CN116777794B - Cornea foreign body image processing method and system - Google Patents
Cornea foreign body image processing method and system Download PDFInfo
- Publication number
- CN116777794B CN116777794B CN202311035548.5A CN202311035548A CN116777794B CN 116777794 B CN116777794 B CN 116777794B CN 202311035548 A CN202311035548 A CN 202311035548A CN 116777794 B CN116777794 B CN 116777794B
- Authority
- CN
- China
- Prior art keywords
- cornea
- pixel point
- image
- layer
- brightness transition
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 210000004087 cornea Anatomy 0.000 title claims abstract description 202
- 238000003672 processing method Methods 0.000 title abstract description 8
- 238000012545 processing Methods 0.000 claims abstract description 21
- 238000012014 optical coherence tomography Methods 0.000 claims abstract description 17
- 238000007781 pre-processing Methods 0.000 claims abstract description 8
- 230000007704 transition Effects 0.000 claims description 86
- 238000000034 method Methods 0.000 claims description 37
- 238000012549 training Methods 0.000 claims description 31
- 230000003044 adaptive effect Effects 0.000 claims description 15
- 238000009499 grossing Methods 0.000 claims description 15
- 238000003706 image smoothing Methods 0.000 claims description 12
- 238000004364 calculation method Methods 0.000 claims description 7
- 206010017012 Foreign body in eye Diseases 0.000 claims description 5
- 238000000605 extraction Methods 0.000 abstract description 3
- 230000009286 beneficial effect Effects 0.000 description 7
- 230000002159 abnormal effect Effects 0.000 description 6
- 230000008859 change Effects 0.000 description 6
- 230000008569 process Effects 0.000 description 6
- 238000012216 screening Methods 0.000 description 4
- 230000001629 suppression Effects 0.000 description 3
- 210000005252 bulbus oculi Anatomy 0.000 description 2
- 238000001514 detection method Methods 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 238000005070 sampling Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 1
- 238000001914 filtration Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
Abstract
The invention discloses a cornea foreign body image processing method and a cornea foreign body image processing system, which relate to the technical field of image processing, and the cornea foreign body image processing method comprises the following steps: collecting cornea images by using an optical coherence tomography scanner, and preprocessing the cornea images to generate smoothed cornea images; extracting a cornea vertex set of the smoothed cornea image; a cornea bright spot area is generated according to the cornea vertex set, and the cornea bright spot area is used as a cornea foreign body existence area. The cornea foreign matter image processing method has the advantages that the cornea images are preprocessed, so that the image definition is improved, noise interference is inhibited, and the cornea vertex extraction and the bright spot area generation are convenient for the subsequent steps; meanwhile, the cornea vertex is accurately extracted through the processing of the brightness parameters, and then the cornea vertex is input into the constructed bright spot area generating network, so that the foreign body area, the foreign body size and the incarceration depth can be quickly locked, and the reference value is provided for medical staff to pre-judge.
Description
Technical Field
The invention relates to the technical field of image processing, in particular to a cornea foreign body image processing method and system.
Background
The cornea is located at the very front of the eyeball and plays an extremely important role in the diopter system of the eyeball. After obtaining the cornea image by the optical coherence tomography, a specific object (such as a foreign object) in the cornea image needs to be identified. It is therefore a current challenge to extract the location of a particular object of a certain class from a cornea image. Conventional cornea image processing generally adopts methods such as convolution and adaptive speckle suppression filtering. The methods have a certain effect on reducing speckle noise of the cornea image, but often cannot give ideal processing results of the optical coherence tomography image, and influence the final foreign body recognition result.
Disclosure of Invention
The invention aims to provide a cornea foreign body image processing method and system.
The embodiment of the invention is realized by the following technical scheme: a method for processing cornea foreign matter image includes the following steps:
collecting cornea images by using an optical coherence tomography scanner, and preprocessing the cornea images to generate smoothed cornea images;
extracting a cornea vertex set of the smoothed cornea image;
a cornea bright spot area is generated according to the cornea vertex set, and the cornea bright spot area is used as a cornea foreign body existence area.
Further, the cornea image is acquired by an optical coherence tomography scanner and preprocessed to generate a smoothed cornea image, comprising the steps of:
the cornea image is acquired by utilizing an optical coherence tomography scanner, and each pixel point of the cornea image is respectively extracted in the cornea imageREdge strength of the channel, inGEdge strength of the channel and the channelBEdge intensity of the channel, generating an edge intensity set;
randomly dividing the edge intensity set into a standard edge intensity subset and a training edge intensity subset;
calculating the smooth edge weight of the image according to the training edge intensity subset;
in the standard edge intensity subset, taking all pixel points corresponding to the edge intensity smaller than the image smoothing edge weight as a pixel point set to be smoothed;
and carrying out smoothing treatment on the pixel point set to be smoothed by using the adaptive window to generate a smoothed cornea image.
The beneficial effects of the above-mentioned further scheme are: in the invention, because the power supply, the detection circuit and the like of the optical coherence tomography scanner can generate noise influence on the acquired cornea image, the acquired cornea image is required to be subjected to smoothing treatment, the brightness of the image can be gradually changed smoothly by utilizing the self-adaptive window, the abrupt gradient is reduced, and the image quality is improved. Before smoothing, the cornea image is divided into a standard edge intensity subset and a training edge intensity subset, and the pixel points to be smoothed in the standard edge intensity subset are determined by using the smoothing edge weight of the training edge intensity subset, so that the pixel points to be smoothed have generalization.
Further, the image smoothing edge weightscThe calculation formula of (2) is as follows:
in the method, in the process of the invention,M R1 representing training edge intensity subsetsRThe maximum edge strength of the channel is set,M G1 representing training edge intensity subsetsGThe maximum edge strength of the channel is set,M B1 representing training edge intensity subsetsBThe maximum edge strength of the channel is set,M R0 representing training edge intensity subsetsRThe minimum edge strength of the channel is set,M G0 representing training edge intensity subsetsGThe minimum edge strength of the channel is set,M B0 representing training edge intensity subsetsBThe minimum edge strength of the channel is set,exp(. Cndot.) means the operation of an exponent,σrepresenting the standard deviation of all edge intensities in the training edge intensity subset.
The beneficial effects of the above-mentioned further scheme are: in the invention, the smooth edge weight of the image is obtained by carrying out mathematical operation on the maximum edge intensity and the minimum edge intensity of each channel, and the standard deviation of all the edge intensities is also used as one of parameters affecting the smooth edge weight, so that the accuracy of the smooth edge weight of the image can be ensured, and meanwhile, the accuracy of pixel points to be smoothed obtained by carrying out size screening on the edge intensities in the subsequent step and the standard edge intensity subset is also ensured.
Further, the specific method for smoothing the pixel point set to be smoothed by the adaptive window comprises the following steps: and taking the pixel point with the minimum gray value in the pixel point set to be smoothed as a smoothed pixel point, taking the sum of the gradient of the smoothed pixel point along 0 degree, the gradient along 45 degrees, the gradient along 90 degrees and the gradient along 145 degrees as the length of the adaptive window, taking the average value of the gradient of the smoothed pixel point along 0 degree, the gradient along 45 degrees, the gradient along 90 degrees and the gradient along 145 degrees as the width of the adaptive window, and carrying out smoothing treatment from left to right by utilizing the adaptive window.
The beneficial effects of the above-mentioned further scheme are: in the invention, the gradient of the pixel point can reflect the change speed of the gray value of the pixel point, so that the sum of the gradients of the pixel point with the minimum gray value along four directions and the average value of the gradients along the four directions are respectively used as the length and the width of the self-adaptive window, the size of the self-adaptive window can be more adapted to the cornea image, and the pixel point at the edge of the cornea image can be ensured to be smoothed.
Further, extracting a corneal vertex set of the smoothed cornea image, comprising the sub-steps of:
taking a pixel point where a centroid in the smoothed cornea image is positioned as a central pixel point, and calculating brightness transition coefficients between other pixel points and the central pixel point;
determining a brightness transition region according to brightness transition coefficients between the rest pixel points and the central pixel point;
extracting four pixel points with the maximum brightness transition coefficient except a brightness transition region from the smoothed cornea image, wherein the four pixel points are a first pixel point, a second pixel point, a third pixel point and a fourth pixel point respectively;
and taking the intersection point of the connecting line between the first pixel point and the center of the brightness transition region and the brightness transition region as a first cornea vertex, taking the intersection point of the connecting line between the second pixel point and the center of the brightness transition region and the brightness transition region as a second cornea vertex, taking the intersection point of the connecting line between the third pixel point and the center of the brightness transition region and the brightness transition region as a third cornea vertex, and taking the intersection point of the connecting line between the fourth pixel point and the center of the brightness transition region and the brightness transition region as a fourth cornea vertex to generate a cornea vertex set.
The beneficial effects of the above-mentioned further scheme are: in the invention, the brightness transition coefficient is obtained by the difference value operation of the pixel point and the central pixel point, so the brightness transition coefficient can represent the brightness change between the pixel point and the central pixel point, and the area with larger brightness change is used as the brightness transition area, namely the area where the bright spots possibly exist. In the non-brightness transition region, the larger the brightness transition coefficient is, the brightness around the pixel point may be abnormal, and the boundary of the brightness transition region may also have the pixel point with abnormal brightness, and the accurate brightness abnormal point can be determined as the cornea vertex by connecting the intersection point of the pixel point and the center of the brightness transition region.
Further, the brightness transition coefficient between the pixel point and the central pixel pointH k The calculation formula of (2) is as follows:
in the method, in the process of the invention,Kthe number of pixels representing a smoothed cornea image,H k represent the firstkThe brightness of the individual pixel points is determined,H 0 representing the brightness of the center pixel.
Further, the method for determining the brightness transition region specifically comprises the following steps: and drawing a circle by taking the pixel point with the smallest brightness transition coefficient of the rest pixel points except the central pixel point as a circle center and taking the Euclidean distance between the pixel point with the smallest brightness transition coefficient and the central pixel point as a radius, and taking the circle as a brightness transition region.
Further, the method for generating the cornea bright spot area specifically comprises the following steps: constructing a bright spot area generating network, inputting a cornea vertex set into the bright spot area generating network, and generating a cornea bright spot area;
the system comprises a bright spot area generating network, a first convolution layer, a second convolution layer, a full connection layer and an output layer, wherein the bright spot area generating network comprises an input layer, a first convolution layer, a second convolution layer, a full connection layer and an output layer;
the input layer is used as an input end of the bright spot area generating network; the first output end of the input layer is connected with the input end of the first convolution layer; the second output end of the input layer is connected with the input end of the second convolution layer; the output end of the first convolution layer is connected with the first input end of the full connection layer; the output end of the second convolution layer is connected with the second input end of the full connection layer; the first output end of the full-connection layer is connected with the first input end of the output layer; the second output end of the full-connection layer is connected with the second input end of the output layer; the output layer serves as the output end of the bright spot area generating network.
The beneficial effects of the above-mentioned further scheme are: in the invention, an input layer is used for inputting a cornea vertex set and a smoothed cornea image into a bright spot area generating network and dividing the smoothed cornea image into a first image block and a second image block; the first convolution layer and the second convolution layer are respectively used for sampling the first image block and the second image block (for example, the image block with the size of 16×16×3 is sampled to obtain a 64×16×16 image block); the full-connection layer determines classification scores of four cornea vertexes in two image blocks by using a classifier, and a mixed non-maximum suppression algorithm is used for screening the smoothed cornea image to obtain a cornea bright spot area.
Further, the loss function of the full connection layerLossThe expression is:
in the method, in the process of the invention,F n representing the first convolution layernThe weight of the individual channels is determined,Nrepresenting the total number of channels of the first convolutional layer,G l representing the first convolution layerlThe weight of the individual channels is determined,Lrepresenting the total number of channels of the second convolutional layer,A 1 representing output via a first convolutional layerIs used for the image length of the (c) image,B 1 representing the width of the image output via the first convolution layer,A 2 representing the length of the image output via the second convolution layer,B 2 representing the width of the image output via the second convolution layer.
The technical scheme of the embodiment of the invention has at least the following advantages and beneficial effects: according to the cornea foreign body image processing method, the cornea image is preprocessed, so that the image definition is improved, noise interference is suppressed, and the cornea vertex extraction and the bright spot area generation are conveniently carried out in the subsequent steps; meanwhile, the cornea vertex is accurately extracted through the processing of the brightness parameters, and then the cornea vertex is input into the constructed bright spot area generating network, so that the foreign object area can be quickly locked, and the reference value is provided for medical staff.
Based on the system, the invention also provides a cornea foreign body image processing system, which comprises a cornea image smoothing unit, a cornea vertex generating unit and a cornea bright spot generating unit;
the cornea image smoothing unit is used for acquiring cornea images by using an optical coherence tomography scanner and preprocessing the cornea images to generate smoothed cornea images;
the cornea vertex generating unit is used for extracting a cornea vertex set of the smoothed cornea image;
the cornea bright spot generating unit is used for generating a cornea bright spot area according to the cornea vertex set and taking the cornea bright spot area as a cornea foreign body existence area.
The technical scheme of the embodiment of the invention has at least the following advantages and beneficial effects: the cornea foreign matter image processing system can accurately judge the area where the foreign matter is located through a series of processing of the cornea images.
Drawings
FIG. 1 is a flowchart of a method for processing a corneal foreign body image according to an embodiment of the present invention;
FIG. 2 is a block diagram of a system for processing an image of a foreign object on the cornea according to an embodiment of the present invention;
fig. 3 is a block diagram of a hot spot area generating network according to an embodiment of the present invention.
Detailed Description
For the purpose of making the objects, technical solutions and advantages of the embodiments of the present invention more apparent, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention, and it is apparent that the described embodiments are some embodiments of the present invention, but not all embodiments of the present invention. The components of the embodiments of the present invention generally described and illustrated in the figures herein may be arranged and designed in a wide variety of different configurations.
As shown in fig. 1, the present invention provides a method for processing a cornea foreign body image, comprising the steps of:
collecting cornea images by using an optical coherence tomography scanner, and preprocessing the cornea images to generate smoothed cornea images;
extracting a cornea vertex set of the smoothed cornea image;
a cornea bright spot area is generated according to the cornea vertex set, and the cornea bright spot area is used as a cornea foreign body existence area.
In the embodiment of the invention, after locking the foreign body area, the medical staff can further determine the depth of the foreign body according to the cornea foreign body area.
In an embodiment of the present invention, a cornea image is acquired by an optical coherence tomography scanner and preprocessed to generate a smoothed cornea image, comprising the steps of:
the cornea image is acquired by utilizing an optical coherence tomography scanner, and each pixel point of the cornea image is respectively extracted in the cornea imageREdge strength of the channel, inGEdge strength of the channel and the channelBEdge intensity of the channel, generating an edge intensity set;
randomly dividing the edge intensity set into a standard edge intensity subset and a training edge intensity subset;
calculating the smooth edge weight of the image according to the training edge intensity subset;
in the standard edge intensity subset, taking all pixel points corresponding to the edge intensity smaller than the image smoothing edge weight as a pixel point set to be smoothed;
and carrying out smoothing treatment on the pixel point set to be smoothed by using the adaptive window to generate a smoothed cornea image.
In the invention, because the power supply, the detection circuit and the like of the optical coherence tomography scanner can generate noise influence on the acquired cornea image, the acquired cornea image is required to be subjected to smoothing treatment, the brightness of the image can be gradually changed smoothly by utilizing the self-adaptive window, the abrupt gradient is reduced, and the image quality is improved. Before smoothing, the cornea image is divided into a standard edge intensity subset and a training edge intensity subset, and the pixel points to be smoothed in the standard edge intensity subset are determined by using the smoothing edge weight of the training edge intensity subset, so that the pixel points to be smoothed have generalization.
In an embodiment of the invention, the image smoothing edge weightscThe calculation formula of (2) is as follows:
in the method, in the process of the invention,M R1 representing training edge intensity subsetsRThe maximum edge strength of the channel is set,M G1 representing training edge intensity subsetsGThe maximum edge strength of the channel is set,M B1 representing training edge intensity subsetsBThe maximum edge strength of the channel is set,M R0 representing training edge intensity subsetsRThe minimum edge strength of the channel is set,M G0 representing training edge intensity subsetsGThe minimum edge strength of the channel is set,M B0 representing training edge intensity subsetsBThe minimum edge strength of the channel is set,exp(. Cndot.) means the operation of an exponent,σrepresenting the standard deviation of all edge intensities in the training edge intensity subset.
In the invention, the smooth edge weight of the image is obtained by carrying out mathematical operation on the maximum edge intensity and the minimum edge intensity of each channel, and the standard deviation of all the edge intensities is also used as one of parameters affecting the smooth edge weight, so that the accuracy of the smooth edge weight of the image can be ensured, and meanwhile, the accuracy of pixel points to be smoothed obtained by carrying out size screening on the edge intensities in the subsequent step and the standard edge intensity subset is also ensured.
In the embodiment of the invention, the specific method for smoothing the pixel point set to be smoothed by the adaptive window comprises the following steps: and taking the pixel point with the minimum gray value in the pixel point set to be smoothed as a smoothed pixel point, taking the sum of the gradient of the smoothed pixel point along 0 degree, the gradient along 45 degrees, the gradient along 90 degrees and the gradient along 145 degrees as the length of the adaptive window, taking the average value of the gradient of the smoothed pixel point along 0 degree, the gradient along 45 degrees, the gradient along 90 degrees and the gradient along 145 degrees as the width of the adaptive window, and carrying out smoothing treatment from left to right by utilizing the adaptive window.
In the invention, the gradient of the pixel point can reflect the change speed of the gray value of the pixel point, so that the sum of the gradients of the pixel point with the minimum gray value along four directions and the average value of the gradients along the four directions are respectively used as the length and the width of the self-adaptive window, the size of the self-adaptive window can be more adapted to the cornea image, and the pixel point at the edge of the cornea image can be ensured to be smoothed.
In an embodiment of the invention, the extraction of a corneal vertex set of a smoothed cornea image comprises the following sub-steps:
taking a pixel point where a centroid in the smoothed cornea image is positioned as a central pixel point, and calculating brightness transition coefficients between other pixel points and the central pixel point;
determining a brightness transition region according to brightness transition coefficients between the rest pixel points and the central pixel point;
extracting four pixel points with the maximum brightness transition coefficient except a brightness transition region from the smoothed cornea image, wherein the four pixel points are a first pixel point, a second pixel point, a third pixel point and a fourth pixel point respectively;
and taking the intersection point of the connecting line between the first pixel point and the center of the brightness transition region and the brightness transition region as a first cornea vertex, taking the intersection point of the connecting line between the second pixel point and the center of the brightness transition region and the brightness transition region as a second cornea vertex, taking the intersection point of the connecting line between the third pixel point and the center of the brightness transition region and the brightness transition region as a third cornea vertex, and taking the intersection point of the connecting line between the fourth pixel point and the center of the brightness transition region and the brightness transition region as a fourth cornea vertex to generate a cornea vertex set.
In the embodiment of the invention, the brightness transition coefficient is obtained by calculating the difference value between the pixel point and the central pixel point, so the brightness transition coefficient can represent the brightness change between the pixel point and the central pixel point, and the area with larger brightness change is used as the brightness transition area, namely the area where the bright spots possibly exist. In the non-brightness transition region, the larger the brightness transition coefficient is, the brightness around the pixel point may be abnormal, and the boundary of the brightness transition region may also have the pixel point with abnormal brightness, and the accurate brightness abnormal point can be determined as the cornea vertex by connecting the intersection point of the pixel point and the center of the brightness transition region.
In the embodiment of the invention, the brightness transition coefficient between the pixel point and the central pixel pointH k The calculation formula of (2) is as follows:
in the method, in the process of the invention,Kthe number of pixels representing a smoothed cornea image,H k represent the firstkThe brightness of the individual pixel points is determined,H 0 representing the brightness of the center pixel.
In the embodiment of the invention, the method for determining the brightness transition region specifically comprises the following steps: and drawing a circle by taking the pixel point with the smallest brightness transition coefficient of the rest pixel points except the central pixel point as a circle center and taking the Euclidean distance between the pixel point with the smallest brightness transition coefficient and the central pixel point as a radius, and taking the circle as a brightness transition region.
In the embodiment of the invention, the generation method of the cornea bright spot area specifically comprises the following steps: constructing a bright spot area generating network, inputting a cornea vertex set into the bright spot area generating network, and generating a cornea bright spot area;
the speckle region generating network comprises an input layer, a first convolution layer, a second convolution layer, a full connection layer and an output layer as shown in fig. 3;
the input layer is used as an input end of the bright spot area generating network; the first output end of the input layer is connected with the input end of the first convolution layer; the second output end of the input layer is connected with the input end of the second convolution layer; the output end of the first convolution layer is connected with the first input end of the full connection layer; the output end of the second convolution layer is connected with the second input end of the full connection layer; the first output end of the full-connection layer is connected with the first input end of the output layer; the second output end of the full-connection layer is connected with the second input end of the output layer; the output layer serves as the output end of the bright spot area generating network.
In the invention, an input layer is used for inputting a cornea vertex set and a smoothed cornea image into a bright spot area generating network and dividing the smoothed cornea image into a first image block and a second image block; the first convolution layer and the second convolution layer are respectively used for sampling the first image block and the second image block (for example, the image block with the size of 16×16×3 is sampled to obtain a 64×16×16 image block); the full-connection layer determines classification scores of four cornea vertexes in two image blocks by using a classifier, and a mixed non-maximum suppression algorithm is used for screening the smoothed cornea image to obtain a cornea bright spot area.
In the embodiment of the invention, the loss function of the full connection layerLossThe expression is:
in the method, in the process of the invention,F n representing the first convolution layernThe weight of the individual channels is determined,Nrepresenting the total number of channels of the first convolutional layer,G l representing the first convolution layerlThe weight of the individual channels is determined,Lrepresenting the total number of channels of the second convolutional layer,A 1 representing the length of the image output via the first convolution layer,B 1 representing the width of the image output via the first convolution layer,A 2 representing the length of the image output via the second convolution layer,B 2 representing the width of the image output via the second convolution layer.
Based on the above method, the invention also provides a system for processing the cornea foreign body image, which comprises a cornea image smoothing unit, a cornea vertex generating unit and a cornea bright spot generating unit as shown in figure 2;
the cornea image smoothing unit is used for acquiring cornea images by using an optical coherence tomography scanner and preprocessing the cornea images to generate smoothed cornea images;
the cornea vertex generating unit is used for extracting a cornea vertex set of the smoothed cornea image;
the cornea bright spot generating unit is used for generating a cornea bright spot area according to the cornea vertex set and taking the cornea bright spot area as a cornea foreign body existence area.
The above is only a preferred embodiment of the present invention, and is not intended to limit the present invention, but various modifications and variations can be made to the present invention by those skilled in the art. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.
Claims (6)
1. A method for processing a cornea foreign matter image, comprising the steps of:
collecting cornea images by using an optical coherence tomography scanner, and preprocessing the cornea images to generate smoothed cornea images;
extracting a cornea vertex set of the smoothed cornea image; the method specifically comprises the following substeps:
taking a pixel point where a centroid in the smoothed cornea image is positioned as a central pixel point, and calculating brightness transition coefficients between other pixel points and the central pixel point; determining a brightness transition region according to brightness transition coefficients between the rest pixel points and the central pixel point; extracting four pixel points with the maximum brightness transition coefficient except a brightness transition region from the smoothed cornea image, wherein the four pixel points are a first pixel point, a second pixel point, a third pixel point and a fourth pixel point respectively; taking the intersection point of the connecting line between the first pixel point and the center of the brightness transition area and the brightness transition area as a first cornea vertex, taking the intersection point of the connecting line between the second pixel point and the center of the brightness transition area and the brightness transition area as a second cornea vertex, taking the intersection point of the connecting line between the third pixel point and the center of the brightness transition area and the brightness transition area as a third cornea vertex, and taking the intersection point of the connecting line between the fourth pixel point and the center of the brightness transition area and the brightness transition area as a fourth cornea vertex to generate a cornea vertex set;
wherein the brightness transition coefficient H between the pixel point and the central pixel point k The calculation formula of (2) is as follows:
wherein K represents the number of pixels of the smoothed cornea image, H k Represents the brightness of the kth pixel, H 0 Representing the brightness of the center pixel;
the method for determining the brightness transition region specifically comprises the following steps: taking the pixel point with the smallest brightness transition coefficient of the rest pixel points except the central pixel point as a circle center, taking the Euclidean distance between the pixel point with the smallest brightness transition coefficient and the central pixel point as a radius, and drawing a circle to be used as a brightness transition region;
generating a cornea bright spot area according to the cornea vertex set, and taking the cornea bright spot area as a cornea foreign body existence area, wherein the cornea bright spot area generation method specifically comprises the following steps: constructing a bright spot area generating network, inputting a cornea vertex set into the bright spot area generating network, and generating a cornea bright spot area; the system comprises a bright spot area generating network, a first convolution layer, a second convolution layer, a full connection layer and an output layer, wherein the bright spot area generating network comprises an input layer, a first convolution layer, a second convolution layer, a full connection layer and an output layer; the input layer is used as an input end of a bright spot area generating network; the first output end of the input layer is connected with the input end of the first convolution layer; the second output end of the input layer is connected with the input end of the second convolution layer; the output end of the first convolution layer is connected with the first input end of the full-connection layer; the output end of the second convolution layer is connected with the second input end of the full-connection layer; the first output end of the full-connection layer is connected with the first input end of the output layer; the second output end of the full-connection layer is connected with the second input end of the output layer; the output layer serves as an output end of the bright spot area generating network.
2. The method for processing a corneal foreign body image according to claim 1, wherein: the method for generating a smoothed cornea image by acquiring a cornea image by using an optical coherence tomography scanner and preprocessing the cornea image comprises the following steps:
collecting cornea images by using an optical coherence tomography, respectively extracting the edge intensity of each pixel point of the cornea images in an R channel, the edge intensity of each pixel point in a G channel and the edge intensity of each pixel point in a B channel, and generating an edge intensity set;
randomly dividing the edge intensity set into a standard edge intensity subset and a training edge intensity subset;
calculating the smooth edge weight of the image according to the training edge intensity subset;
in the standard edge intensity subset, taking all pixel points corresponding to the edge intensity smaller than the image smoothing edge weight as a pixel point set to be smoothed;
and carrying out smoothing treatment on the pixel point set to be smoothed by using the adaptive window to generate a smoothed cornea image.
3. The method for processing a corneal foreign body image according to claim 2, wherein: the calculation formula of the image smoothing edge weight c is as follows:
wherein M is R1 Representing the maximum edge intensity of the R channel in the training edge intensity subset, M G1 Representing the maximum edge intensity of the G-channel in the training edge intensity subset, M B1 Representing the maximum edge intensity of the B channel in the training edge intensity subset, M R0 Representing the minimum edge intensity of the R channel in the training edge intensity subset, M G0 Representing the minimum edge intensity of the G-channel in the training edge intensity subset,M B0 representing the minimum edge intensities of the B-channels in the training edge intensity subset, exp (·) represents the exponential operation, σ represents the standard deviation of all edge intensities in the training edge intensity subset.
4. The method for processing a corneal foreign body image according to claim 2, wherein: the specific method for smoothing the pixel point set to be smoothed by the self-adaptive window comprises the following steps: and taking the pixel point with the minimum gray value in the pixel point set to be smoothed as a smoothed pixel point, taking the sum of the gradient of the smoothed pixel point along 0 degree, the gradient along 45 degrees, the gradient along 90 degrees and the gradient along 145 degrees as the length of the adaptive window, taking the average value of the gradient of the smoothed pixel point along 0 degree, the gradient along 45 degrees, the gradient along 90 degrees and the gradient along 145 degrees as the width of the adaptive window, and carrying out smoothing treatment from left to right by utilizing the adaptive window.
5. The method for processing a corneal foreign body image according to claim 1, wherein: the Loss function Loss expression of the full connection layer is:
wherein F is n Represents the weight of the nth channel in the first convolution layer, N represents the total number of channels of the first convolution layer, G l Representing the weight of the first channel in the second convolution layer, L representing the total number of channels of the second convolution layer, A 1 Representing the length of the image output via the first convolution layer, B 1 Representing the width of the image output via the first convolution layer, A 2 Representing the length of the image output via the second convolution layer, B 2 Representing the width of the image output via the second convolution layer.
6. A processing system of cornea foreign body image is characterized by comprising a cornea image smoothing unit, a cornea vertex generating unit and a cornea bright spot generating unit;
the cornea image smoothing unit is used for acquiring cornea images by using an optical coherence tomography scanner and preprocessing the cornea images to generate smoothed cornea images;
the cornea vertex generating unit is used for extracting a cornea vertex set of the smoothed cornea image; the method comprises the following steps:
taking a pixel point where a centroid in the smoothed cornea image is positioned as a central pixel point, and calculating brightness transition coefficients between other pixel points and the central pixel point; determining a brightness transition region according to brightness transition coefficients between the rest pixel points and the central pixel point; extracting four pixel points with the maximum brightness transition coefficient except a brightness transition region from the smoothed cornea image, wherein the four pixel points are a first pixel point, a second pixel point, a third pixel point and a fourth pixel point respectively; taking the intersection point of the connecting line between the first pixel point and the center of the brightness transition area and the brightness transition area as a first cornea vertex, taking the intersection point of the connecting line between the second pixel point and the center of the brightness transition area and the brightness transition area as a second cornea vertex, taking the intersection point of the connecting line between the third pixel point and the center of the brightness transition area and the brightness transition area as a third cornea vertex, and taking the intersection point of the connecting line between the fourth pixel point and the center of the brightness transition area and the brightness transition area as a fourth cornea vertex to generate a cornea vertex set;
wherein the brightness transition coefficient H between the pixel point and the central pixel point k The calculation formula of (2) is as follows:
wherein K represents the number of pixels of the smoothed cornea image, H k Represents the brightness of the kth pixel, H 0 Representing the brightness of the center pixel;
the method for determining the brightness transition region specifically comprises the following steps: taking the pixel point with the smallest brightness transition coefficient of the rest pixel points except the central pixel point as a circle center, taking the Euclidean distance between the pixel point with the smallest brightness transition coefficient and the central pixel point as a radius, and drawing a circle to be used as a brightness transition region;
the cornea bright spot generating unit is used for generating a cornea bright spot area according to the cornea vertex set and taking the cornea bright spot area as a cornea foreign body existence area, wherein the cornea bright spot area generating method specifically comprises the following steps: constructing a bright spot area generating network, inputting a cornea vertex set into the bright spot area generating network, and generating a cornea bright spot area; the system comprises a bright spot area generating network, a first convolution layer, a second convolution layer, a full connection layer and an output layer, wherein the bright spot area generating network comprises an input layer, a first convolution layer, a second convolution layer, a full connection layer and an output layer; the input layer is used as an input end of a bright spot area generating network; the first output end of the input layer is connected with the input end of the first convolution layer; the second output end of the input layer is connected with the input end of the second convolution layer; the output end of the first convolution layer is connected with the first input end of the full-connection layer; the output end of the second convolution layer is connected with the second input end of the full-connection layer; the first output end of the full-connection layer is connected with the first input end of the output layer; the second output end of the full-connection layer is connected with the second input end of the output layer; the output layer serves as an output end of the bright spot area generating network.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311035548.5A CN116777794B (en) | 2023-08-17 | 2023-08-17 | Cornea foreign body image processing method and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311035548.5A CN116777794B (en) | 2023-08-17 | 2023-08-17 | Cornea foreign body image processing method and system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN116777794A CN116777794A (en) | 2023-09-19 |
| CN116777794B true CN116777794B (en) | 2023-11-03 |
Family
ID=88011840
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311035548.5A Active CN116777794B (en) | 2023-08-17 | 2023-08-17 | Cornea foreign body image processing method and system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116777794B (en) |
Citations (16)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102567737A (en) * | 2011-12-28 | 2012-07-11 | 华南理工大学 | Method for locating eyeball cornea |
| CN103996020A (en) * | 2014-04-10 | 2014-08-20 | 中航华东光电（上海）有限公司 | Head mounted eye tracker detection method |
| CN110010219A (en) * | 2019-03-13 | 2019-07-12 | 杭州电子科技大学 | Optical coherence tomography image retinopathy intelligent checking system and detection method |
| CN110349162A (en) * | 2019-07-17 | 2019-10-18 | 苏州大学 | A kind of more lesion image partition methods of macular edema |
| CN110363782A (en) * | 2019-06-13 | 2019-10-22 | 平安科技（深圳）有限公司 | A kind of area recognizing method based on limb recognition algorithm, device and electronic equipment |
| CN110766656A (en) * | 2019-09-19 | 2020-02-07 | 平安科技（深圳）有限公司 | Method, device, equipment and storage medium for screening abnormality of eyeground macular region |
| WO2020056454A1 (en) * | 2018-09-18 | 2020-03-26 | MacuJect Pty Ltd | A method and system for analysing images of a retina |
| CN111093525A (en) * | 2018-08-07 | 2020-05-01 | 温州医科大学 | Optical coherence tomography image processing method |
| CN111462156A (en) * | 2020-03-30 | 2020-07-28 | 温州医科大学 | Image processing method for acquiring corneal vertex |
| WO2020234640A1 (en) * | 2019-05-20 | 2020-11-26 | Macuject Pty Ltd. | Confidence-based methods and systems for analyzing images of a retina |
| CN112308829A (en) * | 2020-10-27 | 2021-02-02 | 苏州大学 | Self-adaptive network suitable for high-reflection bright spot segmentation in retina optical coherence tomography image |
| CN112669260A (en) * | 2020-12-07 | 2021-04-16 | 上海交通大学 | Method and device for detecting eye fundus image optic disc yellow spots based on deep neural network |
| CN113940812A (en) * | 2021-11-01 | 2022-01-18 | 朴俊杰 | Cornea center positioning method for excimer laser cornea refractive surgery |
| CN116109555A (en) * | 2022-11-16 | 2023-05-12 | 南京博视医疗科技有限公司 | Wavefront facula lattice optimization method and device |
| CN116342636A (en) * | 2023-05-23 | 2023-06-27 | 广东麦特维逊医学研究发展有限公司 | Eye anterior segment OCT image contour fitting method |
| CN116486467A (en) * | 2022-01-13 | 2023-07-25 | 北京七鑫易维信息技术有限公司 | Method, device, equipment and storage medium for determining eye detection frame |
Family Cites Families (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2014040070A1 (en) * | 2012-09-10 | 2014-03-13 | Oregon Health & Science University | Quantification of local circulation with oct angiography |
| US11717155B2 (en) * | 2019-08-21 | 2023-08-08 | Oregon Health & Science University | Identifying retinal layer boundaries |
-
2023
- 2023-08-17 CN CN202311035548.5A patent/CN116777794B/en active Active
Patent Citations (16)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102567737A (en) * | 2011-12-28 | 2012-07-11 | 华南理工大学 | Method for locating eyeball cornea |
| CN103996020A (en) * | 2014-04-10 | 2014-08-20 | 中航华东光电（上海）有限公司 | Head mounted eye tracker detection method |
| CN111093525A (en) * | 2018-08-07 | 2020-05-01 | 温州医科大学 | Optical coherence tomography image processing method |
| WO2020056454A1 (en) * | 2018-09-18 | 2020-03-26 | MacuJect Pty Ltd | A method and system for analysing images of a retina |
| CN110010219A (en) * | 2019-03-13 | 2019-07-12 | 杭州电子科技大学 | Optical coherence tomography image retinopathy intelligent checking system and detection method |
| WO2020234640A1 (en) * | 2019-05-20 | 2020-11-26 | Macuject Pty Ltd. | Confidence-based methods and systems for analyzing images of a retina |
| CN110363782A (en) * | 2019-06-13 | 2019-10-22 | 平安科技（深圳）有限公司 | A kind of area recognizing method based on limb recognition algorithm, device and electronic equipment |
| CN110349162A (en) * | 2019-07-17 | 2019-10-18 | 苏州大学 | A kind of more lesion image partition methods of macular edema |
| CN110766656A (en) * | 2019-09-19 | 2020-02-07 | 平安科技（深圳）有限公司 | Method, device, equipment and storage medium for screening abnormality of eyeground macular region |
| CN111462156A (en) * | 2020-03-30 | 2020-07-28 | 温州医科大学 | Image processing method for acquiring corneal vertex |
| CN112308829A (en) * | 2020-10-27 | 2021-02-02 | 苏州大学 | Self-adaptive network suitable for high-reflection bright spot segmentation in retina optical coherence tomography image |
| CN112669260A (en) * | 2020-12-07 | 2021-04-16 | 上海交通大学 | Method and device for detecting eye fundus image optic disc yellow spots based on deep neural network |
| CN113940812A (en) * | 2021-11-01 | 2022-01-18 | 朴俊杰 | Cornea center positioning method for excimer laser cornea refractive surgery |
| CN116486467A (en) * | 2022-01-13 | 2023-07-25 | 北京七鑫易维信息技术有限公司 | Method, device, equipment and storage medium for determining eye detection frame |
| CN116109555A (en) * | 2022-11-16 | 2023-05-12 | 南京博视医疗科技有限公司 | Wavefront facula lattice optimization method and device |
| CN116342636A (en) * | 2023-05-23 | 2023-06-27 | 广东麦特维逊医学研究发展有限公司 | Eye anterior segment OCT image contour fitting method |
Non-Patent Citations (4)
| Title |
|---|
| Detecting the Optic Disc Boundary in Digital Fundus Images Using Morphological, Edge Detection, and Feature Extraction Techniques;A. Aquino等;《 IEEE Transactions on Medical Imaging》;第29卷(第11期);1860-1869 * |
| Multimode fiber enables control of spatial coherence in Fourier-domain full-field optical coherence tomography for in vivo corneal imaging;Auksorius E等;《Optics Letters》;第46卷(第6期);1413-1416 * |
| Non-contact eye gaze tracking system by mapping of corneal reflections;Yoo D H等;《Fifth IEEE International Conference on Automatic Face Gesture Recognition》;101-106 * |
| 扫频光学相干层析角膜图像轮廓自动提取算法;汪毅等;《物理学报》;第68卷(第20期);109-115 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN116777794A (en) | 2023-09-19 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106339998B (en) | Multi-focus image fusing method based on contrast pyramid transformation | |
| Morales et al. | Automatic detection of optic disc based on PCA and mathematical morphology | |
| Dai et al. | Retinal fundus image enhancement using the normalized convolution and noise removing | |
| WO2018082084A1 (en) | Brain tumor automatic segmentation method by means of fusion of full convolutional neural network and conditional random field | |
| CN106228528B (en) | A kind of multi-focus image fusing method based on decision diagram and rarefaction representation | |
| CN110751636B (en) | Fundus image retinal arteriosclerosis detection method based on improved coding and decoding network | |
| Qureshi et al. | Detection of glaucoma based on cup-to-disc ratio using fundus images | |
| CN113643353B (en) | Measurement method for enhancing resolution of vascular caliber of fundus image | |
| CN109035172A (en) | A kind of non-local mean Ultrasonic Image Denoising method based on deep learning | |
| CN116236150A (en) | Arteriovenous blood vessel image segmentation method based on fundus image | |
| Mathews et al. | EfficientNet for retinal blood vessel segmentation | |
| Salih et al. | Fast optic disc segmentation using FFT-based template-matching and region-growing techniques | |
| Acharya et al. | Swarm intelligence based adaptive gamma corrected (SIAGC) retinal image enhancement technique for early detection of diabetic retinopathy | |
| Singh et al. | An efficient combined approach for medical brain tumour segmentation | |
| Naz et al. | Glaucoma detection in color fundus images using cup to disc ratio | |
| Dikkala et al. | A comprehensive analysis of morphological process dependent retinal blood vessel segmentation | |
| CN116777794B (en) | Cornea foreign body image processing method and system | |
| CN116030042B (en) | Diagnostic device, method, equipment and storage medium for doctor's diagnosis | |
| Kumar et al. | Automatic optic disc segmentation using maximum intensity variation | |
| Purnama et al. | Follicle detection on the usg images to support determination of polycystic ovary syndrome | |
| Zhou et al. | Automatic fovea center localization in retinal images using saliency-guided object discovery and feature extraction | |
| CN111145155A (en) | Meibomian gland recognition method | |
| Shaikha et al. | Optic Disc Detection and Segmentation in Retinal Fundus Image | |
| Utama et al. | Pupil Center Detection Using Radial Symmetry Transform to Measure Pupil Distance in the Eye | |
| Morra et al. | A neural network approach to unsupervised segmentation of single-channel MR images |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |