Search

CN-122025114-A - User health assessment method and system based on sign data

CN122025114ACN 122025114 ACN122025114 ACN 122025114ACN-122025114-A

Abstract

The invention provides a user health assessment method and system based on sign data, and relates to the technical field of data analysis, wherein the method comprises the steps of collecting multi-mode sign data streams in physical examination nodes from a hospital physical examination system, and further constructing a real-time node database and a historical node database; based on the pearson correlation coefficient, carrying out structural feature analysis to obtain structural feature data, constructing a health state evaluation model based on a deep learning algorithm, carrying out primary evaluation to obtain a primary evaluation result, carrying out unstructured feature analysis based on a gray level co-occurrence matrix technology to obtain an unstructured feature set, carrying out similarity analysis on the unstructured feature set to obtain a feature similarity value, carrying out secondary evaluation according to the feature similarity value to obtain a secondary evaluation result, obtaining a health evaluation score according to the primary evaluation result and the secondary evaluation result, and generating a corresponding interpretability report for a user to know the self body health state.

Inventors

  • HU ZHAOMEI
  • YU LIHONG

Assignees

  • 浙江绿丝带润锦健康管理有限公司

Dates

Publication Date
20260512
Application Date
20251201

Claims (9)

  1. 1. A method of user health assessment based on sign data, the method comprising: the method comprises the steps of collecting multi-mode physical sign data streams in each physical examination node from a hospital physical examination system, setting collection periods according to physical examination item types, and constructing a real-time node database and a historical node database based on the multi-mode physical sign data streams of different collection periods; based on the Pearson correlation coefficient, carrying out structural feature analysis on the multi-modal sign data flow in the historical node database to obtain structural feature data of a user to be healthy estimated; based on gray level co-occurrence matrix technology, unstructured feature analysis is carried out on the multi-mode sign data stream, and an unstructured feature set of a user to be subjected to health evaluation is obtained; performing similarity analysis on the unstructured feature set to obtain feature similarity values of the corresponding multi-mode sign data streams; And obtaining the health evaluation score of the user to be subjected to health evaluation according to the primary evaluation result and the secondary evaluation result, further dynamically adjusting the corresponding treatment management scheme, and generating a corresponding interpretable report for the user to know the body health state of the user.
  2. 2. The method of claim 1, wherein the step of collecting the multi-modal physical sign data stream from the hospital physical examination system at each physical examination node comprises: According to the identity verification function of the hospital physical examination system, the data acquisition authority of the interviewee is granted, the multi-mode physical sign data stream of the medical equipment in each physical examination node is acquired according to the data acquisition authority, the multi-mode physical sign data stream comprises numerical data and image data, a uniform acquisition period is set according to the item type, a corresponding acquisition time stamp is marked, and a unique identifier is added for each piece of data.
  3. 3. The method for assessing the health of a user based on vital sign data of claim 2, wherein the process of constructing the real-time node database and the historical node database based on the multi-modal vital sign data streams of different acquisition periods comprises: Setting corresponding fields according to data types and unique identifiers corresponding to the multi-mode physical sign data streams in the corresponding physical examination nodes, and forming a real-time node database and a historical node database according to the setting requirements of the acquisition period; The method comprises the steps of obtaining a multi-modal sign data stream of a physical examination node corresponding to a current acquisition period, sending the multi-modal sign data stream in the current acquisition period to a real-time node database for storage, and sending the multi-modal sign data stream exceeding a preset storage time threshold value and corresponding to the acquisition period to a history node database for storage.
  4. 4. A method of assessing health of a user based on sign data according to claim 3 wherein the step of structural feature analysis of the multi-modal sign data stream in the historical node database based on pearson correlation coefficients comprises: Reading corresponding multi-mode sign data streams from a historical node database, carrying out standardization processing on numerical data in the multi-mode sign data streams, taking any two numerical data subjected to standardization processing, and obtaining a correlation coefficient between any two numerical data according to a calculation formula of a Pearson correlation coefficient; Acquiring two numerical data corresponding to the relevance according to the judging relation between the relevance coefficient and the relevance coefficient threshold value, and acquiring a time sequence characteristic curve of the corresponding numerical data changing along with time according to the corresponding acquisition period and the acquisition time stamp; and comparing the time sequence characteristic curve with the corresponding time sequence characteristic reference curve to obtain a structured characteristic data set of two numerical data corresponding to the relevance.
  5. 5. The method for evaluating health of a user based on sign data according to claim 4, wherein the process of evaluating the multi-modal sign data stream in the real-time node database once to obtain an evaluation result comprises the steps of: Obtaining a plurality of groups of different structured feature data sets of two numerical data with relevance and time sequence feature reference curves corresponding to the numerical data; grouping and marking structured characteristic data sets of two numerical data corresponding to different relevance groups and time sequence characteristic reference curves of the corresponding numerical data, and marking as Is a natural number; Will be Structured feature data sets of two numerical data corresponding to different relevance groups and time sequence feature reference curves of the corresponding numerical data are taken as sample data, and Is smaller than The method comprises the steps of obtaining a sample data mean value by utilizing sample data, recording the sample data mean value as a sample set, taking structured characteristic data sets of two numerical data corresponding to different relativity of other groups and a time sequence characteristic reference curve of the corresponding numerical data as a test set, and forming a training sample set according to the sample set and the test set; The method comprises the steps of establishing a standard state evaluation model based on an LSTM algorithm, inputting a training sample set into the standard state evaluation model, training the standard state evaluation model, and marking the standard state evaluation model after training as a health state evaluation model; and if the corresponding evaluation result of the user to be subjected to health evaluation is determined to be in a health state, performing unstructured feature analysis on the corresponding multi-mode sign data stream, and further obtaining a corresponding secondary evaluation result.
  6. 6. The method for evaluating the health of a user based on sign data according to claim 5, wherein the step of performing unstructured feature analysis on the multi-modal sign data stream based on gray level co-occurrence matrix technique to obtain the unstructured feature set of the user to be evaluated for health comprises: Obtaining image type data in a multi-mode sign data stream corresponding to a user to be subjected to health evaluation from a real-time node database, calculating gray level co-occurrence matrixes around each pixel point in the image type data in a sliding window mode for the image type data, extracting a series of texture characteristic values based on the calculated gray level co-occurrence matrixes, and further constructing an unstructured characteristic set, wherein the texture characteristic values in the unstructured characteristic set comprise contrast, correlation, energy, homogeneity, entropy and inverse variance.
  7. 7. The method of claim 6, wherein the step of performing similarity analysis on the unstructured feature set to obtain feature similarity values for the multi-modal feature data stream comprises: Presetting an unstructured standard feature set corresponding to a healthy user; And obtaining other pixel local areas adjacent to each pixel local area in the image type data in the multi-mode sign data stream of the user to be assessed, and comparing the unstructured feature set of each pixel local area in the image type data in the multi-mode sign data stream of the user to be assessed with the unstructured standard feature set of the adjacent other pixel local areas corresponding to each pixel local area to obtain the feature similarity value of the image type data in the multi-mode sign data stream.
  8. 8. The method for estimating health of a user based on sign data according to claim 7, wherein the step of performing the secondary estimation based on the feature similarity value to obtain the secondary estimation result comprises: setting a standard feature similarity threshold and a similarity analysis error range; If the sum of the feature similarity value and the similarity analysis error range is larger than the standard feature similarity threshold value, judging that the corresponding secondary evaluation result of the user to be subjected to health evaluation is abnormal in health state; If the sum of the feature similarity value and the similarity analysis error range is smaller than or equal to the standard feature similarity threshold value, judging that the corresponding secondary evaluation result of the user to be healthy is healthy; setting an evaluation result weight coefficient, and quantifying a primary evaluation result score and a secondary evaluation result score; And dynamically optimizing a corresponding treatment management scheme according to the health evaluation score, and generating a corresponding interpretability report for the user to know the body health state of the user.
  9. 9. A user health assessment system based on sign data, for implementing the user health assessment method based on sign data as set forth in any one of claims 1 to 8, comprising a data acquisition module, a primary assessment module, a secondary assessment module, a data storage module and a dynamic management module; the data acquisition module is used for acquiring multi-mode physical sign data streams in each physical examination node from the hospital physical examination system, and setting an acquisition period according to the physical examination item type; The system comprises a primary evaluation module, a deep learning algorithm, a health state evaluation model, a primary evaluation module and a secondary evaluation module, wherein the primary evaluation module is used for carrying out structural feature analysis on the multi-modal sign data flow in the historical node database based on the Pearson correlation coefficient to obtain structural feature data of a user to be health evaluated; The secondary evaluation module is used for carrying out unstructured feature analysis on the multi-modal sign data stream based on a gray level co-occurrence matrix technology to obtain an unstructured feature set of a user to be healthy evaluated, carrying out similarity analysis on the unstructured feature set to obtain a feature similarity value of the corresponding multi-modal sign data stream, carrying out secondary evaluation according to the feature similarity value to obtain a secondary evaluation result; The data storage module is used for constructing a real-time node database and a historical node database based on multi-mode sign data streams with different acquisition periods; And the dynamic management module is used for obtaining the health evaluation score of the user to be subjected to health evaluation according to the primary evaluation result and the secondary evaluation result, further dynamically adjusting the corresponding treatment management scheme, and generating a corresponding interpretable report for the user to know the self health state.

Description

User health assessment method and system based on sign data Technical Field The invention relates to the technical field of data analysis, in particular to a user health assessment method and system based on sign data. Background The current health management field relies on single-dimensional sign data or static physical examination reports, and is difficult to capture the dynamic change of the health state of an individual. The prior art has significant limitations in multi-modal data fusion, long-term trend analysis, and personalized intervention scheme generation. For example, the lack of a hierarchical management mechanism for different acquisition cycle data in conventional systems results in insufficient correlation analysis of historical data with real-time data, and failure to effectively identify early signs of potential health risks. In addition, the utilization efficiency of unstructured data is low, and the accurate depicting ability of the evaluation model on complex health states is limited. Therefore, a method and a system for user health assessment based on sign data are provided. Disclosure of Invention In order to solve the technical problems, the invention aims to provide a user health assessment method and system based on sign data. In order to achieve the above purpose, the invention provides a user health assessment method based on sign data, which comprises the following steps: the method comprises the steps of collecting multi-mode physical sign data streams in each physical examination node from a hospital physical examination system, setting collection periods according to physical examination item types, and constructing a real-time node database and a historical node database based on the multi-mode physical sign data streams of different collection periods; based on the Pearson correlation coefficient, carrying out structural feature analysis on the multi-modal sign data flow in the historical node database to obtain structural feature data of a user to be healthy estimated; based on gray level co-occurrence matrix technology, unstructured feature analysis is carried out on the multi-mode sign data stream, and an unstructured feature set of a user to be subjected to health evaluation is obtained; performing similarity analysis on the unstructured feature set to obtain feature similarity values of the corresponding multi-mode sign data streams; And obtaining the health evaluation score of the user to be subjected to health evaluation according to the primary evaluation result and the secondary evaluation result, further dynamically adjusting the corresponding treatment management scheme, and generating a corresponding interpretable report for the user to know the body health state of the user. Further, the process of collecting the multi-modal sign data stream in each physical examination node from the hospital physical examination system comprises: According to the identity verification function of the hospital physical examination system, the data acquisition authority of the interviewee is granted, the multi-mode physical sign data stream of the medical equipment in each physical examination node is acquired according to the data acquisition authority, the multi-mode physical sign data stream comprises numerical data and image data, a uniform acquisition period is set according to the item type, a corresponding acquisition time stamp is marked, and a unique identifier is added for each piece of data. Further, the process of constructing the real-time node database and the historical node database based on the multi-mode sign data streams of different acquisition periods comprises the following steps: Setting corresponding fields according to data types and unique identifiers corresponding to the multi-mode physical sign data streams in the corresponding physical examination nodes, and forming a real-time node database and a historical node database according to the setting requirements of the acquisition period; The method comprises the steps of obtaining a multi-modal sign data stream of a physical examination node corresponding to a current acquisition period, sending the multi-modal sign data stream in the current acquisition period to a real-time node database for storage, and sending the multi-modal sign data stream exceeding a preset storage time threshold value and corresponding to the acquisition period to a history node database for storage. Further, based on the pearson correlation coefficient, the process of performing structural feature analysis on the multi-modal sign data stream in the historical node database comprises the following steps: Reading corresponding multi-mode sign data streams from a historical node database, carrying out standardization processing on numerical data in the multi-mode sign data streams, taking any two numerical data subjected to standardization processing, and obtaining a correlation coefficient between any two numerical data accord