Optical coherence tomography method and system based on multi-dimensional dynamic feature fusion

Abstract

The invention discloses an optical coherence tomography method and system based on multi-dimensional dynamic feature fusion, and relates to the technical field of optical coherence tomography, and the method comprises the steps: S1, obtaining continuous interference signals of OCT scanning; s2, performing preprocessing and fast Fourier transform to generate an intensity signal; s3, carrying out logarithmic domain transformation on the intensity signal and calculating a variance to obtain an LIV value and mapping the LIV value into a pseudo-color image based on the LIV value; s4, performing singular value decomposition on the intensity signal, calculating the power spectral density and dividing a frequency interval; reconstructing to obtain a reconstructed signal and mapping the reconstructed signal into a pseudo-color image based on frequency distribution; s5, calculating an autocorrelation function of the intensity signal to obtain an attenuation curve, performing function fitting to obtain an OCDS value, and mapping the OCDS value into an image based on attenuation characteristics; and S6, fusing the three images into an HSV color space. According to the method, the LIV algorithm, the SVD-PSD algorithm and the improved OCDS algorithm are combined, conjoint analysis is carried out on the OCT signals in the time domain, the frequency domain and the related domain, and comprehensive visual display is carried out in the HSV color space.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of optical coherence tomography, and particularly relates to an optical coherence tomography method and system based on multi-dimensional dynamic feature fusion. BACKGROUND

[0002] Optical coherence tomography (OCT) is a technology that uses a low-coherence light source for imaging. It can provide high-resolution transverse and longitudinal images of tissue structures, and is widely used in medical and biological research, especially in ophthalmology and dermatology. The core of OCT is the low-coherence interference principle, which uses the interaction of light and biological tissue to produce different levels of scattered photons and ballistic photons. Among them, the multi-level scattered photons have large energy attenuation and do not participate in the interference of OCT, even if a small amount of participation, it is also filtered out as noise signal, the main participant is the ballistic photon, its path is fixed, the reflected or backscattered light from the biological tissue returns to the original path, as the sample light, meets the reference light at the beam splitter and interferes, and the final interference signal is received by the detector.

[0003] However, traditional OCT mainly provides morphological information, which is difficult to meet the needs of biological tissue metabolism and function monitoring. Dynamic OCT (DyC-OCT) provides a new paradigm for functional imaging by analyzing dynamic signals at the cellular and subcellular levels. The application of DyC-OCT has been extended from the early time-domain full-field OCT (FF-OCT) to spectral domain and scanning source OCT systems. Currently, DyC-OCT has been widely used in the study of various types of tissues, including ex vivo mouse organs, in vitro tumor spheroids, human biopsy samples, and in vivo human retinas.

[0004] To analyze the dynamic fluctuations of OCT signals, researchers have developed a variety of algorithms. Among them, the logarithmic intensity variation (LIV) and standard deviation (STD) algorithms use the principles in SV-OCT to measure the amplitude of OCT signal fluctuations. These methods highlight high-activity regions in tissue structures that exhibit greater fluctuation amplitudes, while reducing the visibility of fixed structures in DyC-OCT color maps. Although effective, the LIV algorithm cannot distinguish the speed of tissue dynamics. The subsequently developed OCT correlation decay speed (OCDS) analysis overcomes this problem by measuring temporal dynamic motion patterns through a time de-correlation function. By measuring the slope of the autocorrelation curve of each pixel (e.g., OCDS value), a direct quantitative assessment of motion dynamics is obtained. Color maps can be generated using OCDS values. However, the OCDS algorithm lacks sufficient fitting accuracy for fast decay characteristics when dealing with complex signals, making it difficult to fully capture the multi-dimensional features of tissue dynamics. In addition, the power spectral density (PSD) analysis method based on fast Fourier transform (FFT) reveals the frequency and intensity variations of signals through frequency domain analysis, but the PSD signal obtained by directly performing FFT on the time series signal is often cluttered due to the coupling of space-time patterns, limiting the accurate extraction of frequency characteristics.

[0005] Existing DyC-OCT analysis methods are relatively mature in single-dimensional analysis in time domain, frequency domain, or correlation domain, but lack comprehensive quantification and visual expression of multi-dimensional dynamic characteristics such as motion amplitude, frequency, and decay rate. In addition, the linear fitting of the traditional OCDS algorithm cannot accurately describe the fast decay characteristics of the signal, and the PSD analysis is disturbed by the coupling of space-time signals, making it difficult to achieve pure frequency feature extraction. Therefore, there is an urgent need for an analysis method that integrates multiple algorithms, through space-time decomposition, multi-dimensional feature extraction, and comprehensive visualization, to improve the characterization ability of DyC-OCT for micro-scale dynamic characteristics of biological tissues. SUMMARY

[0006] To solve the above problems, the application provides an optical coherence tomography method and system based on multi-dimensional dynamic feature fusion, which comprises the following steps: performing logarithmic domain transformation on an intensity signal and calculating a variance to obtain a logarithmic intensity variation (LIV) value and map the LIV value into a pseudo-color image based on the LIV value; performing singular value decomposition on the intensity signal, calculating a power spectral density, and dividing a frequency interval according to the power spectral density; reconstructing the intensity signal using singular values and the frequency interval to obtain a reconstructed signal; mapping the reconstructed signal into a pseudo-color image based on frequency distribution; calculating an autocorrelation function to obtain an attenuation curve of the intensity signal, performing function fitting on the attenuation curve to obtain an optical coherence decay speed (OCDS) value, and mapping the OCDS value into an image based on attenuation characteristics; combining the logarithmic intensity variation (LIV) algorithm, the singular value decomposition-based power spectral density (SVD-PSD) algorithm, and the improved correlation decay speed (OCDS) algorithm to realize multi-dimensional quantitative analysis of micro-scale dynamic features of biological tissues; and performing joint analysis on the OCT signal in the time domain, the frequency domain, and the correlation domain to simultaneously obtain kinetic parameters such as the motion amplitude, the frequency, and the time decay rate of the tissue, and fuse the information into an HSV color space for comprehensive visual display.

[0007] In one aspect, the optical coherence tomography method based on multi-dimensional dynamic feature fusion comprises the following steps:

[0008] S1, obtaining continuous interference signals of an optical coherence tomography (OCT) obtained by performing multi-frame repeated scanning on a same spatial position of a sample to be measured;

[0009] S2, performing preprocessing and fast Fourier transform on the continuous interference signals to generate an intensity signal containing a time mode;

[0010] S3, performing logarithmic domain transformation on the intensity signal and calculating a variance to obtain a logarithmic intensity variation (LIV) value and map the LIV value into a pseudo-color image based on the LIV value;

[0011] S4, performing singular value decomposition on the intensity signal, calculating a power spectral density, and dividing a frequency interval according to the power spectral density; reconstructing the intensity signal using singular values and the frequency interval to obtain a reconstructed signal; and mapping the reconstructed signal into a pseudo-color image based on frequency distribution;

[0012] S5, calculating an autocorrelation function of the intensity signal to obtain an attenuation curve, performing function fitting on the attenuation curve to obtain an optical coherence decay speed (OCDS) value, and mapping the OCDS value into an image based on attenuation characteristics;

[0013] S6, fusing the pseudo-color image based on the LIV value, the pseudo-color image based on frequency distribution, and the image based on attenuation characteristics into an HSV color space to generate a multi-dimensional optical coherence tomography image; the hue of the multi-dimensional optical coherence tomography image corresponds to a motion frequency, the saturation corresponds to a motion amplitude, and the lightness corresponds to a decay rate.

[0014] Preferably, the step of fusing the LIV-based pseudo-color image, the frequency distribution-based pseudo-color image, and the attenuation characteristic-based image into the HSV color space to generate a multidimensional dynamic optical coherence tomography image specifically involves: mapping the low-frequency, mid-frequency, and high-frequency components in the frequency distribution-based pseudo-color image to blue, green, and red hues, respectively; mapping the fluctuation amplitude reflected by the variance value of the LIV-based pseudo-color image to saturation; and mapping the attenuation rate of the attenuation characteristic-based image to brightness.

[0015] Preferably, the preprocessing and fast Fourier transform of the continuous interference signal to generate an intensity signal containing a time pattern is as follows:

[0016] Valid data points are extracted based on the valid signal index of OCT, while invalid signal points within the frequency sweep cycle of the light source are shielded.

[0017] Remove background noise from the signal and the DC component of the interference signal;

[0018] Use window functions for signal filtering;

[0019] Perform a Fast Fourier Transform on the filtered signal to generate an OCT B-scan intensity signal;

[0020] The first few points of the OCT B-scan intensity signal are extracted as valid signals.

[0021] Preferably, S4 is as follows:

[0022] The time series signals of each A-line in the intensity signal are combined to form a two-dimensional Cascorati matrix A with dimensions M×N; where M represents the number of depth sampling points and N represents the number of time frames.

[0023] Singular value decomposition of matrix A is expressed as:

[0024]

[0025] Where U represents the time pattern matrix, S represents the singular value diagonal matrix, and V represents the spatial pattern matrix; Indicates transpose;

[0026] Perform a Fast Fourier Transform (FFT) on each column vector of matrix U and calculate the power spectral density (PSD), expressed as:

[0027]

[0028] in, Indicates the first The power spectral density of the column; Indicates Fast Fourier Transform; denotes the matrix U the column; denotes the square of the modulo operation;

[0029] According to the power spectral density, the frequency interval is divided into low frequency, medium frequency and high frequency;

[0030] The signal is reconstructed according to the singular value energy distribution and the frequency interval weighting weight, and the reconstructed signal matrix is obtained, which is represented as:

[0031] ,

[0032] wherein, denotes the component of the reconstructed signal matrix in the frequency interval; denotes the component of the time mode matrix in the frequency interval; denotes the component of the singular value diagonal matrix in the frequency interval; denotes the component of the spatial mode matrix in the frequency interval; denotes the frequency interval; denotes the low frequency interval; denotes the medium frequency interval; denotes the high frequency interval;

[0033] The low frequency interval, medium frequency interval and high frequency interval components in the reconstructed signal matrix are respectively mapped to blue, green and red to generate a pseudo-color image based on frequency distribution.

[0034] Preferably, the S5 is specifically as follows:

[0035] The autocorrelation function of the time series signal of each pixel point is calculated, which is represented as:

[0036]

[0037] wherein, denotes the autocorrelation function; is the time delay; denotes the length of the time series signal; denotes the signal value at the n th time point; denotes the signal value at the time point;

[0038] The preset empirical threshold is taken as the segmentation point, and the part of the decay curve of the autocorrelation function with time less than or equal to the preset empirical threshold is taken as the early stage, and the part of the decay curve of the autocorrelation function with time greater than the preset empirical threshold is taken as the late stage;

[0039] An exponential function is fitted to the early section of the decay curve to generate an early OCDS map, expressed as:

[0040]

[0041] wherein, represents the decay exponent; represents the initial amplitude; represents the baseline offset,

[0042] A linear fitting is performed on the late section of the decay curve to extract the slope as a late OCDS value, and a late OCDS map is generated.

[0043] The early OCDS map and the late OCDS map are combined into an image based on the decay characteristics.

[0044] In another aspect, an optical coherence tomography system based on multi-dimensional dynamic feature fusion includes the following:

[0045] An OCT signal acquisition module is configured to acquire continuous interference signals obtained by performing multiple repeated scans on a same spatial position of a sample to be measured by an optical coherence tomography (OCT) system.

[0046] A preprocessing and reconstruction module is configured to perform preprocessing and fast Fourier transform on the continuous interference signals to generate an intensity signal containing a time pattern.

[0047] An LIV processing module is configured to perform a logarithmic domain transformation on the intensity signal and calculate a variance to obtain a logarithmic intensity variation (LIV) value and map the LIV value into a pseudo-color image based on the LIV value.

[0048] An SVD-PSD processing module is configured to perform singular value decomposition on the intensity signal, calculate a power spectral density, and divide a frequency interval according to the power spectral density; reconstruct the intensity signal using the singular value and the frequency interval to obtain a reconstructed signal; and map the reconstructed signal into a pseudo-color image based on a frequency distribution.

[0049] An OCDS processing module is configured to calculate an autocorrelation function of the intensity signal to obtain a decay curve, perform function fitting on the decay curve to obtain an OCDS value, and map the OCDS value into an image based on decay characteristics.

[0050] A feature fusion module is configured to fuse the pseudo-color image based on the LIV value, the pseudo-color image based on the frequency distribution, and the image based on the decay characteristics into an HSV color space to generate a multi-dimensional dynamic optical coherence tomography image; a hue of the multi-dimensional optical coherence tomography image corresponds to a motion frequency, a saturation corresponds to a motion amplitude, and a lightness corresponds to a decay rate.

[0051] Compared with the prior art, the present application has the following beneficial effects:

[0052] (1) The present application obtains the logarithmic intensity variation (LIV) value by performing logarithmic domain transformation on the intensity signal and calculating the variance, and maps it into a pseudo-color image based on LIV; performs singular value decomposition on the intensity signal, calculates the power spectral density, and divides the frequency interval according to the power spectral density; reconstructs the intensity signal using the singular value and the frequency interval to obtain a reconstructed signal; maps the reconstructed signal into a pseudo-color image based on the frequency distribution; calculates the autocorrelation function to obtain the decay curve of the intensity signal, and performs function fitting on the decay curve to obtain the OCDS value, which is mapped into an image based on the attenuation characteristics; the logarithmic intensity variation (LIV) algorithm, the power spectral density based on singular value decomposition (SVD-PSD) algorithm and the improved correlation decay speed (OCDS) algorithm are combined to realize multi-dimensional quantitative analysis of the micro-scale dynamic characteristics of biological tissues; through time-space decomposition and piecewise fitting, the accuracy of dynamic feature extraction is improved, and the method is suitable for functional imaging in OCT biomolecular dynamics research;

[0053] (2) The present application can simultaneously obtain the kinetic parameters such as the motion amplitude, frequency and time decay rate of the tissue by jointly analyzing the OCT signal in the time domain, frequency domain and correlation domain, and fuse these information into the HSV color space for comprehensive visual display. BRIEF DESCRIPTION OF DRAWINGS

[0054] The present application will be further described in detail below with reference to the accompanying drawings;

[0055] Figure 1 The flowchart of the optical coherence tomography method based on multi-dimensional dynamic feature fusion of the embodiment of the present application;

[0056] Figure 2 The flowchart of the optical coherence tomography method based on multi-dimensional dynamic feature fusion of the embodiment of the present application;

[0057] Figure 3 The LIV image of the optical coherence tomography method based on multi-dimensional dynamic feature fusion of the embodiment of the present application;

[0058] Figure 4 The algorithm comparison chart of the optical coherence tomography method based on multi-dimensional dynamic feature fusion of the embodiment of the present application; wherein (a) represents the 2000 frame average image of the traditional OCT signal intensity; (b) represents the power spectral density (PSD) algorithm result image based on fast Fourier transform (FFT); (c) represents the power spectral density (SVD-PSD) algorithm result image based on singular value decomposition;

[0059] Figure 5 The OCDSe image of the optical coherence tomography method based on multi-dimensional dynamic feature fusion of the embodiment of the present application;

[0060] Figure 6 OCDSl image of the method for optical coherence tomography based on multi-dimensional dynamic feature fusion of the embodiment of the application;

[0061] Figure 7 Multi-dimensional dynamic OCT image of the method for optical coherence tomography based on multi-dimensional dynamic feature fusion of the embodiment of the application;

[0062] Figure 8 Structural block diagram of the system for optical coherence tomography based on multi-dimensional dynamic feature fusion of the embodiment of the application. DETAILED DESCRIPTION

[0063] The application is further described below through a specific embodiment. The embodiment takes grape samples as an example to illustrate the implementation process of the method for OCT analysis based on multi-dimensional dynamic feature fusion, but the method is applicable to any biological sample with cell activity, and is not limited to grape.

[0064] As shown in Figure 1 and Figure 2 , the method for optical coherence tomography based on multi-dimensional dynamic feature fusion has the following specific steps:

[0065] S1, obtaining continuous interference signals obtained by repeatedly scanning the same spatial position of a sample to be measured by an optical coherence tomography (OCT) system.

[0066] The purpose of this step is data collection. A SS-OCT system based on a 100 kHz sweep light source is used to repeatedly scan the same spatial position for multiple times to obtain a continuous interference signal data set. The specific settings are as follows:

[0067] Each B-scan contains 1000 A-lines, each A-line contains 3968 sampling points, and 2000 B-scans are collected. The sweep frequency of the light source is 100 kHz, and the rate of a single B-scan is 100 Hz. According to the Nyquist sampling law, the maximum detectable motion frequency is 50 Hz, and the minimum detectable frequency is 0.05 Hz (determined by 2000 B-scans). The SS-OCT system integrates a balanced detector, a fiber type Mach-Zehnder interferometer, a two-dimensional galvanometer beam scanning system, and a synchronous control and high-speed data acquisition module to ensure the stability and high resolution of signal acquisition.

[0068] S2, pre-processing and fast Fourier transform of the continuous interference signals to generate an intensity signal containing a time pattern.

[0069] The purpose of this step is signal pre-processing and reconstruction. The original interference signal is pre-processed and reconstructed, and the specific steps are as follows:

[0070] S21, extracting valid data points according to the valid signal index, and shielding invalid signal points in the light source sweep frequency period to reduce interference on subsequent FFT analysis.

[0071] S22, removing the background noise of the signal and the direct current component of the interference signal.

[0072] S23, performing signal filtering using a Hanning window or a Gaussian window to improve spectral resolution.

[0073] S24, performing fast Fourier transform (FFT) on the filtered signal to generate an OCT B-scan intensity signal.

[0074] S25, extracting the first 512 points from the FFT result of each A-line as valid signals (most of the total 1544 points are non-sample signals), generating a B-scan intensity data set containing time patterns for subsequent dynamic analysis.

[0075] S3, performing logarithmic domain transformation on the intensity signal and calculating variance to obtain a logarithmic intensity variation (LIV) value and map it into a pseudo-color image based on the LIV value.

[0076] The purpose of this step is to analyze the logarithmic intensity variation (LIV). The reconstructed intensity signal is subjected to logarithmic domain transformation, the signal fluctuation amplitude is calculated, and the LIV image is generated, as shown in Figure 3 The specific steps are as follows:

[0077] S31, converting the OCT intensity signal to a logarithmic (dB) scale to separate dynamic and static components, expressed as:

[0078]

[0079] S32, calculating the variance of the logarithmic intensity signal to obtain the LIV value:

[0080]

[0081]

[0082] The LIV value reflects the dynamic change of the scattering intensity inside the tissue. The larger the value, the stronger the tissue activity. A pseudo-color image based on LIV is generated to visually display the tissue activity area.

[0083] S4, performing singular value decomposition on the intensity signal, calculating the power spectral density and dividing the frequency interval according to the power spectral density; using singular values and frequency intervals to reconstruct the intensity signal to obtain a reconstructed signal; and mapping the reconstructed signal into a pseudo-color image based on frequency distribution.

[0084] The purpose of this step is to extract the frequency domain features based on SVD decomposition (SVD-PSD). This algorithm is superior to the traditional single time series signal FFT algorithm, as shown in the comparison chart. Figure 4 The specific steps are as follows:

[0085] S41, the time series signal of each A-line is composed into a two-dimensional Casorati matrix , where M is the number of depth sampling points, and N is the number of time frames.

[0086] S42, singular value decomposition (SVD) is performed on the matrix A, which is represented as:

[0087]

[0088] where U is the time mode matrix, S is the singular value diagonal matrix, and V is the spatial mode matrix.

[0089] S43, perform fast Fourier transform (FFT) on each column vector of the U matrix to calculate the power spectral density (PSD), which is represented as:

[0090]

[0091] S44, according to experience, divide the frequency range: 0-0.05 Hz is low frequency (corresponding to slow motion), 0.05-0.9 Hz is medium frequency (corresponding to medium speed motion), and 0.9-20 Hz is high frequency (corresponding to fast motion).

[0092] S45, weight the reconstructed signal according to the singular value energy distribution and the frequency range, which is represented as:

[0093] ,

[0094] Map the low-frequency component to blue, the medium-frequency component to green, and the high-frequency component to red to generate a pseudo-color PSD image, which shows the spatial distribution of tissue motion frequency (Fig. Figure 4 (c)).

[0095] S5, calculate the autocorrelation function of the intensity signal to obtain the decay curve, and perform function fitting on the decay curve to obtain the OCDS value. Map the OCDS value to an image based on the decay characteristics.

[0096] The purpose of this step is to extract the OCDS decay feature. Calculate the time correlation decay characteristics of the reconstructed intensity signal to generate an OCDS image (as shown in Figs. Figure 5 and Figure 6 The specific steps are as follows:

[0097] S51, calculate the autocorrelation function of the time series signal of each pixel point:

[0098]

[0099] where, is the time delay, reflecting the correlation change of the signal with time delay.

[0100] S52, the correlation decay curve is divided into early stage (short time delay) and late stage (long time delay) according to the empirical threshold.

[0101] S53, in the early stage, an exponential function is used to fit the rapid decay characteristics:

[0102]

[0103] where, is the decay index, reflecting the initial decorrelation rate of the signal, generating the early OCDS map (OCDSe), as shown in Figure 5 .

[0104] S54, in the late stage, a linear fitting is used to extract the overall decay trend, and the slope is defined as the late OCDS value, generating the late OCDS map (OCDSl), as shown in Figure 6 .

[0105] S6, the pseudo-color image based on LIV value, the pseudo-color image based on frequency distribution and the image based on decay characteristics are fused into HSV color space to generate multi-dimensional dynamic optical coherence tomography image.

[0106] The purpose of this step is multi-dimensional feature fusion and visualization; the results of LIV, SVD-PSD and OCDS algorithms are fused into HSV color space to generate multi-dimensional dynamic OCT image, as shown in Figure 7 , and the specific mapping method is as follows:

[0107] A. Hue: the motion frequency information obtained by SVD-PSD algorithm, low frequency (blue), medium frequency (green) and high frequency (red).

[0108] B. Saturation: the motion amplitude information obtained by LIV algorithm, the greater the amplitude, the higher the saturation.

[0109] C. Value: the signal decay rate obtained by the improved OCDS algorithm, the faster the decay rate, the higher the value.

[0110] By HSV color space mapping, a pseudo-color image synthetically reflecting the amplitude, frequency and attenuation characteristics of tissue motion is generated, and the visual expression of multi-dimensional dynamic characteristics is realized. Among them, the SVD-PSD algorithm extracts the frequency domain characteristics to distinguish different physiological motion modes; the LIV algorithm quantifies the amplitude change of the signal to reflect the activity of the tissue; and the improved OCDS algorithm quantifies the motion attenuation rate of the biological tissue by fitting the early OCDS with an exponential function and fitting the late OCDS with a linear function.

[0111] As shown in Figure 8 The application further discloses an optical coherence tomography device based on multi-dimensional dynamic characteristic fusion, which comprises:

[0112] An OCT signal acquisition module 801 is configured to acquire continuous interference signals obtained by performing multi-frame repeated scanning on a same spatial position of a sample to be measured by optical coherence tomography (OCT);

[0113] A preprocessing and reconstruction module 802 is configured to pre-process and perform fast Fourier transform on the continuous interference signals to generate an intensity signal containing a time mode;

[0114] An LIV processing module 803 is configured to perform logarithmic domain transformation on the intensity signal and calculate a variance to obtain a logarithmic intensity variation (LIV) value and map the LIV value into a pseudo-color image based on the LIV value;

[0115] An SVD-PSD processing module 804 is configured to perform singular value decomposition on the intensity signal, calculate a power spectral density and divide a frequency interval according to the power spectral density; reconstruct the intensity signal using the singular value and the frequency interval to obtain a reconstructed signal; and map the reconstructed signal into a pseudo-color image based on frequency distribution;

[0116] An OCDS processing module 805 is configured to calculate an autocorrelation function of the intensity signal to obtain an attenuation curve, perform function fitting on the attenuation curve to obtain an OCDS value, and map the OCDS value into an image based on attenuation characteristics;

[0117] A characteristic fusion module 806 is configured to fuse the pseudo-color image based on the LIV value, the pseudo-color image based on the frequency distribution and the image based on the attenuation characteristics into an HSV color space to generate a multi-dimensional dynamic optical coherence tomography image; and the hue of the multi-dimensional optical coherence tomography image corresponds to a motion frequency, the saturation corresponds to a motion amplitude, and the lightness corresponds to an attenuation rate.

[0118] The specific implementation of the optical coherence tomography system based on multi-dimensional dynamic characteristic fusion is the same as the optical coherence tomography method based on multi-dimensional dynamic characteristic fusion, and will not be repeated here.

[0119] The above merely illustrates the specific embodiments of the present application, but the design concept of the present application is not limited thereto, and any non-essential modification of the present application by using the concept shall be deemed as the infringement of the protection scope of the present application.

Claims

1. An optical coherence tomography method based on multidimensional dynamic feature fusion, characterized in that, Includes the following steps: S1, acquire the continuous interference signal obtained by optical coherence tomography (OCT) of the sample under test by performing multiple repeated scans at the same spatial position; S2 preprocesses and performs fast Fourier transform on the continuous interference signal to generate an intensity signal containing the time pattern. S3, perform a logarithmic domain transformation on the intensity signal and calculate the variance to obtain the logarithmic intensity change LIV value and map it into a pseudo-color image based on the LIV value; S4, perform singular value decomposition on the intensity signal, calculate the power spectral density and divide the frequency interval according to the power spectral density; reconstruct the intensity signal using singular values ​​and frequency intervals to obtain the reconstructed signal; map the reconstructed signal into a pseudo-color image based on frequency distribution; S5, calculate the autocorrelation function of the intensity signal to obtain the attenuation curve, perform function fitting on the attenuation curve to obtain the OCDS value, and map the OCDS value into an image based on the attenuation characteristics; S6, the pseudo-color image based on LIV value, the pseudo-color image based on frequency distribution, and the image based on attenuation characteristics are fused into the HSV color space to generate a multidimensional dynamic optical coherence tomography image; the color of the multidimensional optical coherence tomography image corresponds to the motion frequency, the saturation corresponds to the motion amplitude, and the brightness corresponds to the attenuation rate.

2. The optical coherence tomography method based on multidimensional dynamic feature fusion according to claim 1, characterized in that, The process of fusing the LIV-based pseudo-color image, the frequency distribution-based pseudo-color image, and the attenuation characteristic-based image into the HSV color space to generate a multidimensional dynamic optical coherence tomography image specifically involves: mapping the low-frequency, mid-frequency, and high-frequency components in the frequency distribution-based pseudo-color image to blue, green, and red hues, respectively; mapping the fluctuation amplitude reflected by the variance value of the LIV-based pseudo-color image to saturation; and mapping the attenuation rate of the attenuation characteristic-based image to brightness.

3. The optical coherence tomography method based on multidimensional dynamic feature fusion according to claim 1, characterized in that, The continuous interference signal is preprocessed and subjected to Fast Fourier Transform to generate an intensity signal containing a time pattern, as detailed below: Valid data points are extracted based on the valid signal index of OCT, while invalid signal points within the frequency sweep cycle of the light source are shielded. Remove background noise from the signal and the DC component of the interference signal; Use window functions for signal filtering; Perform a Fast Fourier Transform on the filtered signal to generate an OCT B-scan intensity signal; The first few points of the OCT B-scan intensity signal are extracted as valid signals.

4. The optical coherence tomography method based on multidimensional dynamic feature fusion according to claim 1, characterized in that, The window function is either a Hanning window or a Gaussian window.

5. The optical coherence tomography method based on multidimensional dynamic feature fusion according to claim 1, characterized in that, S4 is specifically as follows: The time series signals of each A-line in the intensity signal are combined to form a two-dimensional Cascorati matrix A with dimensions M×N; where M represents the number of depth sampling points and N represents the number of time frames. Singular value decomposition of matrix A is expressed as: Where U represents the time pattern matrix, S represents the singular value diagonal matrix, and V represents the spatial pattern matrix; Indicates transpose; Perform a Fast Fourier Transform (FFT) on each column vector of matrix U and calculate the power spectral density (PSD), expressed as: in, Indicates the first The power spectral density of the column; Indicates Fast Fourier Transform; Represents the matrix U-th List; This represents the square of the modulus; The frequency range is divided into low frequency, mid frequency and high frequency based on the power spectral density. The reconstructed signal matrix is ​​obtained by weighting the singular value energy distribution and frequency range; it is represented as: ， in, Indicates the reconstructed signal matrix in Components within a frequency range; Indicating the time pattern matrix in Components within a frequency range; Indicates the singular valued diagonal matrix in Components within a frequency range; The spatial pattern matrix is ​​in Components within a frequency range; Indicates the first One frequency range; Indicates the low-frequency range; Indicates the mid-frequency range; Indicates the high-frequency range; The low-frequency, mid-frequency, and high-frequency components in the reconstructed signal matrix are mapped to blue, green, and red, respectively, to generate a pseudo-color image based on frequency distribution.

6. The optical coherence tomography method based on multidimensional dynamic feature fusion according to claim 1, characterized in that, S5 is specifically as follows: The autocorrelation function of the time-series signal for each pixel is calculated and expressed as: in, Represents the autocorrelation function; For time delay; Indicates the length of the time series signal; This represents the signal value at the nth time point; Indicates the first Signal values ​​at each time point; Using a preset empirical threshold as the segmentation point, the portion of the decay curve of the autocorrelation function with a time less than or equal to the preset empirical threshold is taken as the early segment, and the portion of the decay curve of the autocorrelation function with a time greater than the preset empirical threshold is taken as the late segment. An exponential function is used to fit the early segment of the decay curve to generate an early OCDS plot, which is represented as follows: in, Indicates the decay index; Indicates the initial amplitude; Indicates the baseline offset; Linear fitting was applied to the late segment of the decay curve, and the slope was extracted as the late OCDS value to generate a late OCDS map. The early and late OCDS maps are combined into an image based on attenuation characteristics.

7. An optical coherence tomography system based on multidimensional dynamic feature fusion, characterized in that, include: The OCT signal acquisition module is used to acquire continuous interference signals obtained by optical coherence tomography (OCT) of the sample under test through multiple repeated scans at the same spatial position. The preprocessing and reconstruction module is used to preprocess and perform fast Fourier transform on continuous interference signals to generate intensity signals containing time patterns. The LIV processing module is used to perform a logarithmic domain transformation on the intensity signal and calculate the variance to obtain the logarithmic intensity change LIV value and map it into a pseudo-color image based on the LIV value. The SVD-PSD processing module is used to perform singular value decomposition on the intensity signal, calculate the power spectral density, and divide the frequency range according to the power spectral density; reconstruct the intensity signal using singular values ​​and frequency ranges to obtain the reconstructed signal; and map the reconstructed signal into a pseudo-color image based on frequency distribution. The OCDS processing module is used to calculate the autocorrelation function of the intensity signal to obtain the attenuation curve, perform function fitting on the attenuation curve to obtain the OCDS value, and map the OCDS value into an image based on the attenuation characteristics. The feature fusion module is used to fuse a pseudo-color image based on LIV values, a pseudo-color image based on frequency distribution, and an image based on attenuation characteristics into the HSV color space to generate a multidimensional dynamic optical coherence tomography image; the color of the multidimensional optical coherence tomography image corresponds to the motion frequency, the saturation corresponds to the motion amplitude, and the brightness corresponds to the attenuation rate.

Images