Method, device, medium and equipment for acquiring fundus features

By using pre-trained key point detection model and Bezier function fitting curve, the subjectivity problem of vascular arch detection in fundus image analysis is solved, and the accurate identification and analysis of fundus vascular arch is achieved, which improves the objectivity of the analysis and accuracy.

CN114387210BActive Publication Date: 2024-11-08EVISION TECH (BEIJING) CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202111463020.9
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2021-12-03
Publication Date
2024-11-08
Estimated Expiration
2041-12-03

AI Technical Summary

Technical Problem

In the prior art, fundus image analysis has strong artificial subjectivity, which makes it difficult to guarantee the accuracy and consistency of the analysis. Automatic analysis technology has great potential in saving labor costs, but there is a lack of effective methods in the detection and identification of fundus vascular arches.

Method used

By acquiring fundus images, the center point and auxiliary key points of the optic disc are determined using a pre-trained key point detection model, and combined with the Bezier function fitting curve, the fundus vascular arch fitting curve is accurately determined, thereby determining the fundus vascular arch.

Benefits of technology

Accurate detection and identification of fundus vascular arches is achieved, helping to identify eye categories, and providing valuable information for fundus abnormality analysis, improving the objectivity and accuracy of the analysis.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN114387210B_ABST
    Figure CN114387210B_ABST
Patent Text Reader

Abstract

The embodiment of the present invention provides a method, device, medium and equipment for acquiring fundus features, the method comprising: acquiring a fundus image; inputting the fundus image into a pre-trained key point detection model, determining the center point of the optic disc on the fundus image, and obtaining multiple auxiliary key points and at least two target key points; determining a fundus vascular arch fitting curve according to the optic disc center point, multiple auxiliary key points and at least two target key points, and then determining the fundus vascular arch. Accurately determining the vascular arch is not only helpful for identifying eyes, but also helpful for fundus disease analysis.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of fundus image feature analysis and extraction, and in particular to a method, device, medium and equipment for acquiring fundus features. Background Art

[0002] In the prior art, fundus image analysis is mainly performed manually. Since manual analysis is highly subjective, the accuracy and consistency of manual analysis are difficult to guarantee. Automatic analysis technology is not only more objective, but also has great advantages in saving labor costs, so it is increasingly sought after. Among the fundus image features, the vascular arch is one of the basic fundus features. Detecting and identifying it is not only helpful for distinguishing the eye (left and right eye), but also helps to understand fundus abnormalities and provide more valuable information for fundus disease analysis. Summary of the invention

[0003] In view of this, an object of the embodiments of the present invention is to provide a method, device, medium and equipment for acquiring fundus features, so as to determine the fundus vascular arch.

[0004] To achieve the above objectives, in a first aspect, an embodiment of the present invention provides a method for acquiring fundus features, which comprises:

[0005] Acquire fundus images;

[0006] Inputting the fundus image into a pre-trained key point detection model, determining the optic disc center point on the fundus image, and obtaining a plurality of auxiliary key points and at least two target key points;

[0007] A fundus vascular arch fitting curve is determined according to the optic disc center point, the multiple auxiliary key points and the at least two target key points, and the fundus vascular arch is determined by the fundus vascular arch fitting curve.

[0008] In some possible implementations, before acquiring the fundus image, it may also include: acquiring a fundus training image, annotating the fundus training image, obtaining a training sample, inputting the training sample into a deep learning key point detection network for training, and acquiring a key point detection model.

[0009] In some possible implementations, the labeling of the fundus training image to obtain a training sample may specifically include:

[0010] Determining the optic disc center point and the macula center of the fundus training image;

[0011] Determine an upper half of a main blood vessel and a lower half of a main blood vessel in the fundus training image;

[0012] Draw a vertical line through the center of the macula, intersect with the upper half of the main blood vessel to obtain the first target key point, intersect with the upper half of the main blood vessel to obtain the second target key point, or, take the center point of the optic disc as the center and draw a circle with an arbitrary length as the radius, intersect with the lower upper half of the main blood vessel to obtain the first target key point, and intersect with the upper half of the main blood vessel to obtain the second target key point;

[0013] Selecting a first auxiliary key point and a second auxiliary key point between the center point of the optic disc and the lower half of the main blood vessel, and determining the optimal coordinate values ​​of the first auxiliary key point and the second auxiliary key point by Bessel function fitting;

[0014] Selecting a third auxiliary key point and a fourth auxiliary key point between the center point of the optic disc and the upper half of the main blood vessel, and determining the optimal coordinate values ​​of the third auxiliary key point and the fourth auxiliary key point by Bessel function fitting;

[0015] The training sample is obtained according to the first target key point, the second target key point, the first auxiliary key point, the second auxiliary key point, the third auxiliary key point, the fourth auxiliary key point and the center point of the video disc.

[0016] In some possible implementations, the inputting the training samples into the deep learning key point detection network for training may specifically include:

[0017] Step S1: input the training sample into a deep learning key point detection network for convolution to obtain a high-dimensional feature vector graph;

[0018] Step S2: performing deconvolution dimensionality reduction processing on the high-dimensional feature vector image to obtain a one-dimensional feature vector image;

[0019] Step S3: using the one-dimensional feature vector graph to obtain the coordinates of the center point of the optic disc, the first target key point, the second target key point, the first auxiliary key point, the second auxiliary key point, the third auxiliary key point and the fourth auxiliary key point by using a fully connected method;

[0020] Step S4: Calculate the coordinate error values ​​corresponding to the first target key point, the second target key point, the first auxiliary key point, the second auxiliary key point, the third auxiliary key point, the fourth auxiliary key point and the center point of the optic disc and the marked coordinate points;

[0021] The above steps S1 to S4 are iterated continuously until the difference is reduced to a preset coordinate error value.

[0022] In some possible implementations, inputting the fundus image into a pre-trained key point detection model to generate a plurality of auxiliary key points on the fundus image may specifically include:

[0023] Input the training sample into a pre-trained key point detection model for convolution to obtain a high-dimensional feature vector graph;

[0024] Performing deconvolution dimensionality reduction processing on the high-dimensional feature vector graph to obtain a one-dimensional feature vector graph;

[0025] The one-dimensional feature vector graph is used in a fully connected manner to obtain a disc center point, multiple key points and at least two target key points; wherein the multiple auxiliary key points include a first auxiliary key point, a second auxiliary key point, a third auxiliary key point and a fourth auxiliary key point; and the at least two target key points include a first target key point and a second target key point.

[0026] In some possible implementations, determining the fundus vascular arch fitting curve according to the optic disc center point, the multiple auxiliary key points and the at least two target key points, and determining the fundus vascular arch through the fundus vascular arch fitting curve may specifically include:

[0027] Taking the optic disc center point as the starting endpoint and the first target key point as the terminating endpoint, a first vascular arch fitting curve is obtained by Bessel function fitting, wherein the first auxiliary key point and the second auxiliary key point are located between the optic disc center point and the first target key point, and the shape of the first vascular arch fitting curve is adjusted by adjusting the positions of the first auxiliary key point and the second auxiliary key point;

[0028] Taking the optic disc center point as the starting endpoint and the second target key point as the terminating endpoint, a second vascular arch fitting curve is obtained by Bessel function fitting, wherein the third auxiliary key point and the fourth auxiliary key point are located between the optic disc center point and the second target key point, and the shape of the second vascular arch fitting curve is adjusted by adjusting the positions of the third auxiliary key point and the fourth auxiliary key point;

[0029] The first vascular arch fitting curve and the second vascular arch fitting curve are connected and processed to obtain the fundus vascular arch fitting curve, and the fundus vascular arch is determined through the fundus vascular arch fitting curve.

[0030] In some possible implementations, the method may further include: determining the opening angle of the fundus vascular arch based on the optic disc center point, the first target key point, and the second target key point.

[0031] In a second aspect, the present invention provides a device for acquiring fundus features, comprising:

[0032] An acquisition module, for acquiring fundus images;

[0033] A key point detection module, inputting the fundus image into a pre-trained key point detection model, determining the optic disc center point on the fundus image, and obtaining a plurality of auxiliary key points and at least two target key points;

[0034] A determination module determines a fundus vascular arch fitting curve according to the optic disc center point, the multiple auxiliary key points and the at least two target key points, and determines the fundus vascular arch through the fundus vascular arch fitting curve.

[0035] In a third aspect, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein the program, when executed by a processor, implements any method for acquiring fundus features.

[0036] In a fourth aspect, the present invention provides a device for acquiring fundus features, comprising:

[0037] one or more processors;

[0038] A storage device for storing one or more programs;

[0039] When the one or more programs are executed by the one or more processors, the one or more processors implement any one of the above-mentioned methods for acquiring fundus features.

[0040] The above technical solution has the following beneficial effects:

[0041] The embodiment of the present invention inputs the fundus image into a pre-trained key point detection model, determines the optic disc center point on the fundus image, and obtains multiple auxiliary key points and at least two target key points; determines the fundus vascular arch fitting curve according to the optic disc center point, multiple auxiliary key points and at least two target key points, and determines the fundus vascular arch through the fundus vascular arch fitting curve. By performing detection with the pre-trained key point detection model, the vascular arch fitting curve can be accurately determined, and then the fundus vascular arch can be accurately determined, which is not only helpful for helping to identify the eye, but also helpful for understanding fundus abnormalities, and is the basis for subsequent fundus disease analysis. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0043] Figure 1 is a flow chart of a method for acquiring fundus features according to an embodiment of the present invention;

[0044] Figure 2A is a schematic diagram of obtaining training samples according to an embodiment of the present invention;

[0045] Figure 2B is a schematic diagram of obtaining a key point detection model according to an embodiment of the present invention;

[0046] Figure 3 is a schematic diagram of obtaining a vascular arch fitting curve by Bessel function fitting according to an embodiment of the present invention;

[0047] Figure 4 A schematic diagram of measuring the opening angle of fundus vascular arches according to an embodiment of the present invention;

[0048] Figure 5A is a functional block diagram of a device for acquiring fundus features according to an embodiment of the present invention;

[0049] Figure 5B is a functional block diagram of another device for acquiring fundus features according to an embodiment of the present invention;

[0050] Figure 6 is a functional block diagram of a computer-readable storage medium according to an embodiment of the present invention;

[0051] Figure 7 It is a functional block diagram of a device for acquiring fundus features according to an embodiment of the present invention. DETAILED DESCRIPTION

[0052] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0053] The vascular arch is the curve formed by the aortic veins of the fundus. Because of its arc-like shape, it is called a vascular arch. It can be divided into the fundus arterial vascular arch and the fundus venous vascular arch.

[0054] Embodiment 1

[0055] Figure 1 FIG. 1 is a flow chart of a method for acquiring fundus features according to an embodiment of the present invention. Figure 1 As shown, it includes the following steps:

[0056] S110: Acquire fundus image;

[0057] S120: Inputting the fundus image into a pre-trained key point detection model, determining the optic disc center point on the fundus image, and obtaining a plurality of auxiliary key points and at least two target key points;

[0058] S130: Determine a fundus vascular arch fitting curve according to the optic disc center point, multiple auxiliary key points and at least two target key points, and determine the fundus vascular arch through the fundus vascular arch fitting curve.

[0059] In some embodiments, the training process of the key point detection model in step S120 specifically includes:

[0060] S121: Acquire a fundus blood vessel training image;

[0061] S122: Annotate the fundus training image to obtain training samples;

[0062] Figure 2A FIG. 1 is a schematic diagram of obtaining training samples as an example in an embodiment of the present invention. Figure 2A As shown:

[0063] Determine the optic disc center point D and the macula center point H of the fundus training image, determine the upper half of the main blood vessels and the lower half of the main blood vessels in the fundus training image,

[0064] A vertical line is drawn through the macula center point H, intersecting with the lower half L2 of the main blood vessel to obtain the first target key point G, and intersecting with the upper half L1 of the main blood vessel to obtain the second target key point E, or a circle is drawn with the optic disc center point D as the center and any length as the radius, preferably, the radius length is between 0.5PD and 1.5PD (PD is the optic disc diameter), intersecting with the lower half L2 of the main blood vessel to obtain the first target key point G', and intersecting with the upper half of the main blood vessel to obtain the second target key point E';

[0065] Select the first auxiliary key point B and the second auxiliary key point F between the optic disc center point D and the lower half L2 of the main blood vessel, and determine the optimal coordinate values ​​of the first auxiliary key point B and the second auxiliary key point F according to Bessel function fitting;

[0066] A third auxiliary key point A and a fourth auxiliary key point C are selected between the optic disc center point D and the upper half L1 of the main blood vessel, and the optimal coordinate values ​​of the third auxiliary key point A and the fourth auxiliary key point C are determined according to Bessel function fitting;

[0067] The training samples are obtained according to the center point D of the optic disc, the first target key point G, the second target key point E, the first auxiliary key point B, the second auxiliary key point F, the third auxiliary key point A and the fourth auxiliary key point C.

[0068] S123: input the training sample into the deep learning key point detection network for training to obtain a key point detection model;

[0069] In some embodiments, in step S123, the training samples are input into the deep learning key point detection network for training to obtain the key point detection model, which may specifically include:

[0070] Step S1: input the training sample into the deep learning key point detection network for convolution to obtain a high-dimensional feature vector graph;

[0071] Step S2: performing deconvolution dimensionality reduction processing on the high-dimensional feature vector image to obtain a one-dimensional feature vector image;

[0072] Step S3: using a fully connected method to obtain the coordinates of the center point of the optic disc, the first target key point, the second target key point, the first auxiliary key point, the second auxiliary key point, the third auxiliary key point and the fourth auxiliary key point from the one-dimensional feature vector graph;

[0073] Step S4: Calculate the coordinate error values ​​corresponding to the coordinate points of the obtained disc center point, the first target key point, the second target key point, the first auxiliary key point, the second auxiliary key point, the third auxiliary key point and the fourth auxiliary key point;

[0074] The above steps S1 to S4 are continuously iterated until the difference is reduced to a preset coordinate error value, and a key point detection model is obtained.

[0075] Figure 2B FIG. 1 is a schematic diagram of obtaining a key point detection model as an example in an embodiment of the present invention. Figure 2B As shown in the figure: the training samples and the annotated coordinate points are input into the deep learning key point detection network for convolution to obtain a high-dimensional feature vector map (i.e., feature map). The high-dimensional feature vector map is reduced in dimension through the pooling layer to remove class redundant information and extract the most important features. The coordinates of the optic disc center point, the first target key point, the second target key point, the first auxiliary key point, the second auxiliary key point, the third auxiliary key point and the fourth auxiliary key point are obtained through a full connection method, i.e., (x1', y1'), (x2', y2'), (x3', y3')..., the difference between the obtained coordinate points and the annotated coordinate points (x1, y1), (x2, y2), (x3, y3)... is calculated, i.e., |x1'-x1|=Δx1, |y1'-y1|=Δy1..., the above steps are continuously iterated until the difference (Δx1, Δy1...) is reduced to the preset coordinate error.

[0076] In some embodiments, the deep learning key point detection network can be a CNN convolutional neural network.

[0077] In some embodiments, in step S120, the fundus image is input into a pre-trained key point detection model, and an optic disc center point, multiple auxiliary key points and at least two target key points are generated on the fundus image. Specifically:

[0078] The training samples are input into the deep learning key point detection model for convolution to obtain a high-dimensional feature vector graph;

[0079] Perform deconvolution dimensionality reduction processing on the high-dimensional feature vector graph to obtain a one-dimensional feature vector graph;

[0080] The one-dimensional feature vector graph is fully connected to obtain the optic disc center point, multiple auxiliary key points and at least two target key points.

[0081] In some embodiments, the plurality of auxiliary key points include a first auxiliary key point, a second auxiliary key point, a third auxiliary key point and a fourth auxiliary key point, the at least two target key points include a first target key point and a second target key point, the first auxiliary key point and the second auxiliary key point are located between the center point of the optic disc and the first target key point, and the third auxiliary key point and the fourth auxiliary key point are located between the center point of the optic disc and the second target key point;

[0082] In some embodiments, in step 130, determining a fundus vascular arch fitting curve according to the optic disc center point, multiple auxiliary key points and at least two target key points, and determining the fundus vascular arch through the fundus vascular arch fitting curve specifically includes:

[0083] The optic disc center point, the first auxiliary key point among the multiple auxiliary key points, the second auxiliary key point and the first target key point among the at least two target key points are sequentially connected, with the optic disc center point as the starting endpoint and the first target key point as the ending endpoint, and a first vascular arch fitting curve is obtained by Bessel function fitting;

[0084] The optic disc center point, the third auxiliary key point among the multiple auxiliary key points, the fourth auxiliary key point and the second target key point among the at least two target key points are sequentially connected, with the optic disc center point as the starting endpoint and the second target key point as the ending endpoint, and a second vascular arch fitting curve is obtained by Bessel function fitting;

[0085] The first vascular arch fitting curve and the second vascular arch fitting curve are connected and processed to obtain a fundus vascular arch fitting curve, and the fundus vascular arch is determined by the fundus vascular arch fitting curve.

[0086] Figure 3 is a schematic diagram of obtaining a vascular arch fitting curve by Bessel function fitting according to an embodiment of the present invention, such as Figure 3 As shown:

[0087] In this embodiment, the plurality of key points include a first auxiliary key point B, a second auxiliary key point F, a third auxiliary key point A and a fourth auxiliary key point C, a first target key point G and a second target key point E, the first auxiliary key point B and the second auxiliary key point F are located between the center point D of the optic disc and the first target key point G, and the third auxiliary key point A and the fourth auxiliary key point C are located between the center point D of the optic disc and the second target key point E;

[0088] Taking the optic disc center point D as the starting endpoint, the first target key point G as the ending endpoint, the first auxiliary key point B and the second auxiliary key point F as the control points, the first vascular arch fitting curve L1 is obtained by Bessel function fitting, wherein the first auxiliary key point B and the second auxiliary key point F are located between the optic disc center point D and the first target key point G, and the shape of the first vascular arch fitting curve L1 is adjusted by the first auxiliary key point B and the second auxiliary key point F, so that the shape of the first vascular arch fitting curve L1 is consistent with or close to the fundus vascular arch;

[0089] Taking the optic disc center point D as the starting endpoint, the second target key point E as the ending endpoint, and the third auxiliary key point A and the fourth auxiliary key point C as the control points, the second vascular arch fitting curve L2 is obtained by Bessel function fitting, wherein the third auxiliary key point A and the fourth auxiliary key point C are located between the optic disc center point D and the second target key point E, and the shape of the second vascular arch fitting curve L2 is adjusted by the third auxiliary key point A and the fourth auxiliary key point C, so that the second vascular arch fitting curve L2 is consistent with or close to the fundus vascular arch;

[0090] The first vascular arch fitting curve L1 and the second vascular arch fitting curve L2 are connected to obtain the fundus vascular arch fitting curve L.

[0091] In some embodiments, the opening angle of the fundus vascular arch may also be measured based on the optic disc center point, the first target key point, and the second target key point. Figure 4 FIG. 1 is a schematic diagram of measuring the opening angle of a vascular arch according to an embodiment of the present invention as an example. Figure 4 As shown: Connect the center point D of the optic disc with the first target key point G to determine the first line segment L DG , connect the center point D of the optic disc with the second target key point E, and determine the second line segment L DE , the first line segment L DG With the second line segment L DE The angle between them is the opening angle θ of the vascular arch. By measuring the opening angle of the vascular arch, it is convenient to analyze the change trend of the vascular arch.

[0092] The beneficial technical effects of the embodiments of the present invention include:

[0093] The embodiment of the present invention can accurately determine the fundus vascular arch fitting curve, laying the foundation for subsequent fundus abnormality analysis. The embodiment of the present invention detects the vascular arch as the basic feature of the fundus, and through the deep learning network key point detection model, it can accurately determine the fundus vascular arch fitting curve, and then determine the fundus vascular arch, which is not only helpful for helping to identify the eye, but also helpful for understanding the fundus abnormality analysis.

[0094] Embodiment 2

[0095] Figure 5A FIG. 1 is a functional block diagram of a device for acquiring fundus features according to an embodiment of the present invention. Figure 5A As shown, the device 500 includes:

[0096] An acquisition module 510 is used to acquire a fundus image;

[0097] The key point detection module 520 inputs the fundus image into a pre-trained key point detection model, determines the optic disc center point on the fundus image, and obtains multiple auxiliary key points and at least two target key points;

[0098] The determination module 530 determines the fundus vascular arch fitting curve according to the optic disc center, multiple auxiliary key points and at least two target key points.

[0099] In this embodiment, the key point detection module 520 is used to generate an optic disc center point, multiple auxiliary key points and at least two target key points on the fundus image based on a pre-trained key point detection model. Specifically, the training sample is input into the deep learning key point detection model for convolution to obtain a high-dimensional feature vector image; the high-dimensional feature vector image is deconvoluted for dimensionality reduction to obtain a one-dimensional feature vector image; the one-dimensional feature vector image is used in a fully connected manner to obtain the optic disc center point, multiple auxiliary key points and at least two target key points. Preferably, the fundus feature acquisition device 500 also includes a model training module for acquiring a key point detection model. The model training module first acquires a fundus vascular training image, then annotates the fundus training image to obtain a training sample, and finally inputs the training sample into a deep learning key point detection network for training to obtain a key point detection model. Specifically, Figure 2AAs shown, the optic disc center point D and the macula center point H of the fundus training image are determined, the upper half and the lower half of the main blood vessels in the fundus training image are determined, a vertical line is drawn through the macula center point H, and the first target key point G is intersected with the lower half L2 of the main blood vessel, and the second target key point E is intersected with the upper half L1 of the main blood vessel, or a circle is drawn with the optic disc center point D as the center and an arbitrary length as the radius, preferably, the radius length is between 0.5PD and 1.5PD (PD is the optic disc diameter), and the first target key point G' is intersected with the lower half L2 of the main blood vessel, and the second target key point E' is intersected with the upper half of the main blood vessel;

[0100] Select the first auxiliary key point B and the second auxiliary key point F between the optic disc center point D and the lower half L2 of the main blood vessel, and determine the optimal coordinate values ​​of the first auxiliary key point B and the second auxiliary key point F according to Bessel function fitting;

[0101] A third auxiliary key point A and a fourth auxiliary key point C are selected between the optic disc center point D and the upper half L1 of the main blood vessel, and the optimal coordinate values ​​of the third auxiliary key point A and the fourth auxiliary key point C are determined according to Bessel function fitting;

[0102] The training samples are obtained according to the center point D of the optic disc, the first target key point G, the second target key point E, the first auxiliary key point B, the second auxiliary key point F, the third auxiliary key point A and the fourth auxiliary key point C.

[0103] The training samples are input into the deep learning key point detection network for convolution to obtain a high-dimensional feature vector image, the high-dimensional feature vector image is deconvolved to reduce the dimensionality to obtain a one-dimensional feature vector image, and the one-dimensional feature vector image is used in a fully connected manner to obtain the coordinates of the optic disc center point, the first target key point, the second target key point, the first auxiliary key point, the second auxiliary key point, the third auxiliary key point and the fourth auxiliary key point, and the coordinate error values ​​corresponding to the coordinates of the obtained optic disc center point, the first target key point, the second target key point, the first auxiliary key point, the second auxiliary key point, the third auxiliary key point and the fourth auxiliary key point are calculated; the model training module is continuously iterated until the difference is reduced to the preset coordinate error value to obtain the key point detection model.

[0104] In this embodiment, the determination module 530 is used to determine the fundus vascular arch fitting curve according to the optic disc center point, multiple auxiliary key points and at least two target key points. Specifically, the multiple auxiliary key points in this embodiment are a first auxiliary key point, a second auxiliary key point, a third auxiliary key point and a fourth auxiliary key point, and at least two target key points are a first target key point and a second target key point. The first auxiliary key point and the second auxiliary key point are located between the center point of the optic disc and the first eye key point, and the third auxiliary key point and the fourth auxiliary key point are located between the center point of the optic disc and the second target key point. The optic disc center point, the first auxiliary key point, the second auxiliary key point and the first target key point are sequentially connected, with the optic disc center point as the starting endpoint and the first target key point as the ending endpoint, and a first vascular arch fitting curve is obtained by Bessel function fitting. The optic disc center point, the third auxiliary key point, the fourth auxiliary key point and the second target key point are sequentially connected, with the optic disc center point as the starting endpoint and the second target key point as the ending endpoint, and a second vascular arch fitting curve is obtained by Bessel function fitting. The first vascular arch fitting curve and the second vascular arch fitting curve are connected and processed to obtain a fundus vascular arch fitting curve.

[0105] In some embodiments, Figure 5B As shown, after the fundus vascular arch fitting curve is determined, an opening angle measurement module 540 is also included, which is used to measure the opening angle of the vascular arch. Specifically, the center point of the optic disc is connected to the first target key point to determine the first line segment L1, and the center of the optic disc is connected to the second target key point to determine the second line segment L2. The angle between the first line segment L1 and the second line segment L2 is the opening angle θ of the vascular arch.

[0106] The technicians in the relevant field can clearly understand that for the convenience and simplicity of description, only the division of the above-mentioned functional units and modules is used as an example for illustration. In practical applications, the above-mentioned function allocation can be completed by different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiment can be integrated in a processing unit, or each unit can exist physically separately, or two or more units can be integrated in one unit. The above-mentioned integrated unit can be implemented in the form of hardware or in the form of software functional units. In addition, the specific names of the functional units and modules are only for the convenience of distinguishing each other, and are not used to limit the scope of protection of this application. The specific working process of the units and modules in the above-mentioned system can refer to the corresponding process in the aforementioned method embodiment, which will not be repeated here.

[0107] Embodiment 3

[0108] Figure 6600 is a functional block diagram of a computer-readable storage medium according to an embodiment of the present invention. Figure 6 As shown, a computer program 610 is stored in the computer-readable storage medium. When the computer program 610 is executed by the processor, the following is realized:

[0109] Acquire fundus images;

[0110] Input the fundus image into a pre-trained key point detection model, determine the center point of the optic disc on the fundus image, and obtain multiple auxiliary key points and at least two target key points;

[0111] A fundus vascular arch fitting curve is determined according to the optic disc center, multiple auxiliary key points and at least two target key points, and the fundus vascular arch is determined through the fundus vascular arch fitting curve.

[0112] If the integrated module / 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 present invention implements all or part of the processes in the above-mentioned embodiment method, and can also be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium, and the computer program can implement the steps of the above-mentioned various method embodiments when executed by the processor. Among them, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form. The computer-readable medium may include: any entity or device capable of carrying the computer program code, recording medium, U disk, mobile hard disk, disk, optical disk, computer memory, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), electric carrier signal, telecommunication signal and software distribution medium. Of course, there are other ways of readable storage media, such as quantum memory, graphene memory, etc. It should be noted that the content contained in the computer-readable medium can be appropriately increased or decreased according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, computer-readable media does not include electrical carrier signals and telecommunication signals.

[0113] Implementation Four

[0114] Figure 7 FIG. 1 is a functional block diagram of a device for acquiring fundus features according to an embodiment of the present invention. Figure 7 As shown, it includes one or more processors 701 , a communication interface 702 , a memory 703 and a communication bus 704 , wherein the processor 701 , the communication interface 702 , and the memory 703 communicate with each other via the communication bus 704 .

[0115] Memory 703, used for storing computer programs;

[0116] The processor 701 is used to execute the program stored in the memory 703 to achieve:

[0117] Acquire fundus images;

[0118] Input the fundus image into a pre-trained key point detection model, determine the center point of the optic disc on the fundus image, and obtain multiple auxiliary key points and at least two target key points;

[0119] A fundus vascular arch fitting curve is determined according to the optic disc center, multiple auxiliary key points and at least two target key points, and the fundus vascular arch is determined through the fundus vascular arch fitting curve.

[0120] In some embodiments, the processing performed by the processor 701, before acquiring the fundus image, also includes: acquiring a fundus vascular training image, annotating the fundus vascular training image, obtaining training samples, inputting the training samples into a deep learning key point detection network for training, and acquiring a key point detection model.

[0121] In some embodiments, in the process performed by the processor 701, the fundus training image is annotated to obtain a training sample, which specifically includes:

[0122] Determine the optic disc center and macula center of the fundus training image;

[0123] determining an upper half of a main blood vessel and a lower half of a main blood vessel in a fundus training image;

[0124] A vertical line is drawn through the center of the macula, intersecting with the lower half of the main blood vessel to obtain the first target key point, and intersecting with the upper half of the main blood vessel to obtain the second target key point, or a circle is drawn with the center of the optic disc as the center and an arbitrary length as the radius, and the circle intersects with the lower half of the main blood vessel to obtain the first target key point, and intersects with the upper half of the main blood vessel to obtain the second target key point;

[0125] Selecting a first auxiliary key point and a second auxiliary key point between the center of the optic disc and the lower half of the main blood vessel, and determining the optimal coordinate values ​​of the first auxiliary key point and the second auxiliary key point by Bessel function fitting;

[0126] A third auxiliary key point and a fourth auxiliary key point are selected between the center of the optic disc and the upper half of the main blood vessel, and the optimal coordinate values ​​of the third auxiliary key point and the fourth auxiliary key point are determined by Bessel function fitting;

[0127] A training sample is obtained based on the first target key point, the second target key point, the first auxiliary key point, the second auxiliary key point, the third auxiliary key point, the fourth auxiliary key point and the center of the optic disc.

[0128] In some embodiments, in the process performed by the processor 701, the training samples are input into the deep learning key point detection network for training, specifically including:

[0129] Step S1: input the training sample into the deep learning key point detection network for convolution to obtain a high-dimensional feature vector graph;

[0130] Step S2: performing deconvolution dimensionality reduction processing on the high-dimensional feature vector image to obtain a one-dimensional feature vector image;

[0131] Step S3: using a fully connected method to obtain the coordinates of the center point of the optic disc, the first target key point, the second target key point, the first auxiliary key point, the second auxiliary key point, the third auxiliary key point and the fourth auxiliary key point from the one-dimensional feature vector graph;

[0132] Step S4: Calculate the coordinate error values ​​corresponding to the first target key point, the second target key point, the first auxiliary key point, the second auxiliary key point, the third auxiliary key point, the fourth auxiliary key point and the center point of the optic disc and the marked coordinate points;

[0133] The above steps S1 to S4 are iterated continuously until the difference is reduced to a preset coordinate error value.

[0134] In some embodiments, in the process performed by the processor 701, the fundus image is input into a pre-trained key point detection model, and a plurality of auxiliary key points are generated on the fundus image, specifically including:

[0135] The training samples are input into the deep learning key point detection model for convolution to obtain a high-dimensional feature vector graph;

[0136] Perform deconvolution dimensionality reduction processing on the high-dimensional feature vector graph to obtain a one-dimensional feature vector graph;

[0137] The one-dimensional feature vector graph is fully connected to obtain the center point of the optic disc, multiple key points and at least two target key points;

[0138] Wherein, the multiple auxiliary key points include a first auxiliary key point, a second auxiliary key point, a third auxiliary key point and a fourth auxiliary key point; and the at least two target key points include a first target key point and a second target key point.

[0139] In some embodiments, in the processing performed by the processor 701, determining the fundus vascular arch fitting curve according to the optic disc center point, multiple auxiliary key points and at least two target key points specifically includes:

[0140] Taking the optic disc center as the starting endpoint and the first target key point as the ending endpoint, a first vascular arch fitting curve is obtained by Bessel function fitting, wherein the first auxiliary key point and the second auxiliary key point are located between the optic disc center point and the first target key point, and the shape of the first vascular arch fitting curve is adjusted by adjusting the positions of the first auxiliary key point and the second auxiliary key point;

[0141] Taking the optic disc center point as the starting endpoint and the second target key point as the ending endpoint, obtaining the second vascular arch fitting curve by Bessel function fitting, wherein the third auxiliary key point and the fourth auxiliary key point are located between the optic disc center point and the second target key point, and adjusting the position of the third auxiliary key point and the fourth auxiliary key point to adjust the shape of the second vascular arch fitting curve;

[0142] The first vascular arch fitting curve and the second vascular arch fitting curve are connected to obtain a fundus vascular arch fitting curve.

[0143] In some embodiments, the processing performed by the processor 701 further includes: determining the opening angle of the fundus vascular arch based on the optic disc center point, the first target key point and the second target key point.

[0144] The communication bus mentioned in the above electronic device can be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus. The communication bus can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, only one thick line is used in the figure, but it does not mean that there is only one bus or one type of bus. The communication interface is used for communication between the above electronic device and other devices.

[0145] The memory may include a random access memory (RAM) or a non-volatile memory (NVM), such as at least one disk memory. Optionally, the memory may also be at least one storage device located away from the aforementioned processor.

[0146] The above-mentioned processor can be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it can also be a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components.

[0147] The systems, devices, modules or units described in the above embodiments may be implemented by computer chips or entities, or by products with certain functions. A typical implementation device is a computer. Specifically, the computer may be, for example, a personal computer, a laptop computer, a vehicle-mounted human-computer interaction device, a cellular phone, a camera phone, a smart phone, a personal digital assistant, a media player, a navigation device, an email device, a game console, a tablet computer, a wearable device, or a combination of any of these devices.

[0148] Although the present application provides method operation steps as described in the embodiments or flow charts, more or fewer operation steps may be included based on conventional or non-creative means. The order of steps listed in the embodiments is only one way of executing the order of many steps and does not represent the only execution order. When the device or terminal product in practice is executed, it can be executed in sequence or in parallel according to the method shown in the embodiments or the drawings (for example, in a parallel processor or multi-threaded processing environment, or even in a distributed data processing environment).

[0149] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the existence of other identical elements in the process, method, article or device including the elements.

[0150] Each embodiment in this specification is described in a related manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the device, electronic device, and readable storage medium embodiments, since they are basically similar to the method embodiments, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiments.

[0151] The above description is only a preferred embodiment of the present invention and is not intended to limit the protection scope of the present invention. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention are included in the protection scope of the present invention.

Claims

1. A method for acquiring fundus features, characterized in that: include: Acquire fundus images; Input the fundus image into a pre-trained key point detection model, determine the optic disc center point on the fundus image, and obtain multiple auxiliary key points and at least two target key points; wherein the multiple auxiliary key points are points selected between the optic disc center point and the upper half of the main blood vessel, and between the optic disc center point and the lower half of the main blood vessel; the at least two target key points are points where a vertical line through the macula center intersects with the upper half and the lower half of the main blood vessel, or where a circle is drawn with the optic disc center point as the center and an arbitrary length as the radius, and intersects with the upper half and the lower half of the main blood vessel; A fundus vascular arch fitting curve is determined according to the optic disc center point, the multiple auxiliary key points and the at least two target key points, and the fundus vascular arch is determined by the fundus vascular arch fitting curve.

2. The method according to claim 1, characterized in that: Before acquiring the fundus image, the method further includes: Acquire fundus training images; Annotating the fundus training image to obtain a training sample; The training samples are input into a deep learning key point detection network for training to obtain a key point detection model.

3. The method according to claim 2, characterized in that The step of labeling the fundus training image to obtain a training sample specifically includes: Determining the optic disc center point and the macula center point of the fundus training image; Determining an upper half of a main blood vessel and a lower half of a main blood vessel in the fundus training image; A vertical line is drawn through the center of the macula to intersect with the lower half of the main blood vessel to obtain a first target key point, and intersect with the upper half of the main blood vessel to obtain a second target key point, or a circle is drawn with the center of the optic disc as the center and an arbitrary length as the radius, and intersects with the lower half of the main blood vessel to obtain the first target key point, and intersects with the upper half of the main blood vessel to obtain the second target key point; Selecting a first auxiliary key point and a second auxiliary key point between the center point of the optic disc and the lower half of the main blood vessel, and determining the optimal coordinate values ​​of the first auxiliary key point and the second auxiliary key point by Bessel function fitting; Selecting a third auxiliary key point and a fourth auxiliary key point between the center point of the optic disc and the upper half of the main blood vessel, and determining the optimal coordinate values ​​of the third auxiliary key point and the fourth auxiliary key point by Bessel function fitting; The training sample is obtained according to the first target key point, the second target key point, the first auxiliary key point, the second auxiliary key point, the third auxiliary key point, the fourth auxiliary key point and the center point of the optic disc.

4. The method according to claim 2, characterized in that: The step of inputting the training samples into a deep learning key point detection network for training specifically includes: Step S1: inputting the marked key point coordinates and the training samples into a deep learning key point detection network for convolution to obtain a high-dimensional feature vector graph; Step S2: performing deconvolution dimensionality reduction processing on the high-dimensional feature vector image to obtain a one-dimensional feature vector image; Step S3: using the one-dimensional feature vector graph to obtain the coordinates of the center point of the optic disc, the first target key point, the second target key point, the first auxiliary key point, the second auxiliary key point, the third auxiliary key point and the fourth auxiliary key point by using a fully connected method; Step S4: Calculate the coordinate error values ​​corresponding to the first target key point, the second target key point, the first auxiliary key point, the second auxiliary key point, the third auxiliary key point, the fourth auxiliary key point and the center point of the optic disc and the marked coordinate point; The above steps S1 to S4 are iterated continuously until the difference is reduced to a preset coordinate error value.

5. The method according to claim 1, characterized in that The step of inputting the fundus image into a pre-trained key point detection model, determining the optic disc center point on the fundus image, and obtaining a plurality of auxiliary key points and at least two target key points specifically includes: Inputting the fundus image into a pre-trained key point detection model for convolution to obtain a high-dimensional feature vector image; Performing deconvolution dimensionality reduction processing on the high-dimensional feature vector graph to obtain a one-dimensional feature vector graph; The one-dimensional feature vector graph is used in a fully connected manner to obtain the optic disc center point, multiple auxiliary key points and at least two target key points; Wherein, the multiple auxiliary key points include a first auxiliary key point, a second auxiliary key point, a third auxiliary key point and a fourth auxiliary key point; and the at least two target key points include a first target key point and a second target key point.

6. The method according to claim 5, characterized in that The step of determining the fundus vascular arch fitting curve according to the optic disc center point, the multiple auxiliary key points and the at least two target key points specifically includes: Taking the optic disc center point as the starting endpoint and the first target key point as the terminating endpoint, a first vascular arch fitting curve is obtained by Bessel function fitting, wherein the first auxiliary key point and the second auxiliary key point are located between the optic disc center point and the first target key point, and the shape of the first vascular arch fitting curve is adjusted by adjusting the positions of the first auxiliary key point and the second auxiliary key point; Taking the optic disc center point as the starting endpoint and the second target key point as the terminating endpoint, a second vascular arch fitting curve is obtained by Bessel function fitting, wherein the third auxiliary key point and the fourth auxiliary key point are located between the optic disc center point and the second target key point, and the shape of the second vascular arch fitting curve is adjusted by adjusting the positions of the third auxiliary key point and the fourth auxiliary key point; The first vascular arch fitting curve and the second vascular arch fitting curve are connected and processed to obtain the fundus vascular arch fitting curve, and the fundus vascular arch is determined by the fundus vascular arch fitting curve.

7. The method according to claim 6, characterized in that Also includes: The opening angle of the fundus vascular arch is determined based on the optic disc center, the first target key point and the second target key point.

8. A device for acquiring fundus features, characterized in that: include: An acquisition module, for acquiring fundus images; A key point detection module inputs the fundus image into a pre-trained key point detection model, determines the center point of the optic disc on the fundus image, and obtains a plurality of auxiliary key points and at least two target key points; wherein the plurality of auxiliary key points are points selected between the center point of the optic disc and the upper half of the main blood vessel, and between the center point of the optic disc and the lower half of the main blood vessel; the at least two target key points are points where a vertical line through the center point of the macula intersects with the upper half and the lower half of the main blood vessel, or points where a circle is drawn with the center point of the optic disc as the center and with an arbitrary length as the radius, intersecting with the upper half and the lower half of the main blood vessel; A determination module determines a fundus vascular arch fitting curve according to the optic disc center point, the multiple auxiliary key points and the at least two target key points, and determines the fundus vascular arch through the fundus vascular arch fitting curve.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the method for acquiring fundus features as described in any one of claims 1 to 7 is implemented.

10. A device for acquiring fundus features, characterized in that: It includes: one or more processors; A storage device for storing one or more programs; When the one or more programs are executed by the one or more processors, the one or more processors implement the method for acquiring fundus features as described in any one of claims 1-7.