Method and system for predicting fertilized egg or embryo abnormalities through sperm movement analysis

Abstract

The present invention relates to a method and system for predicting fertilized egg or embryo abnormalities through sperm movement analysis. Through the present invention, idiopathic male infertility or abnormalities in sperm-derived fertilized eggs or embryos can be effectively predicted by obtaining sperm movement information from a semen sample of a subject through artificial intelligence model-based analysis and then utilizing the information as abnormality prediction information.

Description

Method and system for predicting abnormalities in fertilized eggs or embryos through sperm motility analysis

[0001] The present invention relates to a method and system for predicting abnormalities in a fertilized egg or embryo through sperm movement analysis, and more particularly, to a method and system for analyzing sperm movement using an artificial intelligence model on a sperm movement image obtained from a semen sample of a subject, and predicting abnormalities in a fertilized egg or embryo fertilized using the subject's semen based on the analysis results.

[0002] Assisted reproductive technologies (ART) are medical techniques that artificially regulate the reproductive process to induce pregnancy in infertile patients, including intrauterine insemination (IUI), in vitro fertilization (IVF), and intracytoplasmic sperm injection (ICSI). Male infertility accounts for approximately 35-40% of all infertility cases, and is mainly diagnosed as oligospermia, aspermia, teratospermia, and azoospermia based on indicators such as sperm count, motility, and morphology.

[0003] These diagnoses are performed based on Computer-Assisted Sperm Analysis (CASA), which can calculate sperm concentration and motility, as well as various kinematic indices such as crossover frequency, curvilinear velocity, linear velocity, mean path velocity, linearity, straightness, and temporal amplitude.

[0004] However, despite these traditional diagnostic criteria, some patients still fail to conceive despite having all sperm parameters within the normal range. This condition is generally defined as "unexplained male infertility," and its incidence has been increasing recently. Unexplained male infertility complicates the decision-making process for ART and can lead to repeated pregnancy failures, increased costs, and psychological distress for patients.

[0005] Meanwhile, aneuploidy, identified through pre-embryo transfer genetic testing (PGT-A) during ART, is known to be a major cause of pregnancy failure and miscarriage. Aneuploidy can be caused by errors in meiosis or early mitotic cell division after fertilization. Recently, it has been reported that sperm abnormalities can contribute to the development of aneuploidy.

[0006] In particular, sperm-derived aneuploidy can cause problems such as paternal genome clustering failure, chromosome segregation errors, and micronuclei formation due to abnormalities in centrosomes, dynein, and microtubule proteins, which negatively impact normal embryo division and implantation. Furthermore, the possibility that key molecules involved in sperm motility and the mechanisms involved in chromosome segregation may be functionally linked has been raised.

[0007] However, the current CASA analysis technology alone cannot sufficiently quantitatively or functionally explain the possibility of sperm-based aneuploidy, and therefore, the development of a new diagnostic tool that can precisely elucidate the correlation between sperm motility patterns and aneuploidy formation is required.

[0008] Recently, AI-based reproductive medicine diagnostic technologies have been actively developed, and clinical applications already exist in the image-based quality assessment of oocytes and embryos. AI, in particular, possesses strengths in learning and classifying complex movement patterns, making it an effective technology for predicting the likelihood of aneuploidy by analyzing subtle changes in sperm motility patterns. However, to date, no AI-based diagnostic technology has been developed that predicts the likelihood of aneuploidy based on sperm motility patterns.

[0009] Accordingly, there is a growing need for new diagnostic algorithms and analysis systems that can identify the causes of unexplained male infertility and improve the success rate of ART. In particular, predicting the likelihood of aneuploidy through quantitative analysis of sperm motility patterns is emerging as a practical technical challenge that can help patients determine treatment strategies and achieve cost and time efficiency.

[0010] The present inventors have confirmed that when sperm movement information obtained by performing a series of artificial intelligence model-based analyses on sperm movement images in a semen sample is used, abnormalities in fertilized eggs or embryos using the semen sample can be effectively predicted.

[0011]

[0012] Accordingly, the present invention provides an abnormality prediction method for predicting abnormalities in a fertilized egg or embryo using the subject's semen.

[0013] In addition, the present invention provides an abnormality prediction system that predicts abnormalities in fertilized eggs or embryos using the subject's semen.

[0014] In addition, the present invention provides a computer program stored in a storage medium that predicts an abnormality in a fertilized egg or embryo using the subject's semen.

[0015] In addition, the present invention provides a computing device for predicting abnormalities in a fertilized egg or embryo using the subject's semen.

[0016] The present invention relates to a method and system for predicting abnormalities in a fertilized egg or embryo through sperm movement analysis. Through the present invention, sperm movement information is obtained from a semen sample of a subject through an artificial intelligence model-based analysis, and then the information is used as abnormality prediction information, thereby effectively predicting abnormalities.

[0017] Hereinafter, the present invention will be described in more detail.

[0018]

[0019] One aspect of the present invention relates to a method for predicting abnormalities in a fertilized egg or embryo, comprising: an image acquisition step of acquiring a sperm movement image from a semen sample of a subject; a sperm movement analysis step of analyzing sperm movement in the sperm movement image to obtain sperm movement pattern information and speed ratio information for each movement pattern; and an abnormality prediction step of predicting abnormalities in a fertilized egg or embryo fertilized using the semen of a subject based on abnormality prediction information including sperm movement pattern information and speed ratio information for each movement pattern.

[0020] The term "subject" in this specification means a paternal individual that is a sexually reproducing animal and has haploid (n) sperm as its germ cells, and sperm motility analysis can be performed on a semen sample obtained from a paternal individual of a mammal, bird, fish or reptile, and mammals may include primates such as chimpanzees, orangutans, gorillas, monkeys including humans, or livestock such as cows, pigs, horses, dogs, cats, rabbits, rats, etc., birds may include chickens, ducks, turkeys, pigeons, parrots, eagles, hawks, etc., fish may include trout, carp, salmon, tuna, goldfish, flounder, sea bass, etc., and reptiles may include iguanas, turtles, snakes, lizards, frogs, chameleons, crocodiles, etc., but are not limited thereto.

[0021] The term "abnormality" in this specification means that some or all of the cells of a specific individual exist in a state that is genetically, chromosomally, morphologically or developmentally different from the normal state. A genetic abnormality in which a mutation occurs in one or more genes in the genome, a chromosomal abnormality in which chromosomes exist aneuploidically or polyploidically or in which some or all of the chromosomes are deleted, duplicated, inverted or translocated, a morphological abnormality in which morphological abnormalities are observed in some or all of the cells, and a developmental abnormality in which some or all of the cells appear abnormal during division or development may all be referred to as "abnormalities." In the present invention, "fertilized egg or embryo abnormality" may be "fertilized egg or embryo aneuploidy."

[0022] The term "aneuploidy" in this specification refers to a state in which some or all of the cells of a specific individual have an abnormality in the number of chromosomes that deviates from the normal number of chromosomes for the species, and is a concept contrasted with "euploidy" which is a state in which there is a normal number of chromosomes. In the field of in vitro fertilization (IVF), testing for aneuploidy in a fertilized egg or embryo for implantation in the maternal individual is an essential element in determining the success of pregnancy and the birth of a normal individual. A fertilized egg or embryo with aneuploidy has an extremely high probability of implantation failure, or even if implantation is successful, it may lead to developmental delay in the embryo, miscarriage, or congenital genetic disease in the newborn.

[0023] In the present invention, sperm movement images can be obtained by photographing the movement of sperm in a semen sample of a subject using an imaging device, and can be obtained in the form of a single image, multiple images, continuous images, or dynamic images. In this case, sperm movement images can be obtained in the form of continuous or dynamic images that can identify sperm movement in a time-series manner to pattern sperm movement, and the obtained images can also be referred to as "image data" or "visual data."

[0024] In one embodiment of the present invention, the sperm movement analysis step may include a sperm tracking step of tracking sperm in a sperm movement image; and a sperm movement classification step of generating sperm movement pattern information by classifying sperm movement in the sperm movement image into one of a plurality of movement patterns.

[0025] In one embodiment of the present invention, the sperm tracking step may include an individual sperm identification step of identifying individual sperm within a sperm movement image; and an identified sperm tracking step of tracking sperm identified in the individual sperm identification step.

[0026] In one embodiment of the present invention, the sperm movement classification step may be to generate sperm movement pattern information by classifying sperm movement into one of two to eight types of movement patterns.

[0027] In one embodiment of the present invention, the sperm movement classification step may be to classify the sperm movement into one movement pattern selected from the group consisting of progressive high-speed linear movement, curved linear movement, progressive curved movement, and circular movement.

[0028] In the present invention, "progressive high-speed linear motion" of sperm refers to a form of motion in which sperm continuously advance in a constant direction along a relatively straight trajectory at high speed, characterized by high linearity and minimal velocity variation. This can be considered an ideal form of motion that allows for efficient access to the egg.

[0029] In the present invention, the "curved linear movement" of sperm refers to a movement in which sperm exhibit a relatively constant directionality while intermittently showing a curved movement trajectory, so that the directionality and motility of sperm can be maintained at a certain level.

[0030] In the present invention, the "progressive curved motion" of sperm refers to a type of motion in which sperm do not have a clear directionality and show a curved, irregular motion trajectory and tend to vibrate greatly to the left and right in the direction of progression. Sperms that perform this type of motion have low efficiency in approaching the egg, but can still be classified as motile sperm.

[0031] In the present invention, the "circular movement" of sperm refers to a form of movement in which sperm repeatedly move along a circular or rotational trajectory within a local area having a displacement below a certain level, which indicates a state in which both sperm motility and directionality are weak, and may be associated with asymmetry of sperm tail movement or structural abnormality of sperm.

[0032] In one embodiment of the present invention, the abnormal prediction information may further include age information of the maternal individual providing the egg or clinical information of the sperm.

[0033] In one embodiment of the present invention, the sperm clinical information may include one or more pieces of information selected from the group consisting of the percentage of progressively motile sperm in semen, the average velocity of motile sperm, the amount of semen, and the subject's age.

[0034] In the present invention, the “proportion of progressively motile sperm in semen” may be a value obtained by dividing the number of sperms whose displacement between the initial position and the final position is greater than a certain level as a result of sperm movement image analysis by the total number of sperms identified in the sperm movement image.

[0035] In the present invention, the “average velocity of motile sperm” may be a value obtained by measuring the average velocity of sperm showing a certain level of motility or higher as a result of sperm movement image analysis.

[0036] In one embodiment of the present invention, the sperm movement analysis step or the abnormality prediction step may be performed through an artificial intelligence model.

[0037] The term "artificial intelligence model" in this specification refers to an information processing algorithm that recognizes a pattern or rule through a series of operations on one or more input values ​​and derives an output value based on the pattern or rule, and may be a non-neural network-based model based on rules, statistics, trees, or similarity, or a neural network-based model based on an artificial neural network (ANN) structure, and may be, for example, a neural network model based on machine learning (ML) or deep learning (DL).

[0038] In the present invention, when the artificial intelligence model is a neural network model based on an artificial neural network structure, the artificial intelligence model may be a neural network model composed of a basic neural network model such as a perceptron or MLP, a convolutional neural network (CNN) model, a recurrent neural network (RNN) model, a spatiotemporal neural network model combined with time series information, and a modification or combination of these models, and when performing a detailed step included in the method for predicting abnormalities in a fertilized egg or embryo of the present invention, an artificial intelligence model in a form most appropriate for the processing data and purpose of the detailed step may be used.

[0039] A neural network model includes an input layer and an output layer, and may include intermediate or hidden layers between the input layer and the output layer depending on the structure or depth of the neural network model. Each layer consists of one or more nodes that constitute each layer and one or more edges that connect the nodes, and complex nonlinear relationships can be modeled through the computational process from the input layer to the output layer. If there are multiple hidden layers between the input layer and the output layer, that is, if the hidden layers are deep, the neural network can be referred to as a "deep neural network." Such a neural network can be referred to as a "network" or a "neural network," and a deep neural network can be referred to as a "deep neural network," which can be used interchangeably.

[0040] Deep neural networks can be classified into, but are not limited to, convolutional neural networks (CNNs), recurrent neural networks (RNNs), autoencoders, generative adversarial networks (GANs), restricted Boltzmann machines (RBMs), deep belief networks (DBNs), Q-networks, U-networks, or Siamese networks, depending on the algorithm of the computational process from the input layer to the output layer.

[0041] In a neural network model, each operation according to the algorithm can be optimized so that the neural network can produce the optimal result or make a judgment under the algorithm that configures the neural network for the result or judgment that the neural network must produce. This process can be referred to as the "learning" of the neural network model. For example, learning can be done in the form of continuously verifying and modifying the weights (W) and biases (b) of the edges connecting the nodes that make up each layer of the neural network, but is not limited to this. The learning of a neural network model can be done in the form of supervised learning where the correct answers are labeled in the training data, unsupervised learning where the correct answers are not labeled in the training data, semi-supervised learning where the correct answers are labeled in some of the training data, reinforcement learning where rules or patterns are specified toward the maximum reward based on data without correct answers, or a hybrid method that combines each learning method.

[0042] In one embodiment of the present invention, the fertilized egg or embryo abnormality may be aneuploidy of the fertilized egg or embryo.

[0043] Another aspect of the present invention relates to a system for predicting abnormalities in a fertilized egg or embryo, comprising: an image acquisition unit for acquiring an image of sperm movement from a semen sample of a subject; a sperm movement analysis unit for analyzing sperm movement in the sperm movement image to obtain sperm movement pattern information and speed ratio information for each movement pattern; and an abnormality prediction unit for predicting abnormalities in a fertilized egg or embryo fertilized using the semen of a subject based on abnormality prediction information including sperm movement pattern information and speed ratio information for each movement pattern.

[0044] In one embodiment of the present invention, the system may further include a storage unit capable of storing any form of information generated or determined by the image acquisition unit, sperm movement analysis unit, or abnormality prediction unit.

[0045] Another aspect of the present invention relates to a computer program stored in a storage medium, which causes a processor to perform an operation for prediction when a file of a fertilized egg or embryo fertilized using the subject's semen is sent, the operation including: an operation of analyzing sperm movement in a sperm movement image to obtain sperm movement pattern information and speed ratio information for each movement pattern; and an operation of predicting an abnormality of a fertilized egg or embryo fertilized using the subject's semen based on abnormality prediction information including sperm movement pattern information and speed ratio information for each movement pattern.

[0046] In the present invention, a computer program may be composed of a series of codes or algorithms, and may independently or in combination command a program processing device, such as a computing device, or be interpreted by the processing device so that the device operates according to the configuration of the computer program. The computer program may be embodied through any type of machine, component, physical device, computer, or storage medium to command the processing device or be interpreted by the processing device. In addition, the computer program may be stored or executed in a distributed manner through a medium without a physical form, such as a network-connected computer system or a cloud server.

[0047] A storage medium may be used to permanently store a computer program or to temporarily store it for execution or download. A storage medium may be a variety of recording or storage means in the form of a single piece of hardware or a combination of several pieces of hardware, and is not limited to media directly connected to a computer system, but may also exist distributed over a network. For example, a storage medium may be configured to store a computer program, including magnetic media such as hard disks, floppy disks, and magnetic tapes, optical recording media such as CD-ROMs and DVDs, magneto-optical media such as floptical disks, and ROM, RAM, flash memory, etc. These storage media may be referred to as "memory" in terms of a device that constitutes a computing device for implementing an algorithm, or as "storage" in terms of the function of a module within a system for implementing an algorithm.

[0048] Another aspect of the present invention relates to a computing device for predicting an abnormality in a fertilized egg or embryo fertilized using semen of a subject, comprising: a processor including one or more cores; wherein the processor analyzes sperm movement in a sperm movement image to obtain sperm movement pattern information and speed ratio information for each movement pattern, and predicts an abnormality in a fertilized egg or embryo fertilized using semen of a subject based on abnormality prediction information including the sperm movement pattern information and speed ratio information for each movement pattern.

[0049] In one embodiment of the present invention, the computing device may further include a memory capable of storing any form of information generated or determined by the processor.

[0050] The present invention relates to a method and system for predicting abnormalities in a fertilized egg or embryo through sperm movement analysis, wherein image data of a semen sample collected for diagnosing infertility of a subject is acquired and analyzed, and the movement pattern of sperm is classified and then used as key information for predicting abnormalities in a fertilized egg or embryo, thereby effectively predicting male infertility of unknown cause or abnormalities in a fertilized egg or embryo derived from sperm.

[0051] FIG. 1 is a schematic diagram illustrating a system for predicting abnormalities in a fertilized egg or embryo according to one embodiment of the present invention.

[0052] Figure 2 is a flowchart illustrating a method for predicting abnormalities in a fertilized egg or embryo according to one embodiment of the present invention.

[0053] FIG. 3 is a drawing showing in more detail a series of processes from sperm movement images to prediction of abnormalities in a fertilized egg or embryo according to one embodiment of the present invention.

[0054] FIG. 4 is a graph showing a receiver operating characteristic (ROC) curve of a binary classification model for pregnancy or non-pregnancy constructed according to one embodiment of the present invention.

[0055] FIG. 5 is a diagram showing the pregnancy prediction performance of a pregnancy or non-pregnancy binary classification model constructed according to one embodiment of the present invention in the form of a confusion matrix.

[0056] An image acquisition step for acquiring an image of sperm movement from a semen sample of a subject;

[0057] A sperm movement analysis step for analyzing sperm movement in the above sperm movement image to obtain sperm movement pattern information and speed ratio information for each movement pattern; and

[0058] An abnormality prediction step for predicting an abnormality in a fertilized egg or embryo using the subject's semen based on abnormality prediction information including the above sperm movement pattern information and speed ratio information for each movement pattern;

[0059] A method for predicting abnormalities in a fertilized egg or embryo, including:

[0060] Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings. The detailed description set forth below, together with the accompanying drawings, is intended to explain exemplary embodiments of the present invention and is not intended to represent the only embodiments in which the present invention may be practiced, and the scope of the present invention is not limited by these examples.

[0061] In some cases, to avoid ambiguity in the concepts of the present invention, well-known structures and devices may be omitted or illustrated in block diagram form focusing on the core functions of each structure and device. Furthermore, the same components are described using the same reference numerals throughout this specification.

[0062] Throughout the specification, whenever a part is said to "include" a component, this does not mean that it excludes other components, but rather that it may include other components, unless otherwise stated.

[0063] Additionally, throughout the specification, when we say that a configuration is "connected" to another configuration, this includes not only the case where it is "directly connected" but also the case where it is "connected with another configuration in between."

[0064] The term "or" is intended to mean an inclusive "or" rather than an exclusive "or." That is, where the connection or use between constructions is not otherwise specified or clear from the context, i.e., if X utilizes A; X utilizes B; or X utilizes both A and B, "X utilizes A or B" can apply to any of these cases, and if X is connected to A; X is connected to B; or X is connected to both A and B, "X is connected to A or B" can apply to any of these cases.

[0065] Additionally, the term "unit" described in the specification refers to a unit that processes at least one function or operation, which may be implemented by hardware, software, or a combination of hardware and software. Furthermore, the terms "one," "one," and similar related terms may be used in the context of describing the present invention to encompass both singular and plural meanings, unless otherwise indicated in the specification or clearly contradicted by the context.

[0066] It should be appreciated that the various illustrative logical blocks, configurations, modules, circuits, means, logics, and algorithm steps described in connection with the embodiments herein may be implemented as electronic hardware, computer software, or combinations of both. To clearly illustrate the interchangeability of hardware and software, various illustrative components, blocks, configurations, means, logics, modules, circuits, and steps have been described above solely in terms of their functionality. Whether such functionality is implemented as hardware or software depends upon the particular application and design constraints imposed on the overall system. Skilled artisans may implement the described functionality in varying ways for each particular application. However, such implementation decisions should not be interpreted as causing a departure from the scope of the present disclosure.

[0067]

[0068] FIG. 1 is a schematic diagram illustrating a system for predicting abnormalities in a fertilized egg or embryo according to one embodiment of the present invention, and FIG. 2 is a flowchart illustrating a method for predicting abnormalities in a fertilized egg or embryo according to one embodiment of the present invention.

[0069] Referring to FIGS. 1 and 2, the abnormality prediction system (100) of the present invention may include an image acquisition unit (10), a sperm movement analysis unit (20), and an abnormality prediction unit (30), and the abnormality prediction method (S100) of the present invention may include an image acquisition step (S10), a sperm movement analysis step (S20), and an abnormality prediction step (S30).

[0070] In the image acquisition unit (10), an image acquisition step (S10) can be performed. Image acquisition can be performed by magnifying and observing the sample at a magnification that can identify and track the shape of sperm in the semen sample of the subject. In the image acquisition step (S10), the image acquisition equipment that can be used to acquire the sperm movement image can be composed of a magnification equipment for magnifying the image and an imaging equipment that can capture the magnified image. As the magnification equipment, an optical microscope, a phase contrast microscope, a fluorescence microscope, a differential interference microscope, etc. can be used, and as the imaging equipment, a digital microscope camera, a high-speed camera, a microscope built-in camera, etc. can be used, but are not limited thereto.

[0071] In the sperm movement analysis unit (20), a sperm movement analysis step (S20) can be performed. In the sperm movement analysis step (S20), sperm movement pattern information and speed ratio information for each movement pattern can be generated through sperm movement analysis in the sperm movement image acquired in the image acquisition step (S10). At this time, the sperm movement pattern information can be generated by classifying the sperm movement pattern by comprehensively considering kinematic factors such as sperm speed, directionality, displacement, and straightness, and for example, the sperm movement patterns in a semen sample can be classified into 2 types, 3 types, 4 types, 5 types, 6 types, 7 types, or 8 types. The speed ratio information for each movement pattern can be generated by calculating the speed ratio of an individual sperm or the average speed ratio of sperm in a sample based on a predetermined reference speed set for each movement pattern.

[0072] The sperm movement analysis step (S20) may include a sperm tracking step (S21) and a sperm movement classification step (S22) in detail. In the sperm tracking step (S21), individual tracking of multiple sperms in a sperm movement image for a semen sample may be performed based on the sperm movement image, and in the sperm movement classification step (S22), sperm movement pattern classification may be performed into one of movement patterns classified into 2 types, 3 types, 4 types, 5 types, 6 types, 7 types, or 8 types based on the movement method of the tracked individual sperms. At this time, the sperm tracking step (S21) may include, in more detail, an individual sperm identification step (not shown) for identifying individual sperms in the sperm movement image, and an identified sperm tracking step (not shown) for tracking the identified individual sperms over a predetermined time or image frame.

[0073] In the sperm movement analysis unit (20), a series of algorithms for performing the sperm movement analysis step (S20) can be performed, and the execution of the algorithms can be performed through an artificial intelligence model.

[0074] The artificial intelligence model for performing sperm motility analysis may be a machine learning-based artificial neural network model, or more specifically, a deep learning-based artificial neural network model. Considering that sperm motility analysis is performed based on sperm motility images acquired from semen samples, the deep learning-based artificial neural network model may be comprised of, or has as its backbone, a convolutional neural network (CNN), but is not limited thereto.

[0075] More specifically, the artificial intelligence model (sperm movement analysis model) for performing the sperm movement analysis step (S20) may include an artificial intelligence model (sperm movement tracking model) for performing the sperm tracking step (S21) and an artificial intelligence model (sperm movement classification model) for performing the sperm movement classification step (S22).

[0076] The sperm tracking model may include an artificial intelligence model (individual sperm identification model) for performing an individual sperm identification step and an artificial intelligence model (identification sperm tracking model) for performing an identification sperm tracking step. The individual sperm identification model may be an object detection model that can detect a class or tag an identification code (ID) for each sperm while identifying the location of an individual sperm within each frame of a sperm movement video, and may be, for example, a YOLO series model including YOLOv5, YOLOv6, YOLOv7, YOLOv8, YOLOv9, YOLOv10 or YOLO-NAS, an SSD model, an EfficientDet model, a DETR model, a DINO model, or a composite, modified or improved model thereof, but is not limited thereto. The sperm tracking model may be an object tracking model that can track identified sperm by assigning an ID to each sperm identified in each frame of a sperm movement image through an individual sperm identification model and then connecting the frames in a time series manner. For example, the model may be a non-deep learning-based SORT model, a deep learning-based Deep SORT model, a FairMOT model, a ByteTrack model, a Norfair model, or a composite, modified, or improved model thereof, but is not limited thereto.

[0077] The sperm movement classification model can be a single time series analysis model or a composite time series analysis model. The single time series analysis model can analyze time series video frame data input from the sperm tracking model based on a single artificial intelligence model, and then recognize and classify the sperm movement pattern in the video. For example, it can be an artificial intelligence model of the 3D CNN series such as the C3D model, I3D model, SlowFast Network model, R(2+1)D model, X3D model, etc., but is not limited thereto. The composite time series analysis model can generate time series analysis data by analyzing time series image frame data input from a sperm tracking model with a CNN-based object classification model, and then input the data into the time series analysis model to classify sperm movement patterns. For example, the object classification model can be a ResNet model including ResNet18, ResNet34, ResNet50, ResNet101, ResNet152, an EfficientNet model including EfficientNet B0, EfficientNet B1, EfficientNet B2, EfficientNet B3, EfficientNet B4, EfficientNet B5, EfficientNet B6, EfficientNet B7, or a composite, modified, or improved model thereof, and the time series analysis model can be a recurrent neural network (RNN) model including LSTM, BiLSTM, GRU, a transformer-based model, or a composite, modified, or improved model thereof, but is not limited thereto.

[0078] The sperm motility analysis model can be trained using pre-prepared sperm motility image data. The sperm motility analysis model can be trained supervised or semi-supervised based on labeled data in which the user inputs sperm motility patterns, or can be trained supervised or semi-supervised based on sperm motility patterns derived from unsupervised learning on image data without sperm motility pattern labeling. For example, when classifying sperm motility patterns using unsupervised learning, the analysis model can establish two to eight motility patterns based on unsupervised learning on sperm motility images, and then perform subsequent supervised learning using labeled image data of PGT-A test results, thereby optimizing the performance of predicting fertilized egg or embryo abnormalities.

[0079] In the abnormality prediction unit (30), an abnormality prediction step (S30) can be performed. In the abnormality prediction step (S30), based on the abnormality prediction information including sperm movement pattern information and speed ratio information for each movement pattern generated in the sperm movement analysis step (S20), an abnormality ratio prediction can be made to determine whether a fertilized egg formed using a subject's semen sample is abnormal or whether an embryo is mosaic.

[0080] At this time, when predicting abnormalities, the reliability and accuracy of prediction may not be sufficient with only sperm movement pattern information and speed ratio information for each movement pattern, and the age information of the maternal individual that provided the egg and sperm clinical information may be utilized as additional abnormality prediction information. Sperm clinical information may include some or all of the measurement elements utilized in conventional CASA analysis, and may include, but is not limited to, the percentage of progressively motile sperm in the semen, the average speed of motile sperm, semen volume, or subject age information.

[0081] In the abnormality prediction unit (30), a series of algorithms can be performed to perform the abnormality prediction step (S30), and the execution of the algorithms can be performed through an artificial intelligence model. For example, based on multivariables including various information including sperm movement pattern information and speed ratio information for each movement pattern, prediction of abnormalities in fertilized eggs or embryos when using a subject's semen sample can be performed through XGBoost, Random Foreset, MLP, and logistic regression models, but is not limited thereto.

[0082] A series of algorithms for performing the sperm movement analysis step (S20) or the abnormality prediction step (S30) in the sperm movement analysis unit (20) or the abnormality prediction unit (30) may be performed through one or more processors (not shown). The processor may be composed of one or more cores and may include a processor for data analysis or deep learning, such as a central processing unit (CPU), a graphics processing unit (GPU), or a tensor processing unit (TPU). In addition, the processor may perform data processing and calculations for learning a neural network model required for driving a system or performing a method according to one embodiment of the present invention.

[0083] The system (100) of the present invention may further include a storage unit (40).

[0084] The system (100) of the present invention may further include a network unit (not shown) for transmitting and receiving data between the image acquisition unit (10), sperm movement analysis unit (20), abnormality prediction unit (30) or storage unit (40), and data transmission and reception through the network unit may be performed in a wired or wireless manner.

[0085]

[0086] Hereinafter, the present invention will be described in more detail with reference to the following examples. However, these examples are intended only to illustrate the present invention, and the scope of the present invention is not limited by these examples.

[0087] Throughout this specification, "%" used to indicate the concentration of a particular substance is (wt / wt)% for solid / solid, (wt / vol)% for solid / liquid, and (vol / vol)% for liquid / liquid, unless otherwise stated.

[0088] Unless otherwise specified, all numbers, values ​​and / or expressions expressing ingredients, reaction conditions and quantities of ingredients used herein are to be understood as being modified in all instances by the term "about" because these numbers are approximations that inherently reflect, among other things, the various uncertainties of measurement encountered in obtaining such values.

[0089] Additionally, when a numerical range is disclosed herein, such range is continuous and includes all values ​​from the minimum value to the maximum value inclusive, unless otherwise specified.

[0090]

[0091] Example 1. Materials and Methods

[0092] Clinical Data Collection and IVF Patient Information

[0093] This study collected clinical data following in vitro fertilization (IVF) treatment through a retrospective analysis. This study was approved by the Institutional Review Board of CHA University, Republic of Korea (Approval Number: GCI 2022-07-008-017). All data collection and analysis procedures, including patient information, were conducted in compliance with the privacy and ethics guidelines outlined by the IRB. Data collection was conducted across the following three categories.

[0094] Sperm motility video data: Sperm motility analysis was performed based on computer-aided sperm analysis (CASA), and dynamic data in the form of video clips were obtained for each sample.

[0095] Aneuploidy analysis data: Aneuploidy information (normal, mosaic, aneuploidy, etc.) of each embryo obtained through Preimplantation Genetic Testing for Aneuploidy (PGT-A) was quantitatively collected.

[0096] Clinical background and outcome data: Various clinical parameters were included, including age of male and female, female anti-Mullerian hormone (AMH) level, body mass index (BMI), basal luteinizing hormone (FSH) level, total number of oocytes retrieved, embryo development rate, oocyte maturity, PGT-A test result (including mosaicism), and clinical pregnancy status.

[0097] The basic data for developing the algorithm of the present invention and verifying its diagnostic performance was constructed based on the comprehensive clinical data described above, and was utilized for data-based modeling to analyze the correlation between sperm movement patterns and the occurrence of aneuploidy.

[0098]

[0099] Deep Learning / Machine Learning-Based Sperm Analysis Data Processing

[0100] In this invention, an AI-based diagnostic algorithm was developed based on clinical data, including sperm motility imaging data and IVF results (PGT-A analysis results and clinical pregnancy status). Deep learning and machine learning techniques were applied to implement the algorithm.

[0101] For sperm detection and tracking, the YOLOv8 model, a state-of-the-art object detection technology, was used. YOLOv8 is a deep learning-based object detection framework that offers both real-time processing performance and high detection accuracy. In this study, it was utilized to simultaneously detect multiple sperm in semen samples and track the movement paths of individual sperm. This model is particularly optimized for small object detection, enabling accurate identification of even the minute shapes of sperm and maintaining stable detection performance under various shooting conditions and sample environments.

[0102]

[0103] Sperm detected by the YOLOv8 model were converted into individual motility path images, and the ResNet18 model was applied to classify sperm motility patterns based on these images. ResNet18 is a deep neural network architecture that incorporates residual learning and can maintain high classification accuracy even with a relatively small number of parameters. The model effectively learned the complex motility patterns of sperm and, based on this, was able to precisely classify sperm into specific motility types.

[0104] In this invention, each sperm was classified into four motility pattern classes using the ResNet18 model. An AI analysis algorithm was then developed based on the velocity ratio and characteristics of each motility pattern. The developed algorithm integrates individual sperm motility information with clinical data to predict the likelihood of aneuploidy and clinical pregnancy success.

[0105]

[0106] Analysis of the Relationship Between AI Score, PGT-A Results, and Pregnancy

[0107] In the present invention, sperm detection and movement path tracking were performed using an artificial intelligence model. Specifically, sperm were automatically detected from sperm movement images and their individual movement paths were tracked, thereby analyzing the functional characteristics of the sperm.

[0108] First, we verified whether the AI ​​model could accurately identify sperm and non-sperm objects in the image. This process evaluated the accuracy and efficiency of sperm detection. This detection performance verification was a crucial step in ensuring the reliability of AI-based diagnosis.

[0109] Subsequently, the motility patterns of the detected sperm were classified and a diagnostic score was calculated. Sperm motility patterns were classified into four classes: Type A (progressive high-speed linear), Type B (curved linear), Type C (progressive curved movement), and Type D (circular). The average speed and velocity distribution ratio of sperm belonging to each class were also calculated. This classification and quantification process enabled a more structural representation of sperm motility characteristics.

[0110] Next, AI analysis scores were calculated based on movement patterns and speed ratios, and these scores were correlated with PGT-A results. Specifically, we assessed whether specific sperm movement patterns were statistically significantly associated with the occurrence of aneuploidy in fertilized eggs.

[0111]

[0112] Example 2. Results

[0113] Development of AI algorithms based on deep learning / machine learning using sperm profiles and clinical data.

[0114] In this invention, an AI-based sperm analysis algorithm was developed using computer-aided sperm analysis-based sperm movement video data obtained from a total of 858 clinical participants. The collected data comprised 4,290 sperm movement video clips and was used to train and evaluate deep learning and machine learning models.

[0115] Next, the present invention evaluated the correlation between sperm-based AI diagnostic scores and aneuploidy test results and clinical pregnancy outcomes. To this end, sperm detection and tracking accuracy was optimized from CASA-based images, and AI-based analysis was performed according to the following procedure. The process of performing AI-based analysis according to these procedures is illustrated in Figure 3.

[0116] (1) Sperm detection and tracking

[0117] The YOLOv8 model was applied to sperm image analysis. This model is optimized for multi-object detection and can track sperm location and movement with high precision.

[0118] (2) Classification of sperm movement patterns

[0119] Based on sperm motility data separated by the YOLOv8 model, the ResNet18 model was applied to classify sperm motility patterns. Motility patterns were classified into four types according to the WHO 6th edition criteria, and kinematic parameters such as velocity and displacement were also analyzed.

[0120] (3) Building a PGT-A prediction model

[0121] A Random Forest algorithm was applied to analyze the relationship between various kinematic characteristics, such as sperm motility patterns and velocity ratios, and the PGT-A results, by constructing multivariate indices. The training data was preprocessed and fed into the classification algorithm, and the model's predictive performance was quantitatively evaluated.

[0122] As a result, the AI-based analysis system of the present invention recorded the following performance:

[0123] Sperm detection accuracy: 95.54%

[0124] Sperm tracking accuracy: 88.48%

[0125] Sperm motility pattern classification accuracy: 97%

[0126] PGT-A result prediction accuracy: 76%

[0127] These results demonstrate that this technology can predict key information related to fertility and genetic health through high-precision analysis based on sperm images, and demonstrate that it can make a practical contribution to diagnosing unexplained male infertility and improving ART success rates.

[0128]

[0129] Correlation between AI-based sperm analysis scores and PGT-A predictions

[0130] In one embodiment of the present invention, the correlation between AI-based sperm analysis results and embryo aneuploidy rate (PGT-A) and clinical pregnancy rate was evaluated.

[0131] First, based on PGT-A data, embryos were dichotomized into two groups: 0% abnormal (completely abnormal chromosomes) and 100% normal (normal chromosomes). Then, sperm motility pattern data were linked to perform a two-way ANOVA to evaluate the cross-correlation between sperm motility patterns and pregnancy rates.

[0132] As a result, no statistically significant correlation was observed between total sperm count or overall motility pattern ratio and pregnancy outcome. However, AI-based analysis confirmed that specific sperm motility patterns (types A and B) showed a significant positive correlation with the pregnancy success rate. In particular, type A and B motility patterns were statistically correlated with a high pregnancy success rate, although their frequency was relatively low compared to types C and D.

[0133] In an embodiment of the present invention, the discriminatory power of the AI ​​model was evaluated through ROC analysis, and as shown in Fig. 4, the prediction model based on type A and type B sperm movement patterns recorded an AUC (Area Under Curve) value of 0.90, showing excellent performance in distinguishing between the pregnant group and the non-pregnant group.

[0134] Additionally, Figure 5 shows the confusion matrix of the XGBoost classification model for pregnancy prediction. The model defined class 0 as the non-pregnant group and class 1 as the pregnant group, and as a result, demonstrated excellent classification performance by accurately predicting 88 cases in the non-pregnant group and 56 cases in the pregnant group.

[0135] Based on these results, the present invention integrated sperm profile information, PGT-A results, and clinical pregnancy outcomes to derive an AI-based diagnostic score. The score was then repeatedly validated for aneuploidy rates and pregnancy prediction performance. This supported the hypothesis that AI-based analysis could make a substantial contribution to improving the predictability of unexplained male infertility and ART success rates.

[0136]

[0137] Example 3. Discussion of Results

[0138] The AI ​​algorithm developed according to the present invention utilizes the latest deep learning-based models, such as YOLOv8 and EfficientNet, to analyze sperm movement images, track the movement paths of individual sperm, and classify them into unique movement patterns. These movement patterns were then used to analyze the correlation with the aneuploidy rate of embryos, and the developed AI diagnostic algorithm achieved a prediction accuracy of 0.9. This is significant in that, unlike most medical AI focused on static image analysis, this technology performs pattern recognition and prediction based on dynamic images.

[0139] One key finding was that a high proportion of sperm with a circular motility pattern significantly increased the rate of aneuploidy in embryos before embryo transfer, whereas sperm with linear and fast-progressive motility patterns were associated with ploidy. This suggests a functional link between sperm motility characteristics and chromosome segregation within the fertilized egg.

[0140] While a direct correlation between paternal age and embryonic aneuploidy rate was not confirmed in the present invention, sperm centrosome proteins are known to be key factors involved in chromosome segregation after fertilization. Sperm centrosome, dynein, and microtubule proteins have been reported to play a functional role in regulating motility, as well as clustering of the paternal genome and chromosome segregation within the fertilized egg.

[0141] While pre-embryo transfer genetic testing (PGT-A) currently focuses primarily on older women or those with recurrent IVF failure, recent studies have highlighted that sperm abnormalities can also contribute to the development of aneuploidy. Specifically, paternal chromosome segregation errors during early mitosis, after pronuclear formation after fertilization, have been reported to be a cause of aneuploidy.

[0142] The correlation between the AI ​​analysis scores of the aneuploidy prediction algorithm disclosed in the present invention and PGT-A results suggests that specific sperm motility profiles may be associated with chromosome missegregation during mitosis within the fertilized egg. Therefore, the AI ​​algorithm according to the present invention can contribute to the diagnosis of unexplained male infertility by predicting the likelihood of aneuploidy in embryos in advance, and is expected to have potential as a clinical decision-making support tool, including determining the need for PGT-A.

[0143] In conclusion, the present invention demonstrates that AI-based analysis of sperm motility patterns can predict the likelihood of aneuploidy in embryos, thereby offering a novel diagnostic approach in the field of reproductive medicine. The analysis findings that progressively motile sperm are associated with aneuploidy and circulating sperm are associated with aneuploidy offer valuable insights into identifying paternal factors for chromosomal errors in fertilized eggs.

[0144] Therefore, the present invention is expected to shift the paradigm of future male infertility diagnosis, contribute to the establishment of more precise personalized treatment plans such as establishing ART strategies and determining whether to perform PGT-A, provide substantial progress in identifying the functional causes of unexplained male infertility through future clinical verification and data expansion, and become an important technological turning point that can improve the success rate of ART.

[0145] 100: System

[0146] 10: Image acquisition section

[0147] 20: Sperm movement analysis department

[0148] 30: Ideal Prediction Department

[0149] 40: Storage

[0150] S10: Image acquisition stage

[0151] S20: Sperm movement analysis stage

[0152] S30: Anomaly Prediction Stage

[0153] The present invention relates to a method and system for predicting abnormalities in a fertilized egg or embryo through sperm movement analysis, and more particularly, to a method and system for analyzing sperm movement using an artificial intelligence model on a sperm movement image obtained from a semen sample of a subject, and predicting abnormalities in a fertilized egg or embryo fertilized using the subject's semen based on the analysis results.

Claims

1. An image acquisition step for acquiring an image of sperm movement from a semen sample of a subject; A sperm movement analysis step for analyzing sperm movement in the above sperm movement image to obtain sperm movement pattern information and speed ratio information for each movement pattern; and An abnormality prediction step for predicting an abnormality in a fertilized egg or embryo using the subject's semen based on abnormality prediction information including the above sperm movement pattern information and speed ratio information for each movement pattern; A method for predicting abnormalities in a fertilized egg or embryo, including:

2. In the first paragraph, the sperm movement analysis step is, A sperm tracking step for tracking sperm within the above sperm movement image; and A sperm movement classification step of generating sperm movement pattern information by classifying sperm movement in the above sperm movement image into one of a plurality of movement patterns; A method for predicting abnormalities in a fertilized egg or embryo, comprising:

3. In the second paragraph, the sperm tracking step is: An individual sperm identification step for identifying individual sperm within the above sperm movement image; and An identification sperm tracking step for tracking sperm identified in the above individual sperm identification step; A method for predicting abnormalities in a fertilized egg or embryo, comprising:

4. A method for predicting abnormalities in a fertilized egg or embryo, wherein the sperm movement classification step in the second paragraph generates sperm movement pattern information by classifying sperm movement into one of two to eight types of movement patterns.

5. A method for predicting abnormalities in a fertilized egg or embryo, wherein the sperm movement classification step, in paragraph 4, classifies sperm movement into one movement pattern selected from the group consisting of progressive high-speed linear movement, curved linear movement, progressive curved movement, and circular movement.

6. A method for predicting abnormalities in a fertilized egg or embryo, wherein the abnormality prediction information in paragraph 1 further includes age information or sperm clinical information of the maternal individual providing the egg.

7. A method for predicting abnormalities in a fertilized egg or embryo, wherein the sperm clinical information in paragraph 6 includes at least one piece of information selected from the group consisting of the proportion of progressively motile sperm in semen, the average velocity of motile sperm, the amount of semen, and the subject's age.

8. A method for predicting abnormalities in a fertilized egg or embryo, wherein the sperm movement analysis step or abnormality prediction step in paragraph 1 is performed using an artificial intelligence model.

9. A method for predicting a fertilized egg or embryo abnormality in paragraph 1, wherein the fertilized egg or embryo abnormality is aneuploidy of the fertilized egg or embryo.

10. An image acquisition unit that acquires an image of sperm movement from a semen sample of the subject; A sperm movement analysis unit that analyzes sperm movement in the above sperm movement image to obtain sperm movement pattern information and speed ratio information for each movement pattern; and An abnormality prediction unit that predicts an abnormality in a fertilized egg or embryo fertilized using the subject's semen based on abnormality prediction information including the above sperm movement pattern information and speed ratio information for each movement pattern; A system for predicting abnormalities in a fertilized egg or embryo, including:

11. The processor performs an operation to predict abnormalities in a fertilized egg or embryo using the subject's semen, The above actions are: An operation of analyzing sperm movement in a sperm movement video to obtain sperm movement pattern information and speed ratio information for each movement pattern; and An operation of predicting an abnormality in a fertilized egg or embryo using the subject's semen based on abnormality prediction information including the above sperm movement pattern information and speed ratio information for each movement pattern; A computer program stored on a storage medium, comprising:

12. A processor comprising one or more cores; The above processor, By analyzing sperm movement in a sperm movement image, sperm movement pattern information and speed ratio information for each movement pattern are obtained, and based on abnormality prediction information including the sperm movement pattern information and speed ratio information for each movement pattern, abnormalities in a fertilized egg or embryo fertilized using the subject's semen are predicted. A computing device for predicting abnormalities in fertilized eggs or embryos using the subject's semen.

Images