Method for identifying and predicting position of cell nucleus of oocyte by using computer image

Abstract

The invention belongs to the technical field of nucleus visual identification, and discloses a method for identifying and predicting the nucleus position of an oocyte by using a computer image, which comprises the following steps of: dividing the oocyte into a mature oocyte and an immature oocyte, and acquiring an oocyte image; dividing the oocyte image into a mature oocyte image and an immature oocyte image; carrying out visual imaging operation on the nucleus of the oocyte, and meanwhile, carrying out image annotation on an oocyte photo; performing data amplification on images containing mature oocytes and cell nucleus positions, and dividing the images into a training set and a verification set; a computer vision framework is trained using the data set. According to the method disclosed by the invention, the cell nucleus position of the mammal oocyte is predicted by utilizing a computer visual identification technology, the enucleation accuracy is improved from 53.84% to 92.30%, and the accuracy of judging the cell nucleus position of the oocyte in a somatic cell nucleus transplantation process is improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of cell nucleus visual recognition, and in particular relates to a method for predicting the position of an oocyte nucleus by using computer image recognition. Background Art

[0002] Traditional cell nucleus visual recognition algorithms are all trained and applied based on adherent cells. Under a microscope, the nuclei of adherent cells have distinct features, and the location of the nucleus can be clearly seen after photographing. However, oocytes are suspension culture cells, and the cell diameter is large. Due to the obstruction of the cytoplasm, the characteristics of the nucleus region are not obvious. In addition, during the somatic cell cloning process, the oocyte is in the metaphase of the second meiotic division (MII stage). During this period, there is no nuclear structure in the traditional sense. The nuclear region lacks a cell membrane, the chromatin is condensed into chromosomes, and the nuclear region is mainly composed of a structure formed by the spindle and chromosomes. These features are different from adherent cells. Therefore, traditional cell nucleus visual recognition methods are not suitable for predicting the position of the oocyte nucleus.

[0003] With the advancement of computer vision technology, mainstream computer vision frameworks are now capable of performing tasks such as object detection and posture recognition. In animal somatic cell nuclear transfer, the success rate of enucleation is crucial for cloning efficiency. Current mainstream enucleation methods suffer from numerous drawbacks, including low efficiency, excessive cytoplasm removal, expensive equipment, and reagent and fluorescence toxicity. The location of the oocyte nucleus exhibits difficult-to-detect visual features. Experienced operators can often use these features to roughly determine the nucleus's location. However, mastering these features can be challenging for junior or intermediate researchers.

[0004] Taking photos of these oocytes and using machine training and computer vision to identify and identify these features is beneficial for precisely locating the nucleus. Furthermore, computer vision technology does not require toxic chemicals, fluorescent illumination, or expensive equipment; a home computer or simple microcontroller can achieve nucleus localization. Combined with a head-mounted display, the microscope image can be projected directly in front of the operator, with the nucleus location superimposed on the image, enabling precise, damage-free enucleation.

[0005] Through the above analysis, the problems and defects of the existing technology are as follows:

[0006] (1) Traditional visual identification methods of cell nuclei are not suitable for predicting the position of oocyte nuclei.

[0007] (2) The current mainstream cell enucleation methods have many defects, such as low efficiency, excessive cytoplasm removal, expensive equipment, reagent toxicity, and fluorescence irradiation toxicity.

[0008] (3) Currently, there is no visual recognition model for the recognition of oocyte nuclei. Summary of the Invention

[0009] To overcome the problems existing in the related art, the present invention discloses a method for predicting the position of the oocyte nucleus using computer image recognition. The technical solution is as follows:

[0010] The present invention is achieved by a method for predicting the position of an oocyte nucleus using computer image recognition, comprising the following steps:

[0011] S1, oocytes were observed under an optical microscope, granulosa cells on the cumulus-oocyte complex were peeled off, and oocytes were divided into mature oocytes and immature oocytes according to the presence or absence of polar bodies;

[0012] S2, performs visualization imaging of the oocyte nucleus, takes photos of the oocyte and annotates the image, marks the location of the nucleus, and rotates and crops the image;

[0013] S3, performs data amplification on images containing mature oocytes and cell nucleus positions, performs grayscale transformation, blurring, rotation, cropping, adding noise, and changing resolution on oocyte images with marked cell nucleus positions; and divides the generated images into training and validation sets;

[0014] S4, uses the generated dataset to train a computer vision framework, and after training, deploys the obtained weight file to a personal computer, server, or microcontroller for prediction of cell nuclei during microsurgery.

[0015] In step S1, images of mature oocytes are collected, and the oocytes are rotated to expose the polar bodies when taking pictures.

[0016] In step S2, the oocyte nucleus is visualized and imaged, including:

[0017] Use nucleic acid dyes to stain the oocytes, illuminate the oocytes under a fluorescence microscope, and determine the location of the nucleus based on the location of the fluorescence; or,

[0018] Live imaging of oocytes using the Spindle View Aurora microscope system, where the nucleus appears bright white within the oocyte cytoplasm; or

[0019] The number of cytoplasmic lipid droplets near the polar body is reduced, and the area appears white or the plasma membrane has bulges, which are identified as the location of the cell nucleus and manually marked; or,

[0020] Injecting the egg with a fluorescently labeled anti-DNA antibody. After the anti-DNA antibody binds to the DNA, the DNA position on the chromosome will show fluorescence under fluorescent excitation, which is then used to indicate the position of the cell nucleus; or

[0021] The eggs are injected with fluorescently labeled anti-histone antibodies. After the antibodies bind to the histones, the histones on the chromosomes show fluorescence under fluorescent excitation, which is then used to indicate the location of the cell nucleus.

[0022] The nucleic acid dyes used for staining oocytes are Hoechst 33342, Hoechst 33258, DAPI, PI, acridine orange, EB, and SYBR series dyes. The concentration of the nucleic acid dyes is 1 μg / ml-20 mg / ml, and the staining time is 1-120 min.

[0023] The antibody used to label DNA is one or more of anti-DNA antibody, anti-DNA methylation antibody, anti-DNA demethylation antibody, anti-DNA acetylation antibody, anti-DNA lactylation antibody, and anti-DNA acylation antibody.

[0024] The antibody used to label histones is one or more of anti-histone antibodies, anti-histone methylation antibodies, anti-histone acetylation antibodies, anti-histone lactylation antibodies, and anti-histone acylation antibodies.

[0025] In step S2, the oocytes are photographed and the images are annotated, including:

[0026] After visualizing the cell nucleus, take fluorescence and bright-field photos of the same location; compare the fluorescence photos and use the dataset generation tools Roboflow Annotate, Labelimg, or Labelme to mark the location of the cell nucleus on the bright-field photos.

[0027] In step S3, data amplification is performed on the image containing the mature oocyte and the position of the cell nucleus, including:

[0028] The mature oocytes after data amplification are formed into an oocyte image set.

[0029] Furthermore, data augmentation includes: changing contrast, changing resolution, image rotation, converting images to grayscale, image blurring, image sharpening, adding noise to images, and image segmentation;

[0030] When the image is processed, the cell nucleus position data marked in step S2 is simultaneously changed relative to the image file itself, but remains at the corresponding position of the egg.

[0031] In step S4, the computer vision framework is the Yolo series framework, including Yolo1-Yolo12; the weights used in training include: Yolon.pt, Yolos.pt, Yolom.pt, Yolol.pt and Yolox.pt.

[0032] Furthermore, the image recognition weight file is deployed on the computer, connected to the camera on the microscope, and the prediction program is started.

[0033] Furthermore, the oocyte is operated to expose the polar body; the computer performs real-time prediction on the collected image to determine the position of the cell nucleus, and the predicted polar body is regarded as the position of the cell nucleus where the angle between the polar body and the cell nucleus is <90° with the center of the oocyte cytoplasm as the center of the circle as the effective prediction.

[0034] In combination with all the above technical solutions, the beneficial effects of the present invention are as follows:

[0035] First, the present invention uses a camera to photograph the oocyte and predict its nucleus, making it easier for researchers to locate the nucleus. The oocyte nucleus is then removed by aspiration or other methods, allowing for subsequent somatic cell nuclear transfer. Using computer image prediction technology to predict the oocyte nucleus, the present invention has increased the accuracy of nucleation removal from 45.7% to 88.2% in dogs, resulting in the successful creation of two somatic cell cloned dogs.

[0036] Second, the present invention utilizes computer vision recognition technology to predict the position of the nucleus in mammalian oocytes, improving the accuracy of determining the position of the oocyte nucleus during somatic cell nuclear transfer. Once implemented, this invention will significantly reduce the training time and costs for laboratory personnel in the somatic cell cloning industry, improving the efficiency of enucleation and somatic cell cloning. It is expected to increase the efficiency of somatic cell cloning by 40%-50% in somatic cell cloning of dogs, cats, and pigs. This invention addresses the lack of a cost-effective and safe way to display the position of the nucleus in the field of animal somatic cell cloning. By utilizing computer vision-assisted technology, the function of displaying the position of the nucleus is achieved at a low cost and without toxicity.

[0037] Third, due to the use of virtual reality technology, the experimenter can operate without leaving the optical eyepiece of the microscope. For a long time, somatic cell cloning personnel have been unable to take their eyes off the microscope eyepiece when working. After long-term work, cervical spondylosis often occurs. The virtual reality technology used in the present invention frees the experimenter from the microscope eyepiece, allowing them to work in other postures, greatly reducing the risk of cervical spondylosis. Due to the use of virtual reality technology, the present invention can operate on real images at the software level, including but not limited to adding various auxiliary information (time, computer prediction information, etc.) to the image, image enhancement technology (using algorithms to enhance images under low illumination conditions to achieve microscopic operations under dark light conditions), and image processing technology (changing the color of fluorescent dark field images. For example, when selecting fluorescent cells, the human eye is less sensitive to red fluorescence. Computer technology can be used to convert red fluorescence into sensitive green fluorescence and project it into the operator's eyes). BRIEF DESCRIPTION OF THE DRAWINGS

[0038] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present disclosure and, together with the description, serve to explain the principles of the present disclosure;

[0039] Figure 1 This is a flow chart of a method for predicting the position of an oocyte nucleus using computer image recognition provided by an embodiment of the present invention;

[0040] Figure 2 This is a schematic diagram of a method for predicting the position of an oocyte nucleus using computer image recognition according to an embodiment of the present invention;

[0041] Figure 3 This is a rendering of two somatic cell cloned dogs obtained using computer vision-assisted enucleation technology provided by an embodiment of the present invention;

[0042] Figure 4 This is an image under VR glasses using the computer vision-assisted nucleation technology provided by an embodiment of the present invention, wherein the content framed in the image is the predicted cell nucleus image. DETAILED DESCRIPTION

[0043] To make the above-mentioned objects, features, and advantages of the present invention more readily apparent, specific embodiments of the present invention are described in detail below with reference to the accompanying drawings. The following description sets forth numerous specific details to facilitate a full understanding of the present invention. However, the present invention can be implemented in many other ways than those described herein, and those skilled in the art may make similar modifications without departing from the scope of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0044] The innovation of the present invention is:

[0045] 1. Using computer vision technology to predict the position of the oocyte nucleus: The present invention uses computer vision technology to predict the position of the oocyte nucleus and use it for somatic cell nuclear transplantation operations.

[0046] 2. Training of computer recognition model: The present invention adopts visual recognition algorithm.

[0047] 3. Image Annotation: The present invention uses image annotation tools to annotate the photographed eggs.

[0048] Example 1, as Figure 1 As shown, the method for predicting the position of the oocyte nucleus using computer image recognition provided by an embodiment of the present invention includes the following steps:

[0049] S1, oocytes were observed under an optical microscope, granulosa cells on the cumulus-oocyte complex were peeled off, and oocytes were divided into mature oocytes and immature oocytes according to the presence or absence of polar bodies;

[0050] Mature canine oocytes were collected in vivo using the Hwang Woo-suk method. After removing the granulosa cells, oocytes with polar bodies were identified as mature oocytes. The mature oocytes were placed under a microscope and manipulated with a micromanipulator to clearly expose the polar bodies. The oocytes were then photographed.

[0051] Sources of oocytes include humans, rodents, dogs, cats, pigs, goats, sheep, cattle, buffalo, horses, donkeys, brown bears, black bears, polar bears, musk deer, deer, mink, and foxes.

[0052] S2, performs visualization imaging of the oocyte nucleus, takes photos of the oocyte and annotates the image, marks the location of the nucleus, and rotates and crops the image;

[0053] The cytoplasmic region near the polar body is manually determined to predict the potential location of the nucleus, and then the potential nucleus location is marked on the image. Five methods are used to confirm the location of the nucleus:

[0054] (1) Use the enucleation needle to aspirate the potential location of the cell nucleus. If the aspirated cytoplasm contains a transparent portion, the location is considered to be the location of the cell nucleus. If the aspirated cytoplasm is black, the potential location does not contain a cell nucleus.

[0055] (2) Stain the oocytes with nucleic acid dyes (including but not limited to Hoechst, DAPI, PI, etc.) at a dye concentration of 1-20 μg / ml. Illuminate the oocytes under a fluorescence microscope and determine the location of the cell nucleus based on the location of the fluorescence.

[0056] (3) The oocytes were imaged using the Spindle View system. The Spindle View system is a commercial device that detects the cell nucleus based on the property of light passing through the cell nucleus and the change in polarization. The cell nucleus appears bright white under the Spindle View system, and the position of the cell nucleus is determined based on this.

[0057] (4) Using fluorescently labeled histone antibodies to stain the oocytes (including but not limited to anti-histone antibodies, histone methylation antibodies, histone acylation antibodies, etc., fluorescent labels include but are not limited to fluorescent labels that are directly chemically modified on the primary antibody or fluorescent labels that are fluorescently labeled with secondary antibodies), irradiating the oocytes under a fluorescence microscope, and determining the position of the cell nucleus based on the location of the fluorescence.

[0058] (5) Using fluorescently labeled DNA antibodies to stain the oocytes (including but not limited to anti-DNA antibodies, DNA methylation antibodies, DNA acylation antibodies, DNA demethylation antibodies, etc., fluorescent labels include but are not limited to fluorescent labels that are directly chemically modified on the primary antibody or fluorescent labels that are fluorescently labeled with a secondary antibody), irradiating the oocytes under a fluorescence microscope, and determining the position of the cell nucleus based on the location of the fluorescence.

[0059] While using the five methods above to determine the location of the cell nucleus, simultaneously capture a brightfield image. Import the brightfield image into the image annotation tool. Based on the nucleus locations determined by the three methods above, mark the corresponding locations in the brightfield image for model training. Use the data generation function in the image generation tool to flip the input image, add noise, adjust contrast, rotate, and adjust resolution, etc. to generate more images for training.

[0060] S3, performs data amplification on images containing mature oocytes and cell nucleus positions, performs grayscale transformation, blurring, rotation, cropping, adding noise, and changing resolution on oocyte images with marked cell nucleus positions; and divides the generated images into training and validation sets;

[0061] Visual recognition technology has developed rapidly in recent years. Through training, it can recognize mainstream objects. However, no model has yet been developed to recognize oocyte nuclei.

[0062] The Yolo series of visual recognition algorithms was developed by Ultralytics. Yolo12, an improvement on Yolo11 by the University at Buffalo and the University of the Chinese Academy of Sciences, inherits the efficiency and speed of the Yolo series and is suitable for real-time image recognition prediction. This version adds a regional attention module, centered around the attention mechanism, establishing a simple and efficient Yolo framework, breaking the dominance of CNN models in the Yolo series.

[0063] Import the training and validation sets generated by the image annotation tool into the model and train it using the Yolo framework. Yolo versions include Yolo1-Yolo12. Models used for training include, but are not limited to, Yolon.pt, Yolos.pt, Yolom.pt, Yolol.pt, and Yolox.pt.

[0064] S4, uses the generated dataset to train a computer vision framework, and after training, deploys the obtained weight file to a personal computer, server, or microcontroller for prediction of cell nuclei during microsurgery.

[0065] Deploy the image recognition algorithm on a computer (including but not limited to computers and servers based on x86, arm, and RISC architecture, with an operating system of one of windows, MAC os, and Linux), connect the camera on the microscope, and start the prediction program. The weight file loaded in the prediction program is the weight file trained in the previous step. Preferably, the weight file is the weight file with the highest accuracy and recall rate tested on the validation set.

[0066] Canine oocytes are manipulated to expose polar bodies. The computer performs real-time predictions on the captured images to determine the position of the cell nucleus. Only the predicted position of the cell nucleus near the polar body is considered a valid prediction. The weight file is directly generated by the visual framework. Specifically, during the process of training the model in the visual framework, a corresponding weight file is generated for each iteration. The visual model recognizes images in the validation set based on this weight file and compares it with the pre-manufactured manually labeled data to obtain the precision and recall data for that training iteration. After multiple iterative training, the weight file with the highest precision and recall is selected as the optimal weight file.

[0067] Based on the location of the nucleus marked by computer vision, a micromanipulator extracts the cytoplasm at the corresponding position. This replaces the traditional method of extracting the cytoplasm near the polar body. This achieves precise enucleation and minimizes cytoplasm loss.

[0068] Using computer image prediction technology to predict oocyte nuclei, the accuracy of enucleation in dogs, for example, increased from 45.7% to 88.2%. Using this enucleation technology, two somatic cell cloned dogs were successfully produced.

[0069] This method uses a camera to capture oocytes and predict their nuclei, making it easier for researchers to locate the nucleus. The oocyte nucleus is then removed through aspiration or other methods, allowing for subsequent somatic cell nuclear transfer. Using computer image prediction technology, this method increased the accuracy of oocyte nucleus removal from 53.84% to 92.30% in dogs, resulting in two successful somatic cell cloned dogs, as shown in Table 1.

[0070] Table 1 Comparison of the efficiency of computer vision-assisted enucleation technology for somatic cell cloning of dogs

[0071]

[0072]

[0073] As shown in Table 1, blind aspiration enucleation was used to produce 25 somatic cell nuclear transfer embryos, resulting in one somatic cell cloned dog. Computer vision-assisted enucleation was used to produce 26 somatic cell nuclear transfer embryos, resulting in two somatic cell cloned dogs. In total, blind aspiration enucleation required 25 embryos per cloned dog, while computer vision-assisted enucleation only required 13 embryos per cloned dog.

[0074] Example 2: The system for predicting the location of an oocyte nucleus using computer image recognition provided by the embodiment of the present invention includes:

[0075] An oocyte image acquisition module is used to acquire and pre-process oocyte images, and to divide oocyte images into mature oocyte images and immature oocyte images;

[0076] The oocyte nucleus determination and image annotation module is used to visualize the oocyte nucleus and take photos and annotate the images of the oocyte at the same time;

[0077] A data amplification module is used to amplify data on images containing mature oocytes and cell nucleus positions, and to divide the generated images into a training set and a validation set;

[0078] The oocyte nucleus prediction module is used to train the computer vision framework using the dataset generated by the data amplification module. After training, the optimal weight file is deployed to a personal computer, server or microcontroller for prediction of the oocyte nucleus during micromanipulation.

[0079] Example 3: The Spindle View system can be used to accurately image cell nuclei without using toxic chemicals, hyperosmotic pressure, or fluorescent illumination.

[0080] Example 4. An embodiment of the present invention provides a real-time positioning device for the nucleus of an oocyte, comprising a processor, wherein the processor is used to execute a computer program, a weight file and a model stored in a memory to realize real-time positioning of the nucleus of an oocyte in an oocyte photo, a video or a camera streaming video.

[0081] Example 5. An embodiment of the present invention provides a head-mounted display device for projecting a microscope image into the operator's eyes in real time and directly displaying the predicted cell nucleus position on the image.

[0082] Hormone testing was performed on dogs in natural estrus. The progesterone content in the serum of dogs in natural estrus was determined by electrochemiluminescence. Ovulation was determined when the content rose to 6-11 ng / ml. The day of ovulation was recorded as day 0. On days 2-5, the dogs underwent abdominal surgery, and canine cumulus oocyte complexes were obtained by fallopian tube flushing.

[0083] The obtained cumulus oocyte complex was placed in an operating solution, and the granulosa cells on the surface were peeled off using a mouth pipette. The canine oocytes were divided into mature, aged, and immature according to the presence or absence of polar bodies. In the present invention, aged and mature oocytes were both treated as mature oocytes.

[0084] Mature eggs are placed in a manipulation solution containing cytochalasin B. The entire manipulation solution droplet and culture dish are placed on an inverted microscope equipped with a camera for real-time video capture and transmission to a computer. Pre-trained weight files are used for real-time nucleus prediction. A micromanipulator is used to flip the egg to expose the polar body.

[0085] Wearing a virtual reality headset, the experimenter observed the field of view under the microscope in real time and flipped the eggs. Once the polar body was exposed and a visually identifiable prediction frame was identified, the cellular components at the predicted frame were extracted based on the position of the predicted frame to achieve enucleation. The blind aspiration method, which does not use computer vision-assisted technology to perform enucleation, directly extracts the cytoplasm near the polar body. AI-assisted enucleation uses computer vision-assisted technology to predict the location of the cell nucleus and extract the corresponding location. Thirteen eggs were enucleated using each of the two techniques. After enucleation, the efficiency of cell nucleus removal was assessed by staining with Hoechst 33342. The results are shown in Table 2.

[0086] In addition, 25 and 26 somatic cell nuclear transfer embryos were generated and transferred using the blind aspiration method and AI-assisted enucleation techniques, respectively. Ultimately, one somatic cell cloned dog was obtained using the blind aspiration method, requiring an average of 25 embryos per cloned dog. Two somatic cell cloned dogs were obtained using the AI-assisted enucleation technique, requiring an average of 13 embryos per cloned dog (Table 1).

[0087] Skin fibroblasts were injected into enucleated oocytes, and somatic cell nuclear transfer embryos were obtained by electrofusion. The somatic cell nuclear transfer embryos were then transplanted into estrus bitches.

[0088] After 62 days of gestation, three somatic cell cloned dogs were finally obtained, one of which underwent enucleation without the use of computer vision prediction technology, and two dogs underwent enucleation using computer enucleation technology.

[0089] The present invention used this system to enucleate canine oocytes. After enucleation, the cytoplasm was stained to determine the enucleation efficiency. The results in Table 1 show that computer vision-assisted enucleation achieved an enucleation rate of 92.30%, while blind aspiration achieved only 53.84%.

[0090] Table 2 Oocyte enucleation rates by different methods

[0091] method Denucleation rate (%) Blind suction 7 / 13(53.84) Computer Vision 12 / 13(92.30)

[0092] The above description is only a preferred specific implementation method of the present invention, but the scope of protection of the present invention is not limited thereto. Any modifications, equivalent substitutions and improvements made by any technician familiar with this technical field within the technical scope disclosed by the present invention and within the spirit and principles of the present invention should be covered by the scope of protection of the present invention.

Claims

1. A method for predicting the position of an oocyte nucleus using computer image recognition, characterized in that: The method comprises the following steps: S1, oocytes were observed under an optical microscope, granulosa cells on the cumulus-oocyte complex were peeled off, and oocytes were divided into mature oocytes and immature oocytes according to the presence or absence of polar bodies; S2, perform visualization imaging of the oocyte nucleus, take photos of the oocyte, annotate the images, and mark the location of the nucleus; S3, performs data amplification on images containing mature oocytes and cell nucleus positions, performs grayscale transformation, blurring, rotation, cropping, adding noise, and changing resolution on oocyte images with marked cell nucleus positions; and divides the generated images into training and validation sets; S4, uses the generated dataset to train a computer vision framework, and after training, deploys the obtained weight file to a personal computer, server, or microcontroller for prediction of cell nuclei during microsurgery.

2. The method for predicting the position of the oocyte nucleus by computer image recognition according to claim 1, wherein: In step S1, images of mature oocytes are collected, and the oocytes are rotated to expose the polar bodies when taking pictures.

3. The method for predicting the position of the oocyte nucleus by computer image recognition according to claim 1, wherein: In step S2, the oocyte nucleus is visualized and imaged, including: Use nucleic acid dyes to stain the oocytes, illuminate the oocytes under a fluorescence microscope, and determine the location of the nucleus based on the location of the fluorescence; or, Live imaging of oocytes using the Spindle View Aurora microscope system, where the nucleus appears bright white within the oocyte cytoplasm; or The number of cytoplasmic lipid droplets near the polar body is reduced, and the area appears white or the plasma membrane has bulges, which are identified as the location of the cell nucleus and manually marked; or, Injecting the egg with a fluorescently labeled anti-DNA antibody. After the anti-DNA antibody binds to the DNA, the DNA position on the chromosome will show fluorescence under fluorescent excitation, which is used to indicate the position of the cell nucleus; or The eggs are injected with fluorescently labeled anti-histone antibodies. After the antibodies bind to the histones, the histones on the chromosomes show fluorescence under fluorescent excitation, which is used to indicate the location of the cell nucleus.

4. The method for predicting the position of the oocyte nucleus by computer image recognition according to claim 3, wherein: The nucleic acid dyes used for staining oocytes are Hoechst 33342, Hoechst 33258, DAPI, PI, acridine orange, EB, and SYBR. The concentration of nucleic acid dyes is 1 μg / ml-20 mg / ml, and the staining time is 1-120 min.

5. The method for predicting the position of the oocyte nucleus by computer image recognition according to claim 1, wherein: In step S2, the oocytes are photographed and the images are annotated, including: After visualizing the cell nucleus, take fluorescence and bright-field photos of the same location; compare the fluorescence photos and use the dataset generation tools Roboflow Annotate, Labelimg, or Labelme to mark the location of the cell nucleus on the bright-field photos.

6. The method for predicting the position of the oocyte nucleus by computer image recognition according to claim 1, wherein: In step S3, data amplification is performed on the image containing the mature oocyte and the position of the cell nucleus, including: The mature oocytes after data amplification are formed into an oocyte image set.

7. The method for predicting the position of the oocyte nucleus by computer image recognition according to claim 6, wherein: Data augmentation includes: changing contrast, changing resolution, image rotation, converting images to grayscale, image blurring, image sharpening, adding noise to images, and image segmentation; When the image is processed, the cell nucleus position data marked in step S2 is simultaneously changed relative to the image file itself, but remains at the corresponding position of the egg.

8. The method for predicting the position of the oocyte nucleus by computer image recognition according to claim 1, wherein: In step S4, the computer vision framework is the Yolo series framework, including Yolo1-Yolo12; the weights used in training include: Yolon.pt, Yolos.pt, Yolom.pt, Yolol.pt and Yolox.pt.

9. The method for predicting the position of the oocyte nucleus by computer image recognition according to claim 7, wherein: Deploy the image recognition weight file on the computer, connect the camera on the microscope, and start the prediction program.

10. The method for predicting the position of the oocyte nucleus by using computer image recognition according to claim 9, characterized in that: The oocyte is operated to expose the polar body; the computer performs real-time prediction on the collected image to determine the position of the cell nucleus. The predicted polar body is centered at the center of the oocyte cytoplasm, and the position of the cell nucleus where the angle between the polar body and the cell nucleus is <90° is considered a valid prediction.

Images