KR20210025427A - Method and device for providing vision correction surgery visualization information - Google Patents

Abstract

The present invention relates to a method of providing a predicted vision image. According to one embodiment of the present invention, a method of providing a predicted vision image after vision correction surgery by using artificial intelligence performed by a computing device, comprises the following steps of: acquiring examination data of an examinee, which includes medical examination and an eyeball characteristic data measurement value; calculating an eyeball characteristic data prediction value, which includes at least one of a vision prediction value and a corneal shape factor prediction value, after vision correction surgery of the examinee by inputting first group data acquired from the examination data of the examinee to a first prediction model; and generating a predicted vision image based on the eyeball characteristic data prediction value, wherein the first prediction model is trained based on at least one among eyeball characteristic data measurement values before surgery of a plurality of patients who underwent vision correction surgery, surgery parameters of vision correction surgery operated to the patients, and eyeball characteristic data measurement values after surgery of the patients.

Description

Visualization information provision method and device for vision correction {METHOD AND DEVICE FOR PROVIDING VISION CORRECTION SURGERY VISUALIZATION INFORMATION}

The present invention relates to a method and apparatus for providing visualization information for vision correction, and more particularly, to a method and apparatus for visualizing and providing information related to vision correction to a subject, etc. using artificial intelligence.

Vision correction techniques such as LASIK and LASEK are attracting a lot of attention from people with poor eyesight regardless of age or sex. Interest in vision correction is increasing day by day to the extent that there are statistics that the population who received vision correction is 100,000 people.

However, it is difficult for a subject who wants to undergo vision correction to determine which vision correction procedure is right for them. The subject should basically select the type of surgery such as LASIK, LASEK, Smile, and lens implantation. In addition, within LASIK, you should choose according to surgical equipment such as I LASIK, DaVinci LASIK, Crystal LASIK, Z LASIK, VIJU LASIK, and OPTI LASIK, and according to the surgical methods such as Contura Vision, Extra LASIK, and Wavefront LASIK. In addition, since the recommended corneal cutting amount varies according to hospitals and doctors, it is more difficult to select, and the reality is that we have no choice but to rely on the words of a doctor or counselor for the quality of vision or side effects after surgery.

One task is to visually provide information related to vision correction to doctors, counselors, and subjects.

Another task is to provide a predicted visual acuity image to a doctor, a counselor, a subject, and the like.

Another task is to provide corneal topography images to doctors, counselors, and subjects.

Another task is to provide a cause for calculating the predicted results to doctors, counselors, and subjects.

The problem to be solved is not limited to the above-described problems, and the problems that are not mentioned will be clearly understood by those of ordinary skill in the art from the present specification and the accompanying drawings.

According to an aspect, in a method of providing a predicted vision image after vision correction using artificial intelligence performed by a computing device, the step of obtaining examination data of a subject-the examination data includes examination data and eye characteristic data measurement values- ; The predicted value of the eye characteristic data after vision correction of the examinee by inputting the first group data obtained from the examination data of the examinee into a first prediction model-The predicted value of the eye characteristic data includes at least one of an acuity predicted value and a corneal shape factor predicted value -Calculating; And generating a predicted visual acuity image based on the predicted value of the eye characteristic data; including, wherein the first predictive model includes a measurement value of eye characteristic data before surgery of a plurality of subjects who have undergone vision correction, and performed on the plurality of subjects to be treated. A method of providing a learned predicted vision image based on at least one of the acquired vision correction surgery parameters and the measured values of eye characteristic data after surgery of the plurality of persons to be treated may be provided.

The solution means of the problem is not limited to the above-described solution means, and solutions that are not mentioned will be clearly understood by those of ordinary skill in the technical field to which the present invention pertains from the present specification and the accompanying drawings. .

According to an embodiment, a method for visually providing information related to vision correction to a doctor, a counselor, a subject, and the like, and an apparatus for performing the same may be provided.

According to another embodiment, a method for providing a predicted vision image to a doctor, a counselor, an examinee, and the like, and an apparatus for performing the same may be provided.

According to another embodiment, a method for providing a corneal topography image to a doctor, a counselor, a subject, and the like, and an apparatus for performing the same may be provided.

According to yet another embodiment, a method for visually providing a cause for calculating a prediction result to a doctor, a counselor, a subject, and the like, and an apparatus for performing the same may be provided.

The effects are not limited to the above-described effects, and effects that are not mentioned will be clearly understood by those of ordinary skill in the art from the present specification and the accompanying drawings.

1 is a diagram of a system for assisting vision correction according to an exemplary embodiment.
2 is a diagram for describing a learning device/prediction device according to an exemplary embodiment.
3 is a diagram illustrating a server device and a client device according to an embodiment.
4 is a diagram illustrating configurations of a server device and a client device according to an embodiment.
5 is a diagram of a model related to vision correction according to an exemplary embodiment.
6 is a diagram illustrating a learning step and a prediction step of a model related to vision correction according to an exemplary embodiment.
7 is a diagram illustrating pre-processing of input/output data according to an exemplary embodiment.
8 is a diagram of a model related to vision correction including submodels connected in series according to an exemplary embodiment.
9 is a diagram of a model related to vision correction including submodels connected in parallel according to an exemplary embodiment.
10 is a diagram of an expected vision image according to an exemplary embodiment.
11 is a diagram of a model for generating a predicted vision image using a filter according to an exemplary embodiment.
12 is a diagram of a corneal topography image according to an exemplary embodiment.
13 is a diagram illustrating a predictive result calculation cause analysis model including a model related to vision correction according to an exemplary embodiment.
14 is a diagram illustrating a reason for calculating a prediction result according to an exemplary embodiment.
15 is a diagram illustrating a serial connection of a model related to vision correction according to an exemplary embodiment.
16 to 19 are diagrams for calculating an output of a model related to vision correction based on an output of a model for predicting a corneal shape factor according to an exemplary embodiment.
20 is a diagram for calculating an output of a model related to vision correction based on an output of a model for predicting the necessity of a custom vision correction according to an embodiment.
21 to 22 are diagrams for calculating an output of a model related to vision correction based on the output of a model proposed for vision correction according to an exemplary embodiment.
23 is a diagram illustrating an output calculation of a model related to vision correction based on an output of a model suggesting a surgical parameter according to an exemplary embodiment.
24 to 25 are diagrams for calculating an output of a model related to vision correction based on the output of the vision prediction model according to an exemplary embodiment.
26 to 27 are diagrams for calculating an output of a model related to vision correction based on an output of a model for predicting a corneal topography image according to an exemplary embodiment.
28 is a diagram for describing a combination of three or more models related to vision correction according to an exemplary embodiment.
29 to 30 are diagrams for calculating an output of a model related to vision correction based on the output of the corneal shape factor prediction model and the vision prediction model, according to an exemplary embodiment.
31 is a diagram for explaining merging of models related to vision correction according to an exemplary embodiment.
32 to 34 are diagrams illustrating an implementation example of merging models related to vision correction according to an embodiment.
35 is a diagram illustrating a method for recommending vision correction according to a first embodiment.
36 is a diagram illustrating a second embodiment of a method for recommending vision correction according to an embodiment.
37 is a diagram of a third embodiment of a method for recommending vision correction according to an embodiment.
38 is a diagram of a fourth embodiment of a method for recommending vision correction according to an embodiment.
39 is a diagram illustrating a fifth embodiment of a method for recommending vision correction according to an embodiment.
40 is a diagram illustrating a sixth embodiment of a method for recommending vision correction according to an embodiment.
41 is a diagram illustrating a method of providing visualization information for vision correction according to a first embodiment.
42 is a diagram illustrating a second embodiment of a method of providing visualization information for vision correction according to an embodiment.
43 is a diagram of a third embodiment of a method of providing visualization information for vision correction according to an embodiment.
44 is a diagram illustrating a fourth embodiment of a method of providing visualization information for vision correction according to an embodiment.

The embodiments described in the present specification are intended to clearly explain the spirit of the present invention to those of ordinary skill in the technical field to which the present invention pertains, so the present invention is not limited by the embodiments described herein, and the present invention The scope of should be construed as including modifications or variations that do not depart from the spirit of the present invention.

The terms used in this specification have been selected as general terms currently widely used in consideration of functions in the present invention, but this varies according to the intention, custom, or the emergence of new technologies of those of ordinary skill in the technical field to which the present invention belongs. I can. However, if a specific term is defined and used in an arbitrary meaning unlike this, the meaning of the term will be separately described. Therefore, the terms used in the present specification should be interpreted based on the actual meaning of the term and the entire contents of the present specification, not a simple name of the term.

The drawings attached to the present specification are for easy explanation of the present invention, and the shapes shown in the drawings may be exaggerated and displayed as necessary to aid understanding of the present invention, so the present invention is not limited by the drawings.

In the present specification, when it is determined that a detailed description of a well-known configuration or function related to the present invention may obscure the subject matter of the present invention, a detailed description thereof will be omitted as necessary. In addition, numbers (eg, first, second, etc.) used in the process of the name of the present specification may mean different configurations, but may correspond to the same configuration.

Hereinafter, a method and apparatus for recommending vision correction to a subject based on examination data will be described. In particular, a method and apparatus for generating a model for recommending vision correction using artificial intelligence and for recommending vision correction to a subject using the generated model will be described.

In addition, a method and apparatus for providing visualized vision correction related information to a subject based on examination data will be described. In particular, using artificial intelligence to provide a predicted vision image after vision correction, predict a corneal topography image, or analyze the cause of the prediction result, create a model that provides vision correction information, and use the generated model to give the subject. A method and apparatus for providing visualization information for vision correction will be described.

In this specification, vision correction is an operation that corrects the vision of the person to be treated, and not only surgery to correct vision through corneal cutting using lasers such as LASIK, LASEK, and SMILE, but also lenses. It should be interpreted broadly, including implantation and other non-laser vision correction.

In addition, in the present specification, the visual acuity includes visual acuity that can be measured based on the judgment of the subject and the visual acuity that can be measured through an eye examination. For example, visual acuity can be measured through an eye chart. Or, visual acuity is higher-order aberrations, which are basic refractive abnormalities such as myopia, farsightedness, and astigmatism, and spherical aberration, coma aberration, and trefoil aberration. aberrations). In addition, visual acuity may include naked vision and corrected vision.

According to an aspect, in a method of providing a predicted vision image after vision correction using artificial intelligence performed by a computing device, the step of obtaining examination data of a subject-the examination data includes examination data and eye characteristic data measurement values- ; The predicted value of the eye characteristic data after vision correction of the examinee by inputting the first group data obtained from the examination data of the examinee into a first prediction model-The predicted value of the eye characteristic data includes at least one of an acuity predicted value and a corneal shape factor predicted value -Calculating; And generating a predicted visual acuity image based on the predicted value of the eye characteristic data; including, wherein the first predictive model includes a measurement value of eye characteristic data before surgery of a plurality of subjects who have undergone vision correction, and performed on the plurality of subjects to be treated. A method of providing a learned predicted vision image based on at least one of the acquired vision correction surgery parameters and the measured values of eye characteristic data after surgery of the plurality of persons to be treated may be provided.

Here, the step of generating the predicted vision image may include calculating or selecting a filter based on the predicted eye characteristic data, and applying the filter to the original image to generate the predicted vision image. .

Here, the eye characteristic data predicted value includes a first eye characteristic data predicted value and a second eye characteristic data predicted value, and the predicted visual acuity image includes a first predicted visual acuity image generated based on the first predicted eye characteristic data and the It is generated based on the predicted value of the second eye characteristic data, and may include a second predicted visual acuity image different from the first predicted visual acuity image.

Here, each of the first predicted eye characteristic data and the second predicted eye characteristic data corresponds to a predicted value of eye characteristic data after standard vision correction and a predicted value of eye characteristic data after custom vision correction, and the first predicted vision image and the second Each of the predicted visual acuity images may correspond to a predicted visual acuity image after standard vision correction and a predicted visual acuity image after custom vision correction.

Here, the predicted vision image may include information on at least one of clarity, light bleeding, contrast sensitivity, night vision, glare, double vision, and afterimages expected after vision correction of the subject.

Here, the method of providing the predicted visual acuity image further comprises: inputting the second group data obtained from the examination data of the subject into a second prediction model to predict the corneal topography image of the subject after vision correction surgery; 2 The predictive model may be learned based on at least one of a preoperative corneal topography image of a plurality of recipients who have undergone vision correction, a vision correction surgery parameter performed on the plurality of recipients, and a postoperative corneal topography image of the plurality of recipients. have.

Here, the method of providing the predicted visual acuity image may further include calculating a dependence of the predicted value of the eye characteristic data on the first group data.

Here, the eye characteristic data predicted value includes a first eye characteristic data predicted value and a second eye characteristic data predicted value, the dependence includes a dependence coefficient corresponding to at least a part of the first group data, and the dependence coefficient is And a first dependency coefficient corresponding to the predicted value of the first eye characteristic data and a second dependency coefficient corresponding to the predicted value of the second eye characteristic data, but different from the first dependency coefficient.

Here, each of the predicted value of the first eye characteristic data and the predicted value of the second eye characteristic data corresponds to a predicted value of eye characteristic data after standard vision correction and a predicted value of eye characteristic data after custom vision correction, and the first dependence coefficient and the second dependence Each of the coefficients may correspond to a dependency of a predicted value of eye characteristic data after standard vision correction on the first group data and a dependency of a predicted value of eye characteristic data after custom vision correction on the first group data.

Here, the method of providing the predicted vision image may further include outputting a dependency coefficient larger than a predetermined value among the dependency coefficients or outputting a predetermined number of dependency coefficients.

Here, the corneal shape factor predicted value may include at least one of an index of height decentration (IHD) predicted value, an index of surface variance (ISV) predicted value, and an index of vertical asymmetry (IVA) predicted value.

Here, the vision predicted value may include at least one of a predicted value of lower-order aberrations and a predicted value of higher-order aberrations.

According to another aspect, there is provided a method for recommending vision correction using artificial intelligence performed by a computing device, the method comprising: acquiring examination data of a subject, the examination data including examination data and eye characteristic data measurement values; Inputting first group data obtained from the examinee's examination data into a first prediction model to predict whether the examinee is suitable for vision correction; Predicting whether or not vision correction can be performed using a laser of the subject by inputting second group data obtained from the examinee's examination data into a second prediction model when the subject is suitable for vision correction; If the subject is capable of correcting vision using a laser, the third group data obtained from the examinee's examination data is input into a third predictive model to be used to determine whether a custom vision correction is necessary, and the subject's standard vision correction is followed. Calculating a corneal shape factor predicted value and a corneal shape factor predicted value after custom vision correction; And, when the subject can perform vision correction using a laser, inputting the fourth group data obtained from the examination data of the subject into a fourth prediction model to propose a vision correction procedure corresponding to the subject. 4 The predictive model provides a method for recommending vision correction, which is learned based on at least one of examination data of a plurality of subjects who have undergone vision correction, vision correction corresponding to the plurality of subjects, and vision after vision correction of the plurality of subjects. I can.

Here, in the proposing of the vision correction procedure, the vision correction procedure may be proposed based on the predicted vision value after the vision correction procedure of the subject.

Here, in the proposing of the vision correction procedure, the vision correction procedure may be proposed by calculating an acuity predicted value after the vision correction procedure of the subject corresponding to each of the plurality of vision correction procedures.

Here, in the step of proposing vision correction, vision correction may be proposed by calculating a plurality of predicted vision values corresponding to a plurality of different viewpoints.

Here, the method of recommending corrective vision may further include predicting whether the subject needs a custom vision correction based on the calculated predicted value of the corneal shape factor after standard vision correction and the predicted value of the corneal shape factor after custom vision correction.It may further include.

Here, in the step of proposing vision correction, vision correction may be proposed based on the calculated predicted value of the corneal shape factor after standard vision correction and the predicted value of the corneal shape factor after custom vision correction.

Here, when the fourth prediction model is learned in consideration of the preferences for vision correction of the plurality of persons to be treated, the fourth group data may include the preferences for vision correction of the subject.

Here, at least one of the first prediction model, the second prediction model, the third prediction model, and the fourth prediction model includes a plurality of submodels, and a result value is calculated based on the results of the plurality of submodels. Can be calculated.

Here, at least one of the first group data, the second group data, the third group data, and the fourth group data includes at least a portion of the examination data of the examinee as it is, or at least a portion of the examination data of the examinee. It may be characterized by including a new type of data calculated from.

Here, the corneal shape factor predicted value may include at least one of an index of height decentration (IHD) predicted value, an index of surface variance (ISV) predicted value, and an index of vertical asymmetry (IVA) predicted value.

Here, the examination data may further include gene information.

According to another aspect, there is provided a method for recommending vision correction using artificial intelligence performed by a computing device, the method comprising: acquiring examination data of a subject, the examination data including examination data and eye characteristic data measurement values; Inputting first group data obtained from the examinee's examination data into a first prediction model to predict whether the examinee is suitable for vision correction; Predicting whether or not vision correction using a laser of the subject is possible by inputting second group data obtained from the examinee's examination data into a second prediction model when the vision correction procedure is suitable for the examinee; Predicting whether the subject needs a custom vision correction by inputting the third group data obtained from the examination data of the subject into a third prediction model when the subject is capable of correcting vision using a laser; And when the subject is capable of correcting vision using a laser, inputting the fourth group data obtained from the examination data of the subject into a fourth prediction model to propose a vision correction procedure corresponding to the subject; including, but the custom The step of predicting the necessity of vision correction may include predicting the necessity of custom vision correction based on the predicted value of the corneal shape factor after the standard vision correction of the subject and the predicted value of the corneal shape factor after the custom vision correction, and the fourth prediction model includes: A method of recommending vision correction may be provided based on at least one of examination data of a plurality of persons undergoing correction, vision correction corresponding to the plurality of persons to be treated, and vision after vision correction of the plurality of persons to be treated.

According to another aspect, there is provided a method for recommending vision correction using artificial intelligence performed by a computing device, the method comprising: acquiring examination data of a subject, the examination data including examination data and eye characteristic data measurement values; Inputting first group data obtained from the examinee's examination data into a first prediction model to predict whether the examinee is suitable for vision correction; Predicting whether or not vision correction can be performed using a laser of the subject by inputting second group data obtained from the examinee's examination data into a second prediction model when the subject is suitable for vision correction; And when the subject is capable of correcting vision using a laser, inputting the third group data obtained from the examination data of the subject into a third prediction model to propose a vision correction procedure corresponding to the subject. The step of proposing correction may include proposing vision correction based on the predicted value of the corneal shape factor after standard vision correction of the subject and the predicted value of the corneal shape factor after custom vision correction, and the third prediction model includes a plurality of patients undergoing vision correction. A method for recommending a vision correction that is learned based on at least one of examination data of, vision correction corresponding to the plurality of persons to be treated, and vision after vision correction of the plurality of persons to be treated may be provided.

According to another aspect, there is provided a method for recommending vision correction using artificial intelligence performed by a computing device, the method comprising: acquiring examination data of a subject, the examination data including examination data and eye characteristic data measurement values; Inputting first group data obtained from the examinee's examination data into a first prediction model to predict whether the examinee is suitable for vision correction; And, when vision correction is appropriate for the subject, inputting the second group data obtained from the examination data of the subject into a second prediction model to propose a vision correction procedure corresponding to the subject; including, but suggesting the vision correction procedure. In the step of, the correction of vision is proposed based on the predicted value of the corneal shape factor after the standard vision correction of the subject and the predicted value of the corneal shape factor after the custom vision correction, and the second prediction model is the examination data of a plurality of subjects who have undergone vision correction. , A method of recommending a vision correction that is learned based on at least one of a vision correction procedure corresponding to the plurality of persons to be treated and visual acuity after the vision correction operation of the plurality of persons to be treated may be provided.

According to another aspect, there is provided a method for recommending vision correction using artificial intelligence performed by a computing device, the method comprising: acquiring examination data of a subject, the examination data including examination data and eye characteristic data measurement values; And inputting group data obtained from the examinee's examination data into a predictive model to propose a vision correction procedure corresponding to the examinee; including, wherein the step of proposing vision correction includes, the cornea after standard vision correction of the examinee A vision correction is proposed based on the shape factor predicted value and the corneal shape factor predicted value after custom vision correction, and the prediction model includes examination data of a plurality of subjects who have undergone vision correction, vision correction corresponding to the plurality of subjects, and the plurality of A method of recommending vision correction that has been learned based on at least one of the eyesight after vision correction of a person to be treated may be provided.

The vision correction assistance system according to an embodiment may include a learning device and a prediction device. Here, the learning device and the prediction device may be a computing device including at least one control unit. Examples of the computing device may include a desktop, a laptop, a tablet PC, and a smartphone, but are not limited thereto.

1 is a diagram of an assistive system for vision correction 10 according to an exemplary embodiment. Referring to FIG. 1, the learning apparatus 100 may learn and/or generate a model for calculating information related to vision correction based on the learning data (hereinafter referred to as “vision correction related model”). Here, the learning data is data necessary for learning and/or generating a model related to vision correction, such as numbers, letters, and images, and there is no limitation on the expression method thereof. For example, the learning data may include examination data and operation parameters of a person who has undergone vision correction.

The prediction device 300 may calculate a prediction result, which is information related to vision correction, based on the vision correction related model and input data generated through the learning device 100. Here, the input data is data that is the basis for calculating the prediction results, such as numbers, characters, and images, and there is no limitation on the expression method.

Specific examples of learning data, input data, prediction results, and models related to vision correction will be described later.

In FIG. 1, the learning device 100 and the prediction device 300 are illustrated as being separate devices, but the learning device 100 and the prediction device 300 may be the same device. For example, it is possible to learn/generate a model related to vision correction in the same device and calculate a prediction result using the model. Alternatively, at least a partial configuration of the learning device 100 and at least a partial configuration of the prediction device 300 may have the same configuration.

In addition, the vision correction assistance system according to an embodiment may include a plurality of learning devices and/or a plurality of prediction devices.

2 is a diagram for describing a learning device/prediction device according to an exemplary embodiment. Referring to FIG. 2, the learning device/prediction device according to an embodiment may include a memory unit 5000 and a control unit 1000.

The learning device/prediction device according to an embodiment may include a control unit 1000 for controlling its operation. The control unit 1000 includes at least one of a central processing unit (CPU), a random access memory (RAM), a graphic processing unit (GPU), one or more microprocessors, and other electronic components capable of processing input data according to predetermined logic. It may include.

The controller 1000 may read system programs and various processing programs stored in the memory unit 5000. For example, the controller 1000 may develop a data processing process for performing a learning and prediction step of a vision correction related model to be described later on the RAM, and may perform various processes according to the developed program. For example, the controller 1000 may learn a model related to vision correction. As another example, the controller 1000 may generate a prediction result using a model related to vision correction.

The learning device/prediction device according to an embodiment may include a memory unit 5000. The memory unit 5000 may store data necessary for learning, a learning model, and a learned model related to vision correction. The memory unit 5000 may store parameters and variables of a model related to vision correction.

The memory unit 5000 is a nonvolatile semiconductor memory, a hard disk, a flash memory, a RAM, a read only memory (ROM), an electrically erasable programmable read-only memory (EEPROM), or other tangible nonvolatile recording medium. It can be implemented as

The memory unit 5000 may store various processing programs, parameters for performing program processing, or processing result data. For example, the memory unit 5000 includes a data processing process program for performing the learning and prediction steps of a model related to vision correction to be described later, a diagnostic process program, parameters for executing each program, and data obtained according to the execution of these programs. (Eg, processed data or prediction results) and the like can be stored.

The learning device/prediction device according to an embodiment may further include a communication unit 9000. The communication unit 9000 may communicate with an external device. For example, the communication unit 9000 of the learning device may communicate with the communication unit 9000 of the prediction device. The communication unit 9000 may perform wired or wireless communication. The communication unit 9000 may perform bi-directional or one-way communication.

The learning device/prediction device illustrated in FIG. 2 is only an example, and the configuration of the learning device/prediction device is not limited thereto.

The vision correction assist system according to an embodiment may include a server device and a client device. 3 is a diagram for describing the server device 500 and the client devices 700a and 700b according to an exemplary embodiment.

The server device 500 according to an embodiment may correspond to the above-described learning device/prediction device. The server device 500 according to an embodiment may learn, store, and/or execute a model related to vision correction.

The client devices 700a and 700b according to an embodiment may correspond to the above-described learning device/prediction device. The client devices 700a and 700b according to an embodiment may learn, store, and/or execute a model related to vision correction.

The client devices 700a and 700b according to an embodiment may acquire a learned model related to vision correction from the server device 500. For example, the client devices 700a and 700b may download a model related to vision correction from the server device 500 through a network.

The server device 500 according to an embodiment may calculate a prediction result based on input data obtained from the client devices 700a and 700b. For example, the client devices 700a and 700b receive the subject's information and transmit it to the server device 500, and the server device 500 may calculate a prediction result through a vision correction-related model based on the subject's information. have. The server device 500 according to an embodiment may obtain input data from a plurality of client devices 700a and 700b.

The server device 500 according to an embodiment may transmit the calculated prediction result to the client devices 700a and 700b. For example, the client devices 700a and 700b may provide a prediction result obtained from the server device 500 to a doctor, a counselor, an examinee, and the like. The server device 500 according to an embodiment may obtain feedback from the client devices 700a and 700b. The server device 500 according to an embodiment may transmit a prediction result to a plurality of client devices 700a and 700b.

The client devices 700a and 700b according to an embodiment may request a prediction result from the server device 500.

The client devices 700a and 700b according to an embodiment may transmit the received data to the server device 500. The client devices 700a and 700b according to an embodiment may change at least a part of the received data and transmit it to the server device 500. The client devices 700a and 700b according to an embodiment may provide a prediction result obtained from the server device 500 to a doctor, a counselor, an examinee, and the like. The client devices 700a and 700b according to an embodiment may change at least a part of the prediction result obtained from the server device 500 and provide it to a doctor, a counselor, a subject, and the like.

3 illustrates a relationship between one server device 500 and two client devices 700a and 700b, but is not limited thereto. Can be applied.

4 is a diagram illustrating the configuration of the server device 500 and the client device 700 according to an embodiment. Referring to FIG. 4, the server device 500 and the client device 700 may include memory units 5000a and 5000b, control units 1000a and 1000b, and communication units 9000a and 9000b. The server device 500 and the client device 700 may transmit and acquire information through the communication units 9000a and 9000b. For example, the client device 700 may acquire a model related to vision correction that has been learned from the communication unit 9000a of the server device 500 through the communication unit 9000b. As another example, the client device 700 may transmit input data to the communication unit 9000a of the server device 500 through the communication unit 9000b, and the server device 500 transmits the prediction result to the client through the communication unit 9000a. It can be transmitted to the communication unit 9000b of the device 700.

As described above, the vision correction-related model is a model that calculates various information that can be considered during and before and after the vision correction procedure.

5 is a diagram of a model M related to vision correction according to an exemplary embodiment. Referring to Figure 5, the vision correction-related model (M) is a surgical suitability prediction model (M10), laser surgery availability prediction model (M11), corneal shape factor prediction model (M12), custom vision correction necessity prediction model (M13), Vision correction proposal model (M14), surgical parameter proposal model (M15), vision prediction model (M16), predicted vision image generation model (M17), corneal topography image prediction model (M18), and predictive result calculation cause analysis model (M19) It may include. A detailed description of each model will be described later.

At least some of the models M related to vision correction may be learned and/or generated in the same learning device. For example, the surgical fit prediction model M10 and the laser surgical fit prediction model M11 may be learned and/or generated in the same learning device. Alternatively, at least some of the models M related to vision correction may be learned and/or generated in different learning devices.

At least some of the models M related to vision correction may be executed in the same prediction device. For example, the surgical fit prediction model M10 and the laser surgical fit prediction model M11 may be executed in the same prediction device. Alternatively, at least some of the models M related to vision correction may be executed by different prediction devices.

The vision correction-related model according to an embodiment may be learned and/or implemented as an artificial intelligence model/algorithm, and the learning and/or implementation method thereof is not limited. For example, models related to vision correction include classification algorithms, regression algorithms, supervised learning, unsupervised learning, reinforcement learning, and support vector machine. ), decision tree, random forest, LASSO, AdaBoost, XGBoost, artificial neural network, etc. Can be.

The model related to vision correction according to an embodiment may be learned and/or generated through a learning step. A vision correction-related model that is learned and/or generated through the learning step may calculate a prediction result through the prediction step.

6 is a diagram for explaining a learning step (S10) and a prediction step (S30) of a model related to vision correction according to an exemplary embodiment. Referring to FIG. 6, the learning step S10 may include a learning data acquisition step S110 and a model training step S150.

The learning data acquisition step S110 may be a step in which the learning device acquires learning data, which is data for learning and/or generating a model related to vision correction.

The model learning step S150 may be a step of learning and/or generating a model related to vision correction. In the model learning step S150, the learning device may train and/or generate the model based on the training data acquired in the training data acquisition step. For example, model parameters constituting a model related to vision correction may be changed in the model learning step S150. According to the change of the parameter, the accuracy of the model may be improved.

The model related to vision correction may be learned based on information related to vision correction of a person who has undergone vision correction. For example, the model is learned based on at least one of examination data before vision correction of a plurality of patients undergoing vision correction, surgery parameters for vision correction performed on the plurality of patients, and examination data after vision correction of the plurality of patients. Can be.

Referring to FIG. 6, the prediction step S30 may include an input data acquisition step S310 and a model execution step S350.

The input data acquisition step S310 may be a step in which the prediction apparatus acquires input data, which is data that can be used to calculate a prediction result using a vision correction related model.

The model execution step S350 may be a step of calculating a prediction result based on the model and input data related to vision correction that are learned and/or generated in the learning step S10. For example, the prediction device may output a prediction result based on input data and model parameters obtained from the learning device.

The input data of the model related to vision correction may include all information acquired about the subject. Alternatively, the input data may include at least some of the information acquired about the subject. In addition, the input data may be the same or different according to a model related to vision correction.

The prediction result of the model related to vision correction may vary according to the learning data. Alternatively, the accuracy of the model related to vision correction may vary according to the learning data.

The prediction result may vary depending on the type of true value included in the training data. For example, the prediction result may vary depending on whether the true value considers the subject's subjective intention. Here, the subjective doctor of the subject may include a preference, such as whether he prefers a specific vision correction procedure, and an ability to pay a cost, such as how much money can be paid for the vision correction procedure. If the subjective intention of the subject is not considered, the predicted result may be a medically suggested result. On the other hand, when the subjective intention of the subject is considered, the predicted result may be the same as the medically suggested result, but may be different.

When the first learning data and the second learning data are obtained from different hospitals, from different doctors, or are learning data at different times, the first model and the first model that are created and/or learned based on the first learning data 2 The second models trained and/or generated based on the training data may be different from each other. As a result, the first prediction result output from the first model and the second prediction result output from the second model may be different from each other. In addition, the accuracy of the first model and the second model may be different from each other.

However, even if the learning data are different, the same model related to vision correction may be learned and/or generated. Alternatively, the same prediction result may be calculated even in the case of a model related to vision correction that is learned and/or generated from different learning data.

Learning data, input data, and prediction results (hereinafter referred to as "input/output data") may include variables such as examination data and surgical parameters. Examination data and surgical parameters can be expressed in various ways such as numbers, letters, images, etc., and there is no limitation on the expression method. In addition, images are not limited in their dimensions, such as 2D or 3D images.

The examination data may include questionnaire data, which is information obtained by asking a question, not through equipment or examination, eye characteristic data, and genetic information, which are information about the eye obtained through equipment or examination.

The questionnaire data includes variables such as gender, age, race, living environment, occupation, income, educational background, education, and family size of the subject, medical history and family history such as hypertension and diabetes, and preference for vision correction. Can include.

Eye characteristic data can include all kinds of eye-related information. For example, the eye characteristic data may include variables such as visual acuity, intraocular pressure, and/or retinal test results.

The eyeball characteristic data may include information on the physical shape of the eyeball. For example, eye characteristic data includes white-to-white (WTW) distance, angle-to-angle (ATA) distance, internal Anterior Chamber Depth (ACD), sulcus-to-sulcus (STS) distance, and pupil size. It can contain variables such as. As another example, the eye characteristic data may include information on the shape of the cornea, such as a corneal shape factor, a corneal topography image, and the like. Here, the corneal shape factor is a number representing the physical shape of the cornea, index of surface variance (ISV), index of vertical asymmetry (IVA), keratoconus index (KI), central keratoconus index (CKI), minimum radius of curvature (Rmin ), index of height asymmetry (IHA), index of height decentration (IHD), central cornea thickness, and the like. In addition, the corneal topography image is an image related to the shape of the cornea, and may include a corneal topography map, an front corneal curvature image, a corneal posterior curvature image, a corneal thickness map, and the like.

Eye characteristic data can be obtained through tomography such as Pentacam, CASIA2, AL-Scan, OQAS, topography, optical coherence tomography (OCT), ultrasound biomicroscopy (UBM) equipment, etc., but is not limited thereto. Information that can be obtained through the above equipment and similar equipment and information derived therefrom may be included in the eye characteristic data.

Genetic information may be obtained through genetic testing or the like. Genetic information can be used to determine whether or not vision correction is suitable for a subject. Alternatively, the genetic information may be used to predict side effects after vision correction surgery. For example, the onset of corneal dystrophy can be predicted through genetic information.

Surgical parameters are variables related to the performance of vision correction, such as LASIK, LASEK, Smile, lens implantation, standard vision correction, custom vision correction, type of vision correction, corneal flap thickness, flap diameter, flap. Variables that can be changed during vision correction, such as side cut angle, corneal cutting profile, eye suction ring operation time, optical zone, hinge position, hinge structure such as hinge angle and hinge width, can be included.

Standard vision correction may mean vision correction for correcting low-order aberrations, and custom vision correction may mean vision correction for correcting low-order and high-order aberrations.

The types of surgical parameters to be considered may differ depending on the operator of standard vision correction and custom vision correction. For example, the types of surgical parameters considered according to the person undergoing custom vision correction may include the types of the surgical parameters considered according to the person undergoing standard vision correction.

The number of surgical parameters that change in response to the operator of standard vision correction may be different from that of custom vision correction. Standard vision correction and custom vision correction can be classified according to the number of surgical parameters that change in response to the person to be treated. The number of surgical parameters (hereinafter referred to as "reference values") that change in response to the operator who distinguishes standard vision correction and custom vision correction may be a predetermined value. For example, when the number of surgical parameters that change in response to the patient to be treated is greater than or equal to a reference value, custom vision correction may be performed, and when the number of surgical parameters is smaller than the reference value, standard vision correction may be performed.

Standard vision correction and custom vision correction may vary depending on the hospital, doctor, and/or timing performing the vision correction procedure. For example, the reference value may vary depending on the hospital, the doctor and/or the timing.

Custom vision correction may include contour vision surgery and wavefront surgery.

When performing custom vision correction, the quality of vision may be improved compared to the case of performing standard vision correction. The quality of vision is a term that comprehensively refers to the good or bad of vision, as well as the acuity, low and high aberration measured through the acuity chart, as well as the clarity of vision, light bleeding, contrast sensitivity, night vision, glare, double vision, afterimages, It can be judged based on other discomfort.

The input/output data may include not only a measured value that can be obtained/measured through a medical examination and a medical examination, but also a predicted value that can be calculated through a model related to vision correction. For example, the eye characteristic data may include not only the measured value of the eye characteristic data of the subject acquired/measured through an examination before vision correction, but also the predicted value of the eye characteristic data after vision correction of the subject calculated through a model related to vision correction. have.

Input/output data may be input to a model related to vision correction through pre-processing. For example, the learning device may pre-process the acquired input/output data and then use it for learning a model related to vision correction. Alternatively, the prediction device may pre-process the acquired input/output data and then input it into a model related to vision correction to calculate a prediction result.

The preprocessing should be broadly interpreted to include all changes applied to the input/output data, and is not limited to the examples disclosed herein.

The preprocessing may include selecting at least some of the variables included in the input/output data, such as feature selection. For example, preprocessing may include t-test, Gini index, information gain, relief, DistAUC, signal to noice, MRMR, Fisher score, Laplacian score, SPEC, and the like.

The preprocessing may include creating a new variable from at least some of the variables included in the input/output data, such as feature extraction. For example, the preprocessing may include principle component analysis, linear discriminant analysis, canonical correlation analysis, singular value decomposition, ISOMAP, locally linear embedding, and the like. As another example, pre-processing may include generating a spectrum from the numerical value or generating an image from the numerical value.

The pre-processing may include a processing method when a variable required by a vision correction-related model (or a learning device and/or a prediction device) is not included in the input/output data, such as processing a missing value. For example, the preprocessing may include treating the missing value as the average of the corresponding variable or treating it as the mode.

Through pre-processing, the accuracy of models related to vision correction can be improved. For example, the accuracy of a model related to vision correction including the feature selection step may be better than that of a model not including it. As another example, the accuracy of a vision correction-related model that generates an image from a numerical value and calculates a prediction result based on the image may be better than that of a vision correction-related model that calculates a prediction result based on the numerical value.

7 is a diagram illustrating pre-processing of input/output data according to an exemplary embodiment. Referring to FIG. 7, input/output data may be input to a model related to vision correction through a pre-processing step (S500). The vision correction-related model may calculate a prediction result based on the pre-processed input/output data (S700).

The model related to vision correction may include a plurality of sub-models (sub-models). The plurality of sub-models may calculate a prediction result based on the input data.

The model related to vision correction may include submodels connected in series and/or in parallel, such as an ensemble model.

The model related to vision correction may include a plurality of submodels connected in series. Here, the fact that the sub-models are connected in series includes calculating the output of at least one sub-model based on the output of at least one sub-model, such as the output of the first sub-model being the input of the second sub-model. can do. Alternatively, the fact that the sub-models are serially connected may include sequentially passing through a plurality of sub-models in order to obtain a prediction result from the input data through a model related to vision correction.

8 is a diagram of a model M related to vision correction including submodels M1 and M2 connected in series according to an exemplary embodiment. Referring to FIG. 8, the model M related to vision correction may include a first sub-model M1 and a second sub-model M2 connected in series. The first sub-model M1 may calculate an output based on input/output data, and the second sub-model M2 may calculate a prediction result based on the output of the first sub-model M1.

The model related to vision correction may include a plurality of submodels connected in parallel. Here, the fact that the sub-models are connected in parallel may include that the output of the sub-model does not affect the output of the other sub-model, such as that the output of the first sub-model does not depend on the output of the second sub-model.

Inputs of a plurality of submodels connected in parallel may be the same. Alternatively, inputs of a plurality of sub-models connected in parallel may be different. For example, the first examination data input to the first sub-model may be different from the second examination data input to the second sub-model.

The first examination data may include at least some other variables compared to the second examination data. For example, the first examination data may include a preference for vision correction, but the second examination data may not include a preference for vision correction.

The first checkup data includes the same type of variable as the second checkup data, but may have different values. For example, the first examination data and the second examination data include corneal thickness, but the method of obtaining the numerical value is different (for example, corneal thickness is measured with different devices), so the numerical value may be different. have.

The vision correction-related model may include an output sub-model that calculates an output based on outputs of a plurality of sub-models connected in parallel. For example, when outputs of a plurality of sub-models are the same, the output sub-model may provide the same output. As another example, when the outputs of the plurality of sub-models are different, the output sub-model may output a result of considering the outputs of the plurality of sub-models at a predetermined ratio or provide a specific output among the plurality of outputs. As another example, the output sub-model may output a result generated based on the output of a plurality of sub-models.

9 is a diagram of a model M related to vision correction including submodels M1 and M2 connected in parallel according to an exemplary embodiment. Referring to FIG. 9, the model M related to vision correction may include a first sub-model M1 and a second sub-model M2 connected in parallel. In addition, the vision correction-related model M may include an output sub-model M3 that calculates an output based on the outputs of the first sub-model M1 and the second sub-model M2. The first sub-model M1 and the second sub-model M2 each calculate a first output and a second output based on input/output data, and the output sub-model M3 is based on the first output and the second output. Prediction results can be calculated.

An example of a model related to vision correction including sub-models connected in parallel may be an ensemble, but is not limited thereto.

Hereinafter, individual examples of models related to vision correction will be described.

The surgical suitability prediction model can predict whether vision correction is appropriate or inappropriate for a subject. The surgical suitability prediction model may predict surgical suitability based on input data.

The suitability of vision correction can mean the medical suitability. Accordingly, the input data of the surgical suitability prediction model may not include the subject's preference for vision correction. Alternatively, the surgical suitability prediction model may predict the operation suitability without considering the patient's preference for vision correction surgery.

Surgical suitability may include whether surgery is possible and whether surgery is necessary. As an example, the surgical suitability prediction model may determine that a subject who does not need surgery because of good vision is unsuitable for surgery. As another example, the surgical suitability prediction model may determine that a subject who is unable to increase visual acuity through vision correction is unsuitable for surgery.

The surgical suitability may be output as a fit/unsuitable operation. Alternatively, the surgical suitability may be output by quantifying or visualizing the suitability for surgery.

The laser surgery prediction model can predict whether a subject can or cannot correct vision using a laser. Here, the vision correction using a laser may mean a surgery for correcting vision by cutting the cornea using a laser. The laser surgery availability prediction model may predict laser surgery availability based on input data.

Whether or not to correct vision using a laser may mean medical or not. Accordingly, the input data of the laser surgery possibility prediction model may not include the subject's preference for vision correction. Alternatively, the laser surgery availability prediction model may predict the laser surgery availability without considering a subject's preference for vision correction surgery.

The laser operation availability may be output as laser operation enabled/disabled. Alternatively, the possibility of laser surgery may be output by quantifying or visualizing the possibility of surgery.

The corneal shape factor prediction model can predict the corneal shape factor after vision correction of the subject. The model can predict one or more corneal shape factors. The corneal shape factor prediction model can predict the corneal shape factor based on input data.

The input data of the corneal shape factor prediction model may include surgical parameters. For example, the input data may include the type of surgery as a surgery parameter. The model may predict a corneal shape factor corresponding to the input surgical parameter. For example, when the input data includes LASIK, LASEC, and Smile as surgical parameters, the output of the model may include a corneal shape factor predicted value after LASIK, a corneal shape factor predicted value after LASIK, and a corneal shape factor predicted value after smile. As another example, when the input data includes standard vision correction and custom vision correction as surgical parameters, the output of the model may include a corneal shape factor predicted value after standard vision correction and a corneal shape factor predicted value after custom vision correction.

The corneal shape factor prediction model can predict a corneal shape factor corresponding to a predetermined surgical parameter regardless of whether the input data includes a surgical parameter. For example, if the model is trained to predict the corneal shape factor after standard vision correction and the corneal shape factor after custom vision correction, even if the input data of the model does not include surgical parameters, the model is the corneal shape factor after standard vision correction. The predicted value and the predicted value of the corneal shape factor after custom vision correction can be output.

Table 1 is an output of a corneal shape factor prediction model according to an embodiment. Referring to Table 1, the corneal shape factor prediction model may output IHD, ISV, and IVA values of a subject. In addition, the corneal shape factor prediction model includes the subject's IHD, ISV, and IVA measurements obtained before vision correction, the predicted IHD, ISV, and IVA values after standard vision correction, and the predicted IHD, ISV, and IVA values after custom vision correction. can do.

Item Now Expected after standard vision correction Expected after custom vision correction IHD 0.015 0.021 0.018 ISV 20.0 27.8 25.0 IVA 0.19 0.23 0.22

The model for predicting the need for custom vision correction can predict whether a subject needs custom vision correction. The custom vision correction need prediction model may predict the need for custom vision correction based on the input data.

The model for predicting the need for custom vision correction can predict the need for custom vision correction without considering the patient's preference for vision correction. Alternatively, the custom vision correction need prediction model may predict the need for custom vision correction in consideration of the subject's vision correction preference. Depending on whether the subject's preference for vision correction is considered, the output of the model for predicting the need for custom vision correction may vary.

The necessity prediction model for custom vision correction may determine whether a subject needs custom vision correction based on eye characteristics data such as corneal shape factors and corneal topography images.

The custom vision correction need prediction model may determine the need for custom vision correction based on the absolute value of the corneal shape factor. For example, a model for predicting the need for a custom vision correction can predict that a subject needs a custom vision correction if the corneal shape factor is outside a certain range.

The custom vision correction need prediction model may determine the need for custom vision correction based on the relative values of the corneal shape factor. For example, the necessity prediction model for custom vision correction may compare the difference between the corneal shape factor after standard vision correction and the corneal shape factor after custom vision correction to determine the need for custom vision correction.

The vision correction proposal model can propose a vision correction procedure corresponding to the subject. For example, the model may output one vision correction procedure. As another example, the model may output a plurality of vision correction techniques. As another example, the model may output a plurality of vision correction procedures together with information on their priorities.

The vision correction proposal model may propose vision correction based on the input data.

Vision correction corresponding to the subject may refer to a vision correction that is calculated by a model proposed for vision correction without considering the patient's preference for vision correction. Alternatively, the vision correction procedure corresponding to the subject may refer to a vision correction procedure calculated by the vision correction proposal model in consideration of the examinee's preference for vision correction. Depending on whether the subject's preference for vision correction is considered, the output of the vision correction proposal model may vary.

The output of the vision correction proposal model may be that the necessity of a custom vision correction procedure is considered. For example, the output may be a standard LASIK, a custom LASIK, a standard LASEK, a custom LASEK, a standard smile, a custom smile, and the like, taking into account the necessity of custom vision correction.

Alternatively, the output of the vision correction proposal model may be that the necessity of a custom vision correction procedure is not considered. For example, the output may be one that does not take into account the necessity of custom vision correction such as LASIK, LASEK, Smile, and lens implantation.

The vision correction proposal model may suggest vision correction based on the quality of vision after vision correction. For example, the model may propose vision correction based on a predicted vision value after vision correction corresponding to a plurality of vision correction techniques.

Table 2 is an output of a model proposed for vision correction according to an embodiment. Referring to Table 2, the proposed model for vision correction may output LASIK, LASEK, and Smile surgery along with priority information such as fitness. In the case of Table 2, values corresponding to smiles are larger than values corresponding to LASIK and LASEK, which may mean that the vision correction proposal model suggests smile as the first priority. In addition to the method of Table 2, information on the priority may be output in various ways to indicate the priority.

Types of vision correction Goodness of fit Lasik 15.45% Lasek 13.30% smile 71.25%

The surgical parameter proposal model can suggest surgical parameters. The model may suggest one or more surgical parameters. The surgical parameter suggestion model can suggest surgical parameters based on input data.

The input data of the surgical parameter proposal model may include data in the form of an image. Here, the image may be an image obtained through measurement and/or inspection using equipment, such as a corneal topography map. Alternatively, the image may be an image generated through interpolation, extrapolation, or artificial intelligence based on the measured value. For example, the image may be an image generated from a corneal shape factor.

The surgical parameter suggestion model may have better results after vision correction surgery performed on the basis of the surgery parameters proposed based on the image than the surgery results after vision correction surgery performed on the basis of the surgical parameters proposed based on the numerical values. For example, the surgical results after vision correction performed based on the surgical parameters suggested by the model based on the corneal topography image are the vision correction performed based on the proposed surgical parameters based on corneal shape factors such as IHD, ISV, and IVA. It may be better than the results of the surgery afterwards. Here, the surgery result may mean the quality of vision after surgery. Alternatively, the surgical result may mean the shape of the cornea after surgery.

The visual acuity prediction model can predict the visual acuity of the subject after corrective vision. The vision prediction model may output a vision prediction value based on input data. The model can predict more than one vision.

Input data of the vision prediction model may include surgical parameters. For example, the input data may include the type of surgery as a surgery parameter. The model may predict visual acuity corresponding to the input surgical parameter. For example, when the input data includes LASIK, LASEK, and Smile as surgical parameters, the output of the model may include a vision prediction value after LASIK, a vision prediction value after LASIK, and a vision prediction value after Smile. As another example, when the input data includes standard vision correction and custom vision correction as surgical parameters, the output of the model may include a vision predicted value after standard vision correction and a vision predicted value after custom vision correction.

The visual acuity prediction model can predict visual acuity corresponding to a predetermined surgical parameter regardless of whether the input data includes a surgical parameter. For example, if the model is trained to predict visual acuity after standard vision correction and visual acuity after custom vision correction, even if the input data of the model does not include surgical parameters, the model will be the predicted visual acuity value after standard vision correction and after custom vision correction. A vision predicted value can be output.

The vision prediction model may predict vision corresponding to a plurality of different visions. For example, the model may predict visual acuity corresponding to the first time and the second time after vision correction. Examples of a plurality of different views include one day, one week, one month, six months, and one year after vision correction, but are not limited thereto.

The vision prediction model may predict a visual acuity recovery rate of a subject based on vision prediction values corresponding to a plurality of different views. For example, the model may predict a visual acuity recovery rate based on the first time and the second time after vision correction.

The predicted visual acuity image generation model can predict the visual field of the subject after corrective vision. The model may generate an image (hereinafter referred to as "expected visual acuity image") visualizing the quality of visual acuity of the subject after corrective vision. As the model outputs the predicted vision image, it is possible to more easily explain the vision correction procedure to the subject. As the model visualizes and outputs the visual field after corrective vision, the subject can more clearly understand the expected result after the corrective vision, and through this, it is possible to receive help in selecting the corrective vision.

The input data of the predicted vision image generation model may include surgical parameters. For example, the input data may include the type of surgery as a surgery parameter. The model may predict an expected visual acuity image corresponding to the input surgical parameter. For example, when the input data includes LASIK, LASEK, and Smile as surgical parameters, the output of the model may include an expected vision image after LASIK, an expected vision image after LASEK, and an expected vision image after Smile. As another example, when the input data includes standard vision correction and custom vision correction as surgical parameters, the output of the model may include an expected vision image after standard vision correction and a predicted vision image after custom vision correction. As another example, when the input data includes a plurality of optical zones, the output of the model may include a plurality of predicted visual acuity images corresponding to the plurality of surgical ranges.

The predicted visual acuity image generation model may generate a predicted visual acuity image corresponding to a predetermined surgical parameter regardless of whether the input data includes the surgical parameter. For example, if the model is trained to generate a predicted visual acuity image after standard vision correction and a predicted visual acuity image after custom vision correction, even if the input data of the model does not include a surgical parameter, the model is a predicted visual acuity image after standard vision correction. And a predicted vision image after the custom vision correction procedure may be output.

The predicted vision image may include information on at least one of clarity, light bleeding, contrast sensitivity, night vision, glare, double vision, and afterimages expected after vision correction of the subject.

10 is a diagram of an expected vision image according to an exemplary embodiment. Referring to FIG. 10A, the predicted vision images I1 and I2 may be expressed by visualizing information on the clarity of vision. Referring to FIG. 10B, the predicted visual acuity images I3, I4, I5, and I6 may be expressed by visualizing information on light bleeding.

The plurality of predicted vision images may correspond to different surgical parameters. Referring to FIG. 10A, the first predicted visual acuity image I1 may correspond to the clarity of visual acuity after a custom vision correction, and the second predicted visual acuity image I2 may correspond to the clarity of visual acuity after the standard visual acuity correction. Referring to FIG. 10B, the third to sixth predicted visual acuity images I3-I6 may be predicted visual acuity images corresponding to different optical zones. For example, the surgical range of the third predicted visual acuity image I3 may be larger than that of the fourth to sixth predicted visual acuity images I4-I6.

The predicted vision image according to an embodiment may be generated through filtering using a filter. Here, filtering is a concept used in a general image processing field, and may mean generating a filtered image by convolving an image and a filter. Examples of filters include an average filter, a weighted average filter, a low-pass filter, a Gaussian filter, a median filter, and a bilateral filter. ), a blurring filter, a high-pass filter, an unsharp masking, a high-boost filter, a sharpening filter, etc. It is not limited thereto.

11 is a diagram of a predicted vision image generation model M17 using a filter according to an exemplary embodiment. Referring to FIG. 11, the predicted vision image generation model M17 may include a first submodel M171 and a second submodel M172.

The first sub-model M171 may calculate and/or select a filter based on input data. For example, the input data may include a predicted value of eye characteristic data after vision correction of the subject, and the first sub-model M171 may calculate and/or select a filter based on the predicted value of the eye characteristic data. As another example, the input data includes a measurement value of eye characteristic data and a surgical parameter of the subject, and the first sub-model M171 may calculate and/or select a filter based on the measurement value of the eye characteristic data and the operation parameter. .

The second sub-model M172 may generate a predicted vision image based on a filter calculated and/or selected by the first sub-model M171. For example, the second sub-model M172 may generate a predicted vision image by applying the filter to an original image. Here, the original image is an image that serves as a basis for generating the predicted eyesight image, and may be input from the outside of the predicted eyesight image generating model M17 or included in the model M17.

The corneal topography image prediction model can predict the corneal topography image after vision correction of the subject. The model can predict one or more corneal topographic images. The corneal topography image prediction model may generate a corneal topography image based on input data.

The input data of the corneal topography image prediction model may include surgical parameters. For example, the input data may include the type of surgery as a surgery parameter. The model can predict the corneal topography image corresponding to the input surgical parameter. For example, when the input data includes LASIK, LASEK, and Smile as surgical parameters, the output of the model may include a corneal topography image after LASIK, a corneal topography image after LASEK, and a corneal topography image after Smile. As another example, when the input data includes standard vision correction and custom vision correction as surgical parameters, the output of the model may include a corneal topography image after standard vision correction and a corneal topography image after custom vision correction.

The corneal topography image prediction model can generate a corneal topography image corresponding to a predetermined surgery parameter regardless of whether the input data includes the surgery parameter. For example, if the model is trained to generate a visual acuity after standard vision correction and a corneal topography image after a custom vision correction, even if the input data of the model does not include a surgical parameter, the model is a corneal topography image and custom after standard vision correction. Corneal topography images can be output after vision correction surgery.

The input data of the corneal topography image prediction model may include a corneal topography image of the subject measured before vision correction. 12 is a diagram of a corneal topography image according to an exemplary embodiment. Referring to FIG. 12, the corneal topography image prediction model M18 may predict the corneal topography image CI2 of the subject after vision correction based on the corneal topography image CI1 of the subject measured before vision correction. In FIG. 12, only the corneal topography image CI1 is input to the model M18, but other input data may be input together.

The predictive result calculation cause analysis model may analyze the cause of the prediction result generated by the vision correction-related model. The model may calculate dependence on input data of a model related to vision correction. Here, the dependence may include how a specific variable of the input data has had an effect on the prediction result.

The prediction result calculation cause analysis model may output the prediction result calculation cause. The model may output one or more reasons for calculating the prediction results. Hereinafter, for convenience of explanation, a case in which the cause of the prediction result is expressed as a numerical value such as the dependence coefficient is described, but the cause of the prediction result is not limited to this, and the expression method such as numerical value, image, text, and combinations thereof is not limited. none.

The predictive result calculation cause analysis model may output at least a part of the dependence coefficient. For example, the model may output all of the calculated dependence coefficients.

The predictive result calculation cause analysis model can output a dependence coefficient that falls within a certain range of the calculated dependence coefficients. For example, the model may output a dependence coefficient greater than a predetermined value among the calculated dependence coefficients. Alternatively, the model may output the calculated dependence coefficient when the absolute value is greater than a predetermined value.

The predictive result calculation cause analysis model can output a certain number of dependence coefficients. For example, the model may output a predetermined number of dependence coefficients.

Predictive result calculation cause analysis model is surgical suitability predictive cause analysis model, laser surgery predictive cause analysis model, corneal shape factor predictive cause analysis model, custom vision correction necessity predictive cause analysis model, vision correction suggested cause analysis model, surgical parameter suggested cause It may include an analysis model, a vision prediction cause analysis model, a predicted vision image generation cause analysis model, and a corneal topography image prediction cause analysis model. For example, the vision prediction cause analysis model may calculate how a variable included in the input data of the vision prediction model has an influence and that the vision prediction model predicts vision. Alternatively, the vision prediction cause analysis model may calculate the dependence of the vision prediction value calculated by the vision prediction model on input data of the vision prediction model.

The predictive result calculation cause analysis model may include at least some of the models related to vision correction. For example, the vision prediction cause analysis model may include a vision prediction model.

13 is a diagram of a predictive result calculation cause analysis model M19 including a vision correction related model M192 according to an exemplary embodiment. Referring to FIG. 27, the prediction result calculation cause analysis model M19 may include an input data disturbance model M191, a vision correction related model M192, and a prediction result analysis model M193.

The input data disturbance model M191 may output disturbed input data based on the input data. The model M191 may output one or more disturbed input data. Here, perturbation of input data may mean changing input data, such as changing values of at least some variables included in the input data. For example, when the variable is a numerical value, disturbance may mean increasing or decreasing the value. Alternatively, when the variable is an image, disturbance may mean increasing or decreasing a pixel value of at least some pixels of the image.

The vision correction related model M192 may output a first prediction result corresponding to the input data and a second prediction result corresponding to the disturbed input data based on the input data and the disturbed input data. For example, when the vision correction-related model M192 included in the prediction result calculation cause analysis model M19 is a vision prediction model, the first prediction result and the second prediction result may be different prediction values. In addition, when there are a plurality of disturbed input data, the model may output a plurality of prediction results corresponding to the plurality of disturbed input data.

The prediction result analysis model M193 may output a reason for calculating the prediction result based on the prediction result. For example, the model M193 may calculate the reason for calculating the prediction result based on a difference between the first prediction result calculated from the undisturbed input data and the second prediction result calculated from the disturbed input data. Specifically, the dependence of the first prediction result on the first variable may be calculated based on a difference between the first prediction result and the second prediction result calculated from the input data in which the first variable is disturbed.

14 is a diagram for a cause of calculating a prediction result according to an exemplary embodiment, and in more detail, a diagram for a cause for predicting vision. Referring to FIG. 14, the reason for calculating the prediction result may be expressed by numerical values and images. Characters such as Cornea_Back_Rmin(V1), Astigmatism(V2), Mono(V3), Nearsightedness(V4), Op_flag(V5) and Pupil_Dia(V6) shown in FIG. 14 may correspond to variables included in the input data. 5.95(N1), -0.5(N2), 0(N3), -1.37(N4), 2(N5), 3.1(N6), etc.expressed to correspond to the above characters correspond to the values of variables included in the input data. Can be.

The dependence of prediction results on variables can be visualized and expressed. For example, the dependence may be expressed by the length, color, and direction of the arrow. Referring to FIG. 14, the length of the arrow may correspond to the absolute value of the dependence coefficient. Also, the direction and color of the arrow may correspond to the sign of the dependence coefficient. 14 shows the predicted visual acuity value (OV) is 1.18, Cornea_Back_Rmin (V1), Astigmatism (V2), Mono (V3) and Nearsightedness (V4) have a positive effect on the predicted visual acuity, and Op_flag (V5) and Pupul_Dia (V6) are It can be interpreted as having a negative effect on the predicted visual acuity.

An example of the model for analyzing the cause of the prediction result calculation may include a Local Interpretable Model-agnostic Explanations (LIME), but is not limited thereto.

Models related to vision correction can be combined with each other. The models may be combined by at least one of a serial connection and a parallel connection.

The fact that the vision correction-related models are serially connected may include calculating an output of at least one other vision correction-related model based on the output of the at least one vision correction-related model.

The fact that the vision correction-related models are connected in parallel may include that the output of the vision correction-related model does not affect the output of the other vision correction-related models.

Hereinafter, an example of a combination of models related to vision correction will be described.

15 is a diagram illustrating a serial connection of a model related to vision correction according to an exemplary embodiment. Referring to FIG. 15, the first vision correction-related model Ma and the second vision correction-related model Mb may be combined by serial connection. The first vision correction-related model Ma may calculate a first prediction result based on the first input data and the second prediction result output by receiving the second input data from the second vision correction-related model Mb. have.

16 to 19 are diagrams for calculating an output of a model related to vision correction based on an output of a model for predicting a corneal shape factor according to an exemplary embodiment.

Referring to FIG. 16, the custom vision correction necessity prediction model M13 is a custom vision correction procedure of the subject based on the corneal shape factor and the first input data that the corneal shape factor prediction model M12 receives and outputs the second input data. The need can be predicted.

Referring to FIG. 17, the vision correction proposal model M14 performs vision correction corresponding to the subject based on the corneal shape factor and the first input data that the corneal shape factor prediction model M12 receives and outputs second input data. I can suggest.

Referring to FIG. 18, the predicted visual acuity image generation model M17 is predicted after vision correction of the subject based on the corneal shape factor and first input data that the corneal shape factor prediction model M12 receives and outputs second input data. It is possible to create a vision image.

Referring to FIG. 19, the corneal topography image prediction model M18 is based on the corneal shape factor and the first input data that the corneal shape factor prediction model M12 receives and outputs the second input data. Terrain images can be predicted.

20 is a diagram for calculating an output of a model related to vision correction based on an output of a model for predicting the necessity of a custom vision correction according to an embodiment. Referring to FIG. 20, the vision correction suggestion model M14 is a vision corresponding to a subject based on the need for custom vision correction and the first input data output by the custom vision correction necessity prediction model M13 input and output second input data. Orthodontics can be suggested.

21 to 22 are diagrams for calculating an output of a model related to vision correction based on the output of a model proposed for vision correction according to an exemplary embodiment.

Referring to FIG. 21, the corneal shape factor prediction model M12 is based on the vision correction procedure and the first input data that the vision correction proposal model M14 receives and outputs the second input data. Factors can be predicted.

Referring to FIG. 22, the vision prediction model M16 calculates an acuity prediction value after the vision correction procedure of the subject based on the vision correction procedure and the first input data that the vision correction proposal model M14 receives and outputs the second input data. can do.

The predicted visual acuity image generation model may generate a predicted visual acuity image of the subject after the corrective vision based on the corrective surgery and the first input data obtained by receiving the second input data and outputting the second input data.

The corneal topography image prediction model may predict the corneal topography image of the subject's vision correction surgery based on the vision correction surgery and the first input data output by the vision correction proposal model receiving the second input data.

23 is a diagram illustrating an output calculation of a model related to vision correction based on an output of a model suggesting a surgical parameter according to an exemplary embodiment.

Referring to FIG. 23, the vision prediction model M16 calculates a vision predicted value after corrective vision of the subject based on the surgical parameters and the first input data output by the surgery parameter suggestion model M15 receiving the second input data. I can.

The corneal shape factor prediction model may predict the corneal shape factor after vision correction of the subject based on the surgical parameter and the first input data output by receiving the second input data from the surgical parameter proposal model.

The predicted visual acuity image generation model may generate a predicted visual acuity image after corrective vision of the subject based on the first input data and the surgical parameter that the surgical parameter suggestion model receives and outputs the second input data.

The corneal topography image prediction model may predict the corneal topography image after vision correction of the subject based on the surgical parameters and the first input data that the surgical parameter suggestion model receives and outputs the second input data.

24 to 25 are diagrams for calculating an output of a model related to vision correction based on the output of the vision prediction model according to an exemplary embodiment.

Referring to FIG. 24, the vision correction proposal model M14 may propose a vision correction procedure corresponding to the subject based on the vision prediction value and the first input data output by the vision prediction model M16 receiving second input data. have.

Referring to FIG. 25, the predicted visual acuity image generation model M17 generates a predicted visual acuity image of the subject after vision correction based on the visual acuity predicted value and the first input data output by the visual acuity prediction model M16 receiving second input data. Can be generated.

26 to 27 are diagrams for calculating an output of a model related to vision correction based on an output of a model for predicting a corneal topography image according to an exemplary embodiment.

Referring to FIG. 26, the custom vision correction necessity prediction model M13 is a custom vision correction procedure of a subject based on the corneal topography image and the first input data that the corneal topography image prediction model M18 receives and outputs the second input data. The need can be predicted.

Referring to FIG. 27, the vision correction proposal model M14 performs vision correction corresponding to the subject based on the corneal topography image and the first input data that the corneal topography image prediction model M18 receives and outputs the second input data. I can suggest.

The corneal shape factor prediction model may predict the corneal shape factor after vision correction of the subject based on the corneal topography image and the first input data that the corneal topography image prediction model receives and outputs the second input data.

The predicted visual acuity image generation model may generate a predicted visual acuity image after vision correction of the subject based on the corneal terrain image and the first input data that the corneal terrain image prediction model receives and outputs second input data.

Three or more models related to vision correction may be combined in series and/or in parallel with each other. 28 is a diagram for describing a combination of three or more models related to vision correction according to an exemplary embodiment. Referring to FIG. 28, as for the first vision correction-related model Ma, the second prediction result obtained by receiving the second input data from the second vision correction-related model Mb, and the third vision correction-related model Mc The third prediction result obtained by receiving the third input data and outputting the first input data and the first prediction result may be calculated by receiving the first input data. Here, the first vision correction-related model Ma may be viewed as being connected in series with the second vision correction-related model Mb and the third vision correction-related model Mc. In addition, the second vision correction-related model (Mb) and the third vision correction-related model (Mc) can be viewed as being connected in parallel.

29 to 30 are diagrams for calculating an output of a model related to vision correction based on the output of the corneal shape factor prediction model and the vision prediction model, according to an exemplary embodiment.

Referring to FIG. 29, in the vision correction proposal model M14, the corneal shape factor prediction model M12 receives the second input data and outputs the corneal shape factor, and the vision prediction model M16 receives the third input data. Based on the output predicted vision value and the first input data, it is possible to propose a vision correction procedure corresponding to the subject.

Referring to FIG. 30, in the predicted vision image generation model M17, the corneal shape factor prediction model M12 receives and outputs the second input data, and the vision prediction model M16 inputs the third input data. Based on the received and output acuity predicted value and first input data, a predicted visual acuity image may be generated after vision correction of the subject.

Models related to vision correction can be merged with each other. The plurality of models related to vision correction may be merged into one model to perform at least some of the functions of the plurality of individual models.

31 is a diagram for explaining merging of models related to vision correction according to an exemplary embodiment. Referring to FIG. 31, the first vision correction-related model and the second vision correction-related model may be merged to form one model Mab. The one model Mab may calculate a prediction result based on input data. Here, the prediction result may include information corresponding to at least one of a first prediction result corresponding to an output of the first vision correction-related model and a second prediction result corresponding to an output of the second vision correction-related model. For example, the prediction result may include at least one of a first prediction result and a second prediction result. Alternatively, the prediction result may include at least one of a first prediction result and a second prediction result.

In FIG. 31, a case of merging two models has been described, but the present invention is not limited thereto, and it is also possible to merge three or more models.

32 to 34 are diagrams illustrating an implementation example of merging models related to vision correction according to an embodiment.

Referring to FIG. 32, a corneal shape factor prediction model and a custom vision correction necessity prediction model may be merged. The merged model M25 may calculate a prediction result based on input data. The prediction result may include information corresponding to at least one of a corneal shape factor and a need for custom vision correction. For example, the prediction result may include at least one of a corneal shape factor and a need for custom vision correction.

Referring to FIG. 33, a model for predicting the necessity of a custom vision correction and a model for suggesting vision correction may be merged. The merged model M27 may calculate a prediction result based on input data. The prediction result may include information corresponding to at least one of necessity of custom vision correction and vision correction. For example, the prediction result may include at least one of necessity of custom vision correction and vision correction.

Referring to FIG. 34, a laser surgery possibility prediction model, a custom vision correction necessity prediction model, and an vision correction proposal model may be merged. The merged model M38 may calculate a prediction result based on input data. The prediction result may include information corresponding to at least one of whether or not laser surgery, need for custom vision correction, and vision correction. For example, the prediction result may include at least one of whether or not laser surgery, the need for custom vision correction, and vision correction.

Hereinafter, an example of a method for recommending vision correction will be described.

The method of recommending vision correction may be implemented using one or more models related to vision correction. When the method is implemented with a plurality of vision correction-related models, whether or not to execute at least one vision correction-related model may depend on a prediction result of at least one other vision correction-related model. For example, whether or not the second vision correction-related model is executed may depend on the prediction result of the first vision correction-related model.

Each step of the method for recommending vision correction to be described later may be performed by a prediction device.

35 is a diagram illustrating a method for recommending vision correction according to a first embodiment.

Referring to FIG. 35, in the method of recommending vision correction according to an embodiment, acquiring examination data of a subject (S1100), predicting whether the subject is suitable for vision correction (S1200), and correcting vision using a laser of the subject are possible. Predicting whether or not (S1300), calculating a predicted value of a corneal shape factor of the subject (S1400), and proposing a vision correction procedure corresponding to the subject (S1500).

Acquiring the examination data of the subject (S1100) may include obtaining, by the computing device, examination data including the examination data and the eye characteristic data measurement values.

Predicting whether the examinee is suitable for vision correction (S1200) may include predicting whether the examinee is suitable for vision correction by inputting the first group data obtained from the examination data of the examinee into the first prediction model. The first prediction model may be a surgical suitability prediction model. The surgical suitability prediction model may predict the suitability for vision correction of the subject based on the first group data.

Predicting whether or not the subject's laser can be used for vision correction (S1300) includes inputting the second group data obtained from the examinee's examination data into a second prediction model to predict whether the subject's laser can be used for vision correction. can do. Whether or not the step (S1300) is performed may depend on whether or not vision correction is appropriate for the subject. For example, the step S1300 may be performed when vision correction is appropriate for the subject. The second predictive model may be a predictive model for whether or not to perform laser surgery. The laser surgery availability prediction model may predict the laser surgery availability of a subject based on the second group data.

In the step of calculating the predicted value of the corneal shape factor of the subject (S1400), the predicted value of the corneal shape factor after standard vision correction of the subject and the corneal shape after custom vision correction by inputting the third group data obtained from the examinee's examination data into the third prediction model. It may include calculating a factor predicted value. Whether or not the step S1400 is performed may depend on whether or not the subject can perform vision correction using a laser. For example, the step S1400 may be performed when the subject is capable of correcting vision using a laser. The third prediction model may be a corneal shape factor prediction model. The third prediction model may predict a corneal shape factor based on the third group data. In this case, the necessity of custom vision correction may be determined based on the corneal shape factor.

Proposing vision correction corresponding to the subject (S1500) may include inputting the fourth group data obtained from the examinee's examination data into the fourth predictive model to propose a corresponding vision correction procedure to the subject. Whether or not the step (S1500) is performed may depend on whether or not the subject can perform vision correction using a laser. For example, the step S1500 may be performed when the subject is capable of correcting vision using a laser. The fourth predictive model may be learned based on at least one of examination data of a plurality of subjects who have undergone vision correction, vision correction corresponding to the plurality of subjects, and vision after vision correction of the plurality of subjects. The fourth prediction model may be a model for suggesting vision correction. The fourth prediction model may propose a vision correction procedure corresponding to the subject based on the fourth group data.

Referring to FIG. 35, in the vision correction, the predicted value of the corneal shape factor and/or the necessity of a custom vision correction may not be considered. For example, the vision correction may include LASIK, LASEK and Smile. In addition, even if laser vision correction is possible, since lens implantation is not impossible, the vision correction may include lens implantation, which is the same in other embodiments and implementations of the present specification. In this case, vision correction in which the predicted value of the corneal shape factor and/or the necessity of custom vision correction is considered may be determined by a doctor and/or a counselor. For example, doctors and/or counselors may use custom vision such as standard LASIK, standard LASEK, standard smile, custom LASIK, custom LASEK, and custom smile based on the corneal shape factor predicted value and vision correction output from the vision correction recommendation method of FIG. Vision correction can be decided in consideration of the need for correction.

36 is a diagram illustrating a second embodiment of a method for recommending vision correction according to an embodiment.

Referring to FIG. 36, the method for recommending vision correction according to an embodiment includes obtaining examination data of a subject (S2100), predicting whether the subject is suitable for vision correction (S2200), and correcting vision using a laser of the subject. Predicting whether or not (S2300), calculating a predicted value of a corneal shape factor of the subject (S2400), and proposing a vision correction procedure corresponding to the subject (S2500).

Since the method of recommending vision correction of FIG. 36 is similar to that of FIG. 35, the difference from FIG. 35 will be mainly described.

Referring to FIG. 36, in the step of suggesting a vision correction procedure corresponding to the subject (S2500), the corneal shape factor predicted value after standard vision correction and the corneal shape factor predicted value after custom vision correction calculated in the step of calculating the corneal shape factor predicted value of the subject. Can suggest vision correction based on. The vision correction procedure may be one in which the necessity of a custom vision correction procedure is considered. For example, the vision correction may include standard LASIK, standard LASEK, standard smile, custom LASIK, custom LASEK and custom smile. In addition, the vision correction may include lens implantation.

37 is a diagram of a third embodiment of a method for recommending vision correction according to an embodiment.

Referring to FIG. 37, in the method of recommending vision correction according to an embodiment, obtaining examination data of a subject (S3100), predicting whether the subject is suitable for vision correction (S3200), and correcting vision using a laser of the subject are possible. Predicting whether or not (S3300), predicting whether the subject needs a custom vision correction (S3400), and suggesting a corresponding vision correction to the subject (S3500).

Since the method of recommending vision correction in FIG. 37 is similar to that of FIG. 35, differences from FIG. 35 will be mainly described.

Referring to FIG. 37, in the step of predicting whether the subject needs custom vision correction (S3400), predicting whether the subject needs custom vision correction by inputting the third group data obtained from the examinee's examination data into the third prediction model. May include. Whether or not the step S3400 is performed may depend on whether or not the subject can perform vision correction using a laser. For example, the step S3400 may be performed when the subject is capable of correcting vision using a laser. The third predictive model may be a predictive model for necessity of custom vision correction. The third prediction model may predict the need for custom vision correction based on the third group data. In the above step, the necessity of custom vision correction may be predicted based on the predicted value of the corneal shape factor after standard vision correction of the subject and the predicted value of the corneal shape factor after custom vision correction.

The step of predicting whether the subject needs a custom vision correction (S3400) and the step of suggesting a corresponding vision correction to the subject (S3500) may be connected. For example, the output of the step S3500 of suggesting a vision correction corresponding to the subject may be calculated based on the output of the step S3400 of predicting whether the subject needs a custom vision correction. Alternatively, the output of the step (S3400) of predicting whether the subject needs a custom vision correction may be calculated based on the output of the step (S3500) of proposing a vision correction procedure corresponding to the subject (S3500).

For example, in the step of proposing a vision correction procedure corresponding to the subject (S3500), vision correction may be proposed based on the necessity of a custom vision correction procedure calculated in the step (S3400) of predicting whether the subject needs a custom vision correction procedure (S3400). For example, the vision correction may include standard LASIK, standard LASEK, standard smile, custom LASIK, custom LASEK and custom smile. In addition, the vision correction may include lens implantation.

As another example, in the step of predicting whether the subject needs a custom vision correction (S3400), a second vision correction procedure may be output based on the first vision correction procedure calculated in the step (S3500) of proposing a vision correction procedure corresponding to the subject. . Here, the first vision correction may not be a necessity of a custom vision correction, and the second vision correction may be a need of a custom vision correction.

As another example, in the step of predicting whether the subject needs a custom vision correction (S3400), whether to execute or not may be determined based on the type of the first vision correction procedure calculated in the step (S3500) of proposing a corresponding vision correction procedure to the subject. . For example, when the first type of vision correction is the first type, a custom vision correction necessity prediction model may not be executed, and when the first vision correction is of the second type, a custom vision correction necessity prediction model may be executed.

The first type and the second type may be classified according to whether the cornea is cut using a laser. For example, the first type may be a non-laser vision correction such as lens implantation, and the second type may be a laser vision correction such as LASIK, LASEK, and Smile.

The first type and the second type may be classified according to whether custom surgery is possible. For example, the first type may be a vision correction procedure in which custom surgery is not possible, and the second type may be a vision correction procedure in which a custom surgery is possible. Here, a predetermined criterion may exist as to whether a custom surgery is possible for vision correction. However, these standards may vary depending on technological advances, hospitals, surgical devices, and doctors' circumstances and judgments. For example, if custom smile surgery is not possible, the first type may include smile and lens implantation, and the second type may include LASIK and LASEK, but if custom smile surgery is possible, the first type includes lens implantation and Two types may include LASIK, LASEK and Smile.

38 is a diagram illustrating a fourth embodiment of a method for recommending vision correction according to an embodiment.

Referring to FIG. 38, in the method of recommending vision correction according to an embodiment, obtaining examination data of a subject (S4100), predicting whether the subject is suitable for vision correction (S4200), and correcting vision using a laser of the subject are possible. Predicting whether or not (S4300) and proposing a corresponding vision correction to the subject (S4400) may be included.

The step of acquiring the examination data of the subject in FIG. 38 (S4100), the step of predicting whether the subject is suitable for vision correction (S4200), and the step of predicting whether the subject's laser can be used for vision correction (S4300) are the same as in FIG. Therefore, a description thereof will be omitted.

Proposing a vision correction procedure corresponding to the examinee (S4400) may include inputting the third group data obtained from the examination data of the examinee into a third predictive model to propose a vision correction procedure corresponding to the examinee.

Whether or not the step S4400 is performed may depend on whether or not the subject can perform vision correction using a laser. For example, the step (S4400) may be performed when the subject is capable of correcting vision using a laser. In the step S4400, vision correction may be proposed based on the predicted value of the corneal shape factor after standard vision correction of the subject and the predicted value of the corneal shape factor after custom vision correction.

The third prediction model may be a model in which a custom vision correction necessity prediction model and an vision correction proposal model are merged. The third predictive model may be learned based on at least one of examination data of a plurality of subjects who have undergone vision correction, vision correction corresponding to the plurality of subjects, and vision after vision correction of the plurality of subjects. The third prediction model may output information on at least one of necessity of custom vision correction and vision correction based on the third group data. For example, the output of the third prediction model may include at least one of necessity of custom vision correction and vision correction.

39 is a diagram illustrating a fifth embodiment of a method for recommending vision correction according to an embodiment.

Referring to FIG. 39, the method for recommending vision correction according to an embodiment includes acquiring examination data of a subject (S5100), predicting whether the subject is suitable for vision correction (S5200), and suggesting corrective vision corresponding to the subject. It may include a step (S5300).

The step of acquiring the examination data of the examinee in FIG. 39 (S5100) and the step of predicting whether the examinee is suitable for vision correction (S5200) are the same as those of FIG. 35, so a description thereof will be omitted.

Proposing a vision correction procedure corresponding to the subject (S5300) may include inputting the second group data obtained from the examinee's examination data into a second prediction model to propose a vision correction procedure corresponding to the subject.

Whether or not the step (S5300) is performed may depend on whether or not vision correction is appropriate for the subject. For example, the step S5300 may be performed when vision correction is appropriate for the subject. In the step S5300, vision correction may be proposed based on the predicted value of the corneal shape factor after standard vision correction of the subject and the predicted value of the corneal shape factor after custom vision correction.

The second prediction model may be a model in which a laser surgery possibility prediction model, a custom vision correction necessity prediction model, and an vision correction proposal model are merged. The second predictive model may be learned based on at least one of examination data of a plurality of subjects who have undergone vision correction, vision correction corresponding to the plurality of subjects, and vision after vision correction of the plurality of subjects. The second prediction model may output information on at least one of whether or not laser surgery, necessity of custom vision correction, and vision correction surgery, based on the second group data. For example, the output of the second prediction model may include at least one of whether or not laser surgery, necessity of custom vision correction, and vision correction.

40 is a diagram of a sixth embodiment of a method for recommending vision correction according to an embodiment.

Referring to FIG. 40, a method for recommending vision correction according to an embodiment may include acquiring examination data of a subject (S6100) and proposing a vision correction procedure corresponding to the subject (S6200).

The step (S6100) of acquiring the examination data of the examinee of FIG. 40 is the same as that of FIG. 35, so a description thereof will be omitted.

Proposing a vision correction procedure corresponding to the examinee (S6200) may include inputting group data obtained from the examination data of the examinee into a predictive model to propose a vision correction procedure corresponding to the examinee. In the step S6200, vision correction may be proposed based on the predicted value of the corneal shape factor after standard vision correction of the subject and the predicted value of the corneal shape factor after custom vision correction.

The predictive model may be a model in which a surgical fit prediction model, a laser surgical fit prediction model, a custom vision correction necessity prediction model, and an vision correction proposal model are merged. The predictive model may be learned based on at least one of examination data of a plurality of persons undergoing vision correction, vision correction corresponding to the plurality of persons to be treated, and vision after vision correction of the plurality of persons to be treated. The predictive model may output information on at least one of whether or not to perform surgery, whether or not to perform laser surgery, necessity for custom vision correction, and corrective vision based on input data. For example, the output of the predictive model may include at least one of whether or not surgery is appropriate, whether or not to perform laser surgery, necessity of custom vision correction, and vision correction.

The combination and/or merging of the above-described method for recommending vision correction and the model related to vision correction is only an example, and in addition, the method for recommending vision correction may be implemented in various ways, or combined and/or combined with the model related to vision correction.

Hereinafter, an embodiment of a method of providing visualization information for vision correction will be described.

The method of providing visualization information for vision correction may be implemented using one or more models related to vision correction. When the method is implemented with a plurality of vision correction-related models, whether or not to execute at least one vision correction-related model may depend on a prediction result of at least one other vision correction-related model. For example, whether or not the second vision correction-related model is executed may depend on the prediction result of the first vision correction-related model.

The method of providing visualization information for vision correction may include a method of providing a predicted vision image, a method of providing a corneal topography image, and a method of providing a cause for calculating a predicted result.

The method of providing the predicted eyesight image may be implemented through a predicted eyesight image generation model. The method of providing a corneal topography image can be implemented through a corneal topography image prediction model. The method of providing the cause of the prediction result calculation may be implemented through a model for analyzing the cause of the prediction result calculation.

Each step of the method for providing visualization information for vision correction to be described later may be performed by a prediction device.

41 is a diagram illustrating a method of providing visualization information for vision correction according to a first embodiment.

Referring to FIG. 41, in the method for providing vision correction information according to an embodiment, acquiring examination data of a subject (S7100), calculating a predicted value of eye characteristic data after vision correction of the subject (S7200), and a predicted visual acuity image It may include the step of generating (S7300). In addition, although not shown, the method of providing visualization information for vision correction according to an embodiment may further include outputting a predicted vision image.

Acquiring the examination data of the subject (S7100) may include obtaining, by the computing device, examination data including the examination data and the eye characteristic data measurement values.

In the step of calculating the predicted value of eye characteristic data after vision correction of the subject (S7200), at least one of the predicted value of visual acuity and the predicted value of corneal shape factor after the patient's vision correction surgery by inputting the first group data obtained from the examination data of the subject into the first prediction model. It may include calculating a predicted value of eye characteristic data including one.

The first predictive model is based on at least one of a pre-operative eye characteristic data measurement value of a plurality of subjects who have undergone vision correction, a vision correction surgery parameter performed on the plurality of subjects, and a measurement value of eye characteristic data after surgery of the plurality of subjects. It can be learned on the basis of. The first prediction model may include at least one of a vision prediction model and a corneal shape factor prediction model. Alternatively, the first prediction model may be a model in which a vision prediction model and a corneal shape factor prediction model are merged. The first prediction model may calculate a predicted value of eye characteristic data after vision correction of the subject based on the first group data.

Generating the predicted visual acuity image (S7300) may include generating a predicted visual acuity image based on the predicted value of eye characteristic data.

42 is a diagram illustrating a second embodiment of a method of providing visualization information for vision correction according to an embodiment.

Referring to FIG. 42, the method of providing visualization information for vision correction according to an embodiment further includes calculating and/or selecting a filter based on a predicted value of eye characteristic data (S7600) and applying the filter to an original image (S7700). Can include.

Calculating and/or selecting a filter based on the predicted value of eye characteristic data (S7600) may be performed by the first sub-model M171 of FIG. 11.

Applying the filter to the original image (S7700) may include applying the filter to the original image to generate a predicted vision image. The step S7700 may be performed by the second sub-model M172 of FIG. 11.

43 is a diagram of a third embodiment of a method of providing visualization information for vision correction according to an embodiment.

Referring to FIG. 43, the method of providing visualization information for vision correction according to an embodiment may further include predicting a corneal topography image after vision correction of a subject (S7400).

The step of predicting the corneal topography image after vision correction of the subject (S7400) may include predicting the corneal topography image after vision correction of the subject by inputting the second group data obtained from the examinee's examination data into the second prediction model. have.

The second predictive model is to be learned based on at least one of preoperative corneal topography images of a plurality of recipients who have undergone vision correction, vision correction surgery parameters performed on the plurality of recipients, and corneal topography images after surgery of the plurality of recipients. I can. The second prediction model may be a corneal topography image prediction model. The second prediction model may predict a corneal topography image of the subject after vision correction based on the second group data.

44 is a diagram of a fourth embodiment of a method of providing visualization information for vision correction according to an embodiment.

Referring to FIG. 44, the method of providing visualization information for vision correction according to an embodiment may further include calculating a dependence of the predicted value of eye characteristic data on the first group data (S7500 ). In addition, although not shown, the method of providing visualization information for vision correction according to an embodiment may further include outputting a dependence coefficient.

The step of outputting the dependency coefficient may include outputting a dependency coefficient larger than a predetermined value among the dependency coefficients or outputting a predetermined number of dependency coefficients.

The method according to the embodiment may be implemented in the form of program instructions that can be executed through various computing devices and recorded in a computer-readable medium. The computer-readable medium may include program instructions, data files, data structures, and the like alone or in combination. The program instructions recorded on the medium may be specially designed and configured for the embodiment, or may be known and usable to those skilled in computer software. Examples of computer-readable recording media include magnetic media such as hard disks, floppy disks, and magnetic tapes, optical media such as CD-ROMs and DVDs, and magnetic media such as floptical disks. -Includes magneto-optical media and hardware devices specially configured to store and execute program instructions such as ROM, RAM, flash memory, and the like. Examples of program instructions include not only machine language codes such as those produced by a compiler, but also high-level language codes that can be executed by a computing device using an interpreter or the like. The hardware device described above may be configured to operate as one or more software modules to perform the operation of the embodiment, and vice versa.

In the above, the configuration and features of the present invention have been described based on the embodiments, but the present invention is not limited thereto, and that various changes or modifications can be made within the spirit and scope of the present invention to those skilled in the art. It is obvious, and therefore, such changes or modifications are found to be within the scope of the appended claims.

10: Vision correction assist system
100: learning device
300: prediction device
500: server device
700: client device
1000: control unit
5000: memory unit
9000: Ministry of Communications

Claims (13)

In the method of providing a predicted vision image after vision correction using artificial intelligence performed by a computing device,
Acquiring examination data of the examinee, the examination data including examination data and eye characteristic data measurement values;
The predicted value of the eye characteristic data after vision correction of the examinee by inputting the first group data obtained from the examination data of the examinee into a first prediction model-The predicted value of the eye characteristic data includes at least one of a predicted visual acuity value and a predicted corneal shape factor. -Calculating; And
Including; generating a predicted vision image based on the predicted eye characteristic data
The first predictive model is at least one of a measurement value of eye characteristic data before surgery of a plurality of subjects who have undergone vision correction, a parameter of a vision correction surgery performed on the plurality of subjects, and a measurement value of eye characteristic data after surgery of the plurality of subjects. Learned on the basis of
How to provide a predicted vision image.
The method of claim 1,
The generating of the predicted visual acuity image includes calculating or selecting a filter based on the predicted value of the eye characteristic data, and applying the filter to the original image to generate the predicted visual acuity image.
How to provide a predicted vision image.
The method of claim 1,
The predicted value of eye characteristic data includes a predicted value of first eye characteristic data and a predicted value of second eye characteristic data,
The predicted visual acuity image is generated based on the first predicted visual acuity image and the second predicted visual acuity image generated based on the predicted value of the first eye characteristic data, Inclusive
How to provide a predicted vision image.
The method of claim 3,
Each of the predicted value of the first eye characteristic data and the predicted value of the second eye characteristic data corresponds to a predicted value of eye characteristic data after standard vision correction and a predicted value of eye characteristic data after custom vision correction,
Each of the first predicted visual acuity image and the second predicted visual acuity image corresponds to an expected visual acuity image after standard vision correction and a predicted visual acuity image after custom vision correction
How to provide a predicted vision image.
The method of claim 1,
The predicted vision image includes information on at least one of clarity, light spread, contrast sensitivity, night vision, glare, double vision, and afterimages expected after vision correction of the subject.
How to provide a predicted vision image.
The method of claim 1,
Predicting the corneal topography image after vision correction of the subject by inputting the second group data obtained from the examination data of the subject into a second prediction model; further comprising,
The second predictive model is learned based on at least one of preoperative corneal topography images of a plurality of recipients who have undergone vision correction, vision correction surgery parameters performed on the plurality of recipients, and postoperative corneal topography images of the plurality of recipients. Made
How to provide a predicted vision image.
The method of claim 1,
Calculating a dependence of the predicted value of eye characteristic data on the first group data;
How to provide a predicted vision image.
The method of claim 7,
The predicted value of eye characteristic data includes a predicted value of first eye characteristic data and a predicted value of second eye characteristic data,
The dependence includes a dependence coefficient corresponding to at least a part of the first group data,
The dependence coefficient includes a first dependency coefficient corresponding to the predicted value of the first eye characteristic data and a second dependency coefficient corresponding to the predicted value of the second eye characteristic data, but different from the first dependency coefficient.
How to provide a predicted vision image.
The method of claim 8,
Each of the predicted value of the first eye characteristic data and the predicted value of the second eye characteristic data corresponds to a predicted value of eye characteristic data after standard vision correction and a predicted value of eye characteristic data after custom vision correction,
Each of the first and second dependence coefficients corresponds to the dependence of the predicted value of eye characteristic data after standard vision correction on the first group data and the dependence of the predicted value of eye characteristic data after custom vision correction on the first group data. felled
How to provide a predicted vision image.
The method of claim 8,
Outputting a dependency coefficient greater than a predetermined value among the dependency coefficients or outputting a predetermined number of dependency coefficients; further comprising
How to provide a predicted vision image.
The method of claim 1,
The corneal shape factor predicted value includes at least one of an index of height decentration (IHD) predicted value, an index of surface variance (ISV) predicted value, and an index of vertical asymmetry (IVA) predicted value.
How to provide a predicted vision image.
The method of claim 1,
The vision predicted value includes at least one of a predicted value of lower-order aberrations and a predicted value of higher-order aberrations.
How to provide a predicted vision image.
A recording medium readable by a computing device storing a program for executing the method according to any one of claims 1 to 12 on a computing device.

Images