An ophthalmic prescription generation method and device, computer equipment and storage medium

Abstract

The application discloses a kind of spectacle prescription generation method, device, computer equipment and storage medium, method includes: obtaining from at least two different types of eye optical measurement equipment original measurement data;Original measurement data is carried out quality evaluation and standardization processing, and the quality score of each data source is generated;According to the data after pre-processing, at least one eye visual optics calculation model is used to calculate initial diopter number;According to quality score, the consistency score between multi-source data and the personalized characteristics of patient, dynamically determine the contribution weight of each data source in fusion process;According to the contribution weight determined, the diopter number from different data sources and / or calculation model is weighted and fused, and generates spectacle prescription.The application significantly improves the accuracy and reliability of spectacle prescription.

Description

Technical Field

[0001] This invention relates to the field of optometry technology, and in particular to a method, apparatus, computer device, and storage medium for generating eyeglass prescriptions. Background Technology

[0002] Traditional prescription generation relies heavily on single or limited ophthalmic measurements (such as refractometer data) and the optometrist's subjective judgment and experience. A single data source may not fully reflect the eye's refractive state, limiting prescription accuracy and making it difficult to achieve the high clinical standard of ±0.25 D. Furthermore, because the optometrist's experience and judgment dominate prescription generation, it is easily influenced by individual differences and fatigue, leading to poor prescription consistency. In addition, traditional methods struggle to make precise adjustments based on individual patient characteristics (such as age, eye habits, and ocular bioparameters), resulting in low prescription personalization.

[0003] Existing AI-assisted refraction systems primarily focus on automated data collection or preliminary prescription recommendations, but generally lack deep integration of multi-source heterogeneous data, dynamic assessment of data quality, and weight optimization mechanisms based on patient characteristics. Some systems only process computer-generated refraction data, neglecting the accurate reflection of key parameters such as axial length and corneal curvature by biometric data, and the reference value of focimeter data on patients' old lens wearing habits and refractive change trends, making it difficult to guarantee the reliability and consistency of prescriptions in complex clinical situations. Summary of the Invention

[0004] The purpose of this invention is to overcome the shortcomings of the prior art and provide a method, apparatus, computer equipment and storage medium for generating eyeglass prescriptions, aiming to solve the problems of low personalization, poor accuracy and reliability of existing eyeglass prescriptions.

[0005] To achieve the above objectives, the present invention adopts the following technical solution: In a first aspect, the present invention provides a method for generating eyeglass prescriptions, comprising: Acquire raw measurement data from at least two different types of eye optical measurement devices; The raw measurement data is subjected to quality assessment and standardization, and quality scores are generated for each data source. Based on the preprocessed data, the initial refractive power is calculated using at least one optometry calculation model. The contribution weight of each data source in the fusion process is dynamically determined based on quality scores, consistency scores among multi-source data, and the patient's individual characteristics. Based on the determined contribution weights, the refractive power values ​​from different data sources and / or calculation models are weighted and fused to generate a prescription for eyeglasses.

[0006] Furthermore, the process of quality assessment and standardization of the raw measurement data, and the generation of quality scores for each data source, includes: For repeated measurements from each data source, outliers are identified and processed using statistical methods based on the criteria of 2 or 1.5 times the standard deviation. Data points that exceed the preset multiple standard deviation range of the mean are marked as outliers. For measurement sequences where the proportion of outliers exceeds a preset threshold, the initial quality score of the data source is reduced or a prompt message is generated suggesting remeasurement. Based on the valid measurement data from each data source after removing outliers, calculate the repeatability score that is inversely correlated with the standard deviation of the measured values; Reliability indicators are extracted from the raw measurement data provided by each eye optical measurement device. The reliability indicators include the number of effective measurements by the computer optometry instrument, the signal strength of the biometry instrument, and the usage of the old glasses provided by the focimeter. The reliability indicators are then converted into a reliability score in the range of 0-1 according to a preset mapping relationship. By comparing similar measurement parameters of the same patient from different data sources, the degree of difference is quantified by the coefficient of variation, and a consistency score is generated based on the degree of difference. The repeatability score, reliability score, and consistency score are weighted and summed to generate a comprehensive quality score for each data source, with a value between 0 and 1.

[0007] Furthermore, the comparison of similar measurement parameters from different data sources for the same patient, quantifying the degree of difference using the coefficient of variation, and generating a consistency score based on the degree of difference, includes: Extract similar measurement parameters from the measurement results of various data sources. These similar measurement parameters include spherical power, cylindrical power, and equivalent spherical power. For each data source, calculate the arithmetic mean and standard deviation of similar measurement parameters, and calculate the coefficient of variation based on the arithmetic mean and standard deviation; The mean values ​​of measurement parameters from different data sources are compared pairwise, and the differences are calculated. A consistency score is generated between the data sources based on the differences.

[0008] Further, the step of calculating the initial refractive power using at least one optometry calculation model based on the preprocessed data includes: Obtain pre-processed ocular bioparameters and patient-specific characteristics; The applicability of Holladay's improved model, Gullstrand's complete eyeball model, Donder's accommodation model, and models that consider the effective lens position were evaluated to determine the candidate model set. Ocular biometrics, patient-specific characteristics, and quality scores are input into a pre-trained machine learning model, which outputs dynamic weights for each candidate model. Each candidate model is run in parallel, and its initial refractive power is calculated independently based on ocular biological parameters. The initial refractive power is weighted and summed according to the dynamic weights to generate a comprehensive initial refractive power.

[0009] Furthermore, the step of inputting ocular biological parameters, patient-specific characteristics, and quality scores into a pre-trained machine learning model and outputting dynamic weights for each candidate model includes: The ocular biological parameters, patient personalized characteristics, and quality scores are concatenated into an input feature vector according to preset dimensions. The ocular biological parameters include at least one of axial length, corneal curvature, anterior chamber depth, corneal diameter, and lens thickness. The patient personalized characteristics include at least one of age, gender, occupation type, eye use habits, medical history, and previous glasses prescription history. The input feature vector is standardized using the Z-score standardization method. The standardized feature vectors are input into a pre-trained machine learning model for forward inference computation; The Softmax function is used to normalize the original weight scores of the model output, so that the sum of the weights of each candidate model is 1 and each weight value is between 0 and 1. Output the weight vector of the dynamic weights of each candidate model.

[0010] Furthermore, the parallel operation of each candidate model independently calculates its initial refractive power based on ocular biological parameters, including: Extract the required ocular biological parameters for each candidate model from the preprocessed data, and start the independent computing threads for each candidate model to run in parallel. The Holladay improved model calculates the baseline refractive power based on axial length, corneal curvature, and anterior chamber depth, and then corrects it by introducing adjustment and age correction terms based on corneal diameter and patient age, outputting the Holladay model refractive power. The Gullstrand complete eye model calculates the actual refractive power of the eye based on the axial length, corneal curvature, anterior chamber depth, and lens thickness, and compares it with the standard refractive power to assess the degree of refractive error, outputting the Gullstrand model refractive power. The Donder accommodation model calculates the remaining accommodation power based on the patient's age and corrects the near refractive error accordingly, outputting the Donder model refractive error. The model considering the effective lens position calculates the effective lens position based on the axial length and corneal diameter, and corrects the refractive power accordingly to correct the power deviation of patients with high myopia, and outputs the refractive power of the ELP model; Collect the diopter values ​​output by each candidate model and encapsulate them into a set of calculation results for each model.

[0011] Furthermore, the dynamic determination of the contribution weight of each data source in the fusion process based on quality scores, consistency scores among multi-source data, and the patient's personalized characteristics includes: Based on the real-time quality scores of each data source, the base weights of each data source are adjusted using an exponential function; Calculate the consistency score among multiple data sources. When the consistency score is higher than the first preset threshold, maintain the existing weight. When the consistency score is between the first preset threshold and the second preset threshold, identify outlier data sources and reduce their weight. When the consistency score is lower than the second preset threshold, use balanced weight allocation or generate a prompt message to transfer to manual review. The quality score, consistency score, patient age, refractive error, axial length, astigmatism status, and eye habits are constructed into a multidimensional feature vector. The multidimensional feature vector is then input into a pre-trained machine learning model that uses historical prescription data as training samples, and the model outputs the adaptation weights of each data source. The adaptation weights are adjusted based on the consistency score, and the final dynamic weights of each data source are output after adjustment, with the sum of all weights being 1.

[0012] Secondly, the present invention also provides an eyeglass prescription generation device, comprising: An acquisition unit is used to acquire raw measurement data from at least two different types of eye optical measurement devices; The processing unit is used to perform quality assessment and standardization on the raw measurement data and generate quality scores for each data source. The calculation unit is used to calculate the initial refractive power based on the preprocessed data using at least one optometry calculation model. The weight optimization unit is used to dynamically determine the contribution weight of each data source in the fusion process based on the quality score, the consistency score between multiple data sources, and the patient's personalized characteristics. The fusion unit is used to weight and fuse refractive power values ​​from different data sources and / or calculation models according to the determined contribution weights to generate a prescription for eyeglasses.

[0013] Thirdly, the present invention also provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement a method for generating an eyeglass prescription as described above.

[0014] Fourthly, the present invention also provides a computer-readable storage medium storing a computer program, the computer program including program instructions, which, when executed by a processor, cause the processor to perform a method for generating an eyeglass prescription as described above.

[0015] The beneficial effects of this invention compared with the prior art are as follows: This invention introduces a dynamic weight optimization algorithm based on data quality, data consistency, and patient characteristics, especially an intelligent weight prediction mechanism based on machine learning models, to achieve adaptive and personalized weight configuration; by integrating multiple advanced optometry calculation models and dynamically selecting and combining them, the accuracy of refractive power calculation is improved; and through data consistency scoring and anomaly handling mechanisms, data conflicts and missing data are effectively addressed, significantly improving the reliability of spectacle prescriptions.

[0016] The above description is merely an overview of the technical solution of the present invention. In order to better understand the technical means of the present invention, it can be implemented according to the contents of the specification. In order to make the above and other objectives, features and advantages of the present invention more obvious and understandable, preferred embodiments are described in detail below. Attached Figure Description

[0017] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 A flowchart illustrating a method for generating eyeglass prescriptions, provided as a specific embodiment of the present invention; Figure 2 A schematic block diagram of an eyeglass prescription generation device provided in a specific embodiment of the present invention; Figure 3 This is a schematic block diagram of a computer device provided for a specific embodiment of the present invention. Detailed Implementation

[0019] The technical solution of the present invention will be clearly and completely described below with reference to specific embodiments. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. 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.

[0020] It should be understood that, when used in this specification and the appended claims, the terms "comprising" and "including" indicate the presence of the described features, integrals, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or collections thereof.

[0021] It should also be understood that the terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to limit the invention. As used in this specification and the appended claims, the singular forms “a,” “an,” and “the” are intended to include the plural forms unless the context clearly indicates otherwise.

[0022] It should also be further understood that the term "and / or" as used in this specification and the appended claims refers to any combination of one or more of the associated listed items and all possible combinations, and includes such combinations.

[0023] like Figure 1 As shown, this embodiment of the invention provides a method for generating eyeglass prescriptions, including the following steps: S10-S50.

[0024] S10. Obtain raw measurement data from at least two different types of eye optical measurement devices.

[0025] The device connects to various measuring devices via a data interface to receive raw measurement data output from those devices. At least two different types of ocular optical measuring devices are selected from any two or all three of the following: a computerized refractometer, a biometer, and a focimeter. The computerized refractometer provides the patient's refractive parameters, specifically spherical power (SPH), cylindrical power (CYL), and cylindrical axis (AXS), along with the number of valid measurements and the device's built-in reliability indicators. The biometer provides the patient's ocular biological parameters, specifically axial length (AL), corneal curvature (K1, K2), anterior chamber depth (ACD), corneal diameter (WTW), and lens thickness (LT), along with the number of measurements and signal quality scores. The focimeter measures the refractive power of the patient's old glasses, specifically the spherical power, cylindrical power, and cylindrical axis, and includes information on the usage of the old glasses.

[0026] While receiving data, record the source device identifier, measurement timestamp, and device model for each measurement data item for subsequent traceability and data management.

[0027] S20. Perform quality assessment and standardization on the raw measurement data, and generate quality scores for each data source.

[0028] In some embodiments, step S20 specifically includes the following steps: S201-S205.

[0029] S201. For repeated measurements provided by each data source, outliers are identified and processed using statistical methods based on the criterion of 2 or 1.5 times the standard deviation. Data points that exceed the preset multiple standard deviation range of the mean are marked as outliers. For measurement sequences where the proportion of outliers exceeds a preset threshold, the initial quality score of the data source is reduced or a prompt message is generated suggesting remeasurement.

[0030] In this embodiment, specifically, the arithmetic mean of multiple measurements from the data source is first calculated. and standard deviation Subsequently, it will exceed Data points within a certain range are marked as outliers, where k is a preset multiple, configurable to 2 or 1.5 depending on clinical needs. When using the 2x standard deviation criterion, values ​​exceeding this range will be considered outliers. Data points falling outside the specified range are considered outliers; when using the 1.5 standard deviation criterion, those exceeding the specified range will be considered outliers. Data points within a certain range are considered outliers.

[0031] Data points marked as outliers are removed from valid measurement data and excluded from subsequent scoring calculations. Simultaneously, the proportion of outliers in the total number of measurements is calculated. If the outlier proportion exceeds a preset threshold (e.g., 30%), it indicates poor measurement stability of the data source, resulting in a lower initial quality score for that data source, or a prompt message is generated and the operator is advised to remeasure via the user interface.

[0032] S202. Based on the valid measurement data from each data source after removing outliers, calculate the repeatability score that is inversely correlated with the standard deviation of the measured values.

[0033] In this embodiment, after removing outliers, a repeatability score is calculated based on the remaining valid measurement data from each data source. The repeatability score measures the consistency of multiple measurements from the same data source and is inversely correlated with the standard deviation of the measurements. Specifically, a smaller standard deviation indicates that the multiple measurement results are more concentrated and the repeatability is better, resulting in a higher repeatability score; a larger standard deviation indicates that the measurement results are more dispersed and the repeatability is worse, resulting in a lower repeatability score.

[0034] The formula for calculating the repeatability score R is: ,in The standard deviation of the effective measurement data is given, where R ranges from 0 to 1.

[0035] S203. Extract reliability indicators from the raw measurement data provided by each eye's optical measurement device. The reliability indicators include the effective number of measurements by the computer optometry instrument, the signal strength of the biometric instrument, and the usage status of the old glasses provided by the focimeter. Convert the reliability indicators into a reliability score in the 0-1 range according to the preset mapping relationship.

[0036] In this embodiment, a reliability indicator is extracted from raw measurement data provided by each measuring device. Different devices provide different types of reliability indicators. Specifically, an automatic refractor provides the number of effective measurements, and a larger number of effective measurements indicates more reliable data.

[0037] A biometer provides signal intensity, which uses a signal-to-noise ratio as a quantitative indicator, reflecting the ratio of a returned signal after an optical signal emitted by the device is reflected by eye tissues to background noise. A higher signal-to-noise ratio indicates a clearer measurement signal and more reliable data. For example, SNR ≥ 1.6 is taken as the minimum threshold for the reliability of a measurement result. When the SNR is lower than 1.6, the current measurement result is rejected and not included in subsequent calculations; reliability is divided into multiple grades according to the SNR value: SNR ≤ 2 corresponds to low reliability, 2 < SNR ≤ 5 corresponds to medium reliability, 5 < SNR ≤ 11 corresponds to high reliability, and SNR > 11 corresponds to extremely high reliability. Different grades are mapped to different reliability scores within the 0-1 interval.

[0038] The usage of the old glasses provided by a lensometer includes at least one of the daily average wearing time and wearing frequency of the old glasses worn by the patient. A longer daily average wearing time and a higher wearing frequency indicate a deeper running-in degree between the power of the old glasses and the daily visual requirements of the patient, and the old glasses have higher reference value for the patient's current refractive state. According to the usage of the old glasses, it is comprehensively converted into a reliability score in the 0-1 interval according to a preset mapping relationship.

[0039] According to the preset mapping relationship, the reliability indicator of each device is converted into a reliability score in the 0-1 interval. Specifically, the mapping relationship can be implemented by means of a piecewise linear function or a table lookup.

[0040] S204: comparing similar measurement parameters from different data sources of the same patient, quantifying the degree of difference thereof by a coefficient of variation, and generating a consistency score according to the degree of difference.

[0041] In some embodiments, step S204 specifically includes the following steps: S2041-S2044.

[0042] S2041: extracting similar measurement parameters from the measurement results of each data source, wherein the similar measurement parameters include sphere power, cylinder power and equivalent sphere power.

[0043] S2042: for each data source, respectively calculating the arithmetic mean and standard deviation of the similar measurement parameters thereof, and calculating the coefficient of variation according to the arithmetic mean and the standard deviation.

[0044] S2043: performing pairwise comparison on the mean values of measurement parameters from different data sources, and calculating difference values thereof.

[0045] S2044. Generate a consistency score between each data source based on the difference.

[0046] For steps S2041-S2044, in this embodiment, similar measurement parameters are first extracted from the measurement results of each data source. For example, the computerized refractometer provides the spherical power SPH_A, and the focimeter provides the spherical power SPH_B of the old lens; both belong to the same type of measurement parameter (both are spherical powers). As another example, the computerized refractometer provides the cylindrical power CYL_A, and the focimeter provides the cylindrical power CYL_B of the old lens; both belong to the same type of measurement parameter.

[0047] For each data source, calculate the arithmetic mean of the same type of measurement parameters. and standard deviation And according to the formula Calculate the coefficient of variation. The coefficient of variation is a standardized metric that measures the dispersion of data. Its advantage lies in eliminating the influence of units and mean size, making it suitable for comparing the dispersion of measurement parameters from different data sources and of different magnitudes in a uniform manner.

[0048] Subsequently, the mean values ​​of measurement parameters from different data sources are compared pairwise, and their differences are calculated. For example, for spherical power, the system calculates... The smaller the difference, the more consistent the measurement results are between different data sources.

[0049] A consistency score is generated between each data source based on the difference. The consistency score is inversely correlated with the difference; that is, the smaller the difference, the higher the consistency score. Specifically, the formula for calculating the consistency score C is: , where δ is a preset normalization factor (e.g., 0.5D), and the value of C ranges from 0 to 1.

[0050] S205. The repeatability score, reliability score and consistency score are weighted and summed to generate a comprehensive quality score for each data source, with a value between 0 and 1.

[0051] In this embodiment, the formula for calculating the overall quality score Q is: Where R is the repeatability score, Rel is the reliability score, C is the consistency score, and α, β, and γ are preset weighting coefficients that satisfy α + β + γ = 1. For example, α = 0.3, β = 0.3, γ = 0.4, meaning that the consistency score is given a higher weight because consistency between different data sources better reflects the authenticity and reliability of the data.

[0052] While generating the comprehensive quality score, the valid measurement data after outlier removal are also standardized to eliminate differences between different units and magnitudes, facilitating subsequent processing by the calculation model. Standardization can be performed using the Min-Max standardization method to transform the data to the [0,1] interval; or using the Z-score standardization method to transform the data into a distribution with a mean of 0 and a standard deviation of 1.

[0053] Through the aforementioned quality assessment and standardization processes, the data quality of each data source can be comprehensively quantified, providing a reliable foundation for subsequent dynamic weight determination. Outlier detection and handling ensure the purity of the input data, repeatability scores reflect measurement stability, reliability scores reflect equipment trustworthiness, consistency scores reflect the degree of mutual corroboration among multi-source data, and the comprehensive quality score provides a quantitative basis for subsequent weight adjustments.

[0054] S30. Based on the preprocessed data, calculate the initial refractive power using at least one optometry calculation model.

[0055] In some embodiments, step S30 specifically includes the following steps: S301-S305.

[0056] S301. Obtain pre-processed ocular bioparameters and patient-specific characteristics.

[0057] Ocular bioparameters include at least one of axial length (AL), corneal curvature (K1, K2), anterior chamber depth (ACD), corneal diameter (WTW), and lens thickness (LT). Patient-specific characteristics include at least one of age, sex, occupation, eye habits, medical history, and previous glasses prescription.

[0058] The applicability of several optometry computational models was then evaluated. These models included the Holladay improved model, the Gullstrand complete eye model, the Donder accommodation model, and a model that considered the effective lens position (ELP).

[0059] S302. Conduct applicability evaluations on the Holladay improved model, the Gullstrand complete eyeball model, the Donder accommodation model, and models that consider the effective lens position to determine the candidate model set.

[0060] The Holladay modified model is suitable for most patients with refractive errors, especially those whose anterior chamber depth, corneal diameter, accommodative ability, and age need to be considered comprehensively. Its priority is increased when patients have accommodative abnormalities or are older.

[0061] The Gullstrand complete eye model assesses the degree of refractive error by calculating the deviation between the actual refractive power of the eye and the standard refractive power. It is suitable for patients who need to accurately assess the overall refractive status of the eye.

[0062] The Donder regulation model calculates regulatory capacity based on the patient's age and is particularly suitable for children, adolescents, and patients requiring local prescriptions. This model is given higher priority when the patient is younger (e.g., ≤18 years old) or requires local prescriptions.

[0063] For models that consider the effective lens position, the model calculates the effective lens position based on the axial length and corneal diameter, and corrects the refractive error accordingly. This model is particularly suitable for patients with high myopia (e.g., refractive error exceeding -6.00D). The priority of this model is increased when the patient has a longer axial length (e.g., ≥26mm) or a higher refractive error.

[0064] Based on the evaluation logic described above, at least one of the four models is selected as a candidate model. For example, by default, all four models are selected as the candidate model set.

[0065] S303. Input ocular biological parameters, patient-specific characteristics, and quality scores into a pre-trained machine learning model and output the dynamic weights of each candidate model.

[0066] In some embodiments, step S303 specifically includes the following steps: S3031-S3035.

[0067] S3031. The ocular biological parameters, patient personalized characteristics and quality scores are concatenated into an input feature vector according to preset dimensions. The ocular biological parameters include at least one of axial length, corneal curvature, anterior chamber depth, corneal diameter and lens thickness. The patient personalized characteristics include at least one of age, gender, occupation type, eye habits, medical history and previous glasses prescription history.

[0068] In this embodiment, ocular biometrics, patient-specific characteristics, and quality scores are first concatenated into an input feature vector according to preset dimensions. Ocular biometrics include at least one of axial length, corneal curvature, anterior chamber depth, corneal diameter, and lens thickness; patient-specific characteristics include at least one of age, gender, occupation type, eye habits, medical history, and previous glasses prescription history.

[0069] For example, the input feature vector contains the following dimensions: age (continuous value), gender (binary encoding), axial length (continuous value), corneal curvature K1 (continuous value), corneal curvature K2 (continuous value), anterior chamber depth (continuous value), corneal diameter (continuous value), lens thickness (continuous value), spherical power (continuous value), cylindrical power (continuous value), equivalent spherical power (continuous value), quality scores for each data source (3 continuous values), occupation type (One-Hot encoding), and eye use habit score (continuous value). The total dimension of the feature vector is approximately 20-30.

[0070] S3032. The input feature vector is standardized using the Z-score standardization method.

[0071] In this embodiment, the input feature vector is standardized using the Z-score standardization method, and the transformation formula is as follows: , where x is the original feature value, μ is the mean of the feature on the training dataset, and σ is the standard deviation of the feature on the training dataset. The mean and standard deviation are pre-calculated and stored by the machine learning model during the training phase and are directly called during actual inference.

[0072] S3033. Input the standardized feature vector into the pre-trained machine learning model for forward inference calculation.

[0073] The machine learning model is one of the following: gradient boosting tree model (such as XGBoost or LightGBM), neural network model, or multilayer perceptron (MLP) model. For example, the gradient boosting tree model performs well in processing tabular data, with high prediction accuracy and a certain degree of interpretability.

[0074] S3034. The Softmax function is used to normalize the original weight scores of the model output, so that the sum of the weights of each candidate model is 1 and each weight value is between 0 and 1.

[0075] In this embodiment, the Softmax function is used to normalize the original weight scores, and the transformation formula is as follows: , where z_i is the original weight score of the i-th candidate model, n is the total number of candidate models, and w_i is the normalized dynamic weight of the i-th candidate model. Softmax normalization ensures that the sum of the weights of all candidate models is 1 and that each weight value is between 0 and 1.

[0076] S3035, Output the weight vector of the dynamic weights of each candidate model.

[0077] In this embodiment, if the candidate models are the Holladay improved model, the Gullstrand complete eye model, the Donder accommodation model, and the ELP model, then the output weight vector is [w_Holladay, w_Gullstrand, w_Donder, w_ELP], which satisfies w_Holladay + w_Gullstrand + w_Donder + w_ELP = 1.

[0078] S304. Run each candidate model in parallel and calculate their initial refractive power independently based on ocular biological parameters.

[0079] In some embodiments, step S304 specifically includes the following steps: S3041-S3046.

[0080] S3041. Extract the required ocular biological parameters for each candidate model from the preprocessed data, and simultaneously start the independent computing threads for each candidate model to enable the models to run in parallel.

[0081] The S3042 and Holladay improved models calculate the baseline refractive power based on axial length, corneal curvature, and anterior chamber depth, and then adjust it according to corneal diameter and patient age by introducing adjustment and age correction terms, outputting the Holladay model refractive power.

[0082] S3043, the Gullstrand complete eye model calculates the actual refractive power of the eyeball based on the axial length, corneal curvature, anterior chamber depth and lens thickness, and compares it with the standard refractive power to assess the degree of refractive error, and outputs the Gullstrand model refractive power.

[0083] S3044: The Donder accommodation model calculates the remaining accommodation power based on the patient's age and corrects the near refractive error accordingly, outputting the Donder model refractive error.

[0084] S3045. The model considering the effective lens position calculates the effective lens position based on the axial length and corneal diameter, and corrects the refractive power accordingly to correct the power deviation of patients with high myopia, and outputs the refractive power of the ELP model.

[0085] S3046. Collect the diopter values ​​output by each candidate model and encapsulate them into a set of calculation results for each model.

[0086] In this embodiment, for steps S3041-S3046, the ocular biological parameters required by each candidate model are first extracted from the preprocessed data. Different models have different requirements for input parameters. Specifically, the Holladay improved model requires at least axial length (AL) and corneal curvature (K1, K2) as input parameters, and optional inputs include anterior chamber depth (ACD) and corneal diameter (WTW). This model further considers anterior chamber depth, corneal diameter, accommodation correction term, and age correction term on the basic formula to provide a more comprehensive refractive power prediction.

[0087] The Gullstrand complete eye model requires at least axial length, corneal curvature, anterior chamber depth, and lens thickness as input parameters. This model is used to calculate the actual refractive power of the eye and compare it to a standard refractive power to assess the degree of refractive error.

[0088] The Donder regulation model requires age as a minimum input parameter. The model calculates regulatory capacity based on the patient's age and uses this information to adjust for near-term prescriptions.

[0089] A model that considers the effective lens position (ELP) requires at least axial length and corneal diameter as input parameters. This model calculates the effective lens position (ELP) based on axial length and corneal diameter and corrects for refractive errors to improve prescription accuracy for patients with high refractive errors.

[0090] Simultaneously, independent computation threads for each candidate model are launched, allowing each model to run concurrently in parallel computation threads. Parallel computation can significantly improve processing efficiency, especially when there are a large number of candidate models, avoiding the time delays caused by serial computation.

[0091] The specific calculation logic for each model is as follows: The Holladay improved model calculates the baseline refractive power based on axial length, corneal curvature, and anterior chamber depth. It then introduces an accommodation correction term based on corneal diameter and an age correction term based on the patient's age to adjust the baseline refractive power, ultimately outputting the Holladay model refractive power.

[0092] The Gullstrand complete eye model calculates the refractive power of each optical interface of the eyeball based on axial length, corneal curvature, anterior chamber depth, and lens thickness using the Gullstrand precision model. The total actual refractive power of the eyeball is obtained by summing the results. The actual refractive power is compared with the preset standard refractive power (such as the average refractive power of emmetropia, 58.64D). The degree of refractive error is assessed based on the deviation between the two, and the Gullstrand model refractive power is output.

[0093] The Donder accommodation model calculates a patient's residual accommodative power based on their age using a pre-defined age-accommodative power mapping relationship. It then corrects the near refractive error based on this residual accommodative power, outputting the Donder model refractive error.

[0094] Models considering the effective lens position (ELP) calculate the effective lens position (ELP) using empirical formulas based on axial length and corneal diameter, for example... Where a, b, and c are regression coefficients obtained by fitting historical data. The refractive power is corrected according to the effective lens position to correct the power deviation caused by the difference in lens position in patients with high myopia, and the ELP model refractive power is output.

[0095] After each model completes its calculation, the refractive power output by each candidate model is collected. The refractive power includes at least the spherical power, cylindrical power, and cylindrical axis. The calculation results of each model are then packaged according to a preset data structure to form a set of calculation results for each model, which can be used in the subsequent weighted fusion step.

[0096] S305. The initial refractive power is weighted and summed according to the dynamic weight to generate a comprehensive initial refractive power.

[0097] In this embodiment, for the spherical power SPH, the comprehensive value For cylindrical lens power CYL, the overall value For cylindrical lens axis AXS, since the axis is an angle value, it cannot be directly weighted averaged (for example, the average of 178° and 2° should be 0° instead of 90°). First, the axis is converted into a vector form (averaged in double angle space), then weighted average is performed, and finally converted back to angle value.

[0098] The weighted fusion method for the cylindrical lens axis is as follows: convert the axial values ​​of each model into vector components in Cartesian coordinates, perform a weighted average of each vector component, and then convert it back to the spherical-cylindrical lens form.

[0099] By employing parallel computing and dynamic weighted fusion of multiple models, the advantages of each optical model can be comprehensively utilized to obtain optimal computational results in different clinical scenarios. The introduction of the AI ​​meta-selector allows model weights to be dynamically adjusted based on the specific characteristics of the patient, rather than remaining fixed, thus achieving truly personalized prescription calculation. The parallel computing mechanism improves processing efficiency and meets the real-time requirements of clinical practice.

[0100] S40. Based on the quality score, the consistency score among multiple data sources, and the patient's individual characteristics, dynamically determine the contribution weight of each data source in the fusion process.

[0101] In some embodiments, step S40 specifically includes the following steps: S401-S404.

[0102] S401. Based on the real-time quality scores of each data source, adjust the basic weights of each data source using an exponential function.

[0103] In this embodiment, the formula for calculating the dynamic weight is: , where w_base_i is the base weight of the i-th data source (which can be statically configured according to the inherent reliability of the data source), Q_i is the overall quality score of the i-th data source (with a value between 0 and 1), and α is the sensitivity parameter, which can be configured according to clinical needs.

[0104] The sensitivity parameter α controls the degree to which the quality score affects the weight. When α=1, the weight is directly proportional to the quality score; when α>1, the weight is more sensitive to changes in the quality score, with a greater increase in the weight of high-quality data sources and a greater decrease in the weight of low-quality data sources; when 0<α<1, the weight is relatively insensitive to changes in the quality score.

[0105] For example, assuming a data source has a base weight of 0.3, an overall quality score of 0.9, and α = 1.5, the adjusted weight is 0.3 × 0.9^1.5 = 0.3 × 0.854 = 0.256. If the overall quality score of this data source is only 0.4, the adjusted weight is 0.3 × 0.4^1.5 = 0.3 × 0.253 = 0.076. It can be seen that when the quality score decreases from 0.9 to 0.4, the weight decreases from 0.256 to 0.076, a significant drop.

[0106] S402. Calculate the consistency score among multiple data sources. When the consistency score is higher than the first preset threshold, maintain the existing weight. When the consistency score is between the first preset threshold and the second preset threshold, identify outlier data sources and reduce their weight. When the consistency score is lower than the second preset threshold, use balanced weight allocation or generate a prompt message to transfer to manual review.

[0107] In this embodiment, the first preset threshold is set to 0.7, and the second preset threshold is set to 0.4. The consistency score ranges from 0 to 1, with a higher score indicating greater consistency in measurement results across different data sources.

[0108] When the consistency score is higher than the first preset threshold (i.e. ≥0.7), it is judged as high consistency, indicating that the measurement results of each data source corroborate each other and have high reliability. The existing weights are maintained and no additional adjustments are made.

[0109] When the consistency score falls between the first and second preset thresholds (i.e., 0.4 ≤ consistency score < 0.7), it is considered moderately consistent, indicating some differences between the data sources. Outlier data sources are identified—those that differ most significantly from the others—and their weights are reduced. For example, the weight of an outlier data source is multiplied by a penalty factor of 0.5, and the weights of other data sources are renormalized proportionally.

[0110] When the consistency score is lower than the second preset threshold (i.e., <0.4), it is judged as low consistency, indicating that there are serious conflicts between the data sources. A balanced weight allocation (i.e., equal weight for each data source) is adopted, or a prompt message is generated and notified to the operator through the user interface to transfer the case to manual review.

[0111] S403. Construct a multidimensional feature vector from the quality score, consistency score, patient age, refractive error, axial length, astigmatism status, and eye habits. Input the multidimensional feature vector into a pre-trained machine learning model that uses historical prescription data as training samples, and output the adaptation weights of each data source.

[0112] In this embodiment, the machine learning model in this step is used to predict the weights of each data source. The input feature vector contains the following dimensions: quality score of each data source (3 consecutive values), consistency score (1 consecutive value), age (consecutive value), refractive error (consecutive value), axial length (consecutive value), astigmatism status (consecutive value), and eye use habit score (consecutive value or classification code). The total dimension of the feature vector is approximately 10-15.

[0113] The machine learning model is a random forest model. The training process is as follows: A large amount of historical prescription data is collected as training samples. Each sample includes patient characteristics, measurements from each data source, quality scores from each data source, and the doctor's final prescription. By analyzing the actual contribution of each data source to the final prescription under different patient characteristics in the historical data, the optimal weights of each data source are deduced as the training objective. The model is trained using supervised learning.

[0114] S404. Adjust the adaptation weights according to the consistency score, and output the final dynamic weights of each data source after adjustment, and the sum of each weight is 1.

[0115] In this embodiment, the correction logic is similar to the weight adjustment based on data consistency: when the consistency score is lower than a preset threshold, the weight of outlier data sources is penalized and reduced.

[0116] After correction, the final dynamic weights of each data source are output, and the sum of all weights is 1. Weight normalization can be performed using a direct normalization method.

[0117] The weighting mechanism based on the three dimensions (quality, consistency, and patient characteristics) adaptively adjusts the contribution weights of each data source: high-quality data sources receive higher weights, outlier data sources inconsistent with other data sources have their weights reduced, and data sources with high matching degrees to patient characteristics receive higher weights. These three mechanisms are independent yet complementary, collectively achieving intelligent and personalized weighting configuration. Compared to traditional methods that use fixed weights or adjust weights based on only a single dimension, the three-dimensional dynamic weighting mechanism significantly improves the rationality of weights and the accuracy of prescriptions.

[0118] S50. Based on the determined contribution weights, the refractive power values ​​from different data sources and / or calculation models are weighted and fused to generate a prescription for eyeglasses.

[0119] Based on the comprehensive refractive error, a final prescription for eyeglasses is generated, which includes spherical power, cylindrical power, and cylindrical axis.

[0120] In summary: By introducing a dynamic weight optimization algorithm based on data quality, data consistency, and patient characteristics, especially an intelligent weight prediction mechanism based on machine learning models, adaptive and personalized weight configuration is achieved; by integrating multiple advanced optometry calculation models and dynamically selecting and combining them, the accuracy of refractive error calculation is improved; and through data consistency scoring and anomaly handling mechanisms, data conflicts and missing data are effectively addressed, significantly enhancing the reliability of prescriptions.

[0121] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0122] This invention also provides an eyeglass prescription generation device, which is used to perform the steps in any of the embodiments of the aforementioned eyeglass prescription generation method. Specifically, please refer to... Figure 2 , Figure 2 A schematic block diagram of a prescription eyeglasses generator 100 provided in an embodiment of this application is shown. The prescription eyeglasses generator 100 specifically includes: The acquisition unit 110 is used to acquire raw measurement data from at least two different types of ophthalmic optical measurement devices; the processing unit 120 is used to perform quality assessment and standardization on the raw measurement data and generate quality scores for each data source; the calculation unit 130 is used to calculate the initial refractive power based on the preprocessed data using at least one ophthalmic optical calculation model; the weight optimization unit 140 is used to dynamically determine the contribution weight of each data source in the fusion process based on the quality score, the consistency score between multi-source data, and the patient's individual characteristics; and the fusion unit 150 is used to perform weighted fusion of refractive power from different data sources and / or calculation models based on the determined contribution weights to generate a prescription for eyeglasses.

[0123] In some embodiments, the processing unit 120 is specifically applied to: for multiple repeated measurement data provided by each data source, using a statistical method based on the 2-times or 1.5-times standard deviation criterion to identify and process outliers, marking data points that exceed the mean by a preset multiple of the standard deviation range as outliers, and for measurement sequences where the proportion of outliers exceeds a preset threshold, reducing the initial quality score of the data source or generating a prompt message to suggest remeasurement; Based on the valid measurement data from each data source after removing outliers, calculate the repeatability score that is inversely correlated with the standard deviation of the measured values; Reliability indicators are extracted from the raw measurement data provided by each eye's optical measurement device. The reliability indicators include the effective number of measurements by the computer optometry instrument, the signal strength of the biometer, and the lens downlight measurement value provided by the focimeter. The reliability indicators are then converted into a reliability score in the 0-1 range according to a preset mapping relationship. By comparing similar measurement parameters of the same patient from different data sources, the degree of difference is quantified by the coefficient of variation, and a consistency score is generated based on the degree of difference. The repeatability score, reliability score, and consistency score are weighted and summed to generate a comprehensive quality score for each data source, with a value between 0 and 1.

[0124] In some embodiments, the processing unit 120 is further specifically applied to: extracting similar measurement parameters from the measurement results of various data sources, the similar measurement parameters including spherical power, cylindrical power and equivalent spherical power; For each data source, calculate the arithmetic mean and standard deviation of similar measurement parameters, and calculate the coefficient of variation based on the arithmetic mean and standard deviation; The mean values ​​of measurement parameters from different data sources are compared pairwise, and the differences are calculated. A consistency score is generated between the data sources based on the differences.

[0125] In some embodiments, the computing unit 130 is specifically used to: acquire preprocessed ocular biological parameters and patient-specific characteristics; The applicability of Holladay's improved model, Gullstrand's complete eyeball model, Donder's accommodation model, and models that consider the effective lens position were evaluated to determine the candidate model set. Ocular biometrics, patient-specific characteristics, and quality scores are input into a pre-trained machine learning model, which outputs dynamic weights for each candidate model. Each candidate model is run in parallel, and its initial refractive power is calculated independently based on ocular biological parameters. The initial refractive power is weighted and summed according to the dynamic weights to generate a comprehensive initial refractive power.

[0126] In some embodiments, the calculation unit 130 is also specifically used to: concatenate ocular biological parameters, patient personalized characteristics and quality scores into an input feature vector according to a preset dimension. The ocular biological parameters include at least one of axial length, corneal curvature, anterior chamber depth, corneal diameter and lens thickness. The patient personalized characteristics include at least one of age, gender, occupation type, eye habits, medical history and previous glasses prescription history. The input feature vector is standardized using the Z-score standardization method. The standardized feature vectors are input into a pre-trained machine learning model for forward inference computation; The Softmax function is used to normalize the original weight scores of the model output, so that the sum of the weights of each candidate model is 1 and each weight value is between 0 and 1. Output the weight vector of the dynamic weights of each candidate model.

[0127] In some embodiments, the computing unit 130 is also specifically used to: extract the eye biological parameters required for each candidate model from the preprocessed data, and simultaneously start the independent computing threads of each candidate model so that each model can run in parallel; The Holladay improved model calculates the baseline refractive power based on axial length, corneal curvature, and anterior chamber depth, and then corrects it by introducing adjustment and age correction terms based on corneal diameter and patient age, outputting the Holladay model refractive power. The Gullstrand complete eye model calculates the actual refractive power of the eye based on the axial length, corneal curvature, anterior chamber depth, and lens thickness, and compares it with the standard refractive power to assess the degree of refractive error, outputting the Gullstrand model refractive power. The Donder accommodation model calculates the remaining accommodation power based on the patient's age and corrects the near refractive error accordingly, outputting the Donder model refractive error. The model considering the effective lens position calculates the effective lens position based on the axial length and corneal diameter, and corrects the refractive power accordingly to correct the power deviation of patients with high myopia, and outputs the refractive power of the ELP model; Collect the diopter values ​​output by each candidate model and encapsulate them into a set of calculation results for each model.

[0128] In some embodiments, the weight optimization unit 140 is specifically applied to: adjusting the basic weights of each data source based on the real-time quality scores of each data source using an exponential function; Calculate the consistency score among multiple data sources. When the consistency score is higher than the first preset threshold, maintain the existing weight. When the consistency score is between the first preset threshold and the second preset threshold, identify outlier data sources and reduce their weight. When the consistency score is lower than the second preset threshold, use balanced weight allocation or generate a prompt message to transfer to manual review. The quality score, consistency score, patient age, refractive error, axial length, astigmatism status, and eye habits are constructed into a multidimensional feature vector. The multidimensional feature vector is then input into a pre-trained machine learning model that uses historical prescription data as training samples, and the model outputs the adaptation weights of each data source. The adaptation weights are adjusted based on the consistency score, and the final dynamic weights of each data source are output after adjustment, with the sum of all weights being 1.

[0129] It should be noted that those skilled in the art can clearly understand that the specific implementation process of the above-mentioned eyeglass prescription generation device 100 and each unit can be referred to the corresponding description in the foregoing method embodiments. For the sake of convenience and brevity, it will not be repeated here.

[0130] The aforementioned prescription generation device can be implemented as a computer program, which can, for example, generate prescriptions for eyeglasses. Figure 3 It runs on the computer device shown.

[0131] Please see Figure 3 , Figure 3 This is a schematic block diagram of a computer device provided in an embodiment of this application. The computer device 700 can be a server, wherein the server can be a standalone server or a server cluster composed of multiple servers.

[0132] like Figure 3 As shown, the computer device includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the steps of the eyeglass prescription generation method described above.

[0133] The computer device 700 includes a processor 720, a memory, and a network interface 750 connected via a system bus 710, wherein the memory may include a non-volatile storage medium 730 and internal memory 740.

[0134] The non-volatile storage medium 730 may store an operating system 731 and a computer program 732. When the computer program 732 is executed, it causes the processor 720 to perform a prescription generation method for eyeglasses.

[0135] The processor 720 provides computing and control capabilities to support the operation of the entire computer device 700.

[0136] The internal memory 740 provides an environment for the operation of the computer program 732 in the non-volatile storage medium 730. When the computer program 732 is executed by the processor 720, the processor 720 can perform the prescription generation method for eyeglasses.

[0137] This network interface 750 is used for network communication, such as sending assigned tasks. Those skilled in the art will understand that... Figure 3 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device 700 to which the present application is applied. The specific computer device 700 may include more or fewer components than shown in the figure, or combine certain components, or have different component arrangements. The processor 720 is used to run program code stored in memory to implement the eyeglass prescription generation method.

[0138] Those skilled in the art will understand that Figure 3 The embodiments of the computer device shown do not constitute a limitation on the specific configuration of the computer device. In other embodiments, the computer device may include more or fewer components than illustrated, or combine certain components, or have different component arrangements. For example, in some embodiments, the computer device may include only memory and a processor. In such embodiments, the structure and function of the memory and processor are different from those shown. Figure 3 The embodiments shown are consistent and will not be repeated here.

[0139] It should be understood that, in the embodiments of this application, the processor 720 may be a Central Processing Unit (CPU), or it may be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor.

[0140] In another embodiment of the present invention, a computer-readable storage medium is provided. This computer-readable storage medium may be a non-volatile computer-readable storage medium. The computer-readable storage medium stores a computer program, wherein when executed by a processor, the computer program implements the eyeglass prescription generation method disclosed in the embodiments of the present invention.

[0141] Those skilled in the art will readily understand that, for the sake of convenience and brevity, the specific working processes of the devices, apparatuses, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here. Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described in terms of function in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this invention.

[0142] In the embodiments provided by this invention, it should be understood that the disclosed devices, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative. For instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. Units with the same function may be grouped into one unit. For example, multiple units or components may be combined or integrated into another device, or some features may be ignored or not executed. In addition, the mutual coupling or direct coupling or communication connection shown or discussed may be indirect coupling or communication connection through some interfaces, devices, or units, or may be electrical, mechanical, or other forms of connection.

[0143] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of the embodiments of the present invention, depending on actual needs.

[0144] Furthermore, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0145] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), magnetic disks, or optical disks.

[0146] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in the present invention, and these modifications or substitutions should all be covered within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for generating eyeglass prescriptions, characterized in that, include: Acquire raw measurement data from at least two different types of eye optical measurement devices; The raw measurement data is subjected to quality assessment and standardization, and quality scores are generated for each data source. Based on the preprocessed data, the initial refractive power is calculated using at least one optometry calculation model. The contribution weight of each data source in the fusion process is dynamically determined based on quality scores, consistency scores among multi-source data, and the patient's individual characteristics. Based on the determined contribution weights, the refractive power values ​​from different data sources and / or calculation models are weighted and fused to generate a prescription for eyeglasses.

2. The method for generating an eyeglass prescription according to claim 1, characterized in that, The process of quality assessment and standardization of the raw measurement data, and the generation of quality scores for each data source, includes: For repeated measurements from each data source, outliers are identified and processed using statistical methods based on the criteria of 2 or 1.5 times the standard deviation. Data points that exceed the preset multiple standard deviation range of the mean are marked as outliers. For measurement sequences where the proportion of outliers exceeds a preset threshold, the initial quality score of the data source is reduced or a prompt message is generated suggesting remeasurement. Based on the valid measurement data from each data source after removing outliers, calculate the repeatability score that is inversely correlated with the standard deviation of the measured values; Reliability indicators are extracted from the raw measurement data provided by each eye optical measurement device. The reliability indicators include the effective number of measurements by the computer optometry instrument, the signal strength of the biometry instrument, and the usage of the old glasses provided by the focimeter. The reliability indicators are then converted into a reliability score in the range of 0-1 according to a preset mapping relationship. By comparing similar measurement parameters of the same patient from different data sources, the degree of difference is quantified by the coefficient of variation, and a consistency score is generated based on the degree of difference. The repeatability score, reliability score, and consistency score are weighted and summed to generate a comprehensive quality score for each data source, with a value between 0 and 1.

3. The method for generating an eyeglass prescription according to claim 2, characterized in that, The comparison of similar measurement parameters from different data sources for the same patient, quantifying the degree of difference using the coefficient of variation, and generating a consistency score based on the degree of difference, includes: Extract similar measurement parameters from the measurement results of various data sources. These similar measurement parameters include spherical power, cylindrical power, and equivalent spherical power. For each data source, calculate the arithmetic mean and standard deviation of similar measurement parameters, and calculate the coefficient of variation based on the arithmetic mean and standard deviation; The mean values ​​of measurement parameters from different data sources are compared pairwise, and the differences are calculated. A consistency score is generated between the data sources based on the differences.

4. The method for generating an eyeglass prescription according to claim 1, characterized in that, The step of calculating the initial refractive power using at least one optometry calculation model based on the preprocessed data includes: Obtain pre-processed ocular bioparameters and patient-specific characteristics; The applicability of Holladay's improved model, Gullstrand's complete eyeball model, Donder's accommodation model, and models that consider the effective lens position were evaluated to determine the candidate model set. Ocular biometrics, patient-specific characteristics, and quality scores are input into a pre-trained machine learning model, which outputs dynamic weights for each candidate model. Each candidate model is run in parallel, and its initial refractive power is calculated independently based on ocular biological parameters. The initial refractive power is weighted and summed according to the dynamic weights to generate a comprehensive initial refractive power.

5. The method for generating an eyeglass prescription according to claim 4, characterized in that, The process of inputting ocular bioparameters, patient-specific characteristics, and quality scores into a pre-trained machine learning model and outputting dynamic weights for each candidate model includes: The ocular biological parameters, patient personalized characteristics, and quality scores are concatenated into an input feature vector according to preset dimensions. The ocular biological parameters include at least one of axial length, corneal curvature, anterior chamber depth, corneal diameter, and lens thickness. The patient personalized characteristics include at least one of age, gender, occupation type, eye use habits, medical history, and previous glasses prescription history. The input feature vector is standardized using the Z-score standardization method. The standardized feature vectors are input into a pre-trained machine learning model for forward inference computation; The Softmax function is used to normalize the original weight scores of the model output, so that the sum of the weights of each candidate model is 1 and each weight value is between 0 and 1. Output the weight vector of the dynamic weights of each candidate model.

6. The method for generating an eyeglass prescription according to claim 4, characterized in that, The parallel operation of each candidate model independently calculates its initial refractive power based on ocular biological parameters, including: Extract the required ocular biological parameters for each candidate model from the preprocessed data, and start the independent computing threads for each candidate model to run in parallel. The Holladay improved model calculates the baseline refractive power based on axial length, corneal curvature, and anterior chamber depth, and then corrects it by introducing adjustment and age correction terms based on corneal diameter and patient age, outputting the Holladay model refractive power. The Gullstrand complete eye model calculates the actual refractive power of the eye based on the axial length, corneal curvature, anterior chamber depth, and lens thickness, and compares it with the standard refractive power to assess the degree of refractive error, outputting the Gullstrand model refractive power. The Donder accommodation model calculates the remaining accommodation power based on the patient's age and corrects the near refractive error accordingly, outputting the Donder model refractive error. The model considering the effective lens position calculates the effective lens position based on the axial length and corneal diameter, and corrects the refractive power accordingly to correct the power deviation of patients with high myopia, and outputs the refractive power of the ELP model; Collect the diopter values ​​output by each candidate model and encapsulate them into a set of calculation results for each model.

7. The method for generating an eyeglass prescription according to claim 1, characterized in that, The contribution weight of each data source in the fusion process is dynamically determined based on quality scores, consistency scores among multi-source data, and the patient's personalized characteristics, including: Based on the real-time quality scores of each data source, the base weights of each data source are adjusted using an exponential function; Calculate the consistency score among multiple data sources. When the consistency score is higher than the first preset threshold, maintain the existing weight. When the consistency score is between the first preset threshold and the second preset threshold, identify outlier data sources and reduce their weight. When the consistency score is lower than the second preset threshold, use balanced weight allocation or generate a prompt message to transfer to manual review. The quality score, consistency score, patient age, refractive error, axial length, astigmatism status, and eye habits are constructed into a multidimensional feature vector. The multidimensional feature vector is then input into a pre-trained machine learning model that uses historical prescription data as training samples, and the model outputs the adaptation weights of each data source. The adaptation weights are adjusted based on the consistency score, and the final dynamic weights of each data source are output after adjustment, with the sum of all weights being 1.

8. A device for generating eyeglass prescriptions, characterized in that, include: An acquisition unit is used to acquire raw measurement data from at least two different types of eye optical measurement devices; The processing unit is used to perform quality assessment and standardization on the raw measurement data and generate quality scores for each data source. The calculation unit is used to calculate the initial refractive power based on the preprocessed data using at least one optometry calculation model. The weight optimization unit is used to dynamically determine the contribution weight of each data source in the fusion process based on the quality score, the consistency score between multiple data sources, and the patient's personalized characteristics. The fusion unit is used to weight and fuse refractive power values ​​from different data sources and / or calculation models according to the determined contribution weights to generate a prescription for eyeglasses.

9. A computer device, characterized in that, The 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 generating an eyeglass prescription as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The storage medium stores a computer program, which includes program instructions that, when executed by a processor, cause the processor to perform a method for generating an eyeglass prescription as described in any one of claims 1 to 7.

Images