Regional Population Health Information Management System Based on Regional Data Mining
Through the deep neural network model based on regional data mining, the differential sequence of physical examination data and the global physical examination feature matrix are used to solve the accuracy of the assessment of population health changes in the geographical environment, and early warning of disease prevention is achieved, and health management is guaranteed. effectiveness.
Patent Information
- Application Number
- CN202210964221.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2022-08-11
- Publication Date
- 2024-12-06
- Estimated Expiration
- 2042-08-11
AI Technical Summary
The prior art is difficult to accurately evaluate changes in population health status in different geographical environments, and there is a lack of effective methods to prevent diseases, which affects the accuracy and effectiveness of health management.
A system based on regional data mining is adopted, and through a deep neural network model, the differential sequence of physical examination data and the global physical examination feature matrix are used to evaluate health status and early warning of disease prevention.
Accurate assessment of changes in regional health status has been achieved, preventing the occurrence of diseases in advance, and protecting people's lives and health.
Smart Images

Figure CN115295159B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of intelligent management of health information, and more specifically, to a regional population health information management system based on regional data mining and a management method thereof. Background Art
[0002] In recent years, with environmental pollution and ecological environment destruction, the health status of people living in specific areas will be affected, but this impact is qualitative, and there is no clear indicator to prove that the environment in a specific area will have an adverse impact on the health of people in that specific area.
[0003] People have long noticed the close relationship between geographical environment and human health. People living in different regions will have different health conditions due to their geographical location, climate change and living habits. However, this change in health can be good or bad. Some will make people healthier and stronger, while others will have adverse effects on people's health and even endanger their lives.
[0004] Therefore, a regional population health information management system based on regional data mining is desired to evaluate changes in the health status of people in the region, thereby preventing the occurrence of diseases in advance and protecting people's lives and health.
[0005] At present, deep learning and neural networks have been widely used in computer vision, natural language processing, text signal processing and other fields. In addition, deep learning and neural networks have also shown a level close to or even beyond that of humans in image classification, object detection, semantic segmentation, text translation and other fields.
[0006] In recent years, the development of deep learning and neural networks has provided solutions and plans for the health information management of regional populations. Summary of the invention
[0007] In order to solve the above technical problems, the present application is proposed. The embodiment of the present application provides a regional population health information management system based on regional data mining and a management method thereof, which uses the first-year physical examination data and the second-year physical examination data of all members in each of multiple regions as input data, and adopts a deep neural network model as a feature extractor to judge whether the health status changes of members in the area to be evaluated meet the predetermined standards. The global physical examination feature association information of the health status change features of the members in each area is used to perform feature mapping and classification. In this way, the health status changes of the members in the area to be evaluated can be accurately evaluated to prevent the occurrence of diseases in advance and protect people's lives and health.
[0008] According to one aspect of the present application, a regional population health information management system based on regional data mining is provided, which includes:
[0009] A physical examination data collection unit, used to obtain the first-year physical examination data and the second-year physical examination data of all members in each of the multiple areas;
[0010] a physical examination data change evaluation unit, used for calculating the difference between the physical examination data of the second year and the physical examination data of the first year according to the physical examination items of each member to obtain a sequence of differential physical examination data corresponding to each member;
[0011] A physical examination change data encoding unit, used for passing the sequence of differential physical examination data of each member through a transformer-based context encoder to obtain a differential physical examination vector corresponding to each member;
[0012] A regional physical examination feature encoding unit, used for two-dimensionally arranging the differential physical examination vectors of all members of each of the regions into a two-dimensional feature matrix and then passing the matrix through a trained first convolutional neural network as a filter to obtain a regional physical examination feature vector corresponding to each region;
[0013] A global physical examination feature encoding unit, used for two-dimensionally arranging the multiple regional physical examination feature vectors of the multiple regions into a two-dimensional regional feature matrix and then passing the two-dimensional regional feature matrix through a trained second convolutional neural network as a feature extractor to obtain a global physical examination feature matrix;
[0014] An area-to-be-evaluated unit, configured to extract a regional physical examination feature vector of the area to be evaluated from the multiple regional physical examination feature vectors of the multiple areas;
[0015] a vector query unit, configured to use the regional physical examination feature vector of the area to be evaluated as a query feature vector and multiply it with the global physical examination feature matrix to obtain a classification feature vector; and
[0016] The health information management result generating unit is used to pass the classification feature vector through a classifier to obtain a classification result, and the classification result is used to indicate whether the change of the health status of the members of the area to be evaluated exceeds a predetermined standard.
[0017] In the above-mentioned regional population health information management system based on regional data mining, the physical examination change data encoding unit includes: an input data construction subunit, which is used to use the embedding layer of the converter-based context encoder to convert the sequence of differential physical examination data of each member into an input vector to obtain a sequence of input vectors; and a context encoding subunit, which is used to use the converter of the converter-based context encoder to perform global context semantic encoding on the sequence of input vectors to obtain the differential physical examination vector corresponding to each member.
[0018] In the above-mentioned regional population health information management system based on regional data mining, the regional physical examination feature encoding unit is further used for: each layer of the first convolutional neural network that has been trained as a filter performs the following on the input data in the forward pass of the layer: convolution processing on the input data to obtain a convolution feature map; mean pooling based on the local feature matrix on the convolution feature map to obtain a pooled feature map; and nonlinear activation on the pooled feature map to obtain an activated feature map; wherein the output of the last layer of the first convolutional neural network is the regional physical examination feature vector corresponding to each region, and the input of the first layer of the first convolutional neural network is the two-dimensional feature matrix.
[0019] In the above-mentioned regional population health information management system based on regional data mining, the global physical examination feature encoding unit is further used for: each layer of the trained second convolutional neural network as a feature extractor performs the following on the input data in the forward pass of the layer: convolution processing on the input data to obtain a convolution feature map; mean pooling based on the local channel dimension on the convolution feature map to obtain a pooled feature map; and nonlinear activation on the pooled feature map to obtain an activation feature map; wherein the output of the last layer of the second convolutional neural network is the global physical examination feature matrix, and the input of the first layer of the second convolutional neural network is the two-dimensional regional feature matrix.
[0020] In the above-mentioned regional population health information management system based on regional data mining, a training module is further included, including: a training data acquisition unit, used to obtain the first-year physical examination data and the second-year physical examination data of all members in each of the multiple regions; a training physical examination data change evaluation unit, used to calculate the difference between the second-year physical examination data and the first-year physical examination data of each member according to the physical examination items to obtain a sequence of differential physical examination data corresponding to each member; a training physical examination change data encoding unit, used to pass the sequence of differential physical examination data of each member through a converter-based context encoder to obtain a differential physical examination vector corresponding to each member; a training regional physical examination feature encoding unit, used to two-dimensionally arrange the differential physical examination vectors of all members in each of the regions into a two-dimensional feature matrix and then pass it through a first convolutional neural network as a filter to obtain a regional physical examination feature vector corresponding to each region; a training global physical examination feature encoding unit, used to two-dimensionally arrange multiple regional physical examination feature vectors of the multiple regions into a two-dimensional feature matrix and then pass it through a first convolutional neural network as a filter A second convolutional neural network as a feature extractor to obtain a global physical examination feature matrix; a training unit for the area to be evaluated, used to extract a regional physical examination feature vector of the area to be evaluated from multiple regional physical examination feature vectors of the multiple areas; a training vector query unit, used to multiply the regional physical examination feature vector of the area to be evaluated with the global physical examination feature matrix as a query feature vector to obtain a classification feature vector; a classification loss unit, used to pass the classification feature vector through a classifier to obtain a classification loss function value; a frugal decomposition encouragement loss unit, used to calculate a frugal decomposition encouragement loss function value based on the global physical examination feature matrix, the frugal decomposition encouragement loss function value based on the global physical examination feature matrix is related to the sum of the global physical examination feature matrix in the row direction corresponding to the regional physical examination feature vector; and a training unit, used to train the first convolutional neural network and the second convolutional neural network using the weighted sum of the frugal decomposition encouragement loss function value based on the global physical examination feature matrix and the classification loss function value as the loss function value.
[0021] In the above regional population health information management system based on regional data mining, the classification loss unit is further used to: use the classifier to process the classification feature vector with the following formula to obtain the classification result, wherein the formula is: softmax{(W n ,B n ):…:(W1,B1)|X}, where W1 to W n is the weight matrix, B1 to B n is the bias vector, X is the classification feature vector; and the cross entropy value between the classification result and the true value is calculated as the classification loss function value.
[0022] In the above regional population health information management system based on regional data mining, the frugal decomposition encouraging loss unit is further used to: calculate the frugal decomposition encouraging loss function value based on the global physical examination feature matrix;
[0023] Wherein, the formula is:
[0024]
[0025] in, represents the sum of the global physical examination feature matrix in the row direction corresponding to the regional physical examination feature vector, ||·||2 represents the binary norm of the feature matrix, and τ is a hyperparameter.
[0026] According to another aspect of the present application, a management method of a regional population health information management system based on regional data mining includes:
[0027] Obtaining first-year physical examination data and second-year physical examination data of all members in each of a plurality of regions;
[0028] Calculating the differences between the second-year physical examination data and the first-year physical examination data of each member according to the physical examination items to obtain a sequence of differential physical examination data corresponding to each member;
[0029] Passing the sequence of differential physical examination data of each member through a transformer-based context encoder to obtain a differential physical examination vector corresponding to each member;
[0030] The differential autopsy vectors of all members of each region are arranged in two dimensions into a two-dimensional feature matrix, and then passed through a first convolutional neural network trained as a filter to obtain a regional autopsy feature vector corresponding to each region;
[0031] Arranging the multiple regional physical examination feature vectors of the multiple regions in two dimensions into a two-dimensional regional feature matrix and then passing the matrix through a trained second convolutional neural network as a feature extractor to obtain a global physical examination feature matrix;
[0032] Extracting a regional physical examination feature vector of the region to be evaluated from the multiple regional physical examination feature vectors of the multiple regions;
[0033] Taking the regional physical examination feature vector of the area to be evaluated as a query feature vector and multiplying it with the global physical examination feature matrix to obtain a classification feature vector; and
[0034] The classification feature vector is passed through a classifier to obtain a classification result, and the classification result is used to indicate whether the change in the health status of the members in the area to be evaluated exceeds a predetermined standard.
[0035] In the management method of the regional population health information management system based on regional data mining, the sequence of differential physical examination data of each member is passed through a converter-based context encoder to obtain a differential physical examination vector corresponding to each member, including: using the embedding layer of the converter-based context encoder to convert the sequence of differential physical examination data of each member into input vectors to obtain a sequence of input vectors; and using the converter of the converter-based context encoder to perform global context semantic encoding on the sequence of input vectors to obtain the differential physical examination vector corresponding to each member.
[0036] In the management method of the regional population health information management system based on regional data mining, the differential physical examination vectors of all members of each of the regions are arranged in two dimensions into a two-dimensional feature matrix and then passed through a first convolutional neural network that has been trained as a filter to obtain regional physical examination feature vectors corresponding to each region, including: each layer of the first convolutional neural network that has been trained as a filter performs the following on the input data in the forward pass of the layer: convolution processing on the input data to obtain a convolutional feature map; mean pooling based on the local feature matrix on the convolutional feature map to obtain a pooled feature map; and nonlinear activation on the pooled feature map to obtain an activated feature map; wherein the output of the last layer of the first convolutional neural network is the regional physical examination feature vector corresponding to each region, and the input of the first layer of the first convolutional neural network is the two-dimensional feature matrix.
[0037] In the management method of the regional population health information management system based on regional data mining, the multiple regional physical examination feature vectors of the multiple regions are two-dimensionally arranged into a two-dimensional regional feature matrix and then passed through a trained second convolutional neural network as a feature extractor to obtain a global physical examination feature matrix, including: each layer of the trained second convolutional neural network as a feature extractor performs the following on the input data in the forward pass of the layer: convolution processing on the input data to obtain a convolution feature map; mean pooling based on the local channel dimension on the convolution feature map to obtain a pooled feature map; and nonlinear activation on the pooled feature map to obtain an activation feature map; wherein the output of the last layer of the second convolutional neural network is the global physical examination feature matrix, and the input of the first layer of the second convolutional neural network is the two-dimensional regional feature matrix.
[0038] In the above-mentioned management method of the regional population health information management system based on regional data mining, a training phase is further included, including: obtaining the first-year physical examination data and the second-year physical examination data of all members in each of the multiple regions; calculating the difference between the second-year physical examination data and the first-year physical examination data of each member according to the physical examination items to obtain a sequence of differential physical examination data corresponding to each member; passing the sequence of differential physical examination data of each member through a converter-based context encoder to obtain a differential physical examination vector corresponding to each member; arranging the differential physical examination vectors of all members in each of the regions in two dimensions into a two-dimensional feature matrix and then passing it through a first convolutional neural network as a filter to obtain a regional physical examination feature vector corresponding to each region; arranging the multiple regional physical examination feature vectors of the multiple regions in two dimensions into a two-dimensional feature matrix and then passing it through a feature extractor as a feature extractor; A second convolutional neural network is used to obtain a global physical examination feature matrix; a regional physical examination feature vector of the area to be evaluated is extracted from the multiple regional physical examination feature vectors of the multiple areas; the regional physical examination feature vector of the area to be evaluated is used as a query feature vector to multiply the global physical examination feature matrix to obtain a classification feature vector; the classification feature vector is passed through a classifier to obtain a classification loss function value; a frugal decomposition encouragement loss function value based on the global physical examination feature matrix is calculated, and the frugal decomposition encouragement loss function value based on the global physical examination feature matrix is related to the sum value of the global physical examination feature matrix in the row direction corresponding to the regional physical examination feature vector; and the first convolutional neural network and the second convolutional neural network are trained using the weighted sum of the frugal decomposition encouragement loss function value based on the global physical examination feature matrix and the classification loss function value as the loss function value.
[0039] In the management method of the regional population health information management system based on regional data mining, the classification feature vector is passed through a classifier to obtain a classification loss function value, including: using the classifier to process the classification feature vector using the following formula to obtain the classification result, wherein the formula is: softmax{(W n ,B n ):…:(W1,B1)|X}, where W1 to W n is the weight matrix, B1 to B n is the bias vector, X is the classification feature vector; and the cross entropy value between the classification result and the true value is calculated as the classification loss function value.
[0040] In the management method of the regional population health information management system based on regional data mining, calculating the value of the frugal decomposition encouragement loss function based on the global physical examination feature matrix includes: calculating the value of the frugal decomposition encouragement loss function based on the global physical examination feature matrix;
[0041] Wherein, the formula is:
[0042]
[0043] in, represents the sum of the global physical examination feature matrix in the row direction corresponding to the regional physical examination feature vector, ||·||2 represents the binary norm of the feature matrix, and τ is a hyperparameter.
[0044] Compared with the prior art, the regional population health information management system and management method based on regional data mining provided by the present application use the first-year physical examination data and the second-year physical examination data of all members in each of multiple regions as input data, and adopt a deep neural network model as a feature extractor to use the global physical examination feature association information of the health status change features of the members in each region to perform feature mapping and classification in the judgment of whether the health status changes of the members in the area to be evaluated meet the predetermined standards. In this way, the health status changes of the members in the area to be evaluated can be accurately evaluated to prevent the occurrence of diseases in advance and protect people's lives and health. BRIEF DESCRIPTION OF THE DRAWINGS
[0045] By describing the embodiments of the present application in more detail in conjunction with the accompanying drawings, the above and other purposes, features and advantages of the present application will become more apparent. The accompanying drawings are used to provide a further understanding of the embodiments of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation of the present application. In the accompanying drawings, the same reference numerals generally represent the same components or steps.
[0046] Figure 1 It is a block diagram of a regional population health information management system based on regional data mining according to an embodiment of the present application.
[0047] Figure 2 It is a block diagram of a training module in a regional population health information management system based on regional data mining according to an embodiment of the present application.
[0048] Figure 3 The present invention is a flowchart of a management method of a regional population health information management system based on regional data mining according to an embodiment of the present application.
[0049] Figure 4 It is a schematic diagram of the architecture of a management method of a regional population health information management system based on regional data mining according to an embodiment of the present application. DETAILED DESCRIPTION
[0050] Below, the exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the exemplary embodiments described here.
[0051] Scenario Overview
[0052] As mentioned earlier, in recent years, with environmental pollution and the destruction of the ecological environment, the health status of people living in specific areas will be affected, but this impact is qualitative, and there is no clear indicator to prove that the environment in a specific area will have an adverse impact on the health status of people in that specific area.
[0053] People have long noticed the close relationship between geographical environment and human health. People living in different regions will have different health conditions due to their geographical location, climate change and living habits. However, this change in health can be good or bad. Some will make people healthier and stronger, while others will have adverse effects on people's health and even endanger their lives.
[0054] Therefore, a regional population health information management system based on regional data mining is desired to evaluate changes in the health status of people in the region, thereby preventing the occurrence of diseases in advance and protecting people's lives and health.
[0055] At present, deep learning and neural networks have been widely used in computer vision, natural language processing, text signal processing and other fields. In addition, deep learning and neural networks have also shown a level close to or even beyond that of humans in image classification, object detection, semantic segmentation, text translation and other fields.
[0056] In recent years, the development of deep learning and neural networks has provided solutions and plans for the health information management of regional populations.
[0057] Based on this, the inventor of the present application considers that people in different regions will have different health conditions due to their geographical location, climate change and living habits. Due to the uncertainty of such changes, there is no basis for the evaluation of such changes, that is, the evaluation of the changes in the health conditions of members in a certain region. This makes it difficult to ensure the accuracy and effectiveness of the evaluation. In addition, the inventor of the present application also considers that there is an implicit correlation between the changes in the health conditions of all members in each region, and there is also a correlation between the changes in the health conditions of all members in each region. Therefore, when evaluating the changes in the health conditions of members in a certain region, it is expected to use these implicit correlation information to make global classification judgments.
[0058] Specifically, in the technical solution of the present application, first, the first-year physical examination data and the second-year physical examination data of all members in each of the multiple regions are obtained. Then, it should be understood that since it is necessary to evaluate the changes in the health status of the members in the region, it is necessary to pay more attention to the change characteristics of the physical examination data, so that the difference between the second-year physical examination data and the first-year physical examination data of each member is further calculated according to the physical examination items to obtain a sequence of differential physical examination data corresponding to each member.
[0059] Furthermore, considering that there is correlation information between the sequences of differential physical examination data of each member, in order to more fully extract this implicit correlation feature information, the sequence of differential physical examination data of each member is further encoded through a converter-based context encoder to extract global high-dimensional semantic features between the sequences of differential physical examination data of each member so as to be more suitable for characterizing the essential characteristics of changes in the health status of each member, thereby obtaining a differential physical examination vector corresponding to each member.
[0060] Then, the differential physical examination vectors of all members of each of the regions are arranged in two dimensions into a two-dimensional feature matrix to integrate the global health implicit change features of all members of each of the regions, and then feature mining is performed in the first convolutional neural network as a filter to extract the high-dimensional implicit correlation feature information of the health change features of all members of each of the regions, thereby obtaining the regional physical examination feature vectors corresponding to each region. It should be understood that feature mining through the first convolutional neural network as a filter can not only mine the implicit feature information of the data, but also reduce the number of parameters, thereby reducing the amount of calculation, preventing overfitting and improving the accuracy of subsequent classification.
[0061] Furthermore, for the regional physical examination feature vectors of each region, in order to improve the accuracy of the assessment of whether the changes in the health status of the members of the region meet the predetermined standards, it is also necessary to use the global physical examination feature association information of the health status change characteristics of the members of each region to perform feature mapping and classification. That is, specifically, in the technical solution of the present application, it is necessary to further arrange the multiple regional physical examination feature vectors of the multiple regions in two dimensions into a two-dimensional regional feature matrix, and then perform feature mining in the second convolutional neural network as a feature extractor to obtain a global physical examination feature matrix.
[0062] When evaluating the health status change of the members of the area to be evaluated, first, the regional physical examination feature vector of the area to be evaluated is extracted from the multiple regional physical examination feature vectors of the multiple areas. Then, the regional physical examination feature vector of the area to be evaluated is used as a query feature vector to be multiplied with the global physical examination feature matrix to map the regional physical examination feature vector of the area to be evaluated to the high-dimensional feature space of the global physical examination feature matrix, thereby obtaining a classification feature vector.
[0063] However, since the regional physical examination feature vector is used as the query feature vector and multiplied with the global physical examination feature matrix to obtain the classification feature vector, in the technical solution of the present application, it is expected that the global physical examination feature matrix can obtain good inter-class expression ability between each regional physical examination feature vector in a direction perpendicular to the direction of the regional physical examination feature vector.
[0064] Therefore, in the technical solution of the present application, a parsimonious decomposition-encouraged loss function based on the global physical examination feature matrix is introduced, which is expressed as:
[0065]
[0066] in, represents the sum of the global physical examination feature matrix in the row direction corresponding to the regional physical examination feature vector, ||·||2 represents the binary norm of the feature matrix, and τ is a hyperparameter.
[0067] That is, the loss function promotes the frugal decomposition of high-dimensional manifolds by grouping the features in the direction of each row arrangement of the inter-class expression and penalizing the overlap of elements in the row direction of the intra-class expression by calculating the distance-based union of the symbolic function, that is, constructing a more solid geometric shape of the high-dimensional manifold based on a set of convex polytopes. In this way, the inter-class dimensional monotonicity of the high-dimensional manifold of the global physical examination feature matrix in the high-dimensional space is improved, thereby improving the inter-class expression ability corresponding to each of the regional physical examination feature vectors, and then improving the accuracy of classification.
[0068] Based on this, the present application proposes a regional population health information management system based on regional data mining, which includes: a physical examination data acquisition unit, used to obtain the first-year physical examination data and the second-year physical examination data of all members in each of multiple regions; a physical examination data change evaluation unit, used to calculate the difference between the second-year physical examination data and the first-year physical examination data of each member according to the physical examination items to obtain a sequence of differential physical examination data corresponding to each member; a physical examination change data encoding unit, used to pass the sequence of differential physical examination data of each member through a converter-based context encoder to obtain a differential physical examination vector corresponding to each member; a regional physical examination feature encoding unit, used to arrange the differential physical examination vectors of all members in each of the regions in two dimensions into a two-dimensional feature matrix and then pass it through a first convolutional neural network trained as a filter a global physical examination feature encoding unit, used for two-dimensionally arranging the multiple regional physical examination feature vectors of the multiple regions into a two-dimensional regional feature matrix and then passing the two-dimensionally arranging the multiple regional physical examination feature vectors of the multiple regions through a second convolutional neural network trained as a feature extractor to obtain a global physical examination feature matrix; an area unit to be evaluated, used for extracting the regional physical examination feature vector of the area to be evaluated from the multiple regional physical examination feature vectors of the multiple regions; a vector query unit, used for multiplying the regional physical examination feature vector of the area to be evaluated as a query feature vector with the global physical examination feature matrix to obtain a classification feature vector; and a health information management result generating unit, used for passing the classification feature vector through a classifier to obtain a classification result, wherein the classification result is used to indicate whether the change in the health status of the members of the area to be evaluated exceeds a predetermined standard.
[0069] After introducing the basic principles of the present application, various non-limiting embodiments of the present application will be described in detail with reference to the accompanying drawings.
[0070] Exemplary Systems
[0071] Figure 1 FIG. 1 is a block diagram of a regional population health information management system based on regional data mining according to an embodiment of the present application. Figure 1As shown, according to the regional population health information management system 200 based on regional data mining according to the embodiment of the present application, it includes: a physical examination data acquisition unit 210, which is used to obtain the first-year physical examination data and the second-year physical examination data of all members in each of the multiple regions; a physical examination data change evaluation unit 220, which is used to calculate the difference between the second-year physical examination data and the first-year physical examination data of each of the members according to the physical examination items to obtain a sequence of differential physical examination data corresponding to each member; a physical examination change data encoding unit 230, which is used to pass the sequence of differential physical examination data of each member through a converter-based context encoder to obtain a differential physical examination vector corresponding to each member; a regional physical examination feature encoding unit 240, which is used to arrange the differential physical examination vectors of all members in each of the regions in two dimensions into a two-dimensional feature matrix and then pass it through a first convolutional neural network trained as a filter. A network is used to obtain regional physical examination feature vectors corresponding to each area; a global physical examination feature encoding unit 250 is used to arrange the multiple regional physical examination feature vectors of the multiple areas in two dimensions into a two-dimensional regional feature matrix and then obtain a global physical examination feature matrix through a second convolutional neural network that has been trained as a feature extractor; an area unit 260 to be evaluated is used to extract the regional physical examination feature vector of the area to be evaluated from the multiple regional physical examination feature vectors of the multiple areas; a vector query unit 270 is used to multiply the regional physical examination feature vector of the area to be evaluated as a query feature vector with the global physical examination feature matrix to obtain a classification feature vector; and a health information management result generation unit 280 is used to pass the classification feature vector through a classifier to obtain a classification result, and the classification result is used to indicate whether the change in the health status of the members of the area to be evaluated exceeds a predetermined standard.
[0072] Specifically, in the embodiment of the present application, the physical examination data acquisition unit 210 and the physical examination data change evaluation unit 220 are used to obtain the first-year physical examination data and the second-year physical examination data of all members in each of the multiple regions, and calculate the difference between the second-year physical examination data and the first-year physical examination data of each member according to the physical examination items to obtain a sequence of differential physical examination data corresponding to each member. As mentioned above, it should be understood that, considering that people in different regional environments will have different health conditions due to their geographical locations, climate changes, and living habits, and due to the uncertainty of such changes, there is no basis for the evaluation of such changes, that is, the evaluation of the health status changes of members in a certain area, which will make it difficult to ensure the accuracy and effectiveness of the evaluation. In addition, considering that there is an implicit correlation between the health status changes of all members in each region, and there is also a correlation between the health status changes of all members in each region, therefore, in the technical solution of the present application, when evaluating the health status changes of members in a certain area, it is expected to use these implicit correlation information for global classification judgment.
[0073] That is, specifically, in the technical solution of the present application, first, the first-year physical examination data and the second-year physical examination data of all members in each of the multiple regions are obtained. Then, it should be understood that since it is necessary to evaluate the changes in the health status of the members in the region, it is necessary to pay more attention to the change characteristics of the physical examination data, so that the difference between the second-year physical examination data and the first-year physical examination data of each member is further calculated according to the physical examination items to obtain a sequence of differential physical examination data corresponding to each member.
[0074] Specifically, in the embodiment of the present application, the physical examination change data encoding unit 230 is used to pass the sequence of differential physical examination data of each member through a converter-based context encoder to obtain a differential physical examination vector corresponding to each member. It should be understood that, considering that there is association information between the sequence of differential physical examination data of each member, therefore, in the technical solution of the present application, in order to more fully extract this implicit association feature information, the sequence of differential physical examination data of each member is further encoded through a converter-based context encoder to extract the global high-dimensional semantic features between the sequence of differential physical examination data of each member to be more suitable for characterizing the essential features of the changes in the health status of each member, thereby obtaining a differential physical examination vector corresponding to each member.
[0075] More specifically, in an embodiment of the present application, the physical examination change data encoding unit includes: an input data construction subunit, used to use the embedding layer of the converter-based context encoder to convert the sequence of differential physical examination data of each member into an input vector to obtain a sequence of input vectors; and a context encoding subunit, used to use the converter of the converter-based context encoder to perform global context semantic encoding on the sequence of input vectors to obtain the differential physical examination vector corresponding to each member.
[0076] Specifically, in an embodiment of the present application, the regional physical examination feature encoding unit 240 is used to arrange the differential physical examination vectors of all members of each of the regions in two dimensions into a two-dimensional feature matrix, and then pass the first convolutional neural network trained as a filter to obtain the regional physical examination feature vectors corresponding to each region. That is, in the technical solution of the present application, the differential physical examination vectors of all members of each of the regions are arranged in two dimensions into a two-dimensional feature matrix to integrate the global health implicit change features of all members of each of the regions, and then perform feature mining in the first convolutional neural network as a filter to extract the high-dimensional implicit correlation feature information of the health change features of all members of each of the regions, thereby obtaining the regional physical examination feature vectors corresponding to each region. It should be understood that by performing feature mining through the first convolutional neural network as a filter, not only can the implicit feature information of the data be mined, but also the number of parameters can be reduced, thereby reducing the amount of calculation, preventing overfitting, and improving the accuracy of subsequent classification.
[0077] More specifically, in an embodiment of the present application, the regional physical examination feature encoding unit is further used to: each layer of the first convolutional neural network that has been trained as a filter performs the following on the input data in the forward pass of the layer: convolution processing on the input data to obtain a convolutional feature map; mean pooling based on the local feature matrix on the convolutional feature map to obtain a pooled feature map; and nonlinear activation on the pooled feature map to obtain an activated feature map; wherein the output of the last layer of the first convolutional neural network is the regional physical examination feature vector corresponding to each region, and the input of the first layer of the first convolutional neural network is the two-dimensional feature matrix.
[0078] Specifically, in an embodiment of the present application, the global physical examination feature encoding unit 250 is used to arrange the multiple regional physical examination feature vectors of the multiple regions in two dimensions into a two-dimensional regional feature matrix, and then pass it through a second convolutional neural network that has been trained as a feature extractor to obtain a global physical examination feature matrix. It should be understood that, further, for the regional physical examination feature vectors of each region, in order to improve the accuracy of the assessment of whether the changes in the health status of the members of the region meet the predetermined standards, it is also necessary to use the global physical examination feature association information of the health status change characteristics of the members of each region to perform feature mapping and classification. That is, specifically, in the technical solution of the present application, it is necessary to further arrange the multiple regional physical examination feature vectors of the multiple regions in two dimensions into a two-dimensional regional feature matrix, and then perform feature mining in a second convolutional neural network that serves as a feature extractor to obtain a global physical examination feature matrix.
[0079] More specifically, in an embodiment of the present application, the global physical examination feature encoding unit is further used to: each layer of the trained second convolutional neural network as a feature extractor performs the following on the input data in the forward pass of the layer: convolution processing on the input data to obtain a convolution feature map; mean pooling based on the local channel dimension on the convolution feature map to obtain a pooled feature map; and nonlinear activation on the pooled feature map to obtain an activation feature map; wherein the output of the last layer of the second convolutional neural network is the global physical examination feature matrix, and the input of the first layer of the second convolutional neural network is the two-dimensional regional feature matrix.
[0080] Specifically, in the embodiment of the present application, the area unit to be evaluated 260, the vector query unit 270 and the health information management result generation unit 280 are used to extract the regional physical examination feature vector of the area to be evaluated from the multiple regional physical examination feature vectors of the multiple areas, and use the regional physical examination feature vector of the area to be evaluated as the query feature vector to multiply with the global physical examination feature matrix to obtain a classification feature vector, and then pass the classification feature vector through the classifier to obtain a classification result, and the classification result is used to indicate whether the change in the health status of the members of the area to be evaluated exceeds a predetermined standard. That is, in the technical solution of the present application, when evaluating the change in the health status of the members of the area to be evaluated, first, extract the regional physical examination feature vector of the area to be evaluated from the multiple regional physical examination feature vectors of the multiple areas. Then, use the regional physical examination feature vector of the area to be evaluated as the query feature vector to multiply with the global physical examination feature matrix to map the regional physical examination feature vector of the area to be evaluated to the high-dimensional feature space of the global physical examination feature matrix, thereby obtaining a classification feature vector. In this way, the classification feature vector can be passed through a classifier to obtain a classification result indicating whether the change in the health status of the members of the area to be evaluated exceeds a predetermined standard.
[0081] However, since the regional physical examination feature vector is used as a query feature vector and multiplied with the global physical examination feature matrix to obtain a classification feature vector, it is expected that the global physical examination feature matrix can obtain good inter-class expression capabilities between the regional physical examination feature vectors in a direction perpendicular to the direction of the regional physical examination feature vector. Therefore, in the technical solution of the present application, a parsimonious decomposition encouraging loss function based on the global physical examination feature matrix is introduced to train the first convolutional neural network and the second convolutional neural network.
[0082] Figure 2 FIG. 1 is a block diagram of a training module in a regional population health information management system based on regional data mining according to an embodiment of the present application. Figure 2As shown, the training module 320 includes: a training data acquisition unit 3201, used to obtain the first-year physical examination data and the second-year physical examination data of all members in each of the multiple regions; a training physical examination data change evaluation unit 3202, used to calculate the difference between the second-year physical examination data and the first-year physical examination data of each member according to the physical examination items to obtain a sequence of differential physical examination data corresponding to each member; a training physical examination change data encoding unit 3203, used to pass the sequence of differential physical examination data of each member through a converter-based context encoder to obtain a differential physical examination vector corresponding to each member; a training regional physical examination feature encoding unit 3204, used to two-dimensionally arrange the differential physical examination vectors of all members in each of the regions into a two-dimensional feature matrix and then pass it through a first convolutional neural network as a filter to obtain a regional physical examination feature vector corresponding to each region; a training global physical examination feature encoding unit 3205, used to two-dimensionally arrange the multiple regional physical examination feature vectors of the multiple regions into a two-dimensional feature matrix and then pass it through a second convolutional neural network as a feature extractor A global physical examination feature matrix is obtained through the network; a training unit 3206 for the area to be evaluated is used to extract a regional physical examination feature vector of the area to be evaluated from multiple regional physical examination feature vectors of the multiple areas; a training vector query unit 3207 is used to multiply the regional physical examination feature vector of the area to be evaluated with the global physical examination feature matrix as a query feature vector to obtain a classification feature vector; a classification loss unit 3208 is used to pass the classification feature vector through a classifier to obtain a classification loss function value; a frugal decomposition encouragement loss unit 3209 is used to calculate a frugal decomposition encouragement loss function value based on the global physical examination feature matrix, and the frugal decomposition encouragement loss function value based on the global physical examination feature matrix is related to the sum value of the global physical examination feature matrix in the row direction corresponding to the regional physical examination feature vector; and a training unit 3210 is used to train the first convolutional neural network and the second convolutional neural network using the weighted sum of the frugal decomposition encouragement loss function value based on the global physical examination feature matrix and the classification loss function value as the loss function value.
[0083] Specifically, in the embodiment of the present application, the classification loss unit 3208 is used to pass the classification feature vector through the classifier to obtain the classification loss function value. That is, in the technical solution of the present application, in the training module, the classification feature vector can be obtained in the above manner, and then the classification feature vector is passed through the classifier to calculate the classification loss function value to assist the training of the first convolutional neural network and the second convolutional neural network.
[0084] More specifically, in the embodiment of the present application, the classification loss unit is further used to: use the classifier to process the classification feature vector according to the following formula to obtain the classification result, wherein the formula is: softmax{(W n ,B n ):…:(W1,B1)|X}, where W1 to W n is the weight matrix, B1 to B n is the bias vector, X is the classification feature vector; and the cross entropy value between the classification result and the true value is calculated as the classification loss function value.
[0085] Specifically, in an embodiment of the present application, the parsimonious decomposition encouraging loss unit 3209 is used to calculate the parsimonious decomposition encouraging loss function value based on the global physical examination feature matrix, and the parsimonious decomposition encouraging loss function value based on the global physical examination feature matrix is related to the sum of the global physical examination feature matrix in the row direction corresponding to the regional physical examination feature vector. It should be understood that since the regional physical examination feature vector is used as the query feature vector and multiplied with the global physical examination feature matrix to obtain the classification feature vector, it is expected that the global physical examination feature matrix can obtain good inter-class expression ability between each regional physical examination feature vector in a direction perpendicular to the direction of the regional physical examination feature vector. Therefore, in the technical solution of the present application, a parsimonious decomposition encouraging loss function based on the global physical examination feature matrix is introduced to train the first convolutional neural network and the second convolutional neural network. That is, the frugal decomposition encourages the loss function value to group the features in the direction of each row arrangement of the inter-class expression, and imposes a penalty on the overlap of elements in the row direction of the intra-class expression, and promotes the frugal decomposition of the high-dimensional manifold by calculating the distance-based union of the symbolic function, that is, constructing a more solid geometric shape of the high-dimensional manifold based on a set of convex polytopes. In this way, the inter-class dimensional monotonicity of the high-dimensional manifold of the global physical examination feature matrix in the high-dimensional space is improved, thereby improving the inter-class expression ability corresponding to each of the regional physical examination feature vectors, and then improving the accuracy of classification.
[0086] More specifically, in the embodiment of the present application, the frugal decomposition encouraging loss unit is further used to: calculate the frugal decomposition encouraging loss function value based on the global physical examination feature matrix;
[0087] Wherein, the formula is:
[0088]
[0089] in, represents the sum of the global physical examination feature matrix in the row direction corresponding to the regional physical examination feature vector, ||·||2 represents the binary norm of the feature matrix, and τ is a hyperparameter.
[0090] In summary, the regional population health information management system 200 based on regional data mining according to the embodiment of the present application is explained, which uses the first-year physical examination data and the second-year physical examination data of all members in each of the multiple regions as input data, and adopts a deep neural network model as a feature extractor to map and classify features based on the global physical examination feature association information of the health status change features of the members in each region in the judgment of whether the health status changes of the members in the region to be evaluated meet the predetermined standards. In this way, the health status changes of the members in the region to be evaluated can be accurately evaluated to prevent the occurrence of diseases in advance and protect people's lives and health.
[0091] As described above, the regional population health information management system 200 based on regional data mining according to the embodiment of the present application can be implemented in various terminal devices, such as a server of a regional population health information management algorithm based on regional data mining, etc. In one example, the regional population health information management system 200 based on regional data mining according to the embodiment of the present application can be integrated into the terminal device as a software module and / or a hardware module. For example, the regional population health information management system 200 based on regional data mining can be a software module in the operating system of the terminal device, or it can be an application developed for the terminal device; of course, the regional population health information management system 200 based on regional data mining can also be one of the many hardware modules of the terminal device.
[0092] Alternatively, in another example, the regional population health information management system 200 based on regional data mining and the terminal device may also be separate devices, and the regional population health information management system 200 based on regional data mining may be connected to the terminal device via a wired and / or wireless network and transmit interactive information in accordance with an agreed data format.
[0093] Exemplary Methods
[0094] Figure 3 The figure shows a flow chart of a management method of a regional population health information management system based on regional data mining. Figure 3As shown, according to the management method of the regional population health information management system based on regional data mining according to the embodiment of the present application, the steps include: S110, obtaining the first-year physical examination data and the second-year physical examination data of all members in each of the multiple regions; S120, calculating the difference between the second-year physical examination data and the first-year physical examination data of each member according to the physical examination items to obtain a sequence of differential physical examination data corresponding to each member; S130, passing the sequence of differential physical examination data of each member through a converter-based context encoder to obtain a differential physical examination vector corresponding to each member; S140, arranging the differential physical examination vectors of all members in each of the regions in two dimensions into a two-dimensional feature matrix and then passing the first convolution neural network trained as a filter The regional physical examination feature vectors corresponding to each area are obtained through the network; S150, the multiple regional physical examination feature vectors of the multiple areas are two-dimensionally arranged into a two-dimensional regional feature matrix, and then a global physical examination feature matrix is obtained through a second convolutional neural network that has been trained as a feature extractor; S160, the regional physical examination feature vector of the area to be evaluated is extracted from the multiple regional physical examination feature vectors of the multiple areas; S170, the regional physical examination feature vector of the area to be evaluated is used as a query feature vector and multiplied with the global physical examination feature matrix to obtain a classification feature vector; and, S180, the classification feature vector is passed through a classifier to obtain a classification result, and the classification result is used to indicate whether the change in the health status of the members of the area to be evaluated exceeds a predetermined standard.
[0095] Figure 4 The figure shows a schematic diagram of the architecture of the management method of the regional population health information management system based on regional data mining according to an embodiment of the present application. Figure 4 As shown, in the network architecture of the management method of the regional population health information management system based on regional data mining, first, the second-year physical examination data of each member is calculated (for example, Figure 4 P2) and first-year physical examination data (e.g. Figure 4 The difference between the physical examination items between the members (P1) is obtained to obtain a sequence of differential physical examination data corresponding to each member (for example, Figure 4 IN as shown in FIG); then, the sequence of differential physical examination data of each member is passed through a context encoder based on a converter (for example, as Figure 4 E1 as shown in FIG. 1 ) to obtain a differential body detection vector corresponding to each member (eg, Figure 4 Then, the differential body detection vectors of all members of each of the regions are arranged in two dimensions into a two-dimensional feature matrix (for example, Figure 4 M1 as shown in FIG) is then passed through a first convolutional neural network (e.g., as shown in FIG) that has been trained as a filter. Figure 4CNN1 as shown in FIG. 1) to obtain the regional physical examination feature vector corresponding to each region (for example, as Figure 4 Then, the plurality of regional physical examination feature vectors of the plurality of regions are arranged in two dimensions into a two-dimensional regional feature matrix (for example, Figure 4 M2) is then passed through a second convolutional neural network (e.g., Figure 4 CNN2 as shown in the figure) to obtain a global physical examination feature matrix (for example, Figure 4 Then, a regional physical examination feature vector of the region to be evaluated is extracted from the multiple regional physical examination feature vectors of the multiple regions (for example, Figure 4 VF3 shown in FIG3 ); then, the regional physical examination feature vector of the area to be evaluated is used as a query feature vector and multiplied with the global physical examination feature matrix to obtain a classification feature vector (for example, Figure 4 and, finally, passing the classified feature vector through a classifier (e.g., Figure 4 The circle S shown in the figure is used to obtain a classification result, which is used to indicate whether the change in the health status of the members in the area to be evaluated exceeds a predetermined standard.
[0096] More specifically, in steps S110 and S120, the first-year physical examination data and the second-year physical examination data of all members in each of the multiple regions are obtained, and the difference between the second-year physical examination data and the first-year physical examination data of each member is calculated according to the physical examination items to obtain a sequence of differential physical examination data corresponding to each member. It should be understood that, considering that people in different regional environments will have different health conditions due to their geographical locations, climate changes, and living habits, and due to the uncertainty of such changes, there is no basis for the evaluation of such changes, that is, the evaluation of the changes in the health conditions of members in a certain region, which will make it difficult to ensure the accuracy and effectiveness of the evaluation. In addition, considering that there is an implicit correlation between the changes in the health conditions of all members in each region, and there is also a correlation between the changes in the health conditions of all members in each region, therefore, in the technical solution of the present application, when evaluating the changes in the health conditions of members in a certain region, it is expected to use these implicit correlation information for global classification judgment.
[0097] That is, specifically, in the technical solution of the present application, first, the first-year physical examination data and the second-year physical examination data of all members in each of the multiple regions are obtained. Then, it should be understood that since it is necessary to evaluate the changes in the health status of the members in the region, it is necessary to pay more attention to the change characteristics of the physical examination data, so that the difference between the second-year physical examination data and the first-year physical examination data of each member is further calculated according to the physical examination items to obtain a sequence of differential physical examination data corresponding to each member.
[0098] More specifically, in step S130, the sequence of differential physical examination data of each member is passed through a converter-based context encoder to obtain a differential physical examination vector corresponding to each member. It should be understood that, considering that there is association information between the sequence of differential physical examination data of each member, in the technical solution of the present application, in order to more fully extract this implicit association feature information, the sequence of differential physical examination data of each member is further encoded through a converter-based context encoder to extract the global high-dimensional semantic features between the sequence of differential physical examination data of each member to be more suitable for characterizing the essential features of the changes in the health status of each member, thereby obtaining a differential physical examination vector corresponding to each member.
[0099] More specifically, in step S140, the differential physical examination vectors of all members of each of the regions are arranged in two dimensions into a two-dimensional feature matrix, and then the regional physical examination feature vectors corresponding to each region are obtained through the first convolutional neural network that has been trained as a filter. That is, in the technical solution of the present application, the differential physical examination vectors of all members of each of the regions are arranged in two dimensions into a two-dimensional feature matrix to integrate the global health implicit change characteristics of all members of each of the regions, and then feature mining is performed in the first convolutional neural network that serves as a filter to extract the high-dimensional implicit correlation feature information of the health change characteristics of all members of each of the regions, thereby obtaining the regional physical examination feature vectors corresponding to each region. It should be understood that by performing feature mining through the first convolutional neural network that serves as a filter, not only can the implicit feature information of the data be mined, but also the number of parameters can be reduced, thereby reducing the amount of calculation, preventing overfitting, and improving the accuracy of subsequent classification.
[0100] More specifically, in step S150, the multiple regional physical examination feature vectors of the multiple regions are two-dimensionally arranged into a two-dimensional regional feature matrix, and then a global physical examination feature matrix is obtained through a second convolutional neural network that has been trained as a feature extractor. It should be understood that, further, for the regional physical examination feature vectors of each region, in order to improve the accuracy of the assessment of whether the changes in the health status of the members of the region meet the predetermined standards, it is also necessary to use the global physical examination feature association information of the health status change characteristics of the members of each region to perform feature mapping and classification. That is, specifically, in the technical solution of the present application, it is necessary to further arrange the multiple regional physical examination feature vectors of the multiple regions in two dimensions into a two-dimensional regional feature matrix, and then perform feature mining through a second convolutional neural network that serves as a feature extractor to obtain a global physical examination feature matrix.
[0101] More specifically, in step S160, step S170 and step S180, the regional physical examination feature vector of the area to be evaluated is extracted from the multiple regional physical examination feature vectors of the multiple areas, and the regional physical examination feature vector of the area to be evaluated is used as a query feature vector to multiply with the global physical examination feature matrix to obtain a classification feature vector, and then the classification feature vector is passed through a classifier to obtain a classification result, and the classification result is used to indicate whether the change in the health status of the members of the area to be evaluated exceeds a predetermined standard. That is, in the technical solution of the present application, when evaluating the change in the health status of the members of the area to be evaluated, first, the regional physical examination feature vector of the area to be evaluated is extracted from the multiple regional physical examination feature vectors of the multiple areas. Then, the regional physical examination feature vector of the area to be evaluated is used as a query feature vector to multiply with the global physical examination feature matrix to map the regional physical examination feature vector of the area to be evaluated to the high-dimensional feature space of the global physical examination feature matrix, thereby obtaining a classification feature vector. In this way, the classification feature vector can be passed through a classifier to obtain a classification result used to indicate whether the change in the health status of the members of the area to be evaluated exceeds a predetermined standard.
[0102] In summary, the management method of the regional population health information management system based on regional data mining based on the embodiment of the present application is explained, which uses the first-year physical examination data and the second-year physical examination data of all members in each of the multiple regions as input data, and adopts a deep neural network model as a feature extractor to use the global physical examination feature association information of the health status change characteristics of the members of the region to be evaluated to perform feature mapping and classification in the judgment of whether the health status changes of the members of the region to be evaluated meet the predetermined standards. In this way, the health status changes of the members of the region to be evaluated can be accurately evaluated to prevent the occurrence of diseases in advance and protect people's lives and health.
[0103] The basic principles of the present application are described above in conjunction with specific embodiments. However, it should be noted that the advantages, strengths, effects, etc. mentioned in the present application are only examples and not limitations, and it cannot be considered that these advantages, strengths, effects, etc. are required by each embodiment of the present application. In addition, the specific details disclosed above are only for the purpose of illustration and ease of understanding, not for limitation, and the above details do not limit the present application to being implemented by adopting the above specific details.
[0104] The block diagrams of the devices, apparatuses, equipment, and systems involved in this application are only illustrative examples and are not intended to require or imply that they must be connected, arranged, and configured in the manner shown in the block diagram. As will be appreciated by those skilled in the art, these devices, apparatuses, equipment, and systems can be connected, arranged, and configured in any manner. Words such as "including", "comprising", "having", etc. are open words, referring to "including but not limited to", and can be used interchangeably with them. The words "or" and "and" used here refer to the words "and / or" and can be used interchangeably with them, unless the context clearly indicates otherwise. The words "such as" used here refer to the phrase "such as but not limited to", and can be used interchangeably with them.
[0105] It should also be noted that in the apparatus, device and method of the present application, each component or each step can be decomposed and / or recombined. Such decomposition and / or recombination should be regarded as equivalent solutions of the present application.
[0106] The above description of the disclosed aspects is provided to enable any person skilled in the art to make or use the present application. Various modifications to these aspects will be readily apparent to those skilled in the art, and the general principles defined herein may be applied to other aspects without departing from the scope of the present application. Therefore, the present application is not intended to be limited to the aspects shown herein, but rather to the widest scope consistent with the principles and novel features disclosed herein.
[0107] The above description has been given for the purpose of illustration and description. In addition, this description is not intended to limit the embodiments of the present application to the forms disclosed herein. Although multiple example aspects and embodiments have been discussed above, those skilled in the art will recognize certain variations, modifications, changes, additions and sub-combinations thereof.
Claims
1. A regional population health information management system based on regional data mining, characterized in that: include: A physical examination data collection unit, used to obtain the first-year physical examination data and the second-year physical examination data of all members in each of the multiple areas; a physical examination data change evaluation unit, used for calculating the difference between the physical examination data of the second year and the physical examination data of the first year according to the physical examination items of each member to obtain a sequence of differential physical examination data corresponding to each member; A physical examination change data encoding unit, used for passing the sequence of differential physical examination data of each member through a transformer-based context encoder to obtain a differential physical examination vector corresponding to each member; A regional physical examination feature encoding unit, used for two-dimensionally arranging the differential physical examination vectors of all members of each of the regions into a two-dimensional feature matrix and then passing the matrix through a trained first convolutional neural network as a filter to obtain a regional physical examination feature vector corresponding to each region; A global physical examination feature encoding unit, used for two-dimensionally arranging the multiple regional physical examination feature vectors of the multiple regions into a two-dimensional regional feature matrix and then passing the two-dimensional regional feature matrix through a trained second convolutional neural network as a feature extractor to obtain a global physical examination feature matrix; An area-to-be-evaluated unit, configured to extract a regional physical examination feature vector of the area to be evaluated from the multiple regional physical examination feature vectors of the multiple areas; A vector query unit, configured to use the regional physical examination feature vector of the area to be evaluated as a query feature vector and multiply it with the global physical examination feature matrix to obtain a classification feature vector; as well as The health information management result generating unit is used to pass the classification feature vector through a classifier to obtain a classification result, and the classification result is used to indicate whether the change of the health status of the members of the area to be evaluated exceeds a predetermined standard.
2. The regional population health information management system based on regional data mining according to claim 1 is characterized in that: The physical examination change data encoding unit comprises: An input data construction subunit, configured to convert the sequence of differential physical examination data of each member into an input vector using the embedding layer of the transformer-based context encoder to obtain a sequence of input vectors; and The context encoding subunit is used to use the converter of the converter-based context encoder to perform global-based context semantic encoding on the sequence of the input vectors to obtain the differential body detection vector corresponding to each member.
3. The regional population health information management system based on regional data mining according to claim 2 is characterized in that: The regional physical examination feature encoding unit is further used for: each layer of the first convolutional neural network that has been trained as a filter performs the following on the input data in the forward transfer of the layer: Perform convolution on the input data to obtain a convolution feature map; Performing mean pooling on the convolution feature map based on a local feature matrix to obtain a pooled feature map; as well as Performing nonlinear activation on the pooled feature map to obtain an activated feature map; Among them, the output of the last layer of the first convolutional neural network is the regional physical examination feature vector corresponding to each region, and the input of the first layer of the first convolutional neural network is the two-dimensional feature matrix.
4. The regional population health information management system based on regional data mining according to claim 3 is characterized in that: The global physical examination feature encoding unit is further used for: each layer of the trained second convolutional neural network as a feature extractor performs the following on the input data in the forward transfer of the layer: Perform convolution on the input data to obtain a convolution feature map; Performing mean pooling based on the local channel dimension on the convolution feature map to obtain a pooled feature map; as well as Performing nonlinear activation on the pooled feature map to obtain an activated feature map; Among them, the output of the last layer of the second convolutional neural network is the global physical examination feature matrix, and the input of the first layer of the second convolutional neural network is the two-dimensional regional feature matrix.
5. The regional population health information management system based on regional data mining according to claim 4 is characterized in that: Further training modules are included, including: A training data collection unit, used for obtaining the first-year physical examination data and the second-year physical examination data of all members in each of the multiple regions; A training physical examination data change evaluation unit, used for calculating the difference between the second year physical examination data and the first year physical examination data of each member according to the physical examination items to obtain a sequence of differential physical examination data corresponding to each member; A training physical examination change data encoding unit, used for passing the sequence of differential physical examination data of each member through a transformer-based context encoder to obtain a differential physical examination vector corresponding to each member; A training area physical examination feature encoding unit is used to arrange the differential physical examination vectors of all members of each area in two dimensions into a two-dimensional feature matrix and then pass it through a first convolutional neural network as a filter to obtain a regional physical examination feature vector corresponding to each area; A global physical examination feature encoding unit is trained to obtain a global physical examination feature matrix by two-dimensionally arranging the multiple regional physical examination feature vectors of the multiple regions into a two-dimensional feature matrix and then passing the two-dimensional feature matrix through a second convolutional neural network as a feature extractor; A training unit for the region to be evaluated is used to extract a regional physical examination feature vector of the region to be evaluated from the multiple regional physical examination feature vectors of the multiple regions; a training vector query unit, configured to use the regional physical examination feature vector of the area to be evaluated as a query feature vector and multiply it with the global physical examination feature matrix to obtain a classification feature vector; and A classification loss unit, used for passing the classification feature vector through a classifier to obtain a classification loss function value; a frugal decomposition encouraging loss unit, configured to calculate a frugal decomposition encouraging loss function value based on the global physical examination feature matrix, wherein the frugal decomposition encouraging loss function value based on the global physical examination feature matrix is related to a summation value of the global physical examination feature matrix in a row direction corresponding to the regional physical examination feature vector; and A training unit is used to train the first convolutional neural network and the second convolutional neural network by using the weighted sum of the frugal decomposition encouragement loss function value based on the global physical examination feature matrix and the classification loss function value as the loss function value.
6. The regional population health information management system based on regional data mining according to claim 5 is characterized in that: The classification loss unit is further used to: The classification feature vector is processed using the classifier according to the following formula to obtain the classification result, wherein the formula is: ;in, arrive is the weight matrix, arrive is the bias vector, is the classification feature vector; and The cross entropy value between the classification result and the true value is calculated as the classification loss function value.
7. The regional population health information management system based on regional data mining according to claim 6 is characterized in that: The frugal decomposition encouraging loss unit is further used to: calculate the frugal decomposition encouraging loss function value based on the global physical examination feature matrix; Wherein, the formula is: ;in, represents summing the global physical examination feature matrix in the row direction corresponding to the regional physical examination feature vector, represents the bi-norm of the feature matrix, is a hyperparameter.
8. A management method for a regional population health information management system based on regional data mining, characterized in that: include: Obtaining first-year physical examination data and second-year physical examination data of all members in each of a plurality of regions; Calculating the differences between the second-year physical examination data and the first-year physical examination data of each member according to the physical examination items to obtain a sequence of differential physical examination data corresponding to each member; Passing the sequence of differential physical examination data of each member through a transformer-based context encoder to obtain a differential physical examination vector corresponding to each member; The differential autopsy vectors of all members of each of the regions are arranged in two dimensions into a two-dimensional feature matrix, and then passed through a trained first convolutional neural network as a filter to obtain a regional autopsy feature vector corresponding to each region; Arranging the multiple regional physical examination feature vectors of the multiple regions in two dimensions into a two-dimensional regional feature matrix and then passing the matrix through a trained second convolutional neural network as a feature extractor to obtain a global physical examination feature matrix; Extracting a regional physical examination feature vector of the region to be evaluated from the multiple regional physical examination feature vectors of the multiple regions; Taking the regional physical examination feature vector of the area to be evaluated as a query feature vector and multiplying it with the global physical examination feature matrix to obtain a classification feature vector; as well as The classification feature vector is passed through a classifier to obtain a classification result, and the classification result is used to indicate whether the change in the health status of the members in the area to be evaluated exceeds a predetermined standard.
9. The management method of the regional population health information management system based on regional data mining according to claim 8 is characterized in that: The step of passing the sequence of differential physical examination data of each member through a transformer-based context encoder to obtain a differential physical examination vector corresponding to each member includes: Using the embedding layer of the transformer-based context encoder, respectively convert the sequence of differential physical examination data of each member into an input vector to obtain a sequence of input vectors; and The converter of the converter-based context encoder is used to perform global context semantic encoding on the sequence of the input vectors to obtain the differential body detection vector corresponding to each member.
10. The management method of the regional population health information management system based on regional data mining according to claim 9 is characterized in that: Further including a training phase, including: Obtaining first-year physical examination data and second-year physical examination data of all members in each of a plurality of regions; Calculating the differences between the second-year physical examination data and the first-year physical examination data of each member according to the physical examination items to obtain a sequence of differential physical examination data corresponding to each member; Passing the sequence of differential physical examination data of each member through a transformer-based context encoder to obtain a differential physical examination vector corresponding to each member; The differential autopsy vectors of all members of each of the regions are arranged in two dimensions into a two-dimensional feature matrix and then passed through a first convolutional neural network as a filter to obtain a regional autopsy feature vector corresponding to each region; Arranging the multiple regional physical examination feature vectors of the multiple regions in two dimensions into a two-dimensional feature matrix and then passing the matrix through a second convolutional neural network as a feature extractor to obtain a global physical examination feature matrix; Extracting a regional physical examination feature vector of the region to be evaluated from the multiple regional physical examination feature vectors of the multiple regions; Taking the regional physical examination feature vector of the area to be evaluated as a query feature vector and multiplying it with the global physical examination feature matrix to obtain a classification feature vector; and Passing the classification feature vector through a classifier to obtain a classification loss function value; Calculating a frugal decomposition encouraging loss function value based on the global physical examination feature matrix, wherein the frugal decomposition encouraging loss function value based on the global physical examination feature matrix is related to a summation value of the global physical examination feature matrix in a row direction corresponding to the regional physical examination feature vector; and The first convolutional neural network and the second convolutional neural network are trained using a weighted sum of the frugal decomposition encouragement loss function value based on the global physical examination feature matrix and the classification loss function value as the loss function value.
Citation Information
Patent Citations
Health trend estimation method and prediction system based on vital sign big data
CN113393934A
Efficient processing method and system for battery lithium membrane
CN114882021A
