Embryo development trajectory mapping system and method fusing morphological features

Abstract

This invention relates to the interdisciplinary field of assisted reproductive technology and artificial intelligence, specifically to a system and method for constructing an embryonic developmental trajectory map that integrates morphological features. By deeply integrating static morphological features with dynamic morphodynamic parameters, a more comprehensive panoramic map of embryonic development with richer information dimensions is constructed, overcoming the one-sidedness of single-dimensional assessment. Automated analysis of the map using an AI model reduces the influence of subjective human factors, enabling the discovery of subtle developmental patterns and potential differences that are difficult to discern with the human eye, significantly improving the accuracy of selecting high-quality embryos. The provided developmental trajectory map allows embryologists to intuitively examine the entire developmental process of the embryo and the performance of each key node, enhancing the interpretability and credibility of the assessment process. Through more scientific embryo selection, it is expected to improve embryo implantation and clinical pregnancy rates, and reduce multiple pregnancy and miscarriage rates.

Description

Technical Field

[0001] This invention relates to the field of interdisciplinary technology of assisted reproductive technology and artificial intelligence, and in particular to a system and method for constructing embryonic development trajectory maps that integrate morphological features. Background Technology

[0002] In in vitro fertilization-embryo transfer (IVF-ET), embryonic development from gamete to blastocyst is a complex and dynamic process. Traditional embryo assessment methods mainly rely on embryologists taking "snapshot" morphological observations of the embryo at specific time points (such as day 2 or 3 after fertilization), such as the number of blastomeres, cell homogeneity, and fragmentation ratio. This method is highly subjective, labor-intensive, and ignores the dynamic process and continuous information of embryonic development.

[0003] In recent years, the application of time-lapse incubators has made it possible to perform continuous, non-invasive imaging of the entire embryonic development process. By analyzing a series of time-series images of the embryo during development, morphodynamic parameters such as cell division time (t2, t3, t4...), cell cycle duration (cc1, cc2, cc3...), and division synchronicity (s2, s3) can be extracted. Studies have shown that these dynamic parameters are closely related to the embryo's final developmental potential and implantation success rate, providing a more accurate quality assessment than observations at a single time point.

[0004] However, existing technologies still have limitations: information is isolated, with most systems analyzing only morphological features (static structure) or morphodynamic parameters (dynamic time series) separately, failing to organically combine the two to construct a unified view that can comprehensively reflect the embryonic developmental state; atlases are missing, lacking a systematic and visual representation of the individual embryonic development process, making it difficult for doctors to intuitively see the "growth trajectory" of a particular embryo throughout the entire developmental cycle and the evolution of its key morphological features; assessment dimensions are limited, with assessment models mostly based on a single or few parameters, failing to fully utilize the advantages of multi-dimensional feature fusion, thus limiting the accuracy and intelligence level of the assessment.

[0005] Therefore, there is an urgent need for a new method and system that can deeply integrate the static morphological characteristics of embryos with dynamic developmental time information to construct an intuitive and comprehensive developmental trajectory map and to accurately assess embryo quality. Summary of the Invention

[0006] The purpose of this invention is to provide a system and method for constructing an embryonic development trajectory map that integrates morphological features, aiming to solve the problems of isolated information, lack of systematic visualization, and single evaluation dimension in the prior art.

[0007] To achieve the above objectives, in a first aspect, the present invention provides an embryonic development trajectory map construction system that integrates morphological features, including a data acquisition and preprocessing module, a feature extraction module, a map construction and analysis module, and a potential assessment and reporting module; The data acquisition and preprocessing module is used to acquire time-series image data of embryo time difference imaging and perform image restoration and enhancement preprocessing. The feature extraction module is used to extract multi-dimensional features of the embryo; The atlas construction and analysis module is used to fuse the extracted morphodynamic parameters and morphological features according to the time series to construct a multidimensional developmental trajectory atlas of a single embryo. The potential assessment and reporting module is used to input the constructed developmental trajectory map into a pre-trained machine learning model, automatically output the developmental potential score or level of the embryo, and generate a visual report containing the map and assessment results.

[0008] The feature extraction module includes a morphological dynamics parameter extraction unit and a morphological feature recognition unit. The morphodynamic parameter extraction unit is used to automatically identify key time nodes in embryonic development and calculate dynamic parameters of cell cycle and division synchronization. The morphological feature recognition unit, based on a deep learning model, automatically segments embryo images, identifies the inner cell mass, trophoblast, and cyst cavity structure of the blastocyst, and quantifies and scores its quality such as morphology, size, and uniformity.

[0009] The multidimensional developmental trajectory map uses developmental time as the horizontal axis and cell division stage, cc2 duration, and inner cell mass area as the vertical axes, respectively, to depict the developmental trajectory of the embryo in the form of multiple curves or heat maps.

[0010] Secondly, a method for constructing an embryonic developmental trajectory map integrating morphological features, used in the embryonic developmental trajectory map construction system integrating morphological features described in the first aspect, includes the following steps: Complete time-lapse image sequences of embryos from fertilization to blastocyst were acquired, and image restoration algorithms were used to deblur the images and improve image quality. Extracting multidimensional features from embryos; By associating all the features of the step with their corresponding developmental time points or stages, a multi-dimensional data matrix is ​​constructed, and a visualized trajectory map is generated based on this matrix. Each embryo's atlas or its high-dimensional feature vector is input into a pre-trained evaluation model to obtain the embryo's potential score. All embryos to be evaluated are sorted by score, and a comparative analysis report is generated, highlighting the embryo with the highest score and its developmental atlas.

[0011] The "extraction of multidimensional features of embryos" includes the following steps: Using a time-series learning model or keyframe recognition algorithm, cell division events are automatically located, key time points t2, t3, t4, t5, and tB are calculated, and morphodynamic parameters cc1, cc2, s2, and s3 are derived. Image segmentation models were used to segment images of key developmental nodes at the pixel level to identify the inner cell mass and trophoblast regions. The area ratio of the inner cell mass to the trophoblast, cell density, and uniformity were calculated, and Gardner scoring quality was assessed in conjunction with expert knowledge.

[0012] The embryonic development trajectory map construction system of this invention integrates morphological features to construct a more comprehensive panoramic map of embryonic development by deeply fusing static morphological features with dynamic morphodynamic parameters, overcoming the limitations of single-dimensional assessment. Automated analysis of the map using an AI model reduces the influence of subjective human factors, enabling the discovery of subtle developmental patterns and potential differences that are difficult to discern with the human eye, significantly improving the accuracy of high-quality embryo selection. The provided developmental trajectory map allows embryologists to intuitively examine the entire developmental process of the embryo and the performance of each key node, enhancing the interpretability and credibility of the assessment process. Through more scientific embryo selection, it is expected to improve embryo implantation and clinical pregnancy rates, while reducing multiple pregnancy and miscarriage rates. Attached Figure Description

[0013] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0014] Figure 1 This is a schematic diagram of the embryonic development trajectory map construction system that integrates morphological features provided by the present invention.

[0015] Figure 2 This is a schematic diagram of the feature extraction module.

[0016] Figure 3 This is an overall flowchart of the method for constructing an embryonic development trajectory map that integrates morphological features, provided by the present invention.

[0017] Figure 4 This invention presents an embryonic development trajectory diagram: cell division and morphological characteristics, showing the trajectory of cell division time and ICM / TE area changes over time.

[0018] Figure 5This is a schematic diagram of the automatic segmentation of embryo morphological features (ICM / TE) in this invention.

[0019] Figure 6 This is a schematic diagram of the interface for an atlas-based embryo potential assessment and comparison report.

[0020] In the figure: 1-Data acquisition and preprocessing module, 2-Feature extraction module, 3-Map construction and analysis module, 4-Potential assessment and reporting module, 21-Morphodynamic parameter extraction unit, 22-Morphological feature recognition unit. Detailed Implementation

[0021] Embodiments of the present invention are described in detail below, examples of which are illustrated in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain the present invention, and should not be construed as limiting the present invention.

[0022] Please see Figures 1 to 2 In a first aspect, the present invention provides an embryonic development trajectory map construction system that integrates morphological features, including a data acquisition and preprocessing module, a feature extraction module, a map construction and analysis module, and a potential assessment and reporting module; The data acquisition and preprocessing module is used to acquire time-series image data of embryo time difference imaging and perform image restoration and enhancement preprocessing. The feature extraction module is used to extract multi-dimensional features of the embryo; The atlas construction and analysis module is used to fuse the extracted morphodynamic parameters and morphological features according to the time series to construct a multidimensional developmental trajectory atlas of a single embryo. The potential assessment and reporting module is used to input the constructed developmental trajectory map into a pre-trained machine learning model, automatically output the developmental potential score or level of the embryo, and generate a visual report containing the map and assessment results.

[0023] Furthermore, the feature extraction module includes a morphological dynamics parameter extraction unit and a morphological feature recognition unit; The morphodynamic parameter extraction unit is used to automatically identify key time nodes in embryonic development and calculate dynamic parameters of cell cycle and division synchronization. The morphological feature recognition unit, based on a deep learning model, automatically segments embryo images, identifies the inner cell mass, trophoblast, and cyst cavity structure of the blastocyst, and quantifies and scores its quality such as morphology, size, and uniformity.

[0024] Furthermore, the multidimensional developmental trajectory map uses developmental time as the horizontal axis and cell division stage, cc2 duration, and inner cell mass area as the vertical axes, respectively, to depict the developmental trajectory of the embryo in the form of multiple curves or heat maps.

[0025] In this embodiment, the data acquisition and preprocessing module acquires time-series image data of embryonic time-varying imaging and performs preprocessing such as image restoration and enhancement. The feature extraction module includes a morphodynamic parameter extraction unit, which automatically identifies key time points in embryonic development (e.g., t2, t3, t4, t5, tB) and calculates dynamic parameters such as cell cycle (cc1, cc2, cc3...) and cell division synchronicity (s2, s3). The morphological feature recognition unit, based on deep learning models (e.g., SA-Net, MTL-ShuffleNet), automatically segments embryonic images, identifies structures such as the inner cell mass (ICM), trophoblast (TE), and cystic cavity of the blastocyst, and quantifies their morphology, size, and uniformity using a quantitative scoring method (e.g., Gardner score). The atlas construction and analysis module fuses the extracted morphodynamic parameters and morphological features according to time series to construct a multi-dimensional developmental trajectory atlas of a single embryo. This atlas can visually display the various characteristic values ​​and their changing trends of the embryo at different developmental stages in a graphical format. Potential assessment and reporting module: This module is used to input the constructed developmental trajectory map into a pre-trained machine learning model (such as support vector machine SVM, deep learning classifier), automatically output the embryo's developmental potential score or grade (such as excellent, good, poor), and generate a visual report containing the map and assessment results.

[0026] Please see Figures 3 to 6 Secondly, a method for constructing an embryonic developmental trajectory map integrating morphological features, used in the embryonic developmental trajectory map construction system integrating morphological features described in the first aspect, includes the following steps: S1 acquires a complete time-lapse image sequence of the embryo from fertilization to blastocyst, and uses an image restoration algorithm to deblur the images and improve image quality; Specifically, continuous images of an embryo from day 1 to day 5 are acquired from a time-lapse incubator (such as the Vitrolife EmbryoScope or the MTL-20-m of this invention). Due to potential liquid level fluctuations or focal drift during culture, the images may be blurred. The system first invokes the image restoration module, which generates a blur intensity feature map based on "fuzzy network strength assessment," and then uses a specific blur-removal master network to iteratively restore the images, obtaining a clear sequence of embryo images.

[0027] S2 extracts multidimensional features of embryos; S21 uses a time-series learning model or keyframe recognition algorithm to automatically locate cell division events, calculate key time points t2, t3, t4, t5, and tB, and derive morphodynamic parameters cc1, cc2, s2, and s3. Specifically, the system employs a precise detection algorithm for each stage of embryonic development. This algorithm includes collaborative network extraction of high-order features, metric learning for feature optimization, and temporal learning for capturing temporal correlations. The system automatically identifies the precise time points of key transitions in the embryo, such as from 1 cell to 2 cells (t2) and from 2 cells to 4 cells (t4), and calculates parameters such as the duration of cc2 (t3-t2) and the synchronicity of s2 (t4-t3).

[0028] S22 uses an image segmentation model to perform pixel-level segmentation of images of key developmental nodes, identifies the inner cell mass and trophoblast regions, calculates the area ratio of the inner cell mass to the trophoblast, cell density, and uniformity quantification indicators, and combines expert knowledge to conduct Gardner score quality assessment.

[0029] Specifically, when the embryo develops to the blastocyst stage (approximately day 5), the system invokes the morphological feature recognition unit. This unit uses **SA-Net (scale-aware network) to segment the blastocyst image (see [reference]). Figure 4 SA-Net can effectively segment different regions such as the zona pellucida, trophoblast (TE), cyst cavity, and inner cell mass (ICM). After segmentation, the system calculates the pixel area ratio and boundary sharpness of the ICM and TE regions, and combines these quantitative indicators to give a Gardner blastocyst quality score (such as 3BB, 4AA, etc.).

[0030] S3 associates all features with their corresponding developmental time points or stages to construct a multi-dimensional data matrix, and generates a visualized trajectory map based on this matrix. Specifically, the system integrates all the data extracted in step S2. For example, for a single embryo, we obtained data such as t2=25h, t4=40h, cc2=8h, ICM / TE area ratio=0.35, and Gardner score=4BB. The system maps these data points onto a time series to generate a multi-dimensional developmental trajectory map (see reference). Figure 3 The atlas clearly shows that the embryo has a short cc2 duration (rapid development) and an ideal ICM / TE area, indicating that it has good developmental potential.

[0031] S4 inputs the atlas or its high-dimensional feature vector of each embryo into the pre-trained evaluation model to obtain the potential score of the embryo. It sorts all the embryos to be evaluated according to the score and generates a comparative analysis report, highlighting the embryo with the highest score and its developmental atlas.

[0032] Specifically, the system pre-trains a potential assessment model based on Support Vector Machine (SVM) or a neural network in the background. The model's training data comes from hundreds of embryos with known clinical outcomes (implantation / non-implantation) and their corresponding morphodynamic and morphological characteristics. After inputting the atlas feature vector of the current embryo into the model, the model outputs a quantified potential score. The system sorts all candidate embryos and generates a report (see [reference]). Figure 5 The report highlights the highest-scoring embryos, which doctors can click to view detailed developmental trajectory maps for final decision-making.

[0033] Beneficial effects: More comprehensive information fusion: For the first time, static morphological features and dynamic morphodynamic parameters are deeply integrated to construct a panoramic atlas of embryonic development with richer information dimensions, overcoming the one-sidedness of single-dimensional assessment.

[0034] More objective and accurate assessment: By using AI models to automatically analyze the patterns, the influence of human subjective factors is reduced, and subtle developmental patterns and potential differences that are difficult to discern with the human eye can be discovered, significantly improving the accuracy of screening high-quality embryos.

[0035] Intuitive visualization: The provided developmental trajectory atlas allows embryologists to visually examine the entire developmental process of the embryo and the performance of each key node, enhancing the interpretability and credibility of the assessment process and facilitating case review and teaching research.

[0036] Improving the success rate of assisted reproduction: By screening embryos more scientifically, it is hoped that the embryo implantation rate and clinical pregnancy rate can be improved, while the multiple pregnancy rate and miscarriage rate can be reduced, ultimately benefiting infertile patients.

[0037] The above-disclosed embodiments are merely preferred embodiments of the embryonic development trajectory map construction system and method integrating morphological features of the present invention. Of course, they should not be construed as limiting the scope of the present invention. Those skilled in the art can understand that implementing all or part of the above embodiments and making equivalent changes in accordance with the claims of the present invention still fall within the scope of the invention.

Claims

1. A system for constructing a map of embryonic development trajectories integrating morphological features, characterized in that, It includes modules for data acquisition and preprocessing, feature extraction, map construction and analysis, and potential assessment and reporting. The data acquisition and preprocessing module is used to acquire time-series image data of embryo time difference imaging and perform image restoration and enhancement preprocessing. The feature extraction module is used to extract multi-dimensional features of the embryo; The atlas construction and analysis module is used to fuse the extracted morphodynamic parameters and morphological features according to the time series to construct a multidimensional developmental trajectory atlas of a single embryo. The potential assessment and reporting module is used to input the constructed developmental trajectory map into a pre-trained machine learning model, automatically output the developmental potential score or level of the embryo, and generate a visual report containing the map and assessment results.

2. The embryonic development trajectory map construction system integrating morphological features as described in claim 1, characterized in that, The feature extraction module includes a morphodynamic parameter extraction unit and a morphological feature recognition unit; The morphodynamic parameter extraction unit is used to automatically identify key time nodes in embryonic development and calculate dynamic parameters of cell cycle and division synchronization. The morphological feature recognition unit, based on a deep learning model, automatically segments embryo images, identifies the inner cell mass, trophoblast, and cyst cavity structure of the blastocyst, and quantifies and scores its quality such as morphology, size, and uniformity.

3. The embryonic development trajectory map construction system integrating morphological features as described in claim 1, characterized in that, The multidimensional developmental trajectory map uses developmental time as the horizontal axis and cell division stage, cc2 duration, and inner cell mass area as the vertical axes, respectively, to depict the developmental trajectory of the embryo in the form of multiple curves or heat maps.

4. A method for constructing an embryonic developmental trajectory atlas incorporating morphological features, used in the embryonic developmental trajectory atlas construction system incorporating morphological features as described in any one of claims 1-3, characterized in that, Includes the following steps: Complete time-lapse image sequences of embryos from fertilization to blastocyst were acquired, and image restoration algorithms were used to deblur the images and improve image quality. Extracting multidimensional features from embryos; By associating all the features of the step with their corresponding developmental time points or stages, a multi-dimensional data matrix is ​​constructed, and a visualized trajectory map is generated based on this matrix. Each embryo's atlas or its high-dimensional feature vector is input into a pre-trained evaluation model to obtain the embryo's potential score. All embryos to be evaluated are sorted by score, and a comparative analysis report is generated, highlighting the embryo with the highest score and its developmental atlas.

5. The fusion morphological feature embryo development trajectory graph construction method of claim 4, wherein, The "Extraction of Multidimensional Features of Embryos" includes the following steps: Using a time-series learning model or keyframe recognition algorithm, cell division events are automatically located, key time points t2, t3, t4, t5, and tB are calculated, and morphodynamic parameters cc1, cc2, s2, and s3 are derived. Image segmentation models were used to segment images of key developmental nodes at the pixel level to identify the inner cell mass and trophoblast regions. The area ratio of the inner cell mass to the trophoblast, cell density, and homogeneity were calculated, and Gardner scoring quality was assessed in conjunction with expert knowledge.

Images