Endometrial data extraction method, device and equipment and readable storage medium

Abstract

The invention discloses an endometrium data extraction method, device and equipment and a readable storage medium, and is applied to the technical field of data processing, and the method comprises the steps: carrying out the processing of original uterus body data, and obtaining a superpixel image; the superpixel image comprises different pixel blocks; different pixel blocks in the super-pixel image are aggregated through iteration, so that the super-pixel image is segmented, and a segmentation result image is obtained; according to attribute information of different blocks in the segmentation result image, performing intima screening on the segmentation result image to obtain a target intima area; and traversing the original uterine body data according to the target intimal region to obtain endometrial volume data. Compared with the prior art that original uterus body data is directly segmented and measured manually, the method has the advantages that the original uterus body data can be intelligently processed, endometrial volume data closely related to the endometrial can be obtained, and the accuracy and efficiency of endometrial receptivity analysis can be improved due to the fact that manual measurement is not needed.

Description

Technical Field

[0001] The present invention relates to the field of data processing technology, and in particular to a method, device, equipment and readable storage medium for extracting endometrial data. Background Art

[0002] Endometrial receptivity refers to the ability of an embryo to be accepted and implanted in the endometrium. In vitro fertilization-embryo transfer (IVF-ET) is a common method for solving infertility problems. The success rate of this method depends to a certain extent on the receptivity of the endometrium. Ultrasound-based endometrial receptivity assessment methods usually perform endometrial segmentation and measurement in two dimensions, or use data obtained by chemical methods to determine endometrial receptivity. Due to the high subjectivity of this method, less experienced doctors often cannot measure it accurately.

[0003] Therefore, the current technical solution for extracting endometrial related data is mainly based on manual extraction, resulting in relatively low accuracy of endometrial data extraction. Summary of the invention

[0004] In view of this, an object of the present invention is to provide a method, device, equipment and readable storage medium for extracting endometrial data, which solve the technical problems of low accuracy and efficiency of endometrial data extraction in the prior art.

[0005] In order to solve the above technical problems, the present invention provides a method for extracting endometrial data, comprising:

[0006] Processing the original uterine body data to obtain a super-pixel image; wherein the super-pixel image includes different pixel blocks;

[0007] Iteratively aggregating the different pixel blocks in the super-pixel image to segment the super-pixel image to obtain a segmentation result image;

[0008] According to the attribute information of different blocks in the segmentation result image, the segmentation result image is screened for intima to obtain a target intima area;

[0009] The original uterine body data is traversed according to the target endometrial region to obtain endometrial volume data.

[0010] Optionally, the processing of the original uterine body data to obtain a super-pixel image includes:

[0011] Performing cutting processing on the original uterine body data to obtain section data;

[0012] Preprocessing the section data to obtain preprocessed section image data; wherein the preprocessing includes at least one of edge extraction, filtering and noise reduction;

[0013] Perform super-pixel segmentation on the image corresponding to the pre-processed section image data to obtain the super-pixel image.

[0014] Optionally, performing intima screening on the segmentation result image according to the attribute information of different blocks in the segmentation result image to obtain a target intima region includes:

[0015] Different blocks in the segmentation result image are screened according to the attribute information to determine the blocks that meet the intima screening conditions, and the target intima area is obtained based on the screened blocks; wherein the attribute information includes at least one of area, minimum bounding box, center of gravity and center coordinates.

[0016] Optionally, screening different blocks in the segmentation result image according to the attribute information to determine blocks that meet the intima screening condition, and obtaining the target intima region based on the screened blocks includes:

[0017] Obtaining the inner membrane attribute threshold corresponding to each of the attribute information;

[0018] Comparing the value of each of the attribute information with the intima attribute threshold, determining the block that meets the intima screening condition according to the comparison result and updating the mark value of the block in the segmentation result image;

[0019] Screening is performed according to the tag value of the block to obtain the target endometrial region.

[0020] Optionally, comparing the value of each of the attribute information with the intima attribute threshold, determining the block that meets the intima screening condition according to the comparison result and updating the label value of the block in the segmentation result image, includes:

[0021] The blocks with an area smaller than the intima area threshold are set to 0;

[0022] and / or setting the blocks whose minimum bounding box perimeter is less than the inner membrane minimum bounding box perimeter threshold to 0;

[0023] and / or setting the block whose mark value at the block center is not equal to the block mark value to 0;

[0024] In the remaining blocks, the block whose center coordinate is the smallest distance from the preset effective center of the image is set to 1, and the remaining blocks are set to 0.

[0025] Optionally, performing intima screening on the segmentation result image to obtain a target intima region includes:

[0026] Preliminarily screening the segmentation result image to obtain candidate intima regions;

[0027] The candidate intimal regions are classified based on machine learning or deep learning to determine the target intimal region.

[0028] Optionally, after the segmentation result image is preliminarily screened to obtain a candidate intima region, the method further includes:

[0029] Extracting features of the candidate intima region to obtain intima region feature data;

[0030] Accordingly, the classifying the candidate intimal regions based on machine learning or deep learning to determine the target intimal region includes:

[0031] The target intimal region is determined by classifying the intimal region feature data based on the machine learning or the deep learning.

[0032] Optionally, the iteratively aggregating the different pixel blocks in the super-pixel image to segment the super-pixel image to obtain a segmentation result image includes:

[0033] Iteratively aggregating the different pixel blocks in the super-pixel image so as to minimize the objective function, thereby segmenting the super-pixel image to obtain the segmentation result image;

[0034] Among them, the objective function is l is the pixel block label value 1≤l≤q, q represents the number of pixel blocks in the superpixel image, c represents the number of types, p represents the pixel, S l Indicates the number of pixels in the lth pixel block area, x p is the grayscale value of the lth pixel block area, v k represents the cluster center, u kl represents the membership coefficient between the pixel block with the label value l and the kth cluster center, m represents the membership coefficient index, k represents the kth cluster center, R l Represents the set of all pixel blocks.

[0035] Optionally, traversing the original uterine body data according to the target endometrial region to obtain endometrial volume data includes:

[0036] The endometrial data extraction method is used to traverse the remaining area of ​​the original uterine body data to obtain the endometrial volume data corresponding to the target endometrial area.

[0037] Optionally, traversing the original uterine body data according to the target endometrial region to obtain endometrial volume data includes:

[0038] The original uterine body data is traversed according to the contour of the target endometrial area to perform adjacent contour detection to obtain the endometrial volume data.

[0039] Optionally, after traversing the original uterine body data according to the target endometrial region to obtain endometrial volume data, the method further includes:

[0040] Performing endometrial analysis according to the endometrial volume data to obtain endometrial target feature data;

[0041] Endometrial receptivity is determined based on the endometrial target characteristic data.

[0042] The present invention also provides an endometrial data extraction device, comprising:

[0043] A super-pixel image determination module is used to process the original uterine body data to obtain a super-pixel image; wherein the super-pixel image includes different pixel blocks;

[0044] An image segmentation module, configured to iteratively aggregate the different pixel blocks in the super-pixel image to segment the super-pixel image to obtain a segmentation result image;

[0045] An intimal screening module, used for screening the intimal region of the segmentation result image according to the attribute information of different blocks in the segmentation result image, to obtain a target intimal region;

[0046] The volume data determination module is used to traverse the original uterine body data according to the target endometrial area to obtain endometrial volume data.

[0047] The present invention also provides an endometrial data extraction device, comprising:

[0048] Memory for storing computer programs;

[0049] A processor is used to implement the steps of the above-mentioned endometrial data extraction method when executing the computer program.

[0050] The present invention also provides a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, the steps of the above-mentioned endometrial data extraction method are implemented.

[0051] It can be seen that the present invention obtains a superpixel image by processing the original uterine body data; iteratively aggregates the different pixel blocks in the superpixel image to divide the superpixel image into different blocks to obtain a segmentation result image; according to the attribute information of different blocks in the segmentation result image, the segmentation result image is screened for endometrium to obtain a target endometrial region; according to the target endometrial region, the original uterine body data is traversed to obtain endometrial volume data. Compared with the current manual segmentation and measurement of the endometrium obtained from the original uterine body data to determine the endometrial receptivity, the present invention can process the original uterine body data to obtain endometrial volume data that is closely related to the endometrial receptivity, thereby directly determining the endometrial receptivity based on the endometrial volume data. Therefore, compared with the manual measurement method, the present invention can intelligently determine the endometrial volume data, thereby improving the accuracy and efficiency of determining the endometrial receptivity.

[0052] In addition, the present invention also provides an endometrial data extraction device, equipment and readable storage medium, which also have the above-mentioned beneficial effects. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying creative work.

[0054] Figure 1 A flow chart of a method for extracting endometrial data provided by an embodiment of the present invention;

[0055] Figure 2 An example flow chart of an endometrial data extraction method provided by an embodiment of the present invention;

[0056] Figure 3 A flow chart of an endometrial data extraction method provided by an embodiment of the present invention;

[0057] Figure 4 A schematic diagram of the structure of an endometrial data extraction device provided by an embodiment of the present invention;

[0058] Figure 5 A schematic structural diagram of an endometrial data extraction device provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0059] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0060] Please refer to Figure 1 , Figure 1 A flowchart of an endometrial data extraction method provided in an embodiment of the present invention. The method can be applied to electronic devices such as ultrasound equipment and personal computers. That is, after the electronic device obtains the uterine body data, the endometrial data is extracted. It should be noted that when the electronic device is not an ultrasound device, it can obtain relevant uterine body data from the ultrasound device, and after the endometrial data is extracted, it can send it to the ultrasound device.

[0061] The method may include:

[0062] S100, processing the original uterine body data to obtain a super-pixel image; wherein the super-pixel image includes different pixel blocks.

[0063] This embodiment does not limit the specific type of the original uterine body data. For example, the three-dimensional volume data corresponding to the uterus. The uterine area can be scanned three-dimensionally by an ultrasonic device to obtain the corresponding volume data, which is the original uterine body data. This embodiment does not limit the specific process of processing the original uterine body data to obtain the super-pixel image. It can be that the image corresponding to the original uterine body data is subjected to super-pixel segmentation to obtain the super-pixel image. For example, the image corresponding to the original uterine body data can be subjected to edge processing, segmented to obtain a gradient map, and the gradient map is iteratively calculated to obtain a super-pixel image; or it can be that the image corresponding to the uterine body data is subjected to edge processing and filtering, and then processed using a watershed-based super-pixel to obtain a super-pixel image. It can be understood that the image corresponding to the original uterine body data can be subjected to image preprocessing, and the image preprocessing includes edge extraction, filtering processing and noise reduction processing to obtain the boundary information of the section of the uterine ultrasound image, which provides a basis for subsequent super-pixel segmentation. Common edge extraction methods include but are not limited to conventional edge detection algorithms such as canny (multi-stage edge algorithm) and sobel (Sobel operator), and multiple algorithms can be used for edge extraction here. After image processing of the image corresponding to the uterine body data, section processing and segmentation processing can be performed to obtain a gradient map, and a super-pixel image is obtained by iterative calculation. Super-pixel segmentation implementation methods include but are not limited to: watershed-based super-pixels, SLIC (super-image segmentation algorithm) super-pixels, etc. The processed image is segmented by a method based on machine learning or deep learning to achieve pixel block detection to segment different pixel blocks, including but not limited to: clustering, support vector machine, convolutional neural network, recurrent neural network, etc. Different label values ​​can be assigned to different pixel blocks. Therefore, the original uterine section image can obtain a super-pixel image containing different pixel blocks through the above steps, and different pixel blocks are labeled with different values. Before obtaining the super-pixel image, the section data will be subjected to a variety of preprocessing to improve the accuracy of super-pixel image acquisition.

[0064] It should be noted that the above-mentioned processing of the original uterine body data to obtain the super-pixel image may include: cutting the original uterine body data to obtain section data; pre-processing the section data to obtain pre-processed section image data; wherein the pre-processing includes at least one of edge extraction, filtering and noise reduction; and super-pixel segmentation of the image corresponding to the pre-processed section image data to obtain the super-pixel image. In this embodiment, since the section data is pre-processed after the original uterine body data is cut to obtain the section data, the section data is pre-processed in a variety of ways, so the comprehensiveness of the processing of the original uterine body data can be improved, and the accuracy of the endometrial receptivity analysis can be improved.

[0065] S101, iteratively aggregating different pixel blocks in a super-pixel image to segment the super-pixel image to obtain a segmentation result image.

[0066] This embodiment iteratively aggregates different pixel blocks in the superpixel image to segment the superpixel image, and when the iteration stop condition is reached, the superpixel image is segmented into different blocks (also called initial intimal regions) to obtain a segmentation result image including the blocks. This process can be understood as a process of preliminarily determining the intimal region in the superpixel image, that is, a process of detecting the intimal region.

[0067] It should be noted that the above-mentioned iteratively aggregating different pixel blocks in the super-pixel image to segment the super-pixel image to obtain a segmentation result image may include: iteratively aggregating different pixel blocks in the super-pixel image to minimize the objective function to segment the super-pixel image into different blocks to obtain a segmentation result image; wherein the objective function is l is the pixel block label value 1≤l≤q, q represents the number of pixel blocks in the superpixel image, c represents the number of types, p represents the pixel, S l Indicates the number of pixels in the lth pixel block area, x p is the gray value of the lth pixel block area, vk represents the cluster center, ukl represents the membership coefficient between the pixel block with the label value l and the kth cluster center, m represents the membership coefficient index, k represents the kth cluster center, R l Represents the set of all pixel blocks.

[0068] For ease of understanding, the specific clustering implementation process provided in this embodiment is described as follows:

[0069] 1) Objective function: The images are divided into c categories by clustering. The objective function is as follows:

[0070]

[0071] Where l is the pixel block label value 1≤l≤q, q represents the number of pixel blocks in the superpixel image, S l Indicates the number of pixels in the lth pixel block area, x p is the grayscale value of the lth pixel block area. k represents the cluster center, u kl represents the membership coefficient between the pixel block with the label value l and the kth cluster center, m represents the membership coefficient index, c represents the number of types, p represents pixels, k represents the kth cluster center, R l Represents the set of all pixel blocks.

[0072] 2) Set the number of iterations to t = 0, the maximum number of iterations to T, the minimum error to η, and randomly initialize the membership coefficients

[0073] 3) Cluster Center Calculation and update, the calculation formula is as follows:

[0074]

[0075] 4) Membership coefficient Calculation and update, the calculation formula is as follows:

[0076]

[0077] 5) If Or t>T the algorithm stops; otherwise, t=t+1 execute step 3)

[0078] Superpixel Image I sp The segmentation result image I is calculated through the above steps cluster , the image is divided into c blocks, and different blocks have different label values. In order to minimize the overall error, this embodiment performs segmentation processing on the super-pixel image by setting an objective function with the minimum overall error to obtain a segmentation result image.

[0079] S102, performing intima screening on the segmentation result image according to the attribute information of different blocks in the segmentation result image to obtain a target intima region.

[0080] This embodiment does not limit the specific method of performing endometrial screening according to the attribute information. The attribute information may be morphological feature information of the block.

[0081] It should be noted that the above-mentioned endometrial screening of the segmentation result image according to the attribute information of different blocks in the segmentation result image to obtain the target endometrial area includes: screening the different blocks in the segmentation result image according to the attribute information to determine the blocks that meet the endometrial screening conditions, and obtaining the target endometrial area based on the screened blocks; wherein the attribute information includes at least one of the area, the minimum bounding box, the center of gravity and the center coordinates. The endometrial screening condition can be a threshold condition that each attribute information needs to meet. This embodiment performs endometrial screening according to the area, the minimum bounding box, the center of gravity and the center coordinates, that is, the segmentation result image is screened for the endometrium according to the endometrial attributes directly related to the attribute information, so that the obtained target endometrial area is more in line with the endometrial screening requirements, and the accuracy of the determination of the target endometrial area is improved, thereby improving the accuracy of the subsequent receptivity analysis.

[0082] It should be further explained that the above-mentioned screening of different blocks in the segmentation result image according to the attribute information to determine the blocks that meet the endothelial screening conditions, and obtaining the target endothelial region based on the screened blocks, may include: obtaining the endothelial attribute threshold corresponding to each attribute information; comparing the value of each attribute information with the endothelial attribute threshold, determining the block that meets the endothelial screening conditions according to the comparison result and updating the label value of the block in the segmentation result image; screening according to the label value of the block to obtain the target endothelial region. This embodiment does not limit the specific method of labeling values. For example, it can be labeled in the form of 1 and 0; it can also be labeled in other ways, as long as different attributes can be distinguished. This embodiment will determine the endothelial attribute thresholds corresponding to different endothelial attributes, comprehensively consider the multi-dimensional characteristics of the endothelial, and thus improve the accuracy of endothelial screening.

[0083] Optionally, a block corresponding to a preset mark value may be determined as a target intima region.

[0084] It should be further explained that the above-mentioned comparison of the value of each attribute information with the endothelial attribute threshold, determining the block that meets the endothelial screening condition according to the comparison result and updating the mark value of the block in the segmentation result image may include: setting the block whose area is smaller than the endothelial area threshold to 0; and / or setting the block whose perimeter of the minimum bounding box is smaller than the endothelial minimum bounding box perimeter threshold to 0; and / or setting the block whose mark value at the position of the block center of gravity is not equal to the block mark value to 0; among the remaining blocks, setting the block whose center coordinate is the smallest distance from the preset image effective center to 1, and the remaining blocks to 0. This embodiment can determine different attribute thresholds and attribute threshold forms according to different attribute information, comprehensively consider the multi-dimensional characteristics of the endothelial, so that the screened endothelial meets the requirements of different dimensions, thereby improving the accuracy of endothelial screening.

[0085] For ease of understanding, the present invention provides a specific method for screening the endometrium, including statistical segmentation result image I clusterThe properties of different marked blocks (i.e., the aforementioned blocks), such as area Area and / or minimum bounding box BoundingBox and / or centroid Centroid and / or center coordinate CenterPt. The area Area is obtained by counting the number of pixels of different marked blocks; the minimum bounding box BoundingBox is obtained by counting the maximum and minimum horizontal and vertical coordinates of the pixels of the marked blocks, BoundingBox (maxX, maxY, minX, minY), and the average value of the maximum and minimum horizontal and vertical coordinates of the minimum bounding box is the center coordinate CenterPt, CenterPt.y = (BoundingBox.maxY + BoundingBox.minY) / 2; the center of gravity Centroid is obtained by counting the pixel distribution of different marked blocks, where i, j represent the horizontal and vertical coordinates, k represents the mark value in the pixel, and the center of gravity of the kth marked block is (Centroid k .x,Centroid k .y):

[0086]

[0087]

[0088] Will I cluster The blocks in the image whose area Area is less than minArea are set to 0, and the blocks whose perimeter lenth of the minimum bounding box BoundingBox is less than minLenth are set to 0. The above two steps can remove blocks with too small areas; the blocks whose mark value at the center of gravity of the block is not equal to the block mark value are set to 0, so that the blocks containing only the edge area can be removed and the blocks containing the middle core part are retained; among the remaining blocks (that is, the blocks that are not set to 0), the block with the smallest distance Dist between the center coordinate CenterPt and the preset image effective center useCenter is set to 1, and the remaining blocks are 0.

[0089] lenth=BoundingBox.maxX+BoundingBox.maxY-BoundingBox.minX-BoundingBox.min Y

[0090]

[0091] The blocks with a label value of 1 were identified as candidate intimal regions.

[0092] At this time, the regions obtained through preliminary screening are analyzed using the following methods, including but not limited to principal component analysis, convolutional neural network, recurrent neural network, etc., to obtain Harr features, Hu moments, skeleton features, corner features, texture features, etc., and calculate the features of different candidate intimal regions. Finally, machine learning or deep learning-based methods, such as Bayesian classifiers, support vector machines, BP neural networks, convolutional neural networks, random forests, etc., are used to classify the candidate intimal regions to determine the final intimal region (i.e., the target intimal region), and the final intimal region is obtained in the image I. cluster The area marked as 1. The image displayed on the display screen of the electronic device can be rendered based on the mark value of the area, for example, the area marked as 1 is rendered as a bright area, and other areas are rendered as dark areas or even black areas.

[0093] It should be noted that the above-mentioned endometrial screening of the segmentation result image to obtain the target endometrial area may include: performing preliminary screening on the segmented image to obtain candidate endometrial areas; classifying the candidate endometrial areas based on machine learning or deep learning to determine the target endometrial area. This embodiment does not limit the specific way of performing preliminary screening. For example, area screening may be performed; or minimum bounding box screening may be performed. This embodiment classifies the candidate endometrial areas based on machine learning or deep learning to determine the target endometrial area. The machine learning and deep learning in this embodiment may be a Bayesian classifier, a support vector machine, a BP neural network, a convolutional neural network, a random forest, and the like. Due to the preliminary screening and classification screening in this embodiment, the target endometrial area obtained is more accurate.

[0094] It should be noted that after the segmented image is preliminarily screened to obtain the candidate endometrial region, it can also include: extracting features from the candidate endometrial region to obtain endometrial region feature data; accordingly, classifying the candidate endometrial region based on machine learning or deep learning to determine the target endometrial region, including: classifying the endometrial region feature data based on machine learning or deep learning to determine the target endometrial region. This embodiment does not limit the specific method of feature extraction. For example, the feature extraction method includes but is not limited to principal component analysis, convolutional neural network, recurrent neural network, etc., to obtain Haar features (Haar features, Haar-like features), Hu moments (image moments), skeleton features, corner features, texture features, etc., and calculate the features of different candidate endometrial regions. Since this embodiment will perform feature extraction and determine the target endometrial region according to the feature extraction results, the accuracy of determining the target endometrial region can be improved.

[0095] S103, traversing the original uterine body data according to the target endometrial region to obtain endometrial volume data.

[0096] This embodiment traverses the original uterine body data according to the target endometrial region in order to obtain endometrial volume data. It can be that the original uterine body data is traversed according to the target endometrial region to read the target volume data associated with the target endometrial region, and these target volume data are determined as endometrial volume data. It can be understood that after determining the target endometrial region, only the two-dimensional framework of the endometrium is known, and its specific spatial form is unknown. Therefore, it is necessary to obtain its three-dimensional data based on the original uterine body data to obtain endometrial volume data, so as to facilitate the subsequent measurement of the endometrium. This step reads the corresponding volume data based on the target endometrial region, which can be understood as obtaining the three-dimensional data corresponding to the two-dimensional data of the target endometrial region based on the original uterine body data to obtain the endometrial volume data.

[0097] Furthermore, the volume data of the remaining area outside the target endometrial area in the original uterine body data may be traversed, and the volume data associated with the target endometrial area may be determined therefrom, and the volume data corresponding to the target endometrial area may constitute the endometrial volume data together. This embodiment does not limit the specific process of traversing the remaining area of ​​the original uterine body data according to the target endometrial area. For example, other areas may be processed according to the method for extracting the target endometrial area to obtain the endometrial volume data. Alternatively, the endometrial volume data may be obtained by processing according to the contour of the target endometrial area.

[0098] It should be noted that the above-mentioned traversing the original uterine body data according to the target endometrial region to obtain the endometrial volume data may include: using the above-mentioned endometrial data extraction method to traverse the remaining area of ​​the original uterine body data, and obtaining the endometrial volume data corresponding to the target endometrial region based on the volume data of the remaining area; or traversing the original uterine body data according to the contour of the target endometrial region to perform adjacent contour detection, and obtaining the endometrial volume data based on the volume data of the detected adjacent contours. This embodiment provides two methods for determining the endometrial volume data, thereby improving the diversity of determining the endometrial volume data.

[0099] It should be noted that after traversing the original uterine body data according to the target endometrial area to obtain the endometrial volume data, it may also include: performing endometrial analysis according to the endometrial volume data to obtain endometrial target feature data; determining endometrial receptivity according to the endometrial target feature data. This embodiment does not limit the specific target feature data. For example, the target feature data may be the endometrial volume; or it may be the size of the endometrial bounding box; or it may be the morphological distribution, endometrial size, endometrial thickness, blood supply and direction information. This embodiment will first extract the target feature data, determine the endometrial receptivity according to the target feature data, and improve the accuracy of determining the endometrial receptivity.

[0100] Compared with the current manual segmentation and measurement of the endometrium, the endometrial data extraction method provided by the embodiment of the present invention can perform intelligent analysis on the endometrium and determine the endometrial volume data related to the endometrium, so that the endometrial receptivity analysis can be performed based on the endometrial volume data obtained by the intelligent method, thereby improving the accuracy and efficiency of the determination of endometrial receptivity. At the same time, the embodiment of the present invention obtains uterine body data, performs image processing and analysis on the uterine body data to extract the endometrial region, and then obtains the endometrial volume data in the uterine body data based on the endometrial region, thereby realizing the accurate acquisition of three-dimensional endometrial data. Compared with two-dimensional data, endometrial volume data contains more characteristic information, which can provide a guarantee for the accurate acquisition of endometrial receptivity.

[0101] In order to make the present invention easier to understand, please refer to Figure 2 , Figure 2 An example flow chart of an endometrial data extraction method provided by an embodiment of the present invention may specifically include:

[0102] S200, receiving three-dimensional uterine body data obtained by performing abdominal ultrasound scanning with ultrasound equipment.

[0103] S201, performing cutting processing and preprocessing on the three-dimensional uterine body data to obtain grayscale section image data; the preprocessing includes at least one of edge extraction, filtering processing and noise reduction processing.

[0104] The uterine body data in this embodiment is endometrial ultrasound three-dimensional volume data. The three-dimensional uterine body data is cut into sections to obtain section images, and then the section images are preprocessed to obtain grayscale section image data. The edge extraction method corresponding to this embodiment may include but is not limited to conventional edge detection algorithms such as canny (multi-stage edge algorithm) and sobel (Sobel operator). A variety of algorithms can be used for edge extraction here.

[0105] S202, segmenting the grayscale section image data to obtain a gradient map.

[0106] S203, iteratively calculating the gradient map to obtain a super-pixel image; wherein the super-pixel image includes different pixel blocks.

[0107] This embodiment uses the gradient map to iterate to perform superpixel segmentation to obtain a superpixel image. The superpixel segmentation implementation includes but is not limited to: watershed-based superpixels, SLIC superpixels, etc. The processed image is segmented by a machine learning or deep learning method to achieve intima detection and segment different intima areas. The methods include but are not limited to: clustering, support vector machine, convolutional neural network, recurrent neural network, etc.

[0108] S204, aggregating different pixel blocks in the super-pixel image to minimize the objective function and obtain a segmented image; wherein the objective function is a function of the overall accumulated error of the image.

[0109] The specific clustering implementation process of this embodiment is as follows:

[0110] 1) Objective function: The images are divided into c categories by clustering. The objective function is as follows:

[0111]

[0112] Where l is the pixel block label value 1≤l≤q, q represents the number of pixel blocks in the superpixel image, S l Indicates the number of pixels in the lth pixel block area, x p is the grayscale value of the lth pixel block area. k represents the cluster center, u kl represents the membership coefficient between the pixel block with the label value l and the kth cluster center, m represents the membership coefficient index, c represents the number of types, p represents pixels, k represents the kth cluster center, R l Represents the set of all pixel blocks.

[0113] 2) Set the number of iterations to t = 0, the maximum number of iterations to T, the minimum error to η, and randomly initialize the membership coefficients

[0114] 3) Cluster Center Calculation and update, the calculation formula is as follows:

[0115]

[0116] 4) Membership coefficient Calculation and update, the calculation formula is as follows:

[0117]

[0118] 5) If Or t>T the algorithm stops; otherwise, t=t+1 execute step 3)

[0119] Superpixel Image I sp The segmented image I is calculated through the above steps. cluster , the image is divided into c blocks, and different blocks have different label values.

[0120] S205, determining the intimal attributes of different marked blocks in the segmented image, and screening the marked blocks according to the intimal attributes to obtain candidate intimal regions; wherein the intimal attributes include at least one of area, minimum bounding box, centroid, and center coordinates.

[0121] This embodiment screens the inner membrane area through the area Area, the minimum bounding box BoundingBox, the center of gravity Centroid and the center coordinate CenterPt. The area Area is obtained by counting the number of pixels of different marking blocks; the minimum bounding box BoundingBox is obtained by counting the maximum and minimum horizontal and vertical coordinates of the pixels of the marking blocks, BoundingBox (maxX, maxY, minX, minY), and the average value of the maximum and minimum horizontal and vertical coordinates of the minimum bounding box is the center coordinate CenterPt, CenterPt.y = (BoundingBox.maxY + BoundingBox.minY) / 2; the center of gravity Centroid is obtained by counting the pixel distribution of different marking blocks, where i, j represent the horizontal and vertical coordinates, k represents the marking value in the pixel, and the center of gravity of the kth marking block is (Centroid k .x,Centroid k .y):

[0122]

[0123]

[0124] Will I cluster The blocks in the image whose area Area is less than minArea are set to 0, the blocks whose perimeter lenth of the minimum bounding box BoundingBox is less than minLenth are set to 0, the blocks whose mark value at the center of gravity is not equal to the block mark value are set to 0, and among the remaining blocks, the block whose center coordinate CenterPt has the smallest distance Dist from the preset image effective center useCenter is set to 1, and the rest of the blocks are 0.

[0125] lenth=BoundingBox.maxX+BoundingBox.maxY-BoundingBox.minX-BoundingBox.min Y

[0126]

[0127] This embodiment performs preliminary screening according to the intima attributes to obtain candidate intima regions. This embodiment can also perform feature extraction on the candidate intima regions, and the feature extraction methods include but are not limited to principal component analysis, convolutional neural network, recurrent neural network, etc., to obtain Harr features, Hu moments, skeleton features, corner features, texture features, etc., and calculate the features of different candidate intima regions, which can be subsequently classified using the features of the candidate intima regions.

[0128] S206, classify the candidate intima regions using a machine learning or deep learning method to determine the final intima region, and mark it as 1 in the image.

[0129] This embodiment uses a method based on machine learning or deep learning, such as Bayesian classifier, support vector machine, BP neural network, convolutional neural network, random forest, etc., to classify the candidate intimal regions to determine the final intimal region, and cluster Marked as 1.

[0130] S207, determining the endometrial image data corresponding to the final endometrial region; wherein, the endometrial image data is obtained by traversing the uterine body data using the above process; or according to the final endometrial region contour, using the active contour model, the geodesic contour model, and the graph cut method to detect adjacent contours, traverse the uterine body data, and obtain the endometrial image data.

[0131] This embodiment uses the above process to traverse the uterine body data to obtain the endometrial image data. It is also possible to detect adjacent contours based on the contour of the endometrial region of the current section using methods such as active contour model, geodesic contour model, and graph cuts, traverse the uterine three-dimensional volume data, and obtain the endometrial image data. dst =SrcImg(I sp = =1), I dst It is the endometrial image data.

[0132] S208: Determine three-dimensional volume data of the endometrium according to the endometrium image data.

[0133] S209, determining endometrial receptivity according to the three-dimensional volume data of the endometrium. Specifically, the process includes: obtaining endometrial target feature data according to the three-dimensional volume data of the endometrium; wherein the endometrial target feature data includes endometrial volume, size of endometrial bounding box, morphological distribution, size, endometrial thickness, blood supply and direction information; and determining endometrial receptivity according to the endometrial target feature data.

[0134] The determined endometrial receptivity result can then be sent to the ultrasound device so that the ultrasound device displays the endometrial receptivity result on a display screen, thereby guiding the ultrasound scanning process and the generation of ultrasound diagnosis results.

[0135] Optional, Figure 3 FIG. 1 is a flow chart of a method for extracting endometrial data provided by an embodiment of the present invention. Figure 3As shown, the endometrial data extraction method includes the following steps: acquiring endometrial ultrasound three-dimensional volume data, determining a section image based on the volume data, performing superpixel segmentation on the section image to obtain a superpixel image, obtaining the endometrium through image preprocessing, endometrial detection, endometrial extraction and endometrial screening of the superpixel image, and finally obtaining three-dimensional volume data of the endometrium, performing three-dimensional endometrial volume analysis, and determining endometrial receptivity.

[0136] An endometrial data extraction device provided by an embodiment of the present invention is introduced below. The endometrial data extraction device described below and the endometrial data extraction method described above can be referred to each other.

[0137] Please refer to Figure 4 , Figure 4 A schematic diagram of the structure of an endometrial data extraction device provided by an embodiment of the present invention may include:

[0138] The super-pixel image determination module 100 is used to process the original uterine body data to obtain a super-pixel image; wherein the super-pixel image includes different pixel blocks;

[0139] An image segmentation module 200 is used to iteratively aggregate the different pixel blocks in the super-pixel image to segment the super-pixel image to obtain a segmentation result image;

[0140] The intima screening module 300 is used to screen the intima of the segmentation result image according to the attribute information of different blocks in the segmentation result image to obtain the target intima area;

[0141] The volume data determination module 400 is used to traverse the original uterine body data according to the target endometrial region to obtain endometrial volume data.

[0142] In an optional embodiment, based on the above embodiment, the super pixel image determination module 100 may include:

[0143] A cutting submodule, used for cutting the original uterine body data to obtain section data;

[0144] A preprocessing submodule, used to preprocess the slice data to obtain preprocessed slice image data; wherein the preprocessing includes at least one of edge extraction, filtering and noise reduction;

[0145] The super-pixel processing submodule is used to perform super-pixel segmentation on the image corresponding to the pre-processed section image data to obtain the super-pixel image.

[0146] In an optional embodiment, based on any of the above embodiments, the endomembrane screening module 300 may include:

[0147] The endocardium screening submodule is used to screen different blocks in the segmentation result image according to the attribute information to determine the blocks that meet the endocardium screening conditions, and obtain the target endocardium area based on the screened blocks; wherein the attribute information includes at least one of the area, minimum bounding box, center of gravity and center coordinates.

[0148] In an optional embodiment, based on the above embodiment, the above inner membrane screening submodule may include:

[0149] A threshold value acquisition unit, used for acquiring an intima attribute threshold value corresponding to each of the attribute information;

[0150] A label value determination unit, used for comparing the value of each of the attribute information with the intima attribute threshold, determining the block satisfying the intima screening condition according to the comparison result and updating the label value of the block in the segmentation result image;

[0151] The endometrium screening unit is used to screen according to the mark value of the block to obtain the target endometrium area.

[0152] In an optional embodiment, based on the above embodiment, the above tag value determination unit may include:

[0153] A first mark value determination subunit is used to set the block whose area is smaller than the intima area threshold to 0;

[0154] and / or, a second marking value determination subunit, configured to set a block whose perimeter of the minimum bounding box is smaller than a threshold of the perimeter of the inner membrane minimum bounding box to 0;

[0155] and / or, a third mark value determination subunit, configured to set the mark value of the block whose mark value at the position of the block center of gravity is not equal to the block mark value to 0;

[0156] The fourth marking value determination subunit is used to set the block whose center coordinates are the smallest distance from the preset effective center of the image to 1 among the remaining blocks, and set the other blocks to 0.

[0157] In an optional embodiment, based on any of the above embodiments, the endomembrane screening module 300 may include:

[0158] A preliminary screening submodule, used for performing preliminary screening on the segmentation result image to obtain candidate intima regions;

[0159] A classification submodule is used to classify the candidate intima region based on machine learning or deep learning to determine the target intima region.

[0160] In an optional embodiment, based on the above embodiment, the endometrial data extraction device may further include:

[0161] A feature extraction module, used to extract features from the candidate intima region to obtain intima region feature data;

[0162] Accordingly, the above classification submodule includes:

[0163] A classification unit is used to classify the intimal region characteristic data based on the machine learning or the deep learning to determine the target intimal region.

[0164] In an optional embodiment, based on any of the above embodiments, the image segmentation module 200 may include:

[0165] An image segmentation submodule, configured to iteratively aggregate the different pixel blocks in the super-pixel image so as to minimize the objective function, so as to segment the super-pixel image and obtain the segmentation result image;

[0166] Among them, the objective function is l is the pixel block label value 1≤l≤q, q represents the number of pixel blocks in the superpixel image, c represents the number of types, p represents the pixel, S l Indicates the number of pixels in the lth pixel block area, x p is the grayscale value of the lth pixel block area, v k represents the cluster center, u kl represents the membership coefficient between the pixel block with the label value l and the kth cluster center, m represents the membership coefficient index, k represents the kth cluster center, R l Represents the set of all pixel blocks.

[0167] In an optional embodiment, based on any of the above embodiments, the volume data determination module 400 may include:

[0168] The first volume data determination submodule is used to traverse the remaining area of ​​the original uterine body data using the above endometrial data extraction method to obtain the endometrial volume data corresponding to the target endometrial area.

[0169] In an optional embodiment, based on any of the above embodiments, the volume data determination module 400 may include:

[0170] The second volume data determination submodule is used to traverse the original uterine body data to perform adjacent contour detection according to the contour of the target endometrial area to obtain the endometrial volume data.

[0171] In an optional embodiment, based on any of the above embodiments, the endometrial data extraction device may further include:

[0172] A characteristic data determination module, configured to perform endometrial analysis according to the endometrial volume data to obtain endometrial target characteristic data;

[0173] A receptivity determination module is used to determine the endometrial receptivity according to the endometrial target characteristic data.

[0174] It should be noted that the order of the modules and units in the above-mentioned endometrial data extraction device can be changed without affecting the logic.

[0175] An endometrial data extraction device provided in an embodiment of the present invention can perform intelligent analysis on the endometrium and determine endometrial volume data related to the endometrium, compared with the current manual segmentation and measurement of the endometrium, so that endometrial receptivity analysis can be performed based on the endometrial volume data obtained by the intelligent method, thereby improving the accuracy and efficiency of determining endometrial receptivity.

[0176] An endometrial data extraction device provided by an embodiment of the present invention is introduced below. The endometrial data extraction device described below and the endometrial data extraction method described above can be referred to each other.

[0177] Please refer to Figure 5 , Figure 5 A schematic diagram of the structure of an endometrial data extraction device provided by an embodiment of the present invention may include:

[0178] A memory 10, used for storing computer programs;

[0179] The processor 20 is used to execute a computer program to implement the above-mentioned endometrial data extraction method.

[0180] The memory 10 , the processor 20 , and the communication interface 30 all communicate with each other via a communication bus 40 .

[0181] In an embodiment of the present invention, the memory 10 is used to store one or more programs, the program may include program code, the program code includes computer operation instructions, and in an embodiment of the present invention, the memory 10 may store a program for implementing the above-mentioned endometrial data extraction method.

[0182] In a possible implementation, the memory 10 may include a program storage area and a data storage area, wherein the program storage area may store an operating system and an application required for at least one function, etc.; the data storage area may store data created during use.

[0183] In addition, the memory 10 may include a read-only memory and a random access memory, and provide instructions and data to the processor. A portion of the memory may also include an NVRAM. The memory stores an operating system and operating instructions, executable modules or data structures, or a subset thereof, or an extended set thereof, wherein the operating instructions may include various operating instructions for implementing various operations. The operating system may include various system programs for implementing various basic tasks and processing hardware-based tasks.

[0184] The processor 20 may be a central processing unit (CPU), an application specific integrated circuit, a digital signal processor, a field programmable gate array or other programmable logic device, a microprocessor or any conventional processor, etc. The processor 20 may call a program stored in the memory 10 .

[0185] The communication interface 30 may be an interface of a communication module, and is used to connect to other devices or systems.

[0186] Of course, it should be noted that Figure 5 The structure shown does not constitute a limitation on the endometrial data extraction device in the embodiment of the present invention. In actual applications, the endometrial data extraction device may include Figure 5 More or fewer components than shown, or combinations of certain components.

[0187] The computer-readable storage medium provided in an embodiment of the present invention is introduced below. The computer-readable storage medium described below and the endometrial data extraction method described above can be referenced to each other.

[0188] The present invention also provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps of the above-mentioned endometrial data extraction method are implemented.

[0189] The computer-readable storage medium may include: a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, and other media that can store program codes.

[0190] In this specification, each embodiment is described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the embodiments can be referred to each other. For the device disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part.

[0191] Professionals may further appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described in the above description according to function. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professionals and technicians may use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.

[0192] Finally, it should be noted that, in this article, relationships such as first and second, etc. are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprises" or any other variations are intended to cover non-exclusive inclusion, so that a process, method, article or device that includes a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device.

[0193] The above is a detailed introduction to the endometrial data extraction method, device, equipment and readable storage medium provided by the present invention. Specific examples are used in this article to illustrate the principles and implementation methods of the present invention. The description of the above embodiments is only used to help understand the method of the present invention and its core idea; at the same time, for those skilled in the art, according to the idea of ​​the present invention, there will be changes in the specific implementation method and application scope. In summary, the content of this specification should not be understood as a limitation on the present invention.

Claims

1. A method for extracting endometrial data, characterized in that: include: Processing the original uterine body data to obtain a super-pixel image; wherein the super-pixel image includes different pixel blocks; Iteratively aggregating the different pixel blocks in the super-pixel image to segment the super-pixel image to obtain a segmentation result image; According to the attribute information of different blocks in the segmentation result image, the segmentation result image is screened for intima to obtain a target intima area; The original uterine body data is traversed according to the target endometrial region to obtain endometrial volume data.

2. The method for extracting endometrial data according to claim 1, characterized in that: The processing of the original uterine body data to obtain a super-pixel image includes: Performing cutting processing on the original uterine body data to obtain section data; Preprocessing the section data to obtain preprocessed section image data; wherein the preprocessing includes at least one of edge extraction, filtering and noise reduction; Perform super-pixel segmentation on the image corresponding to the pre-processed section image data to obtain the super-pixel image.

3. The method for extracting endometrial data according to claim 1, characterized in that: The step of performing intima screening on the segmentation result image according to the attribute information of different blocks in the segmentation result image to obtain a target intima region includes: Different blocks in the segmentation result image are screened according to the attribute information to determine the blocks that meet the intima screening conditions, and the target intima area is obtained based on the screened blocks; wherein the attribute information includes at least one of area, minimum bounding box, center of gravity and center coordinates.

4. The method for extracting endometrial data according to claim 3, characterized in that: The step of screening different blocks in the segmentation result image according to the attribute information to determine blocks that meet the intima screening condition, and obtaining the target intima region based on the screened blocks includes: Obtaining the inner membrane attribute threshold corresponding to each of the attribute information; Comparing the value of each of the attribute information with the intima attribute threshold, determining the block that meets the intima screening condition according to the comparison result and updating the mark value of the block in the segmentation result image; Screening is performed according to the tag value of the block to obtain the target endometrial region.

5. The method for extracting endometrial data according to claim 4, characterized in that: The step of comparing the value of each of the attribute information with the intima attribute threshold, determining the block satisfying the intima screening condition according to the comparison result and updating the label value of the block in the segmentation result image comprises: The blocks with an area smaller than the intima area threshold are set to 0; and / or setting the blocks whose minimum bounding box perimeter is less than the inner membrane minimum bounding box perimeter threshold to 0; and / or setting the block whose mark value at the block center is not equal to the block mark value to 0; In the remaining blocks, the block whose center coordinate is the smallest distance from the preset effective center of the image is set to 1, and the remaining blocks are set to 0.

6. The method for extracting endometrial data according to claim 1, characterized in that: The step of performing intima screening on the segmentation result image to obtain a target intima region includes: Preliminarily screening the segmentation result image to obtain candidate intima regions; The candidate intimal regions are classified based on machine learning or deep learning to determine the target intimal region.

7. The method for extracting endometrial data according to claim 6, characterized in that: After the segmentation result image is preliminarily screened to obtain a candidate intima region, the method further includes: Extracting features of the candidate intima region to obtain intima region feature data; Accordingly, the classifying the candidate intimal regions based on machine learning or deep learning to determine the target intimal region includes: The target intimal region is determined by classifying the intimal region feature data based on the machine learning or the deep learning.

8. The method for extracting endometrial data according to any one of claims 1 to 7, characterized in that: The iteratively aggregating the different pixel blocks in the super-pixel image to segment the super-pixel image to obtain a segmentation result image includes: Iteratively aggregating the different pixel blocks in the super-pixel image so as to minimize the objective function, thereby segmenting the super-pixel image to obtain the segmentation result image; Among them, the objective function is , l is the pixel block label value 1≤l≤q, q represents the number of pixel blocks in the superpixel image, c represents the number of types, p represents the pixel, S l Indicates the number of pixels in the lth pixel block area, x p is the gray value of the lth pixel block area, vk represents the cluster center, ukl represents the membership coefficient between the pixel block with the label value l and the kth cluster center, m represents the membership coefficient index, k represents the kth cluster center, R l Represents the set of all pixel blocks.

9. The method for extracting endometrial data according to any one of claims 1 to 7, characterized in that: The step of traversing the original uterine body data according to the target endometrial region to obtain endometrial volume data includes: Using the above endometrial data extraction method, traverse the remaining area of ​​the original uterine body data to obtain the endometrial volume data corresponding to the target endometrial area; Alternatively, the original uterine body data is traversed according to the contour of the target endometrial area to perform adjacent contour detection to obtain the endometrial volume data.

10. The method for extracting endometrial data according to any one of claims 1 to 7, characterized in that: After traversing the original uterine body data according to the target endometrial region to obtain endometrial volume data, the method further includes: Performing endometrial analysis according to the endometrial volume data to obtain endometrial target feature data; Endometrial receptivity is determined based on the endometrial target characteristic data.

11. An endometrial data extraction device, characterized in that: include: A super-pixel image determination module is used to process the original uterine body data to obtain a super-pixel image; wherein the super-pixel image includes different pixel blocks; An image segmentation module, configured to iteratively aggregate the different pixel blocks in the super-pixel image to segment the super-pixel image to obtain a segmentation result image; An intimal screening module, used for screening the intimal region of the segmentation result image according to the attribute information of different blocks in the segmentation result image, to obtain a target intimal region; The volume data determination module is used to traverse the original uterine body data according to the target endometrial area to obtain endometrial volume data.

12. An endometrial data extraction device, characterized in that: include: Memory for storing computer programs; A processor, configured to implement the steps of the endometrial data extraction method according to any one of claims 1 to 10 when executing the computer program.

13. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the endometrial data extraction method according to any one of claims 1 to 10 are implemented.

Images