Color cell image segmentation method and device and electronic equipment

Abstract

The invention provides a color cell image segmentation method and device and electronic equipment, and the method comprises the steps: carrying out the segmentation of a color cell original image through color deconvolution and a deep learning method, and obtaining a first cell nucleus segmentation image, the color cell original image being Hamp; an E cell image; classifying pixel points in the color cell original image by adopting a clustering algorithm to obtain a classification result corresponding to each pixel point; determining a second cell nucleus segmentation image according to the classification result; fusing the brightness values at the same pixel position in the first cell nucleus segmentation image and the second cell nucleus segmentation image to obtain a fused brightness value corresponding to a pixel at each pixel position; and generating a target cell nucleus segmentation image according to the fusion brightness value corresponding to the pixel at each pixel position. According to the method, cell nucleus segmentation is carried out through a method of combining color deconvolution, deep learning and a clustering algorithm, the false positive phenomenon in the segmentation result is improved, and the segmentation precision is improved.

Description

Technical Field

[0001] The present application relates to the technical field of image segmentation, and in particular to a method, device and electronic equipment for color cell image segmentation. Background Art

[0002] Hematoxylin and eosin (H&E) stains / H&E stains) is the most widely used staining technique in histopathology. Compared with fluorescent cell images (black background, glowing foreground), which are images in which the foreground (nucleus) glows but the background (non-nucleus) does not (black background, glowing foreground), the H&E cell images obtained have a much more complex background. The background of H&E cell images often contains blood clots and granular impurities. Moreover, the background of H&E cell images also has more complex fine cytoplasm patterns, which often interfere with the H&E image segmentation process and are easily "misjudged" as foreground (nucleus) by the image algorithm process.

[0003] The color deconvolution algorithm is an algorithm that can improve the H&E cell image segmentation effect. Its principle is essentially to transform the three basis vectors (R, G, B) in the color space corresponding to each pixel on the color image with a 3×3 matrix with a standard value, so that these three-dimensional vectors are in the new basis vector (R ′ , G ′ , B ′ ) direction, where the pattern of the original cell nucleus region (hematoxylin staining area, blue-purple) is reflected in the new B channel, namely B ′ The background (e.g., eosin-stained areas of the cytoplasm) is only reflected in the new R channel (R ′ ) or the new G channel (G ′ ), even the impurity pattern of blood clots is hardly reflected in B ′ This brings obvious convenience to the subsequent segmentation of the cell nucleus and better "filters out" many patterns except the hematoxylin-stained area.

[0004] Although the color deconvolution method facilitates the subsequent cell nucleus image segmentation, it still cannot fully guarantee the perfect "separation" between the cell nucleus and other background parts. The resulting cell nucleus layer still has some "remaining" cytoplasm, and false positives can still be found in the resulting positive selection, resulting in inaccurate final segmentation results. Summary of the invention

[0005] In view of this, the purpose of the present application is to at least provide a color cell image segmentation method, device and electronic device, which performs cell nucleus segmentation by combining color deconvolution, deep learning and clustering algorithm, improves the false positive phenomenon in the segmentation results, and improves the segmentation accuracy.

[0006] This application mainly includes the following aspects:

[0007] In a first aspect, an embodiment of the present application provides a color cell image segmentation method, the method comprising: using color deconvolution and deep learning methods to segment a color cell original image to obtain a first cell nucleus segmentation image, wherein the color cell original image is an H&E cell image; using a clustering algorithm to classify the pixels in the color cell original image to obtain a classification result corresponding to each pixel; according to the classification result, determining a second cell nucleus segmentation image, wherein the first cell nucleus segmentation image and the second cell nucleus segmentation image are both binary mask images with the cell nucleus as the foreground; fusing the brightness values ​​at the same pixel position in the first cell nucleus segmentation image and the second cell nucleus segmentation image to obtain a fused brightness value corresponding to the pixel at each pixel position, the fused brightness value indicating whether the pixel at the pixel position is a cell nucleus; generating a target cell nucleus segmentation image according to the fused brightness value corresponding to the pixel at each pixel position.

[0008] In a possible implementation, the first cell nucleus segmentation image is determined in the following manner: color deconvolution is performed on the three primary color channels corresponding to each pixel in the color cell original image to obtain a first target image, the three primary color channels corresponding to the pixels in the first target image include designated channels and non-designated channels, and the designated channels are channels carrying cell nucleus patterns; the designated channels are extracted from the first target image to obtain a second target image containing a cell nucleus pattern, and the second target image is a grayscale image; the second target image is segmented using a deep learning method to obtain a first cell nucleus segmentation image, and the first cell nucleus segmentation image is a binary mask image.

[0009] In a possible implementation, the classification result corresponding to each pixel point is obtained in the following manner: the three primary color channels corresponding to the pixel point in the color cell original image are mapped into three-dimensional coordinates; and the three-dimensional coordinates corresponding to each pixel point in the color cell original image are clustered using a clustering algorithm to obtain the classification result corresponding to each pixel point.

[0010] In a possible embodiment, the step of determining a second cell nucleus segmentation image based on the classification result includes: extracting multiple target pixel points based on the classification result corresponding to each pixel point, wherein the target pixel points are classified as cytoplasm or cell nucleus; extracting the position information of each target pixel point in the color cell original image; generating a segmented image to be processed with the classification indicated by the target pixel as the foreground based on the position information corresponding to each target pixel point; performing image optimization processing on the segmented image to be processed to generate a second cell nucleus segmentation image.

[0011] In one possible implementation, the segmented image to be processed is generated in the following manner: a blank image of the same size as the original color cell image is generated; the brightness value of the pixel to be processed at the position corresponding to the target pixel in the blank image is set to 1 to obtain a processed image; the brightness value of the pixel in the processed image other than the pixel to be processed is set to 0 to generate the segmented image to be processed.

[0012] In a possible implementation, if the target pixel is a pixel classified as a cell nucleus, the segmented image to be processed is a cell nucleus segmented image; if the target pixel is a pixel classified as a cytoplasm, the segmented image to be processed is a cytoplasm segmented image.

[0013] In a possible implementation, the second cell nucleus segmentation image is generated in the following manner: if the segmented image to be processed is a cell nucleus segmentation image, an opening operation is performed on the segmented image to obtain a second cell nucleus segmentation image; if the segmented image to be processed is a cell nucleus-cytoplasm segmentation image, a closing operation is performed on the segmented image to obtain a pending cytoplasm segmentation image; and an inversion operation is performed on the pending cytoplasm segmentation image to obtain a second cell nucleus segmentation image.

[0014] In a possible implementation, the fused brightness value corresponding to each pixel position is determined by multiplying the brightness values ​​corresponding to pixels at the same position of the second cell nucleus segmentation image and the first cell nucleus segmentation image to obtain the fused brightness value corresponding to each pixel position.

[0015] In a second aspect, an embodiment of the present application also provides a color cell image segmentation device, the device comprising: a first segmentation module, used to segment the color cell original image using color deconvolution and deep learning methods to obtain a first cell nucleus segmentation image, the color cell original image is an H&E cell image; a clustering module, used to classify the pixels in the color cell original image using a clustering algorithm to obtain a classification result corresponding to each pixel; a second segmentation module, used to determine a second cell nucleus segmentation image based on the classification result, the first cell nucleus segmentation image and the second cell nucleus segmentation image are both binary mask images with the cell nucleus as the foreground; a fusion module, used to fuse the brightness values ​​at the same pixel position in the first cell nucleus segmentation image and the second cell nucleus segmentation image to obtain a fused brightness value corresponding to the pixel at each pixel position, the fused brightness value indicating whether the pixel at the pixel position is a cell nucleus; a generation module, further used to generate a target cell nucleus segmentation image according to the fused brightness value corresponding to the pixel at each pixel position.

[0016] In a third aspect, an embodiment of the present application further provides an electronic device, comprising: a processor, a memory and a bus, wherein the memory stores machine-readable instructions executable by the processor, and when the electronic device is running, the processor and the memory communicate through the bus, and the machine-readable instructions are executed by the processor to execute the steps of the color cell image segmentation method in the above-mentioned first aspect or any possible implementation of the first aspect.

[0017] The embodiments of the present application provide a method, device and electronic device for color cell image segmentation, the method comprising: using color deconvolution and deep learning methods to segment a color cell original image to obtain a first cell nucleus segmentation image, wherein the color cell original image is an H&E cell image; using a clustering algorithm to classify the pixels in the color cell original image to obtain a classification result corresponding to each pixel; determining a second cell nucleus segmentation image according to the classification result, wherein the first cell nucleus segmentation image and the second cell nucleus segmentation image are both binary mask images with the cell nucleus as the foreground; fusing the brightness values ​​at the same pixel position in the first cell nucleus segmentation image and the second cell nucleus segmentation image to obtain a fused brightness value corresponding to the pixel at each pixel position, wherein the fused brightness value indicates whether the pixel at the pixel position is a cell nucleus; generating a target cell nucleus segmentation image according to the fused brightness value corresponding to the pixel at each pixel position.

[0018] The segmentation process of color cell images in this application is mainly divided into three stages. The first stage is to integrate color deconvolution with deep learning to segment the color cell images, so as to optimize the segmentation results in the first stage and screen out most of the non-nucleus images. Then the second stage is to process the color cell images through a clustering algorithm to identify the nucleus-related images in the color cell images. On the basis of the above, the third stage is to integrate the segmentation results of the first stage with the segmentation results of the second stage to obtain the final segmentation results. In the whole process, the final nucleus recognition and segmentation are completed by integrating color deconvolution, deep learning and clustering algorithms, and the relatively complex and difficult-to-identify non-nucleus parts "remaining" in the nucleus image identified after the color deconvolution operation are proposed to improve the segmentation accuracy.

[0019] In order to make the above-mentioned objects, features and advantages of the present application more obvious and easy to understand, preferred embodiments are specifically cited below and described in detail with reference to the attached drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required for use in the embodiments will be briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present application and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without paying creative work.

[0021] Figure 1 A flow chart of a color cell image segmentation method provided in an embodiment of the present application is shown;

[0022] Figure 2 A flow chart for determining a first cell nucleus segmentation image provided by an embodiment of the present application is shown;

[0023] Figure 3 A flow chart of a second cell nucleus segmentation image determination method provided in an embodiment of the present application is shown;

[0024] Figure 4 A functional module diagram of a color cell image segmentation device provided in an embodiment of the present application;

[0025] Figure 5 A schematic structural diagram of an electronic device provided in an embodiment of the present application is shown. DETAILED DESCRIPTION

[0026] To make the purpose, technical scheme and advantages of the embodiments of the present application clearer, the technical scheme in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. It should be understood that the drawings in the present application only serve the purpose of explanation and description and are not used to limit the scope of protection of the present application. In addition, it should be understood that the schematic drawings are not drawn in real proportion. The flowchart used in this application shows the operations implemented according to some embodiments of the present application. It should be understood that the operations of the flowchart can be implemented out of sequence, and the steps without logical context can be reversed in order or implemented simultaneously. In addition, those skilled in the art, under the guidance of the content of the present application, can add one or more other operations to the flowchart, or remove one or more operations from the flowchart.

[0027] In addition, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. The components of the embodiments of the present application described and shown in the drawings here can be arranged and designed in various configurations. Therefore, the following detailed description of the embodiments of the present application provided in the drawings is not intended to limit the scope of the application claimed for protection, but merely represents the selected embodiments of the present application. Based on the embodiments of the present application, all other embodiments obtained by those skilled in the art without making creative work belong to the scope of protection of the present application.

[0028] Color deconvolution is often used for the recognition and segmentation of H&E cell images. Although color deconvolution brings convenience to the subsequent cell nucleus image segmentation, it still cannot fully guarantee the perfect "separation" between the cell nucleus and other background parts. The resulting cell nucleus layer still has some "remaining" cytoplasm, making the final segmented cell nucleus image inaccurate.

[0029] Based on this, the embodiments of the present application provide a color cell image segmentation method, device and electronic device, which performs cell nucleus segmentation by combining color deconvolution, deep learning and clustering algorithm, improves the false positive phenomenon in the segmentation result and improves the segmentation accuracy, as follows:

[0030] See also Figure 1 , Figure 1 FIG. 4 is a flow chart showing a color cell image segmentation method provided in an embodiment of the present application. Figure 1 As shown, the method provided in the embodiment of the present application comprises the following steps:

[0031] S100, using color deconvolution and deep learning methods to segment the color cell original image to obtain a first cell nucleus segmentation image.

[0032] The original colored cell images are H&E cell images.

[0033] S200, using a clustering algorithm to classify the pixels in the original color cell image, and obtaining a classification result corresponding to each pixel.

[0034] S300: Determine a second cell nucleus segmentation image according to the classification result.

[0035] The first cell nucleus segmentation image and the second cell nucleus segmentation image are both binary mask images with the cell nucleus as the foreground.

[0036] S400, merging the brightness values ​​at the same pixel position in the first cell nucleus segmentation image and the second cell nucleus segmentation image to obtain a fused brightness value corresponding to the pixel at each pixel position.

[0037] The fused brightness value indicates whether the pixel beneath the pixel location is a cell nucleus.

[0038] S500, generating a target cell nucleus segmentation image according to the fused brightness value corresponding to the pixel at each pixel position.

[0039] In the specific implementation, in steps S100 to S500, the color cell image is first segmented by color deconvolution and deep learning fusion, most of the non-nucleus part images are screened out, and a first nucleus segmentation image is obtained. Then, the color cell image is processed by a clustering algorithm to identify the nucleus-related images in the color cell image to obtain a second nucleus segmentation image. On the above basis, the first nucleus segmentation image and the second nucleus segmentation image are integrated, and the brightness values ​​of the pixels at the same pixel position are fused to obtain the fused brightness value corresponding to each pixel position. The final target nucleus segmentation image is obtained based on the fused brightness value corresponding to each pixel position. In the whole process, the color deconvolution, deep learning and clustering algorithms are integrated to complete the final nucleus recognition and segmentation, and the relatively complex and difficult-to-identify non-nucleus part "remaining" in the nucleus image identified after the color deconvolution operation is proposed to improve the segmentation accuracy.

[0040] In a preferred embodiment, see Figure 2 , Figure 2 FIG. 4 shows a flow chart of determining a first cell nucleus segmentation image provided by an embodiment of the present application. Figure 2 As shown, step S100 includes:

[0041] S1001. Perform color deconvolution on the three primary color channels corresponding to each pixel in the color cell original image to obtain a first target image.

[0042] The three primary color channels corresponding to the pixels in the first target image include designated channels and non-designated channels, and the designated channels are channels carrying cell nucleus patterns.

[0043] S1002, extracting a designated channel from the first target image to obtain a second target image containing a cell nucleus pattern.

[0044] The second target image is a grayscale image. Specifically, the second target image is a grayscale image with a cell nucleus pattern as the foreground.

[0045] S1003. Segment the second target image using a deep learning method to obtain a first cell nucleus segmentation image.

[0046] The first cell nucleus segmentation image is a binary mask image.

[0047] In one example, in step S1001, each pixel in the color cell image corresponds to a three-primary color channel RGB (including channel R, channel G and channel B), and the different brightness values ​​corresponding to the three primary color channels determine the color displayed by the pixel. In the present application, color deconvolution is performed on the three primary color channels corresponding to each pixel in the color cell image, so that the cell nuclear pattern indicated by the hematoxylin staining area is reflected on the designated channel (i.e., channel B), and the cell nuclear pattern indicated by the eosin staining area is reflected on the non-designated channel (i.e., channel R or channel G), so as to obtain the processed first target image.

[0048] In step S1002, the brightness value corresponding to the designated channel of each pixel in the first target image is extracted or copied to form a second target image containing a cell nucleus pattern.

[0049] In step S1003, the deep learning method includes but is not limited to at least one of the following items: PanopticNet algorithm, watershed algorithm and unet algorithm. Specifically, in an optional implementation, the PanopticNet algorithm and watershed algorithm in deep learning can be used to segment the second target image to obtain a binary mask image of the cell nucleus, that is, a first cell nucleus segmentation image, and the brightness value of each pixel on the first cell nucleus segmentation image is 0 or 1.

[0050] In another optional implementation, the second target image can also be segmented using the UNET algorithm and the watershed algorithm in deep learning to obtain a binary mask image of the cell nucleus, that is, the first cell nucleus segmentation image.

[0051] In a preferred embodiment, step S200 includes:

[0052] The three primary color channels corresponding to the pixels in the original color cell image are mapped into three-dimensional coordinates, and the three-dimensional coordinates corresponding to each pixel in the original color cell image are clustered using a clustering algorithm to obtain the classification result corresponding to each pixel.

[0053] In one example, since the original color cell image is a three-primary color image, each pixel therein displays color through the corresponding brightness values ​​under the three primary color channels R, G and B. Therefore, the brightness values ​​of each pixel under the three channels of R, G, and B are regarded as coordinate points in three-dimensional space. Each pixel has three coordinate values ​​corresponding to it, so that each pixel has a three-dimensional coordinate. The clustering algorithm is the Kmeans algorithm. The Kmeans algorithm is used to perform clustering training and classification on the three-dimensional coordinates corresponding to all pixels on the original color cell image, and a classification label is added to each pixel on the original color cell image to obtain the classification result corresponding to the pixel. Finally, the classification result corresponding to each pixel indicates that the pixel is a cell nucleus pattern or a non-cell nucleus pattern.

[0054] Specifically, it is assumed that the number or type of pixel classifications for the Kmeans algorithm is 3, and the corresponding classification labels are set to 0, 1, and 2, respectively, where the classification label 1 indicates that the pixel is a cell nucleus pattern.

[0055] In a preferred embodiment, see Figure 3 , Figure 3 FIG. 4 is a flow chart showing a second cell nucleus segmentation image determination method provided in an embodiment of the present application. Figure 3 As shown, step S300 includes:

[0056] S3001. Determine multiple target pixels according to the classification result corresponding to each pixel.

[0057] The target pixels were classified as either nucleus or cytoplasm.

[0058] S3002. Extract the position information of each target pixel in the original color cell image.

[0059] S3003. Generate a segmented image to be processed with the classification indicated by the target pixel as the foreground according to the position information corresponding to each target pixel.

[0060] S3004, performing image optimization processing on the segmented image to be processed to generate a second cell nucleus segmented image.

[0061] Specifically, in step S3001 to step S3004, multiple target pixels are extracted according to the classification mark corresponding to each pixel in the color cell original image, and the target pixel is classified as a cell nucleus or cytoplasm. Then, according to the position information of the target pixel in the color cell original image, a segmented image to be processed that is the same as the color cell original image is generated. The segmented image to be processed is a binary mask image. The type of the target pixel determines the type of the generated segmented image to be processed. Specifically, if the target pixel is a pixel classified as a cell nucleus, the segmented image to be processed is a cell nucleus segmented image with the cell nucleus as the foreground. If the target pixel is a pixel classified as a cytoplasm, the segmented image to be processed is a cytoplasm segmented image with the cytoplasm as the foreground.

[0062] Then, according to the type of the segmented image to be processed, the corresponding optimization method is used to optimize the segmented image to be processed, and finally a second cell nucleus segmentation image is obtained. Through image optimization, the accuracy of clustering and segmenting the cell nucleus image can be further improved.

[0063] In a preferred embodiment, step S3003 includes:

[0064] Generate a blank image of the same size as the original color cell image, set the brightness value of the pixel to be processed at the position corresponding to the target pixel in the blank image to 1, and obtain the processed image, and set the brightness value of the pixel in the processed image except the pixel to be processed to 0 to generate the segmented image to be processed.

[0065] Specifically, the segmented image to be processed is a binary image, and the brightness value corresponding to each pixel in the segmented image to be processed is 0 or 1. If the brightness value is 0, it means that the pixel is the background, and if the brightness value is 1, it means that the pixel is the foreground. Specifically, for the segmented image to be processed, the corresponding foreground is the cell nucleus, and for the cell nucleoplasm segmented image to be processed, the corresponding foreground is the cytoplasm.

[0066] In a preferred embodiment, step S3004 includes:

[0067] If the segmented image to be processed is a cell nucleus segmented image, an opening operation is performed on the segmented image to obtain a second cell nucleus segmented image.

[0068] Specifically, an opening operation is performed on the segmented image to be processed, that is, an erosion operation and a dilation operation are performed on the segmented image to be processed in sequence. After the erosion operation, the noise in the segmented image to be processed is removed, but the image is also compressed while removing the noise. Therefore, an dilation operation is performed on the segmented image to be processed after the erosion, and the noise is removed while retaining the original image. It can be seen that the present application can remove the noise interference in the segmented image to be processed by performing an opening operation on the segmented image to be processed, thereby obtaining a second cell nucleus segmentation image without noise interference.

[0069] In another preferred embodiment, step S3004 further includes:

[0070] If the segmented image to be processed is a cell nucleus-cytoplasm segmented image, a closing operation is performed on the segmented image to obtain a cytoplasm segmented image to be determined, and an inversion operation is performed on the cytoplasm segmented image to obtain a second cell nucleus segmented image.

[0071] Specifically, in the present application, when the segmented image to be processed is a cell nucleus-cytoplasm segmentation image, the segmented image to be processed can also be optimized by a closing operation. Specifically, a closing operation is performed on the segmented image to be processed, that is, an expansion operation and an erosion operation are performed on the segmented image to be processed in sequence. The closing operation helps to close the small holes or small black spots inside the cell nucleus part of the image, thereby obtaining a clean cytoplasm segmentation image to be determined. The cytoplasm segmentation image to be determined is then further inverted, that is, the foreground and background in the cytoplasm segmentation image are swapped, and a second cell nucleus segmentation image can be obtained.

[0072] In a preferred embodiment, step S400 includes:

[0073] The brightness values ​​corresponding to the pixels at the same position of the second cell nucleus segmentation image and the first cell nucleus segmentation image are multiplied to obtain a fused brightness value corresponding to each pixel position.

[0074] In a specific embodiment, in steps S400 to S500, the first cell nucleus segmentation image and the second cell nucleus segmentation image are both binary mask images with the cell nucleus as the foreground, that is, in the first cell nucleus segmentation image and the second cell nucleus segmentation image, the brightness value corresponding to each pixel is either 1 or 0. The brightness value corresponding to the pixel is 1, indicating that the pixel belongs to the image foreground and the pixel is the cell nucleus, and the brightness value corresponding to the pixel is 0, indicating that the pixel belongs to the image background and the pixel is not the cell nucleus. Therefore, the present application multiplies the brightness values ​​corresponding to the pixels at the same position in the first cell nucleus segmentation image and the second cell nucleus segmentation image. Similarly, the fused brightness value obtained is The fusion brightness value at the pixel position is either 1 or 0. The fusion brightness value at the pixel position is 1, indicating that the pixels at this pixel position of the first cell nucleus segmentation image and the second cell nucleus segmentation image are both cell nuclei. Then, it is determined that the pixel at this pixel position is the cell nucleus. The fusion brightness value at the pixel position is 0, indicating that at least one pixel at this position of the first cell nucleus segmentation image and the second cell nucleus segmentation image is not a cell nucleus. Then, it is finally determined that the pixel at this pixel position is not a cell nucleus. In this way, the target cell nucleus segmentation image can be generated according to the fusion brightness value corresponding to the pixel at each pixel position. The target cell nucleus segmentation image is also a binary mask image with the cell nucleus as the foreground.

[0075] In this application, based on the combination of color deconvolution and general deep learning algorithms, it is further proposed to add the segmentation results obtained by the clustering algorithm, and use the segmentation results to correct the segmentation results obtained by color deconvolution and general deep learning algorithms, so as to further improve the cell nucleus recognition and segmentation accuracy.

[0076] Based on the same application concept, the embodiments of the present application also provide a color cell image segmentation device corresponding to the color cell image segmentation method provided in the above embodiments. Since the principle of solving the problem by the device in the embodiments of the present application is similar to the color cell image segmentation method in the above embodiments of the present application, the implementation of the device can refer to the implementation of the method, and the repeated parts will not be repeated.

[0077] See also Figure 4 , Figure 4 This is a functional module diagram of a color cell image segmentation device provided in an embodiment of the present application. Figure 4 As shown, the device comprises:

[0078] The first segmentation module 510 is used to segment the color cell original image by using color deconvolution and deep learning method to obtain a first cell nucleus segmentation image, where the color cell original image is a H&E cell image.

[0079] The clustering module 520 is used to classify the pixels in the original color cell image using a clustering algorithm to obtain a classification result corresponding to each pixel.

[0080] The second segmentation module 530 is used to determine a second cell nucleus segmentation image according to the classification result. The first cell nucleus segmentation image and the second cell nucleus segmentation image are both binary mask images with the cell nucleus as the foreground.

[0081] A fusion module 540 is used to fuse the brightness values ​​at the same pixel position in the first cell nucleus segmentation image and the second cell nucleus segmentation image to obtain a fused brightness value corresponding to the pixel at each pixel position, and the fused brightness value indicates whether the pixel at the pixel position is a cell nucleus;

[0082] The generating module 550 is used to generate a target cell nucleus segmentation image according to the fused brightness value corresponding to the pixel at each pixel position.

[0083] Based on the same application idea, please refer to Figure 5 , Figure 5 FIG. 1 is a schematic diagram showing the structure of an electronic device provided by an embodiment of the present application. Figure 5As shown, the electronic device 600 includes: a processor 610, a memory 620 and a bus 630. The memory 620 stores machine-readable instructions executable by the processor 610. When the electronic device 600 is running, the processor 610 and the memory 620 communicate through the bus 630. The machine-readable instructions are executed by the processor 610 when it is running, such as the steps of the color cell image segmentation method provided in any of the above embodiments.

[0084] Based on the same application concept, an embodiment of the present application further provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps of the color cell image segmentation method provided in the above embodiment are executed.

[0085] Those skilled in the art can clearly understand that, for the convenience and simplicity of description, the specific working process of the system and device described above can refer to the corresponding process in the aforementioned method embodiment, and will not be repeated here. In the several embodiments provided in the present application, it should be understood that the disclosed system, device and method can be implemented in other ways. The device embodiments described above are merely schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some communication interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0086] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0087] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0088] If the function is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a non-volatile computer-readable storage medium that is executable by a processor. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions to enable a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), disk or optical disk, etc., various media that can store program codes.

[0089] The above are only specific implementations of the present application, but the protection scope of the present application is not limited thereto. Any technician familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. A color cell image segmentation method, characterized in that: The method comprises: The color cell original image is segmented by using color deconvolution and deep learning method to obtain a first cell nucleus segmentation image, wherein the color cell original image is a H&E cell image; Using a clustering algorithm to classify the pixels in the original color cell image, and obtaining a classification result corresponding to each pixel; Determine a second cell nucleus segmentation image according to the classification result, wherein both the first cell nucleus segmentation image and the second cell nucleus segmentation image are binary mask images with the cell nucleus as the foreground; Fusion of the brightness values ​​at the same pixel position in the first cell nucleus segmentation image and the second cell nucleus segmentation image to obtain a fused brightness value corresponding to the pixel at each pixel position, the fused brightness value indicating whether the pixel at the pixel position is a cell nucleus; According to the fused brightness value corresponding to the pixel at each pixel position, a target cell nucleus segmentation image is generated.

2. The method according to claim 1, characterized in that The first cell nucleus segmentation image is determined by: Performing color deconvolution on the three primary color channels corresponding to each pixel in the color cell original image to obtain a first target image, wherein the three primary color channels corresponding to the pixels in the first target image include designated channels and non-designated channels, and the designated channels are channels carrying cell nucleus patterns; Performing designated channel extraction on the first target image to obtain a second target image containing a cell nucleus pattern, wherein the second target image is a grayscale image; The second target image is segmented by using a deep learning method to obtain the first cell nucleus segmentation image, which is a binary mask image.

3. The method according to claim 1, characterized in that The classification result corresponding to each pixel is obtained in the following way: Mapping the three primary color channels corresponding to the pixel points in the color cell original image into three-dimensional coordinates; A clustering algorithm is used to cluster the three-dimensional coordinates corresponding to each pixel in the original color cell image to obtain a classification result corresponding to each pixel.

4. The method according to claim 1, characterized in that: The step of determining the second cell nucleus segmentation image according to the classification result comprises: Extracting a plurality of target pixels according to the classification result corresponding to each pixel, wherein the target pixels are classified as cell nuclei or cytoplasm; Extracting the position information of each target pixel in the color cell original image; According to the position information corresponding to each target pixel, a segmented image to be processed is generated with the classification indicated by the target pixel as the foreground; Perform image optimization processing on the segmented image to be processed to generate a second cell nucleus segmented image.

5. The method according to claim 4, characterized in that Generate the segmented image to be processed by: Generate a blank image of the same size as the original color cell image; The brightness value of the pixel to be processed at the position corresponding to the target pixel in the blank image is set to 1 to obtain a processed image; The brightness values ​​of the pixels in the processed image other than the pixels to be processed are set to 0 to generate a segmented image to be processed.

6. The method according to claim 4, characterized in that If the target pixel is a pixel classified as a cell nucleus, the segmented image to be processed is a cell nucleus segmented image; if the target pixel is a pixel classified as a cytoplasm, the segmented image to be processed is a cytoplasm segmented image.

7. The method according to claim 6, characterized in that Generate a second nucleus segmentation image by: If the segmented image to be processed is a cell nucleus segmented image, an opening operation is performed on the segmented image to be processed to obtain a second cell nucleus segmented image; If the segmented image to be processed is a cell nucleus-cytoplasm segmented image, a closing operation is performed on the segmented image to be processed to obtain a cell nucleus segmented image to be determined; An inversion operation is performed on the cell nucleus segmentation image to obtain a second cell nucleus segmentation image.

8. The method according to claim 1, characterized in that The fused brightness value corresponding to each pixel position is determined in the following way: The brightness values ​​corresponding to the pixels at the same position of the second cell nucleus segmentation image and the first cell nucleus segmentation image are multiplied to obtain a fused brightness value corresponding to each pixel position.

9. A color cell image segmentation device, characterized in that: The device comprises: A first segmentation module is used to segment the color cell original image by using color deconvolution and deep learning method to obtain a first cell nucleus segmentation image, wherein the color cell original image is a H&E cell image; A clustering module, used to classify the pixels in the original color cell image using a clustering algorithm to obtain a classification result corresponding to each pixel; A second segmentation module is used to determine a second cell nucleus segmentation image according to the classification result, wherein the first cell nucleus segmentation image and the second cell nucleus segmentation image are both binary mask images with the cell nucleus as the foreground; a fusion module, used to fuse the brightness values ​​at the same pixel position in the first cell nucleus segmentation image and the second cell nucleus segmentation image to obtain a fused brightness value corresponding to the pixel at each pixel position, wherein the fused brightness value indicates whether the pixel at the pixel position is a cell nucleus; The generation module is used to generate a target cell nucleus segmentation image according to the fused brightness value corresponding to the pixel at each pixel position.

10. An electronic device, characterized in that: include: A processor, a memory and a bus, wherein the memory stores machine-readable instructions executable by the processor, and when the electronic device is running, the processor and the memory communicate through the bus, and the machine-readable instructions are executed by the processor to execute the steps of the color cell image segmentation method as described in any one of claims 1 to 8.

Images