Method for determining optical cross-section parameters of human cornea and measuring device

Abstract

The present invention relates to the field of corneal measurement technology, and discloses a method for determining optical cross-sectional parameters of human cornea and a measuring device, comprising the following steps: step S101, real-time positioning of the patient's eyeball position, and emitting a first detection pattern according to the patient's eyeball position; step S102, real-time acquisition of the first detection pattern reflected by the patient's eyeball, marking it as a second detection pattern, and real-time judgment of whether a blinking phenomenon is about to occur; step S103, constructing a detection pattern sequence according to the second detection pattern, and the second detection patterns in the detection pattern sequence are arranged in chronological order. Compared with the common detection method of corneal curvature and curvature radius, the present invention uses two straight lines to sweep across the eye, and calculates the curvature and curvature radius of the cornea according to the deformation of the straight lines, which is more intuitive and simpler and more accurate to calculate than a circle. In addition, the present invention positions the eyeball in real time to improve the accuracy of measuring the corneal curvature and curvature radius.

Description

Technical Field

[0001] The present invention relates to the technical field of corneal measurement, and more specifically, to a method for determining optical cross-sectional parameters of human cornea and a measuring device. Background Art

[0002] Measuring corneal curvature and curvature radius is very important for ophthalmic examination. These data can help doctors understand the health of the eyes and provide patients with appropriate treatment or correction plans. The current common measurement method is to use a keratometer to detect the patient's corneal curvature and curvature radius. The keratometer emits several concentric circles to the patient's cornea, observes the reflected pattern, and calculates the patient's corneal curvature and curvature radius by observing the deformation degree of the concentric circles. However, during the detection process, it is difficult for the patient's head and eyes to remain motionless, which will lead to reduced detection accuracy. In addition, it is more complicated to calculate the corneal curvature and curvature radius based on the deformation of the circle, which is not conducive to accurate measurement of the corneal curvature and curvature radius. Therefore, a method for determining the optical cross-section parameters of the human cornea and a measuring device are proposed. Summary of the invention

[0003] The present invention provides a method for determining optical cross-section parameters of human cornea and a measuring device to solve the technical problems in the above-mentioned background technology.

[0004] The present invention provides a method for determining optical cross-section parameters of a human cornea, comprising the following steps:

[0005] Step S101, locating the eyeball position of the patient in real time, and emitting a first detection pattern according to the eyeball position of the patient;

[0006] Step S102, acquiring a first detection pattern of the patient's eyeball reflection in real time, marking it as a second detection pattern, and determining in real time whether a blinking phenomenon is about to occur;

[0007] If so, suspend acquisition;

[0008] If not, continue to obtain;

[0009] Step S103, constructing a detection pattern sequence according to the second detection pattern, wherein the second detection patterns in the detection pattern sequence are arranged in chronological order;

[0010] Step S104: analyzing the patient's corneal curvature and curvature radius according to the detection pattern sequence.

[0011] Further, the steps of locating the patient's eye position are as follows:

[0012] Step S201, collecting image data of the patient's eye position in real time;

[0013] Step S202, constructing an eye positioning model to extract key points in the image data;

[0014] The key points include: the first key point, the second key point, the third key point, the fourth key point and the fifth key point;

[0015] Step S203, connecting the key points;

[0016] Step S204, determining the patient's eyeball position based on the line connecting the key points.

[0017] Furthermore, the training samples used to train the eye positioning model include:

[0018] Obtain a number of eye photos, each eye photo serving as sample data of a training sample;

[0019] The first key point, the second key point, the third key point, the fourth key point and the fifth key point are manually marked for each eye photo, and each marking is used as a sample label for a training sample.

[0020] Furthermore, the step of emitting the first detection pattern is specifically as follows:

[0021] Step S301, constructing a first detection pattern, the first detection pattern is two mutually perpendicular straight lines, namely a first detection line and a second detection line, wherein the first detection line is arranged along a horizontal direction, and the second detection line is arranged along a vertical direction;

[0022] Step S302, after determining the position of the patient's eyeball, setting the position and size of the first detection pattern according to the connection line of the key points;

[0023] Step S303: the first detection line is scanned across the patient's eyes along the vertical direction, and the second detection line is scanned across the patient's eyes along the horizontal direction.

[0024] Furthermore, the steps of setting the position and length of the first detection pattern are specifically as follows:

[0025] Step S401, taking the third key point / the fourth key point as the starting position of the first detection line, the perpendicular line of the end position of the first detection line passes through the fifth key point, the length of the first detection line is set to a, and taking the third key point / the fourth key point as the end position of the first detection line;

[0026] a is a custom parameter;

[0027] Step S402: one end of the first detection line is used as the starting position of the second detection line, the other end of the first detection line is used as the end position of the second detection line, and the length of the second detection line is the travel length of the first detection line.

[0028] Furthermore, the steps for determining whether the patient is about to blink are as follows:

[0029] Step S501, obtaining the first m frames of image data, extracting the positions of the third key point and the fourth key point in the m frames of image data, and constructing an eyelid opening set;

[0030] Step S502, constructing a blink prediction model, and inputting the eyelid opening set into the blink prediction model;

[0031] Step S503, outputting a prediction result of whether the patient will blink when acquiring the next frame of image data;

[0032] m is a custom parameter.

[0033] Furthermore, the training samples used to train the blink prediction model include:

[0034] Obtain eye video data of several patients, set a sliding window in the eye video data, the length of the sliding window is m frames, and each time the sliding window slides, a sample data of a training sample is generated;

[0035] After obtaining the sample data of the training sample, constructing the sample label of the training sample according to the eye opening of the patient in the next frame of eye video data of the sliding window;

[0036] The steps to construct sample labels for training samples are as follows:

[0037] Step S601, obtaining a frame of eye video data with a maximum eyelid opening as first eye video data;

[0038] Step S602, marking the eyelid opening in the first eye video data as 1;

[0039] Step S603, determining whether the eye opening degree of the eye video data of the next frame of the sliding window is less than p;

[0040] If yes, the corresponding sample label is blink;

[0041] If not, the corresponding sample label is no blink;

[0042] p is a custom parameter, preferably, p is 0.7.

[0043] Further, the steps for analyzing the patient's corneal curvature are as follows:

[0044] Step S701, constructing a coordinate system according to the starting positions of the first detection line and the second detection line;

[0045] Step S702, extracting key points according to the second detection pattern, the key points representing the most curved positions of the first detection line and the second detection line, fitting all the key points using the least square method to obtain a curve, the curve representing the contour of the corneal surface;

[0046] Step S703, calculating the corneal curvature in each second detection pattern according to the curve, and the calculation formula is as follows:

[0047] ;

[0048] In the formula, It represents the equation of the curve, a mathematical expression that represents the contour of the corneal surface; It is the first derivative of the curve, which represents the slope of the curve at a certain point; It is the second derivative of the curve, which represents the rate of change of the curvature of the curve at a certain point;

[0049] Step S704, calculating the curvature radius according to the corneal curvature, and the calculation formula is as follows:

[0050] ;

[0051] In the formula, Represents the radius of curvature of the cornea.

[0052] Further, the steps for analyzing the patient's corneal curvature are as follows:

[0053] Step S801, constructing a curvature analysis model;

[0054] Step S802, inputting the detection pattern sequence into the curvature analysis model;

[0055] Step S803, output corneal curvature and curvature radius.

[0056] The present invention also provides a device for measuring optical cross-sectional parameters of human cornea, comprising a detection unit and an analysis unit;

[0057] The detection unit is used to emit a first detection pattern to the patient's eyeball, receive the reflected first detection pattern, and mark it as a second detection pattern; the analysis unit is used to analyze the patient's corneal curvature and curvature radius according to the second detection pattern.

[0058] The beneficial effects of the present invention are as follows: compared with the common detection methods of corneal curvature and curvature radius, the present invention uses two straight lines to sweep across the eye, and calculates the curvature and curvature radius of the cornea based on the deformation of the straight lines, which is more intuitive and the calculation is simpler and more accurate than a circle. In addition, the present invention positions the eyeball in real time, improves the accuracy of measuring the corneal curvature and curvature radius, and can predict whether the patient is about to blink, supporting the patient to maintain a high detection accuracy even when blinking, thereby improving the user experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] Figure 1 It is the overall flow chart of the present invention. DETAILED DESCRIPTION

[0060] The subject matter described herein will now be discussed with reference to example embodiments. It should be understood that the discussion of these embodiments is only to enable those skilled in the art to better understand and implement the subject matter described herein, and the functions and arrangements of the elements discussed may be changed without departing from the scope of protection of the contents of this specification. Each example may omit, replace or add various processes or components as needed. In addition, the features described relative to some examples may also be combined in other examples.

[0061] It should be noted that, unless otherwise defined, the technical terms or scientific terms used in one or more embodiments of the present invention should be understood by people with ordinary skills in the field to which the present invention belongs. The words "first", "second" and similar words used in one or more embodiments of the present invention do not indicate any order, quantity or importance, but are only used to distinguish different components. "Include" or "comprise" and similar words mean that the elements or objects appearing before the word include the elements or objects listed after the word and their equivalents, without excluding other elements or objects. "Connect" or "connected" and similar words are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. "Up", "down", "left", "right" and the like are only used to indicate relative positional relationships. When the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0062] like Figure 1 As shown, the purpose of the first embodiment of the present invention is that when the corneal curvature of the patient is detected, the accuracy requirement is very high, so the patient needs to remain still during the detection. However, it is difficult for a person to remain still, especially the eyeball. Therefore, when the keratometer detects the corneal curvature of the patient, the calculation result will be inaccurate. Therefore, a method for determining the optical cross-section parameters of the human cornea is proposed, which includes the following steps:

[0063] First, since the patient may slightly shake during the test, the patient's eyes need to be positioned. Therefore, in step S101, the patient's eyeball position is positioned in real time, and a first detection pattern is emitted according to the patient's eyeball position.

[0064] The steps to locate the patient's eye position are as follows:

[0065] Step S201, collecting image data of the patient's eye position in real time;

[0066] Step S202, constructing an eye positioning model to extract key points in the image data;

[0067] The key points include: the first key point, the second key point, the third key point, the fourth key point and the fifth key point;

[0068] Among them, the first key point corresponds to the inner corner of the eyeball, the second key point corresponds to the outer corner of the eyeball, the third key point corresponds to the edge of the upper eyelid, the fourth key point corresponds to the edge of the lower eyelid, and the fifth key point corresponds to the center of the pupil of the eyeball;

[0069] Step S203, connecting the key points;

[0070] Step S204, determining the patient's eyeball position based on the line connecting the key points.

[0071] It should be noted that the eye positioning model is built based on the openpose model. In addition, key points can be appropriately added according to actual conditions, such as setting several key points at the upper and lower eyelid positions, so as to build a model that is closer to the actual shape of the eye.

[0072] The training samples used to train the eye localization model include:

[0073] Obtain a number of eye photos, each eye photo serving as sample data of a training sample;

[0074] The first key point, the second key point, the third key point, the fourth key point and the fifth key point are manually marked for each eye photo, and each marking is used as a sample label for a training sample.

[0075] The calculation formula of the loss function of the eye positioning model is as follows:

[0076] ;

[0077] In the formula, Represents the loss value of the eye localization model; Indicates the number of all key points; Indicates the number of key point types; Indicates at location The probability of the occurrence of the c-th key point; Indicates an indicator variable. If the position does contain the c-th type of key points, then ; If the position does not contain the c-th type of key points, then .

[0078] The steps of emitting the first detection pattern are as follows:

[0079] Step S301, constructing a first detection pattern, the first detection pattern is two mutually perpendicular straight lines, namely a first detection line and a second detection line, wherein the first detection line is arranged along a horizontal direction, and the second detection line is arranged along a vertical direction;

[0080] Step S302, after determining the position of the patient's eyeball, setting the position and size of the first detection pattern according to the connection line of the key points;

[0081] Step S303: the first detection line is scanned across the patient's eyes along the vertical direction, and the second detection line is scanned across the patient's eyes along the horizontal direction.

[0082] It should be noted that the first detection line and the second detection line scan the patient's eyes in sequence to avoid the light reflected by the overlapping position of the first detection line and the second detection line when scanning the patient's eyes affecting the detection result.

[0083] The steps for setting the position and length of the first detection pattern are as follows:

[0084] Step S401, taking the third key point / the fourth key point as the starting position of the first detection line, the perpendicular line of the end position of the first detection line passes through the fifth key point, the length of the first detection line is set to a, and taking the third key point / the fourth key point as the end position of the first detection line;

[0085] a is a custom parameter;

[0086] It is worth noting that, since the diameter of the cornea is about 11-12 mm, during detection, the length of the first detection line needs to be slightly larger than the diameter of the cornea. Therefore, preferably, a is 15 mm.

[0087] Step S402: one end of the first detection line is used as the starting position of the second detection line, the other end of the first detection line is used as the end position of the second detection line, and the length of the second detection line is the travel length of the first detection line.

[0088] In addition, when the first detection pattern scans the patient's eyes, the patient is very likely to blink, which will affect the final detection result. In addition, when the patient blinks, the eyeball is likely to move. Therefore, in step S102, the first detection pattern reflected by the patient's eyeball is obtained in real time, marked as the second detection pattern, and whether the blinking phenomenon is about to occur is determined in real time;

[0089] If so, suspend acquisition;

[0090] If not, continue to obtain;

[0091] The steps to determine if a patient is about to blink are as follows:

[0092] Step S501, obtaining the first m frames of image data, extracting the positions of the third key point and the fourth key point in the m frames of image data, and constructing an eyelid opening set;

[0093] Step S502, constructing a blink prediction model, and inputting the eyelid opening set into the blink prediction model;

[0094] Step S503, outputting a prediction result of whether the patient will blink when acquiring the next frame of image data;

[0095] m is a custom parameter, preferably, m is 10.

[0096] Preferably, the blink prediction model is constructed based on a long short-term memory network.

[0097] The training samples used to train the blink prediction model include:

[0098] Obtain eye video data of several patients, set a sliding window in the eye video data, the length of the sliding window is m frames, and each time the sliding window slides, a sample data of a training sample is generated;

[0099] After obtaining the sample data of the training sample, a sample label of the training sample is constructed according to the eye opening of the patient in the next frame of eye video data of the sliding window.

[0100] It should be noted that the time interval between each frame of eye video data and each frame of image data is the same, and the distance between the patient and the detection device is the same.

[0101] The steps to construct sample labels for training samples are as follows:

[0102] Step S601, obtaining a frame of eye video data with a maximum eyelid opening as first eye video data;

[0103] Step S602, marking the eyelid opening in the first eye video data as 1;

[0104] Step S603, determining whether the eye opening degree of the eye video data of the next frame of the sliding window is less than p;

[0105] If yes, the corresponding sample label is blink;

[0106] If not, the corresponding sample label is no blink;

[0107] p is a custom parameter, preferably, p is 0.7.

[0108] It should be noted that the eyelid opening of the first eye video data is marked as 1, where 1 represents the maximum eyelid opening of the patient, which can adapt to patients with different eyelid openings and improve the accuracy of the blink prediction model.

[0109] Step S103, constructing a detection pattern sequence according to the second detection pattern, wherein the second detection patterns in the detection pattern sequence are arranged in chronological order;

[0110] Step S104: analyzing the patient's corneal curvature and curvature radius according to the detection pattern sequence.

[0111] The steps for analyzing a patient's corneal curvature are as follows:

[0112] Step S701, constructing a coordinate system according to the starting positions of the first detection line and the second detection line;

[0113] Step S702, extracting key points according to the second detection pattern, the key points representing the most curved positions of the first detection line and the second detection line, fitting all the key points using the least square method to obtain a curve, the curve representing the contour of the corneal surface;

[0114] Step S703, calculating the corneal curvature in each second detection pattern according to the curve, and the calculation formula is as follows:

[0115] ;

[0116] In the formula, It represents the equation of the curve, a mathematical expression that represents the contour of the corneal surface; It is the first derivative of the curve, which represents the slope of the curve at a certain point; It is the second derivative of the curve, which represents the rate of change of the curvature of the curve at a certain point;

[0117] Step S704, calculating the curvature radius according to the corneal curvature, and the calculation formula is as follows:

[0118] ;

[0119] In the formula, Represents the radius of curvature of the cornea.

[0120] The second embodiment of the present invention aims to analyze the corneal curvature of the patient in the following steps:

[0121] Step S801, constructing a curvature analysis model;

[0122] Step S802, inputting the detection pattern sequence into the curvature analysis model;

[0123] Step S803, output corneal curvature and curvature radius.

[0124] Preferably, the curvature analysis model is constructed based on a long short-term memory network.

[0125] When training the curvature analysis model, several people with different corneal curvature conditions are selected, and their detection pattern sequences are obtained respectively, and each person's detection pattern sequence is used as sample data of a training sample;

[0126] The sample labels of the training samples are obtained through annotation by relevant technical experts.

[0127] The third embodiment of the present invention aims to provide a device for measuring optical cross-section parameters of human cornea, comprising a detection unit and an analysis unit;

[0128] The detection unit is used to emit a first detection pattern to the patient's eyeball, receive the reflected first detection pattern, and mark it as a second detection pattern; the analysis unit is used to analyze the patient's corneal curvature and curvature radius according to the second detection pattern.

[0129] The above describes an embodiment of the present embodiment, but the present embodiment is not limited to the above-mentioned specific implementation methods. The above-mentioned specific implementation methods are merely illustrative and not restrictive. Under the guidance of the present embodiment, ordinary technicians in this field can also make many forms, all of which are protected by the present embodiment.

Claims

1. A method for determining optical cross-section parameters of a human cornea, characterized in that: The following steps are involved: Step S101, locating the eyeball position of the patient in real time, and emitting a first detection pattern according to the eyeball position of the patient; The steps to locate the patient's eye position are as follows: Step S201, collecting image data of the patient's eye position in real time; Step S202, constructing an eye positioning model to extract key points in the image data; The key points include: the first key point, the second key point, the third key point, the fourth key point and the fifth key point; Step S203, connecting the key points; Step S204, determining the patient's eyeball position based on the line connecting the key points; The steps of emitting the first detection pattern are as follows: Step S301, constructing a first detection pattern, the first detection pattern is two mutually perpendicular straight lines, namely a first detection line and a second detection line, wherein the first detection line is arranged along a horizontal direction, and the second detection line is arranged along a vertical direction; Step S302, after determining the position of the patient's eyeball, setting the position and size of the first detection pattern according to the connection line of the key points; Step S303, the first detection line is scanned across the patient's eyes along the vertical direction, and the second detection line is scanned across the patient's eyes along the horizontal direction; The steps for setting the position and length of the first detection pattern are as follows: Step S401, taking the third key point / the fourth key point as the starting position of the first detection line, the perpendicular line of the end position of the first detection line passes through the fifth key point, the length of the first detection line is set to a, and taking the third key point / the fourth key point as the end position of the first detection line; a is a custom parameter; Step S402, taking one end of the first detection line as the starting position of the second detection line, taking the other end of the first detection line as the end position of the second detection line, and the length of the second detection line is the travel length of the first detection line; Step S102, acquiring a first detection pattern of the patient's eyeball reflection in real time, marking it as a second detection pattern, and determining in real time whether a blinking phenomenon is about to occur; If so, suspend acquisition; If not, continue to obtain; Step S103, constructing a detection pattern sequence according to the second detection pattern, wherein the second detection patterns in the detection pattern sequence are arranged in chronological order; Step S104: analyzing the patient's corneal curvature and curvature radius according to the detection pattern sequence.

2. The method for determining the optical cross-section parameters of the human cornea according to claim 1, characterized in that: The training samples used to train the eye localization model include: Obtain a number of eye photos, each eye photo serving as sample data of a training sample; The first key point, the second key point, the third key point, the fourth key point and the fifth key point are manually marked for each eye photo, and each marking is used as a sample label for a training sample.

3. The method for determining the optical cross-section parameters of the human cornea according to claim 1, characterized in that: The steps to determine if a patient is about to blink are as follows: Step S501, obtaining the first m frames of image data, extracting the positions of the third key point and the fourth key point in the m frames of image data, and constructing an eyelid opening set; Step S502, constructing a blink prediction model, and inputting the eyelid opening set into the blink prediction model; Step S503, outputting a prediction result of whether the patient will blink when acquiring the next frame of image data; m is a custom parameter.

4. The method for determining the optical cross-section parameters of the human cornea according to claim 3, characterized in that: The training samples used to train the blink prediction model include: Obtain eye video data of several patients, set a sliding window in the eye video data, the length of the sliding window is m frames, and each time the sliding window slides, sample data of a training sample is generated; After obtaining the sample data of the training sample, constructing the sample label of the training sample according to the eye opening of the patient in the next frame of eye video data of the sliding window; The steps to construct sample labels for training samples are as follows: Step S601, obtaining a frame of eye video data with a maximum eyelid opening as first eye video data; Step S602, marking the eyelid opening in the first eye video data as 1; Step S603, determining whether the eye opening degree of the eye video data of the next frame of the sliding window is less than p, where p is a custom parameter; If yes, the corresponding sample label is blink; If not, the corresponding sample label is no blink.

5. The method for determining the optical cross-section parameters of the human cornea according to claim 1, characterized in that: The steps for analyzing a patient's corneal curvature are as follows: Step S701, constructing a coordinate system according to the starting positions of the first detection line and the second detection line; Step S702, extracting key points according to the second detection pattern, the key points representing the most curved positions of the first detection line and the second detection line, fitting all the key points using the least square method to obtain a curve, the curve representing the contour of the corneal surface; Step S703, calculating the corneal curvature in each second detection pattern according to the curve, and the calculation formula is as follows: ； In the formula, It represents the equation of the curve, a mathematical expression that represents the contour of the corneal surface; It is the first derivative of the curve, which represents the slope of the curve at a certain point; It is the second derivative of the curve, which represents the rate of change of the curvature of the curve at a certain point; Step S704, calculating the curvature radius according to the corneal curvature, and the calculation formula is as follows: ； In the formula, Represents the radius of curvature of the cornea.

6. The method for determining the optical cross-section parameters of the human cornea according to claim 1, characterized in that: The steps for analyzing a patient's corneal curvature are as follows: Step S801, constructing a curvature analysis model; Step S802, inputting the detection pattern sequence into the curvature analysis model; Step S803, output corneal curvature and curvature radius.

Images