Method for automatically measuring meibomian gland quality

Abstract

The invention relates to a method for automatically measuring the quality of meibomian glands. The method comprises the following steps: acquiring a group of continuous grayscale images including eyes and meibomian glands; determining the gray value change rate of each pixel according to the continuous gray images; determining a minimum enclosing rectangle of eyes in each image, forming a foreground area in the minimum enclosing rectangle, and forming a background area outside the minimum enclosing rectangle; according to the gray value change rate of each pixel in the background area, combining the foreground areas to obtain a combined image; preprocessing the merged image to obtain a processed image; wherein the preprocessing comprises median filtering processing, normalization processing and non-uniform illumination removal processing; based on the processed image, meibomian gland quality is measured. According to the method provided by the invention, automatic evaluation of the meibomian gland can be realized based on a group of continuous grayscale images including the eyes and the meibomian gland.

Description

Technical Field

[0001] The present invention relates to the technical field of medical image processing, and in particular to a method for automatically measuring the mass of meibomian glands. Background Art

[0002] MGD (Meibomian Gland Dysfunction) is one of the main causes of evaporative dry eye.

[0003] Current MGD diagnostic methods mainly rely on qualitative evaluation by ophthalmologists and lack automated and quantitative analysis methods.

[0004] Therefore, an automated method to assess the quality and function of the meibomian glands is urgently needed. Summary of the invention

[0005] 1. Technical issues to be resolved

[0006] In order to solve the above problems, the present invention provides a method for automatically measuring the mass of meibomian glands.

[0007] (II) Technical solution

[0008] In order to achieve the above object, the main technical solutions adopted by the present invention include:

[0009] A method for automatically measuring meibomian gland mass, the method comprising:

[0010] Acquire a set of continuous grayscale images including the eye and meibomian glands;

[0011] According to the continuous grayscale image, the grayscale value change rate of each pixel is determined;

[0012] Determine the minimum bounding rectangle of the eye in each image, and form the area within the minimum bounding rectangle into the foreground area, and the area outside the minimum bounding rectangle into the background area;

[0013] According to the gray value change rate of each pixel in the background area, the foreground area is merged to obtain a merged image;

[0014] Preprocessing the merged image to obtain a processed image; wherein the preprocessing includes: median filtering, normalization, and uneven illumination removal;

[0015] Based on the processed images, the meibomian gland mass was measured.

[0016] Optionally, determining the grayscale value change rate of each pixel according to the continuous grayscale image includes:

[0017] Determine the grayscale value of each pixel in each image Among them, x is the image identifier and i is the pixel identifier;

[0018] Determine the mean grayscale value of each pixel Where X is the total number of images;

[0019] Determine the initial change rate of the gray value of each pixel

[0020] Determine the initial rate of change mean Where, I is the total number of pixels;

[0021] Determine the overall rate of change

[0022] Determine the adjustment ratio w of each pixel i ;

[0023] Determine the gray value change rate CG of each pixel i (1) = CG i (0)×(1-w i ).

[0024] Optionally, determine the adjustment ratio w of each pixel i ,include:

[0025] Determine the adjustment degree of each pixel

[0026] like Then i =0;

[0027] like but

[0028] like but

[0029] Among them, max{·} is the maximum value function, and min{·} is the minimum value function.

[0030] Optionally, merging the foreground area according to the gray value change rate of each pixel in the background area to obtain a merged image includes:

[0031] Determine the mean value CG of the gray value change rate of all pixels in the background area avg (1) Maximum value CG max (1) and minimum CG min (1);

[0032] For each pixel in the foreground area, determine the merged grayscale value Among them, x is the image identifier, b is the pixel identifier of the foreground area, is the gray value of image x at pixel b, X is the total number of images, CG b (1) is the gray value change rate of pixel b, min{·} is the minimum value function, α is the preset maximum adjustment value, α∈[0,1];

[0033] According to the combined gray value GV′ of each pixel in the foreground area b Get the merged image.

[0034] Optionally, preprocessing the merged image to obtain a processed image includes:

[0035] Perform median filtering on the merged image based on the m×n kernel to obtain a median filtered image;

[0036] Normalize the median filtered image to obtain a normalized image; the gray value of any pixel c in the normalized image is Among them, GV′ c (MED) is the gray value of any pixel c in the median filtered image, max{·} is the maximum value function, and min{·} is the minimum value function;

[0037] Processing the normalized image with a low-pass filter to obtain a low-pass filtered image;

[0038] The difference between the normalized image and the low-pass filtered image is determined as the processed image.

[0039] Optionally, based on the processed images, the meibomian gland mass is measured, including:

[0040] According to the mask h R processing the processed image and detecting the meibomian gland area in the processed image;

[0041] Based on the meibomian gland area, the meibomian gland mass was measured.

[0042] Optionally, according to the mask h R Processing the processed image to detect the meibomian gland area in the processed image includes:

[0043] Determine the gray value GV of each pixel contained in the meibomian gland area d , where d is the pixel identifier in the meibomian gland area;

[0044] Determine the maximum threshold GV of each pixel d {max} and minimum threshold GV d {min};

[0045] Determine the grayscale processing value of each pixel Among them, GVmax is the maximum pixel threshold, GV min is the minimum pixel threshold, max{·} is the maximum value function, and min{·} is the minimum value function;

[0046] According to the mask h R The image formed by the grayscale processing values ​​is processed to detect the meibomian gland area.

[0047] Optionally,

[0048] Where C is the overall rate of change.

[0049] Optionally,

[0050] Where C is the overall rate of change.

[0051] Optionally,

[0052] Among them, u, v, q are the three dimensions of mask rotation, σ u ,σ v ,σ q is the average deviation of the mask in three dimensions, u R ,v R ,q R are the standardized values ​​of the three dimensions.

[0053] (III) Beneficial effects

[0054] The present invention relates to a method for automatically measuring the quality of meibomian glands, the method comprising: collecting a set of continuous grayscale images including eyes and meibomian glands; determining the grayscale value change rate of each pixel according to the continuous grayscale images; determining the minimum circumscribed rectangle of the eyes in each image, forming the area within the minimum circumscribed rectangle into a foreground area, and forming the area outside the minimum circumscribed rectangle into a background area; merging the foreground area according to the grayscale value change rate of each pixel in the background area to obtain a merged image; preprocessing the merged image to obtain a processed image; wherein the preprocessing comprises: median filtering, normalization and removal of uneven illumination; and measuring the quality of meibomian glands based on the processed image. The method provided by the present invention can realize automatic evaluation of meibomian glands based on a set of continuous grayscale images including eyes and meibomian glands. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] Figure 1 A schematic flow chart of a method for automatically measuring meibomian gland mass provided in one embodiment of the present application;

[0056] Figure 2 A schematic diagram of an image of a set of continuous grayscale images provided in an embodiment of the present application;

[0057] Figure 3 A schematic diagram of meibomian gland area detection provided in one embodiment of the present application. DETAILED DESCRIPTION

[0058] In order to better explain the present invention and facilitate understanding, the present invention is described in detail below through specific implementation modes in conjunction with the accompanying drawings.

[0059] Current MGD diagnostic methods mainly rely on qualitative evaluation by ophthalmologists and lack automated and quantitative analysis methods. Therefore, an automated method is urgently needed to evaluate the quality and function of the meibomian glands.

[0060] In order to solve this problem, the present invention relates to a method for automatically measuring the quality of meibomian glands, the method comprising: collecting a set of continuous grayscale images including eyes and meibomian glands; determining the grayscale value change rate of each pixel according to the continuous grayscale images; determining the minimum bounding rectangle of the eye in each image, forming the foreground area with the area within the minimum bounding rectangle, and forming the background area with the area outside the minimum bounding rectangle; merging the foreground area according to the grayscale value change rate of each pixel in the background area to obtain a merged image; preprocessing the merged image to obtain a processed image; wherein the preprocessing comprises: median filtering, normalization and removal of uneven illumination; based on the processed image, measuring the quality of meibomian glands. The method provided by the present invention can realize the automatic evaluation of meibomian glands based on a set of continuous grayscale images including eyes and meibomian glands.

[0061] See also Figure 1 The implementation process of the method for automatically measuring the mass of meibomian glands provided in this embodiment is as follows:

[0062] 101, collect a set of continuous grayscale images including eyes and meibomian glands.

[0063] For example, a high-precision, high-definition camera is used to capture a video of the user's eyes. During the capture process, the camera's position and parameters cannot change, and the user should try not to change the position. This ensures that the acquisition device and the acquisition object of a set of continuous grayscale images are the same, thereby ensuring the comparability of the subsequent processing of the method for automatically measuring the meibomian gland mass provided by this embodiment. For example, for different images, the content expressed by the same pixel position is the same. For example, an image of a set of continuous grayscale images captured is as follows: Figure 2 shown.

[0064] 102. Determine the grayscale value change rate of each pixel according to the continuous grayscale image.

[0065] The implementation process of this step is:

[0066] 102-1, determine the gray value of each pixel in each image

[0067] Among them, x is the image identifier and i is the pixel identifier.

[0068] For example, the first pixel in the upper left corner is identified as 1, and the pixel 1 is followed by a pixel to the right as pixel 2, and they are numbered in order from left to right and from top to bottom. In this way, the total number of pixels in each image and the positions of pixels with the same identification are the same.

[0069] In step 102-1, the grayscale value of pixel 1 in image 1 is determined. Gray value of pixel 1 in image 2 ..., the gray value of pixel 1 in image X Determine the grayscale value of pixel 2 in image 1, the grayscale value of pixel 2 in image 2, ..., the grayscale value of pixel 2 in image X. ….

[0070] 102-2, determine the grayscale mean of each pixel

[0071] Where X is the total number of images.

[0072] For example, for pixel 1, its grayscale mean

[0073] 102-3, determine the initial change rate of the gray value of each pixel

[0074] For example, for pixel 1, the initial change rate of its gray value is

[0076] Initial change rate CG i (0) represents the change of the gray value of a certain pixel i in each image. Since the position, parameters, and user position of the acquisition device are relatively fixed when acquiring images in step 101, the gray value of a certain pixel i in each image does not change in an ideal situation. That is, in an ideal situation, CG i (0) = 0, that is, CG i The closer (0) is to 0, the more accurate the set of continuous grayscale images acquired in step 101 is.

[0077] 102-4, Determine the initial change rate mean

[0078] Where I is the total number of pixels.

[0079] 102-5, Determine the overall rate of change

[0080] The overall change rate C represents the overall grayscale value change of pixels in a set of continuous grayscale images. The closer C is to 0, the more accurate the set of continuous grayscale images collected in step 101 is.

[0081] 102-6, determine the adjustment ratio w of each pixel i .

[0082] That is, the adjustment degree of each pixel is determined as according to Determine the adjustment ratio w i .

[0083] like Then i =0.

[0084] like but

[0085] like but

[0086] Among them, max{·} is the maximum value function, and min{·} is the minimum value function.

[0087] 102-7, determine the gray value change rate CG of each pixel i (1) = CG i (0)×(1-w i ).

[0088] Adjustment It represents the relationship between the change of the gray value of a certain pixel i in each image and the change of the gray value of the whole pixel in a set of continuous gray images. If the change of the gray value of the whole pixel in a set of continuous gray images is confirmed to be an inevitable difference caused by the error of the device and the user system, then the adjustment degree of a certain pixel i Characterizes the difference between the change of a certain pixel i and the system error, which needs to be processed (i.e. CG i (0)×w i ), so as to ensure the comparability of the grayscale values ​​of each pixel.

[0089] if This means that the change of a certain pixel i is no different from the system error. If the two are the same, there is no need to eliminate them.

[0090] if This means that the change of a certain pixel i is reduced relative to the system error, so it is regressed to the standard error by increasing the difference. The larger the difference, the more regression.

[0091] if This means that the change of a certain pixel i has increased relative to the system error, so it is regressed to the standard error by reducing the difference. The larger the difference, the more regression.

[0092] 103 , determining the minimum bounding rectangle of the eye in each image, forming the area within the minimum bounding rectangle into a foreground area, and forming the area outside the minimum bounding rectangle into a background area.

[0093] This step uses an existing image recognition solution to obtain the minimum bounding rectangle of the eye in each image. Since the position, parameters, and user position of the acquisition device are relatively fixed when the image is acquired in step 101, the minimum bounding rectangle in each image does not change much and can be considered to be unchanged.

[0094] 104 , merging the foreground area according to the gray value change rate of each pixel in the background area to obtain a merged image.

[0095] The merging process is:

[0096] 104-1, determine the average value CG of the gray value change rate of all pixels in the background area avg (1) Maximum value CG max (1) and minimum CG min (1).

[0097] 104-2, for each pixel in the foreground area, determine the combined grayscale value

[0098] Among them, x is the image identifier, b is the pixel identifier of the foreground area, is the gray value of image x at pixel b, X is the total number of images, CG b (1) is the gray value change rate of pixel b, min{·} is the minimum value function, α is the preset maximum adjustment value, α∈[0,1].

[0099] In step 104 - 2 , the image grayscale error caused by uncontrollable factors such as user movement is eliminated, further ensuring the accuracy of subsequent image processing.

[0100] The background area will not change due to human factors, so in step 104-2, the pixels of the foreground area will be adjusted based on the grayscale value of the background area.

[0101] It represents the average gray value of pixel b in the foreground area of ​​all images, which is a basic value. It is the adjustment value of the pixels in the foreground area based on the grayscale value of the background area.

[0102] 104-3, based on the combined gray value GV′ of each pixel in the foreground area b Get the merged image.

[0103] The merged image is actually an image containing only the minimum circumscribed rectangular area of ​​the eye, and the value of each pixel in the image is the merged and adjusted value. It is an image that can accurately reflect the user's eyes and meibomian glands.

[0104] 105, preprocessing the merged image to obtain a processed image.

[0105] Among them, the preprocessing includes: median filtering, normalization and removal of uneven illumination.

[0106] For example, this can be achieved through the following steps:

[0107] 105-1, performing median filtering on the merged image based on the m×n kernel to obtain a median filtered image.

[0108] 105-2, normalize the median filtered image to obtain a normalized image. The grayscale value of any pixel c in the normalized image is

[0110] Among them, GV′ c (MED) is the gray value of any pixel c in the median filtered image, max{·} is the maximum value function, and min{·} is the minimum value function.

[0111] 105-3, performing low-pass filter processing on the normalized image to obtain a low-pass filtered image.

[0112] 105-4, determining the difference between the normalized image and the low-pass filtered image as a processed image.

[0113] The low-pass filtered image characterizes the differences caused by uneven illumination. By determining the difference between the normalized image and the low-pass filtered image as the processed image, the influence of uneven illumination in the normalized image can be removed to obtain an image after the unevenness is removed, that is, the processed image, which can accurately measure the meibomian gland mass.

[0114] 106. Based on the processed images, the meibomian gland mass is measured.

[0115] The implementation process of step 106 is:

[0116] 106-1, according to mask h R The processed image is processed to detect the meibomian gland area in the processed image.

[0117] In order to further remove the influence of the meibomian gland edge area, the implementation process of step 106-1 is as follows:

[0118] 1. Determine the gray value GV of each pixel contained in the meibomian gland area d .

[0119] Where d is the pixel identifier in the meibomian gland area.

[0120] 2. Determine the maximum pixel threshold GV max and pixel minimum threshold GV min .

[0121] in,

[0122] Among them, C is the overall change rate, max{·} is the maximum value function, and min{·} is the minimum value function.

[0123] 3. Determine the grayscale processing value of each pixel

[0124] Grayscale processing value GV" of each pixel d is the gray value GV of each pixel d On this basis, the difference between grayscale values ​​is further increased.

[0125] 4. According to the mask h R The image formed by the grayscale processing values ​​is processed to detect the meibomian gland area.

[0126] Through the above processing, the difference between the grayscale values ​​of each pixel in the image formed by the grayscale processing value can be increased, thereby better increasing the grayscale value difference of the edge of the meibomian gland, more accurately identifying the edge of the meibomian gland, and obtaining the meibomian gland area formed by the edge of the meibomian gland.

[0127] in addition,

[0128] Among them, u, v, q are the three dimensions of mask rotation, σ u ,σ v ,σ q is the average deviation of the mask in three dimensions, u R ,v R ,q R are the standardized values ​​of the three dimensions.

[0129] For example,

[0130] Among them, U, V, and Q are the standard values ​​of the three dimensions of mask rotation.

[0131] By creating a mask h RSequence, the meibomian glands can be separated from the processed image. For example, the mask h R If the obtained value is greater than the preset threshold, the corresponding processed value is 1, otherwise it is 0, and the meibomian gland area is obtained based on the processed value.

[0132] For example Figure 3 It is shown in u =0.05,σ v =0.05,σ q = 0.4 and schematic diagram of the detection results of the meibomian gland area.

[0133] 106-2. Measure the meibomian gland mass according to the meibomian gland area.

[0134] 106-3, according to n M Meibomian gland mass was measured.

[0135] Among them, the quality of meibomian glands includes the proportion of meibomian glands, meibomian gland length, curvature, width, length, number of glands, etc.

[0136] For example, the percentage of meibomian glands is Among them, n M is the number of pixels contained in the meibomian gland area, and N is the number of pixels contained in the eyelid.

[0137] The curvature is the length of the meibomian gland / the distance between the two end points of the meibomian gland.

[0138] The present embodiment provides a method for automatically measuring the quality of meibomian glands, which includes collecting a set of continuous grayscale images including eyes and meibomian glands; determining the grayscale value change rate of each pixel based on the continuous grayscale images; determining the minimum bounding rectangle of the eye in each image, forming the foreground area with the area within the minimum bounding rectangle, and forming the background area with the area outside the minimum bounding rectangle; merging the foreground area based on the grayscale value change rate of each pixel in the background area to obtain a merged image; preprocessing the merged image to obtain a processed image; wherein the preprocessing includes: median filtering, normalization, and removal of uneven illumination; based on the processed image, measuring the quality of meibomian glands. The method provided in the present embodiment can realize automatic evaluation of meibomian glands based on a set of continuous grayscale images including eyes and meibomian glands.

[0139] It should be clear that the present invention is not limited to the specific configuration and processing described above and shown in the figures. For the sake of simplicity, a detailed description of the known method is omitted here. In the above embodiments, several specific steps are described and shown as examples. However, the method process of the present invention is not limited to the specific steps described and shown, and those skilled in the art can make various changes, modifications and additions, or change the order between the steps after understanding the spirit of the present invention.

[0140] It should also be noted that the exemplary embodiments mentioned in the present invention describe some methods or systems based on a series of steps or devices. However, the present invention is not limited to the order of the above steps, that is, the steps can be performed in the order mentioned in the embodiments, or in a different order from the embodiments, or several steps can be performed simultaneously.

[0141] Finally, it should be noted that the above-described embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the present invention. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some or all of the technical features thereof may be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for automatically measuring meibomian gland mass, characterized in that: The method comprises: Acquire a set of continuous grayscale images including the eye and meibomian glands; Determining the grayscale value change rate of each pixel according to the continuous grayscale image; Determine the minimum bounding rectangle of the eye in each image, and form the area within the minimum bounding rectangle into the foreground area, and the area outside the minimum bounding rectangle into the background area; According to the gray value change rate of each pixel in the background area, the foreground area is merged to obtain a merged image; Preprocessing the merged image to obtain a processed image; wherein the preprocessing includes: median filtering, normalization, and uneven illumination removal; Based on the processed images, the meibomian gland mass was measured.

2. The method according to claim 1, characterized in that Determining the grayscale value change rate of each pixel according to the continuous grayscale image includes: Determine the grayscale value of each pixel in each image Among them, x is the image identifier and i is the pixel identifier; Determine the mean grayscale value of each pixel Where X is the total number of images; Determine the initial change rate of the gray value of each pixel Determine the initial rate of change mean Where, I is the total number of pixels; Determine the overall rate of change Determine the adjustment ratio w of each pixel i ; Determine the gray value change rate CG of each pixel i (1) = CG i (0)×(1-w i ).

3. The method according to claim 2, characterized in that The adjustment ratio w of each pixel is determined i ,include: Determine the adjustment degree of each pixel like Then i =0; like but like but Among them, max{·} is the maximum value function, and min{·} is the minimum value function.

4. The method according to claim 1, characterized in that: The foreground area is merged according to the gray value change rate of each pixel in the background area to obtain a merged image, including: Determine the mean value CG of the gray value change rate of all pixels in the background area avg (1) Maximum value CG max (1) and minimum CG min (1); For each pixel in the foreground area, determine the merged grayscale value Among them, x is the image identifier, b is the pixel identifier of the foreground area, is the gray value of image x at pixel b, X is the total number of images, CG b (1) is the gray value change rate of pixel b, min{·} is the minimum value function, α is the preset maximum adjustment value, α∈[0,1]; According to the combined gray value GV′ of each pixel in the foreground area b Get the merged image.

5. The method according to claim 1, characterized in that The preprocessing of the merged image to obtain a processed image includes: Performing median filtering on the merged image based on an m×n kernel to obtain a median filtered image; Normalize the median filtered image to obtain a normalized image; the gray value of any pixel c in the normalized image Among them, GV′ c (MED) is the gray value of any pixel c in the median filtered image, max{·} is the maximum value function, and min{·} is the minimum value function; Processing the normalized image with a low-pass filter to obtain a low-pass filtered image; The difference between the normalized image and the low-pass filtered image is determined as a processed image.

6. The method according to claim 1, characterized in that The method of measuring the meibomian gland mass based on the processed image comprises: According to the mask h R Processing the processed image to detect a meibomian gland area in the processed image; Based on the meibomian gland area, the meibomian gland mass was measured.

7. The method according to claim 6, characterized in that According to the mask h R Processing the processed image to detect the meibomian gland area in the processed image includes: Determine the grayscale value GV of each pixel contained in the meibomian gland area d , where d is the pixel identifier in the meibomian gland area; Determine the maximum threshold GV of each pixel d {max} and minimum threshold GV d {min}; Determine the grayscale processing value of each pixel Among them, GV max is the maximum pixel threshold, GV min is the minimum pixel threshold, max{·} is the maximum value function, and min{·} is the minimum value function; According to the mask h R The image formed by the grayscale processed values ​​is processed to detect the meibomian gland area.

8. The method according to claim 7, characterized in that Where C is the overall rate of change.

9. The method according to claim 7, characterized in that: Where C is the overall rate of change.

10. The method according to claim 6 or 7, characterized in that: Among them, u, v, q are the three dimensions of mask rotation, σ u ,σ v ,σ q is the average deviation of the mask in three dimensions, u R ,v R ,q R are the standardized values ​​of the three dimensions.

Images