A method, medium and system for controlling the sample flow rate of a flow cytometer

Abstract

The present invention provides a method, medium and system for controlling the sample flow rate of a flow cytometer, belonging to the technical field of flow cytometers. The method, medium and system include: selecting the optimal flow rate of the sample in the basic flow rate database as the current flow rate of the flow cytometry experiment according to the type, state, volume, concentration of the sample and the type and concentration of the fluorescent agent; acquiring the fluorescence signals of the flow cytometry experiment in real time, including FSC, SSC and cell fluorescence; establishing a fluorescence feature matrix according to the fluorescence signals; calculating the flow rate error vector of the fluorescence feature matrix by using a neural network; adjusting the current flow rate according to the flow rate error vector to obtain an optimized flow rate; and performing deviation correction calculation on the fluorescence signals before adjusting the flow rate according to the flow rate error vector and updating. Compared with the prior art, the present invention greatly solves the problems of low efficiency and time-consuming manual adjustment of the flow rate and inaccurate flow rate selection in the process of adjusting the flow rate of the sample in the current flow cytometer.

Description

Technical Field

[0001] The present invention belongs to the technical field of flow cytometers, and more particularly, relates to a method, medium, and system for controlling the sample flow rate of a flow cytometer. Background Art

[0002] Flow cytometry is a high-throughput, multi-parameter cell analysis technology. Generally, a flow cytometer is used to perform flow cytometry experiments. Its basic principle is to detect cells injected through fluid injection by optical signals in a flowing state and classify and count them. This technology has been widely applied in the fields of life science, clinical testing, etc.

[0003] In flow cytometry experiments, the cells in the sample are wrapped by sheath fluid to form droplets, which pass through the laser intersection area at a certain speed, exciting fluorescence labels and detecting various optical signals. The flow rate is one of the key parameters affecting the detection effect. If the flow rate is too fast, the passing time of cells will be too short to obtain sufficient optical signals. If the flow rate is too slow, cell aggregation and accumulation will occur, and the signals will overlap and deform, making it impossible to correctly identify each cell. Therefore, selecting an appropriate flow rate is crucial for ensuring the accuracy of flow cytometry.

[0004] Currently, the setting of the flow rate mainly relies on empirical configuration. Different scholars determine the flow rate based on experience such as sample type, cell size, labeling method, etc. This method has the problems of low efficiency and time-consuming manual adjustment of the flow rate, and inaccurate flow rate selection. Summary of the Invention

[0005] In view of this, the present invention provides a method, medium, and system for controlling the sample flow rate of a flow cytometer, which can solve the problems of low efficiency and time-consuming manual adjustment of the flow rate and inaccurate flow rate selection in the current flow cytometer during the flow rate adjustment process.

[0006] The present invention is implemented as follows:

[0007] The first aspect of the present invention provides a method for controlling the sample flow rate of a flow cytometer, which includes the following steps:

[0008] S10. Select the optimal flow rate of the sample as the current flow rate of the flow cytometry experiment in the basic flow rate database according to the type, state, volume, concentration of the sample, and the type and concentration of the fluorescent agent;

[0009] S20. Real-time obtain the fluorescence signals of the flow cytometry experiment, including FSC, SSC, and cell fluorescence;

[0010] S30. Establish a fluorescence feature matrix according to the fluorescence signals;

[0011] S40. Use a neural network to calculate the flow rate error vector of the fluorescence feature matrix;

[0012] S50. Adjust the current flow rate according to the flow rate error vector to obtain an optimized flow rate;

[0013] S60. Perform deviation correction calculation and update on the fluorescence signal before adjusting the flow rate according to the flow rate error vector.

[0014] The technical effects of a flow cytometer sample flow rate control method provided by the present invention are as follows: In step S10, a database storing the optimal flow rates corresponding to different sample conditions and fluorescent agent conditions is established in advance, and the most matching sample record is searched in the database according to the specific parameters of the sample, and its optimal flow rate is selected as the current flow rate, which can realize personalized and precise configuration of the flow rate. Compared with manually configuring the flow rate based on experience, this can automatically find the optimal flow rate for samples of different types, states, volumes, and concentrations, avoiding problems such as cell aggregation, signal weakening, and reduced analysis effect caused by improper flow rate, and ensuring the sensitivity and specificity of flow cytometry detection. At the same time, when the sample state changes, it is also possible to quickly re-search to find a new optimal flow rate, realizing intelligent and dynamic optimization of the flow rate.

[0015] In step S20, the fluorescence intensity signals of each cell in the multi-color fluorescence channels are obtained in real time, and the multi-dimensional fluorescence characteristics of each cell are recorded. This provides basic raw data for subsequent establishment of a fluorescence characteristic matrix to realize cell recognition and counting. Compared with interval sampling, real-time acquisition can obtain the complete time series signal of cells, prevent missing short-term signals, and thus improve the integrity of the acquired signals. Real-time acquisition of fluorescence signals can also feedback and adjust the flow rate to realize closed-loop control of the flow rate.

[0016] In step S30, the multi-color fluorescence signals of all cells are extracted and constructed into a matrix, where the rows represent each cell and the columns represent different fluorescence channels. The matrix representation form is convenient for subsequent operations and analyses such as signal normalization, dimensionality reduction, and classification. Compared with the original signal, the characteristic matrix condenses the comprehensive fluorescence characteristics of each cell, and the distinguishability of the signal can be enhanced by processing the matrix, improving the accuracy of subsequent cell recognition.

[0017] In step S40, a neural network is used to model the fluorescence characteristic matrix, which can realize a complex non-linear mapping from the high-dimensional fluorescence characteristic space to the flow rate error. Compared with methods such as simple linear regression, the neural network can extract the complex internal associations between features through data training, and can also predict the flow rate error more accurately for unknown samples. The obtained error vector reflects the deviation degree between the current flow rate and the optimal flow rate, providing a basis for subsequent flow rate optimization.

[0018] Step S50 constructs a mapping model from the error vector to the flow rate adjustment amount, which can automatically analyze the error vector and calculate the adjustment to be made to the current flow rate, realizing closed-loop control of the flow rate. By updating the parameters of the model online, the flow rate adjustment strategy can be optimized at any time to adapt to sample changes. Compared with manual flow rate adjustment, this can achieve intelligent, dynamic, and precise optimization of the flow rate, significantly improving the performance of flow cytometry detection and analysis.

[0019] Step S60 performs deviation correction calculation and update on the fluorescence signal before adjusting the flow rate according to the flow rate error vector, which can reduce the error of the fluorescence signal before adjusting the flow rate and contribute to obtaining more accurate flow cytometry experimental result data.

[0020] Based on the above technical solutions, a method for controlling the sample flow rate of a flow cytometer according to the present invention can be further improved as follows:

[0021] Among them, the step of selecting the optimal flow rate of the sample in the basic flow rate database as the current flow rate of the flow experiment according to the type, state, volume, concentration of the sample and the type, concentration of the fluorescent agent specifically includes:

[0022] Establish a basic flow rate database, and collect the optimal flow rate data of different types of samples under different states, volumes, concentrations and different types, concentrations of fluorescent agents;

[0023] Before conducting a flow experiment, determine the parameters of this experiment, including the type, state, volume, concentration of the sample and the type and concentration of the fluorescent agent;

[0024] According to the experimental parameters, first try to search for a completely matching record in the database. If there is a completely matching record, use the optimal flow rate of this record as the current flow rate;

[0025] If there is no completely matching record, use a scoring function to calculate the matching degree of all records with the experimental parameters of this time, and select the flow rate of the record with the highest score as the current flow rate.

[0026] Furthermore, the scoring function is obtained by fitting the optimal flow rate data of the existing samples in the database under different states, volumes, concentrations and different types, concentrations of fluorescent agents using a linear regression function.

[0027] Among them, in the step of establishing a fluorescence feature matrix according to the fluorescence signal, the process of normalizing the established fluorescence feature matrix is also included.

[0028] Among them, the flow rate error vector is the difference between the current flow rate and the optimal flow rate.

[0029] Among them, the step of adjusting the current flow rate according to the flow rate error vector to obtain an optimized flow rate specifically includes:

[0030] Input the current flow rate parameter and the flow rate error vector;

[0031] Construct an adjustment function from the error vector to the flow rate adjustment amount;

[0032] Calculate the flow rate adjustment amount according to the adjustment function;

[0033] Update the current flow rate according to the flow rate adjustment amount.

[0034] Furthermore, the adjustment function is a linear regression model or a three-layer fully connected neural network.

[0035] Among them, the step of correcting and updating the fluorescence signal before adjusting the flow rate according to the flow rate error vector specifically includes:

[0036] Input the cell fluorescence feature matrix and the flow rate error vector before adjusting the flow rate;

[0037] Use the fluorescence matrix corresponding to each cell in the sample to establish a linear regression model;

[0038] Fit the relationship between the input cell fluorescence feature matrix and the flow rate error vector, and solve to obtain the parameters of the linear regression model;

[0039] Use the obtained regression model to calculate the fluorescence signal before adjusting the flow rate, and obtain the new corrected fluorescence signal.

[0040] The second aspect of the present invention provides a computer-readable storage medium, wherein program instructions are stored in the computer-readable storage medium, and when the program instructions run, they are used to execute the above-mentioned flow cytometer sample flow rate control method.

[0041] The third aspect of the present invention provides a flow cytometer sample flow rate control system, which includes the above-mentioned computer-readable storage medium.

[0042] Compared with the prior art, the beneficial effects of a flow cytometer sample flow rate control method, medium and system provided by the present invention are:

[0043] 1. By establishing a database for storing the optimal flow rates under different sample conditions, intelligent and precise configuration of the flow rate can be achieved, subjective errors caused by empirical configuration can be avoided, and the accuracy of flow rate setting can be improved.

[0044] 2. Obtain the cell fluorescence signal in real time, record the multi-parameter time series of each cell, and obtain complete and reliable original detection data.

[0045] 3. Construct a fluorescence matrix expressing the multi-dimensional characteristics of cells to improve the subsequent recognition and analysis effects.

[0046] 4. Use a neural network to predict the flow velocity error, achieve an accurate mapping from high-dimensional features to flow velocity adjustment, and provide a basis for closed-loop control.

[0047] 5. Establish a model from the flow velocity error to the adjustment amount through a data-driven method to achieve automatic, dynamic, and precise optimization of the flow velocity.

[0048] 6. Correct the fluorescence signal to eliminate the influence of flow velocity changes and improve the analysis robustness.

[0049] 7. Overall, an intelligent closed-loop control of the flow velocity is achieved. The change in the sample state can trigger the automatic adjustment of the flow velocity to continuously maintain the best detection state.

[0050] Compared with the empirical self-configuration, the present invention greatly improves the intelligence level, dynamic adjustment ability, and accuracy of flow velocity selection and optimization, can significantly improve the performance of flow cytometry analysis, and solves the problems of low efficiency and time-consuming manual adjustment of the flow velocity and inaccurate flow velocity selection in the current flow cytometer during the flow velocity adjustment process of the sample. Brief Description of the Drawings

[0051] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required to be used in the description of the embodiments of the present invention will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0052] Figure 1 It is a flowchart of a method for controlling the sample flow velocity of a flow cytometer provided by the present invention. Detailed Embodiments

[0053] As Figure 1 shown, it is an embodiment of a method for controlling the sample flow velocity of a flow cytometer provided by the first aspect of the present invention. In this embodiment, the following steps are included:

[0054] S10. Select the optimal flow velocity of the sample as the current flow velocity of the flow experiment in the basic flow velocity database according to the type, state, volume, concentration of the sample, and the type and concentration of the fluorescent agent.

[0055] S20. Real-time obtain the fluorescence signals of the flow experiment, including FSC, SSC, and cell fluorescence.

[0056] S30. Establish a fluorescence feature matrix according to the fluorescence signals.

[0057] S40. Use a neural network to calculate the flow velocity error vector of the fluorescence feature matrix.

[0058] S50. Adjust the current flow rate according to the flow rate error vector to obtain an optimized flow rate;

[0059] S60. Perform deviation correction calculation and update on the fluorescence signal before adjusting the flow rate according to the flow rate error vector.

[0060] Each of the above steps is described as follows in the specific implementation process:

[0061] In step S10, we need to select the optimal flow rate of the sample in the basic flow rate database as the current flow rate of the flow cytometry experiment according to the type, state, volume, concentration of the sample, and the type and concentration of the fluorescent agent.

[0062] First, we need to establish a basic flow rate database, which collects and stores the optimal flow rate data of different types of samples under different states, volumes, and concentrations in combination with different types and concentrations of fluorescent agents. The database can be implemented using a relational database such as MySQL or a non-relational database such as MongoDB. Each record in the database includes the following fields:

[0063] Sample type: String, representing the type name of the sample, such as blood sample, cell lysate sample, etc.

[0064] Sample state: String, representing the state of the sample, such as fresh state, frozen state, etc.

[0065] Sample volume: Number, representing the sample volume, with the unit of ml or μl.

[0066] Sample concentration: Number, representing the concentration of the target component in the sample, with the unit depending on the type of sample, such as the number of cells / ml.

[0067] Fluorescent agent type: String, representing the name of the fluorescent agent used.

[0068] Fluorescent agent concentration: Number, representing the usage concentration of the fluorescent agent, with the unit generally being nM or μM.

[0069] Optimal flow rate: Number, representing the optimal flow rate used in combination with this fluorescent agent under the above conditions, with the unit of μl / min.

[0070] The database can be indexed according to the above fields to improve the subsequent query efficiency. The database initialization can be filled by collecting existing sample flow rate data.

[0071] When performing a flow cytometry experiment, we first need to determine the following parameters for this experiment:

[0072] Sample type : String

[0073] Sample state : String

[0074] Sample volume : A number, with the unit of ml or μl

[0075] Sample concentration : A number, and the unit depends on the type of the sample

[0076] Type of fluorescent agent : String

[0077] Concentration of fluorescent agent : A number, with the unit of nM or μM

[0078] If there is a record with a perfect match in the database, the optimal flow rate of this record can be directly obtained as the current flow rate for this flow cytometry experiment.

[0079] If there is no record with a perfect match in the database, approximate matching is required. For this purpose, we can construct a scoring function to evaluate the matching degree between any database record and the parameters of this experiment:

[0080] Define the scoring function as:

[0081]

[0082] Where:

[0083] : Vector of database records, representing the parameters of the experiment

[0084] : Weight of the i-th score item

[0085] : Similarity function for calculating the score of the i-th feature dimension

[0086] : Number of feature dimensions, which is the number of sample parameters in this problem

[0087] For different types of features, different similarity functions are selected , for example:

[0088] Categorical features (type, status, etc.): Adopt the indicator function, where the same is 1 and otherwise 0

[0089] Numerical features (volume, concentration, etc.): Adopt the Gaussian function , representing the numerical distance

[0090] Weight Methods such as information entropy can be used to evaluate the importance of each dimension. The higher the information entropy of a feature, the greater its weight.

[0091] For the training dataset, use methods such as linear regression to learn the weight parameters so that highly similar samples obtain higher scores.

[0092] The objective function is:

[0093]

[0094] where S is the training sample set, sim(x, y) represents the manual similarity annotation of x and y, and the loss function is such as the squared error.

[0095] We can traverse all database records, calculate their matching scores with the current experimental parameters, and then select the optimal flow rate of the record with the highest matching score as the current flow rate. If there are multiple records with the same highest score, then take the average flow rate of these records.

[0096] Through the above method, we can intelligently select the most suitable flow rate value for the current sample and experimental conditions from the basic flow rate database to ensure the optimization of the flow cytometry experiment effect. This flow rate selection method based on the pre-built database and scoring function can handle both exact matching and approximate matching of sample parameters, ensuring that reasonable flow rate values can be found under various experimental conditions, thereby improving the intelligence and reusability of flow cytometry experiments.

[0097] Step S20 is to obtain the fluorescence signal of the flow cytometry experiment in real time. Flow cytometry labels different subpopulations of cells with multi-color fluorescent antibodies and detects their fluorescence signals, thereby analyzing the types and quantities of cells. Then the specific method for obtaining the fluorescence signal in real time is:

[0098] 1) Configure the acquisition channels of the flow cytometer. The flow cytometer can be configured with multiple acquisition channels, and each channel is set to acquire a fluorescence wavelength. For example, channel FL1 acquires the green fluorescence of FITC (wavelength 525nm), channel FL2 acquires the red fluorescence of PE (wavelength 575nm), etc. According to the types of fluorescent antibodies used in this experiment, configure the acquisition channels in advance according to their excitation and emission wavelengths.

[0099] 2) Adjust the flow rate so that the time for a single cell to pass through the laser beam is within the range of 1 - 100 microseconds. If the flow rate is too fast, the passing time of a single cell is too short to obtain sufficient fluorescence signals. If the flow rate is too slow, cell accumulation may cause multiple cells to pass through simultaneously, resulting in distorted signals. It is necessary to select an appropriate flow rate.

[0100] 3) The sample is wrapped by sheath fluid and injected into the laser crossing area through a pipeline. When the cells pass through the laser beam, the fluorescent antibodies are excited to emit fluorescence.

[0101] 4) The fluorescence signal is collected by a photomultiplier tube (PMT). Fluorescence of different wavelengths enters the PMT corresponding to the channel.

[0102] 5) The PMT converts the captured optical signal into an electrical signal, which is amplified and then converted into a digital signal by an ADC.

[0103] 6) The digital signal is input into a computer for storage and analysis. Acquisition software needs to be configured on the computer side, and parameters such as the sampling rate and channels are set.

[0104] 7) The software interface displays in real time the signal curves such as the fluorescence intensity of each channel. The software continuously stores the digital signal as an FCS data file, which contains all-channel fluorescence signal information when each cell passes through.

[0105] 8) For the stored FCS file, post-processing can be performed to extract the fluorescence signal information of all cells and construct a fluorescence signal matrix for further analysis.

[0106] 9) The whole process continues to obtain all-channel fluorescence signal information of each cell in the flow cytometry experiment in real time.

[0107] Through the above series of processes of collection, amplification, conversion, storage and processing by equipment, the real-time acquisition of fluorescence signals in flow cytometry experiments is realized. This provides a basic data source for the subsequent identification and analysis of cell types and quantities.

[0108] The key points are summarized as:

[0109] 1) Configure fluorescence channels

[0110] 2) Adjust the flow rate

[0111] 3) Cells emit fluorescence after being excited by a laser when flowing through.

[0112] 4) The PMT collects and converts the fluorescence signal

[0113] 5) The computer side processes and stores the signal

[0114] 6) Display and store the signal curve in real time

[0115] 7) Continue to obtain the signal of each cell

[0116] Explanation of fluorescence: The optical signals collected by a flow cytometer include scattered light signals and fluorescence signals. The scattered light signals are the so-called FSC (forward scattered light, reflecting the size of cells) and SSC (side scattered light, reflecting the complexity of cells); the fluorescence signal is that when a cell is bound with a fluorophore and is excited by a laser, it will emit a fluorescence signal.

[0117] Specific implementation of S30

[0118] Extract the forward scatter light signal (FSC), side scatter light signal (SSC), and multi-channel fluorescence signals of all cells from the collected FCS files.

[0119] Set the number of optical signal channels collected in the experiment:

[0120] Number of FSC channels (There is only one forward scatter channel)

[0121] Number of SSC channels (There is only one side scatter channel)

[0122] Number of fluorescence channels , a total of fluorescence channels with different wavelengths

[0123] Then the total number of channels

[0124] For the th cell, define its optical characteristics as a -dimensional vector:

[0125]

[0126] Where:

[0127] represents the FSC value of the th cell

[0128] represents the SSC value of the th cell

[0129] represents the fluorescence value of the th cell in the th fluorescence channel

[0130] Construct the optical characteristics matrix of all cells :

[0131] ;

[0132] Where, is the total number of cells.

[0133] is the th row vector of the matrix representing the -dimensional optical characteristics of the

[0134] Normalize the matrix to ensure that different optical characteristics are comparable on the same order of magnitude:

[0135] ;

[0136] Among them, represents the normalized value, are respectively the mean and standard deviation of the elements in the column.

[0137] The matrix after normalization is the final optical feature matrix, which can be used to distinguish different types of cells subsequently.

[0138] Specific implementation of S40:

[0139] Define the flow rate error of a single cell as a scalar value, representing the difference between the current flow rate and the optimal flow rate, denoted as .

[0140] Construct a neural network model, where:

[0141] The number of nodes in the input layer is the number of fluorescence channels

[0142] The number of nodes in the hidden layer is , and the activation function is , such as the sigmoid function

[0143] The number of nodes in the output layer is 1, representing the flow rate error

[0144] The network connection weight matrix is , among which, represents the weight between the th node in the input layer and the th node in the hidden layer, represents the weight between the th node in the hidden layer and the output layer.

[0145] The output expression of the network is:

[0146]

[0147] Among them, is the n-dimensional fluorescence feature vector of the th cell, , is the bias vector between the hidden layer and the output layer.

[0148] 5) Given the training data set , where is the fluorescence feature of the th cell, is the corresponding flow rate error, and use the maximum likelihood estimation to find the network parameters , to minimize the objective function:

[0149]

[0150] is the conditional probability distribution of the flow rate error of the th cell given the network parameters.

[0151] Update the network parameters using the gradient descent method:

[0152]

[0153]

[0154] where is the learning rate, represents the number of network layers.

[0155] Repeat step 6 until the network error converges or reaches the preset number of training times.

[0156] Calculate the predicted value of the flow rate error using the optimized network parameters .

[0157] Construct the predicted error of each cell into a vector which is the flow rate error vector.

[0158] Specific implementation of S50:

[0159] Define the meanings of each variable:

[0160] : Current flow rate

[0161] : Flow rate error vector

[0162] : Number of samples

[0163] : Flow rate adjustment amount

[0164] : New optimized flow rate

[0165] Construct a flow rate adjustment function :

[0166] Input: Flow rate error vector

[0167] Output: Flow rate adjustment amount

[0168] Implemented using a three-layer fully connected neural network :

[0169] Number of nodes in the input layer:

[0170] Number of nodes in hidden layer 1: , activation function

[0171] Number of nodes in hidden layer 2: , activation function

[0172] Number of nodes in the output layer: 1

[0173] The network predicted flow velocity adjustment amount is:

[0174]

[0175] where, are network parameters.

[0176] Use the labeled data to train the network and minimize the loss function:

[0177]

[0178] Predict the flow velocity adjustment amount for the new sample:

[0179]

[0180] Calculate the new flow velocity:

[0181]

[0182] Real-time monitor the signal and continuously optimize the flow velocity.

[0183] This method realizes the closed-loop control and intelligent optimization of the flow velocity by fitting the complex mapping relationship between the error vector of the neural network and the flow velocity adjustment.

[0184] Among them, the method of using a three-layer fully connected neural network to implement the flow velocity adjustment function is as follows:

[0185] Network input layer:

[0186] The input layer contains nodes, which is the same dimension as the flow velocity error vector . The th node receives the th component in the error vector as the input.

[0187] Hidden layer 1:

[0188] The first hidden layer contains nodes. The Each hidden layer node receives the weighted sum of the weighted inputs from each node in the input layer, and then passes through an activation function , expressed as:

[0189]

[0190] where is the input node to the hidden layer node weight, is the activation function, such as ReLU, is the node bias.

[0191] Hidden layer 2:

[0192] The second hidden layer contains nodes. The th node receives the weighted input from the first hidden layer, expressed as:

[0193]

[0194] where is the weight between the two hidden layers, is the activation function, such as sigmoid, is the bias.

[0195] Output layer:

[0196] The output layer contains 1 node, giving the predicted value of the flow rate adjustment:

[0197]

[0198] where is the weight from the second hidden layer to the output layer, is the output bias.

[0199] Network parameters are learned through training to make the prediction close to the true adjustment amount.

[0200] This network models the complex mapping relationship from the error vector to the flow rate adjustment, realizing the intelligent optimization of flow rate control.

[0201] Define variables:

[0202] : Fluorescence feature matrix before flow rate adjustment, where the th row vector represents the fluorescence feature of the th cell

[0203] : Flow rate error vector

[0204] : New fluorescence feature matrix for deviation correction after flow rate adjustment

[0205] Assume that the fluorescence feature and the flow rate error satisfy a linear model:

[0206]

[0207] where are the model parameters, is the random noise.

[0208] Estimate the parameters using the entire cell sample dataset:

[0209]

[0210] where,

[0211]

[0212]

[0213] For the th cell, based on its flow rate error , calculate the new fluorescence feature after deviation correction:

[0214]

[0215] Construct a new matrix from the deviation correction feature vectors of all samples:

[0216]

[0217] In addition to the linear model, a non - linear deviation correction mapping function can also be used:

[0218]

[0219] where is a multi - layer fully - connected neural network model that can learn complex non - linear mappings, where are the network parameters.

[0220] Train the model by minimizing the loss function:

[0221]

[0222] where are the manually labeled features after deviation correction.

[0223] This deviation correction process can reduce the impact of flow rate changes on features and improve the accuracy of subsequent analysis. By training to learn the deviation correction mapping, automatic and robust adjustment of fluorescence features is achieved.

[0224] Specifically, the principle of the present invention can be described as follows:

[0225] 1) Establish an optimal flow rate database that includes different types of samples under various volume and concentration conditions, in combination with parameters such as different types and concentrations of fluorescein. The database can be constructed based on historical experimental data and updated when new experimental data arrives.

[0226] 2) When performing flow cytometry detection, determine the specific parameters of the samples used in this experiment, including the type, status, volume, concentration of the samples, as well as the type and concentration of the fluorescein used.

[0227] 3) According to the parameters of this experiment, search for the most matching sample record in the flow rate database, obtain the corresponding optimal flow rate value, and set it as the flow rate parameter of the instrument for this time.

[0228] 4) Real-time collect the fluorescence intensity signals of each cell in the multi-color fluorescence channels to obtain multi-dimensional time series fluorescence data of the total cell count.

[0229] 5) Organize the collected multi-color fluorescence signals into a matrix form, with rows representing each cell and columns representing different channels. Perform preprocessing such as normalization on the matrix.

[0230] 6) Use a neural network to model the fluorescence matrix and train the network to predict the flow rate error of each cell. Form an error vector.

[0231] 7) According to the flow rate error vector, construct an error-to-flow rate adjustment mapping model. Predict the adjustment amount for new samples and correct the flow rate.

[0232] 8) Iterate the above process in a closed loop to continuously optimize the flow rate to approach the optimal flow rate, thereby improving the performance of flow cytometry detection and analysis.

[0233] 9) Real-time adjust the flow rate to adapt to changes in the sample status and continuously optimize the flow rate.

[0234] Compared with manually configuring the flow rate, the present invention stores the optimal flow rate of the samples in a database and uses a neural network for closed-loop control and automatic optimization of the flow rate, achieving precision and intelligence of the flow rate, and improving the flexibility, reliability, and efficiency of flow cytometry analysis.

[0235] The above is only the specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention can easily think of changes or substitutions, which should all be covered within the protection scope of the present invention. Therefore, the protection scope of the present invention should be subject to the protection scope of the claims.

Claims

1. A method for controlling the sample flow rate of a flow cytometer, characterized in that, It includes the following steps: S10. Select the optimal flow rate of the sample in the basic flow rate database as the current flow rate of the flow cytometry experiment according to the type, state, volume, concentration of the sample, and the type and concentration of the fluorescent agent; S20. Obtain the fluorescence signals of the flow cytometry experiment in real time, including FSC, SSC, and cell fluorescence; S30. Establish a fluorescence feature matrix based on the fluorescence signals; S40. Use a neural network to calculate the flow rate error vector of the fluorescence feature matrix; S50. Adjust the current flow rate according to the flow rate error vector to obtain an optimized flow rate; S60. Perform deviation correction calculation and update on the fluorescence signals before adjusting the flow rate according to the flow rate error vector; Specifically, step S40 is to input the fluorescence feature vectors corresponding to each cell in the sample in the fluorescence feature matrix into a neural network, and the corresponding flow velocity error is output , and the flow velocity errors of each cell are formed into a vector which is the flow velocity error vector; Among them, step S50 specifically is: construct a flow rate adjustment function with the flow rate error vector as the input and the flow rate adjustment amount as the output, and the flow rate adjustment function is implemented by a three-layer fully connected neural network; pass the current flow rate error vector through the trained neural network to calculate the flow rate adjustment amount, and add the current flow rate to the calculated flow rate adjustment amount to obtain the optimized new flow rate, denoted as the optimized flow rate.

2. The flow cytometry sample flow rate control method according to claim 1, wherein, The step of selecting the optimal flow rate of the sample in the basic flow rate database as the current flow rate of the flow cytometry experiment according to the type, state, volume, concentration of the sample, and the type and concentration of the fluorescent agent specifically includes: Establish a basic flow rate database, and collect the optimal flow rate data of different types of samples under different states, volumes, concentrations, and different types and concentrations of fluorescent agents; Before conducting the flow cytometry experiment, determine the parameters of this experiment, including the type, state, volume, concentration of the sample, and the type and concentration of the fluorescent agent; According to the experimental parameters, first try to search for a completely matching record in the database. If there is a completely matching record, use the optimal flow rate of this record as the current flow rate; If there is no completely matching record, use a scoring function to calculate the matching degree of all records with the experimental parameters of this time, and select the flow rate of the record with the highest score as the current flow rate.

3. The flow cytometry sample flow rate control method according to claim 2, wherein, The scoring function is obtained by fitting the optimal flow rate data of the existing samples in the database under different states, volumes, concentrations, and different types and concentrations of fluorescent agents using a linear regression function.

4. The flow cytometry sample flow rate control method according to claim 3, wherein, In the step of establishing a fluorescence feature matrix based on the fluorescence signals, it also includes a process of normalizing the established fluorescence feature matrix.

5. A method for controlling the sample flow rate of a flow cytometer according to claim 4, characterized in that, The flow velocity error represents the difference between the current flow velocity for the corresponding cells and the optimal flow velocity.

6. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores program instructions, and when the program instructions run, they are used to execute the flow cytometry sample flow rate control method according to any one of claims 1-5.

7. A flow cytometry sample flow rate control system, comprising the computer-readable storage medium according to claim 6.

Images