Structured light illumination microscopy method based on principal component analysis

Abstract

The application discloses a kind of structure light illumination microscopic imaging methods based on principal component analysis, and the "dimension reduction" tool based on principal component analysis is used to eliminate the noisy, interfering component irrelevant to illumination in actual imaging, to extract the "first principal component" dominated by illumination parameter, fundamentally eliminate the interference term affecting the accurate estimation of illumination parameter, so as to accurately identify the wave vector and initial phase of sub-pixel accuracy in a simple and efficient way.In addition, the application uses a double window frequency domain mask operator to further suppress interference noise, while reducing the data involved in principal component analysis by nearly 1000 times, greatly improving the calculation efficiency, accuracy and stability of illumination parameter estimation.

Description

Technical Field

[0001] The present invention belongs to the technical field of super-resolution fluorescence microscopy, and specifically provides a structured illumination microscopy based on principal component analysis (PCA-SIM) method for real-time super-resolution observation of living cells. Background Art

[0002] Studying the dynamics of subcellular biological structure and function is crucial for understanding the mechanisms of cell fate determination, uncovering the essence of life phenomena, and exploring the mechanisms of major diseases. In recent years, super-resolution imaging techniques such as stimulated emission depletion (STED), structured illumination microscopy (SIM), and single-molecule localization microscopy (PALM / STORM) have circumvented the Abbe diffraction limit, which for over a century seemed unbreakable. These techniques allow for visualization of biomolecules at the nanoscale and even at the single-molecule level, allowing for unprecedented spatial and temporal resolution in understanding the structure and dynamics of living cells, becoming indispensable tools in life science research. In 2014, the Royal Swiss Academy of Sciences awarded the Nobel Prize in Chemistry to super-resolution fluorescence microscopy, once again demonstrating the importance of this technology in human development and its future direction [Weiss, P.S. Nobel Prizes for super-resolution imaging. (2014)].

[0003] Among many super-resolution imaging techniques, SIM uses structured illumination with sinusoidal intensity modulation to encode high-frequency information of the sample into the detection passband of the optical system in the frequency domain by spatial mixing, thereby achieving super-resolution optical microscopy with a lateral resolution of approximately 100 nm, which breaks the Abbe diffraction limit [Gustafsson, M.G. Surpassing the lateral resolution limit by a factor of two using structured illumination microscopy. J. Microsc. 198, 82–87 (2000).]. Compared with PALM and STED, SIM has unique advantages such as wide field of view, fast imaging, weak photobleaching and phototoxicity, and non-specific requirements for fluorescent dyes. Therefore, it is particularly suitable for long-term dynamic super-resolution imaging of living cells [Li, D. et al. Extended-resolution structured illumination imaging of endocytic and cytoskeletal dynamics. Science 349, (2015).]. Despite its many advantages, SIM still faces some technical challenges. The numerical reconstruction of high-quality super-resolution images depends heavily on the accurate estimation of illumination parameters. Small errors in these parameters can have a significant impact on the reconstruction results, leading to reconstruction artifacts that interfere with quantitative [Müller, M., V.,Hennig,S.,Hübner,W.&Huser,T.Open-source image reconstruction of super-resolution structured illumination microscopy data in ImageJ.Nat.Commun.7,1–6(2016).]. In addition, the "variable" illumination parameters are extremely sensitive to the experimental environment. Unless the imaging conditions are strictly kept stable, it is necessary to extract the illumination parameters frame by frame from the acquired raw observation data before reconstructing each frame of super-resolution image [Huang,X.et al.Fast,long-term,super-resolution imaging with Hessian structured illuminationmicroscopy.Nat.Biotechnol.36,451–459(2018).]. The most widely used parameter estimation method based on iterative cross-correlation (COR) extracts high-precision illumination parameters (wave vector, initial phase, modulation depth) through sub-pixel optimization in the form of real-space phase gradient [Gustafsson, MGL et al. Three-Dimensional Resolution Doubling in Wide-Field Fluorescence Microscopy by Structured Illumination. Biophys. J. 94, 4957–4970 (2008).]. However, the performance of COR is affected by the low signal-to-noise ratio caused by the harsh imaging conditions required to maintain low photobleaching and phototoxicity for long-term imaging of living cells. In addition, the complex and time-consuming iterative optimization process greatly reduces the image reconstruction efficiency of SIM, posing a severe challenge to fast, real-time, long-term, and artifact-free dynamic super-resolution imaging of living cells. Summary of the Invention

[0004] The purpose of the present invention is to provide a structured light illumination microscopy imaging method based on principal component analysis, which provides a fast, real-time, flexible, convenient and low-light-damage observation method for studying the nanoscale subcellular structural characteristics, movement state, interaction and protein function in living cells.

[0005] The technical solution to achieve the purpose of the present invention is: a structured light illumination microscopy imaging method based on principal component analysis, the specific steps are:

[0006] Step 1: Collect the original illumination image of the sample required for structured light super-resolution;

[0007] Step 2: Separate the three spectrum information of the sample;

[0008] Step 3: Extract the central energy of the level 1 spectrum through a double-window frequency domain masking operator;

[0009] Step 4: Get the lighting vector factor;

[0010] Step 5: Extract the principal components of the lighting vector factors;

[0011] Step 6: Accurately estimate the lighting parameters from the principal components of the lighting vector factors;

[0012] Step 7: Perform spectrum separation for precise separation and super-resolution image reconstruction.

[0013] Preferably, the specific method for collecting the original illumination image of the sample required for structured light illumination super-resolution is:

[0014] The three-step phase-shifted sinusoidal illumination images of the sample in three different illumination directions are collected by the structured light illumination imaging system.

[0015] Preferably, the three-step phase-shifted sinusoidal illumination image in any direction is specifically:

[0016]

[0017] Where D represents the captured illumination image, the subscript n represents the number of the three-step phase-shifted image, n = 1 or 2 or 3, and r represents the image space coordinate. represents the convolution operation, S is the sample information, P represents the point spread function of the imaging system, k ex 、 and m are the wave vector, initial phase and modulation depth of the illumination parameters respectively, Indicates the number of phase shift steps.

[0018] Preferably, the three spectrum information of the sample includes level 0 and ±1 spectrum information, and the specific method of separating the three spectrum information of the sample is:

[0019] Step 2.1: Perform Fourier transform on the illumination image obtained in step 1. The transformed spectrum image is represented as:

[0020]

[0021] The superscript ~ represents the Fourier transform of the original object, k represents the frequency coordinate, and the subscripts 0 and ±1 represent the order of the separated sample spectrum. represents the spectrum of the illumination image D, They represent the 0-level spectrum, +1-level spectrum and -1-level spectrum of the sample S without optical transfer function and illumination parameter components, O represents the optical transfer function of the system, j represents the imaginary number symbol, e is the natural base, k ex 、 are the wave vector and initial phase of the illumination parameters respectively;

[0022] Step 2.2: Linearly combine the spectrum images obtained in step 2.1 and preliminarily separate the 0-level and ±1-level spectrum information of the sample:

[0023]

[0024] make C0, C1, C -1 They are the 0-level and ±1-level spectrum information of the initially separated samples respectively.

[0025] Preferably, step 3 extracts the central energy of the level 1 spectrum by a double-window frequency domain masking operator, specifically:

[0026] Step 3.1: Use the integer pixel wave vector to move the sample +1 level spectrum to the center of the image. The moved sample +1 level spectrum information is expressed as:

[0027]

[0028] Where k int represents the wave vector of the integer pixel, k sub represents the residual sub-pixel wave vector;

[0029] Step 3.2: Use the double-window frequency domain mask operator to extract the center spectrum signal of the +1-level spectrum after being shifted by integer pixels. The extracted center energy of the +1-level spectrum is expressed as:

[0030]

[0031] In the formula, NaN represents an invalid point, k x，min Indicates the left or lower boundary of the signal window in the horizontal x direction in the double-window frequency domain mask operator, k y，min The left or lower boundary of the signal window in the vertical y direction in the double-window frequency domain mask operator, k x，max Indicates the right or upper boundary of the signal window in the horizontal x direction in the double-window frequency domain mask operator, k y，max Indicates the right or upper boundary of the signal window in the double-window frequency domain mask operator in the vertical y direction, R indicates the size of the blank window in the double-window frequency domain mask operator in the horizontal and vertical directions, k x 、k y Represents the frequency coordinate along the horizontal x or vertical y direction.

[0032] Preferably, the specific method of obtaining the lighting vector factor in step 4 is:

[0033] Perform inverse Fourier transform on the sample +1 level center spectrum information and obtain its e exponential term:

[0034]

[0035] Where exp represents the exponential function with e as the base, angle represents the function that returns the phase, Represents an inverse Fourier transform operation.

[0036] Preferably, the specific method of extracting the principal component of the lighting vector factor in step 5 is:

[0037] Step 5.1: Perform singular value decomposition on the lighting vector factor. The decomposed lighting vector factor is expressed as:

[0038]

[0039] Where U and V represent the left and right singular matrices of the illumination vector factor, respectively, the superscript T represents the transpose of the matrix, and Λ represents the eigenvalue of the illumination vector factor;

[0040] Step 5.2: Extract the principal components of the lighting vector factors. The extracted principal components are expressed as:

[0041]

[0042] Where, represents a matrix with the first row and first column being 1 and the other elements being 0, k x，min Indicates the left or lower boundary of the signal window in the horizontal x direction in the double-window frequency domain mask operator, k y，min The left or lower boundary of the signal window in the vertical y direction in the double-window frequency domain mask operator, k x，max Indicates the right or upper boundary of the signal window in the horizontal x direction in the double-window frequency domain mask operator, k y，max It represents the right or upper boundary of the signal window in the vertical y direction in the dual-window frequency domain mask operator, and R represents the size of the blank window in the horizontal and vertical directions in the dual-window frequency domain mask operator.

[0043] Preferably, the specific steps of step 6 of estimating the lighting parameters from the principal components of the lighting vector factors are:

[0044] Step 6.1: Phase unwrap the first row of the left singular matrix U and perform a linear fit using the least squares method. The slope of the fit is the residual sub-pixel wave vector k. sub Component in the horizontal direction; Phase unwrap the first column elements of the right singular matrix V, and perform linear fitting by the least squares method. The slope of the fitting is the residual sub-pixel wave vector k sub Component in the vertical direction;

[0045] Step 6.2: Take the principal component of the illumination vector factor U(Λ·M1)V TGet the initial phase at frequency 0

[0046] Step 6.3: Sub-pixel wave vector k obtained in step 6.1 sub The sample +1-level spectrum is sub-pixel shifted, and the modulation index m is obtained by complex linear regression between the deconvolved +1-level spectrum and the 0-level spectrum.

[0047] Preferably, the illumination parameters obtained in step 6 are used to separate and reconstruct the 0-level and ±1-level spectrum information of the sample preliminarily separated in step 2, and the same operation is performed on the other two illumination directions, and a real-time super-resolution image is reconstructed by Wiener deconvolution, specifically:

[0048]

[0049] Where, the subscript i represents the spectral order, i is 0, +1, or -1, the subscript d represents the number of illumination components, and k d，ex represents the wave vector of the d-th illumination direction, represents the i-th level spectrum component of the d-th illumination direction without optical transfer function and illumination parameter components, and w represents the Wiener constant.

[0050] Compared with the existing technology, the present invention has the following significant advantages: for the first time, the present invention realizes the rapid adaptive and precise compensation of illumination parameters and real-time, high-quality dynamic super-resolution imaging of the fine structure of living cells in a complex, low signal-to-noise ratio experimental environment with external interference, providing a fast, real-time, flexible, convenient, low-light-damage observation method for studying the nanoscale subcellular structural characteristics, movement state, interaction and protein function in living cells, which is expected to promote the discovery of new biological phenomena, open up new possibilities for solving problems in cell biology, cancer research, developmental biology and neuroscience, and has great significance for related fields of life sciences.

[0051] The present invention is further described in detail below with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] Figure 1 Flowchart of the present invention.

[0053] Figure 2The experimental results of HeLa cells under different signal-to-noise ratios using the present invention and the traditional method. The cell nucleus and α-actin are labeled with DAPI and FITC, respectively. a Wide-field image and super-resolution image obtained by the present invention. The original image was acquired through a 60X1.42NA objective lens. The α-actin and cell nucleus of HeLa cells were excited by 488nm and 405nm lasers, respectively. bd Wide-field images of the boxed area in Figure a under different signal-to-noise ratios and super-resolution images obtained by the traditional method and the present invention. e Wide-field images under different signal-to-noise ratios. f Structural similarity index measure (SSIM) of the super-resolution results of the traditional method and the present invention at different signal-to-noise ratios relative to the results in Figure b. The experiment was repeated 30 times independently.

[0054] Figure 3 The results of real-time super-resolution reconstruction of COS-7 cell mitochondria at different time points using the present invention. TM Green FM marker. a, d Widefield images and super-resolution images obtained using the present invention. Original images acquired through a 60X 1.42NA objective. b, c Super-resolution images of the area within the blue box in Figure a at different time points. e, f Super-resolution images of the area within the blue and white boxes in Figure d at different time points. DETAILED DESCRIPTION

[0055] A structured light illumination microscopy imaging method based on principal component analysis uses a "dimensionality reduction" tool based on principal component analysis to remove noisy and interfering components in actual imaging that are not related to the illumination phasor, so as to extract the "first principal component" dominated by the illumination parameters, thereby fundamentally eliminating the interference items that affect the accurate estimation of the illumination parameters, thereby accurately identifying the wave vector and initial phase with sub-pixel accuracy in a simple and efficient manner. In addition, a dual-window frequency domain mask operator is used to further suppress interference noise, while reducing the amount of data involved in principal component analysis by nearly a thousand times, greatly improving the computational efficiency, accuracy and stability of illumination parameter estimation. The flowchart of the invented method is as follows: Figure 1 The specific steps are as follows:

[0056] Step 1: Use a structured illumination microscopy (SIM) system to collect three-step phase-shifted sinusoidal illumination images of the sample in three different illumination directions, where the illumination image in a certain illumination direction is represented as:

[0057]

[0058] Where D represents the collected illumination image, the subscript n represents the number of the three-step phase-shifted image, n = 1 or 2 or 3, and r represents the image space coordinate. represents the convolution operation, S is the sample information, P represents the point spread function of the imaging system, k ex 、 and m are the wave vector, initial phase and modulation depth of the illumination parameters respectively, Indicates the number of phase shift steps.

[0059] Step 2: Preliminarily separate the 0-level and ±1-level spectrum information of the sample from the illumination image in step 1. The specific steps are:

[0060] Step 2.1: Perform Fourier transform on the illumination image obtained in step 1. The transformed spectrum image is represented as:

[0061]

[0062] The superscript ~ represents the Fourier transform of the original object, k represents the frequency coordinate, and the subscripts 0 and ±1 are the order of the separated sample spectrum. represents the spectrum of the illumination image D, They represent the 0-level spectrum, +1-level spectrum and -1-level spectrum of the sample S without optical transfer function and illumination parameter components, respectively. O represents the optical transfer function of the system, j represents the imaginary number symbol, and e is the natural base.

[0063] Step 2.2: Linearly combine the spectrum images obtained in step 2.1 and preliminarily separate the 0-level and ±1-level spectrum information of the sample:

[0064]

[0065] Rewrite the three elements on the left side of the above equation as C0, C1, C -1 That is the initially separated 0-level and ±1-level spectrum information of the sample.

[0066] Step 3: Use a double-window frequency domain mask operator to extract the central spectrum information of the sample +1 level spectrum, specifically:

[0067] Step 3.1: Use the integer pixel wave vector to move the sample +1 level spectrum obtained in step 2 to the center of the image. The sample +1 level spectrum information after the movement can be expressed as:

[0068]

[0069] Where k int represents the wave vector of the integer pixel, k sub represents the residual sub-pixel wave vector.

[0070] Step 3.2: Use a double-window frequency domain mask operator to extract the central spectrum signal of the +1-level spectrum after being shifted by integer pixels. The processed sample +1-level spectrum can be expressed as:

[0071]

[0072] Where NaN represents an invalid point, k x，min 、k y，min Indicates the left or lower boundary of the signal window in the horizontal x and vertical y directions in the double-window frequency domain mask operator, k x，max 、k y，max Indicates the right or upper boundary of the signal window in the double-window frequency domain mask operator in the horizontal x and vertical y directions, R indicates the size of the blank window in the double-window frequency domain mask operator in the horizontal and vertical directions, k x 、k y represents the frequency coordinates along the horizontal x and vertical y directions. . The size of the signal window and the blank window, i.e. k x，min 、k y，min 、k x，max 、k y，max The value of R is determined by the formula N 2 P / (k x(y)，max -k x(y)，min +2R+1) 2 The value at which the maximum value is obtained (where N represents the size of the complete spectrum image, and P represents the ratio between the spectrum amplitude in the signal window and the total spectrum amplitude). Step 3 removes interference signals unrelated to the lighting parameters to a certain extent and reduces the amount of data to be processed subsequently by nearly a thousand times, ensuring efficient and high-precision lighting parameter estimation.

[0073] Step 4: Obtain the illumination vector factor using the center spectrum information of the sample +1 level spectrum in step 3, specifically:

[0074] Perform inverse Fourier transform on the sample +1 level center spectrum information obtained in step 3 and obtain its e exponential term:

[0075]

[0076] Where exp represents the exponential function with e as the base, e is the natural base, and angle represents the function that returns the phase. represents the inverse Fourier transform operation. Equation (6) is the obtained illumination vector factor.

[0077] Step 5: Extract the principal components of the lighting vector factors obtained in step 4, specifically:

[0078] Step 5.1: Perform singular value decomposition on the lighting vector factor obtained in step 4. The decomposed lighting vector factor can be expressed as:

[0079]

[0080] Where U and V represent the left and right singular matrices of the illumination vector factor, respectively, the superscript T represents the transpose of the matrix, and Λ represents the eigenvalue of the illumination vector factor.

[0081] Step 5.2: Extract the principal components of the lighting vector factors. The extracted principal components can be expressed as:

[0082]

[0083] In the formula Represents a matrix with the first row and first column element being 1 and the rest being 0. Step 5 removes the noisy and interfering components in the actual imaging that are not related to the illumination phasor, fundamentally eliminating the interference terms that affect the accurate estimation of the illumination parameters and giving the parameter estimation strong noise resistance.

[0084] Step 6: Accurately estimate the lighting parameters based on the principal components of the lighting vector factors obtained in step 5, specifically:

[0085] Step 6.1: Phase unwrap the first row of the left singular matrix U obtained in step 5, and then perform a linear fit using the least squares method. The slope of the fit is the residual sub-pixel wave vector k sub Component in the horizontal direction; similarly, the first column element of the right singular matrix V is phase-unwrapped, and then a linear fit is performed using the least squares method. The slope of the fit is the residual sub-pixel wave vector k sub Component in the vertical direction.

[0086] Step 6.2: Take the principal component of the illumination vector factor U(Λ·M1)V T The initial phase can be obtained by the phase at frequency 0

[0087] Step 6.3: Sub-pixel wave vector k obtained in step 6.1 sub Perform sub-pixel shift on the sample +1-level spectrum, and then obtain the modulation index m by performing complex linear regression between the deconvolved +1-level spectrum and the 0-level spectrum:

[0088]

[0089] The superscript * represents the conjugate of the original object. The sub-pixel wave vector k obtained in steps 6.1, 6.2, and 6.3 is sub , initial phase The non-iterative nature of step 6 gives the parameter estimation an efficient processing speed.

[0090] Step 7: Use the illumination parameters obtained in step 6 to accurately separate and reconstruct the sample's 0-level and ±1-level spectrum information initially separated in step 2. Perform the same operation for the other two illumination directions, and finally reconstruct a real-time super-resolution image through Wiener deconvolution:

[0091]

[0092] Where the subscript i represents the spectral order, i is 0, +1 or -1, the subscript d represents the number of illumination components, and k d，ex represents the wave vector of the d-th illumination direction, represents the i-th order spectral component of the d-th illumination direction without optical transfer function and illumination parameter components, and w represents the Wiener constant (usually determined by experience).

[0093] The high precision, non-iterative reconstruction, robust noise resistance and limited computational complexity make this invention a promising method for fast, long-term, artifact-free super-resolution imaging of living cells. It also realizes for the first time the rapid adaptive and precise compensation of illumination parameters and the real-time, high-quality dynamic super-resolution imaging of the fine structures of living cells in complex, low signal-to-noise ratio experimental environments with external interference.

[0094] Example

[0095] To test the feasibility and real-time performance of the present invention, the method of the present invention (structured illumination microscopy based on principal component analysis, PCA-SIM) was first used to perform super-resolution reconstruction on fixed HeLa cell samples at different signal-to-noise ratios. Figure 2 It can be seen that with the decrease of signal-to-noise ratio or the increase of Gaussian noise power (unit: dBW), the super-resolution image obtained by the traditional iterative cross-correlation parameter estimation method (COR) shows obvious structural distortion and contrast imbalance ( Figure 2 a-2e). In contrast, the method of the present invention effectively suppresses these image artifacts and reconstructs high-quality super-resolution results comparable to those under conventional signal-to-noise ratios. Quantitative data under different signal-to-noise ratios show that the present invention exhibits strong robustness under noisy noise conditions ( Figure 2 f). In terms of speed, the present invention's computational speed is nearly 20 times faster than the traditional cross-correlation method. This example demonstrates the present invention's combined advantages in accuracy, efficiency, and noise immunity, as well as its potential for long-term live cell imaging under weak excitation conditions.

[0096] An important application of the present invention is to observe the dynamic process of subcellular structures in real time under complex or harsh operating environments. To verify this, the present invention was used to observe the mitochondrial dynamic tubulation (MDT) events in living COS-7 cells in real time. Figure 3 a-3b, The elongated and highly dynamic MDT tubules that rapidly extend from mitochondria fuse with other mitochondria and form membrane bridges between them, during which the transfer of certain substances occurs ( Figure 3 b) The MDT tubules gradually thicken and stabilize, eventually forming a mitochondrial network. These experimental phenomena are of great significance for studying the formation of the mitochondrial network and the maintenance of mitochondrial functions (such as mitochondrial DNA integrity and cell apoptosis). This example fully demonstrates the advantages of the present invention in exploring the structure and dynamic processes of living cells.

Claims

1. A structured light illumination microscopy imaging method based on principal component analysis, characterized in that: The specific steps are: Step 1: Collect the original illumination image of the sample required for structured light super-resolution; Step 2: Separate the three spectrum information of the sample; Step 3: Extract the central energy of the level 1 spectrum through the double-window frequency domain masking operator, specifically: Step 3.1: Use the integer pixel wave vector to move the sample + level 1 spectrum to the center of the image. The moved sample + level 1 spectrum information is expressed as: Where k int represents the wave vector of the integer pixel, k sub represents the residual sub-pixel wave vector; Step 3.2: Use the double-window frequency domain mask operator to extract the central spectrum signal of the +1-level spectrum after the integer pixel shift. The extracted +1-level spectrum center energy is expressed as: In the formula, NaN represents an invalid point, k x,min Indicates the left or lower boundary of the signal window in the horizontal x direction in the double-window frequency domain mask operator, k y,min The left or lower boundary of the signal window in the vertical y direction in the double-window frequency domain mask operator, k x,max Indicates the right or upper boundary of the signal window in the horizontal x direction in the double-window frequency domain mask operator, k y,max Indicates the right or upper boundary of the signal window in the double-window frequency domain mask operator in the vertical y direction, R indicates the size of the blank window in the double-window frequency domain mask operator in the horizontal and vertical directions, k x 、k y Represents the frequency coordinate along the horizontal x or vertical y direction; Step 4: Get the lighting vector factor. The specific method is: Perform inverse Fourier transform on the sample + 1 level center spectrum information and obtain its e exponential term: Where exp represents the exponential function with e as the base, angle represents the function that returns the phase, represents the inverse Fourier transform operation; Step 5: Extract the principal components of the lighting vector factors. The specific method is: Step 5.1: Perform singular value decomposition on the lighting vector factor. The decomposed lighting vector factor is expressed as: Where U and V represent the left and right singular matrices of the illumination vector factor, respectively, the superscript T represents the transpose of the matrix, and Λ represents the eigenvalue of the illumination vector factor; Step 5.2: Extract the principal components of the lighting vector factors. The extracted principal components are expressed as: Where, represents a matrix with the first row and first column being 1 and the other elements being 0, k x,min Indicates the left or lower boundary of the signal window in the horizontal x direction in the double-window frequency domain mask operator, k y,min The left or lower boundary of the signal window in the vertical y direction in the double-window frequency domain mask operator, k x,max Indicates the right or upper boundary of the signal window in the horizontal x direction in the double-window frequency domain mask operator, k y,max Indicates the right or upper boundary of the signal window in the dual-window frequency domain mask operator in the vertical y direction, and R indicates the size of the blank window in the dual-window frequency domain mask operator in the horizontal and vertical directions; Step 6: Accurately estimate the lighting parameters from the principal components of the lighting vector factors; Step 7: Perform spectrum separation for precise separation and super-resolution image reconstruction.

2. The structured light illumination microscopy imaging method based on principal component analysis according to claim 1, characterized in that: The specific method for collecting the original illumination image of the sample required for structured light super-resolution is: The three-step phase-shifted sinusoidal illumination images of the sample in three different illumination directions are collected by a structured light illumination imaging system.

3. The structured light illumination microscopy imaging method based on principal component analysis according to claim 2, characterized in that: The three-step phase-shifted sinusoidal illumination image in any direction is specifically: Where D represents the captured illumination image, the subscript n represents the number of the three-step phase-shifted image, n = 1 or 2 or 3, and r represents the image space coordinate. represents the convolution operation, S is the sample information, P represents the point spread function of the imaging system, k ex 、 and m are the wave vector, initial phase and modulation depth of the illumination parameters respectively, Indicates the number of phase shift steps.

4. The structured light illumination microscopy imaging method based on principal component analysis according to claim 1, characterized in that: The three spectrum information of the sample includes the 0-level and ±1-level spectrum information. The specific method of separating the three spectrum information of the sample is: Step 2.1: Perform Fourier transform on the illumination image obtained in step 1. The transformed spectrum image is represented as: The superscript ~ represents the Fourier transform of the original object, k represents the frequency coordinate, and the subscripts 0 and ±1 represent the order of the separated sample spectrum. represents the spectrum of the illumination image D, They represent the 0-level spectrum, +1-level spectrum and -1-level spectrum of the sample S without optical transfer function and illumination parameter components, O represents the optical transfer function of the system, j represents the imaginary number symbol, e is the natural base, k ex 、 are the wave vector and initial phase of the illumination parameters respectively; Step 2.2: Linearly combine the spectrum images obtained in step 2.1 and preliminarily separate the 0-level and ±1-level spectrum information of the sample: make C0, C1, C -1 They are the 0-level and ±1-level spectrum information of the initially separated samples respectively.

5. The structured light illumination microscopy imaging method based on principal component analysis according to claim 1, characterized in that: The specific steps of step 6 to estimate the lighting parameters from the principal components of the lighting vector factors are: Step 6.1: Phase unwrap the first row of the left singular matrix U and perform a linear fit using the least squares method. The slope of the fit is the residual sub-pixel wave vector k. sub Component in the horizontal direction; Phase unwrap the first column elements of the right singular matrix V, and perform linear fitting by the least squares method. The slope of the fitting is the residual sub-pixel wave vector k sub Component in the vertical direction; Step 6.2: Take the principal component of the illumination vector factor U(Λ·M1)V T Get the initial phase at frequency 0 Step 6.3: Sub-pixel wave vector k obtained in step 6.1 sub The sample +1-level spectrum is sub-pixel shifted, and the modulation index m is obtained by complex linear regression between the deconvolved +1-level spectrum and the 0-level spectrum.

6. The structured light illumination microscopy imaging method based on principal component analysis according to claim 1, characterized in that: The illumination parameters obtained in step 6 are used to separate and reconstruct the 0-level and ±1-level spectrum information of the sample initially separated in step 2. The same operation is performed on the other two illumination directions, and a real-time super-resolution image is reconstructed through Wiener deconvolution. Specifically: Where, the subscript i represents the spectral order, i is 0, +1, or -1, the subscript d represents the number of illumination components, and k d,ex represents the wave vector of the d-th illumination direction, represents the i-th level spectrum component of the d-th illumination direction without optical transfer function and illumination parameter components, and w represents the Wiener constant.

Images