An individual health assessment method based on geometric deep learning

Abstract

The application discloses a kind of individual health assessment methods based on geometric deep learning, which comprises the following steps: obtaining the health parameter data of multiple dimensions of the object to be evaluated;Based on graph structure algorithm and graph convolutional neural network, the graph feature parameters corresponding to the health parameter data of multiple dimensions are extracted;Based on geometric mean algorithm, the health scale parameters corresponding to the health parameter data of multiple dimensions are calculated;According to the graph feature parameters and the health scale parameters, based on dynamic time series analysis algorithm, the health assessment result of the object to be evaluated is determined.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of data processing, and in particular to a health assessment method for individuals based on geometric deep learning. BACKGROUND

[0002] With the rapid development of artificial intelligence, big data technology and medical monitoring equipment, personalized health management has gradually become a research hotspot in the medical field. By analyzing the health parameter data of individuals (such as blood pressure, heart rate, blood sugar, etc.), important support can be provided for disease prediction, health management and personalized medicine. At present, more and more health monitoring devices (such as smart bracelets, medical-grade wearable devices) can collect multi-dimensional health parameters and record them in the form of time series. However, how to extract key information from complex, multi-dimensional health data, accurately classify and dynamically predict the health status of individuals, is still a problem to be solved. Traditional health assessment methods are mostly based on mean analysis, linear regression or single classification model. These methods are not capable of capturing the complex relationships of non-linear, multi-parameter associations. At the same time, many methods lack dynamic analysis capability for time series data, resulting in limited performance in health trend prediction and early disease intervention. Especially in chronic disease management (such as Alzheimer's disease, diabetes, etc.), the health status of individuals depends not only on the value of a single parameter, but also on the comprehensive consideration of the complex relationship between multi-dimensional parameters and the trend over time. This complexity poses a great challenge to existing technologies. Therefore, the existing technology has defects and needs to be solved. SUMMARY

[0003] The technical problem to be solved by the present application is to provide a health assessment method and system for individuals based on geometric deep learning, which can accurately capture the complex correlation and time series changes of health data by combining multi-dimensional data, graph convolution and time series analysis, and improve the accuracy and reliability of health assessment.

[0004] To solve the above technical problems, the present application discloses a health assessment method for individuals based on geometric deep learning, which comprises:

[0005] Obtaining multi-dimensional health parameter data of a to-be-evaluated object;

[0006] Extracting graph feature parameters corresponding to the multi-dimensional health parameter data based on a graph structure algorithm and a graph convolutional neural network;

[0007] Calculating health scale parameters corresponding to the multi-dimensional health parameter data based on a geometric mean algorithm;

[0008] Determining a health assessment result of the to-be-evaluated object based on a dynamic time series analysis algorithm according to the graph feature parameters and the health scale parameters.

[0009] As an optional embodiment, in the first aspect of the present invention, the data type of the health parameter data includes at least one of blood glucose, blood pressure, heart rate, body temperature, activity level, number of steps, exercise time, energy consumption, blood oxygen saturation, heart rate variability, electrocardiogram, MRI image features, cognitive test scores and brain metabolism data.

[0010] As an optional embodiment, in the first aspect of the present invention, the extraction of graph feature parameters corresponding to the health parameter data of the multiple dimensions based on the graph structure algorithm and the graph convolutional neural network includes:

[0011] Calculating graph structure information corresponding to the health parameter data of the multiple dimensions based on a graph structure algorithm;

[0012] The health parameter data of multiple dimensions and the graph structure information are input into a trained graph convolutional neural network to obtain output graph feature parameters.

[0013] As an optional embodiment, in the first aspect of the present invention, the calculating of the graph structure information corresponding to the health parameter data of the multiple dimensions based on the graph structure algorithm includes:

[0014] Determining each health parameter data in the multiple dimensions as a parameter node in a preset graph structure;

[0015] For any two parameter nodes, determining the type association between the data types of the health parameter data corresponding to the two parameter nodes; the type association is determined by a preset association rule between different data types;

[0016] Calculating the data similarity of the health parameter data respectively corresponding to the two parameter nodes;

[0017] Determine a historical data sequence of a plurality of historical time points corresponding to the health parameter data respectively corresponding to the two parameter nodes;

[0018] Calculating the change rate similarity between the data change rates of the historical data sequences respectively corresponding to the two parameter nodes;

[0019] Calculating a weighted average of the type association degree, the data similarity, and the change rate similarity to obtain an inter-node relationship value between the two parameter nodes;

[0020] All the parameter nodes and the corresponding inter-node relationship values ​​are output as graph structure information corresponding to the health parameter data of the multiple dimensions.

[0021] As an optional implementation, in the first aspect of the present application, the geometric mean algorithm is used to calculate the health scale parameter corresponding to the plurality of dimensional health parameter data, including:

[0022] For each data type, the geometric mean of all health parameter data corresponding to the data type is calculated;

[0023] According to the average value of all node relationship values corresponding to the parameter nodes corresponding to all health parameter data of the data type, the relationship weight corresponding to the data type is obtained;

[0024] The type scale parameter corresponding to the data type is obtained by calculating the power with the geometric mean as the base and the relationship weight as the exponent;

[0025] The average value of all type scale parameters corresponding to all data types is calculated to obtain the health scale parameter.

[0026] As an optional implementation, in the first aspect of the present application, the dynamic time series analysis algorithm is used to determine the health evaluation result of the to-be-evaluated object according to the graph feature parameter and the health scale parameter, including:

[0027] The health scale parameters corresponding to a plurality of time points of the to-be-evaluated object are sorted in chronological order to obtain a time parameter sequence;

[0028] The time parameter sequence is input into a trained LSTM neural network model to obtain a predicted future health scale parameter corresponding to the to-be-evaluated object; the LSTM neural network is trained by a training data set including a plurality of training time parameter sequences;

[0029] An evaluation algorithm model corresponding to a current evaluation scenario is determined;

[0030] The health evaluation result of the to-be-evaluated object is determined based on the evaluation algorithm model according to the graph feature parameter, the future health scale parameter, and the time parameter sequence.

[0031] As an optional implementation, in the first aspect of the present application, the current evaluation scenario is a chronic management scenario, a sports health evaluation scenario, an acute disease early warning scenario, or an Alzheimer's disease cognitive evaluation scenario; when the current evaluation scenario is a chronic management scenario, the evaluation algorithm model is used to calculate a stability parameter, an abnormality rate parameter, and a trend slope parameter corresponding to the time parameter sequence, to determine the health evaluation result of the to-be-evaluated object; when the current evaluation scenario is a sports health evaluation scenario, the evaluation algorithm model is used to calculate a difference between the health scale parameter of the to-be-evaluated object in a sports state and the health scale parameter in a resting state, to determine the health evaluation result of the to-be-evaluated object; when the current evaluation scenario is an acute disease early warning scenario, the evaluation algorithm model is used to calculate a change rate and a parameter proportion of the health scale parameter exceeding a parameter threshold in the time parameter sequence, to determine the health evaluation result of the to-be-evaluated object; when the current evaluation scenario is an Alzheimer's disease cognitive evaluation scenario, the evaluation algorithm model is used to calculate a change trend of the average value of the health scale parameter corresponding to a plurality of time periods in the time parameter sequence, to determine the health evaluation result of the to-be-evaluated object.

[0032] As an optional implementation, in the first aspect of the present application, the determination of the health evaluation result of the to-be-evaluated object based on the graph feature parameter, the future health scale parameter, and the time parameter sequence based on the evaluation algorithm model comprises:

[0033] inputting the graph feature parameter and the health scale parameter at the nearest time point into the evaluation algorithm model to obtain a current evaluation result;

[0034] inputting the graph feature parameter and the future health scale parameter into the evaluation algorithm model to obtain a future evaluation result;

[0035] calculating a weighted sum average of the current evaluation result and the future evaluation result to obtain the health evaluation result of the to-be-evaluated object; wherein the weighted calculation weight corresponding to the current evaluation result comprises a first weight and a second weight; the weighted calculation weight corresponding to the future evaluation result comprises a third weight and a fourth weight; the first weight and the third weight are both proportional to the difference between the corresponding evaluation result value and a reference evaluation result value; the second weight is inversely proportional to the time difference between the time point corresponding to the current evaluation result and the current time point; the fourth weight is proportional to the number of parameters in the time parameter sequence corresponding to the future health scale parameter corresponding to the future evaluation result; the weighted calculation weight corresponding to the current evaluation result is greater than the weighted calculation weight corresponding to the future evaluation result.

[0036] The second aspect of the embodiment of the present application discloses a health assessment system for individuals based on geometric deep learning, and the system comprises:

[0037] a data acquisition module configured to acquire health parameter data of multiple dimensions of a to-be-assessed object;

[0038] a graph feature extraction module configured to extract graph feature parameters corresponding to the health parameter data of the multiple dimensions based on a graph structure algorithm and a graph convolutional neural network;

[0039] a geometric calculation module configured to calculate health scale parameters corresponding to the health parameter data of the multiple dimensions based on a geometric mean algorithm;

[0040] a health assessment module configured to determine a health assessment result of the to-be-assessed object based on a dynamic time series analysis algorithm according to the graph feature parameters and the health scale parameters.

[0041] As an optional implementation, in the second aspect of the present application, the data types of the health parameter data include at least one of blood glucose, blood pressure, heart rate, body temperature, activity amount, step count, exercise time, energy consumption, blood oxygen saturation, heart rate variability, electrocardiogram, MRI image features, cognitive test scores, and brain metabolism data.

[0042] As an optional implementation, in the second aspect of the present application, the specific manner in which the graph feature extraction module extracts the graph feature parameters corresponding to the health parameter data of the multiple dimensions based on the graph structure algorithm and the graph convolutional neural network comprises:

[0043] calculating graph structure information corresponding to the health parameter data of the multiple dimensions based on the graph structure algorithm;

[0044] inputting the health parameter data of the multiple dimensions and the graph structure information into a trained graph convolutional neural network to obtain output graph feature parameters.

[0045] As an optional implementation, in the second aspect of the present application, the specific manner in which the graph feature extraction module calculates the graph structure information corresponding to the health parameter data of the multiple dimensions based on the graph structure algorithm comprises:

[0046] determining each health parameter data in the health parameter data of the multiple dimensions as a parameter node in a preset graph structure;

[0047] determining a type correlation degree between data types of the health parameter data corresponding to any two parameter nodes; the type correlation degree is determined by a preset correlation degree rule between different data types;

[0048] calculate a data similarity of the health parameter data corresponding to the two parameter nodes respectively;

[0049] determine a historical data sequence of a plurality of historical time points corresponding to the health parameter data corresponding to the two parameter nodes respectively;

[0050] calculate a change rate similarity between the change rates of the historical data sequences corresponding to the two parameter nodes respectively;

[0051] calculate a weighted sum average of the type correlation degree, the data similarity and the change rate similarity to obtain a node relationship value between the two parameter nodes;

[0052] output all the parameter nodes and the corresponding node relationship values as the graph structure information corresponding to the plurality of dimensional health parameter data.

[0053] As an optional implementation, in the second aspect of the present application, the specific manner in which the geometry calculation module calculates the health scale parameter corresponding to the plurality of dimensional health parameter data based on the geometric mean algorithm comprises:

[0054] for each data type, calculate the geometric mean of all the health parameter data corresponding to the data type;

[0055] obtain the relationship weight corresponding to the data type according to the average of all the node relationship values corresponding to the parameter nodes between the parameter nodes corresponding to all the health parameter data corresponding to the data type;

[0056] calculate the power of the relationship weight with the geometric mean as the base number to obtain the type scale parameter corresponding to the data type;

[0057] calculate the average of the type scale parameters corresponding to all the data types to obtain the health scale parameter.

[0058] As an optional implementation, in the second aspect of the present application, the specific manner in which the health evaluation module determines the health evaluation result of the to-be-evaluated object based on the dynamic time series analysis algorithm according to the graph feature parameter and the health scale parameter comprises:

[0059] sort the health scale parameters corresponding to a plurality of time points of the to-be-evaluated object according to time from early to late to obtain a time parameter sequence;

[0060] input the time parameter sequence into a trained LSTM neural network model to obtain a predicted future health scale parameter corresponding to the to-be-evaluated object; the LSTM neural network is trained by a training data set comprising a plurality of training time parameter sequences;

[0061] determine an evaluation algorithm model corresponding to the current evaluation scenario;

[0062] determine a health evaluation result of the to-be-evaluated object based on the evaluation algorithm model according to the graph feature parameter, the future health scale parameter and the time parameter sequence.

[0063] As an optional implementation, in the second aspect of the present application, the current evaluation scenario is a chronic management scenario, a sports health evaluation scenario, an acute disease early warning scenario or an Alzheimer's disease cognitive evaluation scenario; when the current evaluation scenario is a chronic management scenario, the evaluation algorithm model is used to calculate a stability parameter, an abnormality rate parameter and a trend slope parameter corresponding to the time parameter sequence to determine the health evaluation result of the to-be-evaluated object; when the current evaluation scenario is a sports health evaluation scenario, the evaluation algorithm model is used to calculate a difference between the health scale parameter of the to-be-evaluated object in a sports state and the health scale parameter in a resting state to determine the health evaluation result of the to-be-evaluated object; when the current evaluation scenario is an acute disease early warning scenario, the evaluation algorithm model is used to calculate a change rate and a parameter proportion of the health scale parameter exceeding a parameter threshold in the time parameter sequence to determine the health evaluation result of the to-be-evaluated object; when the current evaluation scenario is an Alzheimer's disease cognitive evaluation scenario, the evaluation algorithm model is used to calculate a change trend of an average value of the health scale parameter corresponding to a plurality of time periods in the time parameter sequence to determine the health evaluation result of the to-be-evaluated object.

[0064] As an optional implementation, in the second aspect of the present application, the specific manner in which the health evaluation module determines the health evaluation result of the to-be-evaluated object based on the evaluation algorithm model according to the graph feature parameter, the future health scale parameter and the time parameter sequence includes:

[0065] inputting the graph feature parameter and the health scale parameter at the nearest time point into the evaluation algorithm model to obtain a current evaluation result;

[0066] inputting the graph feature parameter and the future health scale parameter into the evaluation algorithm model to obtain a future evaluation result;

[0067] a weighted sum average of the current evaluation result and the future evaluation result is calculated to obtain a health evaluation result of the to-be-evaluated object, wherein the weighted calculation weight corresponding to the current evaluation result comprises a first weight and a second weight, the weighted calculation weight corresponding to the future evaluation result comprises a third weight and a fourth weight, the first weight and the third weight are both proportional to the difference between the corresponding evaluation result value and a reference evaluation result value, the second weight is inversely proportional to the time difference between the time point corresponding to the current evaluation result and the current time point, the fourth weight is proportional to the number of parameters in the time parameter sequence corresponding to the future health scale parameter corresponding to the future evaluation result, and the weighted calculation weight corresponding to the current evaluation result is greater than the weighted calculation weight corresponding to the future evaluation result.

[0068] The third aspect of the present application discloses another individual health evaluation system based on geometric deep learning, comprising:

[0069] a memory storing executable program codes;

[0070] a processor coupled with the memory;

[0071] The processor invokes the executable program codes stored in the memory to execute part or all of the steps of the individual health evaluation method based on geometric deep learning disclosed in the first aspect of the present application.

[0072] The fourth aspect of the present application discloses a computer storage medium storing computer instructions, which are invoked to execute part or all of the steps of the individual health evaluation method based on geometric deep learning disclosed in the first aspect of the present application.

[0073] Compared with the prior art, the embodiments of the present application have the following beneficial effects:

[0074] The present application can obtain multiple-dimensional health parameter data of a to-be-evaluated object, extract corresponding graph feature parameters by using a graph structure algorithm and a graph convolutional neural network, calculate health scale parameters by using a geometric mean algorithm, and comprehensively analyze these data based on a dynamic time series analysis algorithm, so as to accurately capture the complex correlation and time series change of health data by combining multi-dimensional data and graph convolution and time series analysis, and improve the accuracy and reliability of health evaluation. BRIEF DESCRIPTION OF DRAWINGS

[0075] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following will briefly introduce the drawings needed in the embodiment description. Obviously, the drawings in the following description only show some embodiments of the present application, and for those skilled in the art, other drawings can be obtained without creative effort.

[0076] Figure 1 is a flow diagram of a health assessment method for an individual based on geometric deep learning disclosed by an embodiment of the present application.

[0077] Figure 2 is a structural diagram of a health assessment system for an individual based on geometric deep learning disclosed by an embodiment of the present application.

[0078] Figure 3 is a structural diagram of another health assessment system for an individual based on geometric deep learning disclosed by an embodiment of the present application.

[0079] Figure 4 is an example graph of graph structure information disclosed by an embodiment of the present application. DETAILED DESCRIPTION

[0080] In order to make the technical personnel in the art better understand the present application scheme, the following will combine the drawings in the embodiments of the present application, and the technical solutions in the embodiments of the present application will be described clearly and completely. Obviously, the described embodiments are only some of the embodiments of the present application, not all. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present application.

[0081] The terms "first", "second", etc. in the specification and claims of the present application and the above-mentioned drawings are used to distinguish different objects, not to describe a specific order. In addition, the terms "include" and "have" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, device, product or equipment including a series of steps or units is not limited to the listed steps or units, but can optionally include steps or units not listed, or can optionally include other steps or units inherent to the process, method, product or equipment.

[0082] In this paper, the term "embodiment" means that the specific features, structures or characteristics described in conjunction with the embodiment can be included in at least one embodiment of the present application. The phrase appears in the specification at various places does not necessarily refer to the same embodiment, nor is it independent or alternative to other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0083] The application discloses a kind of individual health assessment method and system based on geometric deep learning, can be extracted by obtaining the health parameter data of multiple dimensions of the object to be evaluated, and using graph structure algorithm and graph convolution neural network corresponding graph feature parameter, it is combined with geometric mean algorithm to calculate health scale parameter, again based on dynamic time series analysis algorithm comprehensive analysis these data, to be able to combine multidimensional data and graph convolution and time series analysis to accurately capture the complex correlation and time series variation of health data, improve the accuracy and reliability of health assessment.The following are described in detail.

[0084] Embodiment one

[0085] Please refer to Figure 1 , Figure 1 It is a kind of flow schematic diagram of individual health assessment method based on geometric deep learning disclosed in the embodiment of the application. Among them, Figure 1 The individual health assessment method based on geometric deep learning described can be applied in data processing system / data processing equipment / data processing server (wherein the server includes local processing server or cloud processing server).As Figure 1 As shown, the individual health assessment method based on geometric deep learning can include the following operations:

[0086] 101, the health parameter data of multiple dimensions of the object to be evaluated is acquired.

[0087] Optionally, health parameter data can come from different types of sensors or records, cover physiology, biochemistry, behavior, cognition and multiple modalities.

[0088] 102, based on graph structure algorithm and graph convolution neural network, the graph feature parameter corresponding to the health parameter data of multiple dimensions is extracted.

[0089] Optionally, the above health parameter data can be constructed as graph structure to represent the association between parameters, and the graph data is input into the pre-trained graph neural network model, and the feature representation reflecting the complex relationship of health parameters is output.

[0090] 103, based on geometric mean algorithm, the health scale parameter corresponding to the health parameter data of multiple dimensions is calculated.

[0091] Optionally, the geometric mean value is calculated for each type of health parameter, and the comprehensive health scale parameter reflecting overall health status is obtained by combining the correlation weight between parameters and power operation to obtain the scale value of each type.

[0092] 104, according to graph feature parameter and health scale parameter, based on dynamic time series analysis algorithm, the health assessment result of the object to be evaluated is determined.

[0093] Optionally, the current and historical health scale parameters are analyzed in time series, combined with the graph features, and the trained time series model (such as long short-term memory network LSTM) is used for health state trend prediction, and then the preset evaluation model is used to give the evaluation result of the current and future health status of the individual.

[0094] It can be seen that the above embodiment can obtain multiple dimension health parameter data of the to-be-evaluated object, extract corresponding graph feature parameters by using a graph structure algorithm and a graph convolutional neural network, calculate a health scale parameter by using a geometric mean algorithm, and comprehensively analyze the data based on a dynamic time series analysis algorithm, so that the complex correlation and time series change of the health data can be accurately captured by combining multi-dimensional data and graph convolution and time series analysis, and the accuracy and reliability of the health evaluation are improved.

[0095] As an optional embodiment, in the above step, the data types of the health parameter data include at least one of blood glucose, blood pressure, heart rate, body temperature, activity amount, step count, exercise time, energy consumption, blood oxygen saturation, heart rate variability, electrocardiogram, MRI image features, cognitive test scores, and brain metabolism data.

[0096] In an actual implementation, preprocessing of the data is also implemented to standardize the health parameters and eliminate the dimensional differences between different parameters, specifically including data cleaning, data standardization, and sliding window processing, wherein the data cleaning implements the elimination of abnormal values (such as values exceeding the physiological range), and the abnormal interval is defined by the IQR method, and the data standardization operation standardizes all health parameters, and the formula is: wherein, P is the health parameter data obtained at time t after standardization, is the mean, is the standard deviation. Specifically, the sliding window processing implements segmentation of the time series data: to ensure that the model captures short-term dynamic changes.

[0097] It can be seen that by limiting the content of the health parameter data in the above optional embodiment, the physiological related features of the to-be-evaluated object are fully represented, so as to facilitate subsequent accurate evaluation of the health status, and to assist in accurately capturing the complex correlation and time series change of the health data by combining multi-dimensional data and graph convolution and time series analysis, and improving the accuracy and reliability of the health evaluation.

[0098] As an optional embodiment, in the above step, the graph structure algorithm and the graph convolutional neural network are used to extract the graph feature parameters corresponding to the multiple dimension health parameter data, including:

[0099] Based on the graph structure algorithm, the graph structure information corresponding to the multiple dimension health parameter data is calculated.

[0100] The multiple-dimension health parameter data and the graph structure information are input into the trained graph convolutional neural network to obtain output graph feature parameters.

[0101] It can be seen that, by the above optional embodiments, the graph structure information of the multiple-dimension health parameter data is calculated based on the graph structure algorithm, and the health parameter data and the graph structure information are input into the trained graph convolutional neural network to extract the graph feature parameters, so that the complex relationship between the health parameters can be effectively captured, the precision of feature extraction is improved, and the complex correlation and the time sequence change of the health data are accurately captured by combining the multi-dimension data and the graph convolution and the time sequence analysis, and the accuracy and the reliability of the health evaluation are improved.

[0102] As an optional embodiment, in the above step, the graph structure information corresponding to the multiple-dimension health parameter data is calculated based on the graph structure algorithm, including:

[0103] Each health parameter data in the multiple-dimension health parameter data is determined as a parameter node in a preset graph structure;

[0104] For any two parameter nodes, a type correlation degree between the data types of the health parameter data corresponding to the two parameter nodes is determined; optionally, the type correlation degree is determined by a preset correlation degree rule between different data types;

[0105] The data similarity of the health parameter data corresponding to the two parameter nodes is calculated;

[0106] The historical data sequences of multiple historical time points corresponding to the health parameter data corresponding to the two parameter nodes are determined;

[0107] The change rate similarity between the data change rates of the historical data sequences corresponding to the two parameter nodes is calculated;

[0108] The weighted sum average value of the type correlation degree, the data similarity and the change rate similarity is calculated to obtain a node relationship value between the two parameter nodes;

[0109] All parameter nodes and corresponding node relationship values are output as the graph structure information corresponding to the multiple-dimension health parameter data.

[0110] For example, a weighted adjacency matrix can be constructed to represent the connection weight between each node (each health indicator) in the graph. The graph structure information will be one of the inputs of the graph neural network model.

[0111] In the above embodiment, a form of constructing graph structure information is implemented: taking health parameters as nodes, generating an adjacency matrix based on correlation, and the similarity between i nodes and j nodes, i.e. the adjacency matrix is: , wherein is the value of the standard deviation of the parameter, and an example of graph structure information can be seen from Figure 4 .

[0112] In a specific embodiment, a graph convolutional neural network is used to extract high-order nonlinear relationships of health parameter data: , wherein: : the adjacency matrix plus the identity matrix; : the node degree matrix; : the node feature of the layer; : the weight matrix of the layer.

[0113] As can be seen, through the above optional embodiments, multiple dimensions of health parameter data are mapped to parameter nodes in a preset graph structure, and the relationship values between nodes are calculated in combination with type association rules, data similarity and historical data change rate similarity, to construct graph structure information of health parameter data, which can accurately model the complex associations between different health parameters, and assist in accurately capturing the complex associations and time series changes of health data by combining multi-dimensional data and graph convolution and time series analysis, to improve the accuracy and reliability of health assessment.

[0114] As an optional embodiment, in the above step, the health scale parameters corresponding to the health parameter data of multiple dimensions are calculated based on the geometric mean algorithm, including:

[0115] For each data type, the geometric mean value of all health parameter data corresponding to the data type is calculated; specifically, the geometric mean value can comprehensively reflect the product characteristics of the overall level of the type parameter, and is less affected by extreme values than simple arithmetic mean, and is suitable for measuring the overall condition of physiological indicators;

[0116] According to the average value of all node relationship values corresponding to the parameter nodes corresponding to all health parameter data of the data type, the relationship weight corresponding to the data type is obtained; specifically, the relationship weight represents the strength of the overall association of the type parameter with other parameters;

[0117] The power of the geometric mean value as the base and the relationship weight as the exponent is calculated to obtain the type scale parameter corresponding to the data type; this step is equivalent to weighting and adjusting the geometric mean level of each type parameter according to its degree of association with the overall health network, to obtain an index that can reflect the contribution of the type health;

[0118] Calculate the average value of the type scale parameters corresponding to all data types to obtain the health scale parameter.

[0119] Specifically, the health scale parameter is the overall health score or index after the fusion of multiple dimensional health data, which reflects the comprehensive level of the individual's health status in various aspects.

[0120] As can be seen, through the above optional embodiments, by calculating the geometric mean of the health parameter data of each data type, combining the relationship weight of the corresponding parameter node, using power operation to determine the type scale parameter, and further calculating the average to obtain the health scale parameter, the internal distribution characteristics of the data and the correlation weight of different data types can be considered comprehensively, the representativeness and robustness of the scale parameter in the health evaluation process are improved, and the complex correlation and time sequence changes of health data are accurately captured by combining multi-dimensional data, graph convolution and time sequence analysis, thereby improving the accuracy and reliability of health evaluation.

[0121] As an optional embodiment, in the above steps, the health evaluation result of the to-be-evaluated object is determined based on a dynamic time sequence analysis algorithm according to the graph feature parameter and the health scale parameter, including:

[0122] The health scale parameters corresponding to the multiple time points of the to-be-evaluated object are sorted in chronological order to obtain a time parameter sequence;

[0123] The time parameter sequence is input into a trained LSTM neural network model to obtain a predicted future health scale parameter corresponding to the to-be-evaluated object; optionally, the LSTM neural network is trained by a training data set including multiple training time parameter sequences;

[0124] Determine the evaluation algorithm model corresponding to the current evaluation scenario;

[0125] Determine the health evaluation result of the to-be-evaluated object based on the evaluation algorithm model according to the graph feature parameter, the future health scale parameter and the time parameter sequence.

[0126] As can be seen, through the above optional embodiments, by time sorting the health scale parameter and inputting it into a trained LSTM neural network model to predict the future health scale parameter, and selecting a suitable evaluation algorithm model according to the current evaluation scenario, the health evaluation result is finally determined based on at least one of the graph feature parameter, the future health scale parameter and the time parameter sequence, thereby accurately capturing the complex correlation and time sequence changes of health data by combining multi-dimensional data, graph convolution and time sequence analysis, and improving the accuracy and reliability of health evaluation.

[0127] As an optional embodiment, in the above steps, the current evaluation scenario is a chronic management scenario, a sports health evaluation scenario, a sudden disease warning scenario or an Alzheimer's disease cognitive evaluation scenario; when the current evaluation scenario is a chronic management scenario, the evaluation algorithm model is used to calculate the stability parameters, abnormality rate parameters and trend slope parameters corresponding to the time parameter sequence to determine the health evaluation result of the object to be evaluated; when the current evaluation scenario is a sports health evaluation scenario, the evaluation algorithm model is used to calculate the difference between the health scale parameters of the object to be evaluated in the exercise state and the health scale parameters in the resting state to determine the health evaluation result of the object to be evaluated; when the current evaluation scenario is a sudden disease warning scenario, the evaluation algorithm model is used to calculate the change rate and parameter proportion of the health scale parameters exceeding the parameter threshold in the time parameter sequence to determine the health evaluation result of the object to be evaluated; when the current evaluation scenario is an Alzheimer's disease cognitive evaluation scenario, the evaluation algorithm model is used to calculate the change trend of the average value of the health scale parameters corresponding to multiple time periods in the time parameter sequence to determine the health evaluation result of the object to be evaluated.

[0128] In a specific embodiment, the stability parameter corresponding to the time parameter sequence is: in, For time point The health parameter data value, is the mean value of the parameter. The anomaly rate parameter corresponding to the time parameter series is the ratio of abnormal data points to the total data points.

[0129] It can be seen that through the above optional embodiments, the type of the current evaluation scenario and the calculation rules of the evaluation algorithm model corresponding to different scenarios are clarified, so as to realize the analysis of the time change pattern of the health scale parameters based on different evaluation algorithm models, so as to extract key indicators such as stability, abnormality rate, trend slope, motion state difference, change rate ratio and average value change trend, so as to accurately evaluate the health status of individuals, and realize the combination of multi-dimensional data, graph convolution and time series analysis to accurately capture the complex correlation and time series changes of health data, thereby improving the accuracy and reliability of health assessment.

[0130] As an optional embodiment, in the above steps, determining the health assessment result of the subject to be assessed based on the assessment algorithm model according to the graph feature parameters, the future health scale parameters, and the time parameter sequence includes:

[0131] Input the graph feature parameters and the health scale parameters at the most recent time point into the evaluation algorithm model to obtain the current evaluation results;

[0132] Input the graph feature parameters and future health scale parameters into the evaluation algorithm model to obtain future evaluation results;

[0133] The weighted sum average of the current evaluation result and the future evaluation result is calculated to obtain the health evaluation result of the to-be-evaluated object; optionally, the weighted calculation weight corresponding to the current evaluation result includes a first weight and a second weight; the weighted calculation weight corresponding to the future evaluation result includes a third weight and a fourth weight; the first weight and the third weight are both proportional to the difference between the corresponding evaluation result value and the reference evaluation result value; the second weight is inversely proportional to the time difference between the time point corresponding to the current evaluation result and the current time point; the fourth weight is proportional to the number of parameters in the time parameter sequence corresponding to the future health scale parameter corresponding to the future evaluation result; and the weighted calculation weight corresponding to the current evaluation result is greater than the weighted calculation weight corresponding to the future evaluation result.

[0134] Specifically, the weight adopted in the weighting can be dynamically adjusted according to the reliability and time correlation of the evaluation result. Preferably, the weight of the current evaluation result can be made greater than the weight of the future evaluation result, so as to give consideration to real-time accuracy and trend prediction value. For example, the weight of the current result can be made proportional to the degree of deviation of the value thereof from the normal reference range (the greater the deviation, the higher the weight to draw attention), and inversely proportional to the time difference between the time point corresponding to the result and the current time point (the more recent the data, the greater the weight); and the weight of the future result can be made proportional to the number of future time points for prediction (the longer the prediction span, the more future indicators to be considered, and the weight is appropriately increased). Through such dynamic weighting, the present application can comprehensively judge the current health status and future change trend of an individual, and obtain a more stable and reliable evaluation conclusion.

[0135] It can be seen that, through the above optional embodiments, the current evaluation result and the future evaluation result are respectively calculated based on the graph feature parameter and the health scale parameter, and the final health evaluation result is determined by weighted sum average, so as to realize accurate judgment of the current health status and reasonable prediction of the future health trend, wherein the weighted calculation weight is dynamically adjusted according to the deviation of the evaluation result, the time interval and the data volume, so that the evaluation result gives consideration to real-time accuracy and long-term trend prediction capability, realizes accurate capture of the complex correlation and time sequence change of health data by combining multi-dimensional data and graph convolution and time sequence analysis, and improves the accuracy and reliability of health evaluation.

[0136] Compared with the prior art, the scheme formed by the above-mentioned embodiments of the present application and the specific implementation schemes thereof have the following beneficial effects:

[0137] 1. Multi-modal data fusion: The method of the present application can fuse and process multiple types of health data, including physiological signals, behavior data and medical images, etc., breaking through the limitations of traditional single-index evaluation. By constructing a health parameter relationship graph and applying a graph neural network to extract features, deep fusion of multi-modal data is realized, and the potential relationship between different indicators is fully tapped.

[0138] 2. Geometric mean: Introducing geometric mean and correlation weight to calculate health scale parameters, avoiding the excessive influence of single abnormal value on overall evaluation, ensuring that each dimension information contributes more balanced to the comprehensive health index. This "geometric mean" calculation improves the robustness and credibility of the health index, and more accurately reflects the overall health status of the individual.

[0139] 3. Dynamic trend prediction: The present application incorporates time series analysis into the evaluation framework, using deep learning models such as LSTM to predict health scale parameters, realizing the forward-looking evaluation of the future health change trend of the individual. Compared with traditional methods that can only describe the current state statically, the present application can discover abnormal trends in health indicators in advance, providing more sufficient reaction time for chronic disease management and disease warning.

[0140] 4. Contextual individual evaluation: By customizing evaluation algorithm models for different application scenarios (such as specific analysis indicators for chronic diseases, exercise, acute events, cognitive impairment, etc.), the method and system of the present application can output more targeted evaluation results to meet the needs of personalized health management. This flexible architecture has wide application value in chronic disease follow-up, exercise guidance, health care for the elderly, etc.

[0141] 5. Commercial application prospects: The method and system of the present application can be combined with real-time health monitoring devices and cloud services, integrated into health management platforms through standardized interfaces (such as API), and realize automatic analysis and feedback of continuous health data of users. Its evaluation results can be used to assist doctors in decision-making, personal health risk warning, customized health intervention programs, etc., with significant commercial potential. The system adopts modular design and is easy to deploy on wearable devices, mobile applications or cloud servers, providing an innovative and feasible technical solution for the field of digital medicine and smart health.

[0142] Embodiment Two

[0143] Please refer to Figure 2 , Figure 2 is a structural schematic diagram of an individual health evaluation system based on geometric deep learning disclosed by an embodiment of the present application. Among them, Figure 2 The individual health evaluation system based on geometric deep learning described can be applied in a data processing system / data processing device / data processing server (wherein the server includes a local processing server or a cloud processing server). As Figure 2 shown, the individual health evaluation system based on geometric deep learning can include:

[0144] The data acquisition module 201 is configured to acquire health parameter data of multiple dimensions of the object to be evaluated.

[0145] The graph feature extraction module 202 is configured to extract graph feature parameters corresponding to the health parameter data in multiple dimensions based on a graph structure algorithm and a graph convolutional neural network.

[0146] The geometric calculation module 203 is configured to calculate health scale parameters corresponding to the health parameter data in multiple dimensions based on a geometric mean algorithm.

[0147] The health assessment module 204 is configured to determine a health assessment result of the to-be-evaluated object based on a dynamic time series analysis algorithm according to the graph feature parameters and the health scale parameters.

[0148] As can be seen, the above embodiments can obtain the health parameter data in multiple dimensions of the to-be-evaluated object, extract the corresponding graph feature parameters by using the graph structure algorithm and the graph convolutional neural network, calculate the health scale parameters by using the geometric mean algorithm, and comprehensively analyze these data based on the dynamic time series analysis algorithm, so as to accurately capture the complex correlation and time series change of the health data by combining the multi-dimensional data and the graph convolution and time series analysis, and improve the accuracy and reliability of the health assessment.

[0149] As an optional embodiment, the data types of the health parameter data include at least one of blood glucose, blood pressure, heart rate, body temperature, activity amount, step count, exercise time, energy consumption, blood oxygen saturation, heart rate variability, electrocardiogram, MRI image features, cognitive test scores, and brain metabolism data.

[0150] As can be seen, by using the above optional embodiments, the content of the health parameter data is limited to comprehensively represent the physiological related features of the to-be-evaluated object, so as to accurately evaluate the health status subsequently, and assist in accurately capturing the complex correlation and time series change of the health data by combining the multi-dimensional data and the graph convolution and time series analysis, and improving the accuracy and reliability of the health assessment.

[0151] As an optional embodiment, the specific manner in which the graph feature extraction module extracts the graph feature parameters corresponding to the health parameter data in multiple dimensions based on the graph structure algorithm and the graph convolutional neural network includes:

[0152] calculating graph structure information corresponding to the health parameter data in multiple dimensions based on the graph structure algorithm;

[0153] inputting the health parameter data in multiple dimensions and the graph structure information into the trained graph convolutional neural network to obtain the output graph feature parameters.

[0154] It can be seen that, by the above optional embodiments, the graph structure information of the health parameter data in multiple dimensions is calculated based on the graph structure algorithm, and the health parameter data and the graph structure information are input into the trained graph convolutional neural network to extract the graph feature parameters, which can effectively capture the complex relationship between the health parameters, improve the accuracy of feature extraction, and assist in accurately capturing the complex correlation and time sequence change of the health data by combining multi-dimensional data and graph convolution and time sequence analysis, and improving the accuracy and reliability of health assessment.

[0155] As an optional embodiment, the specific manner in which the graph feature extraction module calculates the graph structure information corresponding to the health parameter data in multiple dimensions based on the graph structure algorithm includes:

[0156] Each health parameter data in the health parameter data in multiple dimensions is determined as a parameter node in a preset graph structure;

[0157] For any two parameter nodes, the type correlation degree between the data types of the health parameter data corresponding to the two parameter nodes is determined; optionally, the type correlation degree is determined by a preset correlation degree rule between different data types;

[0158] The data similarity of the health parameter data corresponding to the two parameter nodes is calculated;

[0159] The historical data sequence of a plurality of historical time points corresponding to the health parameter data corresponding to the two parameter nodes is determined;

[0160] The change rate similarity between the data change rates of the historical data sequence corresponding to the two parameter nodes is calculated;

[0161] The weighted sum average of the type correlation degree, the data similarity and the change rate similarity is calculated to obtain the inter-node relationship value between the two parameter nodes;

[0162] All parameter nodes and corresponding inter-node relationship values are output as graph structure information corresponding to the health parameter data in multiple dimensions.

[0163] It can be seen that, by the above optional embodiments, the health parameter data in multiple dimensions is mapped to parameter nodes in a preset graph structure, and the inter-node relationship value is calculated in combination with the type correlation degree rule, the data similarity and the historical data change rate similarity to construct the graph structure information of the health parameter data, which can accurately model the complex correlation between different health parameters, and assist in accurately capturing the complex correlation and time sequence change of the health data by combining multi-dimensional data and graph convolution and time sequence analysis, and improving the accuracy and reliability of health assessment.

[0164] As an optional embodiment, the specific manner in which the geometry calculation module calculates the health scale parameter corresponding to the plurality of dimensional health parameter data based on the geometric mean algorithm includes:

[0165] For each data type, calculate the geometric mean of all health parameter data corresponding to the data type;

[0166] According to the average value of all node relationship values corresponding to the parameter nodes corresponding to all health parameter data of the data type, obtain the relationship weight corresponding to the data type;

[0167] Calculate the power of the geometric mean as the base number and the relationship weight as the index to obtain the type scale parameter corresponding to the data type;

[0168] Calculate the average value of the type scale parameters corresponding to all data types to obtain the health scale parameter.

[0169] As can be seen, through the above optional embodiment, the geometric mean of the health parameter data of each data type is calculated, and the type scale parameter is determined by combining the relationship weight of the corresponding parameter node and using power operation, and the health scale parameter is further obtained by averaging, which can comprehensively consider the internal distribution characteristics of the data and the association weight of different data types, improve the representativeness and robustness of the scale parameter in the health evaluation process, and assist in realizing the accurate capture of the complex correlation and time sequence change of health data by combining multi-dimensional data and graph convolution and time sequence analysis, and improving the accuracy and reliability of health evaluation.

[0170] As an optional embodiment, the specific manner in which the health evaluation module determines the health evaluation result of the to-be-evaluated object according to the graph feature parameter and the health scale parameter based on the dynamic time sequence analysis algorithm includes:

[0171] Sort the health scale parameters of the to-be-evaluated object corresponding to a plurality of time points according to time from early to late to obtain a time parameter sequence;

[0172] Input the time parameter sequence into the trained LSTM neural network model to obtain the predicted future health scale parameter corresponding to the to-be-evaluated object; optionally, the LSTM neural network is trained by a training data set including a plurality of training time parameter sequences;

[0173] Determine the evaluation algorithm model corresponding to the current evaluation scene;

[0174] Determine the health evaluation result of the to-be-evaluated object based on the evaluation algorithm model according to the graph feature parameter, the future health scale parameter, and the time parameter sequence.

[0175] It can be seen that, through the above optional embodiments, by time sequencing the health scale parameters and inputting them into the trained LSTM neural network model to predict future health scale parameters, and combining the current evaluation scene to select a suitable evaluation algorithm model, the health evaluation result is finally determined based on at least one of the graph feature parameters, the future health scale parameters and the time parameter sequence, realizing the combination of multi-dimensional data and graph convolution and time series analysis to accurately capture the complex correlation and time series change of health data, and improving the accuracy and reliability of health evaluation.

[0176] As an optional embodiment, the current evaluation scene is a chronic management scene, a sports health evaluation scene, a sudden disease early warning scene or an Alzheimer's disease cognitive evaluation scene; when the current evaluation scene is a chronic management scene, the evaluation algorithm model is used to calculate the stability parameter, the abnormality rate parameter and the trend slope parameter corresponding to the time parameter sequence to determine the health evaluation result of the to-be-evaluated object; when the current evaluation scene is a sports health evaluation scene, the evaluation algorithm model is used to calculate the difference between the health scale parameter of the to-be-evaluated object in the sports state and the health scale parameter in the resting state to determine the health evaluation result of the to-be-evaluated object; when the current evaluation scene is a sudden disease early warning scene, the evaluation algorithm model is used to calculate the change rate and the parameter proportion of the health scale parameter exceeding the parameter threshold in the time parameter sequence to determine the health evaluation result of the to-be-evaluated object; when the current evaluation scene is an Alzheimer's disease cognitive evaluation scene, the evaluation algorithm model is used to calculate the change trend of the average value of the health scale parameters corresponding to multiple time periods in the time parameter sequence to determine the health evaluation result of the to-be-evaluated object.

[0177] It can be seen that, through the above optional embodiments, the type of the current evaluation scene and the calculation rules of the evaluation algorithm models corresponding to different scenes are determined to realize the analysis of the time variation mode of the health scale parameters based on different evaluation algorithm models to extract key indicators such as stability, abnormality rate, trend slope, sports state difference, change rate proportion and average value change trend, thereby accurately evaluating the individual health status, realizing the combination of multi-dimensional data and graph convolution and time series analysis to accurately capture the complex correlation and time series change of health data, and improving the accuracy and reliability of health evaluation.

[0178] As an optional embodiment, the specific way in which the health evaluation module determines the health evaluation result of the to-be-evaluated object based on the graph feature parameters, the future health scale parameters and the time parameter sequence based on the evaluation algorithm model includes:

[0179] inputting the graph feature parameters and the health scale parameter at the nearest time point into the evaluation algorithm model to obtain the current evaluation result;

[0180] inputting the graph feature parameter and the future health scale parameter into an evaluation algorithm model to obtain a future evaluation result;

[0181] calculating a weighted sum average of the current evaluation result and the future evaluation result to obtain a health evaluation result of the to-be-evaluated object; optionally, wherein the weighted calculation weight corresponding to the current evaluation result includes a first weight and a second weight; the weighted calculation weight corresponding to the future evaluation result includes a third weight and a fourth weight; the first weight and the third weight are both proportional to the difference between the corresponding evaluation result value and the reference evaluation result value; the second weight is inversely proportional to the time difference between the time point corresponding to the current evaluation result and the current time point; the fourth weight is proportional to the number of parameters in the time parameter sequence corresponding to the future health scale parameter corresponding to the future evaluation result; and the weighted calculation weight corresponding to the current evaluation result is greater than the weighted calculation weight corresponding to the future evaluation result.

[0182] It can be seen that, through the above optional embodiments, the current evaluation result and the future evaluation result are calculated based on the graph feature parameter and the health scale parameter, and the final health evaluation result is determined by weighted sum average, so as to realize accurate judgment of the current health status and reasonable prediction of the future health trend, wherein the weighted calculation weight is dynamically adjusted according to the deviation of the evaluation result, the time interval and the data quantity, so that the evaluation result takes into account the real-time accuracy and the long-term trend prediction ability, realizes accurate capture of the complex correlation and time sequence change of health data by combining multi-dimensional data, graph convolution and time sequence analysis, and improves the accuracy and reliability of health evaluation.

[0183] Embodiment three

[0184] Please refer to Figure 3 , Figure 3 It is another kind of individual health evaluation system based on geometric deep learning disclosed in the embodiments of the present application. Figure 3 The described individual health evaluation system based on geometric deep learning is applied in a data processing system / data processing device / data processing server (wherein the server includes a local processing server or a cloud processing server). As Figure 3 shown, the individual health evaluation system based on geometric deep learning can include:

[0185] a memory 301 storing executable program codes;

[0186] a processor 302 coupled with the memory 301;

[0187] The processor 302 calls the executable program codes stored in the memory 301 to execute the steps of the individual health evaluation method based on geometric deep learning described in embodiment one.

[0188] Embodiment four

[0189] The embodiment of the present application discloses a computer readable storage medium storing a computer program for electronic data exchange, wherein the computer program causes a computer to execute the steps of the individual health assessment method based on geometric deep learning described in embodiment one.

[0190] Embodiment five

[0191] The embodiment of the present application discloses a computer program product comprising a non-transitory computer readable storage medium storing a computer program, and the computer program is operable to cause a computer to execute the steps of the individual health assessment method based on geometric deep learning described in embodiment one.

[0192] The above describes specific embodiments of the present application, other embodiments are within the scope of the appended claims. In some cases, the acts or steps recited in the claims can be performed in a different order than those described in the embodiments and still achieve desirable results. In addition, the processes depicted in the figures do not necessarily require the particular order shown or sequential order in order to achieve the desired results. In certain implementations, multitasking and parallel processing can be advantageous.

[0193] The systems, apparatuses, modules or units illustrated by the above embodiments can be specifically implemented by a computer chip or entity, or by a product with certain functions. A typical implementation device is a computer. Specifically, the computer may, for example, be a personal computer, a laptop computer, a cellular phone, a camera phone, a smartphone, a personal digital assistant, a media player, a navigation device, an email device, a game console, a tablet computer, a wearable device, or a combination of any of these devices.

[0194] For the convenience of description, the above apparatus is described in various units by function respectively when described. Of course, the functions of each unit can be implemented in the same or more software and / or hardware when implementing the present specification.

[0195] Those skilled in the art will understand that the embodiments of the present specification can be provided as a method, a system, or a computer program product. Therefore, the embodiments of the present specification can take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Moreover, the embodiments of the present specification can take the form of a computer program product implemented on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0196] The specification is presented with reference to flow diagrams and / or block diagrams of methods, apparatus (systems) and computer program products according to embodiments of the specification. It will be understood that each block of the flow diagrams and / or block diagrams, and combinations of blocks in the flow diagrams and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general purpose computer, special purpose computer, embedded processing element or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flow diagrams and / or block diagrams block or blocks. Figure 1 one or more flow or multiple flows and / or blocks Figure 1 one or more flow or multiple flows and / or blocks

[0197] These computer program instructions can also be stored in a computer- readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instructions which implement the function specified in the flow diagrams and / or block diagrams block or blocks. Figure 1 one or more flow or multiple flows and / or blocks Figure 1 one or more flow or multiple flows and / or blocks

[0198] These computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flow diagrams and / or block diagrams block or blocks. Figure 1 one or more flow or multiple flows and / or blocks Figure 1 one or more flow or multiple flows and / or blocks

[0199] In one typical configuration, the computing device includes one or more processors (CPUs), input / output interfaces, network interfaces, and memory.

[0200] The memory can include non-persistent memory and / or volatile memory, such as random access memory (RAM) and / or cache memory, non-volatile memory, such as read-only memory (ROM), EPROM, and / or flash memory, etc. The memory is an example of computer-readable media.

[0201] Computer-readable media includes permanent and non-permanent, movable and non-movable media that can be implemented by any method or technology to store information. The information can be computer-readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, compact disc read-only memory (CD-ROM), digital versatile disc (DVD) or other optical storage, magnetic cassette, magnetic disk storage or other magnetic storage devices, or any other non-transmission medium that can be used to store information accessible by a computing device. According to the definition herein, computer-readable media does not include transitory media such as modulated data signals and carriers.

[0202] It should also be noted that the terms "comprising", "containing", or any other variant thereof are intended to encompass non-exclusive inclusion, such that processes, methods, articles or devices that comprise a list of elements not only include those elements, but also include other elements not expressly listed or inherent to such processes, methods, articles or devices. Without more limitations, the element defined by the statement "comprising a" does not exclude the presence of additional identical elements in the process, method, article or device comprising the element.

[0203] The specification can be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform specific tasks or implement specific abstract data types. The specification can also be practiced in a distributed computing environment, in which tasks are performed by remote processing devices connected by a communication network. In a distributed computing environment, program modules can be located in local and remote computer storage media, including storage devices.

[0204] Each embodiment in the specification is described in a progressive manner, and the same or similar parts between each embodiment can be referred to each other, and each embodiment focuses on the difference from other embodiments. In particular, for system embodiments, since they are basically similar to method embodiments, the description is relatively simple, and the relevant parts can be referred to the part of the method embodiment.

[0205] Finally, it should be noted that: the individual health assessment method and system disclosed in the embodiment of the application disclosed only as the preferred embodiment of the application, only for the description of the technical solutions of the application, not to limit; although the application is described in detail with reference to the foregoing embodiments, those skilled in the art should understand; the technical solutions recorded in the foregoing embodiments can be modified, or some technical features can be replaced; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the application.

Claims

1. A method for individual health assessment based on geometric deep learning, characterized in that: The method comprises: Obtain health parameter data of multiple dimensions of the object to be evaluated; Based on the graph structure algorithm and graph convolutional neural network, graph feature parameters corresponding to the health parameter data of the multiple dimensions are extracted, including: Determining each health parameter data in the multiple dimensions as a parameter node in a preset graph structure; For any two parameter nodes, determining the type association between the data types of the health parameter data corresponding to the two parameter nodes; the type association is determined by a preset association rule between different data types; Calculating the data similarity of the health parameter data respectively corresponding to the two parameter nodes; Determine a historical data sequence of a plurality of historical time points corresponding to the health parameter data respectively corresponding to the two parameter nodes; Calculating the change rate similarity between the data change rates of the historical data sequences respectively corresponding to the two parameter nodes; Calculating a weighted average of the type association degree, the data similarity, and the change rate similarity to obtain an inter-node relationship value between the two parameter nodes; Outputting all the parameter nodes and the corresponding inter-node relationship values ​​as graph structure information corresponding to the health parameter data of the multiple dimensions; Inputting the health parameter data of the multiple dimensions and the graph structure information into a trained graph convolutional neural network to obtain output graph feature parameters; Calculating health scale parameters corresponding to the health parameter data of the multiple dimensions based on a geometric mean algorithm; According to the graph feature parameters and the health scale parameters, based on a dynamic time series analysis algorithm, a health assessment result of the object to be assessed is determined.

2. The individual health assessment method based on geometric deep learning according to claim 1, characterized in that: The data type of the health parameter data includes at least one of blood sugar, blood pressure, heart rate, body temperature, activity level, number of steps, exercise time, energy consumption, blood oxygen saturation, heart rate variability, electrocardiogram, MRI image features, cognitive test scores and brain metabolism data.

3. The individual health assessment method based on geometric deep learning according to claim 1, characterized in that: The calculating of the health scale parameters corresponding to the health parameter data of the multiple dimensions based on the geometric mean algorithm includes: For each data type, calculating the geometric mean of all the health parameter data corresponding to the data type; Obtaining a relationship weight corresponding to the data type according to an average value of all the inter-node relationship values ​​corresponding to the parameter nodes corresponding to all the health parameter data corresponding to the data type; Calculate a power with the geometric mean as the base and the relationship weight as the exponent to obtain a type scale parameter corresponding to the data type; The average value of the type scale parameters corresponding to all the data types is calculated to obtain the health scale parameter.

4. The individual health assessment method based on geometric deep learning according to claim 1, characterized in that: The determining of the health assessment result of the subject to be assessed based on the graph feature parameters and the health scale parameters and a dynamic time series analysis algorithm includes: Sorting the health scale parameters corresponding to multiple time points of the subject to be evaluated from early to late according to time to obtain a time parameter sequence; Inputting the time parameter sequence into a trained LSTM neural network model to obtain predicted future health scale parameters corresponding to the subject to be evaluated; the LSTM neural network is trained using a training data set including multiple training time parameter sequences; Determine the evaluation algorithm model corresponding to the current evaluation scenario; According to the graph feature parameters, the future health scale parameters and the time parameter sequence, based on the evaluation algorithm model, a health evaluation result of the object to be evaluated is determined.

5. The individual health assessment method based on geometric deep learning according to claim 4, characterized in that: The current assessment scenario is a chronic management scenario, a sports health assessment scenario, a sudden illness warning scenario, or an Alzheimer's disease cognitive assessment scenario; when the current assessment scenario is a chronic management scenario, the assessment algorithm model is used to calculate the stability parameter, the abnormality rate parameter, and the trend slope parameter corresponding to the time parameter sequence to determine the health assessment result of the subject to be assessed; When the current evaluation scenario is a sports health evaluation scenario, the evaluation algorithm model is used to calculate the difference between the health scale parameters of the object to be evaluated in a sports state and the health scale parameters in a resting state, so as to determine the health evaluation result of the object to be evaluated; when the current evaluation scenario is a sudden disease warning scenario, the evaluation algorithm model is used to calculate the change rate and parameter proportion of the health scale parameters that exceed the parameter threshold in the time parameter sequence, so as to determine the health evaluation result of the object to be evaluated; when the current evaluation scenario is an Alzheimer's disease cognitive evaluation scenario, the evaluation algorithm model is used to calculate the change trend of the average value of the health scale parameters corresponding to multiple time periods in the time parameter sequence, so as to determine the health evaluation result of the object to be evaluated.

6. The individual health assessment method based on geometric deep learning according to claim 4, characterized in that: The step of determining the health assessment result of the subject to be assessed based on the assessment algorithm model according to the graph feature parameters, the future health scale parameters, and the time parameter sequence includes: Inputting the graph feature parameters and the health scale parameters at the most recent time point into the evaluation algorithm model to obtain a current evaluation result; Inputting the graph feature parameters and the future health scale parameters into the evaluation algorithm model to obtain future evaluation results; Calculate the weighted sum average of the current assessment result and the future assessment result to obtain the health assessment result of the object to be assessed; wherein, the weighted calculation weight corresponding to the current assessment result includes a first weight and a second weight; the weighted calculation weight corresponding to the future assessment result includes a third weight and a fourth weight; the first weight and the third weight are both proportional to the difference between the corresponding assessment result value and the reference assessment result value; the second weight is inversely proportional to the time difference between the time point corresponding to the current assessment result and the current time point; the fourth weight is proportional to the number of parameters in the time parameter sequence corresponding to the future health scale parameter corresponding to the future assessment result; the weighted calculation weight corresponding to the current assessment result is greater than the weighted calculation weight corresponding to the future assessment result.

7. An individual health assessment system based on geometric deep learning, characterized in that: The system comprises: A data acquisition module is used to obtain health parameter data of multiple dimensions of the object to be evaluated; A graph feature extraction module is used to extract graph feature parameters corresponding to the health parameter data of the multiple dimensions based on a graph structure algorithm and a graph convolutional neural network, including: Determining each health parameter data in the multiple dimensions as a parameter node in a preset graph structure; For any two parameter nodes, determining the type association between the data types of the health parameter data corresponding to the two parameter nodes; the type association is determined by a preset association rule between different data types; Calculating the data similarity of the health parameter data respectively corresponding to the two parameter nodes; Determine a historical data sequence of a plurality of historical time points corresponding to the health parameter data respectively corresponding to the two parameter nodes; Calculating the change rate similarity between the data change rates of the historical data sequences respectively corresponding to the two parameter nodes; Calculating a weighted average of the type association degree, the data similarity, and the change rate similarity to obtain an inter-node relationship value between the two parameter nodes; Outputting all the parameter nodes and the corresponding inter-node relationship values ​​as graph structure information corresponding to the health parameter data of the multiple dimensions; Inputting the health parameter data of the multiple dimensions and the graph structure information into a trained graph convolutional neural network to obtain output graph feature parameters; A geometric calculation module, configured to calculate health scale parameters corresponding to the health parameter data of the multiple dimensions based on a geometric mean algorithm; The health assessment module is used to determine the health assessment result of the object to be assessed based on the graph feature parameters and the health scale parameters and a dynamic time series analysis algorithm.

8. An individual health assessment system based on geometric deep learning, characterized in that: The system comprises: a memory storing executable program code; a processor coupled to the memory; The processor calls the executable program code stored in the memory to execute the individual health assessment method based on geometric deep learning as described in any one of claims 1-6.

Images