Blastocyst quality evaluation method and system

Abstract

The present invention relates to a blastocyst quality evaluation method and system. The method comprises: inputting a blastocyst multi-focus image into a pre-trained deep learning model to obtain a blastocyst development stage and quantitative scores for ICM and TE; obtaining a blastocyst quality quantitative score by means of the quantitative scores for ICM and TE; and selecting an optimal blastocyst by means of the blastocyst development stage and the blastocyst quality quantitative score. According to the present invention, the quality of blastocysts of the same grade is distinguished, thereby further quantifying the overall quality of an embryo.

Description

A method and system for evaluating blastocyst quality Technical Field

[0001] The present invention relates to the field of medical technology, and in particular to a blastocyst quality assessment method and system. Background Art

[0002] With changes in the environment, society, economy, lifestyle, and delayed childbearing, among other factors, approximately 8-12% of couples of childbearing age worldwide currently struggle with infertility. As an effective treatment for infertility, in vitro fertilization (IVF) has helped numerous infertile couples successfully conceive, with over 8 million IVF babies born worldwide. IVF involves the removal of multiple eggs from a woman's ovaries and fertilization with sperm in the laboratory. High-quality embryos are then selected from the embryos formed during this process and placed in the uterus for potential implantation. To avoid the health risks and complications associated with multiple pregnancies and births, only a few embryos, ideally just one, are typically transferred. Therefore, selecting embryos suitable for transfer is a critical step in IVF treatment.

[0003] The current standards for embryo evaluation and selection mainly come from the Gardner human embryo grading system, which stipulates that the quality of the blastocyst should be evaluated from three aspects: the blastocyst development stage, the inner cell mass (ICM) and the trophectoderm (TE). The blastocyst development stage is divided into stages 1-6 according to the volume of the blastocyst cavity and the thickness of the zona pellucida. The ICM is divided into three grades AC according to the number of cells and compactness. The TE is also divided into three grades AC according to the number of cells, density, size, etc. Due to the subjectivity of this set of standards and the subjectivity of embryologists who perform embryo evaluation clinically, there are large differences in evaluations between different embryologists. It is now generally recognized in the industry that embryo quality assessment requires objective and quantitative improvements.

[0004] One approach is to use deep learning technology to automatically analyze embryo images to determine the blastocyst developmental stage, ICM, and TE ratings. Because embryos occupy a three-dimensional space, a single image cannot fully capture their entire appearance. Therefore, researchers are using deep learning to process multi-focus images of embryos to assess embryo quality. Multi-focus images focus on different z-axis planes of the embryo, providing a more comprehensive picture of the embryo's entirety.

[0005] However, the results of the embryo evaluation conducted in the above work are consistent with those of embryologists, both of which give the embryo's developmental stage, ICM, and TE rating, such as 4AA, 4BB, etc. The problem of how to distinguish the quality of blastocysts with the same rating and further quantify the overall quality of the embryo remains unresolved.

[0006] Summary of the Invention

[0007] Therefore, the technical problem to be solved by the present invention is to overcome the existing technology.

[0008] To solve the above technical problems, the present invention provides a blastocyst quality assessment method, comprising:

[0009] The multi-focus images of the blastocyst are input into a pre-trained deep learning model to obtain the blastocyst development stage and the quantitative scores of the ICM and the TE;

[0010] Obtaining a quantitative score of blastocyst quality through the quantitative scoring of the ICM and the TE;

[0011] The optimal blastocyst is selected according to the blastocyst development stage and the blastocyst quality quantitative score.

[0012] In one embodiment of the present invention, the pre-trained deep learning model includes:

[0013] The deep learning model is trained using pre-processed multi-focus images of the blastocyst to obtain the probability values ​​of each rating of the ICM and the TE.

[0014] In one embodiment of the present invention, the ratings of the ICM and the TE include:

[0015] The ICM ratings include: A, B, and C:

[0016] The TE ratings include: A, B, and C.

[0017] In one embodiment of the present invention, the pre-processed multi-focus image of the blastocyst comprises:

[0018] Get multiple photos with different focal planes and set them to the same resolution;

[0019] Segmenting the embryos in each of the photos at different focal planes, filling the images to a resolution of a first set size, and then scaling the images to a resolution of a second set size;

[0020] The multiple photos of different focal planes include at least one of the ICM and one of the TE.

[0021] In one embodiment of the present invention, obtaining the rating probability values ​​of the ICM and the TE includes:

[0022] The backbone network in the deep learning model extracts the feature vector of each blastocyst photo at different focal planes;

[0023] Inputting the feature vector into a fully connected network in the deep learning model to obtain the blastocyst development stage and the ratings of the ICM and the TE;

[0024] The blastocyst development stage and the ratings of the ICM and the TE are input into a Softmax layer for normalization to obtain probability values ​​of the ratings of the ICM and the TE.

[0025] In one embodiment of the present invention, obtaining the quantitative scores of the ICM and the TE includes:

[0026] The A-level, B-level, and C-level of the ICM and the TE are weighted and summed according to the set weight values ​​to obtain the quantitative scores of the ICM and the TE.

[0027] In one embodiment of the present invention, obtaining a quantitative score of blastocyst quality includes:

[0028] Obtaining weights of the quantitative scores of the ICM and the TE through a pre-trained logistic regression model;

[0029] The quantitative score of the blastocyst quality is obtained by weighted summation of the quantitative scores of the ICM and the TE.

[0030] In one embodiment of the present invention, the pre-trained logistic regression model includes:

[0031] The weights of the quantitative scores of the ICM and the TE are obtained according to the quantitative scores of the ICM and the TE and the clinical outcomes of the embryos.

[0032] In one embodiment of the present invention, selecting the optimal blastocyst according to the blastocyst development stage and the blastocyst quality quantitative score includes:

[0033] The blastocyst development stage is selected to be greater than or equal to stage three;

[0034] Based on the blastocyst development stage, blastocysts greater than or equal to stage three are selected, and the blastocysts with the highest quantitative quality score are selected.

[0035] The present invention also provides a blastocyst quality assessment system, comprising:

[0036] Photographic device: used to take multi-focus pictures of blastocysts;

[0037] Scoring device: used to determine the blastocyst development stage and score the ICM and TE;

[0038] Selection device: Select the highest blastocyst quality score.

[0039] The above technical solution of the present invention has the following advantages over the prior art:

[0040] The present invention provides a blastocyst quality assessment method and system, comprising: inputting a pre-trained deep learning model into a multi-focus image of the blastocyst to obtain the blastocyst development stage and obtain quantitative scores of ICM and TE; obtaining a quantitative score of the blastocyst quality through the quantitative scores of ICM and TE; selecting the optimal blastocyst based on the blastocyst development stage and the quantitative score of the blastocyst quality; the present invention achieves the ability to distinguish the quality of blastocysts with the same rating, and further quantifies the overall quality of the embryo. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] In order to make the content of the present invention more clearly understood, the present invention is further described in detail below based on specific embodiments of the present invention in conjunction with the accompanying drawings, wherein

[0042] FIG1 is a flow chart of a method for evaluating blastocyst quality according to the present invention;

[0043] FIG2 is a schematic diagram of a multi-focus image of a blastocyst of the present invention;

[0044] FIG3 is a schematic diagram of focal plane photographs of blastocysts of ICE and TE of the present invention;

[0045] FIG4 is a structural diagram of a deep learning model of the present invention;

[0046] FIG5 is a schematic diagram of the blastocyst scoring method of the present invention for distinguishing the quality of two blastocysts. DETAILED DESCRIPTION

[0047] The present invention will be further described below with reference to the accompanying drawings and specific embodiments so that those skilled in the art can better understand the present invention and implement it. However, the embodiments are not intended to limit the present invention.

[0048] Example 1

[0049] 1 , a blastocyst quality assessment method of the present invention includes:

[0050] Multi-focus images of blastocysts are fed into a pre-trained deep learning model to obtain the blastocyst developmental stage and quantitative scores of ICM and TE.

[0051] The quantitative score of blastocyst quality was obtained through the quantitative scoring of ICM and TE;

[0052] The optimal blastocyst was selected based on the blastocyst development stage and blastocyst quality quantitative scoring.

[0053] Pre-trained deep learning models, including:

[0054] The deep learning model was trained using pre-processed multi-focus images of blastocysts to obtain the probability values ​​of each rating for ICM and TE.

[0055] ICM and TE ratings, including:

[0056] ICM ratings include: A, B, and C;

[0057] TE ratings include: A, B, and C.

[0058] Multi-focus images of pre-processed blastocysts, including:

[0059] Get multiple photos with different focal planes and set them to the same resolution;

[0060] Segment the embryos in each image at different focal planes, fill them to the resolution of the first size setting, and then scale them to the resolution of the second size setting;

[0061] The multiple photos at different focal planes include at least one ICM and one TE.

[0062] Obtain the probability values ​​of each rating of ICM and TE, including:

[0063] The backbone network in the deep learning model extracts feature vectors for each blastocyst photo at different focal planes;

[0064] The feature vectors were fed into a fully connected network in the deep learning model to obtain the blastocyst developmental stage and the ratings of ICM and TE.

[0065] The blastocyst development stage and the ratings of ICM and TE are input into the Softmax layer for normalization to obtain the probability values ​​of each rating of ICM and TE.

[0066] Obtain quantitative scores for ICM and TE, including:

[0067] The quantitative scores of ICM and TE are obtained by weighting and summing the A, B and C levels of ICM and TE according to the set weight values.

[0068] Obtain a quantitative score for blastocyst quality, including:

[0069] The weights of the quantitative scores of ICM and TE were obtained through the pre-trained logistic regression model;

[0070] The quantitative score of blastocyst quality was obtained by weighted summation of the quantitative scores of ICM and TE.

[0071] Pre-trained logistic regression models, including:

[0072] The weights of the quantitative scores of ICM and TE were obtained based on the quantitative scores of ICM and TE and the clinical outcomes of the embryos.

[0073] The optimal blastocysts are selected based on the blastocyst development stage and blastocyst quality quantitative scores, including:

[0074] The blastocyst development stage was selected to be greater than or equal to stage three;

[0075] Based on the blastocyst development stage, blastocysts greater than or equal to stage three were selected, and the blastocysts with the highest quantitative quality score were selected.

[0076] Example 2

[0077] The method involves obtaining multi-focus images of blastocysts, using a deep learning model to score the blastocyst developmental stage, inner cell mass, and trophoblast cells, and determining blastocyst quality based on the scores. The multi-focus images consist of at least two images: one focused on the inner cell mass and the other on the trophoblast cells. The inner cell mass and trophoblast cell scores are derived by weighting the probabilities of inner cell mass and trophoblast cells being graded A, B, and C, respectively, as determined by the deep learning model.

[0078] The system includes a device for capturing multi-focus images of blastocysts, a device for scoring the blastocyst developmental stage, inner cell mass, and trophoblast cells using a deep learning model, and a device for determining blastocyst quality based on the scores. The multi-focus images consist of at least two images: one focused on the inner cell mass and the other on the trophoblast cells. The inner cell mass and trophoblast cell scores are derived by weighting the probabilities of inner cell mass and trophoblast cell grades A, B, and C, respectively, as determined by the deep learning model. The blastocyst developmental stage and the grades A, B, and C for the inner cell mass and trophoblast cells are based on the Gardner Human Blastocyst Grading System.

[0079] The method for determining the quality of blastocysts is to determine whether the blastocyst is a fully expanded blastocyst by whether the predicted blastocyst development stage is ≥ stage 3. The quality quantification method for fully expanded blastocysts is the weighted sum of the inner cell mass score and the trophoblast cell score. The weight is obtained by establishing a logistic regression model of the inner cell mass, trophoblast cell score and the clinical outcome of the blastocyst.

[0080] The probabilities of grade A, grade B, and grade C of the inner cell mass and trophoblast cells are obtained by normalizing the output of the last fully connected layer of the deep learning model with a normalized exponential function.

[0081] According to the system of claim 3, the input of the last fully connected layer of the deep learning model is the feature connection obtained by the convolutional neural network backbone network extracting the multi-focus image, and the output is the blastocyst development stage, inner cell mass rating and trophoblast cell rating.

[0082] The device for taking multi-focus images of the blastocyst is an inverted microscope or an embryo time-lapse observation system.

[0083] The system includes:

[0084] A device for taking multi-focus images of blastocysts;

[0085] A device that uses a deep learning model to score the blastocyst developmental stage, inner cell mass, and trophoblast cells;

[0086] A device for determining blastocyst quality based on blastocyst developmental stage, inner cell mass, and trophoblast cell scores, wherein whether the blastocyst is a fully expanded blastocyst is determined by whether the blastocyst developmental stage is ≥ stage 3. For fully expanded blastocysts, the quality quantification method is the weighted sum of the inner cell mass score and the trophoblast cell score;

[0087] and means for selecting a human blastocyst having a highest human blastocyst quality score.

[0088] Example 3

[0089] Quantitative scoring helps to select higher quality blastocysts for transplantation, thereby improving the success rate of IVF (in vitro fertilization) treatment.

[0090] In a first aspect, the present invention relates to a method comprising obtaining multi-focus images of blastocysts, scoring the blastocyst developmental stage, inner cell mass, and trophoblast cells, and determining blastocyst quality based on the scores. Specifically, the present invention can be applied to human blastocysts, and the resulting quantitative blastocyst quality scores can be used to identify and select the most suitable blastocysts for transfer into a female uterus. The selected blastocysts have a higher probability of resulting in a pregnancy and a live birth, thereby improving the success rate of IVF treatment.

[0091] Therefore, in another aspect, the method includes obtaining multi-focus images of blastocysts, using a deep learning model to score the blastocyst development stage, ICM (inner cell mass), and TE (trophoblast), and determining a quantitative score for blastocyst quality based on the score. The multi-focus image contains at least two images, one focused on the ICM and the other focused on the TE; the ICM and TE scores are obtained by weighting the probabilities of the inner cell mass and trophoblast cells being graded A, B, and C, respectively, as determined by the deep learning model, with weights of 3, 2, and 1, wherein the blastocyst development stage and ICM, TE grade A, B, and C standards refer to the Gardner human blastocyst grading system; the blastocyst quality quantitative score is obtained by summing the ICM score and TE score with a weight of 1:1.5.

[0092] Therefore, in another aspect, the method comprises acquiring multifocus images of blastocysts as defined above, thereby obtaining a quantitative score of blastocyst quality, and selecting the blastocyst with the highest quantitative score of blastocyst quality.

[0093] In another aspect, the present invention provides a system comprising a device for implementing the above method. The system can be any suitable system, such as a computer comprising a computer code portion configured to perform the above method. The system can further include a device for obtaining multi-focus images of the embryo at different focal planes, such as an inverted microscope, an embryo time-lapse observation system, or the like.

[0094] In yet another aspect, the invention relates to a data carrier comprising computer code portions constituting means for performing the method as described above.

[0095] One embodiment of the present invention provides a method for quantitatively evaluating blastocyst quality and selecting blastocysts based on multi-focus images of blastocysts, which specifically includes the following steps:

[0096] Referring to Figures 2 and 3, S1: Acquire a multi-focus image of the blastocyst and perform image preprocessing. In this embodiment, the multi-focus image of the blastocyst consists of 11 images at different focal planes, with a resolution of 1280x960. Image preprocessing involves segmenting the embryo in each focal plane, padding them to the same size, such as 500x500, and then scaling them to 300x300. The multi-focus image of the blastocyst should consist of at least two images, focusing on the ICM and TE, respectively.

[0097] As shown in Figure 4, S2: Use a deep learning model to process the multi-focus images of the blastocyst to obtain the probabilities of the blastocyst development stage (DS), ICM, and TE ratings. The backbone network in this deep learning model processes each multi-focus image and extracts feature vectors. In this embodiment, a dense convolutional network (Densenet121) is selected as the backbone network. These feature vectors are concatenated and used as the input of the last layer of fully connected networks. The last layer has three fully connected networks used to predict the blastocyst development stage, ICM, and TE ratings respectively. Their outputs are normalized by a normalized exponential function (Softmax) layer to obtain the probability values ​​belonging to each rating.

[0098] S3: Obtain ICM and TE quantitative scores based on the ICM and TE rating probabilities, and obtain a blastocyst quality quantitative score. The ICM and TE quantitative scores are obtained by weighting and summing the probabilities of the embryo ICM and TE belonging to grade A, grade B, and grade C calculated by the deep learning model according to a weight of 3:2:1. 3:2:1 corresponds to the basic scores of grade A, grade B, and grade C, respectively, and are manually preset. After obtaining the ICM and TE quantitative scores, the ICM and TE quantitative scores are weighted and summed according to a weight of 1:1.5 to obtain the blastocyst quality quantitative score. In this embodiment, the weight of 1:1.5 is obtained by establishing a logistic regression model between the ICM and TE quantitative scores and the blastocyst live birth outcome. After training the logistic regression model to fit the ICM, TE quantitative scores and the embryo clinical outcomes (expressed by 0, 1), the weight coefficients corresponding to the ICM and TE quantitative scores in the logistic regression model are obtained. Technicians can also make adjustments by establishing a logistic regression model between the ICM and TE quantitative scores and other clinical outcomes such as euploidy, implantation, pregnancy, etc.

[0099] As shown in Figure 5, two blastocysts, both rated 4BB, were distinguished by quantitative blastocyst quality scoring. The scoring results indicate that blastocyst 1 is superior to blastocyst 2. Although both blastocysts were rated 4BB, blastocyst 1 has better ICM and TE than blastocyst 2. The figure also shows that blastocyst 1 has a larger ICM area than blastocyst 2, and the number and continuity of TE are also better than blastocyst 2.

[0100] S4: Select the best blastocyst based on the blastocyst development stage and blastocyst quality quantitative score. The best blastocyst should meet the blastocyst development stage ≥ 3.

[0101] On this basis, the blastocyst with the highest quantitative score of blastocyst quality is the optimal blastocyst.

[0102] There are 6 stages of development. The blastocyst that can be transplanted should generally reach stage 4. In some cases, if the patient does not have a stage 4 blastocyst, a stage 3 blastocyst can also be transplanted.

[0103] Example 4

[0104] The present invention also provides a blastocyst quality assessment system, comprising:

[0105] Photographic device: used to take multi-focus pictures of blastocysts;

[0106] Scoring device: used to determine the blastocyst development stage and score the ICM and TE;

[0107] Selection device: Select the highest blastocyst quality score.

[0108] Those skilled in the art will appreciate that the embodiments of the present application can be provided as methods, systems, or computer program products. Therefore, the present application can adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application can adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) that contain computer-usable program code.

[0109] The present application is described with reference to the flow chart and / or block diagram of the method, device (system), and computer program product according to the embodiment of the present application. It should be understood that each flow process and / or box in the flow chart and / or block diagram and the combination of the flow process and / or box in the flow chart and / or block diagram can be realized by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processing machine or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device produce a device for realizing the function specified in one flow chart flow or multiple flows and / or one box or multiple boxes of the block diagram.

[0110] These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce a product including an instruction device that implements the functions specified in one or more processes in the flowchart and / or one or more boxes in the block diagram.

[0111] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, so that the instructions executed on the computer or other programmable device provide steps for implementing the functions specified in one or more processes in the flowchart and / or one or more boxes in the block diagram.

[0112] Obviously, the above embodiments are merely examples for clarity of explanation and are not intended to limit the implementation methods. Those skilled in the art will appreciate that other variations or modifications can be made based on the above description. It is not necessary and impossible to enumerate all implementation methods here. Obvious variations or modifications arising therefrom remain within the scope of protection of the present invention.

Claims

1. A method for evaluating blastocyst quality, characterized in that: include: The multi-focus images of the blastocyst are input into a pre-trained deep learning model to obtain the developmental stage of the blastocyst and obtain the quantitative scores of the ICM and the TE; Obtaining a quantitative score of blastocyst quality through the quantitative scores of the ICM and the TE; The optimal blastocyst is selected according to the blastocyst development stage and the blastocyst quality quantitative score.

2. The method according to claim 1, characterized in that The pre-trained deep learning model includes: The deep learning model is trained using pre-processed multi-focus images of the blastocyst to obtain the rating probability values ​​of the ICM and the TE.

3. The method according to claim 2, characterized in that The ratings of the ICM and the TE include: The ICM ratings include: A, B, and C; The TE ratings include: Grade A, Grade B, and Grade C.

4. The method according to claim 2, characterized in that: The pre-processed multi-focus image of the blastocyst includes: Get multiple photos with different focal planes and set them to the same resolution; Segmenting the embryos in each of the photos of the different focal planes, filling them to a resolution of a first set size, and scaling them to a resolution of a second set size; The multiple photos of different focal planes include at least one of the ICM and one of the TE.

5. The method according to claim 2, characterized in that: The obtaining of each rating probability value of the ICM and the TE includes: The backbone network in the deep learning model extracts the feature vector of each blastocyst photo in different focal planes; Inputting the feature vector into a fully connected network in the deep learning model to obtain the blastocyst development stage and the ratings of the ICM and the TE; The blastocyst development stage and the ratings of the ICM and the TE are input into the Softmax layer for normalization to obtain the probability values ​​of each rating of the ICM and the TE.

6. The method according to claim 3, characterized in that: The obtaining of the quantitative scores of the ICM and the TE comprises: According to the A level, B level, and C level of the ICM and the TE, weighted sums are taken according to set weight values ​​to obtain quantitative scores of the ICM and the TE.

7. The method according to claim 1, characterized in that: The method of obtaining a quantitative score of blastocyst quality comprises: Obtaining weights of the quantitative scores of the ICM and the TE through a pre-trained logistic regression model; The quantitative score of the blastocyst quality is obtained by weighted summation of the quantitative scores of the ICM and the TE.

8. The method according to claim 1, characterized in that: The pre-trained logistic regression model includes: The weights of the quantitative scores of the ICM and the TE are obtained through the quantitative scores of the ICM and the TE and the clinical outcomes of the embryos.

9. The method according to claim 1, characterized in that: The selecting the best blastocyst according to the blastocyst development stage and the blastocyst quality quantitative score comprises: The blastocyst development stage is selected to be greater than or equal to the third stage; Based on the blastocyst development stage, blastocysts greater than or equal to stage three are selected, and the blastocyst with the highest quantitative quality score is selected.

10. A blastocyst quality assessment system, used to implement the blastocyst quality assessment method according to claims 1-9, characterized in that: include: Photographic device: used to take multi-focus pictures of blastocysts; Scoring device: used to determine the blastocyst development stage and score ICM and TE; Selection device: Select the highest blastocyst quality score.

Images