A real-time monitoring culture system for embryo development in assisted reproductive process

Abstract

This invention relates to the field of image analysis technology, specifically to a real-time monitoring and culture system for embryonic development during assisted reproductive technology. It includes: an image data acquisition module for acquiring grayscale images of embryos; an image information content analysis module for determining the splitting information content and sharpness information content of each embryo grayscale image based on changes in the number of edge pixels and differences in pixel grayscale values; a target embryo image screening module for screening target embryo images based on the differences in splitting information content and sharpness information content of the embryo grayscale images; and an embryonic development monitoring module for comparing the similarity features of edge lines in all target embryo images to identify high-information-content embryo images for embryonic development monitoring. This invention can accurately screen embryo images with high reference value, thereby reducing data redundancy and improving the accuracy of embryonic development monitoring.

Description

Technical Field

[0001] This invention relates to the field of image analysis technology, and more specifically to a real-time monitoring and culture system for embryo development during assisted reproductive technology. Background Technology

[0002] Assisted reproductive technologies (ART) are medical techniques that help couples or individuals who cannot conceive naturally to achieve their desire to have children. Common examples include in vitro fertilization (IVF), where embryos are cultured outside the body to a specific stage before being transferred into the uterus for implantation and conception. However, during in vitro embryo culture, the cleavage process of the fertilized egg may become abnormal due to environmental and internal changes. For example, abnormal cleavage timing can prevent effective in vitro culture, thus requiring monitoring of the embryo culture process.

[0003] Existing technologies for real-time monitoring of embryonic development typically employ full-process morphological monitoring. This monitoring method acquires a large amount of information, resulting in excessive redundancy and excessive interference from invalid information. It fails to efficiently extract and distinguish embryonic images with clear information, thereby affecting the monitoring efficiency and accuracy of embryonic development. Summary of the Invention

[0004] To address the technical problem that full-process morphological monitoring generates a large amount of redundant information, with excessive interference from invalid information, making it difficult to efficiently extract and distinguish embryo images with clear information, thus affecting the monitoring efficiency and accuracy of embryo development, this invention aims to provide a real-time embryo development monitoring and culture system in assisted reproductive technology. The specific technical solution adopted is as follows:

[0005] This invention proposes a real-time monitoring and culture system for embryo development during assisted reproductive technology, the system comprising:

[0006] The image data acquisition module is used to acquire a series of grayscale images of the embryos to be monitored in a time sequence.

[0007] The image information content analysis module is used to determine the splitting information content of each embryo grayscale image based on the change in the number of edge pixels; and to determine the clarity information content of each embryo grayscale image based on the grayscale difference of pixels in a local range.

[0008] The target embryo image screening module is used to screen target embryo images from all embryo grayscale images based on the difference in the amount of splitting information between all embryo grayscale images and the amount of clear information in the embryo grayscale images.

[0009] The embryo development monitoring module is used to compare the similar features between the edge lines of all target embryo images to identify high-information-content embryo images; and to monitor the embryo development of the embryos to be monitored based on the high-information-content embryo images.

[0010] Furthermore, the method for obtaining the splitting information includes:

[0011] The absolute value of the difference between the total number of edge pixels in each embryo grayscale image and the previous embryo grayscale image in time sequence is used as the quantity change factor for each embryo grayscale image.

[0012] The time interval between each embryo grayscale image and the previous embryo grayscale image in time sequence is used as the time interval factor corresponding to each embryo grayscale image;

[0013] The ratio between the quantity change factor and the time interval factor corresponding to each embryo grayscale image is used as the splitting information content of each embryo grayscale image.

[0014] Furthermore, the method for obtaining the amount of clear information includes:

[0015] In each grayscale image of an embryo, a preset neighborhood is determined with each edge pixel as the center;

[0016] In the preset neighborhood corresponding to each edge pixel, the absolute value of the difference between the gray values ​​of the center pixel and each neighboring pixel is calculated and used as the gray-level difference factor.

[0017] The sum of all gray-level difference factors of each center pixel is used as the gray-level feature value of each center pixel;

[0018] The mean of the grayscale feature values ​​of all edge pixels in each embryo grayscale image is used as the amount of clear information.

[0019] Furthermore, the method for acquiring the target embryo image includes:

[0020] Based on the differences in the amount of splitting information among all embryo grayscale images, all embryo grayscale images are divided to obtain image subsequences;

[0021] In each image subsequence, the grayscale image of the embryo with the highest clarity information is selected as the target embryo image.

[0022] Furthermore, the method for obtaining the image subsequence includes:

[0023] In a series of embryo grayscale images, the absolute value of the difference between the splitting information content of each embryo grayscale image and the next adjacent embryo grayscale image is negatively correlated and normalized, and the result is used as the consistent feature value.

[0024] In terms of time sequence, the first grayscale image of the embryo is used as the image to be tested, and the image to be tested corresponds to an image set;

[0025] When the consistency feature value between the last embryo grayscale image in the image set corresponding to the image to be tested and the next adjacent embryo grayscale image is greater than or equal to the preset consistency threshold, the next adjacent embryo grayscale image is merged into the image set of the image to be tested and the traversal continues.

[0026] When the consistency feature value between the last embryo grayscale image in the image set corresponding to the image to be tested and the next adjacent embryo grayscale image is less than the preset consistency threshold, the image set corresponding to the image to be tested at this time is taken as an image subsequence, and the next adjacent embryo grayscale image is taken as a new image to be tested and the traversal continues.

[0027] After the traversal is complete, all image subsequences are obtained.

[0028] Furthermore, the method for acquiring the high-information-content embryo image includes:

[0029] The mean gray value of all edge pixels in each target embryo image is used as the gray feature value of each target embryo image;

[0030] For any target embryo image, the absolute value of the difference between the gray-level feature values ​​of the target embryo image and every other target embryo image is used as the gray-level difference factor;

[0031] The mean value of all gray-level difference factors of the target embryo image is taken as the gray-level significant feature value of the target embryo image;

[0032] Analyze the positional similarity between the edge lines of the target embryo images to determine the obvious positional feature values ​​of each target embryo image;

[0033] The sum of the grayscale and positional features of each target embryo image is normalized and used as the splitting feature value of that target embryo image.

[0034] Among all target embryo images, those with a significant division feature value greater than a preset significant division threshold are designated as high-information embryo images.

[0035] Furthermore, the method for obtaining the obvious location feature values ​​includes:

[0036] In each target embryo image, the position coordinates of all edge pixels are fitted with a straight line based on the least squares method, and the slope value of the fitted line is obtained.

[0037] For any target embryo image, the absolute value of the difference between the slope of the fitted line of the target embryo image and that of each other target embryo image is used as the positional difference factor of the target embryo image.

[0038] The mean of all positional difference factors in the target embryo image is taken as the positional characteristic value of the target embryo image.

[0039] Furthermore, the embryonic development monitoring based on high-information-content embryo images includes:

[0040] Each high-information-content embryo image is input into a pre-trained neural network, which outputs the division cycle of each high-information-content embryo image.

[0041] Obtain the time period corresponding to the division cycle of each high-information-content embryo image;

[0042] For any high-information-content embryo image, if the corresponding time point falls within the time period corresponding to the division cycle, the embryo under monitoring is considered to be developing normally at that time point; if the corresponding time point does not fall within the time period corresponding to the division cycle, the embryo under monitoring is considered to be developing abnormally at that time point.

[0043] Furthermore, the method for obtaining the embryo grayscale image includes:

[0044] In terms of time sequence, a series of grayscale images of the embryos to be monitored are acquired;

[0045] Each embryo grayscale image is segmented using a pre-trained semantic segmentation network to obtain the corresponding embryo grayscale image.

[0046] Furthermore, the method for obtaining the edge pixels includes:

[0047] Edge pixels in each embryo grayscale image are obtained based on the Canny operator.

[0048] The present invention has the following beneficial effects:

[0049] First, the image data acquisition module acquires a series of grayscale images of the embryos to be monitored in a time-series manner. During cell division, not every germ cell at every moment contains valid information; images from certain moments may have low reference value due to their relatively uniform morphology. Therefore, in this invention, the amount of information contained in the embryo grayscale images can be analyzed to select high-information-content embryo images for final embryo development monitoring. Given that cell division generates numerous new edges, and the generation of these new edges causes changes in grayscale, the image information content analysis module, to ensure the accuracy and clarity of subsequent image selection, precisely quantifies the division information content of each embryo grayscale image by utilizing the changes in the number of edge pixels, providing a strong basis for subsequent embryo development assessment. Simultaneously, based on the grayscale differences of pixels within a local area, the clarity information content of each embryo grayscale image is determined, ensuring that the images used for monitoring have high clarity. Then, in the target embryo image selection module, representative target embryo images with high information content are selected from all embryo grayscale images based on the characteristics of differences in division information content and the clarity information content. Finally, in the embryo development monitoring module, since analyzing images within the same division stage would also generate data redundancy, we compared the similar features between the edge lines of all target embryo images to identify high-information embryo images. At this point, embryo development monitoring based on these images can effectively reduce data redundancy and improve monitoring accuracy, thereby providing doctors with rich information on embryo development and helping them make timely and accurate judgments based on the embryo development monitoring results. Attached Figure Description

[0050] To more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0051] Figure 1 This is a system block diagram of a real-time monitoring and culture system for embryo development during assisted reproduction provided in one embodiment of the present invention;

[0052] Figure 2 This is a flowchart of a method for acquiring high-information-content embryo images according to an embodiment of the present invention;

[0053] Figure 3 This is a schematic diagram of a high-information embryo image provided in one embodiment of the present invention. Detailed Implementation

[0054] To further illustrate the technical means and effects adopted by the present invention to achieve its intended purpose, the following, in conjunction with the accompanying drawings and preferred embodiments, details the specific implementation, structure, features, and effects of an intelligent monitoring system and method for building suspended platforms proposed according to the present invention. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.

[0055] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0056] The following description, in conjunction with the accompanying drawings, details the specific scheme of the real-time monitoring and culture system for embryonic development in the assisted reproductive process provided by the present invention.

[0057] Please see Figure 1 The diagram shows a system block diagram of a real-time monitoring and culture system for embryo development in assisted reproductive technology according to an embodiment of the present invention. The system includes: an image data acquisition module 101, an image information analysis module 102, a target embryo image screening module 103, and an embryo development monitoring module 104.

[0058] The image data acquisition module 101 is used to acquire a series of grayscale images of the embryos to be monitored in a time sequence.

[0059] Assisted reproductive technology (ART) is a technique that involves artificially intervening in the fertility process of individuals who have difficulty conceiving naturally. The main purpose of this technology is to culture healthy and normal reproductive cells in vitro to achieve a normal pregnancy. However, during the in vitro culture of reproductive cells, the natural environment and changes in the health of the reproductive cells themselves can have an impact, which may lead to abnormal behaviors such as abnormal cleavage timing. If abnormal reproductive cells attempt to implant or result in pregnancy, it may lead to implantation failure or termination of pregnancy. Therefore, it is necessary to monitor cell division behavior during the in vitro culture of reproductive cells.

[0060] Real-time monitoring during cell division is a massive undertaking, and not every germ cell at any given moment contains useful information. There are instances where germ cell information is unclear or low (e.g., no division has occurred or the cell morphology is highly similar to that of cells in images from adjacent time points). Using this information for abnormal germ cell analysis is worthless. Therefore, in this embodiment of the invention, redundant germ cell division processes are removed during real-time monitoring, while embryo image data with high information content is retained, thereby enabling accurate and efficient identification of abnormal germ cell division behavior.

[0061] First, the monitoring equipment (composed of a microscope and a high-definition imaging device) is set up. The specific setup process can be set according to the implementation scenario. Then, the field of view of the high-definition imaging device and the microscope are connected so that the high-definition imaging device can acquire images of the embryos in the field of view of the microscope. The image acquisition time interval of the high-definition imaging device is 10 seconds. Thus, a series of embryo images of the embryos to be monitored in time sequence can be obtained based on the image data acquisition module.

[0062] It should be noted that, in the embodiments of the present invention, when monitoring embryonic development, the focus is on the morphological changes of embryonic development over a period of time. Therefore, it is necessary to acquire a series of embryonic images, and the specific duration of the time period can be adjusted according to the implementation scenario, and is not limited here.

[0063] Since embryo images may contain background information other than the embryo itself, each embryo image can be processed to remove the background information and obtain a grayscale image of the embryo.

[0064] Preferably, in one embodiment of the present invention, the method for obtaining an embryo grayscale image includes:

[0065] In terms of time sequence, a series of embryo images of the embryos to be monitored are processed into grayscale to obtain a series of embryo grayscale images of the embryos to be monitored.

[0066] Finally, each embryo grayscale image is segmented based on a pre-trained semantic segmentation network to obtain the corresponding embryo grayscale image.

[0067] It should be noted that the acquisition of embryonic grayscale images can employ methods such as average grayscale conversion, maximum value method, or weighted grayscale conversion, all of which are well-known techniques to those skilled in the art and will not be limited or elaborated upon here. The specific training process of the semantic segmentation network is briefly described as follows: Images containing embryos are used as training data. Pixels in the embryonic region are labeled as 1, and other pixels (background pixels) are labeled as 0, obtaining label data. The semantic segmentation network adopts an encoder-decoder structure. The training data and label data are normalized and input into the network. The semantic segmentation encoder extracts features from the input data to obtain a feature map. The semantic segmentation decoder samples and transforms the feature map, outputting the semantic segmentation result. The cross-entropy loss function is used to train the network. The training method is a well-known technique, and the specific process will not be elaborated upon.

[0068] In the embodiments of this invention, the collection and acquisition of various embryo data are all authorized by the relevant users, and the process does not violate relevant laws and regulations, nor does it violate public order and good morals.

[0069] The image information content analysis module 102 is used to determine the splitting information content of each embryo grayscale image based on the change in the number of edge pixels; and to determine the clarity information content of each embryo grayscale image based on the grayscale difference of pixels in a local range.

[0070] During the development of germ cells from a fertilized egg to the blastocyst stage, division occurs. The timing of this division process can be used to determine if abnormal division is occurring. During division, the emergence of new cells leads to the appearance of new growth edges for a certain period. However, the number of dividing cells varies at different stages of cell division, resulting in different edge appearances. For most of the time, germ cells are in a chaotic state, lacking clear division characteristics or retaining the cell division state from the previous moment. Therefore, for identifying abnormal germ cell division, image data with distinct differences and clear information should be selected to improve the accuracy and efficiency of abnormal division identification. Thus, this embodiment of the invention analyzes the information content of each embryo grayscale image in the image information content analysis module, thereby helping to select the necessary high-information-content image data in subsequent processes.

[0071] High-information-content image data must be images with obvious division characteristics during the process of germ cell division. At such times, there are often obvious edge features. Therefore, in this embodiment of the invention, the number of edge pixels of each embryo grayscale image is first analyzed to determine the division information content of each embryo grayscale image.

[0072] Preferably, in one embodiment of the present invention, the method for obtaining split information includes:

[0073] When cells divide, a distinct cell membrane appears, exhibiting strong edge features. As the in vitro culture time increases, the number of dividing cells also increases, resulting in more and more edge features. Therefore, the edge pixels in each embryo grayscale image are first obtained based on the Canny operator.

[0074] The absolute value of the difference in the total number of edge pixels between each embryo grayscale image and the previous embryo grayscale image in time sequence is used as the quantity variation factor for each embryo grayscale image. The larger the quantity variation factor, the more obvious the cell division features in the embryo grayscale image, and the more information it contains and the higher its reference value.

[0075] Meanwhile, the time interval between each embryo grayscale image and the preceding embryo grayscale image in time sequence is used as the time interval factor. Then, the ratio between the quantity change factor corresponding to each embryo grayscale image and the time interval factor is used as the division information of each embryo grayscale image. The division information can be used to quantify the rate of cell division. The larger the value, the faster the cell division, and the greater the difference between it and other embryo grayscale images. In this case, the information content is greater, and it should be screened out in subsequent processes for embryo monitoring.

[0076] It should be noted that for the first embryo grayscale image in the time series, since it does not have an adjacent previous embryo grayscale image, its splitting information is set to a preset value. The preset value here is set to the splitting information of the second embryo grayscale image in the time series. The reason is that the second embryo grayscale image is the closest to the first embryo grayscale image in time, and the similarity between the splitting information will be higher. The Canny operator for obtaining edge pixels is a well-known technique, and the specific process will not be described in detail here.

[0077] Based on the aforementioned process, the amount of splitting information in each embryo grayscale image can be obtained. When selecting high-value reference image data in subsequent processes, it is necessary not only to ensure that the image contains a large amount of information, but also to ensure that the embryo splitting image is clear. Clarity means that the germ cells should have obvious boundary features, which helps to clearly identify the morphology of the germ cells. Whether the boundary features are clear or not can be characterized by the grayscale difference of pixels in a local range, thereby calculating the amount of clarity information in each embryo grayscale image.

[0078] Preferably, in one embodiment of the present invention, the method for obtaining clear information includes:

[0079] The grayscale difference within a local area can better reflect the details and clarity of the image. Therefore, in each embryo grayscale image, a preset neighborhood is determined with each edge pixel as the center, and the other pixels except the center pixel are regarded as the neighborhood pixels of the center pixel.

[0080] When the cell membrane is clear, there should be some non-edge neighboring pixels in the neighborhood of the edge pixel where the cell membrane is located that have obvious gray-level differences with the edge pixel. Therefore, in the preset neighborhood corresponding to each edge pixel, the absolute value of the difference between the gray-level values ​​of the center pixel and each neighboring pixel is calculated as the gray-level difference factor. The larger the gray-level difference factor, the more significant the gray-level performance of the center pixel, and the more obvious the detailed features at the center pixel, and the higher the clarity.

[0081] Then, the sum of all grayscale difference factors of each center pixel is used as the grayscale feature value of each center pixel. Based on the above analysis, it can be concluded that the larger the grayscale feature value, the more prominent the grayscale performance of the center pixel and the higher the clarity.

[0082] Finally, in each embryo grayscale image, the features of local grayscale changes of all edge pixels are combined, and the average value of grayscale feature values ​​of all edge pixels is taken as the clarity information content of the embryo grayscale image. At this time, the greater the clarity information content, the more obvious the grayscale performance of edge pixels in the embryo grayscale image, and the richer the detail performance and the higher the clarity of the embryo grayscale image.

[0083] It should be noted that in this embodiment of the present invention, the size of the preset neighborhood is 3×3, and the specific size can be adjusted according to the implementation scenario, and is not limited here.

[0084] The target embryo image screening module 103 is used to screen target embryo images from all embryo grayscale images based on the difference in the amount of splitting information between all embryo grayscale images and the amount of clear information in the embryo grayscale images.

[0085] Based on the aforementioned module, the amount of splitting information and the amount of clarity information in each embryo grayscale image can be obtained, which reflects the splitting information and clarity in the embryo grayscale image. In the process of embryonic development monitoring, real-time monitoring is a huge workload, and not every embryo grayscale image at every moment is of reference value. Therefore, embryo grayscale images with obvious splitting characteristics and clear images should be selected from a large number of embryo grayscale images as target embryo images for subsequent embryonic development monitoring. Therefore, in this module, the target embryo image is selected from all embryo grayscale images based on the difference in the splitting information and the amount of clarity information of the embryo grayscale images.

[0086] Preferably, in one embodiment of the present invention, the method for acquiring the target embryo image includes:

[0087] Using a single moment's embryo grayscale image to judge the performance of germ cells throughout the entire culture process is very limited. Using embryo grayscale images at all moments is highly redundant and requires a greater amount of computation. Therefore, in this embodiment of the invention, the evaluation time limit for each moment's embryo grayscale image is determined based on the initial information content. That is, all embryo grayscale images are divided into subsequences so that embryo grayscale images with the same division performance are grouped into one subsequence, and then the target embryo image is selected based on the amount of clear information.

[0088] First, based on the differences in the amount of splitting information among all embryonic grayscale images, all embryonic grayscale images are divided into image subsequences. The amount of splitting information represents the change in the amount of edge information of newly formed cell membranes over time during cell division. However, there is not a significant change in edge information at every moment. Therefore, in a series of embryonic grayscale images, the absolute value of the difference in splitting information between each embryonic grayscale image and the next adjacent embryonic grayscale image is calculated. The smaller the absolute value of this difference, the more similar the splitting information between the two embryonic grayscale images, and the more likely they are at the same splitting stage, thus indicating higher consistency. Therefore, this absolute value of the difference is negatively correlated and normalized to correct the logical relationship, thereby obtaining the consistency feature value corresponding to each embryonic grayscale image. The larger the consistency feature value, the less significant the splitting difference between the two embryonic grayscale images, and therefore the greater the probability that they are divided into the same subsequence. The negative correlation mapping and normalization process here can be performed using the formula... ,in, Let x represent an exponential function with the natural constant e as the base, and let x represent the independent variable.

[0089] At this point, each pair of adjacent embryo grayscale images has a consistent feature value, which is used to measure the similarity of their splitting.

[0090] Then, the first grayscale image of the embryo in the time series is used as the image to be tested, and the image to be tested corresponds to an image set.

[0091] When the consistency feature value between the last embryo grayscale image in the image set corresponding to the image to be tested is greater than or equal to the preset consistency threshold, it is considered that they may be in the same division stage. Therefore, the next embryo grayscale image is added to the image set of the image to be tested and the traversal continues.

[0092] Conversely, when the consistency feature value between the last embryo grayscale image in the image set corresponding to the image to be tested is less than the preset consistency threshold, it is considered that they may be in different division stages. The image set corresponding to the image to be tested at this time is taken as an image subsequence, and the next adjacent embryo grayscale image is taken as a new image to be tested and the traversal continues.

[0093] To illustrate the aforementioned method for dividing image subsequences, consider four embryonic grayscale images, denoted as embryonic grayscale images 1-4. Initially, embryonic grayscale image 1 corresponds to an image set a, which contains only embryonic grayscale image 1. When the consistency feature value between embryonic grayscale image 1 and embryonic grayscale image 2 is greater than or equal to a preset consistency threshold, they are considered to belong to the same division stage, and therefore embryonic grayscale image 2 is added to image set a. Then, it is further determined whether the consistency feature value between embryonic grayscale image 2 and embryonic grayscale image 3 is greater than or equal to the preset consistency threshold. If not, the first image subsequence a is obtained, which includes embryonic grayscale image 1 and embryonic grayscale image 2. Embryonic grayscale image 3 is then used as a new image to be tested and the process continues. Embryonic grayscale image 3 corresponds to an image set b, which contains only embryonic grayscale image 3. When the consistency feature value between embryo grayscale image 3 and embryo grayscale image 4 is greater than or equal to the preset consistency threshold, embryo grayscale image 4 is merged into image set b. At this time, the traversal is completed, and two image subsequences a and b are obtained. a contains embryo grayscale images 1 and 2, and b contains embryo grayscale images 3 and 4.

[0094] At this point, all image subsequences can be obtained, and the embryo grayscale images in each image subsequence have relatively consistent splitting characteristics.

[0095] Finally, in each image subsequence, the grayscale image of the embryo with the highest clarity and information content is selected as the target embryo image.

[0096] It should be noted that the preset consistency threshold is set to 0.65, and the specific value can be adjusted according to the implementation scenario, without limitation here.

[0097] The embryo development monitoring module 104 is used to compare the similar features between the edge lines of all target embryo images to identify high-information-content embryo images; and to monitor the embryo development of the embryo to be monitored based on the high-information-content embryo images.

[0098] Given that image data with obvious division behavior characteristics and high analytical value should refer to image data whose corresponding division situation is non-repeatable, that is, the division stage or division morphology is unique, and that among the target embryo images with high clarity and obvious division characteristics selected based on the aforementioned module, there may be image data with similar cell division morphology, in order to further improve monitoring efficiency and reduce data redundancy, in this embodiment of the invention, the similarity features between the edge lines of all target embryo images are further compared, thereby selecting high-information-content embryo images from all target embryo images. At this time, the high-information-content embryo images have low similarity to each other, high clarity, and obvious division behavior characteristics. Therefore, using high-information-content embryo images to monitor embryonic development can effectively reduce data redundancy and improve monitoring efficiency.

[0099] First, high-information-content embryo images can be filtered from all target embryo images.

[0100] Preferably, in one embodiment of the present invention, the method for acquiring high-information-content embryo images includes:

[0101] Please see Figure 2 The diagram illustrates a flowchart of a method for acquiring high-information-content embryo images according to an embodiment of the present invention. The method includes the following steps:

[0102] Step S201: Analyze the similarity between the gray values ​​of edge pixels in the target embryo image to determine the obvious gray value features of each target embryo image.

[0103] The mean gray value of all edge pixels in each target embryo image is used as the gray value feature of each target embryo image. The gray value feature can be used as an indicator to describe the overall gray level of the target embryo image.

[0104] For any target embryo image, the absolute value of the difference between the gray-level feature values ​​of the target embryo image and each other target embryo image is used as the gray-level difference factor. The larger the gray-level difference factor, the more obvious the difference between the overall gray-level average value of the target embryo image and other target embryo images, and thus the more special the division stage it is in.

[0105] The mean of all gray-level difference factors of the target embryo image is taken as the gray-level significant feature value of the target embryo image. The larger the gray-level significant feature value, the more obvious the gray-level difference between the target embryo image and all other target embryo images. Therefore, the stronger the obvious features of cell division, the more necessary it is to be screened out as a high-information embryo image.

[0106] Step S202: Analyze the positional similarity between the edge lines of the target embryo images and determine the obvious positional feature values ​​of each target embryo image.

[0107] The position of the edge line reflects the spatial distribution of cell membranes in the target embryo image. Therefore, by analyzing the similarity of positions, the spatial distribution characteristics of cell membranes between target embryo images can be identified.

[0108] In each target embryo image, a two-dimensional rectangular coordinate system is constructed with the lower left corner of the image as the origin, the horizontal direction to the right as the horizontal axis, and the vertical direction upward as the vertical axis, so that each edge pixel in each target embryo image corresponds to a two-dimensional coordinate.

[0109] Then, based on the least squares method, a straight line is fitted to the position coordinates of all edge pixels, and the slope value of the fitted line is obtained. The slope value can reflect the directional characteristics of the spatial distribution of all edge pixels in the target embryo image.

[0110] For any target embryo image, the absolute value of the difference between the slope of the fitted line of the target embryo image and each other target embryo image is used as the position difference factor of the target embryo image. The larger the position difference factor, the greater the difference in the spatial distribution of the edge pixels of the target embryo image and other target embryo images. In this case, the higher its non-repetition, the more it needs to be screened out as a high-information embryo image.

[0111] Finally, the mean of all positional difference factors of the target embryo image is taken as the positional significance value of the target embryo image. The larger the positional significance value, the stronger the obvious cell division feature of the target embryo image, and the higher the probability that the division stage is unique.

[0112] It should be noted that obtaining the fitted line and its slope based on the least squares method are well-known techniques, and the specific process will not be elaborated here.

[0113] Step S203: Combine the obvious grayscale feature values ​​and obvious positional feature values ​​of each target embryo image to determine the obvious division feature values ​​for screening high-information embryo images.

[0114] Analysis shows a positive correlation between the distinct grayscale and positional features of a target embryo image and its probability of being a high-information embryo image. Therefore, the normalized sum of the distinct grayscale and positional features of each target embryo image is used as the distinct cell division feature value. A larger distinct cell division feature value indicates that the target embryo image possesses distinct cell division characteristics and is more likely to be in a unique cell division stage. Normalization is a technique well-known to those skilled in the art, and the normalization function can be linear normalization or standard normalization, etc. Specific normalization methods are not limited here.

[0115] Finally, among all target embryo images, those with a significant cell division feature value greater than a preset significant cell division threshold were designated as high-information-content embryo images. These high-information-content embryo images exhibited significant cell division characteristics. (See also...) Figure 3 It shows a schematic diagram of a high-information embryo image.

[0116] It should be noted that the preset splitting threshold in this embodiment of the present invention is 0.7. The specific value can be adjusted according to the implementation scenario and is not limited here.

[0117] Based on the aforementioned process, high-information-content embryo images can be selected from a series of grayscale embryo images. These images typically contain key information and features during embryonic development, thus enabling embryonic development monitoring of the embryos to be monitored based on these high-information-content embryo images.

[0118] Preferably, in one embodiment of the present invention, monitoring embryonic development based on high-information-content embryo images includes:

[0119] Each high-information-content embryo image is input into a pre-trained neural network, which outputs the division cycle of each high-information-content embryo image (such as fertilized egg, 2-cell stage, 4-cell stage, morula, blastocyst, etc.).

[0120] Obtain the time period corresponding to the division cycle of each high-information embryo image (e.g., the 2-cell stage is 25-30 hours after fertilization).

[0121] Then, for any high-information-content embryo image, if the corresponding time is within the time period corresponding to the division cycle, it is considered that the embryo under monitoring is developing normally at that time and no warning is needed; if the corresponding time is not within the time period corresponding to the division cycle, it is considered that the embryo under monitoring is developing abnormally at that time and a warning is needed, and the doctor should be notified for further judgment.

[0122] It should be noted that the neural network in this embodiment of the invention can be a CNN neural network. A large number of high-quality, high-resolution embryo images are collected. These embryo images should cover typical features of different cell division cycles. The cell division cycles are then labeled, and the labeled dataset is used to train the neural network. The training process of the neural network is a well-known technique, and the specific process will not be elaborated here.

[0123] In summary, this embodiment of the invention first acquires a series of grayscale images of the embryos to be monitored in a temporal sequence using an image data acquisition module. During cell division, not every germ cell possesses valid information, and images from certain moments have low reference value. Therefore, in this embodiment, the amount of information contained in the embryo grayscale images can be analyzed to select high-information-content embryo images for final embryo development monitoring. Given that cell division generates numerous new edges, and the generation of these new edges leads to changes in grayscale, the image information content analysis module, to ensure the accuracy and clarity of subsequent image selection, precisely quantifies the division information content of each embryo grayscale image by utilizing the changes in the number of edge pixels, providing a strong basis for subsequent embryo development assessment. Simultaneously, based on the grayscale differences of pixels within a local area, the clarity information content of each embryo grayscale image is determined, ensuring high-resolution images for monitoring. Then, in the target embryo image selection module, representative target embryo images with high information content are selected from all embryo grayscale images based on the differences in division information content and the clarity information content. Finally, in the embryo development monitoring module, since analyzing images within the same division stage would also generate data redundancy, we compared the similar features between the edge lines of all target embryo images to identify high-information embryo images. At this point, embryo development monitoring based on these images can effectively reduce data redundancy, thereby providing doctors with rich information on embryo development and helping them make timely and accurate judgments based on the embryo development monitoring results.

[0124] It should be noted that the order of the above embodiments of the present invention is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. The processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired result. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0125] The various embodiments in this specification are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

[0126] 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, improvements, etc., made within the principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A real-time monitoring and culture system for embryonic development during assisted reproductive technology, characterized in that, The system includes: The image data acquisition module is used to acquire a series of grayscale images of the embryos to be monitored in a time sequence. The image information content analysis module is used to determine the splitting information content of each embryo grayscale image based on the change in the number of edge pixels; and to determine the clarity information content of each embryo grayscale image based on the grayscale difference of pixels in a local range. The target embryo image screening module is used to screen target embryo images from all embryo grayscale images based on the difference in the amount of splitting information between all embryo grayscale images and the amount of clear information in the embryo grayscale images. The embryo development monitoring module is used to compare the similar features between the edge lines of all target embryo images to identify high-information-content embryo images; and to monitor the embryo development of the embryos to be monitored based on the high-information-content embryo images.

2. The real-time monitoring and culture system for embryonic development in assisted reproductive technology according to claim 1, characterized in that, The method for obtaining the split information includes: The absolute value of the difference between the total number of edge pixels in each embryo grayscale image and the previous embryo grayscale image in time sequence is used as the quantity change factor for each embryo grayscale image. The time interval between each embryo grayscale image and the previous embryo grayscale image in time sequence is used as the time interval factor corresponding to each embryo grayscale image; The ratio between the quantity change factor and the time interval factor corresponding to each embryo grayscale image is used as the splitting information content of each embryo grayscale image.

3. The real-time monitoring and culture system for embryonic development in assisted reproductive technology according to claim 1, characterized in that, The method for obtaining the amount of clear information includes: In each grayscale image of an embryo, a preset neighborhood is determined with each edge pixel as the center; In the preset neighborhood corresponding to each edge pixel, the absolute value of the difference between the gray values ​​of the center pixel and each neighboring pixel is calculated and used as the gray-level difference factor. The sum of all gray-level difference factors of each center pixel is used as the gray-level feature value of each center pixel; The mean of the grayscale feature values ​​of all edge pixels in each embryo grayscale image is used as the amount of clear information.

4. The real-time monitoring and culture system for embryonic development in assisted reproductive technology according to claim 1, characterized in that, The method for acquiring the target embryo image includes: Based on the differences in the amount of splitting information among all embryo grayscale images, all embryo grayscale images are divided to obtain image subsequences; In each image subsequence, the grayscale image of the embryo with the highest clarity information is selected as the target embryo image.

5. The real-time monitoring and culture system for embryonic development in assisted reproductive technology according to claim 4, characterized in that, The method for obtaining the image subsequence includes: In a series of embryo grayscale images, the absolute value of the difference between the splitting information content of each embryo grayscale image and the next adjacent embryo grayscale image is negatively correlated and normalized, and the result is used as the consistent feature value. In terms of time sequence, the first grayscale image of the embryo is used as the image to be tested, and the image to be tested corresponds to an image set; When the consistency feature value between the last embryo grayscale image in the image set corresponding to the image to be tested and the next adjacent embryo grayscale image is greater than or equal to the preset consistency threshold, the next adjacent embryo grayscale image is merged into the image set of the image to be tested and the traversal continues. When the consistency feature value between the last embryo grayscale image in the image set corresponding to the image to be tested and the next adjacent embryo grayscale image is less than the preset consistency threshold, the image set corresponding to the image to be tested at this time is taken as an image subsequence, and the next adjacent embryo grayscale image is taken as a new image to be tested and the traversal continues. After the traversal is complete, all image subsequences are obtained.

6. The real-time monitoring and culture system for embryonic development in assisted reproductive technology according to claim 1, characterized in that, The method for acquiring the high-information-content embryo image includes: The mean gray value of all edge pixels in each target embryo image is used as the gray feature value of each target embryo image; For any target embryo image, the absolute value of the difference between the gray-level feature values ​​of the target embryo image and every other target embryo image is used as the gray-level difference factor; The mean value of all gray-level difference factors of the target embryo image is taken as the gray-level significant feature value of the target embryo image; Analyze the positional similarity between the edge lines of the target embryo images to determine the obvious positional feature values ​​of each target embryo image; The sum of the grayscale and positional features of each target embryo image is normalized and used as the splitting feature value of that target embryo image. Among all target embryo images, those with a significant division feature value greater than a preset significant division threshold are designated as high-information embryo images.

7. The real-time monitoring and culture system for embryonic development in assisted reproductive technology according to claim 6, characterized in that, The method for obtaining the obvious location feature values ​​includes: In each target embryo image, the position coordinates of all edge pixels are fitted with a straight line based on the least squares method, and the slope value of the fitted line is obtained. For any target embryo image, the absolute value of the difference between the slope of the fitted line of the target embryo image and that of each other target embryo image is used as the positional difference factor of the target embryo image. The mean of all positional difference factors in the target embryo image is taken as the positional characteristic value of the target embryo image.

8. The real-time monitoring and culture system for embryonic development in assisted reproductive technology according to claim 1, characterized in that, The embryonic development monitoring based on high-information-content embryo images includes: Each high-information-content embryo image is input into a pre-trained neural network, which outputs the division cycle of each high-information-content embryo image. Obtain the time period corresponding to the division cycle of each high-information-content embryo image; For any high-information-content embryo image, if the corresponding time point falls within the time period corresponding to the division cycle, the embryo under monitoring is considered to be developing normally at that time point; if the corresponding time point does not fall within the time period corresponding to the division cycle, the embryo under monitoring is considered to be developing abnormally at that time point.

9. The real-time monitoring and culture system for embryonic development in assisted reproductive technology according to claim 1, characterized in that, The method for obtaining the grayscale image of the embryo includes: In terms of time sequence, a series of grayscale images of the embryos to be monitored are acquired; Each embryo grayscale image is segmented using a pre-trained semantic segmentation network to obtain the corresponding embryo grayscale image.

10. The real-time monitoring and culture system for embryonic development in assisted reproductive technology according to claim 1, characterized in that, The method for obtaining the edge pixels includes: Edge pixels in each embryo grayscale image are obtained based on the Canny operator.

Images