A cell death detection method and system based on fluorescence technology

Abstract

The present invention relates to the technical field of cell fluorescence analysis, specifically to a method and system for detecting cell death based on fluorescence technology, comprising the following steps: collecting fluorescence signal data of cells, performing standardization processing, calculating the mean value and standard deviation of each fluorescence channel, performing normalization processing on each data point, and obtaining standardized fluorescence data. In the present invention, by normalizing the fluorescence signal data, the method effectively eliminates the data deviation caused by differences in experimental conditions. In the feature extraction link, the features that best represent the changes in cell states are screened out through data conversion techniques such as eigenvalue analysis and matrix construction. The features that are highly correlated with cell death are accurately screened by calculating the correlation coefficient, ensuring the pertinence and effectiveness of the detection model. The parameter optimization and verification of the model are tested through multiple sets of data, which improves the generalization ability of the model, so that it can maintain efficient recognition performance under a variety of experimental conditions.

Description

Technical Field

[0001] The present invention relates to the technical field of cell fluorescence analysis, and in particular to a cell death detection method and system based on fluorescence technology. Background Art

[0002] The field of cell fluorescence analysis technology mainly involves the use of fluorescent markers to detect and analyze various biological properties of cells, including cell activity, cell cycle status, apoptosis and necrosis. This technology relies on the ability of specific fluorescent dyes or fluorescent proteins to specifically bind to cellular components, and observe and quantify fluorescent signals through fluorescence microscopy or flow cytometry. The key to the selection of fluorescent markers lies in their high selectivity for specific cell states or molecules, and their ability to stably emit light in the internal environment of the cell. The high sensitivity and quantitative ability of this technology make it widely used in biomedical research, clinical diagnosis and drug development.

[0003] Among them, the cell death detection method based on fluorescence technology refers to the use of fluorescent markers to identify and measure the proportion of dead cells in cells. This method is particularly important in studying the mechanism of cell death, evaluating the effectiveness of drug treatment, and diagnosing diseases. Dead or apoptotic cells can be marked by specific fluorescent dyes, such as anitol blue or propidium iodide. These dyes have different affinities with living cells and dead cells, so they can be clearly distinguished under a fluorescence microscope or by flow cytometry. This allows researchers to accurately monitor the mortality rate of cells under different conditions, and then analyze the regulatory mechanism of cell death and its role in disease.

[0004] Although existing fluorescence analysis technology has advantages in selectivity and stability, its limitations in actual operation are mainly reflected in the complexity of data processing and experimental operations. Traditional methods rely on intuitive fluorescence intensity analysis and often require manual threshold setting, which may lead to subjectivity and inconsistency in results. In addition, existing technologies have not fully applied modern data processing technology in feature extraction and data utilization, limiting their scope of application in complex biological samples. The dependence and operational complexity of these technologies have increased the time cost and error rate of the experiment to a certain extent, especially when the sample size is large or the state changes are complex, which may lead to a decrease in diagnostic efficiency and accuracy, affecting the advancement of clinical and scientific research work. Summary of the invention

[0005] The purpose of the present invention is to solve the shortcomings of the prior art and to propose a cell death detection method and system based on fluorescence technology.

[0006] In order to achieve the above object, the present invention adopts the following technical solution: a method for detecting cell death based on fluorescence technology, comprising the following steps:

[0007] S1: Collect the fluorescence signal data of the cells, perform standardization, calculate the mean and standard deviation of each fluorescence channel, normalize each data point, and obtain standardized fluorescence data;

[0008] S2: Based on the standardized fluorescence data, the covariance matrix of the data is calculated, the eigenvalues ​​and eigenvectors are solved, the eigenvectors with higher eigenvalues ​​are selected, the characteristic matrix is ​​constructed, and the original data is linearly transformed to obtain the encoded characteristic data;

[0009] S3: using the encoded feature data, calculating the correlation coefficient between each feature and the cell death state, screening features with higher correlation coefficients, using these features and cell state data to establish a mapping relationship between input and output variables, and establishing a preliminary detection model;

[0010] S4: Optimizing the preliminary detection model, randomly dividing the data into multiple groups as training and verification data, adjusting the model parameter values, calculating the error on the verification data, selecting the parameter configuration with the smallest error, and obtaining an optimized cell death detection model.

[0011] As a further solution of the present invention, the step of acquiring the standardized fluorescence data is specifically as follows:

[0012] S111: Collect cell fluorescence signal data from experimental equipment and generate raw fluorescence data sets;

[0013] S112: Extract each channel data from the original fluorescence data set using the formula:

[0014] and ;

[0015] Calculate the mean and standard deviation of fluorescence intensity one by one to generate a statistical feature set;

[0016] in, represents the calculated average value, represents a single data point in the original fluorescence dataset, represents the total number of data points in the original fluorescence dataset, represents the calculated standard deviation, Calculates the average value of all data points The sum of the squares of the differences is used to calculate the standard deviation;

[0017] S113: Using the mean value and standard deviation in the statistical feature set, one by one, compare each data point in the original fluorescence data set To standardize, use the formula:

[0018] ;

[0019] Calculate the z score for each data point to generate normalized fluorescence data;

[0020] in, represents the normalized data points, Represents a single data point in the raw fluorescence dataset.

[0021] As a further solution of the present invention, the step of obtaining the encoding feature data is specifically as follows:

[0022] S211: Calculate the covariance matrix from the standardized fluorescence data, process the data set of each pair of variables, and calculate the mean deviation product using the formula:

[0023] ;

[0024] Generate a covariance matrix;

[0025] in, represents the covariance between two variables, and is a data point, and is the average value, is the sample size;

[0026] S212: Based on the covariance matrix, solve the eigenvalues ​​and eigenvectors, identify and extract multiple eigenvalues ​​and their corresponding eigenvectors through numerical decomposition, ensure that the eigenvectors correctly describe the key variation direction of the data, and generate an eigenvalue list and an eigenvector set;

[0027] S213: Select the eigenvector corresponding to the larger eigenvalue from the eigenvalue list, construct a characteristic matrix, and perform linear transformation using the formula:

[0028] ;

[0029] Perform transformation to generate encoded feature data;

[0030] in, is the transformed data matrix, is the original data matrix, is the eigenvector matrix.

[0031] As a further solution of the present invention, the steps of obtaining the preliminary detection model are specifically as follows:

[0032] S311: Based on the encoded feature data, the correlation coefficient between each feature and the cell death state is calculated using the statistical formula:

[0033] ;

[0034] Get the correlation measure for each feature and generate a list of correlation coefficients;

[0035] in, represents the correlation coefficient, represents the feature data point, represents the cell death status data point, and represents the corresponding average value;

[0036] S312: Based on the correlation coefficient list, screen features with correlation coefficients higher than a set threshold value to ensure that the features have high statistical significance for predicting cell death status, and generate a screening feature list;

[0037] S313: Use the screening feature list and cell state data to build a mapping relationship, and use a logistic regression model to establish the relationship between input and output, using the formula:

[0038] ;

[0039] Build a preliminary detection model;

[0040] in, is the probability output of logistic regression, is the eigenvalue, are the coefficients of the logistic regression model.

[0041] As a further solution of the present invention, the steps for obtaining the optimized cell death detection model are specifically as follows:

[0042] S411: Starting from the preliminary detection model, the entire data set is randomly divided into a training set and a validation set, and a random distribution algorithm is used to ensure uniform distribution of data to generate training and validation data sets;

[0043] S412: adjusting the model parameters on the training and validation data sets, applying a cross-validation method to iteratively optimize the parameters, and generating an adjusted model parameter set by adjusting the learning rate and regularization parameters;

[0044] S413: Calculate the error of the model on the validation set using the adjusted model parameters, and apply the mean square error formula:

[0045] ;

[0046] By comparing the errors under different parameter configurations, the parameter configuration that minimizes the error is selected to generate an optimized cell death detection model;

[0047] in, is the mean square error, is the actual observed value, is the model prediction value, Indicates the sample size.

[0048] A cell death detection system based on fluorescence technology, the cell death detection system based on fluorescence technology is used to perform the cell death detection method based on fluorescence technology, the system comprises:

[0049] The data collection module uses a fluorescence microscope to collect multi-channel fluorescence signals based on cell samples, records the intensity value of each cell in multiple fluorescence channels, and organizes and stores the original fluorescence data;

[0050] The data normalization module receives the raw fluorescence data, calculates the mean value and standard deviation of each fluorescence channel, performs normalization processing on each data point, and generates standardized fluorescence data by subtracting the mean value and dividing by the standard deviation;

[0051] The feature extraction module calculates the covariance matrix based on the standardized fluorescence data, extracts eigenvalues ​​and eigenvectors, selects eigenvectors with higher eigenvalues, applies linear transformation to generate a feature matrix, and obtains coded feature data;

[0052] The detection model module uses the encoded feature data to calculate the correlation coefficient between each feature and the cell death state, screens features with high correlation coefficients, establishes a mapping relationship between input and output variables, and constructs a preliminary detection model;

[0053] The model optimization module groups the data of the preliminary detection model, randomly divides the data into a training group and a validation group, adjusts the model parameters, selects the parameter configuration by calculating the error of the validation group, and obtains the optimized cell death detection model.

[0054] Compared with the prior art, the advantages and positive effects of the present invention are:

[0055] In the present invention, by normalizing the fluorescence signal data, this method effectively eliminates the data deviation caused by differences in experimental conditions and ensures the consistency of subsequent analysis. In the feature extraction stage, the features that best represent changes in cell states are screened out through data conversion techniques such as eigenvalue analysis and matrix construction. By calculating the correlation coefficient, the features that are highly correlated with cell death are accurately screened to ensure the pertinence and effectiveness of the detection model. The parameter optimization and verification of the model are tested through multiple sets of data, which improves the generalization ability of the model, so that it can maintain efficient recognition performance under a variety of experimental conditions, thereby providing a more reliable cell death monitoring tool. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] Figure 1 It is a schematic diagram of the workflow of the present invention;

[0057] Figure 2A flow chart of the steps for obtaining standardized fluorescence data of the present invention;

[0058] Figure 3 A flowchart of the steps for obtaining the encoding feature data of the present invention;

[0059] Figure 4 This is a flow chart of the steps for obtaining the preliminary detection model of the present invention;

[0060] Figure 5 This is a flow chart of the steps for obtaining the optimized cell death detection model of the present invention. DETAILED DESCRIPTION

[0061] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0062] In the description of the present invention, it should be understood that the terms "length", "width", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside" and the like indicate positions or positional relationships based on the positions or positional relationships shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention. In addition, in the description of the present invention, "multiple" means two or more, unless otherwise clearly and specifically defined.

[0063] Example: See Figure 1 The present invention provides a technical solution: a method for detecting cell death based on fluorescence technology, comprising the following steps:

[0064] S1: Collect the fluorescence signal data of the cells, perform standardization, calculate the mean and standard deviation of each fluorescence channel, normalize each data point, and obtain standardized fluorescence data;

[0065] S2: Based on the standardized fluorescence data, the covariance matrix of the data is calculated, the eigenvalues ​​and eigenvectors are solved, the eigenvectors with higher eigenvalues ​​are selected, the feature matrix is ​​constructed, and the original data is linearly transformed to obtain the encoded feature data;

[0066] S3: Using the encoded feature data, calculate the correlation coefficient between each feature and the cell death state, select features with higher correlation coefficients, use these features and cell state data to establish a mapping relationship between input and output variables, and establish a preliminary detection model;

[0067] S4: Optimize the preliminary detection model, randomly divide the data into multiple groups as training and validation data, adjust the model parameter values, calculate the error on the validation data, select the parameter configuration with the smallest error, and obtain the optimized cell death detection model.

[0068] The standardized fluorescence data specifically include fluorescence intensity index, channel deviation, and normalization coefficient. The encoded feature data include principal component analysis results, feature vector sets, and transformation matrices. The preliminary detection model specifically refers to correlation scores, feature selection results, and variable mapping relationships. The optimized cell death detection model includes parameter configuration, verification error, and model stability.

[0069] See also Figure 2 , the specific steps for obtaining standardized fluorescence data are:

[0070] S111: Collect cell fluorescence signal data from experimental equipment and generate raw fluorescence data sets;

[0071] Collect cell fluorescence signal data from the experimental equipment. First, input the data into the fluorescence intensity analysis program, perform preliminary statistical analysis on the fluorescence intensity of each channel, and obtain preliminary intensity values ​​of multiple channels. Then, by comparing the fluorescence intensity of each channel with the standard fluorescence spectrum, identify the characteristics of multi-channel fluorescence signals, and record abnormal signals for subsequent analysis. This process ensures data collection and preliminary analysis, provides input data for the next statistical analysis, and generates the original fluorescence data set.

[0072] S112: Extract each channel data from the original fluorescence data set using the formula:

[0073] and ;

[0074] Calculate the mean and standard deviation of fluorescence intensity one by one to generate a statistical feature set;

[0075] in, represents the calculated average value, represents a single data point in the original fluorescence dataset, represents the total number of data points in the original fluorescence dataset, represents the calculated standard deviation, Calculates the average value of all data points The sum of the squares of the differences is used to calculate the standard deviation;

[0076] formula:

[0077] ;

[0078] Detailed explanation of the formula and the process of formula calculation and derivation:

[0079] Parameter details:

[0080] : Represents the intensity value of a single fluorescence signal data point, which is directly measured from the cell sample by a fluorescence spectrometer.

[0081] : Represents the total number of fluorescence data points, ensuring that all measurements are included in the average calculation.

[0082] Parameter setting basis:

[0083] Each data point Obtained directly from experimental measurements, ensuring the real-time and reliability of data.

[0084] total It is the sample size based on the experimental design to ensure the representativeness and precision of the statistical analysis.

[0085] Formula calculation derivation process (into the example):

[0086] Set up the fluorescence data points collected from the experiment for: ;

[0087] Calculate the average : ;

[0088] Result analysis:

[0089] average value Represents the average fluorescence intensity of this batch of data points. The result reveals the average fluorescence response of the measured cell sample under experimental conditions, which is the basis for subsequent normalization processing and data analysis.

[0090] formula:

[0091] ;

[0092] Detailed explanation of the formula and the process of formula calculation and derivation:

[0093] Parameter details:

[0094] : Represents the intensity value of a single fluorescence signal data point, which is directly measured from the cell sample by a fluorescence spectrometer.

[0095] : represents the average value of the fluorescence data points, which has been calculated previously as .

[0096] : Represents the total number of fluorescence data points.

[0097] Parameter setting basis:

[0098] Each data point It is obtained from experimental measurements to ensure the authenticity of the data.

[0099] average value It is calculated based on all data points and reflects the trend in the sample set.

[0100] total Determined from the experimental design, the multidimensionality and accuracy of statistical analysis are guaranteed.

[0101] Formula calculation derivation process (into the example):

[0102] Using the previous data point: ;

[0103] Known average , calculate the square of the difference between each data point and the mean, and then sum them: ;

[0104] Calculate standard deviation : ;

[0105] Result analysis:

[0106] Standard Deviation It indicates the dispersion of fluorescence data points relative to their mean. The result shows the fluctuation of data points around the mean, and a smaller standard deviation indicates that most data points are concentrated around the mean.

[0107] S113: Use the mean and standard deviation in the statistical feature set to compare each data point in the original fluorescence dataset one by one To standardize, use the formula:

[0108] ;

[0109] Calculate the z score for each data point to generate normalized fluorescence data;

[0110] in, represents the normalized data points, Represents a single data point in the raw fluorescence dataset.

[0111] formula:

[0112] ;

[0113] Detailed explanation of the formula and the process of formula calculation and derivation:

[0114] Parameter details:

[0115] : The intensity value of the original fluorescence data point is directly measured by the fluorescence spectrometer.

[0116] : The average value of the fluorescence data points, calculated as ,in is the total number of data points.

[0117] : Standard deviation of fluorescence data points, calculated as .

[0118] Parameter setting basis:

[0119] Fluorescence data points Acquire directly from experimental equipment to ensure the real-time and accuracy of data.

[0120] average value and standard deviation It is calculated dynamically based on the distribution within the measured data set, reflecting the sample concentration trend and degree of dispersion.

[0121] Formula calculation derivation process (into the example):

[0122] Set the measured fluorescence data points as follows: ;

[0123] Calculate the average : ;

[0124] Calculate standard deviation : ;

[0125] Apply the normalization formula to calculate : ;

[0126] Result analysis:

[0127] Generate normalized fluorescence data Indicates how far each data point deviates from the mean, in units of standard deviation. means that the data point is exactly on the mean, and Indicates that the data point is about standard deviation, Approximately below average standard deviation.

[0128] See also Figure 3 , the specific steps for obtaining the encoding feature data are:

[0129] S211: Calculate the covariance matrix from the standardized fluorescence data, process the data set for each pair of variables, and calculate the mean deviation product using the formula:

[0130] ;

[0131] Generate a covariance matrix;

[0132] in, represents the covariance between two variables, and is a data point, and is the average value, is the sample size;

[0133] formula:

[0134] ;

[0135] Detailed explanation of the formula and the process of formula calculation and derivation:

[0136] Formula used to calculate two variables and The covariance between them measures the degree of linear association between variables. The calculation and derivation process of the parameters are as follows:

[0137] Calculate the average and :

[0138] The following dataset is set up, and , the number of data points .

[0139] ;

[0140] ;

[0141] Compute the difference between each term and the mean:

[0142] ;

[0143] ;

[0144] Compute the sum of the products of the mean deviations:

[0145] ;

[0146] ;

[0147] Calculating covariance :

[0148] ;

[0149] Calculations show that , The covariance between the variables is 12.4, which indicates the strength and direction of the association between the two variables. increase, It also shows an increasing trend;

[0150] S212: Based on the covariance matrix, solve the eigenvalues ​​and eigenvectors, identify and extract multiple eigenvalues ​​and their corresponding eigenvectors through numerical decomposition, ensure that the eigenvectors correctly describe the key variation direction of the data, and generate the eigenvalue list and eigenvector set;

[0151] To extract data from the generated covariance matrix, matrix decomposition technology is used, including determinant operations and eigendecomposition of the matrix, so as to separate each eigenvalue and its corresponding eigenvector. This step not only involves mathematical calculations, but also requires the application of knowledge of matrix theory and linear algebra to ensure that each eigenvalue and eigenvector are correctly extracted from the covariance matrix. The size of the eigenvalue directly affects the selection of subsequent eigenvectors, laying the foundation for data feature extraction.

[0152] S213: Select the eigenvector corresponding to the larger eigenvalue from the eigenvalue list, construct the characteristic matrix, and perform linear transformation using the formula:

[0153] ;

[0154] Perform transformation to generate encoded feature data;

[0155] in, is the transformed data matrix, is the original data matrix, is the eigenvector matrix.

[0156] formula:

[0157] ;

[0158] Detailed explanation of the formula and the process of formula calculation and derivation:

[0159] The formula is used to reduce the dimension of the original data set through the selected principal components (eigenvectors) to generate the current data representation, where represents the original data matrix, represents the matrix consisting of key eigenvectors, is the result matrix, i.e. the data after dimensionality reduction.

[0160] Parameter details:

[0161] : The original data matrix, size is ,in is the number of samples, is the number of features.

[0162] : Eigenvector matrix, size is , obtained by the eigendecomposition of the data covariance matrix, represents the key variation direction of the data, is the number of key features selected.

[0163] : The transformed data matrix has the size , is the data representation after dimensionality reduction.

[0164] Calculation process:

[0165] Get : Obtained through data collection, for example, a set of data includes 100 samples, each sample has 5 features.

[0166] calculate :Calculate first The covariance matrix of , and then perform eigendecomposition and select the largest The eigenvalues ​​correspond to the eigenvectors.

[0167] calculate : Using matrix multiplication Perform calculations.

[0168] Calculation example:

[0169] set up: ;

[0170] ;

[0171] calculate Steps:

[0172] ;

[0173] ;

[0174] ;

[0175] Results Matrix :

[0176] ;

[0177] The results show that after the eigenvector After linear transformation, the original data The dimension is successfully reduced to a current feature space, where each new feature value is a weighted sum of the original features, reflecting the key variability of the data.

[0178] See also Figure 4 ,The specific steps for obtaining the preliminary detection model are:

[0179] S311: Based on the encoded feature data, the correlation coefficient between each feature and the cell death state is calculated using the statistical formula:

[0180] ;

[0181] Get the correlation measure for each feature and generate a list of correlation coefficients;

[0182] in, represents the correlation coefficient, represents the feature data point, represents the cell death status data point, and represents the corresponding average value;

[0183] formula:

[0184] ;

[0185] Detailed explanation of the formula and the process of formula calculation and derivation:

[0186] Parameter definition:

[0187] : The i-th feature data point.

[0188] : The i-th cell death status data point.

[0189] : The average value of the feature data points.

[0190] : Mean value of cell death status data points.

[0191] Calculation process and examples:

[0192] The setup has the following data points:

[0193] Feature data points : [5,7,8];

[0194] Cell Death Status Data Points : [1,3,5];

[0195] Calculate the average:

[0196] ;

[0197] Compute the product of the differences:

[0198]

[0199] Compute the sum of squared differences:

[0200] ;

[0201] ;

[0202] Calculate the correlation coefficient:

[0203] ;

[0204] The calculation process ensures the correlation coefficient The calculation of is based on real and detected data values ​​and follows statistical principles to ensure accuracy. The correlation coefficient indicates the linear relationship between the characteristic data and the cell death status, where High values ​​of indicate strong associations and are suitable for iterative analysis and model building.

[0205] S312: Based on the correlation coefficient list, features with correlation coefficients higher than a set threshold are screened to ensure that the features have high statistical significance for predicting cell death status, and a screening feature list is generated;

[0206] Features with correlation coefficients higher than the set threshold are screened for model construction. The threshold set in the screening process is based on previous research and experimental data, which ensures that only the most predictive features are selected. The screening process is based on statistical principles and is completed by sorting the correlation data of each feature and comparing the threshold.

[0207] S313: Use the screening feature list and cell state data to build a mapping relationship, and use a logistic regression model to establish the relationship between input and output, using the formula:

[0208] ;

[0209] Build a preliminary detection model;

[0210] in, is the probability output of logistic regression, is the eigenvalue, are the coefficients of the logistic regression model.

[0211] formula:

[0212] ;

[0213] Detailed explanation of the formula and the process of formula calculation and derivation:

[0214] formula represents the probability of predicting the cell death state, where is the cardinality of natural logarithms, It is a model parameter, which is obtained by fitting actual data and reflects the influence of differential characteristics on cell death status. A quantitative value representing a differentiated biological feature.

[0215] To calculate the formula, refer to an example and set two features and First, data were collected from clinical trials and logistic regression was used to estimate Setting the actual measurement , indicating the expression level of a certain protein, , indicating the expression level of a cell surface marker. are respectively set to the values ​​estimated from the correlation data, , , .

[0216] Insert the values ​​to calculate:

[0217] ;

[0218] The results show that at a given eigenvalue and Under these conditions, the probability of cells being in a state of death is very high.

[0219] See also Figure 5 , the steps for obtaining the optimized cell death detection model are as follows:

[0220] S411: Starting from the preliminary detection model, the entire data set is randomly divided into a training set and a validation set, and a random distribution algorithm is used to ensure uniform distribution of data to generate training and validation data sets;

[0221] The random distribution algorithm ensures the data balance of the training set and the validation set. During implementation, the algorithm generates a pseudo-random number sequence and divides the data set according to a preset ratio. The random number generator involved in this process is based on the linear congruential method. The method can quickly and effectively generate random number sequences in a big data environment, ensuring the uniformity and randomness of data distribution, and avoiding any potential bias in the data set that affects the generalization ability of the model, thereby ensuring the adaptability and prediction accuracy of the model to new data in the subsequent training and validation process. The stability and reliability of the model parameters are verified through multiple random divisions of the training set and the validation set, and the training and validation data sets are generated.

[0222] S412: adjusting model parameters on the training and validation data sets, applying a cross-validation method to iteratively optimize the parameters, and generating an adjusted model parameter set by adjusting a learning rate and a regularization parameter;

[0223] The adjustment of model parameters is carried out by calculating the loss function after each round of training. The loss function uses the mean square error, which measures the average of the squares of the deviations between the model predictions and the actual observations. The model parameters are optimized by the gradient descent method, which can effectively search for the point with the smallest error in the parameter space and adjust key parameters including the learning rate and regularization coefficient to ensure that the model exhibits optimal learning efficiency and overfitting control in differentiated training cycles. The optimized parameters are cross-validated and iterated multiple times to verify the optimal parameter settings and generate an adjusted model parameter set.

[0224] S413: Calculate the model error on the validation set using the adjusted model parameters, applying the mean square error formula:

[0225] ;

[0226] By comparing the errors under different parameter configurations, the parameter configuration that minimizes the error is selected to generate an optimized cell death detection model;

[0227] in, is the mean square error, is the actual observed value, is the model prediction value, Indicates the sample size.

[0228] formula:

[0229] ;

[0230] Detailed explanation of the formula and the process of formula calculation and derivation:

[0231] Refer to a data set where represents the true value, represents the model prediction value, and the total number of samples is , the true value Measured from actual data, set as a set of integers from 1 to 100, the predicted value As the model output, set each predicted value to be 5 less than the true value, then .

[0232] Calculate MSE:

[0233] ;

[0234] The results show that the square of the average deviation between the model prediction and the actual value is 25, indicating that the model systematically underestimates the deviation by 5 units under the current parameter settings. The indicator helps quantify the accuracy of the model prediction and provides a basis for adjusting the model parameters.

[0235] A cell death detection system based on fluorescence technology, the cell death detection system based on fluorescence technology is used to perform the above-mentioned cell death detection method based on fluorescence technology, the system comprises:

[0236] The data collection module uses a fluorescence microscope to collect multi-channel fluorescence signals based on cell samples, records the intensity value of each cell in multiple fluorescence channels, and organizes and stores the original fluorescence data;

[0237] The data normalization module receives the raw fluorescence data, calculates the mean and standard deviation of each fluorescence channel, and normalizes each data point by subtracting the mean and dividing by the standard deviation to generate standardized fluorescence data;

[0238] The feature extraction module calculates the covariance matrix based on the standardized fluorescence data, extracts the eigenvalues ​​and eigenvectors, selects the eigenvectors with higher eigenvalues, applies linear transformation to generate the feature matrix, and obtains the coded feature data;

[0239] The detection model module uses the encoded feature data to calculate the correlation coefficient between each feature and the cell death state, screens features with high correlation coefficients, establishes a mapping relationship between input and output variables, and constructs a preliminary detection model;

[0240] The model optimization module groups the data of the preliminary detection model, randomly divides the data into a training group and a validation group, adjusts the model parameters, selects the parameter configuration by calculating the error of the validation group, and obtains the optimized cell death detection model.

[0241] The above are only preferred embodiments of the present invention and are not intended to limit the present invention in other forms. Any technician familiar with the profession may use the technical contents disclosed above to change or modify them into equivalent embodiments with equivalent changes and apply them to other fields. However, any simple modification, equivalent change and modification made to the above embodiments based on the technical essence of the present invention without departing from the technical solution of the present invention still falls within the protection scope of the technical solution of the present invention.

Claims

1. A method for detecting cell death based on fluorescence technology, characterized in that: The following steps are involved: Collect the fluorescence signal data of the cells, perform standardization, calculate the mean value and standard deviation of each fluorescence channel, perform normalization on each data point, and obtain standardized fluorescence data; Based on the standardized fluorescence data, the covariance matrix of the data is calculated, the eigenvalues ​​and eigenvectors are solved, the eigenvectors with higher eigenvalues ​​are selected, the characteristic matrix is ​​constructed, and the original data is linearly transformed to obtain the encoded characteristic data; Using the encoded feature data, calculating the correlation coefficient between each feature and the cell death state, screening features with higher correlation coefficients, using these features and cell state data to establish a mapping relationship between input and output variables, and establishing a preliminary detection model; The steps for obtaining the preliminary detection model are specifically as follows: Based on the encoded feature data, the correlation coefficient between each feature and the cell death state was calculated using the statistical formula: Get the correlation measure for each feature and generate a list of correlation coefficients; Among them, r represents the correlation coefficient, Z i Represents the transformed feature data point, y i ’ represents the cell death status data point, and represents the corresponding average value; Based on the correlation coefficient list, features with correlation coefficients higher than a set threshold are screened to ensure that the features have a high statistical significance for predicting cell death status, and a screening feature list is generated; The mapping relationship is constructed using the screening feature list and cell state data, and the relationship between input and output is established using a logistic regression model using the formula: Build a preliminary detection model; Among them, P(y) is the probability output of logistic regression, x1,...,x n are the eigenvalues, β0,...,β n are the coefficients of the logistic regression model; Optimizing the preliminary detection model, randomly dividing the data into multiple groups as training and verification data, adjusting the model parameter values, calculating the error on the verification data, selecting the parameter configuration with the smallest error, and obtaining an optimized cell death detection model; The steps for obtaining the optimized cell death detection model are specifically as follows: Starting from the preliminary detection model, the entire data set is randomly divided into a training set and a validation set, and a random distribution algorithm is used to ensure uniform distribution of data to generate training and validation data sets; Adjusting model parameters on the training and validation data sets, applying a cross-validation method to iteratively optimize the parameters, and generating an adjusted model parameter set by adjusting a learning rate and a regularization parameter; The error of the model is calculated on the validation set using the adjusted model parameters, and the mean square error formula is applied: By comparing the errors under different parameter configurations, the parameter configuration that minimizes the error is selected to generate an optimized cell death detection model; Where MSE is the mean square error, y" i is the actual observed value, is the model prediction value, and n' represents the number of samples.

2. The method for detecting cell death based on fluorescence technology according to claim 1, characterized in that: The steps for obtaining the standardized fluorescence data are specifically as follows: Collect cell fluorescence signal data from experimental equipment to generate raw fluorescence data sets; Extract each channel data from the original fluorescence data set using the formula: and Calculate the mean and standard deviation of fluorescence intensity one by one to generate a statistical feature set; in, represents the calculated average value, x i represents a single data point in the original fluorescence data set, n represents the total number of data points in the original fluorescence data set, s represents the calculated standard deviation, Calculates the average value of all data points The sum of the squares of the differences is used to calculate the standard deviation; Using the mean and standard deviation in the statistical feature set, each data point x in the original fluorescence data set is i To standardize, use the formula: Calculate the z score for each data point to generate normalized fluorescence data; Among them, z i represents the normalized data point, x i Represents a single data point in the raw fluorescence dataset.

3. The method for detecting cell death based on fluorescence technology according to claim 2, characterized in that: The steps for obtaining the encoding feature data are specifically as follows: The covariance matrix was calculated from the normalized fluorescence data and for each pair of variables in the dataset, the mean deviation product was calculated using the formula: Generate a covariance matrix; Among them, σ xy represents the covariance between two variables, y i is a data point, is the average value; Based on the covariance matrix, solve the eigenvalues ​​and eigenvectors, identify and extract multiple eigenvalues ​​and their corresponding eigenvectors through numerical decomposition, ensure that the eigenvectors correctly describe the key variation direction of the data, and generate an eigenvalue list and an eigenvector set; Select the eigenvector corresponding to the larger eigenvalue from the eigenvalue list, construct the characteristic matrix, and perform linear transformation using the formula: Z=XV Perform transformation to generate encoded feature data; Among them, Z is the transformed data matrix, X is the original data matrix, and V is the eigenvector matrix.

4. A cell death detection system based on fluorescence technology, characterized in that: According to any one of claims 1 to 3, the method for detecting cell death based on fluorescence technology comprises: The data collection module uses a fluorescence microscope to collect multi-channel fluorescence signals based on cell samples, records the intensity value of each cell in multiple fluorescence channels, and organizes and stores the original fluorescence data; The data normalization module receives the raw fluorescence data, calculates the mean value and standard deviation of each fluorescence channel, performs normalization processing on each data point, and generates standardized fluorescence data by subtracting the mean value and dividing by the standard deviation; The feature extraction module calculates the covariance matrix based on the standardized fluorescence data, extracts eigenvalues ​​and eigenvectors, selects eigenvectors with higher eigenvalues, applies linear transformation to generate a feature matrix, and obtains coded feature data; The detection model module uses the encoded feature data to calculate the correlation coefficient between each feature and the cell death state, screens features with high correlation coefficients, establishes a mapping relationship between input and output variables, and constructs a preliminary detection model; The model optimization module groups the data of the preliminary detection model, randomly divides the data into a training group and a validation group, adjusts the model parameters, selects the parameter configuration by calculating the error of the validation group, and obtains the optimized cell death detection model.

Images