Intelligent auxiliary system for ophthalmology nursing

Abstract

The invention relates to the technical field of medical care, in particular to an intelligent auxiliary system for ophthalmology nursing, which comprises a health index distribution difference analysis module for acquiring health index data of an ophthalmology patient before nursing and after nursing, comparing the difference between the health index data of the ophthalmology patient before nursing and the health index data of the ophthalmology patient after nursing, and generating a health index transfer feature analysis result. According to the method, the difference between the health index data before and after nursing is analyzed, the multi-dimensional health data can be effectively normalized, and the change characteristics of the multi-dimensional health data can be quantified, so that the specific effect brought by nursing intervention is accurately presented. The quantification of the change characteristics not only enables the dynamic change of the health indexes to be more comparable, but also remarkably improves the accuracy of data processing. In the aspect of dynamic trend prediction, the change trend of the health index in the time sequence is extracted, and the trend change point is divided into a plurality of stages, so that the change node and the trend direction of the health state can be accurately captured.

Description

Technical Field

[0001] The present invention relates to the field of medical care technology, and in particular to an intelligent auxiliary system for ophthalmic care. Background Art

[0002] The field of medical care technology involves the application of technologies such as artificial intelligence, big data, and the Internet of Things, which can achieve real-time monitoring of patients' health conditions, remote care, and precise health intervention, thereby improving the efficiency of medical resource utilization and the quality of patient care.

[0003] Among them, the ophthalmic care intelligent assistance system is a digital care solution based on artificial intelligence and big data technology, which is used for health monitoring, care guidance and recovery management of patients with ophthalmic diseases. The system can provide medical staff with diagnosis and treatment assistance suggestions by intelligently analyzing the patient's eye health data.

[0004] Existing technologies mainly rely on basic analysis and diagnosis and treatment assistance of eye health data, but they are limited in processing complex time series data and capturing dynamic trends in health indicators. Due to the lack of in-depth application of normalized data before and after care, it is often difficult to intuitively present the specific effect differences brought about by nursing interventions. In addition, the changing trends of health data have not been effectively combined with the dynamic characteristics of time series. The diagnosis and treatment recommendations obtained only through static analysis lack the accurate capture of trend change nodes, which may lead to lags in health assessment and prediction. When identifying key change nodes in health data, existing technologies find it difficult to extract key points of data features from continuous change curves, resulting in incomplete extraction of nursing information and affecting the accurate evaluation of nursing effectiveness. Summary of the invention

[0005] The purpose of the present invention is to solve the shortcomings of the prior art and to propose an intelligent auxiliary system for ophthalmic care.

[0006] In order to achieve the above-mentioned object, the present invention adopts the following technical scheme: an intelligent auxiliary system for ophthalmic care comprises:

[0007] The health indicator distribution difference analysis module obtains the health indicator data of the ophthalmic patients before and after nursing care, compares the differences between the health indicator data of the ophthalmic patients before and after nursing care, and generates the health indicator transfer feature analysis results;

[0008] The dynamic trend prediction and segmented modeling module determines the change trend of the health indicator data of ophthalmic patients from before to after care in the time series according to the health indicator transfer feature analysis results, extracts the trend change points therefrom, builds a local fitting model according to the trend change points of each stage, analyzes the change direction of the health indicator data, predicts the future feature changes of the health data in each stage according to the change direction, and generates segmented trend prediction results;

[0009] The health data curvature feature extraction module converts the segmented trend prediction results into a continuous change curve and performs fitting, calculates the curvature values ​​of discrete points in the fitted change curve, compares them with a preset curvature threshold, marks the curvature values ​​exceeding the curvature threshold as key trend nodes, and generates key trend node identification results;

[0010] The health index dynamic update module classifies the key trend nodes in the key trend node identification result, analyzes the distribution law of the key trend nodes in the fitted change curve according to the classification result, and generates reference information for ophthalmic care.

[0011] As a further solution of the present invention, the steps for obtaining the health indicator transfer characteristic analysis results are specifically as follows:

[0012] Based on the patient health index data before and after care, intraocular pressure, retinal thickness and corneal morphology parameters are obtained, and the data of multiple dimensions are normalized to obtain an integrated health index data table;

[0013] Based on the integrated health indicator data table, the formula is used:

[0014] D i =X i,after -X i,before ;

[0015] Calculate the difference value D of the i-th health indicator i , get the indicator difference data;

[0016] Among them, X i,after is the value of the i-th health indicator after care, X i,before is the value of the ith health indicator before care;

[0017] Based on the indicator difference data, the maximum value, minimum value and concentrated distribution range of each type of indicator are counted respectively, and the positive and negative change range of the health indicator is determined in combination with the upper and lower limits of the difference value, and the health indicator transfer characteristic analysis results are generated according to the statistical information.

[0018] As a further solution of the present invention, the step of extracting trend change points is specifically:

[0019] Based on the analysis results of the health indicator transfer characteristics, by sorting the intraocular pressure, retinal thickness and corneal morphology parameter data before and after care, a data sequence is generated in the order of time points, and the change rate and direction of adjacent time points in the time series are analyzed one by one to generate the health indicator time series change trend analysis results;

[0020] Based on the analysis results of the time series change trend of the health indicators, the formula is adopted:

[0021]

[0022] Calculate the maximum absolute value T in the health indicator time series to obtain the trend change point;

[0023] Among them, argmax represents the variable at which the objective function value reaches the maximum value, t is the time point variable, [t1, t2] represents the time point t1 when the care starts and the time point t2 when the care ends, ΔY is the change in the health indicator value, and Δt represents the interval between two adjacent time points in the time series.

[0024] As a further solution of the present invention, the step of obtaining the local fitting model is specifically as follows:

[0025] Dividing the time series into multiple stages according to the trend change points, sorting and grouping the health indicator data in each stage, and generating a staged health indicator data table;

[0026] Based on the staged health index data table, by analyzing the changing rules of the data in each stage and the changing trends of the health indexes in the corresponding stage, a local fitting model is obtained by data fitting.

[0027] As a further solution of the present invention, the step of obtaining the segmented trend prediction result is specifically as follows:

[0028] According to the data in the local fitting model, by extracting the key parameters in the fitting model at each stage, analyzing the change direction of the health indicator in combination with the fitting results, and arranging the change trend in sections, an analysis result of the change direction of the current health indicator is generated;

[0029] Based on the analysis results of the change direction of the current health indicators, the formula is adopted:

[0030] Y pred,i′+1 =Y current,i′ +a i′ ·Δt′+b i′ ·(Δt′) 2 ;

[0031] Calculate the health index value Y at the future i′+1 time point pred,i′+1 , get the segmented trend prediction results;

[0032] Among them, Y current,i′ is the health indicator value at the current time point, a i′ is the linear change slope, b i′ is the quadratic variation coefficient, and Δt′ represents the time length from the current time point to the predicted time point.

[0033] As a further solution of the present invention, the step of obtaining the curvature value of the discrete points in the variation curve after calculation and fitting is specifically as follows:

[0034] The segmented trend prediction results are supplemented with the time points of each data segment and the corresponding health index values ​​by linear interpolation, and converted into a continuous change curve to generate a health index continuous change data set;

[0035] Based on the continuous change data set of health indicators, the formula is adopted:

[0036]

[0037] Calculate the curvature k of the target discrete point and obtain the curvature value of the discrete point;

[0038] Where y1 represents the slope of the fitting curve at the target position, y2 represents the acceleration of the fitting curve at the target position, (1+(y1′) 2 ) 3 / 2 Used to normalize curvature values.

[0039] As a further solution of the present invention, the step of obtaining the key trend node identification result is specifically:

[0040] Comparing the curvature value of the discrete point with a preset curvature threshold point by point, determining whether the curvature at each time point exceeds the threshold through comparison, and recording the data to generate a curvature threshold comparison result;

[0041] Based on the curvature threshold comparison result, the time point exceeding the curvature threshold and the corresponding curvature value data are extracted, and the key trend node identification result is established by marking the target time point as a key trend node and recording the marking information.

[0042] As a further solution of the present invention, the step of obtaining the reference information of ophthalmic care is specifically as follows:

[0043] Based on the key trend node identification result, all key trend nodes in the fitted curve are collected, including the time points and curvature values ​​corresponding to the nodes, and the classification results of the key trend nodes are generated by sorting the curvature values ​​and dividing them into multiple node types according to grouping rules;

[0044] Based on the classification results of the key trend nodes, the time points and curvature values ​​of each category of nodes are extracted, the distribution density of the nodes on the time axis and the time span of the density area are analyzed, the changing trend of the curve and the node-dense sections are marked, and reference information for ophthalmic care is obtained.

[0045] Compared with the prior art, the advantages and positive effects of the present invention are:

[0046] In the present invention, by analyzing the difference between the health indicator data before and after nursing, the multidimensional health data can be effectively normalized and its change characteristics can be quantified, so as to accurately present the specific effects brought by nursing intervention. The quantification of such change characteristics not only makes the dynamic changes of health indicators more comparable, but also significantly improves the accuracy of data processing. In terms of dynamic trend prediction, by extracting the change trend of health indicators in time series and dividing the trend change points into multiple stages, the analysis of time series data is more refined and targeted, and the transition nodes and trend directions of health status can be accurately captured. The segmented modeling of trend changes enables data features to be accurately fitted at each stage, which helps to more reliably predict the change trajectory and development characteristics of future health data, and provide a scientific basis for ophthalmic care. By extracting curvature features, the discrete points in the data change curve are accurately analyzed, and the key trend nodes of health data can be effectively identified. This analysis of the marking and distribution rules of trend nodes makes the nursing reference information more comprehensive and facilitates the formulation of more targeted nursing decisions. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Figure 1 is a system flow chart of the present invention;

[0048] Figure 2 A flow chart of obtaining the health indicator transfer characteristic analysis results of the present invention;

[0049] Figure 3 A flow chart for extracting trend change points of the present invention;

[0050] Figure 4 A flow chart for obtaining a local fitting model for the present invention;

[0051] Figure 5 A flow chart for obtaining segmented trend prediction results of the present invention;

[0052] Figure 6 A flow chart of calculating the curvature value of discrete points in the fitted variation curve according to the present invention;

[0053] Figure 7 A flowchart of obtaining key trend node identification results for the present invention;

[0054] Figure 8 Flowchart for obtaining reference information for ophthalmic care for the present invention. DETAILED DESCRIPTION

[0055] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0056] In the description of the present invention, it should be understood that the terms "length", "width", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside" and the like indicate positions or positional relationships based on the positions or positional relationships shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention. In addition, in the description of the present invention, "multiple" means two or more, unless otherwise clearly and specifically defined.

[0057] See also Figure 1 , an intelligent auxiliary system for ophthalmic care includes:

[0058] The health index distribution difference analysis module obtains the health index data of ophthalmic patients before and after care, including intraocular pressure, retinal thickness and corneal morphology parameters, and performs normalization processing, compares the differences between the health index data of ophthalmic patients before and after care, quantifies the change characteristics of the health index data according to the difference value, and generates the health index transfer characteristic analysis results;

[0059] The dynamic trend prediction and segmented modeling module determines the changing trend of the health indicator data of ophthalmic patients from before to after care in the time series according to the analysis results of the health indicator transfer characteristics, extracts the trend change points from the changing trend, and the trend change points are the trend conversion moments of the health indicators. The time series is divided into multiple stages according to the trend change points, and a local fitting model of the health indicator data corresponding to each stage is constructed. The changing direction of the health indicator data is analyzed according to the data in the local fitting model, and the future characteristic changes of the health data in each stage are predicted according to the changing direction, and the segmented trend prediction results are generated;

[0060] The health data curvature feature extraction module converts the segmented trend prediction results into a continuous change curve, fits the change curve of each segment of data in the continuous change curve, calculates the curvature value of the discrete point in the fitted change curve, compares it with the preset curvature threshold, marks the curvature value exceeding the curvature threshold as a key trend node, and generates a key trend node identification result;

[0061] The health index dynamic update module classifies the key trend nodes in the key trend node identification results, analyzes the distribution law of the key trend nodes in the fitting change curve according to the classification results, and generates reference information for ophthalmic care;

[0062] The results of the health indicator transfer feature analysis include the difference values ​​of the health indicators before and after care, the difference trend change characteristics, and the transfer direction of the health indicators. The segmented trend prediction results include the time position of the trend change point, the local fitting characteristics of the segmented health indicators, and the future change trend information of the staged health indicators. The key trend node identification results include the location distribution of key trend nodes, the curvature value corresponding to the node, and the categories of rapidly changing nodes and stably changing nodes. The reference information of ophthalmic care includes the dynamic change characteristics of key trend nodes, the staged change direction of health indicators, and the indicator priority recommendations for nursing adjustments.

[0063] See also Figure 2 ,The specific steps for obtaining the health indicator transfer feature analysis results are:

[0064] Based on the patient health index data before and after care, intraocular pressure, retinal thickness and corneal morphology parameters are obtained, and the data of multiple dimensions are normalized to obtain an integrated health index data table;

[0065] The health index data of patients before and after care were collected from the ophthalmic care data acquisition system, including intraocular pressure, retinal thickness and corneal morphology parameters. These data were grouped by patients and input into the data processing module. Each group of data was normalized by maximum and minimum values. The normalization steps were as follows: first, the maximum and minimum values ​​of each indicator were filtered out from the original data set, and then normalization was applied to each data item in turn, all data were converted to the [0, 1] interval, and a normalized health index table was generated through the data recording module.

[0066] Based on the integrated health indicator data table, the formula is used:

[0067] D i =X i,after -X i,before ;

[0068] Calculate the difference value D of the i-th health indicator i , get the indicator difference data;

[0069] Among them, X i,after is the value of the ith health indicator after care, collected by ophthalmic examination equipment, such as using a non-contact tonometer to obtain intraocular pressure, a retinal imager to collect retinal thickness, and a corneal topographer to obtain corneal morphology parameters, X i,before It is the value of the ith health indicator before care, and the acquisition method is the same as that after care, through the collection of examination data before care. i represents the category of health indicators, including intraocular pressure, retinal thickness and corneal morphology parameters.

[0070] Assume that the patient's health indicator data before care is measured by the equipment and the following results are obtained: intraocular pressure is 21 mmHg, retinal thickness is 250 μm, and corneal morphology parameter is 0.8. After care, the same equipment is used to measure the following results: intraocular pressure is 18 mmHg, retinal thickness is 240 μm, and corneal morphology parameter is 0.85.

[0071] For intraocular pressure: D 眼压 =X 眼压,after -X 眼压,before =18-21=-3mmHg;

[0072] For retinal thickness: D 视网膜厚度 =X 视网膜厚度,after -X 视网膜厚度,before =240-250=-10μm;

[0073] For corneal morphological parameters: D 角膜形态参数 =X 角膜形态参数,after -X 角膜形态参数,before =0.85-0.8=0.05;

[0074] The results showed that intraocular pressure decreased by 3 mmHg, retinal thickness decreased by 10 μm, and corneal morphological parameters increased by 0.05.

[0075] Based on the indicator difference data, the maximum value, minimum value and concentrated distribution range of each type of indicator are counted respectively, and the positive and negative change range of each health indicator is clarified by combining the upper and lower limits of the difference value. The health indicator transfer characteristic analysis results are generated based on the statistical information;

[0076] First, the difference values ​​of the patient group before and after care were classified and sorted according to health indicators, such as intraocular pressure, retinal thickness and corneal morphological parameters. All patient difference data for each health indicator were summarized separately, and the average change range and the upper and lower limits of the difference range for each type of difference value were counted. The positive and negative change ranges were determined by the maximum and minimum values ​​of the difference data. For example, the change range of intraocular pressure after care was -0.3 to -0.1, indicating that the overall intraocular pressure showed a downward trend. Then, according to the distribution of difference data of all patients, the concentrated distribution range of the difference values ​​was statistically analyzed. For example, the difference values ​​of retinal thickness were concentrated in the range of -0.2 to -0.1, while individual values ​​showed larger deviations. These statistical results were converted into change characteristic range diagrams of each health indicator. The change trend of each type of health indicator was presented by drawing data distribution charts. At the same time, the positive and negative intervals were distinguished by colors to intuitively show the positive or negative impact of care on health indicators. Finally, the health indicator transfer characteristic analysis results were generated.

[0077] See also Figure 3 , the steps to extract trend change points are as follows:

[0078] Based on the analysis results of the health index transfer characteristics, the intraocular pressure, retinal thickness and corneal morphology parameter data before and after care were sorted out, and a data series was generated in the order of time points. The change rate and direction of adjacent time points in the time series were analyzed one by one to generate the analysis results of the change trend of the health index time series.

[0079] By arranging the intraocular pressure, retinal thickness and corneal morphological parameter data before and after care, a data series is generated in the order of time points. For example, the intraocular pressure of a patient before care is 21 mmHg, and the measurement results after care are 19 mmHg on the 7th day, 18 mmHg on the 14th day, and 18 mmHg on the 21st day. The intraocular pressure change in the time series is [21, 19, 18, 18]. The change rate of adjacent time points in the sequence is calculated. For example, the change rate from the 1st day to the 7th day is (-2) / 7=-0.29 mmHg / day, and the change rate from the 7th day to the 14th day is (-1) / 7=-0.14 mmHg / day. This method is used to comprehensively judge the downward trend of intraocular pressure. Similarly, the time series data changes of retinal thickness and corneal morphological parameters are analyzed, and their change rate and direction are extracted to generate a complete change trend analysis report of the time series before and after care.

[0080] Based on the analysis results of the time series change trend of health indicators, the formula is used:

[0081]

[0082] Calculate the maximum absolute value T in the health indicator time series to obtain the trend change point;

[0083] Among them, argmax means finding the variable that makes the objective function value reach the maximum value from the function, t is the time point variable, which means the time point of each data collection in the nursing process, and belongs to the element range [t1, t2] of the time series, which is determined by the preset time interval in the experimental design. For example, if data is collected once a day, t is the time point of each day, [t1, t2] is the start and end range of the time series, which means the time point t1 when the nursing starts and the time point t2 when the nursing ends, ΔY is the change in the health index value, which means the numerical difference of the health index between two adjacent time points before and after the nursing (such as the difference in intraocular pressure), which is obtained by the health index data collection equipment before and after the nursing, such as intraocular pressure is measured by a non-contact tonometer, retinal thickness is measured by a retinal imager, and corneal morphology parameters are measured by a corneal topograph, Δt represents the interval between two adjacent time points in the time series, It is the rate of change of the health indicator, which indicates the magnitude of change of the health indicator per unit time in the time series.

[0084] For example, health indicator data in the time series [t1, t2] are obtained from nursing data collection. For example, the intraocular pressure data of a patient is [21, 20, 19, 18, 18] mmHg, corresponding to the 0th, 7th, 14th, 21st, and 28th days respectively.

[0085] Calculate the change ΔY and time interval Δt of adjacent time points:

[0086] Day 0 to Day 7: ΔY = 20-21 = -1 mmHg, Δt = 7 days;

[0087] Day 7 to Day 14: ΔY = 19-20 = -1 mmHg, Δt = 7 days;

[0088] Day 14 to Day 21: ΔY = 18-19 = -1 mmHg, Δt = 7 days;

[0089] Day 21 to Day 28: ΔY=18-18=0 mmHg, Δt=7 days.

[0090] Calculate the rate of change of health indicators

[0091] Day 0 to Day 7:

[0092] Day 7 to Day 14:

[0093] Day 14 to Day 21:

[0094] Day 21 to Day 28:

[0095] Calculate the absolute value of the rate of change Find the maximum value and the corresponding time point:

[0096] |-0.14| = 0.14 mmHg / day (this rate is the same from day 0 to day 21);

[0097] |0| = 0 mmHg / day (rate from day 21 to day 28).

[0098] The results showed that the maximum change rate |ΔY / Δt|=0.14 mmHg / day occurred from day 0 to day 21, and the trend change point could be selected on day 7 or day 14. For example, the downward trend of intraocular pressure reached its peak on the 14th day of care and then gradually stabilized.

[0099] See also Figure 4 , the steps to obtain the local fitting model are as follows:

[0100] Divide the time series into multiple stages according to the trend change points, organize and group the health indicator data in each stage, and generate a stage-by-stage health indicator data table;

[0101] First, extract trend change points from the time series of health indicators. For example, determine the time point with the largest change rate by calculating the absolute value of the change rate of health indicators, and use it as the trend change point. Then divide the time series into multiple continuous stages according to the position of the change point. For each stage, reorganize the health indicator data within the stage according to the order of time points, and further group the data. For example, the intraocular pressure of a patient changes to [-1, -2] and [-1, -1] in the two stages from day 0 to day 7 and from day 7 to day 14, respectively. Generate a health indicator time series table for each stage. The arrangement of segmented data facilitates the subsequent construction of local fitting models and analysis of characteristic trends. Usually, when arranging segmented time series data, you can group and organize the data through Microsoft Excel, MATLAB or the pandas library in Python. For example, filter time point data by stage in Excel, or use the groupby function of pandas in Python to segment and reconstruct data within the stage. Generate a staged health indicator data table after data grouping, which is convenient for subsequent analysis and modeling.

[0102] Based on the staged health index data table, by analyzing the changing rules of the data in each stage and the changing trends of the health indicators in the corresponding stage, a local fitting model is obtained through data fitting;

[0103] Construct a local fitting model for the health indicator data corresponding to each stage. Select the health indicator data as the input variable in the divided time series stage. Analyze the change law of the data in each stage stage by stage, and use the linear fitting method to model the data of each stage. For example, for a certain stage, the change sequence of intraocular pressure is [21, 20, 19], corresponding to the 0th day, the 7th day, and the 14th day, respectively. First, calculate the change rate between each two time points as [-1, -1]. Then substitute the time point and the change value into the fitting formula Y = aX + b, and calculate the fitting parameters a and b by the least squares method. Among them, Y is the target variable (dependent variable), which represents the value of the health indicator, such as intraocular pressure, retinal thickness or corneal morphology parameters, at each time point. The actual observed value. X is the independent variable (input variable), which represents the time point data, such as the 0th day, the 7th day, the 14th day, etc., corresponding to each time point in the stage time series. a is the slope of the fitting line, which represents the rate of change of the health indicator over time in the stage, reflecting the trend of the health indicator in the time series (such as rising, falling or unchanged). b is the intercept of the fitted line, which indicates the initial value of the health index when the time point is zero, reflecting the baseline level of the health index. The intercept is also automatically generated by the fitting model through the least squares method, reflecting the basic health status within the stage. Finally, the linear fitting model of each stage is obtained for subsequent prediction and analysis of the stage data. The construction of the model can be completed through the CurveFitting tool of MATLAB. The specific process is to first import the stage health indicator data, select the fitting model type (such as linear model or polynomial model), and then run the fitting algorithm to output the model parameters. Finally, the fitting accuracy is verified and the model results are saved. Through the automatic calculation function of MATLAB or Python tools, local fitting models of multiple stages can be quickly constructed and model parameters can be obtained, and finally the data change trend of each stage can be quantitatively analyzed.

[0104] See also Figure 5 , the specific steps for obtaining the segmented trend prediction results are:

[0105] According to the data in the local fitting model, by extracting the key parameters in the fitting model at each stage, combining the fitting results to analyze the change direction of the health indicators, and sorting the change trend by segment, the analysis results of the current change direction of the health indicators are generated;

[0106] First, the local fitting model results of each stage are extracted. For example, for the intraocular pressure data of a certain stage, the slope parameter a is calculated through the linear fitting model Y=aX+b, where the slope a directly reflects the change direction of the health index in this stage. When a>0, the health index shows an increasing trend, and when a<0, the health index shows a decreasing trend. At the same time, the credibility of the model is judged in combination with the fitting error of the model (such as the mean square error MSE). During the specific execution, the slope and error parameters are first obtained through the fitting results of the MATLAB tool, or the corresponding slope value is extracted through the fitting results of the scipy library in Python. For example, the fitting result of the intraocular pressure data of the first stage is a=-0.15, and the MSE is 0.02, indicating that the intraocular pressure in this stage is on a downward trend. The fitting results of all stages are combined to generate an analysis report on the change direction of health indicators.

[0107] Based on the analysis results of the current health indicator change direction, the formula is used:

[0108] Y pred,i′+1 =Y current,i′ +a i′ ·Δt′+b i′ ·(Δt′) 2 ;

[0109] Calculate the health index value Y at the future i′+1 time point pred,i′+1 , get the segmented trend prediction results;

[0110] Among them, Y current,i′ It is the health index value at the current time point, which is directly obtained through health data collection equipment. For example, intraocular pressure is measured by non-contact tonometer, retinal thickness is obtained by retinal imager, and corneal morphological parameters are obtained by corneal topograph. i′ It is the slope of linear change, indicating the rate and direction of change of health indicators in the current time period, reflecting the linear change trend of health indicators. For example, a positive value indicates an increase, and a negative value indicates a decrease. The coefficient of the linear term is calculated by the local fitting model. For example, the scipy library in Python is used to fit the data in the stage (for example, Y = aX + b) and directly extract a i′ , b i′ It is the quadratic variation coefficient, which indicates the acceleration or deceleration of the change of the health indicator, and reflects the nonlinear change trend of the health indicator, such as the degree to which the downward trend of the health indicator slows down or accelerates. The coefficient of the quadratic term is obtained by fitting the time series data with a quadratic polynomial. For example, the numpy.polyfit function in Python is used to fit the data within the stage and extract the quadratic term coefficient. Δt′ is the time interval, which indicates the length of time from the current time point to the predicted time point.

[0111] If we want to predict the intraocular pressure of a patient on the 14th day after care (i.e. the i′+1th time point), assuming that the intraocular pressure on the 7th day is Y current,i′ =19mmHg; linear change slope: a i′ =-0.15; quadratic variation coefficient: b i′ =0.005; time interval: Δt′=7.

[0112] Substitute the formula into the calculation:

[0113] Y pred,i′+1 =19+(-0.15)·7+0.005·7 2 ;

[0114] Calculate the linear change part:

[0115] (-0.15)·7=-1.05;

[0116] Calculate the quadratic change part:

[0117] 0.005·7 2 =0.005·49=0.245;

[0118] Comprehensive calculation of predicted value:

[0119] Y pred,i′+1 =19-1.05+0.245=18.195mmHg;

[0120] The results show that on the 14th day after care, the patient's intraocular pressure is expected to be 18.195 mmHg. By combining the dual trends of linear change and quadratic change, the formula can more accurately describe the changing trajectory of health indicators at future time points. It is especially suitable for scenarios with nonlinear trend changes. For example, intraocular pressure may drop rapidly in the early stage of care and tend to be flat in the later stage. The introduction of the formula effectively solves the error accumulation problem that may be caused by simple linear prediction, and provides a more scientific calculation method for the segmented trend prediction of health indicators.

[0121] See also Figure 6 , the specific steps for obtaining the curvature value of discrete points in the fitting curve are as follows:

[0122] The segmented trend prediction results are used to complete the time points of each data segment and the corresponding health indicator values ​​by linear interpolation, and converted into a continuous change curve to generate a health indicator continuous change data set;

[0123] First, the segmented data is imported into the data table in chronological order to ensure that the time point of each data segment and the corresponding health indicator value are fully recorded. Then, the linear interpolation method is used to interpolate the data between the time points. By extracting the starting point and end point coordinates of each time period, these coordinates are input and linearly interpolated. The specific operation of linear interpolation is: select the starting point coordinates and the end point coordinates, and insert the intermediate values ​​in segments according to the proportional relationship within the time interval, and generate a complete continuously changing data set. Subsequently, all segmented fitting data are input into a data visualization tool (such as MATLAB or Python's matplotlib library) for curve drawing to ensure that each data segment is smoothly connected on the continuously changing curve. Finally, a continuous curve that can reflect the changes of health indicators over time is generated for subsequent change feature analysis.

[0124] Based on the continuous change data set of health indicators, the formula is used:

[0125]

[0126] Calculate the curvature k of the target discrete point and obtain the curvature value of the discrete point;

[0127] Among them, y1 represents the slope of the fitting curve at the target position, reflecting the rate of change of the curve, y2 represents the acceleration of the fitting curve at the target position, reflecting the change of the rate of change of the curve, (1+(y1′) 2 ) 3 / 2 It is used to normalize the curvature value to eliminate the magnitude difference of curvature between different discrete points. y1 and y2 are obtained by fitting the curve equation. The expression can be obtained by taking the first and second order derivatives of the fitting equation. The parameters of the fitting equation can be calculated by the least squares method.

[0128] Assume that the curve equation obtained by fitting the collected health index data is:

[0129] y=-0.01x 3 +0.3x 2 -1.5x+10;

[0130] Find the first-order derivative:

[0131] y1′=3(-0.01)x 2 +2(0.3)x-1.5=-0.03x 2 +0.6x-1.5;

[0132] Find the second-order derivative:

[0133] y2″=6(-0.01)x+2(0.3)=-0.06x+0.6;

[0134] Calculate at x=2:

[0135] y1′=-0.03(2 2 )+0.6(2)-1.5=-0.42

[0136] y2″=-0.06(2)+0.6=0.48;

[0137] Calculate the curvature k:

[0138]

[0139] The result shows that the curvature value at x=2 is 0.3755.

[0140] See also Figure 7 ,The specific steps for obtaining the key trend node identification results are:

[0141] Compare the curvature value of the discrete point with the preset curvature threshold point by point, determine whether the curvature at each time point exceeds the threshold through comparison, record the data, and generate the curvature threshold comparison result;

[0142] Based on the specific scenario where the calculated curvature value k=0.3755 is at time point x=2, the preset curvature threshold k is selected. 阈值 =0.35 for comparison, and compare the calculated value with the threshold point by point. When the curvature value is greater than the threshold, it is judged as a point where the curvature changes dramatically. At this time, k=0.3755>0.35, indicating that there is a significant curvature change in the change curve at time point x=2. According to the comparison result, this point is marked as a candidate key trend point that needs further analysis, and the curvature threshold comparison result is obtained, which clarifies that the curvature value at this time point has exceeded the preset threshold range.

[0143] Based on the curvature threshold comparison result, the time points exceeding the curvature threshold and the corresponding curvature value data are extracted, and the key trend node identification results are established by marking the target time points as key trend nodes and recording the marking information;

[0144] The time points corresponding to the curvature values ​​exceeding the curvature threshold are marked as key trend nodes. Combined with the curvature value k=0.3755 of the compared time point x=2, a marked data set is established to record the curvature changes at the time point, and the time x=2 and the curvature value k=0.3755 are written into the key trend node list. Through this marking process, a clear key trend node identification result is formed, and a complete node list is generated, recording all the marked points exceeding the threshold and their curvature values.

[0145] See also Figure 8 ,The specific steps for obtaining reference information for ophthalmic care are:

[0146] Based on the key trend node identification results, all key trend nodes in the fitted curve are collected, including the time points and curvature values ​​corresponding to the nodes. The curvature values ​​are sorted and divided into multiple node types according to the grouping rules to generate the classification results of the key trend nodes.

[0147] First, all key trend nodes in the fitted curve are collected, including the time points and curvature values ​​corresponding to the nodes. The curvature values ​​are sorted from high to low and the nodes are classified according to the set grouping rules (for example, divided into high curvature intervals, medium curvature intervals, and low curvature intervals). The specific operations include: reading the curvature value array of the node, calculating the quantile of the array, setting the grouping threshold to the 75% quantile and 25% quantile of the curvature value, dividing the nodes above the 75% quantile into high curvature nodes, and those below the 25% quantile into low curvature nodes, and the rest into medium curvature nodes. The classification results of the key trend nodes are obtained through the grouping operation, and the classification results are reorganized according to the time points for further distribution law analysis.

[0148] Based on the classification results of key trend nodes, the time points and curvature values ​​of each category of nodes are extracted, the distribution density of nodes on the time axis and the time span of the density area are analyzed, the change trend of the curve and the node-dense sections are marked, and reference information for ophthalmic care is obtained;

[0149] According to the classification results, the distribution law of nodes in each category is analyzed. Combined with the specific parameters of intraocular pressure, retinal thickness and corneal morphology, the time points of intraocular pressure changes and the corresponding curvature values ​​are first extracted for high curvature nodes to generate a density distribution map of intraocular pressure changes over time. The kernel density estimation method is used to analyze the intraocular pressure changes and mark the time intervals where densely distributed high intraocular pressure nodes are located. For example, the density peak is from 6 am to 10 am, corresponding to the high-risk time period when the intraocular pressure is above 30 mmHg. Then the same density analysis is performed on the retinal thickness changes, and the time points when the retinal thickness changes sharply are extracted and marked. For example, the time point when the retinal thickness changes by more than 50 μm is taken as the key trend node. Combined with the corneal morphology parameters, the time point when the corneal curvature changes sharply is extracted. For example, the rapid change segment where the central corneal thickness decreases from 520 μm to 480 μm is found in the early stage. The node distribution results of the three types of parameters are integrated, and the interaction and trend association on the time axis are analyzed. Finally, a distribution law report of key parameters is generated for intelligent auxiliary decision-making in ophthalmic care.

[0150] The above are only preferred embodiments of the present invention and are not intended to limit the present invention in other forms. Any technician familiar with the profession may use the technical contents disclosed above to change or modify them into equivalent embodiments with equivalent changes and apply them to other fields. However, any simple modification, equivalent change and modification made to the above embodiments based on the technical essence of the present invention without departing from the technical solution of the present invention still falls within the protection scope of the technical solution of the present invention.

Claims

1. An intelligent auxiliary system for ophthalmic care, characterized in that: The system comprises: The health indicator distribution difference analysis module obtains the health indicator data of the ophthalmic patients before and after nursing care, compares the differences between the health indicator data of the ophthalmic patients before and after nursing care, and generates the health indicator transfer feature analysis results; The dynamic trend prediction and segmented modeling module determines the change trend of the health indicator data of ophthalmic patients from before to after care in the time series according to the health indicator transfer feature analysis results, extracts the trend change points therefrom, builds a local fitting model according to the trend change points of each stage, analyzes the change direction of the health indicator data, predicts the future feature changes of the health data in each stage according to the change direction, and generates segmented trend prediction results; The health data curvature feature extraction module converts the segmented trend prediction results into a continuous change curve and performs fitting, calculates the curvature values ​​of discrete points in the fitted change curve, compares them with a preset curvature threshold, marks the curvature values ​​exceeding the curvature threshold as key trend nodes, and generates key trend node identification results; The health index dynamic update module classifies the key trend nodes in the key trend node identification result, analyzes the distribution law of the key trend nodes in the fitted change curve according to the classification result, and generates reference information for ophthalmic care.

2. The intelligent ophthalmic care assistance system according to claim 1, characterized in that: The steps for obtaining the health indicator transfer characteristic analysis results are specifically as follows: Based on the patient health index data before and after care, intraocular pressure, retinal thickness and corneal morphology parameters are obtained, and the data of multiple dimensions are normalized to obtain an integrated health index data table; Based on the integrated health indicator data table, the formula is used: D i =X i,after -X i,before ; Calculate the difference value D of the i-th health indicator i , get the indicator difference data; Among them, X i,after is the value of the i-th health indicator after care, X i,before is the value of the ith health indicator before care; Based on the indicator difference data, the maximum value, minimum value and concentrated distribution range of each type of indicator are counted respectively, and the positive and negative change range of the health indicator is determined in combination with the upper and lower limits of the difference value, and the health indicator transfer characteristic analysis results are generated according to the statistical information.

3. The intelligent ophthalmic care assistance system according to claim 2, characterized in that: The steps of extracting trend change points are specifically as follows: Based on the analysis results of the health indicator transfer characteristics, by sorting the intraocular pressure, retinal thickness and corneal morphology parameter data before and after care, a data sequence is generated in the order of time points, and the change rate and direction of adjacent time points in the time series are analyzed one by one to generate the health indicator time series change trend analysis results; Based on the analysis results of the time series change trend of the health indicators, the formula is adopted: Calculate the maximum absolute value T in the health indicator time series to obtain the trend change point; Among them, argmax represents the variable at which the objective function value reaches the maximum value, t is the time point variable, [t1, t2 represents the time point t1 when the care starts and the time point t2 when the care ends, ΔY is the change in the health indicator value, and Δt represents the interval between two adjacent time points in the time series.

4. The intelligent ophthalmic care assistance system according to claim 3, characterized in that: The steps of obtaining the local fitting model are specifically as follows: Dividing the time series into multiple stages according to the trend change points, sorting and grouping the health indicator data in each stage, and generating a staged health indicator data table; Based on the staged health index data table, by analyzing the changing rules of the data in each stage and the changing trends of the health indexes in the corresponding stage, a local fitting model is obtained by data fitting.

5. The intelligent ophthalmic care assistance system according to claim 4, characterized in that: The steps for obtaining the segmented trend prediction results are specifically as follows: According to the data in the local fitting model, by extracting the key parameters in the fitting model at each stage, analyzing the change direction of the health indicator in combination with the fitting results, and arranging the change trend in sections, an analysis result of the change direction of the current health indicator is generated; Based on the analysis results of the change direction of the current health indicators, the formula is adopted: Y pred,i′+1 =Y current,i′ +a i′ ·Δt′+b i′ ·(Δt′) 2 ; Calculate the health index value Y at the future i′+1 time point pred,i′+1 , get the segmented trend prediction results; Among them, Y current,i′ is the health indicator value at the current time point, a i′ is the linear change slope, b i′ is the quadratic variation coefficient, and Δt′ represents the time length from the current time point to the predicted time point.

6. The intelligent ophthalmic care assistance system according to claim 5, characterized in that: The step of obtaining the curvature value of the discrete points in the variation curve after calculation and fitting is specifically as follows: The segmented trend prediction results are supplemented with the time points of each data segment and the corresponding health index values ​​by linear interpolation, and converted into a continuous change curve to generate a health index continuous change data set; Based on the continuous change data set of health indicators, the formula is adopted: Calculate the curvature k of the target discrete point and obtain the curvature value of the discrete point; Where y1 represents the slope of the fitting curve at the target position, y2 represents the acceleration of the fitting curve at the target position, (1+(y1′) 2 ) 3 / 2 Used to normalize curvature values.

7. The intelligent ophthalmic care assistance system according to claim 6, characterized in that: The steps for obtaining the key trend node identification result are specifically as follows: Comparing the curvature value of the discrete point with a preset curvature threshold point by point, determining whether the curvature at each time point exceeds the threshold through comparison, and recording the data to generate a curvature threshold comparison result; Based on the curvature threshold comparison result, the time point exceeding the curvature threshold and the corresponding curvature value data are extracted, and the key trend node identification result is established by marking the target time point as a key trend node and recording the marking information.

8. The intelligent auxiliary system for ophthalmic care according to claim 7, characterized in that: The steps for obtaining the reference information of ophthalmic care are specifically as follows: Based on the key trend node identification result, all key trend nodes in the fitted curve are collected, including the time points and curvature values ​​corresponding to the nodes, and the classification results of the key trend nodes are generated by sorting the curvature values ​​and dividing them into multiple node types according to grouping rules; Based on the classification results of the key trend nodes, the time points and curvature values ​​of each category of nodes are extracted, the distribution density of the nodes on the time axis and the time span of the density area are analyzed, the changing trend of the curve and the node-dense sections are marked, and reference information for ophthalmic care is obtained.

Images