Pathological image processing method and system based on focus target area selection

Abstract

The embodiment of the invention relates to the technical field of pathological image processing, and particularly discloses a pathological image processing method and system based on focus target area selection. According to the embodiment of the invention, the method comprises the steps: obtaining a target pathological image, and preprocessing the image; performing fuzzy feature analysis, and extracting a suspicious focus image; carrying out fine segmentation processing, and extracting a target focus image; performing focus classification, grading and quantitative analysis to obtain a focus analysis report; and based on the focus analysis report, creating a target focus image and a background focus image, and synthesizing and displaying a multi-dimensional focus image. Image preprocessing, fuzzy feature analysis, fine segmentation processing, lesion classification, grading and quantitative analysis can be performed on a target pathological image, a target lesion image and a background lesion image are created, a multi-dimensional lesion image is synthesized and visually displayed, standardized processing of the pathological image is realized, the completeness of pathological image processing is improved, and the accuracy of pathological image processing is improved. And the processing process is completely automatic, the labeling cost is reduced, and popularization and use are promoted.

Description

Technical Field

[0001] The present invention belongs to the technical field of pathological image processing, and in particular, relates to a pathological image processing method and system based on lesion target area selection. Background Art

[0002] Pathological image processing refers to the process of analyzing and processing images in pathology using computer technology. In the medical field, especially in pathology, pathological image processing technology is widely used in the digital analysis of cell and tissue samples to assist diagnosis, research and teaching.

[0003] In the prior art, the pathology image processing process is usually relatively simple, and there is no standardized processing flow, which easily leads to incomplete pathology image processing. In addition, the selection process of the lesion target area usually requires manual processing, and the cost of manual labeling is high, which affects its promotion and use. Summary of the invention

[0004] The purpose of the embodiments of the present invention is to provide a pathology image processing method and system based on lesion target area selection, aiming to solve the problems raised in the background technology.

[0005] To achieve the above objectives, the embodiments of the present invention provide the following technical solutions: A pathological image processing method based on lesion target area selection, the method specifically comprises the following steps: Acquire a target pathological image, perform image preprocessing on the target pathological image, and generate a standard pathological image; Performing fuzzy feature analysis on the standard pathological image, determining a suspicious lesion area from the standard pathological image, and extracting a suspicious lesion image; Performing fine segmentation processing on the suspicious lesion image, determining the lesion segmentation boundary, and extracting the target lesion image from the suspicious lesion image; Extracting lesion features from the target lesion image to obtain lesion feature data, and performing lesion classification, grading and quantitative analysis based on the lesion feature data to obtain a lesion analysis report; Based on the lesion analysis report, a target lesion image and a background lesion image are created, and a multi-dimensional lesion image is synthesized and visually displayed.

[0006] As a further limitation of the technical solution of the embodiment of the present invention, the acquiring of the target pathological image, performing image preprocessing on the target pathological image, and generating a standard pathological image specifically include the following steps: Acquire target pathological images; De-noising the target pathological image to generate a denoised pathological image; Performing contrast enhancement on the denoised pathological image to generate a contrast pathological image; The contrast pathological image is standardized to generate a standard pathological image.

[0007] As a further limitation of the technical solution of the embodiment of the present invention, the fuzzy feature analysis of the standard pathological image, determining the suspicious lesion area from the standard pathological image, and extracting the suspicious lesion image specifically include the following steps: Performing overall fuzzy feature analysis on the standard pathological image to obtain fuzzy feature data; Based on the fuzzy feature data, locating candidate lesions on the standard pathological image to determine suspicious lesion areas; A suspicious lesion image corresponding to the suspicious lesion area is extracted from the standard pathological image.

[0008] As a further limitation of the technical solution of the embodiment of the present invention, the fine segmentation processing of the suspicious lesion image, determining the lesion segmentation boundary, and extracting the target lesion image from the suspicious lesion image specifically includes the following steps: Performing quality analysis on the suspicious lesion image to determine the lesion image quality; Selecting a target cutting algorithm according to the lesion image quality; Based on the target cutting algorithm, fine segmentation processing is performed on the suspicious lesion image to determine the lesion segmentation boundary; From the suspicious lesion image, the target lesion image corresponding to the lesion segmentation boundary is extracted.

[0009] As a further limitation of the technical solution of the embodiment of the present invention, the extracting of lesion features from the target lesion image, obtaining lesion feature data, and performing lesion classification, grading and quantitative analysis based on the lesion feature data to obtain a lesion analysis report specifically includes the following steps: Extracting lesion features from the target lesion image to obtain lesion feature data; Performing lesion classification analysis according to the lesion characteristic data to obtain a lesion classification result; Performing lesion grading analysis according to the lesion characteristic data to obtain a lesion grading result; Performing lesion quantitative analysis according to the lesion characteristic data to obtain lesion quantification results; The lesion classification result, the lesion grading result and the lesion quantification result are integrated to obtain a lesion analysis report.

[0010] As a further limitation of the technical solution of the embodiment of the present invention, the creating of the target lesion image and the background lesion image based on the lesion analysis report, and synthesizing and visually displaying the multi-dimensional lesion image specifically comprises the following steps: Based on the lesion analysis report, creating a target lesion image and a background lesion image; Determining a display position and a display ratio of the target lesion image; According to the display position and the display ratio, multi-dimensionally processing the target lesion image and the background lesion image is performed to synthesize a multi-dimensional lesion image; The multi-dimensional lesion image is visually displayed.

[0011] A pathological image processing system based on lesion target area selection, the system comprises a pathological image preprocessing unit, a fuzzy feature analysis unit, a fine segmentation processing unit, a lesion feature extraction unit and a multi-dimensional visualization display unit, wherein: A pathological image preprocessing unit, used for acquiring a target pathological image, performing image preprocessing on the target pathological image, and generating a standard pathological image; A fuzzy feature analysis unit, used to perform fuzzy feature analysis on the standard pathological image, determine the suspicious lesion area from the standard pathological image, and extract the suspicious lesion image; A fine segmentation processing unit, used to perform fine segmentation processing on the suspicious lesion image, determine the lesion segmentation boundary, and extract the target lesion image from the suspicious lesion image; A lesion feature extraction unit is used to extract lesion features from the target lesion image, obtain lesion feature data, and perform lesion classification, grading and quantitative analysis based on the lesion feature data to obtain a lesion analysis report; The multi-dimensional visualization display unit is used to create a target lesion image and a background lesion image based on the lesion analysis report, and synthesize and visualize the multi-dimensional lesion image.

[0012] As a further limitation of the technical solution of the embodiment of the present invention, the pathological image preprocessing unit specifically includes: An image acquisition module, used for acquiring a target pathological image; A noise removal module, used to remove noise from the target pathological image to generate a denoised pathological image; A contrast enhancement module, used for performing contrast enhancement on the denoised pathological image to generate a contrast pathological image; The standardization processing module is used to perform standardization processing on the contrast pathology image to generate a standard pathology image.

[0013] As a further limitation of the technical solution of the embodiment of the present invention, the fuzzy feature analysis unit specifically includes: A fuzzy feature analysis module, used for performing overall fuzzy feature analysis on the standard pathological image to obtain fuzzy feature data; A lesion candidate positioning module, used to perform lesion candidate positioning on the standard pathological image based on the fuzzy feature data to determine a suspicious lesion area; The suspicious lesion image extraction module is used to extract the suspicious lesion image corresponding to the suspicious lesion area from the standard pathological image.

[0014] As a further limitation of the technical solution of the embodiment of the present invention, the fine segmentation processing unit specifically includes: A quality analysis module, used to perform quality analysis on the suspicious lesion image to determine the quality of the lesion image; A cutting algorithm selection module, used for selecting a target cutting algorithm according to the quality of the lesion image; A fine segmentation processing module, used to perform fine segmentation processing on the suspicious lesion image based on the target cutting algorithm to determine the lesion segmentation boundary; The target lesion image extraction module is used to extract the target lesion image corresponding to the lesion segmentation boundary from the suspicious lesion image.

[0015] Compared with the prior art, the present invention has the following beneficial effects: The embodiment of the present invention obtains the target pathological image and performs image preprocessing; performs fuzzy feature analysis to extract suspicious lesion images; performs fine segmentation processing to extract target lesion images; performs lesion classification, grading and quantitative analysis to obtain a lesion analysis report; based on the lesion analysis report, creates a target lesion image and a background lesion image, synthesizes and displays a multi-dimensional lesion image. The target pathological image can be subjected to image preprocessing, fuzzy feature analysis, fine segmentation processing, lesion classification, grading and quantitative analysis, creates a target lesion image and a background lesion image, synthesizes and visually displays a multi-dimensional lesion image, realizes standardized processing of pathological images, improves the integrity of pathological image processing, and the processing process is fully automatic, reduces annotation costs, and promotes popularization and use. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present invention.

[0017] Figure 1 A flow chart of a method provided by an embodiment of the present invention is shown.

[0018] Figure 2 A flow chart of generating a standard pathological image in the method provided in an embodiment of the present invention is shown.

[0019] Figure 3 A flow chart of extracting suspicious lesion images in the method provided by an embodiment of the present invention is shown.

[0020] Figure 4 A flow chart of extracting a target lesion image in a method provided by an embodiment of the present invention is shown.

[0021] Figure 5 A flow chart of obtaining a lesion analysis report in the method provided in an embodiment of the present invention is shown.

[0022] Figure 6 A flow chart of synthesizing a multi-dimensional lesion image in the method provided in an embodiment of the present invention is shown.

[0023] Figure 7 The application architecture diagram of the system provided by the embodiment of the present invention is shown.

[0024] Figure 8 The structure block diagram of the pathological image preprocessing unit in the system provided by the embodiment of the present invention is shown.

[0025] Fig. 9 The structure block diagram of the fuzzy feature analysis unit in the system provided by the embodiment of the present invention is shown.

[0026] Fig.10 The structure block diagram of the fine segmentation processing unit in the system provided by the embodiment of the present invention is shown. DETAILED DESCRIPTION

[0027] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0028] It is understandable that in the prior art, the pathology image processing process is usually relatively simple, and there is no standardized processing flow, which easily leads to incomplete pathology image processing. In addition, the selection process of the lesion target area usually requires manual processing, and the cost of manual labeling is high, which affects its promotion and use.

[0029] To solve the above problems, the embodiment of the present invention obtains a target pathological image, performs image preprocessing on the target pathological image, and generates a standard pathological image; performs fuzzy feature analysis on the standard pathological image, determines the suspicious lesion area from the standard pathological image, and extracts the suspicious lesion image; performs fine segmentation processing on the suspicious lesion image, determines the lesion segmentation boundary, and extracts the target lesion image from the suspicious lesion image; performs lesion feature extraction on the target lesion image, obtains lesion feature data, and performs lesion classification, grading and quantitative analysis based on the lesion feature data to obtain a lesion analysis report; based on the lesion analysis report, creates a target lesion image and a background lesion image, synthesizes and visualizes a multi-dimensional lesion image. The target pathological image can be preprocessed, fuzzy feature analyzed, fine segmented, lesion classified, graded and quantitatively analyzed, creates a target lesion image and a background lesion image, synthesizes and visualizes a multi-dimensional lesion image, realizes standardized processing of pathological images, improves the integrity of pathological image processing, and the processing process is fully automatic, reduces the annotation cost, and promotes popularization and use.

[0030] Figure 1 A flow chart of a method provided by an embodiment of the present invention is shown.

[0031] Specifically, a pathological image processing method based on lesion target area selection comprises the following steps: Step S101 : acquiring a target pathological image, performing image preprocessing on the target pathological image, and generating a standard pathological image.

[0032] In an embodiment of the present invention, a denoised pathology image is generated by removing random noise and salt and pepper noise using Gaussian filtering and / or median filtering. Then, the contrast of the denoised pathology image is adjusted using methods such as histogram equalization or CLAHE to generate a contrast pathology image. After that, the contrast pathology image is standardized in size, resolution, etc. to generate a standard pathology image.

[0033] Specifically, Figure 2 A flow chart of generating a standard pathological image in the method provided in an embodiment of the present invention is shown.

[0034] Among them, in the preferred embodiment provided by the present invention, the acquiring of the target pathological image, performing image preprocessing on the target pathological image, and generating a standard pathological image specifically include the following steps: Step S1011, acquiring a target pathological image; Step S1012, removing noise from the target pathological image to generate a denoised pathological image; Step S1013, performing contrast enhancement on the denoised pathological image to generate a contrast pathological image; Step S1014: performing standardization processing on the contrast pathology image to generate a standard pathology image.

[0035] Furthermore, the pathological image processing method based on lesion target area selection also includes the following steps: Step S102, performing fuzzy feature analysis on the standard pathological image, determining a suspicious lesion area from the standard pathological image, and extracting a suspicious lesion image.

[0036] In an embodiment of the present invention, fuzzy feature data is obtained by performing an overall texture, color, shape and other fuzzy feature analysis on a standard pathological image. Based on the fuzzy feature data, methods such as sliding windows, region growing, and superpixel segmentation are used to locate candidate lesions in the standard pathological image, determine suspicious lesion areas, and then extract a suspicious lesion image corresponding to the suspicious lesion area from the standard pathological image.

[0037] Specifically, Figure 3 A flow chart of extracting suspicious lesion images in the method provided by an embodiment of the present invention is shown.

[0038] Among them, in the preferred embodiment provided by the present invention, the fuzzy feature analysis is performed on the standard pathological image, and the suspicious lesion area is determined from the standard pathological image, and the extraction of the suspicious lesion image specifically includes the following steps: Step S1021, performing overall fuzzy feature analysis on the standard pathological image to obtain fuzzy feature data; Step S1022, based on the fuzzy feature data, locating candidate lesions on the standard pathological image to determine suspicious lesion areas; Step S1023: extracting a suspicious lesion image corresponding to the suspicious lesion area from the standard pathological image.

[0039] Furthermore, performing overall fuzzy feature analysis on the standard pathological image to obtain fuzzy feature data specifically includes the following steps: The standard pathological image is blurred using the Gaussian blur kernel to obtain the blurred image. The corresponding process has the following relationship: ; in, represents a standard pathological image, represents the Gaussian blur kernel, represents the blurred image, represents the image coordinates, Indicates the current offset position on the image. Indicates that the image is at position The pixel value of Indicates The center is The Gaussian kernel is at the offset position The value of According to the blurred image, a color histogram is constructed to obtain color features. The corresponding process has the following relationship: ; in, represents the color histogram, represents the Dirac function, represents an interval of the color histogram, Represents the total number of pixels in the image. Indicates that the blurred image is at position The pixel value of The gray-level co-occurrence matrix is ​​defined using the blurred image, and the corresponding process has the following relationship: ; in, Indicates distance, Indicates the angle, Indicates distance and angle The gray-level co-occurrence matrix under Represents two different values ​​of gray levels in an image; According to the obtained gray level co-occurrence matrix, the contrast texture features, homogeneity texture features and energy texture features are calculated respectively. The corresponding process has the following relationship: ; ; ; in, represents contrast texture features, Represents homogeneous texture features, Represents energy texture features; Use the Sobel operator to perform edge detection on the blurred image in the horizontal and vertical directions to obtain the edge detection result. The corresponding process has the following relationship: ; in, represents the edge detection result, Represent the horizontal Sobel operator and the vertical Sobel operator respectively; According to the edge detection results, the sum of edge intensity, edge density and edge direction histogram are calculated respectively. The corresponding process has the following relationship: ; ; ; in, represents the sum of edge strengths, represents the edge density, represents the edge direction histogram, Indicates that the edge detection result is at position The strength value of Indicates that the edge is at position The direction angle at Indicates the specified direction angle interval; Combining color histogram, contrast texture features, homogeneity texture features, energy texture features, edge intensity sum, edge density and edge direction histogram, the fuzzy feature data is obtained. The corresponding process has the following relationship: ; in, Represents fuzzy feature data.

[0040] Furthermore, based on the fuzzy feature data, the candidate lesion locations are performed on the standard pathological image, and determining the suspicious lesion area specifically includes the following steps: Set multiple window sizes, and divide the fuzzy feature data according to different window sizes based on the sliding window operation. The corresponding process has the following relationship: ; in, Indicates the position of the sliding window at the kth scale area, Indicates the coordinates of the upper left corner of the sliding window. Indicates the size of the window; The average value of all features in the divided window is taken as the comprehensive feature of the window. The corresponding process has the following relationship: ; in, Indicates the comprehensive characteristics of the window; The window comprehensive features of different windows are fused to obtain a multi-scale comprehensive feature vector. The corresponding process has the following relationship: ; in, represents the multi-scale comprehensive feature vector, represents the number of scales, represents the weight of the kth scale; The K-means clustering algorithm is used to classify similar regions in the multi-scale comprehensive feature vector into the same category. The corresponding process has the following relationship: ; in, represents the set of clusters after clustering, represents the number of clusters, Represents the K-means clustering algorithm; Given a fuzzy feature template representing typical lesion features, the matching degree between each cluster and the fuzzy feature template is calculated. The corresponding process has the following relationship: ; in, Indicates the matching degree, represents the fuzzy feature template, Represents a distance metric operation; Set the matching threshold and select the windows contained in the clusters with high matching degree as seed points for region growing. The corresponding process has the following relationship: ; ; in, represents the initial seed point, represents the matching threshold, represents the growth area, Represents a growth operation; The SLIC algorithm is used to perform superpixel segmentation on the standard pathological image to obtain a superpixel set. The corresponding process has the following relationship: ; in, represents a superpixel set, represents the number of superpixels, represents the compactness parameter, represents the SLIC algorithm; By merging the superpixel set with the superpixels that intersect the growing area, the candidate area is updated and expanded, and the suspicious lesion area is determined. The corresponding process has the following relationship: ; in, Indicates suspicious lesion areas. Represents the empty set.

[0041] Furthermore, the pathological image processing method based on lesion target area selection also includes the following steps: Step S103, performing fine segmentation processing on the suspicious lesion image, determining the lesion segmentation boundary, and extracting the target lesion image from the suspicious lesion image.

[0042] In an embodiment of the present invention, the quality of the suspicious lesion image is determined by performing a quality analysis on the suspicious lesion image, and then a target cutting algorithm is selected based on the lesion image quality, which may be Graph Cut, Level Set, U-Net, Mask R-CNN, etc., and then based on the target cutting algorithm, the suspicious lesion image is finely segmented to determine the lesion segmentation boundary, and then the target lesion image corresponding to the lesion segmentation boundary is extracted from the suspicious lesion image.

[0043] Specifically, Figure 4 A flow chart of extracting a target lesion image in a method provided by an embodiment of the present invention is shown.

[0044] Among them, in the preferred embodiment provided by the present invention, the fine segmentation processing of the suspicious lesion image, determining the lesion segmentation boundary, and extracting the target lesion image from the suspicious lesion image specifically includes the following steps: Step S1031, performing quality analysis on the suspicious lesion image to determine the lesion image quality; Step S1032, selecting a target cutting algorithm according to the lesion image quality; Step S1033, based on the target cutting algorithm, fine segmentation processing is performed on the suspicious lesion image to determine the lesion segmentation boundary; Step S1034: extracting a target lesion image corresponding to the lesion segmentation boundary from the suspicious lesion image.

[0045] Furthermore, extracting the target lesion image corresponding to the lesion segmentation boundary from the suspicious lesion image specifically includes the following steps: Generate a binary mask using the finely segmented boundary, multiply the mask by the suspicious lesion image pixel by pixel, and extract the lesion area image containing only the lesion area; Adaptive histogram equalization is used to adjust the contrast of the lesion area image, Laplace sharpening is used to improve the edge clarity and detail performance of the lesion area, so as to obtain an enhanced lesion area image, and color correction is performed on the enhanced lesion area image to obtain a corrected lesion area image; The Patch-based texture synthesis method fills the missing parts in the lesion area of ​​the corrected lesion area image to ensure the integrity of the area, and then uses the Gabor filter to enhance the texture features to obtain the target lesion image.

[0046] Furthermore, the pathological image processing method based on lesion target area selection also includes the following steps: Step S104, extracting lesion features from the target lesion image to obtain lesion feature data, and performing lesion classification, grading and quantitative analysis based on the lesion feature data to obtain a lesion analysis report.

[0047] In an embodiment of the present invention, lesion features such as texture, shape, size, and grayscale distribution are extracted from the target lesion image to obtain lesion feature data, and then lesion classification analysis is performed based on the lesion feature data using machine learning algorithms such as support vector machine (SVM) and random forest to obtain lesion classification results, lesion grading analysis is performed to obtain lesion grading results, and lesion quantification analysis is performed to obtain lesion quantification results. A lesion analysis report is obtained by integrating the lesion classification results, lesion grading results, and lesion quantification results.

[0048] Specifically, Figure 5 A flow chart of obtaining a lesion analysis report in the method provided in an embodiment of the present invention is shown.

[0049] Among them, in the preferred embodiment provided by the present invention, the lesion feature extraction is performed on the target lesion image to obtain lesion feature data, and lesion classification, grading and quantitative analysis are performed according to the lesion feature data to obtain a lesion analysis report, which specifically includes the following steps: Step S1041, extracting lesion features from the target lesion image to obtain lesion feature data; Step S1042, performing lesion classification analysis based on the lesion feature data to obtain a lesion classification result; Step S1043, performing lesion grading analysis according to the lesion characteristic data to obtain a lesion grading result; Step S1044, performing lesion quantitative analysis according to the lesion characteristic data to obtain a lesion quantification result; Step S1045, combining the lesion classification result, the lesion grading result and the lesion quantification result to obtain a lesion analysis report.

[0050] Furthermore, the pathological image processing method based on lesion target area selection also includes the following steps: Step S105, based on the lesion analysis report, creating a target lesion image and a background lesion image, synthesizing and visually displaying a multi-dimensional lesion image.

[0051] In an embodiment of the present invention, a target lesion image and a background lesion image are created based on a lesion analysis report, and then the display position and display ratio of the target lesion image are determined. Then, according to the display position and display ratio, the target lesion image and the background lesion image are multi-dimensionally processed to synthesize a multi-dimensional lesion image. When there is a display requirement, the multi-dimensional lesion image is visualized.

[0052] It can be understood that the target lesion image is a target lesion image combined with a partial lesion analysis report; the background lesion image is a standard pathological image combined with a partial lesion analysis report.

[0053] Specifically, Figure 6 A flow chart of synthesizing a multi-dimensional lesion image in the method provided in an embodiment of the present invention is shown.

[0054] Among them, in the preferred embodiment provided by the present invention, the creation of the target lesion image and the background lesion image based on the lesion analysis report, and the synthesis and visual display of the multi-dimensional lesion image specifically include the following steps: Step S1051, creating a target lesion image and a background lesion image based on the lesion analysis report; Step S1052, determining the display position and display ratio of the target lesion image; Step S1053, performing multi-dimensional processing on the target lesion image and the background lesion image according to the display position and the display ratio to synthesize a multi-dimensional lesion image; Step S1054: Visually display the multi-dimensional lesion image.

[0055] Furthermore, Figure 7 The application architecture diagram of the system provided by the embodiment of the present invention is shown.

[0056] Among them, in another preferred embodiment provided by the present invention, a pathological image processing system based on lesion target area selection includes: The pathological image preprocessing unit 101 is used to acquire a target pathological image, perform image preprocessing on the target pathological image, and generate a standard pathological image.

[0057] In an embodiment of the present invention, the pathological image preprocessing unit 101 uses Gaussian filtering and / or median filtering to remove random noise and salt and pepper noise from the target pathological image to generate a denoised pathological image, and then uses methods such as histogram equalization or CLAHE to adjust the contrast of the denoised pathological image to generate a contrast pathological image. Thereafter, the contrast pathological image is standardized in size, resolution, etc. to generate a standard pathological image.

[0058] Specifically, Figure 8 The structure block diagram of the pathological image preprocessing unit 101 in the system provided by the embodiment of the present invention is shown.

[0059] Among them, in the preferred embodiment provided by the present invention, the pathological image preprocessing unit 101 specifically includes: An image acquisition module 1011 is used to acquire a target pathological image; A noise removal module 1012 is used to remove noise from the target pathological image to generate a denoised pathological image; A contrast enhancement module 1013 is used to perform contrast enhancement on the denoised pathological image to generate a contrast pathological image; The standardization processing module 1014 is used to perform standardization processing on the contrast pathology image to generate a standard pathology image.

[0060] Furthermore, the pathological image processing system based on lesion target area selection also includes: The fuzzy feature analysis unit 102 is used to perform fuzzy feature analysis on the standard pathological image, determine the suspicious lesion area from the standard pathological image, and extract the suspicious lesion image.

[0061] In an embodiment of the present invention, the fuzzy feature analysis unit 102 obtains fuzzy feature data by performing an overall texture, color, shape and other fuzzy feature analysis on the standard pathological image, and based on the fuzzy feature data, uses methods such as sliding windows, region growing, and superpixel segmentation to locate candidate lesions in the standard pathological image, determine suspicious lesion areas, and then extract suspicious lesion images corresponding to the suspicious lesion areas from the standard pathological image.

[0062] Specifically, Fig. 9 The structure block diagram of the fuzzy feature analysis unit 102 in the system provided by the embodiment of the present invention is shown.

[0063] Among them, in the preferred embodiment provided by the present invention, the fuzzy feature analysis unit 102 specifically includes: A fuzzy feature analysis module 1021 is used to perform overall fuzzy feature analysis on the standard pathological image to obtain fuzzy feature data; A lesion candidate positioning module 1022 is used to perform lesion candidate positioning on the standard pathological image based on the fuzzy feature data to determine a suspicious lesion area; The suspicious lesion image extraction module 1023 is used to extract the suspicious lesion image corresponding to the suspicious lesion area from the standard pathological image.

[0064] Furthermore, the pathological image processing system based on lesion target area selection also includes: The fine segmentation processing unit 103 is used to perform fine segmentation processing on the suspicious lesion image, determine the lesion segmentation boundary, and extract the target lesion image from the suspicious lesion image.

[0065] In an embodiment of the present invention, the fine segmentation processing unit 103 determines the quality of the lesion image by performing a quality analysis on the suspicious lesion image, and then selects a target cutting algorithm based on the lesion image quality, which may be Graph Cut, Level Set, U-Net, Mask R-CNN, etc., and then performs fine segmentation processing on the suspicious lesion image based on the target cutting algorithm to determine the lesion segmentation boundary, and then extracts the target lesion image corresponding to the lesion segmentation boundary from the suspicious lesion image.

[0066] Specifically, Fig.10 It shows a structural block diagram of the fine segmentation processing unit 103 in the system provided by an embodiment of the present invention.

[0067] In a preferred embodiment of the present invention, the fine segmentation processing unit 103 specifically includes: The quality analysis module 1031 is used to perform quality analysis on the suspicious lesion image to determine the quality of the lesion image; A cutting algorithm selection module 1032 is used to select a target cutting algorithm according to the quality of the lesion image; A fine segmentation processing module 1033 is used to perform fine segmentation processing on the suspicious lesion image based on the target cutting algorithm to determine the lesion segmentation boundary; The target lesion image extraction module 1034 is used to extract the target lesion image corresponding to the lesion segmentation boundary from the suspicious lesion image.

[0068] Furthermore, the pathological image processing system based on lesion target area selection also includes: The lesion feature extraction unit 104 is used to extract lesion features from the target lesion image, obtain lesion feature data, and perform lesion classification, grading and quantitative analysis based on the lesion feature data to obtain a lesion analysis report.

[0069] In the embodiment of the present invention, the lesion feature extraction unit 104 obtains lesion feature data by extracting lesion features such as texture, shape, size, and grayscale distribution from the target lesion image, and then performs lesion classification analysis based on the lesion feature data through machine learning algorithms such as support vector machine (SVM) and random forest to obtain lesion classification results, performs lesion grading analysis to obtain lesion grading results, and performs lesion quantification analysis to obtain lesion quantification results. A lesion analysis report is obtained by integrating the lesion classification results, lesion grading results, and lesion quantification results.

[0070] The multi-dimensional visualization display unit 105 is used to create a target lesion image and a background lesion image based on the lesion analysis report, and synthesize and visualize a multi-dimensional lesion image.

[0071] In an embodiment of the present invention, the multidimensional visualization display unit 105 creates a target lesion image and a background lesion image based on the lesion analysis report, and then determines the display position and display ratio of the target lesion image, and then performs multidimensional processing on the target lesion image and the background lesion image according to the display position and the display ratio to synthesize a multidimensional lesion image. When there is a display requirement, the multidimensional visualization display unit 105 visualizes the multidimensional lesion image.

[0072] It should be understood that, although each step in the flow chart of each embodiment of the present invention is shown in sequence according to the indication of the arrow, these steps are not necessarily performed in sequence according to the order indicated by the arrow. Unless there is a clear explanation in this article, the execution of these steps does not have a strict order restriction, and these steps can be performed in other orders. Moreover, at least a portion of the steps in each embodiment may include a plurality of sub-steps or a plurality of stages, and these sub-steps or stages are not necessarily performed at the same time, but can be performed at different times, and the execution order of these sub-steps or stages is not necessarily performed in sequence, but can be performed in turn or alternately with at least a portion of other steps or sub-steps or stages of other steps.

[0073] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the program can be stored in a non-volatile computer-readable storage medium. When the program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application may include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

[0074] The technical features of the above-described embodiments may be arbitrarily combined. To make the description concise, not all possible combinations of the technical features in the above-described embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0075] The above-mentioned embodiments only express several implementation methods of the present invention, and the description thereof is relatively specific and detailed, but it cannot be understood as limiting the scope of the patent of the present invention. It should be pointed out that, for ordinary technicians in this field, several variations and improvements can be made without departing from the concept of the present invention, which all belong to the protection scope of the present invention. Therefore, the protection scope of the patent of the present invention shall be subject to the attached claims.

[0076] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A pathological image processing method based on lesion target area selection, characterized in that: The method specifically comprises the following steps: Acquire a target pathological image, perform image preprocessing on the target pathological image, and generate a standard pathological image; Performing fuzzy feature analysis on the standard pathological image, determining a suspicious lesion area from the standard pathological image, and extracting a suspicious lesion image; Performing fine segmentation processing on the suspicious lesion image, determining the lesion segmentation boundary, and extracting the target lesion image from the suspicious lesion image; Extracting lesion features from the target lesion image to obtain lesion feature data, and performing lesion classification, grading and quantitative analysis based on the lesion feature data to obtain a lesion analysis report; Based on the lesion analysis report, a target lesion image and a background lesion image are created, and a multi-dimensional lesion image is synthesized and visually displayed.

2. The pathological image processing method based on lesion target area selection according to claim 1 is characterized in that: The step of acquiring a target pathological image, performing image preprocessing on the target pathological image, and generating a standard pathological image specifically comprises the following steps: Acquire target pathological images; Denoising the target pathological image to generate a denoised pathological image; Performing contrast enhancement on the denoised pathological image to generate a contrast pathological image; The contrast pathological image is standardized to generate a standard pathological image.

3. The pathological image processing method based on lesion target area selection according to claim 2 is characterized in that: The step of performing fuzzy feature analysis on the standard pathological image, determining a suspicious lesion area from the standard pathological image, and extracting a suspicious lesion image specifically comprises the following steps: Performing overall fuzzy feature analysis on the standard pathological image to obtain fuzzy feature data; Based on the fuzzy feature data, locating candidate lesions on the standard pathological image to determine suspicious lesion areas; A suspicious lesion image corresponding to the suspicious lesion area is extracted from the standard pathological image.

4. The pathological image processing method based on lesion target area selection according to claim 3 is characterized in that: Performing overall fuzzy feature analysis on the standard pathological image to obtain fuzzy feature data specifically includes the following steps: The standard pathological image is blurred using the Gaussian blur kernel to obtain the blurred image. The corresponding process has the following relationship: ； in, represents a standard pathological image, represents the Gaussian blur kernel, represents the blurred image, represents the image coordinates, Indicates the current offset position on the image. Indicates that the image is at position The pixel value of Indicates The center is The Gaussian kernel is at the offset position The value of According to the blurred image, a color histogram is constructed to obtain color features. The corresponding process has the following relationship: ； in, represents the color histogram, represents the Dirac function, represents an interval of the color histogram, Represents the total number of pixels in the image. Indicates that the blurred image is at position The pixel value of The gray-level co-occurrence matrix is ​​defined using the blurred image, and the corresponding process has the following relationship: ； in, Indicates distance, Indicates the angle, Indicates distance and angle The gray-level co-occurrence matrix under Represents two different values ​​of gray levels in an image; According to the obtained gray level co-occurrence matrix, the contrast texture features, homogeneity texture features and energy texture features are calculated respectively. The corresponding process has the following relationship: ； ； ； in, represents contrast texture features, Represents homogeneous texture features, Represents energy texture features; Use the Sobel operator to perform edge detection on the blurred image in the horizontal and vertical directions to obtain the edge detection result. The corresponding process has the following relationship: ； in, represents the edge detection result, Represent the horizontal Sobel operator and the vertical Sobel operator respectively; According to the edge detection results, the sum of edge intensity, edge density and edge direction histogram are calculated respectively. The corresponding process has the following relationship: ； ； ； in, represents the sum of edge strengths, represents the edge density, represents the edge direction histogram, Indicates that the edge detection result is at position The strength value of Indicates that the edge is at position The direction angle at Indicates the specified direction angle range; Combining color histogram, contrast texture features, homogeneity texture features, energy texture features, edge intensity sum, edge density and edge direction histogram, the fuzzy feature data is obtained. The corresponding process has the following relationship: ； in, Represents fuzzy feature data.

5. The pathological image processing method based on lesion target area selection according to claim 4 is characterized in that: Based on the fuzzy feature data, the candidate lesion locations are performed on the standard pathological image, and the suspicious lesion area is determined, which specifically includes the following steps: Set multiple window sizes, and divide the fuzzy feature data according to different window sizes based on the sliding window operation. The corresponding process has the following relationship: ； in, Indicates the position of the sliding window at the kth scale area, Indicates the coordinates of the upper left corner of the sliding window. Indicates the size of the window; The average value of all features in the divided window is taken as the comprehensive feature of the window. The corresponding process has the following relationship: ； in, Indicates the comprehensive characteristics of the window; The window comprehensive features of different windows are fused to obtain a multi-scale comprehensive feature vector. The corresponding process has the following relationship: ； in, represents the multi-scale comprehensive feature vector, represents the number of scales, represents the weight of the kth scale; The K-means clustering algorithm is used to classify similar regions in the multi-scale comprehensive feature vector into the same category. The corresponding process has the following relationship: ； in, represents the set of clusters after clustering, represents the number of clusters, Represents the K-means clustering algorithm; Given a fuzzy feature template representing typical lesion features, the matching degree between each cluster and the fuzzy feature template is calculated. The corresponding process has the following relationship: ； in, Indicates the matching degree, represents the fuzzy feature template, Represents a distance metric operation; Set the matching threshold and select the windows contained in the clusters with high matching degree as seed points for region growing. The corresponding process has the following relationship: ； ； in, represents the initial seed point, represents the matching threshold, represents the growth area, Represents a growth operation; The SLIC algorithm is used to perform superpixel segmentation on the standard pathological image to obtain a superpixel set. The corresponding process has the following relationship: ； in, represents a superpixel set, represents the number of superpixels, represents the compactness parameter, represents the SLIC algorithm; By merging the superpixel set with the superpixels that intersect the growing area, the candidate area is updated and expanded, and the suspicious lesion area is determined. The corresponding process has the following relationship: ； in, Indicates suspicious lesion areas. Represents the empty set.

6. The pathological image processing method based on lesion target area selection according to claim 5, characterized in that: The step of performing fine segmentation processing on the suspicious lesion image, determining the lesion segmentation boundary, and extracting the target lesion image from the suspicious lesion image specifically comprises the following steps: Performing quality analysis on the suspicious lesion image to determine the lesion image quality; Selecting a target cutting algorithm according to the lesion image quality; Based on the target cutting algorithm, fine segmentation processing is performed on the suspicious lesion image to determine the lesion segmentation boundary; From the suspicious lesion image, the target lesion image corresponding to the lesion segmentation boundary is extracted.

7. The pathological image processing method based on lesion target area selection according to claim 6 is characterized in that: Extracting the target lesion image corresponding to the lesion segmentation boundary from the suspicious lesion image specifically includes the following steps: Generate a binary mask using the finely segmented boundary, multiply the mask by the suspicious lesion image pixel by pixel, and extract the lesion area image containing only the lesion area; Adaptive histogram equalization is used to adjust the contrast of the lesion area image, Laplace sharpening is used to improve the edge clarity and detail performance of the lesion area, so as to obtain an enhanced lesion area image, and color correction is performed on the enhanced lesion area image to obtain a corrected lesion area image; The Patch-based texture synthesis method fills the missing parts in the lesion area of ​​the corrected lesion area image to ensure the integrity of the area, and then uses the Gabor filter to enhance the texture features to obtain the target lesion image.

8. The pathological image processing method based on lesion target area selection according to claim 7, characterized in that: The extracting of lesion features from the target lesion image, acquiring lesion feature data, and performing lesion classification, grading and quantitative analysis based on the lesion feature data to obtain a lesion analysis report specifically comprises the following steps: Extracting lesion features from the target lesion image to obtain lesion feature data; Performing lesion classification analysis according to the lesion characteristic data to obtain a lesion classification result; Performing lesion grading analysis according to the lesion characteristic data to obtain a lesion grading result; Performing lesion quantitative analysis according to the lesion characteristic data to obtain lesion quantification results; The lesion classification result, the lesion grading result and the lesion quantification result are integrated to obtain a lesion analysis report.

9. The pathological image processing method based on lesion target area selection according to claim 8, characterized in that: The step of creating a target lesion image and a background lesion image based on the lesion analysis report, and synthesizing and visually displaying a multi-dimensional lesion image specifically comprises the following steps: Based on the lesion analysis report, creating a target lesion image and a background lesion image; Determining a display position and a display ratio of the target lesion image; According to the display position and the display ratio, multi-dimensionally processing the target lesion image and the background lesion image is performed to synthesize a multi-dimensional lesion image; The multi-dimensional lesion image is visually displayed.

10. A pathological image processing system based on lesion target area selection, the system applying the pathological image processing method based on lesion target area selection according to any one of claims 1 to 9, characterized in that: The system includes a pathological image preprocessing unit, a fuzzy feature analysis unit, a fine segmentation processing unit, a lesion feature extraction unit and a multi-dimensional visualization display unit, wherein: A pathological image preprocessing unit, used for acquiring a target pathological image, performing image preprocessing on the target pathological image, and generating a standard pathological image; A fuzzy feature analysis unit, used to perform fuzzy feature analysis on the standard pathological image, determine the suspicious lesion area from the standard pathological image, and extract the suspicious lesion image; A fine segmentation processing unit, used to perform fine segmentation processing on the suspicious lesion image, determine the lesion segmentation boundary, and extract the target lesion image from the suspicious lesion image; A lesion feature extraction unit is used to extract lesion features from the target lesion image, obtain lesion feature data, and perform lesion classification, grading and quantitative analysis based on the lesion feature data to obtain a lesion analysis report; The multi-dimensional visualization display unit is used to create a target lesion image and a background lesion image based on the lesion analysis report, and synthesize and visualize the multi-dimensional lesion image.

Images