CN115910328A - Orthokeratology mirror fitting system based on artificial intelligence analysis - Google Patents
Orthokeratology mirror fitting system based on artificial intelligence analysis Download PDFInfo
- Publication number
- CN115910328A CN115910328A CN202211734057.5A CN202211734057A CN115910328A CN 115910328 A CN115910328 A CN 115910328A CN 202211734057 A CN202211734057 A CN 202211734057A CN 115910328 A CN115910328 A CN 115910328A
- Authority
- CN
- China
- Prior art keywords
- data
- component
- model
- label
- result
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
- Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
- Y02P90/30—Computing systems specially adapted for manufacturing
Landscapes
- Eye Examination Apparatus (AREA)
Abstract
The invention relates to the technical field of testing and matching of a plastic cornea mirror, in particular to a system for testing and matching the plastic cornea mirror based on artificial intelligence analysis, which comprises an input data assembly, a data acquisition assembly and a data processing assembly, wherein the input data assembly is used for inputting a page for front-end operation and realizing the substitution of visual light data of a patient; the application database component inputs data into the application database, stores the data and waits for the model component to call; the model prediction component automatically detects whether the application database component completes complete data entry; the result storage component stores the result after model calculation and returns the result to the application data component; and the front-end display component is used for displaying the result to a software interface and providing comparison reference of the film selection parameters for the doctor to check and match the corneal plastic lens. With the help of the intelligent AI diagnosis and treatment tool, the professional requirements of the angle plastic examination on doctors can be properly relaxed, the examination and treatment difficulty is obviously reduced, and the examination and treatment capacity and the service level of doctors, particularly interns, are improved.
Description
Technical Field
The invention relates to the technical field of testing and matching of a orthokeratology mirror, in particular to a system for testing and matching the orthokeratology mirror based on artificial intelligence analysis.
Background
According to the results of China children and teenagers myopia published by the State health Commission, the method comprises the following steps: in 2020, the total myopia rate of children and teenagers in China is 52.7%. Wherein the number of children aged 6 is 14.3%, the number of pupils is 35.6%, the number of junior and middle school students is 71.1%, and the number of senior and middle school students is 80.5%. The myopia spot-check results of children and teenagers in 6-18 years old are embarrassing and show a cross-grade doubling growth trend. Along with the serious condition of myopia of teenagers, the number of Chinese ophthalmology diagnosis and treatment people is continuously increased. However, in the field of medical ophthalmology, the objective diagnosis and treatment requirements of eye diseases are seriously mismatched with high-quality doctors and medical resources. At present, about 4 million ophthalmologists exist in China, but the problem of efficient allocation of medical resources is urgently needed to be solved in the face of hundreds of millions of children and teenagers. The testing amount of the orthokeratology lens is increased year by year, the market has higher and higher requirements on testing and matching personnel, the technology of the testing and matching personnel can be gradually improved through training and learning modes, the time consumption is too long, the traditional training cannot effectively transfer the testing and matching practical operation experience, and the culture period of testing and matching experts is long. The corner plastic fitting also faces a plurality of professional difficulties in specific operation: such as: the angle plastic testing has high requirement on the specialty of doctors, large difference of each brand, upgrading testing difficulty, incapability of testing one step, overhigh labor cost, cross infection hidden danger of repeated try-on of patients and the like.
Disclosure of Invention
The invention provides a cornea shaping mirror fitting system based on artificial intelligence analysis, which can appropriately relax the professional requirements of a doctor on the cornea shaping fitting with the help of an intelligent AI diagnosis and treatment tool, obviously reduce the fitting difficulty, and improve the fitting capability and the service level of the doctor, particularly a intern.
In order to achieve the purpose, the invention provides the following technical scheme: a cornea moulding mirror fitting system based on artificial intelligence analysis comprises an input data component, a front-end operation input page, an AI corner moulding fitting system and a control module, wherein the input data component is used for logging in the AI corner moulding fitting system which automatically substitutes the visual light data of a patient according to an examination record; the application database component inputs data into an application database, stores the data and waits for the model component to call; the model prediction component automatically detects whether the application database component completes complete data entry, if the detection data is complete, the model prediction component calls a trained model interface, and transmits necessary parameters such as a sphere lens, a cylinder lens, an axis, a cornea diameter and the like to perform data prediction to obtain a fitting parameter; the result storage component stores the result after model calculation and returns the result to the application data component; the front-end display assembly is used for displaying results to a software interface for a user to check and providing comparison reference of film selection parameters for a doctor to check and match the corneal plastic lens.
Preferably, the modeling method of the model prediction component is as follows:
s1, collecting FK, SK and e values of a corneal topography, a sphere lens, astigmatism and axis numerical value of a diopter, a front depth, an eye axis, gender and age data; the indexes include: amplitude reduction, AC, CP, diameter;
s2, splitting the data set of the S1 into a training set and a test set, wherein the test set accounts for 30%;
s3, modeling four labels in the index by applying an ETR (extract-transform-average), adaboost and GBDT (guaranteed bit rate) integrated algorithm model respectively; adaboost regression algorithm is applied to label amplitude reduction; applying an ETR regression algorithm to the label AC; the label CP applies GBDT classification algorithm; applying a GBDT regression algorithm to the label diameter; and selecting a mean square error MSE and a decision coefficient R2 which are commonly used in a machine learning regression task as performance evaluation indexes of the shaping mirror model by using the label AC, the amplitude reduction and the diameter, wherein the mean square error MSE is used for grasping the difference between a predicted value and a real value of each sample.
Preferably, the mean square error MSE and the decision coefficient R2 are calculated as follows:
wherein
Indicates the predicted value, y i Are true values.
The invention has the beneficial effects that: with the help of the intelligent AI diagnosis and treatment tool, the professional requirements of the angle plastic testing pair doctors can be properly relaxed, the testing difficulty is obviously reduced, and the testing and matching capability and the service level of the doctors, particularly the interns, are improved. In the time level, the former angle plastic fitting physician can only fit a plurality of patients every day, the fitting efficiency is low, but under the assistance of the intelligent AI diagnosis and treatment tool, the fitting time of a single patient can be reduced by more than 50%, the working intensity of the physician is greatly reduced, and the physician can provide services for more patients every day.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, it is obvious that the drawings in the following description are only some embodiments of the present invention, and for those skilled in the art, other drawings can be obtained according to the drawings without creative efforts.
FIG. 1 is a general work flow diagram of the system of the present invention;
FIG. 2 is a comparison of the predicted value and the true value of the test set 'reduced amplitude';
FIG. 3 is a comparison of predicted values and true values for a test set 'AC' of the present invention;
FIG. 4 is a graph of the diameter data distribution of the tag of the present invention;
FIG. 5 is a comparison of predicted values and true values for 'diameter' of a test set in accordance with the present invention;
FIG. 6 shows the comparison between the predicted value and the actual value of the test set 'CP' according to the present invention.
Detailed Description
The technical solutions of the present invention will be described clearly and completely with reference to the accompanying drawings, and it is to be understood that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
An orthokeratology mirror fitting system based on artificial intelligence analysis, comprising:
the input data component is a front-end operation input page and is used for logging in an AI corner plastic fitting system, and the AI corner plastic fitting system automatically substitutes the visual light data of the patient according to the check record;
the application database component inputs data into an application database, stores the data and waits for the model component to call;
the model prediction component automatically detects whether the application database component completes complete data entry, if the detection data is complete, the model prediction component calls a trained model interface and transmits necessary parameters such as a sphere lens, a cylinder lens, an axis, a cornea diameter and the like to perform data prediction to obtain a fitting parameter;
the result storage component stores the result after model calculation and returns the result to the application data component;
the front-end display assembly is used for displaying results to a software interface for a user to check and providing comparison reference of film selection parameters for a doctor to check and match the corneal plastic lens.
The modeling method of the model prediction component comprises the following steps:
s1, collecting FK, SK and e values of a corneal topography, a sphere lens of a dioptric power, astigmatism and an axis numerical value, a front depth, an eye axis, gender and age data; the indexes include: amplitude reduction, AC, CP, diameter;
s2, splitting the data set of the S1 into a training set and a test set, wherein the test set accounts for 30%;
s3, modeling four labels in the index by applying an ETR (extract-transform-average), adaboost and GBDT (guaranteed bit rate) integrated algorithm model respectively; adaboost regression algorithm is applied to label amplitude reduction; applying an ETR regression algorithm to the label AC; the label CP applies GBDT classification algorithm; applying a GBDT regression algorithm to the label diameter;
and selecting a mean square error MSE and a decision coefficient R2 which are commonly used in a machine learning regression task as performance evaluation indexes of the shaping mirror model by using the label AC, the amplitude reduction and the diameter, wherein the mean square error MSE is used for grasping the difference between a predicted value and a real value of each sample. If the fitting effect of the model on a part of data is good, but the fitting effect on a part of data is poor, the overall MSE calculation result of the model is good, but the overall generalization capability of the model is not enough, so that the model does not grasp the characteristics of the part of data, the phenomenon can be well measured by using the R2 performance index, and the closer R2 is to 1, the better the fitting of the model to the overall data is.
The mean square error MSE and the decision coefficient R2 are calculated as follows:
wherein
Indicates the predicted value, y i Are true values.
And directly selecting the accuracy as an evaluation index of the model performance for the label CP converted into the classification.
The label reduction prediction results are as follows:
r2 and MSE of the Adaboost model on the training set are 0.973 and 0.048 respectively, and R2 and MSE on the testing set are 0.935 and 0.088 respectively, which shows that the Adaboost model completely learns the relationship between the label reduction and the feature.
The label reduction is distributed in discrete values in actual inspection, so discrete processing needs to be carried out on a predicted continuous value, and according to a value which is possibly generated in the actual label reduction, a predicted value of the label reduction is filed in the discrete value of the label reduction closest to the actual label reduction. After modeling and prediction are carried out by adopting an Adaboost model, the R2 and the MSE of the processed label amplitude reduction in a training set are respectively 0.957 and 0.076, and the R2 and the MSE in a testing set are respectively 0.928 and 0.099. The prediction result of the discretized Adaboost model on the label reduction is still good. Fig. 2 shows the result difference between the discretized predicted value and the true value after modeling by using the Adaboost model.
The predicted result of tag AC is as follows:
the generalization ability of the ETR model can be ensured by double random sampling of the ETR model on samples and characteristics, the learning ability of the model is enhanced to a certain extent, and the data relationship between the label AC and each characteristic can be better grasped, and fig. 3 shows the result difference between the predicted value and the true value of the test set after modeling by adopting the ETR model.
The predicted label diameter is as follows:
r2 and MSE of the GBDT model on a training set are 0.757 and 0.048 respectively, R2 and MSE on a test set are 0.626 and 0.026 respectively, but compared with the prediction results of tag AC and tag amplitude reduction, the prediction results of tag AC and tag amplitude reduction R2 are both above 0.9, but the MSE of the prediction results of the diameter is smaller than that of the AC and amplitude reduction mainly because of the problem of data distribution, FIG. 4 is a data distribution diagram of the tag diameter, as can be seen from the diagram, the numerical change difference of the diameter is small and concentrated, 3200 samples are mainly distributed between 10.2 and 11.2, and the number of repeated values is too large, and the model is difficult to learn the specific distribution characteristics of the data, so that R2 is low, and the variation interval is small, so that the MSE of the model prediction results is small. The GBDT model can flexibly process continuous and discrete data, and has stronger robustness to abnormal values, so that the prediction result on the diameter of the label is better than that of other models. FIG. 5 shows the difference between the predicted value and the true value of the test set after the GDBT model is adopted for modeling.
The predicted results of the label CP are as follows:
the label CP needs to be subjected to other processing, the original label CP data distribution is in a discrete state, the data numerical values represent a small number of categories and are relatively concentrated, and therefore the prediction of the label CP cannot be treated as continuous value regression, and the prediction of the label CP needs to be converted from regression to classification. And performing oversampling on the minority class by using a Borderline-SMOTE method, wherein the Borderline-SMOTE is an oversampling algorithm improved on the basis of SMOTE, and only the minority class samples on the boundary are used for synthesizing a new sample, so that the class distribution of the sample is improved.
The data set before and after the label CP is over-sampled is also divided into a training set and a testing set according to the proportion of (7:3), and a GBDT classifier suitable for the characteristics of discrete data is used for modeling. The accuracy is used as the performance evaluation index of the model, the accuracy of the model on a training set is as high as 0.983, the accuracy of the model on a testing set is 0.961, and the prediction result of the model is close to 1. FIG. 6 shows the difference between the predicted value and the true value of the test set after the GDBT model is adopted for modeling.
The data utilized above can be implemented by a orthokeratology fitting table, the fields of which include: gender, age, flat K-value, steep K-value, e-value, corneal diameter, sphere, cylinder, axis, anterior chamber depth, etc.; and the corresponding relation of the fields is as follows: different sample data are corresponding to each other, but it is possible that the fitting parameters are consistent, that is, multiple cases may correspond to one fitting parameter.
As a specific embodiment: as shown in figure 1, the orthokeratology mirror fitting system based on artificial intelligence analysis
The system comprises a 1, a user input component; 2. an application database component; 3. a model prediction component; 4. a result storage component; 5. front end display assembly
The user input assembly 1 is used for feature data collection;
the application database component 2 is used for the system to automatically input data into the application database and save, waiting for the model component to make a call.
The model prediction component 3 is used for automatically detecting whether the application database component completes complete data entry, and if the detected data is complete, the component calls the trained model interface to perform data prediction to obtain the fitting parameters (amplitude reduction, AC, CP and diameter).
The result storage component 4 is used to store the results of the model calculations, i.e. return the results to the application data component, save the new data results, which will also pass the result data to the next component.
The front end display component 5 is used for receiving the selection parameter result of the result storage component and displaying the result to the software interface for the user to view.
The user terminal 1 comprises a mobile phone, a tablet computer, a personal computer and the like, so that the diversity of the terminal is ensured, and different use requirements are met.
The working principle and the using process of the invention are as follows:
in use, a user logs in the user input component 1 after passing identity verification, and inputs characteristic items (FK, SK, e value, sphere, astigmatism, axis, front depth, eye axis, sex, age); selecting a corresponding brand, integrating the characteristic items by the application database component 2, automatically inputting the data into an application database and storing the data, waiting for calling the model component, calling the application database component 2 by the model prediction component 3 to automatically detect whether the application database component completes complete data entry, calling a trained model interface by the component if the detected data is complete, performing data prediction to obtain fitting parameters (amplitude reduction, AC, CP and diameter), storing a result after model calculation by the result storage component 4, returning the result to the application database component, storing a new data result, transmitting the result data to a chip selection parameter result of the storage component of the front-end display component 5 by the component, and displaying the result to a software interface for a user to check.
Based on the scheme, with the help of An Intelligent (AI) diagnosis and treatment tool, the system has the advantages that the professional requirements of angle plastic testing and matching physicians can be properly relaxed, the testing and matching difficulty is obviously reduced, and the testing and matching capability and the service level of the physicians, particularly the practicing physicians, are improved. In the time level, the former angle plastic fitting doctor can only fit a plurality of patients every day, the fitting efficiency is low, but under the assistance of An Intelligent (AI) diagnosis and treatment tool, the fitting time of a single patient can be reduced by more than 50%, the working strength of the doctor is greatly reduced, and the service can be provided for more patients every day.
The above description is only for the specific embodiments of the present invention, but the scope of the present invention is not limited thereto, and any person skilled in the art can easily conceive of the changes or substitutions within the technical scope of the present invention, and all the changes or substitutions should be covered within the scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims.
Claims (3)
1. A orthokeratology lens fitting system based on artificial intelligence analysis, comprising:
the input data component is a front-end operation input page and is used for logging in an AI corner fitting system, and the AI corner fitting system automatically substitutes the visual data of the patient according to the examination record;
the application database component inputs data into an application database, stores the data and waits for the model component to call;
the model prediction component automatically detects whether the application database component completes complete data entry, if the detection data is complete, the model prediction component calls a trained model interface and transmits necessary parameters such as a sphere lens, a cylinder lens, an axis, a cornea diameter and the like to perform data prediction to obtain a fitting parameter;
the result storage component stores the result after model calculation and returns the result to the application data component;
the front-end display assembly is used for displaying results to a software interface for a user to check and providing comparison reference of film selection parameters for a doctor to check and match the corneal plastic lens.
2. The orthokeratology mirror fitting system based on artificial intelligence analysis, as claimed in claim 1, wherein: the modeling method of the model prediction component comprises the following steps:
s1, collecting FK, SK and e values of a corneal topography, a sphere lens of a dioptric power, astigmatism and an axis numerical value, a front depth, an eye axis, gender and age data; the indexes include: amplitude reduction, AC, CP, diameter;
s2, splitting the data set of the S1 into a training set and a test set, wherein the test set accounts for 30%;
s3, modeling four labels in the index by applying an ETR (extract-transform-average), adaboost and GBDT (guaranteed bit rate) integrated algorithm model respectively; adaboost regression algorithm is applied to label amplitude reduction; applying an ETR regression algorithm to the label AC; the label CP applies GBDT classification algorithm; applying a GBDT regression algorithm to the label diameter;
the label AC, the amplitude reduction and the diameter are used for selecting a mean square error MSE and a decision coefficient R2 which are commonly used by a machine learning regression task as performance evaluation indexes of the shaping mirror model, and the mean square error MSE is used for grasping the difference between a predicted value and a true value of each sample.
3. The orthokeratology mirror fitting system based on artificial intelligence analysis, as claimed in claim 2, wherein: the mean square error MSE and the decision coefficient R2 are calculated according to the following formula:
wherein
Indicates the predicted value, y i Are true values. />
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211734057.5A CN115910328A (en) | 2022-12-23 | 2022-12-23 | Orthokeratology mirror fitting system based on artificial intelligence analysis |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211734057.5A CN115910328A (en) | 2022-12-23 | 2022-12-23 | Orthokeratology mirror fitting system based on artificial intelligence analysis |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115910328A true CN115910328A (en) | 2023-04-04 |
Family
ID=86480890
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211734057.5A Pending CN115910328A (en) | 2022-12-23 | 2022-12-23 | Orthokeratology mirror fitting system based on artificial intelligence analysis |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115910328A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116564540A (en) * | 2023-07-11 | 2023-08-08 | 天津师范大学 | Cornea shaping mirror parameter prediction method and system based on ensemble learning algorithm |
-
2022
- 2022-12-23 CN CN202211734057.5A patent/CN115910328A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116564540A (en) * | 2023-07-11 | 2023-08-08 | 天津师范大学 | Cornea shaping mirror parameter prediction method and system based on ensemble learning algorithm |
| CN116564540B (en) * | 2023-07-11 | 2023-09-08 | 天津师范大学 | Cornea shaping mirror parameter prediction method and system based on ensemble learning algorithm |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107680683A (en) | A kind of AI eye healths appraisal procedure | |
| CN108198620A (en) | A kind of skin disease intelligent auxiliary diagnosis system based on deep learning | |
| EP3783619A1 (en) | Human body health assessment method and system based on sleep big data | |
| CN109841267A (en) | A kind of clinical ophthalmology data collection system and method | |
| CN112700858B (en) | Early warning method and device for myopia of children and teenagers | |
| CN112472048B (en) | Method for realizing neural network for identifying pulse condition of cardiovascular disease patient | |
| CN111951965B (en) | Panoramic health dynamic monitoring and predicting system based on time sequence knowledge graph | |
| CN115910328A (en) | Orthokeratology mirror fitting system based on artificial intelligence analysis | |
| CN110335681A (en) | One kind being used for senile dementia early warning system and method for early warning | |
| CN110473631B (en) | Intelligent sleep monitoring method and system based on real world research | |
| Kabari et al. | Neural networks and decision trees for eye diseases diagnosis | |
| CN115985515A (en) | Amblyopia correction effect prediction method, device and equipment based on machine learning | |
| CN112185564B (en) | Ophthalmic disease prediction method based on structured electronic medical record and storage device | |
| CN112891164A (en) | Amblyopia rehabilitation training concentration degree assessment method in virtual reality environment | |
| CN117338234A (en) | Diopter and vision joint detection method | |
| Kumar et al. | Analysis of CNN model based classification of diabetic retinopathy diagnosis | |
| CN110960207A (en) | Tree model-based atrial fibrillation detection method, device, equipment and storage medium | |
| CN114202772B (en) | Reference information generation system and method based on artificial intelligence and intelligent medical treatment | |
| Oza et al. | Glaucoma detection using convolutional neural networks | |
| Guijarro-Berdiñas et al. | Empirical evaluation of a hybrid intelligent monitoring system using different measures of effectiveness | |
| CN113273959B (en) | Portable diabetic retinopathy diagnosis and treatment instrument | |
| Li et al. | Multi-label constitution identification based on tongue image in traditional Chinese medicine | |
| Zhang et al. | A classification method of arrhythmia based on adaboost algorithm | |
| Wang et al. | Promote Retinal Lesion Detection for Diabetic Retinopathy Stage Classification | |
| Nanarkar et al. | A Survey on Classification and identification of Arrhythmia using Machine Learning techniques |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |