WO2020155415A1 - 基于特征空间的光学相干层析的三维血流造影方法及系统 - Google Patents

基于特征空间的光学相干层析的三维血流造影方法及系统 Download PDF

Info

Publication number
WO2020155415A1
WO2020155415A1 PCT/CN2019/083491 CN2019083491W WO2020155415A1 WO 2020155415 A1 WO2020155415 A1 WO 2020155415A1 CN 2019083491 W CN2019083491 W CN 2019083491W WO 2020155415 A1 WO2020155415 A1 WO 2020155415A1
Authority
WO
WIPO (PCT)
Prior art keywords
signal
blood flow
oct
dimensional
noise ratio
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Ceased
Application number
PCT/CN2019/083491
Other languages
English (en)
French (fr)
Inventor
李鹏
黄璐哲
付奕铭
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Zhejiang University
Original Assignee
Zhejiang University
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Zhejiang University filed Critical Zhejiang University
Priority to US17/628,579 priority Critical patent/US11510574B2/en
Publication of WO2020155415A1 publication Critical patent/WO2020155415A1/zh
Anticipated expiration legal-status Critical
Ceased legal-status Critical Current

Links

Images

Classifications

    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/0059—Measuring for diagnostic purposes; Identification of persons using light, e.g. diagnosis by transillumination, diascopy, fluorescence
    • A61B5/0062—Arrangements for scanning
    • A61B5/0066—Optical coherence imaging
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/02—Detecting, measuring or recording for evaluating the cardiovascular system, e.g. pulse, heart rate, blood pressure or blood flow
    • A61B5/02007—Evaluating blood vessel condition, e.g. elasticity, compliance
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/02—Detecting, measuring or recording for evaluating the cardiovascular system, e.g. pulse, heart rate, blood pressure or blood flow
    • A61B5/026—Measuring blood flow
    • A61B5/0261—Measuring blood flow using optical means, e.g. infrared light
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
    • A61B5/7203—Signal processing specially adapted for physiological signals or for diagnostic purposes for noise prevention, reduction or removal
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
    • A61B5/7235—Details of waveform analysis
    • A61B5/7264—Classification of physiological signals or data, e.g. using neural networks, statistical classifiers, expert systems or fuzzy systems

Definitions

  • the present invention generally relates to the field of biomedical imaging, and more specifically to optical coherence tomography (Optical Coherence Tomography, OCT) and optical coherence tomography (OCT Angiography, OCTA) associated blood flow contrast method and Blood flow classification method based on time series model.
  • OCT optical Coherence Tomography
  • OCT Angiography optical coherence tomography
  • Blood flow is an important indicator to measure physiological function and pathological status.
  • the commonly used clinical angiography technique requires intravenous injection of exogenous markers.
  • the possible side effects make it unsuitable for long-term and frequent tracking of human blood flow.
  • OCTA an angiography technology developed on the basis of optical coherence tomography technology, replaces traditional exogenous fluorescent markers with endogenous blood flow movement. Its non-invasive and label-free features, as well as its effects on biological tissues
  • the ability of the microvascular network to perform clear and reliable three-dimensional imaging has enabled this technology to develop rapidly since its invention, and has been applied in the research of fundus, cerebral cortex blood vessels and skin imaging.
  • OCTA blood flow images In order to obtain OCTA blood flow images, it is usually necessary to repeat sampling (repeated A-line scan or B-frame scan) at each spatial location of biological tissues at a certain time interval, and quantify each by analyzing the time dynamics of OCT scattering signals The exercise intensity at the position is classified according to the quantified exercise intensity to classify the blood flow signal and the static tissue signal.
  • the reported motion metrics of OCTA are mainly based on the calculation of difference, variance or decorrelation between adjacent A-line scans (or between adjacent B-scan frames).
  • the OCTA motion measurement index based on decorrelation calculation uses signal intensity and phase information at the same time, so the reliability of the classification result is higher, and it is less affected by the Doppler angle and the overall intensity of the light source.
  • decorrelation measures the similarity between adjacent B-scan frames, it is less affected by changes in the overall light source intensity.
  • the movement index of the vascular flow of the OCTA especially the movement index of the decorrelation coefficient
  • signal strength such as in the deep tissue area
  • random noise will gradually occupy the main component, and its time-varying characteristics are similar to the real blood flow area, thus leading to decorrelation artifacts.
  • the decorrelation coefficient movement index it is impossible to distinguish the randomness of noise and decorrelation caused by the movement of red blood cells. Therefore, areas with weak signal-to-noise ratio are easily misjudged as blood flow signal areas, which seriously affects the contrast of blood flow images.
  • the common solution is to set an empirical intensity threshold and generate an intensity mask to remove all low signal-to-noise ratio signals.
  • an intensity mask due to the complex dependence between the de-correlation coefficient and the signal intensity, a simple intensity mask will result in high classification error rate and low motion contrast.
  • the present invention proposes a three-dimensional blood flow imaging method and system based on feature space optical coherence tomography, which uses the signal-to-noise ratio and decorrelation features of OCT signals Yes, using a multivariate time series model for signal-to-noise ratio-adapted three-dimensional blood flow imaging can accurately distinguish between blood flow and static areas mixed with noise in OCT images.
  • It includes a collector step for collecting OCT scattered signals of scattered signal samples in a three-dimensional space
  • Including a classifier step combining the signal-to-noise ratio of the OCT scattering signal and the decorrelation coefficient to construct a two-dimensional feature space to realize the classification of dynamic blood flow signals and static tissues, and then realize the correction using the signal-to-noise ratio;
  • the scattering signal sample is a biological tissue sample, and the biological tissue sample may be human and animal skin, brain tissue, and eyes, for example.
  • the OCT scattering signal used to collect the scattering signal sample in the three-dimensional space specifically includes: performing OCT scanning imaging of the scattering signal sample in the three-dimensional space, re-sampling the same spatial position and nearby positions at S different time points and using multiple channels Repeat sampling.
  • Multi-channel refers to the multiple polarization states and multiple scanning incident angles of the light emitted by the OCT light source.
  • the three-dimensional OCT scanning imaging of the scattered signal samples is performed, and the same spatial position and nearby positions are repeatedly sampled at S different time points, using one of the following methods:
  • Time-domain OCT imaging method that changes the optical path of the reference arm by scanning
  • a spectral domain OCT imaging method that uses a spectrometer to record spectral interference signals
  • a swept-frequency OCT imaging method that uses a swept-frequency light source to record spectral interference signals.
  • the combination of the signal-to-noise ratio of the OCT scattering signal and the decorrelation coefficient to construct a two-dimensional signal-to-noise ratio reciprocal-decrelation coefficient (ID) feature space to realize the classification of dynamic blood flow signals and static tissues includes:
  • the OCT scattering signal is divided into dynamic blood flow signal and noise and static tissue signal.
  • the noise signal is the signal of the sample-free area.
  • the dynamic blood flow signal and the static tissue signal respectively correspond to the blood flow area and the static tissue area.
  • the step S1 is specifically:
  • the first-order and zero-order autocovariance are used to calculate and analyze the OCT scattering signal, and the moving average or Gaussian average is performed in three dimensions, such as three-dimensional space, time, channel, etc., to obtain two of the intensity and decorrelation coefficient of each OCT scattering signal Features, as shown in Figures 4(a) and 4(b).
  • the features of the decorrelation coefficient include: the amplitude part of the OCT scattering signal obtained by scanning at the same spatial position at S different time points or calculating the decorrelation coefficient for the OCT scattering signal including the amplitude and phase, as the OCT blood of each OCT scattering signal Flow contrast, which is used as OCTA blood flow information.
  • the classifier shown in Figure 4(c) is built in the feature space; the blood flow angiogram is generated according to the classification results of the classifier as shown in Figure 4. (d).
  • the signal X(m,t) collected by the OCT scattering signal is the backscattered light signal A(m,t) and random noise n(m,t) from the biological tissue sample ) Is superimposed, where m and t respectively refer to the spatial position and time of the voxel.
  • the time series A, n establishes the following conditions: 1. n(m, t) is spatially independent white noise, and is not related to the backscattered light signal A(m, t); 2. A(m ,t) is a stable, ergodic time series that is independent and identically distributed in space, and it is expected to exist, with finite variance (second-order moment); 3. A(m,t) is partially correlated in time, that is, the first-order self The covariance exists, and the value range of the modulus of the first-order autocorrelation coefficient k is [0,1].
  • the step S2 is specifically: combining the intensity of the OCT scattering signal and the decorrelation coefficient to construct a signal-to-noise ratio reciprocal-decrelation coefficient (ID) feature space, and calculating the signal-to-noise ratio of the OCT scattering signal by numerical simulation according to the multivariate time series model-decorrelation
  • the asymptotic relationship of the coefficients (Equation (5)), the OCT scattering signal is projected into the reciprocal signal-to-noise ratio-decorrelation coefficient (ID) feature space, as shown in Figure 4(c), to obtain the volume in the static region and the noise region Prime distribution (Equation (6)) and its statistical characteristics (Equation (7, 8)).
  • the asymptotic relationship between the signal-to-noise ratio and the decorrelation coefficient of the OCT scattering signal is calculated by numerical simulation according to the multivariate time series model.
  • the intensity and uniform distribution of A are simulated according to the generally considered Rayleigh distribution, and the white noise is set as Gaussian complex random noise to generate several simulated pictures.
  • the final result will be The analog signal is projected on the signal-to-noise ratio-decrelation coefficient feature space.
  • the variance of the decorrelation is inversely proportional to the average parameter, and the inverse proportional function parameters can be obtained by fitting the squared coefficient of variation of the sample and the average parameter.
  • the linear classifier parameters derived from the model are determined by the parameters obtained by the simulation and the noise level.
  • the OCT scattering signal is divided into dynamic blood flow signal and static tissue noise signal.
  • the decorrelation value for display or directly set it to 1 for display.
  • the linear classifier refers to: through theoretical formulas, according to When the algorithm is executed, the multi-dimensional average kernel size (including the size in space, time and the number of channels) directly obtains a linear classifier in the two-dimensional feature space, and the classifier is universal and has nothing to do with other parameters.
  • the improved classifier refers to the linear classifier obtained by adjusting the slope or intercept (overall rotation or translation) of the classifier on the basis of the aforementioned linear classifier.
  • the classification result based on the classifier generates a blood flow contrast map (25), and the static tissue signal with noise signal has been removed from the classification result, and the remaining signal is used as a dynamic blood flow signal, and its decorrelation coefficient value is used to form blood Flow radiography, as shown in Figure 3.
  • performing operations such as Gaussian filtering, median filtering, Hessian filtering, etc. to remove noise, Improve the effect of blood flow imaging.
  • a three-dimensional blood flow imaging system based on feature space 1.
  • An OCT optical coherence tomography detection device for collecting OCT scattering signals of scattering signal samples in a three-dimensional space
  • One or more signal processors are used to obtain and analyze the intensity information of the OCT scattered signal and the OCTA blood flow information, and combine the intensity information of the scattered signal and the OCTA blood flow information to classify the dynamic blood flow signal and the static tissue background.
  • the OCT optical coherence tomography detection device uses one of the following:
  • low-coherence light source Including low-coherence light source, interferometer and detector, low-coherence light source and detector are respectively connected to the interferometer;
  • low-coherence light source interferometer and spectrometer
  • low-coherence light source and spectrometer are respectively connected to the interferometer
  • swept-bandwidth-spectrum light source includes a swept-bandwidth-spectrum light source, an interferometer and a detector, and the swept-bandwidth-spectrum light source and the detector are respectively connected to the interferometer.
  • a visible light indicating device is selectively configured to indicate the position of the OCT detection beam and guide the placement position of the detection target.
  • the method of the present invention introduces a multivariate time series model, describes the OCT signal with two characteristics of signal-to-noise ratio and decorrelation coefficient (ID), establishes an asymptotic relationship between the two characteristics, and demonstrates the asymptoticity through numerical simulation and theoretical deduction
  • ID signal-to-noise ratio and decorrelation coefficient
  • the method of the present invention is based on a multivariate time series model, supports the effectiveness and feasibility of the proposed classifier, and, compared with other methods, greatly eliminates the influence of motion artifacts caused by noise.
  • the present invention proposes a theoretical model for describing OCT signals using multivariate time series, and gives the asymptotic relationship between the two characteristics of the signal (signal-to-noise ratio and decorrelation coefficient), combining numerical simulation and theory
  • the existing method mainly based on a uniform intensity mask
  • the existing method can solve the problem of random noise bands by determining the boundary between the blood flow signal and other signals that are more in line with the actual signal distribution.
  • the problem of unsatisfactory suppression effect of coming artifacts improves the quality of blood flow images.
  • the present invention has the following significant advantages:
  • OCTA based on decorrelation calculation due to the dependence between the decorrelation coefficient of the OCT scattering signal and the signal-to-noise ratio, the decorrelation artifacts introduced by random noise in the low SNR area cannot be distinguished from the decorrelation caused by blood flow movement.
  • the common solution is to set an empirical threshold for intensity masking, which is equivalent to using an intensity (signal-to-noise ratio) threshold to classify blood flow signals and other signals in the ID feature space, and the signal-to-correlation coefficient and signal-to-noise ratio are more different.
  • the complex dependency relationship makes the boundary between the actual blood flow signal and other signals very different from these two straight lines, resulting in a higher misclassification rate.
  • the present invention proposes a mathematical model describing the characteristics of the OCT signal, and derives the asymptotic relationship of the ID characteristics of the signal; further analysis combined with numerical simulation gives a formula for calculating the variance of the asymptotic distribution, thus Its asymptotic distribution establishes a linear classifier in the ID feature space adapted to the signal-to-noise ratio, which can improve the classification accuracy and obtain higher blood flow contrast.
  • the effect of traditional intensity masks relies on empirical threshold selection, which lacks appropriate basis.
  • the classifier constructed by the present invention is based on a solid theoretical foundation and determined through a large number of numerical simulation verifications;
  • the decision parameters of the classifier are only determined by the average kernel, without other complicated calibration of other parameters of the system. And the calculation process is extremely efficient, which is an order of magnitude improvement compared to other methods based on machine learning;
  • the present invention proposes to establish the theoretical basis of the classifier, so it is more reliable; at the same time, since the static and noise regions are completely removed, the visibility and overall contrast of the angiogram under all signal-to-noise ratios can be improved , After a large number of samples, the performance is significantly better than traditional methods.
  • Figure 1 is a schematic diagram of the method of the present invention
  • Figure 2 is a schematic diagram of the device of the present invention.
  • Figure 3 is a schematic diagram of a device according to an embodiment of the present invention.
  • Figure 4 is an exemplary result diagram of the present invention on a blood flow template
  • Figure 4 (a) is the (strength) structure diagram of the blood flow template and the curve diagram of the change of the signal-to-noise ratio of the template with depth position;
  • Figure 4(b) is the decorrelation coefficient diagram of the blood flow template, and the dashed ellipse marks the motion artifacts caused by the low signal-to-noise ratio;
  • Figure 4(c) shows the projection of the voxels of the blood flow template in the reciprocal signal-to-noise ratio-decorrelation coefficient (ID) feature space, where darker points represent static voxels, and lighter points represent fast moving dynamic bodies The light-colored dots represent slow moving dynamic voxels.
  • the solid line shown in the dashed rectangular box is the asymptotic relationship between the signal-to-noise ratio and the decorrelation coefficient, the dashed line is the proposed classifier, the horizontal dashed line represents the traditional intensity threshold classifier, and the dashed ellipse represents the proposed method compared to the traditional The angiography method can distinguish more voxel parts;
  • Figure 4(d) is the angiogram of the proposed method (ID-OCTA) on the blood flow template, and the black dotted line is the depth map that can be imaged by the traditional angiography method on the template;
  • Figure 4(e) is an angiogram of the traditional angiography method (cmOCT) on the blood flow template;
  • Figure 4(f) is the receiver operating curve of the proposed classifier and the traditional intensity threshold classifier.
  • the curve marked by the diamond is the receiver operating curve of the proposed classifier, and the curve marked by the circle is The curve is the receiver operating curve of the traditional intensity threshold classifier;
  • Fig. 5 is an exemplary result diagram of a living human skin blood flow imaging experiment according to an embodiment of the present invention.
  • Figure 5(a) shows the result of the proposed angiography method (ID-OCTA) on the skin tissues of healthy humans
  • Figure 5(b) The contrast result of the traditional intensity threshold method (cmOCT) on the skin tissue of healthy human living body;
  • Figure 5(d) An angiographic cross-sectional view of the proposed angiography method (ID-OCTA) at the position of the dotted line;
  • Figure 5(e) An angiographic cross-sectional view of the traditional intensity threshold method (cmOCT) at the position of the dotted line;
  • Figure 5(f) A cross-sectional view of angiography at the dotted line using a higher intensity threshold
  • Figure 6 is a quantitative analysis comparison diagram of the present invention compared with the traditional method.
  • Figure 6(a) The blood vessel area of healthy human living skin tissue artificially marked for quantitative analysis, the light color indicates the blood vessel area, and the remaining black area is the background noise area map;
  • the proposed angiography method is a statistical histogram of the blood vessel and noise area of the angiography result on the skin tissue of healthy human living body;
  • Figure 6(e) The statistical histogram of the blood vessel and noise area of the contrast result of the traditional intensity threshold method (cmOCT) on the skin tissue of healthy human body;
  • the angiography method (ID-OCTA) proposed in FIG. 7 uses the angiography result graph on the skin tissue of healthy human body with the corrected signal-to-noise ratio and the uncorrected decorrelation coefficient.
  • Fig. 7(b) The proposed angiography method (ID-OCTA) uses the angiography result of the decorrelation coefficient without correction of the signal-to-noise ratio.
  • each operation is described as a plurality of discrete operations, but the order of description does not represent the order of implementing the operations.
  • the sample measurement space is represented by x-y-z three-dimensional coordinates based on the spatial direction.
  • the depth z direction is the direction along the incident optical axis
  • the x-y plane is the plane perpendicular to the optical axis, where x and y are orthogonal, and x represents the OCT transverse fast scanning direction, and y represents the slow scanning direction.
  • S, A, n, etc. represent variables, which are only used to facilitate discussion, and are not intended to limit the application of the embodiment of the present invention, and can be any values such as 1, 2, 3, etc.
  • the discussion on the multi-channel, polarization state and other average dimensions of the OCT system is omitted here, and only the space-time dimension is taken as an example.
  • the actual execution steps are the same as the following operations in the space-time dimension.
  • the method of the present invention is shown in Fig. 1.
  • the second is the signal classification part, which combines the intensity of the scattered signal and the time decorrelation coefficient to construct a two-dimensional feature space to realize the classification of dynamic blood flow signals and static tissue background2.
  • the decorrelation coefficient is obtained through decorrelation operation processing on the OCT scattering signal.
  • the decorrelation operation includes calculation of complex OCT scattering signals including amplitude and phase obtained by scanning S different time points, and the decorrelation coefficient is calculated. Among them, the correlation calculation of the complex signal can obtain higher motion contrast.
  • C represents the first-order autocovariance
  • I represents the zero-order autocovariance
  • D represents the decorrelation coefficient, as the OCTA blood flow information
  • * represents the conjugate of the complex number
  • X(m,t) is a certain space
  • the position (z, x) is the complex signal at time t
  • K represents the total number of high-dimensional average kernels in the xyz space when the decorrelation coefficient is calculated
  • k represents the high-dimensional average kernels in the xyz space when the decorrelation coefficient is calculated
  • S represents the total number of high-dimensional average cores in the time dimension t taken when the decorrelation coefficient is calculated, that is, the number of adjacent B-scan frames in the OCT scan
  • s represents the high value taken when the decorrelation coefficient is calculated
  • the above formula (2) is usually also the following formula (4) to calculate the intensity feature I (zero-order autocovariance):
  • the first-order and zero-order autocovariances of each voxel are calculated using the above formula, and averaged in various dimensions such as time and space to obtain the decorrelation value of all voxels in the scan volume of the entire scattering signal sample , Can improve the calculation speed.
  • the noise source is mainly shot noise, which is considered to be approximately constant in the entire scanning volume, which can be determined by scanning air data or calibrating the system in advance.
  • s 2 is the noise level of the OCT system.
  • the noise level is obtained by calculating the intensity of the data blank area in the OCT scattering signal or calibrating in advance for the OCT system.
  • r is an order autocorrelation coefficients A scattered light signal
  • the first-order autocorrelation coefficient ⁇ 1 represents OCT signal X
  • the reciprocal of the local ISNR of SNR the reciprocal of the local ISNR of SNR.
  • the decorrelation coefficient D and the intensity I obey an asymptotic normal distribution.
  • the noise at each position of the OCT system is considered to be consistent, and the intensity I and the signal-to-noise ratio are in a corresponding relationship. Therefore, the reciprocal of the signal-to-noise ratio and the decorrelation coefficient are It is an asymptotic linear relationship.
  • the image area of OCT scattering signal is divided into blood flow area, static area and noise area.
  • c v is the coefficient of variation of the decorrelation coefficient of each voxel
  • G represents the parameter of the inverse proportional relationship between the square of the coefficient of variation of the decorrelation coefficient and the high-dimensional average kernel size
  • G ⁇ 1.5 is a constant obtained by numerical simulation, and its Will not be affected by the noise level of OCT equipment.
  • a classifier 23 is constructed to classify the OCT scattered signal 24 in the two-dimensional feature space of the inverse SNR-decrelation coefficient.
  • the asymptotic distribution of D is normal.
  • the boundary of the asymptotic relationship between the signal-to-noise ratio and the decorrelation coefficient is obtained, and then the blood vessel mold is analogous.
  • the method uses the boundary as the classifier, that is, divides the two sides of the boundary into different classes, which are used to remove all static and noisy voxels.
  • D c represents the classification boundary of the decorrelation coefficient in the classifier.
  • the result of preliminary classification is processed by morphological filtering or Gaussian filtering from the perspective of graphics to obtain the final OCTA blood flow image, as shown in Figure 5(a).
  • D * represents the first-order decorrelation coefficient corrected according to the signal-to-noise ratio.
  • the decorrelation coefficient corrected by the signal-to-noise ratio is used as the motion metric to eliminate the influence of noise on the motion metric, as shown in Figure 7(a).
  • the present invention uses the decorrelation coefficient corrected by the signal-to-noise ratio as the motion metric according to formula (9) to eliminate the influence of noise on the motion metric, which can better remove the influence of noise. Reflect the intensity of blood flow changes over time. The effect of comparing the uncorrected decorrelation coefficient is shown in Figure 7.
  • FIG. 2 shows a schematic diagram of the structure of the acquisition device of the OCT blood flow contrast technology based on the ID feature space in the present invention.
  • the main structure of the low-coherence interferometry part of the device is an interferometer, which is composed of 11-23.
  • the light source 11 is connected to the input end on the side of the beam splitter 12, and the light emitted by the light source 11 is divided into two parts by the beam splitter 12: one of the light beams enters the reference arm of the interferometer through a polarization controller 13 and passes through the reference The arm collimator 14 irradiates the reference plane mirror 15; another beam of light enters the sample arm through another polarization controller 13, and specifically focuses on the sample 21 to be tested through the collimator lens 16 and the optical path of the scanning device.
  • the light beam is reflected by the two-dimensional scanning galvanometer group 17 and 18, the "4f" lens group 54, 55 and the dichroic mirror 19, and then is focused on the sample 21 to be tested by the objective lens 20.
  • the lens groups 54 and 55 are composed of two lenses 54 and 55 arranged on the same optical axis.
  • the lens groups 54 and 55 are designed to ensure the beam center of the two-dimensional scanning galvanometer mirror and the beam of the dichroic mirror during scanning. The center is fixed, so that the beam in the OCT sample arm does not affect the imaging characteristics of the objective lens during scanning.
  • the light reflected by each of the reference arm and the sample arm returns to the output of the beam splitter 12 through the original path, and is received by the interference signal detection device 22 after interference.
  • the interference signal detection device 22 is then connected to the signal processor module and the calculation unit 23 .
  • two polarization controllers 13 are used to adjust the polarization state of the light beam to maximize the signal interference effect.
  • the specific implementation is also provided with a visible light indicating device.
  • the visible light indicating device includes a low-power visible light source 25, a collimator lens 24, and a filter 52.
  • the visible light for indicating passes through the collimator lens 24, the filter 52, and the dichroic mirror in sequence. 19 and the focusing objective lens 20 to the sample 21 to be tested.
  • a feature space-based three-dimensional blood flow imaging system device shown in FIG. 2 specifically includes:
  • the light source 11 uses broadband low-coherence light
  • the plane mirror 15 can move along the optical axis
  • the interference signal detection device 22 is a one-point detector.
  • the interference signals of the two arms are detected by the point detector 22, and the low-coherence interference detection of the scattered signals in the z-direction at a certain spatial depth, so as to obtain the depth spatial dimension sample volume .
  • the light source 11 uses broadband low-coherence light
  • the plane mirror 15 is fixed
  • the interference signal detection device 22 uses a spectrometer.
  • the interference signal passes through the line camera in the spectrometer 22 while recording the interference spectrum.
  • the Fourier analysis method is used to analyze the interference spectrum signal, and the scattering information in the depth z direction is obtained in parallel, so as to obtain the sample volume of the depth space dimension.
  • the light source 11 is a swept frequency light source
  • the plane mirror 15 is fixed
  • the interference signal detection device 22 is a point detector.
  • the point detector 22 records the low-coherence interference spectrum of the swept-frequency light source in time-division. Sampling Fourier analysis of the interference spectrum signal, parallel acquisition of the depth z direction scattering information, so as to obtain the depth space dimension of the sampling volume.
  • the OCT scanning imaging method involved in the description of Figure 1 can be combined to analyze the relative movement of blood flow and surrounding tissues to generate OCTA blood flow motion contrast, and perform graphic filtering on the OCTA blood flow contrast image to generate Enhanced OCTA blood flow movement imaging.
  • FIG. 3 shows an exemplary embodiment using the present invention.
  • a three-dimensional OCT angiography system based on ID feature space including a broadband low-coherence light source 26, an optical circulator 27, a fiber coupler 28 with a splitting ratio of 50:50, a first polarization controller 29, a first fiber collimator 30, Focusing lens 36, plane mirror 37, second polarization controller 38, second fiber collimation device 39, two-dimensional scanning galvanometer combination 40 and 41, dichroic mirror 42, focusing objective lens 43, third fiber collimation device 45.
  • the broadband low-coherence light source 26 shown in this example uses a super-luminescent diode light source with a center wavelength of 1325nm and a bandwidth of 100nm.
  • the focusing objective 43 uses an achromatic double cemented lens with a focal length of 30mm.
  • the high-speed line scan camera 48 uses a 2048 pixel unit. Line scan camera.
  • the light emitted by the low-coherence broadband light source 26 used in this example device passes through the optical circulator 27 and enters the input end of the fiber coupler 28 with a splitting ratio of 50:50.
  • the light emitted from the fiber coupler 28 is Divided into two sub-beams: one of them is connected to the first fiber collimating device 30 in the reference arm through the first polarization controller 29 through an optical fiber, and then irradiates the plane mirror 37 after passing through the collimating and focusing lens 36;
  • the beam of light is connected to the second fiber collimating device 39 of the sample arm through the second polarization controller 38 through the optical fiber, and after collimation, it passes through the two scanning galvanometers 40, 41, the "4f" lens group 56, 57 and the dichroic After the mirror 42 is reflected, it is focused on the sample 44 by the focusing objective 43 and scattered by back reflection.
  • the lens groups 56 and 57 are designed to ensure the beam center of the two-dimensional scanning galvanometer mirror and the reflection of the dichroic mirror during scanning.
  • the beam center of the surface is fixed and does not change.
  • the light reflected by the plane mirror 37 in the reference arm interferes with the backscattered light of the sample under test in the sample arm at the fiber coupler 28, and the interference light is detected and recorded by the spectrometer 45 ⁇ 48, and then the signal processor module and The calculation unit 49 collects and performs signal analysis and processing.
  • the spectrometer includes devices 45 to 48 connected in sequence, wherein device 45 is a fiber coupler, 46 is a grating, and 47 is a converging lens, which focuses the light separated by the grating dispersion onto the linear array detector shown in 48.
  • the specific implementation is also provided with a visible light indicating device.
  • the visible light indicating device includes a visible light indicating light source 50 and a collimating lens 51.
  • the visible light indicating light source 50 emits visible light for indicating after passing through the collimating lens 51, the dichroic mirror 42 and the focusing objective lens 43.
  • the sample to be tested 44 is also provided with a visible light indicating device.
  • the visible light indicating device includes a visible light indicating light source 50 and a collimating lens 51.
  • the visible light indicating light source 50 emits visible light for indicating after passing through the collimating lens 51, the dichroic mirror 42 and the focusing objective lens 43.
  • the sample to be tested 44 is also provided with a visible light indicating device.
  • Figure 4 shows the experimental results of using the blood flow template (simulation) obtained in this embodiment.
  • Figures 4(a) and 4(b) respectively correspond to the representative cross-sectional OCT structure diagram and OCTA blood flow image of the phantom.
  • Figure 4(a) indicates the static, dynamic and noise signal areas, and the phantom is imaged
  • the classification characteristics of known signals, such as data facilitate the quantification of classification effects.
  • Figure 4(c) projects the static, dynamic, and noise signals in the reciprocal signal-to-noise ratio-decorrelation coefficient (ID) feature space.
  • ID reciprocal signal-to-noise ratio-decorrelation coefficient
  • TPR true positive rate
  • FPR false positive rate
  • the proposed ID feature classifier can image deeper regions under the same false positive rate, and the difference in depth between the two can be clearly indicated by the black dotted line. Using these two indicators as two coordinate axes, the receiver operating characteristic curve (ROC) is drawn.
  • the TPR is 80%
  • the ID-OCTA method only has a TPR of 9.9%, but the traditional intensity threshold method reaches 75.7%.
  • FIG. 5 shows the results of blood flow imaging of living human skin using this embodiment.
  • Figure 5(a) is the result of blood flow contrast using the proposed ID-OCTA method
  • Figure 5(b) is the traditional intensity threshold method, in which the threshold is set to approximately twice the system noise level.
  • the superficial large blood vessels have high blood flow velocity and large decorrelation coefficient, so they can be clearly observed under the two methods.
  • the traditional intensity threshold cannot clear a large number of static areas, resulting in high background noise in blood flow imaging. This makes it difficult to see the tiny capillaries. As shown in the area shown in the box and its enlarged detail, the background full of noise reduces the visibility of some capillaries.
  • Figure 5(c) shows the results of increasing the intensity threshold.
  • the intensity threshold By increasing the intensity threshold to about six times the noise level, the background noise of the resulting angiogram can be suppressed to a level similar to the proposed ID-OCTA method. But obviously it also excludes a lot of blood vessel areas with low signal-to-noise ratio.
  • the low intensity threshold method causes many static signals to be misclassified as blood flow signals and remains, which weakens the contrast of tiny capillaries; while the high intensity threshold error excludes a large number of main concentrations Blood vessels with low signal-to-noise ratio in deep regions.
  • Figure 5 (d), (e) and (f) are the cross-sectional angiograms of Figure 5 (a), (b) and (c), respectively, with arrows pointing out some detailed comparisons.
  • the blood vessel in Figure 4(f) at the yellow arrow was mistakenly removed as noise by the high-intensity threshold, while the low-intensity threshold error in Figure 5(e) left a large number of noise regions. Therefore, through the comparison of the in vivo imaging results shown in Fig. 5, the blood flow image obtained by the method proposed by the present invention and the result of the traditional method have a significant enhancement in the visibility of tiny capillaries, and the whole angiogram has Significant contrast improvement.
  • the present invention uses contrast-noise-ratio (CNR) to measure the improvement of the proposed algorithm.
  • CNR contrast-noise-ratio
  • Figure 7 shows the angiographic results of the decorrelation coefficient after the signal-to-noise ratio is corrected.
  • the residual noise can be corrected more thoroughly and the contrast of the angiogram can be further improved;
  • Areas with signal-to-noise ratio (such as deep blood vessels) can show more realistic motion contrast, help users make correct interpretations of the results, and help subsequent hemodynamic quantification work.
  • the above experimental comparison results fully illustrate: the use of the feature space-based three-dimensional blood flow imaging method of the present invention can improve the accuracy of blood flow signal classification, achieve effective enhancement of blood flow contrast, and improvement of blood flow image quality. Its prominent technical effect.

Landscapes

  • Health & Medical Sciences (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Molecular Biology (AREA)
  • General Health & Medical Sciences (AREA)
  • Veterinary Medicine (AREA)
  • Public Health (AREA)
  • Animal Behavior & Ethology (AREA)
  • Biophysics (AREA)
  • Pathology (AREA)
  • Biomedical Technology (AREA)
  • Heart & Thoracic Surgery (AREA)
  • Medical Informatics (AREA)
  • Surgery (AREA)
  • Physiology (AREA)
  • Artificial Intelligence (AREA)
  • Signal Processing (AREA)
  • Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
  • Psychiatry (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Radiology & Medical Imaging (AREA)
  • Cardiology (AREA)
  • Evolutionary Computation (AREA)
  • Fuzzy Systems (AREA)
  • Mathematical Physics (AREA)
  • Hematology (AREA)
  • Vascular Medicine (AREA)
  • Investigating Or Analysing Materials By Optical Means (AREA)
  • Measuring Pulse, Heart Rate, Blood Pressure Or Blood Flow (AREA)
  • Eye Examination Apparatus (AREA)

Abstract

一种基于特征空间的光学相干层析的三维血流造影方法及系统，通过采集器（1）收集三维空间内散射信号样本的OCT散射信号；通过理论建立的分类器（2）结合OCT散射信号的局部信噪比和去相关系数构建二维特征空间，实现动态血流信号与静态组织的分类，具体包括：采用一阶和零阶自协方差对OCT散射信号计算分析，得到各OCT散射信号信噪比和去相关的两个特征；构建信噪比倒数-去相关系数二维特征空间；基于多元时间序列理论，构建ID空间的线性分类器，去除静态周围组织背景。基于特征空间的光学相干层析的三维血流造影方法及系统，能显著抑制系统噪声对血流成像的影响，提高血流图像的对比度，特别是深层组织的血管可见度，提高血流量化的准确度。

Description

基于特征空间的光学相干层析的三维血流造影方法及系统 技术领域
本发明大体涉及生物医学成像领域，且更具体地涉及光学相干层析成像技术(Optical Coherence Tomography，OCT)和光学相干层析血流造影术(OCT Angiography，OCTA)相关联的血流造影方法和基于时间序列模型的血流分类方法。
背景技术
血流是衡量生理功能和病理状态的重要指标，目前在临床上常用的血管造影技术需要静脉注射外源性标记物，可能引发的副作用使其不适合对人体血流进行长期的、频繁的跟踪检测以及部分病人的身体状况。近年来，以光学相干层析技术为基础发展起来的血管造影技术OCTA，以内源性的血流运动代替传统的外源荧光标记物，其非侵入性、无标记的特点，以及对生物组织内的微血管网络进行清晰、可靠的三维成像的能力，使得该技术在被发明以来得到了很快的发展，并在眼底、脑皮层血管和皮肤成像的研究中得到了应用。
为了获取OCTA血流图像，通常需要在生物组织的每个空间位置、以一定的时间间隔进行重复采样(重复的A线扫描或B帧扫描)，通过分析OCT散射信号的时间动态来量化每个位置处的运动强度，根据量化得到的运动强度来对血流信号和静态组织信号进行分类。目前已报道的OCTA的运动度量指标，主要是基于相邻的A线扫描间(或相邻B扫描帧间)的差分、方差或去相关计算。其中基于去相关计算的OCTA运动度量指标，由于其同时利用信号的强度和相位信息，因此分类结果的可靠性更高，受多普勒角度和光源整体强度变化的影响更小。同时，由于去相关衡量的是相邻B扫描帧间的相似度，因此受整体光源强度变化的影响小。
但是，OCTA的血管流动的运动指标，特别是去相关系数的运动指标，对原始的OCT散射信号的噪声水平具有显著的依赖性。随着信号强度的衰减(如在深层组织区域)，随机噪声将逐渐占据主要成分，其时变特性与真实血流区域相似，因此导致去相关伪影。在去相关系数运动指标上，无法区分噪声的随机性和血红细胞的运动导致的去相关，因此信噪比较弱的区域容易被误判为血流信号区域，严重影响血流图像的对比度。常见的解决办法是设置一个经验性的强度阈值，生成强度掩膜来去掉所有低信噪比的信号。但是，由于去相关系 数和信号强度间存在着复杂的依赖关系，简单的强度掩膜会导致高分类错误率和低运动对比度。
现有的一些OCTA分类方法大多缺少较强的理论支撑，受到模版参数与实际生物组织差异的干扰，不能明显地区分动静态信号的准确分类；或者以复杂的估计量修正信噪比对去相关的影响，运算复杂且无法去除静态区域导致的背景噪声。
发明内容
为了解决背景技术中存在的问题，针对现有技术的不足，本发明提出了一种基于特征空间的光学相干层析的三维血流造影方法及系统，利用OCT信号的信噪比和去相关特征的，采用多元时间序列模型进行信噪比适应的三维血流成像，能够精确地区分OCT图像中血流和混有噪声的静态区域。
本发明的目的是通过如下技术方案实现的：
包括采集器步骤，用于收集三维空间内散射信号样本的OCT散射信号；
包括分类器步骤，结合OCT散射信号的信噪比与去相关系数构建二维特征空间实现动态血流信号与静态组织的分类，进而实现利用信噪比的修正；
包括成像造影步骤，基于分类器的分类结果生成血流造影图。
所述的散射信号样本为生物组织样本，生物组织样本例如可以为人和动物皮肤、脑组织、眼睛。
所述用于收集三维空间内散射信号样本的OCT散射信号，具体包括：对散射信号样本进行三维空间的OCT扫描成像，相同空间位置及其附近位置在S个不同时间点重复采样以及使用多通道进行重复采样。
多通道是指OCT光源发出光的多偏振态、多扫描入射角度等。
所述对散射信号样本进行三维空间的OCT扫描成像，相同空间位置及其附近位置在S个不同时间点重复采样，采用以下方式之一：
通过扫描改变参考臂光程的时间域OCT成像方法；
利用光谱仪记录光谱干涉信号的光谱域OCT成像方法；
利用扫频光源记录光谱干涉信号的扫频OCT成像方法。
所述结合OCT散射信号的信噪比与去相关系数构建二维信噪比倒数-去相关性系数(ID)特征空间，实现动态血流信号与静态组织的分类，具体包括：
S1、采用一阶和零阶自协方差对OCT散射信号计算分析，并在时间、空间、通道等多个维度上作滑动平均或高斯平均，得到各OCT散射信号的信噪比和去相关系数的两个特征；
S2、建立信噪比-去相关系数特征空间，将体素映射至信噪比-去相关系数(ID) 特征空间；用多元时间序列模型建立线性分类器，在信噪比倒数-去相关系数特征空间中将OCT散射信号分为动态血流信号和静态组织噪声信号，并使用经信噪比修正替换去相关系数生成血管造影图。
所述的OCT散射信号分为动态血流信号和噪声与静态组织信号，噪声信号即为无样品区域的信号，动态血流信号和静态组织信号分别对应血流区域和静态组织区域。
所述步骤S1具体为：
采用一阶和零阶自协方差对OCT散射信号计算分析，并在三维空间、时间、通道等多个维度上做滑动平均或高斯平均，得到各OCT散射信号的强度和去相关系数的两个特征，如图4(a)和4(b)所示。
所述去相关系数的特征包括：对同一空间位置S个不同时间点扫描得到的OCT散射信号的幅度部分或对包含幅度和相位的OCT散射信号计算去相关系数，作为各OCT散射信号的OCTA血流对比度，即作为OCTA血流信息。
去相关运算中，按照高维平均核(包含时间、空间、偏振、通道等维度)大小，在特征空间建立图4(c)的分类器；根据分类器的分类结果生成血流造影如图4(d)。
所述的多元时间序列模型的具体定义为：OCT散射信号采集到的信号X(m,t)是由生物组织样本的后向散射光信号A(m,t)和随机噪声n(m,t)叠加而成，其中m,t分别指体素的空间位置和时间。一般地，时间序列A,n建立如下条件：1、n(m,t)是空间上互相独立的白噪声，并且与后向散射光信号A(m,t)不相关；2、A(m,t)是空间上相互独立同分布的平稳、遍历时间序列，并且其期望存在，具有有限的方差(二阶矩)；3、A(m,t)在时间上部分相关，即一阶自协方差存在，一阶自相关系数k的模的取值范围为[0,1]。
所述步骤S2具体为：结合OCT散射信号强度与去相关系数构建信噪比倒数-去相关系数(ID)特征空间，根据多元时间序列模型通过数值模拟计算OCT散射信号的信噪比-去相关系数的渐近关系(式(5))，将OCT散射信号投影到信噪比倒数-去相关系数(ID)特征空间中，如图4(c)所示，获得静态区域和噪声区域中体素分布(式(6))及其统计特征(式(7，8))。
具体实施中，根据多元时间序列模型，通过数值模拟计算OCT散射信号的信噪比-去相关系数的渐近关系。在模拟过程中，根据通常认为的瑞利分布对A的强度和均匀分布对其相位进行模拟，并设定白噪声为高斯分布的复随机噪声，以此生成数张模拟图片，将得到的最终模拟信号投影在信噪比-去相关系数特征空间上。通过计算模拟数据的变异系数，根据模型，去相关的方差与平均参数 成反比例关系，拟合其样本平方变异系数和平均参数可以得到反比例函数的参数。进而，根据该模型推导的线性分类器参数由模拟得到的参数和噪声水平决定。在信噪比-去相关系数特征空间中将OCT散射信号分为动态血流信号和静态组织噪声信号。对于分离出来的血流信号，使用去相关值进行显示，或者直接将其置为1进行显示。
在信噪比倒数-去相关系数二维特征空间中对OCT散射信号进行分类(23)，包括直接的线性分类器及其改进版本，其中，线性分类器，是指：通过理论公式，根据算法执行时的多维平均核大小(包括空间上，时间上的大小以及通道数目)，直接在二维特征空间得到线性分类器，并且该分类器具有普适性，在与其他参数无关。改进分类器是指在前述线性分类器的基础上，改变分类器的斜率或截距(整体旋转或平移)调整得到的线性分类器。
所述基于分类器的分类结果生成血流造影图(25)，分类结果中已经将带有噪声信号的静态组织信号分去除，剩余的信号作为动态血流信号，使用其去相关系数值形成血流造影图，如图3。
所述基于分类器的分类结果生成血流造影图(25)之后进一步包括：在得到基于分类器的分类结果的血流造影图后，进行高斯滤波、中值滤波、Hessian滤波等操作去除噪声，提升血流造影图效果。
二、一种基于特征空间的三维血流成像系统：
一OCT光学相干层析探测装置，用于对三维空间内的散射信号样本的OCT散射信号进行采集；
一个或多个信号处理器，用于获取并分析OCT散射信号的强度信息和OCTA血流信息，并结合散射信号的强度信息和OCTA血流信息对动态血流信号和静态组织背景进行分类。
所述的一OCT光学相干层析探测装置是采用以下的一种：
包括低相干光源、干涉仪和探测器，低相干光源和探测器分别连接到干涉仪；
或者包括低相干光源、干涉仪和光谱仪，低相干光源和光谱仪分别连接到干涉仪；
或者包括扫频宽光谱光源、干涉仪和探测器，扫频宽光谱光源和探测器分别连接到干涉仪。
所述的一OCT光学相干层析探测装置中选择地配置一个可见光指示装置，用于指示OCT探测光束的位置，指导探测目标的放置位置。
本发明方法中引入多元时间序列模型，以信噪比和去相关系数(ID)两个 特征描述OCT信号，建立了两个特征之间的渐近关系，并通过数值模拟和理论推导表明渐近分布的方差仅由高维平均核大小和局部信噪比决定，进而建立起简单易行的，信噪比适应的ID特征空间上的分类器。
本发明方法以多元时间序列模型为基础，支撑了所提出的分类器的有效性和可行性，并且相比于其他方法，极大的消除了噪声产生的运动伪影的影响。
本发明的有益效果和创新点如下：
对比已有技术，本发明提出了使用多元时间序列描述OCT信号的理论模型，并给出了信号的两个特征(信噪比和去相关系数)之间的渐近关系，结合数值模拟与理论基础，建立信噪比适应的分类器，通过对于更符合实际信号分布的血流信号和其他信号的分界线的确定，来解决现有方法(以均一的强度掩膜为主)对随机噪声带来的伪影的抑制效果不理想的问题，提升血流图像的质量。
本发明对比已有技术具有以下显著优势：
1.基于去相关计算的OCTA，由于OCT散射信号的去相关系数和信噪比间的依赖关系，低信噪比区域随机噪声引入的去相关伪影无法和血流运动引入的去相关区分。常用解决办法是设置经验性阈值进行强度掩膜，相当于在ID特征空间中用一个强度(信噪比)阈值来分类血流信号和其他信号，而信号的去相关系数和信噪比间更复杂的依赖关系使得实际血流信号和其它信号的分界与这两条直线有很大差异，造成较高的误分类率。本发明基于多元时间序列理论，提出了描述OCT信号特征的数学模型，并推导出了信号的ID特征的渐近关系；进一步的分析结合数值模拟给出了计算渐近分布方差的公式，从而根据其渐近分布建立一信噪比适应的ID特征空间线性分类器，因此能够提高分类准确率，获得更高的血流对比度。
2.传统的强度掩膜的效果依赖于经验性的阈值选取，缺乏合适的依据。本发明构建的分类器是基于坚实的理论基础，并通过大量的数值模拟验证而确定的；
3.分类器的决定参数仅由平均核决定，无需其他复杂的对于系统其他参数的标定。并且计算过程效率极高，相比于其他基于机器学习的方法有着数量级的提升；
3.对比现有的方法，本发明提出建立分类器的理论依据，因此更加可靠；同时由于完全去除了静态和噪声区域，能够提升血管造影图在所有信噪比下的可视度和整体对比度，经过大量样本验证表现显著优于传统方法。
附图说明
图1为本发明方法的示意图；
图2为本发明装置的示意图；
图3为本发明实施例的装置的示意图；
图4为本发明在血流模版上的示例性结果图；
图4(a)为血流模版的(强度)结构图以及该模版的信噪比随深度位置的变化曲线图；
图4(b)为血流模版的去相关系数图，虚线椭圆标记了由于低信噪比导致的运动伪影图；
图4(c)为血流模版的体素在信噪比倒数-去相关系数(ID)特征空间的投影情况，其中，深色点表示静态体素，较浅的点代表快速运动的动态体素，浅色点代表慢速运动的动态体素。虚线矩形框内所示的实线是信噪比-去相关系数的渐近关系，虚线是所提出的分类器，水平虚线代表传统的强度阈值分类器，虚线椭圆表示所提出的方法相比传统血管造影方法所能区分更多的体素部分；
图4(d)为所提出的方法(ID-OCTA)在血流模版上的的血管造影图，黑色虚线是在该模版上传统血管造影方法所能成像的深度图；
图4(e)为传统血管造影方法(cmOCT)在血流模版上的血管造影图；
图4(f)为所提出的分类器与传统的强度阈值分类器的受试者工作曲线图，其中菱形标识所在的曲线为所提出的分类器的受试者工作曲线，圆形标识所在的曲线为传统的强度阈值分类器的受试者工作曲线图；
图5为本发明实施例的活体人类皮肤血流成像实验示例性结果图；
图5(a)为所提出的血管造影方法(ID-OCTA)在健康人体活体皮肤组织上的造影结果图；
图5(b)传统的强度阈值方法(cmOCT)在健康人体活体皮肤组织上的造影结果图；
图5(c)使用较高的强度阈值在健康人体活体皮肤组织上的造影结果；
图5(d)所提出的血管造影方法(ID-OCTA)在虚线位置处的血管造影截面图；
图5(e)传统的强度阈值方法(cmOCT)在虚线位置处的血管造影截面图；
图5(f)使用较高的强度阈值在虚线位置处的血管造影截面图；
图6为本发明相比传统方法的量化分析对比图。
图6(a)用于量化分析而人工标记的健康人体活体皮肤组织血管区域，浅色表示的为血管区域，其余黑色区域为背景噪声区域图；
图6(b)所提出的血管造影方法(ID-OCTA)在健康人体活体皮肤组织上的造影结果图；
图6(c)传统的强度阈值方法(cmOCT)在健康人体活体皮肤组织上的造影结果；
图6(d)所提出的血管造影方法(ID-OCTA)在健康人体活体皮肤组织上的造影结果的血管与噪声区域的统计直方图；
图6(e)传统的强度阈值方法(cmOCT)在健康人体活体皮肤组织上的造影结果的血管与噪声区域的统计直方图；
图7所提出的血管造影方法(ID-OCTA)使用经信噪比修正和未经修正的去相关系数的在健康人体活体皮肤组织上的血管造影结果图。
图7(a)所提出的血管造影方法(ID-OCTA)使用经信噪比修正的去相关系数的血管造影结果图；
图7(b)所提出的血管造影方法(ID-OCTA)使用未经信噪比修正的去相关系数的血管造影结果图。
图中：1-一种采集器；2-一种分类器；21-得到OCTA信号特征；22-将OCT散射信号投影在信噪比倒数-去相关系数(ID)二维特征空间中；23-构建分类器在(ID)特征空间中对OCT散射信号进行分类，其中231-根据(5)式数值地计算静态总体的分布方差，232-建立线性分类器，24-根据建立的分类器对数据进行分类；25-根据分类器的分类结果生成血流造影图。
具体实施方式
下面将结合附图对本发明的具体实施方式做详细说明，附图形成本发明的一部分。需要注意的是，这些说明及示例仅仅为示例性，不能被理解为限制了本发明的范围，本发明的保护范围由随附的权利要求书限定，任何在本发明权利要求基础上的改动都是本发明的保护范围。
本发明的实施例如下：
为了便于理解本发明的实施例，将各操作描述成多个离散的操作，但是描述的顺序不代表实施操作的顺序。
本描述中针对样品测量空间采用基于空间方向的x-y-z三维坐标表示。这种描述仅仅用于促进讨论，而不意欲限制本发明的实施例的应用。其中：深度z方向为沿入射光轴的方向；x-y平面为垂直于光轴的平面，其中x与y正交，且x表示OCT横向快扫描方向，y表示慢扫描方向。
上述S，A，n等表示变量，仅仅用于促进讨论，而不意欲限制本发明的实施例的应用，可以是1,2,3等任一数值。为了表述简便起见，此处略去对于OCT系统多通道、偏振态等其他平均维度的讨论，仅以时空维度为示例。其实际执行步骤与下述在时空维度上的操作相同。
本发明方法如图1所示，首先为信号采集部分，对组织样本进行OCT三维扫描成像，相同或相邻空间位置在T个不同时间点重复采样1。其次为信号分类部分，结合散射信号强度与时间去相关系数构建二维特征空间，实现动态血流信号与静态组织背景的分类2。
其具体步骤是：
1)分析血流和周围组织的相对运动，得到各OCT散射信号的OCTA血流ID特征(强度和去相关系数)21。
去相关系数是对OCT散射信号通过去相关运算处理获得，去相关运算包括对S个不同时间点扫描得到的包含幅度和相位的复数OCT散射信号计算，计算获得去相关系数。其中对复数信号的相关计算可以获得更高的运动对比度。
对于血流和周围组织中的某一局部区域，针对每个体素用其相邻的S个OCT扫描的B扫描帧(x-z平面)进行平均计算(即使用高维平均核进行卷积)获得每个体素的一阶与零阶自协方差和去相关系数：
Figure PCTCN2019083491-appb-000001
Figure PCTCN2019083491-appb-000002
Figure PCTCN2019083491-appb-000003
其中，C表示一阶自协方差，I表示零阶自协方差，作为强度，D表示去相关系数，作为OCTA血流信息；*表示复数的共轭，X(m,t)是某一个空间位置(z,x)在时刻t的复数信号；K表示进行去相关系数计算时所取高维平均核在x-y-z空间的总数；k表示进行去相关系数计算时所取高维平均核在x-y-z空间的序数；S表示进行去相关系数计算时所取高维平均核在时间维度t上的总数，即为OCT扫描的相邻B扫描帧的帧数量；s表示进行去相关系数计算时所取高维平均核在时间维度t上的序数。
在计算中，上式(2)通常也下式(4)计算强度特征I(零阶自协方差)：
Figure PCTCN2019083491-appb-000004
计算过程中，先对于每个体素采用上述公式计算其一阶和零阶自协方差，在时间和空间等各个维度上进行平均从而得到整个散射信号样品的扫描体积内所有体素的去相关值，能够提升运算速度。
2)在OCT系统中，噪声来源主要为散粒噪声，视为在整个扫描体积内近似不变，可以通过扫描空气的数据或事先对系统进行标定而确定。
然后采用以下公式计算每个体素的局部信噪比SNR，作为信噪比：
Figure PCTCN2019083491-appb-000005
其中，s 2是OCT系统的噪声水平。
通过计算OCT散射信号中数据空白区域的强度或者针对OCT系统事先标定得到噪声水平。
3)结合上述OCT探测得到的信噪比和基于去相关计算得到的OCTA血流信息建立信噪比-去相关系数(ID)二维特征空间，并将OCT散射信号投影在特征空间中22。
根据通过时间序列模型的结果，获得如下公式表示的信噪比和去相关系数之间的渐近关系，构建信噪比倒数-去相关系数二维特征空间：
Figure PCTCN2019083491-appb-000006
其中，r是后向散射光信号A的一阶自相关系数，ρ 1表示OCT信号X的一阶自相关系数，iSNR表示局部信噪比SNR的倒数。→表示收敛，平均体素的数量足够多能实现收敛。
此外，去相关系数D和强度I服从渐近正态分布，本发明中认为OCT系统的各个位置噪声保持一致，强度I和信噪比成对应关系，因此信噪比倒数和去相关系数之间为渐近线性关系。
特别的，对于静态和噪声区域，r＝1则
D→iSNR(6)
即对于静态与噪声体素，其局部信噪比的倒数收敛于去相关系数。OCT散射信号的图像区域分为血流区域、静态区域和噪声区域。
并且特别地，对于静态区域和噪声区域，通过模拟计算给出渐近分布的方差水平，且其只与局部信噪比SNR和高维平均核大小N＝S×K决定。
采用以下公式计算获得每个体素的去相关系数的变异系数c v：
Figure PCTCN2019083491-appb-000007
其中，c v是每个体素的去相关系数的变异系数，G表示去相关系数的变异系数平方-高维平均核大小的反比例关系的参数，G≈1.5是通过数值模拟得到的常 数，并且其不会随OCT设备噪声水平的影响。
4)之后构建分类器23，在信噪比倒数-去相关系数二维特征空间中对OCT散射信号进行分类24。
在时间序列理论中证明了D的渐近分布是正态的，一般按照拉依达准则(3σ原则)，获得信噪比和去相关系数之间的渐近关系的边界，再类比血管铸型方法以边界作为分类器，即将边界两侧分为不同类，用于将静态和噪声的体素全部去除。
建立以下公式表示的分类器计算获得分类边界：
D c＝(1+3c v)iSNR  (8)
其中，D c表示分类器中去相关系数的分类边界。
具体是中如图4(a)所示，原点位于左上角情况下，分类边界上方的区域为带有噪声的静态组织，分类边界下方的区域为动态血流。
并且该范围可以由使用者对于分类器性能的不同要求所调整。对初步分类的结果，再从图形角度进行形态学滤波或Gaussian滤波等处理,得到最终的OCTA血流图像，如图5(a)所示。
此外，根据式(4)知道对于血流区域，去相关系数随着信号的信噪比而变化，信噪比越低，去相关系数越大。
5)分类后，采用以下公式计算后向散射光信号A(m,t)的一阶去相关系数，并替代去相关系数，能消除噪声的影响，实现利用信噪比修正去相关系数：
Figure PCTCN2019083491-appb-000008
其中，D *表示根据信噪比修正后的一阶去相关系数。
根据式(9)，以经信噪比修正过后的去相关系数作为运动度量，以消除噪声对运动度量的影响，如图7(a)。
相比于传统的去相关系数，本发明根据式(9)以经信噪比修正过后的去相关系数作为运动度量，以消除噪声对运动度量的影响，能更好的去除噪声的影响，真实地反映血流时间上变化的强度。其对比未经修正的去相关系数的效果如图7所示。
图2示出的是本发明中基于ID特征空间的OCT血流造影技术的采集装置结构示意图。该装置的低相干干涉测量部分的主体结构为一干涉仪，由11～23构成。光源11连接到分束器12一侧的输入端，光源11发出的光被分束器12分成两部分光束：其中的一束光经一偏振控制器13进入到干涉仪的参考臂，通 过参考臂准直镜14照射于参考的平面反射镜15上；另一束光经另一偏振控制器13进入到样品臂，具体经过准直透镜16和扫描装置光路聚焦到待测样品21上。其中扫描装置光路中，光束经过二维扫描振镜组17、18，“4f”透镜组54、55和二向色镜19的反射后，经过物镜20聚焦在待测样品21上。其中透镜组54、55是由两个透镜54、55同光轴布置组成，透镜组54、55的设计是为了确保扫描时二维扫描振镜镜面的光束中心和二相色镜反射面的光束中心固定不变，使得OCT样品臂中的光束在扫描时不影响物镜的成像特性。
然后，参考臂和样品臂各自反射回的光经原路返回到分束器12输出，发生干涉后由干涉信号探测装置22接收，干涉信号探测装置22再连接到信号处理器模块与计算单元23。对于光纤型光路，采用两个偏振控制器13调整光束的偏振态，最大化信号干涉效果。
具体实施还设置有可见光指示装置，可见光指示装置包括低功率可见光光源25、准直透镜24和滤光片52，用于指示的可见光依次经过准直透镜24、滤光片52、二向色镜19和聚焦物镜20后到待测样品21。
依据低相干干涉探测信号的不同方式，图2所示的一种基于特征空间的三维血流成像系统装置具体包括：
1)时间域测量装置。光源11采用宽带低相干光，平面反射镜15可沿光轴方向移动，干涉信号探测装置22为一点探测器。通过移动平面反射镜15改变参考臂光程，两臂的干涉信号由点探测器22探测到，对某一空间深度的z方向的散射信号的低相干干涉探测，从而得到深度空间维度的采样体。
2)光谱域测量装置。光源11采用宽带低相干光，平面反射镜15固定不动，干涉信号探测装置22采用光谱仪。干涉信号经过光谱仪22中的线阵相机同时记录干涉光谱。采用傅里叶分析方法分析干涉光谱信号，并行获取深度z方向的散射信息，从而得到深度空间维度的采样体。
3)扫频测量装置。光源11采用扫频光源，平面反射镜15固定不动，干涉信号探测装置22采用点探测器。点探测器22分时记录扫频光源的低相干干涉光谱。采样傅里叶分析干涉光谱信号，并行获取深度z方向的散射信息，从而得到深度空间维度的采样体。
对于上述不同的测量装置，可分别结合图1叙述中所涉及的OCT扫描成像方式，分析血流与周围组织的相对运动生成OCTA血流运动造影，并对OCTA血流造影图像进行图形滤波，生成增强的OCTA血流运动造影。
图3示出的是利用本发明的一个示例性实施例。基于ID特征空间的三维OCT血管造影系统，包括宽带低相干光源26、光环形器27、分光比为50:50的 光纤耦合器28、第一偏振控制器29、第一光纤准直器件30、聚焦透镜36、平面反射镜37、第二偏振控制器38、第二光纤准直器件39、二维扫描振镜组合40和41、二向色镜42、聚焦物镜43、第三光纤准直器件45、光栅46、聚焦透镜47、高速线阵相机48、信号处理器模块与计算单元49、可见光指示光源50、准直透镜51、“4f”透镜组56和57。
本示例所示的宽带低相干光源26采用中心波长为1325nm、带宽为100nm的超发光二极管光源，聚焦物镜43采用焦距为30mm的消色差双胶合透镜，高速线阵相机48采用由2048像素单元组成的线阵扫描相机。其中由本示例装置所使用的低相干宽带光源26发出的光，经过光环行器27后进入到分光比为50:50的光纤耦合器28一侧的输入端，从光纤耦合器28出射的光被分成两部分子光束：其中一束光通过光纤经过第一偏振控制器29连接至参考臂中的第一光纤准直器件30，经过准直和聚焦透镜36后照射到平面反射镜37；另一束光通过光纤经过第二偏振控制器38连接至样品臂部分的第二光纤准直器件39，准直后经过两个扫描振镜40、41、“4f”透镜组56、57和二向色镜42反射后，由聚焦物镜43聚焦到被测样品44上并背向反射散射，其中透镜组56、57的设计是为了确保扫描时二维扫描振镜镜面的光束中心和二相色镜反射面的光束中心固定不变化。由参考臂中平面反射镜37反射的光与样品臂中被测样品背向散射的光在光纤耦合器28处干涉，干涉光经过光谱仪45～48探测并被记录，而后由信号处理器模块与计算单元49采集并作信号分析处理。光谱仪包括依次连接的器件45～48，其中器件45为光纤耦合器，46为光栅，47为汇聚透镜，将光栅色散分出的光聚焦到48所示的线阵探测器上。
具体实施还设置有可见光指示装置，可见光指示装置包括可见光指示光源50、准直透镜51，可见光指示光源50发出用于指示的可见光经过准直透镜51、二向色镜42和聚焦物镜43后到待测样品44。
图4示出的是利用本实施例中得到的血流模版(仿体)实验结果。图4(a)和4(b)分别对应仿体代表性的横断面的OCT结构图和OCTA血流图像，在图4(a)中指示了静态、动态和噪声信号的区域，仿体成像数据这种已知信号的分类的特性，方便量化分类效果。图4(c)将静态、动态和噪声信号投影在信噪比倒数-去相关系数(ID)特征空间中，显然动态和静态信号有着明显的区分。用所提出的ID分类器和传统的均匀强度阈值作用在ID特征空间，以区分血流信号和组织噪声信号，在相同的分别得到图4(d)和4(e)的分类结果。通过查准率(true positive rate,TPR)和假阳性率(false positive rate,FPR)量化血流信号分类的准确性，定量评价基于机器学习构建分类器的血流分类方法相对 传统方法的优势。其中TPR的定义为被分类成血流信号的真实血流信号的像素数目在所有被分类成血流信号的像素数目的占比，TPR越高，意味着分类越准确。FPR的定义是被分类为血流信号的噪声信号的像素数目占总噪声信号的像素数目的比例，FPR越低，反映分类器得到的假阳性样本更少。因此，对于一个分类器来说，TPR越高越好，FPR越低越好，动态和静态信号的区分得更加明显，能得到对比度更高的血流图像。由图可见，所提出的ID特征分类器在相同的假阳性率下，能够成像更深的区域，两者在深度上的差异可以由黑色虚线所清晰地表明。以这两个指标作为两个坐标轴，画出了受试者工作特征曲线(receiver operating characteristic curve,ROC)。在TPR为80％时，ID-OCTA方法仅有9.9％的TPR但传统的强度阈值方法却达到了75.7％。
此外，对健康人类正常皮肤区域进行了活体三维血流成像的实验。图5展示了利用本实施例对活体人类皮肤的血流成像结果。图5(a)为使用所提出的ID-OCTA方法的血流造影结果，图5(b)是传统的强度阈值方法，其中阈值大约设置为系统噪声水平的两倍。相比而言，浅层的大血管由于具有高的血流速度和大的去相关系数，因此在两种方法下能够清楚地观察到。但传统的强度阈值由于不能清除大量的静态区域，导致血流造影图的背景噪声高。这会使得微小的毛细血管难以看清，如方框所示区域及其放大的细节图所示，充满噪声的背景降低了部分毛细血管的可视度。图5(c)分别展示了提高强度阈值的结果。通过提高强度阈值至大约六倍的噪声水平，得到的血管造影图背景噪声可以抑制到与所提出的ID-OCTA方法相似的水平。但显然其同时排除了大量信噪比较低的血管区域。通过正面血管投影图的对比可以看到，低的强度阈值方法导致很多静态信号被误分类为血流信号残留下来，削弱了微小的毛细血管的对比度；而高的强度阈值错误排除了大量主要集中在深部区域的低信噪比的血管。图5(d)，(e)和(f)分别是图5(a)，(b)和(c)的截面血管造影图，同时用箭头指出了一些细节的对比。对比之下，黄色箭头处图4(f)中血管被高强度阈值错误地当作噪声所去除了，而图5(e)低强度阈值错误保留的大量噪声区域。因此，通过图5中所显示的活体成像结果的对比，本发明提出的方法得到的血流图像和传统方法的结果在对于微小毛细血管的可见度上有了显著的增强，对于整个血管造影图有着显著的对比度提升。
进一步地，为了量化所提出的ID-OCTA方法带来的对比度的提升，本发明使用对比度-噪声比(contrast-noise-ratio,CNR)来衡量所提出算法的提升。首先，如图6(a)所示，标记出了ID-OCTA和传统方法结果的血管正面投影图的血管区域和背景噪声区域。通过观察图6(b)和6(c)，很明显地传统的阈值方法 背景更加嘈杂，图6(d)和6(e)分别是血管和背景噪声区域的统计直方图，进一步证实了所提出方法的对比度提升。CNR由下式计算得到：
Figure PCTCN2019083491-appb-000009
其中，
Figure PCTCN2019083491-appb-000010
分别是血管区域和背景噪声区域的去相关系数的均值，σ n是背景噪声区域的标准差。通过在6个不同的样本上进行测试并采用成对数据t检验，ID-OCTA的CNR(5.96±1.00)显著地高于传统阈值方法(5.03±0.89)，显著性水平为5.75*10 -4，这表明所提出的方法确实地增强了血管造影结果的对比度。
图7展现了经信噪比修正后的去相关系数的血管造影结果，与未经修正的去相关系数对比，能将残余的噪声修正地更加彻底，进一步提高血管造影图的对比度；同时针对低信噪比的区域(如深层血管)能够展现更加真实的运动对比度，帮助使用者对于结果作出正确的解读，也有助于后续的血液动力学量化工作。上述实验对比结果充分说明：利用本发明所涉及的基于特征空间的三维血流成像方法，能够提升血流信号分类的准确率，实现血流对比度的有效增强，和血流图像质量的提升，具有其突出显著的技术效果。

Claims (10)

  1. 一种基于特征空间的光学相干层析的三维血流造影方法，其特征在于包括：包括采集器(1)步骤，用于收集三维空间内散射信号样本的OCT散射信号；包括分类器(2)步骤，结合OCT散射信号的信噪比与去相关系数构建二维特征空间实现动态血流信号与静态组织的分类，进而实现利用信噪比的修正；包括成像造影步骤，基于分类器的分类结果生成血流造影图(25)。
  2. 根据权利要求1所述的一种基于特征空间的光学相干层析的三维血流造影方法一种信号采集方法，其特征在于：所述用于收集三维空间内散射信号样本的OCT散射信号，具体包括：对散射信号样本进行三维空间的OCT扫描成像，相同空间位置及其附近位置在S个不同时间点重复采样以及使用多通道进行重复采样。
  3. 根据权利要求2所述的一种基于特征空间的三维血流成像方法，其特征在于：所述对散射信号样本进行三维空间的OCT扫描成像，相同空间位置及其附近位置在S个不同时间点重复采样，采用以下方式之一：通过扫描改变参考臂光程的时间域OCT成像方法；利用光谱仪记录光谱干涉信号的光谱域OCT成像方法；利用扫频光源记录光谱干涉信号的扫频OCT成像方法。
  4. 根据权利要求1所述的一种基于特征空间的三维血流成像方法，其特征在于：所述结合OCT散射信号的信噪比与去相关系数构建二维信噪比倒数-去相关性系数特征空间，实现动态血流信号与静态组织的分类，具体包括：
    S1、采用一阶和零阶自协方差对OCT散射信号计算分析，并在时间、空间、通道等多个维度上作滑动平均或高斯平均，得到各OCT散射信号的信噪比和去相关系数的两个特征；
    S2、建立信噪比-去相关系数特征空间，将体素映射至信噪比倒数-去相关系数(ID)特征空间；用多元时间序列模型建立线性分类器，在信噪比倒数-去相关系数特征空间中将OCT散射信号分为动态血流信号和静态组织噪声信号，并使用经信噪比修正替换去相关系数生成血管造影图。
  5. 根据权利要求4所述的一种基于特征空间的三维血流成像方法，其特征在于：所述步骤S1具体为：采用一阶和零阶自协方差对OCT散射信号计算分析，并在三维空间、时间、通道等多个维度上做滑动平均或高斯平均，得到各OCT散射信号的强度和去相关系数的两个特征。
  6. 根据权利要求4所述的一种基于特征空间的三维血流成像方法，其特征在于：所述去相关系数的特征包括：对同一空间位置S个不同时间点扫描得到 的OCT散射信号的幅度部分或对包含幅度和相位的OCT散射信号计算去相关系数，作为各OCT散射信号的OCTA血流对比度，即作为OCTA血流信息。
  7. 根据权利要求4所述的一种基于特征空间的三维血流成像方法，其特征在于：所述步骤S2具体为：结合OCT散射信号强度与去相关系数构建信噪比倒数-去相关系数特征空间，根据多元时间序列模型通过数值模拟计算OCT散射信号的信噪比-去相关系数的渐近关系，将OCT散射信号投影到信噪比倒数-去相关系数特征空间中，获得静态区域和噪声区域中体素分布及其统计特征。
  8. 用于实施权利要求1～1任一所述方法的基于特征空间的三维血流成像系统，包括：
    一OCT光学相干层析探测装置，用于对三维空间内的散射信号样本的OCT散射信号进行采集；
    一个或多个信号处理器，用于获取并分析OCT散射信号的强度信息和OCTA血流信息，并结合散射信号的强度信息和OCTA血流信息对动态血流信号和静态组织背景进行分类。
  9. 根据权利要求1所述的基于特征空间的三维血流成像系统，其特征在于：所述的一OCT光学相干层析探测装置是采用以下的一种：
    包括低相干光源、干涉仪和探测器，低相干光源和探测器分别连接到干涉仪；
    或者包括低相干光源、干涉仪和光谱仪，低相干光源和光谱仪分别连接到干涉仪；
    或者包括扫频宽光谱光源、干涉仪和探测器，扫频宽光谱光源和探测器分别连接到干涉仪。
  10. 根据权利要求8或9所述的基于特征空间的三维血流成像系统，其特征在于：所述的一OCT光学相干层析探测装置中选择地配置一个可见光指示装置，用于指示OCT探测光束的位置，指导探测目标的放置位置。
PCT/CN2019/083491 2019-01-31 2019-04-19 基于特征空间的光学相干层析的三维血流造影方法及系统 Ceased WO2020155415A1 (zh)

Priority Applications (1)

Application Number Priority Date Filing Date Title
US17/628,579 US11510574B2 (en) 2019-01-31 2019-04-19 Three-dimensional (3D) optical coherence tomography angiography (OCTA) method and system based on feature space

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201910097827.1 2019-01-31
CN201910097827.1A CN109907731B (zh) 2019-01-31 2019-01-31 基于特征空间的光学相干层析的三维血流造影方法

Publications (1)

Publication Number Publication Date
WO2020155415A1 true WO2020155415A1 (zh) 2020-08-06

Family

ID=66961179

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2019/083491 Ceased WO2020155415A1 (zh) 2019-01-31 2019-04-19 基于特征空间的光学相干层析的三维血流造影方法及系统

Country Status (3)

Country Link
US (1) US11510574B2 (zh)
CN (1) CN109907731B (zh)
WO (1) WO2020155415A1 (zh)

Cited By (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112116704A (zh) * 2020-09-11 2020-12-22 同济大学 基于光学相干层析成像的皮下微血管分割与三维重建方法
CN112535465A (zh) * 2020-11-03 2021-03-23 佛山科学技术学院 一种基于片层光的三维血流速度成像方法及装置
CN113706567A (zh) * 2021-07-19 2021-11-26 浙江大学 一种结合血管形态特征的血流成像量化处理方法与装置
CN115778318A (zh) * 2022-11-09 2023-03-14 山东大学 基于双光谱仪探测的可见光oct系统及图像重建方法
CN116392068A (zh) * 2023-04-11 2023-07-07 北京理工大学 一种光学相干血流扫描成像方法

Families Citing this family (23)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110522438B (zh) * 2019-07-31 2022-04-22 华中科技大学苏州脑空间信息研究院 计算血流速度的方法、装置、介质及血流成像方法和系统
CN110693457B (zh) * 2019-10-14 2020-10-16 浙江大学 一种基于光学相干技术的组织活性检测的方法与系统
CN110956107B (zh) * 2019-11-21 2023-05-30 佛山科学技术学院 一种基于oct扫描系统的三维血管类型区分方法
CN110974242B (zh) * 2019-12-26 2023-02-10 浙江福祉科创有限公司 一种用于可穿戴式设备的步态异常程度评价方法及该设备
CN111543971B (zh) * 2020-04-14 2021-08-06 浙江大学 时空自适应样本系综去相关运算的血流量化方法与系统
CN111436909B (zh) * 2020-04-28 2023-03-28 北京信息科技大学 一种活体组织的光学相干层析成像系统及方法
CN111436910B (zh) * 2020-04-28 2023-07-18 北京信息科技大学 一种活体组织的光学相干层析多模态成像装置及方法
CN112057049B (zh) * 2020-09-14 2021-08-10 浙江大学 一种基于多维度特征空间的光学相干血流造影方法与系统
CN112294260B (zh) * 2020-10-10 2022-04-05 浙江大学 一种磁兼容的光学脑功能成像方法与装置
CN112396622B (zh) * 2020-11-24 2023-10-31 浙江大学 基于多维特征空间的微血流图像分割量化方法和系统
CN113331809B (zh) * 2021-05-20 2023-02-14 浙江大学 基于mems微型振镜的腔道内三维血流成像方法和装置
CN113712527A (zh) * 2021-07-19 2021-11-30 浙江大学 一种基于幅度去相关的三维血流成像方法与系统
CN114037769B (zh) * 2021-09-22 2022-12-09 图湃(北京)医疗科技有限公司 光学相干层析血管造影方法、装置、电子设备及存储介质
CN113729593B (zh) * 2021-09-28 2022-11-01 上海交通大学 基于多角度散射随机矩阵的3d内窥镜用血流成像方法
CN116263411A (zh) * 2021-12-15 2023-06-16 香港中文大学 高速单帧无标记细胞断层扫描
CN114947727A (zh) * 2022-05-24 2022-08-30 浙江大学 一种视网膜神经血管耦合功能的检测方法与装置
CN115067911B (zh) * 2022-05-30 2025-04-29 浙江大学嘉兴研究院 一种基于gpu实时处理的octa图像优化方法与装置
CN115436325B (zh) * 2022-07-28 2025-09-16 南昌航空大学 一种布里渊-octa-散斑多模态弹性成像系统装置
CN115517710B (zh) * 2022-09-30 2025-04-04 飞依诺科技股份有限公司 基于调整平面波发射角度数的造影成像方法、系统及存储介质
CN115758053A (zh) * 2022-11-07 2023-03-07 南通先进通信技术研究院有限公司 一种基于扩散相关光谱的血流量估算方法
CN116616738A (zh) * 2023-05-22 2023-08-22 上海美沃精密仪器股份有限公司 一种结合局部信噪比的强度octa成像方法与装置
CN116725492B (zh) * 2023-07-11 2023-12-12 江苏金视传奇科技有限公司 一种基于光学相干层析成像的血管成像方法及其系统
CN118446971B (zh) * 2024-04-28 2025-04-18 康那生物技术(山东)有限公司 基于图像分析的血清异常状态识别方法、系统及存储介质

Citations (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102764139A (zh) * 2012-07-12 2012-11-07 复旦大学 基于特征空间分析和区域判别的医学超声波束形成方法
CN102988082A (zh) * 2011-09-09 2013-03-27 美国西门子医疗解决公司 医学超声剪切波成像中的分类预处理
CN105574861A (zh) * 2015-12-14 2016-05-11 上海交通大学 无标记的血流成像方法及系统
CN107452029A (zh) * 2017-07-31 2017-12-08 中国医学科学院生物医学工程研究所 一种光学微血管血流成像方法
CN108491886A (zh) * 2018-03-29 2018-09-04 重庆大学 一种基于卷积神经网络的多元时序数据的分类方法
CN108670239A (zh) * 2018-05-22 2018-10-19 浙江大学 一种基于特征空间的三维血流成像方法与系统
US20180317851A1 (en) * 2017-05-08 2018-11-08 Oregon Health & Science University Bulk motion subtraction in optical coherence tomography angiography
US20180344149A1 (en) * 2017-06-01 2018-12-06 Santec Corporation Flow imaging in an optical coherence tomography (oct) system

Family Cites Families (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US9883810B2 (en) * 2012-02-03 2018-02-06 Oregon Health & Science University In vivo optical flow imaging
US20140073917A1 (en) * 2012-09-10 2014-03-13 Oregon Health & Science University Quantification of local circulation with oct angiography
CN108042126B (zh) * 2017-12-08 2022-03-22 中国医学科学院生物医学工程研究所 一种改进的激光散斑衬比血流成像方法

Patent Citations (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102988082A (zh) * 2011-09-09 2013-03-27 美国西门子医疗解决公司 医学超声剪切波成像中的分类预处理
CN102764139A (zh) * 2012-07-12 2012-11-07 复旦大学 基于特征空间分析和区域判别的医学超声波束形成方法
CN105574861A (zh) * 2015-12-14 2016-05-11 上海交通大学 无标记的血流成像方法及系统
US20180317851A1 (en) * 2017-05-08 2018-11-08 Oregon Health & Science University Bulk motion subtraction in optical coherence tomography angiography
US20180344149A1 (en) * 2017-06-01 2018-12-06 Santec Corporation Flow imaging in an optical coherence tomography (oct) system
CN107452029A (zh) * 2017-07-31 2017-12-08 中国医学科学院生物医学工程研究所 一种光学微血管血流成像方法
CN108491886A (zh) * 2018-03-29 2018-09-04 重庆大学 一种基于卷积神经网络的多元时序数据的分类方法
CN108670239A (zh) * 2018-05-22 2018-10-19 浙江大学 一种基于特征空间的三维血流成像方法与系统

Cited By (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112116704A (zh) * 2020-09-11 2020-12-22 同济大学 基于光学相干层析成像的皮下微血管分割与三维重建方法
CN112116704B (zh) * 2020-09-11 2024-04-26 同济大学 基于光学相干层析成像的皮下微血管分割与三维重建方法
CN112535465A (zh) * 2020-11-03 2021-03-23 佛山科学技术学院 一种基于片层光的三维血流速度成像方法及装置
CN113706567A (zh) * 2021-07-19 2021-11-26 浙江大学 一种结合血管形态特征的血流成像量化处理方法与装置
CN115778318A (zh) * 2022-11-09 2023-03-14 山东大学 基于双光谱仪探测的可见光oct系统及图像重建方法
CN115778318B (zh) * 2022-11-09 2024-05-07 山东大学 基于双光谱仪探测的可见光oct系统及图像重建方法
CN116392068A (zh) * 2023-04-11 2023-07-07 北京理工大学 一种光学相干血流扫描成像方法

Also Published As

Publication number Publication date
CN109907731B (zh) 2021-06-01
US11510574B2 (en) 2022-11-29
CN109907731A (zh) 2019-06-21
US20220248960A1 (en) 2022-08-11

Similar Documents

Publication Publication Date Title
CN109907731B (zh) 基于特征空间的光学相干层析的三维血流造影方法
CN108670239B (zh) 一种基于特征空间的三维血流成像方法与系统
CN107595250B (zh) 基于运动与图形混合对比度的血流成像方法与系统
US7995814B2 (en) Dynamic motion contrast and transverse flow estimation using optical coherence tomography
CN107788950B (zh) 基于自适应阈值分割的血流成像方法与系统
USRE46412E1 (en) Methods and systems for performing angle-resolved Fourier-domain optical coherence tomography
CN105559756B (zh) 基于全空间调制谱分割角度复合的微血管造影方法与系统
KR102046309B1 (ko) 화상생성장치, 화상생성방법, 및 기억매체
CN112057049B (zh) 一种基于多维度特征空间的光学相干血流造影方法与系统
CN208837916U (zh) 一种血流成像系统
CN110693457B (zh) 一种基于光学相干技术的组织活性检测的方法与系统
CN106073700A (zh) 图像生成方法和图像生成装置
CN106137134B (zh) 多角度复合的血流成像方法及系统
CN112396622B (zh) 基于多维特征空间的微血流图像分割量化方法和系统
CN113331809B (zh) 基于mems微型振镜的腔道内三维血流成像方法和装置
CN104545872B (zh) 基于线性相关系数来重构三维微血流分布的方法及装置
CN113706567B (zh) 一种结合血管形态特征的血流成像量化处理方法与装置
CN106491078A (zh) 去除血流图像中组织抖动噪声的方法及装置
CN111543971B (zh) 时空自适应样本系综去相关运算的血流量化方法与系统
CN113712527A (zh) 一种基于幅度去相关的三维血流成像方法与系统
CN115067911B (zh) 一种基于gpu实时处理的octa图像优化方法与装置
CN117770804A (zh) 一种基于octa的非侵入连续血糖测量方法与装置
CN117999456A (zh) 用于样本成像的光学相干断层扫描系统和方法
US12629018B2 (en) Using the dynamic forward scattering signal for optical coherence tomography based flow quantification

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 19912695

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 19912695

Country of ref document: EP

Kind code of ref document: A1