OCT (Optical Coherence Tomography) measurement imaging method based on characteristic value calculation and multi-signal classification spectrum estimation

Abstract

The invention discloses an OCT (Optical Coherence Tomography) measurement imaging method based on characteristic value calculation and multi-signal classification spectrum estimation. The method comprises the following steps: firstly, performing frequency spectrum scanning on a sample through frequency domain OCT; carrying out photoelectric conversion and multi-point sampling on the obtained frequency domain interference fringes by utilizing a photoelectric conversion device; after interference spectrum signals are obtained, autocorrelation matrix estimation and eigenvalue calculation are carried out, and the sizes of signal subspaces in eigenvalues are obtained through analysis; performing multi-signal classification power spectrum estimation on the autocorrelation matrix to obtain a power spectrum carrying spatial information, thereby obtaining OCT longitudinal intensity distribution through logarithmic transformation and calibration; and finally, realizing OCT image reconstruction through transverse scanning. According to the method, the resolution limitation of OCT longitudinal intensity distribution obtained through traditional discrete Fourier transform is broken through, and the OCT image with the higher resolution can be generated.

Description

Technical Field

[0001] The present invention is designed to be a technology in the field of optical measurement, specifically a measurement method of optical coherence tomography (OCT); more specifically, a measurement imaging method of optical coherence tomography based on eigenvalue calculation and multi-signal classification spectrum estimation. Background Art

[0002] Optical coherence tomography technology uses the principle of low-coherence interference and a low-coherence light source to obtain the reflectivity information of the sample to be tested at different depths, and then obtains a two-dimensional or three-dimensional structural image of the sample to be tested by scanning. This method does not cause damage to the sample to be tested, and has the characteristics of high resolution and non-invasiveness. In the existing OCT theory, the reconstruction method from the interference spectrum to the image is based on the inverse discrete Fourier transform (IDFT). The axial resolution of the image produced by this method is approximately equal to the coherence length of the light source and is inversely proportional to the bandwidth of the light source. In recent years, the main way to improve the resolution of OCT is to increase the bandwidth of the system and select a shorter wavelength light source. Both directions face problems, such as dispersion introduced by large bandwidth and high loss and low penetration caused by short wavelength. Summary of the invention

[0003] In view of the problems and defects existing in the existing IDFT reconstruction method of OCT, the present invention proposes an OCT measurement imaging method based on eigenvalue calculation and multi-signal classification spectrum estimation, which breaks the limitation of traditional discrete Fourier transform on axial resolution and obtains a reconstructed image with higher resolution.

[0004] The present invention is achieved through the following technical solutions:

[0005] The present invention relates to an OCT measurement imaging method based on eigenvalue calculation and multi-signal classification spectrum estimation. Firstly, a frequency domain OCT is used to perform spectrum scanning on a sample to obtain frequency domain interference fringes. Then, a photoelectric conversion device is used to perform photoelectric conversion and multi-point sampling on the obtained frequency domain interference fringes to obtain an interference spectrum. Eigenvalue calculation and multi-signal classification spectrum estimation are performed on the interference spectrum to obtain the OCT longitudinal intensity distribution. Finally, a frequency domain OCT system is used for lateral scanning to obtain multiple OCT longitudinal intensity distributions, thereby realizing two-dimensional or three-dimensional reconstruction of the OCT image.

[0006] The specific steps of calculating the eigenvalues ​​of the interference spectrum and estimating the multi-signal classification spectrum to obtain the OCT longitudinal intensity distribution include:

[0007] The m-order autocorrelation matrix of the interference spectrum is estimated to obtain an autocorrelation matrix signal with a size of (m+1)*(m+1).

[0008] Perform eigenvalue decomposition on the autocorrelation matrix signal to obtain an eigenvalue vector with a length of (m+1) and its corresponding eigenvector matrix.

[0009] The eigenvalue vector is segmented using the threshold method, and the eigenvalues ​​exceeding the custom threshold are counted to obtain the effective dimension of the subspace. Then, it is multiplied by the custom coefficient to obtain the estimated value of the number of signal sources in the interference signal.

[0010] The multi-signal classification spectrum is estimated for the autocorrelation matrix signal, and the order of the multi-signal classification spectrum estimation is specified by the number of signal sources obtained by calculation, so as to obtain the power spectrum of different frequencies in the signal subspace.

[0011] Finally, the power spectra of different frequencies were logarithmically transformed and calibrated to obtain the OCT longitudinal intensity distribution.

[0012] Technical Effects

[0013] The present invention makes up for the defects of the prior art and is a potential alternative to the inverse discrete Fourier transform. Compared with the prior art, the present invention can reconstruct an OCT image with a higher resolution. In terms of noise, since the method has the characteristics of separating the signal subspace and the noise subspace, after obtaining a relatively accurate number of signal sources, only the signal subspace is reconstructed. Therefore, a good noise suppression effect is produced in the reconstructed OCT image. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1 It is a flow chart of the present invention.

[0015] Figure 2 In the example of the present invention, a photoelectric conversion device is used to perform photoelectric conversion on the frequency domain interference fringes and the interference spectrum obtained by multi-point sampling.

[0016] Figure 3 It is the autocorrelation interference signal of the spectrum intensity obtained by calculating the characteristic value of the interference spectrum in the example of the present invention.

[0017] Figure 4 The eigenvalues ​​obtained by calculating the eigenvalues ​​of the autocorrelation interference signal in the example of the present invention have been arranged from large to small.

[0018] Figure 5 It is the OCT longitudinal intensity distribution obtained by performing eigenvalue calculation on the interference spectrum and multi-signal classification spectrum estimation in the example of the present invention.

[0019] Figure 6 The longitudinal intensity distribution of the interference spectrum after Fourier transform is used for comparison.

[0020] Figure 7This is an OCT reconstructed image that performs eigenvalue calculation and multi-signal classification spectrum estimation on the interference spectrum in an example of the present invention.

[0021] Figure 8 The OCT reconstructed image is obtained by Fourier transforming the interference spectrum for comparison. DETAILED DESCRIPTION

[0022] In this embodiment, OCT imaging is performed on a finger skin sample, and a frequency domain OCT system is used to perform spectrum scanning on the sample. Then, a photoelectric conversion device is used to perform photoelectric conversion and multi-point sampling on the obtained frequency domain interference fringes to obtain an interference spectrum. Eigenvalue calculation and multi-signal classification spectrum estimation are performed on the interference spectrum to obtain the OCT longitudinal intensity distribution. Finally, the lateral scanning of the frequency domain OCT system is performed to achieve two-dimensional reconstruction of the OCT image.

[0023] The important characteristics of the frequency domain OCT system used in this example are that the spectrum range of the swept light source is 1260nm-1360nm, and the obtained frequency domain interference fringes are photoelectrically converted and multi-point sampled by the photoelectric balanced detector. For each scanning position, 2048 points of the interference spectrum are collected. Figure 2 As shown, this is the interference spectrum collected by the photoelectric conversion device.

[0024] The 1600-order autocorrelation matrix of the interference spectrum is estimated. Using transposition, the interference spectrum is multiplied with the transposed interference spectrum to obtain the matrix-represented autocorrelation interference signal. Figure 3 As shown in the figure, in the process of eigenvalue calculation, Figure 2 The interference spectrum uses the autocorrelation interference signal obtained by transposing and self-scaling. The size of the signal is selected to be 1601*1601.

[0025] Perform eigenvalue decomposition on the autocorrelation interference signal and obtain an eigenvalue vector with a length of 1601. Perform eigenvalue decomposition on the autocorrelation interference signal and arrange the obtained eigenvalues ​​from large to small. Figure 4 As shown in the figure, in the process of eigenvalue calculation, Figure 3 The result of calculating the eigenvalue of the autocorrelation interference signal and arranging it from large to small. In order to clearly show the threshold segmentation effect, the eigenvalue has been expressed logarithmically.

[0026] The feature vector is segmented using the threshold method, and the feature values ​​exceeding the custom threshold are counted to obtain the effective dimension of the subspace, which is then multiplied by the custom coefficient to obtain the estimated value of the number of signal sources in the interference signal. In this example, the custom threshold is 0, the custom coefficient is set to 1 / 10, and the estimated value of the number of signal sources in the interference signal is 83.

[0027] The multi-signal classification spectrum estimation is used for the autocorrelation interference signal. The number of signal sources obtained by calculation is used to determine the order of the multi-signal classification spectrum estimation, and the power spectrum of different frequencies in the signal subspace is obtained.

[0028] The power spectra of different frequencies are logarithmically transformed and calibrated to obtain the OCT longitudinal intensity distribution. Figure 5 As shown in FIG, the longitudinal intensity distribution of the OCT scan is shown. Figure 6 The result shown is the result of applying Fourier transform to the interference signal.

[0029] Finally, the scanning of the OCT imaging area is completed to obtain the entire image. Figure 7 As shown in FIG. 1 , an image obtained by performing OCT imaging of a finger skin sample according to an embodiment of the present invention is shown. Figure 8 Shown is an image obtained by OCT imaging using Fourier transform.

Claims

1. An OCT measurement imaging method based on eigenvalue calculation and multi-signal classification spectrum estimation, characterized in that: Includes steps: S1. Scan the sample spectrum through the frequency domain OCT system to obtain frequency domain interference fringes; S2. Using a photoelectric conversion device to perform photoelectric conversion and multi-point sampling on the obtained frequency domain interference fringes to obtain an interference spectrum; S3. Calculate the eigenvalues ​​of the interference spectrum and estimate the multi-signal classification spectrum to obtain the OCT longitudinal intensity distribution; S4. Through the lateral scanning of the frequency domain OCT system and repeating S2-S3, multiple OCT longitudinal intensity distributions are obtained, thereby realizing two-dimensional and three-dimensional reconstruction of the OCT image.

2. The OCT measurement imaging method based on eigenvalue calculation and multi-signal classification spectrum estimation according to claim 1, characterized in that: The step S3 specifically includes: S3.1 estimates the m-th order autocorrelation matrix of the interference spectrum to obtain an autocorrelation matrix signal with a size of (m+1)*(m+1). S3.2 performs eigenvalue decomposition on the autocorrelation matrix signal to obtain an eigenvalue vector with a length of (m+1) and its corresponding eigenvector matrix. S3.3 uses the threshold method to segment the eigenvalue vector, counts the eigenvalues ​​that exceed the custom threshold, obtains the effective dimension of the subspace, and then multiplies it by the custom coefficient to obtain the estimated value of the number of signal sources in the interference signal. S3.4 performs multi-signal classification spectrum estimation on the autocorrelation matrix signal obtained in S3.1, uses the number of signal sources calculated in S3.3 to specify the order of the multi-signal classification spectrum estimation, and obtains the power spectrum of different frequencies in the signal subspace. S3.5 performs logarithmic transformation and calibration on the power spectra of different frequencies obtained in S3.4 to obtain the OCT longitudinal intensity distribution.

Images