JP2024005549A - Time lapse image discriminator, learning method of time lapse image discriminator, and embryo discrimination device and embryo discrimination method - Google Patents

Abstract

To provide a time lapse image discriminator, a learning method of the time lapse image discriminator, and an embryo discrimination device and an embryo discrimination method that enhance discrimination precision for an embryo.SOLUTION: A time lapse image discriminator consists of an input layer to which a time lapse image obtained by imaging an embryo of a patient to be discriminated is input, an output layer to which a discrimination result of the embryo is output, and an intermediate layer which calculates the discrimination result to be output to the output layer by using the time lapse image input to the input layer and parent body information related to the parent body of the patient. The intermediate layer is learnt through a neural network, a learning time lapse image for a learning embryo of a learning patient is input to the input layer in the learning of the intermediate layer, and learning parent body information related to the parent body of the learning patient is input to the intermediate layer.SELECTED DRAWING: Figure 4

Description

The present invention relates to a time-lapse image discriminator, a learning method for a time-lapse image discriminator, an embryo discrimination device, and an embryo discrimination method.

The number of pregnancies delivered using assisted reproductive technology (ART) in Japan has been increasing in recent years. In 2019, the number of deliveries using ART exceeded 60,000, which is approximately 7% of the annual number of deliveries. In ART, multiple collected eggs are cultured until they become blastocysts in an embryo culture device (incubator), and the embryos selected as good embryos from among the cultured multiple embryos (fertilized eggs) are implanted into the uterus. It is a treatment for infertility.

Selection of good embryos is carried out by an embryo cultivator who visually observes the embryos immediately after fertilization (early embryos) and 5 days after fertilization (blastocysts). The accuracy of evaluation of embryo selection by the embryologist's visual inspection depends on the embryologist's knowledge and experience. Therefore, a uniform and highly accurate embryo evaluation method is required.

On the other hand, it is not uncommon for good embryos that are implanted in the uterus to end up in miscarriage. Many of the causes of miscarriage are chromosomal abnormalities (karyotype abnormalities). It is known that the probability of chromosomal abnormalities increases with the age of the patient. Therefore, for patients who have reached a certain age, evaluating chromosomal abnormalities is more important than evaluating the quality of the embryo (whether it is a good embryo or not) in determining the success or failure of obtaining a live birth from embryo transfer. Therefore, in recent years, Preimplantation Genetic Testing for Aneuploidy (PGT-A), which examines the number of chromosomes in the embryo before it is transferred to the uterus, has been conducted for cases of habitual miscarriage and repeated ART failures. ing. However, PGT-A has issues such as invasiveness to embryos and high testing costs.

Until now, inventions related to embryo culture devices (time-lapse incubators) that image embryos being cultured at predetermined intervals to generate time-lapse images, and inventions related to embryo evaluation using machine-learning models using time-lapse images have been made. have been proposed (for example, see Patent Documents 1 and 2, and Non-Patent Documents 1-3).

Patent No. 6956302 specification Patent No. 6414310 specification

Jorgen Berntsen et.al., Robust and generalizable embryo selection based on artificial intelligence and time-lapse image sequences, PLOS ONE, 2022. Tran et.al., Deep learning as a predictive tool for fetal heart pregnancy following time-lapse incubation and blastocyst transfer, Human Reproduction, 2018. Khosravi et.al., Deep learning enables robust assessment and selection of human blastocyst after in vitro fertilization, NPJ Digital Medicine, 2019.

An object of the present invention is to provide a time-lapse image discriminator, a learning method for a time-lapse image discriminator, an embryo discrimination device, and an embryo discrimination method that improve embryo discrimination accuracy (prediction accuracy).

The time-lapse image discriminator according to the present invention includes an input layer into which a time-lapse image of an embryo of a patient to be discriminated is input, an output layer to which a discrimination result of the embryo is output, and a time-lapse image input into the input layer. and maternal information related to the patient's mother, and an intermediate layer that calculates a discrimination result to be output to the output layer, the intermediate layer is a neural network. In the learning of the middle layer, the learning time-lapse image of the learning embryo of the learning patient is input to the input layer, and the learning maternal information related to the mother of the learning patient is input to the middle layer. characterized by being done.

The present invention improves the accuracy of embryo discrimination.

FIG. 1 is a functional block diagram showing an embodiment of an embryo discrimination device according to the present invention. FIG. 2 is a schematic diagram showing an example of a conventional neural network model. FIG. 2 is a schematic diagram showing another example of a conventional neural network model. FIG. 2 is a schematic diagram showing an example of a neural network model used as a time-lapse image classifier according to the present invention. FIG. 3 is a schematic diagram showing another example of a neural network model used as the time-lapse image discriminator. FIG. 7 is a schematic diagram showing yet another example of a neural network model used as the time-lapse image discriminator. FIG. 7 is a schematic diagram showing yet another example of a neural network model used as the time-lapse image discriminator.

Embodiments of a time-lapse image classifier, a learning method for a time-lapse image classifier, an embryo classifier, and an embryo classifier method according to the present invention will be described below with reference to the drawings.

●Configuration of embryo discrimination device●
FIG. 1 is a functional block diagram showing an embodiment of an embryo discrimination device (hereinafter referred to as "this device") according to the present invention.

The present device 1 is an embryo culturing device (time-lapse incubator) that captures a time-lapse image of an embryo while cultivating an embryo of a patient to be determined. The present device 1 also functions as a time-lapse image discriminator in which a time-lapse image discriminator (hereinafter referred to as "main discriminator") according to the present invention operates.

This discriminator is a model of a trained neural network that has been machine learned using supervised learning data using the learning method for a time-lapse image discriminator according to the present invention (hereinafter referred to as "this learning method"). It is. Details of this learning method will be described later. The neural network is, for example, a convolutional neural network, which is said to have high image discrimination accuracy. The learning data includes time-lapse images and maternal information of embryos for each learning patient, and progress information of embryos for each learning patient. The embryo progress information is information that is to be discriminated (predicted) by this discriminator, and includes, for example, whether the embryo has implanted successfully or not, and the presence or absence of chromosomal abnormalities (whether or not it is a euploid embryo).

This discriminator identifies good embryos with a high implantation rate based on time-lapse images and maternal information of embryos (fertilized eggs) of patients to be discriminated, that is, patients who wish to give birth through assisted reproductive technology. or whether it is a chromosomal abnormality with a high miscarriage rate.

In the following description, it is assumed that the object of discrimination by this discriminator is whether the embryo is a good embryo or not.

Note that, instead of operating in a time-lapse incubator, this discriminator may operate in an information processing device such as a personal computer. The information processing device on which this discriminator operates functions as a time-lapse image discriminator. The present discriminator discriminates the patient's embryo (predicts the progress of the embryo) in cooperation with the hardware resources of the information processing device in which the present discriminator operates.

Here, the hardware resources of the information processing device on which this discriminator operates are, for example, processors such as a CPU (Central Processing Unit), MPU (Micro Processing Unit), and DSP (Digital Signal Processor), and storage media. This storage medium stores the main discriminator, a time-lapse image of the patient to be discriminated, maternal information, and the like. The processor on which this discriminator operates uses information stored in a storage medium included in the information processing device on which this discriminator operates, using each of the below-mentioned means (input section, acquisition section, discriminating section, output section).

The device 1 includes a storage section 2, an imaging section 3, an input section 4, an acquisition section 5, a discrimination section 6, and an output section 7.

The storage unit 2 stores information and the like used by the present discriminator and the present apparatus 1 to realize the embryo discrimination method (hereinafter referred to as "the present method") according to the present invention. Details of the information used by this device 1 to implement this method will be described later.

The storage unit of the device 1 may be, for example, a portable storage medium such as an HDD (Hard Disk Drive), an SSD (Solid State Drive), or a flash memory, or other non-temporary storage media, or a RAM (Random Access Memory) or other non-transitory storage medium. Temporary recording media, etc.

The imaging unit 3 captures a time-lapse image of an embryo (cultivated embryo from fertilization to blastocyst) that is being cultured in the device 1. The imaging unit 3 captures time-lapse images at predetermined time intervals (for example, every 10 minutes). The captured time-lapse image is stored in the storage unit 2. The plurality of time-lapse images stored in the storage unit 2 are displayed on the display (not shown) of the device 1 in the order in which they were captured, as if they were a moving image. The operator of the device 1 (for example, an embryo cultivator) can check the changes over time in the displayed time-lapse image.

The input unit 4 inputs maternal information of the patient to be determined. The input unit 4 is information input means included in the device 1, and is, for example, a keyboard, a mouse, a touch panel, or the like. The input unit 4 is operated by the operator of the device 1 . The patient's maternal information is, for example, information regarding the patient's age (information indicating age or age) or information regarding the patient's body (for example, medical history). The maternal information input from the input section 4 is stored in the storage section 2.

The acquisition unit 5 acquires a time-lapse image of the patient's embryo and maternal information stored in the storage unit 2.

The discrimination unit 6 discriminates the patient's embryo based on the time-lapse image and maternal information of the patient's embryo that the acquisition unit 5 acquired from the storage unit 2 and the discriminator stored in the storage unit 2.

The output unit 7 outputs the determination result determined by the determination unit 6. The output unit 7 outputs the determination results, for example, by storing the determination results in the storage unit 2, displaying the determination results on a display (not shown) of the device 1, or connecting to the device 1 via a communication network. For example, the determination result is transmitted to an information processing device (not shown) that is to be used.

●Traditional neural network model●
FIG. 2 is a schematic diagram showing an example of a conventional neural network model.

A neural network is composed of an input layer, an intermediate layer (hidden layer), and an output layer. The neural network here is a discriminator that uses time-lapse images of embryos of patients to be discriminated to determine whether or not the embryos are good embryos. That is, the input layer is a layer into which a time-lapse image is input. The output layer is a layer in which the determination result (whether the embryo is a good embryo or not) is output. The intermediate layer is a layer that connects the input layer and the output layer. A middle layer is a collection of multiple individual middle layers. The individual intermediate layer includes, from the input layer to the output layer, a first individual intermediate layer, a second individual intermediate layer, a third individual intermediate layer, ..., an (n-1) individual intermediate layer, an n-th individual intermediate layer. Contains n individual intermediate layers.

In the figure, neurons (units) that make up the input layer are shown as black circles, neurons that make up the intermediate layer are shown as white circles (solid lines), and neurons that make up the output layer are shown as white circles (dashed lines).

The number of neurons constituting the input layer depends on the resolution of the time-lapse image. When the resolution of the time-lapse image is 1280×1024 px, the number of neurons forming the input layer is 1,310,720 (=1280×1024). That is, each neuron forming the input layer is provided with information regarding the brightness value and RGB for each pixel forming the time-lapse image.

Note that the neurons forming the input layer may include neurons corresponding to so-called bias. In this case, the number of neurons forming the input layer is the number corresponding to the resolution of the time-lapse image plus the number of biases.

The number of neurons in the output layer corresponds to the number of discrimination results. The determination result here is whether the embryo is a good embryo or not. That is, the number of neurons in the output layer is 2 (“good embryo”, “not good embryo”).

Neural network In the learning process, it is set as appropriate based on evaluation values such as AUC (Area Under the Curve).

FIG. 3 is a schematic diagram showing another example of a conventional neural network model.
This figure shows that neurons indicated by black squares have been added to the input layer of the model shown in FIG. 2. The black square neurons include maternal information of the patient to be determined. That is, the difference between the model shown in FIG. 3 and the model shown in FIG. 2 is whether parent information is included in the input layer (model in FIG. 3) or not (model in FIG. 2). In other words, the model shown in FIG. 2 determines whether a patient's embryo is a good embryo or not based only on the time-lapse image of the patient's embryo. On the other hand, the model shown in FIG. 3 determines whether a patient's embryo is a good embryo or not based on a time-lapse image of the patient's embryo and maternal information.

The number of neurons constituting the input layer of the model shown in Figure 3 is (without considering bias) the number depending on the resolution of the time-lapse image (the number of black circles), and the number of neurons containing the parent information (the number of black squares). ) and the sum of . Here, as described above, when the resolution of the time-lapse image is 1280×1024 px, the number of black circles is 1,310,720. On the other hand, the number of black squares is 1.

In this way, when the time-lapse image and maternal information are input to the input layer and discrimination is performed, the amount of maternal information is overwhelmingly smaller than the amount of information of the time-lapse image. Therefore, in the learning process of the neural network, the matrix information is difficult to be reflected in the learning results. Similarly, in the discrimination process using a trained neural network, maternal information is unlikely to be reflected in the discrimination results.

On the other hand, the quality of the embryo (whether it is a good embryo or not) and the presence or absence of chromosomal abnormalities are important in determining the success or failure of obtaining a live birth from embryo transfer. The quality of the embryo and the presence or absence of chromosomal abnormalities can be influenced by the mother's condition, such as the patient's age and medical history. Therefore, when identifying embryos using a neural network, it is desirable to take into account information related to the patient's mother.

●Configuration of time-lapse image classifier●
As explained below, this classifier improves the accuracy of discrimination by reflecting mother information, which has an overwhelmingly smaller amount of information than the time-lapse image, in the learning results and discrimination results together with the time-lapse image.

FIG. 4 is a schematic diagram showing an example of a neural network model used as the main discriminator.
This figure shows that neurons indicated by black squares are added to a part of the individual intermediate layer (n-1th individual intermediate layer) of the intermediate layer of the model shown in FIG. The black square neurons are the same as the black square neurons shown in FIG. 3, and include maternal information of the patient to be discriminated. In other words, the difference between the model shown in FIG. 4 and the model shown in FIG. model) or is. That is, the model shown in FIG. 3 determines whether a patient's embryo is a good embryo or not based on the time-lapse image of the patient's embryo and maternal information input to the input layer. On the other hand, the model shown in Figure 4 determines whether or not a patient's embryo is a good embryo using a time-lapse image of the patient's embryo that is input to the input layer and an intermediate layer (n-1 individual intermediate layer). It is determined based on the patient's maternal information that is input.

Here, in the neural network model used as the main discriminator in Figure 4, during the learning process using training data, time-lapse images of embryos for each training patient are input to the input layer, and The parent information is input to the (n-1)th individual hidden layer and learned.
Similarly, in the neural network model used as the main discriminator in Figure 4, during the discrimination process using the data of the patient to be discriminated, the time-lapse image of the patient's embryo is input to the input layer, and the patient's maternal information is input to the input layer. It is input to the (n-1)th individual intermediate layer and discriminated.

In this way, in the learning process and discrimination process of the neural network model used as the main classifier, by inputting the matrix information, which has an overwhelmingly smaller amount of information compared to the time-lapse image, into the intermediate layer, the matrix information can be used together with the time-lapse image. Compared to the case where information is input to the input layer (the model in FIG. 3), the matrix information is more likely to be reflected in the learning results and discrimination results. This is because, when the maternal information is input to the input layer together with the time-lapse image, the contents of the maternal information are treated to the same extent as one pixel of the time-lapse image. On the other hand, as in this discriminator, when inputting the matrix information to the individual intermediate layers constituting the intermediate layer, especially to the individual intermediate layer close to the output layer where the information of the time-lapse image input to the input layer is aggregated, the matrix The content of the information is emphasized and reflected in learning results and discrimination results. In other words, this discriminator obtains learning results and discrimination results that reflect maternal information that influences the success or failure of obtaining a live birth from embryo implantation. In other words, the present discriminator can improve the accuracy of embryo discrimination.

Note that in the neural network model used as the present discriminator, the matrix information is input to at least one individual intermediate layer among the plurality of individual intermediate layers forming the intermediate layer.

FIG. 5-7 is a schematic diagram showing another example of a neural network model used as the main classifier.

FIG. 5 shows that common base information is input to two individual intermediate layers (the (n-1) individual intermediate layer and the n-th individual intermediate layer) in the learning process and the discrimination process. Multiple individual intermediate layers into which common mother information is input, and positions within the individual intermediate layer where mother information is input (arrangement of time-lapse image neurons and mother information neurons composing the individual intermediate layer: the figure shows , which indicates that the base information is input at the beginning of each individual intermediate layer (at the top of the page) is appropriately set during the learning process.

FIG. 6 shows that common base information is input to two locations in one individual intermediate layer (n-1th individual intermediate layer) in the learning process and the discrimination process. The number of pieces of matrix information input into one individual intermediate layer and the position within the individual intermediate layer at which the matrix information is input are appropriately set during the learning process.

FIG. 7 shows that common base information is input to all the individual intermediate layers constituting the intermediate layer and the input layer in the learning process and the discrimination process. The number of base information input to each individual intermediate layer and the position within the individual intermediate layer at which the base information is input are appropriately set in the learning process.

In addition, in the neural network model used as this discriminator, among the individual intermediate layers that constitute the intermediate layer, the individual intermediate layer into which maternal information is input is based on the time-lapse image and maternal information of the patient to be discriminated. It may be selected as the main discriminator. The method of selecting the individual intermediate layer into which the parent information is input is appropriately set during the learning process using the learning data. That is, for example, in the learning process, when it is evaluated that age has a small effect on the discrimination result in the case of a young person, and that age has a large effect on the discrimination result in the case of a middle-aged and elderly person, the An individual intermediate layer into which parent information is input may be selected. That is, for example, as the patient's age increases, the individual intermediate layer into which maternal information is input may be selected to be located closer to the output layer.

●Summary●
According to the embodiment described above, in the learning process of the neural network model used as the main discriminator, among the time-lapse images of each patient for learning as input data and maternal information, information is compared to the time-lapse images. The amount of base information that is overwhelmingly small is input to any one of the plurality of individual intermediate layers that constitute the intermediate layer. Therefore, compared to the case where the maternal information is input to the input layer together with the time-lapse image, the maternal information is more likely to be reflected in the learning results and the discrimination results. That is, the present discriminator and the present device obtain learning results and discrimination results that reflect maternal information that influences the success or failure of obtaining a live birth from embryo implantation. In other words, the present discriminator and the present device can improve the embryo discrimination accuracy (prediction accuracy).

As mentioned above, this discriminator and this device determine whether the embryo is a good embryo with a high implantation rate or whether it is a chromosomal abnormality with a high miscarriage rate based on the time-lapse image of the patient's embryo and maternal information. It is possible to determine (predict) whether or not there is. In particular, the present discriminator and present device that can determine whether or not there is a chromosomal abnormality can serve as an alternative to PGT-A, which has been the only testing method up until now. As mentioned above, PGT-A has issues such as invasiveness to embryos and high testing costs. By using the present discriminator or the present device for discrimination as an alternative to PGT-A, the number of cases leading to PGT-A can be reduced.

Furthermore, in the embodiment described above, time-lapse images and maternal information of embryos for each patient for learning, as well as progress information of embryos for each patient for learning, were used as learning data for the present discriminator. Alternatively, the learning data of the present discriminator may include the PGT-A test results (embryo chromosome number) of the embryo for each learning patient. That is, the present discriminator may be trained using time-lapse images of embryos, maternal information, PGT-A test results, and progress information of embryos for each learning patient. In this case, this discriminator determines whether the embryo is good or not, or whether it has chromosomal abnormality, based on the time-lapse image of the patient's embryo, maternal information, and PGT-A test results. , is determined. Here, in the learning process or discrimination process of this discriminator, the layer into which the test results of PGT-A are input is the same intermediate layer as the intermediate layer into which the mother information is input (including the input layer in FIG. 7), or , is an intermediate layer (including the input layer in FIG. 7) that is different from the intermediate layer into which the parent information is input. The position in the individual intermediate layer to which the PGT-A test result is input is set as appropriate in the learning process of this discriminator, similar to the determination of the position in the individual intermediate layer to which the parent information is input described above.

●Characteristics of this discriminator, this learning method, this device, and this method●
The features of the present discriminator, the present learning method, the present device, and the present method described so far will be summarized below.

●Characteristics of this classifier This classifier is
an input layer into which a time-lapse image of an embryo of a patient to be determined is input;
an output layer in which the embryo discrimination results are output;
an intermediate layer that calculates the discrimination result to be output to the output layer using the time-lapse image input to the input layer and maternal information related to the patient's mother;
A time-lapse image classifier comprising:
The intermediate layer is trained by a neural network,
In learning the intermediate layer,
A learning time-lapse image of a learning embryo of a learning patient is input to the input layer,
Learning maternal information related to the learning patient's mother is input to the intermediate layer;
It is characterized by

In this classifier,
The maternal information includes information regarding the age of the patient;
But that's fine.

In this classifier,
The intermediate layer includes a plurality of individual intermediate layers,
In learning the intermediate layer, the learning base information is input to at least one of the plurality of individual intermediate layers,
It can be anything.

In this classifier,
Among the plurality of individual intermediate layers, the individual intermediate layer into which the maternal information of the patient to be determined is input is selected based on the time-lapse image or the maternal information of the patient to be determined;
It can be anything.

In this classifier,
In learning the intermediate layer, the learning base information is input to the input layer,
It can be anything.

●Characteristics of this learning method This learning method is
an input layer into which a time-lapse image of an embryo of a patient to be determined is input;
an output layer in which the embryo discrimination results are output;
an intermediate layer that calculates the discrimination result to be output to the output layer using the time-lapse image input to the input layer and maternal information related to the patient's mother;
A learning method for a time-lapse image classifier comprising:
The intermediate layer is trained by a neural network,
In learning the intermediate layer,
inputting a training time-lapse image of a training embryo of a training patient into the input layer;
inputting learning maternal information related to the learning patient's mother into the intermediate layer;
consisting of
It is characterized by

●Features of this device This device has the following features:
a storage unit that stores a discriminator learned based on a learning time-lapse image for each learning patient and learning mother information related to the mother for each learning patient;
an acquisition unit that acquires a time-lapse image of an embryo of a patient to be identified and maternal information related to the patient's mother;
a discrimination unit that discriminates the embryo based on the time-lapse image acquired by the acquisition unit, the maternal information, and the discriminator;
an output unit that outputs the discrimination result determined by the discrimination unit;
It has
The discriminator is a main discriminator,
It is characterized by

●Characteristics of this method This method:
a storage unit that stores a discriminator trained based on a learning time-lapse image for each learning patient and learning mother information related to the mother for each learning patient;
An embryo discrimination method carried out on an embryo discrimination device comprising:
The embryo discrimination device includes:
an acquisition step of acquiring a time-lapse image of an embryo of a patient to be identified and maternal information related to the patient's mother;
a discrimination step of discriminating the embryo based on the time-lapse image acquired in the acquisition step, the maternal information, and the discriminator;
an output step for outputting the discrimination result determined in the discrimination step;
It has
The discriminator is a main discriminator,
It is characterized by

1 Embryo discrimination device 2 Storage unit 3 Imaging unit 4 Input unit 5 Acquisition unit 6 Discrimination unit 7 Output unit

Claims (8)

an input layer into which a time-lapse image of an embryo of a patient to be determined is input;
an output layer in which the embryo discrimination results are output;
an intermediate layer that calculates the discrimination result to be output to the output layer using the time-lapse image input to the input layer and maternal information related to the patient's mother;
A time-lapse image classifier comprising:
The intermediate layer is trained by a neural network,
In learning the intermediate layer,
A learning time-lapse image of a learning embryo of a learning patient is input to the input layer,
Learning maternal information related to the learning patient's mother is input to the intermediate layer;
A time-lapse image discriminator characterized by: The maternal information is information regarding the age of the patient.
The time-lapse image discriminator according to claim 1. The intermediate layer includes a plurality of individual intermediate layers,
In learning the intermediate layer, the learning base information is input to at least one of the plurality of individual intermediate layers,
The time-lapse image discriminator according to claim 1. Among the plurality of individual intermediate layers, the individual intermediate layer into which the maternal information of the patient to be determined is input is selected based on the time-lapse image or the maternal information of the patient to be determined;
The time-lapse image discriminator according to claim 3. In learning the intermediate layer, the learning base information is input to the input layer,
The time-lapse image discriminator according to claim 3. an input layer into which a time-lapse image of an embryo of a patient to be determined is input;
an output layer in which the embryo discrimination results are output;
an intermediate layer that calculates the discrimination result to be output to the output layer using the time-lapse image input to the input layer and maternal information related to the patient's mother;
A learning method for a time-lapse image classifier comprising:
The intermediate layer is trained by a neural network,
In learning the intermediate layer,
inputting a training time-lapse image of a training embryo of a training patient into the input layer;
inputting learning maternal information related to the learning patient's mother into the intermediate layer;
consisting of
A learning method for a time-lapse image classifier characterized by the following. a storage unit that stores a discriminator learned based on a learning time-lapse image for each learning patient and learning mother information related to the mother for each learning patient;
an acquisition unit that acquires a time-lapse image of an embryo of a patient to be identified and maternal information related to the patient's mother;
a discrimination unit that discriminates the embryo based on the time-lapse image acquired by the acquisition unit, the maternal information, and the discriminator;
an output unit that outputs the discrimination result determined by the discrimination unit;
It has
The discriminator is a time-lapse image discriminator according to claim 1,
An embryo discrimination device characterized by: a storage unit that stores a discriminator trained based on a learning time-lapse image for each learning patient and learning mother information related to the mother for each learning patient;
An embryo discrimination method carried out on an embryo discrimination device comprising:
The embryo discrimination device includes:
an acquisition step of acquiring a time-lapse image of an embryo of a patient to be identified and maternal information related to the patient's mother;
a discrimination step of discriminating the embryo based on the time-lapse image acquired in the acquisition step, the maternal information, and the discriminator;
an output step for outputting the discrimination result determined in the discrimination step;
It has
The discriminator is a time-lapse image discriminator according to claim 1,
An embryo identification method characterized by the following.