Vascular dynamics video registration quantification method and device based on vascular skeleton

Abstract

The invention discloses a vascular dynamics video registration quantification method and device based on a vascular skeleton. The method comprises the steps of collecting a blood vessel signal of a target area, generating a dynamic blood vessel video according to the blood vessel signal, obtaining a reference image according to the dynamic blood vessel video, then obtaining a corresponding dense deformation field according to the reference image and the dynamic blood vessel video, and modifying the dense deformation field to obtain a modified dense deformation field. And obtaining relative change of the blood vessel information according to the modified dense deformation field and the dynamic blood vessel video, and finally obtaining a dynamic change curve of the blood vessel along with time according to the relative change of the blood vessel information. According to the method, non-rigid rapid registration of the pixel-level vascular dynamics video is achieved, meanwhile, dynamic change information of blood vessels is reserved, and accurate quantification of vascular dynamics response is achieved.

Description

Technical Field

[0001] The present invention belongs to the field of biomedical imaging, and particularly relates to a method and device for vascular dynamics video registration and quantification based on vascular skeletons. Background Art

[0002] Neurovascular coupling (NVC) is crucial for ensuring continuous and sufficient blood supply to maintain the function and health of organs and tissues. In the early stages of important diseases such as diabetes mellitus (DM) and Parkinson's disease (PD), this mechanism may be disrupted. Detecting the NVC mechanism by recording vascular hyperemia under external stimuli contributes to the early diagnosis of such diseases.

[0003] However, the recording time of blood vessels usually lasts for several minutes. Due to involuntary movements such as breathing, the captured vascular dynamics videos often exhibit significant imaging field-of-view shifts, and this shift is accompanied by non-rigid deformations of the blood vessels. To accurately extract vascular dynamics information from the videos for NVC assessment, an effective registration technique is needed to eliminate inter-frame shifts and vascular deformations while retaining changes in vascular dynamics responses. Among the existing registration methods so far, rigid registration methods, such as feature point-based transformations, have low registration accuracy and cannot compensate for non-rigid deformations of blood vessels, affecting the precise quantification of vascular dynamics, i.e., pixel-level changes occurring in individual capillaries. Non-rigid registration methods, such as B-splines and optical flow, can achieve higher registration accuracy, but tend to scale the blood vessels during the registration process, resulting in the loss of valuable vascular dynamics information in the videos.

[0004] Therefore, to make up for the deficiencies of existing registration methods, there is an urgent need for a registration technique that can achieve non-rigid dynamic registration of hemodynamic videos while retaining hemodynamic information in the videos, helping to achieve precise quantification of hemodynamic information and promoting the clinical application of vascular function imaging. Summary of the Invention

[0005] The purpose of the present invention is to address the deficiencies of the prior art and propose a method and device for vascular dynamics video registration and quantification based on vascular skeletons. By using vascular skeletons to modify the deformation field from the image to be registered to the reference image, it ensures that the blood vessel diameters remain unchanged during the registration process while retaining the non-rigid motion compensation ability of the original deformation field, achieving non-rigid registration and quantification of vascular images that retain vascular dynamic change information.

[0006] The purpose of the present invention is achieved through the following technical solutions:

[0007] I. A method for vascular dynamics video registration and quantification based on vascular skeletons

[0008] It includes the following steps: S1. Collect the vascular signals of the target area at different time points through a collection device, obtain a corresponding number of frames of vascular images according to the vascular signals at different time points, and combine all frames of vascular images into a dynamic vascular video.

[0009] S2. Obtain a reference image according to the dynamic vascular video, sequentially use each frame of vascular image in the dynamic vascular video as a to-be-registered image, and sequentially generate a dense deformation field from the to-be-registered image to the reference image according to the to-be-registered image and the reference image.

[0010] S3. Obtain the vascular map and skeleton map corresponding to each frame of vascular image according to the dynamic vascular video respectively, and modify the corresponding dense deformation field according to the vascular map and skeleton map corresponding to each frame of vascular image, so as to obtain the modified dense deformation fields from all to-be-registered images to the reference image.

[0011] S4. Apply each modified dense deformation field to the corresponding vascular image to obtain all registered vascular images, obtain the relative change of vascular information according to the reference image and all registered vascular images, obtain the global response map of the vascular function state of each frame of registered vascular image according to the relative change of vascular information, and obtain the dynamic change curve of the blood vessels over time according to the global response maps of all registered vascular images.

[0012] The step S2 includes:

[0013] Use the vascular image of one frame in the dynamic vascular video as the reference image;

[0014] Or perform an averaging process on the vascular images of some or all frames in the dynamic vascular video, and use the obtained averaged vascular image as the reference image;

[0015] Or use a rigid or non-rigid registration method to register the vascular images of some or all frames in the dynamic vascular video, obtain the corresponding registered dynamic vascular video, then perform an averaging process on the vascular images of some or all frames in the registered dynamic vascular video, and use the obtained averaged vascular image as the reference image.

[0016] The step S2 further includes:

[0017] Use the optical flow field from the to-be-registered image to the reference image as the dense deformation field;

[0018] Or use B-spline curve fitting to obtain the deformation field from the to-be-registered image to the reference image as the dense deformation field.

[0019] The step S3 is specifically:

[0020] S31. According to the dynamic vascular video, obtain the pixel values of each frame of vascular image in the dynamic vascular video and the vascular morphology in each frame of vascular image.

[0021] S32. Obtain the vascular graph and the skeleton graph corresponding to each frame of vascular image respectively by using the binarization method and the morphological skeleton method according to the pixel value and the vascular morphology.

[0022] S33. Obtain the pixel positions where the vessels are located according to the vascular graph, and obtain the pixel positions where the skeletons are located according to the skeleton graph. In each dense deformation field, use the pixel positions where the vessels are located in the vascular graph to modify the displacement values of the corresponding pixel positions of the vessels in the obtained dense deformation field to the displacement values of the positions corresponding to the nearest skeletons in the corresponding positions in the skeleton graph in the dense deformation field, so as to obtain the modified dense deformation field, and further obtain the modified dense deformation fields between all the images to be registered and the reference image.

[0023] The binarization method in the step S32 is set according to the following formula:

[0024]

[0025] where I(x, y) represents the pixel value of the point where the pixel (x, y) is located, T(x, y) represents the binarization threshold of the pixel (x, y), m(x, y) and s(x, y) respectively represent the local mean and the local standard deviation of the pixel (x, y) in the preset n×n neighborhood, R represents the dynamic range of the standard deviation, k, p and q are all constants, and e is a constant.

[0026] The morphological skeleton method in the step S32 is set according to the following formula:

[0027]

[0028] where S represents the skeleton, represents the erosion operation, ⊕ represents the dilation operation, A represents the input binarized image, and B represents the structure element used.

[0029] The modified dense deformation field obtained in the step S33 is specifically set according to the following formula:

[0030]

[0031] where b j represents the index j pixel position where the vessel is located in the vascular graph, s j represents the pixel position of the nearest skeleton in the corresponding skeleton graph to b j , s i represents the index i pixel position of the nearby skeletons in the corresponding skeleton graph to b j , V(b j ) represents the displacement value of the pixel position where the vessel is located at b j , V(s j)Indicates the location s of blood vessels j The displacement value of the pixel position, B represents the set of pixel positions where blood vessels are located in the blood vessel image, S represents the set of pixel positions where the skeleton is located in the skeleton image, ||*||1 represents the L1 norm, and argmin represents taking the minimum value.

[0032] The skeleton in step S33 is the corresponding skeleton in the skeleton image or the skeleton of the skeleton image corresponding to the local depth in the field of view.

[0033] Step S4 is specifically as follows:

[0034] S41. Apply the modified dense deformation field to each frame of blood vessel image in the corresponding dynamic blood vessel video to obtain each frame of registered blood vessel image, and obtain all frames of registered blood vessel images, thereby obtaining the registered dynamic blood vessel video.

[0035] S42. According to the registered dynamic blood vessel video, obtain the blood vessel information of all pixel positions where blood vessels are located on each frame of registered blood vessel image.

[0036] S43. According to the reference image, obtain the blood vessel information of all pixel positions where blood vessels are located on the reference image.

[0037] S44. Compare and process the blood vessel information of each pixel position where blood vessels are located on the registered blood vessel image with the blood vessel information of the corresponding pixel position where blood vessels are located on the reference image respectively, to obtain the relative change of the blood vessel information of all pixel positions where blood vessels are located.

[0038] The comparison and processing can be operations such as taking the difference and taking the square root after taking the difference.

[0039] S45. Take the relative change of the blood vessel information of each pixel position where blood vessels are located as the local response map of the blood vessel function state of the corresponding pixel position, thereby obtaining the local response map of the blood vessel function state of all pixel positions.

[0040] S46. Perform average or summation processing on all or part of the local response maps to obtain the global response map of the blood vessel function state of the registered blood vessel image.

[0041] S47. Obtain the dynamic change curve of the blood vessel over time according to the global response map of the blood vessel function state of all registered blood vessel images.

[0042] The blood vessel information in step S4 includes blood vessel diameter, blood flow velocity, and blood flow rate.

[0043] II. A blood vessel dynamics video registration quantization device based on blood vessel skeleton

[0044] The device includes: an optical coherence tomography imaging device for performing dynamic OCTA detection and imaging on the target area;

[0045] Or a fluorescence blood flow imaging device for dynamically detecting and imaging fluorescence signals in a target area;

[0046] Or a laser speckle imaging device for dynamically detecting and imaging speckle signals in a target area;

[0047] Or an ultrasonic localization microscopy imaging device for dynamically detecting and imaging ultrasonic signals in a target area;

[0048] Or a magnetic resonance device for dynamically detecting and imaging magnetic signals in a target area;

[0049] Or an X-ray computed tomography device for dynamically detecting and imaging ray signals in a target area;

[0050] Or a photoacoustic imaging device for dynamically detecting and imaging photoacoustic signals in a target area;

[0051] One or more processors for generating a vascular map and a skeleton map, obtaining a dense deformation field, a modified dense deformation field between a to-be-registered image and a reference image, generating a local response map and a global response map, and realizing quantitative registration of a vascular dynamics video based on a vascular skeleton.

[0052] The optical coherence tomography device described above is one of the following:

[0053] Including a low-coherence light source, an interferometer, and a detector;

[0054] Or including a low-coherence light source, an interferometer, and a spectrometer;

[0055] Or including a swept-source broad-spectrum light source, an interferometer, and a detector.

[0056] The beneficial effects of the present invention are as follows:

[0057] 1. Compared with the existing non-rigid vascular image registration technology, since it does not consider vascular information, the registered video no longer contains complete vascular dynamic change information. While realizing non-rigid registration of a dynamic vascular video, the present invention retains the dynamic change information of blood vessels in the dynamic vascular video, and also modifies the original dense deformation field based on the vascular skeleton, and realizes non-rigid registration of the dynamic vascular video with the dynamic change information of blood vessels retained by ensuring that the relative positions of the blood vessel walls remain unchanged.

[0058] 2. Compared with existing rigid dynamic registration techniques, since their registration is obtained through rigid transformation, while the blood vessels in dynamic blood vessel videos will undergo non-rigid distortion transformation, there is a greater impact on the accurate quantification of blood vessel dynamics. The present invention is optimized and improved based on non-rigid registration techniques, and the registration process is still non-rigid transformation, achieving an accurate assessment of the functional response information at each spatial position in the blood vessel video. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] Figure 1 is a flowchart of the method of the present invention;

[0060] Figure 2 is a schematic diagram of the device of the present invention;

[0061] Figure 3 is a schematic diagram of the device of an exemplary embodiment of the present invention;

[0062] Figure 4 is a timing diagram of light stimulation of an exemplary embodiment of the present invention;

[0063] Figure 5 is a flowchart of the method for registering and quantifying blood vessel dynamics video based on blood vessel skeleton of an exemplary embodiment of the present invention;

[0064] Figure 6 is a response diagram and a quantitative comparison diagram of the blood vessel functional state before and after registration of an exemplary embodiment of the present invention.

[0065] Wherein: 11. Light source; 12. Beam splitter; 13. Reference arm polarization controller; 14. Reference arm collimating lens; 15. Reference arm focusing lens; 16. Reflection lens; 17. Sample arm polarization controller; 18. Sample arm collimating lens; 19. Scanning module; 20. Sample arm focusing lens; 21. Sample external stimulation module; 22. Sample; 23. Signal detector; 24. Signal processing unit; 31. Superluminescent diode broadband light source; 32. Fiber optic coupler; 33. Reference arm polarization controller; 34. Reference arm collimating lens; 35. Reference arm dispersion compensator; 36. Reference arm doublet lens; 37. Reference arm dispersion matching lens; 38. Reference arm mirror; 39. Sample arm collimating lens; 40. Sample arm scanning module; 41. Dichroic mirror; 42. White light emitting diode; 43. Sample arm doublet lens; 44. Sample arm eyepiece; 45. Sample eye; 46. Interference light collimating lens; 47. Grating beam splitter; 48. Focusing lens; 49. Linear array detector; 50. Computing unit. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0066] The following will describe in detail the specific embodiments of the present invention in conjunction with the accompanying drawings, which form a part of the present invention. It should be noted that these descriptions and examples are merely exemplary and should not be construed as limiting the scope of the present invention. The protection scope of the present invention is defined by the appended claims, and any modification based on the claims of the present invention falls within the protection scope of the present invention.

[0067] To facilitate the understanding of the embodiments of the present invention, each operation is described as a plurality of discrete operations. However, the described order does not represent the order of performing the operations.

[0068] In this description, a three-dimensional x-y-z coordinate representation based on spatial directions is adopted for the sample measurement space. This description is only used to facilitate the discussion and is not intended to limit the application of the embodiments of the present invention. Among them: 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.

[0069] The method of the present invention is as Figure 1 shown and includes the following steps:

[0070] S1. Collect vascular signals of the target area at different time points through a collection device, obtain a corresponding number of frames of vascular images according to the vascular signals at different time points, and combine all frames of vascular images into a dynamic vascular video.

[0071] S2. Obtain a reference image according to the dynamic vascular video, sequentially use each frame of vascular image in the dynamic vascular video as a to-be-registered image, and sequentially generate a dense deformation field from the to-be-registered image to the reference image according to the to-be-registered image and the reference image.

[0072] Step S2 includes: using a vascular image of one of the frames in the dynamic vascular video as the reference image; or performing an averaging process on some or all of the frames of vascular images in the dynamic vascular video, and using the obtained averaged vascular image as the reference image; or using a rigid or non-rigid registration method to register some or all of the frames of vascular images in the dynamic vascular video to obtain a corresponding registered dynamic vascular video, and then performing an averaging process on some or all of the frames of vascular images in the registered dynamic vascular video, and using the obtained averaged vascular image as the reference image.

[0073] In a specific implementation, the first frame of the dynamic vascular video is used as the reference image for subsequent dynamic registration.

[0074] Step S2 further includes: using the optical flow field from the to-be-registered image to the reference image as the dense deformation field; or using B-spline curve fitting to obtain the deformation field from the to-be-registered image to the reference image as the dense deformation field.

[0075] In specific implementation, the optical flow method is used to calculate the optical flow field from the image to be registered to the reference image as the dense deformation field.

[0076] S3. For each frame of the vascular image in the dynamic vascular video, obtain the corresponding vascular map and skeleton map. According to the vascular map and skeleton map corresponding to each frame of the vascular image, modify the dense deformation field between the corresponding image to be registered and the reference image to obtain the modified dense deformation field between all images to be registered and the reference image.

[0077] In specific implementation, first extract the superficial vascular plexus in the dynamic vascular video, then use the Phansalker method to obtain the binarization threshold, perform binarization processing on the OCTA video, and then use the Zhang–Suen method to extract the vascular map and skeleton map.

[0078] S31. According to the dynamic vascular video, obtain the pixel values of each frame of the vascular image in the dynamic vascular video and the vascular morphology in each frame of the vascular image.

[0079] S32. According to the pixel values and vascular morphology, use the binarization method and the morphological skeleton method to obtain the corresponding vascular map and skeleton map for each frame of the vascular image respectively.

[0080] The binarization method is set according to the following formula:

[0081]

[0082]

[0083] Among them, I(x, y) represents the pixel value of the point where the pixel (x, y) is located, T(x, y) represents the binarization threshold of the pixel (x, y), m(x, y) and s(x, y) respectively represent the local mean and local standard deviation of the pixel (x, y) in the preset n×n neighborhood. In specific implementation, n is 20, R represents the dynamic range of the standard deviation. In specific implementation, the normalized image is equal to 0.5, k, p, and q are all constants. In specific implementation, k, p, and q are taken as 0.25, 2, and 10 respectively, and e is a constant.

[0084] The morphological skeleton method is set according to the following formula:

[0085]

[0086] Among them, S represents the skeleton, represents the erosion operation, ⊕ represents the dilation operation, A represents the input binarized image, B represents the structural element used. In specific implementation, B uses a 3*3 matrix as the structural element.

[0087] S33. Obtain the pixel positions of blood vessels based on the blood vessel map, and obtain the pixel positions of the skeleton based on the skeleton map. In each dense deformation field, use the pixel positions of blood vessels in the blood vessel map to modify the displacement value of the corresponding pixel position of the blood vessel in the obtained dense deformation field to the displacement value of the position corresponding to the pixel position of the skeleton closest to it in the skeleton map in the dense deformation field, so as to obtain the modified dense deformation field, and further obtain the modified dense deformation fields between all the images to be registered and the reference image.

[0088] The obtained modified dense deformation field is specifically set according to the following formula:

[0089]

[0090] Among them, b j represents the index j pixel position of the blood vessel in the blood vessel map, s j represents the pixel position of the skeleton closest to b j in the corresponding skeleton map, s i represents the index i pixel position of the nearby skeleton of b j in the corresponding skeleton map, V(b j ) represents the displacement value of the pixel position where the blood vessel is located at b j , V(s j ) represents the displacement value of the pixel position where the blood vessel is located at s j , B represents the set of pixel positions of blood vessels in the blood vessel map, S represents the set of pixel positions of skeletons in the skeleton map, |‖*‖‖1 represents the L1 norm, argmin represents taking the minimum value, and both i and j are indices.

[0091] In specific implementation, in the dense deformation field, for each pixel position where a blood vessel is located, within the range of 20*20 pixels around it, find the pixel position of the skeleton closest to it. After the search is completed, modify the displacement value of each pixel position where a blood vessel is located to the displacement value of the pixel position of the skeleton closest to it found.

[0092] The skeleton in step S33 is the corresponding skeleton in the skeleton image or the skeleton of the skeleton image corresponding to the local depth within the field of view.

[0093] S4. Apply each modified dense deformation field to the corresponding blood vessel image to obtain all registered blood vessel images. Obtain the relative change of blood vessel information based on the reference image and all registered blood vessel images. Obtain the global response map of the blood vessel function state of each registered blood vessel image based on the relative change of blood vessel information. Obtain the dynamic change curve of the blood vessel over time based on the global response maps of all registered blood vessel images.

[0094] The vascular information in step S4 includes vessel diameter, blood flow velocity based on decorrelation, and blood flow rate.

[0095] In a specific implementation, when obtaining the vessel diameter at a certain pixel position where the blood vessel is located, the sum of the number of all pixel points on the inner diameter of the blood vessel at this pixel position is used as the vessel diameter at this pixel position.

[0096] S41. Apply the modified dense deformation field to each frame of the blood vessel image in the corresponding dynamic blood vessel video to obtain each frame of registered blood vessel image, and obtain all frames of registered blood vessel images, so as to obtain the registered dynamic blood vessel video.

[0097] S42. According to the registered dynamic blood vessel video, obtain the vascular information at all pixel positions where the blood vessels are located on each frame of the registered blood vessel image.

[0098] S43. According to the reference image, obtain the vascular information at all pixel positions where the blood vessels are located on the reference image;

[0099] S44. Compare and process the vascular information at each pixel position where the blood vessels are located on the registered blood vessel image with the vascular information at the corresponding pixel positions where the blood vessels are located on the reference image respectively, to obtain the relative change of the vascular information at all pixel positions where the blood vessels are located.

[0100] The comparison and processing can be operations such as taking the difference and taking the square root after taking the difference. Any method that can reflect the relative change relationship is acceptable for the comparison and processing.

[0101] S45. Take the relative change of the vascular information at each pixel position where the blood vessels are located as the local response map of the vascular function state at the corresponding pixel position, so as to obtain the local response map of the vascular function state at all pixel positions.

[0102] S46. Perform averaging or summing processing on all or part of the local response maps to obtain the global response map of the vascular function state of the registered blood vessel image.

[0103] S47. Obtain the dynamic change curve of the blood vessels over time according to the global response map of the vascular function state of all registered blood vessel images.

[0104] A vascular dynamics video registration and quantification device based on a vascular skeleton according to the present invention includes: an optical coherence tomography device for performing dynamic OCTA detection and imaging on a target area; or a fluorescence angiography device for performing dynamic fluorescence signal detection and imaging on the target area; or a laser speckle imaging device for performing dynamic speckle signal detection and imaging on the target area; or an ultrasound localization microscopy imaging device for performing dynamic ultrasound signal detection and imaging on the target area; or a magnetic resonance device for performing dynamic magnetic signal detection and imaging on the target area; or an X-ray computed tomography device for performing dynamic ray signal detection and imaging on the target area; or a photoacoustic imaging device for performing dynamic photoacoustic signal detection and imaging on the target area; one or more processors for generating a vascular map and a skeleton map, obtaining a dense deformation field, a modified dense deformation field, generating a local response map and a global response map between the image to be registered and the reference image, and realizing vascular dynamics video registration and quantification based on the vascular skeleton.

[0105] Figure 2 It is a schematic diagram taking the optical coherence tomography device as an example. The main structure of the low-coherence interferometric measurement part of this device is an interferometer, which is composed of 11-20 and 22-24. The detection light emitted by the light source 11 is split into two beams of light by the beam splitter 12: one beam of light is in the reference arm. After passing through the reference arm polarization controller 13 and the reference arm collimating lens 14, it is converged on the reflection lens 16 by the reference arm focusing lens 15, and then returns to the beam splitter along the original path; the other beam of light enters the sample arm and passes through the sample arm polarization controller 17, the sample arm collimating lens 18, the scanning module 19 and the sample arm focusing lens 20 in sequence, and then converges on the sample 22, and then returns to the beam splitter along the original path. The two beams of light returned from the reference arm and the sample arm interfere in the beam splitter 12, and then are transmitted to the signal detector 23. The detection signal generated by the detector is transmitted to the signal processing unit 24. The reference arm polarization airer 13 and the sample arm polarization controller 17 are used to adjust the polarization state of the light beam to maximize the interference signal of the two arms.

[0106] According to different ways of low-coherence interference detection signals, Figure 2 The optical coherence tomography device shown specifically includes:

[0107] 1) Time-domain measurement device. The light source 11 uses broadband low-coherence light, the mirror 16 can move along the optical axis direction, and the interference signal detection device 23 is a point detector. By moving the mirror 16 to change the optical path of the reference arm, the interference signal of the two arms is detected by the point detector 23, realizing low-coherence interference detection of the scattering signal in the z direction of the sample space depth, and combining the scanning module 19 to obtain depth-resolved three-dimensional OCT volume data.

[0108] 2) Spectral domain measurement device. The light source 11 uses broadband low-coherence light. The plane mirror 16 is fixed. The interference signal detection device 23 uses a spectrometer. The high-speed linear array camera in the spectrometer is used to simultaneously detect the interference signals in the depth direction of the sample. The Fourier transform method is used in the signal processing unit 23 to analyze the interference spectrum signals, and the scattering information in the depth z direction of the sample is obtained in parallel. Combining with the scanning module 19, depth-resolved three-dimensional OCT volume data is obtained.

[0109] 3) Swept-source measurement device. The light source 11 uses a swept-source. The plane mirror 16 is fixed. The interference signal detection device 23 uses a point detector. As the light frequency emitted by the light source changes, the point detector records the low-coherence interference signals at different frequencies in a time-sharing manner to form a low-coherence interference signal spectrum. The Fourier transform method is used in the signal processing unit 23 to analyze the interference spectrum signals, and the scattering information in the depth z direction is obtained in parallel. Combining with the scanning module 19, depth-resolved three-dimensional OCT volume data is obtained.

[0110] For the above different measurement devices, the OCTA imaging method can be respectively combined to analyze the relative motion differences between blood flow and surrounding tissues to generate OCTA blood flow motion angiography. By continuously collecting the target area under visible light stimulation, a dynamic OCTA blood flow video is obtained.

[0111] Figure 3 An exemplary embodiment of the present invention is disclosed. A non-rigid dynamic registration quantization device based on a vascular skeleton includes a superluminescent diode broadband light source 31; a 75:25 fiber coupler 32; a reference arm polarization controller 33; a reference arm collimating lens 34; a reference arm dispersion compensator 35; a reference arm doublet lens 36; a reference arm dispersion matching lens 37; a reference arm mirror 38; a sample arm collimating lens 39; a sample arm scanning module 40; a dichroic mirror 41; a white light emitting diode 42; a sample arm doublet lens 43; a sample arm eyepiece 44; a sample eye 45; a detection light collimating lens 46; a grating spectrometer 47; a focusing lens 48; a linear array detector 49; a calculation unit 50.

[0112] The superluminescent diode broadband light source 31 is connected to one of the branches at one end of the 75:25 fiber coupler 32. The other branch at this end of the 75:25 fiber coupler 32 is connected to the calculation unit 50 after passing through the interference light collimating mirror 46, the grating beam splitter 47, the focusing lens 48, and the linear array detector 49. One of the branches at the other end of the 75:25 fiber coupler 32 is connected to the reference arm. The reference arm includes a reference arm polarization controller 33, a reference arm collimating lens 34, a reference arm dispersion compensator 35, a reference arm doublet lens 36, a reference arm dispersion matching lens 37, and a reference arm mirror 38. This branch at this end of the 75:25 fiber coupler 32 is connected to one end of the reference arm polarization controller 33. The other end of the reference arm polarization controller 33 is connected to the incident end of the reference arm collimating lens 34. A reference arm dispersion compensator 35, a reference arm doublet lens 36, and a reference arm dispersion matching lens 37 are sequentially arranged between the reference arm collimating lens 34 and the reference arm reflecting lens 38. The dispersion compensator 35 is close to the reference arm collimating lens 34, and the reference arm dispersion matching lens 37 is close to the reference arm reflecting lens 38. The other branch at this end of the 75:25 fiber coupler 32 is connected to the sample arm. The sample arm includes a sample arm collimating lens 39, a sample arm two-dimensional scanning device 40, a sample arm dichroic mirror 41, a sample arm doublet lens 43, a sample arm eyepiece 44, and a sample eye 45. A sample arm OCT two-dimensional scanning device 40, a sample arm dichroic mirror 41, a sample arm doublet lens 43, and a sample arm eyepiece 44 are sequentially arranged between the sample arm collimating lens 39 and the sample eye 45. The sample arm eyepiece 44 coincides with the focal point of the sample arm doublet lens 43, so as to irradiate the surface of the sample eye 45 with parallel light and converge on the retina through the eyeball. The stimulating light emitted by the white light-emitting diode 42 is coupled into the sample arm detection optical path through the sample arm dichroic mirror 41 and is transmitted to the sample eye 45 after passing through the sample arm doublet lens 43 and the sample arm eyepiece 44.

[0113] A vascular dynamics video registration and quantification device based on a blood vessel skeleton according to this embodiment. The central wavelength of the superluminescent diode broadband light source 31 is 840 nm, and the bandwidth is 120 nm. The OCT signal of the sample is collected by the high-speed linear array detector 49, and the signal is transmitted to the calculation unit 50. The parallel light emitted by the sample arm collimating lens 39 is incident on the sample arm OCT scanning module 40. Subsequently, the emitted parallel light passes through the sample arm dichroic mirror 41 and then enters the sample arm doublet lens 43. The sample arm OCT scanning module 40 consists of two planar scanning galvanometers. The range of OCT signal collection is controlled by adjusting the angles of the two planar scanning galvanometers. The detection light emitted by the superluminescent diode broadband light source 31 is transmitted through an optical fiber to the 75:25 fiber coupler 32 and is divided into two beams of light that enter the reference arm and the sample arm respectively. Among them, 75% of the detection light enters the reference arm. After passing through the polarization controller 33, it is transmitted to the reference arm collimating lens 34. After being collimated, it passes through the dispersion compensation module 35, the reference arm doublet lens 36, and the reference arm dispersion matching lens 37, and converges on the reference arm mirror 38. Subsequently, this beam of light returns to the 75:25 fiber coupler 32 along the original path. 25% of the detection light enters the sample arm. After passing through the sample arm collimating mirror 39 and the scanning galvanometer 40, it sequentially passes through the sample arm dichroic mirror 41, the sample arm doublet lens 43, and the eyepiece 44 and enters the sample eye 45. The light beam in the sample eye 45 is focused on the retina by the eye. The visible light stimulation light source 42 is a white light-emitting diode. The emitted stimulation light is coupled into the sample arm detection optical path through the sample arm dichroic mirror 41, and after passing through the sample arm doublet lens 43 and the eyepiece 44, it is transmitted to the sample eye 45 to achieve visual stimulation. Subsequently, the detection light beam carrying the fundus information of the sample eye 45 returns to the 75:25 fiber coupler 32 along the original path. The backscattered light returned from the reference arm and the sample arm interferes in the 75:25 fiber coupler 32. The generated interference light sequentially passes through the detection module collimating mirror 46 and the grating spectrometer 47, and is converged on the linear array detector 49 by the focusing lens 48 to achieve signal detection and recording. Subsequently, the signal processing module 50 collects and further processes it.

[0114] In specific implementation, the timing of visible light stimulation is as Figure 4 shown. There is no visible light stimulation energy in the t1 stage, visible light stimulation is applied to the retina in the t2 stage, and the t3 stage is the recovery period without visible light stimulation. Each stimulation pulse width is 100 ms, the frequency is 10 Hz, the amplitude is 2.7 V, and the stimulation time is 30 s, that is, 3000 pulse sequences.

[0115] Figure 5 It is a flowchart of a vascular dynamics video registration and quantification method based on a blood vessel skeleton.

[0116] Figure 6Response results of the vascular function status obtained before and after dynamic vascular video registration. Among them, Figure 6 a and Figure 6 b are the global response maps of the vascular function status obtained from the dynamic vascular videos before and after registration respectively. It can be seen that due to the offset of the position of the blood vessels before registration, it is impossible to calculate the local response map. After registration, while the spatial positions of the blood vessels are aligned, the vascular dynamics information is retained, and the response values on individual blood vessels can be accurately calculated.

[0117] Figure 6 a and Figure 6 The changes of the marked positions in b over time are shown in Figure 6 c, where from left to right are the baseline phase, the stimulation phase, and the recovery phase. I represents before registration, and II represents after registration. It can be seen that during the whole experiment, the spatial positions of the blood vessels shift significantly, while the spatial positions of the blood vessels remain unchanged after registration, and obvious blood vessel dilation and contraction phenomena can be observed.

[0118] Figure 6 d gives the quantitative results of the relative changes in blood flow of the large retinal blood vessels and capillaries in the experiment, and it is found that compared with the baseline segment, the flickering light stimulation induces an increase in the diameter of the retinal blood vessels, and gradually returns to near the baseline level after turning off the light stimulation.

[0119] The above experimental results fully illustrate that: the method and device for vascular dynamics video registration and quantification based on the vascular skeleton proposed by the present invention achieve accurate registration while retaining the change information between dynamic videos, realize the accurate quantification of the local position response of blood vessels, and ensure the accuracy of the quantitative evaluation of neurovascular function.

[0120] Finally, it should be noted that the above embodiments and descriptions are only used to illustrate the technical solutions of the present invention and are not intended to limit them. Those of ordinary skill in the art should understand that the technical solutions of the present invention can be modified or equivalently replaced. Without departing from the spirit and scope of the disclosure of the technical solutions of the present invention, they should all be covered by the protection scope of the claims of the present invention.

Claims

1. A method for quantifying video registration of vascular dynamics based on vascular skeletons, characterized in that, It includes the following steps: S1. Collect the vascular signals of the target area at different time points, obtain a corresponding number of frames of vascular images according to the vascular signals at different time points, and combine all frames of vascular images into a dynamic vascular video; S2. Obtain a reference image according to the dynamic vascular video, sequentially use each frame of vascular image in the dynamic vascular video as a to-be-registered image, and sequentially generate a dense deformation field from the to-be-registered image to the reference image according to the to-be-registered image and the reference image; S3. Obtain the vascular map and the skeleton map corresponding to each frame of vascular image according to the dynamic vascular video respectively, and modify the corresponding dense deformation field according to the vascular map and the skeleton map corresponding to each frame of vascular image, so as to obtain the modified dense deformation field from all to-be-registered images to the reference image; S4. Apply each modified dense deformation field to the corresponding vascular image to obtain all registered vascular images, obtain the relative change of vascular information according to the reference image and all registered vascular images, obtain the global response map of each registered vascular image according to the relative change of vascular information, and obtain the dynamic change curve of blood vessels over time according to the global response maps of all registered vascular images.

2. The vascular dynamics video registration and quantification method based on vascular skeleton according to claim 1, wherein The step S2 includes: Use the vascular image of one frame in the dynamic vascular video as the reference image; Or perform an averaging process on the vascular images of some or all frames in the dynamic vascular video, and use the obtained averaged vascular image as the reference image; Or use a rigid or non-rigid registration method to register the vascular images of some or all frames in the dynamic vascular video, obtain the corresponding registered dynamic vascular video, then perform an averaging process on the vascular images of some or all frames in the registered dynamic vascular video, and use the obtained averaged vascular image as the reference image.

3. A method for quantifying vascular dynamics video registration based on a vascular skeleton according to claim 1, characterized in that The step S2 further includes: Use the optical flow field from the to-be-registered image to the reference image as the dense deformation field; Or use B-spline curve fitting to obtain the deformation field from the to-be-registered image to the reference image as the dense deformation field.

4. A method for vascular dynamics video registration and quantification based on a vascular skeleton according to claim 1, characterized in that The step S3 is specifically: S31. According to the dynamic vascular video, obtain the pixel values of each frame of vascular image in the dynamic vascular video and the vascular morphology in each frame of vascular image; S32. Use the binary method and the morphological skeleton method respectively to obtain the vascular map and the skeleton map corresponding to each frame of vascular image according to the pixel values and the vascular morphology; S33. In each dense deformation field, use the pixel positions of the blood vessels in the vascular map to modify the displacement values of the corresponding pixel positions of the blood vessels in the obtained dense deformation field to the displacement values of the pixel positions corresponding to the nearest skeleton in the skeleton map at the corresponding positions in the dense deformation field, obtain the modified dense deformation field, and further obtain the modified dense deformation field from all to-be-registered images to the reference image.

5. A method for vascular dynamics video registration and quantization based on vascular skeleton according to claim 4, characterized in that: The binary method in the step S32 is set according to the following formula: Wherein, I(x, y) represents the pixel value of the point where the pixel (x, y) is located, T(x, y) represents the binarization threshold of the pixel (x, y), m(x, y) and s(x, y) respectively represent the local mean and local standard deviation of the pixel (x, y) within a preset neighborhood, R represents the dynamic range of the standard deviation, k, p, and q are all constants, and e is a constant; The morphological skeleton method in the step S32 is set according to the following formula: Among them, S represents the skeleton, represents the erosion operation, represents the dilation operation, A represents the input binary image, and B represents the structuring element used.

6. A method for vascular dynamics video registration quantization based on vascular skeleton according to claim 4, characterized in that: The modified dense deformation field obtained in the step S33 is specifically set according to the following formula: Among them, b j represents the index j pixel position where the blood vessel is located in the blood vessel map, s j represents the pixel position of the nearest skeleton to b j in the corresponding skeleton map, s i represents the index i pixel position of the nearby skeleton to b j in the corresponding skeleton map, V(b j ) represents the displacement value of the pixel position where the blood vessel is located at b j , V(s j ) represents the displacement value of the pixel position where the blood vessel is located at s j , B represents the set of pixel positions where the blood vessels are located in the blood vessel map, S represents the set of pixel positions where the skeletons are located in the skeleton map, ||*||1 represents the L1 norm, and argmin represents taking the minimum value; The skeleton in the step S33 is the corresponding skeleton in the skeleton image or the skeleton of the skeleton image corresponding to the local depth within the field of view.

7. A method for vascular dynamics video registration quantification based on vascular skeleton according to claim 1, characterized in that, The step S4 is specifically as follows: S41. Apply the modified dense deformation field to each frame of the corresponding vascular image to obtain each frame of registered vascular image, and obtain all frames of registered vascular images, thereby obtaining a registered dynamic vascular video; S42. According to the registered dynamic vascular video, obtain the vascular information at all pixel positions where the blood vessels are located on each frame of the registered vascular image; S43. According to the reference image, obtain the vascular information at all pixel positions where the blood vessels are located on the reference image; S44. Compare and process the vascular information at each pixel position where the blood vessels are located on the registered vascular image with the vascular information at the corresponding pixel position where the blood vessels are located on the reference image respectively to obtain the relative change of the vascular information at all pixel positions where the blood vessels are located; S45. Take the relative change of the vascular information at each pixel position where the blood vessels are located as the local response map at the corresponding pixel position, thereby obtaining the local response maps at all pixel positions; S46. Perform averaging or summation processing on all or part of the local response maps to obtain the global response map of the registered vascular image; S47. Obtain the dynamic change curve of the blood vessels over time according to the global response maps of all registered vascular images.

8. A method for vascular dynamics video registration quantization based on vascular skeleton according to claim 1, characterized in that: The vascular information in the step S4 includes blood vessel diameter, blood flow velocity, and blood flow rate.

9. A vascular dynamics video registration and quantification device based on vascular skeleton for implementing the vascular dynamics video registration and quantification method based on vascular skeleton according to any one of claims 1-8, characterized in that, Including: An optical coherence tomography imaging device for performing dynamic OCTA detection and imaging on a target area; Or a fluorescence angiography imaging device for performing dynamic fluorescence signal detection and imaging on a target area; Or a laser speckle imaging device for performing dynamic speckle signal detection and imaging on a target area; Or an ultrasonic localization microscopy imaging device for performing dynamic ultrasonic signal detection and imaging on a target area; Or a magnetic resonance device for performing dynamic magnetic signal detection and imaging on a target area; Or an X-ray computed tomography device for performing dynamic ray signal detection and imaging on a target area; Or a photoacoustic imaging device for performing dynamic photoacoustic signal detection and imaging on a target area; One or more processors for generating vascular graphs and skeleton graphs, obtaining a dense deformation field between a to-be-registered image and a reference image, modifying the dense deformation field, generating local response graphs and global response graphs, and realizing vascular dynamics video registration quantification based on vascular skeletons.

10. The vascular dynamics video registration and quantification device based on a vascular skeleton according to claim 9, characterized in that: The optical coherence tomography device described above is one of the following: Comprising a low-coherence light source, an interferometer, and a detector; Or comprising a low-coherence light source, an interferometer, and a spectrometer; Or comprising a swept-source broad-spectrum light source, an interferometer, and a detector.

Images