A statistical gating method for multiple scattering components in frequency domain OCT and extraction of tissue optical parameters

Abstract

This invention relates to a method for statistical gating of multiple scattering components and extraction of tissue optical parameters using frequency-domain OCT. The method includes: performing multiple consecutive samplings at a single sampling location within an imaging region using frequency-domain OCT; constructing a matrix using the sampled k-space data and calculating its covariance matrix as a random matrix description; modeling this description as a low-rank perturbation of a Wishart random matrix; statistically gating the ballistic component and multiple scattering components based on the statistical characteristics of the low-rank perturbation of the Wishart random matrix, reconstructing the ballistic signal matrix and the multiple scattering component matrix respectively; and calculating the tissue absorption coefficient and tissue anisotropy parameters using inverse Fourier transform and the extended Huygens-Fresnel formula, based on the ballistic component information in the ballistic signal matrix and the multiple scattering component information in the multiple scattering component matrix. Compared with existing technologies, this invention can extract multiple scattering components in frequency-domain OCT and obtain the tissue absorption coefficient and tissue anisotropy parameters.

Description

Technical Field

[0001] This invention relates to the field of optical coherence tomography (OCT) technology, and in particular to a method for statistical gating of multiple scattering components in frequency domain OCT and extraction of tissue optical parameters. Background Technology

[0002] Optical coherence tomography (OCT) is a novel high-resolution tomographic imaging modality. Its working principle is primarily based on interferometric spectroscopy and fast Fourier transform (FFT). It obtains longitudinal depth information by measuring the spectrum of the interference signal and performing a FFT on it. Frequency-domain OCT collects beams representing different depths of the sample and then calculates the intensity information of the sample depth through an inverse Fourier transform, thus achieving parallel acquisition of the sample's axial information.

[0003] To ensure the imaging quality of tissue structures, it is necessary to extract and reconstruct the scattering signals in frequency-domain OCT. Existing technologies, such as the Chinese patent ZL202110867339.1, propose a method for extracting ballistic scattering signals from visible light OCT. This method establishes a random matrix description in k-space and uses the maximum eigenvalue of the Wishart random matrix to reconstruct the ballistic scattering components. Although this scheme can extract the ballistic scattering components, due to the interaction effect between the ballistic scattering components and multiple scattering components, it cannot accurately extract single scattering classifications or multiple scattering components. Furthermore, this scheme cannot obtain tissue functional information (such as tissue absorption coefficients and tissue anisotropy parameters). Summary of the Invention

[0004] The purpose of this invention is to overcome the shortcomings of the prior art by providing a method for statistical gating of multiple scattering components in frequency domain OCT and extraction of tissue optical parameters. This method can extract multiple scattering components in frequency domain OCT and further calculate and obtain the tissue absorption coefficient and tissue anisotropy parameters using the multiple scattering components.

[0005] The objective of this invention can be achieved through the following technical solution: a method for statistical gating of multiple scattering components in frequency domain OCT and extraction of tissue optical parameters, comprising the following steps:

[0006] S1. For a single sampling position within the imaging area, use frequency domain OCT to continuously sample multiple times. Use the k-space data obtained from the sampling to construct a matrix and calculate its covariance matrix as a description of the random matrix. Model this description as a low-rank perturbation of the Wishart random matrix.

[0007] S2. Based on the statistical characteristics of the low-rank perturbation of the Wishart random matrix, perform statistical gating on the ballistic component and the multiple scattering component, and reconstruct the ballistic signal matrix and the multiple scattering component matrix respectively;

[0008] S3. According to the ballistic component information in the ballistic signal matrix and the multiple scattering component information in the multiple scattering component matrix, through Fourier inverse transform, and combined with the extended Huygens-Fresnel formula, calculate the tissue absorption coefficient and the tissue anisotropy parameter.

[0009] Further, the specific process of step S1 includes: For a single sampling position in the x-y plane of the imaging area, that is, a single A-line, perform L samplings. Each sampling obtains the k-space data of a single A-line, with a total of M data points. Then construct a matrix K of M rows and L columns, K = K L , L ,

[0014] , L +K M , each column of the matrix K is the k-space data of a single A-line obtained by one measurement, K B is the ballistic component, which is a low-rank matrix of M rows and L columns, K M is the multiple scattering component, which is a full-rank matrix of M rows and L columns;

[0010] The optimal parameter c is achieved by controlling the number of samplings. Where c = L / M, 0 < c < ∞, that is, the aspect ratio of the matrix. Construct a matrix from the k-space data and find its covariance matrix to describe and model the low-rank perturbation of the Wishart random matrix, that is, obtain The symbol represents the conjugate transpose of the matrix.

[0011] Further, the low-rank perturbation C of the Wishart random matrix is a matrix of L rows and L columns, C = C L +C M , where C L is a low-rank matrix of L rows and L columns,

[0012] C M is a full-rank Wishart random matrix of L rows and L columns,

[0013] Further, the process of reconstructing the ballistic signal matrix in step S2 includes:

[0014] Based on the low-rank perturbation of the Wishart random matrix, use the L eigenvalues {s1, s2,..., S L} of the matrix C to solve the quadratic equation to obtain {θ1, θ2,..., θ L}, and select from {θ1, θ2,..., θ L} those greater than eigenvalues ​​{θ1, θ2, ..., θ q}, to be used as an estimate of the characteristic values ​​of the ballistic components;

[0015] For eigenvalues ​​{θ1, θ2, ..., θ q Taking the square root yields singular values. And construct matrix S using the main diagonal elements. B Singular value decomposition of the K matrix Reconstructing the ballistic component matrix

[0016] Furthermore, the quadratic equation in one variable is: θ 2 +(σ 2 -σ 2 c-λ)θ+σ 2 When cs = 0, take the positive root from the two roots.

[0017] Furthermore, the process of reconstructing the multiple scattering component matrix in step S2 includes:

[0018] Based on the Tracy-Widom theorem, the multiple scattering component K is solved using the minimum eigenvalue. M The variance statistics parameter σ of matrix elements 2 ;

[0019] Due to the multiple scattering component K M The corresponding Wishart random matrix C with Gaussian statistics M C M The eigenvalues ​​of the matrix follow an unnormalized Marchenko-Pastur distribution. Therefore, sampling according to the Marchenko-Pastur statistical distribution yields a complete estimate of the eigenvalues ​​of the multiple scattering components {λ}. M1 , λ M2 ,...,λ ML};

[0020] The complete estimate obtained [λ] M1 , λ M2 ,...,λ ML Taking the square root yields singular values. And construct matrix S using the main diagonal elements. M Reconstructing the multiple scattering component matrix The first column of the matrix is ​​taken as the multiple scattering component in the A-line at that location.

[0021] Furthermore, the variance statistical parameter σ 2 Specifically:

[0022]

[0023] Among them, s L The smallest non-zero eigenvalue of the C matrix obtained from the measured data. <y>The value range is (-1000 to +1000), and it is obtained by numerical calculation based on the Tracy-Widom theorem.

[0024] Furthermore, the Marchenko-Pastur statistical distribution is specifically as follows:

[0025]

[0026] Among them, C M The matrix has L eigenvalues, namely {λ1, λ2, ..., λ3}. L }, where p(λ) is the probability density function of the eigenvalues. and These represent the maximum and minimum values ​​of the range that the non-zero eigenvalues ​​can take, respectively.

[0027] Further, step S3 includes the following steps:

[0028] S31. Perform an inverse Fourier transform on the multiple scattering components to obtain the time-domain multiple scattering component A. M (z), taking the square of its modulus yields the multiple scattering intensity information I. M (z)=|A M (z)| 2 , where z is the depth coordinate along the axis;

[0029] S32. Based on the multiple scattering intensity information and the characteristics of the multiple scattering invariant of light in tissue, calculate the tissue absorption coefficient parameter diagram:

[0030]

[0031]

[0032] Among them, I M0 (z) is the proportionality constant, and its value is a real number not less than zero. <s>The physical dimensions of the coherent volume. <s>Depending on the spatial resolution of the OCT system, l z For axial resolution, l x and l y Horizontal resolution;

[0033] S33. Perform an inverse Fourier transform on the ballistic components to obtain the time-domain ballistic component A. B (z) and its corresponding intensity information, B (z)=|A B (z)| 2 Then, the scattering intensity information is calculated:

[0034]

[0035] Among them, I B0 (z) is the proportionality constant, and its value range is a real number not less than zero;

[0036] S34. Perform an inverse Fourier transform directly on the k-space data of the original single A-line to obtain A(z). Take the square of the modulus of A(z) to obtain the original scattering intensity information I(z)=|A(z)|. 2 By utilizing the tissue absorption coefficient μa(z), the influence of absorption effects in the original scattering intensity information is removed, thus obtaining a scattering signal without absorption effects.

[0037] S35, Scattering signal I without absorption effect d (z), fitted using the extended Huygens-Fresnel formula, yields... The anisotropy parameter g(z) of the tissue is then calculated.

[0038] Furthermore, the extended Huygens-Fresnel formula in step S35 is specifically as follows:

[0039]

[0040]

[0041] The specific anisotropy parameters of the tissue are as follows:

[0042]

[0043] Where I0 is the incident light intensity, ω0 is the incident beam waist, and z R Let z be the Rayleigh length of the Gaussian beam. f Let n be the beam focusing depth and n be the imaging medium.

[0044] Compared with the prior art, the present invention has the following advantages:

[0045] This invention targets a single sampling location within an imaging region. It utilizes k-space data from frequency-domain OCT to establish a statistical description of a random matrix. The k-space data is used to construct a matrix, and its covariance matrix is ​​calculated. This matrix is ​​then used as a low-rank perturbation of the Wishart random matrix. Based on this low-rank perturbation, the ballistic signal matrix and the multiple scattering component matrix are reconstructed. Then, using the ballistic component information from the ballistic signal matrix and the multiple scattering component information from the multiple scattering component matrix, the tissue absorption coefficient and tissue anisotropy parameters are calculated through inverse Fourier transform and the extended Huygens-Fresnel formula. Thus, based on the k-space statistical gating method, multiple scattering components are extracted from frequency-domain OCT. Furthermore, the tissue absorption coefficient and tissue anisotropy parameters are calculated using these multiple scattering components. Compared to existing technologies, this invention can accurately extract multiple scattering components, implement statistical gating for each multiple scattering component individually, and calculate the absorption coefficient and tissue anisotropy parameters for each voxel.

[0046] This invention designs a matrix from k-space data and calculates its covariance matrix to model a low-rank perturbation of a Wishart random matrix as a random matrix description. Then, it designs a scheme to reconstruct the ballistic component matrix and the multiple scattering component matrix, which can accurately extract single scattering classification and achieve the purpose of extracting multiple scattering components.

[0047] This invention, based on the ballistic component information in the ballistic signal matrix, the multiple scattering component information in the multiple scattering component matrix, and the original k-space data, uses inverse Fourier transform to obtain multiple scattering intensity information, tissue absorption coefficient parameter map, ballistic component scattering intensity information, original scattering intensity information, and scattering signal without absorption effect, respectively. Then, by fitting the extended Huygens-Fresnel formula, the tissue anisotropy parameters can be solved, thereby ensuring the accurate extraction of tissue optical parameters. Attached Figure Description

[0048] Figure 1 This is a schematic diagram of the method flow of the present invention;

[0049] Figures 2a-2c This is a schematic diagram illustrating the actual application process of the present invention. Detailed Implementation

[0050] The present invention will now be described in detail with reference to the accompanying drawings and specific embodiments.

[0051] like Figure 1 As shown, a method for statistical gating of multiple scattering components in frequency domain OCT and extraction of tissue optical parameters includes the following steps:

[0052] S1. Use frequency-domain OCT to continuously sample a single sampling position within the imaging region multiple times. Construct a matrix using the sampled k-space data and calculate its covariance matrix as a random matrix description. Model this description as a low-rank perturbation of a Wishart random matrix.

[0053] S2. Based on the statistical characteristics of the low-rank perturbation of the Wishart random matrix, perform statistical gating on the ballistic component and the multiple-scattering component, and reconstruct the ballistic signal matrix and the multiple-scattering component matrix respectively.

[0054] S3. According to the ballistic component information in the ballistic signal matrix and the multiple-scattering component information in the multiple-scattering component matrix, through Fourier inverse transform, and combined with the extended Huygens-Fresnel formula, calculate the tissue absorption coefficient and the tissue anisotropy parameter.

[0055] Apply the above scheme to the actual situation, as Figures 2a-2c shown, the specific content includes:

[0056] Step 1. For a single sampling position (a single A-line) within the imaging region (x-y plane), establish a random matrix statistical description using the k-space data of frequency-domain OCT.

[0057] Perform L samplings on the same sampling position. Each sampling obtains the k-space data of a single A-line, and there are M data points in total. Then construct a matrix K with M rows and L columns. By controlling the number of samplings, the optimal parameter c = L / M (0 < c < ∞), that is, the aspect ratio of the matrix, is achieved. Construct a matrix from the k-space data and calculate its covariance matrix to model the random matrix description as a low-rank perturbation of a Wishart random matrix, that is e where *H* represents the conjugate transpose of the matrix. Symbol *H* represents the conjugate transpose of the matrix.

[0058] where K = K B + K M , and each column is the k-space data of a single A-line obtained from one measurement. K B is the ballistic component, which is a low-rank matrix with M rows and L columns; K M is the multiple-scattering component, which is a full-rank matrix with M rows and L columns.

[0059] C is a matrix with L rows and L columns, is a low-rank matrix with L rows and L columns, is a full-rank Wishart random matrix with L rows and L columns.

[0060] It should be noted that in practical applications, the construction of the C matrix is According to the characteristics of matrix operations, the construction of the C matrix can also be carried out in the following way. Then the subsequent steps can be adjusted accordingly.

[0061] In addition, when performing a scan, different schemes can be adopted to construct the k-space spatiotemporal matrix corresponding to a single A-line: (1) Perform multiple scans on the same A-line (this method has a long sampling time, high spatial clarity, and is suitable for high-resolution imaging);

[0062] (2) Adjacent A-lines within the same B-Scan (sampling time does not increase, but some spatial resolution is lost);

[0063] (3) Spatially adjacent A-lines in adjacent B-scans;

[0064] (4) Combining the above sampling methods (1) to (3), such as sampling the same A-line multiple times and also using adjacent A-lines.

[0065] Step 2: Let the L eigenvalues ​​of matrix C be {s1, s2, ..., s...} L Arrange the eigenvalues ​​(from largest to smallest) into the following formula to solve for {θ1, θ2, ..., θ...} L }

[0066] θ 2 +(σ 2 -σ 2 c-λ)θ+σ 2 cs=0

[0067] It should be noted that this is a quadratic equation in one variable, and the positive root is taken from the two roots.

[0068] Step 3: From {θ1, θ2, ..., θ L } Filter out all values ​​greater than The values ​​are obtained to get {θ1, θ2, ..., θ q }, to be used as an estimate of the characteristic values ​​of the ballistic components.

[0069] Step 4: Apply the eigenvalues ​​{θ1, θ2, ..., θ3} obtained in Step 3. q Taking the square root yields the singular value. And construct matrix S using the main diagonal elements. B Singular value decomposition of the K matrix. Reconstructing the ballistic component matrix

[0070] Step 5: Based on the Tracy-Widom theorem, solve for the multiple scattering component K using the minimum eigenvalue. M The variance statistics parameter σ of matrix elements 2 .

[0071]

[0072] Among them, s L The smallest non-zero eigenvalue of the C matrix obtained from the measured data. <y>The value range is (-1000 to +1000), and the numerical calculation is based on the Tracy-Widom theorem.

[0073] Step 6, Multiple scattering component K M The corresponding Wishart random matrix C with Gaussian statistics M C M The statistical distribution of the eigenvalues ​​of the matrix follows an unnormalized Marchenko-Pastur distribution.

[0074]

[0075] Among them, C M The matrix has L eigenvalues, namely {λ1, λ2, ..., λ3}. L p(λ) is the probability density function of the eigenvalues. and These represent the maximum and minimum values ​​that a non-zero eigenvalue can take.

[0076] Step 7: Statistical sampling of the statistical distribution established in Step 6 yields a complete estimate of the eigenvalues ​​of a set of multiple scattering components {λ}. M1 , λ M2 ,...,λ ML }

[0077] It should be noted that in step 7, if the statistical distribution is sampled only once, only one set of eigenvalues ​​for multiple scattering components is obtained, which may result in sampling error. Therefore, in practice, the statistical distribution can be sampled independently multiple times to generate multiple sets of eigenvalue estimates and then the average of these multiple sets can be calculated, thereby significantly reducing sampling error.

[0078] Step 8: Take the square root of the eigenvalues ​​obtained in Step 7 to obtain singular values. And construct matrix S using the main diagonal elements. M Reconstructing the multiple scattering component matrix The first column of the matrix is ​​taken as the multiple scattering component in the A-line at that location.

[0079] Step 9: Perform an inverse Fourier transform on the multiple scattering components to obtain the time-domain multiple scattering component A. M (z), taking the square of its modulus yields the multiple scattering intensity information I. M (z)=|A M (z)| 2 , where z is the depth coordinate along the axis.

[0080] Step 10: Combining the multiple scattering intensity information with the characteristics of the multiple scattering invariant of light in tissue, calculate the tissue absorption coefficient parameter diagram:

[0081]

[0082] Among them, I M0 (z) is the proportionality coefficient, and its value is a real number not less than zero. In practical applications, its value can be obtained through standard product testing or Monte Carlo simulation. (s>) is the physical size of the coherent volume. Depending on the spatial resolution of the OCT system (l z For axial resolution, l x and l y (Horizontal resolution).

[0083] Step 11: Perform an inverse Fourier transform on the ballistic components to obtain the time-domain ballistic component A. B (z), and its corresponding intensity information I B (z)=|A B (z)| 2 Then, the scattering intensity information is calculated:

[0084]

[0085] Among them, I B0 (z) is the proportionality coefficient, and its value is a real number not less than zero. In practical applications, its value can be obtained through standard product testing or Monte Carlo simulation.

[0086] Step 12: Perform an inverse Fourier transform on the original k-space data of a single A-line to obtain A(z). Take the square of the modulus of A(z) to obtain the original scattering intensity information I(z) = |A(z)|. 2 Using the tissue absorption coefficient μ a (z) Removes the influence of absorption effect from the original scattering intensity information to obtain a scattering signal without absorption effect.

[0087] Step 13: Scattering signal I without absorption effect d (z), fitted using the following extended Huygens-Fresnel formula, yields...

[0088]

[0089] Where I0 is the incident light intensity. ω0 is the waist of the incident beam, z R Let z be the Rayleigh length of the Gaussian beam. f Let n be the beam focusing depth and n be the imaging medium.

[0090] Step 14: Solve for the anisotropy parameter g(z) of the microstructure:

[0091]

[0092] Step 15: Repeat steps 1 to 14 for sampling positions along the x or y direction within the imaging area to achieve B-scan imaging;

[0093] By repeating steps 1 to 14 at all sampling locations in the imaging area, three-dimensional imaging can be achieved.

[0094] Example 1

[0095] This embodiment focuses on spectral domain OCT, applying the aforementioned technical solution. Spectral domain OCT utilizes a broadband near-infrared light source. In a B-scan (512 positions along the x-direction), L = 16 measurements are performed at each position. After completing the B-scan, each set of A-line data (16 measurements) is used as a column vector to construct matrix K. Steps 1-14 are then applied to calculate the multiple scattering components |A| along the A-line. M (z)|, tissue absorption coefficient μ a (z) and tissue anisotropy parameter g(z). After processing all 512 locations, the multiple scattering component map, tissue absorption coefficient map, and tissue anisotropy parameter map corresponding to this B-san can be constructed.

[0096] Example 2

[0097] This embodiment focuses on visible light OCT, applying the aforementioned technical solution. Visible light OCT uses a broadband or supercontinuum light source in the visible light band. A B-scan (512 positions along the x-direction) is completed. For each A-line, matrix K is constructed using its own data and the data from 16 adjacent A-lines (a total of 17 A-line data) as column vectors. Steps 1-8 are used to calculate the spatial data of the multiple scattering components k along the A-line. In step 9, a short-time Fourier inverse transform with a Gaussian window is used to obtain the A-wavelengths corresponding to the 12 center wavelengths. M (z). A for each wavelength M (z) All samples undergo the processing described in step 10. The processing results are then fitted with the absorption spectra of oxygenated and hypoxic hemoglobin to determine the ratio between the two types of hemoglobin and to calculate blood oxygen saturation. Once all 512 positions have been processed, the blood oxygen saturation map corresponding to that B-san can be constructed.

[0098] In summary, compared with existing technologies, which have errors in extracting ballistic components and cannot accurately extract multiple scattering components, this solution can achieve statistical gating of multiple scattering components separately.

[0099] In the ballistic components obtained by existing technology, the scattering coefficient and absorption coefficient are coupled together, making it impossible to calculate the absorption coefficient for each voxel. However, this solution can calculate the absorption coefficient for each voxel.

[0100] Existing technologies cannot calculate tissue anisotropy parameters for each voxel, while this solution can calculate tissue anisotropy parameters for each voxel.< / y> < / s> < / s> < / y>

Claims

1. A method for statistical gating of multiple scattering components in frequency domain OCT and extraction of tissue optical parameters, characterized in that, Includes the following steps: S1. For a single sampling position within the imaging area, use frequency domain OCT to continuously sample multiple times. Use the k-space data obtained from the sampling to construct a matrix and calculate its covariance matrix as a description of the random matrix. Model this description as a low-rank perturbation of the Wishart random matrix. S2. Based on the statistical characteristics of the low-rank perturbation of the Wishart random matrix, statistical gating is performed on the ballistic component and the multiple scattering component to reconstruct the ballistic signal matrix and the multiple scattering component matrix, respectively. S3. Based on the ballistic component information in the ballistic signal matrix and the multiple scattering component information in the multiple scattering component matrix, the tissue absorption coefficient and tissue anisotropy parameter are calculated by inverse Fourier transform and combined with the extended Huygens-Fresnel formula. The specific process of step S1 includes: for a single sampling position in the xy plane of the imaging region, i.e., a single A-line, performing... Each sampling session yields k-space data for a single A-line, totaling [number] samples. Data points, and then construct OK Column matrix , ,matrix Each column represents the k-space data of a single A-line obtained from a single measurement. It is a ballistic component, and it is OK A low-rank matrix of columns. For multiple scattering components, it is OK A full-rank matrix with columns; Optimal parameters can be achieved by controlling the number of samplings. ,in, The aspect ratio of the matrix is ​​used to construct a matrix from the k-space data and calculate its covariance matrix. This covariance matrix is ​​then used as a description of the random matrix and modeled as a low-rank perturbation of the Wishart random matrix. , The symbol represents the conjugate transpose of a matrix; The process of reconstructing the ballistic signal matrix in step S2 includes: Based on the low-rank perturbation of the Wishart random matrix, using the matrix of L indivual Solve ,from Filter out those greater than eigenvalues , to be used as an estimate of the characteristic values ​​of the ballistic components; For eigenvalues Square root to obtain singular values And construct a matrix using the main diagonal elements. ,right Singular value decomposition of a matrix Reconstruct the ballistic component matrix; The quadratic equation in one variable is: When solving, take the positive root from the two roots; The process of reconstructing the multiple scattering component matrix in step S2 includes: Based on the Tracy-Widom theorem, the multiple scattering components are solved using the minimum eigenvalue. Variance statistics of matrix elements ; Due to multiple scattering components Corresponding to the Wishart random matrix with Gaussian statistics , The eigenvalues ​​of the matrix follow an unnormalized Marchenko–Pastur distribution. Therefore, sampling according to the Marchenko–Pastur statistical distribution yields a complete estimate of the eigenvalues ​​of the multiple scattering components. ; The complete estimate obtained Square root to obtain singular values And construct a matrix using the main diagonal elements. Reconstructing the multiple scattering component matrix The first column of the matrix is ​​taken as the multiple scattering component in the A-line at that location; Step S3 includes the following steps: S31. Perform an inverse Fourier transform on the multiple scattering components to obtain the time-domain multiple scattering components. The intensity information of multiple scattering is obtained by taking the square of its modulus. ,in The depth coordinates are along the axial direction; S32. Based on the multiple scattering intensity information and the characteristics of the multiple scattering invariant of light in tissue, calculate the tissue absorption coefficient parameter diagram: ， ， in, This is the proportionality constant, and its value range is a real number not less than zero. The physical dimensions of the coherent volume. Depending on the spatial resolution of the OCT system, For axial resolution, and Horizontal resolution; S33. Perform an inverse Fourier transform on the ballistic components to obtain the time-domain ballistic components. and its corresponding intensity information Then, the scattering intensity information is calculated: ， in, This is the proportionality constant, and its value range is a real number not less than zero; S34. Perform an inverse Fourier transform directly on the k-space data of the original single A-line to obtain... ,Pick The square of the modulus yields the original scattering intensity information. Using tissue absorption coefficient The absorption effect is removed from the original scattering intensity information to obtain a scattering signal without absorption effect. ; S35. Scattered signals without absorption effect The extended Huygens-Fresnel formula was used for fitting to obtain... Then, the anisotropy parameters of the tissue are calculated. ; The extended Huygens-Fresnel formula in step S35 is as follows: ， ， The specific anisotropy parameters of the tissue are as follows: ， in, For the incident light intensity, The waist of the incident beam Let Rayleigh be the length of the Gaussian beam. The depth of focus of the beam. It serves as the imaging medium.

2. The method for statistical gating of frequency domain OCT multiple scattering components and extraction of tissue optical parameters according to claim 1, characterized in that, The low-rank perturbation of the Wisart random matrix for OK A matrix of columns, ,in, for OK A low-rank matrix of columns. ; for OK A full-rank Wishart random matrix of columns. .

3. The method for statistical gating of frequency domain OCT multiple scattering components and extraction of tissue optical parameters according to claim 1, characterized in that, The variance statistics parameter Specifically: ， in, Obtained from measured data The smallest non-zero eigenvalue of a matrix. The value range is (-1000~+1000), which is obtained by numerical calculation based on the Tracy-Widom theorem.

4. The method for statistical gating of frequency domain OCT multiple scattering components and extraction of tissue optical parameters according to claim 3, characterized in that, The Marchenko–Pastur statistical distribution is specifically as follows: ， in, The matrix has Each eigenvalue, i.e. , The probability density function of the eigenvalues. and These represent the maximum and minimum values ​​of the range that the non-zero eigenvalues ​​can take, respectively.

Images