Method for measuring strabismus degree of rotary strabismus patient based on fundus photo

Abstract

The invention relates to the technical field of strabismus auxiliary measurement, in particular to a method for measuring the strabismus degree of a rotary strabismus patient based on a fundus photograph, and the method comprises the steps: collecting a fundus image and carrying out the gray processing, and obtaining a rotary strabismus image; according to a gray value difference condition among eight neighborhood pixel points of each pixel point in the rotation type squint image, obtaining a fundus structure difference degree of each pixel point; obtaining the number of neighborhood similar structures of each pixel point according to the difference distribution between the fundus structure dissimilarity degrees of each pixel point and all eight neighborhood pixel points; obtaining an adaptive smoothing parameter of each pixel point according to the number of the neighborhood similar structures; performing filtering processing on the rotary strabismus image according to the self-adaptive smoothing parameter of each pixel point to obtain a rotary strabismus measurement image, detecting a circle in the rotary strabismus measurement image to obtain a macular and an optic disc, and calculating the rotary strabismus degree. The strabismus degree measuring method and device can improve the strabismus degree measuring accuracy.

Description

Technical Field

[0001] The present application relates to the technical field of strabismus auxiliary measurement, and in particular to a method for measuring the strabismus degree of a patient with rotational strabismus based on fundus photographs. Background Art

[0002] Rotational strabismus is a complex eye movement abnormality, which manifests as a rotational tilt of one or both eyes around its anterior-posterior axis toward the temporal or nasal side. This condition often coexists with hypertropia and hypotropia, mainly due to abnormal vertical muscle function of the eye, such as excessive or weakened superior or inferior rectus muscle function. Rotational strabismus not only affects the motor function of the eye and has a significant impact on visual function, but may also cause discomfort such as headaches and nausea. Rotational strabismus is generally detected by measuring the angle between the macula and the middle and lower 1 / 3 of the optic disc. The accuracy of rotational strabismus measurement is mainly limited by the positioning accuracy of the macula and optic disc of the eye. At present, the positioning of the macula and optic disc of the eye is usually achieved by processing and identifying fundus images.

[0003] Although the relative position changes of the macula and optic disc can be determined through fundus photographs, their positions cannot be accurately determined due to the clarity of the photographs, and the rotation angle of the eyeball cannot be accurately quantified. As a result, in cases of small rotation angles or mild rotational strabismus, the image may not be sensitive enough and difficult to be directly observed in the photographs. In order to clearly identify the macula and optic disc, the fundus images are generally denoised through filtering algorithms to enhance the clarity of the images, and the enhanced images are used to identify the macula and optic disc. However, since the same structures in the fundus images of patients have the characteristics of connection and similarity, when the same filtering parameters are used for denoising the entire fundus image, it is easy to reduce the clarity of the image at the edge of the macula and optic disc, affecting the accuracy of strabismus measurement. Summary of the invention

[0004] In order to solve the above technical problems, the present application provides a method for measuring the strabismus of patients with rotational strabismus based on fundus photographs to solve the existing problems.

[0005] The method for measuring the strabismus of patients with rotational strabismus based on fundus photographs of the present application adopts the following technical solution: One embodiment of the present application provides a method for measuring the strabismus of a patient with rotational strabismus based on a fundus photograph, the method comprising the following steps: S1, collecting fundus images and performing grayscale processing to obtain rotational oblique vision images; S2, according to the gray value difference between the eight neighboring pixels of each pixel in the rotational oblique image, obtain the neighborhood distribution sequence of each pixel; according to the gray value difference between the pixels before and after all adjacent pixels in the neighborhood distribution sequence of each pixel, obtain the fundus structure difference of each pixel; S3, according to the difference distribution between the fundus structure dissimilarity of each pixel and all its eight neighboring pixels, the number of similar structures in the neighborhood of each pixel is obtained; according to the number of similar structures in the neighborhood of each pixel, the adaptive smoothing parameter of each pixel in the filtering process is obtained; S4, filtering the rotational strabismus image according to the adaptive smoothing parameter of each pixel point to obtain a rotational strabismus measurement image, detecting the circle in the rotational strabismus measurement image to obtain the macula and the optic disc; and measuring the strabismus degree according to the positions of the macula and the optic disc.

[0006] Furthermore, the method for acquiring the neighborhood distribution sequence of each pixel point includes: Any pixel point in the rotational oblique image is recorded as the target point; the pixel point with the smallest grayscale value among the eight neighboring pixel points of the target point is obtained and recorded as the starting point; starting from the starting point, all eight neighboring pixel points are arranged in the sorting direction to obtain the neighborhood distribution sequence of the target point.

[0007] Furthermore, the method for obtaining the sorting direction includes: Respectively obtain two pixel points adjacent to the starting point in the eight neighborhood pixel points, and record them as two adjacent points of the starting point; obtain the difference between the grayscale values ​​of the starting point and each of its adjacent points; take the adjacent point with the smallest difference as the direction point, wherein if the differences between two adjacent points are the same, select any adjacent point as the direction point; if the direction point is located in the clockwise direction of the starting point, the sorting direction is clockwise; otherwise, the sorting direction is counterclockwise.

[0008] Furthermore, the method for obtaining the fundus structure difference of each pixel point includes: For the neighborhood distribution sequence of each pixel point, the fundus structure recognition degree between each pair of adjacent pixel points is obtained according to the gray value difference between all pixel points before and after each pair of adjacent pixel points in the neighborhood distribution sequence; In the neighborhood distribution sequence, the first and second pixels in a pair of adjacent pixels with the highest fundus structure recognition are recorded as the front structure discrimination point and the back structure discrimination point, respectively. According to the gray value difference between all the pixels before the front structure determination point and all the pixels after the back structure determination point, the fundus structure difference of each pixel is obtained.

[0009] Furthermore, the method for obtaining the fundus structure recognition degree between each pair of adjacent pixel points includes: recording the fundus structure recognition degree between the i-th and i+1-th pixel points in the neighborhood distribution sequence as , ;in, , They represent the grayscale values ​​of the j-th and k-th pixels in the neighborhood distribution sequence respectively; i represents the sequence number of the previous pixel between the current two adjacent pixels in the neighborhood distribution sequence; n represents the number of pixels in the neighborhood distribution sequence; || is the absolute value symbol.

[0010] Furthermore, the grayscale value difference between all pixels before the front structure determination point and all pixels after the back structure determination point is used to obtain the fundus structure difference of each pixel, including: The mean grayscale value of all pixels from the first pixel point to the front structure discrimination point in the neighborhood distribution sequence is obtained, which is recorded as the first structure feature value; the mean grayscale value of all pixels from the back structure discrimination point to the last pixel point in the neighborhood distribution sequence is obtained, which is recorded as the second structure feature value; the difference between the grayscale value of each pixel point and the first structure feature value and the second structure feature value is obtained respectively, which is recorded as the first difference and the second difference; the minimum value between the first difference and the second difference is taken as the fundus structure difference degree of each pixel point.

[0011] Furthermore, the method for obtaining the number of similar structures in the neighborhood of each pixel point includes: The segmentation threshold is obtained according to the difference between the fundus structure of all pixels and their eight neighboring pixels; The number of fundus structures between each pixel and all its eight neighboring pixels whose dissimilarity is less than the segmentation threshold is counted and recorded as the number of similar neighborhood structures of each pixel.

[0012] Furthermore, the method for obtaining the segmentation threshold comprises: Obtaining the difference in fundus structure difference between each pixel point and each of its eight neighboring pixels as the fundus structure dissimilarity between each pixel point and each of its eight neighboring pixels; The Otsu threshold analysis is performed on the fundus structure dissimilarity between all pixels and all their eight neighboring pixels to obtain the segmentation threshold.

[0013] Further, the method of obtaining the adaptive smoothing parameter of each pixel in the filtering process according to the number of similar structures in the neighborhood of each pixel includes: recording the adaptive smoothing parameter of the ath pixel in the rotational oblique image as , , where Indicates the preset initial smoothing parameter; Indicates the preset adjustment factor; Represents the number of similar structures in the neighborhood of the a-th pixel in the rotational oblique image; Represents the sigmoid normalization function.

[0014] Furthermore, the method for obtaining the macula and optic disc comprises: Substitute the adaptive smoothing parameters of each pixel in the rotational squint image into the NLM filtering algorithm to denoise the rotational squint image and obtain the rotational squint measurement image. The circles in the rotational strabismus measurement image are obtained; the inner area of ​​the circle with the second largest diameter is taken as the macula; and the inner area of ​​the circle with the smallest diameter is taken as the optic disc.

[0015] This application has at least the following beneficial effects: The present application analyzes the influence of the connection and similarity characteristics of the same structure in the fundus image on the filtering effect. First, the characteristics of different eyeball structures in the fundus image are analyzed. According to the certain continuity of the same structure, the local grayscale difference of the pixel point is used to obtain the fundus structure difference of each pixel point, which reflects the position of the pixel point in the fundus structure and preliminarily judges the noise situation; further analyzes the similarity of the fundus structure difference between the pixel point of the same structure and the pixel points in its neighborhood, and obtains the number of neighborhood similar structures of each pixel point. The more the number of neighborhood similar structures, the more obvious the fundus structure characteristics of the pixel point; in order to eliminate the interference of noise, the adaptive smoothing parameter of each pixel point is determined according to the number of neighborhood similar structures of the pixel point, and each pixel point in the image is locally filtered according to the adaptive smoothing parameter, which improves the clarity of the image after filtering, can accurately identify the position of the macula and the optic disc, and makes the strabismus measurement result more accurate. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the technical solutions and advantages in the embodiments of the present application or the prior art, the drawings required for use in the embodiments or the prior art descriptions are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0017] Figure 1 A flowchart of the steps of the method for measuring the strabismus of patients with rotational strabismus based on fundus photographs provided in the present application; Figure 2 A block diagram for obtaining fundus structure differences provided in accordance with an embodiment of the present application. DETAILED DESCRIPTION

[0018] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. In the absence of conflict, the embodiments of the present application and the technical features in the embodiments can be combined with each other. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0019] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs.

[0020] The specific scheme of the method for measuring the strabismus of patients with rotational strabismus based on fundus photographs provided by the present application is described in detail below with reference to the accompanying drawings.

[0021] An embodiment of the present application provides a method for measuring the strabismus of a patient with rotational strabismus based on a fundus photograph. Specifically, the following method for measuring the strabismus of a patient with rotational strabismus based on a fundus photograph is provided. Figure 1 , the method comprises the following steps: S1, collecting fundus images and performing grayscale processing to obtain rotational oblique vision images.

[0022] The impact of rotational strabismus on visual function is mainly reflected in the disruption of visual fusion. Visual fusion refers to the process of integrating the visual information received by the two eyes into a clear image in the brain. When the rotation angle of the eyes is abnormal, the images seen by the two eyes are no longer precisely aligned, resulting in difficulty or inability to perform visual fusion. This fusion disorder not only causes diplopia (seeing repeated images), but may also further induce a series of systemic symptoms, such as headaches, nausea, vomiting, etc., and in severe cases even affect the quality of daily life. At present, the main methods used to measure the strabismus of rotational strabismus are as follows: (1) Maddox rod test: The Maddox rod test is a simple and commonly used subjective measurement method. The test uses a Maddox rod containing a red parallel line to refract light into a line. The patient reports the condition of rotational strabismus by observing the position relationship between the red line and the sight mark. This method can evaluate the vertical and horizontal deviations of the eye, and can also roughly measure the rotational deviation.

[0023] (2) Hess screen test: The Hess screen test is a tool commonly used to evaluate eye movement disorders, especially when evaluating strabismus. This method requires the patient to sit in front of a gridded screen and record the response of each eye to different sight marks by moving the eyes in different directions. By analyzing the tracking trajectory of the eyes to the visual target, the presence and degree of rotational strabismus can be evaluated.

[0024] (3) Synoptophore: Synoptophore is a device used to measure binocular coordination ability. It tests the patient's eye movement response by providing different visual targets to each eye and records the changes in eye position. Synoptophore can not only measure horizontal and vertical strabismus, but also be used to evaluate rotational strabismus. It is one of the commonly used clinical examination methods.

[0025] (4) Fundus photography: Judging from the relative position of the macula and the optic disc in the fundus photograph, the macula is not completely horizontally aligned with the center point of the optic disc, but is slightly lower than the center of the optic disc and is located in the middle and lower third of the optic disc. If the position of the macula relative to the optic disc is abnormal (such as the macula is significantly moved up or down), it may indicate rotational strabismus of the eyeball.

[0026] The most widely used method is to locate the macula and optic disc through fundus photographs, and measure the strabismus based on the abnormal position of the macula relative to the optic disc. This application aims to accurately determine the position of the macula and optic disc through fundus photographs, and then measure the patient's rotational strabismus based on this.

[0027] Specifically, a specific implementation scenario of the embodiment of the present application is a fundus image filtering scenario during the strabismus measurement process of rotational strabismus; in the detection of rotational strabismus in patients, it is usually necessary to locate the macula and optic disc of the eyeball, and measure the strabismus by measuring the angle between the macula and the lower 1 / 3 of the optic disc.

[0028] Therefore, the present embodiment first acquires the fundus image by a non-mydriatic fundus color camera. The shooting method is: the patient's lower jaw is placed on the jaw rest of the detector, and the forehead is ensured to be close to the head frame to maintain the stability of the head and avoid any movement during the shooting process. The original fundus image acquired is an RGB color image. In order to better assist in the detection of strabismus, the present embodiment adopts the weighted average method to convert the color image into a grayscale image, which is recorded as a rotational strabismus image. The weighted average method is a well-known technology, and the specific calculation process will not be repeated.

[0029] In the process of positioning the macula and optic disc of the eyeball according to the rotational strabismus image, the positioning accuracy is limited by the image clarity. In the process of filtering the rotational strabismus image, since the same structures in the patient's fundus image have the characteristics of connection and similarity, using the same filtering parameters for the entire rotational strabismus image will easily lead to a decrease in the clarity of the image at the edge of the macula and optic disc. Therefore, this embodiment uses the local distribution characteristics of the pixel points to analyze the similarity characteristics of the local structure, judge the recognition degree of the fundus structure of the pixel points; and adaptively determine the filtering parameters according to the recognition degree of the fundus structure of the pixel points.

[0030] S2, according to the gray value difference between the eight neighboring pixels of each pixel in the rotational oblique vision image, the neighborhood distribution sequence of each pixel is obtained; according to the gray value difference between the pixels before and after all adjacent pixels in the neighborhood distribution sequence of each pixel, the fundus structure difference of each pixel is obtained.

[0031] The fundus images taken include the blood vessels, retina, macula and optic disc of the fundus. The local features of the fundus images have high similarity features. When the pixel points in the rotational oblique image are noise points, the pixel values ​​of the noise points are greatly different from those of the local neighboring pixels. When denoising, the denoising intensity should be enhanced for the position of the noise points.

[0032] The eight-neighborhood pixel points of each pixel point of the rotational oblique image are obtained, and the mean filling method is used to fill the insufficient pixel points in the window for the edge pixels of the image. Since the pixel points of the rotational oblique image are highly similar in the same structure, the main structures are blood vessels, retina, macula of the eyeball, and optic disc. There are obvious differences between different structures, such as the obvious difference in grayscale values ​​between the retina and retinal blood vessels.

[0033] If the pixels are formed by the same structure, the difference between the grayscale values ​​of the pixels is small. In order to distinguish the composition of the local neighborhood of the pixels, the eight-neighborhood pixels of each pixel are sorted. The sorting method is: any pixel in the rotational oblique image is recorded as the target point; the pixel with the smallest grayscale value among the eight-neighborhood pixels of the target point is obtained and recorded as the starting point; starting from the starting point, all the eight-neighborhood pixels are arranged in the sorting direction to obtain the neighborhood distribution sequence of the target point.

[0034] The method for obtaining the sorting direction is as follows: respectively obtain two pixel points adjacent to the starting point in the eight neighboring pixel points, and record them as two adjacent points of the starting point; obtain the difference between the grayscale values ​​of the starting point and each of its adjacent points; use the adjacent point with the smallest difference as the direction point, wherein, if the difference between the two adjacent points is the same, select any adjacent point as the direction point; if the direction point is located in the clockwise direction of the starting point, the sorting direction is clockwise; otherwise, the sorting direction is counterclockwise. The two pixel points adjacent to the starting point represent the two pixel points closest to the starting point.

[0035] The direction point represents the direction of the gradual change of grayscale starting from the starting point. Therefore, the adjacent points with the smallest difference are selected as direction points and arranged in the sorting direction. The obtained neighborhood distribution sequence reflects the gradual change characteristics of the grayscale around the central pixel.

[0036] In this embodiment, the difference is calculated by using the absolute value of the difference, and the obtained neighborhood distribution sequence reflects the fundus structure characteristics of the pixel position.

[0037] Since the fundus image contains structures such as the lens, blood vessels, retina, macula, and optic disc, there are obvious differences between the various structures, and the same structure has a certain physical connection in the image, so the difference between the grayscale values ​​of the pixels of the rotational oblique image and the pixels of its surrounding neighborhood is small. When there is noise in the neighborhood of the central pixel or the central pixel is the edge of the structure such as the blood vessel, macula, and optic disc edge in the rotational oblique image, the grayscale value difference between the central pixel and only some of the neighboring pixels is small. In this case, there is a certain difference between the pixels on both sides of the neighborhood distribution sequence. Therefore, according to the grayscale value difference on both sides of the adjacent pixels in the neighborhood distribution sequence, the neighborhood distribution sequence is divided to identify the fundus structure.

[0038] First, the fundus structure recognition degree between adjacent pixels is calculated based on the grayscale value difference on both sides of adjacent pixels in the neighborhood distribution sequence. For each pixel’s neighborhood distribution sequence, the fundus structure recognition degree between the i-th and i+1-th pixels in the neighborhood distribution sequence is recorded as , ;in, , They represent the grayscale values ​​of the j-th and k-th pixels in the neighborhood distribution sequence respectively; i represents the sequence number of the previous pixel between the current two adjacent pixels in the neighborhood distribution sequence; n represents the number of pixels in the neighborhood distribution sequence, n=8; || is the absolute value symbol.

[0039] When the difference between the pixels on both sides of the neighborhood distribution sequence is greater, that is, the absolute value of the difference between the mean of the first i pixels and the mean of the last ni pixels is greater, it means that there are obvious differences on both sides of the neighborhood distribution sequence. This difference is caused by different types of fundus structures. Therefore, the fundus structure recognition degree between the i-th and i+1-th pixels in the neighborhood distribution sequence is The larger the value of .

[0040] The position with the highest fundus structure recognition in the neighborhood distribution sequence indicates that the pixels on both sides of the position are significantly different, and the two sides of the sequence are formed by different structures. In order to distinguish different structures of the fundus, the neighborhood distribution sequence is segmented by selecting the position with the highest fundus structure recognition in the neighborhood distribution sequence. Specifically, in the neighborhood distribution sequence, the first and second pixels in a pair of adjacent pixels with the highest fundus structure recognition are recorded as the front structure discrimination point and the back structure discrimination point, respectively; The mean grayscale value of all pixels from the first pixel point to the front structure discrimination point in the neighborhood distribution sequence is obtained, which is recorded as the first structure feature value; the mean grayscale value of all pixels from the back structure discrimination point to the last pixel point in the neighborhood distribution sequence is obtained, which is recorded as the second structure feature value; the difference between the grayscale value of each pixel point and the first structure feature value and the second structure feature value is obtained respectively, which is recorded as the first difference and the second difference; the minimum value between the first difference and the second difference is taken as the fundus structure difference degree of each pixel point.

[0041] For example, the fundus structure recognition degree between the qth and q+1th pixels in the neighborhood distribution sequence is the largest, then the grayscale value average of the first q pixels is the first structural feature value, and the grayscale value average of the last nq pixels is the second structural feature value.

[0042] When the pixel in the rotational oblique image is a normal pixel, there is a certain similarity between the pixel and its local neighborhood pixels. Since the pixel may exist at the edge of the structure in the fundus image, there are also some dissimilar pixels between the pixel in this part and the neighborhood. Therefore, the minimum value of the difference between the central pixel and different structures is taken as the fundus structure difference of the central pixel.

[0043] The block diagram for obtaining the difference of fundus structure is as follows: Figure 2 shown.

[0044] S3, according to the difference distribution between the fundus structure dissimilarity of each pixel and all its eight neighboring pixels, the number of similar neighborhood structures of each pixel is obtained; according to the number of similar neighborhood structures of each pixel, the adaptive smoothing parameter of each pixel in the filtering process is obtained.

[0045] For normal pixels, there are pixels with similar or identical structures to the pixel in the neighborhood of the pixel. The more similar the structures between two pixels are, the smaller the difference between the fundus structure differences between the pixels. The more similar pixels there are in the neighborhood, the more normal the pixel is. Therefore, the structural characteristics of the pixel are judged based on the difference between the fundus structure differences between the central pixel and its eight neighborhood pixels.

[0046] The difference between the fundus structure difference between each pixel point and its eight neighboring pixel points is obtained as the fundus structure dissimilarity between each pixel point and its eight neighboring pixel points.

[0047] The smaller the difference in fundus structure between the central pixel and its neighboring pixels, the closer the structural features of the two pixels are, and the closer the two pixels are to each other. The more pixels with similar structural features, the more likely the central pixel is to be a normal pixel. The denoising intensity should be reduced for the central pixel to ensure the detailed properties of the pixel.

[0048] The fundus structure dissimilarity of all pixels and all eight neighboring pixels is analyzed by Otsu threshold to obtain the segmentation threshold; specifically, the fundus structure dissimilarity of all pixels and all eight neighboring pixels is used as the input of OTSU method, and the output is the segmentation threshold of fundus structure dissimilarity. Usually, the grayscale difference between pixels in the same structure is small, so that the fundus structure dissimilarity between pixels of the same structure is small. Therefore, the fundus structure dissimilarity of all pixels whose fundus structure dissimilarity is less than or equal to the threshold is obtained, and they are formed into a fundus structure similarity set. The OTSU method is a well-known technology, and the specific calculation process is not repeated here.

[0049] Furthermore, the number of fundus structures with a dissimilarity less than the segmentation threshold between each pixel and all its eight neighboring pixels is counted, and recorded as the number of neighborhood similar structures of each pixel. The larger the number of neighborhood similar structures of a pixel, the more similar points the data point has locally.

[0050] The larger the number of similar structures in the neighborhood of a pixel point, the more the pixel points are distributed in the same structure of the fundus image, and the pixel points are smoother in the local area. When denoising the pixel point, a smaller denoising intensity should be selected to retain more structural information, so that the detection of the patient's rotational strabismus can be more accurate.

[0051] For the NLM filtering algorithm, the smoothing parameter h is used to control the degree of noise suppression and the degree of retention of image details in the algorithm's denoising process. If more structural features need to be retained, the smoothing parameter h needs to be reduced so that the pixel points have more detailed attributes during the NLM filtering process; conversely, if the pixel points are noise points, a larger smoothing parameter h should be used to enhance the filtering effect.

[0052] Therefore, this embodiment calculates the adaptive smoothing parameter of the pixel point by the number of similar structures in the neighborhood of the pixel point. The adaptive smoothing parameter of the ath pixel point in the rotational oblique image is recorded as , , where represents the preset initial smoothing parameter in the NLM algorithm, and its value in this scheme is 12; A represents the preset adjustment factor, and its value is 9; Represents the number of similar structures in the neighborhood of the a-th pixel in the rotational oblique image; Represents the sigmoid normalization function.

[0053] When the adaptive smoothing parameter of the pixel in the rotational oblique image The larger the value is, the greater the number of similar structures in the neighborhood of the pixel point within its local range. The smaller the value, the more likely the pixel is a noise point. The filtering effect should be increased for this pixel so that the pixel value returns to the normal range.

[0054] S4, filtering the rotational strabismus image according to the adaptive smoothing parameter of each pixel point to obtain a rotational strabismus measurement image, detecting the circle in the rotational strabismus measurement image to obtain the macula and the optic disc; and measuring the strabismus degree according to the positions of the macula and the optic disc.

[0055] The rotational squint image is used as the input of the NLM algorithm, and the adaptive smoothing parameter of the pixel point is used as the smoothing parameter of each pixel point in the filtering process of the NLM algorithm, wherein the side length of the algorithm's search window is 21, and the side length of the neighborhood window is 3, and the output is the filtered rotational squint image, which is recorded as the rotational squint measurement image. The NLM algorithm is a well-known technology, and the specific calculation process is not repeated here.

[0056] For the rotational strabismus measurement image calculated in the above steps, Hough transform is used to fit the circles in the rotational strabismus measurement image, wherein the largest circle is the eyeball, the second largest circle is the macula, and the smallest circle is the optic disc. Hough transform is a well-known technology, and the specific calculation process is not repeated here.

[0057] Finally, the strabismus is measured according to the position of the macula and the optic disc. The calculation of the rotational strabismus is a well-known technique, specifically including: in the normal fundus image without rotational strabismus, the position of the macula is at the lower 1 / 3 of the optic disc. In order to measure the patient's rotational strabismus, the enhanced rotational strabismus measurement image is imported into the Auto CAD professional drawing software, and the center position of the second largest circle is taken as the center position of the macula. The length of the smallest circle is measured by professional drawing software and divided into three equal parts. The position of the lower 1 / 3 mark point of the smallest circle is taken as the position of the lower 1 / 3 point of the optic disc. The center of the macular circle is connected to form a measurement straight line. The angle between the measurement straight line and the horizontal line at the lower 1 / 3 of the optic disc in the enhanced rotational strabismus measurement image is the rotational strabismus of the patient's eye.

[0058] The doctor can obtain the patient's rotational strabismus through the above steps as an aid to measure the patient's rotational strabismus.

[0059] In summary, judging rotational strabismus through fundus photographs is a non-invasive examination method that does not require direct contact with the eyeball and does not involve any complex equipment. It is only necessary to determine whether there is rotation of the eyeball based on the position of the macula and optic disc in the fundus photograph. This process is painless to the patient and is easy to operate, suitable for most people. In this embodiment, the processing method for analyzing the characteristics of the fundus photographs improves the recognition accuracy of the macula and optic disc, obtains a more accurate position of the two, and thereby improves the accuracy of the strabismus calculated by the macula and optic disc.

[0060] Through the above description of the implementation method in combination with the accompanying drawings, technical personnel in the relevant field can understand that for the convenience and simplicity of description, only the division of the above-mentioned functional modules is used as an example. In actual applications, the above-mentioned functions can be assigned to different functional modules as needed, that is, the internal structure of the device can be divided into different functional modules to complete all or part of the functions described above.

[0061] The above contents are only specific implementation methods of the present application, but the protection scope of the present application is not limited thereto. Any technician familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be covered by the protection scope of the present application.

Claims

1. A method for measuring the strabismus of patients with rotational strabismus based on fundus photographs, characterized in that: The method comprises the following steps: S1, collecting fundus images and performing grayscale processing to obtain rotational oblique vision images; S2, according to the gray value difference between the eight neighboring pixels of each pixel in the rotational oblique image, obtain the neighborhood distribution sequence of each pixel; according to the gray value difference between the pixels before and after all adjacent pixels in the neighborhood distribution sequence of each pixel, obtain the fundus structure difference of each pixel; S3, according to the difference distribution between the fundus structure dissimilarity of each pixel and all its eight neighboring pixels, the number of similar structures in the neighborhood of each pixel is obtained; according to the number of similar structures in the neighborhood of each pixel, the adaptive smoothing parameter of each pixel in the filtering process is obtained; S4, filtering the rotational strabismus image according to the adaptive smoothing parameter of each pixel point to obtain a rotational strabismus measurement image, detecting the circle in the rotational strabismus measurement image to obtain the macula and the optic disc; and measuring the strabismus degree according to the positions of the macula and the optic disc.

2. The method for measuring the strabismus of a patient with rotational strabismus based on fundus photographs according to claim 1, characterized in that: The method for acquiring the neighborhood distribution sequence of each pixel point includes: Any pixel point in the rotational oblique image is recorded as the target point; the pixel point with the smallest grayscale value among the eight neighboring pixel points of the target point is obtained and recorded as the starting point; starting from the starting point, all eight neighboring pixel points are arranged in the sorting direction to obtain the neighborhood distribution sequence of the target point.

3. The method for measuring the strabismus of a patient with rotational strabismus based on fundus photographs according to claim 2, characterized in that: The method for obtaining the sorting direction includes: Respectively obtain two pixel points adjacent to the starting point in the eight neighborhood pixel points, and record them as two adjacent points of the starting point; obtain the difference between the grayscale values ​​of the starting point and each of its adjacent points; take the adjacent point with the smallest difference as the direction point, wherein if the differences between two adjacent points are the same, select any adjacent point as the direction point; if the direction point is located in the clockwise direction of the starting point, the sorting direction is clockwise; otherwise, the sorting direction is counterclockwise.

4. The method for measuring the strabismus of a patient with rotational strabismus based on fundus photographs according to claim 1, characterized in that: The method for obtaining the fundus structure difference of each pixel point includes: For the neighborhood distribution sequence of each pixel point, the fundus structure recognition degree between each pair of adjacent pixel points is obtained according to the gray value difference between all pixel points before and after each pair of adjacent pixel points in the neighborhood distribution sequence; In the neighborhood distribution sequence, the first and second pixels in a pair of adjacent pixels with the highest fundus structure recognition are recorded as the front structure discrimination point and the back structure discrimination point, respectively. According to the gray value difference between all the pixels before the front structure determination point and all the pixels after the back structure determination point, the fundus structure difference of each pixel is obtained.

5. The method for measuring the strabismus of a patient with rotational strabismus based on fundus photographs according to claim 4, characterized in that: The method for obtaining the fundus structure recognition degree between each pair of adjacent pixel points comprises: recording the fundus structure recognition degree between the i-th and i+1-th pixel points in the neighborhood distribution sequence as , ;in, , They represent the grayscale values ​​of the j-th and k-th pixels in the neighborhood distribution sequence respectively; i represents the sequence number of the previous pixel between the current two adjacent pixels in the neighborhood distribution sequence; n represents the number of pixels in the neighborhood distribution sequence; || is the absolute value symbol.

6. The method for measuring the strabismus of a patient with rotational strabismus based on fundus photographs according to claim 4, characterized in that: The method of obtaining the fundus structure difference of each pixel point according to the gray value difference between all the pixel points before the front structure determination point and all the pixel points after the back structure determination point comprises: The mean grayscale value of all pixels from the first pixel point to the front structure discrimination point in the neighborhood distribution sequence is obtained, which is recorded as the first structure feature value; the mean grayscale value of all pixels from the back structure discrimination point to the last pixel point in the neighborhood distribution sequence is obtained, which is recorded as the second structure feature value; the difference between the grayscale value of each pixel point and the first structure feature value and the second structure feature value is obtained respectively, which is recorded as the first difference and the second difference; the minimum value between the first difference and the second difference is taken as the fundus structure difference degree of each pixel point.

7. The method for measuring the strabismus of a patient with rotational strabismus based on fundus photographs according to claim 1, characterized in that: The method for obtaining the number of similar structures in the neighborhood of each pixel point includes: The segmentation threshold is obtained according to the difference between the fundus structure of all pixels and their eight neighboring pixels; The number of fundus structures between each pixel and all its eight neighboring pixels whose dissimilarity is less than the segmentation threshold is counted and recorded as the number of similar neighborhood structures of each pixel.

8. The method for measuring the strabismus of a patient with rotational strabismus based on fundus photographs according to claim 7, characterized in that: The method for obtaining the segmentation threshold comprises: Obtaining the difference in fundus structure difference between each pixel point and each of its eight neighboring pixel points as the fundus structure dissimilarity between each pixel point and each of its eight neighboring pixel points; The Otsu threshold analysis is performed on the fundus structure dissimilarity between all pixels and all their eight neighboring pixels to obtain the segmentation threshold.

9. The method for measuring the strabismus of a patient with rotational strabismus based on fundus photographs according to claim 1, characterized in that: The method of obtaining the adaptive smoothing parameter of each pixel point in the filtering process according to the number of similar structures in the neighborhood of each pixel point includes: recording the adaptive smoothing parameter of the ath pixel point in the rotational oblique image as , , where Indicates the preset initial smoothing parameter; Indicates the preset adjustment factor; Represents the number of similar structures in the neighborhood of the a-th pixel in the rotational oblique image; Represents the sigmoid normalization function.

10. The method for measuring the strabismus of a patient with rotational strabismus based on fundus photographs according to claim 1, characterized in that: The method for obtaining the macula and optic disc comprises: Substitute the adaptive smoothing parameters of each pixel in the rotational squint image into the NLM filtering algorithm to denoise the rotational squint image and obtain the rotational squint measurement image. The circles in the rotational strabismus measurement image are obtained; the inner area of ​​the circle with the second largest diameter is taken as the macula; and the inner area of ​​the circle with the smallest diameter is taken as the optic disc.

Images