CN103903231A - Method for removing auto-fluorescence interference in multi-spectra excitation fluorescence imaging - Google Patents

Abstract

The invention discloses a method for removing autofluorescence interference in multi-spectral excited fluorescence imaging, which is characterized in that it comprises the following steps: collecting multiple multi-spectral original fluorescence images on the surface of a living body; selecting from the images according to the size of the living body Rectangular region of interest; construct the combined image and calculate its correlation matrix, and perform eigenvalue decomposition on the correlation matrix; calculate the initial value of iteration; iteratively solve the autofluorescence image; remove the autofluorescence image. The advantages of the present invention are: only one wavelength of excitation light source is required, and there is no requirement for the excitation efficiency of fluorescent probes at different wavelengths; the collected multi-spectral original fluorescence images are cut without loss of fluorescence image information At the same time, the number of pixels participating in the processing is reduced, and the amount of calculation is reduced; the initial value of the iteration is obtained through eigenvalue decomposition, and no in vitro experiments are required to measure the spectral curve of the fluorescence signal, which reduces the complexity of implementation while ensuring the calculation accuracy.

Description

A Method for Removing Autofluorescence Interference in Multispectral Excited Fluorescence Imaging

technical field

The invention relates to an image preprocessing method for excited fluorescence imaging, in particular to a method for removing autofluorescence interference in multispectral excited fluorescence imaging, and belongs to the technical field of image preprocessing.

Background technique

In multi-spectral excited fluorescence imaging, the fluorescent probe in the organism is excited by an external light source, and the emitted fluorescence signal of the fluorescent probe reaches the surface of the organism through scattering and absorption in the organism. Multi-spectral excited fluorescence imaging uses devices such as optical fibers or cameras to collect fluorescent signals on the surface of organisms filtered by band-pass filter groups, which can conduct qualitative and quantitative research on biological processes in vivo at the cellular and molecular levels. It is widely used in fields such as diagnosis, biodistribution research of probes, and in vivo metabolism tracking of targeted drugs.

Autofluorescence mainly comes from the fluorescence produced by living organisms' own intestines, fat, skin and other tissues excited by external light sources. It is relatively strong in the visible light band (400nm-700nm), and when it is serious, it will drown the useless target fluorescence signal. interference.

In the prior art, the autofluorescence interference is removed by collecting two fluorescence images with different excitation wavelengths or a plurality of different emission wavelengths, and adopting an image processing method.

Chinese invention patent application, publication number: CN102096914A, discloses a method for removing autofluorescence interference in bioluminescent images. In this method, two excited fluorescence images of different wavelength bands were collected, and the interference of autofluorescence was removed by methods such as cluster analysis and seed flooding. This method requires that the excitation efficiency of the dye in the excited fluorescence image 1 is higher than that of the fluorescent dye in the excited fluorescence image 2, and the excitation wavelength of the excitation light in the excited fluorescence image 1 is greater than that in the excited fluorescence image 2.

I.Mars“Nonnegative matrixfactorization:a blind spectra separation method for in vivofluorescent optical imaging”,Journal of BiomedicalOptics,15(5),056009,2010公开了一种多光谱激发荧光成像中的自体荧光去除方法。该方法通过采集多幅不同发射波长的荧光图像，以体外实验得到的荧光探针的光谱曲线作为迭代初值，采用加入正则化约束项的非负矩阵分解方法，去除自体荧光。Literature Anne-sophie Montcuquet, Lionel Hervé, Fabrice Navarro, Jean-Marc Dinten,

I.Mars "Nonnegative matrixfactorization: a blind spectrum separation method for in vivofluorescent optical imaging", Journal of Biomedical Optics, 15(5), 056009, 2010 discloses a method for removing autofluorescence in multispectral excited fluorescence imaging. In this method, multiple fluorescent images with different emission wavelengths are collected, and the spectral curve of the fluorescent probe obtained from in vitro experiments is used as an initial iteration value, and a non-negative matrix decomposition method with regularization constraints is used to remove autofluorescence.

The defects in the implementation of the above two types of methods are as follows: the first type of method requires two excitation light sources with different wavelengths, and the excitation efficiency of the fluorescent probe required to be used is higher than that of the short wavelength at the long wavelength; The second type of method uses the spectral curve of the fluorescent probe obtained from the in vitro experiment as the iterative initial value, and the user needs to perform an in vitro measurement experiment, which is inconvenient to operate; at the same time, since the region of interest usually only occupies part of the pixels in the image, this type of method directly To process the collected multispectral images, the region of interest is not selected and cropped, which requires a large amount of computation.

Contents of the invention

In order to solve the deficiencies of the prior art, the purpose of the present invention is to provide a method for removing autofluorescence interference that has no special requirements on excitation light sources and fluorescent probes, and does not require in vitro measurement of fluorescence signal spectral curves.

In order to achieve the above object, the present invention adopts the following technical solutions:

A method for removing autofluorescence interference in multispectral excited fluorescence imaging, comprising the following steps:

(1) Acquisition of multi-spectral raw fluorescence images: use multi-spectral excitation fluorescence imaging equipment to collect J multi-spectral raw fluorescence images of biological body surfaces under the same excitation wavelength, denoted as A _j , j=1,2,... ,J;

(2) Image cropping: According to the size of the organism, select a rectangular region of interest from the J multispectral raw fluorescence images collected in step (1), and crop the multispectral raw fluorescence images. The length and length of the cropped fluorescence images are The width is the same as the length and width of the organism in the image, denoted as B _j , B _j includes P rows and Q columns of pixels, j=1,2,...,J;

(3) Eigenvalue decomposition: construct a combined image C, the jth row of the combined image C is composed of row vectors arranged by the image Bj, the combined image C includes J rows and L columns of pixels, L=P×Q, and the combined image C is calculated Correlation matrix D of the correlation matrix D, perform eigenvalue decomposition on the correlation matrix D, take the eigencolumn vector u corresponding to the largest eigenvalue, and set the elements less than 0 in u to 0, and keep the rest of the elements unchanged to obtain v;

(4) Calculate the initial value of iteration: According to the vector v obtained in step (3), calculate the initial value of iteration a ₀ and s ₀ by the following formula:

a ₀ =1 (1)

s ₀ =v ^T (2)

In formula (1), 1 means a full 1-column vector containing L elements;

In formula (2), T represents the transpose operation;

(5) Iteratively solve the autofluorescence image: take a ₀ and s ₀ in step (4) as initial values, and use the following formula to perform K iterations:

${\mathrm{<span\; class="notranslate">\; a</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; =</span>\frac{{\mathrm{<span\; class="notranslate">\; a</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; i</span>}}{<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; Bs</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}^{\mathrm{<span\; class="notranslate">\; T</span>}}<span\; class="notranslate">\; )</span>}_{\mathrm{<span\; class="notranslate">\; i</span>}}}{{\mathrm{<span\; class="notranslate">\; a</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; i</span>}}{\mathrm{<span\; class="notranslate">\; the\; s</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; the\; s</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}^{\mathrm{<span\; class="notranslate">\; T</span>}}}<span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 3</span>}<span\; class="notranslate">\; )</span>$

${\mathrm{<span\; class="notranslate">\; the\; s</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; =</span>\frac{{\mathrm{<span\; class="notranslate">\; a</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; i</span>}}{<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; a</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}}^{\mathrm{<span\; class="notranslate">\; T</span>}}\mathrm{<span\; class="notranslate">\; B</span>}<span\; class="notranslate">\; )</span>}_{\mathrm{<span\; class="notranslate">\; i</span>}}}{{\mathrm{<span\; class="notranslate">\; a</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}}^{\mathrm{<span\; class="notranslate">\; T</span>}}{\mathrm{<span\; class="notranslate">\; a</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}}{\mathrm{<span\; class="notranslate">\; a</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; i</span>}}}<span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 4</span>}<span\; class="notranslate">\; )</span>$

表示列向量

的第i个元素；

表示行向量

的第i个元素；In formula (3) and formula (4), k=1,2,...,K; a _k,i represents the i-th element of the column vector a _k obtained in the k-th iteration;

Represents a column vector

The i-th element of ;

represents a row vector

The i-th element of ;

Calculate E=a _K s _K according to the results obtained in the K-th iteration, and reconstruct each row of the aforementioned E into an image including P rows and Q columns of pixels to obtain J autofluorescence images, denoted as F _j , j= 1,2,...,J;

(6) Remove the autofluorescence image: Subtract the autofluorescence image F _j obtained in step (5) from the cropped J fluorescence images B _j to obtain J fluorescence images G _j including only target information:

G _j =B _j -F _j (5)

In the formula, j=1,2,...,J.

The aforementioned method for removing autofluorescence interference in multispectral excited fluorescence imaging is characterized in that, in step (1), J emission filters are used to collect the aforementioned multispectral raw fluorescence images.

The aforementioned method for removing autofluorescence interference in multi-spectral excited fluorescence imaging is characterized in that the aforementioned J emission filters have the same bandwidth, and the center wavelength intervals of adjacent filters are equal to the filter bandwidth.

The aforementioned method for removing autofluorescence interference in multispectral excited fluorescence imaging is characterized in that, in step (1), when collecting the aforementioned J multispectral raw fluorescence images, the integration time of the acquisition device and the posture of the imaging organism remain unchanged .

The aforementioned method for removing autofluorescence interference in multispectral excited fluorescence imaging is characterized in that, in step (3), the aforementioned correlation matrix D is calculated according to the following formula:

$\mathrm{<span\; class="notranslate">\; D.</span>}<span\; class="notranslate">\; =</span>\frac{\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; L</span>}}{\mathrm{<span\; class="notranslate">\; CC</span>}}^{\mathrm{<span\; class="notranslate">\; T</span>}}<span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 6</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; .</span>$

The benefits of the present invention are:

1. Only one wavelength of excitation light source is required, and there is no requirement for the excitation efficiency of fluorescent probes at different wavelengths;

2. The present invention cuts the collected multi-spectral original fluorescence image, reduces the number of pixels involved in the processing without losing the information of the fluorescence image, and reduces the amount of computation;

3. The present invention obtains the iterative initial value through eigenvalue decomposition, and does not need to conduct in vitro experiments in advance to measure the spectral curve of the fluorescence signal, which reduces the implementation complexity while ensuring the calculation accuracy.

Description of drawings

Fig. 1 is the main flow diagram of the method of the present invention;

Fig. 2 is a diagram of simulation experiment results obtained by using the method of the present invention.

Detailed ways

The method provided by the present invention collects multiple emission fluorescence images under the same excitation wavelength and cuts the images, utilizes the characteristics of different spectral curves of useful fluorescence signals and autofluorescence signals in the process of multi-spectral excitation fluorescence imaging, and provides an iterative initial method through eigenvalue decomposition. Values, autofluorescence was removed using the non-negative matrix factorization method.

The present invention will be specifically introduced below in conjunction with the accompanying drawings and specific embodiments.

Referring to Fig. 1, the method for removing autofluorescence in the multispectral excited fluorescence imaging of the present invention comprises the following steps:

1. Acquisition of multispectral raw fluorescence images

Taking organisms with fluorescent probes in the body as imaging objects, using multi-spectral excitation fluorescence imaging equipment to collect images, specifically, by switching the emission filters of the multi-spectral excitation fluorescence imaging equipment, collecting images under the same excitation wavelength and different emission wavelengths The raw multispectral fluorescence images of J biological surface, denoted as A _j , j=1,2,...,J.

The number of emission filters is J, and the bandwidths of the J emission filters are the same, and the center wavelength interval of adjacent filters is equal to the bandwidth of the filters.

When acquiring J multispectral raw fluorescence images, the integration time of the acquisition equipment and the pose of the imaging organism remain unchanged.

2. Image cropping

Usually, the imaging field of view of multi-spectral excited fluorescence imaging equipment is larger than the imaging object, and only some pixels of the collected multi-spectral raw fluorescence image contain valid information. The size of the organism, select a rectangular region of interest from the J multispectral raw fluorescence images collected in step 1, and crop the multispectral raw fluorescence image. The length and width of the cropped fluorescence image are the same as the length and width of the organism in the image Same, denoted as B _j , B _j includes pixels in rows P and columns Q, j=1,2,...,J.

3. Eigenvalue decomposition

(1) Construct the combined image C. The jth row of the combined image C is composed of row vectors arranged by the image B _j . The combined image C includes J rows and L columns of pixels, L=P×Q.

The specific construction steps of the combined image C are as follows:

The cropped fluorescent image B _j is composed of pixels from the first row to the Jth row of pixels into a row of pixels, the number of pixels L is: L=P×Q, J rows of pixels are obtained from J images, and the J row of pixels starts from the top Arranged downwards to form a composite image including J rows and L columns of pixels.

(2) Calculate the correlation matrix D of the combined image C, and the correlation matrix D is calculated according to the following formula:

$\mathrm{<span\; class="notranslate">\; D.</span>}<span\; class="notranslate">\; =</span>\frac{\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; L</span>}}{\mathrm{<span\; class="notranslate">\; CC</span>}}^{\mathrm{<span\; class="notranslate">\; T</span>}}<span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 6</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; .</span>$

(3) Decompose the eigenvalue of the correlation matrix D, take the eigencolumn vector u corresponding to the largest eigenvalue, set u and the elements less than 0 in u to 0, and keep the rest of the elements unchanged to obtain v.

4. Calculate the initial value of iteration

According to the vector v obtained in step 3, the iteration initial values a ₀ and s ₀ are calculated by the following formula:

a ₀ =1 (1)

s ₀ =v ^T (2)

In formula (1), 1 represents a full 1-column vector containing L elements;

In formula (2), T represents the transposition operation.

5. Iteratively solve the autofluorescence image

Taking a ₀ and s ₀ in step 4 as initial values, use the following formula to perform K iterations:

${\mathrm{<span\; class="notranslate">\; a</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; =</span>\frac{{\mathrm{<span\; class="notranslate">\; a</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; i</span>}}{<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; Bs</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}^{\mathrm{<span\; class="notranslate">\; T</span>}}<span\; class="notranslate">\; )</span>}_{\mathrm{<span\; class="notranslate">\; i</span>}}}{{\mathrm{<span\; class="notranslate">\; a</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; i</span>}}{\mathrm{<span\; class="notranslate">\; the\; s</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; the\; s</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}^{\mathrm{<span\; class="notranslate">\; T</span>}}}<span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 3</span>}<span\; class="notranslate">\; )</span>$

${\mathrm{<span\; class="notranslate">\; the\; s</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; =</span>\frac{{\mathrm{<span\; class="notranslate">\; a</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; i</span>}}{<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; a</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}}^{\mathrm{<span\; class="notranslate">\; T</span>}}\mathrm{<span\; class="notranslate">\; B</span>}<span\; class="notranslate">\; )</span>}_{\mathrm{<span\; class="notranslate">\; i</span>}}}{{\mathrm{<span\; class="notranslate">\; a</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}}^{\mathrm{<span\; class="notranslate">\; T</span>}}{\mathrm{<span\; class="notranslate">\; a</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}}{\mathrm{<span\; class="notranslate">\; a</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; i</span>}}}<span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 4</span>}<span\; class="notranslate">\; )</span>$

表示列向量

的第i个元素；表示行向量

的第i个元素。In formula (3) and formula (4), k=1,2,...,K; a _k,i represents the i-th element of the column vector a _k obtained in the k-th iteration;

Represents a column vector

The i-th element of ; represents a row vector

The i-th element of .

Calculate E=a _K s _K based on the results obtained in the Kth iteration, and reconstruct each row of E into an image including P rows and Q columns of pixels to obtain J autofluorescence images, denoted as F _j , j=1 ,2,...,J.

Specifically, the construction steps of J autofluorescence images _Fj are as follows:

Divide each row of pixels of E into P parts of Q pixels, and arrange the P parts of Q pixels from top to bottom to obtain a P×Q two-dimensional image. E includes J lines, and a total of J pictures are obtained Autofluorescence image.

6. Remove autofluorescence images

Subtract the autofluorescence image F _j obtained in step 5 from the cropped J fluorescence images B _j to obtain J fluorescence images G _j that only include target information:

G _j =B _j -F _j (5)

In the formula, j=1,2,...,J.

In order to better understand the present invention, the effects of the present invention will be further described below in combination with simulation experiments. The simulation experiments use nude mice with fluorescent probes in their bodies as imaging objects.

1. Collect multispectral raw fluorescence images.

The nude mice labeled with the fluorescent probe were irradiated with an excitation light source matching the absorption peak wavelength of the fluorescent probe, and the original fluorescence images of the six central wavelengths near the emission peak of the fluorescent probe on the surface of the nude mice were collected using a multi-spectral excitation fluorescence device. , denoted as A _j , j=1,2,…,6.

It should be noted that the multi-spectral excitation fluorescence equipment uses a high-sensitivity CCD camera for image acquisition, and the number of pixels is 1024×1024. Therefore, each original fluorescence image collected consists of 1024×1024 pixels.

2. Image cropping.

According to the size of the nude mice, a rectangular region of interest of 725×270 was selected from the six original _fluorescence images collected _in step 1, and the original fluorescence images were cropped. Column of pixels, j=1,2,…,6.

3. Eigenvalue decomposition.

Combine the pixels from the 1st row to the 725th row of the cropped fluorescent image B _j into a row of pixels, the number of pixels is 725×270=195750, and 6 rows of pixels are obtained from 6 images, and the 6 rows of pixels are from top to bottom Arranged to form a combined image containing 6 rows and 195,750 columns of pixels, denoted as C, and the correlation matrix D of the combined image C is calculated according to the following formula:

$\mathrm{<span\; class="notranslate">\; D.</span>}<span\; class="notranslate">\; =</span>\frac{\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; 195750</span>}}{\mathrm{<span\; class="notranslate">\; CC</span>}}^{\mathrm{<span\; class="notranslate">\; T</span>}}$

Among them, T represents the transposition operation, and the eigenvalue decomposition is performed on the correlation matrix D, and the characteristic column vector u corresponding to the largest eigenvalue is taken, and the elements less than 0 in u are set to 0, and the remaining elements remain unchanged to obtain v.

4. Calculate the initial value of iteration.

The iteration initial values a ₀ and s ₀ are calculated by the following formula:

a ₀ =1, s ₀ =v ^T

Among them, 1 means a full 1-column vector containing 195750 elements, and T means a transpose operation.

5. Iteratively solve the autofluorescence image.

Taking a ₀ and s ₀ in step 4 as initial values, use the following formula to perform 20 iterations:

${\mathrm{<span\; class="notranslate">\; a</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; =</span>\frac{{\mathrm{<span\; class="notranslate">\; a</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; i</span>}}{<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; Bs</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}^{\mathrm{<span\; class="notranslate">\; T</span>}}<span\; class="notranslate">\; )</span>}_{\mathrm{<span\; class="notranslate">\; i</span>}}}{{\mathrm{<span\; class="notranslate">\; a</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; i</span>}}{\mathrm{<span\; class="notranslate">\; the\; s</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; the\; s</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}^{\mathrm{<span\; class="notranslate">\; T</span>}}}$

${\mathrm{<span\; class="notranslate">\; the\; s</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; =</span>\frac{{\mathrm{<span\; class="notranslate">\; a</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; i</span>}}{<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; a</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}}^{\mathrm{<span\; class="notranslate">\; T</span>}}\mathrm{<span\; class="notranslate">\; B</span>}<span\; class="notranslate">\; )</span>}_{\mathrm{<span\; class="notranslate">\; i</span>}}}{{\mathrm{<span\; class="notranslate">\; a</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}}^{\mathrm{<span\; class="notranslate">\; T</span>}}{\mathrm{<span\; class="notranslate">\; a</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}}{\mathrm{<span\; class="notranslate">\; a</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; i</span>}}}$

表示列向量

的第i个元素；

表示行向量

的第i个元素。In the formula, k=1,2,...,20; a _k,i represents the ith element of the column vector a _k obtained in the kth iteration;

Represents a column vector

The i-th element of ;

represents a row vector

The i-th element of .

Calculate E=a _K s _K according to the results obtained in the Kth iteration, and reconstruct each row of E into an image including 725 rows and 270 columns of pixels to obtain 6 autofluorescence images, which are denoted as F ₁ , F ₂ ,..., _F6 .

6. Remove the autofluorescence image, subtract the autofluorescence image obtained in step 5 from the 6 cropped fluorescence images according to the following formula, and obtain 6 fluorescence images including only target information:

G _j =B _j -F _j , j=1,2,...,6.

Figure 2 shows the results of two sets of simulation experiments. In the figure, a1, a2,...,a6 show six cropped multi-spectral excitation fluorescence images, each fluorescence image includes 725×270 pixels, including target fluorescence signals and autofluorescence signals , it can be seen that the autofluorescence signal interferes with the target fluorescence signal to some extent. Figures b1, b2, ..., b6 show the fluorescence images after the autofluorescence is removed by the method of the present invention, and it can be seen that the autofluorescence is effectively removed.

This shows that method of the present invention has following advantage:

1. Only one wavelength of excitation light source is required, and there is no requirement for the excitation efficiency of fluorescent probes at different wavelengths;

2. Crop the collected multi-spectral original fluorescence image, reduce the number of pixels involved in the processing without losing the information of the fluorescence image, and reduce the amount of calculation;

3. The iterative initial value is obtained through eigenvalue decomposition, and no in vitro experiments are required to measure the spectral curve of the fluorescent signal, which reduces the complexity of implementation while ensuring calculation accuracy.

It should be noted that the above embodiments do not limit the present invention in any form, and all technical solutions obtained by means of equivalent replacement or equivalent transformation fall within the protection scope of the present invention.

Claims (5)

1. in multispectral fluorescence excitation imaging, remove the method that autofluorescence is disturbed, it is characterized in that, comprise the following steps:

(1), gather multispectral original fluoroscopic image: utilize multispectral fluorescence excitation imaging device to gather the multispectral original fluoroscopic image of the J width organism surface under same excitation wavelength, be designated as A _j, j=1,2 ..., J;

(2), image cropping: according to the size of biosome, J the multispectral original fluoroscopic image gathering from step (1), choose rectangle area-of-interest, multispectral original fluoroscopic image is carried out to cutting, and the long and wide length with biosome in image of fluoroscopic image after cutting and wide identical, is designated as B _j, B _jcomprise the capable Q row of P pixel, j=1,2 ..., J;

(3), Eigenvalues Decomposition: build combination image C, the j of combination image C is capable of image B _jthe row vector being arranged in forms, and combination image C comprises the capable L row of J pixel, L=P × Q, the correlation matrix D of calculation combination image C, carries out Eigenvalues Decomposition to correlation matrix D, gets eigenvalue of maximum characteristic of correspondence column vector u, and in u, be less than 0 element and be set to 0, all the other elements are constant, obtain v;

(4), calculate iterative initial value: the vector v obtaining according to step (3), calculate iterative initial value a by following formula ₀and s ₀:

a ₀=1 （1）

s ₀=v ^T （2）

In formula (1), 1 represents complete 1 column vector that comprises L element;

In formula (2), T represents transposition computing;

(5), iterative Autofluorescence imaging: with a in step (4) ₀and s ₀as initial value, utilize following formula to carry out iteration K time:

$a_{k,i}=\frac{a_{k-1,i}{\left({\mathrm{Bs}}_{k-1}^T\right)}_i}{a_{k-1,i}{s}_{k-1}{s}_{k-1}^T}---\left(3\right)$

$s_{k,i}=\frac{a_{k,i}{\left(a_k^{T}B\right)}_i}{a_k^T{a}_k{a}_{k-1,i}}---\left(4\right)$

In formula (3) and formula (4), k=1,2 ..., K; a _k,irepresent the column vector a that the k time iteration obtains _ki element;

represent column vector

i element;

represent row vector i element;

The result obtaining according to the K time iteration is calculated E=a _ks _k, and become a width to comprise the image of the capable Q row of P pixel each line reconstruction of described E, and obtain J width Autofluorescence imaging, be designated as F _j, j=1,2 ..., J;

(6), remove Autofluorescence imaging: the J width fluoroscopic image B obtaining from cutting _jin deduct the Autofluorescence imaging F being obtained by step (5) _j, obtain the fluoroscopic image G that J width only comprises target information _j:

G _j=B _j-F _j （5）

In formula, j=1,2 ..., J.

2. in multispectral fluorescence excitation imaging according to claim 1, remove the method that autofluorescence is disturbed, it is characterized in that, in step (1), adopt J transmitting optical filter to gather described multispectral original fluoroscopic image.

3. in multispectral fluorescence excitation imaging according to claim 2, remove the method that autofluorescence is disturbed, it is characterized in that, described J transmitting optical filter bandwidth centre wavelength interval identical, adjacent optical filter equates with optical filter bandwidth.

4. in multispectral fluorescence excitation imaging according to claim 1, remove the method that autofluorescence is disturbed, it is characterized in that, in step (1), while gathering described J multispectral original fluoroscopic image, collecting device integral time and imaging biosome attitude all remain unchanged.

5. in multispectral fluorescence excitation imaging according to claim 1, remove the method that autofluorescence is disturbed, it is characterized in that, in step (3), described correlation matrix D calculates according to following formula:

$D=\frac{1}{L}{\mathrm{CC}}^T---\left(6\right).$

Images