Model construction method and system for predicting sustainable pregnancy of embryo, electronic equipment and storage medium

Abstract

The invention provides a model construction method and system for predicting the sustainable pregnancy of an embryo, electronic equipment and a storage medium, the prediction method integrates morphological information, development information and genetic information of the embryo, a model for predicting the sustainable pregnancy of the embryo is constructed based on a random forest algorithm and an XGBoost algorithm weighted average fusion method, and the model is used for predicting the sustainable pregnancy of the embryo. According to the method, embryos with higher sustainable pregnancy can be screened, and the embryo which is most likely to be successful is selected for transplantation, so that the success probability of pregnancy is increased.

Description

Technical Field

[0001] The present invention relates to the technical field of embryo sustainable pregnancy prediction, and in particular to a model construction method, system, electronic device and storage medium for predicting embryo sustainable pregnancy. Background Art

[0002] Assisted reproductive technology is an important technology in modern medicine, often used to help couples who are unable to conceive naturally due to physiological reasons. These technologies include in vitro fertilization (IVF) and in vitro embryo culture. During the IVF process, a woman's eggs and a man's sperm are combined in the laboratory to form fertilized eggs. These fertilized eggs are then cultured in a specific nutrient solution to promote their growth and development. Doctors usually culture multiple fertilized eggs. During the culture process, doctors can monitor the development of the embryos and select the embryos most likely to succeed for transplantation.

[0003] Embryo screening methods can ensure that the most promising embryos are transplanted, increase the chance of successful pregnancy, and thus reduce the economic burden of patients and the risk of complications caused by multiple egg retrieval and transplantation. Commonly used embryo screening methods are as follows. Morphological evaluation is a commonly used embryo screening method. The quality is evaluated by observing the morphological characteristics of embryos at different developmental stages under a microscope. One commonly used method is the Gardner evaluation method proposed by Gardner et al. (Gardner, DK, & Schoolcraft, WB; Culture and transfer of human blastocysts; Current Opinion in Obstetrics and Gynecology; 1999 11(3), 307). This method mainly targets blastocysts and conducts a comprehensive evaluation of the blastocyst cavity, inner cell mass (ICM), and trophoblast (TE). However, the morphological rating obtained by this method has been shown to have little correlation with the chromosomal euploidy of the embryo. However, chromosomal abnormalities of the embryo are an important factor leading to transplantation failure. Therefore, relying solely on morphological evaluation may result in the transplantation of embryos with chromosomal abnormalities, which may lead to transplantation failure. Basile et al. proposed a timelapse evaluation method (Basile, N., Nogales, MDC, Bronet, F., Florensa, M., Riqueiros, M., Rodrigo, L., García-Velasco, J., & Meseguer, M.; Increasing the probability of selecting chromosomally normal embryos by time-lapse morphokinetics analysis; Fertility and Sterility; 2014, 101(3), 699-704.e1.). This method uses a microscope camera to continuously record the embryonic development process and then uses a machine learning algorithm to evaluate the dynamic process of embryonic development. This method requires expensive timelapse equipment, and many hospitals do not have the financial strength to configure such equipment. Therefore, this method is limited in use in hospitals. Genetic methods, such as preimplantation genetic testing (PGT), will test the genome of the embryo to check whether there are abnormalities in the copy number of chromosomes, and screen euploid embryos for transplantation. Conventional PGT is for embryos in the blastocyst stage, and several cells need to be removed from the trophoblast of the blastocyst for testing. This process is traumatic to the embryo. Many doctors have doubts about this, suspecting that this invasive sampling may affect the subsequent development of the embryo.The non-invasive PGT proposed by L. Chen et al. is to obtain the culture medium during the embryo culture process (Chen, L., Li, W., Liu, Y., Peng, Z., Cai, L., Zhang, N., Xu, J., Wang, L., Teng, X., Yao, Y., Zou, Y., Ma, M., Liu, J., Lu, S., Sun, H., & Yao, B.; Non-invasive embryo selection strategy for clinical IVF to avoid wastage of potentially competent embryos; Reproductive BioMedicine Online; 2022, 45(1), 26–34.)(L. Chen et al., 2022), and then detect the DNA in the culture medium. Among them, the non-invasive method is becoming more and more popular among doctors because it does not harm the embryo. The commonly used methods in clinical practice are morphological scoring and PGT / non-invasive PGT. Morphology can evaluate the developmental status of the embryo, and PGT / non-invasive PGT can evaluate whether the embryo's chromosomes are abnormal. However, both methods only evaluate one aspect of the embryo. There is currently no method to comprehensively evaluate the embryo's developmental status and chromosomal status. Summary of the invention

[0004] In order to solve the above technical problems, the purpose of the present invention is to provide a model construction method, system, electronic device and storage medium for predicting the sustainable pregnancy of embryos, which can integrate morphological information, embryo development information, and embryo genetic information, and construct a model for predicting the sustainable pregnancy of embryos through machine learning methods to better screen embryos and improve the success rate of conception.

[0005] The present invention first provides a method for constructing a model for predicting the sustainable pregnancy of an embryo, which comprises the following steps:

[0006] S101: Divide the embryo samples into a training set and an independent validation set, and evaluate the embryo's sustainable pregnancy potential;

[0007] S102: Select the morphological characteristics of the embryo, the developmental time characteristics of the embryo, and the genomic information of the embryo;

[0008] S103: Construct a prediction model for embryo sustainable pregnancy based on the weighted average fusion of random forest algorithm and XGBoost algorithm.

[0009] In the above-mentioned model construction method for predicting the sustainable pregnancy of an embryo, preferably, the morphological characteristics of the embryo in S102 are scores obtained by evaluating according to the Gardner scoring method, and the morphological characteristics of the embryo include:

[0010] shape_1: Gardner scoring method for the blastocyst cavity;

[0011] shape_2: Gardner scoring method for inner cell mass;

[0012] shape_3: Gardner scoring method for trophoblast;

[0013] shape_quality: Convert shape_1, shape_2, and shape_3 scores into three levels: Good Fair Poor.

[0014] More preferably, the shape_1 scoring includes: a score of 1 indicating an early blastocyst whose blastocyst cavity is less than half of the total volume of the embryo, a score of 2 indicating a blastocyst cavity greater than half of the total volume of the embryo, a score of 3 indicating a complete blastocyst cavity, a score of 4 indicating an extended blastocyst whose blastocyst cavity has expanded beyond the initial embryo volume, a score of 5 indicating a hatching blastocyst, and a score of 6 indicating a fully hatched blastocyst;

[0015] The shape_2 scores include: a score of A indicates that many cells are closely arranged, a score of B indicates that a small number of cells are slightly loosely arranged, and a score of C indicates very few cells;

[0016] The shape_3 scoring includes: a score of A indicates that many cells are closely arranged, a score of B indicates that a small number of cells are slightly loosely arranged, and a score of C indicates that there are very few or even only single cells.

[0017] In the above-mentioned model construction method for predicting the sustainable pregnancy of embryos, preferably, the development time characteristic of the embryo in S102 is blastocyst_day, which indicates the time for the embryo to develop from a fertilized egg to a blastocyst.

[0018] In the above-mentioned model construction method for predicting the sustainable pregnancy of an embryo, preferably, the genomic information of the embryo in S102 includes:

[0019] For the copy number variation results with 10M:0M resolution, the threshold for reporting mosaic ratio is 30%;

[0020] 10M_mos: copy number variation results at 10 resolution, the threshold for reporting mosaic ratio is 50%;

[0021] Arm: copy number variation results at chromosome arm level resolution, with a mosaic reporting threshold of 30%;

[0022] Arm_mos: copy number variation results at chromosome arm level resolution, with a mosaic reporting threshold of 50%;

[0023] Chrom: copy number variation results at chromosome level resolution, with a mosaic reporting threshold of 30%;

[0024] Chrom_mos: copy number variation results at chromosome level resolution, with a mosaic reporting threshold of 50%;

[0025] NormalNum: the number of times euploid is reported in the cases of 10M, 10M_mos, Arm, Arm_mos, Chrom and Chrom_mos;

[0026] ChromNum: aneuploid chromosome number;

[0027] Mos: maximum mosaic ratio of aneuploidy;

[0028] Type: the size of the abnormal fragment with the maximum mosaic ratio;

[0029] Sex: whether the sex chromosomes are abnormal;

[0030] log_MT_copy_number: mitochondrial copy number after log2 conversion;

[0031] AI_Prob: The probability that the embryo is euploid, the value is 0-100;

[0032] AI_Rating: Embryo rating, the value is A>B>C.

[0033] In the above-mentioned model construction method for predicting the sustainable pregnancy of an embryo, preferably, the step S101 comprises:

[0034] S1011: The embryo samples were cultured to the blastocyst stage to obtain blastocysts, and the time from the fertilized egg to the blastocyst was recorded;

[0035] S1012: performing copy number variation detection and analysis on the culture medium of the embryo sample, outputting the genomic information of the embryo, evaluating the blastocyst according to the Gardner morphological evaluation method, and recording the morphological evaluation score;

[0036] S1013: Select embryos for transplantation and record whether the pregnancy continues after the embryo transplantation, wherein the criterion for determining the continued pregnancy is that the embryo is normally pregnant, that is, after 10-12 weeks, the fetus is still developing normally in the uterus and has a fetal heartbeat.

[0037] In the above-mentioned model construction method for predicting the sustainable pregnancy of an embryo, preferably, the step S103 comprises:

[0038] S1031: using the data of the training set to train a random forest model and an XGBoost model respectively, to obtain an optimal random forest model, whose hyper parameters are max_features=0.6, max_depth=6, and an optimal XGBoost model, whose hyper parameters are eta=0.075, max_depth=7, subsample=0.9, colsample_bytree=0.6;

[0039] S1032: Use the training set to train the weight values ​​of the random forest optimal model and the XGBoost optimal model to obtain a weighted fusion model of random forest and XGBoost, wherein the random forest weight value is 0.5 and the XGBoost weight value is 0.5.

[0040] The present application also provides a system for predicting the sustainable pregnancy of an embryo, the system comprising an embryo sample pre-processing module, an embryo processing feature processing module and a model prediction module:

[0041] The pre-processing module of the embryo sample is configured to score the morphology of the embryo after culturing the embryo into a blastocyst according to the Gardner scoring rule and to score the expansion of the blastocyst cavity, the inner cell mass and the trophoblast of the blastocyst; or, when the embryo does not develop well and cannot develop into a blastocyst or the development is abnormal and cannot get the three scores, only record the morphological characteristics of the embryo and do not perform subsequent analysis on the embryo; record the number of days from the embryo to the blastocyst as the development time of the embryo; extract DNA from the embryo culture medium and construct a sequencing library for sequencing, and then perform copy number variation analysis and non-invasive embryo implantation potential screening artificial intelligence scoring, which is recorded as the genomic characteristics of the embryo;

[0042] The embryo processing feature processing module is configured to process the features in the embryo pre-processing module, including converting the embryo development time into a floating point type, converting the morphological features according to the above method, and converting the embryo genome features according to the above method to obtain the converted features;

[0043] The model prediction module is configured to input the converted features into the above-constructed model for predicting the sustainable pregnancy of the embryo, and the model outputs the probability of the sustainable pregnancy of the embryo.

[0044] The present application also provides an electronic device, comprising:

[0045] Processor and memory;

[0046] The memory is used to store computer operation instructions;

[0047] The processor is used to complete the method for predicting the sustainable pregnancy of an embryo by calling the computer operation instruction, and the method for predicting the sustainable pregnancy of an embryo includes:

[0048] After the embryo is cultured into a blastocyst, the morphology of the embryo is scored according to the Gardner scoring rule, and the expansion of the blastocyst cavity, the inner cell mass and the trophoblast of the blastocyst are scored; or, if the embryo does not develop well and does not develop into a blastocyst or the development is abnormal and the three scores cannot be obtained, only the morphological characteristics of the embryo are recorded, and no subsequent analysis of the embryo is performed; the number of days from the embryo to the blastocyst is recorded as the development time of the embryo; DNA is extracted from the culture medium of the embryo and a sequencing library is constructed for sequencing, and then copy number variation analysis and non-invasive embryo implantation potential screening artificial intelligence score are performed, which is recorded as the genomic characteristics of the embryo;

[0049] Processing the features in the embryo pre-processing module, including converting the embryo development time into a floating point type, converting the morphological features according to the above method, and converting the embryo genome features according to the above method to obtain the converted features;

[0050] The converted features are input into the above-constructed model for predicting the sustainable pregnancy of the embryo, and the model outputs the probability of the sustainable pregnancy of the embryo.

[0051] The present application also provides a computer-readable storage medium for storing computer instructions. When the computer instructions are executed by a processor, a method for predicting the sustainable pregnancy of an embryo is completed; the method for predicting the sustainable pregnancy of an embryo includes:

[0052] After the embryo is cultured into a blastocyst, the morphology of the embryo is scored according to the Gardner scoring rule, and the expansion of the blastocyst cavity, the inner cell mass and the trophoblast of the blastocyst are scored; or, if the embryo does not develop well and does not develop into a blastocyst or the development is abnormal and the three scores cannot be obtained, only the morphological characteristics of the embryo are recorded, and no subsequent analysis of the embryo is performed; the number of days from the embryo to the blastocyst is recorded as the development time of the embryo; DNA is extracted from the culture medium of the embryo and a sequencing library is constructed for sequencing, and then copy number variation analysis and non-invasive embryo implantation potential screening artificial intelligence score are performed, which is recorded as the genomic characteristics of the embryo;

[0053] Processing the features in the embryo pre-processing module, including converting the embryo development time into a floating point type, converting the morphological features according to the above method, and converting the embryo genome features according to the above method to obtain the converted features;

[0054] The model prediction module is configured to input the converted features into the above-constructed model for predicting the sustainable pregnancy of the embryo, and the model outputs the probability of the sustainable pregnancy of the embryo.

[0055] The technical solution of the present invention has the following beneficial technical effects:

[0056] The method, system, electronic device and storage medium for constructing a prediction model for embryo sustainable pregnancy provided by the present invention integrate morphological information, embryo development information and embryo genetic information to construct a prediction model for embryo sustainable pregnancy based on weighted fusion of random forest and XGBoost machine learning methods. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] Figure 1 The conversion principle for shape_quality;

[0058] Figure 2 The architecture of SNB_model;

[0059] Figure 3 The receiver operating characteristic curve of the model in the independent test set;

[0060] Figure 4 Analysis process for the embryo ongoing pregnancy prediction model. DETAILED DESCRIPTION

[0061] In order to have a clearer understanding of the technical features, purposes and beneficial effects of the present invention, the technical solution of the present invention is now described in detail below, but it should not be construed as limiting the applicable scope of the present invention.

[0062] Embodiment 1:

[0063] This embodiment provides a method for constructing a prediction model for embryo sustainable pregnancy, and the process is as follows:

[0064] 1. Sample processing:

[0065] (1) 1024 embryo samples were collected, of which 808 were used for model training and 266 were used as independent validation sets;

[0066] (2) Cultivate the embryo sample to the blastocyst stage and record the time from fertilized egg to blastocyst;

[0067] (3) Then the embryo sample culture medium is subjected to copy number variation detection and analysis using Yikon's non-invasive embryo implantation potential screening (NICS) product to output the embryo's genomic information;

[0068] (4) evaluating the blastocysts according to the Gardner morphological evaluation method and recording the morphological evaluation scores;

[0069] (5) Select embryos for transplantation according to conventional methods and record whether the pregnancy continues after embryo transplantation. The criterion for continued pregnancy is that the embryo is normal, that is, after 10-12 weeks, the fetus is still developing normally in the uterus and has a fetal heartbeat.

[0070] 2. Select features:

[0071] (1) Morphological characteristics: The morphological information of the embryo is evaluated according to the Gardner scoring method, which evaluates the blastocyst stage embryo in three aspects: the expansion of the blastocyst cavity, the inner cell mass (ICM) and the trophoblast (TE). The specific selected morphological characteristics and embryo morphological characteristics processing are shown in Table 1:

[0072] Table 1

[0073] Morphological characteristics of embryos explain shape Gardner scoring method for overall embryo scoring shape_1 Scoring of the blastocyst cavity by Gardner scoring method shape_2 Gardner scoring method for inner cell mass (ICM) shape_3 Gardner scoring method for trophoblast (TE) shape23 Gardner scoring method for the overall score of the inner cell mass (ICM) and trophoblast (TE) shape_quality Convert Gardner score to three levels: Good, Fair, Poor

[0074] Among them, the scores of shape_1 include: a score of 1 indicates an early blastocyst with a blastocyst cavity less than half of the total volume of the embryo, a score of 2 indicates a blastocyst cavity larger than half of the total volume of the embryo, a score of 3 indicates a complete blastocyst cavity, a score of 4 indicates an extended blastocyst with the blastocyst cavity expanding beyond the initial embryo volume, a score of 5 indicates a hatching blastocyst, and a score of 6 indicates a fully hatched blastocyst;

[0075] The scores for shape_2 include: a score of A indicates that many cells are closely packed, a score of B indicates that a small number of cells are slightly loosely packed, and a score of C indicates very few cells;

[0076] The shape_3 scoring includes: a score of A indicates that many cells are closely arranged, a score of B indicates that a small number of cells are slightly loosely arranged, and a score of C indicates that there are very few or even only single cells.

[0077] Shape_quality, its conversion principle was proposed by Munné, S et al. (Munné, S., Kaplan, B., Frattarelli, JL, Child, T., Nakhuda, G., Shamma, FN, Silverberg, K., Kalista, T., Handyside, AH, Katz-Jaffe, M., Wells, D., Gordon, T., Stock-Myer, S., Willman, S., Acacio, B., Lavery, S., Carby, A., Boostanfar, R., Forman, R., … Alfarawati, S.; Preimplantation genetic testing for aneuploidy versus morphology as selection criteria for single frozen-thawed embryo transfer in good-prognosis patients: A multicenter randomized clinical trial; Fertility and Sterility; 2019, 112(6), 1071-1079.e7.), the conversion principle of shape_quality can be found in Figure 1 , embryos with poor development were not included in the analysis, and only embryos with 3 morphological scores were analyzed.

[0078] (2) Embryo developmental time characteristics: The embryo developmental time characteristics selected for constructing the model are blastocyst_day, which is the time it takes for the embryo to develop from the fertilized egg to the blastocyst.

[0079] (3) Genomic information of embryos: The genomic information of embryos is the information output by the process after copy number variation detection using Yikon's NICS product. The information output by NICS-AI includes two: AI_Prob is the probability that the embryo is euploid, and AI_Rating is the rating of the embryo according to AI_Prob. AI_Prob is greater than or equal to 0.94 and is rated A, AI_Prob is between 0.7-0.94 and is rated B, and AI_Prob is less than or equal to 0.7 and is rated B. The specific selected genomic information features are shown in Table 2:

[0080] Table 2

[0081]

[0082] 3. Model construction:

[0083] The model construction adopts the model fusion method. The reason for choosing the fusion model is that the Ensembl method has shown relatively high performance and stability in practical applications. The base model of the fusion model selects two widely used ensemble learning methods, random forest and xgboost. These methods have good performance and stable results. The fused machine learning algorithm model includes two models, one is the model built by the random forest algorithm, and the other is the model built by the XGBoost algorithm. The averaging method is used for fusion, that is, the probability prediction value of continued pregnancy output by the model is the weighted average of the probability prediction value of continued pregnancy of the random forest model and the probability prediction value of continued pregnancy of the XGBoost model. The architecture of the trained model (SNB_model) is shown in Figure 2 .

[0084] The model training steps are as follows. The random forest model and the XGBoost model are trained with the training set respectively. The hyperparameter ranges set during the random forest model training are shown in Table 3, and the hyperparameter ranges set during the XGBoost model training are shown in Table 4:

[0085] Table 3

[0086] Random Forest Hyperparameters Parameter value range max_features 0.1-1.0 max_depth 2-20 n_estimators 1000

[0087] Table 4

[0088] XGBoost Hyperparameters Parameter value range eta 0.01-0.15 max_depth 2-20 subsample 0.1-1.0 colsample_bytree 0.1-1.0 n_estimatros 100

[0089] After the two models are trained, the model weights are also determined as hyperparameters through training. The weight value range of the model fusion is shown in Table 5.

[0090] Table 5

[0091] Fusion Model Weight value range Random Forest Model 0.0-1.0 XGBoost Model 0.0-1.0

[0092] The hyperparameters of SNB_model are as follows:

[0093] The optimal hyperparameters for the random forest model are: max_features=0.6, max_depth=6;

[0094] The optimal hyperparameters for the XGBoost model are: eta = 0.075, max_depth = 7, subsample = 0.9, colsample_bytree = 0.6;

[0095] The weights of model fusion are: the optimal weight value of random forest is 0.5, and the optimal weight value of XGBoost is 0.5.

[0096] 4. Evaluation Model

[0097] The 10-fold cross validation method was used to evaluate the model. For each model determined by the hyperparameter combination, the training set was divided into 10 parts. 9 parts were used for training first, and the remaining 1 part was used for testing. This cycle was repeated 10 times until every part of the data was used for testing. The test results of the 10 parts of the data were then used as the evaluation results of the model. The receiver operating characteristic (ROC) curve results of the model are shown in Figure 3 , where Blastocyst_day is the ROC of the Blatocyst day feature (the time it takes for the embryo to develop from the fertilized egg to the blastocyst) to predict the continued pregnancy of the embryo, Morphology is the ROC of the morphological feature shape_quality to predict the continued pregnancy of the embryo, NICSA is the ROC of AI_Prob output by NICS-AI to predict the continued pregnancy of the embryo, and SNB_model is the ROC of the model constructed in this application. The detailed model performance evaluation results are shown in Table 6.

[0098] Table 6

[0099]

[0100] Embodiment 2:

[0101] This embodiment provides a system (device) and method for predicting the fetal sustainable pregnancy. The fetal sustainable pregnancy prediction model analysis process is shown in Figure 4 The system includes an embryo sample pre-processing module, an embryo processing feature processing module and a model prediction module:

[0102] The pre-processing module of the embryo sample is configured to score the morphology of the embryo after culturing the embryo to the blastocyst according to the Gardner scoring rule and score the expansion of the blastocyst cavity, the inner cell mass (ICM) and the trophoblast (TE) of the blastocyst; or, when the embryo does not develop well and does not develop into a blastocyst or the development is abnormal and does not get the three scores, only record the morphological characteristics of the embryo and do not perform subsequent analysis on the embryo; record the number of days from the embryo to the blastocyst as the development time of the embryo; extract DNA from the culture medium of the embryo and construct a sequencing library for sequencing, then perform CNV analysis and NICS-AI scoring, record the CNV analysis results and NICS-AI scores, and record them as the genomic characteristics of the embryo. For specific steps, refer to patent CN112582022B;

[0103] The embryo processing feature processing module is configured to process the features in the embryo pre-processing module, specifically including converting the embryo development time into a floating point type, performing morphological feature conversion according to the method described in Example 1, and performing embryo genome feature conversion according to the method described in Example 1 to obtain converted features;

[0104] A model prediction module is configured to input the converted features into the model for predicting the sustainable pregnancy of the embryo constructed in Example 1, and the model outputs the probability of sustainable pregnancy of the embryo;

[0105] It should be noted that the above modules correspond to the steps described in Example 1, and the examples and application scenarios implemented by the above modules and the corresponding steps are the same, but are not limited to the contents disclosed in the above Example 1. It should be noted that the above modules, as part of the system, can be executed in a computer system such as a set of computer executable instructions.

[0106] In further embodiments, there is also provided:

[0107] An electronic device includes a processor and a memory; the memory is used to store computer operation instructions; the processor is used to complete an embryo sustainable pregnancy prediction method by calling the computer operation instructions, and the embryo sustainable pregnancy prediction method includes:

[0108] The embryo samples were divided into a training set and an independent validation set, and the embryo sustainable pregnancy was evaluated;

[0109] Select the morphological characteristics of the embryos, the developmental time characteristics of the embryos, and the genomic information of the embryos;

[0110] Construct a prediction model for embryo sustainable pregnancy based on the weighted average fusion of random forest algorithm and XGBoost algorithm;

[0111] The pregnancy sustainability of the embryo samples provided by the patient is predicted based on the prediction model of embryo sustainable pregnancy.

[0112] It should be understood that in this embodiment, the processor may be a central processing unit CPU, and the processor may also be other general-purpose processors, digital signal processors DSP, application-specific integrated circuits ASIC, off-the-shelf programmable gate arrays FPGA or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc.

[0113] The memory may include a read-only memory and a random access memory, and provide instructions and data to the processor. A portion of the memory may also include a non-volatile random access memory. For example, the memory may also store information about the device type.

[0114] A computer-readable storage medium is used to store computer instructions. When the computer instructions are executed by a processor, a method for predicting the sustainable pregnancy of an embryo is completed; the method for predicting the sustainable pregnancy of an embryo comprises:

[0115] 808 samples out of 1024 embryo samples were used as training sets, and 266 samples were used as independent validation sets to evaluate the embryo sustainable pregnancy.

[0116] Select the morphological characteristics of the embryos, the developmental time characteristics of the embryos, and the genomic information of the embryos;

[0117] Construct a prediction model for embryo sustainable pregnancy based on the weighted average fusion of random forest algorithm and XGBoost algorithm;

[0118] The pregnancy sustainability of the embryo samples provided by the patient is predicted based on the prediction model of embryo sustainable pregnancy.

[0119] The method can be directly implemented as a hardware processor, or can be implemented by a combination of hardware and software modules in the processor. The software module can be located in a mature storage medium in the field such as a random access memory, a flash memory, a read-only memory, a programmable read-only memory, or an electrically erasable programmable memory, a register, etc. The storage medium is located in the memory, and the processor reads the information in the memory and completes the steps of the above method in combination with its hardware. To avoid repetition, it will not be described in detail here.

[0120] Those skilled in the art will appreciate that the units, i.e., algorithm steps, of the various examples described in conjunction with this embodiment can be implemented in electronic hardware or in a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0121] Although the above describes the specific implementation mode of the present invention in conjunction with the accompanying drawings, it is not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art on the basis of the technical solution of the present invention without creative work are still within the scope of protection of the present invention.

Claims

1. A method for constructing a model for predicting the sustainable pregnancy of an embryo, comprising the following steps: S101: Divide the embryo samples into a training set and an independent validation set, and evaluate the embryo's sustainable pregnancy potential; S102: Select the morphological characteristics of the embryo, the developmental time characteristics of the embryo, and the genomic information of the embryo; S103: Construct a prediction model for embryo sustainable pregnancy based on the weighted average fusion of random forest algorithm and XGBoost algorithm.

2. The method for constructing a model for predicting embryo sustainable pregnancy according to claim 1, wherein: The morphological characteristics of the embryo in S102 are scores obtained by evaluating according to the Gardner scoring method, and the morphological characteristics of the embryo include: shape_1: Gardner scoring method for the blastocyst cavity; shape_2: Gardner scoring method for inner cell mass; shape_3: Gardner scoring method for trophoblast; shape_quality: Convert shape_1, shape_2, and shape_3 scores into three levels: Good Fair Poor.

3. The method for constructing a model for predicting embryo sustainable pregnancy according to claim 2, wherein: The shape_1 scoring includes: a score of 1 for an early blastocyst whose blastocyst cavity is less than half of the total volume of the embryo, a score of 2 for a blastocyst cavity larger than half of the total volume of the embryo, a score of 3 for a complete blastocyst cavity, a score of 4 for an extended blastocyst whose blastocyst cavity has expanded beyond the initial embryo volume, a score of 5 for a hatching blastocyst, and a score of 6 for a fully hatched blastocyst; The shape_2 scores include: a score of A indicates that many cells are closely arranged, a score of B indicates that a small number of cells are slightly loosely arranged, and a score of C indicates very few cells; The shape_3 scoring includes: a score of A indicates that many cells are closely arranged, a score of B indicates that a small number of cells are slightly loosely arranged, and a score of C indicates that there are very few or even only single cells.

4. The method for constructing a model for predicting embryo sustainable pregnancy according to claim 1, wherein: The embryo development time characteristic in S102 is blastocyst_day, which indicates the time it takes for the embryo to develop from a fertilized egg to a blastocyst.

5. The method for constructing a model for predicting embryo sustainable pregnancy according to claim 1, wherein: The genomic information of the embryo in S102 includes: 10M: The copy number variation results with 10M resolution, the reporting threshold of mosaic ratio is 30%; 10M_mos: copy number variation results at 10M resolution, the threshold for reporting mosaic ratio is 50%; Arm: copy number variation results at chromosome arm level resolution, with a mosaic reporting threshold of 30%; Arm_mos: copy number variation results at chromosome arm level resolution, with a mosaic reporting threshold of 50%; Chrom: copy number variation results at chromosome level resolution, with a mosaic reporting threshold of 30%; Chrom_mos: copy number variation results at chromosome level resolution, with a mosaic reporting threshold of 50%; NormalNum: the number of times euploid is reported in the cases of 10M, 10M_mos, Arm, Arm_mos, Chrom and Chrom_mos; ChromNum: aneuploid chromosome number; Mos: maximum mosaic ratio of aneuploidy; Type: the size of the abnormal fragment with the maximum mosaic ratio; Sex: whether the sex chromosomes are abnormal; log_MT_copy_number: mitochondrial copy number after log2 conversion; AI_Prob: The probability that the embryo is euploid, the value is 0-100; AI_Rating: Embryo rating, the value is A>B>C.

6. The method for constructing a model for predicting embryo sustainable pregnancy according to claim 1, wherein: The step of S101 includes: S1011: The embryo samples were cultured to the blastocyst stage to obtain blastocysts, and the time from the fertilized egg to the blastocyst was recorded; S1012: performing copy number variation detection and analysis on the culture medium of the embryo sample, outputting the genomic information of the embryo, evaluating the blastocyst according to the Gardner morphological evaluation method, and recording the morphological evaluation score; S1013: Select embryos for transplantation and record whether the pregnancy continues after the embryo transplantation, wherein the criterion for determining the continued pregnancy is that the embryo is normally pregnant, that is, after 10-12 weeks, the fetus is still developing normally in the uterus and has a fetal heartbeat.

7. The method for constructing a model for predicting embryo sustainable pregnancy according to claim 1, wherein: The step of S103 includes: S1031: using the data of the training set to train a random forest model and an XGBoost model respectively, to obtain an optimal random forest model, whose hyper parameters are max_features=0.6, max_depth=6, and an optimal XGBoost model, whose hyper parameters are eta=0.075, max_depth=7, subsample=0.9, colsample_bytree=0.6; S1032: Use the training set to train the weight values ​​of the random forest optimal model and the XGBoost optimal model to obtain a weighted fusion model of random forest and XGBoost, wherein the random forest weight value is 0.5 and the XGBoost weight value is 0.

5.

8. A system for predicting the sustainable pregnancy of an embryo, the system comprising an embryo sample pre-processing module, an embryo processing feature processing module and a model prediction module: The pre-processing module of the embryo sample is configured to score the morphology of the embryo after culturing the embryo into a blastocyst according to the Gardner scoring rule and to score the expansion of the blastocyst cavity, the inner cell mass and the trophoblast of the blastocyst; or, when the embryo does not develop well and cannot develop into a blastocyst or the development is abnormal and cannot get the three scores, only record the morphological characteristics of the embryo and do not perform subsequent analysis on the embryo; record the number of days from the embryo to the blastocyst as the development time of the embryo; extract DNA from the embryo culture medium and construct a sequencing library for sequencing, and then perform copy number variation analysis and non-invasive embryo implantation potential screening artificial intelligence scoring, which is recorded as the genomic characteristics of the embryo; The embryo processing feature processing module is configured to process the features in the embryo pre-processing module, including converting the embryo development time into a floating point type, converting the morphological features according to the method according to claim 2 or 3, and converting the embryo genome features according to the method according to claim 5, to obtain the converted features; The model prediction module is configured to input the converted features into a model for predicting the sustainable pregnancy of an embryo constructed according to any one of claims 1 to 7, and the model outputs the probability of sustainable pregnancy of the embryo.

9. An electronic device comprising: Processor and memory; The memory is used to store computer operation instructions; The processor is used to complete the method for predicting the sustainable pregnancy of an embryo by calling the computer operation instruction, and the method for predicting the sustainable pregnancy of an embryo includes: After the embryo is cultured into a blastocyst, the morphology of the embryo is scored according to the Gardner scoring rule, and the expansion of the blastocyst cavity, the inner cell mass and the trophoblast of the blastocyst are scored; or, if the embryo does not develop well and does not develop into a blastocyst or the development is abnormal and the three scores cannot be obtained, only the morphological characteristics of the embryo are recorded, and no subsequent analysis of the embryo is performed; the number of days from the embryo to the blastocyst is recorded as the development time of the embryo; DNA is extracted from the culture medium of the embryo and a sequencing library is constructed for sequencing, and then copy number variation analysis and non-invasive embryo implantation potential screening artificial intelligence score are performed, which is recorded as the genomic characteristics of the embryo; Processing the features in the embryo pre-processing module, including converting the embryo development time into a floating point type, converting the morphological features according to the method according to claim 2 or 3, and converting the embryo genome features according to the method according to claim 5, to obtain converted features; The converted features are input into a model for predicting the sustainable pregnancy of an embryo constructed according to any one of claims 1 to 7, and the model outputs the probability of the sustainable pregnancy of the embryo.

10. A computer-readable storage medium for storing computer instructions, wherein when the computer instructions are executed by a processor, a method for predicting the sustainable pregnancy of an embryo is completed; The method for predicting the sustainable pregnancy of the embryo comprises: After the embryo is cultured into a blastocyst, the morphology of the embryo is scored according to the Gardner scoring rule, and the expansion of the blastocyst cavity, the inner cell mass and the trophoblast of the blastocyst are scored; or, if the embryo does not develop well and does not develop into a blastocyst or the development is abnormal and the three scores cannot be obtained, only the morphological characteristics of the embryo are recorded, and no subsequent analysis of the embryo is performed; the number of days from the embryo to the blastocyst is recorded as the development time of the embryo; DNA is extracted from the culture medium of the embryo and a sequencing library is constructed for sequencing, and then copy number variation analysis and non-invasive embryo implantation potential screening artificial intelligence score are performed, which is recorded as the genomic characteristics of the embryo; Processing the features in the embryo pre-processing module, including converting the embryo development time into a floating point type, converting the morphological features according to the method according to claim 2 or 3, and converting the embryo genome features according to the method according to claim 5, to obtain converted features; The converted features are input into a model for predicting the sustainable pregnancy of an embryo constructed according to any one of claims 1 to 7, and the model outputs the probability of the sustainable pregnancy of the embryo.

Images