A method, system, device, and computer-readable storage medium for predicting true intraocular pressure after corneal refractive surgery based on a machine learning model
Patent Information
- Application Number
- CN202611209530.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-08-11
- Publication Date
- 2026-09-15
Smart Images
Figure CN122744701A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of ophthalmic medical technology, and more specifically to a method, system, device, and computer-readable storage medium for predicting the true intraocular pressure after corneal refractive surgery based on a machine learning model. Background Technology
[0002] Corneal refractive surgeries (such as FS-LASIK and SMILE) correct refractive errors by altering the shape of the cornea. While the surgery does not directly affect aqueous humor circulation or the volume of the eye's contents, it significantly alters the central corneal thickness, curvature, and biomechanical properties. Currently, the measurements taken by routinely used non-contact tonometers (NCT), Goldmann applanation tonometers, and iCare rebound tonometers are significantly affected by these corneal parameters, resulting in postoperative intraocular pressure values that are generally lower than the actual intraocular pressure. This can lead to serious clinical problems such as: 1. Increased risk of missed diagnosis: Most people who undergo corneal refractive surgery are young and middle-aged adults, and the risk of developing primary open-angle glaucoma in the future always exists. The "normal" or "low" intraocular pressure measured after surgery may mask the actual increase or borderline high intraocular pressure, leading to missed diagnosis of ocular hypertension and even early glaucoma, thus delaying intervention.
[0003] 2. Serious clinical challenges: Currently, there is a lack of universally accepted and accurate formulas or assessment methods for correcting postoperative intraocular pressure. Physicians can only rely on experience for rough judgments or need to use invasive and complex correction references, making it difficult to conveniently and accurately assess patients' postoperative intraocular pressure status during routine follow-ups.
[0004] Currently, some studies have attempted to correct intraocular pressure based on a single or a few corneal parameters (such as central corneal thickness, CCT) using linear regression formulas. However, corneal refractive surgery alters the multi-layered and multi-dimensional comprehensive characteristics of the cornea, and simple linear models lack accuracy and have poor individualized prediction effects, thus failing to be widely applied in clinical practice.
[0005] Therefore, how to comprehensively analyze multiple parameters to accurately predict the true intraocular pressure after corneal refractive surgery is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention
[0006] In view of this, the present invention provides a method, system, device and computer-readable storage medium for predicting the true intraocular pressure after corneal refractive surgery based on a machine learning model. It can integrate diverse ocular biometric data of patients to construct a high-precision prediction model, so as to realize individualized, non-invasive and convenient prediction of the true intraocular pressure after surgery, assist clinicians in timely detection of postoperative intraocular pressure abnormalities and reduce the risk of missed diagnosis of ocular hypertension and glaucoma.
[0007] To achieve the above objectives, the present invention adopts the following technical solution: A method for predicting true intraocular pressure after corneal refractive surgery based on a machine learning model includes: Collect patient data including preoperative ocular parameters and postoperative reference intraocular pressure values; The collected patient data is preprocessed to construct a dataset; the dataset includes a training set and a test set. Using the training set, a postoperative true intraocular pressure prediction model based on machine learning was trained. The trained postoperative real intraocular pressure prediction model was validated based on the test set, and then external validation was performed using multi-center independent data to obtain the validated postoperative real intraocular pressure prediction model. The preoperative ocular parameters of the patient to be tested are processed and then input into the postoperative true intraocular pressure prediction model to obtain the true intraocular pressure value of the patient to be tested.
[0008] Optionally, the preoperative ocular parameters may include at least two or more of the following: age, sex, preoperative non-contact intraocular pressure measurement, refractive parameters, spherical power, cylindrical power, equivalent spherical power, algebraic sum of spherocytic and cylindrical power, preoperative average corneal curvature, average radius of curvature, central corneal thickness, anterior chamber depth, corneal diameter, corneal volume, anterior chamber volume, corneal diameter, anterior chamber angle, pupil diameter, axial length, anterior chamber depth, lens thickness, and horizontal corneal diameter.
[0009] Optionally, the postoperative reference intraocular pressure value is collected by measuring the intraocular pressure three times using the same standard tonometer at a specific postoperative time point and obtaining the average value of the intraocular pressure as the postoperative reference intraocular pressure value at that time point.
[0010] Optionally, the preprocessing includes cleaning the collected patient data, processing missing values and outliers, and standardizing or normalizing continuous features.
[0011] Optionally, the machine learning algorithm may be one or a combination of two of the following: random forest, decision tree, support vector machine, and artificial neural network.
[0012] Optionally, the trained postoperative real intraocular pressure prediction model is validated based on the test set. Specifically, root mean square error, mean absolute error, and coefficient of determination are used as evaluation indicators to assess the consistency and accuracy between the predicted value of the postoperative real intraocular pressure prediction model and the actual postoperative reference intraocular pressure value.
[0013] A system for predicting true intraocular pressure after corneal refractive surgery based on a machine learning model includes: The data acquisition module collects patient data including preoperative ocular parameters and postoperative reference intraocular pressure values; The data preprocessing module preprocesses the collected patient data to construct a dataset, which includes a training set and a test set. The model training and validation module uses the training set to train a postoperative real intraocular pressure prediction model based on machine learning; it validates the trained postoperative real intraocular pressure prediction model based on the test set, and then performs multi-center independent data external validation to obtain the validated postoperative real intraocular pressure prediction model. The output module processes the preoperative ocular parameters of the patient and inputs them into the postoperative true intraocular pressure prediction model to obtain the true intraocular pressure value of the patient.
[0014] An electronic device includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the computer program, implements a method for predicting true intraocular pressure after corneal refractive surgery based on a machine learning model.
[0015] A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements a method for predicting true intraocular pressure after corneal refractive surgery based on a machine learning model.
[0016] As can be seen from the above technical solutions, compared with the prior art, the present invention discloses a method, system, device, and computer-readable storage medium for predicting the true intraocular pressure after corneal refractive surgery based on a machine learning model, which has the following beneficial effects: 1. High prediction accuracy: This invention breaks through the limitations of single parameter correction, and comprehensively utilizes multi-dimensional features that reflect corneal morphology, refractive state, and ocular structure. It captures the complex nonlinear relationship between these features and postoperative intraocular pressure through machine learning algorithms, thereby providing more accurate and individualized predictions than traditional linear formulas.
[0017] 2. High clinical applicability: The required input parameters are all routine preoperative examination items, requiring no additional special examinations, making it easy to promote in clinical practice. The prediction process is fast and non-invasive, and can be directly integrated into hospital information systems or portable devices.
[0018] 3. Significant diagnostic value: It provides clinicians with an objective and quantitative reference value for postoperative intraocular pressure. By dynamically comparing the "predicted value" with the "measured value," it can effectively identify patients whose measured intraocular pressure "appears normal but is actually high-risk," greatly reducing the probability of missed glaucoma diagnosis and improving the quality of long-term postoperative follow-up.
[0019] 4. Iterative optimization: The model framework constructed in this invention has good scalability. With the inclusion of more centers and larger sample data, the model can be continuously iterated and optimized, so that its prediction performance can be continuously improved and adapted to the differences of different populations and devices. Attached Figure Description
[0020] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.
[0021] Figure 1 This is a schematic diagram of the method flow provided by the present invention; Figure 2 This is a schematic diagram of the system structure provided by the present invention; Figure 3 This is a schematic diagram illustrating the principle of the method provided by the present invention. Detailed Implementation
[0022] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0023] This invention discloses a method for predicting true intraocular pressure after corneal refractive surgery based on a machine learning model, such as... Figure 1 and Figure 3 As shown, it includes: Collect patient data including preoperative ocular parameters and postoperative reference intraocular pressure values; The collected patient data is preprocessed to construct a dataset; the dataset includes a training set and a test set. Using the training set, a postoperative true intraocular pressure prediction model based on machine learning was trained. The trained postoperative real intraocular pressure prediction model was validated based on the test set, and then external validation was performed using multi-center independent data to obtain the validated postoperative real intraocular pressure prediction model. The preoperative ocular parameters of the patient to be tested are processed and then input into the postoperative true intraocular pressure prediction model to obtain the true intraocular pressure value of the patient to be tested.
[0024] Specifically, data acquisition and preprocessing: Complete data from a batch of patients who have undergone corneal refractive surgery and whose glaucoma has been ruled out preoperatively will be collected to form a modeling dataset. The data will include at least: Preoperative parameters: Patient basic information: age, gender, preoperative non-contact intraocular pressure measurement; Refractive parameters: spherical power, cylindrical power, equivalent spherical power, spherocytic sum; Corneal topography parameters: preoperative mean corneal curvature (Km), mean radius of curvature (Rm), central corneal thickness (CCT), anterior chamber depth, corneal diameter, corneal volume, anterior chamber volume, corneal diameter, anterior chamber angle, pupil diameter. Ocular biometric parameters (obtained via Lenstar device): axial length (AL), anterior chamber depth (ACD), lens thickness (LT), horizontal corneal diameter (WTW), etc.
[0025] Specifically, it is divided into five categories: (1) Basic information: including age, eye type and gender (see Appendix Table 1, serial numbers 1-3); (2) Refractive error related parameters: refractive error indicators obtained from routine examinations (serial numbers 5-8); (3) Corneal morphological parameters: 53 morphological parameters of the Pentacam anterior segment analysis system (serial numbers 10-53); (4) Intraocular structural parameters: intraocular structural parameters obtained from the Lenstar 900 biometer (serial numbers 54-56); (5) Corneal biomechanics: biomechanical indicators collected by Corvis ST (serial numbers 57-93); All feature codes correspond to Table 1.
[0026] Table 1. Full names and definitions of 109 relevant parameters.
[0027] Target value: The intraocular pressure is measured three times at a specific postoperative time point (3 months postoperatively) using the same standard tonometer (such as a non-contact tonometer), and the average value of the intraocular pressure is obtained as the postoperative reference intraocular pressure value at that time point (i.e. the target to be predicted).
[0028] The collected patient data is cleaned, missing and outlier values are removed, and continuous features are standardized or normalized.
[0029] Specifically, this also includes feature engineering: constructing a preoperative predictive feature set from the preoperative parameters. This may include original features as well as new features derived through calculation, such as: corneal curvature radius, expected corneal ablation depth / thickness change (estimated based on surgical method and correction degree), and the ratio of axial length to corneal curvature.
[0030] Model building and training: The preprocessed dataset is divided into a training set and a test set according to a preset ratio. Using the training set, a machine learning algorithm is used to train a postoperative true intraocular pressure (IOP) prediction model that takes the preoperative predicted feature set as input and the postoperative reference IOP value as output.
[0031] The machine learning algorithm can be selected from one or a combination of the following: Random Forest, Decision Tree (GBDT, XGBoost, LightGBM), Support Vector Machine (SVR), Artificial Neural Network (ANN), or Linear Regression and its regularized variants (Lasso, Ridge).
[0032] During training, cross-validation is used to optimize model parameters in order to achieve the best prediction performance.
[0033] Model Validation and Evaluation: The trained postoperative real intraocular pressure (IOP) prediction model was validated using a reserved test set. Root mean square error (RMSE), mean absolute error (MAE), and coefficient of determination (R²) were used as the main evaluation metrics to assess the consistency and accuracy between the predicted values of the postoperative real IOP prediction model and the actual postoperative reference IOP values.
[0034] Model Application: The validated postoperative true intraocular pressure (IOP) prediction model was applied clinically. For a patient scheduled for or who has already undergone corneal refractive surgery, preoperative parameters were collected and input into the postoperative true IOP prediction model. The model output was the predicted IOP value for the patient during the stable postoperative period. By comparing this predicted IOP value with the actual postoperative IOP measurement, clinicians could more accurately determine the patient's true IOP status: if the actual measurement was significantly lower than the predicted value, it could be considered a measurement decrease caused by surgery; if the actual measurement was close to or even higher than the predicted value, the possibility of elevated true IOP should be highly suspected, and follow-up and glaucoma screening should be strengthened.
[0035] The following specific examples further illustrate the method of the present invention.
[0036] Example 1: Prediction Model Construction
[0037] This study retrospectively collected data from 7856 patients who underwent corneal lenticule extraction (KLEx) at the Refractive Surgery Center of Tianjin Eye Hospital. Inclusion criteria: preoperative intraocular pressure <21 mmHg, excluding glaucoma. Age, sex, and preoperative parameters including IOP, SE, Km, CCT, AL, ACD, LT, and WTW were collected.
[0038] The target variable was the average intraocular pressure (IOP) measured three times by the same non-contact tonometer at 3 months post-surgery for each patient.
[0039] After data cleaning, 2830 data points were randomly divided into a training set and 4974 data points were used as a test set. Z-score standardization was applied to continuous features.
[0040] A random forest regression model was trained using Python's Scikit-learn library with the training set data. Hyperparameters such as the number of trees and maximum depth in the model were optimized through grid search and 5-fold cross-validation.
[0041] The model performance was evaluated on the test set, and the RMSE was 0.78 mmHg, the MAE was 0.62 mmHg, and the R² was 0.86, indicating that the model has high prediction accuracy.
[0042] Example 2: Clinical Predictive Application
[0043] A 25-year-old male patient is scheduled to undergo KLEx surgery to correct myopia. Preoperative examination: IOP: 17.8 mmHg, SE-5.00D, Km43.5D, CCT 545μm, AL 25.2mm, ACD 3.3mm.
[0044] Input the above preoperative parameters into the trained prediction model (as obtained in Example 1).
[0045] The model outputs a predicted intraocular pressure of 12.8 mmHg for the patient 3 months post-surgery.
[0046] Three months after the surgery, the patient underwent a follow-up examination, and the actual intraocular pressure measured by NCT was 12 mmHg.
[0047] Clinical analysis: The doctor recorded the patient's "predicted reference intraocular pressure" as 12.8 mmHg, a difference of 5 mmHg from the preoperative intraocular pressure. In future annual follow-ups, if the actual measured intraocular pressure rises to 16.8 mmHg or higher, although still within the "normal range" (<21 mmHg), it is close to the individualized predicted reference value, suggesting the need for enhanced monitoring and more comprehensive glaucoma screening (such as fundus optic disc nerve scanning, visual field testing, etc.) to achieve early warning of hypertension or glaucoma risk.
[0048] A system for predicting true intraocular pressure after corneal refractive surgery based on a machine learning model, such as Figure 2 As shown, it includes: The data acquisition module collects patient data including preoperative ocular parameters and postoperative reference intraocular pressure values; The data preprocessing module preprocesses the collected patient data to construct a dataset, which includes a training set and a test set. The model training and validation module uses the training set to train a postoperative real intraocular pressure prediction model based on machine learning; it validates the trained postoperative real intraocular pressure prediction model based on the test set, and then performs multi-center independent data external validation to obtain the validated postoperative real intraocular pressure prediction model. The output module processes the preoperative ocular parameters of the patient and inputs them into the postoperative true intraocular pressure prediction model to obtain the patient's true intraocular pressure value. It also receives and displays the patient's actual postoperative intraocular pressure value and compares the patient's true intraocular pressure value with the actual measured value.
[0049] An electronic device includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the computer program, implements a method for predicting true intraocular pressure after corneal refractive surgery based on a machine learning model.
[0050] A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements a method for predicting true intraocular pressure after corneal refractive surgery based on a machine learning model.
[0051] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the apparatus disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the description is relatively simple; relevant parts can be referred to the method section.
[0052] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.
Claims
1. A method for predicting true intraocular pressure after corneal refractive surgery based on a machine learning model, characterized in that, include: Collect patient data including preoperative ocular parameters and postoperative reference intraocular pressure values; The collected patient data is preprocessed to construct a dataset; The dataset includes a training set and a test set; Using the training set, a postoperative true intraocular pressure prediction model based on machine learning was trained. The trained postoperative real intraocular pressure prediction model was validated based on the test set, and then external validation was performed using multi-center independent data to obtain the validated postoperative real intraocular pressure prediction model. The preoperative ocular parameters of the patient to be tested are processed and then input into the postoperative true intraocular pressure prediction model to obtain the true intraocular pressure value of the patient to be tested.
2. The method for predicting true intraocular pressure after corneal refractive surgery based on a machine learning model according to claim 1, characterized in that, The preoperative ocular parameters include at least several of the following: age, gender, preoperative non-contact intraocular pressure measurement, refractive parameters, spherical power, cylindrical power, equivalent spherical power, algebraic sum of spherocytic and cylindrical power, preoperative mean corneal curvature, mean radius of curvature, central corneal thickness, anterior chamber depth, corneal diameter, corneal volume, anterior chamber volume, corneal diameter, anterior chamber angle, pupil diameter, axial length, anterior chamber depth, lens thickness, and horizontal corneal diameter.
3. The method for predicting true intraocular pressure after corneal refractive surgery based on a machine learning model according to claim 1, characterized in that, The specific method for collecting the postoperative reference intraocular pressure (IOP) value is as follows: IOP is measured three times using the same standard tonometer at a specific postoperative time point, and the average IOP value is obtained as the postoperative reference IOP value at that time point.
4. The method for predicting true intraocular pressure after corneal refractive surgery based on a machine learning model according to claim 1, characterized in that, The preprocessing includes cleaning the collected patient data, processing missing and outlier values, and standardizing or normalizing continuous features.
5. The method for predicting true intraocular pressure after corneal refractive surgery based on a machine learning model according to claim 1, characterized in that, The machine learning algorithm selected is one or a combination of two of the following: random forest, decision tree, support vector machine, and artificial neural network.
6. The method for predicting true intraocular pressure after corneal refractive surgery based on a machine learning model according to claim 1, characterized in that, The trained postoperative real intraocular pressure prediction model was validated based on the test set. Specifically, root mean square error, mean absolute error, and coefficient of determination were used as evaluation indicators to assess the consistency and accuracy between the predicted value of the postoperative real intraocular pressure prediction model and the actual postoperative reference intraocular pressure value.
7. A system for predicting true intraocular pressure after corneal refractive surgery based on a machine learning model, characterized in that, A method for predicting true intraocular pressure after corneal refractive surgery based on a machine learning model, as described in any one of claims 1-6, comprises: The data acquisition module collects patient data including preoperative ocular parameters and postoperative reference intraocular pressure values; The data preprocessing module preprocesses the collected patient data to construct a dataset, which includes a training set and a test set. The model training and validation module uses the training set to train a postoperative real intraocular pressure prediction model based on machine learning; it validates the trained postoperative real intraocular pressure prediction model based on the test set, and then performs multi-center independent data external validation to obtain the validated postoperative real intraocular pressure prediction model. The output module processes the preoperative ocular parameters of the patient and inputs them into the postoperative true intraocular pressure prediction model to obtain the true intraocular pressure value of the patient.
8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements a method for predicting the true intraocular pressure after corneal refractive surgery based on a machine learning model, as described in claims 1-6.
9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements a method for predicting the true intraocular pressure after corneal refractive surgery based on a machine learning model, as described in any one of claims 1-6.