Cow blastocyst development potential evaluation method and system based on unmarked imaging and lipid metabolism characteristics

Abstract

The invention belongs to the technical field of animal embryo engineering, and particularly relates to a cattle blastocyst development potential evaluation method and system based on unmarked imaging and lipid metabolism characteristics. The method comprises the following steps: performing laser unmarked multi-modal imaging on a cattle blastocyst to obtain a multi-modal fusion image including cattle blastocyst morphology, metabonomics and matrix structuring; quantitatively extracting morphological kinetic parameters, metabolic activity parameters and matrix structure parameters from the multi-modal fusion image, and constructing a multi-dimensional feature set; and training an evaluation model constructed based on a machine learning algorithm by adopting the multi-dimensional feature set, and obtaining a development potential score of the cattle blastocyst to be evaluated based on the trained evaluation model. According to the method, high-precision, non-invasive and objective prediction of the blastocyst in-vivo development outcome before transplantation is realized.

Description

Technical Field

[0001] This invention belongs to the field of animal embryo engineering technology, specifically to the field of bovine breeding and bovine embryo screening technology, and specifically relates to a method and system for assessing the developmental potential of bovine blastocysts based on label-free imaging and lipid metabolism characteristics. Background Technology

[0002] The protection and utilization of superior livestock and poultry germplasm resources in my country is the "chip" of the seed industry. Cattle, as a large livestock breed, especially endangered local breeds, face enormous challenges in the efficient protection and utilization of their genetic resources. In vitro embryo production is a core means to address this challenge, and its efficiency is a key factor determining its widespread application. In the field of bovine genetic breeding, assisted reproductive technologies such as in vitro fertilization (IVF) have been widely used in the breeding of superior cattle breeds, but factors such as batch-to-batch instability mean that its efficiency still has considerable room for improvement. Somatic cell nuclear transfer (SCNT) technology is a key means of propagating top-quality bulls and preserving and reviving endangered breeds, but its extremely low cloning efficiency is a major bottleneck restricting its application. Currently, blastocysts are mainly used for bovine embryo transfer. Therefore, the developmental capacity of blastocysts directly determines whether an embryo can develop in vivo. Accurate assessment of blastocyst developmental capacity is a crucial link in improving breeding efficiency and promoting the commercialization of superior embryos.

[0003] The industry's assessment of bovine blastocyst developmental capacity primarily relies on traditional morphological observation, which involves examining the blastocyst's morphological integrity, cell arrangement, and inner cell mass size under a microscope. Traditional morphological observation is heavily influenced by human experience; differing evaluation standards among personnel due to experience lead to highly subjective and reproducible results, making it difficult to accurately distinguish between defective and superior embryos. This consumes significant human and material resources, increasing the cost and risk of screening commercial embryos. Molecular markers or staining methods are sometimes used to assist in assessment by detecting specific gene expression or biochemical indicators. While these methods provide deeper information, they require invasive treatment of the blastocyst, potentially damaging embryo viability and affecting subsequent implantation survival rates. Furthermore, they are complex and time-consuming, failing to meet the demands of efficient screening in large-scale production. Furthermore, some staining agents may remain in the embryo, posing a potential threat to the safety of bovine breeding. For example, patent CN201811473357.6 (Embryonic Staining Method, Method for Assessing Embryonic Development Capacity and Reagent Kit) discloses a method for detecting embryonic development capacity through embryo staining, which cannot avoid a series of damage problems to the embryo caused by embryonic staining markers. Embryo biopsy is another method for blastocyst testing. Embryo biopsy technology requires "removing" a small piece of embryonic "material" for genetic testing to diagnose whether the embryo is abnormal and to select healthy embryos for transfer. Embryo biopsy is also an invasive procedure. It is necessary to ensure that a sufficient number of cells are obtained without causing excessive damage to the embryo. Therefore, the operation requirements are very high. The complex operation technique increases the various problems caused by operational errors, including embryonic structural damage, embryonic developmental arrest, and high testing costs. There are some indirect detection methods in the existing technology, but these methods are not accurate enough. For example, CN202310190332.X (A non-invasive detection method for screening embryos with good developmental potential) discloses a detection technology for non-invasively screening embryo development ability through embryo culture medium, but it cannot avoid the problem of accuracy difference caused by indirect analysis through embryo nutrient solution.

[0004] It is evident that existing embryo development screening and detection technologies suffer from the aforementioned shortcomings. Therefore, there is an urgent need for a more accurate, scientific, economical, and non-destructive technology for detecting bovine embryo developmental capabilities. Summary of the Invention

[0005] To address the problems existing in the prior art, this invention provides a method and system for assessing the developmental potential of bovine blastocysts based on label-free imaging and lipid metabolism characteristics. The aim is to evaluate the developmental capacity of bovine blastocysts using laser label-free multimodal imaging technology, providing technical support for the screening and application of superior bovine embryos prepared in vitro.

[0006] To achieve the above objectives, the present invention provides the following solution: A method for assessing the developmental potential of bovine blastocysts based on label-free imaging and lipid metabolism characteristics, the method comprising: Laser label-free multimodal imaging of bovine blastocysts was performed to obtain multimodal fusion images including bovine blastocyst morphology, metabolomics, and matrix structure. Quantitatively extract morphodynamic parameters, metabolic activity parameters, and matrix structure parameters from multimodal fused images to construct a multidimensional feature set; A multi-dimensional feature set was used to train an evaluation model based on a machine learning algorithm. Based on the trained evaluation model, the developmental potential score of the bovine blastocyst to be evaluated was obtained.

[0007] Preferably, the method for performing label-free laser multimodal imaging on bovine blastocysts to obtain multimodal fused images including bovine blastocyst morphology, metabolomics, and matrix structure includes: Three-dimensional scanning of bovine blastocysts was performed using a femtosecond label-free scanning microscopy system, and three types of optical signals, namely second harmonic scattering, autofluorescence and coherent anti-Stokes Raman scattering, were acquired simultaneously. Preprocessing of three types of optical signals, including wavelet denoising, spatial registration, and attention-weighted fusion, generates multimodal fusion images encompassing bovine blastocyst morphology, metabolomics, and matrix structure.

[0008] Preferred methods for extracting morphodynamic parameters include: Blastocyst diameter was calculated by extracting the blastocyst contour using the Canny edge detection algorithm and then employing the minimum circumcircle method. ; Among them, D BL A is the diameter of the blastocyst. BL The area of ​​the blastocyst region; Inner cell mass area and shape factor: ; Among them, S ICM A is a factor relating the area and shape of the inner cell mass. ICM L represents the area of ​​the inner cell mass. ICM The perimeter of the inner cell mass outline; Trophoblast cell density: ; Among them, C TE For the density of trophoblast cell arrangement, A gap A represents the intercellular area of ​​the trophoblast. TE,total This represents the total area of ​​the nutrient layer.

[0009] Preferred methods for extracting metabolic activity parameters include: Average fluorescence intensity: ; Among them, I avg Let I(x,y) be the average fluorescence intensity, I(x,y) be the fluorescence value of pixel (x,y), and N be the fluorescence intensity. ROI The total number of pixels in the region of interest (ROI) is the total number of pixels in the ROI. Fluorescence lifetime was obtained by fitting and solving a double exponential decay model. ; Where τ1 and τ2 are fluorescence lifetime parameters, f(t) represents the function of fluorescence intensity changing with time, A1 is the amplitude coefficient of the first component, t represents time, and A2 is the amplitude coefficient of the second component, and satisfies A1+A2=1.

[0010] Preferably, the method for extracting matrix structure parameters includes: Collagen fiber orientation consistency is calculated using structural tensors: ; Where, λ 1,i , λ 2,i Let D be the eigenvalue of the structural tensor of the i-th collagen fiber. ori To ensure the uniformity of collagen fiber orientation, Indicates the total number of collagen fibers; Mesh structure complexity, calculated based on fractal dimension: ; Where F represents the mesh structure complexity, and N(r) represents the number of collagen fiber meshes covering the scale r.

[0011] The present invention also provides a bovine blastocyst developmental potential assessment system based on label-free imaging and lipid metabolism characteristics. The system is used to implement the aforementioned method and includes: an image acquisition module, a data processing module, and an assessment decision module. The image acquisition module is used to perform laser label-free multimodal imaging of bovine blastocysts to acquire multimodal fused images including bovine blastocyst morphology, metabolomics, and matrix structure. The data processing module is used to quantitatively extract morphodynamic parameters, metabolic activity parameters, and matrix structure parameters from multimodal fused images and construct a multidimensional feature set. The evaluation decision module is used to train an evaluation model based on a machine learning algorithm using a multi-dimensional feature set, and to obtain a developmental potential score for the bovine blastocyst to be evaluated based on the trained evaluation model.

[0012] Preferably, the image acquisition module includes a signal acquisition unit and an image acquisition unit; The signal acquisition unit is used to perform three-dimensional scanning of bovine blastocysts using a femtosecond label-free scanning microscopy system, and simultaneously acquire three types of optical signals: second harmonic scattering, autofluorescence, and coherent anti-Stokes Raman scattering. The image acquisition unit is used to preprocess the three types of optical signals by wavelet denoising, spatial registration, and attention-weighted fusion to generate multimodal fusion images including bovine blastocyst morphology, metabolomics, and matrix structure.

[0013] Preferred methods for extracting morphodynamic parameters include: Blastocyst diameter was calculated by extracting the blastocyst contour using the Canny edge detection algorithm and then employing the minimum circumcircle method. ; Among them, D BL A is the diameter of the blastocyst. BL The area of ​​the blastocyst region; Inner cell mass area and shape factor: ; Among them, S ICM A is a factor relating the area and shape of the inner cell mass. ICM L represents the area of ​​the inner cell mass. ICM The perimeter of the inner cell mass outline; Trophoblast cell density: ; Among them, C TE For the density of trophoblast cell arrangement, A gap A represents the intercellular area of ​​the trophoblast. TE,total This represents the total area of ​​the nutrient layer.

[0014] Preferred methods for extracting metabolic activity parameters include: Average fluorescence intensity: ; Among them, I avg Let I(x,y) be the average fluorescence intensity, I(x,y) be the fluorescence value of pixel (x,y), and N be the fluorescence intensity. ROI The total number of pixels in the region of interest (ROI) is the total number of pixels in the ROI. Fluorescence lifetime was obtained by fitting and solving a double exponential decay model. ; Where τ1 and τ2 are fluorescence lifetime parameters, f(t) represents the function of fluorescence intensity changing with time, A1 is the amplitude coefficient of the first component, t represents time, and A2 is the amplitude coefficient of the second component, and satisfies A1+A2=1.

[0015] Preferably, the method for extracting matrix structure parameters includes: Collagen fiber orientation consistency is calculated using structural tensors: ; Where, λ 1,i , λ 2,i Let D be the eigenvalue of the structural tensor of the i-th collagen fiber. ori To ensure the uniformity of collagen fiber orientation, Indicates the total number of collagen fibers; Mesh structure complexity, calculated based on fractal dimension: ; Where F represents the mesh structure complexity, and N(r) represents the number of collagen fiber meshes covering the scale r.

[0016] Compared with the prior art, the beneficial effects of the present invention are as follows: 1. Pioneering a new paradigm of non-invasive and precise assessment: For the first time, laser label-free multimodal imaging technology is combined with big data-driven machine learning models to achieve quantitative, objective, and reproducible assessment of the developmental potential of bovine blastocysts, completely changing the traditional model that relies on the subjective experience of embryologists.

[0017] 2. A qualitative leap in prediction accuracy: Based on an image-pregnancy outcome pairing database of over 1000 blastocysts, the constructed evaluation model can uncover deep biological features invisible to the human eye. Backtesting has shown that this system's prediction accuracy for high-potential blastocysts is significantly higher than traditional morphological methods.

[0018] 3. Non-invasive and damage-free throughout the entire process: From imaging and feature extraction to evaluation and decision-making, no staining, fixation or physical contact with the embryo is required throughout the entire process. High-quality embryos after evaluation can be used for transplantation immediately, maximizing the viability and commercial value of the embryo.

[0019] 4. Deep integration of multimodal information brings disruptive insights: This invention breaks through the bottleneck of single morphological assessment. By fully integrating morphological and metabolic information, it has for the first time constructed a comprehensive evaluation system for blastocyst quality at the in vivo and three-dimensional levels, providing a brand-new scientific tool for understanding the laws of embryonic development. Attached Figure Description

[0020] To more clearly illustrate the technical solution of the present invention, the drawings used in the embodiments are 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.

[0021] Figure 1This is a schematic diagram of the method for assessing the developmental potential of bovine blastocysts based on label-free imaging and lipid metabolism characteristics, as described in an embodiment of the present invention. Figure 2 This is a schematic diagram of a bovine blastocyst developmental potential assessment system module based on label-free imaging and lipid metabolism characteristics, according to an embodiment of the present invention. Detailed Implementation

[0022] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0023] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0024] Example 1 like Figure 1 As shown, this invention provides a method for assessing the developmental potential of bovine blastocysts based on label-free imaging and lipid metabolism characteristics, comprising: Laser label-free multimodal imaging of bovine blastocysts was performed to obtain multimodal fusion images including bovine blastocyst morphology, metabolomics, and matrix structure. Quantitatively extract morphodynamic parameters, metabolic activity parameters, and matrix structure parameters from multimodal fused images to construct a multidimensional feature set; A multi-dimensional feature set was used to train an evaluation model based on a machine learning algorithm. Based on the trained evaluation model, the developmental potential score of the bovine blastocyst to be evaluated was obtained.

[0025] The specific implementation process of this invention is as follows: First, blastocysts required for assessing the developmental potential of bovine blastocysts are prepared. The specific process includes: In vitro fertilization (IVF) blastocyst preparation: Ovaries are obtained from slaughterhouses, and cumulus-oocyte complexes are aspirated from the follicles and cultured for maturation. The matured cumulus-oocyte complexes are then treated with hyaluronidase to remove granulosa cells. The oocytes, after granulosa cell removal, are then co-incubated with sperm. Successfully fertilized embryos are selected for in vitro culture to develop into IVF blastocysts.

[0026] Somatic cell nuclear transfer blastocyst preparation: Fibroblasts are isolated and cultured from bovine ear margin tissue to serve as donor cells for somatic cell nuclear transfer. The nucleus of the donor cell contains almost all the genetic material of the recipient animal, which is crucial for achieving genetic consistency in cloned animals using somatic cell nuclear transfer technology. The donor cell also plays a key role in the development and differentiation of the blastocyst. Ovaries are obtained from the slaughterhouse, and cumulus-oocyte complexes are aspirated from the follicles and matured. After maturation, granulosa cells in the cumulus-oocyte complexes are removed using hyaluronidase. The oocytes are then enucleated using a blind aspiration method with an oocyte-holding needle and an enucleation needle. A single donor cell is then aspirated using an injection needle and injected into the enucleated oocyte. Electrical stimulation is used to initiate the fusion and development of the reconstructed embryo. The activated reconstructed embryo is then transferred to embryo culture medium and allowed to develop into a blastocyst.

[0027] The bovine blastocysts prepared by the above-mentioned in vitro fertilization and somatic cell nuclear transfer together constitute the sample group of 1000 bovine blastocysts in the modeling and verification of the embodiments of the present invention.

[0028] Furthermore, methods for obtaining multimodal fusion images of bovine blastocysts, including bovine blastocyst morphology, metabolomics, and matrix structure, through laser label-free multimodal imaging include: Three-dimensional scanning of bovine blastocysts was performed using a femtosecond label-free scanning microscopy system, and three types of optical signals, namely second harmonic scattering, autofluorescence and coherent anti-Stokes Raman scattering, were acquired simultaneously. Preprocessing of three types of optical signals, including wavelet denoising, spatial registration, and attention-weighted fusion, generates multimodal fusion images encompassing bovine blastocyst morphology, metabolomics, and matrix structure.

[0029] Specifically, before transplantation, each bovine blastocyst was subjected to multi-channel signal acquisition using a commercial FI-100 femtosecond label-free scanning microscopy system placed in the sample carrier, generating high-resolution multimodal images.

[0030] The multi-channel signal acquisition module in the femtosecond label-free scanning microscopy system is configured with the following parameters: wavelength λ = 800 ± 10 nm, pulse width τ = 120 ± 5 fs, and repetition frequency f = 80 ± 2 MHz. It synchronously triggers the scanning surface control unit and the multi-channel signal detection array within the system. To ensure spatiotemporal synchronization, a single hardware trigger line is designed, with the main control module simultaneously sending start commands to both units, replacing the asynchronous transmission of software commands. A three-dimensional helical scanning trajectory is employed, using a piezoelectric drive platform (positioning accuracy ≤ 0.1 μm) to achieve full blastocyst coverage, simultaneously acquiring three types of optical signals: 1. Second harmonic signal (SHG): The detection wavelength λSHG=λ / 2 is separated by a bandpass filter module with a center wavelength of 400nm and a bandwidth of 10nm, reflecting the nonlinear optical characteristics of the matrix structure such as collagen fibers in the transparent zone; 2. Autofluorescence signal (AF): Acquired in the 450-650 nm band and analyzed by a 5 nm resolution spectral dispersive module, corresponding to the fluorescence distribution of intracellular lipids, mitochondria and other metabolic components; 3. Coherent anti-Stokes Raman scattering signal (CARS): frequency modulated by the difference between the pump light and the Stokes light (frequency difference Δν = 2850 cm⁻¹). -1 It specifically detects the vibrational modes of lipid molecules.

[0031] Preprocessing is performed on three types of optical signals (raw image signals): an adaptive wavelet thresholding denoising algorithm is used to suppress noise, and the threshold function T(·) is defined as: ; Where ω is the wavelet coefficient and σ is the noise standard deviation (estimated by the median absolute deviation method σ=1.4826×MAD, where MAD is the median absolute deviation).

[0032] Using the SHG image as a reference, a feature point matching algorithm is employed to spatially align the AF image and the CARS image. The registration transformation matrix M is optimized through least squares iterative optimization. ; in, , These are the feature point coordinates of the reference image and the image to be registered, respectively, with a registration error ≤ 1μm.

[0033] Attention-weighted fusion is performed using a multimodal feature image fusion network structure, mainly including: 1. Feature extraction layer with 3 parallel convolution branches (each layer contains 64 3×3 convolution kernels, stride 1, ReLU activation) to extract deep features from SHG, AF, and CARS images respectively; 2. Attention weight layer: The weight coefficient ω of each modality's deep features is calculated through a fully connected layer (128 nodes). i (satisfy ), to obtain weighted features, the weight update formula is: ω i =softmax(s i ), where s i Score the importance of features; 3. The output fusion layer compresses the dimensions of the weighted features through 1×1 convolution (32 kernels) to generate a high-resolution multimodal fusion image with a resolution of 2048×2048 pixels and a pixel size of 0.25μm×0.25μm, which includes bovine blastocyst morphology, metabolomics and matrix structure.

[0034] Furthermore, the imaged bovine blastocysts are transferred to surrogate cows in estrus at the same time for in vivo development. The pregnancy is confirmed by ultrasound 60 days post-transfer. This confirmation effectively avoids the need for continued high-cost pregnancy maintenance programs on non-pregnant cows, improving breeding efficiency, and simultaneously enabling real-time monitoring of the health status of both the surrogate cows and the bovine embryos.

[0035] Furthermore, morphodynamic parameters, metabolic activity parameters, and matrix structure parameters were quantitatively extracted from the multimodal fusion images to construct a multidimensional feature set, including: (1) morphodynamic parameters: blastocyst diameter, inner cell mass area and shape factor, trophoblast cell arrangement density, etc.; (2) metabolic activity parameters: average fluorescence intensity, fluorescence lifetime, and regional distribution of intracellular lipid imaging and mitochondrial imaging; (3) matrix structure parameters: density, orientation consistency, and mesh structure complexity of collagen fibers around the zona pellucida. The specific calculation method is as follows: (1) Morphological and dynamic parameters Blastocyst diameter, determined using the Canny edge detection algorithm (threshold T1=0.15I). max T2=0.3I max I max Extract the blastocyst contour (using the maximum grayscale value of the image) and calculate the blastocyst diameter using the least circumscribed circle method: ; Here, threshold T1 represents a lower threshold among all edge points found by the Canny edge detection algorithm, used to recover edge points connected to strong edge points but with slightly weaker intensity to form a complete contour; threshold T2 represents a higher threshold among all edge points found by the Canny edge detection algorithm, used to filter out the most certain and obvious edge points. And D... BL A is the diameter of the blastocyst. BL The area of ​​the blastocyst region; Inner cell mass area and shape factor: ; Among them, S ICM A is a factor relating the area and shape of the inner cell mass. ICM L represents the area of ​​the inner cell mass. ICM The perimeter of the inner cell mass outline; Trophoblast cell density: ; Among them, C TE For the density of trophoblast cell arrangement, A gap A represents the intercellular area of ​​the trophoblast. TE,total This represents the total area of ​​the nutrient layer.

[0036] (2) Metabolic activity parameters Average fluorescence intensity: ; Among them, I avg Let I(x,y) be the average fluorescence intensity, I(x,y) be the fluorescence value of pixel (x,y), and N be the fluorescence intensity. ROI The ROI is the total number of pixels in the region of interest. In image analysis, the ROI refers to a specific region that researchers define as the focus of their analysis based on their research objectives. Fluorescence lifetime was obtained by fitting and solving a double exponential decay model. ; Where τ1 and τ2 are fluorescence lifetime parameters, f(t) represents the function of fluorescence intensity changing with time, describing the decay law of fluorescence signal in the time dimension, A1 is the amplitude coefficient of the first component, representing the proportion of this component in the total fluorescence decay, and t represents time, i.e., the time variable in the fluorescence signal decay process. A2 is the amplitude coefficient of the second component, also representing the proportion (weight) of this component in the total fluorescence decay, and satisfies A1+A2=1.

[0037] (3) Matrix structural parameters Collagen fiber orientation consistency is calculated using structural tensors: ; Where, λ 1,i , λ 2,i Let λ be the eigenvalue of the structural tensor of the i-th collagen fiber. 1,i ≥λ 2,i ), D ori To ensure the uniformity of collagen fiber orientation, D ori The closer to 1, the more consistent the orientation. This represents the total number of collagen fibers analyzed, which is the sum of the number of collagen fibers involved in the calculation of collagen fiber orientation consistency. Mesh structure complexity, calculated based on fractal dimension: ; Where F represents the mesh structure complexity, and N(r) represents the number of collagen fiber meshes covering the scale r.

[0038] Furthermore, a multi-dimensional feature set is used to train an evaluation model constructed based on a machine learning algorithm. Based on the trained evaluation model, methods for obtaining the developmental potential score of the bovine blastocyst to be evaluated include: The multi-dimensional feature set constructed above is used to train and apply an evaluation model based on machine learning algorithms. This mainly includes three parts: data preprocessing, evaluation model construction, and model prediction and embryo screening. The detailed construction process is as follows: (1) Data preprocessing Z-score normalization is applied to the features in the multi-dimensional feature set: ; in, Here, x represents the standardized features, and μ represents the features in the multi-dimensional feature set. x σ is the characteristic mean. x The standard deviation of the features is used. Collinear features are screened by variance inflation factor (VIF<5), and finally the M-dimensional core features are retained.

[0039] (2) Evaluation model construction Autosampled training set generation: For a dataset D containing D embryo samples, construct T decision trees, each with a training set D. t Generated through sampling with replacement. The probability P of a single embryo not being selected in the single-tree training set. out for: ; Where N represents the number of times a single embryo is sampled with replacement; Unselected embryo samples are treated as out-of-bag (OOB) data and predicted using decision trees that were not included in the training. The prediction error rate of all out-of-bag samples is calculated, i.e., OOB error. OOB error can replace test set error and is used to determine the optimal number of decision trees. As the number of trees in the forest model increases, the OOB error will gradually decrease and tend to stabilize. When the error no longer decreases significantly, the optimal number of trees can be determined, avoiding overfitting or resource waste caused by too many trees, thereby achieving the evaluation of the model's generalization performance.

[0040] Decision tree node splitting strategy: For each node in each tree, M feature subsets are randomly selected from the multi-dimensional feature set, and the optimal splitting feature and splitting point are chosen based on Gini impurity. For a node dataset S containing two classes ("well-developed" / "poorly developed"), the Gini impurity is calculated as follows: Gini(S) = 1 - ; in, P K For category K The proportion of samples in the nodes.

[0041] The split gain is: ΔGini = Gini(S) - |S left ∣ / ∣S∣Gini(S left ) -∣S right ∣ / ∣S∣Gini(S right ); Among them, S left S represents the set of samples corresponding to the left child node generated after the decision tree splits. right This represents the set of samples corresponding to the right child node generated after the decision tree splits; Choose the splitting method with the largest ΔGini value to maximize the "purity" of the child nodes.

[0042] Model training termination condition: When the number of node samples is less than or equal to the minimum number of node samples or the tree depth reaches the maximum depth, the splitting stops and the leaf node outputs the prediction result with the highest category proportion.

[0043] (3) Model prediction and embryo screening For the new embryo sample x to be evaluated, the random forest generates the final prediction using a majority voting mechanism: ; Among them, h t (x) represents the predicted category of the new embryo sample x to be evaluated by the t-th tree (0 represents "poor development", 1 represents "good development"), t=1,2,...,T.

[0044] Simultaneously, the probability that the new embryo sample to be evaluated belongs to the "well-developed" category can be calculated as follows: ; Where I(·) is an indicator function (1 when the condition is met, 0 otherwise).

[0045] Embryos with good developmental potential are selected based on the predicted category and probability threshold.

[0046] In summary, this invention provides a non-invasive, objective, quantitative, and high-precision method for assessing the developmental potential of bovine blastocysts. By modeling and validating over 1000 bovine blastocysts, it achieves high-precision, non-invasive, and objective prediction of the in vivo developmental outcome of blastocysts before transplantation. This method can effectively screen blastocysts with high developmental potential before embryo transfer, thereby improving the pregnancy rate and calving rate.

[0047] This invention is the first to clearly define and verify the necessity and superiority of a "multi-dimensional feature set" that integrates morphological, metabolic, and matrix parameters for predicting the developmental capacity of bovine blastocysts. This invention employs a label-free embryo developmental capacity assessment technique, avoiding the series of damage problems caused by embryo staining markers, thus providing better protection for the tested embryos. This invention analyzes the impact of embryonic structure directly, avoiding the accuracy differences caused by indirect analysis through embryo nutrient solutions, thereby improving the accuracy of the detection.

[0048] Example 2 like Figure 2 As shown, based on the same inventive concept, the present invention also provides a bovine blastocyst developmental potential assessment system based on label-free imaging and lipid metabolism characteristics, used to implement the method described in the foregoing embodiments. The system includes: an image acquisition module, a data processing module, and an assessment decision module. The image acquisition module is used to perform laser label-free multimodal imaging of bovine blastocysts to acquire multimodal fused images including bovine blastocyst morphology, metabolomics, and matrix structure. The data processing module is used to quantitatively extract morphodynamic parameters, metabolic activity parameters, and matrix structure parameters from multimodal fused images and construct a multidimensional feature set. The evaluation decision module is used to train an evaluation model based on a machine learning algorithm using a multi-dimensional feature set, and to obtain a developmental potential score for the bovine blastocyst to be evaluated based on the trained evaluation model.

[0049] Furthermore, the image acquisition module includes a signal acquisition unit and an image acquisition unit; The signal acquisition unit is used to perform three-dimensional scanning of bovine blastocysts using a femtosecond label-free scanning microscopy system, and simultaneously acquire three types of optical signals: second harmonic scattering, autofluorescence, and coherent anti-Stokes Raman scattering. The image acquisition unit is used to preprocess the three types of optical signals by wavelet denoising, spatial registration, and attention-weighted fusion to generate multimodal fusion images including bovine blastocyst morphology, metabolomics, and matrix structure.

[0050] Furthermore, methods for extracting morphological dynamic parameters include: Blastocyst diameter was calculated by extracting the blastocyst contour using the Canny edge detection algorithm and then employing the minimum circumcircle method. ; Among them, D BL A is the diameter of the blastocyst. BL The area of ​​the blastocyst region; Inner cell mass area and shape factor: ; Among them, S ICM A is a factor relating the area and shape of the inner cell mass. ICM L represents the area of ​​the inner cell mass. ICM The perimeter of the inner cell mass outline; Trophoblast cell density: ; Among them, C TE For the density of trophoblast cell arrangement, A gap A represents the intercellular area of ​​the trophoblast. TE,total This represents the total area of ​​the nutrient layer.

[0051] Furthermore, methods for extracting metabolic activity parameters include: Average fluorescence intensity: ; Among them, I avg Let I(x,y) be the average fluorescence intensity, I(x,y) be the fluorescence value of pixel (x,y), and N be the fluorescence intensity. ROI The ROI is the total number of pixels in the region of interest. In image analysis, the ROI refers to a specific region that researchers define as the focus of their analysis based on their research objectives. Fluorescence lifetime was obtained by fitting and solving a double exponential decay model. ; Where τ1 and τ2 are fluorescence lifetime parameters, f(t) represents the fluorescence intensity as a function of time, describing the decay law of the fluorescence signal in the time dimension, A1 is the amplitude coefficient of the first component, representing the proportion of this component in the total fluorescence decay, and t represents time, i.e., the time variable in the fluorescence signal decay process. A2 is the amplitude coefficient of the second component, which also represents the proportion (weight) of this component in the total fluorescence decay, and satisfies A1+A2=1.

[0052] Furthermore, methods for extracting matrix structure parameters include: Collagen fiber orientation consistency is calculated using structural tensors: ; Where, λ 1,i , λ 2,i Let D be the eigenvalue of the structural tensor of the i-th collagen fiber. ori To ensure the uniformity of collagen fiber orientation, D ori The closer to 1, the more consistent the orientation. This represents the total number of collagen fibers analyzed, which is the sum of the number of collagen fibers involved in the calculation of collagen fiber orientation consistency. Mesh structure complexity, calculated based on fractal dimension: ; Where F represents the mesh structure complexity, and N(r) represents the number of collagen fiber meshes covering the scale r.

[0053] The embodiments described above are merely preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Various modifications and improvements made to the technical solutions of the present invention by those skilled in the art without departing from the spirit of the present invention should fall within the protection scope defined by the claims of the present invention.

Claims

1. A method for assessing the developmental potential of bovine blastocysts based on label-free imaging and lipid metabolism characteristics, characterized in that, The method includes: Laser label-free multimodal imaging of bovine blastocysts was performed to obtain multimodal fusion images including bovine blastocyst morphology, metabolomics, and matrix structure. Quantitatively extract morphodynamic parameters, metabolic activity parameters, and matrix structure parameters from multimodal fused images to construct a multidimensional feature set; A multi-dimensional feature set was used to train an evaluation model based on a machine learning algorithm. Based on the trained evaluation model, the developmental potential score of the bovine blastocyst to be evaluated was obtained.

2. The method according to claim 1, characterized in that, Methods for obtaining multimodal fusion images of bovine blastocysts, including morphology, metabolomics, and matrix structure, through label-free laser imaging include: Three-dimensional scanning of bovine blastocysts was performed using a femtosecond label-free scanning microscopy system, and three types of optical signals, namely second harmonic scattering, autofluorescence and coherent anti-Stokes Raman scattering, were acquired simultaneously. Preprocessing of three types of optical signals, including wavelet denoising, spatial registration, and attention-weighted fusion, generates multimodal fusion images encompassing bovine blastocyst morphology, metabolomics, and matrix structure.

3. The method according to claim 1, characterized in that, Methods for extracting morphological dynamic parameters include: Blastocyst diameter was calculated by extracting the blastocyst contour using the Canny edge detection algorithm and then employing the minimum circumcircle method. ； Among them, D BL A is the diameter of the blastocyst. BL Area of ​​the blastocyst region; Inner cell mass area and shape factor: ； Among them, S ICM A is a factor relating the area and shape of the inner cell mass. ICM L represents the area of ​​the inner cell mass. ICM The perimeter of the inner cell mass outline; Trophoblast cell density: ； Among them, C TE For the density of trophoblast cell arrangement, A gap A represents the intercellular area of ​​the trophoblast. TE,total This represents the total area of ​​the nutrient layer.

4. The method according to claim 1, characterized in that, Methods for extracting metabolic activity parameters include: Average fluorescence intensity: ； Among them, I avg Let I(x,y) be the average fluorescence intensity, I(x,y) be the fluorescence value of pixel (x,y), and N be the fluorescence intensity. ROI The total number of pixels in the region of interest (ROI) is the total number of pixels in the ROI. Fluorescence lifetime was obtained by fitting and solving a double exponential decay model. ； Where τ1 and τ2 are fluorescence lifetime parameters, f(t) represents the function of fluorescence intensity changing with time, A1 is the amplitude coefficient of the first component, t represents time, and A2 is the amplitude coefficient of the second component, and satisfies A1+A2=1.

5. The method according to claim 1, characterized in that, Methods for extracting matrix structure parameters include: Collagen fiber orientation consistency is calculated using structural tensors: ； Where, λ 1,i , λ 2,i Let D be the eigenvalue of the structural tensor of the i-th collagen fiber. ori To ensure the uniformity of collagen fiber orientation, Indicates the total number of collagen fibers; Mesh structure complexity, calculated based on fractal dimension: ； Where F represents the mesh structure complexity, and N(r) represents the number of collagen fiber meshes covering the scale r.

6. A bovine blastocyst developmental potential assessment system based on label-free imaging and lipid metabolism characteristics, said system being used to implement the method described in any one of claims 1-5, characterized in that, The system includes: an image acquisition module, a data processing module, and an evaluation and decision-making module; The image acquisition module is used to perform laser label-free multimodal imaging of bovine blastocysts to acquire multimodal fused images including bovine blastocyst morphology, metabolomics, and matrix structure. The data processing module is used to quantitatively extract morphodynamic parameters, metabolic activity parameters, and matrix structure parameters from multimodal fused images and construct a multidimensional feature set. The evaluation decision module is used to train an evaluation model based on a machine learning algorithm using a multi-dimensional feature set, and to obtain a developmental potential score for the bovine blastocyst to be evaluated based on the trained evaluation model.

7. The system according to claim 6, characterized in that, The image acquisition module includes a signal acquisition unit and an image acquisition unit; The signal acquisition unit is used to perform three-dimensional scanning of bovine blastocysts using a femtosecond label-free scanning microscopy system, and simultaneously acquire three types of optical signals: second harmonic scattering, autofluorescence, and coherent anti-Stokes Raman scattering. The image acquisition unit is used to preprocess the three types of optical signals by wavelet denoising, spatial registration, and attention-weighted fusion to generate multimodal fusion images including bovine blastocyst morphology, metabolomics, and matrix structure.

8. The system according to claim 6, characterized in that, Methods for extracting morphological dynamic parameters include: Blastocyst diameter was calculated by extracting the blastocyst contour using the Canny edge detection algorithm and then employing the minimum circumcircle method. ； Among them, D BL A is the diameter of the blastocyst. BL Area of ​​the blastocyst region; Inner cell mass area and shape factor: ； Among them, S ICM A is a factor relating the area and shape of the inner cell mass. ICM L represents the area of ​​the inner cell mass. ICM The perimeter of the inner cell mass outline; Trophoblast cell density: ； Among them, C TE For the density of trophoblast cell arrangement, A gap A represents the intercellular area of ​​the trophoblast. TE,total This represents the total area of ​​the nutrient layer.

9. The system according to claim 6, characterized in that, Methods for extracting metabolic activity parameters include: Average fluorescence intensity: ； Among them, I avg Let I(x,y) be the average fluorescence intensity, I(x,y) be the fluorescence value of pixel (x,y), and N be the fluorescence intensity. ROI The total number of pixels in the region of interest (ROI) is the total number of pixels in the ROI. Fluorescence lifetime was obtained by fitting and solving a double exponential decay model. ； Where τ1 and τ2 are fluorescence lifetime parameters, f(t) represents the function of fluorescence intensity changing with time, A1 is the amplitude coefficient of the first component, t represents time, and A2 is the amplitude coefficient of the second component, and satisfies A1+A2=1.

10. The system according to claim 6, characterized in that, Methods for extracting matrix structure parameters include: Collagen fiber orientation consistency is calculated using structural tensors: ； Where, λ 1,i , λ 2,i Let D be the eigenvalue of the structural tensor of the i-th collagen fiber. ori To ensure the uniformity of collagen fiber orientation, Indicates the total number of collagen fibers; Mesh structure complexity, calculated based on fractal dimension: ； Where F represents the mesh structure complexity, and N(r) represents the number of collagen fiber meshes covering the scale r.

Images