A method for predicting corneal orthokeratology parameters based on artificial neural network

Abstract

The embodiment of the present invention discloses a method for predicting corneal reshaping parameters based on an artificial neural network, and relates to the technical field of orthokeratology lenses. The method comprises: obtaining an initial sample data set, deleting target samples that do not contain necessary parameters, and obtaining a basic sample data set; preprocessing the basic sample data set to obtain a final sample data set; training the initial artificial neural network model to obtain a preliminary model; using the preliminary model to forward propagate the input layer data of the basic sample data set to obtain a calculation result and compare it with the result of the basic sample to obtain a final sample data set, and using the final sample data set to train the preliminary model to obtain a prediction model; inputting actual eye detection data into the prediction model to obtain corresponding corneal reshaping parameters. The present invention splits the data to be predicted into multiple dimensions, constructs different prediction models according to each data to be predicted, and then performs distributed training and uses them in combination.

Description

Technical Field

[0001] The embodiments of the present invention relate to the technical field of orthokeratology lenses, and specifically to a method for predicting orthokeratology parameters based on artificial neural networks. Background Art

[0002] Orthokeratology lenses are a type of specially designed hard contact lens that is worn while sleeping to change the shape of the corneal surface and generate peripheral out-of-focus points to improve naked eye vision and control myopia. Currently, orthokeratology lenses are fitted mainly through basic examinations and fluorescent staining adaptation, while fluorescent staining adaptation is achieved by staining the tears with sodium fluorescein while wearing the lenses. The distribution of tears under the lens is then analyzed based on experience using the colored fluorescein, and the parameters are repeatedly adjusted to achieve the ideal state.

[0003] This method requires highly professional medical personnel to complete, and it is very time-consuming. In the evaluation stage during the debugging process, the number of lenses tried mostly does not match the corneal morphology of the wearer. In addition to the high degree of discomfort, there are also certain safety risks to corneal health.

[0004] Artificial intelligence technology learns a large number of past cases and establishes a mathematical model that relates corneal morphology to adaptation parameters. Using artificial intelligence technology and algorithms to quickly find the best matching results will greatly improve clinical efficiency and improve the adverse effects of prolonged conventional adaptation on corneal health.

[0005] Conventional neural networks can establish mathematical models with multiple input layer data and output layer data, and use neural networks and weights to reflect the overall correlation. However, many data such as corneal examination data and corneal reshaping adaptation parameters are closely related to certain data of corneal morphology, and some data are affected by the correlation of multiple parameters, and the degree of influence varies.

[0006] In addition to considering data relevance, we also need to consider the integrity of the data of the wearers used in training on all related data. That is to say, for example, a certain data related to corneal reshaping adaptation data is related to multiple examination parameters, but some parameters are not examined by all wearers, and thus cannot be applied to the neural network prediction model, or the corneal reshaping parameters that are completely different from the actual ones are obtained. For example, the toric amount of the lens is mainly determined by the amount of corneal astigmatism of the patient, which is also the basic data that everyone needs to check, but the amount of astigmatism is a low-dimensional parameter and cannot accurately quantify the coverage, symmetry, irregular shape, etc. of the astigmatism. The difference in sagittal height on the steep and flat meridians can be used to better evaluate the toric amount of the corneal astigmatism corresponding to the corneal reshaping doctor, but this parameter is not available to all examiners or their examination equipment. Therefore, it is crucial to integrate this part of the parameters that are not available in every training data into the training and to more accurately affect the model training and learn it. Summary of the invention

[0007] The purpose of an embodiment of the present invention is to provide a corneal reshaping parameter prediction method based on an artificial neural network, which splits the data parameters to be predicted to construct multiple different models, each model performs its own personalized data indicator association, and is combined after distributed training to ensure the accuracy of the prediction.

[0008] In order to solve the above problems, the first aspect of the embodiment of the present invention discloses a method for predicting corneal orthokeratology parameters based on an artificial neural network, comprising: Acquire an initial sample data set, and determine the parameters of the initial sample data set. If the target sample in the initial sample data set does not contain the necessary parameters, delete the target sample to obtain a basic sample data set. The necessary parameters include orthokeratology lens brand, flat e value, visible iris transverse diameter, flat K value, steep K value, diopter, intraocular pressure, toric state, positioning arc value, toric value, lens total diameter, and lens drop. Preprocessing the basic sample data set to obtain a preliminary sample data set, wherein each sample in the preliminary sample data set is recorded as a basic sample; Constructing an initial artificial neural network model, inputting the orthokeratology lens brand, flat e value, visible iris transverse diameter, flat K value, steep K value, diopter, intraocular pressure, toric curvature state and positioning arc value in the basic sample into the initial artificial neural network model, and comparing the toric curvature value, positioning arc value, total lens diameter and lens drop with the toric curvature prediction value, positioning arc prediction value, total lens diameter prediction value and lens drop prediction value output by the initial artificial neural network model, completing the preliminary training of the initial artificial neural network model, and obtaining a preliminary model; Input the orthokeratology lens brand, flat e value, visible iris transverse diameter, flat K value, steep K value, diopter, intraocular pressure, toric curvature state and positioning arc value in the basic sample into the preliminary model, compare the toric curvature value, positioning arc value, total lens diameter and lens drop with the toric curvature prediction value, positioning arc prediction value, total lens diameter prediction value and lens drop prediction value output by the preliminary model, filter out discrete data to obtain a final sample data set, use the final sample data set to perform final training on the initial artificial neural network model or the preliminary model to obtain a prediction model; The actual eye detection data is input into the prediction model to obtain the corneal shaping parameters corresponding to the actual eye detection data.

[0009] As a preferred embodiment, in the first aspect of the embodiment of the present invention, an initial artificial neural network model is constructed, and the initial artificial neural network model is trained to obtain a preliminary model, including: Constructing a first initial artificial neural network model for predicting the loop curvature value, and obtaining a first preliminary model after training; Constructing a second initial artificial neural network model for predicting the positioning arc value, and obtaining a second preliminary model after training; Constructing a third initial artificial neural network model for predicting the total diameter value of the lens, and obtaining a third preliminary model after training; A fourth initial artificial neural network model is constructed to predict the lens drop value, and a fourth preliminary model is obtained after training.

[0010] As a preferred embodiment, in the first aspect of the embodiment of the present invention, the loop curvature value is predicted, and a first preliminary model is obtained after training, including: Inputting the flat e value, flat K value and steep K value in the basic sample into the first initial artificial neural network model; Compare the first loop curvature prediction value output by the first initial artificial neural network model with the loop curvature value in the basic sample, as well as the first loop curvature state prediction value output by the first initial artificial neural network model with the loop curvature state in the basic sample, and use the gradient descent algorithm to update the parameters of the first initial artificial neural network model until the training of the first initial artificial neural network model is completed to obtain a first preliminary model.

[0011] As a preferred embodiment, in the first aspect of the embodiment of the present invention, the positioning arc value is predicted, and a second prediction model is obtained after training, including: Inputting the flat K value, steep K value, flat e value, visible iris transverse diameter, diopter, orthokeratology lens brand and toricity state in the basic sample into the second initial artificial neural network model; The first positioning arc prediction value output by the second initial artificial neural network model is compared with the positioning arc value in the basic sample, and the parameters of the second initial artificial neural network model are updated using a gradient descent algorithm until the training of the second initial artificial neural network model is completed to obtain a second preliminary model.

[0012] As a preferred embodiment, in the first aspect of the embodiment of the present invention, the total diameter value of the lens is predicted, and a third prediction model is obtained after training, including: Input the visible iris transverse diameter, flat K value, steep K value, flat e value, diopter, orthokeratology lens brand, toric curvature state and positioning arc value in the basic sample into the third initial artificial neural network model; The predicted value of the first lens total diameter output by the third initial artificial neural network model is compared with the total diameter value of the lens in the basic sample, and the parameters of the third initial artificial neural network model are updated using a gradient descent algorithm until the training of the third initial artificial neural network model is completed to obtain a third preliminary model.

[0013] As a preferred embodiment, in the first aspect of the embodiment of the present invention, the lens drop value is predicted, and a fourth prediction model is obtained after training, including: Inputting the flat K value, diopter, intraocular pressure and positioning arc value in the basic sample into the fourth initial artificial neural network model; The first lens degradation prediction value output by the fourth initial artificial neural network model is compared with the lens degradation in the basic sample, and the parameters of the fourth initial artificial neural network model are updated using a gradient descent algorithm until the training of the fourth initial artificial neural network model is completed to obtain a fourth preliminary model.

[0014] As a preferred embodiment, in the first aspect of the embodiment of the present invention, the orthokeratology lens brand, flat e value, visible iris transverse diameter, flat K value, steep K value, diopter, intraocular pressure, toric state and positioning arc value in the basic sample are input into the preliminary model, and the positioning arc value, toric value, total lens diameter and lens drop are compared with the positioning arc prediction value, toric prediction value, total lens diameter prediction value and lens drop prediction value output by the preliminary model, and the final sample data set is obtained after the discrete data is screened out, including: Input the flat e value, the flat K value and the steep K value in the basic sample into the first preliminary model to obtain a second torus prediction value output by the first preliminary model; Input the flat K value, steep K value, flat e value, visible iris transverse diameter, diopter, orthokeratology lens brand and toric curvature state in the basic sample into the second preliminary model to obtain the second positioning arc prediction value output by the second preliminary model; Input the visible iris transverse diameter, flat K value, steep K value, flat e value, diopter, orthokeratology lens brand, toric curvature state and positioning arc value in the basic sample into the third preliminary model to obtain the predicted value of the second lens total diameter output by the third preliminary model; Input the flat K value, diopter, intraocular pressure and positioning arc value in the basic sample into the fourth preliminary model to obtain the second lens drop prediction value output by the fourth preliminary model; The relationship between the torus value, positioning arc value, total lens diameter and lens drop in the basic sample and the second torus prediction value, second positioning arc prediction value, second lens total diameter prediction value and second lens drop prediction value are compared respectively, the target basic sample is filtered out, and the final sample data set is obtained.

[0015] As a preferred embodiment, in the first aspect of the embodiment of the present invention, the relationship between the loop value, positioning arc value, lens total diameter and lens drop in the basic sample and the second loop value prediction value, the second positioning arc prediction value, the second lens total diameter prediction value and the second lens drop prediction value is respectively compared, and the target basic sample is filtered out to obtain a final sample data set, including: When any one or more of the following situations exist in the basic sample, the basic sample is defined as the target basic sample: Case 1: The absolute value of the difference between the second loop curvature prediction value and the corresponding loop curvature value in the basic sample is greater than 0.75; Case 2: The absolute value of the difference between the second positioning arc prediction value and the corresponding positioning arc value in the basic sample is greater than 0.62; Case 3: the absolute value of the difference between the predicted value of the second lens total diameter and the corresponding lens total diameter in the basic sample is greater than 0.3; Scenario 4: The absolute value of the difference between the predicted value of the second lens drop and the lens drop in the corresponding basic sample is greater than 1; The target basic samples are screened out to obtain a final sample data set.

[0016] As a preferred embodiment, in the first aspect of the embodiment of the present invention, the initial artificial neural network model or the preliminary model is finally trained using the final sample data set to obtain a prediction model, including: Using the final sample data set to train the first preliminary model, the second preliminary model, the third preliminary model and the fourth preliminary model respectively, to obtain a first prediction model, a second prediction model, a third prediction model and a fourth prediction model respectively; Alternatively, the first initial artificial neural network model, the second initial artificial neural network model, the third initial artificial neural network model and the fourth initial artificial neural network model are trained respectively using the final sample data set to obtain a first prediction model, a second prediction model, a third prediction model and a fourth prediction model respectively.

[0017] As a preferred embodiment, in the first aspect of the embodiment of the present invention, the actual eye detection data is input into the prediction model to obtain the corneal shaping parameters corresponding to the actual eye detection data, including: Inputting the actual flat K value, the actual steep K value, and the actual flat e value in the actual eye detection data into the first prediction model, and outputting the loop curvature value prediction data and the loop curvature state prediction data; Input the actual flat K value, the actual steep K value, the actual flat e value, the actual visible iris transverse diameter, the actual diopter, the orthokeratology lens brand and the toric curvature state prediction data output by the first prediction model in the actual eye detection data into the second prediction model, and output the positioning arc prediction data; Input the actual visible iris transverse diameter, actual flat K value, actual steep K value, actual flat e value, actual diopter, orthokeratology lens brand in the actual eye detection data, as well as the toric state prediction data output by the first prediction model and the positioning arc prediction data output by the second prediction model into the third prediction model, and output the total lens diameter prediction data; Input the actual flat K value, actual refractive power, actual intraocular pressure in the actual eye detection data and the positioning arc prediction data output by the second prediction model into the fourth prediction model, and output the lens drop prediction data; The toric curvature value prediction data, positioning arc prediction data, lens total diameter prediction data and lens drop prediction data constitute the corneal shaping parameters for preparing orthokeratology lenses.

[0018] A second aspect of an embodiment of the present invention discloses a corneal shaping parameter prediction device based on an artificial neural network, comprising: an acquisition unit, used for acquiring an initial sample data set and determining parameters of the initial sample data set, and if a target sample in the initial sample data set does not contain necessary parameters, deleting the target sample to obtain a basic sample data set, wherein the necessary parameters include a flat e value, a visible iris transverse diameter, a flat K value, a steep K value, a diopter, a toric state, a positioning arc value, a toric value, a total lens diameter, and a lens drop; A preprocessing unit, used for preprocessing the basic sample data set to obtain a final sample data set, where each sample in the sample data set is recorded as a basic sample; A training unit is used to construct an initial artificial neural network model, input the flat e value, visible iris horizontal diameter, flat K value, steep K value, diopter, toric state and positioning arc value in the basic sample into the initial artificial neural network model, compare the positioning arc value, toric value, total lens diameter and lens drop with the positioning arc prediction value, toric prediction value, total lens diameter prediction value and lens drop prediction value output by the initial artificial neural network model, complete the training of the initial artificial neural network model, and obtain a prediction model; The prediction unit is used to input the actual eye detection data into the prediction model to obtain the corneal shaping parameters corresponding to the actual eye detection data.

[0019] A third aspect of an embodiment of the present invention discloses an electronic device, comprising a memory, a processor, and a computer program stored in the memory and capable of running on the processor. When the processor executes the computer program, the steps of the corneal shaping parameter prediction method based on artificial neural network disclosed in the first aspect of the embodiment of the present invention are implemented.

[0020] A fourth aspect of an embodiment of the present invention discloses a computer-readable storage medium storing a computer program, wherein the computer program enables a computer to execute the steps of the corneal shaping parameter prediction method based on an artificial neural network disclosed in the first aspect of the embodiment of the present invention.

[0021] The fifth aspect of the embodiments of the present invention discloses a computer program product. When the computer program product is run on a computer, the computer is enabled to execute the steps of the corneal shaping parameter prediction method based on artificial neural network disclosed in the first aspect of the embodiments of the present invention.

[0022] A sixth aspect of an embodiment of the present invention discloses an application publishing platform, which is used to publish a computer program product. When the computer program product runs on a computer, the computer executes the steps of the corneal shaping parameter prediction method based on an artificial neural network disclosed in the first aspect of the embodiment of the present invention.

[0023] Compared with the prior art, the embodiments of the present invention have the following beneficial effects: 1. The present invention splits the data to be predicted into multiple dimensions, constructs different prediction models according to each data to be predicted, and each prediction model performs its own personalized data indicator association, and then performs distributed training and uses them in combination. The distributed training of multiple prediction models allows each prediction data to be adjusted separately, and each prediction model does not affect each other, while the established data association still exists in the most concise and core form.

[0024] 2. The present invention establishes a specified association between the eye examination parameters (input data) and each corneal reshaping adaptation parameter (output data of each prediction model) in the prediction model, and performs optimal processing on the associated data, thereby ensuring the association and influence of each parameter, thereby greatly reducing the loss function of the model training and significantly improving the accuracy of the data.

[0025] 3. The present invention sets necessary and non-essential data, which not only ensures that the model can absorb the influence of non-essential data, but also ensures that the model can be trained and used normally without necessary parameters.

[0026] 4. In the preprocessing stage, additional manual algorithms are involved based on clinical experience to verify and correct the model data so that the data output by the model conforms to clinical standards and empirical values. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] Figure 1 It is a flow chart of a method for predicting corneal shaping parameters based on an artificial neural network disclosed in an embodiment of the present invention; Figure 2 It is a schematic diagram of the structure of a corneal shaping parameter prediction device based on an artificial neural network disclosed in an embodiment of the present invention; Figure 3 It is a schematic diagram of the structure of an electronic device disclosed in an embodiment of the present invention. DETAILED DESCRIPTION

[0028] This specific implementation manner is merely an explanation of the embodiments of the present invention, and it is not a limitation of the embodiments of the present invention. After reading this specification, those skilled in the art can make non-creative modifications to the embodiments as needed, but as long as it is within the scope of the claims of the embodiments of the present invention, it is protected by the patent law.

[0029] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of them. Based on the embodiments in the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the embodiments of the present invention.

[0030] The term "comprise" and any variations thereof in the specification and claims of the present application are intended to cover non-exclusive inclusions. For example, a process, method, system, product or apparatus comprising a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to these processes, methods, products or apparatuses.

[0031] In the embodiments of the present invention, words such as "exemplary" or "for example" are used to indicate examples, illustrations or explanations. Any embodiment or design described as "exemplary" or "for example" in the embodiments of the present invention should not be interpreted as being more preferred or more advantageous than other embodiments or designs. Specifically, the use of words such as "exemplary" or "for example" is intended to present related concepts in a specific way.

[0032] The present invention splits the data to be predicted into multiple dimensions, constructs different prediction models according to each data to be predicted, and each prediction model performs its own personalized data indicator association, and then performs distributed training and uses them in combination. The distributed training of multiple prediction models allows each prediction data to be adjusted separately, and each prediction model does not affect each other, while the established data association still exists in the most concise and core form, which is described in detail below in conjunction with the accompanying drawings.

[0033] Embodiment 1

[0034] See also Figure 1 , Figure 1 FIG. 1 is a flow chart of a method for predicting corneal reshaping parameters based on an artificial neural network disclosed in an embodiment of the present invention. Figure 1 As shown, the corneal shaping parameter prediction method based on artificial neural network includes: S110, obtaining an initial sample data set, and determining the parameters of the initial sample data set. If the target sample in the initial sample data set does not contain necessary parameters, the target sample is deleted to obtain a basic sample data set. The necessary parameters include orthokeratology lens brand, flat e value, visible iris transverse diameter, flat K value, steep K value, refractive power, intraocular pressure, toric curvature status, positioning arc value, toric curvature value, total lens diameter and lens drop.

[0035] In the embodiment of the present invention, the parameters of the orthokeratology lens designed mainly for VST are predicted, and other design methods such as CRT are not within the protection scope of the embodiment of the present invention.

[0036] It should be noted that the order of step S110 and step S120 in the embodiment of the present invention can be swapped, that is, the initial sample data set can be preprocessed first, and then the samples that do not contain the necessary parameters can be deleted. In other embodiments, before step S110 and step S120, each sample of the initial sample data set can also be preprocessed, and the preprocessing includes but is not limited to data deduplication, screening, manual error correction, null value assignment, etc.

[0037] In a preferred embodiment of the present invention, there are four main orthokeratology parameters that need to be predicted, namely: positioning arc value (AC, also known as curvature value), toric value (CP), lens total diameter (DIA) and lens drop (RP). The following determines the corresponding necessary parameters based on the correlation between these orthokeratology parameters and the examination parameters.

[0038] Please refer to Table 1. It can be seen from Table 1 that the detection parameters that have a greater impact on the value of the loop curvature, namely the necessary parameters, include the flat K value, the steep K value, and the flat e value, and the non-essential parameters include AC and Sag height difference. It should be noted that it is more accurate to analyze the difference between the two meridians from the height difference of the corresponding half chord (determined based on the flat K value and the steep K value), and in other embodiments, if there is a Sag height difference, the CP value can be directly calculated by the Sag height difference.

[0039] Table 1. Related parameters affecting the value of loop curvature

[0040] Please refer to Table 2. It can be seen from Table 2 that the detection parameters that have a greater impact on the positioning arc value, namely the necessary parameters, include the flat e value, the visible iris diameter (HVID, which directly affects the total diameter of the lens), the flat K value, the steep K value, the refractive power, and the toric state. In addition, the specific definitions of the positioning arcs of VST-designed orthokeratology lenses of different brands may not be the same, but the rules of the same brand are the same. Therefore, the brand of orthokeratology lenses will also have a greater impact on the prediction of the positioning arc value.

[0041] Table 2. Related parameters affecting positioning arc value

[0042] Please refer to Table 3. It can be seen from Table 3 that the detection parameters that have a greater impact on the total diameter of the lens, i.e., the necessary parameters, include the visible iris transverse diameter, flat e-value, flat K value, steep K value, refractive power, toric curvature status, and positioning arc value (the positioning arc value and its influencing parameters, flat e-value, flat K value, steep K value, refractive power, toric curvature status, and orthokeratology lens brand will have a greater impact on the total diameter of the lens).

[0043] Table 3. Parameters affecting the total diameter of the lens

[0044] Please refer to Table 4. It can be seen from Table 4 that the detection parameters that have a greater impact on the lens reduction, namely the necessary parameters, include flat K value, refractive power, intraocular pressure and positioning arc value, and the non-essential parameters include intraocular pressure value and central corneal thickness value.

[0045] Table 4. Related parameters affecting lens drop

[0046] Based on the influence of the above-mentioned parameters that need to be predicted, the brand of orthokeratology lenses, flat e-value, visible iris transverse diameter, flat K value, steep K value, diopter, intraocular pressure, toric curvature status, positioning arc value, toric curvature value, total lens diameter and lens reduction are determined to be necessary parameters. Among them, the brand of orthokeratology lenses, flat e-value, visible iris transverse diameter, flat K value, steep K value, diopter and intraocular pressure are mainly used as input parameters, the positioning arc value and toric curvature status can be used as input parameters as well as comparison parameters, and the toric curvature value, total lens diameter and lens reduction are mainly used as comparison parameters, that is, they are compared with the output results of the model to update the internal parameters of the model.

[0047] Therefore, if a sample in an initial sample data set does not contain any one or more of the above necessary parameters, the sample is recorded as a target sample and cannot complete the training of the model and needs to be deleted.

[0048] S120 , preprocessing the basic sample data set to obtain a preliminary sample data set, where each sample in the preliminary sample data set is recorded as a basic sample.

[0049] When training artificial neural networks for parameter prediction of orthokeratology lenses, a large amount of clinical case data is used for training. Due to the diversity of human eye parameters, the difficulty in standardizing and evaluating corneal biomechanical parameters, and the complexity of multi-dimensional eye examination parameter combinations, the optimal fitting effect of orthokeratology lenses will generally vary to a certain extent. In other words, not all case experiences are optimal. Due to the particularity and unquantifiable characteristics of factors such as corneal biomechanics and eyelid tension, not all patients can achieve the best fit. At the same time, the fitting effect is also closely related to the scale of the hospital and the number and professionalism of professionals. Therefore, the data used for training requires very complex preprocessing, and various extreme corneal morphologies, anti-rule and anti-regular fitting parameters, etc. need to be screened, and it is necessary to ensure that the proportion is not too high and the overall fitting data is not positively affected.

[0050] In a preferred embodiment of the present invention, the preprocessing is to filter the sample data (referred to as samples to be screened) in the basic sample data set that exceeds the physiological index and the set regular threshold, thereby obtaining a preliminary selection data set.

[0051] Please refer to Table 5, which shows the screening of theoretical abnormal data. When the sample to be screened falls within any unacceptable range shown in Table 5, the sample to be screened is discarded.

[0052] Table 5. Theoretical abnormal data screening table

[0053] Specifically, if the sample to be screened has one or more of the following conditions, the sample to be screened is filtered out: Case 1: FK<38 or FK>48; Case 2: em<0 or em>1; Case 3: AC<38 or AC>48; If the sample to be screened has one or more of the following parameters and one or more of the following situations, the sample to be screened is filtered out: Situation 4: IOP ≤ 7 or IOP ≥ 24; Scenario 5: ; Case 6: SK<39 or SK>50; Case 7: ΔK<0 or ΔK>5; Scenario 8: HVID<9 or HVID>15; Case 9: RP<0 or RP>6; Case 10: DIA<9 or DIA>12; Scenario 11: CP<0 or CP>4; Please refer to Table 6, which shows the screening of regular abnormal data. When the sample to be screened falls within the unacceptable range of any item in Table 6, the sample to be screened is eliminated.

[0054] Table 6. Regular abnormal data screening table

[0055] If one or more of the following situations exist, the sample to be screened is filtered out: Case 1: When 0.55≤em≤1, ; Case 2: When 0≤em≤0.45, ; Case 2: When 0.45<em<0.55, ; Among them, FK is the flat K value, em is the flat e value, and AC is the positioning arc value.

[0056] If the sample to be screened still has spherical power and drop, or astigmatism and toric curvature, or corneal diameter and total diameter, and one or more of the following situations exist, the sample to be screened is filtered out: Case 4: RP+FK-AC-F<-0.25 or RP+FK-AC-F>1.25; Case 5: HVID-DIA<0.4 or HVID-DIA>2.2; Scenario 6: ; Among them, F is the spherical power, RP is the reduction, HVID is the corneal diameter, DIA is the total diameter, CP is the toric curvature, and ΔK is the astigmatism degree.

[0057] Through the above-mentioned theoretical anomaly processing and regular anomaly processing, various extreme corneal morphologies, adaptation parameters that go against the rules and laws, etc. can be eliminated to obtain a preliminary sample data set, which can be used to train the neural network model.

[0058] S130, constructing an initial artificial neural network model, inputting the orthokeratology lens brand, flat e value, visible iris transverse diameter, flat K value, steep K value, diopter, intraocular pressure, toric curvature status and positioning arc value in the basic sample into the initial artificial neural network model, comparing the toric curvature value, positioning arc value, total lens diameter and lens drop with the toric curvature prediction value, positioning arc prediction value, total lens diameter prediction value and lens drop prediction value output by the initial artificial neural network model, completing the preliminary training of the initial artificial neural network model and obtaining a preliminary model.

[0059] In a preferred embodiment of the present invention, the data to be predicted is split into multiple dimensions, and different prediction models are constructed according to each data to be predicted. Each prediction model performs its own personalized data indicator association, and then performs distributed training and uses them in combination. The distributed training of multiple prediction models allows each prediction data to be adjusted separately, and each prediction model does not affect each other, while the established data association still exists in the most concise and core form.

[0060] First, it is necessary to conduct preliminary training on each initial artificial neural network model to obtain each preliminary model, and then use each preliminary model to screen the preliminary sample data set again, eliminate the abnormal fitting data, and obtain the final sample data set. Use the final sample data set to train each preliminary model to obtain each prediction model.

[0061] Specifically, a first initial artificial neural network model, a second initial artificial neural network model, a third initial artificial neural network model and a fourth initial artificial neural network model can be constructed respectively, and used to predict the torus value, the positioning arc value, the total diameter value of the lens and the lens drop value respectively. After training, the first preliminary model, the second preliminary model, the third preliminary model and the fourth preliminary model are obtained respectively.

[0062] Based on the necessary parameters corresponding to the above prediction parameters, the preliminary training of each initial artificial neural network model is completed, specifically: The loop curvature value is predicted and the first preliminary model is obtained after training, including: Inputting the flat e value, flat K value and steep K value in the basic sample into the first initial artificial neural network model; Compare the first loop curvature prediction value output by the first initial artificial neural network model with the loop curvature value in the basic sample, as well as the first loop curvature state prediction value output by the first initial artificial neural network model with the loop curvature state in the basic sample, and use the gradient descent algorithm to update the parameters of the first initial artificial neural network model until the training of the first initial artificial neural network model is completed to obtain a first preliminary model.

[0063] In other embodiments, the positioning arc value in the basic sample can also be input into the first initial artificial neural network model together with the flat K value, steep K value, and flat e value in the basic sample to train the first initial artificial neural network model.

[0064] It should be noted that if non-essential parameters are trained together with necessary parameters, the prediction model finally obtained in actual application, that is, when the actual eye detection data is input into the prediction model to obtain the corneal shaping parameters corresponding to the actual eye detection data, the actual eye detection data should also include these non-essential parameters, otherwise the output result may not be accurate, or these prediction models trained with non-essential parameters are divided into two types, one is a prediction model obtained by training purely with necessary parameters, and the other is a prediction model obtained by training necessary parameters and non-essential parameters together. The following other prediction models involving non-essential parameters are similar.

[0065] The positioning arc value is predicted and the second prediction model is obtained after training, including: Input the flat K value, steep K value, flat e value, visible iris horizontal diameter, diopter, orthokeratology lens brand and toric state in the basic sample into the second initial artificial neural network model, wherein the toric state is 0 or 1. When the toric state is 0, it means that there is no toric value in the basic sample or the toric value is 0. When the toric state is 1, it means that the toric value in the basic sample is non-zero. The first positioning arc prediction value output by the second initial artificial neural network model is compared with the positioning arc value in the basic sample, and the parameters of the second initial artificial neural network model are updated using a gradient descent algorithm until the training of the second initial artificial neural network model is completed to obtain a second preliminary model.

[0066] The total diameter of the lens is predicted, and the third prediction model is obtained after training, including: Input the visible iris transverse diameter, flat K value, steep K value, flat e value, diopter, orthokeratology lens brand, toric curvature state and positioning arc value in the basic sample into the third initial artificial neural network model; The predicted value of the first lens total diameter output by the third initial artificial neural network model is compared with the total diameter value of the lens in the basic sample, and the parameters of the third initial artificial neural network model are updated using a gradient descent algorithm until the training of the third initial artificial neural network model is completed to obtain a third preliminary model.

[0067] The predicted value of the total diameter of the lens needs to be magnified accordingly before it can be compared with the total diameter value of the lens in the basic sample. In the embodiment of the present invention, the magnification method is: DIA 原 ×10-90, DIA 原 The predicted value of the total diameter of the first lens output by the third initial artificial neural network model.

[0068] The lens drop value is predicted, and the fourth prediction model is obtained after training, including: Inputting the flat K value, diopter, intraocular pressure and positioning arc value in the basic sample into the fourth initial artificial neural network model; The first lens degradation prediction value output by the fourth initial artificial neural network model is compared with the lens degradation in the basic sample, and the parameters of the fourth initial artificial neural network model are updated using a gradient descent algorithm until the training of the fourth initial artificial neural network model is completed to obtain a fourth preliminary model.

[0069] When there is a central corneal thickness value in the basic sample, the central corneal thickness value in the basic sample can also be input into the fourth initial artificial neural network model together with the flat K value, refractive power, intraocular pressure and positioning arc value in the basic sample to train the fourth initial artificial neural network model.

[0070] Training each initial artificial neural network model mainly includes the following steps: 1. Divide the final sample data set into a training set and a validation set.

[0071] 2. Set the random seed, for example 42, to ensure the consistency of training results.

[0072] 3. Construct each initial artificial neural network model: In a preferred embodiment of the present invention, a multi-layer perceptron (MLP) model and a relu activation function are used, a random normal distribution is used for initialization of the weights, the learning rate is set to 0.001, and the Adam optimizer is used to adjust the gradient descent algorithm to converge quickly, thereby reducing the training time.

[0073] 4. Perform model training: epochs: The number of complete passes through the training data set is 350. batch size: The training data is divided into smaller segments, each with 32 data items.

[0074] 5. Model saving: Convert the model to tflite format (this format is convenient for apps to use the model). Save the tflite model to a specified folder.

[0075] 6. Model Validation: Load the validation set data, perform model inference, obtain the model output, calculate the error, and adjust the number of neural network layers, number of neurons, learning rate, epochs, and batch size based on the error.

[0076] Taking AC prediction as an example, in order to solve the accuracy and efficiency problems existing in traditional methods, and to make the process of predicting AC more automated and accurate in practical applications, it is helpful for research and application in the field of ophthalmic medicine.

[0077] 1. Improved computing performance and efficiency: Model structure optimization: Keras provides a simple way to stack layers to achieve complex model architectures. This modular design enables us to quickly build and optimize models, improving the computational efficiency of training and reasoning.

[0078] Automatic differentiation: The automatic differentiation mechanism in TensorFlow can significantly speed up the gradient calculation during model training and improve computing performance.

[0079] 2. Improved prediction accuracy: Feature learning: Deep neural networks can automatically learn complex feature representations from input data and can more accurately capture the underlying patterns in the data when processing multi-dimensional features (such as em, HVID, FK, SK, refraction, CP).

[0080] Nonlinear modeling capabilities: By adding multiple hidden layers and activation functions, the model can effectively capture nonlinear relationships in the data, thereby improving the accuracy of predictions.

[0081] 3. Improved stability: Regularization technology: During the model building process, we use random normal distribution to initialize the weights to prevent overfitting, enhance the model's predictive ability on unknown data, and improve stability.

[0082] Optimization algorithm: Using the Adam algorithm can effectively adjust model parameters and improve the convergence speed and stability of the model.

[0083] 4. Improved user experience Response speed: Optimized deep learning models can make predictions in a shorter time, thereby increasing the system's response speed and improving the user experience.

[0084] Precision control: Ensure the reliability and accuracy of the system in actual application through accurate AC prediction.

[0085] S140. Input the orthokeratology lens brand, flat e value, visible iris transverse diameter, flat K value, steep K value, diopter, intraocular pressure, toric curvature status and positioning arc value in the basic sample into the preliminary model, compare the toric curvature value, positioning arc value, total lens diameter and lens drop with the toric curvature prediction value, positioning arc prediction value, total lens diameter prediction value and lens drop prediction value output by the preliminary model, filter out discrete data to obtain the final sample data set, use the final sample data set to perform final training on the preliminary model to obtain a prediction model.

[0086] In order to ensure the accuracy of the prediction model, in a preferred embodiment of the present invention, the preliminary sample data set is also screened for fitting abnormal data, that is, after eliminating the fitting abnormal data (also called discrete data), the preliminary model is re-trained using the final sample data set, so that the prediction model is more in line with reality.

[0087] Specifically, the relevant data of each basic sample are respectively input into the first preliminary model, the second preliminary model, the third preliminary model and the fourth preliminary model, and the second torus prediction value, the second positioning arc prediction value, the second lens total diameter prediction value and the second lens drop prediction value are respectively obtained through the forward propagation algorithm, and then the relationship between the torus value, positioning arc value, lens total diameter and lens drop in the basic sample and the second torus prediction value, the second positioning arc prediction value, the second lens total diameter prediction value and the second lens drop prediction value are respectively compared, and the target basic samples with fitting abnormalities are filtered out to obtain the final sample data set.

[0088] Specifically, the flat e value, the flat K value, and the steep K value in the basic sample are input into the first preliminary model to obtain the second loop curvature prediction value output by the first preliminary model.

[0089] The flat K value, steep K value, flat e value, visible iris transverse diameter, refractive power, orthokeratology lens brand and toric curvature status in the basic sample are input into the second preliminary model to obtain the second positioning arc prediction value output by the second preliminary model.

[0090] The visible iris transverse diameter, flat K value, steep K value, flat e value, diopter, orthokeratology lens brand, toric curvature status and positioning arc value in the basic sample are input into the third preliminary model to obtain the predicted value of the second lens total diameter output by the third preliminary model.

[0091] The flat K value, refractive power, intraocular pressure and positioning arc value in the basic sample are input into the fourth preliminary model to obtain the second lens drop prediction value output by the fourth preliminary model.

[0092] When any one or more of the following situations exist in the basic sample, the basic sample is defined as the target basic sample: Case 1: The absolute value of the difference between the second loop curvature prediction value and the corresponding loop curvature value in the basic sample is greater than 0.75; Case 2: The absolute value of the difference between the second positioning arc prediction value and the corresponding positioning arc value in the basic sample is greater than 0.62; Case 3: the absolute value of the difference between the predicted value of the second lens total diameter and the corresponding lens total diameter in the basic sample is greater than 0.3; Scenario 4: The absolute value of the difference between the predicted value of the second lens drop and the lens drop in the corresponding basic sample is greater than 1; The target basic samples are screened out to obtain a final sample data set.

[0093] After the final sample data set is obtained, the final sample data set is used to train each preliminary model in a similar manner to the above step S130 to obtain the final first prediction model, the second prediction model, the third prediction model and the fourth prediction model. Of course, in other embodiments, the final sample data set can also be used to train each initial artificial neural network model to obtain each prediction model.

[0094] S150, inputting actual eye detection data into the prediction model to obtain corneal shaping parameters corresponding to the actual eye detection data.

[0095] In practical applications, the above four prediction models are used in combination, and the four prediction models obtained by training with necessary parameters are taken as an example. Therefore, the actual eye detection data must include the brand of orthokeratology lenses, flat e value, visible iris transverse diameter, flat K value, steep K value, refractive power, and intraocular pressure. The input parameter toric curvature state required by the second prediction model and the third prediction model, and the input parameter positioning arc value required by the third prediction model and the fourth prediction model can be obtained through the first prediction model and the second prediction model respectively. Therefore, the prediction order is toric curvature value prediction, positioning arc prediction, and lens total diameter prediction and lens reduction prediction (the order of the latter two can be swapped).

[0096] Specifically, the process of completing the actual prediction may include: 1. Input the actual flat K value, the actual steep K value, and the actual flat e value in the actual eye detection data into the first prediction model, and output the torus value prediction data and the torus state prediction data. It should be noted that the characterization method of the final torus value prediction data is F+N, where N is the specific value of the torus value prediction data output by the second prediction model minus the diopter, and F represents the diopter, which does not display the specific value, but only displays the letter F.

[0097] 2. Input the actual flat K value, actual steep K value, actual flat e value, actual visible iris diameter, actual refractive power, orthokeratology lens brand in the actual eye detection data, and the toric curvature state prediction data output by the first prediction model into the second prediction model, and output the positioning arc prediction data.

[0098] 3. The actual visible iris transverse diameter, actual flat K value, actual steep K value, actual flat e value, actual diopter, orthokeratology lens brand in the actual eye detection data, as well as the toric state prediction data output by the first prediction model and the positioning arc prediction data output by the second prediction model are input into the third prediction model to output the total lens diameter prediction data. The total lens diameter prediction data output by the third prediction model also needs to be magnified according to the above magnification to obtain the final total lens diameter prediction data.

[0099] 4. The actual flat K value, actual refractive power, actual intraocular pressure in the actual eye detection data and the positioning arc prediction data output by the second prediction model are input into the fourth prediction model to output the lens drop prediction data.

[0100] The above-mentioned toric curvature value prediction data, positioning arc prediction data, lens total diameter prediction data and lens reduction prediction data constitute the corneal reshaping parameters for preparing orthokeratology lenses. For different brands, it may be necessary to make corresponding changes based on the above-mentioned prediction data, and these data themselves can cover the parameter requirements of most brand companies for the preparation of orthokeratology lenses. Embodiment 2

[0101] Please refer to Figure 2 As shown, Figure 2 FIG. 1 is a schematic diagram of a corneal shaping parameter prediction system based on an artificial neural network disclosed in an embodiment of the present invention. Figure 2 As shown, the corneal reshaping parameter prediction system based on artificial neural network includes: The acquisition unit 210 is used to acquire an initial sample data set and determine the parameters of the initial sample data set. If the target sample in the initial sample data set does not contain the necessary parameters, the target sample is deleted to obtain a basic sample data set. The necessary parameters include orthokeratology lens brand, flat e value, visible iris transverse diameter, flat K value, steep K value, diopter, intraocular pressure, toric state, positioning arc value, toric value, lens total diameter and lens drop; A preprocessing unit 220 is used to preprocess the basic sample data set to obtain a preliminary sample data set, where each sample in the preliminary sample data set is recorded as a basic sample; The preliminary training unit 230 is used to construct an initial artificial neural network model, input the orthokeratology lens brand, flat e value, visible iris transverse diameter, flat K value, steep K value, diopter, intraocular pressure, toric curvature state and positioning arc value in the basic sample into the initial artificial neural network model, compare the toric curvature value, positioning arc value, total lens diameter and lens drop with the toric curvature prediction value, positioning arc prediction value, total lens diameter prediction value and lens drop prediction value output by the initial artificial neural network model, complete the preliminary training of the initial artificial neural network model, and obtain a preliminary model; The final training unit 240 is used to input the orthokeratology lens brand, flat e value, visible iris transverse diameter, flat K value, steep K value, diopter, intraocular pressure, toric curvature state and positioning arc value in the basic sample into the preliminary model, compare the toric curvature value, positioning arc value, total lens diameter and lens drop with the toric curvature prediction value, positioning arc prediction value, total lens diameter prediction value and lens drop prediction value output by the preliminary model, filter out discrete data to obtain a final sample data set, and use the final sample data set to perform final training on the preliminary model or the initial artificial neural network model to obtain a prediction model; The prediction unit 250 is used to input the actual eye detection data into the prediction model to obtain the corneal shaping parameters corresponding to the actual eye detection data. Embodiment 3

[0102] See also Figure 3 , Figure 3A schematic diagram of an electronic device that can be used to implement an embodiment of the present invention is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processing, cellular phones, smart phones, wearable devices (such as helmets, glasses, watches, etc.) and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the embodiments of the present invention described herein or / and claimed.

[0103] like Figure 3 As shown, the electronic device includes at least one processor 310, and a memory connected to the at least one processor 310, such as a ROM (read-only memory) 320, a RAM (random access memory) 330, etc., wherein the memory stores a computer program that can be executed by at least one processor, and the processor 310 can perform various appropriate actions and processes according to the computer program stored in the ROM 320 or the computer program loaded from the storage unit 380 to the random access memory RAM 330. In the RAM 330, various programs and data required for the operation of the electronic device can also be stored. The processor 310, the ROM 320, and the RAM 330 are connected to each other via a bus 340. The I / O (input / output) interface 350 is also connected to the bus 340.

[0104] A number of components in the electronic device are connected to the I / O interface 350, including: an input unit 360, such as a keyboard, a mouse, etc.; an output unit 370, such as various types of displays, speakers, etc.; a storage unit 380, such as a disk, an optical disk, etc.; and a communication unit 390, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 390 allows the electronic device to exchange information / data with other devices through a computer network such as the Internet or / and various telecommunication networks.

[0105] The processor 310 may be any general or / and dedicated processing component with processing and computing capabilities. Some examples of the processor 310 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various dedicated artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, a digital signal processor (DSP), and any appropriate processor, controller, microcontroller, etc. The processor 310 executes one or more steps of a corneal shaping parameter prediction method based on an artificial neural network described in the first embodiment above.

[0106] In some embodiments, a method for predicting corneal shaping parameters based on an artificial neural network may be implemented as a computer program, which is tangibly contained in a computer-readable storage medium, such as a storage unit 380. In some embodiments, part or all of the computer program may be loaded into or / and installed on an electronic device via a ROM 320 and / or a communication unit 390. When the computer program is loaded into the RAM 330 and executed by the processor 310, one or more steps of a method for predicting corneal shaping parameters based on an artificial neural network described in the first embodiment above may be executed. Alternatively, in other embodiments, the processor 310 may be configured to execute a method for predicting corneal shaping parameters based on an artificial neural network in any other appropriate manner (e.g., by means of firmware).

[0107] Various embodiments of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chips (SOCs), load programmable logic devices (CPLDs), computer hardware, firmware, software, or / and combinations thereof. These various embodiments may include: being implemented in one or more computer programs, which may be executed or / and interpreted on a programmable system including at least one programmable processor, which may be a special purpose or general purpose programmable processor, which may receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.

[0108] The computer programs for implementing the methods of the embodiments of the present invention may be written in any combination of one or more programming languages. These computer programs may be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, so that when the computer programs are executed by the processor, the functions / operations specified in the flow charts and / or block diagrams are implemented. The computer programs may be executed entirely on the machine, partially on the machine, partially on the machine and partially on a remote machine as a stand-alone software package, or entirely on a remote machine or server.

[0109] In the context of an embodiment of the present invention, a computer-readable storage medium may be a tangible medium that may contain or store a computer program for use by or in conjunction with an instruction execution system, device, or equipment. A computer-readable storage medium may include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, device, or equipment, or any suitable combination of the foregoing. Alternatively, a computer-readable storage medium may be a machine-readable signal medium. A more specific example of a machine-readable storage medium may include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0110] To provide interaction with a user, the systems and techniques described herein may be implemented on an electronic device having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and a pointing device (e.g., a mouse or trackball) through which the user can provide input to the electronic device. Other types of devices may also be used to provide interaction with the user; for example, the feedback provided to the user may be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user may be received in any form (including acoustic input, voice input, or tactile input).

[0111] The systems and techniques described herein may be implemented in a computing system that includes backend components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes frontend components (e.g., a user computer with a graphical user interface or a web browser through which a user can interact with implementations of the systems and techniques described herein), or a computing system that includes any combination of such backend components, middleware components, or frontend components. The components of the system may be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: a local area network (LAN), a wide area network (WAN), a blockchain network, and the Internet.

[0112] A computing system may include a client and a server. The client and the server are generally remote from each other and usually interact through a communication network. The client and server relationship is generated by computer programs running on the corresponding computers and having a client-server relationship with each other. The server may be a cloud server, also known as a cloud computing server or cloud host, which is a host product in the cloud computing service system to solve the defects of difficult management and weak business scalability in traditional physical hosts and VPS services.

[0113] The above is a detailed introduction to a corneal shaping parameter prediction method based on an artificial neural network disclosed in the present invention. Specific examples are used in this article to illustrate the principles and implementation methods of the present invention. The description of the above embodiments is only used to help understand the method of the present invention and its core idea; at the same time, for those skilled in the art, according to the idea of ​​the present invention, there will be changes in the specific implementation methods and application scopes. In summary, the content of this specification should not be understood as a limitation on the present invention.

Claims

1. A method for predicting corneal orthokeratology parameters based on artificial neural network, characterized in that: It includes: Acquire an initial sample data set, and determine the parameters of the initial sample data set. If the target sample in the initial sample data set does not contain the necessary parameters, delete the target sample to obtain a basic sample data set. The necessary parameters include orthokeratology lens brand, flat e value, visible iris transverse diameter, flat K value, steep K value, diopter, intraocular pressure, toric state, positioning arc value, toric value, lens total diameter, and lens drop. Preprocessing the basic sample data set to obtain a preliminary sample data set, wherein each sample in the preliminary sample data set is recorded as a basic sample; Constructing an initial artificial neural network model, inputting the orthokeratology lens brand, flat e value, visible iris transverse diameter, flat K value, steep K value, diopter, intraocular pressure, toric curvature state and positioning arc value in the basic sample into the initial artificial neural network model, and comparing the toric curvature value, positioning arc value, total lens diameter and lens drop with the toric curvature prediction value, positioning arc prediction value, total lens diameter prediction value and lens drop prediction value output by the initial artificial neural network model, completing the preliminary training of the initial artificial neural network model, and obtaining a preliminary model; Input the orthokeratology lens brand, flat e value, visible iris transverse diameter, flat K value, steep K value, diopter, intraocular pressure, toric curvature state and positioning arc value in the basic sample into the preliminary model, compare the toric curvature value, positioning arc value, total lens diameter and lens drop with the toric curvature prediction value, positioning arc prediction value, total lens diameter prediction value and lens drop prediction value output by the preliminary model, filter out discrete data to obtain a final sample data set, use the final sample data set to perform final training on the initial artificial neural network model or the preliminary model to obtain a prediction model; The actual eye detection data is input into the prediction model to obtain the corneal shaping parameters corresponding to the actual eye detection data.

2. The method for predicting corneal orthokeratology parameters based on artificial neural network according to claim 1, characterized in that: Constructing an initial artificial neural network model and training the initial artificial neural network model to obtain a preliminary model includes: Constructing a first initial artificial neural network model for predicting the loop curvature value, and obtaining a first preliminary model after training; Constructing a second initial artificial neural network model for predicting the positioning arc value, and obtaining a second preliminary model after training; Constructing a third initial artificial neural network model for predicting the total diameter value of the lens, and obtaining a third preliminary model after training; A fourth initial artificial neural network model is constructed to predict the lens drop value, and a fourth preliminary model is obtained after training.

3. The method for predicting corneal orthokeratology parameters based on artificial neural network according to claim 2, characterized in that: The loop curvature value is predicted and the first preliminary model is obtained after training, including: Inputting the flat e value, flat K value and steep K value in the basic sample into the first initial artificial neural network model; Compare the first loop curvature prediction value output by the first initial artificial neural network model with the loop curvature value in the basic sample, as well as the first loop curvature state prediction value output by the first initial artificial neural network model with the loop curvature state in the basic sample, and use the gradient descent algorithm to update the parameters of the first initial artificial neural network model until the training of the first initial artificial neural network model is completed to obtain a first preliminary model.

4. The method for predicting corneal orthokeratology parameters based on artificial neural network according to claim 2, characterized in that: The positioning arc value is predicted and the second prediction model is obtained after training, including: Inputting the flat K value, steep K value, flat e value, visible iris transverse diameter, diopter, orthokeratology lens brand and toricity state in the basic sample into the second initial artificial neural network model; The first positioning arc prediction value output by the second initial artificial neural network model is compared with the positioning arc value in the basic sample, and the parameters of the second initial artificial neural network model are updated using a gradient descent algorithm until the training of the second initial artificial neural network model is completed to obtain a second preliminary model.

5. The method for predicting corneal orthokeratology parameters based on artificial neural network according to claim 2, characterized in that: The total diameter of the lens is predicted, and the third prediction model is obtained after training, including: Input the visible iris transverse diameter, flat K value, steep K value, flat e value, diopter, orthokeratology lens brand, toric curvature state and positioning arc value in the basic sample into the third initial artificial neural network model; The predicted value of the first lens total diameter output by the third initial artificial neural network model is compared with the total diameter value of the lens in the basic sample, and the parameters of the third initial artificial neural network model are updated using a gradient descent algorithm until the training of the third initial artificial neural network model is completed to obtain a third preliminary model.

6. The method for predicting corneal orthokeratology parameters based on artificial neural network according to claim 2, characterized in that: The lens drop value is predicted, and the fourth prediction model is obtained after training, including: Inputting the flat K value, diopter, intraocular pressure and positioning arc value in the basic sample into the fourth initial artificial neural network model; The first lens degradation prediction value output by the fourth initial artificial neural network model is compared with the lens degradation in the basic sample, and the parameters of the fourth initial artificial neural network model are updated using a gradient descent algorithm until the training of the fourth initial artificial neural network model is completed to obtain a fourth preliminary model.

7. The method for predicting corneal orthokeratology parameters based on artificial neural network according to claim 2, characterized in that: The orthokeratology lens brand, flat e value, visible iris transverse diameter, flat K value, steep K value, diopter, intraocular pressure, toric curvature state and positioning arc value in the basic sample are input into the preliminary model, and the positioning arc value, toric curvature value, total lens diameter and lens drop are compared with the positioning arc prediction value, toric curvature prediction value, total lens diameter prediction value and lens drop prediction value output by the preliminary model, and the final sample data set is obtained after the discrete data is screened out, including: Input the flat e value, the flat K value and the steep K value in the basic sample into the first preliminary model to obtain a second torus prediction value output by the first preliminary model; Input the flat K value, steep K value, flat e value, visible iris transverse diameter, diopter, orthokeratology lens brand and toric curvature state in the basic sample into the second preliminary model to obtain the second positioning arc prediction value output by the second preliminary model; Input the visible iris transverse diameter, flat K value, steep K value, flat e value, diopter, orthokeratology lens brand, toric curvature state and positioning arc value in the basic sample into the third preliminary model to obtain the predicted value of the second lens total diameter output by the third preliminary model; Input the flat K value, diopter, intraocular pressure and positioning arc value in the basic sample into the fourth preliminary model to obtain the second lens drop prediction value output by the fourth preliminary model; The relationship between the torus value, positioning arc value, total lens diameter and lens drop in the basic sample and the second torus prediction value, second positioning arc prediction value, second lens total diameter prediction value and second lens drop prediction value are compared respectively, the target basic sample is filtered out, and the final sample data set is obtained.

8. The method for predicting corneal orthokeratology parameters based on artificial neural network according to claim 7, characterized in that: The relationship between the loop value, positioning arc value, lens total diameter and lens drop in the basic sample and the second loop value prediction value, second positioning arc prediction value, second lens total diameter prediction value and second lens drop prediction value are respectively compared, and the target basic sample is filtered out to obtain the final sample data set, including: When any one or more of the following situations exist in the basic sample, the basic sample is defined as the target basic sample: Case 1: The absolute value of the difference between the second loop curvature prediction value and the corresponding loop curvature value in the basic sample is greater than 0.75; Case 2: The absolute value of the difference between the second positioning arc prediction value and the corresponding positioning arc value in the basic sample is greater than 0.62; Case 3: the absolute value of the difference between the predicted value of the second lens total diameter and the corresponding lens total diameter in the basic sample is greater than 0.3; Scenario 4: The absolute value of the difference between the predicted value of the second lens drop and the lens drop in the corresponding basic sample is greater than 1; The target basic samples are screened out to obtain a final sample data set.

9. The method for predicting corneal orthokeratology parameters based on artificial neural network according to claim 7, characterized in that: The initial artificial neural network model or the preliminary model is finally trained using the final sample data set to obtain a prediction model, including: Using the final sample data set to train the first preliminary model, the second preliminary model, the third preliminary model and the fourth preliminary model respectively, to obtain a first prediction model, a second prediction model, a third prediction model and a fourth prediction model respectively; Alternatively, the first initial artificial neural network model, the second initial artificial neural network model, the third initial artificial neural network model and the fourth initial artificial neural network model are trained respectively using the final sample data set to obtain a first prediction model, a second prediction model, a third prediction model and a fourth prediction model respectively.

10. The method for predicting corneal orthokeratology parameters based on artificial neural network according to claim 9, characterized in that: Inputting actual eye detection data into the prediction model to obtain corneal shaping parameters corresponding to the actual eye detection data includes: Inputting the actual flat K value, the actual steep K value, and the actual flat e value in the actual eye detection data into the first prediction model, and outputting the loop curvature value prediction data and the loop curvature state prediction data; Input the actual flat K value, the actual steep K value, the actual flat e value, the actual visible iris transverse diameter, the actual diopter, the orthokeratology lens brand and the toric curvature state prediction data output by the first prediction model in the actual eye detection data into the second prediction model, and output the positioning arc prediction data; Input the actual visible iris transverse diameter, actual flat K value, actual steep K value, actual flat e value, actual diopter, orthokeratology lens brand in the actual eye detection data, as well as the toric state prediction data output by the first prediction model and the positioning arc prediction data output by the second prediction model into the third prediction model, and output the total lens diameter prediction data; Input the actual flat K value, actual refractive power, actual intraocular pressure in the actual eye detection data and the positioning arc prediction data output by the second prediction model into the fourth prediction model, and output the lens drop prediction data; The toric curvature value prediction data, positioning arc prediction data, lens total diameter prediction data and lens drop prediction data constitute the corneal shaping parameters for preparing orthokeratology lenses.

Images