A three-dimensional full-eye retina blood vessel map construction method

Abstract

The application discloses a kind of three-dimensional full eye retina blood vessel atlas construction methods.The steps include: obtaining original retina blood vessel data and pre-processing;Extract the key point of blood vessel surface and construct the three-dimensional grid model of smoothing and refinement;Screen the vertex of the surface where retina deep blood vessel is located, and calculate the geodesic distance and orientation from the position of optic nerve to each vertex, construct three-dimensional latitude and longitude coordinate model;According to projection method, project three-dimensional latitude and longitude coordinate to two-dimensional plane;Feature extraction is carried out to the retina blood vessel at the surface vertex, combined with the extracted feature and three-dimensional latitude and longitude coordinate model, obtain three-dimensional latitude and longitude feature atlas model.The three-dimensional full eye retina blood vessel atlas construction method disclosed in the application overcomes the problem of cross-globe sample matching through three-dimensional latitude and longitude coordinate model, establishes a standard three-dimensional retina blood vessel atlas template;Projection method is used to improve the visualization effect and the convenience of feature atlas calculation.

Description

Technical Field

[0001] The present invention relates to the technical field of medical image processing, and in particular to a method for constructing a three-dimensional whole-eye retinal vascular atlas. Background Art

[0002] Eye atlas not only provides a basic research tool for understanding complex visual functions, but also offers a new perspective and research method for understanding the intrinsic mechanisms of eye diseases, discovering biological markers for early diagnosis and efficacy evaluation of eye diseases, and establishing clinical personalized precision treatment.

[0003] Currently, there are no atlas templates specifically designed for retinal vascular imaging. Among traditional retinal vascular imaging methods, OCT lacks specific labeling capabilities, suffers from low resolution, and has a limited field of view. Confocal microscopy, on the other hand, requires the eyeball to be imaged to be cut open into a "four-leaf clover" structure, disrupting the integrity of the retinal vessels. Therefore, only light-sheet fluorescence microscopy can achieve complete whole-eye retinal vascular imaging, offering the advantages of high resolution and rapid, large-volume imaging.

[0004] However, different eyeballs have different shapes and are easily deformed during sample processing, making it impossible to directly match the retina's original standard hemispherical shape across samples. Therefore, it is necessary to develop a mapping method to map the retinal blood vessels of the entire eye to a unified spatial template for alignment, so as to achieve accurate mapping of the retinal blood vessels. Summary of the Invention

[0005] In order to solve the problems in the background technology, the present invention provides a method for constructing a three-dimensional whole-eye retinal vascular map, which solves the problem of matching difficulties caused by retinal surface deformation and other reasons in cross-sample matching, establishes a standard three-dimensional retinal vascular map template, and uses the projection method to improve the visualization effect and the convenience of feature map calculation, providing a reliable solution for the accurate drawing of retinal vascular maps.

[0006] The technical solution adopted in the present invention is:

[0007] The method for constructing a three-dimensional full-eye retinal vascular atlas comprises the following steps:

[0008] S1) Acquire a three-dimensional image of retinal blood vessels.

[0009] S2) preprocessing the retinal vascular three-dimensional image to obtain a preprocessed retinal vascular three-dimensional image. The preprocessing includes binarization, vascular cavity filling, skeletonization, and morphological processing in sequence. The morphological processing specifically includes: performing morphological processing using a voxel connected domain method, and retaining the connected domain with the largest volume.

[0010] S3) performing surface key point extraction and three-dimensional point cloud anomaly detection on the preprocessed retinal blood vessel three-dimensional image in sequence to obtain retinal blood vessel three-dimensional point cloud data.

[0011] Preferably, the rolling ball method is used to extract surface key points.

[0012] Preferably, an isolated point cloud detection method based on statistical distribution is used to perform three-dimensional point cloud anomaly detection.

[0013] S4) using a three-dimensional meshing method to establish a three-dimensional mesh model of the retinal blood vessels based on the three-dimensional point cloud data of the retinal blood vessels obtained in step S3: using a triangulation meshing method to convert surface key points in the three-dimensional point cloud data of the retinal blood vessels into a triangular mesh composed of vertices and triangular facets to obtain an original triangular mesh model, and then sequentially performing surface smoothing and refinement processing on the original triangular mesh model to obtain a three-dimensional mesh model of the retinal blood vessels.

[0014] S5) First, surface mesh vertices are extracted from the three-dimensional mesh model of the retinal blood vessels created in step S4, and a three-dimensional surface model is constructed: the triangular mesh model is rotated so that the optic nerve is at the top, the XY plane is divided into a plurality of xy subplanes, and the mesh vertex with the maximum z value in each xy subplane is used as the surface mesh vertex. The three-dimensional surface model is generated based on all surface mesh vertices. The surface mesh vertices are mesh vertices on the surface where the deep blood vessels are located.

[0015] Secondly, based on the 3D surface model, the geodesic distance and vector angle from the optic nerve to each surface mesh vertex are calculated, converted into longitude and latitude, and a 3D longitude and latitude coordinate model is constructed: the fast marching algorithm is used to calculate the geodesic distance from the optic nerve to each surface mesh vertex;

[0016] Get the maximum geodesic distance and use the following formula to calculate the equatorial radius:

[0017] R=2L max / π

[0018] Where R represents the equatorial radius, L max represents the maximum geodesic distance;

[0019] The geodesic distance of each surface mesh vertex is processed using the following formula to obtain the latitude of the surface mesh vertex:

[0020] lon i =π / 2-L i / R

[0021] Where, lon i Indicates the latitude of the i-th surface mesh vertex, L irepresents the geodesic distance of the i-th surface mesh vertex, and R represents the equatorial radius;

[0022] At the same time, for the projection point of each surface mesh vertex on the XY plane, the vector angle pointing from the optic nerve location to the surface mesh vertex is obtained, and the vector angle is converted into longitude.

[0023] S6) performing feature extraction on the three-dimensional surface model, combining the extracted features with the three-dimensional longitude and latitude coordinate model to obtain a three-dimensional longitude and latitude feature map model: performing feature extraction on the three-dimensional surface model to obtain a feature value at each surface grid vertex, taking the combination of the feature value at each surface grid vertex and the longitude and latitude of the surface grid vertex in the three-dimensional longitude and latitude coordinate model as a data point, and constructing a three-dimensional longitude and latitude feature map model.

[0024] In this step, the eigenvalues ​​at all surface mesh vertices constitute the eigenvalue distribution in the three-dimensional surface model and / or the three-dimensional latitude and longitude feature map model.

[0025] Preferably, the characteristics include one or more of vascular density, vascular tortuosity, vascular branch length, and vascular branch point density.

[0026] S7) processing a plurality of different retinal vascular three-dimensional images according to steps S2 to S6 to obtain a three-dimensional latitude and longitude feature map model corresponding to each of the three-dimensional retinal vascular images, mapping all of the three-dimensional latitude and longitude feature map models to the same spatial template by a spatial interpolation method for alignment to obtain aligned three-dimensional latitude and longitude feature map models corresponding to each of the three-dimensional retinal vascular images, and then averaging all of the aligned three-dimensional latitude and longitude feature map models to obtain an integrated three-dimensional latitude and longitude feature map model.

[0027] The spatial template is the upper hemisphere of a unit sphere positioned using a latitude and longitude grid.

[0028] The averaging process is performed according to the following formula:

[0029]

[0030] Where n is the total number of retinal vascular three-dimensional images, m is j is the eigenvalue distribution in the aligned 3D latitude and longitude feature map model corresponding to the j-th retinal vascular 3D image, and M is the average eigenvalue distribution in the integrated 3D latitude and longitude feature map model.

[0031] Among them, the method of the present invention can achieve cross-eye sample matching. When the multiple different retinal vascular three-dimensional images are complete retinal vascular three-dimensional images from different retinal samples collected by light-sheet fluorescence microscopy, the method of the present invention can construct an accurate integrated three-dimensional latitude and longitude feature map model.

[0032] Furthermore, the construction method further includes the following step: projecting the three-dimensional latitude and longitude coordinate model obtained in step S5 onto a two-dimensional plane for display using a projection method. This step can also replace the above-mentioned step S6.

[0033] Furthermore, the construction method further includes the following steps: using a projection method to project the three-dimensional latitude and longitude feature map model obtained in step S6 onto a two-dimensional plane for display.

[0034] Furthermore, the construction method further includes the following steps: using a projection method to project the integrated three-dimensional latitude and longitude feature map model obtained in step S7 onto a two-dimensional plane for display.

[0035] Preferably, the projection is performed using the Lambert equal-area cylindrical projection method.

[0036] The beneficial effects of the present invention are:

[0037] The method for constructing a three-dimensional whole-eye retinal vascular map described in the present invention overcomes the problem of cross-eye sample matching through a three-dimensional latitude and longitude coordinate model, establishes a standard three-dimensional retinal vascular map template, and uses a projection method to improve visualization effects and the convenience of feature map calculation. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] Figure 1 This is a flow chart of the method for constructing a retinal vascular atlas of the present invention.

[0039] Figure 2 It is the original three-dimensional image data of retinal blood vessels.

[0040] Figure 3 This is the three-dimensional retinal vascular map after binarization.

[0041] Figure 4 This is a three-dimensional retinal vascular map after skeletonization and morphological processing.

[0042] Figure 5 This is the key point map of the blood vessel surface after removing abnormal points.

[0043] Figure 6 This is a triangular mesh model of three-dimensional retinal blood vessels.

[0044] Figure 7 Schematic diagram of vector angle.

[0045] Figure 8 is the latitude and longitude coordinate model of the three-dimensional retinal blood vessels.

[0046] Figure 9 It is a two-dimensional plane map after the latitude and longitude coordinate model is projected by Lambert equal-area cylindrical projection. DETAILED DESCRIPTION

[0047] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0048] The specific embodiments of the present invention are as follows:

[0049] Example 1

[0050] The process of the present invention is as follows Figure 1 The steps of the method of the present invention will be further described below in conjunction with this embodiment:

[0051] S1) Acquire a complete three-dimensional image of retinal blood vessels, wherein the three-dimensional image of retinal blood vessels is preferably a 3D grayscale image composed of a plurality of voxels.

[0052] Preferably, a complete three-dimensional image of the retinal vasculature is acquired by light-sheet fluorescence microscopy.

[0053] In the embodiment of the present invention, a light sheet fluorescence microscope is used to collect images of retinal samples as retinal blood vessel three-dimensional images. The original three-dimensional image of the retinal blood vessels is as follows: Figure 2 As shown, the size is 2048×2048×1350, and the actual size of a single voxel is 1.625μm×1.625μm×2.00μm.

[0054] S2) performing binarization, vascular cavity filling, skeletonization, and morphological processing on the retinal vascular three-dimensional image in sequence to obtain a preprocessed retinal vascular three-dimensional image.

[0055] Preferably, an adaptive threshold method is used to perform binarization processing to segment the retinal blood vessels and the background in the three-dimensional image data.

[0056] Preferably, a deep learning-based vascular cavity filling network is used to perform vascular cavity filling processing.

[0057] Preferably, the skeletonization process is performed using the 3D skeletonization method of K3M.

[0058] Preferably, a voxel connected domain method is used for morphological processing. During the processing, the largest connected domain is retained to reduce the influence of noise points on key point extraction.

[0059] In the embodiment of the present invention, the pre-processing process is specifically as follows:

[0060] S2.1) binarizing the 3D retinal image data using an adaptive threshold method to segment the retinal blood vessels from the background in the 3D retinal blood vessel image data;

[0061] S2.2) Since large blood vessels only have large signal intensity and grayscale values ​​at the vessel wall, there are cavities inside the blood vessels. Therefore, after binarization, the blood vessel cavities are filled with the help of a deep learning network to ensure that the complete binary blood vessels can be extracted. After the blood vessel cavities are filled, the binary image of the retinal blood vessels is as follows: Figure 3 As shown;

[0062] S2.3) extracting the vascular skeleton using the K3M 3D skeletonization method from the data after vascular cavity filling;

[0063] S2.4) The size of the retinal vascular skeleton image is 2048×2048×1350. The skeleton is then morphologically processed using the voxel connected domain method, retaining the largest connected object to obtain the pre-processed retinal vascular three-dimensional image data, as shown in Figure 4 shown.

[0064] S3) Surface key point extraction and 3D point cloud anomaly detection are sequentially performed on the preprocessed retinal vascular 3D image to obtain retinal vascular 3D point cloud data. By using an isolated point cloud detection method based on statistical distribution, the accuracy of surface extraction can be improved and the impact of noise on surface 3D model construction can be reduced.

[0065] Preferably, the rolling ball method is used to extract surface key points of the pre-processed three-dimensional image data of the retinal blood vessels.

[0066] Preferably, because key points at the retinal edge are typically sparsely distributed, they are not included in the scope of outlier detection. Therefore, for the original 3D point cloud data obtained after surface key points, 3D point cloud anomaly detection is performed only on the non-edge area. The non-edge area refers to the area near the optic nerve where blood vessels are densely distributed.

[0067] Optionally, the non-edge area can be set as follows: with the location of the optic nerve as the highest point and center position, the area with a Z-axis coordinate value higher than a preset Z-axis threshold is set as the non-edge area; or, with the location of the optic nerve as the highest point and center position, on the xy plane, a circular area with the location of the optic nerve as the center and within a preset radius threshold is set as the non-edge area.

[0068] Preferably, an isolated point cloud detection method based on statistical distribution is used to perform three-dimensional point cloud anomaly detection.

[0069] In the embodiment of the present invention, this step is specifically as follows: the rolling ball method is used to extract the key points of the pre-processed retinal blood vessel three-dimensional image data, and the data is saved in a point cloud format. The isolated point cloud detection method based on statistical distribution is used to remove key points with abnormal spatial positions in non-edge areas, wherein the number of surrounding points is set to 35, the standard deviation multiple is set to 2, and the key points in the surrounding areas with a distribution range exceeding two times the standard deviation are determined to be abnormal points. The non-edge area is the area where the Z-axis coordinate value is greater than 1200. The key points after the abnormal point removal are as follows: Figure 5 shown.

[0070] S4) Using a three-dimensional meshing method, a three-dimensional mesh model of the retinal blood vessels is established based on the three-dimensional point cloud data of the retinal blood vessels obtained in step S3.

[0071] Preferably, the three-dimensional mesh model adopts a triangular mesh model, which is generated by a triangulation method and mainly consists of vertices and triangular facets.

[0072] Furthermore, during this process, the original 3D mesh model undergoes surface smoothing to reduce abnormal protrusions. The surface is then refined, with the original triangular facets subdivided to increase the number of facets. The smoothed and refined 3D mesh model is then used for the next step of processing. Smoothing and refining the original triangular mesh model improves the accuracy of geodetic distance calculations.

[0073] In the embodiment of the present invention, this step is specifically as follows: for the key points of the surface after the outliers are removed, the triangulation method is used to convert them into a triangular mesh model composed of vertices and triangular facets. Then the triangular mesh model is subjected to surface smoothing processing, and the number of smoothing iterations is set to 5. Then the surface is refined, and the original triangular facets are subdivided to increase the number of facets, and the number of iterations is set to 2. The triangular mesh model is as follows Figure 6 shown.

[0074] S5) Extracting mesh vertices on the surface of deep-layer blood vessels from the three-dimensional mesh model of retinal blood vessels created in step S4 as surface mesh vertices, and constructing a three-dimensional surface model. Based on the three-dimensional surface model, the geodesic distance and vector angle from the optic nerve location to each surface mesh vertex are calculated, converted to longitude and latitude, and a three-dimensional longitude and latitude coordinate model is constructed.

[0075] This step is specifically:

[0076] S5.1) Filtering the vertices of the three-dimensional mesh model of the deep retinal blood vessels from the three-dimensional mesh model: Rotate the triangular mesh model so that the optic nerve is at the top, so that the upper surface of the triangular mesh model corresponds to the surface of the deep retinal blood vessels; Divide the XY plane into several xy subplanes, and define the mesh vertex with the maximum z value in each xy subplane as the mesh vertex on the surface of the deep blood vessels, i.e., the surface mesh vertex;

[0077] Constructing a 3D surface model: generating a 3D surface model based on all surface mesh vertices;

[0078] S5.2) Calculate the geodesic distances and vector angles of the mesh vertices, convert them to longitude and latitude, and construct a 3D longitude and latitude coordinate model based on the 3D surface model:

[0079] The fast marching algorithm is used to calculate the geodesic distance from each surface mesh vertex to the location of the optic nerve;

[0080] The maximum geodesic distance is processed using the following formula to obtain the equatorial radius:

[0081] R=2L max / π

[0082] Where R represents the equatorial radius, L max represents the maximum geodesic distance;

[0083] The latitude is obtained by taking the geodesic distance of each surface mesh vertex and using the following formula:

[0084] lon i =π / 2-L i / R

[0085] Where, lon i Indicates the latitude of the i-th surface mesh vertex, L i represents the geodesic distance of the i-th surface mesh vertex, and R represents the equatorial radius;

[0086] At the same time, for the projection point of each surface mesh vertex on the XY plane, obtain the vector angle pointing from the optic nerve location to the surface mesh vertex, and convert the vector angle into longitude;

[0087] Construct a three-dimensional latitude and longitude coordinate model: Generate a three-dimensional latitude and longitude coordinate model based on all surface mesh vertices and their corresponding longitude and latitude.

[0088] In the embodiment of the present invention, this step is specifically as follows:

[0089] S5.1) First, rotate the triangular mesh model so that the optic nerve is at the top, and the entire retinal vasculature is an inverted bowl. The surface where the deep retinal vessels are located is the top surface. Divide the XY plane into several small squares of 15×15 to cover the projection of the entire retinal vasculature on the XY plane. Then, find the mesh vertex with the highest Z value in each square. This vertex is the surface mesh vertex. Traverse all small squares to obtain all surface mesh vertices.

[0090] S5.2) Consider the entire retina as the northern hemisphere of the Earth, the location of the optic nerve as the "North Pole," and the location of the edge of the retina as the equator. First, find the longest geodesic distance L from the "North Pole" to the edge of the retina. max , then the radius of the sphere is:

[0091] R=2L max / π

[0092] Where R represents the equatorial radius, L max represents the maximum geodesic distance;

[0093] The geodesic distance from the North Pole to any surface mesh vertex is L i , then the corresponding latitude is:

[0094] lon i =π / 2-L i / R

[0095] Where, lon i Indicates the latitude of the i-th surface mesh vertex, L i represents the geodesic distance of the i-th surface mesh vertex, and R represents the equatorial radius;

[0096] Get the azimuth of each surface mesh vertex. The azimuth is specifically the vector angle from the "North Pole" to the surface mesh vertex on the XY plane. In this embodiment, Figure 7 As shown in the figure, the vector angle is the angle between the vector pointing from the optic nerve to the surface mesh vertex and the positive direction of the x-axis. When the surface mesh vertex is above the x-axis, the vector angle is positive and ranges from 0° to 180°; when the surface mesh vertex is below the x-axis, the vector angle is negative and ranges from 0° to -180°. When converted to longitude and latitude, angles between 0° and 180° are east longitude (W), and between 0° and -180° are west longitude (E).

[0097] The schematic diagram of the three-dimensional latitude and longitude coordinate model obtained in this embodiment is as follows Figure 8 shown.

[0098] S6) The three-dimensional longitude and latitude coordinate model obtained in step S5 is projected into two-dimensional space using the Lambert equal-area cylindrical projection method to establish a position mapping relationship between the retinal blood vessel surface and the two-dimensional plane to enhance the visualization effect and the convenience of feature map calculation. The three-dimensional longitude and latitude coordinate model obtained in this embodiment ( Figure 8 ) after being projected onto a two-dimensional plane. Figure 9 shown.

[0099] It should be noted that Figure 8 and Figure 9 These are just schematics. In practice, the surface mesh vertex distribution is quite random and will not exactly follow the shape of the longitude and latitude grid.

[0100] It can be seen that the three-dimensional latitude and longitude coordinate model obtained by the method of the present invention can be projected onto a two-dimensional plane using Lambert equal-area cylindrical projection, thereby enhancing the visualization effect and the convenience of feature map calculation.

[0101] Furthermore, the present invention also demonstrates in the following embodiments a method for constructing a feature map for a single retinal sample (Example 2) or multiple retinal samples (Examples 3 and 4) by combining the three-dimensional surface model and the three-dimensional latitude and longitude coordinate model obtained in step S5.

[0102] Example 2

[0103] This embodiment constructs a feature map for a three-dimensional image of blood vessels of a single retinal sample.

[0104] In this embodiment, steps S1 to S5 are the same as those in embodiment 1. The difference is that step S6 is to extract features from the three-dimensional surface model, and combine the extracted features with the three-dimensional longitude and latitude coordinate model to obtain a three-dimensional longitude and latitude feature map model.

[0105] Optionally, the extracted features are one or more of features such as blood vessel density, blood vessel tortuosity, blood vessel branch length, and blood vessel branch point density.

[0106] This embodiment takes the extracted feature of blood vessel branch point density as an example, and the specific process is as follows:

[0107] S6.1) For each surface mesh vertex in the three-dimensional surface model obtained in step S5 of Example 1, with any surface mesh vertex as the center of a sphere and a radius r = 45 as the radius, count the number N of vascular branch points within the radius r, where the spatial positions of the vascular branch points are obtained through the preprocessed three-dimensional image of the retinal blood vessels obtained in step S2.4 of Example 1.

[0108] Calculate the volume of the sphere The blood vessel branch point density value N / V at the vertex of the surface mesh is obtained.

[0109] S6.2) Calculate the blood vessel branch point density value of each surface mesh vertex according to step S6.1) to obtain the blood vessel branch point density value distribution of a single retinal sample at each surface mesh vertex position.

[0110] S6.3) The blood vessel branch point density value of each surface mesh vertex and its longitude and latitude in the three-dimensional latitude and longitude coordinate model are taken as a data point, and a three-dimensional latitude and longitude feature map model is generated based on the data points corresponding to all surface mesh vertices.

[0111] Furthermore, the method of the present invention can also be used to integrate the three-dimensional latitude and longitude feature map models of multiple retinal samples to obtain a unified integrated three-dimensional latitude and longitude feature map model. The integration process specifically refers to: obtaining the three-dimensional latitude and longitude feature map models of multiple retinal samples, that is, the vascular branch point density value distribution combined with the three-dimensional latitude and longitude coordinate model, mapping these three-dimensional latitude and longitude feature map models to the same spatial template and aligning them, and then calculating the average value of the vascular branch point density at each latitude and longitude position in the spatial template through averaging processing to obtain the feature map corresponding to the vascular branch point density.

[0112] Next, the present invention demonstrates, through Examples 3 and 4, an implementation scheme for integrating three-dimensional latitude and longitude feature map models of multiple retinal samples to obtain an integrated three-dimensional latitude and longitude feature map model.

[0113] Example 3

[0114] In this embodiment, the same process as steps S1 to S5 of embodiment 1 is used to obtain the three-dimensional longitude and latitude coordinate models corresponding to the three-dimensional vascular images of multiple different retinal samples, and the same process as embodiment 2 is used to obtain the three-dimensional longitude and latitude feature map model corresponding to each retinal sample. The three-dimensional longitude and latitude feature map models corresponding to all retinal samples are then mapped to the same spatial template for alignment.

[0115] The spatial template refers to the upper hemisphere of the unit sphere positioned using a longitude and latitude grid (all longitude and latitude combinations with a longitude range of -180° to 180° and an interval of 1°, and a latitude range of 0° to 90° and an interval of 1°).

[0116] The specific process of mapping alignment is:

[0117] According to Example 2, the vascular branch point density value distribution of each retinal vascular three-dimensional image at each surface grid vertex position is obtained. Combined with step S5 of Example 1, it can be seen that each surface grid vertex corresponds to a longitude and latitude coordinate. By spatial interpolation, the vascular branch point density value distribution m at all grid points of the longitude and latitude grid (all longitude and latitude combinations in the longitude range of -180° to 180° with an interval of 1 degree and the latitude range of 0° to 90° with an interval of 1°) is calculated. m is a 361x91 matrix, and m lon,lat Represents the density of blood vessel branch points at each longitude and latitude (lon, lat), thereby achieving the mapping alignment of the three-dimensional longitude and latitude coordinate model of a single retinal sample to the spatial template.

[0118] The eigenvalue distributions in the three-dimensional latitude and longitude feature map models corresponding to each retinal sample are mapped and aligned in the same way, and finally the aligned three-dimensional latitude and longitude feature map models corresponding to each retinal sample are obtained.

[0119] Example 4

[0120] Taking the eigenvalue distribution in the aligned three-dimensional latitude and longitude feature map model as the mapped aligned eigenvalue distribution, this embodiment obtains an integrated three-dimensional latitude and longitude feature map model by averaging the mapped aligned eigenvalue distributions corresponding to all retinal samples.

[0121] The specific process is:

[0122] For a single retinal sample, after the feature extraction of Example 2 and the mapping alignment of Example 3, the eigenvalue distribution at all grid points of the longitude and latitude grid (longitude range -180° to 180°, interval 1 degree, latitude range 0° to 90°, interval 1°) is obtained, that is, the eigenvalue distribution after mapping alignment, which is recorded as m j , j represents the serial number of the retinal sample.

[0123] After aligning the eigenvalue distributions of the mappings corresponding to multiple samples, an integrated three-dimensional latitude and longitude feature map model is obtained after averaging. The eigenvalue distribution in the integrated three-dimensional latitude and longitude feature map model is expressed as:

[0124]

[0125] Where n is the number of retinal samples, m j is the eigenvalue distribution at all longitude and latitude grid points in the aligned three-dimensional longitude and latitude feature map model corresponding to the j-th retinal sample, and M is the average eigenvalue distribution at all longitude and latitude grid points in the integrated three-dimensional longitude and latitude feature map model.

[0126] As can be seen, the 3D latitude and longitude coordinate model constructed by the present method can uniformly map retinal vessels of varying shapes into a standardized 3D spatial template, thereby resolving the problem of cross-sample matching. Furthermore, by mapping all retinal vessels in the eye into a unified spatial template for alignment, accurate retinal vascular mapping can be achieved.

[0127] In summary, the method for constructing a three-dimensional full-eye retinal vascular map described in the present invention overcomes the problem of cross-eye sample matching through a three-dimensional latitude and longitude coordinate model, establishes a standard three-dimensional retinal vascular map template, and uses the projection method to improve visualization effects and the convenience of feature map calculation.

[0128] The above specific embodiments are used to illustrate the present invention rather than to limit the present invention. Any modifications and changes made to the present invention within the spirit of the present invention and the protection scope of the claims shall fall within the protection scope of the present invention.

[0129] The above description is only a preferred embodiment of the present invention. Therefore, any equivalent changes or modifications made according to the structure, characteristics and principles described in the scope of the patent application of the present invention are included in the scope of the patent application of the present invention.

Claims

1. A method for constructing a three-dimensional whole-eye retinal vascular atlas, characterized in that: The following steps are involved: S1) obtaining a three-dimensional image of retinal blood vessels; S2) preprocessing the retinal vascular three-dimensional image to obtain a preprocessed retinal vascular three-dimensional image; S3) performing surface key point extraction and 3D point cloud anomaly detection on the preprocessed retinal vascular 3D image in sequence to obtain retinal vascular 3D point cloud data; S4) using a three-dimensional meshing method to establish a three-dimensional mesh model of the retinal blood vessels based on the three-dimensional point cloud data of the retinal blood vessels obtained in step S3; S5) extracting surface mesh vertices from the three-dimensional mesh model of the retinal blood vessels established in step S4, and constructing a three-dimensional surface model; Based on the three-dimensional surface model, the geodesic distance and vector angle from the location of the optic nerve to each surface grid vertex are calculated, converted into longitude and latitude, and a three-dimensional longitude and latitude coordinate model is constructed; The specific process of constructing a three-dimensional latitude and longitude coordinate model based on the three-dimensional surface model is as follows: The fast marching algorithm is used to calculate the geodesic distance from the location of the optic nerve to each surface mesh vertex; Get the maximum geodesic distance and use the following formula to calculate the equatorial radius: R=2L max / π Where R represents the equatorial radius, L max represents the maximum geodesic distance; The geodesic distance of each surface mesh vertex is processed using the following formula to obtain the latitude of the surface mesh vertex: lon i =π / 2-L i / R Where, lon i Indicates the latitude of the i-th surface mesh vertex, L i represents the geodesic distance of the i-th surface mesh vertex, and R represents the equatorial radius; At the same time, for the projection point of each surface mesh vertex on the XY plane, obtain the vector angle pointing from the optic nerve location to the surface mesh vertex, and convert the vector angle into longitude; S6) performing feature extraction on the three-dimensional surface model, combining the extracted features with the three-dimensional latitude and longitude coordinate model to obtain a three-dimensional latitude and longitude feature map model; S7) processing a plurality of different retinal vascular three-dimensional images according to steps S2 to S6 to obtain a three-dimensional latitude and longitude feature map model corresponding to each of the three-dimensional retinal vascular images, mapping all of the three-dimensional latitude and longitude feature map models to the same spatial template by a spatial interpolation method for alignment to obtain aligned three-dimensional latitude and longitude feature map models corresponding to each of the three-dimensional retinal vascular images, and then averaging all of the aligned three-dimensional latitude and longitude feature map models to obtain an integrated three-dimensional latitude and longitude feature map model; Among them, the spatial template is the upper hemisphere of the unit sphere positioned by the latitude and longitude grid; The averaging process is performed according to the following formula: Where n is the total number of retinal vascular three-dimensional images, m is j is the eigenvalue distribution in the aligned 3D latitude and longitude feature map model corresponding to the j-th retinal vascular 3D image, and M is the average eigenvalue distribution in the integrated 3D latitude and longitude feature map model.

2. The method for constructing a three-dimensional whole-eye retinal vascular atlas according to claim 1, characterized in that: In step S5, the surface mesh vertices are mesh vertices on the surface where the deep blood vessels are located; the process of constructing a three-dimensional surface model based on the three-dimensional mesh model is specifically as follows: rotating the triangular mesh model so that the optic nerve is at the top, dividing the XY plane into several xy sub-planes, and using the mesh vertex with the maximum z value on each xy sub-plane as the surface mesh vertex, and generating a three-dimensional surface model based on all the surface mesh vertices.

3. The method for constructing a three-dimensional whole-eye retinal vascular atlas according to claim 1, characterized in that: The step S6 specifically comprises: performing feature extraction on the three-dimensional surface model to obtain a feature value at each surface mesh vertex, taking the feature value at each surface mesh vertex and the combination of the longitude and latitude of the surface mesh vertex in the three-dimensional latitude and longitude coordinate model as a data point, and constructing a three-dimensional latitude and longitude feature map model.

4. The method for constructing a three-dimensional whole-eye retinal vascular atlas according to claim 1 or 3, characterized in that: The characteristics include one or more of vascular density, vascular tortuosity, vascular branch length, and vascular branch point density.

5. The method for constructing a three-dimensional whole-eye retinal vascular atlas according to claim 1, characterized in that: The construction method further includes: using a projection method to project the three-dimensional latitude and longitude coordinate model obtained in step S5, the three-dimensional latitude and longitude feature map model obtained in step S6, and / or the integrated three-dimensional latitude and longitude feature map model obtained in step S7 onto a two-dimensional plane for display.

6. The method for constructing a three-dimensional whole-eye retinal vascular atlas according to claim 5, characterized in that: Projected using the Lambert equal-area cylindrical projection.

7. The method for constructing a three-dimensional whole-eye retinal vascular atlas according to claim 1, characterized in that: In step S3, a rolling ball method is used to extract surface key points; and / or an isolated point cloud detection method based on statistical distribution is used to perform three-dimensional point cloud anomaly detection.

8. The method for constructing a three-dimensional whole-eye retinal vascular atlas according to claim 1, characterized in that: Step S4 specifically comprises: using a triangulation meshing method to convert the surface key points in the retinal blood vessel three-dimensional point cloud data into a triangular mesh composed of vertices and triangular facets to obtain an original triangular mesh model, and then performing surface smoothing and refinement processing on the original triangular mesh model in sequence to obtain a three-dimensional mesh model of the retinal blood vessels.

Images