A method, device and medium for analyzing red eye level combined with vascular morphology analysis

Abstract

The present invention provides a method, device and medium for analyzing eye redness level combined with vascular morphology analysis, and relates to the field of image processing technology. The method includes inputting a target eye image into a pre-trained segmentation model to obtain a conjunctival region and a corneal region, determining a limbal region based on the conjunctival region and the corneal region, extracting red pixel regions in the conjunctival region and the limbal region and performing skeletonization processing to obtain a skeletonized mask image, expanding the skeletonized mask image and performing an XOR operation with the red pixel region to obtain a bleeding region; determining a vascular region in the red pixel region based on the bleeding region and measuring a vascular morphology value, determining an eye redness level based on the area ratio of the red pixel region in the conjunctival region and the limbal region, and further correcting the eye redness level through the vascular morphology value. The present application can accurately separate blood vessels and determine the eye redness level in combination with the vascular morphology, and the eye redness analysis is more comprehensive and scientific.

Description

Technical Field

[0001] The present invention relates to the technical field of ophthalmic image processing, and in particular to a method, device and medium for analyzing eye redness level combined with vascular morphology analysis. Background Art

[0002] Traditional red eye analysis relies on naked eye observation by doctors, who use their experience to determine the severity. However, human eye observation is often highly subjective and arbitrary, and the accuracy is not high. Traditional image processing technology uses threshold segmentation to analyze the red area ratio to determine the red eye level. However, due to the influence of lighting, equipment color difference and photographer's technology, the background color of the ocular surface image is not constant. Simply analyzing the red ratio through color extraction often has a high misidentification rate, and the accuracy of red eye analysis is not high.

[0003] The Chinese invention patent document with application number CN202311743538.7 discloses a method, device and medium for determining the level of red eyes. The method obtains a red eye area image by segmenting a target eye image with ocular surface congestion; performs blood vessel segmentation on the red eye area image to determine the blood vessel ratio in the red eye area image; determines a red eye index based on the red eye area image; and obtains the red eye level of the target eye image through a red eye level classification model based on the blood vessel ratio and the red eye index. The method can combine the red eye index and the blood vessel ratio to perform red eye level grading, which can improve the accuracy of red eye level grading. However, although the method adds blood vessel segmentation to the red eye analysis, the method still only considers the area ratio of the blood vessels, and the evaluation index is still relatively single, and the accuracy of red eye level grading needs to be improved. Summary of the invention

[0004] The embodiments of the present application provide a method, device and medium for analyzing redness level of eyes combined with vascular morphology analysis, which are used to solve the technical problems of single redness analysis rating index and poor accuracy of redness grading.

[0005] The specific technical solutions provided by the embodiments of the present application are introduced below.

[0006] In a first aspect, a method for analyzing eye redness level combined with vascular morphology analysis is provided, comprising:

[0007] Inputting the target eye image into a pre-trained segmentation model for segmentation to obtain a conjunctival region and a corneal region, and determining a limbal region based on the conjunctival region and the corneal region;

[0008] Extracting red pixel areas in the conjunctival area and the limbal area and performing skeletonization processing to obtain a skeletonized mask image, dilating the skeletonized mask image, and performing an XOR operation with the red pixel area to obtain a bleeding area;

[0009] Determine a blood vessel region in the red pixel region except the bleeding region based on the bleeding region, and measure a blood vessel morphology value of the blood vessel region, wherein the blood vessel morphology value includes one or more of a blood vessel width, a blood vessel topology structure, and a blood vessel density;

[0010] The eye redness level is determined based on the area of ​​the red pixel region, and the eye redness level is further corrected by the blood vessel morphology value.

[0011] In a possible embodiment, the method for analyzing red eye grade combined with vascular morphology analysis is characterized by further comprising:

[0012] If it is determined that the blood vessel width is greater than a first preset threshold, or the blood vessel topology is a reticular blood vessel and the blood vessel density is greater than a second preset threshold, the eye redness level is increased.

[0013] In a possible embodiment, the step of measuring the width of the blood vessel specifically includes:

[0014] The vessel tree mask is refined to obtain the centerline of each vessel;

[0015] respectively calculating the distance from a point on the centerline of the blood vessel to the nearest background and determining the distance as the radius of the blood vessel;

[0016] Multiply the vessel radius by two to get the vessel width.

[0017] In a possible embodiment, measuring the blood vessel topology specifically includes:

[0018] Detecting key points in the blood vessel region, the key points including endpoints having only one neighbor pixel, branch points having two neighbor pixels, and intersection points having at least three neighbor pixels;

[0019] Starting from each endpoint, the vessel centerline was traced to the next branching point or intersection, and the path length and direction change were recorded;

[0020] The connectivity of the blood vessels, the number of branch points, the length of the blood vessels and the angle of the blood vessel branches were calculated to obtain the vascular topology.

[0021] In a possible embodiment, measuring blood vessel density specifically includes:

[0022] Calculate the total number of all non-zero pixels on the blood vessel tree mask to obtain a first value;

[0023] The first value is divided by the total area of ​​the vascular tree mask to obtain the vascular density.

[0024] In a possible embodiment, measuring blood vessel density specifically includes:

[0025] Refine the vessel tree mask;

[0026] Traverse the refined image, calculate the total number of all non-zero pixels, and obtain a second value;

[0027] The second value is divided by the total area of ​​the vascular tree mask to obtain the vascular density.

[0028] In a possible embodiment, after extracting the red pixel regions in the conjunctival region and the corneal limbus region and before skeletonizing the red pixel regions, the method further includes: completing the red pixel regions using a region growing algorithm.

[0029] In a second aspect, a device for analyzing redness level of eyes combined with vascular morphology analysis is provided, comprising:

[0030] A memory for storing program instructions;

[0031] A processor is used to call the program instructions stored in the memory to implement the eye redness grade analysis method combined with vascular morphology analysis as described in any technical solution in the first aspect.

[0032] In a third aspect, a computer-readable storage medium is provided, wherein the computer-readable storage medium stores a program code, and the program code is used to implement the method for analyzing redness level of eyes combined with vascular morphology analysis as described in any technical solution in the first aspect.

[0033] The beneficial effects of the present application include at least: the present application provides a method for analyzing the redness level combined with the vascular morphology analysis, which accurately separates the blood vessels and measures the vascular morphology by skeletonizing, dilating, and performing XOR operations on the red pixel area (redness area), and determines the redness level by combining the vascular morphology and the redness area ratio. In addition to analyzing the redness area ratio and the vascular area ratio, the method also fully considers the vascular morphology characteristics such as vascular width, vascular topology, and vascular density, making the redness analysis more comprehensive and scientific. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] Figure 1 A flow chart of a method for analyzing redness level of eyes combined with vascular morphology analysis provided in an embodiment of the present application.

[0035] Figure 2 This is a flowchart for implementing step S101. DETAILED DESCRIPTION

[0036] In order to more clearly understand the purpose, features and advantages of the invention of the present application, the present application is further described in detail below in conjunction with the accompanying drawings and specific embodiments. In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present application can also be implemented in other ways different from those described herein, and therefore, the present application is not limited to the limitations of the specific embodiments disclosed below.

[0037] The embodiment of the present application is applied to image analysis of red eye images. For a more detailed and comprehensive analysis of red eye diseases, the image of eye congestion can be segmented and the vascular area in the red eye can be separated before vascular morphology measurement is performed. The red eye grade can be determined in combination with the vascular morphology value. The red eye evaluation index is more detailed and comprehensive.

[0038] The embodiments of the present application are further described in detail below in conjunction with the drawings in the specification and through specific implementation methods. It should be understood that the embodiments described herein are only used to illustrate and explain the present application, and are not used to limit the present application, and in the absence of conflict, the embodiments of the present application and the features in the embodiments can be combined with each other.

[0039] Before introducing a method for analyzing the level of redness in eyes combined with vascular morphology analysis provided by an embodiment of the present application, in order to facilitate understanding, the technical background of the embodiment of the present application is first introduced in detail. In addition to blood vessels (more commonly a single thick blood vessel, or multiple fine blood vessels crisscrossing into a network), congestion in the eye also has uniform redness and (diffuse) bleeding with unclear blood vessels. This type of bleeding often indicates that the inflammation is more severe or lasts for a long time in clinical practice, and may even be accompanied by tissue infiltration, edema, proliferation or degeneration. In the prior art, the analysis of redness in the eye is often to convert the image into an HSV color mode or a LAB color mode, and identify the red pixels in the conjunctival area by reading the H channel in HSV, or the A and B channels in LAB. However, the ocular surface image is greatly affected by light, equipment color difference, photographer technology, etc., and it is difficult to achieve uniform brightness and fixed color. The use of traditional solutions has a high misrecognition rate (for example, the background is detected as a blood vessel, or the blood vessel cannot be detected). Some existing methods for red eye analysis have added blood vessel segmentation. For example, the Chinese invention patent "A method, device and medium for determining the level of red eyes" with application number CN202311743538.7 discloses: A method for determining the level of red eyes, including: segmenting a target eye image with ocular surface congestion to obtain a red eye area image; performing blood vessel segmentation on the red eye area image to determine the blood vessel ratio in the red eye area image; determining a red eye index based on the red eye area image; and obtaining the red eye level of the target eye image based on the blood vessel ratio and the red eye index through a red eye level classification model, wherein the red eye level classification model is obtained by training a training sample set including blood vessel ratio features, red eye index features and corresponding red eye level labels. However, this method still only considers the area ratio of blood vessels, and the evaluation index is still relatively simple, which needs to be improved.

[0040] like Figure 1 As shown, a method for analyzing eye redness level combined with vascular morphology analysis provided by an embodiment of the present application is provided. The method comprises the following steps:

[0041] Step S101, inputting the target eye image into a pre-trained segmentation model for segmentation to obtain a conjunctival region and a corneal region, and determining a limbal region based on the conjunctival region and the corneal region;

[0042] In some specific implementations of this step, a photograph of an eye with ocular surface congestion is taken to obtain an ocular surface congestion target eye image to be detected, and the target eye image contains a red eye area caused by vascular dilation and / or diffuse bleeding. Figure 2 Step S101 may specifically include the following steps:

[0043] Step S1011, obtaining a target eye image and performing normalization preprocessing such as brightness and saturation;

[0044] Step S1012, input the pre-processed image data into a pre-trained neural network segmentation model, the model may use GAN, U-NET, YOLO, etc., and output the segmented conjunctival area and corneal area;

[0045] Step S1013, obtaining the limbus region by performing image dilation and image XOR operation on the conjunctiva region and the cornea region.

[0046] Before step S1013 , step S101 may further include obtaining an accurate conjunctival region by edge fitting to improve the accuracy of segmentation.

[0047] Step S102, extracting the red pixel areas in the conjunctival area and the corneal limbus area and performing skeletonization processing to obtain a skeletonized mask image, dilating the skeletonized mask image, and performing an XOR operation with the red pixel area to obtain a bleeding area;

[0048] In some specific implementations of this step, the method of extracting the red pixel area in the conjunctival area and the limbal area may include:

[0049] The first method is to convert the image into HSV color mode or LAB color mode, and identify the red pixels in the conjunctival area by reading the H channel in HSV or the A and B channels in LAB to obtain the red pixel area.

[0050] The second method is to convert the image into grayscale color mode, obtain the blood vessel area through Sobel / Scharr edge detection, and obtain the red pixel area through double threshold connection.

[0051] The third method is to convert the image into a grayscale color mode, obtain the blood vessel area through a morphology-based blood vessel extraction method, and obtain the red pixel area.

[0052] The fourth method is a blood vessel extraction method based on an active contour model ("snake line" model) to obtain the blood vessel area and obtain the red pixel area.

[0053] In some specific embodiments of the present application, before extracting the red pixel region, the method may further include: enhancing the blood vessel region by a Frangi filter algorithm based on the Hessian matrix, so that the extraction of the red pixel region becomes easier, more comprehensive and more accurate.

[0054] Step S103, determining a blood vessel region in the red pixel region except the bleeding region based on the bleeding region, and measuring a blood vessel morphology value of the blood vessel region, wherein the blood vessel morphology value includes one or more of a blood vessel width, a blood vessel topology structure, and a blood vessel density;

[0055] Through step S102 and step S103, the blood vessel area in the red pixel area can be separated.

[0056] In some specific implementations of step S103, measuring the blood vessel width of the blood vessel region may include:

[0057] The vessel width is calculated according to the vessel edge coordinates. A specific implementation of determining the vessel edge coordinates may include: marking each vessel in a clockwise direction as the first vessel, the second vessel, ..., the nth vessel, and obtaining the vessel edge coordinates after marking.

[0058] In some other specific implementations of step S103, measuring the blood vessel width of the blood vessel region may include:

[0059] The blood vessel mask is thinned to obtain the blood vessel centerline, the distance from the point on the blood vessel centerline to the nearest background is calculated and the distance is determined as the blood vessel radius, and the blood vessel radius is multiplied by 2 to obtain the blood vessel width. The specific implementation is as follows:

[0060] The vessel tree mask is thinned to obtain the centerline of each vessel. This can be achieved by morphological operations such as erosion until all vessel lines become as thin as possible.

[0061] A distance transform is applied to the points on the centerline of the vessel mask to obtain the distance of each point to the nearest background point.

[0062] For each point on the refined centerline, the minimum distance value can be found around it, that is, the radius of the blood vessel at that position.

[0063] By calculating twice the distance values ​​at these points, the vessel width can be obtained.

[0064] Measuring blood vessel width can help determine angiogenesis and inflammatory responses. For patients undergoing treatment, tracking changes in blood vessel width can be used to assess treatment efficacy or disease progression.

[0065] In some specific implementations of step S103, measuring the vascular topology may specifically include:

[0066] Detecting key points in the blood vessel region, the key points including endpoints having only one neighbor pixel, branch points having two neighbor pixels, and intersection points having at least three neighbor pixels;

[0067] Starting from each endpoint, trace along the centerline of the blood vessel to the next branch point or intersection, record the path length and direction change; calculate the connectivity of the blood vessels, the number of branch points, the length of the blood vessels, and the angle of the blood vessel branches to obtain the blood vessel topology. The specific implementation method of measuring the blood vessel topology is as follows:

[0068] First: detect key points, including (endpoints: have only one neighbor pixel. Branch points: have two neighbor pixels. Intersection points: have three or more neighbor pixels.)

[0069] Second: Starting from each endpoint, trace along the centerline of the vessel to the next branch point or intersection, recording information such as path length and direction change.

[0070] Third: Calculate the connectivity of blood vessels: Determine whether the vascular network is fully connected.

[0071] Fourth: Branching pattern: Analyze the number and distribution of branch points.

[0072] Fifth: Vessel length: Calculate the length of each vessel segment.

[0073] Sixth: Angle analysis: calculate the angle of blood vessel branches to understand the curvature of the blood vessels.

[0074] Understanding the vascular topology under normal and pathological conditions helps reveal the mechanisms of disease development. Certain ocular surface diseases can lead to abnormal changes in the vascular network, such as vascular dilation in dry eye and neovascularization in diabetic retinopathy. Analyzing the topology of blood vessels can assist in the diagnosis of these diseases.

[0075] In some specific implementations of step S103, measuring the blood vessel density may specifically include:

[0076] Calculating the area ratio of blood vessel pixels in the conjunctival region and the limbal region to obtain a first density;

[0077] The second density is obtained by calculating the blood vessel length per unit area in the conjunctival region and the limbal region,

[0078] A blood vessel density is obtained based on the first density and / or the second density.

[0079] Specifically, the vascular density is divided into two types: vascular pixel density and vascular length density. Among them, the vascular pixel density is defined as the ratio of the number of vascular pixels to the total area. The calculation steps include:

[0080] First: Calculate the total number of all non-zero pixels on the vessel tree mask.

[0081] Second: Divide this value by the total area of ​​the mask.

[0082] The blood vessel length density is defined as the blood vessel length per unit area, and the calculation steps include:

[0083] First: refine the vessel tree mask.

[0084] Second: traverse the refined image and calculate the total number of all non-zero pixels.

[0085] Third: Divide this value by the total area of ​​the mask.

[0086] Changes in ocular surface vascular density can be used as a symptom indicator for a variety of diseases, such as diabetic retinopathy, hypertensive retinopathy, and abnormal vascular distribution in dry eye. These diseases can cause abnormal proliferation or atrophy of blood vessels, thereby changing vascular density. For patients who have been diagnosed, regular monitoring of changes in vascular density can help evaluate treatment efficacy and disease progression.

[0087] Step S104, determining the eye redness level based on the area ratio of the red pixel area in the conjunctival area and the limbal area, and further correcting the eye redness level by the vascular morphology value.

[0088] In a possible embodiment, the method for analyzing eye redness level in combination with blood vessel morphology analysis further includes:

[0089] If it is determined that the blood vessel width is greater than a first preset threshold, the blood vessel topology is a reticular blood vessel, and the blood vessel density is greater than a second preset threshold, the eye redness level is increased.

[0090] In a possible embodiment, the method for analyzing eye redness level in combination with blood vessel morphology analysis further includes: if it is determined that the area of ​​the bleeding region is greater than a third preset threshold, then increasing the eye redness level.

[0091] In a possible embodiment, the method for analyzing red eye level combined with vascular morphology analysis further includes: determining that the brightness information of the red pixel area is greater than a fourth preset threshold, then increasing the red eye level; or determining that the saturation information of the red pixel area is greater than a fifth preset threshold, then increasing the red eye level.

[0092] For ease of understanding, the following is an example of the process of determining the eye redness level. Traditional eye redness levels are generally divided into five levels from 0 to 4. For example, by calculating the area ratio of the red pixel area in the conjunctival region, the eye redness level of conjunctival congestion is determined to be level 1; by calculating the area ratio of the red pixel area in the limbus region, the eye redness level of ciliary congestion is determined to be level 2.

[0093] If it is determined that the blood vessel width is greater than or equal to the preset blood vessel width threshold, the redness level is increased (for example, 0.5 levels are added to the redness level determined according to the area ratio. In this example, the redness level is determined to be (1+2) / 2+0.5=2 levels. This application does not limit the calculation scheme for increasing the redness level and will not give examples one by one).

[0094] In another embodiment of the present application, after extracting the red pixel areas in the conjunctival area and the corneal limbus area and before skeletonizing the red pixel areas, it further includes: using a region growing algorithm to complete the red pixel areas.

[0095] In some specific implementations of the embodiments of the present application, if there are multiple different regions in the image that need to be segmented, the above process can be repeated to select a new seed point for each region of interest and perform region growing.

[0096] The method for analyzing red eyes level combined with vascular morphology analysis provided in the embodiment of the present application can accurately separate the blood vessels in the red eyes, measure the vascular morphology, and determine the red eyes level based on the vascular morphology value, making the red eyes analysis more comprehensive and scientific.

[0097] In order to remove interference and improve the accuracy of red eye analysis, in a possible embodiment, before extracting the red pixel area in the conjunctival area and the corneal limbus area, it also includes: removing the eyelash occlusion area and the pigmentation area in the conjunctival area and the corneal limbus area by screening the brightness value and / or saturation value.

[0098] In a possible embodiment, the eye redness level analysis method combined with vascular morphology analysis further includes: acquiring a target eye image with eye congestion and performing preprocessing, wherein the preprocessing includes brightness normalization and / or saturation normalization.

[0099] In a possible embodiment, the eye redness level analysis method combined with vascular morphology analysis further includes: constructing and training a segmentation model for segmenting the conjunctival region and the corneal region, wherein the segmentation model adopts a GAN, U-NET or YOLO network structure.

[0100] Another embodiment of the present application provides a device for analyzing redness of eyes combined with vascular morphology analysis, comprising:

[0101] A memory for storing program instructions;

[0102] A processor is used to call the program instructions stored in the memory to implement the eye redness grade analysis method combined with vascular morphology analysis as described in any of the above embodiments.

[0103] In another embodiment of the present application, a computer-readable storage medium is provided, wherein the computer-readable storage medium stores a program code, and the program code is used to implement the eye redness grade analysis method combined with vascular morphology analysis as described in any of the above embodiments.

[0104] All or part of the steps in the various methods of the above embodiments can be completed by controlling the relevant hardware through a program, and the program can be stored in a readable storage medium, and the storage medium includes a read-only memory (ROM), a random access memory (RAM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), a one-time programmable read-only memory (OTPROM), an electronically erasable programmable read-only memory (EEPROM), a compact disc (CD-ROM) or other optical disc storage, magnetic disk storage, magnetic tape storage, or any other readable medium that can be used to carry or store data.

[0105] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for analyzing red eyes level combined with vascular morphology analysis, characterized in that: include: Inputting the target eye image into a pre-trained segmentation model for segmentation to obtain a conjunctival region and a corneal region, and determining a limbal region based on the conjunctival region and the corneal region; Extracting red pixel areas in the conjunctival area and the limbal area and performing skeletonization processing to obtain a skeletonized mask image, dilating the skeletonized mask image, and performing an XOR operation with the red pixel area to obtain a bleeding area; Determine a blood vessel region in the red pixel region except the bleeding region based on the bleeding region, and measure a blood vessel morphology value of the blood vessel region, wherein the blood vessel morphology value includes one or more of a blood vessel width, a blood vessel topology structure, and a blood vessel density; The eye redness level is determined based on the area ratio of the red pixel area in the conjunctival area and the corneal limbus area, and the eye redness level is further corrected by the vascular morphology value.

2. The method for analyzing red eyes level combined with vascular morphology analysis according to claim 1, characterized in that: Also includes: If it is determined that the blood vessel width is greater than a first preset threshold, or the blood vessel topology is a reticular blood vessel and the blood vessel density is greater than a second preset threshold, the eye redness level is increased.

3. The method for analyzing red eyes level combined with vascular morphology analysis according to claim 1, characterized in that: The steps for measuring blood vessel width include: The vessel tree mask is refined to obtain the centerline of each vessel; respectively calculating the distance from a point on the centerline of the blood vessel to the nearest background and determining the distance as the radius of the blood vessel; Multiply the vessel radius by two to get the vessel width.

4. The method for analyzing red eyes level combined with vascular morphology analysis according to claim 1, characterized in that: Measuring vascular topology specifically includes: Detecting key points in the blood vessel region, the key points including endpoints having only one neighbor pixel, branch points having two neighbor pixels, and intersections having at least three neighbor pixels; Starting from each endpoint, the vessel centerline was traced to the next branching point or intersection, and the path length and direction change were recorded; The connectivity of the blood vessels, the number of branch points, the length of the blood vessels and the angle of the blood vessel branches were calculated to obtain the vascular topology.

5. The method for analyzing red eyes level combined with vascular morphology analysis according to claim 1, characterized in that: Measurement of vascular density specifically includes: Calculate the total number of all non-zero pixels on the blood vessel tree mask to obtain a first value; The first value is divided by the total area of ​​the vascular tree mask to obtain the vascular density.

6. The method for analyzing red eyes level combined with vascular morphology analysis according to claim 1, characterized in that: Measurement of vascular density specifically includes: Refine the vessel tree mask; Traverse the refined image, calculate the total number of all non-zero pixels, and obtain a second value; The second value is divided by the total area of ​​the vascular tree mask to obtain the vascular density.

7. The method for analyzing red eyes grade combined with vascular morphology analysis according to claim 1, characterized in that: After extracting the red pixel regions in the conjunctival region and the limbal region, and before skeletonizing the red pixel regions, the method further includes: using a region growing algorithm to complete the red pixel regions.

8. A device for analyzing redness level of eyes combined with blood vessel morphology analysis, characterized in that: include: A memory for storing program instructions; A processor, configured to call the program instructions stored in the memory to implement the eye redness grade analysis method combined with vascular morphology analysis as described in any one of claims 1 to 7.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a program code, and the program code is used to implement the eye redness grade analysis method combined with vascular morphology analysis as described in any one of claims 1 to 7.

Images