Method, device and storage medium for automatically extracting multiple fluorescence reference spectra

Abstract

This application relates to the field of fluorescence microscopy imaging technology, and particularly to a method and system for automatically extracting multiple fluorescence reference spectra. The method includes: inputting a stack of lambda spectral images; performing preprocessing of effective data and data redistribution on the stack of lambda spectral images; constructing a training variational autoencoder network and training the training variational autoencoder; generating an implicit space vector and clustering the implicit space vector; and obtaining multiple fluorescence reference spectra. This application can achieve the automatic extraction of multiple fluorescence reference spectra when the dye information is unknown or the experimental environment deviates greatly from the ideal environment.

Description

Technical Field

[0001] The present application relates to the field of fluorescence microscopy imaging technology, and particularly relates to a method, device and storage medium for automatically extracting multiple fluorescence reference spectra. Background Art

[0002] Optical microscopy imaging is an important technical means for cell research. With the continuous progress of fluorescence labeling technology and microscopy instruments, the generation and imaging of complex biological samples containing multiple fluorophores have gradually become possible. Among them, in order to accurately quantify the data of multiple fluorescence channels simultaneously, the reliable splitting of spectral signals of different fluorescence channels is particularly important.

[0003] The single-channel reference spectral curve is the basis of the multiple fluorescence spectral splitting technology. The existing methods for obtaining reference spectra mainly include: based on dye databases, making single-fluorescence samples, and based on lambda spectral imaging. The method based on dye databases is to find the dye spectral curve matching the actual sample from the official platform and directly use it as the reference spectrum of the fluorescent dye. The method based on single-fluorescence samples requires making a fluorescence sample containing only a single cell structure for each channel, then performing spectral imaging on each sample, and then obtaining its reference spectrum. The method based on lambda spectral imaging can generate a reference spectrum by manually selecting a region or by constructing a feature space, such as principal component analysis (PCA), sine / cosine transform, etc. After obtaining the reference spectrum through the above methods and combining with the existing spectral splitting technology, such as linear splitting, the signal of a single separated fluorescence channel can be obtained.

[0004] Among them, the method based on dye databases is the simplest and requires knowing the dye information in advance. However, due to the changes in the environment such as temperature and humidity in the actual experiment, there will be a certain deviation between the theoretical reference spectrum and the actual spectral curve, resulting in inaccurate splitting results. Although the spectral curve obtained by making a single-fluorescence sample is relatively accurate, its production process is complex, and there is a high implementation difficulty for some sub-organelles that are closely related in structure. In the method based on lambda spectral imaging, manually selecting a region requires artificial discrimination of the pixel region of a single sample in the image, and this process depends on the experience of the operator and the structure of the sample to be detected. When different dyes are closely related in space or there is strong background signal interference, the reference spectrum obtained by this method will have a large deviation from the real curve. At present, although conventional means such as the commonly used principal component analysis (PCA) and low-order harmonic analysis based on sine / cosine transform can realize the automatic extraction of the reference spectrum, they depend on the selection of the feature space, and due to the low dimension of the generated eigenvectors, their adaptability to multiple channels still needs to be improved.

[0005] In summary, when the dye information is unknown or the experimental environment deviates significantly from the ideal environment, how to automatically extract the multiple fluorescence reference spectra is a difficult problem faced at present. Summary of the Invention

[0006] This application provides a method and system for automatically extracting multiple fluorescence reference spectra, which can automatically extract the multiple fluorescence reference spectra when the dye information is unknown or the experimental environment deviates significantly from the ideal environment. The present application provides the following technical solutions:

[0007] In a first aspect, this application provides a method for automatically extracting multiple fluorescence reference spectra, the method comprising:

[0008] Input a stack of lambda spectral images;

[0009] Perform effective data preprocessing and data redistribution on the stack of lambda spectral images;

[0010] Construct a training variational autoencoder network and train the training variational autoencoder;

[0011] Generate an implicit space vector and perform clustering on the implicit space vector;

[0012] Obtain multiple fluorescence reference spectra.

[0013] In a specific feasible embodiment, the performing effective data preprocessing and data redistribution on the stack of lambda spectral images includes:

[0014] Calculate the mean μ m and the standard deviation σ m ;

[0015] Rearrange the data structure of the stack of lambda spectral images, and the new arranged data format is L*M, where L represents the total number of data samples and M is the length of the feature vector in the lambda direction;

[0016] Select valid samples from the data samples; among them, for each sample m, which can be represented as a one-dimensional vector with a length of M, calculate its mean μ i and the standard deviation σ i ; if μ i is greater than 0.2*μ m , and σ i is greater than 0.2*σ m , then it is considered a valid sample and is retained, otherwise the sample is excluded; the selected valid data samples are L e *M, where L e is the number of valid data samples;

[0017] Normalize the valid data sample L e *M.

[0018] In a specific feasible implementation, constructing the training variational autoencoder network and training the training variational autoencoder includes:

[0019] Construct the variational autoencoder network structure, where the variational autoencoder network structure includes an encoding Encoder module and a decoding Decoder module;

[0020] Construct the loss function loss = loss_recon + loss_dist;

[0021] Input the valid data sample L e *M norm , and train the variational autoencoder network parameters;

[0022] If the loss function loss is less than the set threshold or the set number of training times is reached, the training is completed.

[0023] In a specific feasible implementation, the loss function loss consists of two parts:

[0024] represents the degree of difference between the vector m' generated by the decoding Decoder module and the input sample m, where ||·||2 represents the vector two-norm;

[0025] Among them, represents a normal distribution with an expectation of z μ and a variance of , N(0, I) represents a standard normal distribution with an expectation of 0 and a variance of I, I represents the identity matrix, and D KL represents the Kullback–Leibler divergence between two probability distributions, that is, the degree of difference between the distribution of the implicit z vector output by the encoding Encoder module and the standard normal distribution.

[0026] In a specific feasible implementation, before constructing the loss function loss = loss_recon + loss_dist, it also includes:

[0027] Input the sample m into the encoding Encoder module, and output the implicit vector z μ and z σ ;

[0028] Sample according to the Gaussian distribution and output the sampled vector z s ;

[0029] The sampled vector zs Input decoding Decoder module, outputting the generated vector m'.

[0030] In a specific feasible implementation, the generation of the implicit space vector includes:

[0031] Taking the encoding Encoder module of the variational autoencoder that has completed training;

[0032] For the valid data sample L e *M norm , generating the implicit space vector set L e *Z; where for each sample m, inputting it into the encoding Encoder module of the variational autoencoder and outputting the implicit vector z μ , whose vector length is Z.

[0033] In a specific feasible implementation, the clustering of the implicit space vector includes:

[0034] For the mixed fluorescence spectrum sample with the number of channels N, setting the clustering category to N;

[0035] Sampling the clustering method to cluster the implicit space vector;

[0036] Generating the clustering sample clusters L1*Z, L2*Z, L3*Z,..., L N *Z, and the corresponding sample cluster centroids z1, z2, z3,..., z N , where L1 + L2 + L3 +... + L N = L e .

[0037] In a specific feasible implementation, the acquisition of the multiple fluorescence reference spectra includes:

[0038] Taking a single sample cluster L i *Z and its sample cluster centroid z i ;

[0039] For the single sample cluster L i *Z, searching for the K nearest neighbor points corresponding to the centroid z i to form the implicit space nearest neighbor set L knn *Z, where K = 0.3*L;

[0040] For the implicit space nearest neighbor set L knn *Z, backtracking its corresponding nearest neighbor set L e *M in the valid data sample L knn *M;

[0041] For the data sample space nearest neighbor set L knn *M, calculating M i=(M1 + M2 + M3 + … + M Lknn ) / L knn ,M i is the reference spectrum to be obtained;

[0042] For the clustering sample clusters L1*Z, L2*Z, L3*Z, …, L N *Z, perform the above operations on each sample cluster L i *Z (i = 1 to N) in sequence, and the reference spectra of all fluorescence channels can be obtained.

[0043] In a second aspect, the present application provides an electronic device, which includes a processor and a memory; a program is stored in the memory, and the program is loaded and executed by the processor to implement a method for automatically extracting multiple fluorescence reference spectra.

[0044] In a third aspect, the present application provides a computer-readable storage medium, in which a program is stored, and when the program is executed by a processor, it is used to implement a method for automatically extracting multiple fluorescence reference spectra.

[0045] In summary, the beneficial effects of the present application at least include:

[0046] 1) It can be used for automatically obtaining multiple fluorescence reference spectra with unknown dye information;

[0047] 2) When the deviation between the experimental environment and the ideal environment is large, more accurate results can be obtained by using this method;

[0048] 3) Without the intervention of experimental personnel, it can realize the automatic acquisition of multiple fluorescence reference spectra;

[0049] 4) The variational auto-encoding technology used to generate implicit space vectors is based on a mixture Gaussian model and has strong interpretability. The implicit space vectors learned through an unsupervised generation model have higher robustness for obtaining multiple fluorescence (number of fluorescence channels > 3) reference spectra.

[0050] Generate a multi-dimensional implicit space through variational auto-encoding technology, and then cluster or classify the implicit space to realize the automatic extraction of multiple fluorescence reference spectra. When the dye information is unknown or the deviation between the experimental environment and the ideal environment is large, more ideal results can be obtained by using this method.

[0051] The above description is only an overview of the technical solution of the present application. In order to be able to understand the technical means of the present application more clearly and implement it according to the content of the specification, the following takes the preferred embodiments of the present application and combines with the drawings to describe in detail as follows. Description of the Drawings

[0052] Figure 1It is a schematic flowchart of a method for automatically extracting a multi - fluorescence reference spectrum provided by an embodiment of the present application.

[0053] Figure 2 It is a schematic structural diagram of a variational auto - encoder provided by an embodiment of the present application.

[0054] Figure 3 It is a schematic flowchart of a system for obtaining a multi - fluorescence reference spectrum provided by an embodiment of the present application.

[0055] Figure 4 It is a block diagram of an electronic device for automatically extracting a multi - fluorescence reference spectrum provided by an embodiment of the present application. Detailed implementation manners

[0056] The following will further describe in detail the specific implementation manners of the present application in conjunction with the accompanying drawings and embodiments. The following embodiments are used to illustrate the present application, but not to limit the scope of the present application.

[0057] Optionally, the method for automatically extracting a multi - fluorescence reference spectrum provided by each embodiment of the present application is taken as an example for illustration in an electronic device. The electronic device is a terminal or a server. The terminal can be a mobile phone, a computer, a tablet computer, a scanner, an electronic eye, a surveillance camera, etc. The type of the electronic device is not limited in this embodiment.

[0058] Refer to Figure 1 , which is a schematic flowchart of a method for automatically extracting a multi - fluorescence reference spectrum provided by an embodiment of the present application. The method includes at least the following steps:

[0059] S100. Input the lambda - spectrum image stack.

[0060] The lambda - spectrum image stack is provided by a fluorescence imaging device with spectral imaging function, such as a laser scanning confocal microscope equipped with a spectral imaging module.

[0061] Specifically, the size of the lambda - spectrum image stack is X * Y * M, where X is the horizontal resolution of the imaging device. For an imaging device equipped with a CMOS photosensitive device, its horizontal resolution is the horizontal resolution of the CMOS device. For an imaging device equipped with a PMT photosensitive device, its horizontal resolution is the horizontal scanning resolution of the sample. Y is the vertical resolution of the imaging device. For an imaging device equipped with a CMOS photosensitive device, its vertical resolution is the vertical resolution of the CMOS device. For an imaging device equipped with a PMT photosensitive device, its vertical resolution is the vertical scanning resolution of the sample. M is the number of sampling points of the spectrum along the wavelength lambda direction, that is, the length of the eigenvector in the lambda direction.

[0062] S200. Data pre - processing and redistribution.

[0063] For effectively screening valid data and redistributing data for the lambda spectral image stack in S100 above, the specific processes of preprocessing and redistribution include:

[0064] S201. Calculate the average value μ of the lambda spectral image stack in S100 above m and the standard deviation σ m .

[0065] S202. Rearrange the data structure of the lambda spectral image stack in S100 above. The new arranged data format is L*M, where L = X*Y represents the total number of data samples, and M is the length of the feature vector in the lambda direction.

[0066] S203. Screen out valid samples from the data samples in S202 above.

[0067] Specifically, for each sample m, which can be represented as a one-dimensional vector with a length of M, calculate its average value μ i and the standard deviation σ i . If μ i is greater than 0.2*μ m , and σ i is greater than 0.2*σ m , then it is considered a valid sample and is retained; otherwise, the sample is excluded. The screened valid data samples are L e *M, where L e is the number of valid data samples.

[0068] S204. Normalize the valid data samples L e *M in S203 above.

[0069] Specifically, for each sample m, search for the maximum component m max of this sample, and normalize this sample m norm = m / m max . The normalized sample is Le*M norm .

[0070] S300. Train a variational autoencoder.

[0071] Referring to Figure 2 , which is a schematic diagram of the variational autoencoder structure provided by an embodiment of the present application, its training process is as follows:

[0072] S301. Construct a variational autoencoder network structure, mainly including an encoding Encoder module and a decoding Decoder module:

[0073] Optionally, the Encoder module can be composed of a linear layer, an activation function layer, and a linear layer;

[0074] Optionally, the Decoder module can be composed of a linear layer, an activation function layer, and a linear layer.

[0075] Among them, the data forward propagation process, that is, the process from data input to output, includes:

[0076] S3011. The sample m is input into the Encoder module, and the implicit vector z μ and z σ .

[0077] S3012. Sampling is performed according to the Gaussian distribution to output the sampling vector z s .

[0078] S3013. The sampling vector z s is input into the Decoder module, and the generated vector m' is output.

[0079] S302. Construct the loss function loss = loss_recon + loss_dist.

[0080] Among them, the loss function loss consists of two parts:

[0081] represents the degree of difference between the generated vector m' by the Decoder module and the input sample m, where ||·||2 represents the vector two-norm.

[0082] , where, represents the normal distribution with an expectation of z μ and a variance of , N(0, I) represents the standard normal distribution with an expectation of 0 and a variance of I, I represents the identity matrix, and D KL represents the Kullback–Leibler divergence (abbreviated as KL divergence) between two probability distributions, that is, the degree of difference between the distribution of the implicit z vector output by the Encoder module and the standard normal distribution.

[0083] S303. Input the valid data sample L e *M norm in S204 above to train the variational autoencoder network parameters.

[0084] S304. If the loss function loss in S302 above is less than the set threshold or reaches the set number of training times, the training is completed.

[0085] S400. Generate the implicit space vector.

[0086] The specific process includes:

[0087] S401. Take the encoding Encoder module of the variational autoencoder that has been trained in the above S300.

[0088] S402. For the valid data samples L e *M norm in the above S204, generate a set of implicit space vectors L e *Z.

[0089] For each sample m, input it into the encoding Encoder module of the variational autoencoder in S300, and output an implicit vector z μ , whose vector length is Z.

[0090] S500. Implicit space clustering.

[0091] The specific process includes:

[0092] S501. For the mixed fluorescence spectrum samples with the number of channels being N, set the number of clustering categories to N.

[0093] S502. Use a sampling clustering method to cluster the implicit space vectors.

[0094] Optionally, the selected clustering method can be k-cluster clustering;

[0095] Optionally, the selected clustering method can be t-sne clustering, and this clustering method can perform dimensionality reduction on the data before clustering.

[0096] S503. Generate clustering sample clusters L1*Z, L2*Z, L3*Z, …, L N *Z, and the corresponding sample cluster centroids z1, z2, z3, …, z N , where L1 + L2 + L3 + … + L N = L e .

[0097] S600. Obtain the multi-fluorescence reference spectrum.

[0098] Referring to Figure 3 , which is a schematic flow diagram of the process for obtaining the multi-fluorescence reference spectrum provided by an embodiment of the present application. Specifically, for the clustering sample clusters L1*Z, L2*Z, L3*Z, …, L N *Z described in the above S503, calculate the reference spectrum corresponding to each sample cluster L i *Z (i = 1 to N) in sequence. The process includes:

[0099] S601. Take a single sample cluster L i*Z and the centroid z of its sample clusters i (i = 1 to N).

[0100] S602. For the above single sample cluster L i *Z, search for the centroid z i The corresponding K-nearest neighbor points are formed into an implicit space neighbor set L knn *Z, where K = 0.3 * L i .

[0101] S603. For the implicit space neighbor set L in S602 above knn *Z, trace back its corresponding neighbor set L in the valid data sample L e *M knn *M.

[0102] S604. For the data sample space neighbor set L in S603 above knn *M, calculate M i =(M1 + M2 + M3 + … + M Lknn ) / L knn , M i is the reference spectrum to be obtained;

[0103] S605. For the clustering sample clusters L1*Z, L2*Z, L3*Z, …, L in S503 above N *Z, for each sample cluster L i *Z (i = 1 to N) execute steps S601 - S604 in sequence, and the reference spectra of all fluorescence channels can be obtained.

[0104] In summary, aiming at the deficiencies of the existing technology, this application designs a method for automatically extracting multiple fluorescence reference spectra. By using variational autoencoding technology to generate a multi-dimensional implicit space, and then clustering or classifying the implicit space, the automatic extraction of multiple fluorescence reference spectra is realized. When the dye information is unknown or the experimental environment deviates greatly from the ideal environment, relatively ideal results can be obtained by using this method. In addition, this method does not require the intervention of experimental personnel. The variational autoencoding technology used to generate the implicit space is based on a mixture Gaussian model, and any mixed spectral distribution is regarded as a linear superposition of several independent reference spectral distributions, which is highly interpretable and makes the subsequent classification and extraction of reference spectra more robust.

[0105] Figure 4 It is a block diagram of an electronic device provided by an embodiment of this application. This device at least includes a processor 401 and a memory 402.

[0106] The processor 401 may include one or more processing cores, such as a 4-core processor, an 8-core processor, etc. The processor 401 may be implemented in at least one hardware form of DSP (Digital Signal Processing), FPGA (Field-Programmable Gate Array), or PLA (Programmable Logic Array). The processor 401 may also include a main processor and a coprocessor. The main processor is a processor used to process data in the wake state, also known as the CPU (Central Processing Unit); the coprocessor is a low-power processor used to process data in the standby state. In some embodiments, the processor 401 may be integrated with a GPU (Graphics Processing Unit), and the GPU is responsible for rendering and drawing the content to be displayed on the display screen. In some embodiments, the processor 401 may further include an AI (Artificial Intelligence) processor, and the AI processor is used to process computational operations related to machine learning.

[0107] The memory 402 may include one or more computer-readable storage media, and the computer-readable storage media may be non-transitory. The memory 402 may further include high-speed random access memory and non-volatile memory, such as one or more disk storage devices and flash storage devices. In some embodiments, the non-transitory computer-readable storage media in the memory 402 is used to store at least one instruction, and the at least one instruction is used to be executed by the processor 401 to implement the method for automatically extracting multiple fluorescence reference spectra provided in the method embodiments of the present application.

[0108] In some embodiments, the electronic device may further optionally include: a peripheral device interface and at least one peripheral device. The processor 401, the memory 402, and the peripheral device interface may be connected through a bus or signal lines. Each peripheral device may be connected to the peripheral device interface through a bus, signal lines, or a circuit board. Schematically, the peripheral devices include but are not limited to: a radio frequency circuit, a touch display screen, an audio circuit, and a power supply, etc.

[0109] Of course, the electronic device may also include fewer or more components, and this embodiment does not limit this.

[0110] Optionally, the present application also provides a computer-readable storage medium, and a program is stored in the computer-readable storage medium. The program is loaded and executed by the processor to implement the method for automatically extracting multiple fluorescence reference spectra in the above method embodiments.

[0111] Optionally, the present application also provides a computer product, which includes a computer-readable storage medium. A program is stored in the computer-readable storage medium and is loaded and executed by a processor to implement the method for automatically extracting a multiple fluorescence reference spectrum in the above method embodiment.

[0112] The technical features of the above embodiments can be combined arbitrarily. For the sake of concise description, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, it should be considered as the scope recorded in this specification.

[0113] The above embodiments only represent several implementation manners of the present application. The description is relatively specific and detailed, but it should not be construed as a limitation on the scope of the invention patent. It should be noted that for those of ordinary skill in the art, without departing from the concept of the present application, several modifications and improvements can be made, and these all belong to the protection scope of the present application. Therefore, the protection scope of the patent of the present application shall be subject to the appended claims.

Claims

1. A method for automatically extracting a multi-fluorescence reference spectrum, characterized in that, The method includes: inputting a stack of lambda spectral images; performing effective data preprocessing and data redistribution on the stack of lambda spectral images; constructing a training variational autoencoder network and training the training variational autoencoder; generating implicit space vectors and clustering the implicit space vectors; obtaining multiple fluorescence reference spectra, including: Take a single sample cluster L i *Z and its sample cluster centroid z i ; For a single sample cluster L i *Z, search for the centroid z i The corresponding K nearest neighbor points form an implicit spatial neighbor set L knn *Z, where K = 0.3 * L; For the implicit spatial neighbor set L knn *Z, backtrack its corresponding neighbor set L e *M in the valid data sample L knn *M; For the set of nearest neighbors L in the data sample space knn *M, calculate M i =(M1 + M2 + M3 + … + M Lknn ) / L knn , M i is the reference spectrum to be obtained; For the clustered sample clusters L1*Z, L2*Z, L3*Z, …, L N *Z, for each sample cluster L i *Z, where i = 1 to N, perform search, backtracking, and calculation operations, and the reference spectra of all fluorescence channels can be obtained. N is the number of channels.

2. The method for automatically extracting a multiple fluorescence reference spectrum according to claim 1, wherein The performing effective data preprocessing and data redistribution on the stack of lambda spectral images includes: Calculate the mean value μ of the stack of lambda spectral images m and the standard deviation σ m ; rearranging the data structure of the stack of lambda spectral images, and the new arranged data format is L*M, where L represents the total number of data samples and M is the length of the feature vector in the lambda direction; Select valid samples from the data samples; for each sample m, which can be represented as a one-dimensional vector with a length of M, calculate its mean μ i and standard deviation σ i ; if μ i is greater than 0.2 * μ m , and σ i is greater than 0.2 * σ m , then it is considered a valid sample and retained; otherwise, the sample is excluded; the filtered valid data samples are L e *M, where L e is the number of valid data samples; Normalize the valid data sample L e *M.

3. The method for automatically extracting a multiple fluorescence reference spectrum according to claim 2, wherein The constructing a training variational autoencoder network and training the training variational autoencoder includes: constructing a variational autoencoder network structure, and the variational autoencoder network structure includes an encoding Encoder module and a decoding Decoder module; constructing a loss function loss = loss_recon + loss_dist; Input the valid data sample L e *M norm , and train the variational autoencoder network parameters; If the loss function loss is less than a set threshold or reaches a set number of training times, the training is completed.

4. The method for automatically extracting a multiple fluorescence reference spectrum according to claim 3, wherein The loss function loss consists of two parts: Indicates the degree of difference between the vector m' generated by the decoding Decoder module and the input sample m, where ||·||2 represents the vector two-norm; Among them, represents the expected value of z μ , and the variance is of the normal distribution, N(0, I) represents the standard normal distribution with an expected value of 0 and a variance of I, I represents the identity matrix, and D KL represents the Kullback–Leibler divergence between two probability distributions, that is, the degree of difference between the implicit z-vector distribution output by the encoding Encoder module and the standard normal distribution.

5. The method for automatically extracting a multiple fluorescence reference spectrum according to claim 3, characterized in that Before the constructing the loss function loss = loss_recon + loss_dist, it further includes: The sample m is input into the Encoder module for encoding, and the implicit vector z is output μ and z σ ; According to the Gaussian distribution perform sampling and output the sampling vector z s ; Sampling vector z s Input the decoding Decoder module and output the generated vector m'.

6. The method for automatically extracting a multiple fluorescence reference spectrum according to claim 3, wherein The generating implicit space vectors includes: taking the encoding Encoder module of the trained variational autoencoder; For valid data samples L e *M norm , generate an implicit space vector set L e *Z; for each sample m, input it into the encoding Encoder module of the variational autoencoder and output an implicit vector z μ , whose vector length is Z.

7. The method for automatically extracting a multi-fluorescence reference spectrum according to claim 6, characterized in that The clustering the implicit space vectors includes: for a mixed fluorescence spectrum sample with N channels, setting the number of clustering categories to N; sampling a clustering method to cluster the implicit space vectors; Generate clustering sample clusters L1*Z, L2*Z, L3*Z, …, L N *Z, and the corresponding sample cluster centroids z1, z2, z3, …, z N , where L1 + L2 + L3 + … + L N = L e .

8. An electronic device, characterized in that, The device includes a processor and a memory; a program is stored in the memory, and the program is loaded and executed by the processor to implement a method for automatically extracting multiple fluorescence reference spectra according to any one of claims 1 to 7.

9. A computer-readable storage medium, characterized in that, A program is stored in the storage medium, and when the program is executed by a processor, it is used to implement a method for automatically extracting multiple fluorescence reference spectra according to any one of claims 1 to 7.

Images