Fungus focus sketching method for fungal keratitis under microscope

Abstract

The invention relates to the technical field of image recognition, in particular to a fungal focus sketching method for fungal keratitis under a microscope, which comprises the following steps: acquiring microscopic images of epithelium, matrix layer and endodermis of different depths of a cornea sample; analyzing the influence range of fungal spores according to the gray level distribution characteristics of the microscopic image of the epithelium, and obtaining an undetermined focus connected domain of the epithelium; analyzing the interweaving condition of the highlight part according to the edge distribution characteristics of the microscopic image of the matrix layer to obtain an undetermined focus connected domain of the matrix layer; analyzing regularity and permeability of cells according to transparency characteristics of the microscopic image of the endodermis to obtain a to-be-determined focus connected domain of the endodermis; fusing the to-be-determined focus connected regions of the epithelial layer, the matrix layer and the endothelial layer, and sketching a fungal focus region corresponding to the cornea sample. The influence of nerve fibers on fungal lesion recognition can be avoided, and the accuracy of a fungal lesion sketching result is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of image recognition, and particularly to a method for delineating fungal lesions of fungal keratitis under a microscope. Background Art

[0002] Fungal keratitis is an eye disease caused by fungal infection, which usually leads to corneal inflammation and may seriously affect vision or even cause vision loss when severe. The typical features of this disease include damage to the corneal surface (such as small ulcers or erosions), corneal edema, and corneal opacity. These lesions will hinder the normal transmission of light and thus affect visual function. To help medical students better understand the pathological features of fungal keratitis, in teaching, a corneal swept-source optical coherence tomography (SL-OCT) is usually used to collect optical anatomical images to obtain clear corneal lesion images and assist medical students in intuitively learning the pathological manifestations of fungal infections.

[0003] However, in corneal lesion images, fungal hyphae are mainly distributed in the corneal stromal layer, and their morphology and orientation are highly similar to those of corneal nerve fibers. An example diagram of nerve fibers is as shown in Figure 1 shown, and an example diagram of fungal hyphae is as shown in Figure 2 shown. This similarity may cause confusion for medical students when observing and identifying fungal lesions. Especially when accurately delineating and defining fungal lesions, it is easy to be interfered by corneal nerve fibers. This interference will not only affect the accurate identification and range assessment of fungal lesions by medical students, but also may lead to errors in lesion identification during the teaching process, thereby affecting the ability cultivation of medical students. Summary of the Invention

[0004] In order to solve the technical problem that in the above corneal lesion images, the morphology and orientation of fungal hyphae are highly similar to those of corneal nerve fibers, resulting in confusion for medical students when observing and identifying fungal lesions, the purpose of the present invention is to provide a method for delineating fungal lesions of fungal keratitis under a microscope, and the specific technical solution adopted is as follows: An embodiment of the present invention provides a method for delineating fungal lesions of fungal keratitis under a microscope, and the method includes the following steps: Obtain microscopic images of the epithelium, stroma, and endothelium at different depths of a corneal sample; Analyze the influence range of fungal spores according to the gray-scale distribution characteristics of the microscopic image of the epithelium to obtain a pending lesion connected domain of the epithelium; Analyze the interlacing situation of the highlight part according to the edge distribution characteristics of the microscopic image of the stroma to obtain a pending lesion connected domain of the stroma; Analyze the regularity and permeability of cells based on the transparency characteristics of the microscopic images of the endothelium, and obtain the to-be-determined lesion connected regions of the endothelium. Fuse the to-be-determined lesion connected regions of the epithelium, stroma, and endothelium, and delineate the fungal lesion regions corresponding to the corneal samples.

[0005] Further, the analyzing the influence range of fungal spores according to the gray-scale distribution characteristics of the microscopic images of the epithelium to obtain the to-be-determined lesion connected regions of the epithelium includes: Perform superpixel segmentation on the microscopic images of the epithelium to obtain each superpixel cluster; perform fitting analysis on each superpixel block in the superpixel cluster to obtain the fitting evaluation value of each superpixel cluster; Classify each superpixel cluster according to the gray-scale value situation of each superpixel cluster; denote the superpixel cluster with the highest gray-scale as the spore cluster; Obtain the to-be-determined lesion connected regions of the epithelium according to the fitting evaluation values of each superpixel cluster and each spore cluster.

[0006] Further, the obtaining the to-be-determined lesion connected regions of the epithelium according to the fitting evaluation values of each superpixel cluster and each spore cluster includes: Obtain the surrounding regions of all the spore clusters in the microscopic images of the epithelium, and continuously adjust the area size of the surrounding regions through an optimization algorithm until the area of the surrounding region is the largest and the cumulative value of the fitting evaluation values of all the superpixel clusters within the surrounding region is the smallest, and determine the final surrounding region as the to-be-determined lesion connected regions of the epithelium; wherein, the surrounding region is the region containing all the spore clusters.

[0007] Further, the analyzing the interlacing situation of the highlight parts according to the edge distribution characteristics of the microscopic images of the stroma to obtain the to-be-determined lesion connected regions of the stroma includes: Perform edge detection on the microscopic images of the stroma to obtain each edge pixel point, and perform threshold segmentation on the microscopic images of the stroma to obtain each high-threshold pixel point; Select the edge pixel points from all the high-threshold pixel points as the feature pixel points; Determine the first significant degree of the hyphal characteristics of each feature line according to each feature pixel point; wherein, the first significant degree of the hyphal characteristics is used to characterize the straightness of the feature line; Determine the second significant degree of the hyphal characteristics of each feature line according to the mutual interlacing characteristics between each feature line; Combine the first significant degree of the hyphal characteristics and the second significant degree of the hyphal characteristics of each feature line, and screen out the to-be-determined hyphal segments; Take the smallest connected region containing all the to-be-determined hypha segments in the microscopic image of the substrate layer as the to-be-determined lesion connected region of the substrate layer.

[0008] Further, the first significant degree of the hypha characteristics of each feature line determined according to each feature pixel point includes: Connect adjacent feature pixel points within the eight-neighborhood range to form feature lines, obtaining a number of feature lines; For any feature line, obtain the connection line length between the two end points of the feature line, and compare the feature line length with the connection line length to obtain the deviation degree; Draw a perpendicular line from each feature pixel point on the feature line to the connection line length segment, the direction of the perpendicular line is perpendicular to the connection line length segment, and determine the sharpness according to the difference in the number of perpendicular lines on both sides of the connection line length segment; Obtain the edge chain code of the feature pixel points on the feature line, and obtain the tortuosity according to the difference between adjacent code numbers in the edge chain code; According to the deviation degree, the sharpness and the tortuosity of the feature line, obtain the first significant degree of the hypha characteristics of the feature line; Wherein, the deviation degree, the sharpness and the tortuosity are all negatively correlated with the first significant degree of the hypha characteristics.

[0009] Further, the second significant degree of the hypha characteristics of each feature line determined according to the mutual intersection characteristics between each feature line includes: For any feature line, record the feature line closest to the end point of the feature line as the first reference line, and record the feature line second closest to the end point of the feature line as the second reference line; Extend the feature line in the direction of the to-be-determined reference line, and then connect it to the to-be-determined reference line to obtain a new feature line; wherein, the to-be-determined reference line is the first reference line or the second reference line; Obtain each inflection point on the to-be-determined reference line, and then obtain the included angle formed by the connection between the two end points of the to-be-determined reference line and any inflection point, and take the smallest included angle as the target included angle corresponding to the to-be-determined reference line; Determine the second significant degree of the hypha characteristics of the feature line according to the uniform similarity degree of the gray level distribution between the feature line and the new feature line, and the target included angles corresponding to the first reference line and the second reference line of the feature line.

[0010] Further, the second significant degree of the hypha characteristics of the feature line determined according to the uniform similarity degree of the gray level distribution between the feature line and the new feature line, and the target included angles corresponding to the first reference line and the second reference line of the feature line includes: Obtain the first gray variance of the feature line, the second gray variance of the new feature line when the to-be-determined reference line is the first reference line, and the third gray variance of the new feature line when the to-be-determined reference line is the second reference line; Determine the first similarity according to the difference between the first gray variance and the second gray variance, and determine the second similarity according to the difference between the first gray variance and the third gray variance; Determine the second significant degree of the hyphal characteristics of the feature line through the first similarity, the second similarity, and the target included angle corresponding to the first reference line and the second reference line.

[0011] Further, the determining the second significant degree of the hyphal characteristics of the feature line through the first similarity, the second similarity, and the target included angle corresponding to the first reference line and the second reference line includes: Calculate the ratio of the second similarity to the average value of the target included angles corresponding to the second reference line, and calculate the ratio of the first similarity to the average value of the target included angles corresponding to the first reference line. After calculating the difference between the two ratios, perform normalization processing to determine the second significant degree of the hyphal characteristics of the feature line.

[0012] Further, the analyzing the regularity and permeability of cells according to the transparency characteristics of the microscopic image of the endodermis to obtain the to-be-determined lesion connected region of the endodermis includes: Obtain the high-threshold region of the microscopic image of the endodermis, perform connected region detection on the high-threshold region to obtain each cell connected region; perform hexagonal fitting on each cell connected region to obtain a loss value and a fitting side length; Determine the deformation coefficient of each cell connected region according to the loss value of the cell connected region and the difference between the fitting side length and the average side length; wherein, the average side length is the average value of the fitting side lengths of all cell connected regions; Perform image segmentation on the microscopic image of the endodermis according to the distribution of pixel points outside the cell connected regions in the microscopic image of the endodermis and the deformation coefficient to obtain the to-be-determined lesion connected region of the endodermis.

[0013] Further, the performing image segmentation on the microscopic image of the endodermis according to the distribution of pixel points outside the cell connected regions in the microscopic image of the endodermis and the deformation coefficient to obtain the to-be-determined lesion connected region of the endodermis includes: Take the pixel points outside the cell connected regions in the microscopic image of the endodermis as the to-be-detected pixel points, and determine the minimum distance value between the to-be-detected pixel points and all cell connected regions in the microscopic image of the endodermis; Determine a window area centered on the pixel to be detected, and determine each maximum point within the window area; wherein, the maximum point is the pixel with the largest gray value within its eight-neighborhood range. Calculate the information entropy of the distance between each maximum point within the window area and the pixel to be detected, and obtain the fungal infection characteristic value of each pixel to be detected according to the minimum distance value, the information entropy, and the deformation coefficient. Perform threshold segmentation on the fungal infection characteristic values of each pixel to be detected in the microscopic image of the endothelium to obtain high-threshold pixels. Merge the high-threshold pixels, the cell connectivity regions adjacent to the high-threshold pixels, and the cell connectivity regions surrounded by the high-threshold pixels to obtain the pending lesion connectivity region of the endothelium.

[0014] The present invention has the following beneficial effects: Compared with the existing method of outlining fungal lesions by ignoring the influence of corneal nerve fibers, the present invention provides a method for outlining fungal lesions of fungal keratitis under a microscope. By analyzing the distinguishing features between fungal hyphae and nerve fibers in fiber images of different depth layers, the pending lesion connectivity regions of the epithelium, stroma, and endothelium are determined, and then the pending lesion connectivity regions of the three layers are fused to outline the fungal lesion area corresponding to the corneal sample, which can achieve accurate identification and range assessment of fungal lesions, and provide more reliable analysis materials for the teaching of fungal keratitis. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] In order to more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required to be used in the description of the embodiments or the prior art. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0016] Figure 1 It is an example diagram of nerve fibers; Figure 2 It is an example diagram of fungal hyphae; Figure 3 It is a flowchart of the steps of a method for outlining fungal lesions of fungal keratitis under a microscope according to an embodiment of the present invention; Figure 4 It is an example diagram of the microscopic image of the epithelium; Figure 5 It is an example diagram of the microscopic image of the epithelium after filtering and morphological processing in an embodiment of the present invention; Figure 6 It is a flowchart of the steps of step S 2 in an embodiment of the present invention; Figure 7An example diagram of the result after superpixel segmentation of the microscopic image of the epithelial layer in the embodiment of the present invention; Figure 8 A step flowchart of step S3 in the embodiment of the present invention; Figure 9 A step flowchart of step S4 in the embodiment of the present invention; Figure 10 An example diagram of the fungal lesion area corresponding to the corneal sample in the embodiment of the present invention. Detailed implementation manners

[0017] In order to further elaborate on the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the following specifically describes in detail the specific implementation manners, structures, features and effects of the technical solutions proposed according to the present invention in combination with the accompanying drawings and preferred embodiments. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. In addition, the specific features, structures or characteristics in one or more embodiments can be combined in any suitable form.

[0018] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the technical field to which the present invention belongs.

[0019] The application scenario targeted by the present invention can be: Under the microscope, fungal hyphae are mainly distributed in the corneal stroma layer, and their morphology and orientation are highly similar to those of corneal nerve fibers. Analyzing according to the morphological characteristics of hyphae by traditional methods may misjudge some nerve fibers as fungal hyphae, resulting in interference in accurately delineating and defining fungal lesions, thereby affecting the accurate identification and scope assessment of fungal lesions and bringing challenges to medical students to accurately identify fungal lesions.

[0020] An embodiment of the present invention provides a method for delineating fungal lesions of fungal keratitis under a microscope, as Figure 3 shown, including the following steps: S1, obtaining microscopic images of the epithelial layer, stroma layer and endothelium layer at different depths of the corneal sample.

[0021] In order to delineate the fungal lesions of fungal keratitis, it is first necessary to collect microscopic images of the corneal sample. The image property of the microscopic image here is a grayscale image, and each pixel point in the grayscale image has its corresponding grayscale value.

[0022] Specifically, the eye is correctly positioned under the microscope. Special brackets or devices are usually used during positioning to ensure the stability of the eye. To avoid the influence caused by eye discomfort or involuntary blinking, local anesthetic eye drops are usually instilled into the eye to keep the eye calm and reduce interference. The microscope scans different depths of the cornea layer by layer using optical sectioning technology. Each layer of corneal image is obtained by laser exciting different layers of the cornea and detecting the light signals reflected or scattered from the corresponding layers. The light signals after laser irradiation are captured by a detector, which receives the reflected signals and converts them into digital images. Image processing is performed on the digital images to obtain microscopic images of the epithelium, stroma, and endothelium. Among them, image processing includes but is not limited to filtering, morphological processing, and grayscale processing.

[0023] The above microscope can be a confocal microscope. The confocal microscope uses laser as the light source. The laser used is usually a low-power single-wavelength laser (for example, 488nm or 785nm). The laser irradiates the surface of the cornea through the optical system of the microscope. It should be noted that the laser beam is only concentrated at the focal point and does not spread to the surrounding area.

[0024] It should be noted that according to the intensity and nature of the reflected light, the confocal microscope can obtain information about different layers of the cornea. Since each layer is scanned independently, the acquired images are continuous section images, reflecting the details and hierarchical structure of the cornea. Therefore, microscopic images of the epithelium, stroma, and endothelium at different depths of the corneal sample are obtained.

[0025] So far, microscopic images of the epithelium, stroma, and endothelium of the corneal sample have been obtained in this embodiment.

[0026] It should be noted that when fungal keratitis occurs in the corneal sample, fungal infections usually involve multiple layers of the cornea, especially the epithelium, stroma, and endothelium. Therefore, obtaining microscopic images of different depth layers of the corneal sample as basic image data to analyze the image characteristics of fungal hyphae in different layers to obtain the pending lesion connected regions of each layer, and finally performing superposition analysis on the pending lesion connected regions identified in the three layers. Compared with only identifying the fungal lesion connected region of one layer, this embodiment can outline a more accurate fungal lesion connected region for medical students to conduct teaching analysis.

[0027] S2. Analyze the influence range of fungal spores according to the gray distribution characteristics of the microscopic image of the epithelium to obtain the pending lesion connected region of the epithelium.

[0028] The fiber image of the epithelial layer contains epithelial cells. The epithelial layer with fungal infection often shows rough, discontinuous, and irregular morphology. Especially in the lesion area, a large range of epithelial defects or ulcers may be observed. Therefore, when fungal infection occurs, some epithelial cells in some areas will fall off, resulting in some local fungal spores and irregular changes at the cell edges in the epithelial layer. And the fungal spores will appear in a highlighted form between the cells. Therefore, by analyzing the influence range of the fungal spores, the undetermined lesion connected regions of the epithelial layer can be obtained. Among them, an example diagram of the microscopic image of the epithelial layer is as Figure 4 shown, and an example diagram of the microscopic image of the epithelial layer after filtering and morphological processing is as Figure 5 shown.

[0029] The above step S2 can be implemented through Figure 6 the steps S21 to S23 shown as follows: S21, perform superpixel segmentation on the microscopic image of the epithelial layer to obtain each superpixel cluster; perform fitting analysis on each superpixel block in the superpixel cluster to obtain the fitting evaluation value of each superpixel cluster.

[0030] Here, the fitting evaluation value can characterize the regularity of the superpixel distribution in the superpixel cluster. The larger the fitting evaluation value, the more regular the superpixel distribution in the cluster. By segmenting the microscopic image of the epithelial layer into regions with similar color characteristics, the image structure can be simplified and the efficiency of subsequent image processing can be improved. In addition, step S2 is to obtain the undetermined lesion connected regions of the epithelial layer. Superpixel clustering can usually better retain the boundary information of the image, which is helpful for connected region analysis; the fungal infection of the epithelial layer usually shows a large range of defects or ulcers, which results in poor regularity of the image pixel distribution. Therefore, when determining the undetermined lesion connected regions of the epithelial layer, the regularity of the superpixel distribution of its lesion connected regions should be poor, that is, small.

[0031] In this embodiment, the implementation process of superpixel segmentation is a prior art and will not be elaborated in detail here; the chi-square test is used for fitting the superpixel cluster. The fitting effect is evaluated by calculating the chi-square statistic between the color distribution of each superpixel and the expected distribution to obtain the fitting evaluation value. The level of the fitting evaluation value can be determined by the chi-square statistic. The better the fitting effect, the more regular the superpixel distribution within the superpixel cluster. The process of realizing the fitting of the superpixel cluster through the chi-square test is a prior basis and will not be described in detail here. Of course, the implementer can also adopt other fitting methods to realize, and the fitting method for the superpixel cluster is not specifically limited, such as Jensen-Shannon divergence or Bhattacharyya distance.

[0032] Among them, an example diagram of the result of performing superpixel segmentation on the microscopic image of the epithelial layer is as Figure 7 shown, Figure 7It is a labeled graph of superpixel regions, and the labeled colors do not represent the actual grayscale of the superpixels.

[0033] S22. Classify each superpixel cluster according to the grayscale value situation to obtain various superpixel clusters; Denote the superpixel cluster with the highest grayscale as the spore cluster.

[0034] Here, the spore cluster refers to the cluster with the largest cumulative sum of all superpixel grayscales. The larger the grayscale value, the more likely it is a fungal spore presented in a highlighted form. Therefore, the superpixel cluster with the highest grayscale is denoted as the spore cluster.

[0035] Using the grayscale similarity threshold, superpixel clusters with similar grayscales can be grouped together to obtain several superpixel clusters. Specifically, calculate the grayscale value differences between pairwise superpixel clusters, and group the two superpixel clusters with grayscale value differences less than the grayscale similarity threshold together. Among them, the grayscale similarity threshold and the number of categories can be set by the implementer according to experience and actual situations, without specific limitations.

[0036] S23. Obtain the undetermined lesion connected region of the epithelial layer according to the fitting evaluation values of each superpixel cluster and each spore cluster.

[0037] Specifically, obtain the surrounding region of all spore clusters in the microscopic image of the epithelial layer, and continuously adjust the area size of the surrounding region through an optimization algorithm until the area of the surrounding region is the largest and the cumulative sum of the fitting evaluation values of all superpixel clusters within the surrounding region is the smallest, then determine the final surrounding region as the undetermined lesion connected region of the epithelial layer.

[0038] Among them, the surrounding region is the region containing all spore clusters, which needs to meet the condition that the area of the surrounding region is as large as possible to cover more possible lesion areas; the cumulative sum of the fitting evaluation values of all superpixel clusters within the surrounding region is as small as possible to ensure a higher consistency within the region. The optimization algorithm can be a greedy algorithm, dynamic programming, genetic algorithm, etc.

[0039] It should be noted that for superpixel clusters where some superpixel blocks are not within the surrounding region while some are, the fitting evaluation value calculation is re - performed based on the partial superpixel blocks within the surrounding region.

[0040] So far, the undetermined lesion connected region of the epithelial layer has been obtained in this embodiment.

[0041] S3. Analyze the interlacing situation of the highlighted part according to the edge distribution characteristics of the microscopic image of the stromal layer to obtain the undetermined lesion connected region of the stromal layer.

[0042] When fungal keratitis appears in the corneal sample, obvious hyphal branches appear in the stromal layer. In more extreme cases, the fungal hyphae may penetrate the stroma into the endothelium, causing damage or dysfunction of the endothelial cells. Therefore, stromal layer recognition of fungal manifestations is required when identifying fungi.

[0043] The stromal layer is the most common layer in corneal infections. In the stromal layer, fungal infections are manifested as obvious long and slender branched fungal hyphae. The fungal hyphae show obvious curved or branched morphologies, and the reflection of the hyphae is relatively obvious. Therefore, according to the edge distribution characteristics of the microscopic image of the stromal layer, the interlacing situation of the highlight part is analyzed to obtain the connected domain of the pending lesions in the stromal layer.

[0044] The above step S3 can be implemented by Figure 8 the steps S31 to S33 shown as follows: S31, perform edge detection on the microscopic image of the stromal layer to obtain each edge pixel point, and perform threshold segmentation on the microscopic image of the stromal layer to obtain each high-threshold pixel point; select the edge pixel points from all the high-threshold pixel points as feature pixel points.

[0045] Here, the feature pixel points are pixel points containing two pixel properties, namely high-threshold pixel points and edge pixel points. Through the feature pixel points, the distribution of the edge pixel points in the highlighted area of the fibrous image of the stromal layer can be analyzed, and it can be used to represent the interlacing characteristics of the highlight part.

[0046] In this embodiment, the edge detection method is canny edge detection, and the threshold segmentation method is Otsu threshold. The implementation processes of both are prior arts and will not be described in detail here. Of course, the implementer can also adopt other edge detection and threshold segmentation methods, which are not specifically limited here.

[0047] S32, determine the first significant degree of the hyphal characteristics of each feature line according to each feature pixel point; determine the second significant degree of the hyphal characteristics of each feature line according to the mutual interlacing characteristics between each feature line.

[0048] Here, the first significant degree of the hyphal characteristics is used to represent the rapid and straight degree of the trend of the feature line. The greater the first significant degree of the hyphal characteristics, the more the feature line conforms to the branching characteristics of the fungal hyphae; the second significant degree of the hyphal characteristics is used to represent the degree of mutual interlacing of the feature lines. The greater the second significant degree of the hyphal characteristics, the more the feature lines conform to the characteristics of mutual accumulation between the fungal hyphae.

[0049] First, determining the first significant degree of the hyphal characteristics of each feature line according to each feature pixel point includes: In the first step, connect the adjacent feature pixel points within the eight-neighborhood range to form feature lines, and obtain a number of feature lines.

[0050] In this embodiment, the purpose of obtaining the feature line is to analyze the straight or curved state of the branches of the highlighted part in the microscopic image of the substrate layer. The number of feature pixel points included in the feature line shall not be less than 5. That is, for a feature line that is too short in length, this embodiment does not perform feature analysis on it; adjacent feature pixel points refer to the feature pixel points that are closely adjacent to each other one after another. There is a possibility that the feature line is the edge of a fungal hypha or a nerve fiber edge.

[0051] In the second step, for any feature line, obtain the connection line length between the two end points of the feature line, and compare the feature line length with the connection line length to obtain the deviation degree.

[0052] In this embodiment, the length refers to the number of pixel points. The deviation degree can be the difference between the feature line length and the connection line length, or the ratio of the feature line length to the connection line length; the greater the feature line length is than the connection line length, the greater the degree of deviation of the feature line from a straight line, that is, the greater the deviation degree of the feature line, and the smaller the first significant degree of the hypha feature will be.

[0053] In the third step, draw a perpendicular line from each feature pixel point on the feature line to the connection line length segment. The direction of the perpendicular line is perpendicular to the connection line length segment, and determine the sharpness according to the difference in the number of perpendicular lines on both sides of the connection line length segment.

[0054] In this embodiment, for the number of perpendicular lines on both sides of the connection line length segment, if the connection line length segment is parallel to the horizontal line, then count the number of perpendicular lines on the upper and lower sides of the connection line length segment; if the connection line length segment is perpendicular to the horizontal line, then count the number of perpendicular lines on the left and right sides of the connection line length segment. Take the absolute value of the difference in the number of perpendicular lines on both sides as the sharpness. The sharpness can characterize whether the feature line bends towards one side, and the greater the degree of bending towards one side, the greater the sharpness.

[0055] In the fourth step, obtain the edge chain code of the feature pixel points on the feature line, and obtain the tortuosity according to the difference between adjacent code numbers in the edge chain code.

[0056] In this embodiment, the process of determining the tortuosity through the edge chain code is a prior art and is not within the protection scope of the present invention, so it will not be elaborated in detail here.

[0057] In the fifth step, according to the deviation degree, sharpness and tortuosity of the feature line, obtain the first significant degree of the hypha feature of the feature line.

[0058] In this embodiment, the deviation degree, sharpness and tortuosity are all negatively correlated with the first significant degree of the hypha feature.

[0059] As an example, the calculation formula for the first significant degree of the hypha feature of the i-th feature line can be: In the formula, represents the first significant degree of the hyphal characteristics of the i-th characteristic line, represents the deviation degree of the i-th characteristic line, represents the sharpness of the i-th characteristic line, represents the tortuosity of the i-th characteristic line, and m represents a non-zero constant with an empirical value of 0.01, which is used to prevent the denominator of the fraction from being zero.

[0060] In the calculation formula of the first significant degree of hyphal characteristics, the greater the deviation degree, the greater the bending degree of the characteristic line; the greater the sharpness, the greater the bending degree of the characteristic line; the greater the tortuosity, the greater the bending degree of the characteristic line. And the greater the bending degree, the less the characteristic line conforms to the straight and sharp morphological characteristics of fungal hyphae, and the smaller the first significant degree of hyphal characteristics. The deviation degree, sharpness, and tortuosity are usually dimensionless indicators used to describe the characteristics of curves, so there is no clear unit.

[0061] It should be noted that quantifying the straight and sharp characteristics of the characteristic line from three different aspects can more comprehensively analyze the bending degree of the characteristic line, making the numerical accuracy of the first significant degree of hyphal characteristics representing the hyphal morphological characteristics higher, and further making the reliability of the undetermined lesion connectivity domain of the substrate layer obtained based on the first significant degree stronger.

[0062] Secondly, according to the interlaced characteristics between each characteristic line, determine the second significant degree of the hyphal characteristics of each characteristic line.

[0063] First of all, it should be noted that since some nerve fibers may have a situation similar to the straight and sharp characteristics of hyphae, detecting hyphae only based on the first significant degree of hyphal characteristics will be inaccurate. Therefore, in order to improve the recognition accuracy of the undetermined lesion connectivity domain of the substrate layer, it is necessary to further analyze the fungal hyphal characteristics of the characteristic line. For the fungal hyphae in the substrate layer, they show many branches, disordered arrangement, and bamboo joint-like or dendritic-like. The refractive index and morphology of hyphae of different strains are different. Therefore, due to the disordered arrangement and branching of fungal hyphae, a situation of mutual accumulation between fungal hyphae is formed, resulting in the situation of interlacing between hyphae.

[0064] Here, the second significant degree of hyphal characteristics is used to represent the degree of interlacing of the characteristic line. The more serious the interlacing degree, the more obvious the mutual accumulation characteristics between fungal hyphae are formed, and the higher the possibility that the characteristic line belongs to fungal hyphae.

[0065] In the first step, for any characteristic line, the characteristic line closest to the end point of the characteristic line is recorded as the first reference line, and the characteristic line second closest to the end point of the characteristic line is recorded as the second reference line.

[0066] It should be noted that the first reference line and the second reference line for obtaining the characteristic line are used to analyze the intersection relationship between the characteristic line and other characteristic lines close to it, so as to quantify whether the characteristic line is in an intersecting state and the existing degree of intersection, providing a data analysis basis for the subsequent quantification of the intersection coefficient.

[0067] In this embodiment, each end point of each characteristic line has its corresponding first reference line and second reference line. The distance between the end point and other characteristic lines except the characteristic line where it is located can be obtained by calculating the existing point-line distance, and no detailed description will be given here. If a certain end point corresponds to two characteristic lines with the same distance, either of the two characteristic lines can be used as the first reference line and the other as the second reference line.

[0068] In the second step, extend the characteristic line in the direction of the to-be-determined reference line, and then connect it to the to-be-determined reference line to obtain a new characteristic line. Among them, the to-be-determined reference line is the first reference line or the second reference line.

[0069] In this embodiment, the extension of the characteristic line means extending along the directions of the to-be-determined reference lines at both end points until the characteristic line is connected to the to-be-determined reference lines at both of its ends, obtaining the extended characteristic line. The extended characteristic line after completion is used as the new characteristic line, and each characteristic line can correspond to two different new characteristic lines.

[0070] In the third step, obtain each inflection point on the to-be-determined reference line, and then obtain the included angle formed by the connection between the two end points of the to-be-determined reference line and any inflection point. The minimum included angle is used as the target included angle corresponding to the to-be-determined reference line.

[0071] Specifically, perform a second derivative operation on the to-be-determined reference line to detect inflection points, and each inflection point on the to-be-determined reference line can be obtained; for any inflection point, obtain the included angle formed by the connection between the two end points of the to-be-determined reference line and this inflection point, and then obtain all the included angles corresponding to the to-be-determined reference line. The minimum included angle is used as the target included angle corresponding to the to-be-determined reference line. Among them, the implementation processes of the second derivative operation and inflection point detection are both existing technologies and not within the protection scope of the present invention, and no detailed elaboration will be given here.

[0072] When determining the target included angle, the smaller the included angle corresponding to the to-be-determined reference line, the greater the possibility of intersection between the fungal hyphae. Therefore, when determining the target included angle, the minimum included angle is used as the target included angle corresponding to the to-be-determined reference line.

[0073] Fourthly, determine the second significant degree of the hyphal characteristics of the feature line according to the uniform similarity degree of the gray-scale distribution between the feature line and the new feature line, and the target included angle corresponding to the first reference line and the second reference line of the feature line.

[0074] Here, the uniform gray-scale distribution is used to analyze the disordered situation of the arrangement of fungal hyphae, and the target included angle can analyze the forking situation of fungal hyphae. By combining the interleaved feature situations in these two aspects, the second significant degree of the hyphal characteristics of the feature line can be determined.

[0075] As an optional implementation manner, determining the second significant degree of the hyphal characteristics of the feature line includes: The first sub-step is to obtain the first gray-scale variance of the feature line, the second gray-scale variance of the new feature line when the pending reference line is the first reference line, and the third gray-scale variance of the new feature line when the pending reference line is the second reference line.

[0076] In this embodiment, through the gray-scale values of each feature pixel point on the feature line and the new feature line, the gray-scale variance can be calculated. For the convenience of distinction, the gray-scale variance corresponding to the feature line is denoted as the first gray-scale variance, the gray-scale variance corresponding to the new feature line when the pending reference line is the first reference line is denoted as the second gray-scale variance, and the gray-scale variance corresponding to the new feature line when the pending reference line is the second reference line is denoted as the third gray-scale variance.

[0077] The second sub-step is to determine the first similarity according to the difference between the first gray-scale variance and the second gray-scale variance, and determine the second similarity according to the difference between the first gray-scale variance and the third gray-scale variance.

[0078] In this embodiment, by using the exponential function exp(-) with the natural constant as the base to process the absolute value of the difference between the two gray-scale variances, the uniform similarity situation of the gray-scale distribution can be obtained. Of course, the implementer can also use other methods to quantify the similarity, which is not specifically limited here. For example, the Bhattacharyya distance can be used to analyze the distribution similarity of two groups of gray-scale sets.

[0079] The third sub-step is to determine the second significant degree of the hyphal characteristics of the feature line through the first similarity, the second similarity, and the target included angle corresponding to the first reference line and the second reference line.

[0080] In this embodiment, the greater the first similarity corresponding to the first reference line and the smaller the target included angle, while the smaller the second similarity corresponding to the second reference line and the larger the target included angle, it indicates that the possibility of intersection between the first reference line and the second reference line of the feature line is greater, and the second significant degree of the hyphal characteristics of the feature line is also greater.

[0081] Specifically, calculate the ratio of the second similarity to the average value of the target included angle corresponding to the second reference line, and calculate the ratio of the first similarity to the average value of the target included angle corresponding to the first reference line. After calculating the difference between the two ratios, perform normalization processing to determine the second significant degree of the hyphal characteristics of the feature line.

[0082] As an example, the calculation formula for the second significant degree of the hyphal characteristics of the i-th feature line can be: ; In the formula, represents the second significant degree of the hyphal characteristics of the i-th feature line, represents the hyperbolic function, which is used to implement data normalization processing, represents the second similarity, represents the average value of the target included angle corresponding to the second reference line, represents the first similarity, represents the average value of the target included angle corresponding to the first reference line, represents the absolute value function.

[0083] In the calculation formula for the second significant degree of the hyphal characteristics, for the i-th feature line, the greater the second similarity corresponding to the second reference line and the smaller the average value of the target included angle, that is, the greater, and the smaller the second similarity corresponding to the first reference line and the greater the average value of the target included angle, that is, the smaller, it indicates that the possibility of intersection between the first reference line and the second reference line is greater, and the second significant degree of the hyphal characteristics is greater.

[0084] S33. Combine the first significant degree and the second significant degree of the hyphal characteristics of each feature line to screen out the pending hyphal segments; take the smallest connected domain in the microscopic image of the substrate layer that contains all the pending hyphal segments as the pending lesion connected domain of the substrate layer.

[0085] In this embodiment, for any feature line, calculate the product of the first significant degree and the second significant degree of the hyphal characteristics of the feature line as the hyphal determination index, and obtain the hyphal determination indexes of each feature line; set a determination threshold as the hyphal determination criterion, and take the empirical value as 0.7. The implementation can set the size of the determination threshold according to specific actual situations and requirements; take the feature lines with the hyphal determination index greater than the determination threshold of 0.7 as the pending hyphal segments. After obtaining all the pending hyphal segments corresponding to the substrate layer, perform connected domain analysis on all the pending hyphal segments, and take the smallest connected domain as the pending lesion connected domain of the substrate layer, that is, the fungal infection area of the substrate layer.

[0086] So far, this embodiment has obtained the pending lesion connected domain of the substrate layer.

[0087] S4. Analyze the regularity and permeability of cells based on the transparency characteristics of the microscopic image of the endothelium to obtain the to-be-determined lesion connected regions of the endothelium.

[0088] When fungal infection penetrates deep into the endothelium, more severe lesions and structural damages usually exist. Since the main function of the endothelium is to maintain the transparency of the cornea, when the infection penetrates deep into the endothelium, the infection will affect the fluid balance and then cause corneal opacity. When the corneal opacity is severe, both the light penetration and the reflected light will decrease significantly, which will significantly affect the imaging quality of the microscopic image of the endothelium. Therefore, it is necessary to analyze the regularity and permeability of cells based on the transparency characteristics of the microscopic image of the endothelium to obtain the to-be-determined lesion connected regions of the endothelium.

[0089] The above step S4 can be implemented through Figure 9 the steps S41 to S43 shown as follows: S41. Obtain the high-threshold region of the microscopic image of the endothelium, perform connected region detection on the high-threshold region to obtain each cell connected region; perform hexagon fitting on each cell connected region to obtain the loss value and the fitting side length.

[0090] In this embodiment, when fungal infection penetrates deep into the endothelium, the structures of some cells are damaged. Therefore, it is necessary to obtain the loss value and the fitting side length of each cell connected region to facilitate quantifying the damage degree of the cell structure, and then obtain the to-be-determined lesion connected regions of the endothelium.

[0091] Specifically, use Otsu threshold segmentation to segment the microscopic image of the endothelium to obtain the high-threshold region; perform connected region detection on the high-threshold region, and several high-threshold connected regions can be obtained. Take the high-threshold connected regions as cell connected regions; for each cell connected region, use the least squares method to perform hexagon fitting on each cell connected region to obtain the loss value of each cell connected region and the total side length of the hexagon, and take the total side length of the hexagon as the fitting side length of the cell connected region.

[0092] Among them, the implementation process of Otsu threshold segmentation and the acquisition of the loss value and the fitting side length in the implementation process of the least squares method are all prior arts and are not within the protection scope of the present invention, so no detailed description will be given here.

[0093] S42. Determine the deformation coefficient of each cell connected region according to the difference between the loss value of the cell connected region, the fitting side length and the average side length; where the average side length is the average value of the fitting side lengths of all cell connected regions.

[0094] In this embodiment, both the loss value and the difference value are numerical values representing the degree of loss. The difference between the fitted side length and the average side length is normalized to obtain the normalized value of the difference, such as the norm function. Therefore, the product of the normalized values of the loss value and the difference value can be used as the deformation coefficient of the corresponding cell connected domain. The larger the loss value and the larger the difference between the fitted side length and the average side length, the more serious the deformation loss of the cell connected domain and the larger the deformation coefficient.

[0095] S43. According to the distribution of pixel points outside the cell connected domain in the microscopic image of the endothelium and the deformation coefficient, perform image segmentation on the microscopic image of the endothelium to obtain the to-be-determined lesion connected domain of the endothelium.

[0096] Here, the pixel points outside the cell connected domain in the microscopic image of the endothelium may be infection pixel points outside the cells. Since the edema caused by the infection in the inner layer of the skin will lead to uneven moisture in the endothelium, it will cause uneven scattering of light in the corneal endothelium, resulting in an uneven brightness situation. Therefore, analyze the light structure characteristics of the area around the pixel points to be detected.

[0097] As an optional implementation manner, obtaining the to-be-determined lesion connected domain of the endothelium includes: The first step is to use the pixel points outside the cell connected domain in the microscopic image of the endothelium as the pixel points to be detected, and determine the minimum distance value between the pixel points to be detected and all cell connected domains in the microscopic image of the endothelium.

[0098] In this embodiment, for any pixel point to be detected, calculate the distance between the pixel point to be detected and each cell connected domain, and determine the minimum distance value from all distances to obtain the minimum distance value corresponding to the pixel point to be detected. Among them, the distance between the pixel point to be detected and the cell connected domain can be obtained by finding the position on the boundary of the connected domain closest to the pixel point to be detected and calculating the distance between these two points.

[0099] The second step is to determine a window area centered on the pixel point to be detected, and determine each maximum value point within the window area; among them, the maximum value point is the pixel point with the largest gray value within its eight-neighborhood range.

[0100] In this embodiment, the size of the window area is determined according to the size of the microscopic image of the endothelium. The larger the size of the microscopic image, the larger the size of the window area. The window area can be set to . For any pixel point, if the gray value of the pixel point is the largest within its eight-neighborhood range, then use this pixel point as the maximum value point to obtain all the maximum value points within the window area.

[0101] In the third step, calculate the information entropy of the distances between each maximum value point in the window area and the pixel points to be detected, and obtain the fungal infection characteristic value of each pixel point to be detected according to the minimum distance value, information entropy, and deformation coefficient.

[0102] In this embodiment, calculate the distances between each maximum value point in the window area and the pixel points to be detected, sort the distances, and then calculate the information entropy based on the sorted distances. Among them, the calculation process of information entropy is prior art and not within the protection scope of the present invention, so it will not be elaborated in detail here; the greater the information entropy, it can be characterized that the distance changes between the maximum value points and the pixel points to be detected in the image are more diverse, and the image structure is more complex.

[0103] As an example, the calculation formula for the fungal infection characteristic value of the j-th pixel point to be detected can be: ; in the formula, represents the fungal infection characteristic value of the j-th pixel point to be detected, th represents the hyperbolic function, which is used to normalize the data so that the value range of the fungal infection characteristic value is limited between 0 and 1, represents the minimum distance value of the j-th pixel point to be detected, represents the deformation coefficient of the j-th pixel point to be detected, represents the information entropy of the j-th pixel point to be detected.

[0104] In the calculation formula of the fungal infection characteristic value, the smaller the minimum distance value , it indicates that the j-th pixel point to be detected is closer to the cell connectivity domain, and the greater the possibility of its being infected, and the greater the fungal infection characteristic value of the j-th pixel point to be detected; the larger the deformation coefficient , it represents that the deformation loss of the cell connectivity domain corresponding to the j-th pixel point to be detected is greater, and the greater the influence of fungal infection; the greater the information entropy , it represents that the structural change in the surrounding area of the j-th pixel point to be detected is more complex, the light change difference is greater, and it is more likely to be the scattering caused by uneven brightness, which conforms to the uneven scattering of light in the corneal endothelium caused by fungal infection.

[0105] In the fourth step, perform threshold segmentation on the fungal infection characteristic values of each pixel point to be detected in the microscopic image of the endothelium to obtain high-threshold pixel points.

[0106] In this embodiment, perform Otsu threshold segmentation on the fungal infection characteristic values of each pixel point to be detected in the microscopic image of the endothelium to obtain high-threshold pixel points, so as to screen out the pixel points to be detected with more serious fungal infections, and then perform connectivity domain analysis based on the screened high-threshold pixel points.

[0107] Step 5: Merge the high-threshold pixel points, the cell-connected domains adjacent to the high-threshold pixel points, and the cell-connected domains surrounded by the high-threshold pixel points to obtain the pending lesion-connected domain of the endothelium.

[0108] In this embodiment, the merging process refers to performing connected domain analysis on the high-threshold pixel points, the cell-connected domains adjacent to the high-threshold pixel points, and the cell-connected domains surrounded by the high-threshold pixel points, and finally obtaining a connected domain, which is used as the pending lesion-connected domain of the endothelium.

[0109] Among them, the cell-connected domains adjacent to the high-threshold pixel points and the cell-connected domains surrounded by the high-threshold pixel points are both areas affected by fungal infection, so they need to be merged.

[0110] So far, this embodiment has obtained the pending lesion-connected domain of the endothelium.

[0111] S5. Merge the pending lesion-connected domains of the epithelium, stroma, and endothelium, and delineate the fungal lesion area corresponding to the corneal sample.

[0112] In this embodiment, the microscopic images of the epithelium, stroma, and endothelium are enlarged and reduced in the same proportion to obtain adjusted microscopic images; the pending lesion-connected domains in the three adjusted microscopic images are superimposed to obtain a superimposed lesion-connected domain, which is used as the fungal lesion area corresponding to the corneal sample, and the fungal lesion area in the image is marked. Among them, an example diagram of the fungal lesion area corresponding to the corneal sample is as Figure 10 shown.

[0113] So far, this embodiment has completed the accurate delineation of the fungal lesion area, that is, accurately identified the fungal lesion area.

[0114] The above-described embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that: they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements on some of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the scope of the technical solutions of the various embodiments of the present invention, and should all be included in the protection scope of the present invention.

Claims

1. A method for delineating fungal lesions of fungal keratitis under a microscope, characterized in that, Including the following steps: Obtain microscopic images of the epithelium, stroma, and endothelium at different depths of the corneal sample; Analyze the influence range of fungal spores based on the gray-scale distribution characteristics of the microscopic images of the epithelium to obtain the undetermined lesion connected regions of the epithelium; Analyze the interlacing situation of the highlight parts based on the edge distribution characteristics of the microscopic images of the stroma to obtain the undetermined lesion connected regions of the stroma; Analyze the regularity and permeability of cells based on the transparency characteristics of the microscopic images of the endothelium to obtain the undetermined lesion connected regions of the endothelium; Fuse the undetermined lesion connected regions of the epithelium, stroma, and endothelium to delineate the fungal lesion area corresponding to the corneal sample.

2. The method for delineating fungal lesions of fungal keratitis under a microscope according to claim 1, wherein The analyzing the influence range of fungal spores based on the gray-scale distribution characteristics of the microscopic images of the epithelium to obtain the undetermined lesion connected regions of the epithelium includes: Performing superpixel segmentation on the microscopic images of the epithelium to obtain each superpixel cluster; performing fitting analysis on each superpixel block in the superpixel cluster to obtain the fitting evaluation value of each superpixel cluster; Classifying each superpixel cluster according to the gray-scale value situation of each superpixel cluster; recording the superpixel cluster with the highest gray-scale as the spore cluster; Obtaining the undetermined lesion connected regions of the epithelium according to the fitting evaluation values of each superpixel cluster and each spore cluster.

3. The method for delineating fungal lesions of fungal keratitis under a microscope according to claim 2, wherein, The obtaining the undetermined lesion connected regions of the epithelium according to the fitting evaluation values of each superpixel cluster and each spore cluster includes: Obtaining the surrounding area of all the spore clusters in the microscopic images of the epithelium, continuously adjusting the area size of the surrounding area through an optimization algorithm until the area of the surrounding area is the largest and the cumulative value of the fitting evaluation values of all the superpixel clusters within the surrounding area is the smallest, and determining the final surrounding area as the undetermined lesion connected regions of the epithelium; wherein, the surrounding area is the area containing all the spore clusters.

4. A method for delineating fungal lesions of fungal keratitis under a microscope according to claim 1, characterized in that, The analyzing the interlacing situation of the highlight parts based on the edge distribution characteristics of the microscopic images of the stroma to obtain the undetermined lesion connected regions of the stroma includes: Performing edge detection on the microscopic images of the stroma to obtain each edge pixel point, and performing threshold segmentation on the microscopic images of the stroma to obtain each high-threshold pixel point; Selecting the edge pixel points from all the high-threshold pixel points as feature pixel points; Determining the first significant degree of the hyphal characteristics of each feature line according to each feature pixel point; wherein, the first significant degree of the hyphal characteristics is used to characterize the straightness degree of the feature line; Determining the second significant degree of the hyphal characteristics of each feature line according to the mutual intersection characteristics between each feature line; Combining the first significant degree and the second significant degree of the hyphal characteristics of each feature line to screen out the undetermined hyphal segments; Taking the smallest connected region in the microscopic images of the stroma that contains all the undetermined hyphal segments as the undetermined lesion connected regions of the stroma.

5. A method for delineating fungal lesions of fungal keratitis under a microscope according to claim 4, characterized in that, The determining the first significant degree of the hyphal characteristics of each feature line according to each feature pixel point includes: Connecting adjacent feature pixel points within the eight-neighborhood range to form feature lines, obtaining a plurality of feature lines; For any characteristic line, obtain the length of the line connecting the two endpoints of the characteristic line, and compare the length of the characteristic line with the length of the connecting line to obtain the deviation degree; From each characteristic pixel point on the characteristic line, draw a perpendicular line to the line segment of the connecting line length. The direction of the perpendicular line is perpendicular to the line segment of the connecting line length. Determine the sharpness according to the difference in the number of perpendicular lines on both sides of the line segment of the connecting line length; Obtain the edge chain code of the characteristic pixel points on the characteristic line, and obtain the tortuosity according to the difference between adjacent code numbers in the edge chain code; According to the deviation degree, the sharpness, and the tortuosity of the characteristic line, obtain the first significant degree of the hypha characteristics of the characteristic line; Among them, the deviation degree, the sharpness, and the tortuosity are all negatively correlated with the first significant degree of the hypha characteristics.

6. A method for delineating fungal lesions of fungal keratitis under a microscope according to claim 4, characterized in that The determining the second significant degree of the hypha characteristics of each characteristic line according to the mutual intersection characteristics between the characteristic lines includes: For any characteristic line, record the characteristic line that is the first closest to the endpoint of the characteristic line as the first reference line, and record the characteristic line that is the second closest to the endpoint of the characteristic line as the second reference line; Extend the characteristic line in the direction of the to-be-determined reference line, and then connect it to the to-be-determined reference line to obtain a new characteristic line; wherein, the to-be-determined reference line is the first reference line or the second reference line; Obtain each inflection point on the to-be-determined reference line, and then obtain the included angle formed by the connection between the two endpoints of the to-be-determined reference line and any inflection point, and take the minimum included angle as the target included angle corresponding to the to-be-determined reference line; Determine the second significant degree of the hypha characteristics of the characteristic line according to the uniform similarity degree of the gray-scale distribution between the characteristic line and the new characteristic line, the first reference line of the characteristic line, and the target included angle corresponding to the second reference line.

7. A method for delineating fungal lesions of fungal keratitis under a microscope according to claim 6, characterized in that, The determining the second significant degree of the hypha characteristics of the characteristic line according to the uniform similarity degree of the gray-scale distribution between the characteristic line and the new characteristic line, the first reference line of the characteristic line, and the target included angle corresponding to the second reference line includes: Obtain the first gray-scale variance of the characteristic line, the second gray-scale variance of the new characteristic line when the to-be-determined reference line is the first reference line, and the third gray-scale variance of the new characteristic line when the to-be-determined reference line is the second reference line; Determine the first similarity according to the difference between the first gray-scale variance and the second gray-scale variance, and determine the second similarity according to the difference between the first gray-scale variance and the third gray-scale variance; Determine the second significant degree of the hypha characteristics of the characteristic line through the first similarity, the second similarity, and the target included angle corresponding to the first reference line and the second reference line.

8. A method for delineating fungal lesions of fungal keratitis under a microscope according to claim 7, characterized in that, The determining the second significant degree of the hypha characteristics of the characteristic line through the first similarity, the second similarity, and the target included angle corresponding to the first reference line and the second reference line includes: Calculate the ratio of the second similarity to the average target included angle corresponding to the second reference line, and calculate the ratio of the first similarity to the average target included angle corresponding to the first reference line. After calculating the difference between the two ratios, perform normalization processing to determine the second significant degree of the hypha characteristics of the characteristic line.

9. A method for delineating fungal lesions of fungal keratitis under a microscope according to claim 1, characterized in that, Analyze the regularity and permeability of cells based on the transparency characteristics of the microscopic image of the endodermis to obtain the to-be-determined lesion connected domain of the endodermis, including: Obtain the high-threshold region of the microscopic image of the endodermis, perform connected domain detection on the high-threshold region to obtain each cell connected domain; perform hexagonal fitting on each cell connected domain to obtain a loss value and the fitted side length. Determine the deformation coefficient of each cell connected domain according to the loss value of the cell connected domain, the difference between the fitted side length and the average side length; where the average side length is the average of the fitted side lengths of all cell connected domains. Perform image segmentation on the microscopic image of the endodermis according to the distribution of pixel points outside the cell connected domain in the microscopic image of the endodermis and the deformation coefficient to obtain the to-be-determined lesion connected domain of the endodermis.

10. A method for delineating fungal lesions of fungal keratitis under a microscope according to claim 9, characterized in that, The performing image segmentation on the microscopic image of the endodermis according to the distribution of pixel points outside the cell connected domain in the microscopic image of the endodermis and the deformation coefficient to obtain the to-be-determined lesion connected domain of the endodermis includes: Take the pixel points outside the cell connected domain in the microscopic image of the endodermis as the to-be-detected pixel points, and determine the minimum distance value between the to-be-detected pixel points and all cell connected domains in the microscopic image of the endodermis. Determine a window region centered on the to-be-detected pixel points, and determine each maximum value point within the window region; where the maximum value point is the pixel point with the largest gray value within its eight-neighborhood range. Calculate the information entropy of the distance between each maximum value point within the window region and the to-be-detected pixel points, and obtain the fungal infection characteristic value of each to-be-detected pixel point according to the minimum distance value, the information entropy, and the deformation coefficient. Perform threshold segmentation through the fungal infection characteristic values of each to-be-detected pixel point in the microscopic image of the endodermis to obtain high-threshold pixel points. Perform a merging process on the high-threshold pixel points, the cell connected domains adjacent to the high-threshold pixel points, and the cell connected domains surrounded by the high-threshold pixel points to obtain the to-be-determined lesion connected domain of the endodermis.

Images