Spectrum splitting method and device

Through the non-negative matrix decomposition NMF method, the problem of not being able to calculate the abundance matrix under the unknown spectral feature column vector is solved, spectral splitting is realized, and signal quantitative analysis capabilities of small animals are improved.

CN119067965APending Publication Date: 2024-12-03RWD LIFE SCI CO LTD
View PDF 4 Cites 0 Cited by

Patent Information

Application Number
CN202411538417.3
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-10-31
Publication Date
2024-12-03

AI Technical Summary

Technical Problem

The prior art cannot calculate the abundance matrix H in the case of unknown spectral feature column vectors, resulting in the inability to perform spectral splitting.

Method used

The non-negative matrix decomposition NMF method is used to randomly initialize the spectral feature matrix and abundance matrix, and perform preset matrix decomposition based on the known spectral feature column vector until the preset termination conditions are met, spectral splitting is achieved.

Benefits of technology

When some or all of the spectral feature column vectors in the spectral feature matrix are unknown, spectral splitting is successfully achieved, pure fluorescence signals are obtained, and the quantitative analysis ability of fluorescence imaging in small animals is improved.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119067965A_ABST
    Figure CN119067965A_ABST
Patent Text Reader

Abstract

The spectrum splitting method provided by the embodiment of the invention comprises the following steps: acquiring m fluorescent images with the same size to form a first fluorescent image matrix; initializing a first spectral feature matrix of m rows and r columns by using a random positive decimal, wherein r is a positive integer greater than 1; initializing a first abundance matrix by using a random positive decimal; and performing preset non-negative matrix factorization (NMF) according to the first fluorescence image matrix, the first spectral feature matrix, the first abundance matrix and the k spectral feature column vectors until a preset termination condition is met. A first spectral feature matrix in the embodiment of the invention comprises r spectral feature column vectors, NMF is preset according to a first fluorescence image matrix, k known spectral feature column vectors, an initialized first spectral feature matrix and a first abundance matrix, k is smaller than r, and r is smaller than r. And under the condition that part or all of the spectral feature column vectors in the first spectral feature matrix are unknown, spectral splitting is realized.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of image processing, and in particular to a spectrum splitting method and a device thereof. Background Art

[0002] In recent years, the research on small animal fluorescence imaging has become more and more extensive. In the process of small animal fluorescence imaging, spectral unmixing is a very important step. Through spectral unmixing, pure and useful fluorescence signals can be obtained, and the interference of spontaneous fluorescence of biological tissues or spontaneous fluorescence of equipment parts can be removed, which is of great significance for the quantitative analysis of small animal fluorescence signals.

[0003] Figure 1 This is a schematic diagram of spectrum splitting through matrix operations. Among them, the fluorescence image matrix V is m rows and n columns, indicating that m fluorescence images are arranged in the form of row vectors, and the number of pixels in each fluorescence image is n; the spectral feature matrix W is m rows and r columns, and r spectral feature column vectors are arranged in the form of column vectors. Each spectral feature column vector includes m elements, and the number of columns r indicates that there are r kinds of fluorescent substances emitting fluorescence signals in the fluorescence imaging process, that is, r spectral feature column vectors; the abundance matrix H is r rows and n columns, indicating the weight coefficient of each spectral feature column vector at each pixel, that is, the fluorescence signal intensity. In the matrix operation V=W*H, W*H indicates that r m-dimensional spectral feature column vectors are formed into a fluorescence image matrix through different linear combinations. Spectral splitting requires calculating the abundance matrix H.

[0004] In the prior art, matrix decomposition methods such as the least squares method are usually used to calculate the abundance matrix H. However, these methods cannot calculate the abundance matrix H without knowing the spectral feature column vectors (that is, the spectral feature column vectors in the spectral feature matrix W are partially or completely unknown). Summary of the invention

[0005] The embodiment of the present invention provides a spectrum splitting method, aiming to solve the problem in the prior art that spectrum splitting cannot be performed when part or all of the spectrum feature column vectors are unknown.

[0006] In a first aspect, a spectrum splitting method is provided, comprising: Acquire m fluorescence images of the same size to form a first fluorescence image matrix with m rows and n columns, where m and n are positive integers greater than or equal to 1, and n represents the number of pixels in each fluorescence image; Initialize the first spectrum feature matrix with m rows and r columns with random positive decimals, where r is a positive integer greater than 1; Initialize the first abundance matrix with r rows and n columns with random positive decimals; A preset non-negative matrix decomposition NMF is performed according to the first fluorescence image matrix, the first spectral feature matrix, the first abundance matrix and k spectral feature column vectors until a preset termination condition is met, where k is an integer greater than or equal to 0 and k is less than r.

[0007] In a second aspect, a spectrum splitting device is provided, comprising: an acquisition unit, used for acquiring m fluorescence images of the same size to form a first fluorescence image matrix of m rows and n columns, where m and n are respectively positive integers greater than or equal to 1, and n represents the number of pixels of each fluorescence image; A first initialization unit is used to initialize a first spectrum feature matrix of m rows and r columns with a random positive decimal, where r is a positive integer greater than 1; A second initialization unit is used to initialize a first abundance matrix with r rows and n columns using random positive decimals; The first matrix decomposition unit is used to perform a preset non-negative matrix decomposition NMF according to the first fluorescence image matrix, the first spectral feature matrix, the first abundance matrix and k spectral feature column vectors until a preset termination condition is met, where k is an integer greater than or equal to 0 and k is less than r.

[0008] In the embodiment of the present invention, the first spectral feature matrix includes r spectral feature column vectors. A preset non-negative matrix decomposition NMF is performed based on the first fluorescence image matrix, k known spectral feature column vectors, the first spectral feature matrix initialized by random positive decimals, and the first abundance matrix. k is less than r, and spectral splitting is achieved when some or all of the spectral feature column vectors in the first spectral feature matrix are unknown. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] The above and / or additional aspects and advantages of the present invention will become apparent and easily understood from the following description of the embodiments in conjunction with the accompanying drawings, in which: Figure 1 It is a schematic diagram showing spectrum splitting through matrix operations; Figure 2 is a flow chart of the spectrum splitting method provided in Example 1 of the present invention; Figure 3 is a flow chart of initializing a first spectral feature matrix provided in Embodiment 1 of the present invention; Figure 4 is a flowchart of a preset NMF provided in Embodiment 1 of the present invention; Figure 5 is a schematic diagram of the spectrum splitting effect provided by the first embodiment of the present invention; Figure 6 This is a structural block diagram of a spectrum splitting device provided in Embodiment 2 of the present invention. DETAILED DESCRIPTION

[0010] Embodiments of the present invention are described in detail below, and examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar modules or modules with the same or similar functions. The embodiments described below by reference to the accompanying drawings are exemplary and are only used to explain the present invention, and should not be construed as limitations of the present invention. On the contrary, embodiments of the present invention include all changes, modifications and equivalents that fall within the spirit and connotation of the appended claims.

[0011] In the process of small animal fluorescence imaging, several known fluorescent substances are first injected into different parts of the experimental animal's body, and then the fluorescent signals emitted by these fluorescent substances are used to form images to observe the experimental animal. In addition, some biological tissues of the experimental animal may spontaneously fluoresce, and parts of the imaging equipment may also spontaneously fluoresce. In this case, the fluorescent signals emitted by the fluorescent substances injected into the experimental animal's body during imaging are useful signals, and the spontaneous fluorescent signals of the experimental animal and the imaging equipment are interference signals.

[0012] For example, if two known fluorescent substances are injected into experimental animals, the spectral feature column vectors of the two fluorescent substances can be obtained from a spectral library or a library built by previous experiments, that is, the spectral feature column vectors corresponding to the two fluorescent substances are known; however, the types of spontaneous fluorescence signals of experimental animals and imaging equipment are usually unclear, that is, their spectral feature column vectors are unknown. Therefore, among all the column vectors of the spectral feature matrix required for spectral splitting, two spectral feature column vectors are known, and the remaining one or more spectral feature column vectors are unknown.

[0013] The first spectral feature matrix in the embodiment of the present invention includes r spectral feature column vectors. A preset non-negative matrix factorization (NMF) is performed based on the first fluorescence image matrix, k known spectral feature column vectors, the first spectral feature matrix initialized by random positive decimals, and the first abundance matrix. k is less than r, and spectral splitting is achieved when some or all of the spectral feature column vectors in the first spectral feature matrix are unknown.

[0014] Figure 2 : is a flow chart of the spectrum splitting method provided in Example 1 of the present invention. Figure 2 As shown, the method includes: Step S201: acquiring m fluorescence images of the same size to form a first fluorescence image matrix with m rows and n columns, where m and n are positive integers greater than or equal to 1, and n represents the number of pixels in each fluorescence image.

[0015] In an embodiment of the present invention, obtaining multiple fluorescent images of the same size can be a plurality of grayscale images or color images continuously taken of the same experimental animal in an imaging device within a certain period of time, and then converting the color image into a grayscale image. The same size means that the pixel arrangement of the above-mentioned multiple fluorescent images has the same number of rows and columns. For example, 8 fluorescent images are taken, and each fluorescent image has 5 rows and 30 columns. For each fluorescent image, all its pixels are converted into row vectors in the same order, and all row vectors are combined to form a first fluorescent image matrix. For example, the first fluorescent image matrix formed is 8 rows and 150 columns.

[0016] Step S202: performing principal component analysis (PCA) on the first fluorescence image matrix to determine the number r of spectral feature column vectors contained in the first fluorescence image matrix, where r is a positive integer greater than 1.

[0017] As an embodiment of the present invention, if the number of unknown spectral feature column vectors cannot be determined, that is, it cannot be determined whether the experimental animal is autofluorescent, whether the imaging device is autofluorescent, or whether there are other sources of autofluorescence, step S202 is also included before spectral splitting, and PCA is performed on the first fluorescent image matrix to determine the number r of spectral feature column vectors contained in the first fluorescent image matrix. PCA adopts existing technology and will not be repeated here. If the number of unknown spectral feature column vectors has been determined, step S202 is skipped and step S203 is executed.

[0018] Step S203: Initialize a first spectrum feature matrix with m rows and r columns with random positive decimals, where r is a positive integer greater than 1.

[0019] In the embodiment of the present invention, the captured fluorescent image and the image corresponding to each fluorescent substance to be obtained after spectrum decomposition need to be expressed in the form of grayscale values, and the grayscale value is greater than or equal to 0. Therefore, when using the matrix decomposition method for spectrum decomposition, in addition to simple matrix operations, non-negativity constraints need to be considered, so that the results obtained are meaningful.

[0020] Figure 3 , which is a flowchart of initializing the first spectral feature matrix provided by the first embodiment of the present invention. As an embodiment of the present invention, adapted to the preset NMF in the following steps, the method of initializing the first spectral feature matrix is ​​as follows: Figure 3 As shown, the method includes: Step S301: All elements of the first spectrum feature matrix with m rows and r columns are set to random positive decimals.

[0021] Step S302: putting k spectral feature column vectors into corresponding k columns in the first spectral feature matrix respectively.

[0022] In an embodiment of the present invention, all elements of the first spectral feature matrix of m rows and r columns are set to random positive decimals, and then k spectral feature column vectors are respectively placed in the corresponding k columns of the first spectral feature matrix, where k is an integer greater than or equal to 0, and k is less than r. In the first spectral feature matrix, each column represents a spectral feature column vector. The spectral feature column vector corresponding to each fluorescent substance corresponds to a column in the first spectral feature matrix. In subsequent steps, the column in the spectral feature matrix always represents the spectral feature column vector corresponding to the fluorescent substance. If it has been determined that the first fluorescent image matrix contains 3 spectral feature column vectors, the spectral feature column vectors of the two known fluorescent substances injected into the experimental animal are placed in the first and second columns of the first spectral feature matrix, and the third column of the first spectral feature matrix represents the spectral feature column vectors corresponding to other spontaneous fluorescence. When k is equal to 0, the known spectral feature column vectors are not placed in the first spectral feature matrix.

[0023] As another embodiment of the present invention, k spectral feature column vectors may be placed into corresponding k columns of the first spectral feature matrix respectively, and then all remaining elements in the first spectral feature matrix may be set to random positive decimals.

[0024] Step S204: Initialize a first abundance matrix of r rows and n columns with random positive decimals.

[0025] Step S205: performing a preset NMF according to the first fluorescence image matrix, the first spectral feature matrix, the first abundance matrix and the k spectral feature column vectors until a preset termination condition is met.

[0026] The existing NMF is: first initialize the spectral feature matrix W and the abundance matrix H, then calculate the spectral feature matrix W of the next round based on the spectral feature matrix W and the abundance matrix H of this round, and calculate the abundance matrix H of the next round based on the spectral feature matrix W and the abundance matrix H of this round, and stop iterative updating when the Euclidean distance between the result of the next round W*H and the fluorescence image matrix V is less than a preset threshold, thereby completing the matrix decomposition and obtaining the final abundance matrix H.

[0027] Figure 4 : is a flowchart of the preset NMF provided in the first embodiment of the present invention. As an embodiment of the present invention, the preset NMF is different from the existing NMF. After calculating the spectral feature matrix W of the next round, the corresponding columns in the spectral feature matrix W are reset using the known spectral feature column vectors to obtain the final spectral feature matrix W of the next round. The method of the preset NMF is as follows: Figure 4 As shown, the method includes: Step S401: Calculate a second spectral feature matrix according to the first abundance matrix and non-negative matrix factorization (NMF), and reset corresponding k columns in the second spectral feature matrix with k spectral feature column vectors.

[0028] Step S402: Calculate a second abundance matrix according to the first spectral feature matrix and non-negative matrix factorization (NMF).

[0029] In an embodiment of the present invention, the second spectral feature matrix W is calculated based on the initialized first spectral feature matrix W, the first abundance matrix H and the existing NMF formula, and then the k known spectral feature column vectors are reset to the corresponding k columns in the second spectral feature matrix W to obtain the next round of second spectral feature matrix W; the second abundance matrix H is calculated based on the initialized first spectral feature matrix W, the first abundance matrix H and the existing NMF formula.

[0030] Step S403: Calculate a second fluorescence image matrix according to the second spectral feature matrix and the second abundance matrix.

[0031] Step S404: judging whether a preset termination condition is satisfied according to the Euclidean distance between the first fluorescent image matrix and the second fluorescent image matrix.

[0032] In the embodiment of the present invention, the product of the second spectral feature matrix W of the next round and the second abundance matrix H of the next round is calculated to obtain the second fluorescence image matrix, and then the Euclidean distance between the second fluorescence image matrix and the first fluorescence image matrix is ​​calculated to determine whether the preset termination condition is met. If the first fluorescence image matrix is ​​V1, the second fluorescence image matrix is ​​V2, which is a matrix reconstructed by W and H, represented by V2=W*H, the Euclidean distance between V2 and V1 is calculated as follows: , i represents the i-th row in the matrix, and j represents the j-th column in the matrix. The iterative update is stopped when the Euclidean distance between the two is less than a preset threshold. Optionally, the number of iterative updates is preset, and the iterative update is stopped when the number of iterative updates reaches a preset value. If the preset termination condition is not met, the preset NMF is executed cyclically, and the second spectral feature matrix, the second abundance matrix, and the second fluorescence image matrix are used for iterative update during the cycle until the preset termination condition is met.

[0033] After matrix decomposition is completed, the final abundance matrix H of r rows and n columns is obtained, and each row vector thereof represents the fluorescence signal intensity of a fluorescent substance at each pixel. In the opposite order of converting all pixels of each fluorescent image into row vectors in step S201, each row vector of the abundance matrix H is converted into an image of the same size as the fluorescent image, that is, the image corresponding to each fluorescent substance, thereby obtaining the required pure fluorescence signal.

[0034] Figure 5Schematic diagram of the spectrum splitting effect provided by the first embodiment of the present invention. For example, the first fluorescent image matrix contains three spectral feature column vectors, including spectral feature column vectors of two known fluorescent substances injected into the experimental animal and spectral feature column vectors corresponding to other autofluorescence. Figure 5 As shown, the left and middle images are images corresponding to two known fluorescent substances injected into the experimental animals, which are useful signals; the right image is an image corresponding to the autofluorescence of the experimental animals, which is an interference signal.

[0035] Step S206: performing non-negative matrix decomposition (NMF) according to the first fluorescence image matrix, the second spectral feature matrix and the second abundance matrix until a preset termination condition is met.

[0036] If the purity of a known fluorescent substance is low or it has deteriorated, its actual spectral feature column vector will no longer correspond to the spectral feature column vector obtained from the spectral library or the library built from previous experiments, which will affect the results of spectral splitting.

[0037] As one embodiment of the present invention, the spectral feature column vector corresponding to the deteriorated fluorescent substance is regarded as unknown, and the above-mentioned spectral splitting step is performed. As another embodiment of the present invention, the spectral feature column vector obtained from the spectral library or the library built by the past experiment of the deteriorated fluorescent substance is still used as the known spectral feature column vector to perform the above-mentioned spectral splitting step, and then step S206 is performed. In an embodiment of the present invention, after completing the above-mentioned preset NMF, the existing NMF is performed with the iteratively updated second spectral feature matrix and the second abundance matrix as initial values, wherein the known spectral feature column vector is no longer used to reset the corresponding columns in the second spectral feature matrix. Both of the above-mentioned two embodiments can not only realize accurate spectral splitting, but also update the spectral feature column vector corresponding to the deteriorated fluorescent substance. After the spectral feature column vector corresponding to the deteriorated fluorescent substance is updated, it can replace the original corresponding spectral feature column vector in the spectral library or the library built by the experiment for use in subsequent experiments.

[0038] The first spectral feature matrix in the embodiment of the present invention includes r spectral feature column vectors. The NMF is preset according to the first fluorescence image matrix, k known spectral feature column vectors, the first spectral feature matrix initialized by random positive decimals, and the first abundance matrix. k is less than r. Spectral splitting is achieved when some or all of the spectral feature column vectors in the first spectral feature matrix are unknown.

[0039] Figure 6 The structure block diagram of the spectrum splitting device provided in the second embodiment of the present invention is as follows: Figure 6As shown, the device includes: an acquisition unit 61, a first initialization unit 62, a second initialization unit 63 and a first matrix decomposition unit 64.

[0040] The acquisition unit 61 is used to acquire m fluorescence images of the same size to form a first fluorescence image matrix with m rows and n columns, where m and n are positive integers greater than or equal to 1, and n represents the number of pixels in each fluorescence image.

[0041] The first initialization unit 62 is used to initialize a first spectrum feature matrix having m rows and r columns with a random positive decimal, where r is a positive integer greater than 1.

[0042] The second initialization unit 63 is used to initialize the first abundance matrix with r rows and n columns using random positive decimals.

[0043] The first matrix decomposition unit 64 is used to perform a preset non-negative matrix decomposition NMF according to the first fluorescence image matrix, the first spectral feature matrix, the first abundance matrix and k spectral feature column vectors until a preset termination condition is met, where k is an integer greater than or equal to 0 and k is less than r.

[0044] Preferably, the first initialization unit 62 includes a random subunit and a value assignment subunit.

[0045] The random subunit is used to set all elements of the first spectrum feature matrix with m rows and r columns to random positive decimals.

[0046] The assignment subunit is used to put the k spectral feature column vectors into the corresponding k columns in the first spectral feature matrix respectively, where k is an integer greater than or equal to 0 and k is less than r.

[0047] Preferably, the first matrix decomposition unit 64 includes a first calculation subunit, a second calculation subunit, a third calculation subunit and a judgment subunit.

[0048] The first calculation subunit is used for calculating the second spectral feature matrix according to the first abundance matrix and non-negative matrix factorization NMF, and resetting the corresponding k columns in the second spectral feature matrix with k spectral feature column vectors.

[0049] The second calculation subunit is used to calculate the second abundance matrix according to the first spectral feature matrix and non-negative matrix factorization NMF.

[0050] The third calculation subunit is used to calculate a second fluorescence image matrix according to the second spectral feature matrix and the second abundance matrix.

[0051] The judging subunit is used to judge whether a preset termination condition is satisfied according to the Euclidean distance between the first fluorescent image matrix and the second fluorescent image matrix.

[0052] Furthermore, the spectrum splitting device also includes a second matrix decomposition unit, which is used to perform non-negative matrix decomposition NMF according to the first fluorescence image matrix, the second spectrum feature matrix and the second abundance matrix until a preset termination condition is met.

[0053] Furthermore, the spectrum splitting device also includes a principal component analysis unit, which is used to perform principal component analysis PCA on the first fluorescence image matrix to determine the number r of spectral feature column vectors contained in the first fluorescence image matrix, where r is a positive integer greater than 1.

[0054] The first spectral feature matrix in the embodiment of the present invention includes r spectral feature column vectors. The NMF is preset according to the first fluorescence image matrix, k known spectral feature column vectors, the first spectral feature matrix initialized by random positive decimals, and the first abundance matrix. k is less than r. Spectral splitting is achieved when some or all of the spectral feature column vectors in the first spectral feature matrix are unknown.

[0055] Although the embodiments of the present invention have been shown and described above, it is to be understood that the above embodiments are exemplary and are not to be construed as limitations of the present invention. A person skilled in the art may change, modify, replace and vary the above embodiments within the scope of the present invention.

Claims

1. A spectrum splitting method, characterized in that: The method comprises: Acquire m fluorescence images of the same size to form a first fluorescence image matrix with m rows and n columns, where m and n are positive integers greater than or equal to 1, and n represents the number of pixels in each fluorescence image; Initialize the first spectrum feature matrix with m rows and r columns with random positive decimals, where r is a positive integer greater than 1; Initialize the first abundance matrix with r rows and n columns with random positive decimals; A preset non-negative matrix decomposition NMF is performed according to the first fluorescence image matrix, the first spectral feature matrix, the first abundance matrix and k spectral feature column vectors until a preset termination condition is met, where k is an integer greater than or equal to 0 and k is less than r.

2. The method according to claim 1, characterized in that Initializing the first spectrum feature matrix of m rows and r columns with random positive decimals includes: Set all elements of the first spectrum feature matrix with m rows and r columns to random positive decimals; The k spectral feature column vectors are respectively placed into the corresponding k columns of the first spectral feature matrix, where k is an integer greater than or equal to 0 and less than r.

3. The method according to claim 1, characterized in that Performing a preset non-negative matrix decomposition NMF according to the first fluorescence image matrix, the first spectral feature matrix, the first abundance matrix, and k spectral feature column vectors until a preset termination condition is met includes: Calculate a second spectral feature matrix according to the first abundance matrix and non-negative matrix factorization (NMF), and reset corresponding k columns in the second spectral feature matrix with k spectral feature column vectors; Calculate a second abundance matrix based on the first spectral feature matrix and non-negative matrix factorization (NMF); calculating a second fluorescence image matrix according to the second spectral feature matrix and the second abundance matrix; Whether a preset termination condition is satisfied is determined according to the Euclidean distance between the first fluorescent image matrix and the second fluorescent image matrix.

4. The method according to claim 3, characterized in that Also includes: Non-negative matrix decomposition (NMF) is performed according to the first fluorescence image matrix, the second spectral feature matrix and the second abundance matrix until a preset termination condition is met.

5. The method according to any one of claims 1 to 4, characterized in that: Before initializing the first spectrum feature matrix of m rows and r columns with random positive decimals, the method further includes: A principal component analysis (PCA) is performed on the first fluorescent image matrix to determine the number r of spectral feature column vectors included in the first fluorescent image matrix, where r is a positive integer greater than 1.

6. A spectrum splitting device, characterized in that: The device comprises: an acquisition unit, used for acquiring m fluorescence images of the same size to form a first fluorescence image matrix of m rows and n columns, where m and n are respectively positive integers greater than or equal to 1, and n represents the number of pixels of each fluorescence image; A first initialization unit is used to initialize a first spectrum feature matrix of m rows and r columns with a random positive decimal, where r is a positive integer greater than 1; A second initialization unit is used to initialize a first abundance matrix with r rows and n columns using random positive decimals; The first matrix decomposition unit is used to perform a preset non-negative matrix decomposition NMF according to the first fluorescence image matrix, the first spectral feature matrix, the first abundance matrix and k spectral feature column vectors until a preset termination condition is met, where k is an integer greater than or equal to 0 and k is less than r.

7. The device according to claim 6, characterized in that The first initialization unit comprises: A random subunit, used for setting all elements of the first spectrum feature matrix with m rows and r columns to random positive decimals; The assignment subunit is used to put k spectral feature column vectors into corresponding k columns in the first spectral feature matrix respectively, where k is an integer greater than or equal to 0 and less than r.

8. The device according to claim 6, characterized in that The first matrix decomposition unit comprises: A first calculation subunit is used to calculate a second spectral feature matrix according to the first abundance matrix and non-negative matrix factorization (NMF), and reset corresponding k columns in the second spectral feature matrix with k spectral feature column vectors; A second calculation subunit is used to calculate a second abundance matrix according to the first spectral feature matrix and non-negative matrix decomposition NMF; a third calculation subunit, configured to calculate a second fluorescence image matrix according to the second spectral feature matrix and the second abundance matrix; The judging subunit is used to judge whether a preset termination condition is satisfied according to the Euclidean distance between the first fluorescent image matrix and the second fluorescent image matrix.

9. The device according to claim 8, characterized in that It also includes a second matrix decomposition unit, which is used to perform non-negative matrix decomposition NMF according to the first fluorescence image matrix, the second spectral feature matrix and the second abundance matrix until a preset termination condition is met.

10. The device according to any one of claims 6 to 9, characterized in that: It also includes a principal component analysis unit, which is used to perform principal component analysis PCA on the first fluorescent image matrix to determine the number r of spectral feature column vectors contained in the first fluorescent image matrix, where r is a positive integer greater than 1.

Citation Information

Patent Citations

  • Optical function imaging method combining spatial regularization and semi-blind spectrum unmixing

    CN113066142A

  • Spectral image unmixing method based on weighted non-negative matrix factorization and application thereof

    CN113793646A

  • Multi-channel fluorescence microscopic image spectrum splitting method based on non-negative matrix factorization

    CN114820472A

  • Hyperspectral imaging system with mixed unmixing

    CN118450841A