A stem cell quality evaluation system and method

Abstract

The application relates to the field of stem cell biotechnology and discloses a stem cell quality evaluation system and method, a stem cell quality evaluation system, which comprises a data acquisition module, which acquires local environment data and image data of stem cells within a preset range; a morphology identification module, which obtains cell morphology data of all stem cells within the preset range through a cell morphology identification model; a graph structure data construction module, which constructs graph structure data; a scoring and labeling module, which determines the proliferation capacity, genetic stability capacity and surface marker expression capacity of stem cells after a preset time length, and obtains the quality score of the stem cells through artificial labeling; and a model training module, which trains a stem cell quality scoring model through an optimization algorithm; the cell morphology data of the stem cells are extracted through a multi-branch network of the cell morphology identification model, and the interaction between the stem cells is comprehensively considered through the stem cell quality scoring model, so that the accuracy of stem cell quality evaluation is improved.

Description

Technical Field

[0001] The present invention relates to the field of stem cell biotechnology, and more particularly to a stem cell quality evaluation system and method. Background Art

[0002] Stem cells are a special type of cell with the ability to self-renew and multidirectionally differentiate. They hold broad application prospects in regenerative medicine, tissue engineering, and cell therapy. To ensure their safety and efficacy in clinical and industrial applications, stem cell quality assessment is a critical step. Traditional stem cell quality assessment relies primarily on a series of experimental assays, including but not limited to measuring cell proliferation capacity (e.g., population doubling time), genetic stability (e.g., chromosome karyotype analysis), and surface marker expression (e.g., CD molecule detection by flow cytometry). While these methods can reflect the functional status of stem cells, they are limited by long testing cycles, high experimental costs, and high technical barriers, making real-time monitoring of stem cell status difficult.

[0003] With the development of artificial intelligence, stem cell quality assessment is now performed through intelligent image analysis. For example, high-resolution microscopy is used to collect cell image data, and techniques such as multi-scale convolution and attention mechanisms are used to extract image features from this data. These features are then converted into stem cell quality scores using nonlinear mapping methods (such as multi-layer perceptrons and support vector machines). However, these methods generally analyze individual stem cells as isolated entities, ignoring interactions between stem cells, such as paracrine effects (stem cells secrete cytokines that influence the differentiation direction and proliferation rate of neighboring cells) and cell contact regulation (cell contact also affects the proliferation rate of neighboring cells). This leads to biased stem cell quality assessments. Summary of the Invention

[0004] The present invention provides a stem cell quality evaluation system and method to solve the technical problems in the above-mentioned background technology.

[0005] The present invention provides a stem cell quality evaluation system, comprising:

[0006] A data acquisition module, which is used to collect local environmental data and image data of stem cells within a preset range;

[0007] Local environmental data include: pH, temperature, nutrient concentration, and number of stem cells within a preset range;

[0008] A morphology recognition module, which is used to obtain cell morphology data of all stem cells within a preset range through a cell morphology recognition model;

[0009] The cell morphology data includes cell area, aspect ratio, vacuole rate, edge sharpness and whether adherent;

[0010] a graph structure data construction module configured to construct graph structure data according to cell morphology data of all stem cells within a preset range;

[0011] The graph structure data is composed of nodes and edges between the nodes;

[0012] The nodes include core nodes and neighbor nodes, the core nodes are represented by cell morphology data of stem cells at the center of the preset range, and the neighbor nodes are represented by cell morphology data of other stem cells;

[0013] The construction of the edges between the nodes must satisfy a preset condition;

[0014] a score labeling module configured to determine the proliferation ability, genetic stability ability and surface marker expression ability of the stem cells after a preset time length, and obtain a quality score of the stem cells through artificial labeling;

[0015] a model training module configured to use the graph structure data of the stem cells and local environment data within the preset range as sample data, use the quality score of the stem cells after the preset time length as a sample label, and train a stem cell quality score model by using the sample data and the sample label through an optimization algorithm.

[0016] Further, the preset time length is a self-defined parameter, the preset range represents a circular region generated with the stem cell as the center and a preset distance as the radius, and the preset distance is a self-defined parameter.

[0017] Further, the preset condition includes that the Euclidean distance between the corresponding stem cells is less than a distance threshold, and the correlation coefficient between the corresponding stem cells is greater than or equal to a coefficient threshold, wherein the distance threshold and the coefficient threshold are self-defined parameters.

[0018] Further, the cell morphology recognition model is composed of five branches with the same structure but not sharing weight parameters, the input of the five branches is image data of the stem cell, and the output of the five branches is cell area, aspect ratio, vacuole rate, edge sharpness and whether adherent respectively;

[0019] Each branch is composed of a region division layer, a linear projection layer, a position encoding layer, a feature extraction layer and a first classifier;

[0020] The region division layer is configured to divide the image data into N square regions, and each square region is represented by a first vector with a dimension of P square C, wherein N = H * W / P square, H, W and C represent the height, width and channel number of the image data respectively, and P represents the side length of the square region, and P is a self-defined parameter;

[0021] The linear projection layer is used to convert each square region into a second vector representation with a dimension number of D, where D is a self-defined parameter, and the calculation formula of the linear projection layer is as follows:

[0022]

[0023] wherein and respectively represent the second vector and the first vector of the i-th square region, represents a weight matrix, represents a bias vector, and e represents a natural constant;

[0024] The position encoding layer is used to embed a position vector into the second vector of each square region to obtain a third vector, and the calculation formula of the position vector is as follows:

[0025]

[0026] wherein L represents a maximum value of a self-increasing order, and D is a self-defined parameter, represents two dimension values of the k-th self-increasing order of the position vector of the i-th square region, p represents a normalized position number, and norm represents a Min-Max normalization method;

[0027] The feature extraction layer is used to perform feature extraction on a sequence composed of the third vectors of the N square regions to obtain a fourth vector, and the fourth vector is taken as an input of the first classifier.

[0028] Further, the calculation formula of the extraction layer includes:

[0029]

[0030]

[0031]

[0032] wherein represents the fourth vector output by the feature extraction layer, and M represents an attention matrix, and respectively represent a weight matrix and a bias vector corresponding to the attention matrix, Q, K and V respectively represent a query matrix, a key matrix and a value matrix, which are obtained by multiplying a sequence composed of the third vectors of the N square regions by a corresponding weight matrix, and the sizes are all NxA, wherein A is a self-defined parameter, represents a mask matrix obtained by performing a mask operation on the query matrix and the key matrix, , and ​​​​​respectively represent the element value of the nth row and the a-th column of the mask matrix, the query matrix and the key matrix, hash represents a hash function, T represents a transpose operation, represents point-by-point multiplication, Swish represents a Swish activation function, and softmax represents a softmax activation function.

[0033] Further, the number of newly proliferated cells obtained by cell counting is used as the proliferation ability, the number of newly proliferated cells with normal chromosomes obtained by karyotype analysis is used as the genetic stability ability, and the number of newly proliferated cells with positive flow cytometry results is used as the surface marker expression ability.

[0034] Further, the stem cell quality score model comprises a graph structure data analysis layer, a splicing layer and a second classifier.

[0035] The graph structure data analysis layer is configured to update the graph structure data of the stem cell.

[0036] The splicing layer is configured to splice and normalize the vector of the updated core node with the local environment data within a preset range to obtain a combined vector, and the number of dimensions of the combined vector is a custom parameter.

[0037] The second classifier inputs the combined vector, and the class space of the second classifier represents the quality score of the stem cell.

[0038] The graph structure data analysis layer is constructed based on a graph attention network model.

[0039] Further, the stem cell quality score model is trained by an optimization algorithm, comprising the following steps:

[0040] In step S201, the parameters of the stem cell quality score model are randomly generated as the encoding of the individuals of the initialization population.

[0041] In step S202, the loss values of all individuals in the initialization population are calculated by a loss function.

[0042] The calculation formula of the loss function is the mean square error between the value output by the stem cell quality score model using the parameters corresponding to the encoding of the individual and the sample label.

[0043] In step S203, a non-linear adjustment factor is generated according to the current iteration number and the maximum iteration number, and it is determined whether the non-linear adjustment factor is less than a preset factor threshold. If yes, step S204 is entered, otherwise step S205 is entered.

[0044] The calculation formula of the non-linear adjustment factor Factor is as follows:

[0045] ;

[0046] where t represents the current iteration number t. represents the maximum iteration number, and a preset factor threshold is a user-defined parameter;

[0047] In step S204, the current iteration number is incremented by 1, and the encoding of all individuals in the initialization population is updated by using the first updating strategy.

[0048] In step S205, the current iteration number is incremented by 1, and the encoding of all individuals in the initialization population is updated by using the second updating strategy.

[0049] In step S206, it is determined whether the current iteration number is greater than or equal to the maximum iteration number or the minimum value of the loss values of the individuals in the initialization population is less than or equal to a preset loss threshold, and if so, the iteration is terminated, the encoding of the individual with the minimum loss value is taken as the parameter of the stem cell quality score model, otherwise, the step S202 is returned to continue execution.

[0050] The maximum iteration number and the preset loss threshold are both user-defined parameters.

[0051] Further, the calculation formula of the optimization algorithm comprises:

[0052] The calculation formula of the first updating strategy comprises:

[0053] ;

[0054] ;

[0055] wherein 1≤g≤G, 1≤t≤ , and respectively represent the encoding of the gth individual at the current iteration number t and t+1, represents the encoding of a random individual at the current iteration number t, and R represents a correlation coefficient, represents a first random number with a value range of 0 to 1;

[0056] The calculation formula of the second updating strategy is as follows:

[0057] ;

[0058] wherein represents the encoding of the individual with the minimum loss value in the initialization population at the current iteration number t, represents a second random number with a value range of 0 to 1.

[0059] The present application provides a stem cell quality evaluation method, comprising the following steps:

[0060] In step S301, local environment data and image data of stem cells in a preset range are collected.

[0061] Step S302, obtaining cell morphology data of all stem cells in the preset range through the cell morphology recognition model;

[0062] Step S303, constructing graph structure data according to the cell morphology data of all stem cells in the preset range;

[0063] Step S304, measuring the proliferation ability, genetic stability ability and surface marker expression ability of the stem cells after a preset time length, and obtaining the quality score of the stem cells through artificial labeling;

[0064] Step S305, taking the graph structure data of the stem cells and the local environment data in the preset range as sample data, taking the quality score of the stem cells after the preset time length as sample label, and training the stem cell quality score model by using the sample data and the sample label through the optimization algorithm.

[0065] The beneficial effects of the present application are that the present application extracts the cell morphology data of the stem cells through the multi-branch network of the cell morphology recognition model, and comprehensively considers the interaction between the stem cells through the stem cell quality score model, thereby improving the accuracy of the stem cell quality evaluation. BRIEF DESCRIPTION OF DRAWINGS

[0066] Fig. 1 is a schematic diagram of a stem cell quality evaluation system of the present application;

[0067] Fig. 2 is a flowchart of training the stem cell quality score model by the optimization algorithm of the present application;

[0068] Fig. 3 is a flowchart of a stem cell quality evaluation method of the present application.

[0069] In the figure: data acquisition module 101, morphology recognition module 102, graph structure data construction module 103, score labeling module 104, model training module 105. DETAILED DESCRIPTION

[0070] The subject matter described herein will now be discussed with reference to example implementations. It should be understood that the discussion of these implementations is merely meant to provide a better understanding of the subject matter described herein and can be changed in function and arrangement without departing from the scope of the present description. Each of the various examples can omit, substitute or add various procedures or components as appropriate. In addition, features described in relation to some examples can also be combined in other examples.

[0071] It should be noted that the technical terms or scientific terms used in one or more embodiments of the present application should be understood as the general meaning understood by those skilled in the art in the field of the present application, unless otherwise defined. The terms "first", "second", and the like used in one or more embodiments of the present application do not represent any order, number, or importance, but are only used to distinguish different components. The terms "include" or "contain" and the like mean that the elements or objects appearing before the terms encompass the elements or objects listed after the terms and their equivalents, and do not exclude other elements or objects. The terms "connected" or "connected" and the like are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect. The terms "up", "down", "left", "right", and the like are only used to represent relative positional relationships, and when the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0072] As shown in Figs. 1-3 , a stem cell quality evaluation system comprises:

[0073] A data acquisition module 101 is configured to acquire local environment data and image data of stem cells within a preset range;

[0074] The local environment data includes pH value, temperature, nutrient concentration, and the number of stem cells within the preset range;

[0075] A morphology recognition module 102 is configured to obtain cell morphology data of all stem cells within the preset range through a cell morphology recognition model;

[0076] The cell morphology data includes cell area, aspect ratio, vacuole rate, edge definition, and whether adherent;

[0077] A graph structure data construction module 103 is configured to construct graph structure data according to the cell morphology data of all stem cells within the preset range;

[0078] The graph structure data is composed of nodes and edges between nodes;

[0079] The nodes include core nodes and neighbor nodes, the core nodes are represented by the cell morphology data of the stem cells at the center of the preset range, and the neighbor nodes are represented by the cell morphology data of other stem cells;

[0080] The construction of the edges between the nodes must meet the preset conditions;

[0081] A score annotation module 104 is configured to determine the proliferation capacity, genetic stability capacity, and surface marker expression capacity of stem cells after a preset time, and obtain the quality score of stem cells through artificial annotation;

[0082] a model training module 105, configured to take the graph structure data of the stem cells and the local environment data within a preset range as sample data, take the quality score of the stem cells after a preset time length as sample label, and train the stem cell quality score model by an optimization algorithm using the sample data and the sample label.

[0083] In an embodiment of the present application, the preset time length is a custom parameter, for example, the preset time length is set to 48 hours, and the preset range represents a circular region with the stem cell as the center and a preset distance as the radius, wherein the preset distance is a custom parameter, for example, the preset distance is set to 200 microns.

[0084] In an embodiment of the present application, the preset condition includes: the Euclidean distance between the corresponding stem cells is less than a distance threshold, wherein the distance threshold is a custom parameter, for example, the distance threshold is set to 50 microns; and the correlation coefficient between the corresponding stem cells is greater than or equal to a coefficient threshold, wherein the coefficient threshold is a custom parameter, for example, the coefficient threshold is set to 0.7, and the correlation coefficient can be any one of Pearson correlation coefficient, Spearman correlation coefficient or cosine similarity.

[0085] In an embodiment of the present application, the cell morphology recognition model is composed of 5 branches with the same structure but not sharing weight parameters, and the input of the 5 branches is the image data of the stem cells, and the output is the cell area, the aspect ratio, the vacuole rate, the edge definition and whether adherent, respectively.

[0086] Each branch is composed of a region division layer, a linear projection layer, a position encoding layer, a feature extraction layer and a first classifier.

[0087] The region division layer is configured to divide the image data into N square regions, and each square region is represented by a first vector with a dimension number of P²C, wherein N=H×W / P², H, W and C represent the height, width and channel number of the image data, respectively, and P represents the side length of the square region, and P is a custom parameter, for example, P is set to H / 16.

[0088] It should be noted that the image data of the stem cells can also be preprocessed before being input into the cell morphology recognition model, such as denoising, uniform size, etc.

[0089] The linear projection layer is configured to convert each square region into a second vector with a dimension number of D, wherein D is a custom parameter, for example, D is set to 16, and the calculation formula of the linear projection layer is as follows:

[0090] ;

[0091] wherein and represent the second vector and the first vector of the i-th square region, respectively. denotes a weight matrix, has a size of P 2 C x D, denotes a bias vector, has a dimension number of D, and e denotes a natural constant;

[0092] The position encoding layer is used to obtain a third vector from a second vector embedding position vector of each square region, and the calculation formula of the position vector is as follows:

[0093] ;

[0094] wherein L denotes a maximum value of the self-increasing order, and is a self-defined parameter, for example, L is set to 8, denotes a 2-dimension value of the kth self-increasing order of the position vector of the ith square region, p denotes a normalized position number, norm denotes a Min-Max normalization method, the position vector has a dimension number of 2L, and the third vector has a dimension number of D+2L;

[0095] It should be noted that the position vector can also be obtained by fixed sine cosine position encoding, and the specific calculation formula includes:

[0096] ;

[0097] ;

[0098] wherein 1≤i≤N, 1≤j≤D, and denote the 2j and 2j+1 dimension values of the position vector of the ith square region, respectively, and D denotes a dimension number of the second vector, that is, the dimension numbers of the position vector and the second vector are the same;

[0099] The feature extraction layer is used to obtain a fourth vector by performing feature extraction on a sequence composed of the third vectors of the N square regions, and the fourth vector is taken as an input of the first classifier.

[0100] It should be noted that the first classifiers of the above-mentioned 5 branches are all constructed based on a multilayer perceptron, the activation function of the first classifier whose category space is the cell area, the length-width ratio, the vacuole rate and the edge definition is a PReLU activation function, and the activation function of the first classifier whose category space is whether adherent is a Sigmoid activation function, which will not be repeated here.

[0101] It should be noted that the cell area is related to the proliferation ability of the cell; the length-width ratio reflects the morphological symmetry of the cell, normal stem cells usually maintain symmetrical morphology, and irregular length-width ratio may mean that the stem cell is differentiated or in a stress state; the vacuole rate represents the proportion of transparent areas of the cell, and a high vacuole rate may indicate that the cell is under certain physiological stress, such as insufficient nutrition, oxidative stress, etc., which has a certain influence on the health and differentiation ability of the cell; the edge definition reflects the integrity and structural health of the cell membrane; the adhesion ability of the stem cell reflects its vitality and proliferation state, and generally healthy stem cells can adhere to the culture substrate.

[0102] In an embodiment of the present application, the calculation formula of the feature extraction layer comprises:

[0103] ;

[0104] wherein represents a fourth vector output by the feature extraction layer, M represents an attention matrix with a size of NxA, and respectively represent a weight matrix and a bias vector corresponding to the attention matrix, has a size of 1xN, has a dimension number of A, then the fourth vector has a dimension number of A, and Swish represents a Swish activation function;

[0105] ;

[0106] ;

[0107] wherein Q, K and V respectively represent a query matrix, a key matrix and a value matrix, are all obtained by multiplying a sequence (NxD) of N square areas by a corresponding weight matrix, and all have a size of NxA, that is, the size of the weight matrix is DxA, wherein A is a self-defined parameter, for example, A is set to 16, represents that a mask operation is performed on the query matrix and the key matrix to obtain a mask matrix, that is, the size of the mask matrix is NxA, , and respectively represent the element value of the nth row and the a th column of the mask matrix, the query matrix and the key matrix, hash represents a hash function, T represents a transpose operation, represents point-by-point multiplication, and softmax represents a softmax activation function.

[0108] It should be noted that the parameters (weight matrix, bias vector) in the cell morphology recognition model are all learnable superparameters, and the sample labels used to train the cell morphology recognition model are obtained by manual annotation. In the training process, the difference between the sample label and the value output by the cell morphology recognition model can be specified as a loss function, and a gradient optimizer (such as AdaGrad, RMSProp, etc.) is used to update the parameters in the cell morphology recognition model in reverse, so that the loss is minimized, and the convergence of the model is completed. In addition, the addition of the mask operation is mainly for information screening and reducing data calculation.

[0109] In an embodiment of the present application, the number of newly proliferated cells obtained by cell counting method is used as the proliferation ability, the number of newly proliferated cells with normal chromosomes obtained by karyotype analysis is used as the genetic stability ability, and the number of newly proliferated cells with positive flow cytometry results is used as the surface marker expression ability.

[0110] It should be noted that in addition to manual annotation, automatic annotation can also be achieved through normalization and weighting operations. The corresponding weighting coefficients can be set by humans. In addition, the flow cytometry results obtained by flow cytometry are related to CD molecule types. Positive indicates that the stem cell expresses a marker (CD molecule), and negative indicates that the stem cell does not express a marker. For example, the positive markers of mesenchymal stem cells include CD73 and CD90, while CD34 and CD45 are negatively expressed. If CD90 is positive, it indicates that the stem cell quality is good, and if CD45 is positive, it indicates that the stem cell quality is abnormal. This will not be repeated here.

[0111] In an embodiment of the present application, the stem cell quality scoring model is composed of a graph structure data analysis layer, a splicing layer and a second classifier;

[0112] The graph structure data analysis layer is used to update the graph structure data of the stem cell;

[0113] The splicing layer is used to splice and normalize the vector of the updated core node with the local environment data within the preset range to obtain a combined vector. The number of dimensions of the combined vector is a self-defined parameter. For example, the number of dimensions of the combined vector is set to 32.

[0114] The second classifier inputs the combined vector, and the class space of the second classifier represents the quality score of the stem cell;

[0115] It should be noted that the activation function of the second classifier can also be designed as a PReLU activation function, and the training method of the stem cell quality scoring model and the cell morphology recognition model is the same.

[0116] In an embodiment of the present application, the graph structure data analysis layer is constructed based on a GAT (graph attention network) model, and can also be constructed based on a GCN (graph convolution network) model, which is not described herein.

[0117] In an embodiment of the present application, as shown in Fig. 2 The stem cell quality score model is trained by an optimization algorithm, including the following steps:

[0118] Step S201, randomly generating parameters of the stem cell quality score model as encodings of individuals of an initialization population;

[0119] Step S202, calculating loss values of all individuals in the initialization population by a loss function;

[0120] The calculation formula of the loss function is the mean square error between the value output by the stem cell quality score model using the parameters corresponding to the encodings of the individuals and the sample label;

[0121] Step S203, generating a non-linear adjustment factor according to the current iteration number and the maximum iteration number, and determining whether the non-linear adjustment factor is less than a preset factor threshold value, if yes, entering step S204, otherwise entering step S205;

[0122] The calculation formula of the non-linear adjustment factor Factor is as follows:

[0123] ;

[0124] where t represents the current iteration number t, max represents the maximum iteration number, and the preset factor threshold value is a self-defined parameter, for example, the preset factor threshold value is set to 1;

[0125] Step S204, increasing the current iteration number by 1, and updating the encodings of all individuals in the initialization population by a first updating strategy;

[0126] Step S205, increasing the current iteration number by 1, and updating the encodings of all individuals in the initialization population by a second updating strategy;

[0127] Step S206, determining whether the current iteration number is greater than or equal to the maximum iteration number or the minimum value of the loss values of the individuals in the initialization population is less than or equal to a preset loss threshold value, if yes, terminating the iteration, and taking the encoding of the individual with the minimum loss value as the parameters of the stem cell quality score model, otherwise returning to step S202 for continuous execution;

[0128] The maximum iteration number and the preset loss threshold value are both self-defined parameters, for example, the maximum iteration number is set to 100, and the preset loss threshold value is set to 0.01.

[0129] In an embodiment of the present application, the calculation formula of the optimization algorithm includes:

[0130] The calculation formula of the first updating strategy comprises:

[0131]

[0132]

[0133] wherein 1≤g≤G, 1≤t≤T, and T represents the total number of iterations, and respectively represent the code of the gth individual at the current iteration number t and t+1, represent the code of a random individual at the current iteration number t, and R represents a correlation coefficient, represent a first random number with a value range of 0 to 1;

[0134] The calculation formula of the second updating strategy is as follows:

[0135]

[0136] wherein represent the code of an individual with the minimum loss value in the initialized population at the current iteration number t, represent a second random number with a value range of 0 to 1.

[0137] In an embodiment of the present application, as shown in Fig. 3 a stem cell quality evaluation method comprises the following steps:

[0138] Step S301, collecting local environment data and image data of stem cells in a preset range;

[0139] Step S302, obtaining cell morphology data of all stem cells in the preset range through a cell morphology recognition model;

[0140] Step S303, constructing graph structure data according to the cell morphology data of all stem cells in the preset range;

[0141] Step S304, measuring the proliferation capacity, genetic stability capacity and surface marker expression capacity of stem cells after a preset time length, and obtaining the quality score of stem cells through artificial labeling;

[0142] Step S305, taking the graph structure data of stem cells and the local environment data in the preset range as sample data, taking the quality score of stem cells after the preset time length as a sample label, and training a stem cell quality score model through an optimization algorithm by using the sample data and the sample label.

[0143] ​​​​The above describes the embodiments of the present embodiment, but the present embodiment is not limited to the above-described specific embodiments, and the above-described specific embodiments are only illustrative but not restrictive, and those skilled in the art can make many forms under the inspiration of the present embodiment, which all belong to the protection of the present embodiment.

Claims

1. A stem cell quality evaluation system, characterized by comprising: The method comprises the following steps: a data acquisition module is used to acquire local environment data and image data of stem cells within a preset range; the local environment data comprises pH value, temperature, nutrient concentration and the number of stem cells within the preset range; a morphology recognition module is used to obtain cell morphology data of all stem cells within the preset range through a cell morphology recognition model; the cell morphology data comprises cell area, aspect ratio, vacuole rate, edge definition and whether adherent; a graph structure data construction module is used to construct graph structure data according to the cell morphology data of all stem cells within the preset range; the graph structure data is composed of nodes and edges between the nodes; the nodes comprise core nodes and neighbor nodes, the core nodes are represented by the cell morphology data of stem cells at the center of the preset range, and the neighbor nodes are represented by the cell morphology data of other stem cells; the edges between the nodes must satisfy a preset condition; a score labeling module is used to determine the proliferation ability, genetic stability and surface marker expression ability of stem cells after a preset time, and obtain the quality score of the stem cells through artificial labeling; a model training module is used to take the graph structure data of the stem cells and the local environment data within the preset range as sample data, take the quality score of the stem cells after the preset time as sample label, and train a stem cell quality score model by using the sample data and the sample label through an optimization algorithm; the preset condition comprises that the Euclidean distance between the corresponding stem cells is less than a distance threshold, wherein the distance threshold is a self-defined parameter; and the correlation coefficient between the corresponding stem cells is greater than or equal to a coefficient threshold, wherein the coefficient threshold is a self-defined parameter; the cell morphology recognition model is composed of five branches with the same structure but not sharing weight parameters, the inputs of the five branches are all the image data of the stem cells, and the outputs are respectively the cell area, the aspect ratio, the vacuole rate, the edge definition and whether adherent; each branch is composed of a region division layer, a linear projection layer, a position encoding layer, a feature extraction layer and a first classifier; the stem cell quality score model is composed of a graph structure data analysis layer, a splicing layer and a second classifier; the graph structure data analysis layer is used to update the graph structure data of the stem cells; the splicing layer is used to splice and normalize the vector of the updated core node with the local environment data within the preset range to obtain a combined vector, and the dimension number of the combined vector is a self-defined parameter; the second classifier inputs the combined vector, and the class space of the second classifier represents the quality score of the stem cells; the graph structure data analysis layer is constructed based on a graph attention network model.

2. The stem cell quality evaluation system according to claim 1, wherein the preset time is a self-defined parameter, the preset range represents a circular region generated with the stem cell as the center and a preset distance as the radius, and the preset distance is a self-defined parameter.

3. The stem cell quality evaluation system of claim 1, wherein the region division layer is used to divide the image data into N square regions, and each square region is represented by a first vector with a dimension number of P²C, wherein N=H×W / P², H, W and C respectively represent the height, width and channel number of the image data, and P represents the side length of the square region, and P is a self-defined parameter; The linear projection layer is used to convert each square region into a second vector representation with a dimension number of D, where D is a self-defined parameter, and the calculation formula of the linear projection layer is as follows: ； wherein and denote the second and first vector of the i-th square region, respectively, denotes a weight matrix, denotes a bias vector, e denotes the natural constant; The position encoding layer is used to obtain a third vector by embedding a position vector into the second vector of each square region, and the calculation formula of the position vector is as follows: ； where L represents the maximum value of the auto-increments, and is a user-defined parameter, represents the 2-dimensional values of the kth auto-increment of the position vector of the ith square region, p represents the normalized position number, and norm represents the Min-Max normalization method; The feature extraction layer is used to obtain a fourth vector by performing feature extraction on a sequence composed of the third vectors of the N square regions, and the fourth vector is taken as the input of the first classifier.

4. The stem cell quality evaluation system of claim 1, wherein The calculation formula of the feature extraction layer includes: ； ； ； wherein denotes the fourth vector output by the feature extraction layer, M denotes the attention matrix, and denote the weight matrix and the bias vector corresponding to the attention matrix respectively, Q, K and V denote the query matrix, the key matrix and the value matrix respectively, are all obtained by multiplying the corresponding weight matrix by a sequence of third vectors composed of N square areas, and have a size of NxA, wherein A is a self-defined parameter, denotes a mask operation on the query matrix and the key matrix to obtain a mask matrix, , and denote the element value of the nth row and the a th column of the mask matrix, the query matrix and the key matrix respectively, hash denotes a hash function, T denotes a transpose operation, denotes point-by-point multiplication, Swish denotes a Swish activation function, and softmax denotes a softmax activation function.

5. The stem cell quality evaluation system of claim 1, wherein The number of newly proliferated cells is obtained by cell counting as the proliferation capacity, the number of newly proliferated cells with normal chromosomes is obtained by karyotype analysis as the genetic stability capacity, and the number of newly proliferated cells with positive flow cytometry results is obtained as the surface marker expression capacity.

6. The stem cell quality evaluation system of claim 1, wherein The stem cell quality scoring model is trained by an optimization algorithm, including the following steps: Step S201, randomly generate parameters of the stem cell quality scoring model as the encoding of individuals in the initialization population; Step S202, calculate the loss values of all individuals in the initialization population by a loss function; The calculation formula of the loss function is the mean square error between the value output by the stem cell quality scoring model using the parameters corresponding to the encoding of the individual and the sample label; Step S203, generate a non-linear adjustment factor according to the current iteration number and the maximum iteration number, and determine whether the non-linear adjustment factor is less than a preset factor threshold, if yes, go to step S204, otherwise go to step S205; The calculation formula of the non-linear adjustment factor Factor is as follows: ； where t represents the current iteration number t, represents the maximum iteration number, and the preset factor threshold is a user-defined parameter. Step S204, increase the current iteration number by 1, and update the encoding of all individuals in the initialization population by a first update strategy; Step S205, increase the current iteration number by 1, and update the encoding of all individuals in the initialization population by a second update strategy; Step S206, determine whether the current iteration number is greater than or equal to the maximum iteration number or the minimum value of the loss values of the individuals in the initialization population is less than or equal to a preset loss threshold, if yes, terminate the iteration, and take the encoding of the individual with the minimum loss value as the parameters of the stem cell quality scoring model, otherwise return to step S202 for continuous execution; Wherein the maximum iteration number and the preset loss threshold are both self-defined parameters.

7. The stem cell quality evaluation system according to claim 6, wherein The calculation formula of the optimization algorithm includes: The calculation formula of the first update strategy includes: ； ； where 1 < g < G, 1 < t < T , and denote the encoding of the gth individual at the current iteration number t and t+1, respectively, denotes the encoding of a random individual at the current iteration number t, and R denotes the correlation coefficient, denotes a first random number with a value ranging between 0 and 1; The calculation formula of the second update strategy is as follows: ； wherein represents the encoding of the individual with the minimum loss value in the initialization population for the current iteration number t, represents a second random number with a value range between 0 and 1.

8. A method for evaluating the quality of stem cells, characterized by, Execute a stem cell quality evaluation system as claimed in any one of claims 1 to 7, including the following steps: Step S301, collect local environmental data and image data of stem cells within a preset range; Step S302, obtain cell morphology data of all stem cells within the preset range by a cell morphology recognition model; Step S303, construct graph structure data according to the cell morphology data of all stem cells within the preset range; Step S304, measure the proliferation capacity, genetic stability capacity and surface marker expression capacity of stem cells after a preset time length, and obtain the quality score of stem cells by manual annotation; Step S305, taking the graph structure data of the stem cell and the local environment data within the preset range as sample data, taking the quality score of the stem cell after the preset time length as sample label, and training the stem cell quality score model by using the sample data and the sample label through an optimization algorithm.

Images