Embryo quality evaluation method, device and equipment and computer readable storage medium

Abstract

The invention provides an embryo quality evaluation method, device and equipment and a computer readable storage medium, and relates to the technical field of medical treatment. The method comprises the following steps: firstly, obtaining a three-dimensional reconstruction model of a to-be-evaluated embryo in a preset cultivation stage; secondly, performing feature extraction on the three-dimensional reconstruction model to obtain initial interaction features of trophoblast cells and inner cell clusters in the embryo to be evaluated; thirdly, performing feature mining on the initial interaction features to obtain target interaction features of the trophoblast cells and the inner cell clusters; and finally, according to the target interaction characteristics, performing quality evaluation on the to-be-evaluated embryo by adopting a preset embryo quality evaluation model to obtain a predicted value of a quality evaluation parameter of the to-be-evaluated embryo. Therefore, the embryo with the most developmental potential can be more accurately selected clinically.

Description

Technical Field

[0001] The present invention relates to the field of medical technology, and in particular to an embryo quality assessment method, device, equipment and computer-readable storage medium. Background Art

[0002] With the rapid development of computer vision and assisted reproductive technology, the application of artificial intelligence in the field of assisted reproduction is becoming increasingly in-depth. How to simulate the spatial structure of the blastocyst through three-dimensional reconstruction technology without damaging the embryo, thereby revealing information that cannot be observed from a two-dimensional perspective, has always been one of the focuses of embryology experts.

[0003] Since its introduction into the field of embryo culture, time-difference culture technology has brought revolutionary changes to assisted reproductive technology, overturned the traditional working mode of embryologists, and greatly improved work efficiency. With the help of time-difference culture technology, not only can more embryo characteristics be obtained, but it also facilitates the formation of a variety of embryo scoring systems. In addition, the mature application of time-difference culture technology in the field of assisted reproduction has laid a solid foundation for non-invasive three-dimensional reconstruction of embryos. Through the embryo three-dimensional reconstruction technology based on time-difference culture imaging data, the three-dimensional spatial structure of the blastocyst can be accurately simulated, which provides a new way to deeply explore the characteristics of the blastocyst period and lays a data foundation for building a quality assessment model.

[0004] Shan, G. (2022) applied the 3D reconstruction technology of blastocyst surface based on surface reconstruction to extract some 3D trophoblast information, including the number of trophoblast cells (TE), cell density, cell size, and the size of inner cell mass (ICM) (2D features), and created a scoring system for predicting PGT results. The 3D reconstruction method proposed by Shan, G. (2022) can be briefly summarized as unfolding a 3D spherical structure. The specific implementation scheme is: fix the embryo with two needles, then rotate it at a fixed angle each time, match the pixel position of the acquired image, and segment the trophoblast and inner cell mass, so as to extract the features of the inner cell mass and trophoblast cells, and construct a PGT prediction model based on the extracted features. This 3D reconstruction method is more inclined to obtain more complete trophoblast cell features, but it does not further study and measure the relationship between the two important subjects of the inner cell mass and trophoblast cells in the three-dimensional spatial structure. Summary of the invention

[0005] The purpose of the present invention is to provide an embryo quality assessment method, device, equipment and computer-readable storage medium to address the deficiencies in the above-mentioned prior art, so as to help clinicians more accurately select embryos with the greatest developmental potential.

[0006] To achieve the above purpose, the technical solution adopted in the embodiment of the present application is as follows:

[0007] In a first aspect, an embodiment of the present application provides an embryo quality assessment method, comprising obtaining a three-dimensional reconstructed model of the embryo to be assessed at a preset cultivation stage; performing feature extraction on the three-dimensional reconstructed model to obtain initial interaction characteristics of trophoblast cells and inner cell mass in the embryo to be assessed; performing feature mining on the initial interaction characteristics to obtain target interaction characteristics of the trophoblast cells and inner cell mass; and performing quality assessment on the embryo to be assessed using a preset embryo quality assessment model based on the target interaction characteristics to obtain predicted values ​​of quality assessment parameters of the embryo to be assessed.

[0008] In one embodiment, the feature mining of the initial interaction feature to obtain the target interaction feature of the trophoblast cells and inner cell mass includes: adjusting the self-distribution of the initial interaction feature to obtain a first interaction feature; the target interaction feature includes the first interaction feature.

[0009] In one embodiment, the adjusting the self-distribution of the initial interaction feature to obtain the first interaction feature includes: using a first preset algorithm to perform computational processing on the initial interaction feature to obtain the first interaction feature.

[0010] In one embodiment, the adjusting the self-distribution of the initial interaction feature to obtain the first interaction feature includes: determining a nonlinear relationship expression of the initial interaction feature; using the nonlinear relationship expression to perform a polynomial expansion on the initial interaction feature to obtain the first interaction feature.

[0011] In one embodiment, feature mining is performed on the initial interaction features to obtain target interaction features of the trophoblast cells and inner cell mass, including: based on the interaction between at least two of the initial interaction features, feature combination of different initial interaction features is performed to obtain a second interaction feature; the target interaction feature includes the second interaction feature.

[0012] In one embodiment, based on the interaction between at least two of the initial interaction features, feature combination is performed on different initial interaction features to obtain a second interaction feature, including: based on the interaction between the first initial interaction feature and the second initial interaction feature, using a second preset algorithm to perform a logistic regression operation to obtain the second interaction feature.

[0013] In one embodiment, based on the interaction between at least two of the initial interaction features, the features of different initial interaction features are combined to obtain a second interaction feature, including: discretizing the third initial interaction feature to obtain at least two types of classification data; and arranging and combining the classification data, other classification data and the initial interaction features to obtain the second interaction feature.

[0014] In one embodiment, before performing quality assessment on the embryo to be assessed based on the target interaction feature using a preset embryo quality assessment model to obtain a predicted value of the quality assessment parameter of the embryo to be assessed, the method further includes: performing feature extraction on clinical embryo data to obtain a clinical initial interaction feature data set; performing feature mining on each clinical initial interaction feature in the clinical initial interaction feature data set to obtain a plurality of clinical interaction features; screening the plurality of clinical interaction features to obtain a clinical target interaction feature; performing feature analysis on the clinical target interaction feature based on the clinical initial interaction feature data set to obtain a correlation between the clinical target interaction feature and the preset clinical quality assessment parameter; constructing sample data based on the correlation; and performing model training on a preset basic embryo quality assessment model using the sample data to obtain the embryo quality assessment model.

[0015] In a second aspect, an embodiment of the present application further provides an embryo quality assessment device, comprising: an acquisition module, configured to acquire a three-dimensional reconstructed model of the embryo to be assessed at a preset cultivation stage; a feature extraction module, configured to perform feature extraction on the three-dimensional reconstructed model to obtain initial interaction characteristics of trophoblast cells and inner cell mass in the embryo to be assessed; a feature mining module, configured to perform feature mining on the initial interaction characteristics to obtain target interaction characteristics of the trophoblast cells and inner cell mass; a prediction module, configured to perform quality assessment on the embryo to be assessed using a preset embryo quality assessment model based on the target interaction characteristics to obtain predicted values ​​of quality assessment parameters of the embryo to be assessed.

[0016] In one embodiment, the feature mining module is configured to: adjust the self-distribution of the initial interaction feature to obtain a first interaction feature; and the target interaction feature includes the first interaction feature.

[0017] In one embodiment, the feature mining module is configured to: use a first preset algorithm to perform computational processing on the initial interaction feature to obtain the first interaction feature.

[0018] In one embodiment, the feature mining module is configured to: determine a nonlinear relationship expression of the initial interaction feature; and use the nonlinear relationship expression to perform a polynomial expansion on the initial interaction feature to obtain the first interaction feature.

[0019] In one embodiment, the feature mining module is configured to: perform feature combination on different initial interaction features according to the interaction between at least two of the initial interaction features to obtain a second interaction feature; the target interaction feature includes the second interaction feature.

[0020] In one embodiment, the feature mining module is configured to: perform a logistic regression operation using a second preset algorithm based on the interaction between the first initial interaction feature and the second initial interaction feature to obtain the second interaction feature.

[0021] In one embodiment, the feature mining module is configured to: discretize the third initial interaction feature to obtain at least two categories of classification data; and arrange and combine the classification data, other classification data and the initial interaction feature to obtain the second interaction feature.

[0022] In one embodiment, the embryo quality assessment device also includes a model training module, which is configured to perform feature extraction on clinical embryo data to obtain a clinical initial interaction feature data set; perform feature mining on each clinical initial interaction feature in the clinical initial interaction feature data set to obtain multiple clinical interaction features; screen the multiple clinical interaction features to obtain clinical target interaction features; perform feature analysis on the clinical target interaction features based on the clinical initial interaction feature data set to obtain a correlation between the clinical target interaction features and the preset clinical quality assessment parameters; construct sample data based on the correlation; and use the sample data to perform model training on a preset basic embryo quality assessment model to obtain the embryo quality assessment model.

[0023] In a third aspect, an embodiment of the present application further provides a computer device, comprising: a processor, a storage medium and a bus, wherein the storage medium stores program instructions executable by the processor, and when the computer device is running, the processor communicates with the storage medium via the bus, and the processor executes the program instructions to perform the steps of any one of the above methods.

[0024] In a fourth aspect, an embodiment of the present application further 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 any one of the above methods are executed.

[0025] The beneficial effects of the present application are as follows: first, a three-dimensional reconstruction model of the embryo to be evaluated in a preset cultivation stage is obtained; second, feature extraction is performed on the three-dimensional reconstruction model to obtain the initial interaction characteristics of the trophoblast cells and the inner cell mass in the embryo to be evaluated; third, feature mining is performed on the initial interaction characteristics to obtain the target interaction characteristics of the trophoblast cells and the inner cell mass; finally, according to the target interaction characteristics, a preset embryo quality assessment model is used to perform quality assessment on the embryo to be evaluated to obtain the predicted value of the quality assessment parameter of the embryo to be evaluated. In this way, by mining the initial interaction characteristics, the target interaction characteristics are obtained, and quality assessment is performed based on the target interaction characteristics, so that it can help the clinic to more accurately select the embryo with the most developmental potential. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for use in the embodiments are briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present invention and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without creative work.

[0027] Figure 1 A schematic diagram of a process for evaluating embryo quality provided in an embodiment of the present application;

[0028] Figure 2 A schematic diagram of a process for evaluating embryo quality provided in an embodiment of the present application;

[0029] Figure 3 A schematic diagram of a process for evaluating embryo quality provided in an embodiment of the present application;

[0030] Figure 4 A schematic diagram of a process for evaluating embryo quality provided in an embodiment of the present application;

[0031] Figure 5 A schematic diagram of the structure of an embryo quality assessment device provided in an embodiment of the present application;

[0032] Figure 6 A schematic diagram of the structure of a computer device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0033] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions 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.

[0034] Therefore, the following detailed description of the embodiments of the present application provided in the accompanying drawings is not intended to limit the scope of the present application for which protection is sought, but merely represents selected embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in the field without creative work are within the scope of protection of the present application.

[0035] In the description of the present application, it should be noted that if the terms "upper", "lower", etc. appear, the orientation or position relationship indicated is based on the orientation or position relationship shown in the drawings, or is the orientation or position relationship in which the product of the application is usually placed when used. It is only for the convenience of describing the present application and simplifying the description, and does not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation on the present application.

[0036] In addition, the terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0037] It should be noted that, in the absence of conflict, the features in the embodiments of the present application may be combined with each other.

[0038] Figure 1 A schematic diagram of a method for evaluating embryo quality provided in an embodiment of the present application; Figure 1 As shown, the method includes:

[0039] Step 110: Obtain a three-dimensional reconstructed model of the embryo to be evaluated at a preset culture stage.

[0040] The three-dimensional reconstruction model can be obtained by performing three-dimensional spatial reconstruction on the blastocyst image data through three-dimensional spatial reconstruction technology.

[0041] 3D space reconstruction technology is a technology that uses a camera to scan or photograph an object or scene and uses a computer algorithm to convert the 2D photo data into an interactive 3D model. This technology mainly relies on deep learning methods, using the powerful learning and fitting capabilities of deep neural networks to process RGB or RGBD images to restore the 3D model of the object.

[0042] The preset culture stage can be the blastocyst stage. In actual operation, blastocyst image data with 11 or more focal planes under the time-lapse culture system (Hoffman imaging) can be selected for three-dimensional spatial reconstruction. It should be noted that when selecting blastocyst image data, the inner cell mass must be within the field of view, the number of focal planes where the inner cell mass is located must be greater than 3, and there must be fewer fragments (too many fragments will affect the accuracy of the three-dimensional reconstruction model).

[0043] Step 120: extract features from the three-dimensional reconstructed model to obtain initial interaction features between trophoblast cells and inner cell mass in the embryo to be evaluated.

[0044] Among them, the initial interaction characteristics of the trophoblast cells and the inner cell mass are characteristics defined and proposed in an embryo quality assessment method provided in an embodiment of the present application, and are used to represent the association relationship and mutual relationship between the inner cell mass and the trophoblast cells; for example: the distance between the inner cell mass and the trophoblast cells, the contact area between the inner cell mass and the trophoblast cells, and the contact ratio between the inner cell mass and the trophoblast cells;

[0045] The contact ratio is the ratio of the contact area to the surface area of ​​the inner cell mass of the three-dimensional reconstruction model; the distance and contact ratio are both used to characterize the degree of squeezing between the inner cell mass and the trophoblast cells.

[0046] Furthermore, if the first three-dimensional quadrant and the second three-dimensional quadrant are the same quadrant, it can be represented by 1, otherwise, it can be represented by 0. If the adjacent quadrant of the first three-dimensional quadrant and the second three-dimensional quadrant are the same quadrant, it can also be represented by 1, otherwise, it can be represented by 0. Of course, it can also be represented by other forms, such as letters, text, etc., which are not limited here.

[0047] Generally, the quadrant in which the center of gravity of the inner cell mass is located determines the quadrant in which the inner cell mass is located; for example: if the center of gravity of the inner cell mass is in the first quadrant, then the quadrant in which the inner cell mass is located is the first quadrant.

[0048] In actual operation, an xyz three-dimensional coordinate system can be established with the center of gravity of the blastocyst as the center of the circle, and the space of the three-dimensional reconstructed model can be divided into 8 quadrants.

[0049] This step can also extract the characteristics of trophoblast cells and inner cell mass. The characteristics of trophoblast cells and inner cell mass are consistent with the characteristics extracted in related technologies, such as: the number of trophoblast cells, the density of trophoblast cells, the surface area of ​​the inner cell mass, the volume of the inner cell mass, etc.

[0050] Step 130: Perform feature mining on the initial interaction features to obtain target interaction features of trophoblast cells and inner cell mass.

[0051] Among them, feature mining can be understood as further processing the initial interaction features to obtain other features of the initial interaction features. Here, feature mining is mainly carried out from two perspectives: the distribution of the features themselves and the interaction between features. Specifically, feature mining from the distribution of the features themselves can mine the nonlinear relationship between the initial interaction features, and can also mine the high-dimensional features of the initial interaction features. Feature mining from the interaction between features can be a combination of different initial interaction features.

[0052] In actual operation, the continuous initial interaction features can also be discretized to mine the initial interaction features.

[0053] Step 140: Based on the target interaction characteristics, a preset embryo quality assessment model is used to perform quality assessment on the embryo to be assessed, and a predicted value of a quality assessment parameter of the embryo to be assessed is obtained.

[0054] Among them, quality assessment models include decision tree prediction models, logistic regression prediction models, etc.

[0055] In actual operation, the Gardner scoring system can be used as a performance baseline, and the cell distribution characteristics defined and proposed by an embryo quality assessment method provided in an embodiment of the present application can be added to the Gardner scoring system to establish a decision tree prediction model for performance comparison.

[0056] The Gardner scoring system is a method used to evaluate the potential of embryonic development. It is mainly based on the expansion state of the blastocyst, the development of the inner cell mass and the trophectoderm. This system was proposed by David Gardner and is currently one of the most commonly used blastocyst scoring methods. Specifically, the Gardner scoring system includes the following aspects:

[0057] 1. Expansion degree of blastocyst cavity: According to the size of blastocyst cavity and whether it is hatched, the development of blastocyst is divided into six periods, from early chambered blastocyst to hatched blastocyst.

[0058] 2. Grading of inner cell mass: According to the number of cells and the density of arrangement, the inner cell mass is divided into three grades: A, B, and C. Grade A indicates a large number of cells with dense arrangement, while grade C indicates a small number of cells.

[0059] 3. Grading of trophoblast cells: Trophoblast cells are divided into three grades: A, B, and C, also based on the number of cells and the density of arrangement.

[0060] The Gardner scoring system is widely used and is mainly used to evaluate the quality and developmental potential of embryos. Blastocysts with higher scores (such as 4AA, 4BB, etc.) are considered high-quality embryos with a higher transplant success rate. In addition, the developmental stage of the blastocyst, the grading of the inner cell mass and trophoblast cells are also used to determine whether the embryo is suitable for transplantation or cryopreservation.

[0061] Through the Gardner scoring system, doctors can more accurately assess the quality of embryos, thereby improving the success rate of IVF. At the same time, the system also provides patients with more choices and hope, making assisted reproductive technology more scientific and effective.

[0062] The embryo quality assessment method provided in the embodiment of the present application, first, obtains a three-dimensional reconstruction model of the embryo to be assessed in a preset cultivation stage; secondly, extracts features from the three-dimensional reconstruction model to obtain the initial interaction features of the trophoblast cells and the inner cell mass in the embryo to be assessed; thirdly, performs feature mining on the initial interaction features to obtain the target interaction features of the trophoblast cells and the inner cell mass; finally, according to the target interaction features, a preset embryo quality assessment model is used to perform quality assessment on the embryo to be assessed, and the predicted value of the quality assessment parameter of the embryo to be assessed is obtained. In this way, by mining the initial interaction features, the target interaction features are obtained, and quality assessment is performed based on the target interaction features, so that it can help the clinic to more accurately select the embryo with the most developmental potential.

[0063] The following is an explanation of the feature mining process in the above embodiment with examples.

[0064] The above step 130 may further include the following steps:

[0065] The self-distribution of the initial interaction feature is adjusted to obtain a first interaction feature; the target interaction feature includes the first interaction feature.

[0066] This step is the feature mining from the distribution of the features themselves. Here, the nonlinear relationship between the initial interaction features can be mined, and the high-dimensional features of the initial interaction features can also be mined.

[0067] The following is an example to explain the feature mining process in the above embodiment. The above steps may further include the following steps:

[0068] The first preset algorithm is used to perform computational processing on the initial interaction feature to obtain a first interaction feature.

[0069] Among them, this step is to mine the high-dimensional features of the initial interaction features in the feature mining from the distribution angle of the features themselves.

[0070] The first preset algorithm may include: addition, multiplication, division, exponential function, etc.

[0071] Exemplarily, the contact area between ICM and TE*the distance between ICM and TE, log(contact ratio between ICM and TE)+log(distance between ICM and TE), wherein the use of the exponential function can avoid the situation where the value after multiplication of smaller numbers is smaller (ie, numerical underflow).

[0072] The following is an explanation of the feature mining process in the above embodiment with examples. Figure 2 This is a schematic diagram of the process of the embryo quality assessment method provided in the embodiment of the present application. Figure 2 As shown, the above steps may further include the following steps 210 and 220:

[0073] Step 210: Determine the nonlinear relationship expression of the initial interaction characteristics.

[0074] Among them, this step is to mine the nonlinear relationship between the initial interaction features in the feature mining from the distribution angle of the features themselves.

[0075] For example, the nonlinear relationship expression can be the distance between ICM and TE 2 , the distance between ICM and TE 3 , ICM and TE contact ratio 2 , ICM and TE contact ratio 3 , the distance between ICM and TE*the contact ratio between ICM and TE, the distance between ICM and TE 2 *Contact ratio of ICM and TE 2 wait.

[0076] Step 220: Use a nonlinear relationship expression to perform polynomial expansion on the initial interaction characteristic to obtain a first interaction characteristic.

[0077] Among them, this step is to perform polynomial expansion on the nonlinear relationship expression of the initial interaction characteristics. In actual operation, the high-order term characteristics of the initial interaction characteristics can be obtained.

[0078] For example, the distance between ICM and TE 2 Expand to get the distance between ICM and TE*the distance between ICM and TE; the distance between ICM and TE 2 *Contact ratio of ICM and TE 2 Expand to get the distance between ICM and TE * the distance between ICM and TE * the contact ratio between ICM and TE * the contact ratio between ICM and TE, etc.

[0079] The following is an explanation of the feature mining process in the above embodiment with examples.

[0080] The above step 130 may further include the following steps:

[0081] According to the interaction between at least two initial interaction features, feature combination is performed on different initial interaction features to obtain a second interaction feature; the target interaction feature includes the second interaction feature.

[0082] Among them, this step is the above-mentioned feature mining from the perspective of interaction between features.

[0083] The feature combination can be to define the interaction between ICM and TE in the initial interaction feature as a combination of two or three different partial weights, that is, a comprehensive measure of the interaction in three-dimensional space; or it can be to arrange and combine different initial interaction features.

[0084] The following is an example to explain the feature mining process in the above embodiment. The above steps may further include the following steps:

[0085] According to the interaction between the first initial interaction feature and the second initial interaction feature, a logistic regression operation is performed using a second preset algorithm to obtain a second interaction feature.

[0086] Among them, this step, namely, defining the interaction between ICM and TE in the initial interaction characteristics as a combination of two or three different partial weights, is a comprehensive measure of the interaction in three-dimensional space.

[0087] Exemplarily, the contact ratio between ICM and TE*0.1+the distance between ICM and TE*0.9, the contact ratio between ICM and TE*the distance between ICM and TE*the ratio of the number of TE cells in the quadrant where the ICM is located, and then the values ​​are assigned by logistic regression.

[0088] The following is an explanation of the feature mining process in the above embodiment with examples. Figure 3 This is a schematic diagram of the process of the embryo quality assessment method provided in the embodiment of the present application. Figure 3 As shown, the above steps may further include the following steps 310 and 320:

[0089] Step 310: discretize the third initial interaction feature to obtain at least two types of classification data.

[0090] Among them, step 310 and step 320 are the above-mentioned arrangement and combination of different initial interaction features.

[0091] Among the initial interaction features, some feature data are continuous feature data, and some feature data are discrete feature data. In actual operation, the continuous feature data can be processed into discrete feature data through decision trees, quantiles or other discretization methods to obtain classified data.

[0092] Step 320: Arrange and combine the classification data, other classification data and the initial interaction feature to obtain a second interaction feature.

[0093] Among them, the classification data is the discrete feature data after the continuous feature data in the initial interaction feature in the above step 310 is discretized; for example: the number of trophoblast cells in the first three-dimensional quadrant where the inner cell mass is located, the proportion of the number of trophoblast cells in the first three-dimensional quadrant, the total number of trophoblast cells, the density of trophoblast cells; the volume of the inner cell mass and the surface area of ​​the inner cell mass are the continuous feature data in the initial interaction feature.

[0094] Other classification data are discrete feature data in the initial interaction features, such as: the first spatial position relationship between the first three-dimensional quadrant and the second three-dimensional quadrant with the largest number of trophoblast cells, and the second spatial position relationship between the first three-dimensional quadrant or its adjacent quadrant and the second three-dimensional quadrant.

[0095] The following is an explanation of the training process of the embryo quality assessment model in the above embodiment with reference to examples. Figure 4 This is a schematic diagram of the process of the embryo quality assessment method provided in the embodiment of the present application. Figure 4 As shown, the embryo quality assessment method provided in the embodiment of the present application may further include the following steps 410 to 460:

[0096] Step 410: extract features from clinical embryo data to obtain a clinical initial interaction feature data set.

[0097] Among them, the three-dimensional reconstruction technology can be used to simulate the three-dimensional reconstruction model corresponding to the clinical embryo data based on the clinical embryo data, and then the features of the three-dimensional reconstruction model can be extracted to obtain the clinical initial interaction feature data set.

[0098] The feature extraction method can adopt the feature extraction method in the related technology, which will not be described here.

[0099] Step 420: Perform feature mining on each clinical initial interaction feature in the clinical initial interaction feature data set to obtain multiple clinical interaction features.

[0100] Among them, feature mining can be understood as further processing the initial interaction features to obtain other features of the initial interaction features. Here, feature mining is mainly carried out from two perspectives: the distribution of the features themselves and the interaction between features. Specifically, feature mining from the distribution of the features themselves can mine the nonlinear relationship between the initial interaction features, and can also mine the high-dimensional features of the initial interaction features. Feature mining from the interaction between features can be a combination of different initial interaction features.

[0101] This step is essentially the same as the above step 130 and will not be described again here.

[0102] Step 430: Screen multiple clinical interaction features to obtain clinical target interaction features.

[0103] Among them, screening can be based on features, and interaction features (i.e., features including at least two dimensions of inner cell mass and trophoblast cells) are selected as target interaction features; for example: the number of trophoblast cells in the first three-dimensional quadrant where the inner cell mass is located, the proportion of trophoblast cells in the first three-dimensional quadrant, the first spatial position relationship between the first three-dimensional quadrant and the second three-dimensional quadrant with the largest number of trophoblast cells, and the second spatial position relationship between the first three-dimensional quadrant or the adjacent quadrant of the first three-dimensional quadrant and the second three-dimensional quadrant; the distance between the inner cell mass and the trophoblast cells, the contact area between the inner cell mass and the trophoblast cells, and the contact proportion between the inner cell mass and the trophoblast cells.

[0104] Step 440: Perform feature analysis on the clinical target interaction features based on the clinical initial interaction feature data set to obtain the correlation between the clinical target interaction features and the preset clinical quality assessment parameters.

[0105] Among them, the preset clinical quality assessment parameters include true clinical pregnancy rate, true clinical live birth rate, true clinical live birth rate, etc.

[0106] Feature analysis can be understood as the correlation analysis between features and preset clinical quality assessment parameters. Specifically, it is to analyze the correlation between clinical target interaction features and preset clinical quality assessment parameters.

[0107] In actual operation, the Mannwhitney variance can be used to analyze continuous feature data, and the chi-square test can be used to analyze discrete feature data. Generally, the number of trophoblast cells in the first three-dimensional quadrant where the inner cell mass is located, the proportion of the number of trophoblast cells in the first three-dimensional quadrant, the total number of trophoblast cells, and the density of trophoblast cells; the volume of the inner cell mass and the surface area of ​​the inner cell mass are continuous data, and the first spatial position relationship between the first three-dimensional quadrant and the second three-dimensional quadrant with the largest number of trophoblast cells, and the second spatial position relationship between the first three-dimensional quadrant or its adjacent quadrant and the second three-dimensional quadrant are discrete data. Continuous data can be discretized into discrete data, for example: discretization of continuous data by equal-width discretization, equal-frequency discretization, quantile discretization, cluster discretization, custom discretization, discretization based on statistical parameters, discretization based on decision trees, decision tree binning, and other discretization methods.

[0108] The feature analysis results are as follows:

[0109] 1) Combined feature 1: ICM and TE contact ratio * ICM and TE distance, that is, the distance between ICM and TE is used as the weight and combined with the contact ratio. Considering the influence of both factors at the same time, the live birth blastocyst has a larger combined feature 1 (p = 0.02);

[0110] 2) Combined feature 2: ICM and TE contact ratio * 0.1 + ICM and TE distance * 0.9. It is believed that the relationship between the ICM and TE in three-dimensional space is composed of the contact ratio and the spatial distance. Among them, the spatial distance ratio is larger, and the blastocysts of pregnancy (p = 0.02) and live birth (p = 0.05) have a larger index;

[0111] 3) Combined feature 3: contact area between ICM and TE * distance between ICM and TE * percentage of trophoblast cells in the first three-dimensional quadrant, that is, the distance between ICM and TE and the percentage of trophoblast cells in the first three-dimensional quadrant were used as weights, combined with the contact area, and the influence of the three factors was considered at the same time. The live blastocysts had a larger combined feature 3 (p = 0.03);

[0112] 4) Combined feature 4: grouped according to whether the number of TE cells in the ICM quadrant is greater than 16. If greater than 16, the positional relationship between ICM and TE in three-dimensional space = contact area between ICM and TE (normalized to 0 to 1) + distance between ICM and TE; if less than 16, the positional relationship between ICM and TE in three-dimensional space = contact area between ICM and TE (normalized to 0 to 1) * 0.4 + distance between ICM and TE * 0.6. The value of combined feature 4 was greater in pregnant (p < 0.001) and live birth (p < 0.001) embryos;

[0113] 5) Combined feature 5: According to whether the number of TE cells in the quadrant where the ICM is located is greater than 16 groups and the contact area between ICM and TE is greater than 1196.4, the blastocysts are divided into three categories: good, medium, and poor; the blastocysts defined as good have the highest true clinical pregnancy rate and true clinical live birth rate, as shown in Table 1 below:

[0114]

[0115] Table 1

[0116] 6) Combined feature 6: According to whether the number of TE cells in the ICM quadrant is greater than 16 groups and the distance between ICM and TE is greater than 0.561, the blastocysts are divided into three categories: good, medium, and poor; the blastocysts defined as good have the highest true clinical pregnancy rate and true clinical live birth rate, as shown in Table 2 below:

[0117]

[0118] Table 2

[0119] The characteristics of trophoblast cells include: the total number of trophoblast cells, the density of trophoblast cells; the characteristics of inner cell mass include: the volume of inner cell mass, the surface area of ​​inner cell mass.

[0120] Among them, the characteristics of trophoblast cells and the characteristics of inner cell mass are the characteristics that are extracted in the related technology, and therefore, can be regarded as basic characteristics. Here, the density of trophoblast cells = the total number of trophoblast cells / the surface area of ​​blastocyst*1000.

[0121] Furthermore, before performing feature analysis, the data can be cleaned and then subjected to correlation analysis. If there is a linear relationship between the data, feature analysis will not be performed on the data; if there is a nonlinear relationship between the data, feature analysis will be performed on the data. Specifically, in the data cleaning stage, abnormal data and data that are not the above-mentioned feature data can be deleted; in the correlation analysis stage, a scatter plot between features can be drawn to determine whether there is a linear relationship between the feature data. Among them, correlation analysis is mainly used to analyze the difference between new features and original features. For example, the number of trophoblast cells in the first three-dimensional quadrant where the inner cell mass is located and the proportion of the number of trophoblast cells in the first three-dimensional quadrant can be regarded as new features, and the number of trophoblast cells and the density of trophoblast cells can be regarded as original features.

[0122] Step 450: construct sample data according to the association relationship.

[0123] Among them, the characteristic data in the association relationship can be used as the characteristic sample, the preset clinical quality assessment parameters in the association relationship can be used as the labeled sample of the above-mentioned characteristic sample, and the characteristic sample and the labeled sample can be used as sample data.

[0124] When the preset clinical quality assessment parameter is one, the association relationship is one group, and each feature sample has one labeled sample; when the preset clinical quality assessment parameters are two, the association relationship is two groups, and each feature sample has two labeled samples, and so on.

[0125] Step 460: Use the sample data to perform model training on the preset basic embryo quality assessment model to obtain an embryo quality assessment model.

[0126] Among them, all data can be divided into training data and test data in chronological order. For example, the data from 2018 to 2020 are used as training data, and the data from 2021 are used as test data. The training data is first used to train the embryo quality assessment model to be trained, and then the test data is used to test the embryo quality assessment model to determine its evaluation effect.

[0127] In actual operation, the performance of the embryo quality assessment model has been improved due to the addition of association relationships (at least two-dimensional spatial mutual features and the association relationship between the preset clinical quality assessment parameters); the specific model performance is as follows:

[0128] 1) Real clinical pregnancy rate: The auc of the Gardner model in the training set and test set was 0.622 and 0.615; adding "ICM and TE contact area * ICM and TE distance" to the Gardner parameters improved the model performance to 0.687 and 0.638. The parameters "ICM and TE contact area * ICM and TE distance * the proportion of trophoblast cells in the first three-dimensional quadrant" and "combined feature 6" all showed different degrees of performance improvement, and the results are summarized in Table 3;

[0129] 2) Actual clinical live birth rate: The auc of the Gardner model in the training set and test set was 0.603 and 0.624; adding "the contact ratio of ICM and TE*0.1+the distance between ICM and TE*0.9" to the Gardner parameters improved the model performance to 0.641 and 0.626. The following features have the same improvement effect: the contact area of ​​ICM and TE*the distance between ICM and TE*the proportion of the number of trophoblast cells in the first three-dimensional quadrant, and the combined features 4 and 5. The results are summarized in Table 4;

[0130] 3) Actual clinical live birth rate: The auc of the Gardner model in the training set and the test set were 0.565 and 0.604 respectively. Based on the Gardner parameters, the contact proportion of ICM and TE*0.1+the distance between ICM and TE*0.9, the contact area of ​​ICM and TE*the distance between ICM and TE*the proportion of the number of trophoblast cells in the first three-dimensional quadrant, and the combined features 4, 5 and 6 were added, and the model performance was improved.

[0131] The results are summarized in Table 5.

[0132]

[0133] Table 3

[0134]

[0135] Table 4

[0136]

[0137] Table 5

[0138] The embryo quality assessment method provided in the embodiment of the present application, first, feature extraction is performed on clinical embryo data to obtain a clinical initial interaction feature data set; secondly, feature mining is performed on each clinical initial interaction feature in the clinical initial interaction feature data set to obtain multiple clinical interaction features; again, multiple clinical interaction features are screened to obtain clinical target interaction features; then, according to the clinical initial interaction feature data set, feature analysis is performed on the clinical target interaction features to obtain the correlation between the clinical target interaction features and the preset clinical quality assessment parameters; then, according to the correlation, sample data is constructed; finally, the preset basic embryo quality assessment model is trained using the sample data to obtain an embryo quality assessment model. In this way, by adding the correlation between the clinical target interaction features and the preset clinical quality assessment parameters in the sample data, the performance of the embryo quality assessment model is improved, so that the quality assessment of the embryo to be assessed can be more accurately performed.

[0139] After introducing the embryo quality assessment method according to the exemplary embodiment of the present disclosure, next, reference is made to Figure 5 An embryo quality assessment device 500 according to an exemplary embodiment of the present disclosure will be described.

[0140] refer to Figure 5The embryo quality assessment device 500 includes: an acquisition module 510, configured to obtain a three-dimensional reconstruction model of the embryo to be assessed at a preset cultivation stage; a feature extraction module 520, configured to perform feature extraction on the three-dimensional reconstruction model to obtain initial interaction features of trophoblast cells and inner cell mass in the embryo to be assessed; a feature mining module 530, configured to perform feature mining on the initial interaction features to obtain target interaction features of trophoblast cells and inner cell mass; a prediction module 540, configured to perform quality assessment on the embryo to be assessed using a preset embryo quality assessment model based on the target interaction features to obtain a predicted value of a quality assessment parameter of the embryo to be assessed.

[0141] In one embodiment, the feature mining module 530 is configured to: adjust the self-distribution of the initial interaction feature to obtain the first interaction feature; the target interaction feature includes the first interaction feature.

[0142] In one embodiment, the feature mining module 530 is configured to: use a first preset algorithm to perform computational processing on the initial interaction feature to obtain a first interaction feature.

[0143] In one embodiment, the feature mining module 530 is configured to: determine a nonlinear relationship expression of an initial interaction feature; and use the nonlinear relationship expression to perform a polynomial expansion on the initial interaction feature to obtain a first interaction feature.

[0144] In one embodiment, the feature mining module 530 is configured to: perform feature combination on different initial interaction features according to the interaction between at least two initial interaction features to obtain a second interaction feature; the target interaction feature includes the second interaction feature.

[0145] In one embodiment, the feature mining module 530 is configured to: perform a logistic regression operation using a second preset algorithm according to the interaction between the first initial interaction feature and the second initial interaction feature to obtain a second interaction feature.

[0146] In one embodiment, the feature mining module 530 is configured to: discretize the third initial interaction feature to obtain at least two types of classification data; and arrange and combine the classification data, other classification data, and the initial interaction feature to obtain the second interaction feature.

[0147] In one embodiment, the embryo quality assessment device 500 also includes a model training module, which is configured to perform feature extraction on clinical embryo data to obtain a clinical initial interaction feature data set; perform feature mining on each clinical initial interaction feature in the clinical initial interaction feature data set to obtain multiple clinical interaction features; screen multiple clinical interaction features to obtain clinical target interaction features; perform feature analysis on the clinical target interaction features based on the clinical initial interaction feature data set to obtain a correlation between the clinical target interaction features and preset clinical quality assessment parameters; construct sample data based on the correlation; and use the sample data to perform model training on a preset basic embryo quality assessment model to obtain an embryo quality assessment model.

[0148] The above-mentioned device is used to execute the method provided by the aforementioned embodiment, and its implementation principle and technical effect are similar, which will not be repeated here.

[0149] The above modules may be one or more integrated circuits configured to implement the above methods, such as one or more application specific integrated circuits (ASICs), or one or more microprocessors, or one or more field programmable gate arrays (FPGAs). For another example, when a module is implemented in the form of a processing element scheduling program code, the processing element may be a general-purpose processor, such as a central processing unit (CPU) or other processor that can call program code. For another example, these modules may be integrated together and implemented in the form of a system-on-a-chip (SOC).

[0150] Figure 6 A schematic diagram of a computer device provided in an embodiment of the present application. The device can be integrated into a terminal device or a chip of a terminal device. The terminal can be a computing device with data processing capabilities.

[0151] The device includes: a processor 601 , a storage medium 602 and a bus 603 .

[0152] The storage medium 602 stores program instructions executable by the processor 601. When the computer device 600 is running, the processor 601 communicates with the storage medium 602 via the bus 603. The processor 601 executes the program instructions to perform the above method embodiment. The specific implementation method and technical effect are similar and will not be repeated here.

[0153] Optionally, the present invention also provides a program product, such as a computer-readable storage medium, comprising a program, which is used to execute the above method embodiment when executed by a processor.

[0154] In the several embodiments provided by the present invention, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0155] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0156] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of hardware plus software functional units.

[0157] The above-mentioned integrated unit implemented in the form of a software functional unit can be stored in a computer-readable storage medium. The above-mentioned software functional unit is stored in a storage medium, including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) or a processor (English: processor) to perform some steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (English: Read-Only Memory, abbreviated: ROM), random access memory (English: Random Access Memory, abbreviated: RAM), disk or optical disk and other media that can store program codes.

[0158] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art who is familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.

Claims

1. A method for evaluating embryo quality, characterized in that: include: Obtain a three-dimensional reconstruction model of the embryo to be evaluated at a preset cultivation stage; Extracting features from the three-dimensional reconstructed model to obtain initial interaction features between trophoblast cells and inner cell mass in the embryo to be evaluated; Performing feature mining on the initial interaction features to obtain target interaction features between the trophoblast cells and the inner cell mass; According to the target interaction characteristics, a preset embryo quality assessment model is used to perform quality assessment on the embryo to be assessed, and a predicted value of the quality assessment parameter of the embryo to be assessed is obtained.

2. The method according to claim 1, characterized in that: The feature mining of the initial interaction features to obtain the target interaction features of the trophoblast cells and the inner cell mass includes: The self-distribution of the initial interaction feature is adjusted to obtain a first interaction feature; the target interaction feature includes the first interaction feature.

3. The method according to claim 2, characterized in that: The adjusting the self-distribution of the initial interaction feature to obtain the first interaction feature includes: A first preset algorithm is used to perform computational processing on the initial interaction feature to obtain the first interaction feature.

4. The method according to claim 2, characterized in that: The adjusting the self-distribution of the initial interaction feature to obtain the first interaction feature includes: Determining a nonlinear relationship expression of the initial interaction characteristics; The nonlinear relationship expression is used to perform polynomial expansion on the initial interaction characteristic to obtain the first interaction characteristic.

5. The method according to claim 1, characterized in that: Feature mining is performed on the initial interaction features to obtain target interaction features of the trophoblast cells and the inner cell mass, including: According to the interaction between at least two of the initial interaction features, feature combination is performed on different initial interaction features to obtain a second interaction feature; the target interaction feature includes the second interaction feature.

6. The method according to claim 5, characterized in that: The step of combining different initial interaction features according to the interaction between at least two initial interaction features to obtain a second interaction feature comprises: According to the interaction between the first initial interaction feature and the second initial interaction feature, a logistic regression operation is performed using a second preset algorithm to obtain the second interaction feature.

7. The method according to claim 5, characterized in that: The step of combining different initial interaction features according to the interaction between at least two initial interaction features to obtain a second interaction feature comprises: discretizing the third initial interaction feature to obtain at least two categories of classification data; The classification data, other classification data and the initial interaction feature are arranged and combined to obtain the second interaction feature.

8. The method according to claim 1, characterized in that: Before performing quality assessment on the embryo to be assessed using a preset embryo quality assessment model according to the target interaction characteristics to obtain a predicted value of a quality assessment parameter of the embryo to be assessed, the method further includes: Extract features from clinical embryo data to obtain a clinical initial interaction feature dataset; performing feature mining on each clinical initial interaction feature in the clinical initial interaction feature data set to obtain a plurality of clinical interaction features; screening the multiple clinical interaction features to obtain clinical target interaction features; According to the clinical initial interaction feature data set, feature analysis is performed on the clinical target interaction feature to obtain a correlation between the clinical target interaction feature and the preset clinical quality assessment parameter; Constructing sample data according to the association relationship; The sample data is used to perform model training on a preset basic embryo quality assessment model to obtain the embryo quality assessment model.

9. An embryo quality assessment device, characterized in that: include: An acquisition module is configured to acquire a three-dimensional reconstruction model of the embryo to be evaluated at a preset cultivation stage; A feature extraction module is configured to extract features from the three-dimensional reconstruction model to obtain initial interaction features between trophoblast cells and inner cell mass in the embryo to be evaluated; A feature mining module is configured to perform feature mining on the initial interaction features to obtain target interaction features between the trophoblast cells and the inner cell mass; The prediction module is configured to perform quality assessment on the embryo to be assessed according to the target interaction characteristics using a preset embryo quality assessment model to obtain a predicted value of a quality assessment parameter of the embryo to be assessed.

10. A computer device, characterized in that: include: A processor, a storage medium and a bus, wherein the storage medium stores program instructions executable by the processor, and when the computer device is running, the processor communicates with the storage medium via the bus, and the processor executes the program instructions to perform the steps of the embryo quality assessment method as described in any one of claims 1 to 8.

Images