Retinal Map Construction Method and Its Device, Computer Equipment, Storage Medium

Abstract

The method and device for constructing a retinal atlas, computer device, and storage medium proposed in the embodiments of the present application obtain initial retinal images of different subjects, perform hierarchical processing on the initial retinal images to obtain an overall thickness image and multiple hierarchical thickness images, perform image registration on the overall thickness image to obtain a deformation vector field, perform image transformation on the hierarchical thickness images according to the deformation vector field to obtain target retinal images, combine the target retinal images corresponding to different subjects to obtain a retinal thickness image tensor, decompose the retinal thickness image tensor to obtain eigenvectors corresponding to each pixel point in the target retinal image, perform clustering on all pixel points in the target retinal image based on the eigenvectors to obtain a clustering result, and mark the pixel points corresponding to the target retinal image in the hierarchical thickness images according to the clustering result to obtain a target retinal atlas, which can improve the accuracy of retinal partitioning.

Description

Technical Field

[0001] The present invention relates to the technical field of medical image processing, and particularly relates to a method and device for constructing a retinal map, a computer device, and a storage medium. Background Art

[0002] In the related art, the ETDRS grid method is used to partition the retina in the en-face direction. The ETDRS grid method realizes retinal partitioning by drawing circular regions in the en-face direction of the retina and further dividing the circular regions into multiple sub-regions. However, the circular regions and sub-regions have fixed sizes, and there are certain limitations in terms of the accuracy and robustness of feature extraction. Summary of the Invention

[0003] The main purpose of the embodiments of the present application is to propose a method and device for constructing a retinal map, a computer device, and a storage medium, which can improve the accuracy and robustness of feature extraction and achieve precise partitioning in the en-face direction of the retina.

[0004] To achieve the above object, a first aspect of the embodiments of the present application proposes a method for constructing a retinal map, the method comprising:

[0005] Obtaining initial retinal images of different subjects;

[0006] Performing layer processing on the initial retinal images to obtain an overall thickness image and a plurality of layer thickness images;

[0007] Performing image registration on the overall thickness image to obtain a deformation vector field;

[0008] Performing image transformation on the layer thickness images according to the deformation vector field to obtain target retinal images;

[0009] Combining the target retinal images corresponding to different subjects to obtain a retinal thickness image tensor;

[0010] Decomposing the retinal thickness image tensor to obtain a feature vector corresponding to each pixel point in the target retinal image, the feature vector being composed of pixel points at the same pixel position in the target retinal images of different subjects;

[0011] Performing clustering on all pixel points in the target retinal image based on the feature vector to obtain a clustering result;

[0012] Marking the pixel points in the layer thickness images corresponding to the target retinal image according to the clustering result to obtain a target retinal map corresponding to the layer thickness images.

[0013] In some embodiments, after performing layer processing on the initial retinal image to obtain an overall thickness image and a plurality of layer thickness images, the retinal atlas construction method further includes:

[0014] Performing image denoising on the overall thickness image to obtain a denoised overall thickness image;

[0015] Performing image denoising on the layer thickness images to obtain denoised layer thickness images.

[0016] In some embodiments, after performing layer processing on the initial retinal image to obtain an overall thickness image and a plurality of layer thickness images, the retinal atlas construction method further includes:

[0017] Performing image enhancement on the overall thickness image to obtain an enhanced overall thickness image;

[0018] Performing image enhancement on the layer thickness images to obtain enhanced layer thickness images.

[0019] In some embodiments, the performing image registration on the overall thickness image to obtain a deformation vector field includes:

[0020] Performing thickness averaging calculation on the overall thickness images of different subjects to obtain a template retinal image;

[0021] Performing polar coordinate space transformation on the template retinal image to obtain a polar coordinate template image;

[0022] Performing polar coordinate space transformation on the overall thickness image to obtain a polar coordinate thickness image;

[0023] Performing image registration on the polar coordinate template image and the polar coordinate thickness image to obtain a polar coordinate space deformation vector field;

[0024] Performing image space transformation on the polar coordinate space deformation vector field to obtain a deformation vector field.

[0025] In some embodiments, the performing image registration on the polar coordinate template image and the polar coordinate thickness image to obtain a polar coordinate space deformation vector field includes:

[0026] Scanning the polar coordinate template image according to the angle in the polar coordinate space to obtain a first thickness curve; the first thickness curve is the thickness information corresponding to each angle of the polar coordinate template image;

[0027] Scanning the polar coordinate thickness image according to the angle in the polar coordinate space to obtain a second thickness curve; the second thickness curve is the thickness information corresponding to each angle of the polar coordinate thickness image;

[0028] Extract the consistency features from the first thickness curve and the second thickness curve to obtain target feature points;

[0029] Perform image registration on the polar coordinate template image and the polar coordinate thickness image in the polar coordinate space according to the target feature points to obtain a polar coordinate space deformation vector field.

[0030] In some embodiments, clustering all pixel points in the target retinal image based on the feature vectors to obtain a clustering result, including:

[0031] Determine the clustering center;

[0032] Calculate the similarity between the feature vector corresponding to each pixel point in the target retinal image and the clustering center;

[0033] Cluster all pixel points in the target retinal image according to the similarity to obtain a clustering result.

[0034] In some embodiments, marking the pixel points in the layered thickness image according to the clustering result to obtain a target retinal map corresponding to the layered thickness image, including:

[0035] Mark the pixel points in the layered thickness image corresponding to the target retinal image according to the clustering result to obtain an initial retinal map;

[0036] Perform post-processing on the initial retinal map to remove outliers from the initial retinal map to obtain a target retinal map.

[0037] A second aspect of the embodiments of the present application proposes a retinal map construction device, the device includes:

[0038] An acquisition module, configured to acquire initial retinal images of different subjects;

[0039] A layering module, configured to perform layering processing on the initial retinal image to obtain an overall thickness image and a plurality of layered thickness images;

[0040] An image registration module, configured to perform image registration on the overall thickness image to obtain a deformation vector field;

[0041] An image transformation module, configured to perform image transformation on the layered thickness image according to the deformation vector field to obtain a target retinal image;

[0042] An image combination module, configured to combine the target retinal images corresponding to different subjects to obtain a retinal thickness image tensor;

[0043] A decomposition module, configured to decompose the retinal thickness image tensor to obtain an eigenvector corresponding to each pixel point in the target retinal image, where the eigenvector is composed of pixel points at the same pixel position in the target retinal images of different subjects;

[0044] A clustering module, configured to cluster all pixel points in the target retinal image based on the eigenvector to obtain a clustering result;

[0045] A retinal map construction module, configured to mark pixel points in the layered thickness image corresponding to the target retinal image according to the clustering result to obtain a target retinal map corresponding to the layered thickness image.

[0046] In a third aspect of the embodiments of the present application, a computer device is provided. The computer device includes a memory and a processor. Wherein, a program is stored in the memory, and when the program is executed by the processor, the processor is configured to execute the retinal map construction method according to any one of the embodiments of the first aspect of the present application.

[0047] In a fourth aspect of the embodiments of the present application, a storage medium is provided. The storage medium is a computer-readable storage medium, and the storage medium stores computer-executable instructions, and the computer-executable instructions are configured to cause a computer to execute the retinal map construction method according to any one of the embodiments of the first aspect of the present application.

[0048] The method and device for constructing a retinal map, computer device, and storage medium provided by the embodiments of the present application obtain initial retinal images of different subjects, perform hierarchical processing on the initial retinal images to obtain an overall thickness image and multiple hierarchical thickness images, perform image registration on the overall thickness image to obtain a deformation vector field, perform image transformation on the hierarchical thickness images according to the deformation vector field to obtain target retinal images, combine the target retinal images corresponding to different subjects to obtain a retinal thickness image tensor, decompose the retinal thickness image tensor to obtain a feature vector corresponding to each pixel point in the target retinal image, the feature vector is composed of pixel points at the same pixel position in the target retinal images of different subjects, perform clustering on all pixel points in the target retinal image based on the feature vector to obtain a clustering result, and mark the pixel points corresponding to the target retinal image in the hierarchical thickness image according to the clustering result to obtain a target retinal map corresponding to the hierarchical thickness image. By performing hierarchical processing on the initial retinal images, the embodiments of the present application can take into account the differences in the anatomical structures of different layers of the retina. Performing image registration on the overall thickness image can eliminate the individual differences in the retinal morphology of different subjects and avoid the influence of individual differences in retinal morphology on the retinal partitioning result. By combining the target retinal images of different subjects, the retinal thickness information between different regions of different subjects can be introduced, and the feature vectors corresponding to the pixel points of the retinal image can be clustered according to the correlation of the retinal thickness information. The retinal partitioning result is obtained according to the clustering result, improving the accuracy of retinal partitioning. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] Figure 1 is a flowchart of the method for constructing a retinal map provided by the embodiments of the present application;

[0050] Figure 2 is a schematic diagram of the relationship between the retinal thickness image tensor and the feature vector of the method for constructing a retinal map provided by the embodiments of the present application;

[0051] Figure 3 is Figure 1 a flowchart of the specific method of step S120 in

[0052] Figure 4 is Figure 1 a flowchart of the specific method of step S120 in

[0053] Figure 5 is Figure 1 a flowchart of the specific method of step S130 in

[0054] Figure 6 is Figure 5 a flowchart of the specific method of step S540 in

[0055] Figure 7 is Figure 1 a flowchart of the specific method of step S170 in

[0056] Figure 8 is Figure 1 a flowchart of the specific method of step S180 in

[0057] Figure 9 a module structure diagram of the retinal map construction device provided by the embodiments of the present application. Detailed implementation manners

[0058] In order to make the objectives, technical solutions and advantages of the present invention clearer, the present application will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application, and are not used to limit the present application.

[0059] It should be noted that although functional module division is performed in the device schematic diagram and the logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in a different order from the module division in the device or the order in the flowchart. Terms such as "first" and "second" in the specification, claims and the above drawings are used to distinguish similar objects, and do not necessarily need to describe a specific order or sequence.

[0060] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the technical field to which the present invention belongs. The terms used herein are only for the purpose of describing the embodiments of the present invention, and are not intended to limit the present invention.

[0061] The retina plays a crucial role in the formation of vision. In the eye axis direction, the retina can be divided into ten layers. In the en-face direction, the retina can be divided into different regions such as the fovea, parafovea, and perifoveal area. Currently, fundus imaging technologies such as OCT technology can clearly distinguish different layers of the retina, but cannot clearly divide different regions of the retina in the en-face direction. How to accurately partition the retina in the en-face direction is a major challenge for retinal feature extraction.

[0062] To address this issue, the commonly used method at present is the ETDRS grid method. By drawing a circular area in the en-face direction of the retina and further dividing the circular area into different sub-areas, the center of the circular area is aligned with the center of the macula, and the radius of the circular area and the scope of each sub-area are of fixed sizes. The ETDRS grid method has solved to a certain extent the problem of difficult partitioning in the en-face direction of the retina. However, this method of partitioning with a fixed shape does not strictly follow the anatomical structure in the en-face direction of the retina, and the fixed sizes do not consider the individual differences in the retinal structure. Therefore, this method has certain limitations in terms of the accuracy and robustness of feature extraction.

[0063] Brain atlases are a commonly used method for partitioning the brain. The retina and the brain have structural similarities. Drawing on the method of brain atlases to construct a retinal atlas is expected to achieve more accurate partitioning in the en-face direction of the retina. Currently, the construction of brain atlases mainly involves registering multi-person brain images to generate a brain image template, and then manually or automatically segmenting different structures of the brain on the brain image template. The implementation of this method is mainly based on the fact that different subjects have similar brain structures, and different brain regions can be clearly distinguished on brain images. However, on current retinal images, most regions in the en-face direction of the retina cannot be clearly distinguished. Therefore, the method for constructing brain atlases cannot be directly used to generate a retinal atlas.

[0064] Based on this, the main objective of the embodiments of this application is to propose a method for constructing a retinal atlas. By performing image registration on the overall thickness image, it is possible to eliminate the influence of individual differences in retinal morphology of different subject objects on the retinal partitioning result. By combining the target retinal images corresponding to different subject objects, it is possible to introduce the retinal thickness information between different regions of different subject objects. By clustering the feature vectors corresponding to the pixel points of the retinal image based on the correlation of the retinal thickness information, and obtaining the partitioning result of the anatomical structures of different layers of the retina according to the clustering result, and using the partitioning result as the retinal atlas, it is possible to take into account the differences in the anatomical structures of different layers of the retina and solve the problem of difficult partitioning in the en-face direction of the retina. Using a fixed shape and size to partition the retina is difficult to reflect the true anatomical information of the retina and is also easily affected by individual differences in retinal morphology. The embodiments of this application can improve the accuracy of retinal partitioning by eliminating individual differences in retinal morphology and introducing differences in the anatomical structures of different layers of the retina.

[0065] Referring to Figure 1 , according to the method for constructing a retinal atlas in the first aspect embodiments of this application, the method for constructing a retinal atlas includes but is not limited to steps S110 to S180.

[0066] Step S110: Obtain the initial retinal images of different subjects.

[0067] Step S120: Perform layer processing on the initial retinal images to obtain the overall thickness image and multiple layer thickness images.

[0068] Step S130: Perform image registration on the overall thickness image to obtain the deformation vector field.

[0069] Step S140: Perform image transformation on the layer thickness images according to the deformation vector field to obtain the target retinal images; Step S150: Combine the target retinal images corresponding to different subjects to obtain the retinal thickness image tensor.

[0070] Step S160: Decompose the retinal thickness image tensor to obtain the eigenvector corresponding to each pixel point in the target retinal image, where the eigenvector is composed of the pixel points at the same pixel position in the target retinal images of different subjects.

[0071] Step S170: Cluster all the pixel points in the target retinal image based on the eigenvectors to obtain the clustering result.

[0072] Step S180: Mark the pixel points in the layer thickness images corresponding to the target retinal image according to the clustering result to obtain the target retinal atlas corresponding to the layer thickness images.

[0073] In step S110, the initial retinal images of different subjects are obtained, where the initial retinal images are OCT images.

[0074] In step S120, the initial retinal images are layer processed using explorer or a trained deep learning model to achieve the segmentation of different layers of the retina, and the overall thickness en-face image and multiple layer thickness en-face images are obtained, where the deep learning model can be U-Net.

[0075] In step S130, in order to eliminate the differences in the overall retinal morphology between different subjects, the overall thickness en-face image of each subject is registered to a unified template to obtain the deformation vector field of the overall thickness en-face image of each subject registered to the template.

[0076] In step S140, in order to register the en-face image of the layer thickness, an image transformation is performed on the en-face image of the layer thickness corresponding to the subject according to the deformation vector field generated by registering the overall thickness en-face image of the subject to the template, to obtain a target retinal image, where the target retinal image is the registered en-face image of the layer thickness of the subject. If the layer thickness image is the en-face image of the k-th layer thickness, an image transformation is performed on the en-face image of the k-th layer thickness according to the deformation vector field to obtain a target retinal image.

[0077] In step S150, as Figure 2 shown, the target retinal images of subject 1 to subject M are combined to obtain an image sequence, and this image sequence is used as a retinal thickness image tensor, where the target retinal image is the registered en-face image of the k-th layer thickness. Combining the registered en-face images of the k-th layer thickness of different subjects can introduce individual differences in the retinal anatomical structures of different subjects to achieve retinal en-face direction partitioning and obtain the k-th layer retinal atlas.

[0078] In step S160, the retinal thickness image tensor is decomposed into multiple feature vectors. The relationship between the feature vectors and the retinal thickness image tensor is as Figure 2 shown. Each pixel point in the target retinal image corresponds to a feature vector. The feature vector corresponding to a certain pixel point in the k-th layer target retinal image is composed of the pixel points at the same pixel position as this pixel point in the registered en-face images of the k-th layer thickness of different subjects.

[0079] In step S170, based on the correlation of the thickness information between different regions of the retinal images of different subjects, all pixel points in the target retinal image are clustered based on the feature vectors to obtain a clustering result.

[0080] In step S180, the pixel points in the en-face image of the k-th layer thickness corresponding to the k-th layer target retinal image are marked according to the clustering result to obtain the k-th layer retinal atlas corresponding to the en-face image of the k-th layer thickness.

[0081] The method for constructing a retinal map proposed in the embodiments of the present application includes obtaining initial retinal images of different subjects, performing layer-by-layer processing on the initial retinal images to obtain an overall thickness image and multiple layer thickness images, performing image registration on the overall thickness image to obtain a deformation vector field, performing image transformation on the layer thickness images according to the deformation vector field to obtain target retinal images, combining the target retinal images corresponding to different subjects to obtain a retinal thickness image tensor, decomposing the retinal thickness image tensor to obtain eigenvectors corresponding to each pixel point in the target retinal image, where the eigenvectors are composed of pixel points at the same pixel position in the target retinal images of different subjects, clustering all pixel points in the target retinal image based on the eigenvectors to obtain a clustering result, and marking the pixel points in the layer thickness images corresponding to the target retinal image according to the clustering result to obtain a target retinal map corresponding to the layer thickness images. By performing layer-by-layer processing on the initial retinal images, the embodiments of the present application can take into account the differences in the anatomical structures of different layers of the retina. Performing image registration on the overall thickness image can eliminate the individual differences in the retinal morphology of different subjects and avoid the influence of individual differences in retinal morphology on the results of retinal partitioning. By combining the target retinal images of different subjects, the retinal thickness information between different regions of different subjects can be introduced, and all pixel points in the target retinal image can be clustered according to the correlation of the retinal thickness information. The retinal partitioning result is obtained according to the clustering result, improving the accuracy of retinal partitioning.

[0082] In some embodiments, as Figure 3 shown, after step S120, it specifically includes but is not limited to steps S310 to S320.

[0083] Step S310: Denoise the overall thickness image to obtain a denoised overall thickness image;

[0084] Step S320: Denoise the layer thickness images to obtain denoised layer thickness images.

[0085] In steps S310 to S320, in order to remove the noise in the overall thickness image and the layer thickness images and avoid the influence of the noise on the results of retinal partitioning, methods such as Gaussian denoising and non-local means denoising are used to denoise the overall thickness image and the layer thickness images.

[0086] In some embodiments, as Figure 4 shown, after step S120, it specifically includes but is not limited to steps S410 to S420.

[0087] Step S410: Enhance the overall thickness image to obtain an enhanced overall thickness image;

[0088] In step S420, the layer thickness image is subjected to image enhancement to obtain an enhanced layer thickness image.

[0089] In steps S410 to S420, in order to enhance detailed information such as the edges of the overall thickness image and the layer thickness image, histogram equalization is used to perform image enhancement on the overall thickness image and the layer thickness image.

[0090] In some embodiments, as Figure 5 shown, step S130 specifically includes but is not limited to steps S510 to S550.

[0091] In step S510, thickness averaging calculation is performed on the overall thickness images of different subjects to obtain a template retinal image;

[0092] In step S520, polar coordinate space transformation is performed on the template retinal image to obtain a polar coordinate template image;

[0093] In step S530, polar coordinate space transformation is performed on the overall thickness image to obtain a polar coordinate thickness image;

[0094] In step S540, the polar coordinate template image and the polar coordinate thickness image are subjected to image registration to obtain a polar coordinate space deformation vector field;

[0095] In step S550, the polar coordinate space deformation vector field is subjected to image space transformation to obtain a deformation vector field.

[0096] In step S510, the thicknesses of the corresponding pixel points in the overall thickness images of different subjects are added to obtain the thickness sum value of the pixel point. The ratio of the thickness sum value to the number of subjects is used as the average thickness value of the template retinal image at the pixel point. When the average thickness values of all pixel points are calculated, the template retinal image is obtained.

[0097] In steps S520 to S550, the retinal thickness is distributed centered on the macula. Compared with the original image space, the polar coordinate space is easier to process features with a similar central distribution. Therefore, in the embodiments of the present application, polar coordinate space transformation is used to transform the template retinal image and the overall thickness image from the original image space to the polar coordinate space to obtain a polar coordinate template image and a polar coordinate thickness image. In order to eliminate the influence of individual differences in retinal morphology of different subjects on retinal partitioning, in the polar coordinate space, the polar coordinate template image and the polar coordinate thickness image are subjected to image registration to generate a polar coordinate space deformation vector field, and the polar coordinate space deformation vector field is transformed back to the original image space to obtain a deformation vector field.

[0098] In some embodiments, as Figure 6As shown, step S540 specifically includes, but is not limited to, steps S610 to S640.

[0099] Step S610: Scan the polar coordinate template image according to the angle in the polar coordinate space to obtain a first thickness curve; the first thickness curve is the thickness information corresponding to each angle of the polar coordinate template image.

[0100] Step S620: Scan the polar coordinate thickness image according to the angle in the polar coordinate space to obtain a second thickness curve; the second thickness curve is the thickness information corresponding to each angle of the polar coordinate thickness image.

[0101] Step S630: Extract the consistency features of the first thickness curve and the second thickness curve to obtain target feature points.

[0102] Step S640: Perform image registration on the polar coordinate template image and the polar coordinate thickness image in the polar coordinate space according to the target feature points to obtain a polar coordinate space deformation vector field.

[0103] In steps S610 to S640, scan the polar coordinate template image and the polar coordinate thickness image along the angle dimension of the polar coordinate space, extract the thickness curve information corresponding to each angle, obtain the first thickness curve and the second thickness curve, extract the feature points with consistency in the first thickness curve and the second thickness curve as target feature points, and perform image registration on the polar coordinate template image and the polar coordinate thickness image in the polar coordinate space according to the target feature points to generate a polar coordinate space deformation vector field.

[0104] In some embodiments, as Figure 7 shown, step S170 specifically includes, but is not limited to, steps S710 to S730.

[0105] Step S710: Determine the cluster centers.

[0106] Step S720: Calculate the similarity between the feature vector corresponding to each pixel point in the target retinal image and the cluster centers.

[0107] Step S730: Cluster all the pixel points in the target retinal image according to the similarity to obtain a clustering result.

[0108] In step S710, the clustering method used is k-means clustering to determine the number of cluster centers and the pixel coordinates of each cluster center in the image space.

[0109] In step S720, calculate the Euclidean distance from the feature vector corresponding to each pixel point in the target retinal image to each cluster center. The larger the distance, the smaller the similarity between the feature vector corresponding to the pixel point in the target retinal image and the cluster center; the smaller the distance, the greater the similarity between the feature vector corresponding to the pixel point in the target retinal image and the cluster center.

[0110] In step S730, if the similarity between the feature vector corresponding to a pixel point in the target retinal image and a cluster center is less than a preset threshold, it indicates that the feature vector corresponding to the pixel point in the target retinal image is not similar to the cluster center, that is, the feature vector corresponding to the pixel point does not belong to the cluster center. If the similarity between the feature vector of the target retinal image pixel point and the cluster center is greater than or equal to the preset threshold, it indicates that the feature vector corresponding to the pixel point in the target retinal image is similar to the cluster center, and the feature vector corresponding to the pixel point in the target retinal image is clustered into the pixel point set to which the cluster center belongs. Through k-means clustering, multiple feature vectors are classified into different sets to cluster all pixel points in the target retinal image, thereby partitioning the retina to obtain a retinal map.

[0111] In some embodiments, as Figure 8 shown, step S180 specifically includes but is not limited to steps S810 to S820.

[0112] Step S810, mark the pixel points in the stratified thickness image corresponding to the target retinal image according to the clustering result to obtain an initial retinal map;

[0113] Step S820, post-process the initial retinal map to remove outliers from the initial retinal map to obtain the target retinal map.

[0114] In steps S810 to S820, the feature vectors of the target retinal image pixel points belong to different cluster centers. Different marks are assigned to the pixel points in the stratified thickness image corresponding to the target retinal image according to the different cluster centers to obtain an initial retinal map, that is, an initial retinal partition. The initial retinal partition is post-processed to remove outliers from the retinal partition, that is, the incorrect partition results in the retinal partition, to obtain the final retinal partition, and the final retinal partition is used as the target retinal map.

[0115] The following describes in detail the method for constructing a retinal map according to the embodiments of the present invention with a specific embodiment. It should be understood that the following description is only an exemplary illustration and not a specific limitation of the invention.

[0116] On the OCT image, different layers of the retina are segmented, the overall thickness information of the retina and the thickness information of each layer are calculated, an overall thickness image and a layered thickness image are generated, the thickness of the overall thickness images of different subjects is averaged to obtain a template retina image, the template retina image is subjected to a polar coordinate space transformation to obtain a polar coordinate template image, the overall thickness image is subjected to a polar coordinate space transformation to obtain a polar coordinate thickness image, the polar coordinate template image is scanned according to the angle in the polar coordinate space to obtain a first thickness curve, the polar coordinate thickness image is scanned according to the angle in the polar coordinate space to obtain a second thickness curve, the first thickness curve and the second thickness curve are subjected to consistency feature extraction to obtain target feature points, the polar coordinate template image and the polar coordinate thickness image are image-registered in the polar coordinate space according to the target feature points to obtain a polar coordinate space deformation vector field, the polar coordinate space deformation vector field is transformed back to the original image space to obtain a deformation vector field, the layered thickness image is image-transformed according to the deformation vector field to obtain a target retina image, the target retina images corresponding to different subjects are combined to obtain a retina thickness image tensor, the retina thickness image tensor is decomposed to obtain a feature vector corresponding to each pixel point in the target retina image, the feature vector is composed of pixel points at the same pixel position in the target retina images of different subjects, all pixel points in the target retina image are clustered based on k-means clustering according to all feature vectors to obtain a clustering result, the pixel points in the layered thickness image corresponding to the target retina image are marked according to the clustering result to obtain an initial retina atlas, and the initial retina atlas is post-processed to remove outliers from the initial retina atlas to obtain a target retina atlas.

[0117] The embodiment of the present application also provides a retina atlas construction device, such as Figure 9As shown, the retinal map construction device can implement the above-mentioned retinal map construction method. The device includes an acquisition module 910, a layering module 920, an image registration module 930, an image transformation module 940, an image combination module 950, a decomposition module 960, a clustering module 970, and a retinal map construction module 980. Among them, the acquisition module 910 is used to acquire initial retinal images of different subjects; the layering module 920 is used to perform layering processing on the initial retinal images to obtain an overall thickness image and multiple layered thickness images; the image registration module 930 is used to perform image registration on the overall thickness image to obtain a deformation vector field; the image transformation module 940 is used to perform image transformation on the layered thickness images according to the deformation vector field to obtain target retinal images; the image combination module 950 is used to combine the target retinal images corresponding to different subjects to obtain a retinal thickness image tensor; the decomposition module 960 is used to decompose the retinal thickness image tensor to obtain the feature vectors corresponding to each pixel point in the target retinal image, and the feature vectors are composed of the pixel points at the same pixel position in the target retinal images of different subjects; the clustering module 970 is used to cluster all pixel points in the target retinal image based on the feature vectors to obtain a clustering result; the retinal map construction module 980 is used to mark the pixel points in the layered thickness images corresponding to the target retinal image according to the clustering result to obtain the target retinal map corresponding to the layered thickness image.

[0118] The retinal map construction device of the embodiment of the present application is used to execute the retinal map construction method in the above-mentioned embodiment, and its specific processing process is the same as that of the retinal map construction method in the above-mentioned embodiment, and will not be elaborated here one by one.

[0119] The retinal atlas construction device proposed in the embodiments of the present application obtains the initial retinal images of different subjects, performs hierarchical processing on the initial retinal images to obtain the overall thickness image and multiple hierarchical thickness images, performs image registration on the overall thickness image to obtain the deformation vector field, performs image transformation on the hierarchical thickness images according to the deformation vector field to obtain the target retinal images, combines the target retinal images corresponding to different subjects to obtain the retinal thickness image tensor, decomposes the retinal thickness image tensor to obtain the feature vector corresponding to each pixel point in the target retinal image, the feature vector is composed of the pixel points at the same pixel position in the target retinal images of different subjects, performs clustering on all pixel points in the target retinal image based on the feature vector to obtain the clustering result, and marks the pixel points in the hierarchical thickness images corresponding to the target retinal image according to the clustering result to obtain the target retinal atlas corresponding to the hierarchical thickness images. By performing hierarchical processing on the initial retinal images, the embodiments of the present application can take into account the differences in the anatomical structures of different layers of the retina. Performing image registration on the overall thickness image can eliminate the individual differences in the retinal morphology of different subjects and avoid the influence of individual differences in retinal morphology on the retinal partitioning result. By combining the target retinal images of different subjects, the retinal thickness information between different regions of different subjects can be introduced, and all pixel points in the target retinal image are clustered according to the correlation of the retinal thickness information, and the retinal partitioning result is obtained according to the clustering result, improving the accuracy of retinal partitioning.

[0120] The embodiments of the present application further provide a computer device, including:

[0121] At least one processor, and,

[0122] A memory communicatively connected to the at least one processor; wherein,

[0123] The memory stores instructions, and the instructions are executed by the at least one processor so that when the at least one processor executes the instructions, it implements the retinal atlas construction method according to any one of the embodiments of the first aspect of the present application.

[0124] This computer device includes: a processor, a memory, an input / output interface, a communication interface, and a bus.

[0125] The processor can be implemented in a general-purpose CPU (Central Processing Unit), a microprocessor, an application-specific integrated circuit (ASIC), or one or more integrated circuits, etc., and is used to execute relevant programs to implement the technical solutions provided by the embodiments of the present application;

[0126] A memory can be implemented in the form of a ROM (Read Only Memory), a static storage device, a dynamic storage device, or a RAM (Random Access Memory), etc. The memory can store an operating system and other application programs. When implementing the technical solutions provided in the embodiments of this specification through software or firmware, the relevant program codes are stored in the memory and are called by the processor to execute the retinal map construction method according to any one of the embodiments of the first aspect of this application;

[0127] An input / output interface for implementing information input and output;

[0128] A communication interface for implementing communication interaction between this device and other devices. Communication can be achieved through wired means (such as USB, network cable, etc.) or wireless means (such as mobile network, WIFI, Bluetooth, etc.); and

[0129] A bus for transmitting information between various components of the device (such as a processor, a memory, an input / output interface, and a communication interface);

[0130] Among them, the processor, the memory, the input / output interface, and the communication interface are communicatively connected to each other inside the device through the bus.

[0131] The embodiments of this application also provide a storage medium, which is a computer-readable storage medium. The computer-readable storage medium stores computer-executable instructions, and the computer-executable instructions are used to cause a computer to execute the retinal map construction method according to any one of the embodiments of the first aspect of this application.

[0132] As a non-transitory computer-readable storage medium, the memory can be used to store non-transitory software programs and non-transitory computer-executable programs. In addition, the memory can include high-speed random access memory, and can also include non-transitory memory, such as at least one magnetic disk storage device, a flash memory device, or other non-transitory solid-state storage devices. In some embodiments, the memory optionally includes a memory remotely set relative to the processor, and these remote memories can be connected to the processor through a network. Examples of the above networks include, but are not limited to, the Internet, an enterprise intranet, a local area network, a mobile communication network, and combinations thereof.

[0133] The embodiments described in the embodiments of this application are for more clearly illustrating the technical solutions of the embodiments of this application, and do not constitute a limitation on the technical solutions provided by the embodiments of this application. Those skilled in the art can know that with the evolution of technology and the emergence of new application scenarios, the technical solutions provided by the embodiments of this application are equally applicable to similar technical problems.

[0134] Those skilled in the art can understand that Figures 1 to 8 the technical solutions shown in

[0135] do not constitute a limitation to the embodiments of the present application, and may include more or fewer steps than those shown in the figures, or combine certain steps, or different steps.

[0136] Those of ordinary skill in the art can understand that all or some of the steps in the methods disclosed above, and the functional modules / units in the systems and devices, can be implemented as software, firmware, hardware, and their appropriate combinations.

[0137] The terms "first", "second", "third", "fourth", etc. (if any) in the specification of the present application and the above-mentioned drawings are used to distinguish similar objects, and do not have to be used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged under appropriate circumstances, so that the embodiments of the present application described here can be implemented in an order different from those illustrated or described here. In addition, the terms "comprising" and "having" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or device that includes a series of steps or units does not have to be limited to those steps or units clearly listed, but may include other steps or units not clearly listed or inherent to these processes, methods, products, or devices.

[0138] It should be understood that in the present application, "at least one (item)" means one or more, and "a plurality" means two or more. The "and / or" is used to describe the relationship between associated objects and indicates that three relationships can exist. For example, "A and / or B" can mean: only A exists, only B exists, and both A and B exist at the same time. Here, A and B can be singular or plural. The character " / " generally means that the associated objects before and after are in an "or" relationship. "At least one (one) of the following" or its similar expressions refer to any combination of these items, including any combination of single items (ones) or plural items (ones). For example, at least one (one) of a, b, or c can mean: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", where a, b, and c can be single or multiple.

[0139] In several embodiments provided by the present application, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For example, the division of the units is only a logical function division. In actual implementation, there may be other division methods. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the displayed or discussed couplings or direct couplings or communication connections to each other can be through some interfaces. The indirect couplings or communication connections of the devices or units can be in electrical, mechanical or other forms.

[0140] The units described as separate components may or may not be physically separated. The components displayed as units may or may not be physical units, that is, they may be located in one place or distributed to multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0141] In addition, in each embodiment of the present application, the functional units can be integrated into one processing unit, or each unit can exist physically alone, or two or more units can be integrated into one unit. The above-mentioned integrated units can be implemented in the form of hardware or in the form of software functional units.

[0142] If the above-mentioned integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, in essence, or the part that contributes to the prior art, or all or part of this technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes multiple instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in each embodiment of the present application. And the aforementioned storage medium includes: various media such as USB flash drives, mobile hard disks, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical discs that can store programs.

[0143] The preferred embodiments of the embodiments of the present application have been described above with reference to the accompanying drawings, and thus do not limit the scope of rights of the embodiments of the present application. Any modifications, equivalent replacements, and improvements made by those skilled in the art without departing from the scope and essence of the embodiments of the present application shall be within the scope of rights of the embodiments of the present application.

Claims

1. A method for constructing a retinal map, characterized in that, The method includes: Obtaining initial retinal images of different subjects; Performing layer processing on the initial retinal images to obtain an overall thickness image and multiple layer thickness images; Performing image registration on the overall thickness image to obtain a deformation vector field; Performing image transformation on the layer thickness images according to the deformation vector field to obtain target retinal images; Combining the target retinal images corresponding to different subjects to obtain a retinal thickness image tensor; Decomposing the retinal thickness image tensor to obtain a feature vector corresponding to each pixel point in the target retinal image, where the feature vector is composed of pixel points at the same pixel position in the target retinal images of different subjects; Clustering all pixel points in the target retinal image based on the feature vector to obtain a clustering result; Marking the pixel points in the layer thickness images corresponding to the target retinal image according to the clustering result to obtain a target retinal atlas corresponding to the layer thickness images; The performing image registration on the overall thickness image to obtain a deformation vector field includes: Performing thickness averaging calculation on the overall thickness images of different subjects to obtain a template retinal image; Performing polar coordinate space transformation on the template retinal image to obtain a polar coordinate template image; Performing polar coordinate space transformation on the overall thickness image to obtain a polar coordinate thickness image; Performing image registration on the polar coordinate template image and the polar coordinate thickness image to obtain a polar coordinate space deformation vector field; Performing image space transformation on the polar coordinate space deformation vector field to obtain a deformation vector field; The performing image registration on the polar coordinate template image and the polar coordinate thickness image to obtain a polar coordinate space deformation vector field includes: Scanning the polar coordinate template image according to the angle in the polar coordinate space to obtain a first thickness curve; the first thickness curve is the thickness information corresponding to each angle of the polar coordinate template image; Scanning the polar coordinate thickness image according to the angle in the polar coordinate space to obtain a second thickness curve; the second thickness curve is the thickness information corresponding to each angle of the polar coordinate thickness image; Performing consistency feature extraction on the first thickness curve and the second thickness curve to obtain target feature points; Performing image registration on the polar coordinate template image and the polar coordinate thickness image in the polar coordinate space according to the target feature points to obtain a polar coordinate space deformation vector field.

2. The method for constructing a retinal map according to claim 1, wherein After the performing layer processing on the initial retinal images to obtain an overall thickness image and multiple layer thickness images, the retinal atlas construction method further includes: Performing image denoising on the overall thickness image to obtain a denoised overall thickness image; Performing image denoising on the layer thickness images to obtain denoised layer thickness images.

3. The method for constructing a retinal map according to claim 1, wherein After the performing layer processing on the initial retinal images to obtain an overall thickness image and multiple layer thickness images, the retinal atlas construction method further includes: Performing image enhancement on the overall thickness image to obtain an enhanced overall thickness image; Perform image enhancement on the layered thickness image to obtain an enhanced layered thickness image.

4. The method for constructing a retinal map according to any one of claims 1 to 3, characterized in that Clustering all pixel points in the target retinal image based on the feature vector to obtain a clustering result, including: Determine the clustering center; Calculate the similarity between the feature vector corresponding to each pixel point in the target retinal image and the clustering center; Cluster all pixel points in the target retinal image according to the similarity to obtain a clustering result.

5. The method for constructing a retinal map according to any one of claims 1 to 3, characterized in that, Marking the pixel points in the layered thickness image corresponding to the target retinal image according to the clustering result to obtain the target retinal atlas corresponding to the layered thickness image, including: Mark the pixel points in the layered thickness image corresponding to the target retinal image according to the clustering result to obtain an initial retinal atlas; Perform post-processing on the initial retinal atlas to remove outliers from the initial retinal atlas to obtain the target retinal atlas.

6. A retinal map construction device, characterized in that, The device includes: An acquisition module for acquiring initial retinal images of different subjects; A layering module for performing layering processing on the initial retinal image to obtain an overall thickness image and a plurality of layered thickness images; An image registration module for performing image registration on the overall thickness image to obtain a deformation vector field; An image transformation module for performing image transformation on the layered thickness image according to the deformation vector field to obtain a target retinal image; An image combination module for combining the target retinal images corresponding to different subjects to obtain a retinal thickness image tensor; A decomposition module for decomposing the retinal thickness image tensor to obtain the feature vector corresponding to each pixel point in the target retinal image, where the feature vector is composed of pixel points at the same pixel position in the target retinal images of different subjects; A clustering module for clustering all pixel points in the target retinal image based on the feature vector to obtain a clustering result; A retinal atlas construction module for marking the pixel points in the layered thickness image corresponding to the target retinal image according to the clustering result to obtain the target retinal atlas corresponding to the layered thickness image; Performing image registration on the overall thickness image to obtain a deformation vector field, including: Performing thickness average calculation on the overall thickness images of different subjects to obtain a template retinal image; Performing polar coordinate space transformation on the template retinal image to obtain a polar coordinate template image; Performing polar coordinate space transformation on the overall thickness image to obtain a polar coordinate thickness image; Performing image registration on the polar coordinate template image and the polar coordinate thickness image to obtain a polar coordinate space deformation vector field; Performing image space transformation on the polar coordinate space deformation vector field to obtain a deformation vector field; Performing image registration on the polar coordinate template image and the polar coordinate thickness image to obtain a polar coordinate space deformation vector field, including: Scanning the polar coordinate template image according to the angle in the polar coordinate space to obtain a first thickness curve; the first thickness curve is the thickness information corresponding to each angle of the polar coordinate template image; Scanning the polar coordinate thickness image according to the angle in the polar coordinate space to obtain a second thickness curve; the second thickness curve is the thickness information corresponding to each angle of the polar coordinate thickness image; Performing consistency feature extraction on the first thickness curve and the second thickness curve to obtain target feature points; Performing image registration on the polar coordinate template image and the polar coordinate thickness image in the polar coordinate space according to the target feature points to obtain a polar coordinate space deformation vector field.

7. Computer device, characterized in that The computer device includes a memory and a processor, wherein a program is stored in the memory, and when the program is executed by the processor, the processor is used to execute: The method for constructing a retinal atlas according to any one of claims 1 to 5.

8. A storage medium, the storage medium being a computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program, and when the computer program is executed by a computer, the computer is used to execute: The method for constructing a retinal atlas according to any one of claims 1 to 5.

Images