Nucleus segmentation method and device based on contour characteristics, equipment and storage medium

Abstract

This application provides a method, apparatus, device, and storage medium for cell nucleus segmentation based on contour characteristics. The method includes: acquiring an image to be segmented; preprocessing the image to be segmented to obtain a preprocessed image; using a convolutional filter on the preprocessed image to determine a denoised image; determining effective cell nucleus contours based on the denoised image; and expanding the effective cell nucleus contours based on adjacent pixels to obtain segmented cell nuclei. This scheme can achieve high accuracy on real cervical cell images (i.e., the BSMMU dataset) while maintaining a fairly high recall rate.

Description

Technical Field

[0001] This invention belongs to the field of cell segmentation technology, and specifically relates to a method, device, electronic device and storage medium for cell nucleus segmentation based on contour characteristics. Background Technology

[0002] The first stage of the automated computer-assisted cervical cancer screening process is the segmentation of cervical cells. This requires first detecting and identifying the cervical cell nuclei, and then using these nuclei to segment the cytoplasm into overlapping and non-overlapping segments. Because each segmented nucleus represents a cell, the result of this step (i.e., its effectiveness) directly affects the final cytoplasmic segmentation result.

[0003] To address this problem, Jung et al. proposed a method based on a Bayesian classifier. They used distance transformation and the EM (Expectation-Maximization) algorithm, and then used Bayesian classification to segment overlapping kernels. Keenan et al. applied grayscale thresholding, contour tracking algorithm and morphological operations to segment and classify cell kernels. Plissiti et al. proposed a method based on a physical deformation model. Ushizima et al. used an improved version of the local thresholding method proposed by Phansalkar et al. to segment cell kernels. Lu et al. [6] segmented cell kernels by finding the maximum stable extreme value region (MSER). Lee et al. segmented cell kernels by performing local thresholding and removing outliers based on features such as average intensity, roundness and size. Saha et al. used fuzzy c-means clustering constrained by a circular shape function (CSF) to detect cell kernels. Subsequently, Saha et al. also proposed a neck cell kernel segmentation method based on pairwise region contrast and image gradient contour evaluation that fuses oversegmented SLIC superpixel regions. Their recent work proposed segmenting cell kernels by using superpixels generated by pairwise region contrast and gradient boundary merging statistical region merging (SRM) algorithm. Braz et al. used deep learning convolutional networks to detect and segment cell nuclei from Pap smear images. Tareef et al. proposed a novel method based on local features and shape-guided deformation, and their recent paper introduces a method based on multi-channel fast watersheds to segment cell nuclei and cytoplasm in cervical cytology images.

[0004] All the studies on cell nucleus segmentation in cervical cytology images mentioned above performed quite well on the ISBI (International Symposium on Biomedical Imaging) dataset. However, since most studies were conducted based on the ISBI dataset challenge, the performance of the corresponding methods on other independent (real-world) datasets remains to be explored. Preliminary experiments show that while the methods mentioned above perform well on the ISBI dataset, they perform below standard on datasets containing real cervical cytology images (BSMMU (Bangabandhu Sheikh Mujib Medical University) dataset). Summary of the Invention

[0005] The purpose of the embodiments in this specification is to provide a method, apparatus, electronic device, and storage medium for cell nucleus segmentation based on contour characteristics.

[0006] To solve the above-mentioned technical problems, the embodiments of this application are implemented in the following ways:

[0007] In a first aspect, this application provides a cell nucleus segmentation method based on contour characteristics, the method comprising:

[0008] Obtain the image to be segmented;

[0009] Preprocess the image to be segmented to obtain the preprocessed image;

[0010] The preprocessed image is then filtered using a convolutional filter to determine the denoised image;

[0011] Determine the effective cell nucleus contours based on the denoised image;

[0012] Based on the adjacent pixels of the effective cell nucleus outline, the effective cell nucleus outline is expanded to obtain the segmented cell nucleus.

[0013] In one embodiment, the preprocessed image is used with a convolutional filter to determine a denoised image, including:

[0014] The preprocessed image is convolved with convolution kernels in different directions in the convolution filter to obtain the corresponding convolution image;

[0015] Convolutional images in different directions are superimposed to obtain a response image;

[0016] The preprocessed image is subtracted from the response image to obtain the denoised image.

[0017] In one embodiment, the convolutional kernels in the convolutional filter include eight-directional convolutional kernels, namely left, right, lower left, upper left, lower, upper, lower, lower right, and upper right.

[0018] In one embodiment, determining the effective cell nucleus contour based on the denoised image includes:

[0019] Based on the contour detection method, extract all detected contours from the denoised image;

[0020] Delete contours whose strength is less than the minimum strength.

[0021] Delete contours whose dimensions are smaller than the minimum dimension or larger than the maximum dimension.

[0022] Delete contours whose axial length ratio is less than the minimum axial length ratio.

[0023] The remaining contours are used as valid cell nucleus contours.

[0024] In one embodiment, the effective cell nucleus contour is expanded based on adjacent pixels of the effective cell nucleus contour to obtain the segmented cell nucleus, including:

[0025] Determine the average intensity of all pixels in the effective cell nucleus outline;

[0026] Determine the strength range based on the average strength;

[0027] If the intensity values ​​of adjacent pixels are within the intensity range, the adjacent pixels are determined to be valid pixels;

[0028] The effective cell kernel contour is expanded by incorporating the effective pixels as part of the effective cell kernel contour.

[0029] The expansion process continues until the preset conditions are met, at which point the expansion ends and the segmented cell nuclei are obtained.

[0030] In one embodiment, the preset conditions include at least one of the following:

[0031] All valid pixels in the direct neighborhood of the effective cell nucleus outline boundary are identified.

[0032] The total size of the expanded effective nuclear outline is greater than the preset maximum nuclear size;

[0033] The robustness of the expanded effective cell nucleus outline is less than the preset robustness.

[0034] In one embodiment, the image to be segmented is preprocessed to obtain a preprocessed image, including:

[0035] If the image to be segmented is an RGB image, convert it to a grayscale image;

[0036] A Gaussian filter is used to smooth the grayscale image, resulting in a smooth image.

[0037] An adaptive thresholding algorithm is used to threshold the smoothed image to obtain the preprocessed image.

[0038] Secondly, this application provides a cell nucleus segmentation device based on contour characteristics, the device comprising:

[0039] The acquisition module is used to acquire the image to be segmented;

[0040] The preprocessing module is used to preprocess the image to be segmented to obtain the preprocessed image;

[0041] The denoising module is used to determine the denoised image by applying a convolutional filter to the preprocessed image.

[0042] The determination module is used to determine the effective cell nucleus contours based on the denoised image;

[0043] An extension module is used to extend the effective cell nucleus contour based on the adjacent pixels of the effective cell nucleus contour to obtain the segmented cell nucleus.

[0044] Thirdly, this application provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the cell nucleus segmentation method based on contour characteristics as described in the first aspect.

[0045] Fourthly, this application provides a readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the contour-based cell nucleus segmentation method as described in the first aspect.

[0046] As can be seen from the technical solutions provided in the embodiments of this specification above, the solution is:

[0047] It achieves very high precision and recall on the standard ISBI dataset. Furthermore, it balances precision and recall on real cervical cytology images, achieving high precision on real cervical cell images (i.e., the BSMMU dataset) while maintaining a reasonably high recall. Attached Figure Description

[0048] To more clearly illustrate the technical solutions in the embodiments or prior art of this specification, the drawings used in the description of the embodiments or prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this specification. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0049] Figure 1 A schematic flowchart of the contour-based cell nucleus segmentation method provided in this application;

[0050] Figure 2 Schematic diagrams of convolution kernels in different directions provided in this application;

[0051] Figure 3 The image shows the recognition effect achieved using the method described in this application.

[0052] Figure 4 A schematic diagram of the structure of the cell nucleus segmentation device based on contour characteristics provided in this application;

[0053] Figure 5 A schematic diagram of the structure of the electronic device provided in this application. Detailed Implementation

[0054] To enable those skilled in the art to better understand the technical solutions in this specification, the technical solutions in the embodiments of this specification will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this specification, and not all embodiments. Based on the embodiments in this specification, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of this specification.

[0055] In the following description, specific details such as particular system architectures and techniques are set forth for illustrative purposes and not for limitation, in order to provide a thorough understanding of the embodiments of this application. However, those skilled in the art will understand that this application may also be implemented in other embodiments without these specific details. In other instances, detailed descriptions of well-known systems, apparatuses, circuits, and methods have been omitted so as not to obscure the description of this application with unnecessary detail.

[0056] Various modifications and variations can be made to the specific embodiments described in this application without departing from the scope or spirit of this application, as will be apparent to those skilled in the art. Other embodiments derived from this application will be obvious to those skilled in the art. This application specification and embodiments are merely exemplary.

[0057] The terms “include,” “including,” “have,” “contain,” etc., used in this article are all open-ended terms, meaning that they include but are not limited to.

[0058] Unless otherwise specified, "parts" in this application refers to parts by weight.

[0059] The present invention will be further described in detail below with reference to the accompanying drawings and embodiments.

[0060] Reference Figure 1 It illustrates a flowchart of the cell nucleus segmentation method based on contour characteristics provided in the embodiments of this application.

[0061] like Figure 1 As shown, a cell nucleus segmentation method based on contour characteristics may include:

[0062] S110. Obtain the image to be segmented.

[0063] Specifically, the image to be segmented can be a skin cancer cell image from the ISBI dataset, a cervical cancer cell image from the BSMMU dataset, or other cell images.

[0064] S120. Preprocess the image to be segmented to obtain a preprocessed image, including:

[0065] If the image to be segmented is an RGB image, convert it to a grayscale image;

[0066] A Gaussian filter is used to smooth the grayscale image, resulting in a smooth image.

[0067] An adaptive thresholding algorithm is used to threshold the smoothed image to obtain the preprocessed image.

[0068] Specifically, the cell images in the ISBI dataset are already grayscale images, so no grayscale conversion is needed. However, the cell images in the BSMMU dataset are RGB images. Since the green channel of an RGB image contains the most information, the best way to convert them to grayscale is to use only the green channel. Therefore, the RGB images in the BSMMU dataset are converted to grayscale images simply by obtaining the intensity values ​​of the green channel.

[0069] Then, a Gaussian filter is used to smooth the grayscale image, resulting in a smoothed image.

[0070] Because the cell nucleus is the darkest visible region of its cytoplasm, an adaptive thresholding algorithm can be used to divide a smoothed image into portions where pixel grayscale values ​​are greater than or equal to a threshold and portions where pixel grayscale values ​​are less than a threshold. Understandably, OpenCV's built-in adaptive thresholding function can be used.

[0071] Since different cells and different images may correspond to different thresholds, the threshold can be adjusted by adjusting the window size wsize and the offset "C".

[0072] For datasets with magnified atomic nuclei, a larger window size (wsize) is needed, while for datasets with smaller nuclei, a smaller window size (wsize) is needed. Because the nuclei are magnified more in the ISBI dataset, a larger window size (wsize) is required. Conversely, the nuclei are shrunk more in the BSMMU dataset, thus requiring a smaller window size (wsize).

[0073] The window size, wsize, can be automatically adjusted based on the size of the identified cell nuclei. If the identified cell nuclei are generally larger than the initial value, wsize will increase; conversely, it will decrease.

[0074] The offset value "C" is used to subtract from the average or weighted average of the gray values ​​calculated within the window. For datasets where the contrast between the cell nucleus and the rest of the image increases progressively, a smaller offset value "C" is needed.

[0075] Due to overlapping cells, surface noise, and artifacts, thresholded images still contain varying degrees of unwanted regions, especially in the BSMMU dataset (real images of the cervix).

[0076] S130. The preprocessed image is used to determine the denoised image using a convolutional filter, which may include:

[0077] The preprocessed image is convolved with convolution kernels in different directions in the convolution filter to obtain the corresponding convolution image;

[0078] Convolutional images in different directions are superimposed to obtain a response image;

[0079] The preprocessed image is subtracted from the response image to obtain the denoised image.

[0080] Among them, such as Figure 2 As shown, the convolutional filter contains convolutional kernels in eight directions. Figure 2 From left to right and from top to bottom, the textures are: left side texture, right side texture, bottom left side texture, top left side texture, bottom side texture, top side texture, bottom right side texture, and top right side texture.

[0081] Specifically, using convolutional filters can reduce the number of unwanted regions.

[0082] A threshold can be set to determine whether a pixel belongs to an edge. If a pixel is greater than or equal to the threshold, it is considered an edge; if it is less than the threshold, it is not. The threshold can be set according to actual needs. The response image contains various edges detected in the image.

[0083] Since the convolutional filter responds weakly to uniform, dense, convex regions than to irregularly shaped, non-convex regions, subtracting the response image from the preprocessed image can remove much of the unwanted noise generated by the irregular shape.

[0084] However, this step reduces the size of the region containing the actual cell nucleus, which needs to be corrected in subsequent steps.

[0085] S140. Determining the effective cell nucleus contour based on the denoised image may include:

[0086] Based on the contour detection method, extract all detected contours from the denoised image;

[0087] Delete contours whose strength is less than the minimum strength.

[0088] Delete any outline whose size is smaller than the minimum size or larger than the maximum size.

[0089] Delete contours whose axial length ratio is less than the minimum axial length ratio.

[0090] The remaining contours are used as valid cell nucleus contours.

[0091] Specifically, OpenCV's built-in contour detection function can be used to extract all detected contours from the denoised image and examine their properties one by one. Contour properties include the contour's size, robustness, and aspect ratio. Robustness is the ratio of the contour's area to its convex hull area, and the aspect ratio is the ratio of the contour's major axis to its minor axis. The contour area can be counted in pixels. Since atomic nuclei are uniformly shaped, solid convex regions, they have relatively high robustness, with most having a size greater than 0.9.

[0092] Any contour with a minimum robustness will be removed from the denoised image. The minimum robustness can be set according to specific needs. Typically, a minimum robustness of 0.8 is recommended to ensure effective identification and a very low false positive rate. While a lower minimum robustness may identify more cell nuclei, the probability of false positives will be higher.

[0093] Outlines with dimensions smaller than the minimum or larger than the maximum size will also be deleted; that is, outlines that are too small or too large. The minimum and maximum sizes can be set according to actual needs, and depend on the scaling level of the cell kernels in the dataset.

[0094] It will also delete contours with an aspect ratio less than the minimum aspect ratio, because contours with a small aspect ratio are too long and not suitable for a cell nucleus. The minimum aspect ratio can be set according to actual needs.

[0095] S150. Based on the adjacent pixels of the effective cell nucleus contour, expand the effective cell nucleus contour to obtain the segmented cell nucleus, including:

[0096] Determine the average intensity of all pixels in the effective cell nucleus outline;

[0097] Determine the strength range based on the average strength;

[0098] If the intensity values ​​of adjacent pixels are within the intensity range, the adjacent pixels are determined to be valid pixels;

[0099] The effective cell kernel contour is expanded by incorporating the effective pixels as part of the effective cell kernel contour.

[0100] The expansion process continues until the preset conditions are met, at which point the expansion ends and the segmented cell nuclei are obtained.

[0101] The preset conditions include at least one of the following:

[0102] All valid pixels in the direct neighborhood of the effective cell nucleus outline boundary are identified.

[0103] The total size of the expanded effective nuclear outline is greater than the preset maximum nuclear size;

[0104] The robustness of the expanded effective cell nucleus outline is less than the preset robustness.

[0105] Specifically, expanding the effective nuclear outline is mainly used to restore the size of the shrunken nuclear region in step S130.

[0106] This step is an iterative algorithm that checks each neighboring pixel of a valid cell kernel contour to determine if it belongs to that contour. The validity of a pixel is measured by its intensity value.

[0107] For a given effective cell nucleus contour, first calculate the average intensity of all pixels already belonging to that region (meaning the average intensity within the contour range). If a neighboring pixel of the effective cell nucleus contour has an intensity value within a certain range of the average intensity in that region (i.e., an intensity range), then that pixel is determined to be an effective pixel and becomes part of the effective cell nucleus contour, thereby expanding the effective cell nucleus contour.

[0108] The intensity range is the difference between the average intensity of the effective cell nucleus contour and the intensity of the adjacent pixels considered. This intensity range can be set according to actual needs. For datasets with low contrast between the cell nucleus and the outer cytoplasm, the intensity range value should be smaller; otherwise, it should be larger.

[0109] All valid pixels among all adjacent pixels in the direct neighborhood of the effective cell nucleus outline boundary are found; that is, no valid pixels can be found in the direct neighborhood of the effective cell nucleus outline boundary.

[0110] The preset solidity can be set according to actual needs. It is generally set to be lower than the usual solid value of effective cell nuclei (0.8), for example, set to 0.75.

[0111] The contour-based cell nucleus segmentation method provided in this application can balance precision and recall in real cervical cytology images. The method achieves very high precision and recall on the standard ISBI dataset. Subsequently, the method was tested on the BSMMU dataset. Although the results on real cervical cell images (i.e., the BSMMU dataset) are not as outstanding, they are still better than other algorithms, achieving high precision while maintaining a fairly high recall.

[0112] The method presented in this application can depict the boundaries of cell nuclei relatively well. While the method by Phoulady et al. can detect the presence of cell nuclei well, it sometimes fails to accurately depict the boundaries of cell nuclei in PAP smear images. This is due to over-dilation at the end of the algorithm, resulting in a nuclear boundary that exceeds the actual boundary. The method presented in this application is more conservative in this regard. Instead of general dilation in the nuclear recovery stage, this application examines boundary pixels and expands the nuclear boundary based on their average intensity, followed by further checks through a robustness check. This allows the method presented in this application to depict cell nucleus boundaries more accurately at the pixel level.

[0113] The method presented in this application is more robust to outliers. Real-world Papanicolaou smear images, unlike those in the ISBI dataset, contain many outliers (artifacts, surface cells). These outliers often appear very similar to actual atomic nuclei, but in reality, they can lead to incorrect identification results. Compared to the algorithm by Phulady et al., the method presented in this application is more robust to these types of outliers. Their algorithm only uses minimum size and stability to filter out noise and outliers, which is insufficient in filtering out all outliers due to their low accuracy. On the other hand, the method presented in this application has a separate noise removal stage, followed by a more rigorous check of contour characteristics (including minimum size, robustness, etc.). Therefore, compared to the algorithm by Phulady et al., the final output image of the method presented in this application contains far fewer outliers.

[0114] Experimental verification

[0115] All experiments were conducted on a machine running Windows 10 Professional Edition (64-bit) equipped with an Intel Core i9 CPU @ 3.6GHz (8 cores, 16 threads) and 64GB of RAM. Python (version 3.7.9) and the OpenCV library (version 4.5.2) for Python were used in the experiments.

[0116] Verification has shown that the proposed method can effectively segment cell nuclei on both the ISBI and BSMMU datasets. The recognition results are as follows: Figure 3 As shown:

[0117] Figure 3 In the image, a and c are images from the ISBI dataset, b and d are the corresponding segmentation results, e and g are images from the BSMMU dataset, and f and h are the corresponding segmentation results. All of them can effectively identify the cell nucleus.

[0118] Statistical calculations show that the method described in this application achieves an accuracy of 97.8% on the ISBI dataset, compared to 96.1% for the previous algorithm, representing a 1% improvement in accuracy. On the BSMMU dataset, the method described in this application achieves an accuracy of 77%, compared to 64.3% for the conventional algorithm, representing a 13% improvement in accuracy.

[0119] Reference Figure 4 The diagram illustrates a schematic representation of a contour-based cell nucleus segmentation apparatus according to an embodiment of this application.

[0120] like Figure 4 As shown, the contour-based cell nucleus segmentation device 400 may include:

[0121] The acquisition module 410 is used to acquire the image to be segmented;

[0122] Preprocessing module 420 is used to preprocess the image to be segmented to obtain a preprocessed image;

[0123] The denoising module 430 is used to determine the denoised image by applying a convolutional filter to the preprocessed image;

[0124] The determination module 440 is used to determine the effective cell nucleus contours based on the denoised image;

[0125] The extension module 450 is used to extend the effective cell nucleus contour based on the adjacent pixels of the effective cell nucleus contour to obtain the segmented cell nucleus.

[0126] Optionally, the noise reduction module 430 is also used for:

[0127] The preprocessed image is convolved with convolution kernels in different directions in the convolution filter to obtain the corresponding convolution image;

[0128] Convolutional images in different directions are superimposed to obtain a response image;

[0129] The preprocessed image is subtracted from the response image to obtain the denoised image.

[0130] Optionally, the convolutional kernels in the convolutional filter include convolutional kernels in eight directions, namely left, right, lower left, upper left, lower, upper, lower, lower right, and upper right.

[0131] Optionally, the determining module 440 is also used for:

[0132] Based on the contour detection method, extract all detected contours from the denoised image;

[0133] Delete contours whose strength is less than the minimum strength.

[0134] Delete any outline whose size is smaller than the minimum size or larger than the maximum size.

[0135] Delete contours whose axial length ratio is less than the minimum axial length ratio.

[0136] The remaining contours are used as valid cell nucleus contours.

[0137] Optionally, the expansion module 450 is also used for:

[0138] Determine the average intensity of all pixels in the effective cell nucleus outline;

[0139] Determine the strength range based on the average strength;

[0140] If the intensity values ​​of adjacent pixels are within the intensity range, the adjacent pixels are determined to be valid pixels;

[0141] The effective cell kernel contour is expanded by incorporating the effective pixels as part of the effective cell kernel contour.

[0142] The expansion process continues until the preset conditions are met, at which point the expansion ends and the segmented cell nuclei are obtained.

[0143] Optionally, the preset conditions include at least one of the following:

[0144] All valid pixels in the direct neighborhood of the effective cell nucleus outline boundary are identified;

[0145] The total size of the expanded effective nuclear outline is greater than the preset maximum nuclear size;

[0146] The robustness of the expanded effective nuclear outline is less than the preset robustness.

[0147] Optionally, the preprocessing module 420 is also used for:

[0148] If the image to be segmented is an RGB image, convert it to a grayscale image;

[0149] A Gaussian filter is used to smooth the grayscale image, resulting in a smooth image.

[0150] An adaptive thresholding algorithm is used to threshold the smoothed image to obtain the preprocessed image.

[0151] This embodiment provides a cell nucleus segmentation device based on contour characteristics, which can perform the above-described method. Its implementation principle and technical effect are similar, and will not be repeated here.

[0152] Figure 5 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention. Figure 5 The diagram shows a structural schematic of an electronic device 300 suitable for implementing embodiments of this application.

[0153] like Figure 5 As shown, the electronic device 300 includes a central processing unit (CPU) 301, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 302 or a program loaded from a storage section 308 into a random access memory (RAM) 303. The RAM 303 also stores various programs and data required for the operation of the device 300. The CPU 301, ROM 302, and RAM 303 are interconnected via a bus 304. An input / output (I / O) interface 305 is also connected to the bus 304.

[0154] The following components are connected to I / O interface 305: an input section 306 including a keyboard, mouse, etc.; an output section 307 including a cathode ray tube (CRT), liquid crystal display (LCD), etc., and speakers, etc.; a storage section 308 including a hard disk, etc.; and a communication section 309 including a network interface card such as a LAN card, modem, etc. The communication section 309 performs communication processing via a network such as the Internet. Drive 310 is also connected to I / O interface 306 as needed. Removable media 311, such as a disk, optical disk, magneto-optical disk, semiconductor memory, etc., are installed on drive 310 as needed so that computer programs read from them can be installed into storage section 308 as needed.

[0155] In particular, according to embodiments of this disclosure, the above references Figure 1The described process can be implemented as a computer software program. For example, embodiments of this disclosure include a computer program product comprising a computer program tangibly embodied on a machine-readable medium, the computer program containing program code for performing the above-described contour-based cell nucleus segmentation method. In such embodiments, the computer program can be downloaded and installed from a network via communication section 309, and / or installed from removable medium 311.

[0156] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of the present invention. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing the specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0157] The units or modules described in the embodiments of this application can be implemented in software or hardware. The described units or modules can also be located in a processor. The names of these units or modules do not necessarily constitute a limitation on the unit or module itself.

[0158] The systems, devices, modules, or units described in the above embodiments can be implemented by computer chips or entities, or by products with certain functions. A typical implementation device is a computer. Specifically, a computer can be, for example, a personal computer, a laptop computer, a mobile phone, a smartphone, a personal digital assistant, a media player, a navigation device, an email device, a game console, a tablet computer, a wearable device, or any combination of these devices.

[0159] In another aspect, this application also provides a storage medium, which may be the storage medium included in the aforementioned apparatus in the above embodiments; or it may be a standalone storage medium not assembled into the device. The storage medium stores one or more programs, which are used by one or more processors to execute the contour-based cell nucleus segmentation method described in this application.

[0160] Storage media, including both permanent and non-permanent, removable and non-removable media, can be used to store information by any method or technology. Information can be computer-readable instructions, data structures, program modules, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic magnetic disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include transient computer-readable media, such as modulated data signals and carrier waves.

[0161] It should be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.

[0162] The various embodiments in this specification are described in a progressive manner. Similar or identical parts between embodiments can be referred to interchangeably. Each embodiment focuses on describing the differences from other embodiments. In particular, the system embodiments are basically similar to the method embodiments, so the description is relatively simple; relevant parts can be referred to the descriptions in the method embodiments.

Claims

1. A cell nucleus segmentation method based on contour characteristics, characterized in that, The method includes: Obtain the image to be segmented; The image to be segmented is preprocessed to obtain a preprocessed image; The preprocessed image is convolved using convolution kernels in eight directions of the convolution filter to obtain the corresponding convolution image. The convolution images in the eight directions are superimposed to obtain the response image. The preprocessed image is subtracted from the response image to determine the denoised image. The eight directions are left, right, lower left, upper left, lower, upper, lower, lower right, and upper right. Based on the denoised image, the effective cell nucleus contours are determined; Expanding the effective cell nucleus contour based on neighboring pixels to obtain segmented cell nuclei includes: determining the average intensity of all pixels in the effective cell nucleus contour; determining an intensity range based on the average intensity; identifying neighboring pixels as effective pixels if their intensity values ​​fall within the intensity range; expanding the effective cell nucleus contour as part of the effective cell nucleus contour; and ending the expansion until a preset condition is met to obtain segmented cell nuclei. The preset condition includes at least one of the following: all effective pixels in the direct neighborhood of the effective cell nucleus contour boundary are found; the total size of the expanded effective cell nucleus contour is greater than a preset maximum nucleus size; and the robustness of the expanded effective cell nucleus contour is less than a preset robustness.

2. The method according to claim 1, characterized in that, Determining the effective cell nucleus contour based on the denoised image includes: According to the contour detection method, all detected contours are extracted from the denoised image; Delete the contours whose strength is less than the minimum strength. Delete the contours whose dimensions are smaller than the minimum dimension or larger than the maximum dimension. Delete the contours whose axial length ratio is less than the minimum axial length ratio. The remaining contours are used as valid cell nucleus contours.

3. The method according to any one of claims 1, characterized in that, The preprocessing of the image to be segmented to obtain a preprocessed image includes: If the image to be segmented is an RGB image, convert the image to be segmented into a grayscale image; A Gaussian filter is used to smooth the grayscale image to obtain a smoothed image. The smoothed image is thresholded using an adaptive thresholding algorithm to obtain the preprocessed image.

4. A cell nucleus segmentation device based on contour characteristics, characterized in that, The device includes: The acquisition module is used to acquire the image to be segmented; The preprocessing module is used to preprocess the image to be segmented to obtain a preprocessed image; The denoising module is used to convolve the preprocessed image using convolution kernels in eight directions of the convolution filter to obtain the corresponding convolution image. The convolution images in the eight directions are superimposed to obtain the response image. The preprocessed image is subtracted from the response image to determine the denoised image. The eight directions are left, right, lower left, upper left, lower, upper, lower, lower right, and upper right. A determination module is used to determine the effective cell nucleus contours based on the denoised image; An extension module is used to extend the effective cell nucleus contour based on the adjacent pixels of the effective cell nucleus contour to obtain a segmented cell nucleus. This includes: determining the average intensity of all pixels in the effective cell nucleus contour; determining an intensity range based on the average intensity; identifying adjacent pixels as effective pixels if their intensity values ​​fall within the intensity range; extending the effective cell nucleus contour as part of the effective cell nucleus contour; and ending the extension process until a preset condition is met, thus obtaining a segmented cell nucleus. The preset condition includes at least one of the following: all effective pixels in the direct neighborhood of the effective cell nucleus contour boundary are found; the total size of the extended effective cell nucleus contour is greater than a preset maximum nucleus size; and the robustness of the extended effective cell nucleus contour is less than a preset robustness.

5. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the cell nucleus segmentation method based on contour characteristics as described in any one of claims 1-3.

6. A readable storage medium having a computer program stored thereon, characterized in that, When executed by a processor, the program implements the contour-based cell nucleus segmentation method as described in any one of claims 1-3.

Images