闭运算计算,将所述第一图像集中图像细胞之间的小裂缝进行填充,清除所述第一图像集中图像内部的空洞,使得所述第一图像集中前景细胞与所述第一图像集中背景区进行分离,生成异常细胞标注集。
传统的细胞图像分割方法大致分为两类:基于区域的分割方法以及基于边缘的分割方法,基于区域的分割方法的基本原理是通过把具有相似特征的相邻区域归为一类来实现分割,基于边缘的分割方法一般通过把灰度级或者结构具有突变的地方作为边缘来进行分割,本申请采取基于边缘的分割方法通过自适应阈值分割算法进行图像分割,得到异常细胞标注集。
所述异常生成模块105、将所述异常细胞标注集映射到所述细胞病理图片上,得到异常细胞标注图。
本申请实施例中,所述异常细胞标注集用于对目标细胞通过深度学***滑的方法进行边缘调整。
需要强调的是,为进一步保证上述异常细胞标注集的私密和安全性,上述异常细胞标注集还可以存储于一区块链的节点中。
如图3所示,是实现本申请异常细胞自动标注方法的电子设备的结构示意图。
所述电子设备1可以包括处理器10、存储器11和总线,还可以包括存储在所述存储器11中并可在所述处理器10上运行的计算机程序,如异常细胞自动标注程序12。
其中,所述存储器11至少包括一种类型的可读存储介质,所述可读存储介质包括闪 存、移动硬盘、多媒体卡、卡型存储器(例如:SD或DX存储器等)、磁性存储器、磁盘、光盘等。所述存储器11在一些实施例中可以是电子设备1的内部存储单元,例如该电子设备1的移动硬盘。所述存储器11在另一些实施例中也可以是电子设备1的外部存储设备,例如电子设备1上配备的插接式移动硬盘、智能存储卡(Smart Media Card,SMC)、安全数字(Secure Digital,SD)卡、闪存卡(Flash Card)等。进一步地,所述存储器11还可以既包括电子设备1的内部存储单元也包括外部存储设备。所述存储器11不仅可以用于存储安装于电子设备1的应用软件及各类数据,例如异常细胞自动标注程序的代码等,还可以用于暂时地存储已经输出或者将要输出的数据。
所述处理器10在一些实施例中可以由集成电路组成,例如可以由单个封装的集成电路所组成,也可以是由多个相同功能或不同功能封装的集成电路所组成,包括一个或者多个中央处理器(Central Processing unit,CPU)、微处理器、数字处理芯片、图形处理器及各种控制芯片的组合等。所述处理器10是所述电子设备的控制核心(Control Unit),利用各种接口和线路连接整个电子设备的各个部件,通过运行或执行存储在所述存储器11内的程序或者模块(例如执行异常细胞自动标注程序等),以及调用存储在所述存储器11内的数据,以执行电子设备1的各种功能和处理数据。
所述总线可以是外设部件互连标准(peripheral component interconnect,简称PCI)总线或扩展工业标准结构(extended industry standard architecture,简称EISA)总线等。该总线可以分为地址总线、数据总线、控制总线等。所述总线被设置为实现所述存储器11以及至少一个处理器10等之间的连接通信。
图3仅示出了具有部件的电子设备,本领域技术人员可以理解的是,图3示出的结构并不构成对所述电子设备1的限定,可以包括比图示更少或者更多的部件,或者组合某些部件,或者不同的部件布置。
例如,尽管未示出,所述电子设备1还可以包括给各个部件供电的电源(比如电池),优选地,电源可以通过电源管理装置与所述至少一个处理器10逻辑相连,从而通过电源管理装置实现充电管理、放电管理、以及功耗管理等功能。电源还可以包括一个或一个以上的直流或交流电源、再充电装置、电源故障检测电路、电源转换器或者逆变器、电源状态指示器等任意组件。所述电子设备1还可以包括多种传感器、蓝牙模块、Wi-Fi模块等,在此不再赘述。
进一步地,所述电子设备1还可以包括网络接口,可选地,所述网络接口可以包括有线接口和/或无线接口(如WI-FI接口、蓝牙接口等),通常用于在该电子设备1与其他电子设备之间建立通信连接。
可选地,该电子设备1还可以包括用户接口,用户接口可以是显示器(Display)、输入单元(比如键盘(Keyboard)),可选地,用户接口还可以是标准的有线接口、无线接口。可选地,在一些实施例中,显示器可以是LED显示器、液晶显示器、触控式液晶显示器以及OLED(Organic Light-Emitting Diode,有机发光二极管)触摸器等。其中,显示器也可以适当的称为显示屏或显示单元,用于显示在电子设备1中处理的信息以及用于显示可视化的用户界面。
应该了解,所述实施例仅为说明之用,在专利申请范围上并不受此结构的限制。
所述电子设备1中的所述存储器11存储的异常细胞自动标注程序12是多个指令的组合,在所述处理器10中运行时,可以实现:
获取细胞病理图片,对所述细胞病理图片进行预设次数的高斯卷积平滑处理,生成多个细胞病理图片,根据所述细胞病理图片得到图像金字塔;
对所述图像金字塔通过霍夫变换圆检测方法进行检测及拟合,得到低倍拟合感兴趣区域;
将所述低倍拟合感兴趣区域映射到获取的所述细胞病理图片上并进行图像放大操作, 生成拟合高倍图像,并获取所述拟合高倍图像中感兴趣区域的所有图像坐标;
利用所述图像坐标将所述拟合高倍图像中的感兴趣区域分割出来,生成分割高倍图像;
对所述分割高倍图像通过自适应阈值分割算法进行标注切割,得到异常细胞标注集;
将所述异常细胞标注集映射到所述细胞病理图片上,得到异常细胞标注图。
具体地,所述处理器10对上述指令的具体实现方法可参考图1对应实施例中相关步骤的描述,在此不赘述。
进一步地,所述电子设备1集成的模块/单元如果以软件功能单元的形式实现并作为独立的产品销售或使用时,可以存储在一个计算机可读存储介质中。所述计算机可读存储介质可以是非易失性,也可以是易失性。所述计算机可读存储介质可以包括:能够携带所述计算机程序代码的任何实体或装置、记录介质、U盘、移动硬盘、磁碟、光盘、计算机存储器、只读存储器(ROM,Read-Only Memory)。
进一步地,所述计算机可读存储介质可主要包括存储程序区和存储数据区,其中,存储程序区可存储操作***、至少一个功能所需的应用程序等;存储数据区可存储根据区块链节点的使用所创建的数据等。
本申请所指区块链是分布式数据存储、点对点传输、共识机制、加密算法等计算机技术的新型应用模式。区块链(Blockchain),本质上是一个去中心化的数据库,是一串使用密码学方法相关联产生的数据块,每一个数据块中包含了一批次网络交易的信息,用于验证其信息的有效性(防伪)和生成下一个区块。区块链可以包括区块链底层平台、平台产品服务层以及应用服务层等。
在本申请所提供的几个实施例中,应该理解到,所揭露的设备,装置和方法,可以通过其它的方式实现。例如,以上所描述的装置实施例仅仅是示意性的,例如,所述模块的划分,仅仅为一种逻辑功能划分,实际实现时可以有另外的划分方式。
所述作为分离部件说明的模块可以是或者也可以不是物理上分开的,作为模块显示的部件可以是或者也可以不是物理单元,即可以位于一个地方,或者也可以分布到多个网络单元上。可以根据实际的需要选择其中的部分或者全部模块来实现本实施例方案的目的。
另外,在本申请各个实施例中的各功能模块可以集成在一个处理单元中,也可以是各个单元单独物理存在,也可以两个或两个以上单元集成在一个单元中。上述集成的单元既可以采用硬件的形式实现,也可以采用硬件加软件功能模块的形式实现。
对于本领域技术人员而言,显然本申请不限于上述示范性实施例的细节,而且在不背离本申请的精神或基本特征的情况下,能够以其他的具体形式实现本申请。
因此,无论从哪一点来看,均应将实施例看作是示范性的,而且是非限制性的,本申请的范围由所附权利要求而不是上述说明限定,因此旨在将落在权利要求的等同要件的含义和范围内的所有变化涵括在本申请内。不应将权利要求中的任何附关联图标记视为限制所涉及的权利要求。
此外,显然“包括”一词不排除其他单元或步骤,单数不排除复数。***权利要求中陈述的多个单元或装置也可以由一个单元或装置通过软件或者硬件来实现。第二等词语用来表示名称,而并不表示任何特定的顺序。
最后应说明的是,以上实施例仅用以说明本申请的技术方案而非限制,尽管参照较佳实施例对本申请进行了详细说明,本领域的普通技术人员应当理解,可以对本申请的技术方案进行修改或等同替换,而不脱离本申请技术方案的精神和范围。

Claims (20)

  1. 一种异常细胞自动标注方法,其中,所述方法包括:
    获取细胞病理图片,对所述细胞病理图片进行预设次数的高斯卷积平滑处理,生成多个细胞病理图片,根据所述细胞病理图片得到图像金字塔;
    对所述图像金字塔进行检测及拟合,得到低倍拟合感兴趣区域;
    将所述低倍拟合感兴趣区域映射到获取的所述细胞病理图片上并进行图像放大操作,生成拟合高倍图像,并获取所述拟合高倍图像中感兴趣区域的所有图像坐标;
    利用所述图像坐标将所述拟合高倍图像中的感兴趣区域分割出来,生成分割高倍图像;
    对所述分割高倍图像通过自适应阈值分割算法进行标注切割,得到异常细胞标注集;
    将所述异常细胞标注集映射到所述细胞病理图片上,得到异常细胞标注图。
  2. 如权利要求1所述的异常细胞自动标注方法,其中,所述对所述细胞病理图片进行预设次数的高斯卷积平滑处理,生成多个细胞病理图片,根据所述细胞病理图片得到图像金字塔,包括:
    步骤A:将所述获取的细胞病理图片执行扩大操作,得到第a组第b层图像,其中,a、b的初始值均为1,并将所述第a组第b层图像利用高斯卷积函数进行高斯卷积的平滑操作,得到第a组第b+1层图像,其中,所述高斯卷积函数为:
    Figure PCTCN2020098969-appb-100001
    其中σ表示平滑因子,G (x,y,σ)表示x与y的卷积,x、y代表图像坐标;
    步骤B:将所述平滑因子σ乘以比例系数k,得到新的平滑因子kσ,通过所述新的平滑因子kσ对所述第a组第b+1层图像利用上述高斯卷积函数进行高斯卷积的平滑操作,得到第a组第b+2层图像,并重复该步骤,直到直到得到第a组第L层图像,其中L为预定义值;
    步骤C:对所述第a组第L层图像进行降采样操作,得到第a+1组第b层图像,对所述第a+1组第b层图像利用上述高斯卷积函数进行高斯卷积的平滑操作,得到第a+1组第b+1层图像,重复所述步骤B,直至得到第a+1组第L层图像;
    步骤D:对所述步骤C得到的结果进行所述步骤C和步骤B的循环操作,直至得到第O组第L层图像,其中O为预定义值;
    步骤E:将从所述第a组第b层图像到所述第O组第L层图像组合起来,生成所述图像金字塔。
  3. 如权利要求1所述的异常细胞自动标注方法,其中,所述对所述图像金字塔通过霍夫变换圆检测方法进行检测及拟合,得到所述低倍拟合感兴趣区域,包括:
    将所述图像金字塔进行灰度转化,生成灰度图像;
    对所述灰度图像执行二值化处理,得到轮廓图像;
    从所述轮廓图像中采用随机选取的方法,得到候选区域;
    利用霍夫圆变换检测方法识别出所述候选区域中的感兴趣区域,对所述感兴趣区域进行拟合,得到所述低倍拟合感兴趣区域。
  4. 如权利要求3所述的异常细胞自动标注方法,其中,所述利用霍夫变换圆检测方法识别出所述感兴趣区域并拟合,得到所述低倍拟合感兴趣区域,包括:
    在所述候选区域的图像空间中通过边缘检测算法对所述候选区域的边缘进行检测,得到n个边缘像素点集;
    将所述n个边缘像素点集,以预定义值r为半径映射到参数空间:
    Figure PCTCN2020098969-appb-100002
    其中,r为预定义值,θ∈[0,2π),x与y表示所述n个边缘像素点集相应的坐标(x,y),(a,b)表示所述参数空间中参考点坐标;
    将所述参数空间中所有坐标点进行统计,对θ进行遍历,使得图像空间上的边缘像素点映射到参数空间为一个圆时,将这个圆作为霍夫圆,通过所述霍夫圆构成所述感兴趣区域;
    将所述感兴趣区域进行拟合,生成低倍拟合感兴趣区域。
  5. 如权利要求1所述的异常细胞自动标注方法,其中,所述映射是将低分辨图像划分为多个子区域,通过子区域坐标与倍数相乘,将各个所述子区域添加至高分辨率图像中,所述倍数为所述高分辨率图像相对于低分辨率图像的放大倍数。
  6. 如权利要求1所述的异常细胞自动标注方法,其中,所述对所述分割高倍图像通过自适应阈值分割算法进行标注切割,得到异常细胞标注集,包括:
    将所述分割高倍图像的每一个像素i通过公式:
    Figure PCTCN2020098969-appb-100003
    转换为0-1之间的实数,得到归一化值h;
    利用预定义的伽马校正补偿值将所述归一化值h进行预补偿计算,得到预补常值f;
    将所述预补常值f通过公式f*256-0.5进行反归一化计算,得到所述分割高倍图像的校正图像;
    将所述校正图像的图像灰度级和所述校正图像中坐标的像素点灰度级构成二元组,计算所有二元组的均值和方差,通过所述均值和方差建立二维最大类间方差模型,对所述二维最大类间方差模型通过自适应粒子群集算法进行计算,生成所述校正图像的最佳阈值;
    利用所述最佳阈值对所述校正图像进行分割,生成背景前景分割图像;
    将所述背景前景分割图像进行开运算与闭运算处理,生成异常细胞标注集。
  7. 如权利要求6所述的异常细胞自动标注方法,其中,所述将所述背景前景分割图像进行开运算与闭运算处理,生成异常细胞标注集,包括:
    对所述背景前景分割图像通过形态学算法依次进行开运算计算,将分割的所述图像中孤立的小点去除,生成第一图像集;
    将所述第一图像集,通过形态学算法依次进行闭运算计算,将所述第一图像集中图像细胞之间的小裂缝进行填充,清除所述第一图像集中图像内部的空洞,使得所述第一图像集中前景细胞与所述第一图像集中背景区进行分离,生成异常细胞标注集。
  8. 一种电子设备,其中,所述电子设备包括:
    至少一个处理器;以及,
    与所述至少一个处理器通信连接的存储器;其中,
    所述存储器存储有可被所述至少一个处理器执行的指令,所述指令被所述至少一个处理器执行,以使所述至少一个处理器能够执行如下所述的异常细胞自动标注方法:
    获取细胞病理图片,对所述细胞病理图片进行预设次数的高斯卷积平滑处理,生成多个细胞病理图片,根据所述细胞病理图片得到图像金字塔;
    对所述图像金字塔进行检测及拟合,得到低倍拟合感兴趣区域;
    将所述低倍拟合感兴趣区域映射到获取的所述细胞病理图片上并进行图像放大操作,生成拟合高倍图像,并获取所述拟合高倍图像中感兴趣区域的所有图像坐标;
    利用所述图像坐标将所述拟合高倍图像中的感兴趣区域分割出来,生成分割高倍图像;
    对所述分割高倍图像通过自适应阈值分割算法进行标注切割,得到异常细胞标注集;
    将所述异常细胞标注集映射到所述细胞病理图片上,得到异常细胞标注图。
  9. 如权利要求8所述的电子设备,其中,所述对所述细胞病理图片进行预设次数的 高斯卷积平滑处理,生成多个细胞病理图片,根据所述细胞病理图片得到图像金字塔,包括:
    步骤A:将所述获取的细胞病理图片执行扩大操作,得到第a组第b层图像,其中,a、b的初始值均为1,并将所述第a组第b层图像利用高斯卷积函数进行高斯卷积的平滑操作,得到第a组第b+1层图像,其中,所述高斯卷积函数为:
    Figure PCTCN2020098969-appb-100004
    其中σ表示平滑因子,G (x,y,σ)表示x与y的卷积,x、y代表图像坐标;
    步骤B:将所述平滑因子σ乘以比例系数k,得到新的平滑因子kσ,通过所述新的平滑因子kσ对所述第a组第b+1层图像利用上述高斯卷积函数进行高斯卷积的平滑操作,得到第a组第b+2层图像,并重复该步骤,直到直到得到第a组第L层图像,其中L为预定义值;
    步骤C:对所述第a组第L层图像进行降采样操作,得到第a+1组第b层图像,对所述第a+1组第b层图像利用上述高斯卷积函数进行高斯卷积的平滑操作,得到第a+1组第b+1层图像,重复所述步骤B,直至得到第a+1组第L层图像;
    步骤D:对所述步骤C得到的结果进行所述步骤C和步骤B的循环操作,直至得到第O组第L层图像,其中O为预定义值;
    步骤E:将从所述第a组第b层图像到所述第O组第L层图像组合起来,生成所述图像金字塔。
  10. 如权利要求8所述的电子设备,其中,所述对所述图像金字塔通过霍夫变换圆检测方法进行检测及拟合,得到所述低倍拟合感兴趣区域,包括:
    将所述图像金字塔进行灰度转化,生成灰度图像;
    对所述灰度图像执行二值化处理,得到轮廓图像;
    从所述轮廓图像中采用随机选取的方法,得到候选区域;
    利用霍夫圆变换检测方法识别出所述候选区域中的感兴趣区域,对所述感兴趣区域进行拟合,得到所述低倍拟合感兴趣区域。
  11. 如权利要求10所述的电子设备,其中,所述利用霍夫变换圆检测方法识别出所述感兴趣区域并拟合,得到所述低倍拟合感兴趣区域,包括:
    在所述候选区域的图像空间中通过边缘检测算法对所述候选区域的边缘进行检测,得到n个边缘像素点集;
    将所述n个边缘像素点集,以预定义值r为半径映射到参数空间:
    Figure PCTCN2020098969-appb-100005
    其中,r为预定义值,θ∈[0,2π),x与y表示所述n个边缘像素点集相应的坐标(x,y),(a,b)表示所述参数空间中参考点坐标;
    将所述参数空间中所有坐标点进行统计,对θ进行遍历,使得图像空间上的边缘像素点映射到参数空间为一个圆时,将这个圆作为霍夫圆,通过所述霍夫圆构成所述感兴趣区域;
    将所述感兴趣区域进行拟合,生成低倍拟合感兴趣区域。
  12. 如权利要求8所述的电子设备,其中,所述映射是将低分辨图像划分为多个子区域,通过子区域坐标与倍数相乘,将各个所述子区域添加至高分辨率图像中,所述倍数为所述高分辨率图像相对于低分辨率图像的放大倍数。
  13. 如权利要求8所述的电子设备,其中,所述对所述分割高倍图像通过自适应阈值分割算法进行标注切割,得到异常细胞标注集,包括:
    将所述分割高倍图像的每一个像素i通过公式:
    Figure PCTCN2020098969-appb-100006
    转换为0-1之间的实数,得到归一化值h;
    利用预定义的伽马校正补偿值将所述归一化值h进行预补偿计算,得到预补常值f;
    将所述预补常值f通过公式f*256-0.5进行反归一化计算,得到所述分割高倍图像的校正图像;
    将所述校正图像的图像灰度级和所述校正图像中坐标的像素点灰度级构成二元组,计算所有二元组的均值和方差,通过所述均值和方差建立二维最大类间方差模型,对所述二维最大类间方差模型通过自适应粒子群集算法进行计算,生成所述校正图像的最佳阈值;
    利用所述最佳阈值对所述校正图像进行分割,生成背景前景分割图像;
    将所述背景前景分割图像进行开运算与闭运算处理,生成异常细胞标注集。
  14. 一种计算机可读存储介质,其中,所述计算机可读存储介质上存储有异常细胞自动标注程序,所述异常细胞自动标注程序可被一个或者多个处理器执行,以实现如下所述的异常细胞自动标注方法的步骤:
    图像金字塔生成步骤,用于获取细胞病理图片,对所述细胞病理图片进行预设次数的高斯卷积平滑处理,生成多个细胞病理图片,根据所述细胞病理图片得到图像金字塔;
    检测拟合步骤,用于对所述图像金字塔进行检测及拟合,得到低倍拟合感兴趣区域;
    映射分割步骤,用于将所述低倍拟合感兴趣区域映射到获取的所述细胞病理图片上并进行图像放大操作,生成拟合高倍图像,并获取所述拟合高倍图像中感兴趣区域的所有图像坐标,再利用所述图像坐标将所述拟合高倍图像中的感兴趣区域分割出来,生成分割高倍图像;
    标注切割步骤,用于对所述分割高倍图像通过自适应阈值分割算法进行标注切割,得到异常细胞标注集;
    异常生成步骤,用于将所述异常细胞标注集映射到所述细胞病理图片上,得到异常细胞标注图。
  15. 如权利要求14所述的计算机可读存储介质,其中,所述对所述细胞病理图片进行预设次数的高斯卷积平滑处理,生成多个细胞病理图片,根据所述细胞病理图片得到图像金字塔,包括:
    步骤A:将所述获取的细胞病理图片执行扩大操作,得到第a组第b层图像,其中,a、b的初始值均为1,并将所述第a组第b层图像利用高斯卷积函数进行高斯卷积的平滑操作,得到第a组第b+1层图像,其中,所述高斯卷积函数为:
    Figure PCTCN2020098969-appb-100007
    其中σ表示平滑因子,G (x,y,σ)表示x与y的卷积,x、y代表图像坐标;
    步骤B:将所述平滑因子σ乘以比例系数k,得到新的平滑因子kσ,通过所述新的平滑因子kσ对所述第a组第b+1层图像利用上述高斯卷积函数进行高斯卷积的平滑操作,得到第a组第b+2层图像,并重复该步骤,直到直到得到第a组第L层图像,其中L为预定义值;
    步骤C:对所述第a组第L层图像进行降采样操作,得到第a+1组第b层图像,对所述第a+1组第b层图像利用上述高斯卷积函数进行高斯卷积的平滑操作,得到第a+1组第b+1层图像,重复所述步骤B,直至得到第a+1组第L层图像;
    步骤D:对所述步骤C得到的结果进行所述步骤C和步骤B的循环操作,直至得到第O组第L层图像,其中O为预定义值;
    步骤E:将从所述第a组第b层图像到所述第O组第L层图像组合起来,生成所述图 像金字塔。
  16. 如权利要求14所述的计算机可读存储介质,其中,所述对所述图像金字塔通过霍夫变换圆检测方法进行检测及拟合,得到所述低倍拟合感兴趣区域,包括:
    将所述图像金字塔进行灰度转化,生成灰度图像;
    对所述灰度图像执行二值化处理,得到轮廓图像;
    从所述轮廓图像中采用随机选取的方法,得到候选区域;
    利用霍夫圆变换检测方法识别出所述候选区域中的感兴趣区域,对所述感兴趣区域进行拟合,得到所述低倍拟合感兴趣区域。
  17. 如权利要求16所述的计算机可读存储介质,其中,所述利用霍夫变换圆检测方法识别出所述感兴趣区域并拟合,得到所述低倍拟合感兴趣区域,包括:
    在所述候选区域的图像空间中通过边缘检测算法对所述候选区域的边缘进行检测,得到n个边缘像素点集;
    将所述n个边缘像素点集,以预定义值r为半径映射到参数空间:
    Figure PCTCN2020098969-appb-100008
    其中,r为预定义值,θ∈[0,2π),x与y表示所述n个边缘像素点集相应的坐标(x,y),(a,b)表示所述参数空间中参考点坐标;
    将所述参数空间中所有坐标点进行统计,对θ进行遍历,使得图像空间上的边缘像素点映射到参数空间为一个圆时,将这个圆作为霍夫圆,通过所述霍夫圆构成所述感兴趣区域;
    将所述感兴趣区域进行拟合,生成低倍拟合感兴趣区域。
  18. 如权利要求14所述的计算机可读存储介质,其中,所述映射是将低分辨图像划分为多个子区域,通过子区域坐标与倍数相乘,将各个所述子区域添加至高分辨率图像中,所述倍数为所述高分辨率图像相对于低分辨率图像的放大倍数。
  19. 如权利要求14所述的计算机可读存储介质,其中,所述对所述分割高倍图像通过自适应阈值分割算法进行标注切割,得到异常细胞标注集,包括:
    将所述分割高倍图像的每一个像素i通过公式:
    Figure PCTCN2020098969-appb-100009
    转换为0-1之间的实数,得到归一化值h;
    利用预定义的伽马校正补偿值将所述归一化值h进行预补偿计算,得到预补常值f;
    将所述预补常值f通过公式f*256-0.5进行反归一化计算,得到所述分割高倍图像的校正图像;
    将所述校正图像的图像灰度级和所述校正图像中坐标的像素点灰度级构成二元组,计算所有二元组的均值和方差,通过所述均值和方差建立二维最大类间方差模型,对所述二维最大类间方差模型通过自适应粒子群集算法进行计算,生成所述校正图像的最佳阈值;
    利用所述最佳阈值对所述校正图像进行分割,生成背景前景分割图像;
    将所述背景前景分割图像进行开运算与闭运算处理,生成异常细胞标注集。
  20. 一种异常细胞自动标注装置,其中,所述装置包括:
    图像金字塔生成模块,用于获取细胞病理图片,对所述细胞病理图片进行预设次数的高斯卷积平滑处理,生成多个细胞病理图片,根据所述细胞病理图片得到图像金字塔;
    检测拟合模块,用于对所述图像金字塔进行检测及拟合,得到低倍拟合感兴趣区域;
    映射分割模块,用于将所述低倍拟合感兴趣区域映射到获取的所述细胞病理图片上并进行图像放大操作,生成拟合高倍图像,并获取所述拟合高倍图像中感兴趣区域的所有图像坐标,再利用所述图像坐标将所述拟合高倍图像中的感兴趣区域分割出来,生成分割高倍图像;
    标注切割模块,用于对所述分割高倍图像通过自适应阈值分割算法进行标注切割,得到异常细胞标注集;
    异常生成模块,用于将所述异常细胞标注集映射到所述细胞病理图片上,得到异常细胞标注图。
PCT/CN2020/098969 2020-04-27 2020-06-29 异常细胞自动标注方法、装置、电子设备及存储介质 WO2021217851A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202010348583.2A CN111652845A (zh) 2020-04-27 2020-04-27 异常细胞自动标注方法、装置、电子设备及存储介质
CN202010348583.2 2020-04-27

Publications (1)

Publication Number Publication Date
WO2021217851A1 true WO2021217851A1 (zh) 2021-11-04

Family

ID=72348336

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2020/098969 WO2021217851A1 (zh) 2020-04-27 2020-06-29 异常细胞自动标注方法、装置、电子设备及存储介质

Country Status (2)

Country Link
CN (1) CN111652845A (zh)
WO (1) WO2021217851A1 (zh)

Cited By (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114092478A (zh) * 2022-01-21 2022-02-25 合肥中科类脑智能技术有限公司 一种异常检测方法
CN114862765A (zh) * 2022-04-13 2022-08-05 湖南科技大学 一种细胞分层图像处理方法
CN115201092A (zh) * 2022-09-08 2022-10-18 珠海圣美生物诊断技术有限公司 细胞扫描图像的获取方法和装置
CN116580397A (zh) * 2023-07-12 2023-08-11 北京大学 病理图像识别方法、装置、设备及存储介质
CN116580041A (zh) * 2023-05-30 2023-08-11 山东第一医科大学附属眼科研究所(山东省眼科研究所、山东第一医科大学附属青岛眼科医院) 基于voronoi图的角膜内皮细胞边界分割方法和装置
CN116703917A (zh) * 2023-08-07 2023-09-05 广州盛安医学检验有限公司 一种女性生殖道细胞病理智能分析***
CN116739949A (zh) * 2023-08-15 2023-09-12 武汉互创联合科技有限公司 一种胚胎图像的卵裂球边缘增强处理方法
CN117078698A (zh) * 2023-08-22 2023-11-17 山东第一医科大学第二附属医院 一种基于深度学习的外周血管影像辅助分割方法及***
CN117129029A (zh) * 2023-10-26 2023-11-28 昂坤视觉(北京)科技有限公司 一种芯片检测方法及***

Families Citing this family (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112102351A (zh) * 2020-10-14 2020-12-18 平安科技(深圳)有限公司 医学图像分析方法、装置、电子设备及可读存储介质
CN113237881B (zh) * 2021-05-10 2023-09-12 博奥生物集团有限公司 一种特定细胞的检测方法、装置和病理切片检测***
CN114549603B (zh) * 2022-02-28 2022-10-18 赛维森(广州)医疗科技服务有限公司 细胞病理图像的标注坐标转换方法、***、设备及介质
CN116363127B (zh) * 2023-05-31 2023-07-28 临沂中科芯华新材料科技有限公司 基于图像处理的全降解塑料制品质量检测方法

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US10169685B2 (en) * 2014-07-07 2019-01-01 The Regents Of The University Of California Automatic segmentation and quantitative parameterization of brain tumors in MRI
CN110363747A (zh) * 2019-06-14 2019-10-22 平安科技(深圳)有限公司 智能化异常细胞判断方法、装置及计算机可读存储介质
CN110415212A (zh) * 2019-06-18 2019-11-05 平安科技(深圳)有限公司 异常细胞检测方法、装置及计算机可读存储介质
CN110570352A (zh) * 2019-08-26 2019-12-13 腾讯科技(深圳)有限公司 图像标注方法、装置、***及细胞标注方法

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US10169685B2 (en) * 2014-07-07 2019-01-01 The Regents Of The University Of California Automatic segmentation and quantitative parameterization of brain tumors in MRI
CN110363747A (zh) * 2019-06-14 2019-10-22 平安科技(深圳)有限公司 智能化异常细胞判断方法、装置及计算机可读存储介质
CN110415212A (zh) * 2019-06-18 2019-11-05 平安科技(深圳)有限公司 异常细胞检测方法、装置及计算机可读存储介质
CN110570352A (zh) * 2019-08-26 2019-12-13 腾讯科技(深圳)有限公司 图像标注方法、装置、***及细胞标注方法

Cited By (15)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114092478A (zh) * 2022-01-21 2022-02-25 合肥中科类脑智能技术有限公司 一种异常检测方法
CN114862765A (zh) * 2022-04-13 2022-08-05 湖南科技大学 一种细胞分层图像处理方法
CN114862765B (zh) * 2022-04-13 2024-04-16 湖南科技大学 一种细胞分层图像处理方法
CN115201092A (zh) * 2022-09-08 2022-10-18 珠海圣美生物诊断技术有限公司 细胞扫描图像的获取方法和装置
CN116580041A (zh) * 2023-05-30 2023-08-11 山东第一医科大学附属眼科研究所(山东省眼科研究所、山东第一医科大学附属青岛眼科医院) 基于voronoi图的角膜内皮细胞边界分割方法和装置
CN116580397A (zh) * 2023-07-12 2023-08-11 北京大学 病理图像识别方法、装置、设备及存储介质
CN116580397B (zh) * 2023-07-12 2023-11-10 北京大学 病理图像识别方法、装置、设备及存储介质
CN116703917B (zh) * 2023-08-07 2024-01-26 广州盛安医学检验有限公司 一种女性生殖道细胞病理智能分析***
CN116703917A (zh) * 2023-08-07 2023-09-05 广州盛安医学检验有限公司 一种女性生殖道细胞病理智能分析***
CN116739949A (zh) * 2023-08-15 2023-09-12 武汉互创联合科技有限公司 一种胚胎图像的卵裂球边缘增强处理方法
CN116739949B (zh) * 2023-08-15 2023-11-03 武汉互创联合科技有限公司 一种胚胎图像的卵裂球边缘增强处理方法
CN117078698A (zh) * 2023-08-22 2023-11-17 山东第一医科大学第二附属医院 一种基于深度学习的外周血管影像辅助分割方法及***
CN117078698B (zh) * 2023-08-22 2024-03-05 山东第一医科大学第二附属医院 一种基于深度学习的外周血管影像辅助分割方法及***
CN117129029A (zh) * 2023-10-26 2023-11-28 昂坤视觉(北京)科技有限公司 一种芯片检测方法及***
CN117129029B (zh) * 2023-10-26 2024-01-05 昂坤视觉(北京)科技有限公司 一种芯片检测方法及***

Also Published As

Publication number Publication date
CN111652845A (zh) 2020-09-11

Similar Documents

Publication Publication Date Title
WO2021217851A1 (zh) 异常细胞自动标注方法、装置、电子设备及存储介质
WO2021189912A1 (zh) 图像中目标物的检测方法、装置、电子设备及存储介质
WO2021189901A1 (zh) 图像分割方法、装置、电子设备及计算机可读存储介质
WO2020253508A1 (zh) 异常细胞检测方法、装置及计算机可读存储介质
WO2022121156A1 (zh) 图像中目标物检测方法、装置、电子设备及可读存储介质
CN110059697B (zh) 一种基于深度学习的肺结节自动分割方法
WO2021189913A1 (zh) 图像中目标物的分割方法、装置、电子设备及存储介质
WO2021151338A1 (zh) 医学影像图片分析方法、装置、电子设备及可读存储介质
WO2022213465A1 (zh) 基于神经网络的图像识别方法、装置、电子设备及介质
WO2022247005A1 (zh) 图像中目标物识别方法、装置、电子设备及存储介质
WO2021151272A1 (zh) 细胞图像分割方法、装置、电子设备及可读存储介质
CN110705583A (zh) 细胞检测模型训练方法、装置、计算机设备及存储介质
WO2022105179A1 (zh) 生物特征图像识别方法、装置、电子设备及可读存储介质
CN112699775A (zh) 基于深度学习的证件识别方法、装置、设备及存储介质
WO2021189855A1 (zh) 基于ct序列的图像识别方法、装置、电子设备及介质
WO2020248848A1 (zh) 智能化异常细胞判断方法、装置及计算机可读存储介质
WO2022141858A1 (zh) 行人检测方法、装置、电子设备及存储介质
CN112507934A (zh) 活体检测方法、装置、电子设备及存储介质
WO2021189856A1 (zh) 证件校验方法、装置、电子设备及介质
WO2021151277A1 (zh) 目标物损伤程度判定方法、装置、电子设备及存储介质
CN111476760B (zh) 医学图像的生成方法、装置、电子设备及介质
WO2021168703A1 (zh) 字符处理及字符识别方法、存储介质和终端设备
CN116433704A (zh) 一种基于中心点的细胞核分割方法及相关设备
CN115984588A (zh) 图像背景相似度分析方法、装置、电子设备及存储介质
CN109191452A (zh) 一种基于主动学习的腹腔ct图像腹膜转移自动标记方法

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 20932941

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 20932941

Country of ref document: EP

Kind code of ref document: A1