Method and system for non-invasive evaluation of embryo morphology

Abstract

The invention belongs to the technical field of embryo culture, and discloses a method for non-invasive evaluation of embryo morphology, which utilizes machine learning to photograph and compare embryos at different stages in a culture cycle to obtain and evaluate the embryo state, and specifically comprises the following steps: firstly, carrying out full-cycle culture on a plurality of standard embryos, dividing the culture cycle into a plurality of stages, shooting a plurality of images conforming to embryo morphological standards in each stage, labeling the images, and inputting the images into an AI model for identification to obtain a standard identification model for each stage; then, directly inputting a plurality of images at different stages in training embryo culture into the standard recognition model for training recognition, and performing annotation adjustment according to a judgment result after training to obtain a trained recognition model; the trained recognition model is used for recognizing the embryo image in the culturing process, and the completeness comparison with the completeness of the standard embryo model state in the same stage is used as a score value for scoring.

Description

Technical Field

[0001] The present invention belongs to the technical field of in vitro embryo culture, and in particular relates to a method and system for non-invasively evaluating embryo morphology. Background Art

[0002] In vitro fertilization refers to the process of combining a woman's egg and a man's sperm in a laboratory dish. It is usually performed during a woman's ovulation period. First, superovulation drugs are used to induce the simultaneous maturation of multiple follicles. Then, the eggs are removed from the ovaries through egg retrieval surgery. At the same time, sperm samples are collected. After processing and optimization, the eggs and sperm are placed together in a dish containing a specific culture medium. They are fertilized under conditions that simulate the in vivo environment. After a period of culture, the fertilized eggs will develop into embryos. At this time, embryo transplantation can be performed as needed or culture can continue to reach a more advanced stage, such as the blastocyst stage. During the entire process, temperature, humidity, pH and other conditions need to be precisely controlled to ensure the healthy development of the embryo.

[0003] During the embryo culture process, the development of the embryo will be judged based on morphological characteristics. For the identification of embryo morphological characteristics, there are currently two main methods: manual and image recognition. Among them, the manual method is to observe the development of the embryo at different stages of embryo development. The defect of this method is that regular removal of embryos for manual observation may change the development environment of the embryo and interfere with the normal development of the embryo. And the efficiency of manual observation is low and the generalizability is not strong. The image recognition method is mainly to identify the embryo morphology through embryo pictures and recognition software. Existing studies generally only use embryo pictures to identify a single morphological feature. The defect of this method is that the time factor of embryo development is not considered, and it is simply recognized by pictures. And only a single morphological feature is considered, and different models need to be used to identify different morphological features. Summary of the invention

[0004] In order to solve the problems existing in the prior art, the present invention provides a method and system for non-invasively evaluating embryo morphology, which can provide a more intelligent and rapid embryo development evaluation for the existing in vitro embryo culture process, thereby efficiently evaluating the development cycle of each embryo, thereby comprehensively understanding the development of each embryo and providing support for other research.

[0005] The technical solution adopted by the present invention is:

[0006] In a first aspect, the present invention provides a method for non-invasively evaluating embryo morphology, which uses machine learning to photograph and compare embryos at different stages of the culture cycle to obtain and evaluate the embryo status, and the specific steps are as follows:

[0007] First, a number of standard embryos were cultured throughout the entire cycle, and the culture cycle was divided into several stages. Several images that met the embryo morphology standards were taken at each stage, and the images were annotated and input into the AI ​​model for recognition to obtain a standard recognition model for each stage.

[0008] Then, a number of images of different stages of training embryo culture are directly input into the standard recognition model for training and recognition, and annotation and adjustment are performed based on the judgment results after training to obtain a trained recognition model;

[0009] The trained recognition model is used to identify the images of embryos in culture, and the images are scored based on the completeness comparison with the state of the standard embryo model at the same stage.

[0010] It should be noted that conventional embryo definition refers to the culture process of fertilized eggs 3-8 weeks after human fertilization. The in vitro embryo culture time in this application generally reaches the blastocyst stage and can be transplanted or cryopreserved, and the cycle is relatively short and controlled within 1-2 weeks after the fertilized egg is fertilized. The culture process includes fertilization, culture, and then monitoring and transfer. First, the fertilized egg is obtained by in vitro fertilization (IVF) or other assisted reproductive technologies. The fertilized egg is placed in a culture dish containing a special culture medium, which provides all the nutrition and hormones required for embryo growth. The method in this application is mainly used in the process of taking pictures and monitoring the embryos cultured in the culture dish, without changing the state of the culture dish, only the embryo in the incubator is taken pictures, and then the state of the embryo at this time is judged according to the image taken.

[0011] The culture dish for culturing embryos can be a conventional device, and a camera capable of taking photos can be installed on the outside to automatically control the embryo to take photos. The culture dish is usually placed in a constant temperature box to ensure that it is in an optimal temperature environment, and the camera can be placed at any observable position to take photos of the embryo at regular intervals.

[0012] This application uses AI model training to automatically identify and judge several groups of acquired embryo photos. It will train a standard recognition model and have judgment criteria for the morphology of each node of embryo culture. The so-called standard embryo model refers to the best real embryo state for each period obtained based on morphological analysis, and the AI ​​model is used to extract features and learn from the pictures of these standard embryo models to form a standard model for comparison and evaluation of other embryo morphologies.

[0013] In combination with the first aspect, the present invention provides a first implementation manner of the first aspect, in which the integrity comparison is used as a scoring value, the integrity has several reference characteristics, the reference characteristics include cell number, cell uniformity, fragmentation rate, zona pellucida, cell division rate, and development rate, and a relationship coefficient is set for each reference characteristic. The scoring value is a weighted average of the reference characteristics.

[0014] In combination with the first aspect, the present invention provides a second implementation of the first aspect. During the recognition process, images of the embryos being cultured are taken every 15-30 minutes, and the trained recognition model is used to make judgments and scores.

[0015] In combination with the second implementation of the first aspect, the present invention provides a third implementation of the first aspect, wherein the specific steps of training to obtain the recognition model are as follows:

[0016] Step 100. First, a transparent constant temperature culture device is used to culture several groups of embryos, and a camera assembly is arranged inside or outside the constant temperature culture device to regularly capture an image of each embryo, and the image is pre-processed and stored in a database with a time tag;

[0017] Step 200. Extract and merge features of each embryo image in the database to obtain input features, then use EfficientNet as the convolutional neural network architecture of the backbone network to form an EfficientNet model, input the merged input features into the EfficientNet model to obtain output content;

[0018] Step 300. Then training is carried out, and several targets are predicted through several branches at the output end of the backbone network, including all reference characteristics, and then the task branch definitions are awakened, namely Output1=Linear1, Output2=Linear2, Output3=Linear3, and then the loss function in the model training process is the weighted sum of the losses of each task, and each task loss uses cross entropy loss. After obtaining the model output, the AI ​​recognition system includes a post-processing process to convert the model output into corresponding embryonic morphological features, and then complete the training to obtain the recognition model.

[0019] In combination with the third embodiment of the first aspect, the present invention provides a fourth embodiment of the first aspect, wherein the image preprocessing is a data cleaning process for removing invalid data from the obtained embryo image.

[0020] In combination with the third implementation of the first aspect, the present invention provides a fifth implementation of the first aspect, wherein when the image is subjected to feature extraction and merging, the pixels of the image are first normalized, and the formula is as follows:

[0021] f_image=(x-x_min) / (x_max-x_min)

[0022] Where x is the original pixel value, x_min and x_max are the minimum and maximum pixel values ​​respectively;

[0023] Then convert the image capture time information into the time from the start of embryo culture using the following formula:

[0024] f_time=current_time-start_time

[0025] The image information is then extracted through convolution operations, the time information is extracted through linear layers, and the extracted image and time information are merged to form the input features of the backbone model.

[0026] In combination with the first aspect or several embodiments of the first aspect, the present invention provides a sixth embodiment of the first aspect, wherein the image comprises a microscopic magnified image of the embryo.

[0027] In the second aspect, the present invention also provides a system, which is used as a constant temperature culture device in the above-mentioned method of non-invasively evaluating embryo morphology, including a box body, the box body has an operation panel and a placement chamber, a plurality of mounting seats are arranged in the placement chamber, each mounting seat is connected to a culture box, the embryo is cultured by the culture dish in the culture box, and the image of the embryo in the culture dish is captured by an inverted microlens arranged in the culture box.

[0028] In combination with the second aspect, the present invention provides a first implementation of the second aspect, wherein a gas circulation system is provided in the box, the gas circulation system comprises a gas cylinder, a circulation pump and a pipeline, the pipeline is provided with a first circulation gas nozzle at the installation position, and the bottom of the culture box has a second circulation gas nozzle corresponding to the first circulation gas nozzle;

[0029] The culture box has a culture cavity, the culture dish is placed at a lens of a microlens in the culture cavity, the culture cavity is also provided with an LED light on the top of the culture dish, and the second circulation gas nozzle is connected to the culture cavity to form a gas circulation;

[0030] A heating plate for temperature control is also provided in the culture chamber. Each culture box is connected to the mounting base for power supply and data. The operation panel controls the temperature and gas environment of each culture box.

[0031] In combination with the first embodiment of the second aspect, the present invention provides a second embodiment of the second aspect, wherein the culture box comprises a bottom box and a top cover hinged to each other, the LED lamp is arranged on the inner side of the top cover, and a transparent placement area is provided in the middle of the end surface of the bottom box on one side close to the top cover, the placement area is for the microlens arranged inside the bottom box to receive external light, and the LED lamp provides a light source vertically toward the placement area and passes through the culture dish arranged in the placement area; a camera for assisting the microlens to find the embryo is also provided in the top cover.

[0032] The beneficial effects of the present invention are:

[0033] (1) The present invention summarizes the results of multiple layers of images to more accurately confirm the morphology of the embryo at the current time, and scores it according to the standard embryo morphology at the corresponding time, so that the development of the embryo can be intuitively understood;

[0034] (2) The present invention improves detection efficiency by regularly taking pictures of embryos outside the culture dish using a camera device and identifying them using a trained AI model, while avoiding the risk of changing the culture environment by manually opening the culture dish;

[0035] (3) The present invention provides a plurality of detachable independent culture box devices, so that a plurality of embryos can be cultured and observed independently, and pictures can be automatically collected and uploaded to the system for analysis. The temperature and gas environment can be controlled through a unified box, so as to ensure that each embryo has the same and stable culture environment;

[0036] (4) The present invention also obtains a microscopic image of the embryo by setting an electronically inverted microlens in each culture box, and cooperates with an additional external camera to determine the specific position of the embryo, so as to capture the image by automatically adjusting the objective lens or adjusting the position of the culture dish, thereby facilitating accurate and rapid image capture of the embryo. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1 is a front view of a system for non-invasively assessing embryo morphology in an embodiment of the present invention;

[0038] Figure 2 is a top view of a system for non-invasively assessing embryo morphology in an embodiment of the present invention;

[0039] Figure 3 is an axonometric diagram of a system for non-invasively assessing embryo morphology in an embodiment of the present invention;

[0040] Figure 4 The present invention Figure 3 A is a local enlarged schematic diagram;

[0041] Figure 5 4 is a flow chart of a method for non-invasively evaluating embryo morphology in an embodiment of the present invention.

[0042] In the figure: 1-box, 2-culture box, 3-operation panel, 4-top cover, 5-camera, 6-LED light, 7-bottom box, 8-microscope lens. DETAILED DESCRIPTION

[0043] The present invention will be further explained below in conjunction with the accompanying drawings and specific embodiments.

[0044] In order to make the purpose, technical solution and advantages of the embodiments of the present application clearer, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, rather than all the embodiments. The components of the embodiments of the present application described and shown in the drawings here can be arranged and designed in various different configurations.

[0045] Therefore, the following detailed description of the embodiments of the present application provided in the accompanying drawings is not intended to limit the scope of the present application for which protection is sought, but merely represents selected embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in the field without creative work are within the scope of protection of the present application.

[0046] It should be noted that similar reference numerals and letters denote similar items in the following drawings, and therefore, once an item is defined in one drawing, further definition and explanation thereof is not required in subsequent drawings.

[0047] In the description of this application, it should be noted that if the terms "center", "up", "down", "left", "right", "vertical", "horizontal", "inside", "outside", etc. appear, the orientation or position relationship indicated is based on the orientation or position relationship shown in the drawings, or the orientation or position relationship in which the product of the application is usually placed when in use. It is only for the convenience of describing this application and simplifying the description, and does not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operate in a specific orientation, and therefore cannot be understood as a limitation on this application. In addition, if the terms "first", "second", etc. appear in the description of this application, they are only used to distinguish the description and cannot be understood as indicating or implying relative importance.

[0048] In addition, if the terms "horizontal" or "vertical" appear in the description of this application, it does not mean that the components are required to be absolutely horizontal or suspended, but can be slightly tilted. For example, "horizontal" only means that its direction is more horizontal than "vertical", and does not mean that the structure must be completely horizontal, but can be slightly tilted.

[0049] In the description of this application, it should also be noted that, unless otherwise clearly specified and limited, the terms "set", "install", "connect", and "connect" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be indirectly connected through an intermediate medium, or it can be the internal connection of two components. For ordinary technicians in this field, the specific meanings of the above terms in this application can be understood according to specific circumstances.

[0050] Embodiment 1:

[0051] This embodiment discloses a method for non-invasively evaluating embryo morphology. In the embryo culture of in vitro fertilized eggs, the morphological characteristics are used to judge the developmental status of the embryo throughout its development cycle, thereby providing a complete analysis path for the overall development of the embryo.

[0052] It should be noted that the method of this embodiment relies on the existing in vitro fertilized egg culture equipment, and cultivates several embryos in a fully transparent culture dish in an environment suitable for embryonic development. On the basis of not changing the internal environment of the culture dish, a constant temperature device is set on the outside to keep the temperature constant. At the same time, a gas environment maintenance device is also provided to ensure that the gas environment in the culture equipment or a single culture dish is constant.

[0053] The culture device also includes a microscope, through which the state of each embryo is observed. In this embodiment, an electronic microscope device is used, which can automatically take pictures of the embryos in each culture dish and number them, and send the digital images to the processing module of the culture device for processing and imaging.

[0054] Specifically, the method of this embodiment is as follows:

[0055] First, the standard is determined through the above-mentioned culture equipment and the established culture environment, that is, the AI ​​model is learned according to the standard morphological feature image. The above-mentioned culture equipment can be used to cultivate embryos that can meet the standard model of morphological characteristics at all stages, and then photos are taken through a microscope into the AI ​​model to determine a standard recognition model. In this process, manual adjustments and annotations can be made to obtain the best standard recognition model.

[0056] Then, a number of common embryo culture images are obtained through the above-mentioned culture equipment, and then the images are trained using a standard recognition model, and after each image obtains a recognition result, it is annotated and adjusted to obtain a trained recognition model.

[0057] Finally, the trained recognition model is used in the embryo culture process to score the embryo photos obtained by taking pictures of each embryo at a certain periodic interval, and the scoring is based on the embryo integrity.

[0058] Furthermore, in the integrity comparison as a scoring value, the integrity has several reference characteristics, including cell number, cell uniformity, fragmentation rate, zona pellucida, cell division rate, and development rate. A relationship coefficient is set for each reference characteristic, and the scoring value is a weighted average of the reference characteristics.

[0059] Furthermore, a closed incubator is used as a culture device, each of which is equipped with several culture dishes. An image of each culture dish is then taken through a movable microscope, and the image is then automatically identified and scored using an AI model.

[0060] Specifically, refer to Figure 5 The method of this embodiment is as follows:

[0061] First, several groups of embryos are cultured, and a camera assembly is set inside or outside the constant temperature culture device to regularly capture images of each embryo. The images are preprocessed, and the images are mirror-cropped during the preprocessing process to remove invalid data. The cropped images are saved in a database with time tags.

[0062] Then, the image of each embryo in the database is subjected to feature extraction and merging to obtain input features. When the image is subjected to feature extraction and merging, the pixels of the image are first normalized. The formula is as follows:

[0063] f_image=(x-x_min) / (x_max-x_min)

[0064] Where x is the original pixel value, x_min and x_max are the minimum and maximum pixel values ​​respectively;

[0065] Then convert the image capture time information into the time from the start of embryo culture using the following formula:

[0066] f_time=current_time-start_time

[0067] The image information is then extracted through convolution operations, the time information is extracted through linear layers, and the extracted image and time information are merged to form the input features of the backbone model.

[0068] Then, the convolutional neural network architecture of EfficientNet is used as the backbone network to form the EfficientNet model, and the combined input features are input into the EfficientNet model to obtain the output content;

[0069] Then training is carried out, and several targets are predicted through several branches at the output end of the backbone network, including all reference characteristics, and then the task branch definitions are awakened, namely Output1=Linear1, Output2=Linear2, Output3=Linear3, and then the loss function in the model training process is the weighted sum of the losses of each task, and the loss of each task uses cross entropy loss. After obtaining the model output, the AI ​​recognition system includes a post-processing process to convert the model output into corresponding embryonic morphological features, and then complete the training to obtain the recognition model.

[0070] Furthermore, this embodiment also provides a more efficient and intelligent integrated culture device for the above method, referring to Figure 1 - Figure 2 shows a stand-alone device, as follows:

[0071] The invention comprises an integrated box body 1, which is formed by a shell of a polymer medical material, and has a space inside for arranging corresponding parts. The box body 1 itself is a rectangular structure, and has a large volume placement slot in the main part.

[0072] The device also includes a plurality of independent culture boxes 2, which are placed in the placement slot in an orderly manner. Specifically, a groove is provided at the bottom of the culture box 2, and a mounting seat is provided in the placement slot corresponding to each culture box 2. When setting, the groove of the culture box 2 is aligned with the mounting seat to achieve a snap connection, and the culture box 2 is detachably connected to the box body 1. When the culture box 2 is installed, its top height protrudes a certain part of the placement slot, which is convenient for manual operation to grab and separate it from the box body 1.

[0073] A gas circulation system is arranged in the box 1, and the gas circulation system comprises a gas cylinder, a circulation pump and a pipeline. The pipeline is provided with a first circulation gas nozzle at the installation position, and a second circulation gas nozzle corresponding to the first circulation gas nozzle is arranged in the groove at the bottom of the culture box 2.

[0074] The gas cylinder contains a mixed gas at a certain pressure, which is added according to the gaseous environment required by the embryo, specifically containing 6%±0.5% CO2, 5%±0.5% O2 and 89%±0.5% N2. Different from the existing air composition, the proportion of oxygen is reduced, while the proportion of carbon dioxide and nitrogen is increased.

[0075] For this box 1 device with a gas circulation system, there are two settings to keep the gaseous environment in the culture environment stable. That is, a transparent sealing cover is provided outside the placement groove to cover all the culture boxes 2, so as to form a closed environment at the placement groove of the box 1. In this case, there is no need to provide a gas nozzle structure connected between the culture box 2 and the mounting seat, and it is only necessary to keep the environmental gas in the sealing cover stable.

[0076] Another arrangement is as described above, the placement tank is an open environment, and the culture box 2 has a culture cavity, the culture dish is placed in the culture cavity, and the second circulation gas nozzle is connected to the culture cavity to form a gas circulation in the culture cavity. This method has a higher stability because the environmental volume it needs to protect is smaller.

[0077] The entire gas circulation system maintains a stable gas environment by circulating gas with the environment in each culture chamber. A component detection sensor is installed at the gas pump or any pipeline. Once the gas environment changes, the mixed gas with a certain pressure in the gas cylinder is released and the internal gas is discharged, thereby achieving a stable effect.

[0078] Further, refer to Figure 3 and Figure 4 , this embodiment further limits the culture box 2.

[0079] The culture box 2 includes two rotatably connected parts, namely, a top cover 4 and a bottom box 7. A pressable elastic member is provided on the top cover 4, and the elastic member is always pushed outward by a built-in spring, and the edge of the elastic member is set by the top cover 4 to limit the position, so that it can be pressed inward to a certain extent.

[0080] There is a hook at the lower part of the elastic member, and a slot for inserting the hook is provided at the corresponding position of the bottom box 7. When the hook is inserted into the slot, it will be pushed toward one side of the rotating shaft by force and finally form a locking limit. At this time, the top cover 4 is buckled on the bottom shell and a culture chamber is formed through a circumferentially arranged sealing ring. Pressing the elastic member from the outside can disengage the locking limit between the top cover 4 and the bottom box 7 and flip the top cover 4 upward.

[0081] The top cover 4 has an inwardly recessed area on the inside, in which an LED lamp 6 is provided, and a transparent area is provided at the position of the bottom box 7 corresponding to the LED lamp 6, and the transparent area is the culture dish for placing the embryo in the above content. The culture dish can be a separate structure, which is directly placed in the transparent area, or a sealed cover body is provided to buckle it to form an area for placing the embryo. An inverted microlens 8 is provided inside the bottom box 7 corresponding to the transparent area. The so-called microlens 8 refers to a microscope structure with the lens facing upward, which is built-in with an objective lens, but is not provided with an eyepiece. The object image is projected onto the sensor for imaging through an objective lens with a magnification, which can save space. After the top cover 4 is buckled, the LED lamp 6 can face the microlens 8, and the light passes through the embryo and is poured into the microlens 8, thereby forming a complete observation imaging path.

[0082] It is worth noting that, since the volume of the culture box 2 is relatively small and the size of the embryo remains constant, the required magnification range does not vary much, so the objective lens in the culture box 2 adopts a fixed magnification, generally a magnification of 30x-50x.

[0083] Since the objective lens with fixed magnification is used, the embryo itself may move when it develops in a certain space. In order to see the shape of the embryo clearly in the photo, it occupies at least 30% of the area of ​​the captured image. And in this magnification, once the embryo moves to a certain extent, it is likely that the microlens 8 will not be able to obtain the embryo image in the picture when shooting next time.

[0084] In order to solve this technical problem, this embodiment further provides an auxiliary camera 5 in the top cover 4. The camera 5 itself is not used to capture embryo images, but can capture macro magnified images of a certain magnification through its special focal length.

[0085] In this embodiment, the culture box 2 and the incubator are connected to each other through a gas nozzle to realize gas circulation, and are also provided with signal contacts and conductive contacts to realize power supply and data transmission. An operation panel 3 is provided on one side of the box body 1, and a circuit board with a processing module is provided inside the operation panel 3, and a touch-operated display screen is provided on the operation panel 3.

[0086] Based on the structure of the box body 1 and the culture box 2, each culture box 2 is interactively connected with the box body 1 when inserted into the placement slot of the box body 1, and the box body 1 can obtain real-time data of each culture box 2, including images taken by the microlens 8 and images taken by the camera 5.

[0087] The specific principles are as follows:

[0088] After placing the embryo on the culture dish in the transparent area, close the top cover 4, and the LED light 6 will be controlled by the processing module to light up at a fixed time, and the brightness will be adjusted according to the needs. First, adjust the brightness of the LED light 6 to a low brightness state, and then obtain the picture through the camera 5. When the top cover 4 is closed, the focus center of the camera 5 is directly opposite to the center of the culture dish, and the maximum picture edge obtained is exactly the edge of the entire culture dish. A circle of reflection rings is set around the periphery of the microlens 8 at the bottom of the culture dish. When the LED light 6 is on, a certain amount of light can be irradiated on the reflection ring to form floodlight, brightening the entire space.

[0089] At this time, after the camera 5 obtains the picture, it finds the position of the embryo therein after algorithm processing, and then positions the center of the embryo by forming a coordinate axis or a cross scale in the picture. It should be noted that, in order to adjust, the entire microlens 8 (i.e., the object lens and the sensor) can be connected with the controlled movement (generally realized by the electromagnetic actuation motor in the lens anti-shake and lens automatic zoom technology) between the bottom box 7. Or a displacement mechanism (generally realized by a lead screw or other micro-motor) is provided at the culture dish, which can adjust the planar position relationship between the culture dish where the embryo is located and the microlens 8. Due to the position of the microlens 8 itself being determined, after the position of the embryo is obtained by the camera 5, the position difference of the focusing picture of the microlens 8 can be calculated, so as to fine-tune the microlens 8 or the culture dish to realize relative position compensation, and then open the microlens 8 to obtain the picture and take pictures, and then complete the first photo position calibration process.

[0090] The present invention is not limited to the above optional implementations, and anyone can derive other various forms of products under the enlightenment of the present invention. The above specific implementations should not be understood as limiting the scope of protection of the present invention. The scope of protection of the present invention should be based on the definition in the claims, and the description can be used to interpret the claims.

Claims

1. A method for non-invasively evaluating embryo morphology, using machine learning to compare and photograph embryos at different stages of the culture cycle to obtain and evaluate the embryo status, characterized in that: The specific steps are as follows: First, a number of standard embryos were cultured throughout the entire cycle, and the culture cycle was divided into several stages. Several images that met the embryo morphology standards were taken at each stage, and the images were annotated and input into the AI ​​model for recognition to obtain a standard recognition model for each stage. Then, a number of images of different stages of training embryo culture are directly input into the standard recognition model for training and recognition, and annotation and adjustment are performed based on the judgment results after training to obtain a trained recognition model; The trained recognition model is used to identify the images of embryos in culture, and the images are scored based on the completeness comparison with the state of the standard embryo model at the same stage.

2. A method for non-invasively assessing embryo morphology according to claim 1, characterized in that: In the integrity comparison as a scoring value, the integrity has several reference characteristics, including cell number, cell uniformity, fragmentation rate, zona pellucida, cell division rate, and development rate. A relationship coefficient is set for each reference characteristic, and the scoring value is a weighted average of the reference characteristics.

3. A method for non-invasively assessing embryo morphology according to claim 1, characterized in that: During the recognition process, images of the embryos being cultured are taken every 15-30 minutes, and the trained recognition model is used to make judgments and scores.

4. A method for non-invasively assessing embryo morphology according to claim 2, characterized in that: The specific steps for training and obtaining the recognition model are as follows: Step 100. First, a transparent constant temperature culture device is used to culture several groups of embryos, and a camera assembly is arranged inside or outside the constant temperature culture device to regularly capture an image of each embryo, and the image is pre-processed and stored in a database with a time tag; Step 200. Extract and merge features of each embryo image in the database to obtain input features, then use EfficientNet as the convolutional neural network architecture of the backbone network to form an EfficientNet model, input the merged input features into the EfficientNet model to obtain output content; Step 300. Then training is carried out, and several targets are predicted through several branches at the output end of the backbone network, including all reference characteristics, and then the task branch definitions are awakened, namely Output1=Linear1, Output2=Linear2, Output3=Linear3, and then the loss function in the model training process is the weighted sum of the losses of each task, and each task loss uses cross entropy loss. After obtaining the model output, the AI ​​recognition system includes a post-processing process to convert the model output into corresponding embryonic morphological features, and then complete the training to obtain the recognition model.

5. A method for non-invasively assessing embryo morphology according to claim 4, characterized in that: in, Image preprocessing is a data cleaning process that removes invalid data from the acquired embryo images.

6. A method for non-invasively assessing embryo morphology according to claim 4, characterized in that: in, When extracting and merging features of an image, the pixels of the image are first normalized. The formula is as follows: f_image=(x-x_min) / (x_max-x_min) Where x is the original pixel value, x_min and x_max are the minimum and maximum pixel values ​​respectively; Then convert the image capture time information into the time from the start of embryo culture using the following formula: f_time=current_time-start_time The image information is then extracted through convolution operations, the time information is extracted through linear layers, and the extracted image and time information are merged to form the input features of the backbone model.

7. A method for non-invasively assessing embryo morphology according to any one of claims 1 to 6, characterized in that: The images are microscopic magnifications of embryos.

8. A system, characterized in that: The device is used as a constant temperature culture device in a method for non-invasively evaluating embryo morphology in claim 4 above, comprising a housing (1), the housing (1) having an operation panel (3) and a placement chamber, a plurality of mounting seats being arranged in the placement chamber, each mounting seat being connected to a culture box (2), the embryo being cultured by a culture dish in the culture box (2), and the image of the embryo in the culture dish being captured by an inverted microlens (8) arranged in the culture box (2).

9. A system according to claim 8, characterized in that: The box (1) is provided with a gas circulation system, the gas circulation system comprising a gas cylinder, a circulation pump and a pipeline, the pipeline is provided with a first circulation gas nozzle at the installation position, and the bottom of the culture box (2) is provided with a second circulation gas nozzle corresponding to the first circulation gas nozzle; The culture box (2) has a culture cavity inside, the culture dish is placed at a lens of a microlens (8) in the culture cavity, the culture cavity is also provided with an LED light (6) at the top of the culture dish, and the second circulation gas nozzle is connected to the culture cavity to form a gas circulation; A heating plate for temperature control is also provided in the culture chamber. Each culture box (2) is connected to the mounting base for power supply and data. The operating panel (3) controls the temperature and gas environment of each culture box (2).

10. A system according to claim 9, characterized in that: The culture box (2) comprises a bottom box (7) and a top cover (4) which are hinged to each other. The LED lamp (6) is arranged on the inner side of the top cover (4). A transparent placement area is arranged in the middle of the end surface of one side of the bottom box (7) close to the top cover (4). The placement area is for a microlens (8) arranged inside the bottom box (7) to receive external light. The LED lamp (6) provides a light source which is vertically oriented toward the placement area and passes through a culture dish arranged in the placement area. A camera (5) for assisting the microlens (8) in finding the embryo is also arranged inside the top cover (4).

Images