Method and system for identifying microvascular invasion area of ​​liver cancer based on pathological slice images

Through the identification method based on pathological slice images, image processing technology is used to extract and segment the closed contour area of ​​blood vessels, which solves the existing problems of low diagnosis efficiency and poor accuracy of liver cancer, and achieves efficient and accurate identification of microvascular invasion areas without artificial participation and Tumor cell count.

CN118135620BActive Publication Date: 2024-11-26GUANGZHOU KUNYUAN FANGQING MEDICAL TECH CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202410240646.0
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-03-04
Publication Date
2024-11-26
Estimated Expiration
2044-03-04

AI Technical Summary

Technical Problem

The existing liver cancer diagnosis methods are inefficient and have poor accuracy, rely on manual operations and are prone to incorrect diagnosis results due to personnel omissions or cell obstruction, overlap, and misalignment.

Method used

Based on the identification method of pathological section images, by obtaining digital pathological section images, extracting image features, segmenting and classifying the vascular closed contour area, determining whether it is located outside the liver cancer tumor area, locate the liver cancer microvascular invasion area and calculate the tumor cell quantity, using HSV color Image processing is carried out by technologies such as range segmentation method and canny edge segmentation method.

Benefits of technology

No manual participation is required, which improves the accuracy and efficiency of identification of microvascular invasion areas of liver cancer and ensures the accuracy of diagnostic results.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118135620B_ABST
    Figure CN118135620B_ABST
Patent Text Reader

Abstract

The present invention provides a method and system for identifying liver cancer microvascular invasion areas based on pathological slice images, which mainly include: extracting image features from digital pathological slice images to determine whether there is a closed blood vessel contour area; marking the closed blood vessel contour area that is determined to be located outside the liver cancer tumor area as the target blood vessel area; and performing a comprehensive analysis on all target blood vessel areas to obtain the liver cancer microvascular invasion area and the amount of tumor cells invaded by the liver cancer microvascular invasion. By extracting the closed blood vessel contour area and judging whether it is located outside the liver cancer tumor area, the target blood vessel area can be effectively confirmed, and then by analyzing all target blood vessel areas, the liver cancer microvascular invasion area can be effectively and accurately located, and the amount of invaded tumor cells can be accurately known based on the data obtained by the analysis. The whole process does not require manual participation, which solves the problem of low efficiency and poor accuracy in the existing liver cancer diagnosis.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of image processing, and in particular to a method and system for identifying a liver cancer microvascular invasion region based on a pathological section image. Background Art

[0002] Scientifically identifying high-risk groups for liver cancer and formulating stratified monitoring plans are the most critical links in early detection, early diagnosis and improving the overall survival rate of liver cancer.

[0003] At present, the diagnostic methods for confirming liver cancer mainly include imaging examination, hematological examination and pathological examination. Through imaging examination or pathological examination, the location, number, size of tumor or cancer cells and their relationship with surrounding organs and blood vessels can be obtained.

[0004] However, existing imaging or pathological examinations are usually performed manually by doctors, which is a complex, time-consuming, labor-intensive process that is highly dependent on the doctor's diagnostic experience. In addition, manual operations can easily lead to errors in target cell counting due to human negligence or cell occlusion, overlap, misalignment, etc., resulting in incorrect diagnostic results. Summary of the invention

[0005] The purpose of the present invention is to provide a method and system for identifying the microvascular invasion area of ​​liver cancer based on pathological section images, so as to solve the problem of low efficiency and poor accuracy of existing liver cancer diagnosis.

[0006] In order to solve the above technical problems, the present invention provides a method for identifying the microvascular invasion area of ​​liver cancer based on pathological slice images, comprising:

[0007] Acquire digital pathology slide images;

[0008] Extract image features from digital pathology slice images to determine whether there is a closed blood vessel contour area;

[0009] Segment and classify each blood vessel closed contour area to obtain the cell nucleus contour information;

[0010] Using the cell nucleus contour information, determining whether the closed blood vessel contour region is located outside the liver cancer tumor region, and if so, marking the closed blood vessel contour region as a target blood vessel region;

[0011] All target vascular regions were comprehensively analyzed to locate the areas of microvascular invasion of liver cancer, and the amount of tumor cells in microvascular invasion of liver cancer was obtained based on the cell nuclear contour information.

[0012] Optionally, in the method for identifying liver cancer microvascular invasion areas based on pathological slice images, the method for extracting image features from digital pathological slice images to determine whether there is a closed vascular contour area includes:

[0013] Obtain all suspected closed vascular contour areas in digital pathology slice images;

[0014] Analyze each suspected closed blood vessel contour area to determine whether it contains obvious image features of suspected blood vessel boundaries;

[0015] If there are obvious image features of suspected blood vessel boundaries, then judging whether the suspected blood vessel closed contour region contains obvious image features of suspected cancer cells;

[0016] If the image features of obvious suspected cancer cells are contained, the suspected closed blood vessel contour region is determined as a region containing a closed blood vessel contour.

[0017] Optionally, in the method for identifying liver cancer microvascular invasion areas based on pathological slice images, the method for obtaining all suspected closed blood vessel contour areas in the digital pathological slice images includes:

[0018] The digital pathology slice image is processed into blocks to obtain a plurality of sub-slice images;

[0019] Each sub-slice image is processed using the HSV color range segmentation method to obtain a sub-mask image corresponding to the sub-slice image;

[0020] Concatenate all sub-mask images to obtain a full-view mask image;

[0021] Obtain the suspected closed blood vessel contour area from the full-view mask image.

[0022] Optionally, in the method for identifying the microvascular invasion area of ​​liver cancer based on the pathological section image, in the HSV color range segmentation method adopted, the lower limit of the numerical values ​​of the three HSV channels is [0,0,180] and the upper limit is [180,80,255]; the cv2.findCountor function is used to obtain the suspected vascular closed contour area from the full-field mask image.

[0023] Optionally, in the method for identifying liver cancer microvascular invasion areas based on pathological slice images, the method for analyzing each suspected closed blood vessel contour area to determine whether it contains obvious image features of suspected blood vessel boundaries includes:

[0024] The HSV color range segmentation method is used to obtain the mask image of vascular wall cells from the suspected vascular closed contour area;

[0025] The canny edge segmentation method is used to obtain the vascular edge image from the suspected vascular closed contour area;

[0026] The mask image of the vascular wall cells and the vascular edge image are used to determine whether the suspected vascular closed contour area contains obvious image features of the suspected vascular boundary.

[0027] Optionally, in the method for identifying the microvascular invasion area of ​​liver cancer based on the pathological section image, in the HSV color range segmentation method, the lower limit of the values ​​of the three HSV channels is [26, 43, 46] and the upper limit is [34, 255, 255]; and the method of obtaining a vascular edge image from a suspected vascular closed contour area using the canny edge segmentation method includes:

[0028] Perform grayscale and Gaussian smoothing on the suspected closed blood vessel contour area to obtain a grayscale image;

[0029] Calculate the gradient and gradient direction of the grayscale image to obtain the initial edge;

[0030] Utilize non-maximum suppression and double threshold screening to extract target edges from initial edges;

[0031] The target edge is tracked by using the hysteresis boundary tracking to obtain the blood vessel edge image.

[0032] Optionally, in the method for identifying liver cancer microvascular invasion areas based on pathological slice images, if the image features of the suspected blood vessel boundary are obvious, the method for determining whether the suspected blood vessel closed contour area contains obvious image features of suspected cancer cells includes:

[0033] The HSV color range segmentation method is used to process each suspected blood vessel closed contour region containing obvious image features of suspected blood vessel boundaries to obtain a regional mask image corresponding to the suspected blood vessel closed contour region, wherein the lower limit of the values ​​of the three HSV channels in the HSV color range segmentation method is [0,0,0] and the upper limit is [145,255,140];

[0034] The regional mask image is used to determine whether the suspected closed blood vessel contour region contains obvious image features of suspected cancer cells.

[0035] Optionally, in the method for identifying liver cancer microvascular invasion areas based on pathological slice images, the method for segmenting and classifying each blood vessel closed contour area to obtain cell nucleus contour information includes:

[0036] Build and train the instance segmentation algorithm model;

[0037] The instance segmentation algorithm model is used to segment and classify the closed contour area of ​​the blood vessel to obtain the contour information of the cell nucleus around and inside the blood vessel area.

[0038] Optionally, in the method for identifying the liver cancer microvascular invasion area based on the pathological section image, the cell nucleus contour information includes tumor cell distribution information; and the method for using the cell nucleus contour information to determine whether the blood vessel closed contour area is located outside the liver cancer tumor area includes:

[0039] Obtaining a tumor cell mask according to tumor cell distribution information;

[0040] Taking the closed contour area of ​​the blood vessel as the center, the mask is extended in eight directions to obtain eight blood vessel contour masks;

[0041] If within the preset distance range, all blood vessel contour masks and tumor cell masks have a union, then the blood vessel closed contour region is judged to be located inside the liver cancer tumor region, otherwise the blood vessel closed contour region is judged to be located outside the liver cancer tumor region.

[0042] In order to solve the above technical problems, the present invention further provides a system for identifying liver cancer microvascular invasion areas based on pathological section images, which is used to implement the method for identifying liver cancer microvascular invasion areas based on pathological section images as described in any one of the above items. The system for identifying liver cancer microvascular invasion areas based on pathological section images comprises:

[0043] An image acquisition module, used for acquiring digital pathology slice images;

[0044] A feature extraction module is used to extract image features from the digital pathology slice image to determine whether the digital pathology slice image contains a closed blood vessel contour area;

[0045] An analysis and processing module is used to process each closed blood vessel contour area to obtain the target blood vessel area, and to perform a comprehensive analysis on all target blood vessel areas to obtain the liver cancer microvascular invasion area and the amount of tumor cells invaded by the liver cancer microvascular invasion;

[0046] The report generation module is used to organize the microvascular invasion area of ​​liver cancer and the amount of tumor cells invaded by microvascular invasion of liver cancer to generate an identification report and output it.

[0047] The method and system for identifying the microvascular invasion area of ​​liver cancer based on pathological slice images provided by the present invention include: obtaining a digital pathological slice image; extracting image features from the digital pathological slice image to determine whether it contains a closed vascular contour area; segmenting and classifying each closed vascular contour area to obtain cell nucleus contour information; using the cell nucleus contour information to determine whether the closed vascular contour area is located outside the liver cancer tumor area, and if so, marking the closed vascular contour area as a target vascular area; performing a comprehensive analysis on all target vascular areas to locate the microvascular invasion area of ​​liver cancer, and obtaining the amount of tumor cells invaded by the microvascular invasion of liver cancer according to the cell nucleus contour information. By extracting the closed vascular contour area and determining whether it is located outside the liver cancer tumor area, the target vascular area can be effectively confirmed, and then by analyzing all target vascular areas, the microvascular invasion area of ​​liver cancer can be effectively and accurately located, and the amount of invaded tumor cells can be accurately known according to the data obtained by the analysis, and the whole process does not require manual participation, which solves the problem of low efficiency and poor accuracy of existing liver cancer diagnosis. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Figure 1 A flow chart of a method for identifying liver cancer microvascular invasion areas based on pathological slice images provided in this embodiment;

[0049] Figure 2 A schematic diagram of the structure of the instance segmentation algorithm model provided in this embodiment;

[0050] Figure 3 A schematic diagram of the structure of a system for identifying liver cancer microvascular invasion areas based on pathological slice images provided in this embodiment. DETAILED DESCRIPTION

[0051] The following is a further detailed description of the method and system for identifying the microvascular invasion area of ​​liver cancer based on pathological section images proposed by the present invention in conjunction with the accompanying drawings and specific embodiments. It should be noted that the accompanying drawings are all in a very simplified form and use non-precise proportions, which are only used to conveniently and clearly assist in explaining the purpose of the embodiments of the present invention. In addition, the structure shown in the accompanying drawings is often a part of the actual structure. In particular, the emphasis that each drawing needs to show is different, and sometimes different proportions are used.

[0052] It should be noted that the terms "first", "second", etc. in the specification, claims and drawings of the present invention are used to distinguish similar objects in order to describe the embodiments of the present invention, rather than to describe a specific order or sequence. It should be understood that the structures used in this way can be interchanged under appropriate circumstances. In addition, the terms "include" and "have" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0053] This embodiment provides a method for identifying the microvascular invasion area of ​​liver cancer based on pathological slice images. Figure 1 As shown, including:

[0054] S1, obtain digital pathology slice images;

[0055] S2, extracting image features from the digital pathology slice image to determine whether it contains a closed vascular contour area;

[0056] S3, performing segmentation and classification processing on each closed contour area of ​​the blood vessel to obtain the contour information of the cell nucleus;

[0057] S4, using the cell nucleus contour information to determine whether the closed blood vessel contour region is located outside the liver cancer tumor region, and if so, marking the closed blood vessel contour region as a target blood vessel region;

[0058] S5, all target vascular regions are comprehensively analyzed to locate the microvascular invasion area of ​​liver cancer, and the amount of tumor cells in the microvascular invasion of liver cancer is obtained based on the cell nuclear contour information.

[0059] The method for identifying the microvascular invasion area of ​​liver cancer based on pathological section images provided in this embodiment can effectively confirm the target vascular area by extracting the closed contour area of ​​the blood vessels and judging whether it is located outside the liver cancer tumor area. Then, by analyzing all the target vascular areas, the microvascular invasion area of ​​liver cancer can be effectively and accurately located, and the amount of invaded tumor cells can be accurately known based on the data obtained from the analysis. The whole process does not require human intervention, which solves the problem of low efficiency and poor accuracy in the existing liver cancer diagnosis.

[0060] Specifically, in this embodiment, step S1, obtaining a digital pathology slice image, can be obtained from reports provided by various imaging devices. The method of obtaining the digital pathology slice image is well known to those skilled in the art, and will not be elaborated in this application.

[0061] Further, in this embodiment, step S2, the method of extracting image features from the digital pathology slice image to determine whether there is a closed blood vessel contour area includes:

[0062] S21, acquiring all suspected closed blood vessel contour regions in the digital pathology slice image.

[0063] Specifically, in this embodiment, the method for obtaining all suspected closed blood vessel contour areas in the digital pathology slice image includes:

[0064] First, the digital pathology slice image is processed into blocks to obtain a plurality of sub-slice images. In this embodiment, the 10-fold image layer of the digital pathology slice image can be processed into blocks to obtain a plurality of sub-slice images with a high degree of detail.

[0065] Then, each sub-slice image is processed using the HSV color range segmentation method to obtain a sub-mask image corresponding to the sub-slice image. In this embodiment, in the HSV color range segmentation method used, the lower limit of the values ​​of the three HSV channels is [0, 0, 180] and the upper limit is [180, 80, 255], so that a mask image that is more consistent with the color area of ​​the vascular cavity can be obtained. Among them, HSV color refers to hue (Hue), saturation (Saturation) and brightness (Value). The specific implementation process of the HSV color range segmentation method is well known to those skilled in the art, and this application will not repeat it.

[0066] Then, all the sub-mask images are spliced ​​to obtain a full-view mask image. In this embodiment, the sub-mask images can be spliced ​​to obtain a full-view mask image according to the corresponding relationship between the sub-mask images and the sub-slice images and the connection order between the sub-slice images.

[0067] Finally, the suspected closed contour area of ​​the blood vessel is obtained from the full-view mask image. In this embodiment, the cv2.findCountor function is used to obtain the suspected closed contour area of ​​the blood vessel from the full-view mask image. Among them, the cv2.findCountor function is a commonly used function for image contour search, and its specific implementation method is also well known to those skilled in the art, and this application will not be repeated.

[0068] In this way, the HSV color range segmentation method can effectively intercept the mask area of ​​the vascular cavity, thereby effectively improving the segmentation accuracy of the vascular closed contour area; the cv2.findCountor function can effectively and accurately extract the suspected vascular closed contour area.

[0069] S22, analyzing each suspected blood vessel closed contour region to determine whether it contains obvious image features of the suspected blood vessel boundary.

[0070] Specifically, in this embodiment, the method of analyzing each suspected closed blood vessel contour region to determine whether it contains obvious image features of suspected blood vessel boundaries includes:

[0071] First, the HSV color range segmentation method is used to obtain the mask image of the vascular wall cells from the suspected closed contour area of ​​the blood vessel. In this embodiment, in the HSV color range segmentation method used in this step, the lower limit of the values ​​of the three HSV channels is [26, 43, 46] and the upper limit is [34, 255, 255], so that the mask image of the vascular wall cells can be effectively obtained; similarly, the cv.findContours function can be used to find the cell contour, calculate its cell aspect ratio, and its minimum distance to the vascular wall, and then obtain a part of the characteristic values ​​that can be used to judge whether it is close to the smooth attribute of the vascular wall.

[0072] Of course, the vascular contour in the suspected vascular closed contour area can also be Gaussian smoothed to obtain a smooth vascular cavity contour; by traversing each contour point and calculating the contour direction angle and distance, a part of the characteristic values ​​that can be used to judge whether it is close to the smoothness property of the vascular wall can be obtained.

[0073] Then, the canny edge segmentation method is used to obtain the vascular edge image from the suspected closed vascular contour area. In this embodiment, the suspected closed vascular contour area is grayed and Gaussian smoothed to obtain a gray image. In this way, the number of channels can be reduced by graying, and the noise can be reduced by Gaussian smoothing, thereby improving the processing efficiency and accuracy; the gradient and gradient direction of the gray image are calculated to obtain the initial edge. For example, if the intensity changes drastically at a certain point in the gray image, an edge will be formed there, and the edge is defined as the initial edge; non-maximum suppression and double threshold screening are used to extract the target edge from the initial edge, wherein the image that is not an edge is removed by NMS (Non-maximum-supperession). The pixel points are removed, and then the set threshold boundary is used to screen the pixel points to obtain the target edge; the target edge is tracked by using the hysteresis boundary tracking to obtain the vascular edge image. Since the effects of strong boundaries and weak boundaries are quite different, the weak boundaries connected to the strong boundaries are considered as boundaries in this embodiment, and other weak boundaries are suppressed. At the same time, the hysteresis threshold (including high threshold and low threshold) is used to track the edge. For example, if the important edge in the image is a continuous curve, a larger threshold is first used to include the real edge as much as possible, and then the lower threshold is gradually increased, so that a tracking can be achieved to obtain a complete curve, which is the vascular edge.

[0074] Finally, the mask image of the vascular wall cells and the vascular margin image are used to determine whether the suspected vascular closed contour region contains obvious image features of the suspected vascular boundary. In other words, if the suspected vascular closed contour region contains a vascular margin, it can be considered that it contains obvious image features of the suspected vascular boundary.

[0075] S23: If the image features of the suspected blood vessel boundary are obvious, then determine whether the suspected blood vessel closed contour region contains obvious image features of suspected cancer cells.

[0076] Specifically, in this embodiment, the method for determining whether a suspected closed blood vessel contour region having an obvious image feature of a suspected blood vessel boundary contains obvious image features of suspected cancer cells includes:

[0077] First, the HSV color range segmentation method is used to process each suspected closed blood vessel contour region containing obvious image features of suspected blood vessel boundaries to obtain a regional mask image corresponding to the suspected closed blood vessel contour region. In this step, the HSV color range segmentation method is used, and the lower limit of the value of the three HSV channels is [0,0,0] and the upper limit is [145,255,140].

[0078] Then, the regional mask image is used to determine whether the suspected closed contour area of ​​the blood vessel contains obvious image features of suspected cancer cells. The size of the tumor can be determined by calculating the regional mask area through the cv2.count_non_zero() function, so that it can be determined whether it contains obvious image features of suspected cancer cells based on the size of the tumor. In other words, if there is a tumor within a certain size range, it is determined to contain obvious image features of suspected cancer cells.

[0079] S24: If the image contains obvious suspected cancer cells, the suspected blood vessel closed contour region is determined as a blood vessel closed contour region.

[0080] In this way, the closed contour area containing blood vessels can be accurately locked, so that subsequent analysis and processing can be carried out to obtain the area of ​​microvascular invasion of liver cancer and the amount of tumor cells invaded by microvascular invasion of liver cancer.

[0081] Furthermore, in this embodiment, in step S3, the method of performing segmentation and classification processing on each closed blood vessel contour region to obtain cell nucleus contour information includes:

[0082] S31, build and train the instance segmentation algorithm model.

[0083] Specifically, in this embodiment, the instance segmentation algorithm model (Mask R-CNN) used is based on the Fast R-CNN framework, and a fully connected segmentation network is added after the basic feature network, changing the original two tasks (classification + regression) into three tasks (classification + regression + segmentation). Figure 2 As shown in the figure, Mask R-CNN uses the same two stages as Faster R-CNN, and uses FCN for semantic segmentation for each proposal box of Fast R-CNN.

[0084] In this embodiment, the outlines of tumor cells and stromal cells in a certain number of 512x512 sub-images can be manually annotated, and the original image and the annotated image can be input into Mask R-CNN for training iteration until convergence, so as to obtain a model that can perform cell nucleus segmentation and classification on the sub-images. The specific implementation method of building and training Mask R-CNN is well known to those skilled in the art, and will not be described in detail in this application.

[0085] Since Mask R-CNN is a deep learning model for object detection and instance segmentation, which can generate a binary mask of the object, Mask R-CNN can be used to achieve accurate instance segmentation and improve the segmentation accuracy of the cell nucleus contour.

[0086] S32, using the instance segmentation algorithm model to perform segmentation and classification processing on the closed contour area of ​​the blood vessel to obtain the contour information of the cell nuclei around the blood vessel area and inside the blood vessel.

[0087] Specifically, by segmenting and classifying the closed contour area of ​​the blood vessel through Mask R-CNN, the contour distribution information of tumor cells, stromal cells, etc. around and inside the blood vessel area can be obtained.

[0088] Furthermore, in this embodiment, step S4 uses the cell nucleus contour information to determine whether the closed blood vessel contour region is located outside the liver cancer tumor region. If so, the closed blood vessel contour region is identified as a target blood vessel region, which includes:

[0089] S41, obtaining a tumor cell mask according to the tumor cell distribution information.

[0090] Specifically, the HSV color range segmentation method can also be used to process the sub-slice image in the tumor cell distribution information, thereby obtaining a tumor cell mask image.

[0091] S42, extending the mask in eight directions with the closed contour area of ​​the blood vessel as the center to obtain eight blood vessel contour masks.

[0092] Specifically, the eight directions are evenly divided within the range of 360°, that is, the angle between every two directions is 45°. Of course, in other embodiments, fewer extension directions or more extension directions may be set, wherein fewer extension directions mean faster processing efficiency but a certain loss of accuracy, while more extension directions mean higher judgment accuracy but lower processing efficiency. By comparison, setting eight directions can ensure both judgment accuracy and processing efficiency, achieving a relative balance between the two.

[0093] S43, if within the preset distance range, all blood vessel contour masks and tumor cell masks have a union, then it is determined that the blood vessel closed contour region is located inside the liver cancer tumor region, otherwise it is determined that the blood vessel closed contour region is located outside the liver cancer tumor region.

[0094] Specifically, the preset distance range can be set according to actual needs; and, determining whether there is a union between the blood vessel contour mask and the tumor cell mask can be implemented using the cv2.bitwise_and function, the specific implementation method of which is well known to those skilled in the art and will not be elaborated in this application.

[0095] Furthermore, in this embodiment, in step S5, all target blood vessel regions are comprehensively analyzed to locate the liver cancer microvascular invasion region, and the amount of tumor cells invaded by the liver cancer microvascular invasion is obtained based on the cell nucleus contour information.

[0096] Specifically, by integrating all target vascular areas, the microvascular invasion area of ​​liver cancer can be located; through the above-mentioned image processing process and the segmentation and classification process of Mask R-CNN, the mask position of the tumor cell nucleus in the tumor area inside the vascular wall can be obtained, and then through the clustering method, the number of distribution centers of tumor cells can be calculated, which is the amount of tumor cells invaded by the microvessels.

[0097] In this way, by obtaining the mask image through the HSV color range segmentation method and using Mask R-CNN for segmentation and classification processing, the microvascular invasion area of ​​liver cancer can be effectively and accurately located, and the amount of invaded tumor cells can be accurately known based on the data obtained from the analysis. The whole process does not require human intervention, which solves the problem of low efficiency and poor accuracy in existing liver cancer diagnosis.

[0098] This embodiment also provides a liver cancer microvascular invasion region recognition system based on pathological slice images, which is used to implement the liver cancer microvascular invasion region recognition method based on pathological slice images as described above. Figure 3As shown, the recognition system for liver cancer microvascular invasion area based on pathological slice images includes: an image acquisition module, used to acquire digital pathological slice images; a feature extraction module, used to extract image features from the digital pathological slice images to determine whether the digital pathological slice images contain a vascular closed contour area; an analysis and processing module, used to process each vascular closed contour area to obtain a target vascular area, and to perform a comprehensive analysis on all target vascular areas to obtain liver cancer microvascular invasion areas and the amount of tumor cells invaded by liver cancer microvascular invasion; a report generation module, used to organize liver cancer microvascular invasion areas and the amount of tumor cells invaded by liver cancer microvascular invasion to generate and output an identification report.

[0099] The liver cancer microvascular invasion area recognition system based on pathological section images provided in this embodiment extracts the closed contour area of ​​blood vessels through the feature extraction module and the analysis and processing module, and determines whether it is located outside the liver cancer tumor area, so as to effectively confirm the target blood vessel area, and then by analyzing all target blood vessel areas, the liver cancer microvascular invasion area can be effectively and accurately located, and the amount of invaded tumor cells can be accurately known based on the data obtained from the analysis. The whole process does not require human intervention, which solves the problem of low efficiency and poor accuracy in existing liver cancer diagnosis.

[0100] It should be noted that the various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the various embodiments can be referenced to each other. In addition, the different parts between the various embodiments can also be used in combination with each other, and the present invention is not limited to this.

[0101] The present embodiment provides a method and system for identifying the microvascular invasion area of ​​liver cancer based on pathological slice images, including: obtaining a digital pathological slice image; extracting image features from the digital pathological slice image to determine whether it contains a closed vascular contour area; performing segmentation and classification processing on each closed vascular contour area to obtain cell nucleus contour information; using the cell nucleus contour information to determine whether the closed vascular contour area is located outside the liver cancer tumor area, and if so, marking the closed vascular contour area as a target vascular area; performing a comprehensive analysis on all target vascular areas to locate the microvascular invasion area of ​​liver cancer, and obtaining the amount of tumor cells invaded by the microvascular invasion of liver cancer according to the cell nucleus contour information. By extracting the closed vascular contour area and determining whether it is located outside the liver cancer tumor area, the target vascular area can be effectively confirmed, and then by analyzing all target vascular areas, the microvascular invasion area of ​​liver cancer can be effectively and accurately located, and the amount of invaded tumor cells can be accurately known based on the data obtained by the analysis, and the whole process does not require manual participation, which solves the problem of low efficiency and poor accuracy in the existing liver cancer diagnosis.

[0102] The above description is only a description of the preferred embodiments of the present invention, and is not intended to limit the scope of the present invention. Any changes or modifications made by a person skilled in the art in the field of the present invention based on the above disclosure shall fall within the scope of protection of the claims.

Claims

1. A method for identifying microvascular invasion areas of liver cancer based on pathological slice images, characterized in that: include: Acquire digital pathology slide images; Extract image features from digital pathology slice images to determine whether there is a closed blood vessel contour area; Segment and classify each blood vessel closed contour area to obtain the cell nucleus contour information; Using the cell nucleus contour information, determining whether the closed blood vessel contour region is located outside the liver cancer tumor region, and if so, marking the closed blood vessel contour region as a target blood vessel region; Comprehensive analysis of all target vascular regions was performed to locate the microvascular invasion area of ​​liver cancer, and the amount of tumor cells in the microvascular invasion of liver cancer was obtained based on the cell nuclear contour information; The method for extracting image features from a digital pathology slice image to determine whether a blood vessel closed contour area is contained comprises: Obtain all suspected closed vascular contour areas in digital pathology slice images; Analyze each suspected closed blood vessel contour area to determine whether it contains obvious image features of suspected blood vessel boundaries; If there are obvious image features of suspected blood vessel boundaries, then judging whether the suspected blood vessel closed contour region contains obvious image features of suspected cancer cells; If the image contains obvious suspected cancer cells, the suspected blood vessel closed contour region is determined as containing a blood vessel closed contour region; The method of analyzing each suspected closed blood vessel contour region to determine whether it contains obvious image features of suspected blood vessel boundaries includes: The HSV color range segmentation method is used to obtain the mask image of vascular wall cells from the suspected vascular closed contour area; The canny edge segmentation method is used to obtain the vascular edge image from the suspected vascular closed contour area; The mask image of the vascular wall cells and the vascular edge image are used to determine whether the suspected vascular closed contour area contains obvious image features of the suspected vascular boundary.

2. The method for identifying the microvascular invasion area of ​​liver cancer based on pathological slice images according to claim 1, characterized in that: The method for obtaining all suspected closed blood vessel contour areas in a digital pathology slice image comprises: The digital pathology slice image is processed into blocks to obtain a plurality of sub-slice images; Each sub-slice image is processed using the HSV color range segmentation method to obtain a sub-mask image corresponding to the sub-slice image; Concatenate all sub-mask images to obtain a full-view mask image; Obtain the suspected closed blood vessel contour area from the full-view mask image.

3. The method for identifying the microvascular invasion area of ​​liver cancer based on pathological section images according to claim 2, characterized in that: In the HSV color range segmentation method used, the lower limit of the values ​​of the three HSV channels is [0, 0, 180] and the upper limit is [180, 80, 255]. The cv2.findCountor function is used to obtain the suspected closed contour area of ​​the blood vessel from the full-view mask image.

4. The method for identifying the microvascular invasion area of ​​liver cancer based on pathological section images according to claim 3, characterized in that: In the HSV color range segmentation method, the lower limit of the values ​​of the three HSV channels is [26, 43, 46] and the upper limit is [34, 255, 255]; and the method of obtaining a vascular edge image from a suspected vascular closed contour area using the canny edge segmentation method includes: Perform grayscale and Gaussian smoothing on the suspected closed blood vessel contour area to obtain a grayscale image; Calculate the gradient and gradient direction of the grayscale image to obtain the initial edge; Utilize non-maximum suppression and double threshold screening to extract target edges from initial edges; The target edge is tracked by using the hysteresis boundary tracking to obtain the blood vessel edge image.

5. The method for identifying liver cancer microvascular invasion area based on pathological slice images according to claim 1, characterized in that: If the image feature of the suspected blood vessel boundary is obvious, the method for determining whether the suspected blood vessel closed contour region contains obvious image features of suspected cancer cells includes: The HSV color range segmentation method is used to process each suspected closed blood vessel contour region containing obvious image features of suspected blood vessel boundaries to obtain a regional mask image corresponding to the suspected closed blood vessel contour region, wherein the lower limit of the values ​​of the three HSV channels in the HSV color range segmentation method is [0, 0, 0] and the upper limit is [145, 255, 140]; The regional mask image is used to determine whether the suspected closed blood vessel contour region contains obvious image features of suspected cancer cells.

6. The method for identifying liver cancer microvascular invasion areas based on pathological slice images according to claim 1, characterized in that: The method of performing segmentation and classification processing on each blood vessel closed contour area to obtain cell nucleus contour information includes: Build and train the instance segmentation algorithm model; The instance segmentation algorithm model is used to segment and classify the closed contour area of ​​the blood vessel to obtain the contour information of the cell nucleus around and inside the blood vessel area.

7. The method for identifying liver cancer microvascular invasion area based on pathological slice images according to claim 1, characterized in that: The cell nucleus contour information includes tumor cell distribution information; the method for using the cell nucleus contour information to determine whether the blood vessel closed contour area is located outside the liver cancer tumor area includes: Obtaining a tumor cell mask according to tumor cell distribution information; Taking the closed contour area of ​​the blood vessel as the center, the mask is extended in eight directions to obtain eight blood vessel contour masks; If within the preset distance range, all blood vessel contour masks and tumor cell masks have a union, then the blood vessel closed contour region is judged to be located inside the liver cancer tumor region, otherwise the blood vessel closed contour region is judged to be located outside the liver cancer tumor region.

8. A system for identifying liver cancer microvascular invasion areas based on pathological slice images, used to implement the method for identifying liver cancer microvascular invasion areas based on pathological slice images as claimed in any one of claims 1 to 7, characterized in that: The liver cancer microvascular invasion area recognition system based on pathological slice images includes: An image acquisition module, used for acquiring digital pathology slice images; A feature extraction module is used to extract image features from the digital pathology slice image to determine whether the digital pathology slice image contains a closed blood vessel contour area; An analysis and processing module is used to process each closed blood vessel contour area to obtain the target blood vessel area, and to perform a comprehensive analysis on all target blood vessel areas to obtain the liver cancer microvascular invasion area and the amount of tumor cells invaded by the liver cancer microvascular invasion; The report generation module is used to organize the microvascular invasion area of ​​liver cancer and the amount of tumor cells invaded by microvascular invasion of liver cancer to generate an identification report and output it.