A method, medium and system for reducing diffraction effect of flow cytometer

Abstract

The invention provides a method, a medium and a system for reducing diffraction effect of a flow cytometer, belonging to the technical field of flow cytometers, and comprising: obtaining parameters in a flow experiment process and establishing a fluorescence matrix; establishing linear regression relationships between sample cell concentration, detection sample liquid flow rate, incident light intensity and FSC, SSC and cell fluorescence respectively; based on the FSC linear regression relationship, the SSC linear regression relationship and the cell fluorescence linear regression relationship, screening a signal whose difference with any regression relationship is greater than a threshold value, and recording it as a first nonlinear fluorescence signal; changing the sample flow rate and incident light intensity of a detection point to obtain a second and a third nonlinear fluorescence signal; fusing the experimental parameters and the nonlinear fluorescence signal, and analyzing them by using a diffraction effect analysis model to obtain a diffraction effect equation; and performing inverse operations on the FSC matrix, the SSC matrix and the cell fluorescence matrix by using the obtained diffraction effect equation respectively to obtain a final result.

Description

Technical Field

[0001] The present invention belongs to the technical field of flow cytometers, and in particular, relates to a method, a medium and a system for reducing diffraction effect of a flow cytometer. Background Art

[0002] Flow cytometry is an important single-cell analysis technology that is widely used in cytological research, clinical testing, and drug screening. In flow cytometry, when a cell sample passes through a laser beam at a certain speed, forward scattered light (FSC) and side scattered light (SSC), as well as fluorescent signals, are generated. These optical signals contain information such as cell size, internal structure, and specific markers. However, signal detection is easily affected by diffraction effects. The diffraction effect refers to the diffraction and refraction of light in the gap area when there is a gap between cells, forming a secondary diffraction peak at the detector end, causing false signals to appear in the background that should have been in the dark area. This will cause erroneous additions to the fluorescent signal and reduce the detection sensitivity. Common diffraction effects include scattered light diffraction and fluorescence diffraction effects. Traditional solutions include mechanical methods and image processing methods, such as adjusting the cell flow rate, increasing laser focus, spiral focus, and eliminating diffraction background through image processing algorithms. However, these methods are not very targeted and the elimination effect is limited.

[0003] As the sensitivity of flow cytometer detection increases, the impact of diffraction effect becomes increasingly significant. It is difficult to completely eliminate the diffraction effect by optimizing hardware equipment alone. The software algorithm has poor adaptability to various types of cell samples and cannot achieve the desired elimination effect. Therefore, a new method is urgently needed to accurately detect and effectively eliminate the diffraction effect in flow cytometers to improve the accuracy of cell sorting and detection. Summary of the invention

[0004] In view of this, the present invention provides a method, medium and system for reducing diffraction effect in a flow cytometer, which can solve the technical problem in the prior art that when using a flow cytometer to detect samples, the diffraction effect has a great influence on the detection results.

[0005] The present invention is achieved in that:

[0006] A first aspect of the present invention provides a method for reducing diffraction effect in a flow cytometer, comprising the following steps:

[0007] S10, obtaining experimental parameters during the flow cytometer experiment, including sample cell concentration, sample flow rate at the detection point, incident light intensity, and fluorescence signals obtained from the experiment, including FSC, SSC, and cell fluorescence, and establishing FSC matrix, SSC matrix, and cell fluorescence matrix;

[0008] S20, according to the experimental parameters, establish linear regression relationships between sample cell concentration, detection sample liquid flow rate, incident light intensity and FSC, SSC and cell fluorescence, respectively, recorded as FSC linear regression relationship, SSC linear regression relationship and cell fluorescence linear regression relationship;

[0009] S30, based on the FSC linear regression relationship, the SSC linear regression relationship, and the cell fluorescence linear regression relationship, screening a fluorescence signal whose difference from any regression relationship is greater than a preset threshold, and recording it as a first nonlinear fluorescence signal;

[0010] S40, changing the sample flow rate at the detection point so that the cell gap formed by the sample flow rate at the detection point is kept for a short time, reducing the diffraction effect formed by the cell gap, and repeating steps S20 to S30 to obtain a second nonlinear fluorescence signal with the sample flow rate at the detection point changed; changing the incident light intensity, repeating steps S20 to S30 to obtain a third nonlinear fluorescence signal with the sample flow rate at the detection point changed;

[0011] S50, fusing the experimental parameters and the first nonlinear fluorescence signal into a first parameter matrix, fusing the experimental parameters after changing the flow rate of the sample at the detection point and the second nonlinear fluorescence signal into a second parameter matrix, and fusing the experimental parameters after changing the incident light intensity and the third nonlinear fluorescence signal into a third parameter matrix;

[0012] S60, using a pre-trained diffraction effect analysis model, inputting the first parameter matrix, the second parameter matrix and the third parameter matrix, to obtain a diffraction effect equation;

[0013] S70, respectively performing inverse operations on the FSC matrix, SSC matrix and cell fluorescence matrix using the obtained diffraction effect equation, and using the obtained results as the final output FSC matrix, SSC matrix and cell fluorescence matrix.

[0014] On the basis of the above technical solution, a method for reducing diffraction effect of a flow cytometer of the present invention can also be improved as follows:

[0015] The step of establishing the linear regression relationship between the sample cell concentration, the flow rate of the detection sample liquid, the incident light intensity and FSC, SSC and cell fluorescence respectively according to the experimental parameters is specifically:

[0016] Using the experimental parameters as independent variables;

[0017] The obtained FSC matrix, SSC matrix and cell fluorescence matrix are used as dependent variables;

[0018] A linear regression algorithm is used to establish and fit the linear regression relationship between the independent variable and the dependent variable.

[0019] Furthermore, the step of screening the fluorescent signal whose difference from any regression relationship is greater than a preset threshold is specifically:

[0020] Calculate the predicted value, the predicted value is obtained by substituting the obtained FSC linear regression relationship into the experimental parameters, and the difference between the predicted value and the fluorescence signal is taken as the residual;

[0021] Setting a preset threshold, wherein the threshold is used to set a residual threshold based on experimental experience;

[0022] The screening signal is to compare the absolute value of the residual of each signal with a preset threshold, and screen out the signal whose absolute value of the residual is greater than the preset threshold.

[0023] Furthermore, the preset threshold is set to the mean value of the residual minus 3 times the standard deviation. If the preset threshold is less than zero, the mean value of the residual minus the standard deviation is directly used.

[0024] The steps of establishing and training the diffraction effect analysis model specifically include:

[0025] Get training samples, specifically:

[0026] Step 1: Perform a flow cytometry experiment using flow fluorescent microspheres, and use steps S10 to S50 to obtain the first, second and third parameter matrices of the experiment;

[0027] Step 2: Change the experimental parameters, conduct multiple groups of experiments, and obtain the first, second, and third parameter matrices of the multiple groups of experiments;

[0028] Step 3, using the flow cytometry experimental parameters performed by the flow cytometry fluorescent microspheres to calculate the standard output fluorescence signal;

[0029] Step 4: Calculate the diffraction effect equations corresponding to the first, second and third parameter matrices of multiple groups of experiments according to the standard fluorescence signal;

[0030] The neural network is trained, specifically: a neural network is trained using the training samples to obtain a diffraction effect analysis model.

[0031] Furthermore, the neural network is a BP neural network.

[0032] Among them, the time for the diffraction effect to form is related to the wavelength of the incident light; the intercellular time is related to the gap length and flow rate; the number of gaps is related to the total volume, single cell volume, and concentration; the core flow diameter is related to the volume flow rate and core flow rate. The above adjustments shorten the intercellular time or reduce the number of gaps, thereby weakening the generation of visible diffraction effects.

[0033] Furthermore, the specific structure of the BP neural network includes:

[0034] An input layer, used for inputting data of the first parameter matrix, the second parameter matrix and the third parameter matrix;

[0035] The first hidden layer and the second hidden layer are used to extract the data features of the parameter matrix and connect the input layer and the output layer;

[0036] The output layer is used to output the predicted diffraction effect equation results;

[0037] The input layer and the first hidden layer, the first hidden layer and the second hidden layer, and the second hidden layer and the output layer are fully connected structures;

[0038] The activation functions of the first hidden layer and the second hidden layer are Sigmoid functions, and the activation function of the output layer is a linear function;

[0039] The BP neural network is trained through an error back propagation algorithm to minimize the loss function of the predicted output and realize the extraction of the diffraction effect equation.

[0040] Among them, the first parameter matrix, the second parameter matrix, and the third parameter matrix are all constructed by horizontally splicing the independent variable parameter matrix and the dependent variable parameter matrix.

[0041] A second aspect of the present invention provides a computer-readable storage medium, wherein the computer-readable storage medium stores program instructions, and when the program instructions are executed, they are used to execute the above-mentioned method for reducing diffraction effect in a flow cytometer.

[0042] A third aspect of the present invention provides a system for reducing diffraction effects in a flow cytometer, which includes the above-mentioned computer-readable storage medium.

[0043] Compared with the prior art, the method, medium and system for reducing diffraction effect in a flow cytometer provided by the present invention have the following beneficial effects:

[0044] 1) It can accurately detect nonlinear signals affected by diffraction effects in flow cytometry experiments, avoid missed or misjudgment, and improve recognition accuracy.

[0045] 2) By changing key parameters, the diffraction signal is re-acquired to obtain more comprehensive diffraction information, which is beneficial to the training of the neural network model.

[0046] 3) The application of neural network modeling can realize the learning and approximation of the complex nonlinear relationship of the diffraction effect and obtain a high-precision diffraction effect prediction equation.

[0047] 4) Perform batch elimination through matrix operations to avoid signal processing one by one, ensure the consistency of results, and save a lot of repetitive work.

[0048] 5) This method can quantitatively detect the diffraction effect formed by cell gaps and eliminate it in a targeted manner, and has good adaptability to different samples and conditions.

[0049] 6) Greatly improved the accuracy of flow cytometry experimental results, providing reliable guarantee for cell function research and clinical testing.

[0050] In summary, the technical solution of the present invention solves the technical problem in the prior art that when using flow cytometry to detect samples, the diffraction effect has a significant impact on the detection results. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments of the present invention will be briefly introduced below. Obviously, the accompanying drawings in the following description are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative labor.

[0052] Figure 1 A flow chart of the method provided by the present invention. DETAILED DESCRIPTION

[0053] In order to make the purpose, technical solution and advantages of the embodiments of the present invention more clear, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention.

[0054] The following is a specific implementation of a method for reducing diffraction effect in a flow cytometer provided by the first aspect of the present invention:

[0055] S10, obtaining experimental parameters: obtaining experimental parameters during the flow cytometer experiment, including sample cell concentration, sample flow rate at the detection point, incident light intensity, and fluorescence signals obtained from the experiment, including FSC, SSC, and cell fluorescence, and establishing the FSC matrix, SSC matrix, and cell fluorescence matrix. The specific implementation method is as follows:

[0056] 1) Prepare the experimental samples: Select the cell samples to be tested and adjust the concentration of the cell suspension to a concentration level suitable for flow cytometry analysis (usually 1x106-1x107 cells / ml). Use a cell counting plate or the counting function of the flow cytometer to determine the concentration of the cell sample. Record the concentration data of the cell sample.

[0057] 2) Set the experimental parameters of the flow cytometer: According to the purpose of the experiment, set the light source wavelength and light source intensity parameters of the flow cytometer. Among them, the forward scatter signal detection (FSC) corresponds to the laser wavelength of 488nm, the side scatter signal detection (SSC) corresponds to the laser wavelength of 488nm, and the cell fluorescence signal detection corresponds to different fluorescence laser wavelengths. Record the incident light intensity data of each channel.

[0058] 3) Adjust the flow rate: Set the speed at which the flow cytometer sampling nozzle sucks in the sample, that is, the flow rate of the cells when they pass through the light source. This flow rate will directly affect the time each cell is excited to emit light, which is related to the level of fluorescence intensity. Record the set flow rate data.

[0059] 4) Perform sample detection: inject the cell sample with adjusted concentration into the sampler and start flow cytometry analysis. Record the fluorescence signal detection results, including FSC, SSC and fluorescence intensity data of different fluorescence channels.

[0060] 5) Construct a signal matrix: Arrange the FSC, SSC, and fluorescence intensity data of different cell fluorescence channels obtained during the detection process and construct them into a matrix form, where the rows correspond to different cell samples and the columns represent the signal intensity of different channels. Record the obtained FSC matrix, SSC matrix, and fluorescence matrix.

[0061] S20, establishing regression relationship: according to the experimental parameters, establishing linear regression relationships between sample cell concentration, detection sample liquid flow rate, incident light intensity and FSC, SSC and cell fluorescence, respectively, which are respectively recorded as FSC linear regression relationship, SSC linear regression relationship and cell fluorescence linear regression relationship. The specific implementation methods are as follows:

[0062] 1) Organize the experimental parameter data: extract the cell concentration data, flow rate data and incident light intensity data recorded in S10 as independent variables; extract the FSC, SSC and fluorescence intensity matrix obtained in S10 as dependent variables.

[0063] 2) Establish a regression model: Use a linear regression algorithm, take cell concentration data as the independent variable, and FSC as the dependent variable, establish a linear regression relationship between the two, and obtain the FSC linear regression equation. Similarly, establish a linear regression relationship between SSC and cell concentration, and obtain the SSC linear regression equation. Take flow rate as the independent variable and each cell fluorescence as the dependent variable, establish a linear regression relationship between fluorescence intensity and flow rate, and obtain the linear regression equation of each cell fluorescence.

[0064] 3) Model testing: Use model testing methods (such as R square, p value, etc.) to evaluate the goodness of fit of each linear regression equation to ensure that the established linear regression relationship can fully reflect the linear correlation between variables.

[0065] 4) Relationship recording: the linear regression equation between FSC and cell concentration is recorded as FSC linear regression relationship, the linear regression equation between SSC and cell concentration is recorded as SSC linear regression relationship, and the linear regression equation between each cell fluorescence and flow velocity is recorded as cell fluorescence linear regression relationship.

[0066] The regression relationship formula for construction is as follows:

[0067] 1.FSC linear regression relationship:

[0068] FSC = α1 × C + b1;

[0069] Among them, FSC represents the forward scatter signal, C represents the cell concentration, and a1 and b1 are the regression coefficients to be determined.

[0070] This formula represents the linear regression relationship between FSC and cell concentration C.

[0071] 2.SSC linear regression relationship:

[0072] SSC=a2×C+b2;

[0073] Among them, SSC represents the side scatter signal, C represents the cell concentration, and a2 and b2 are the regression coefficients to be determined.

[0074] This formula represents the linear regression relationship between SSC and cell concentration C.

[0075] 3. Cell fluorescence linear regression relationship:

[0076] F = a3 × V + b3;

[0077] Among them, F represents a certain cell fluorescence signal, V represents flow velocity, and a3 and b3 are regression coefficients to be determined.

[0078] This formula represents the linear regression relationship between cell fluorescence F and flow velocity V.

[0079] The above method steps use the linear regression model to establish the linear relationship between variables, and the fitting effect of the linear regression equation is guaranteed by model testing. The recorded linear regression relationship provides a basis for the subsequent identification of diffraction effects. The implementation process uses the linear regression algorithm in statistics and machine learning to find the best fitting linear regression equation model between the given independent variables and dependent variables based on the data. This method can effectively establish a linear correlation model between experimental parameters and signals.

[0080] S30,

[0081] Obtaining a first nonlinear fluorescence signal: Based on the FSC linear regression relationship, the SSC linear regression relationship, and the cell fluorescence linear regression relationship, screening a fluorescence signal whose difference from any regression relationship is greater than a preset threshold, and recording it as the first nonlinear fluorescence signal. The specific implementation method is as follows:

[0082] 1) Calculate the predicted value: Substitute the FSC linear regression relationship obtained in S20 into the cell concentration data to calculate the predicted value of FSC; substitute the SSC linear regression relationship into the cell concentration to calculate the predicted value of SSC; substitute the cell fluorescence linear regression relationship into the flow rate data to calculate the predicted value of each cell fluorescence.

[0083] 2) Calculate residuals: Subtract the predicted values ​​from the actual observed values ​​recorded in S10 to calculate the residuals (residuals) of FSC, SSC and fluorescence of each cell. The residuals reflect the deviation between the predicted values ​​and the actual values.

[0084] 3) Setting the threshold: According to experimental experience, the threshold of the residual is set, such as 20% of the predicted value. If it is greater than the threshold, it means that there is a large deviation between the predicted value and the actual observed value, which does not conform to the linear relationship.

[0085] 4) Screening signals: Compare the absolute value of the residual of each signal with the set threshold, and screen out the signals whose absolute value of the residual is greater than the threshold. These signals have a large deviation from the linear regression relationship and are recorded as the first nonlinear fluorescence signal.

[0086] 5) Recording signals: Recording the first nonlinear fluorescence signal that is screened out and does not conform to the linear regression relationship, in preparation for subsequent analysis of the diffraction effect.

[0087] Generally, the preset threshold is set to the mean value of the residual minus 3 times the standard deviation. If the preset threshold is less than zero, the mean value of the residual minus the standard deviation is directly used.

[0088] In addition, a method for calculating the preset threshold may be:

[0089] 1) Collect multiple sets of experimental data, including independent variables (cell concentration, flow rate, etc.) and dependent variables (FSC, SSC, fluorescence signal, etc.).

[0090] 2) For each set of data, establish a linear regression model of the independent variables and dependent variables.

[0091] 3) Calculate the residual (the difference between the observed value and the predicted value) for each set of data.

[0092] 4) Take the absolute value of all residuals and calculate the mean μ and standard deviation σ.

[0093] 5) The preset threshold is set as:

[0094] T=μ+3σ, where the preset threshold is the mean value of the residual plus 3 times the standard deviation.

[0095] 6. That is, when the absolute value of the residual of a certain signal is greater than the preset threshold T, it is determined to be a nonlinear signal and may be affected by the diffraction effect.

[0096] This method is simple and reliable. By analyzing the overall residual distribution of multiple groups of data and using the statistical principle of "mean + 3σ", it is possible to identify abnormal signals that significantly deviate from linearity outside the normal variation range.

[0097] Among them, the mean value μ represents the normal variation center, the standard deviation σ represents the normal variation range, and “μ+3σ” covers about 99.7% of normal situations. Residuals beyond this interval are likely to come from abnormal nonlinear signals.

[0098] 3 standard deviations are commonly used judgment thresholds in statistics, and can also be adjusted to 2σ or 4σ as needed. This calculation method can provide a simple and effective way to determine the preset threshold.

[0099] By calculating the residual and setting the threshold, the fluorescence signals that deviate greatly from the established linear regression relationship and do not conform to the linear relationship can be effectively screened out. These signals are likely to be affected by the diffraction effect and have nonlinear errors. This method utilizes the residual concept in regression analysis and effectively realizes the detection of nonlinear signals by setting the threshold, laying the foundation for the subsequent identification of diffraction effects.

[0100] S40,

[0101] Changing parameters to obtain the second and third nonlinear fluorescence signals: changing the sample flow rate at the detection point so that the cell gap formed by the sample flow rate at the detection point maintains the time period, reducing the diffraction effect formed by the cell gap, and repeatedly performing steps S20 to S30 to obtain the second nonlinear fluorescence signal with the sample flow rate at the detection point changed; changing the incident light intensity, and repeatedly performing steps S20 to S30 to obtain the third nonlinear fluorescence signal with the sample flow rate at the detection point changed. The specific implementation method is as follows:

[0102] 1) Adjusting the flow rate: According to the flow rate value that shortens the intercellular gap time calculated in the previous step, the flow rate parameters are adjusted in the flow cytometer to shorten the holding time of the intercellular gap and reduce the diffraction effect at the gap.

[0103] 2) Repeat the test: Use the adjusted flow rate to repeat the flow cytometric test of the sample and record the various parameters obtained from the test, including the cell concentration after the adjusted flow rate, flow rate parameters, incident light intensity, etc.

[0104] 3) Model building: Based on the recorded parameters under the adjusted flow rate, step S20 is repeated to re-establish the linear regression model between the variables.

[0105] 4) Screening signals: According to the re-established linear regression model, step S30 is repeated to re-screen the nonlinear fluorescence signals that have a large deviation from the linear model.

[0106] 5) Recording signals: Recording the new nonlinear fluorescence signal detected under the condition of adjusting the flow rate as the second nonlinear fluorescence signal.

[0107] 6) Adjust the intensity of incident light: While keeping other parameters unchanged, adjust the intensity parameters of the incident light source and make the change.

[0108] 7) Repeat detection: using the adjusted incident light intensity, repeat steps S20-S30 to obtain a third nonlinear fluorescence signal.

[0109] By adjusting the flow rate parameters to reduce the cell gap time, the diffraction generated in the gap can be effectively reduced. The parameter modeling and signal acquisition processes are repeated to obtain the diffraction signal under the flow rate adjustment condition. In combination with the change in the incident light intensity, a more comprehensive nonlinear signal is obtained to provide sample data under different conditions for building a diffraction effect model.

[0110] By adjusting the flow rate parameters to reduce the cell gap time, the diffraction generated in the gap can be effectively reduced. The principle is as follows:

[0111] It takes a certain amount of time for the diffraction effect to occur. According to the diffraction principle, when the laser irradiates the cell sample, the cell will scatter, and since there is no scatterer in the cell gap, the light wave will diffract in the gap area. This diffraction wave needs a certain propagation time to form a detectable diffraction effect.

[0112] According to the calculation formula of the diffraction wave propagation time τ, τ is related to the propagation speed of the light wave in the medium, and τ will increase with the expansion of the gap area. At the same time, the light wave propagation time in the cell gap is related to the gap length and flow velocity.

[0113] One way to reduce the intercellular time is to increase the flow rate so that the cells flow quickly through the intercellular area. This will prevent the light waves in the intercellular space from having enough time to diffract significantly, thus weakening the diffraction effect. If the time is less than the propagation time of the diffracted wave τ, no significant diffraction will occur in the intercellular space, thus effectively reducing the diffraction effect.

[0114] Therefore, by accurately calculating the diffraction time τ, setting the flow rate value to be less than τ, and actually adjusting the flow rate parameters, the gap time can be effectively controlled in theory to reduce the probability of diffraction effect at the gap. This is the scientific principle basis for adjusting the flow rate to reduce the diffraction effect.

[0115] S50,

[0116] Parameter signal fusion: The experimental parameters are fused with the first nonlinear fluorescence signal into a first parameter matrix, the experimental parameters after the sample flow rate of the detection point is changed are fused with the second nonlinear fluorescence signal into a second parameter matrix, and the experimental parameters after the incident light intensity is changed are fused with the third nonlinear fluorescence signal into a third parameter matrix. The specific implementation methods are as follows:

[0117] 1) Constructing the first parameter matrix: Integrate the cell concentration, flow rate, incident light intensity recorded in S10 and the first nonlinear fluorescence signal screened in S30 to construct a parameter matrix, where rows represent different cell samples and columns represent different parameters.

[0118] 2) Constructing a second parameter matrix: integrating the cell concentration, flow rate, incident light intensity and the corresponding screened second nonlinear fluorescence signal after the flow rate adjustment in S40 to construct a parameter matrix.

[0119] 3) Constructing a third parameter matrix: integrating the cell concentration, flow rate, incident light intensity after the incident light intensity is adjusted in S40, and the corresponding screened third nonlinear fluorescence signal to construct a parameter matrix.

[0120] 4) Matrix fusion: The above three parameter matrices are fused horizontally and in parallel to obtain the final model input matrix, which is used to train the diffraction effect prediction model.

[0121] Assume that all cell sample parameters constitute matrix X and nonlinear fluorescence signals constitute matrix Y, then

[0122] First parameter matrix: M1 = [X, Y];

[0123] Second parameter matrix: M2 = [X', Y'];

[0124] The third parameter matrix: M3 = [X ″ , Y ″ ];

[0125] Among them, X, X', X ″ Corresponding to the parameters under original, variable flow rate and variable light intensity conditions,

[0126] Y, Y', Y ″ Corresponding to the nonlinear fluorescence signal obtained by screening.

[0127] The construction of the parameter matrix can centrally integrate variables and responses, making it easier to input the model for training. Parameter matrix fusion can increase the sample size and information volume, improving the effect of model training. This method is beneficial to the storage of variable information and the compatibility of the input model through matrix integration and fusion.

[0128] S60,

[0129] Model calculation: Using the pre-trained diffraction effect analysis model, inputting the first parameter matrix, the second parameter matrix and the third parameter matrix, the specific implementation method of obtaining the diffraction effect equation is as follows:

[0130] 1) Construct a neural network model: Select a multi-layer feedforward neural network as the structural framework of the diffraction effect analysis model, design the number of network layers and neurons according to the input and output dimensions, and construct the network model.

[0131] 2) Obtain training samples, specifically:

[0132] Step 1: Perform a flow cytometry experiment using flow fluorescent microspheres, and use steps S10 to S50 to obtain the first, second and third parameter matrices of the experiment;

[0133] Step 2: Change the experimental parameters, conduct multiple groups of experiments, and obtain the first, second, and third parameter matrices of the multiple groups of experiments;

[0134] Step 3, using the flow cytometry experimental parameters performed by the flow cytometry fluorescent microspheres to calculate the standard output fluorescence signal;

[0135] Step 4: According to the standard fluorescence signal, the diffraction effect equations corresponding to the first, second and third parameter matrices of multiple groups of experiments are calculated by linear regression, which can be expressed as:

[0136] I corrected =I observed -f(C, V, L);

[0137] Among them, I corrected represents the corrected fluorescence signal, I observed is the observed raw fluorescence signal, and the function f(C, V, L) represents the diffraction effect equation, which includes variables such as cell concentration C, flow velocity V and incident light intensity L.

[0138] 3) Network training: Use the training data to train the constructed BP neural network model, and gradually adjust the network weights and threshold parameters through the error back propagation algorithm so that the predicted output of the network model can approach the standard output and achieve model optimization.

[0139] 4) Model evaluation: Use the validation set to evaluate the network model, and determine the convergence and effectiveness of the network model training through indicators such as loss function and accuracy. Obtain a reliable diffraction effect analysis model.

[0140] 5) Input experimental data: The first, second and third parameter matrices constructed in S50 are used as input and fed into the trained neural network model.

[0141] 6) Model prediction: The neural network model performs forward calculation and prediction on the input matrix data and outputs the parameters of the diffraction effect equation.

[0142] Among them, the specific structure of the BP neural network used in the diffraction effect analysis model is as follows:

[0143] 1) Input layer: The input layer contains m input nodes, where m is the number of input variables (independent variables such as cell concentration, flow rate, and incident light intensity).

[0144] 2) Hidden layer: Set up two hidden layers, the first hidden layer contains n1 nodes, and the second hidden layer contains n2 nodes. The number of hidden layer nodes n1 and n2 is determined according to the input dimension, output dimension and model complexity, and the default value of the BP neural network can be used directly.

[0145] 3) Output layer: The output layer contains k nodes, where k is the number of output variables (parameters of the diffraction effect equation).

[0146] 4) Connection method: The input layer is fully connected to the first hidden layer, the first hidden layer is fully connected to the second hidden layer, and the second hidden layer is fully connected to the output layer.

[0147] 5) Activation function: The hidden layer nodes use the ReLU activation function, and the output layer nodes use the linear activation function.

[0148] 6) Network structure: A three-layer fully connected network structure with two hidden layers added in the middle to increase the network's ability to fit complex mapping relationships.

[0149] 7) Network training: Use the error back propagation algorithm to optimize the network parameters through gradient descent, minimize the loss function, and make the predicted output close to the actual output.

[0150] 8) Model evaluation: Use indicators such as mean square error, R square, and accuracy to evaluate the model effect.

[0151] The fully connected feedforward neural network structure includes an input layer, multiple hidden layers, and an output layer, and approximates complex diffraction effect relationships through nonlinear mapping. Network training and evaluation ensure the model effect, and finally the network is deployed to achieve efficient prediction. This structure can meet the functional requirements of the diffraction effect model.

[0152] This method extracts the diffraction effect equation by constructing and training a neural network model. Neural networks can perform complex nonlinear mapping, learn implicit laws from large amounts of data, construct nonlinear relationships between diffraction effects and multiple variables, and achieve high-precision predictions.

[0153] S70, diffraction effect correction: The FSC matrix, SSC matrix and cell fluorescence matrix are respectively subjected to inverse calculations using the obtained diffraction effect equations, and the specific implementation methods of using the obtained results as the final output FSC matrix, SSC matrix and cell fluorescence matrix are as follows:

[0154] 1) Extraction matrix: Extract the FSC matrix, SSC matrix and multiple cell fluorescence intensity matrices recorded during the detection process from the original results of the flow cytometry experiment.

[0155] 2) Inverse operation equation: The diffraction effect equation obtained in S60 is subjected to inverse operation to obtain a correction equation that can be used to correct the diffraction effect.

[0156] 3) Matrix operation: Substitute the extracted FSC matrix and SSC matrix into the correction equation for operation respectively, remove the influence of diffraction effect by matrix-matrix operation, and obtain the corrected result matrix.

[0157] 4) Repeat the operation: Similarly, the modified equation is used to perform matrix operations on the cell fluorescence matrix to obtain a modified fluorescence result matrix.

[0158] 5) Output results: The corrected FSC matrix, SSC matrix and cell fluorescence matrix are used as the final output results to provide accurate flow cytometry experimental analysis results.

[0159] Matrix operations can correct the diffraction effects of a batch of signal results at the same time, avoiding the complexity of signal-by-signal operations. Matrix operations can also ensure the consistency of processing. This method uses the inverse operation characteristics of the matrix to eliminate the error introduced by diffraction as much as possible and improve the accuracy of the results.

[0160] A second aspect of the present invention provides a computer-readable storage medium, wherein the computer-readable storage medium stores program instructions, and when the program instructions are executed, they are used to execute the above-mentioned method for reducing diffraction effect in a flow cytometer.

[0161] A third aspect of the present invention provides a system for reducing diffraction effects in a flow cytometer, which includes the above-mentioned computer-readable storage medium.

[0162] Specifically, the principle of the present invention is:

[0163] 1) Establishing a linear regression relationship between variables can effectively detect nonlinear signals that do not conform to the linear relationship. These signals are likely to be diffraction signals.

[0164] 2) Change the flow rate and incident light intensity to repeatedly obtain diffraction signals. Since the flow rate is adjusted, the cell gap is maintained for a short time, which will interfere with the diffraction effect. Following the law of controlled variables, the influence of different variables on the signal can be separated.

[0165] 3) By utilizing the powerful nonlinear fitting and generalization capabilities of neural networks, a complex relationship model of diffraction effects containing multiple variables can be established.

[0166] 4) Matrix operations can be performed on large batches of signals at the same time, avoiding tedious signal-by-signal processing and ensuring consistency of results.

[0167] 5) By using the inverse operation of the diffraction effect equation, the diffraction effect component predicted by the equation can be effectively eliminated and the detection results can be accurately corrected.

[0168] 6) This scheme makes full use of statistics, neural networks, matrix operations and other theories, adopts the control variable method and inverse operation algorithm, and scientifically and systematically realizes the detection and elimination of diffraction effects.

[0169] 7) The whole method can form a closed-loop feedback control, which can further improve the accuracy of subsequent experimental parameter optimization by eliminating the diffraction effect.

[0170] The above description is only a specific implementation mode of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

Claims

1. A method for reducing diffraction effect in a flow cytometer, characterized in that: The steps include: S10, obtaining experimental parameters in the flow cytometer experiment process, wherein the experimental parameters include sample cell concentration, sample flow rate at the detection point, incident light intensity, and fluorescence signals obtained in the experiment, wherein the fluorescence signals include FSC, SSC, and cell fluorescence, and establishing FSC matrix, SSC matrix, and cell fluorescence matrix; S20. According to the experimental parameters, establish linear regression relationships between sample cell concentration, detection sample liquid flow rate, incident light intensity, FSC, SSC and cell fluorescence, respectively, which are recorded as FSC linear regression relationship, SSC linear regression relationship and cell fluorescence linear regression relationship, respectively; specifically: adopt a linear regression algorithm, take cell concentration data as an independent variable, take FSC as a dependent variable, establish a linear regression relationship between the two, and obtain an FSC linear regression equation; establish a linear regression relationship between SSC and cell concentration, and obtain an SSC linear regression equation; take flow rate as an independent variable, take each cell fluorescence as a dependent variable, establish a linear regression relationship between fluorescence intensity and flow rate, and obtain each cell fluorescence linear regression equation; S30, based on the FSC linear regression relationship, the SSC linear regression relationship, and the cell fluorescence linear regression relationship, screening a fluorescence signal whose difference from any regression relationship is greater than a preset threshold, and recording it as a first nonlinear fluorescence signal; Specifically: 1) Calculate the predicted value: Substitute the FSC linear regression equation obtained in S20 into the cell concentration data to calculate the predicted value of FSC; Substitute the SSC linear regression equation into the cell concentration to calculate the predicted value of SSC; Substitute the cell fluorescence linear regression equation into the flow rate data to calculate the predicted value of each cell fluorescence; 2) Calculate the residuals: subtract the predicted values ​​from the actual observed values ​​recorded in S10 to calculate the residuals of FSC, SSC and fluorescence of each cell; 3) Setting the threshold: Setting the threshold of the residual according to experimental experience. If the residual is greater than the threshold, it means that there is a large deviation between the predicted value and the actual observed value, which does not conform to the linear relationship; 4) Screening signals: comparing the absolute value of the residual of each signal with the threshold, screening out the signal whose absolute value of the residual is greater than the threshold, and recording it as the first nonlinear fluorescence signal; S40, changing the sample flow rate at the detection point so that the cell gap retention time formed by the sample flow rate at the detection point is shortened, reducing the diffraction effect formed by the cell gap, and repeating steps S20-S30 to obtain a second nonlinear fluorescence signal with the sample flow rate at the detection point changed; changing the incident light intensity, repeating steps S20-S30 to obtain a third nonlinear fluorescence signal with the sample flow rate at the detection point changed; S50, fusing the experimental parameters and the first nonlinear fluorescence signal into a first parameter matrix, fusing the experimental parameters after changing the flow rate of the sample at the detection point and the second nonlinear fluorescence signal into a second parameter matrix, and fusing the experimental parameters after changing the incident light intensity and the third nonlinear fluorescence signal into a third parameter matrix; S60, using a pre-trained diffraction effect analysis model, inputting the first parameter matrix, the second parameter matrix and the third parameter matrix, to obtain a diffraction effect equation; S70, respectively performing inverse operations on the FSC matrix, SSC matrix and cell fluorescence matrix using the obtained diffraction effect equation, and using the obtained results as the final output FSC matrix, SSC matrix and cell fluorescence matrix.

2. A method for reducing diffraction effect in a flow cytometer according to claim 1, characterized in that: The step of establishing the linear regression relationship between the sample cell concentration, the flow rate of the detection sample liquid, the incident light intensity and the FSC, SSC and cell fluorescence respectively according to the experimental parameters is specifically: Using the experimental parameters as independent variables; The obtained FSC matrix, SSC matrix and cell fluorescence matrix are used as dependent variables; A linear regression algorithm is used to establish and fit the linear regression relationship between the independent variable and the dependent variable.

3. A method for reducing diffraction effect in a flow cytometer according to claim 2, characterized in that: The step of screening the fluorescent signal whose difference from any regression relationship is greater than a preset threshold is specifically: Calculate the predicted value, the predicted value is obtained by substituting the obtained FSC linear regression relationship into the experimental parameters, and the difference between the predicted value and the fluorescence signal is taken as the residual; Setting a preset threshold, wherein the threshold is used to set a residual threshold based on experimental experience; The screening signal is to compare the absolute value of the residual of each signal with a preset threshold, and screen out the signal whose absolute value of the residual is greater than the preset threshold.

4. A method for reducing diffraction effect in a flow cytometer according to claim 3, characterized in that: The preset threshold is set to the mean value of the residual minus 3 times the standard deviation. If the preset threshold is less than zero, the mean value of the residual minus the standard deviation is directly used.

5. A method for reducing diffraction effect in a flow cytometer according to claim 1, characterized in that: The steps of establishing and training the diffraction effect analysis model specifically include: Get training samples, specifically: Step 1: Perform a flow cytometry experiment using flow fluorescent microspheres, and use steps S10 to S50 to obtain the first, second, and third parameter matrices of the experiment; Step 2: Change the experimental parameters, conduct multiple groups of experiments, and obtain the first, second, and third parameter matrices of the multiple groups of experiments; Step 3, using the flow cytometry experimental parameters performed by the flow cytometry fluorescent microspheres to calculate the standard output fluorescence signal; Step 4: Calculate the diffraction effect equations corresponding to the first, second and third parameter matrices of multiple groups of experiments according to the standard fluorescence signal; The neural network is trained, specifically: a neural network is trained using the training samples to obtain a diffraction effect analysis model.

6. A method for reducing diffraction effect in a flow cytometer according to claim 5, characterized in that: The neural network is a BP neural network.

7. A method for reducing diffraction effect in a flow cytometer according to claim 6, characterized in that: The specific structure of the BP neural network includes: An input layer, used for inputting data of the first parameter matrix, the second parameter matrix and the third parameter matrix; The first hidden layer and the second hidden layer are used to extract the data features of the parameter matrix and connect the input layer and the output layer; The output layer is used to output the predicted diffraction effect equation results; The input layer and the first hidden layer, the first hidden layer and the second hidden layer, and the second hidden layer and the output layer are fully connected structures; The activation functions of the first hidden layer and the second hidden layer are Sigmoid functions, and the activation function of the output layer is a linear function; The BP neural network is trained through an error back propagation algorithm to minimize the loss function of the predicted output and realize the extraction of the diffraction effect equation.

8. A method for reducing diffraction effect in a flow cytometer according to claim 1, characterized in that: The first parameter matrix, the second parameter matrix and the third parameter matrix are all constructed by horizontally splicing the independent variable parameter matrix and the dependent variable parameter matrix.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores program instructions, and when the program instructions are executed, they are used to execute the method for reducing diffraction effect in a flow cytometer according to any one of claims 1 to 8.

10. A system for reducing diffraction effect in a flow cytometer, characterized in that: A computer-readable storage medium comprising the computer-readable storage medium of claim 9.

Images