Myopia risk assessment method, device and system based on artificial intelligence

Abstract

The present disclosure relates to the technical field of artificial intelligence, and discloses a myopia risk assessment method, device and system based on artificial intelligence. The method comprises the following steps: acquiring vision-related historical data of a target user; extracting time sequence features in the vision-related historical data to obtain a time sequence feature set; acquiring daily average outdoor illumination time and daily average near-distance eye use time of the target user within the same time range to obtain environment and behavior monitoring data; aligning and correlating the trend features, fluctuation features and periodic features in the time sequence feature set with the environment and behavior monitoring data to obtain an eye use feature vector; performing myopia risk assessment on the target user based on the eye use feature vector; and generating an eye use assessment report according to the result of the myopia risk assessment and the environment and behavior monitoring data. The present disclosure can improve the efficiency of myopia risk assessment based on artificial intelligence.

Description

Technical Field

[0001] This invention relates to the field of artificial intelligence technology, and in particular to an artificial intelligence-based method, system, and smart terminal for myopia risk assessment. Background Technology

[0002] Traditional myopia risk assessment methods rely heavily on fragmented vision test data, failing to fully explore the dynamic trends and cyclical patterns of vision changes. They also neglect the impact of environmental and behavioral factors such as outdoor lighting and close-range eye use on the occurrence and development of myopia. This results in a single assessment dimension, making it difficult to comprehensively reflect the user's true myopia risk status and leading to insufficient accuracy in the assessment results.

[0003] Existing assessment technologies lack scientific, standardized alignment and weighting mechanisms in data processing and fusion. The synergistic effect of multi-dimensional features is not effectively considered. At the same time, the assessment logic is simplistic and lacks the ability to coordinate and arbitrate conflicts between different judgment results. This makes the assessment process inefficient, unable to quickly generate targeted eye assessment reports, and difficult to meet users' actual needs for early detection and intervention of myopia risks. Summary of the Invention

[0004] This disclosure provides a method, system, and smart terminal for myopia risk assessment based on artificial intelligence.

[0005] Firstly, this disclosure provides an artificial intelligence-based method for myopia risk assessment, including:

[0006] S1. Obtain historical vision-related data of the target user;

[0007] S2. Extract the temporal features reflecting the dynamic trend and periodic pattern of vision changes from the vision-related historical data to obtain the temporal feature set of the target user;

[0008] S3. Obtain the average daily outdoor light exposure duration and average daily close-range eye use duration of the target user within the same time range to obtain the environmental and behavioral monitoring data of the target user;

[0009] S4. The trend features, fluctuation features and periodic features in the time series feature set are standardized, aligned and weighted with the environmental and behavioral monitoring data to obtain the eye use feature vector of the target user;

[0010] S5. Assess the myopia risk of the target user based on the eye use feature vector;

[0011] S6. Based on the results of the myopia risk assessment and the environmental and behavioral monitoring data, generate an eye use assessment report for the target user.

[0012] In a preferred embodiment, the step of extracting time-series features reflecting the dynamic trends and cyclical patterns of vision changes from the historical vision-related data to obtain the time-series feature set of the target user includes:

[0013] A sliding window is used to divide the refractive error sequence in the historical visual acuity data to obtain time subsequences;

[0014] The slope of the linear fit between data points in the time subsequence is used as a trend feature of the long-term change direction of visual acuity within the target user's window.

[0015] The standard deviation of the data points in the time subsequence from the mean of the subsequence is used as the fluctuation characteristic of the short-term fluctuation intensity of visual acuity within the target user's window.

[0016] Identify the characteristic patterns of refractive error sequences and seasonal cycles in the vision-related historical data to obtain periodic features;

[0017] The trend features, fluctuation features, and periodic features are combined to obtain the time-series feature set of the target user.

[0018] In a preferred embodiment, identifying characteristic patterns in the refractive error sequence of the vision-related historical data that are associated with seasonal cycles to obtain periodic features includes:

[0019] Identify the potential periodic components present in the refractive power sequence;

[0020] The potential periodic components are matched with a preset seasonal periodic template;

[0021] When the potential periodic component successfully matches the seasonal periodic template, the period length, amplitude, and phase parameters of the seasonal periodic template are extracted.

[0022] Based on the period length, amplitude, and phase parameters, a periodic feature describing the seasonal variation of refractive power is generated.

[0023] In a preferred embodiment, obtaining the target user's average daily outdoor light exposure duration and average daily near-vision usage duration within the same time range to obtain the target user's environmental and behavioral monitoring data includes:

[0024] Obtain the original light intensity monitoring data and original screen usage distance monitoring data of the target user;

[0025] Identify and accumulate the time periods exceeding a preset threshold in the original light intensity monitoring data to obtain the average daily outdoor light exposure duration of the target user;

[0026] Identify and accumulate the time periods in the original screen usage distance monitoring data where the usage distance is less than a preset threshold to obtain the average daily near-field eye use time of the target user;

[0027] The average daily outdoor light exposure duration and the average daily near-field eye use duration are combined to obtain the environmental and behavioral monitoring data of the target user.

[0028] In a preferred embodiment, the step of standardizing, aligning, and weighting the trend features, fluctuation features, and periodic features in the time-series feature set with the environmental and behavioral monitoring data to obtain the target user's eye use feature vector includes:

[0029] The trend features, fluctuation features, and periodic features in the time series feature set are normalized respectively to obtain standardized time series sub-features;

[0030] The average daily outdoor light duration and the average daily near-field eye use duration are normalized to obtain standardized environmental and behavioral sub-features.

[0031] Based on the prior knowledge of the impact of the standardized temporal sub-features and the standardized environmental and behavioral sub-features on myopia risk, corresponding correlation weight coefficients are assigned to the sub-features.

[0032] Using the aforementioned correlation weight coefficients, the standardized temporal sub-features and the standardized environmental and behavioral sub-features are weighted and calculated to obtain the weighted feature components.

[0033] All the weighted feature components are concatenated in a predetermined order to obtain the eye feature vector of the target user.

[0034] In a preferred embodiment, the step of assigning corresponding correlation weight coefficients to the sub-features based on prior knowledge of the impact of the standardized temporal sub-features and the standardized environmental and behavioral sub-features on myopia risk includes:

[0035] Based on the consensus of ophthalmology, an expert experience rule base was established to map the relationship between feature combinations and risk contribution.

[0036] The standardized temporal sub-features and the standardized environmental and behavioral sub-features are used as input conditions and submitted to the rule engine of the expert experience rule base.

[0037] Traverse the expert experience rule base and filter out multiple candidate rules that match the input conditions;

[0038] The multiple candidate rules are conflict-resolved to generate sub-feature corresponding association weight coefficients for the target user.

[0039] In a preferred embodiment, the step of assessing the myopia risk of the target user based on the eye use feature vector includes:

[0040] The eye use feature vector is simultaneously input into multiple heterogeneous primary risk assessors to obtain multiple initial risk levels for the target user;

[0041] Identify the conflicts and consensuses in the conclusions drawn from the multiple initial risk levels under different judgment logics;

[0042] Based on the aforementioned conflicts and consensus, the initial risk levels where conflicts exist are coordinated and arbitrated.

[0043] Based on the results of coordination and arbitration, a myopia risk assessment result is generated for the target user.

[0044] In a preferred embodiment, the primary risk assessor includes:

[0045] The first type of primary risk assessor based on rule reasoning includes: loading a preset set of myopia risk assessment rules, comparing and performing logical operations on each component of the eye use feature vector with the threshold conditions in the set of myopia risk assessment rules, generating a rule matching score, and outputting the corresponding initial risk level based on the rule matching score.

[0046] The second type of primary risk assessor based on case similarity includes: retrieving several historical cases that are closest to the eye use feature vector in the historical case feature vector space, and determining and outputting the corresponding initial risk level through a voting mechanism based on the risk level marked by the several historical cases;

[0047] The third type of primary risk assessor based on statistical distribution includes: mapping the eye use feature vector to a feature distribution model composed of the features of historical high-risk users and low-risk users, calculating the probability that the eye use feature vector belongs to the high-risk feature cluster, and outputting the corresponding initial risk level based on the probability.

[0048] Compared with the prior art, the present invention has the following beneficial effects:

[0049] 1. This invention extracts three types of time-series features—trend, fluctuation, and cycle—from historical vision-related data, integrates environmental behavioral data such as average daily outdoor light exposure duration and near-field eye use duration, and comprehensively captures the patterns and influencing factors of vision changes through standardized alignment and correlation weighting based on ophthalmological consensus. Furthermore, by using conflict coordination and arbitration of multiple heterogeneous primary decision-makers, the bias of single-dimensional or single-logic assessments is significantly reduced, making the results more consistent with the user's actual risk situation.

[0050] 2. This invention utilizes artificial intelligence to automate the entire process of data extraction, feature fusion, and risk assessment, avoiding the tedious and time-consuming manual intervention. The generated eye use assessment report combines risk results with environmental behavior monitoring data to provide users with targeted references, helping to detect and intervene in myopia early, while balancing assessment efficiency and practical application guidance. Attached Figure Description

[0051] The present disclosure will be described in more detail below based on embodiments and with reference to the accompanying drawings:

[0052] Figure 1 The flowchart of an artificial intelligence-based myopia risk assessment method according to Embodiment 1 of the present invention is shown.

[0053] Figure 2 The diagram shows a functional block diagram of an artificial intelligence-based myopia risk assessment device according to Embodiment 2 of the present invention.

[0054] Figure 3 The diagram shows the system composition of the artificial intelligence-based myopia risk assessment method according to Embodiment 3 of the present invention. Detailed Implementation

[0055] To enable those skilled in the art to better understand the technical solutions of this disclosure, and to fully understand and implement the process of how this disclosure applies technical means to solve technical problems and achieve corresponding technical effects, the technical solutions in the embodiments of this disclosure will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this disclosure, not all embodiments. The embodiments of this disclosure and the various features within them can be combined with each other without conflict, and the resulting technical solutions are all within the protection scope of this disclosure. All other embodiments obtained by those skilled in the art based on the embodiments of this disclosure without creative effort should fall within the protection scope of this disclosure.

[0056] It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although a logical order is shown in the flowchart, in some cases the steps shown or described may be executed in a different order than that shown here.

[0057] Example 1

[0058] Figure 1 This is a flowchart illustrating an artificial intelligence-based myopia risk assessment method provided in an embodiment of this disclosure. Figure 1 As shown, a smart device control method includes:

[0059] In this embodiment of the invention, S1. Obtain historical vision-related data of the target user;

[0060] Complete the unique identity verification of the target user, collect the target user's legally valid identity information, and match this information with the identity information database of the vision data storage provider field by field. If the matching result is completely consistent, the identity verification is completed. If the verification is not completed, all subsequent operations are terminated, thereby ensuring that the retrieved vision-related historical data accurately corresponds to the target user.

[0061] Identify the legal storage provider for vision-related historical data and complete the compliant authorization process for data retrieval. The legal storage provider for vision-related historical data is limited to professional ophthalmology medical institutions, regular optometry centers, and physical examination institutions with vision testing qualifications. Submit the target user's identity verification certificate and a written data retrieval authorization document signed by the target user to such storage provider. The authorization document should clearly indicate that the target data to be retrieved is the target user's vision-related historical data, and specify the purpose, scope, and duration of data retrieval. The storage provider will review the authenticity, completeness, and validity of the authorization document. After the review is approved, a data retrieval acceptance receipt will be issued, thereby establishing a compliant basis for obtaining the target user's vision-related historical data.

[0062] Submit a list of vision-related historical data for the target user to a compliance-approved data storage provider. The list should clearly specify the types of vision-related historical data to be obtained, including all vision-related historical test records such as uncorrected visual acuity test data, corrected visual acuity test data, refractive error test data, intraocular pressure test data, and fundus examination data. The storage provider should also be required to provide the test time, test location, and test personnel information for each data item to ensure that the retrieved vision-related historical data is a complete historical record of the target user.

[0063] According to the submitted retrieval list, the data management personnel of the vision-related historical data storage provider simultaneously retrieve the target user's vision-related historical data from its official electronic data storage system and paper archive storage area. In the electronic data storage system, all vision-related historical test electronic records uniquely bound to the target user's identity are directly extracted. In the paper archive storage area, the corresponding vision-related historical test paper archives are searched one by one according to the target user's identity information. The vision-related historical data in the paper archives are manually transcribed page by page and item by item to ensure that the transcribed content is completely consistent with the text and values ​​of the original paper archive records, without any data omissions or information deviations.

[0064] A full consistency check is performed on the target user's vision-related historical data retrieved from the storage provider and manually transcribed. The electronic vision-related historical data extracted from the electronic data storage system is compared item by item and content by content with the manually transcribed paper vision-related historical data. The focus is on verifying whether core information such as test time, test value, and test conclusion are completely matched. If any inconsistencies are found, the vision-related historical data storage provider is immediately notified, and the provider's professional testing personnel will review and confirm the original test files until all information in the electronic data and the paper-transcribed data are completely consistent. This ensures that the verified vision-related historical data is complete and error-free.

[0065] The target users' vision-related historical data that has completed consistency verification is standardized and organized. All vision-related historical data are arranged in chronological order of testing time. Each test data item is labeled with the corresponding test type and test background information. The organized vision-related historical data is integrated into a unified written data document and electronic data document. The document retains only the target users' vision-related historical data and corresponding related information, and removes all redundant content that is not related to vision testing. This ensures that the organized document can clearly and completely present the target users' vision-related historical data.

[0066] Once the acquisition and confirmation of the target user's vision-related historical data are completed, the compiled written and electronic documents of the vision-related historical data will be submitted to the target user for item-by-item verification. After the target user confirms all the vision-related historical data in the document, they will sign the data acquisition confirmation document. The vision-related historical data storage provider will also stamp and file the confirmation document. The confirmation document will clearly record the scope, quantity, and compilation status of the target user's vision-related historical data acquired this time. The signing of this document signifies that the target user's vision-related historical data has been successfully acquired.

[0067] S2. Extract the temporal features reflecting the dynamic trend and periodic pattern of vision changes from the vision-related historical data to obtain the temporal feature set of the target user;

[0068] In this embodiment of the invention, the step of extracting time-series features reflecting the dynamic trend and periodic pattern of vision changes from the vision-related historical data to obtain the time-series feature set of the target user includes:

[0069] A sliding window is used to divide the refractive error sequence in the historical visual acuity data to obtain time subsequences;

[0070] The slope of the linear fit between data points in the time subsequence is used as a trend feature of the long-term change direction of visual acuity within the target user's window.

[0071] The standard deviation of the data points in the time subsequence from the mean of the subsequence is used as the fluctuation characteristic of the short-term fluctuation intensity of visual acuity within the target user's window.

[0072] Identify the characteristic patterns of refractive error sequences and seasonal cycles in the vision-related historical data to obtain periodic features;

[0073] The trend features, fluctuation features, and periodic features are combined to obtain the time-series feature set of the target user.

[0074] The identification of characteristic patterns in the refractive error sequence of the vision-related historical data that are associated with seasonal cycles yields periodic features, including:

[0075] Identify the potential periodic components present in the refractive power sequence;

[0076] The potential periodic components are matched with a preset seasonal periodic template;

[0077] When the potential periodic component successfully matches the seasonal periodic template, the period length, amplitude, and phase parameters of the seasonal periodic template are extracted.

[0078] Based on the period length, amplitude, and phase parameters, a periodic feature describing the seasonal variation of refractive power is generated.

[0079] A sliding window partitioning operation is performed on the refractive error sequence of the target user. First, the refractive error sequence of the target user is sorted sequentially from earliest to latest according to the detection time corresponding to the refractive error data, ensuring that all data within the refractive error sequence are in a continuous and ordered state along the time dimension. Then, a fixed range of window boundaries is defined, which can accommodate a fixed number of continuous refractive error data. The starting position of the sorted refractive error sequence is used as the initial placement position of the window, so that the window contains a fixed number of continuous refractive error data at the beginning of the refractive error sequence. Next, the window is shifted backward along the time axis of the refractive error sequence, moving only the position of one refractive error data at each shift. After the shift, the fixed number of continuous refractive error data at the new position is redefined by the window boundary. The shifting and defining operations are continuously performed until the rear boundary of the window covers the last refractive error data of the refractive error sequence, at which point all operations are stopped. Each ordered data set containing a fixed number of continuous refractive error data defined on the refractive error sequence in this way is a time subsequence obtained by partitioning the refractive error sequence through a sliding window.

[0080] Linear fitting and trend feature extraction are performed on a single time subsequence. First, the horizontal dimension of the two-dimensional reference plane is determined as the detection time, and the vertical dimension is the diopter value. The detection time and corresponding diopter value of each diopter data point in the time subsequence are used as a set of coordinates. All diopter data points are accurately marked on the corresponding coordinate positions of the two-dimensional reference plane. Then, a straight line is drawn in the two-dimensional reference plane. During the drawing process, the tilt angle and position of the line are adjusted so that the sum of the vertical distances from the line to each diopter data point in the two-dimensional reference plane is minimized. At this point, the linear fitting of all data points in the time subsequence is completed. The straight line formed by the fitting is the linear fitting line. The tilt of the linear fitting line is the linear fitting slope. The linear fitting slope is directly determined as the trend feature of the long-term change direction of vision within the target user window corresponding to the time subsequence.

[0081] To calculate the standard deviation and extract fluctuation characteristics for a single time subsequence, the following steps are taken: First, the total number of refractive error data points in the time subsequence is counted. Then, the specific values ​​of all refractive error data points in the time subsequence are summed one by one to obtain the sum of all refractive error values. This sum is divided by the total number of refractive error data points to obtain the subsequence mean. Next, the difference between the specific value of each refractive error data point in the time subsequence and the subsequence mean is calculated sequentially. Each calculated difference is squared to eliminate the positive or negative attribute of the difference. Then, all the squared values ​​are summed to obtain the sum of squares. This sum of squares is divided by the total number of refractive error data points in the time subsequence to obtain a new calculation result. Finally, the square root of this result is performed to restore the dimensions of the value. The final value obtained is the standard deviation between the data points in the time subsequence and the subsequence mean. This standard deviation is directly determined as the fluctuation characteristic of the short-term visual acuity fluctuation intensity within the target user window corresponding to the time subsequence.

[0082] To identify potential periodic components in the complete refractive error sequence of the target user, the refractive error sequence is first arranged in ascending order of detection time. The horizontal dimension of the two-dimensional plane is defined as detection time, and the vertical dimension as refractive error value. The detection time and corresponding refractive error value of each data point in the sorted refractive error sequence are used as a coordinate system. All refractive error data points are precisely marked on their corresponding coordinate positions on the two-dimensional plane. Then, adjacent refractive error data points in the two-dimensional plane are connected sequentially with straight lines to form a continuous refractive error variation curve that conforms to the changing pattern of all data points. Subsequently, this refractive error variation curve is analyzed... A comprehensive observation and analysis is conducted, starting from the beginning and ending points of the curve, to identify recurring patterns of fluctuation. This confirms that the pattern appears repeatedly in a fixed pattern. The time span required for the repeated fluctuation pattern to complete a full rise, fall, and recovery is then precisely extracted. Simultaneously, the maximum and minimum values ​​of the refractive error in a single complete fluctuation are extracted to clarify the range of numerical variation. The refractive error variation portion with a fixed recurring pattern identified and extracted through these operations represents the potential periodic component in the refractive error sequence.

[0083] The identified potential periodic components are matched with a preset seasonal periodic template. First, the core reference features contained in the preset seasonal periodic template are sorted out, including the template's pre-defined seasonal refractive error fluctuation reference pattern, the template's set reference time span for completing one full fluctuation of the fluctuation reference pattern, the template's set seasonal time interval range corresponding to the fluctuation reference pattern, and the template's preset seasonal typical refractive error fluctuation value range. Then, all features of the potential periodic components are extracted, including the actual repeating fluctuation pattern presented by the potential periodic components, the actual time span for the potential periodic components to complete one full repeating fluctuation, the time interval range corresponding to the actual fluctuation of the potential periodic components, and the actual refractive error fluctuation value range of the potential periodic components. Next, the features of the potential periodic components and the seasonal periodic template are compared one by one dimension. First, the fluctuation pattern is compared to see if it is consistent. Then, the actual time span is compared to see if it matches the reference time span. Then, the actual time interval is compared to see if it matches the seasonal time interval set by the template. Finally, the actual fluctuation value range is compared to see if it matches the template's preset typical fluctuation value range. This completes the matching operation of all dimensions, which is the complete process of matching the potential periodic components with the preset seasonal periodic template.

[0084] When the potential periodic component is completely consistent with all dimensional features of the preset seasonal periodic template, it is determined that the potential periodic component and the seasonal periodic template are successfully matched. At this time, the period length is extracted from the preset seasonal periodic template. The period length is the standard time span required for the pre-set refractive error fluctuation reference pattern of the corresponding season to complete one complete repeating fluctuation. Then, the amplitude is extracted from the template. The amplitude is the difference between the maximum and minimum values ​​of the refractive error value in the pre-set refractive error fluctuation reference pattern of the corresponding season. It can intuitively reflect the magnitude of the seasonal refractive error fluctuation set by the template. Next, the phase parameter is extracted from the template. The phase parameter includes the starting time node of the pre-set refractive error fluctuation reference pattern of the corresponding season and the initial refractive error value corresponding to the starting time node. The period length, amplitude and phase parameters in the template are completely extracted, and all operations of this step are completed.

[0085] Based on the period length, amplitude, and phase parameters extracted from the template, periodic features are generated. First, using the extracted period length as the core, the time pattern of the target user's refractive error completing a full fluctuation with seasonal changes is clarified, and the time period attribute of the fluctuation is determined. Then, using the extracted amplitude as the core, the magnitude of the fluctuation of the target user's refractive error with seasonal changes is clarified, reflecting the range of numerical changes in refractive error within the seasonal cycle. Next, using the extracted phase parameter as the core, the starting time node and the initial refractive error value of the target user's refractive error with seasonal changes are clarified, reflecting the initial state at the beginning of the fluctuation. Subsequently, the time pattern reflected by the period length, the fluctuation magnitude reflected by the amplitude, and the initial state reflected by the phase parameter are organically integrated to form a set of information that can comprehensively, accurately, and systematically describe the overall pattern and specific characteristics of the target user's refractive error with seasonal changes. This set of information is the periodic feature describing the seasonal changes in refractive error.

[0086] The trend features, fluctuation features, and periodic features are combined to obtain a time-series feature set. First, all time subsequences obtained by sliding window division are sorted out, and the trend features of the long-term change direction of vision within the target user window corresponding to each time subsequence are extracted. All trend features are centrally organized to form a trend feature set. Then, the fluctuation features of the short-term fluctuation intensity of vision within the target user window corresponding to each time subsequence are extracted, and all fluctuation features are centrally organized to form a fluctuation feature set. Next, the generated periodic features describing the seasonal changes of refractive power are organized to ensure that the information of the periodic features is complete and without omissions. Subsequently, the organized trend feature set, fluctuation feature set, and periodic feature set are comprehensively summarized, and all information of the three types of features is included in a unified feature set. This feature set can simultaneously reflect the long-term change direction of vision, short-term fluctuation intensity, and seasonal change pattern of refractive power of the target user. It is a complete set containing three core time-series features, and this complete set is the time-series feature set of the target user.

[0087] S3. Obtain the average daily outdoor light exposure duration and average daily close-range eye use duration of the target user within the same time range to obtain the environmental and behavioral monitoring data of the target user;

[0088] In this embodiment of the invention, obtaining the target user's average daily outdoor light exposure duration and average daily near-vision usage duration within the same time range to obtain the target user's environmental and behavioral monitoring data includes:

[0089] Obtain the original light intensity monitoring data and original screen usage distance monitoring data of the target user;

[0090] Identify and accumulate the time periods exceeding a preset threshold in the original light intensity monitoring data to obtain the average daily outdoor light exposure duration of the target user;

[0091] Identify and accumulate the time periods in the original screen usage distance monitoring data where the usage distance is less than a preset threshold to obtain the average daily near-field eye use time of the target user;

[0092] The average daily outdoor light exposure duration and the average daily near-field eye use duration are combined to obtain the environmental and behavioral monitoring data of the target user.

[0093] The original records of light intensity monitoring and screen usage distance monitoring of the target user are extracted from the local storage modules of the light intensity monitoring device and the screen usage distance monitoring device used by the target user. During the extraction process, the two types of original monitoring records are sorted and arranged in order from morning to night according to the actual monitoring time of the monitoring devices. This ensures that the extracted original records of light intensity monitoring contain the specific time of each monitoring and the corresponding light intensity value, and the extracted original records of screen usage distance monitoring contain the specific time of each monitoring and the corresponding screen usage distance value. Both types of original records retain all data from the monitoring process without any omissions, and the monitoring time dimension of the two types of data remains completely corresponding. The original records of light intensity monitoring and screen usage distance monitoring extracted and sorted by the above operations are the original light intensity monitoring data and the original records of screen usage distance monitoring extracted and sorted by the above operations are the original screen usage distance monitoring data.

[0094] The original light intensity monitoring data of the target user was analyzed to confirm that each record contained the corresponding monitoring time and light intensity value. Each light intensity value in the original light intensity monitoring data was compared with a preset threshold to accurately mark all monitoring data points whose light intensity values ​​exceeded the preset threshold. Adjacent marked monitoring data points were connected in chronological order of monitoring time to form one or more consecutive time periods in which the light intensity exceeded the preset threshold. For each consecutive time period, the complete time span from the start time to the end time of monitoring was calculated. The time spans of all consecutive time periods were added together to obtain the total outdoor light duration of the target user within the monitoring period. The total number of complete monitoring days corresponding to the original light intensity monitoring data was counted. The total outdoor light duration was then evenly distributed according to the total number of monitoring days to obtain the average daily outdoor light duration of the target user.

[0095] The original screen usage distance monitoring data of the target user was analyzed to confirm that each record contained the corresponding monitoring time and screen usage distance value. Each screen usage distance value in the original screen usage distance monitoring data was compared with a preset threshold to accurately mark all monitoring data points whose screen usage distance value was less than the preset threshold. Adjacent marked monitoring data points were connected in chronological order of monitoring time to form one or more consecutive time periods in which the screen usage distance was less than the preset threshold. The complete time span from the start time to the end time of each consecutive time period was calculated. The time spans of all consecutive time periods were added together to obtain the total near-field eye use time of the target user within the monitoring period. The total number of complete monitoring days corresponding to the original screen usage distance monitoring data was counted. The total near-field eye use time was averaged according to the total number of monitoring days to obtain the target user's average daily near-field eye use time.

[0096] The system integrates the calculated daily average outdoor light exposure duration and daily average near-field eye use duration of the target users, assigning unique information names to both. This ensures that the two types of duration information are clearly distinguishable and unambiguous. The complete set of information, which includes both the daily average outdoor light exposure duration and the daily average near-field eye use duration, is then directly identified as the environmental and behavioral monitoring data for the target users.

[0097] S4. The trend features, fluctuation features and periodic features in the time series feature set are standardized, aligned and weighted with the environmental and behavioral monitoring data to obtain the eye use feature vector of the target user;

[0098] In this embodiment of the invention, the step of standardizing, aligning, and weighting the trend features, fluctuation features, and periodic features in the time-series feature set with the environmental and behavioral monitoring data to obtain the target user's eye use feature vector includes:

[0099] The trend features, fluctuation features, and periodic features in the time series feature set are normalized respectively to obtain standardized time series sub-features;

[0100] The average daily outdoor light duration and the average daily near-field eye use duration are normalized to obtain standardized environmental and behavioral sub-features.

[0101] Based on the prior knowledge of the impact of the standardized temporal sub-features and the standardized environmental and behavioral sub-features on myopia risk, corresponding correlation weight coefficients are assigned to the sub-features.

[0102] Using the aforementioned correlation weight coefficients, the standardized temporal sub-features and the standardized environmental and behavioral sub-features are weighted and calculated to obtain the weighted feature components.

[0103] All the weighted feature components are concatenated in a predetermined order to obtain the eye feature vector of the target user.

[0104] The step involves assigning corresponding correlation weight coefficients to the sub-features based on prior knowledge of the impact of the standardized temporal sub-features and the standardized environmental and behavioral sub-features on myopia risk, including:

[0105] Based on the consensus of ophthalmology, an expert experience rule base was established to map the relationship between feature combinations and risk contribution.

[0106] The standardized temporal sub-features and the standardized environmental and behavioral sub-features are used as input conditions and submitted to the rule engine of the expert experience rule base.

[0107] Traverse the expert experience rule base and filter out multiple candidate rules that match the input conditions;

[0108] The multiple candidate rules are conflict-resolved to generate sub-feature corresponding association weight coefficients for the target user.

[0109] We analyze the time-series features of the target users and extract all feature values ​​for trend, fluctuation, and periodic features. We then perform normalization on each of these three types of features. First, we determine the maximum and minimum values ​​among all feature values ​​for each feature type and calculate the difference between them. Next, we subtract the minimum value from each feature value in that type and divide the result by the difference to obtain the standardized value for each feature value. All standardized values ​​obtained after this process for trend features constitute the standardized trend time-series sub-feature; all standardized values ​​obtained after this process for fluctuation features constitute the standardized fluctuation time-series sub-feature; and all standardized values ​​obtained after this process for periodic features constitute the standardized periodic time-series sub-feature. These three types of standardized sub-features together form the standardized time-series sub-feature.

[0110] Feature values ​​of the target user's average daily outdoor light exposure duration and average daily near-field eye use duration are extracted. Normalization processing is performed on the two types of durations separately. During processing, the maximum and minimum values ​​of all feature values ​​under a single type of duration are first determined, and the difference between the maximum and minimum values ​​is calculated. Then, the minimum value is subtracted from each feature value in that type of duration, and the result is divided by the difference to obtain the standardized value corresponding to each feature value. All standardized values ​​of the average daily outdoor light exposure duration after this processing constitute the standardized outdoor light exposure sub-feature, and all standardized values ​​of the average daily near-field eye use duration after this processing constitute the standardized near-field eye use sub-feature. The two types of standardized sub-features together constitute the standardized environmental and behavioral sub-features.

[0111] This study compiles clinically validated and industry-recognized medical research findings, clinical experience, and industry technical standards related to vision in the ophthalmology field to form an ophthalmological medical consensus. The feature types included in the rule base are standardized temporal sub-features and standardized environmental and behavioral sub-features. All feature dimensions of these two types are analyzed and combined to form different feature combinations. The study analyzes the specific impact of each feature combination on vision risk based on the ophthalmological medical consensus, clarifies the unique mapping relationship between each feature combination and its corresponding risk contribution, and assigns a unique rule expression and rule identifier to each defined mapping relationship. All feature combinations, corresponding risk contributions, mapping relationships, and rule identifiers are organized according to a unified storage format and incorporated into a dedicated rule storage module, thus constructing an expert experience rule base that maps feature combinations to risk contributions.

[0112] All feature information of the standardized time-series sub-features and standardized environment and behavior sub-features is collected. According to the input data format preset by the rule engine of the expert experience rule base, the feature information of the two types of sub-features is matched and converted to ensure that the converted feature information is completely consistent with the data requirements of the rule engine. The standardized time-series sub-features and standardized environment and behavior sub-features after format conversion are integrated into a unified input condition data package. This data package is submitted to the rule engine of the expert experience rule base through the rule engine's dedicated data access channel. The rule engine performs full parsing of the received input condition data package and stores the parsed feature information according to feature type in the temporary data module of the rule engine, ensuring that all feature information is complete and without omissions.

[0113] The rule engine reads the rule content sequentially from the rule storage module of the expert experience rule base, starting with the first rule, according to the rule identifier order. It extracts the preset feature combination conditions in each rule and performs a precise feature-by-feature and dimension-by-dimensional comparison with the input condition feature information stored in the temporary data module to confirm whether the input condition completely contains all the elements of the rule feature combination conditions and whether the attributes of the feature elements are completely consistent with the rule settings. If the comparison result shows that the elements and attributes are completely matched, the rule is marked as a candidate rule and stored in the candidate rule module. If the comparison result shows that the elements or attributes are not matched, the rule is skipped and the next rule is read. The reading, comparison, and marking operations are continuously executed until the rule engine completes a complete traversal of all rules in the expert experience rule base. All the rules stored in the candidate rule module are the multiple candidate rules that match the input conditions.

[0114] The rule identifiers, feature combination conditions, and corresponding risk contributions of all candidate rules are extracted from the candidate rule module. The overlap of feature combinations and the conflict of risk contributions among all candidate rules are analyzed. Based on the consensus of ophthalmology, the candidate rules with conflicts are judged hierarchically in terms of the reliability of medical basis and the universality of clinical application. The core rules and secondary rules among the conflicting candidate rules are identified. The feature combination and risk contribution mapping relationship of the core rules are retained according to the principle of core rules taking priority. The content of secondary rules that conflict with the core rules is removed. For all candidate rules retained after conflict resolution, a unique weight value is assigned to each sub-feature in each rule according to the size of its corresponding risk contribution. This ensures that each standardized time-series sub-feature and standardized environment and behavior sub-feature of the target user has a corresponding unique weight value. All sub-features and their corresponding unique weight values ​​are integrated to form the sub-feature corresponding association weight coefficient for the target user.

[0115] Extract the weight values ​​corresponding to each standardized time-series sub-feature and standardized environment and behavior sub-feature from the corresponding association weight coefficients of the sub-features. Multiply all the standardized values ​​of each standardized sub-feature by the weight values ​​corresponding to that sub-feature to complete the weighted calculation operation of a single sub-feature. In order, integrate all the values ​​obtained after the weighted calculation of each sub-feature to form the weighted value set corresponding to that sub-feature. Each sub-feature information containing the weighted value set is a weighted feature component. After the above operation, all standardized time-series sub-features and standardized environment and behavior sub-features obtain the corresponding weighted feature components.

[0116] The predetermined feature splicing order in the expert experience rule base is determined. All weighted values ​​are extracted from each weighted feature component in sequence according to the predetermined order. All extracted weighted values ​​are continuously and orderly arranged and spliced ​​in the predetermined order so that all weighted values ​​form a continuous and uninterrupted numerical sequence with a fixed order. This continuous and orderly numerical sequence is the eye feature vector of the target user.

[0117] S5. Assess the myopia risk of the target user based on the eye use feature vector;

[0118] In this embodiment of the invention, the step of assessing the myopia risk of the target user based on the eye use feature vector includes:

[0119] The eye use feature vector is simultaneously input into multiple heterogeneous primary risk assessors to obtain multiple initial risk levels for the target user;

[0120] Identify the conflicts and consensuses in the conclusions drawn from the multiple initial risk levels under different judgment logics;

[0121] Based on the aforementioned conflicts and consensus, the initial risk levels where conflicts exist are coordinated and arbitrated.

[0122] Based on the results of coordination and arbitration, a myopia risk assessment result is generated for the target user.

[0123] The primary risk assessor includes:

[0124] The first type of primary risk assessor based on rule reasoning includes: loading a preset set of myopia risk assessment rules, comparing and performing logical operations on each component of the eye use feature vector with the threshold conditions in the set of myopia risk assessment rules, generating a rule matching score, and outputting the corresponding initial risk level based on the rule matching score.

[0125] The second type of primary risk assessor based on case similarity includes: retrieving several historical cases that are closest to the eye use feature vector in the historical case feature vector space, and determining and outputting the corresponding initial risk level through a voting mechanism based on the risk level marked by the several historical cases;

[0126] The third type of primary risk assessor based on statistical distribution includes: mapping the eye use feature vector to a feature distribution model composed of the features of historical high-risk users and low-risk users, calculating the probability that the eye use feature vector belongs to the high-risk feature cluster, and outputting the corresponding initial risk level based on the probability.

[0127] The system organizes all components of the target user's eye use feature vector and performs field adaptation and data format conversion according to the preset feature vector input formats of the first type of primary risk assessor based on rule reasoning, the second type based on case similarity, and the third type based on statistical distribution. This ensures that the converted eye use feature vector can be fully recognized and parsed by each type of primary risk assessor. The format-converted eye use feature vector is then simultaneously transmitted to the feature receiving modules of the three heterogeneous primary risk assessors. Each of the three types of primary risk assessors independently performs risk level determination. The first type of primary risk assessor based on rule reasoning first loads the preset myopia risk assessment rule set, and then compares each component of the eye use feature vector with the threshold conditions in the myopia risk assessment rule set. The system performs item comparison and logical operations to generate a rule matching score. Based on this score, it outputs the corresponding initial risk level. The second type of primary risk assessor, based on case similarity, first retrieves several historical cases that are closest to the eye use feature vector in the historical case feature vector space. Then, based on the risk levels marked on these historical cases, it determines and outputs the corresponding initial risk level through a voting mechanism. The third type of primary risk assessor, based on statistical distribution, first maps the eye use feature vector to a feature distribution model composed of the features of historical high-risk users and low-risk users. Then, it calculates the probability that the eye use feature vector belongs to the high-risk feature cluster. Subsequently, it outputs the corresponding initial risk level based on this probability. All the initial risk levels output by the three types of primary risk assessors together constitute multiple initial risk levels for the target user.

[0128] This study identifies the types of primary risk assessors corresponding to multiple initial risk levels, clarifying the assessment logic for each initial risk level as rule-based reasoning, case similarity-based assessment, and statistical distribution-based assessment. The initial risk level results output by each of the three types of assessors are compared one by one to confirm whether the conclusions of all initial risk levels are completely consistent. If all conclusions are identical, they are determined to be consensus conclusions under different assessment logics. If there are different conclusions, the assessment logics corresponding to the inconsistent initial risk levels are marked one by one, clarifying the conflicting assessment logic types and corresponding risk level conclusions. Simultaneously, the specific content of the consensus conclusions and their corresponding assessment logics are analyzed. This process comprehensively identifies the conflicts and consensuses of conclusions for multiple initial risk levels under different assessment logics, forming a complete conflict and consensus identification result.

[0129] The identification results of conflicts and consensuses are extracted, and the consensus conclusions formed under all judgment logics are retained. For the identified conflict conclusions, a coordinated arbitration judgment basis is established based on the core principles of ophthalmological clinical diagnosis and treatment, taking into account the technical characteristics of different judgment logics, in combination with the clinical judgment principles of ophthalmological medical consensus and myopia risk assessment. The judgment process of each primary risk judge that formed the conflict conclusions is retrospectively checked to confirm whether there are data processing deviations in the feature analysis, rule matching, case retrieval, and probability calculation stages when processing eye feature vectors. Erroneous conflict conclusions caused by data processing deviations are eliminated. For the remaining reasonable conflict conclusions, the importance level of the conflict risk under different judgment logics is determined according to the coordinated arbitration judgment basis, and the reference weight of each conflict conclusion is determined. The unified risk level conclusion after coordinated arbitration is determined in order of reference weight from high to low. In this way, the coordination and arbitration of the initial risk levels with conflicts are completed, and a complete coordinated arbitration result is formed.

[0130] The results of coordination and arbitration are compiled. If the identification results contain only consensus conclusions and no conflict conclusions, the consensus conclusion is directly determined as the core assessment result. If a unified risk level conclusion is formed after coordination and arbitration, the unified conclusion after coordination and arbitration is determined as the core assessment result. At the same time, the process of identifying conflicts and consensus, the judgment basis and handling process of coordination and arbitration are used as supplementary assessment information. The corresponding judgment basis and supplementary explanations are marked for the core assessment result to ensure the integrity and traceability of the assessment result. The integrated core assessment result and supplementary assessment information together constitute the complete assessment content, which is the myopia risk assessment result of the target user.

[0131] In this embodiment of the invention, S6. Based on the results of the myopia risk assessment and the environmental and behavioral monitoring data, an eye use assessment report for the target user is generated.

[0132] We analyzed the myopia risk assessment results and environmental and behavioral monitoring data of the target users, and extracted all supplementary assessment information, including the core assessment results, conflict and consensus identification process, and coordination and arbitration judgment basis, from the myopia risk assessment results. At the same time, we extracted the specific numerical information of the average daily outdoor light exposure time and the average daily near-vision time from the environmental and behavioral monitoring data. We checked the completeness of both types of data one by one to confirm that all core information was complete and uniquely corresponded to the target user. Then, we initially sorted the two types of data according to the classification method of assessment interpretation and data presentation to prepare data for the subsequent report content.

[0133] A standardized content framework for eye use assessment reports was established, comprising a basic information module, a risk assessment results module, an environmental and behavioral monitoring and analysis module, a comprehensive assessment conclusion module, and a personalized eye use recommendation module. Each module has a clearly defined scope of content presentation. The basic information module presents basic information about the user and the report; the risk assessment results module presents and interprets the myopia risk assessment results; the environmental and behavioral monitoring and analysis module analyzes environmental and behavioral monitoring data; the comprehensive assessment conclusion module forms an overall judgment; and the personalized eye use recommendation module develops targeted adjustment suggestions. These modules are logically interconnected and progressively build upon each other.

[0134] In the basic information module of the eye use assessment report, the unique identification information corresponding to the target user is accurately entered, the official generation time of the eye use assessment report is clearly marked, and the scope of application of the report is clearly stated as a professional assessment of the target user's current vision status and eye use behavior. It is only valid for the myopia risk assessment results and environmental and behavioral monitoring data obtained in this case, ensuring that the content of the basic information module is accurate and has unique identification.

[0135] The risk assessment results module of the eye use assessment report fully presents the core assessment results of the myopia risk assessment. It elaborates on the identification details of conflicts and consensuses in the formation of the core assessment results, the ophthalmological medical consensus and clinical judgment principles on which the coordination and arbitration are based, and the specific coordination and arbitration process. At the same time, it combines the recognized myopia risk grading standards in the ophthalmology field to provide a professional interpretation of the core assessment results, clarifying the natural vision development trend and potential vision health impact corresponding to the risk level, ensuring that the interpretation of the risk assessment results is professional and in line with clinical practice.

[0136] The environmental and behavioral monitoring and analysis module of the eye use assessment report fully presents the specific data on the target user's average daily outdoor light exposure time and average daily near-vision time. Combining the healthy eye use standards defined in the ophthalmological consensus, the two types of time data are compared item by item with the corresponding health standards, clearly indicating whether the average daily outdoor light exposure time meets the basic requirements for healthy eye use and whether the average daily near-vision time exceeds the reasonable range for healthy eye use. At the same time, based on the conclusions of ophthalmological clinical research, the intrinsic correlation between these two types of environmental and behavioral factors, namely average daily outdoor light exposure time and average daily near-vision time, and the target user's myopia risk assessment results is analyzed, clearly illustrating the specific impact of these two types of factors on the formation of the current myopia risk level.

[0137] In the comprehensive assessment conclusion module of the eye use assessment report, a comprehensive judgment is made by combining the professional interpretation of the risk assessment results module with the comparative analysis results of the environmental and behavioral monitoring and analysis module. The report systematically summarizes the target user's current myopia risk status, the core environmental and behavioral factors affecting this risk status, and clearly points out the advantages and problems of the target user's current eye use behavior in accordance with healthy eye use standards. A comprehensive assessment conclusion that is both professional and comprehensive is formed. This conclusion must be highly consistent with all the analysis content in the preceding text and have no logical deviations.

[0138] In the personalized eye care advice module of the eye care assessment report, based on the myopia risk status and eye care behavior problems identified in the comprehensive assessment conclusion, and in conjunction with ophthalmological consensus and healthy eye care guidelines, specific and actionable adjustment measures are formulated to optimize the average daily outdoor light exposure time. The implementation standards and long-term adherence requirements of these measures are clearly defined. At the same time, specific and implementable control plans are formulated to manage the average daily near-field eye use time, specifying the implementation methods and frequency of these plans. In addition, based on the target user's current myopia risk level, corresponding requirements for regular vision monitoring and daily vision protection measures are formulated to ensure that all recommendations are aligned with the target user's actual monitoring data and are targeted and practical.

[0139] We conducted a comprehensive review of all modules in the eye use assessment report, verifying word by word whether all data presented in the report was completely consistent with the original information of the myopia risk assessment results and environmental and behavioral monitoring data, and whether there were any data deviations or information errors. We also checked whether the content of each module was complete, logically coherent, and whether the professional interpretations were in line with the consensus of ophthalmology. At the same time, we removed all redundant statements and repetitive information from the report to ensure that the report content was concise and accurate.

[0140] Following the pre-set standardized content framework, the basic information module, risk assessment results module, environmental and behavioral monitoring and analysis module, comprehensive assessment conclusion module, and personalized eye care advice module, which have completed content filling and full-dimensional verification, are systematically integrated to form a complete, logically progressive, and coherent overall document, which is the eye care assessment report for the target user.

[0141] Example 2

[0142] like Figure 2 As shown in the figure, this embodiment also provides a functional block diagram of an artificial intelligence-based myopia risk assessment device.

[0143] The artificial intelligence-based myopia risk assessment device 100 described in this embodiment can be installed in a smart terminal. Depending on the functions implemented, the artificial intelligence-based myopia risk assessment device 100 may include a historical data acquisition module 101, a feature extraction module 102, an environmental and behavioral data acquisition module 103, a vector construction module 104, a risk assessment module 105, and a report generation module 106. The modules described in this invention can also be referred to as units, which are a series of computer program segments that can be executed by the smart terminal processor and perform a fixed function, stored in the smart terminal's memory.

[0144] In this embodiment, the functions of each module / unit are as follows:

[0145] The historical data acquisition module 101 is used to acquire vision-related historical data of the target user;

[0146] The feature extraction module 102 is used to extract time-series features from the vision-related historical data that reflect the dynamic trend and periodic pattern of vision changes, and obtain the time-series feature set of the target user;

[0147] The environmental and behavioral data acquisition module 103 is used to acquire the average daily outdoor light exposure duration and average daily close-range eye use duration of the target user within the same time range, and obtain the environmental and behavioral monitoring data of the target user.

[0148] The vector construction module 104 is used to standardize, align, and weight the trend features, fluctuation features, and periodic features in the time-series feature set with the environmental and behavioral monitoring data to obtain the eye use feature vector of the target user.

[0149] The risk assessment module 105 is used to assess the myopia risk of the target user based on the eye use feature vector.

[0150] The report generation module 106 is used to generate an eye use assessment report for the target user based on the results of the myopia risk assessment and the environmental and behavioral monitoring data.

[0151] In detail, each module in the artificial intelligence-based myopia risk assessment device 100 described in this embodiment of the invention uses the same technical means as the artificial intelligence-based myopia risk assessment method described in Embodiments 1 and 2, and can produce the same technical effect, which will not be repeated here.

[0152] Example 3

[0153] like Figure 3As shown, this embodiment also provides a system, which may include a processor 10, a memory 11, a communication bus 12 and a communication interface 13, and may also include a computer program stored in the memory 11 and capable of running on the processor 10, such as an artificial intelligence-based myopia risk assessment program.

[0154] In some embodiments, the processor 10 may be composed of integrated circuits, such as a single packaged integrated circuit or multiple integrated circuits with the same or different functions, including combinations of one or more central processing units (CPUs), microprocessors, digital processing chips, graphics processors, and various control chips. The processor 10 is the control unit of the system, connecting various components of the system through various interfaces and lines. It executes programs or modules stored in the memory 11 (e.g., executing an AI-based myopia risk assessment program) and calls data stored in the memory 11 to perform various system functions and process data.

[0155] The memory 11 includes at least one type of medium, including flash memory, portable hard drive, multimedia card, card-type memory (e.g., SD or DX memory), magnetic memory, disk, optical disk, etc. In some embodiments, the memory 11 can be an internal storage unit of the system, such as the system's portable hard drive. In other embodiments, the memory 11 can be an external storage device of the system, such as a plug-in portable hard drive, Smart Media Card (SMC), Secure Digital (SD) card, Flash Card, etc. Furthermore, the memory 11 can include both internal and external storage units. The memory 11 can be used not only to store application software and various types of data installed on the system, such as the code of an AI-based myopia risk assessment program, but also to temporarily store data that has been output or will be output.

[0156] The communication bus 12 can be a peripheral component interconnect (PCI) bus or an extended industry standard architecture (EISA) bus, etc. This bus can be divided into an address bus, a data bus, a control bus, etc. The bus is configured to enable communication between the memory 11 and at least one processor 10, etc.

[0157] The communication interface 13 is used for communication between the aforementioned system and other systems, including a network interface and a user interface. Optionally, the network interface may include a wired interface and / or a wireless interface (such as a Wi-Fi interface, Bluetooth interface, etc.), typically used to establish communication connections between the system and other systems. The user interface may be a display, an input unit (such as a keyboard), or optionally, a standard wired or wireless interface. Optionally, in some embodiments, the display may be an LED display, a liquid crystal display, a touch-sensitive liquid crystal display, or an OLED (Organic Light-Emitting Diode) touchscreen, etc. The display may also be appropriately referred to as a screen or display unit, used to display information processed in the system and to display a visual user interface.

[0158] The figure only shows a system with components. Those skilled in the art will understand that the structure shown in the figure does not constitute a limitation on the system and may include fewer or more components than shown, or combine certain components, or have different component arrangements.

[0159] For example, although not shown, the system may also include a power supply (such as a battery) to power the various components. Preferably, the power supply can be logically connected to the at least one processor 10 through a power management system, thereby enabling functions such as charging management, discharging management, and power consumption management through the power management system. The power supply may also include one or more DC or AC power supplies, a recharging system, a power fault detection circuit, a power converter or inverter, a power status indicator, and other arbitrary components. The system may also include various sensors, Bluetooth modules, Wi-Fi modules, etc., which will not be described in detail here.

[0160] It should be understood that the embodiments described are for illustrative purposes only and are not limited to this structure in the scope of the patent application.

[0161] The memory 11 in the system stores an artificial intelligence-based myopia risk assessment program, which is a combination of multiple instructions. When run in the processor 10, it can achieve the following:

[0162] S1. Obtain historical vision-related data of the target user;

[0163] S2. Extract the temporal features reflecting the dynamic trend and periodic pattern of vision changes from the vision-related historical data to obtain the temporal feature set of the target user;

[0164] S3. Obtain the average daily outdoor light exposure duration and average daily close-range eye use duration of the target user within the same time range to obtain the environmental and behavioral monitoring data of the target user;

[0165] S4. The trend features, fluctuation features and periodic features in the time series feature set are standardized, aligned and weighted with the environmental and behavioral monitoring data to obtain the eye use feature vector of the target user;

[0166] S5. Assess the myopia risk of the target user based on the eye use feature vector;

[0167] S6. Based on the results of the myopia risk assessment and the environmental and behavioral monitoring data, generate an eye use assessment report for the target user.

[0168] Specifically, the specific implementation method of the processor 10 for the above instructions can be referred to the description of the relevant steps in the corresponding embodiment of the accompanying drawings, and will not be repeated here.

[0169] Furthermore, if the modules / units integrated into the system are implemented as software functional units and sold or used as independent products, they can be stored in a medium. The medium can be volatile or non-volatile. For example, the medium may include: any entity or system capable of carrying the computer program code, a recording medium, a USB flash drive, a portable hard drive, a magnetic disk, an optical disk, a computer memory, or a read-only memory (ROM).

[0170] In the several embodiments provided by this invention, it should be understood that the disclosed systems, methods, and approaches can be implemented in other ways. For example, the system embodiments described above are merely illustrative; for instance, the division of modules is only a logical functional division, and other division methods may be used in actual implementation.

[0171] The modules described as separate components may or may not be physically separate. The components shown as modules 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 modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0172] Furthermore, the functional modules 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 in the form of hardware plus software functional modules.

[0173] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0174] The embodiments of this application can acquire and process relevant data based on artificial intelligence technology. Artificial intelligence (AI) refers to the theories, methods, technologies, and application systems that use digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use that knowledge to obtain optimal results.

[0175] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A myopia risk assessment method based on artificial intelligence, characterized in that, The method includes: S1. Obtain historical vision-related data of the target user; S2. Extract the temporal features reflecting the dynamic trend and periodic pattern of vision changes from the vision-related historical data to obtain the temporal feature set of the target user; S3. Obtain the average daily outdoor light exposure duration and average daily close-range eye use duration of the target user within the same time range to obtain the environmental and behavioral monitoring data of the target user; S4. The trend features, fluctuation features and periodic features in the time series feature set are standardized, aligned and weighted with the environmental and behavioral monitoring data to obtain the eye use feature vector of the target user; S5. Assess the myopia risk of the target user based on the eye use feature vector; S6. Based on the results of the myopia risk assessment and the environmental and behavioral monitoring data, generate an eye use assessment report for the target user.

2. The myopia risk assessment method based on artificial intelligence as described in claim 1, characterized in that, The extraction of time-series features reflecting the dynamic trends and cyclical patterns of vision changes from the historical vision-related data yields a time-series feature set for the target user, including: A sliding window is used to divide the refractive error sequence in the historical visual acuity data to obtain time subsequences; The slope of the linear fit between data points in the time subsequence is used as a trend feature of the long-term change direction of vision within the target user's window. The standard deviation of the data points in the time subsequence from the mean of the subsequence is used as the fluctuation characteristic of the short-term fluctuation intensity of visual acuity within the target user's window. Identify the characteristic patterns of refractive error sequences and seasonal cycles in the historical vision-related data to obtain periodic features; The trend features, fluctuation features, and periodic features are combined to obtain the time-series feature set of the target user.

3. The myopia risk assessment method based on artificial intelligence as described in claim 2, characterized in that, The identification of characteristic patterns in the refractive error sequence of the vision-related historical data that are associated with seasonal cycles yields periodic features, including: Identify the potential periodic components present in the refractive power sequence; The potential periodic components are matched with a preset seasonal periodic template; When the potential periodic component successfully matches the seasonal periodic template, the period length, amplitude, and phase parameters of the seasonal periodic template are extracted. Based on the period length, amplitude, and phase parameters, a periodic feature describing the seasonal variation of refractive power is generated.

4. The myopia risk assessment method based on artificial intelligence as described in claim 1, characterized in that, The process of obtaining the target user's average daily outdoor light exposure duration and average daily near-vision usage duration within the same time range to acquire the target user's environmental and behavioral monitoring data includes: Obtain the original light intensity monitoring data and original screen usage distance monitoring data of the target user; Identify and accumulate the time periods exceeding a preset threshold in the original light intensity monitoring data to obtain the average daily outdoor light exposure duration of the target user; Identify and accumulate the time periods in the original screen usage distance monitoring data where the usage distance is less than a preset threshold to obtain the average daily near-field eye use time of the target user; The average daily outdoor light exposure duration and the average daily near-field eye use duration are combined to obtain the environmental and behavioral monitoring data of the target user.

5. The myopia risk assessment method based on artificial intelligence as described in claim 4, characterized in that, The trend features, fluctuation features, and periodic features in the time-series feature set are standardized, aligned, and weighted with the environmental and behavioral monitoring data to obtain the target user's eye use feature vector, including: The trend features, fluctuation features, and periodic features in the time series feature set are normalized respectively to obtain standardized time series sub-features; The average daily outdoor light duration and the average daily near-field eye use duration are normalized to obtain standardized environmental and behavioral sub-features. Based on the prior knowledge of the impact of the standardized temporal sub-features and the standardized environmental and behavioral sub-features on myopia risk, corresponding correlation weight coefficients are assigned to the sub-features. Using the aforementioned correlation weight coefficients, the standardized temporal sub-features and the standardized environmental and behavioral sub-features are weighted and calculated to obtain the weighted feature components. All the weighted feature components are concatenated in a predetermined order to obtain the eye feature vector of the target user.

6. The myopia risk assessment method based on artificial intelligence as described in claim 5, characterized in that, The step involves assigning corresponding correlation weight coefficients to the sub-features based on prior knowledge of the impact of the standardized temporal sub-features and the standardized environmental and behavioral sub-features on myopia risk, including: Based on the consensus of ophthalmology, an expert experience rule base was established to map the relationship between feature combinations and risk contribution. The standardized temporal sub-features and the standardized environmental and behavioral sub-features are used as input conditions and submitted to the rule engine of the expert experience rule base. Traverse the expert experience rule base and filter out multiple candidate rules that match the input conditions; The multiple candidate rules are conflict-resolved to generate sub-feature corresponding association weight coefficients for the target user.

7. The myopia risk assessment method based on artificial intelligence as described in claim 1, characterized in that, The process of assessing the myopia risk of the target user based on the eye use feature vector includes: The eye use feature vector is simultaneously input into multiple heterogeneous primary risk assessors to obtain multiple initial risk levels for the target user; Identify the conflicts and consensuses in the conclusions drawn from the multiple initial risk levels under different judgment logics; Based on the aforementioned conflicts and consensus, the initial risk levels where conflicts exist are coordinated and arbitrated. Based on the results of coordination and arbitration, a myopia risk assessment result is generated for the target user.

8. The myopia risk assessment method based on artificial intelligence as described in claim 7, characterized in that, The primary risk assessor includes: The first type of primary risk assessor based on rule reasoning includes: loading a preset set of myopia risk assessment rules, comparing and performing logical operations on each component of the eye use feature vector with the threshold conditions in the set of myopia risk assessment rules, generating a rule matching score, and outputting the corresponding initial risk level based on the rule matching score. The second type of primary risk assessor based on case similarity includes: retrieving several historical cases that are closest to the eye use feature vector in the historical case feature vector space, and determining and outputting the corresponding initial risk level through a voting mechanism based on the risk level marked by the several historical cases; The third type of primary risk assessor based on statistical distribution includes: mapping the eye use feature vector to a feature distribution model composed of the features of historical high-risk users and low-risk users, calculating the probability that the eye use feature vector belongs to the high-risk feature cluster, and outputting the corresponding initial risk level based on the probability.

9. A myopia risk assessment device based on artificial intelligence, characterized in that, The system for implementing the artificial intelligence-based myopia risk assessment method according to claims 1-8 includes: The historical data acquisition module is used to acquire historical vision-related data of the target user; The feature extraction module is used to extract time-series features from the vision-related historical data that reflect the dynamic trend and periodic pattern of vision changes, and obtain the time-series feature set of the target user; The environmental and behavioral data acquisition module is used to acquire the average daily outdoor light exposure duration and average daily close-range eye use duration of the target user within the same time range, so as to obtain the environmental and behavioral monitoring data of the target user. The vector construction module is used to standardize, align, and weight the trend features, fluctuation features, and periodic features in the time-series feature set with the environmental and behavioral monitoring data to obtain the eye use feature vector of the target user. The risk assessment module is used to assess the myopia risk of the target user based on the eye use feature vector. The report generation module is used to generate an eye use assessment report for the target user based on the results of the myopia risk assessment and the environmental and behavioral monitoring data.

10. A system comprising a memory, a processor, and a computer program stored in the memory, characterized in that, The processor executes the computer program to implement the steps of the artificial intelligence-based myopia risk assessment method according to any one of claims 1 to 8.

Images