A method for rating the corrective effect of corneal topography after orthokeratology.

Abstract

This invention provides a method for rating the corrective effect of corneal topography after orthokeratology. It utilizes a mobile image acquisition system to collect corneal topography images of the user one month after lens wear. In a data segmentation and processing system, a deep learning model segments the effective optical zone and pupil within the corneal topography image. Medical indicators such as eccentricity, defocus, and effective defocus range are calculated and rated according to evaluation criteria. The results are finally output to visual hardware. This invention provides a method for rating the corrective effect of corneal topography after orthokeratology. By leveraging the segmentation capabilities of a deep learning model, the effective optical zone and pupil in the corneal topography image can be accurately segmented, and the corneal topography can be rated based on calculated medical indicators. This avoids subjective interference from the operator during lens fitting, and overall, this invention can significantly reduce the fitting workload of optometrists.

Description

Technical Field

[0001] This invention belongs to the field of optometric testing, and in particular relates to a method for rating the corrective effect of corneal topography after orthokeratology. Background Technology

[0002] In recent years, with the rise of various electronic products and increasingly busy academic workloads, teenagers have been spending more time using their eyes, leading to a year-on-year increase in the incidence of myopia and a trend towards younger ages. Statistics show that in 2020, the overall myopia rate among Chinese teenagers was 52.7%, seriously affecting the normal learning and lives of primary and secondary school students and placing a huge burden on society, becoming a global concern. Orthokeratology lenses are specially designed, high-oxygen-permeable rigid corneal contact lenses. When worn overnight, the positive pressure of the lens and the negative pressure formed by the tear film under the lens reshape the cornea, thus correcting refractive errors by changing the corneal geometry, thereby temporarily reducing myopia or corneal astigmatism. Therefore, orthokeratology lenses are widely used for myopia control among teenagers.

[0003] Corneal topography is an essential examination during the fitting of orthokeratology lenses. It provides a visual morphological analysis, with different colors at each point reflecting the refractive power at that location. During the pre-fitting examination, it can screen for suitable candidates, such as filtering out individuals with keratoconus, limbal degeneration, or irregular corneas that are too flat or steep. It can also select orthokeratology lens parameters that match the wearer's cornea based on information such as flatness or steepness. During follow-up examinations, corneal topography can also provide valuable information on the reshaping area and the reshaping effect. Annual follow-up examinations are recommended after orthokeratology lens use, requiring a 4-6 week break from lens wear to allow the cornea to return to its original shape. All these functions of corneal topography help doctors develop subsequent treatment plans for users. However, previous studies have only focused on basic morphological information from corneal topography and have not conducted in-depth analysis of commonly used medical indicators. In fact, when fitting orthokeratology lenses to different users, optometrists need to consider multiple factors, such as whether the lens's eccentricity is reasonable, whether there is a certain amount of defocus within the pupil, and the relative positional relationship between the pupil and the effective optical zone. These medical indicators together determine the quality of corneal topography correction.

[0004] Currently, there is no objective evaluation method for the fitting effect of orthokeratology lenses. Optometrists often analyze the morphology and curvature of important areas of corneal topography based on clinical experience, which leads to a certain degree of subjectivity in the fitting results. Furthermore, because optometrists' clinical experience varies, it is difficult to guarantee that the optimal fitting result can be obtained for each fitting based on multiple refractive indicators and corneal topography information. In addition, statistics show that the axial elongation of the eye varies among users with different types of corneal topography in the coming year. Therefore, fitting the appropriate type of corneal topography for the user can not only better correct vision but also inhibit axial elongation to a certain extent. In conclusion, the method described in this study helps to improve the fitting effect of orthokeratology lenses, thereby improving the myopia control effect. Summary of the Invention

[0005] In view of this, the present invention aims to propose a method for rating the corrective effect of corneal topography after orthokeratology, in order to solve the problem that personal factors have a great influence on the lens fitting process and the quality is uncontrollable in the existing technology.

[0006] To achieve the above objectives, the technical solution of the present invention is implemented as follows:

[0007] A method for rating the corrective effect of corneal topography after orthokeratology involves using a mobile image acquisition system to collect corneal topography images of the user one month after wearing the lenses. In a data segmentation and processing system, a deep learning model is used to segment the effective optical zone and pupil in the corneal topography image, calculate indicators such as eccentricity, defocus amount, and effective defocus contact range, and then output the rating of the corneal topography image to visual hardware according to specified conditions.

[0008] Furthermore, the method for acquiring corneal topography in the mobile image acquisition system is as follows:

[0009] S1. Using a mobile image acquisition device, corneal topography of the wearer can be acquired according to the user's needs. The device can directly access the stored images or directly capture images using a corneal topography instrument.

[0010] S2. Organize all corneal topography maps collected from the wearer's follow-up examination one month later, and clean out incomplete topography maps caused by factors such as the wearer closing their eyes.

[0011] S3. Under the guidance of an optometrist, use the labelme annotation method to annotate the effective optical zone and the range of the pupil.

[0012] Furthermore, the data segmentation system requires the construction of a deep learning model, which can be divided into convolutional layers, pooling layers, and activation layers. After obtaining the corneal topography map to be segmented, the image data information is input through multiple channels and convolved. Each convolution in the convolutional layer produces one channel of output, thus obtaining multiple channels of output. Convolution is defined as:

[0013]

[0014] Where z[x,y] is the calculation result, g[x,y] is the input data, f[x,y] is the convolution kernel, and * indicates the convolution operation.

[0015] Furthermore, each value of the input channel needs to be modified by an activation function to obtain the same output size. This calculation process is completed in the activation layer, specifically using the ReLU function: f(x) = max(0,x).

[0016] Furthermore, the output data needs to be sampled proportionally in the pooling layer, and the sampling method can be expressed as: f(X) = max([X]).

[0017] like Figure 3 As shown, after deep learning model, effective optical zone and pupil segmentation map can be generated; input corneal topography map of the user after wearing orthokeratology lenses for one month, after multiple convolution and activation processes, the image is pooled to obtain multi-dimensional feature map, the feature map is upsampled multiple times to obtain the corresponding segmentation map, and then the network structure and parameters are optimized according to the actual image situation.

[0018] Furthermore, the aforementioned data processing system calculates indicators such as eccentricity, defocus amount, and effective defocus contact range, and classifies the corneal topography according to specified conditions. The specific steps are as follows:

[0019] A1. Extract the effective optical zone and pupil outline in the HSV space according to the specified color.

[0020] A2. Calculate the circumscribed rectangle of the effective optical zone and the pupil outline, and obtain the position of their center points. The distance between their center points is the eccentricity.

[0021] A3. Regions corresponding to each color inside the pupil, segmented within the HSV space.

[0022] A4. Take the intersection of the pupil and the effective optical zone, and perform a difference operation between the area obtained in step 3 and this intersection to obtain the range and area of ​​each color in the defocused area (non-optical zone inside the pupil).

[0023] A5. In the corneal topography map, each color corresponds to a refractive power. Multiply the area of ​​each color by its corresponding refractive power and then sum the results. This value is the defocus amount.

[0024] A6. Binarize the segmented pupils and use the four-neighbor algorithm to obtain the coordinates of all pupil boundary points and the coordinates of defocus region boundary points.

[0025] A7. The number of points where the pupil boundary coordinates coincide with the defocus area boundary coordinates is denoted as n1, and the total number of all coordinate points at the pupil boundary is denoted as n2. n1 / n2 is the effective defocus contact range.

[0026] A8. The above three indicators will be used to perform a topographic rating of the corneal reshaping effect according to the conditions specified by the fitting physician. The specific conditions are as follows:

[0027] Level 1: 0mm ≤ eccentricity < 1mm, effective defocus contact range > 3 / 4, defocus amount > 30.

[0028] Level 2: 0mm ≤ eccentricity < 1mm, effective defocus contact range between 1 / 4 and 3 / 4, defocus amount > 30.

[0029] Level 3: Eccentricity > 1mm, defocusing amount > 30.

[0030] Level 4: Eccentricity <0.5mm, effective defocus contact range <1 / 4, defocus amount <30. Attached Figure Description

[0031] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an undue limitation of the invention. In the drawings:

[0032] Figure 1 This is a schematic diagram of the operation of the application unit described in the embodiment of the present invention.

[0033] Figure 2 This is a schematic diagram of the deep learning mesh structure described in an embodiment of the present invention.

[0034] Figure 3 This is a schematic diagram of the segmented image according to an embodiment of the present invention.

[0035] Figure 4 This is a flowchart of the eccentricity calculation as described in an embodiment of the present invention.

[0036] Figure 5 This is a schematic diagram of color extraction in the defocused area according to an embodiment of the present invention.

[0037] Figure 6 This is a flowchart illustrating the calculation of the effective defocus contact range as described in an embodiment of the present invention.

[0038] Figure 7 This is a schematic diagram of corneal topography at various levels as described in the embodiments of the present invention. Detailed Implementation

[0039] It should be noted that, unless otherwise specified, the embodiments and features described in the present invention can be combined with each other.

[0040] In the description of this invention, it should be understood that the terms "center," "longitudinal," "lateral," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," and "outer," etc., indicating orientations or positional relationships based on the orientations or positional relationships shown in the accompanying drawings, are only for the convenience of describing the invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the invention. Furthermore, the terms "first," "second," etc., are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, a feature defined with "first," "second," etc., may explicitly or implicitly include one or more of that feature. In the description of this invention, unless otherwise stated, "a plurality of" means two or more.

[0041] In the description of this invention, it should be noted that, unless otherwise explicitly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art will understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0042] The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0043] like Figure 1 As shown in this embodiment, a method for rating the corrective effect of corneal topography after orthokeratology includes a mobile image acquisition system, a data segmentation system, a data processing system, and a personalized user system. The mobile image acquisition system acquires corneal topography after one month of wear. Then, a deep learning model is used to identify the effective optical zone and pupil in the corneal topography and calculate indicators such as eccentricity, defocus, and effective defocus contact range. The corneal topography is rated according to the specified conditions, thereby helping operators to diagnose wearers more quickly and effectively.

[0044] The mobile image acquisition system directly collects corneal topography information from the wearer using a corneal topography instrument.

[0045] The data segmentation system initially collects corneal topography maps from a large number of wearers. Under the guidance of professional optometrists, the effective optical zone and pupil are labeled using the LabelMe annotation method. The labeled files and their corresponding corneal topography maps are then input into the U-net neural network for training, thereby segmenting the effective optical zone and pupil in the corneal topography maps.

[0046] The data processing system identifies the effective optical zone and pupil in the corneal topography map and calculates indicators such as eccentricity, defocus amount, and effective defocus contact range, and rates the corneal topography map according to specified conditions.

[0047] The personalized user system returns information such as the effective optical zone and pupil to the user interface based on the operator's requirements and usage habits, thereby helping the operator to make more effective diagnoses for the wearer.

[0048] When a user installs the corneal reshaping vision correction effect rating software on a mobile client, the mobile image acquisition system requires granting the method certain operating permissions; that is, the method needs to obtain image usage permissions from the mobile device. The specific steps are as follows:

[0049] S1. Using a mobile image acquisition device, corneal topography of the wearer can be acquired according to the user's needs. The stored images can be accessed directly or the image can be captured directly using a corneal topography instrument.

[0050] S2. Organize all corneal topography maps collected from the wearer's follow-up examination one month later, and clean out incomplete topography maps caused by factors such as the wearer closing their eyes.

[0051] S3. Under the guidance of an optometrist, use the labelme annotation method to annotate the effective optical zone and the range of the pupil.

[0052] The data segmentation system requires the construction of a deep learning model. For example... Figure 2 Specifically, it can be divided into convolutional layers, pooling layers, and activation layers. After obtaining the corneal topography map to be segmented, the image data is input through multiple channels and convolved. Each convolution in the convolutional layer produces one channel of output, thus obtaining multiple channels of output. Convolution is defined as:

[0053]

[0054] Where z[x,y] is the calculation result, g[x,y] is the input data, f[x,y] is the convolution kernel, and * indicates the convolution operation.

[0055] Furthermore, each value of the input channel needs to be modified by an activation function to obtain the same output size. This calculation process is completed in the activation layer, specifically using the ReLU function: f(x) = max(0,x).

[0056] Furthermore, the output data needs to be sampled proportionally in the pooling layer, and the sampling method can be expressed as: f(X) = max([X]).

[0057] like Figure 3 As shown, after processing by a deep learning model, segmentation maps of the effective optical zone and pupil can be generated. The input is a corneal topography map of the user one month after wearing orthokeratology lenses. After multiple convolution and activation processes, the image undergoes pooling to obtain a multi-dimensional feature map. The feature map is then upsampled multiple times to obtain the corresponding segmentation map. Finally, the network structure and parameters are optimized based on the actual image conditions.

[0058] The data processing system calculates indicators such as eccentricity, defocus amount, and effective defocus contact range, and classifies the corneal topography according to specified conditions. The specific steps are as follows:

[0059] A1. Extract the effective optical zone and pupil outline in the HSV space according to the specified color.

[0060] A2, such as Figure 4 As shown, calculate the circumscribed rectangle of the effective optical zone and the pupil contour, obtain the position of their center points, and calculate the distance between their center points as the eccentricity.

[0061] A3, such as Figure 5 As shown, the regions corresponding to each color inside the pupil are segmented in the HSV space.

[0062] A4. Take the intersection of the pupil and the effective optical zone, and perform a difference operation between the area obtained in step 3 and this intersection to obtain the range and area of ​​each color in the defocused area (non-optical zone inside the pupil).

[0063] A5. In the corneal topography map, each color corresponds to a refractive power. Multiply the area of ​​each color by its corresponding refractive power and then sum the results. This value is the defocus amount.

[0064] A6. Binarize the segmented pupils and use the four-neighbor algorithm to obtain the coordinates of all pupil boundary points and the coordinates of defocus region boundary points.

[0065] A7, such as Figure 6 As shown, the number of points where the pupil boundary coordinates coincide with the defocus area boundary coordinates is denoted as n1, and the number of all coordinate points at the pupil boundary is denoted as n2. n1 / n2 is the effective defocus contact range.

[0066] A8, such as Figure 7As shown, the above three indicators are used to rate the corneal topography visual acuity correction effect according to the conditions specified by the optometrist. The specific conditions are as follows:

[0067] Level 1: 0mm ≤ eccentricity < 1mm, effective defocus contact range > 3 / 4, defocus amount > 30.

[0068] Level 2: 0mm ≤ eccentricity < 1mm, effective defocus contact range between 1 / 4 and 3 / 4, defocus amount > 30.

[0069] Level 3: Eccentricity > 1mm, defocusing amount > 30.

[0070] Level 4: Eccentricity <0.5mm, effective defocus contact range <1 / 4, defocus amount <30.

[0071] The personalized user system allows direct interaction with optometrists, who can directly observe the corneal topography visual acuity correction rating. Furthermore, if an optometrist perceives significant issues with the corneal topography visual acuity correction rating calculation, such as inaccurate segmentation of the effective optical zone or pupil leading to deviations in the rating, they can correct these issues using the effective optical zone or pupil annotation file. If the optometrist's correction is correct after a second discussion, the corrected topography annotation file can be used as a training set to retrain the model, thereby optimizing it. More accurate segmentation of the effective optical zone and pupil facilitates more precise calculations of the corneal topography visual acuity correction rating.

[0072] By leveraging the segmentation capabilities of deep learning models, the effective optical zone and pupil in corneal topography can be accurately segmented, and a corneal topography vision correction effect rating can be calculated, thereby evaluating the fitting effect and avoiding interference from subjective factors when doctors fit users. Overall, this invention can greatly reduce the fitting pressure on optometrists and help them fit users more effectively and faster.

[0073] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method of performing a corrective effect rating of a corneal topography after orthokeratology, characterized by: The corneal topography of the user after wearing for one month is collected by using a mobile terminal image acquisition system, the effective optical zone and pupil in the corneal topography are segmented in the data segmentation system and the data processing system through a deep learning model, the eccentricity, defocus amount and effective defocus range medical indexes are calculated, and the evaluation standard is used for grading, and finally the calculation result is output to a visual hardware; The data processing system calculates the eccentricity, defocus amount and effective defocus contact range indexes, and classifies the corneal topography according to the specified conditions, and the specific steps are as follows: A1, extracting the outline of the effective optical zone and pupil according to the specified color in the hsv space; A2, calculating the circumscribed rectangle of the effective optical zone and pupil outline, and obtaining the position of the center points of the two, and calculating the distance between the center points of the two, that is, the eccentricity; A3, segmenting the regions corresponding to each color inside the pupil in the hsv space; A4, taking the intersection of the pupil and the effective optical zone, and performing difference operation on the region obtained in step 3 and the intersection, that is, obtaining the range and area of each color corresponding to the defocus region (non-optical zone inside the pupil); A5, each color in the corneal topography corresponds to one diopter, and the area of each color is multiplied by the corresponding diopter and then summed to obtain the defocus amount; A6, binarizing the segmented pupil, and obtaining all the pupil boundary point coordinates and defocus region boundary point coordinates by using the four-neighborhood algorithm; A7, the number of overlapping points of the pupil boundary point coordinates and the defocus region boundary point coordinates is n1, the number of all pupil boundary point coordinates is n2, and n1 / n2 is the effective defocus contact range; A8, the three indexes in step 7 are graded according to the conditions specified by the optometrist for the visual correction effect of the corneal topography, and the conditions are as follows: 1st level: 0mm≤eccentricity<1mm, effective defocus contact range>3 / 4, defocus amount>30; 2nd level: 0mm≤eccentricity<1mm, effective defocus contact range is between 1 / 4 and 3 / 4, defocus amount>30; 3rd level: eccentricity>1mm, defocus amount>30; 4th level: eccentricity<0.5mm, effective defocus contact range<1 / 4, defocus amount<30.

2. The method of claim 1, wherein: The method for collecting corneal topography by using a mobile terminal image acquisition system is as follows: S1, using a mobile terminal image acquisition device, collecting the corneal topography of the wearer according to the user's demand, which can directly access the stored image or directly use the corneal topography instrument to take the image; S2, sorting all the collected corneal topography of the wearer after one month of review, and cleaning the incomplete topography caused by the wearer's closed eyes; S3, using the labelme labeling method, labeling the range of the effective optical zone and pupil under the guidance of the optometrist.

3. The method of claim 1, wherein: The data segmentation system needs to construct a deep learning model, which can be divided into convolution layer, pooling layer and activation layer; wherein, after obtaining the corneal topography to be segmented, the information of the image data is input and convolved through multiple channels, and each convolution of the convolution layer produces an output of one channel, so that multiple channel outputs can be obtained, wherein the convolution is defined as: Wherein, z[x, y] is the calculation result, g[x, y] is the input data, f[x, y] is the convolution kernel, and * represents the convolution operation.

4. The method of claim 3, wherein: Each value of the input channel needs to be modified by the activation function, so as to obtain the output of the same size, and the calculation process is completed in the activation layer, and the Relu function can be used: f(x) = max(0, x).

5. The method of claim 3, wherein: The output data needs to be proportionally sampled in the pooling layer, and the sampling method can be represented as: f(X) = max([X]); After the deep learning model, the segmentation map of the effective optical zone and the pupil can be generated; the corneal topography after the user wears the molding mirror for one month is input, the image is subjected to multiple convolution processes and activation processes, and then the multi-dimensional feature map is obtained through the pooling operation, the feature map is subjected to multiple up-sampling to obtain the corresponding segmentation map, and then the network structure and parameters are optimized according to the actual situation of the image.

Images