CN113393934B - Health trend estimation method and prediction system based on vital sign big data - Google Patents

Health trend estimation method and prediction system based on vital sign big data Download PDF

Info

Publication number
CN113393934B
CN113393934B CN202110633034.4A CN202110633034A CN113393934B CN 113393934 B CN113393934 B CN 113393934B CN 202110633034 A CN202110633034 A CN 202110633034A CN 113393934 B CN113393934 B CN 113393934B
Authority
CN
China
Prior art keywords
feature map
feature
vital sign
weighting
data
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active
Application number
CN202110633034.4A
Other languages
Chinese (zh)
Other versions
CN113393934A (en
Inventor
陈静
王德
陈教托
葛东飞
谢尚托
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Yijin Hangzhou Health Technology Co ltd
Original Assignee
Yijin Hangzhou Health Technology Co ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Yijin Hangzhou Health Technology Co ltd filed Critical Yijin Hangzhou Health Technology Co ltd
Priority to CN202110633034.4A priority Critical patent/CN113393934B/en
Publication of CN113393934A publication Critical patent/CN113393934A/en
Application granted granted Critical
Publication of CN113393934B publication Critical patent/CN113393934B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
    • G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
    • G16H50/00—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics
    • G16H50/30—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for calculating health indices; for individual health risk assessment
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
    • A61B5/7235—Details of waveform analysis
    • A61B5/7264—Classification of physiological signals or data, e.g. using neural networks, statistical classifiers, expert systems or fuzzy systems
    • A61B5/7267—Classification of physiological signals or data, e.g. using neural networks, statistical classifiers, expert systems or fuzzy systems involving training the classification device
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
    • A61B5/7271—Specific aspects of physiological measurement analysis
    • A61B5/7275—Determining trends in physiological measurement data; Predicting development of a medical condition based on physiological measurements, e.g. determining a risk factor
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
    • G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks
    • G—PHYSICS
    • G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
    • G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
    • G16H50/00—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics
    • G16H50/80—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for detecting, monitoring or modelling epidemics or pandemics, e.g. flu

Landscapes

  • Engineering & Computer Science (AREA)
  • Health & Medical Sciences (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Public Health (AREA)
  • Medical Informatics (AREA)
  • Artificial Intelligence (AREA)
  • Physics & Mathematics (AREA)
  • Biomedical Technology (AREA)
  • Data Mining & Analysis (AREA)
  • General Health & Medical Sciences (AREA)
  • Theoretical Computer Science (AREA)
  • Pathology (AREA)
  • Biophysics (AREA)
  • Evolutionary Computation (AREA)
  • Molecular Biology (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Heart & Thoracic Surgery (AREA)
  • Veterinary Medicine (AREA)
  • Primary Health Care (AREA)
  • Epidemiology (AREA)
  • Mathematical Physics (AREA)
  • Databases & Information Systems (AREA)
  • Animal Behavior & Ethology (AREA)
  • Physiology (AREA)
  • Psychiatry (AREA)
  • Signal Processing (AREA)
  • General Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • Surgery (AREA)
  • Computing Systems (AREA)
  • Fuzzy Systems (AREA)
  • Computational Linguistics (AREA)
  • Software Systems (AREA)
  • Evolutionary Biology (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Measuring And Recording Apparatus For Diagnosis (AREA)

Abstract

The application discloses a health trend estimation method and a prediction system based on vital sign big data. The health trend estimation method is used for mining statistical characteristics from existing data of vital signs, predicting future vital sign data based on historical and current vital sign data, and comprehensively monitoring the vital signs of a user in a contactless mode and comprehensively reflecting the physical state of the user.

Description

Health trend estimation method and prediction system based on vital sign big data
Technical Field
The present application relates to the field of artificial intelligence, and more particularly, to a health trend estimation method based on vital sign big data, a health trend prediction system based on vital sign big data, and an electronic device.
Background
At present, the product about in the aspect of user's sign intelligent monitoring is the monitoring of single sign for the majority, medical equipment can't be in succession, real-time supervision goes out the body temperature of disease, the rhythm of the heart, blood pressure, body signs such as blood oxygen, abnormal conditions is difficult by in time discovery, the nursing system of partial hospital and various large-scale medical instrument of measuring vital sign can satisfy the demand of comprehensive sign monitoring, but all need medical personnel to carry out actual closely contact with the patient and just can measure, unable large-scale quick accuracy is monitored the crowd, be unfavorable for epidemic situation quick response and the prevention and control treatment that probably takes place, and frequently, a large amount of manpowers of closely manual measurement consumption, there is great cross infection risk.
Therefore, an optimized health trend estimation and prediction technical scheme based on vital sign big data is expected.
At present, deep learning and neural networks have been widely applied in the fields of computer vision, natural language processing, text signal processing, and the like. In addition, deep learning and neural networks also exhibit a level close to or even exceeding that of humans in the fields of image classification, object detection, semantic segmentation, text translation, and the like.
In recent years, deep learning and the development of neural networks provide new solutions and solutions for estimating and predicting exposure to health trends based on vital sign big data.
Disclosure of Invention
The present application is proposed to solve the above-mentioned technical problems. The embodiment of the application provides a health trend estimation method based on vital sign big data, a health trend prediction system based on vital sign big data and electronic equipment, wherein statistical characteristics are mined from existing data of vital signs, future vital sign data are predicted based on historical and current vital sign data, and therefore a contactless mode is adopted to comprehensively monitor the vital signs of a user and comprehensively reflect the body state of the user.
According to one aspect of the application, a health trend estimation method based on vital sign big data is provided, and comprises the following steps:
a training phase comprising:
obtaining a plurality of aspects of vital sign data of a plurality of time periods arranged along a time sequence and constructing the plurality of aspects of vital sign data of each time period into a plurality of data matrixes; and
using data matrixes of adjacent time periods as input matrixes for training and real values respectively to train the convolutional neural network until the parameters of the convolutional neural network are converged; and a prediction phase;
obtaining a plurality of aspects of vital sign data arranged along a time series and constructing the plurality of aspects of vital sign data into an input matrix;
inputting the input matrix into the convolutional neural network trained in a training stage to obtain a first feature map, wherein the scale of the first feature map is t × s × c, t is a time dimension, s is a sample dimension, and c is the number of channels of the convolutional neural network;
calculating Softmax-like classification function values of the respective feature matrices of the first feature map in a time dimension relative to the first feature map as first weighting values to obtain first weighting vectors, wherein the Softmax-like classification function values are weighted sums of natural index function values raised by negative values of the feature values of the respective positions in the feature matrices divided by weighted sums of natural index function values raised by negative values of the feature values of the respective positions in the first feature map;
calculating, as a second weighting value, an inverse of each feature matrix of the first feature map in a sample dimension with respect to a Softmax-like function value of the first feature map to obtain a second weighting vector, wherein the Softmax-like function value is a weighted sum of natural index function values raised by negative values of feature values of respective positions in the feature matrix divided by a weighted sum of natural index function values raised by negative values of feature values of respective positions in the first feature map;
weighting the first feature map in a time dimension and in a sample dimension with the first weighting vector and the second weighting vector, respectively, to obtain a second feature map; and passing the second feature map through a classifier to obtain a classification result, wherein the classification result is used for representing a health tendency estimation result.
In the health trend estimation method based on vital sign big data, training a convolutional neural network by using data matrices of adjacent time periods as an input matrix for training and a real value respectively until parameters of the convolutional neural network converge includes: inputting the training input matrix into a convolutional neural network to obtain a training feature map; and training the convolutional neural network based on a mean square error loss function value between the training feature map and the real value until parameters of the convolutional neural network converge.
In the above health trend estimation method based on vital sign big data, calculating Softmax-like classification function values of respective feature matrices of the first feature map in a time dimension with respect to the first feature map as first weighting values to obtain a first weighting vector, including: calculating Softmax-like classification function values of the respective feature matrices of the first feature map in the time dimension relative to the first feature map as first weighting values to obtain a first weighting vector, with the following formula: p is a radical of formulaa,a∈t=∑xi∈Rs*cexp(-xi)/ ∑yi∈Rt*s*cexp(-yi), wherein xi represents the eigenvalue of each position in the eigen matrix, yi represents the eigenvalue of each position in the first eigen map.
In the above health trend estimation method based on vital sign big data, calculating, as a second weighting value, an inverse of each feature matrix of the first feature map in a sample dimension with respect to a Softmax-like function value of the first feature map to obtain a second weighting vector, including: calculating the reciprocal of each feature matrix of the first feature map in the sample dimension relative to the Softmax-like function value of the first feature map as a second weighting value to obtain a second weighting vector according to the following formula: p is a radical of formulab,b∈s=α*1/[∑xi∈Rt*cexp(-xi)/ ∑yi∈Rt*s*cexp(-yi)]Where α is for pb,b∈sAdjusted to [0,1 ]]Xi represents a feature value of each position in the feature matrix, and yi represents a feature value of each position in the first feature map.
In the above health trend estimation method based on vital sign big data, weighting the first feature map from the time dimension and the sample dimension with the first weighting vector and the second weighting vector, respectively, to obtain a second feature map, includes: calculating the weight of each feature matrix of the first feature map in the time dimension by using the first weight vector; and calculating the weighting of each feature matrix of the first feature map matrix weighted by the first weighting vector on a sample dimension by using the second weighting vector to obtain the second feature map.
In the above health trend estimation method based on vital sign big data, the second feature map is passed through a classifier to obtain a classification result, and the classification result is used for representing a health trend estimation result, including: passing the second feature map through one or more fully-connected layers to encode the second feature map through the one or more fully-connected layers to obtain a classified feature vector; and inputting the classification feature vector into a Softmax classification function to obtain the classification result.
In the above health trend estimation method based on vital sign big data, the classification result includes: the health trend becomes better, stable and worse.
In the health trend estimation method based on vital sign big data, the convolutional neural network is a deep residual error network.
According to another aspect of the present application, there is provided a health trend prediction system based on vital sign big data, including: a training module comprising: the training data unit is used for obtaining a plurality of aspects of vital sign data of a plurality of time periods and arranging the aspects of vital sign data of each time period along a time sequence to construct a plurality of data matrixes;
the training unit is used for respectively taking the data matrixes of the adjacent time periods obtained by the data acquisition and construction unit as input matrixes and real values for training the convolutional neural network until the parameters of the convolutional neural network are converged; and a prediction module comprising: the detection data unit is used for obtaining the vital sign data of a plurality of aspects arranged along a time sequence and constructing the vital sign data of the aspects into an input matrix;
a first feature map generation unit, configured to input the input matrix obtained by the detection data unit into the convolutional neural network trained in the training stage to obtain a first feature map, where a scale of the first feature map is t × s × c, t is a time dimension, s is a sample dimension, and c is a number of channels of the convolutional neural network;
a first weighting vector generation unit configured to calculate, as a first weighting value, a Softmax-like classification function value of each feature matrix of the first feature map generated by the first feature map generation unit with respect to the first feature map in a time dimension, to obtain a first weighting vector, where the Softmax-like classification function value is a weighted sum of natural exponent function values raised to powers of negative values of feature values of respective positions in the feature matrix divided by a weighted sum of natural exponent function values raised to powers of negative values of feature values of respective positions in the first feature map;
a second weighting vector generation unit configured to calculate, as a second weighting value, a reciprocal of a Softmax-like function value of each feature matrix of the first feature map generated by the first feature map generation unit in a sample dimension with respect to the first feature map, to obtain a second weighting vector, wherein the Softmax-like function value is a weighted sum of natural exponent function values raised to the negative values of the feature values at the respective positions in the feature matrix divided by a weighted sum of natural exponent function values raised to the negative values of the feature values at the respective positions in the first feature map;
a second feature map generation unit configured to weight the first feature map from a time dimension and a sample dimension with the first weighting vector generated by the first weighting vector generation unit and the second weighting vector generated by the second weighting vector generation unit, respectively, to obtain a second feature map; and the classification result generating unit is used for enabling the second feature map generated by the second feature map generating unit to pass through a classifier to obtain a classification result, and the classification result is used for representing a health trend estimation result.
In the above health trend prediction system based on vital sign big data, the training unit is further configured to: inputting the training input matrix into a convolutional neural network to obtain a training feature map; and training the convolutional neural network based on a mean square error loss function value between the training feature map and the real value until parameters of the convolutional neural network converge.
In the above health trend prediction system based on vital sign big data, the first weight vector generation unit is further configured to: calculating Softmax-like classification function values of the respective feature matrices of the first feature map in the time dimension relative to the first feature map as first weighting values to obtain a first weighting vector, with the following formula: p is a radical of formulaa,a∈t=∑xi∈Rs*cexp(-xi)/ ∑yi∈Rt*s*cexp(-yi), wherein xi represents the eigenvalue of each position in the eigen matrix, yi represents the eigenvalue of each position in the first eigen map.
In the above health trend prediction system based on vital sign big data, the second weight vector generation unit is further configured to: calculating the reciprocal of each feature matrix of the first feature map in the sample dimension relative to the Softmax-like function value of the first feature map as a second weighting value to obtain a second weighting vector according to the following formula: p is a radical ofb,b∈s=α*1/[∑xi∈Rt*cexp(-xi)/ ∑yi∈Rt*s*cexp(-yi)]Where α is for pb,b∈sAdjusted to [0,1 ]]Xi represents a feature value of each position in the feature matrix, and yi represents a feature value of each position in the first feature map.
In the health trend prediction system based on vital sign big data, the second feature map generation unit is further configured to: calculating the weight of each feature matrix of the first feature map in the time dimension by using the first weight vector; and calculating the weighting of each feature matrix of the first feature map matrix weighted by the first weighting vector on a sample dimension by using the second weighting vector to obtain the second feature map.
In the health trend prediction system based on vital sign big data, the classification result generating unit is further configured to: passing the second feature map through one or more fully-connected layers to encode the second feature map through the one or more fully-connected layers to obtain a classified feature vector; and inputting the classification feature vector into a Softmax classification function to obtain the classification result.
In the above health trend prediction system based on vital sign big data, the classification result generating unit is further configured to: obtaining a classification result, the classification result comprising: the health trend becomes better, stable and worse.
In the health trend prediction system based on vital sign big data, the convolutional neural network is a deep residual error network.
According to still another aspect of the present application, there is provided an electronic apparatus including: a processor; and a memory having stored therein computer program instructions which, when executed by the processor, cause the processor to perform the vital sign big data based health trend estimation method as described above.
According to yet another aspect of the present application, a computer readable medium is provided, having stored thereon computer program instructions, which, when executed by a processor, cause the processor to perform the vital sign big data based health trend estimation method as described above.
According to the health trend estimation method based on vital sign big data, the health trend prediction system based on vital sign big data and the electronic equipment, the feature extractor and the predictor are constructed based on the deep learning neural network model, so that the aims of quickly and accurately reflecting and predicting the health condition and health trend of a user are fulfilled.
Drawings
The above and other objects, features and advantages of the present application will become more apparent by describing in more detail embodiments of the present application with reference to the attached drawings. The accompanying drawings are included to provide a further understanding of the embodiments of the application and are incorporated in and constitute a part of this specification, illustrate embodiments of the application and together with the description serve to explain the principles of the application. In the drawings, like reference numbers generally indicate like parts or steps.
Fig. 1 illustrates a scene schematic diagram of a health trend estimation method based on vital sign big data according to an embodiment of the present application.
Fig. 2 illustrates a flow chart of a vital sign big data based health trend estimation method according to an embodiment of the present application.
Fig. 3 illustrates an architecture diagram of a training phase in a health trend estimation method based on vital sign big data according to an embodiment of the present application.
Fig. 4 illustrates an architectural diagram of a prediction stage in a health trend estimation method based on vital sign big data according to an embodiment of the present application.
Fig. 5 illustrates a flowchart of training a convolutional neural network until parameters of the convolutional neural network converge by using data matrices of adjacent time periods as input matrices for training and real values, respectively, in a health trend estimation method based on vital sign big data according to an embodiment of the present application.
Fig. 6 illustrates a flow chart of weighting the first feature map from a time dimension and from a sample dimension with the first weighting vector and the second weighting vector respectively to obtain a second feature map in the vital sign big data based health trend estimation method according to the embodiment of the present application.
Fig. 7 illustrates a block diagram of a vital signs big data based health trend prediction system according to an embodiment of the present application.
FIG. 8 illustrates a block diagram of an electronic device in accordance with an embodiment of the present application.
Detailed Description
Hereinafter, example embodiments according to the present application will be described in detail with reference to the accompanying drawings. It should be understood that the described embodiments are only some embodiments of the present application and not all embodiments of the present application, and that the present application is not limited by the example embodiments described herein.
As mentioned previously, most of the products related to the intelligent monitoring of user signs are single sign monitoring at present, and cannot reflect the physical conditions of users relatively comprehensively, and although the nursing system of hospitals and various large medical instruments for measuring vital signs can meet the requirements of comprehensive sign monitoring, medical staff and patients are required to be in actual close contact to measure, so that the large-scale rapid and accurate monitoring of people groups is impossible, and the quick response to the epidemic situation which may occur and the timely prevention and control treatment are not facilitated.
Therefore, an optimized health trend estimation and prediction technical scheme based on vital sign big data is expected.
In particular, the applicant of the present application considers that health trend estimation based on vital sign big data, including body temperature, blood oxygen, respiration rate, heart rate, etc., essentially mining statistical features from the existing data of vital signs, and performing future vital sign data prediction based on historical and current vital sign data, so that the above object can be achieved by constructing a feature extractor and a predictor through a neural network model based on deep learning.
Therefore, in the technical solution of the present application, it is first required to train a deep neural network to extract high-dimensional associated features based on existing vital sign big data, which may train a single neural network to extract intra-class features of the vital sign data along a class of a time dimension and inter-class features along a sample dimension by using a concept similar to "inter-frame warped field" in image processing. That is, first, vital sign data arranged in time series of aspects, such as body temperature data, blood oxygen data, respiration rate data, heart rate data as described above, are obtained, and an input data matrix is constructed, and then input to a convolutional neural network to obtain a first feature map.
Here, the convolutional neural network is trained using vital sign big data, that is, the data matrices of multiple time periods constructed as described above are obtained, and the data matrices of adjacent time periods are used as the input matrix and the real value for training, respectively, to train the convolutional neural network. For example, the input matrix may be input to a convolutional neural network to obtain a training feature map, and the convolutional neural network may be trained based on a mean square error loss function between the training feature map and a true value, so that the convolutional neural network can be trained to extract intra-class features of the vital feature data along a time dimension and inter-class features along a sample dimension.
Then, for the obtained first feature map, assuming that its time dimension is t and sample dimension is s, the first feature map may be represented as t × s × c, where c represents a channel of the convolutional neural network, and the TF-IDF concept based on the intra-class features and the inter-class features is used to adjust the weight of the feature value of each location in the first feature map. That is, a Softmax-like classification function of each s-c matrix in the t dimension with respect to the first feature map is calculated as a first weighting value, denoted as pa,a∈t=∑xi∈Rs*cexp(-xi)/ ∑yi∈Rt*s*cexp(-yi), and calculating the reciprocal of each s-c matrix in the s-dimension relative to the Softmax-like classification function of the first feature map as a second weighting value, denoted as pb,b∈s=α*1/[∑xi∈Rt*cexp(-xi)/ ∑yi∈Rt*s*cexp(-yi)]Where α is for pb,b∈sAdjusted to [0,1 ]]The coefficient is normalized by the maximum value within the interval range of (a).
In this way, the first signature is weighted from the time dimension and the sample dimension with a first weighting value and a second weighting value, respectively, to obtain a second signature, i.e. each s c matrix in the t dimension is multiplied by the first weighting value and each s c matrix in the s dimension is multiplied by the second weighting value. Then, the second feature map is used for passing through the classifier, and a classification result for representing the health tendency estimation result can be obtained.
Fig. 1 illustrates a scene schematic diagram of a health trend estimation method based on vital sign big data according to an embodiment of the present application. As shown in fig. 1, in this application scenario, in a training phase, parameters of a measured person, such as body temperature, heart rate, respiration rate, blood oxygen, etc., are acquired as training data by a plurality of sensors (e.g., C as illustrated in fig. 1); the training data is then input into a server (e.g., S as illustrated in fig. 1) deployed with a vital sign big data based health trend estimation algorithm, wherein the server is capable of training a convolutional neural network with a data matrix constructed from the training data based on the vital sign big data health trend estimation algorithm.
After training is completed, in a prediction phase, acquiring parameters such as body temperature, heart rate, respiratory rate, blood oxygen and the like of a tested person as detection data through a plurality of sensors (for example, as indicated by C in figure 1); the detected data is then input into a server (e.g., S as illustrated in fig. 1) deployed with a vital sign big data based health trend estimation algorithm, wherein the server is capable of processing the detected data with the vital sign big data based health trend estimation algorithm to generate a health trend estimation result.
Having described the general principles of the present application, various non-limiting embodiments of the present application will now be described with reference to the accompanying drawings.
An exemplary method:
fig. 2 illustrates a flow chart of a vital sign big data based health trend estimation method according to an embodiment of the present application. As shown in fig. 2, a health trend estimation method based on vital sign big data according to an embodiment of the present application includes: a training phase comprising: s110, obtaining the vital sign data of multiple aspects arranged along a time sequence in multiple time periods and constructing the vital sign data of the multiple aspects in each time period into multiple data matrixes; s120, using the data matrixes of the adjacent time periods as input matrixes for training and real values respectively to train the convolutional neural network until the parameters of the convolutional neural network are converged; and, a prediction phase comprising: s130, obtaining the vital sign data of a plurality of aspects arranged along a time sequence and constructing the vital sign data of the plurality of aspects into an input matrix; s140, inputting the input matrix into the convolutional neural network trained in the training stage to obtain a first feature map, where a scale of the first feature map is t × S × c, t is a time dimension, S is a sample dimension, and c is a number of channels of the convolutional neural network; s150, calculating Softmax-like classification function values of the respective feature matrices of the first feature map in the time dimension with respect to the first feature map as first weighting values to obtain first weighting vectors, wherein the Softmax-like classification function values are weighted sums of natural exponent function values raised to the power of negative values of feature values of respective positions in the feature matrices and divided by weighted sums of natural exponent function values raised to the power of negative values of feature values of respective positions in the first feature map; s160, calculating, as a second weighting value, a reciprocal of each feature matrix of the first feature map in a sample dimension with respect to a Softmax-like function value of the first feature map, to obtain a second weighting vector, wherein the Softmax-like function value is a weighted sum of natural exponent function values raised to powers of negative values of feature values of respective positions in the feature matrix divided by a weighted sum of natural exponent function values raised to powers of negative values of feature values of respective positions in the first feature map; s170, weighting the first feature map from a time dimension and a sample dimension by the first weighting vector and the second weighting vector respectively to obtain a second feature map; and S180, passing the second feature map through a classifier to obtain a classification result, wherein the classification result is used for representing a health trend estimation result.
Fig. 3 illustrates an architecture diagram of a training phase in a health trend estimation method based on vital sign big data according to an embodiment of the present application. As shown IN fig. 3, IN the training phase, IN the network architecture, a training data set (e.g., IN0 illustrated IN fig. 3) is first obtained, the training data set includes multiple aspects of vital sign data arranged along a time series of multiple time periods, then the multiple aspects of vital sign data of each time period of the training data set are constructed into multiple data matrices (e.g., M1 illustrated IN fig. 3), and then a convolutional neural network (e.g., CNN illustrated IN fig. 3) is trained by using the data matrices of adjacent time periods as input matrices for training (e.g., M2 illustrated IN fig. 3) and real values respectively until the parameters of the convolutional neural network converge.
Fig. 4 illustrates an architectural diagram of a prediction stage in a health trend estimation method based on vital sign big data according to an embodiment of the present application. As shown IN fig. 4, IN the prediction phase, IN the network structure, first, a training data set (e.g., IN0 as illustrated IN fig. 4) is obtained and the vital sign data of the aspects of the respective time periods of the training data set are constructed into a plurality of input matrices (e.g., M2 as illustrated IN fig. 4). Then, the input matrix is input into the convolutional neural network (e.g., CNN as illustrated in fig. 4) trained by the training phase to obtain a first feature map (e.g., F1 as illustrated in fig. 4). Next, Softmax-like classification function values of the respective feature matrices of the first feature map in the time dimension with respect to the first feature map are calculated as first weighting values to obtain a first weighting vector (e.g., V1 as illustrated in fig. 4). Next, the reciprocal of each feature matrix of the first feature map in a sample dimension with respect to the Softmax-like function value of the first feature map is calculated as a second weighting value to obtain a second weighting vector (e.g., V2 as illustrated in fig. 4). Then, the first feature map (e.g., F1 as illustrated in fig. 4) is weighted from a time dimension and a sample dimension with the first weighting vector (e.g., V1 as illustrated in fig. 4) and the second weighting vector (e.g., V2 as illustrated in fig. 4) respectively to obtain a second feature map (e.g., F2 as illustrated in fig. 4), and then the second feature map (e.g., F2 as illustrated in fig. 4) is passed through a classifier to obtain a classification result, which is used to represent a health tendency estimation result.
More specifically, in the training phase, in step S110, multiple aspects of vital sign data arranged along a time series of multiple time periods are obtained and constructed into multiple data matrices. As described above, in the technical solution of the present application, it is first required to train the deep neural network to extract the high-dimensional associated features based on the existing vital sign big data, that is, first obtain the vital sign data arranged along the time series in multiple aspects, such as body temperature data, blood oxygen data, respiration rate data, and heart rate data, and construct the data matrix.
More specifically, in the training phase, in step S120, the convolutional neural network is trained using the data matrices of adjacent time periods as the input matrix for training and the true value, respectively, until the parameters of the convolutional neural network converge. As described above, in the technical solution of the present application, the convolutional neural network is trained using vital sign big data, that is, the data matrix constructed as described above in a plurality of time periods is obtained, and the data matrices in adjacent time periods are used as an input matrix and a true value for training, respectively, to train the convolutional neural network. That is, the input data matrix and the true values of the vital sign data construct that obtains the plurality of aspects arranged along the time series are trained by the input convolutional neural network such that the neural network extracts the intra-class features of the vital sign data along the time dimension and the inter-class features along the sample dimension.
As shown in fig. 5, in the health trend estimation method based on vital sign big data according to the embodiment of the present application, the training the convolutional neural network using the data matrices of adjacent time periods as the input matrix for training and the real value respectively until the parameters of the convolutional neural network converge includes: s210, inputting the input matrix for training into a convolutional neural network to obtain a training characteristic diagram; and S220, training the convolutional neural network based on the function value of the mean square error loss between the training feature map and the real value until the parameters of the convolutional neural network are converged. That is, the convolutional neural network is first trained by an input matrix to obtain a training feature map, and then trained by a mean square error loss function of the input training feature map and a true value so that parameters of the convolutional neural network converge.
Preferably, in the embodiment of the present application, the convolutional neural network is implemented as a depth residual network. Compared with the traditional convolutional neural network, the deep residual error network is an optimized network structure provided on the basis of the traditional convolutional neural network, and mainly solves the problem that the gradient disappears in the training process. The depth residual error network introduces a residual error network structure, the network layer can be made deeper through the residual error network structure, and the problem of gradient disappearance can not occur. The residual error network uses the cross-layer link thought of a high-speed network for reference, breaks through the convention that the traditional neural network only can provide N layers as input from the input layer of the N-1 layer, enables the output of a certain layer to directly cross several layers as the input of the later layer, and has the significance of providing a new direction for the difficult problem that the error rate of the whole learning model is not reduced and inversely increased by superposing multiple layers of networks.
In summary, according to the training phase and the illustration in the health trend estimation method based on vital sign big data of the present application, in order to enable the convolutional neural network to be trained to extract features of vital sign data within a class along a time dimension and between classes along a sample dimension, in each training process, the data matrices of a plurality of time periods constructed as described above are obtained, and the data matrices of adjacent time periods are respectively used as an input matrix and a true value for training to train the convolutional neural network until the parameters of the convolutional neural network converge. After training is completed, a prediction phase is entered.
More specifically, in the prediction phase, in step S130, a plurality of aspects of vital sign data arranged along a time series are obtained and constructed as an input matrix. That is, in the prediction stage, the vital sign data of the plurality of aspects including body temperature data, blood oxygen data, respiration rate data, heart rate data, etc. arranged along time are constructed as an input matrix.
More specifically, in the prediction phase, in step S140, the input matrix is input into the convolutional neural network trained in the training phase to obtain a first feature map, where a scale of the first feature map is t × S × c, t is a time dimension, S is a sample dimension, and c is a number of channels of the convolutional neural network. That is, a first feature map is obtained by inputting the input matrix obtained in S130 into the trained convolutional neural network, and for the obtained first feature map, assuming that the time dimension is t and the sample dimension is S, the first feature map may be represented as t × S × c, where c represents a channel of the convolutional neural network.
In particular, in the embodiment of the present application, the intra-class and inter-class features of the vital sign data along the time dimension and along the sample dimension are extracted by using the concept similar to the "inter-frame distortion field" in image processing.
More specifically, in the prediction phase, Softmax-like classification function values of the respective feature matrices of the first feature map in the time dimension with respect to the first feature map are calculated as first weighting values to obtain first weighting vectors, wherein the Softmax-like classification function values are weighted sums of natural index function values that are raised to the power of negative values of feature values of the respective positions in the feature matrices and divided by weighted sums of natural index function values that are raised to the power of negative values of feature values of the respective positions in the first feature map in step S150. That is, the weight of the feature value of each position in the first feature map is adjusted using the TF-IDF concept based on the intra-class features and the inter-class features, and the Softmax-like classification function of each s × c matrix in the t dimension with respect to the first feature map is calculated as the first weighting value.
Specifically, in this embodiment of the present application, Softmax-like classification function values of respective feature matrices of the first feature map in the time dimension with respect to the first feature map are calculated as a first weighting value to obtain a first weighting vector, with the following formula: p is a radical ofa,a∈t=∑xi∈Rs*cexp(-xi)/ ∑yi∈Rt*s*cexp(-yi), wherein xi represents the eigenvalue of each position in the eigen matrix, yi represents the eigenvalue of each position in the first eigen map.
More specifically, in the prediction phase, in step S160, the reciprocal of each feature matrix of the first feature map in the sample dimension with respect to the Softmax-like function value of the first feature map is calculated as a second weighting value to obtain a second weighting vector, wherein the Softmax-like function value is a weighted sum of natural exponent function values raised to the power of negative values of feature values of respective positions in the feature matrix divided by a weighted sum of natural exponent function values raised to the power of negative values of feature values of respective positions in the first feature map. That is, the Softmax-like classification function of each s × c matrix in the t dimension with respect to the first feature map is calculated as the first weighting value, and the reciprocal of the Softmax-like classification function of each s × c matrix in the s dimension with respect to the first feature map is calculated as the second weighting value.
Specifically, in this embodiment of the present application, the reciprocal of the Softmax-like function value of each feature matrix of the first feature map in the sample dimension with respect to the second feature map is calculated as a second weighting value to obtain a second weighting vector according to the following formula: p is a radical ofb,b∈s=α*1/[∑xi∈Rt*cexp(-xi)/ ∑yi∈Rt*s*cexp(-yi)]Where α is for pb,b∈sAdjusted to [0,1 ]]Xi represents a feature value of each position in the feature matrix, and yi represents a feature value of each position in the first feature map.
More specifically, in the prediction phase, in step S170, the first feature map is weighted from the time dimension and the sample dimension with the first weighting vector and the second weighting vector, respectively, to obtain a second feature map. That is, the first feature map is weighted from the time dimension and the sample dimension with a first weighting value and a second weighting value, respectively, to obtain the second feature map, i.e., each s × c matrix in the t dimension is multiplied by the first weighting value, and each s × c matrix in the s dimension is multiplied by the second weighting value.
As shown in fig. 6, in the vital sign big data-based health trend estimation method according to the embodiment of the present application, weighting the first feature map from the time dimension and the sample dimension with the first weighting vector and the second weighting vector, respectively, to obtain a second feature map includes: s310, calculating the weight of each feature matrix of the first feature map on the time dimension by using the first weight vector; and S320, calculating the weighting of each feature matrix of the first feature map matrix weighted by the first weighting vector on a sample dimension by using the second weighting vector to obtain the second feature map.
More specifically, in the prediction phase, in step S180, the second feature map is passed through a classifier to obtain a classification result, which is used to represent a health tendency estimation result. Further, the classification result includes: the health trend becomes better, stable and worse.
In summary, the health trend estimation method based on vital sign big data according to the embodiment of the present application is illustrated, in the training process, the data matrixes in the plurality of time periods constructed as described above are obtained, and the data matrixes in the adjacent time periods are respectively used as the input matrix and the real value for training to train the convolutional neural network until the parameters of the convolutional neural network converge. In the inference stage, an input matrix is input into a trained convolutional neural network to obtain a first feature map, then a Softmax-like classification function value of each feature matrix of the first feature map in a time dimension relative to the first feature map is calculated as a first weighted value to obtain a first weighted vector, then the reciprocal of the Softmax-like classification function value of each feature matrix of the first feature map in a sample dimension relative to the first feature map is calculated as a second weighted value to obtain a second weighted vector, then the first feature map is weighted in the time dimension and the sample dimension with the first weighted vector and the second weighted vector respectively to obtain a second feature map, and finally the second feature map is passed through a classifier to obtain a classification result, wherein the classification result is used for representing a health tendency estimation result.
An exemplary system:
fig. 7 illustrates a block diagram of a vital sign big data based health trend prediction system according to an embodiment of the present application.
As shown in fig. 7, a vital sign big data-based health trend prediction system 700 according to an embodiment of the present application includes: a training module 710 comprising: a training data unit 711, configured to obtain multiple aspects of vital sign data of multiple time periods arranged along a time series and construct the multiple aspects of vital sign data of each time period into multiple data matrices; a training unit 712, configured to train the convolutional neural network by using the data matrices of the adjacent time periods obtained by the training data unit 711 as an input matrix for training and a real value respectively until parameters of the convolutional neural network converge; and, a prediction module 720, comprising: a detection data unit 721 for obtaining the vital sign data of the plurality of aspects arranged along the time series and constructing the vital sign data of the plurality of aspects as an input matrix; a first feature map generating unit 722, configured to input the input matrix obtained by the detection data unit 721 into the convolutional neural network trained in the training stage to obtain a first feature map, where a scale of the first feature map is t × s × c, t is a time dimension, s is a sample dimension, and c is a number of channels of the convolutional neural network; a first weighting vector generating unit 723 configured to calculate, as a first weighting value, a Softmax-like classification function value of each feature matrix of the first feature map generated by the first feature map generating unit 722 in a time dimension with respect to the first feature map, to obtain a first weighting vector, where the Softmax-like classification function value is a weighted sum of natural exponent function values raised to negative values of feature values of respective positions in the feature matrix divided by a weighted sum of natural exponent function values raised to negative values of feature values of respective positions in the first feature map; a second weighting vector generating unit 724 configured to calculate, as a second weighting value, an inverse of a Softmax-like function value of each feature matrix of the first feature map generated by the first feature map generating unit 722 in a sample dimension with respect to the first feature map, to obtain a second weighting vector, where the Softmax-like function value is a weighted sum of natural exponent function values raised by negative values of feature values of respective positions in the feature matrix divided by a weighted sum of natural exponent function values raised by negative values of feature values of respective positions in the first feature map; a second feature map generating unit 725 configured to weight the first feature map from a time dimension and a sample dimension with the first weighting vector generated by the first weighting vector generating unit 723 and the second weighting vector generated by the second weighting vector generating unit 724, respectively, to obtain a second feature map; and a classification result generating unit 726 for passing the second feature map generated by the second feature map generating unit 725 through a classifier to obtain a classification result, wherein the classification result is used for representing a health tendency estimation result.
In an example, in the above prediction system 700, the training unit 712 is further configured to: inputting the training input matrix into a convolutional neural network to obtain a training feature map; and training the convolutional neural network based on a mean square error loss function value between the training feature map and the real value until parameters of the convolutional neural network converge.
In an example, in the above prediction system 700, the first weighting vector generating unit 723 is further configured to: calculating Softmax-like classification function values of the respective feature matrices of the first feature map in the time dimension relative to the first feature map as first weighting values to obtain a first weighting vector, with the following formula: p is a radical of formulaa,a∈t=∑xi∈Rs*cexp(-xi)/ ∑yi∈Rt*s*cexp(-yi), wherein xi represents the eigenvalue of each position in the eigen matrix, yi represents the eigenvalue of each position in the first eigen map.
In an example, in the prediction system 700 described above, the second weight vector generating unit 724 is further configured to: calculating the reciprocal of each feature matrix of the first feature map in the sample dimension relative to the Softmax-like function value of the first feature map as a second weighting value to obtain a second weighting vector according to the following formula: p is a radical ofb,b∈s=α*1/[∑xi∈Rt*cexp(-xi)/ ∑yi∈Rt*s*cexp(-yi)]Where α is for pb,b∈sAdjusted to [0,1 ]]Xi represents a feature value of each position in the feature matrix, and yi represents a feature value of each position in the first feature map.
In one example, in the above prediction system 700, the second feature map generation unit 725 is further configured to: calculating the weight of each feature matrix of the first feature map in the time dimension with the first weight vector generated by the first weight vector generation unit 723; and calculating the weight of each feature matrix of the first feature map matrix weighted by the first weighting vector in the sample dimension by using the second weighting vector generated by the second weighting vector generation unit 724 to obtain the second feature map.
In an example, in the above prediction system 700, the classification result generating unit 726 is further configured to: passing the second feature map generated by the second feature map generation unit 725 through one or more fully-connected layers to encode the second feature map through the one or more fully-connected layers to obtain a classified feature vector; and inputting the classification feature vector into a Softmax classification function to obtain the classification result.
In an example, in the above prediction system 700, the classification result generating unit 726 is further configured to: obtaining a classification result, the classification result comprising: the health trend becomes better, stable and worse.
In one example, in the prediction system 700 described above, the convolutional neural network is a deep residual network.
Here, it can be understood by those skilled in the art that the specific functions and operations of the respective units and modules in the prediction system 700 described above have been introduced in detail in the above description of the vital sign big data based health tendency estimation method with reference to fig. 1 to 6, and thus, a repeated description thereof will be omitted.
As described above, the prediction system 700 according to the embodiment of the present application can be implemented in various terminal devices, such as a server of a health trend estimation method based on vital sign big data. In one example, the prediction system 700 according to embodiments of the present application may be integrated into a terminal device as a software module and/or a hardware module. For example, the prediction system 700 can be a software module in the operating system of the terminal device, or can be an application developed for the terminal device; of course, the prediction system 700 could equally be one of many hardware modules of the terminal device.
Alternatively, in another example, the prediction system 700 and the terminal device may be separate devices, and the prediction system 700 may be connected to the terminal device through a wired and/or wireless network and transmit the interaction information according to an agreed data format.
An exemplary electronic device:
next, an electronic apparatus according to an embodiment of the present application is described with reference to fig. 8.
FIG. 8 illustrates a block diagram of an electronic device in accordance with an embodiment of the present application.
As shown in fig. 8, the electronic device 10 includes one or more processors 11 and memory 12.
The processor 11 may be a Central Processing Unit (CPU) or other form of processing unit having data processing capabilities and/or instruction execution capabilities, and may control other components in the electronic device 10 to perform desired functions.
Memory 12 may include one or more computer program products that may include various forms of computer-readable storage media, such as volatile memory and/or non-volatile memory. The volatile memory may include, for example, Random Access Memory (RAM), cache memory (cache), and/or the like. The non-volatile memory may include, for example, Read Only Memory (ROM), hard disk, flash memory, etc. One or more computer program instructions may be stored on the computer readable storage medium and executed by the processor 11 to implement the vital sign big data-based health trend estimation method of the various embodiments of the present application described above and/or other desired functions. Various content such as training image sets, prediction results, and the like may also be stored in the computer-readable storage medium.
In one example, the electronic device 10 may further include: an input device 13 and an output device 14, which are interconnected by a bus system and/or other form of connection mechanism (not shown).
The input device 13 may include, for example, a keyboard, a mouse, and the like.
The output device 14 can output various information including prediction results to the outside. The output devices 14 may include, for example, a display, speakers, a printer, and a communication network and its connected remote output devices, among others.
Of course, for simplicity, only some of the components of the electronic device 10 relevant to the present application are shown in fig. 8, and components such as buses, input/output interfaces, and the like are omitted. In addition, the electronic device 10 may include any other suitable components depending on the particular application.
Exemplary computer program products and computer-readable storage media:
in addition to the above-described methods and devices, embodiments of the present application may also be a computer program product comprising computer program instructions which, when executed by a processor, cause the processor to perform the steps in the functions of the vital sign big data based health trend estimation method according to various embodiments of the present application described in the above-mentioned "exemplary methods" section of this specification.
The computer program product may be written with program code for performing the operations of embodiments of the present application in any combination of one or more programming languages, including an object oriented programming language such as Java, C + + or the like and conventional procedural programming languages, such as the "C" programming language or similar programming languages. The program code may execute entirely on the user's computing device, partly on the user's device, as a stand-alone software package, partly on the user's computing device and partly on a remote computing device, or entirely on the remote computing device or server.
Furthermore, embodiments of the present application may also be a computer-readable storage medium having stored thereon computer program instructions, which, when executed by a processor, cause the processor to perform the steps of the vital sign big data based health trend estimation method described in the "exemplary method" section above in this specification.
The computer-readable storage medium may take any combination of one or more readable media. The readable medium may be a readable signal medium or a readable storage medium. A readable storage medium may include, for example, but not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, or device, or any combination of the foregoing. More specific examples (a non-exhaustive list) of the readable storage medium include: an electrical connection having one or more wires, a portable disk, a hard disk, a Random Access Memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.
The foregoing describes the general principles of the present application in conjunction with specific embodiments, however, it is noted that the advantages, effects, etc. mentioned in the present application are merely examples and are not limiting, and they should not be considered essential to the various embodiments of the present application. Furthermore, the foregoing disclosure of specific details is provided for purposes of illustration and understanding only, and is not intended to limit the application to the details which are set forth in order to provide a thorough understanding of the present application.
The block diagrams of devices, apparatuses, systems referred to in this application are only given as illustrative examples and are not intended to require or imply that the connections, arrangements, configurations, etc. must be made in the manner shown in the block diagrams. These devices, apparatuses, devices, systems may be connected, arranged, configured in any manner, as will be appreciated by those skilled in the art. Words such as "including," "comprising," "having," and the like are open-ended words that mean "including, but not limited to," and are used interchangeably herein. As used herein, the words "or" and "refer to, and are used interchangeably with, the word" and/or, "unless the context clearly dictates otherwise. The word "such as" is used herein to mean, and is used interchangeably with, the phrase "such as but not limited to".
It should also be noted that in the devices, apparatuses, and methods of the present application, the components or steps may be decomposed and/or recombined. These decompositions and/or recombinations are to be considered as equivalents of the present application.
The previous 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 generic principles defined herein may be applied to other aspects without departing from the scope of the application. Thus, the present application is not intended to be limited to the aspects shown herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.
The foregoing description has been presented for purposes of illustration and description. Furthermore, the description is not intended to limit embodiments of the application to the form disclosed herein. While a number of example aspects and embodiments have been discussed above, those of skill in the art will recognize certain variations, modifications, alterations, additions and sub-combinations thereof.

Claims (9)

1. A health trend estimation method based on vital sign big data is characterized by comprising the following steps:
a training phase comprising:
obtaining a plurality of aspects of vital sign data of a plurality of time periods arranged along a time sequence and constructing the plurality of aspects of vital sign data of each time period into a plurality of data matrixes; and
using data matrixes of adjacent time periods as input matrixes for training and real values respectively to train the convolutional neural network until the parameters of the convolutional neural network are converged; and
a prediction stage:
obtaining a plurality of aspects of vital sign data arranged along a time series and constructing the plurality of aspects of vital sign data into an input matrix;
inputting the input matrix into the convolutional neural network trained in a training stage to obtain a first feature map, wherein the scale of the first feature map is t × s × c, t is a time dimension, s is a sample dimension, and c is the number of channels of the convolutional neural network;
calculating, as a first weighting value, a Softmax-like classification function value of each feature matrix of the first feature map in a time dimension with respect to the first feature map to obtain a first weighting vector, wherein the Softmax-like classification function value is a weighted sum of natural index function values raised by negative values of the feature values of each position in the feature matrix divided by a weighted sum of natural index function values raised by negative values of the feature values of each position in the first feature map;
calculating, as a second weighting value, an inverse of each feature matrix of the first feature map in a sample dimension with respect to a Softmax-like function value of the first feature map to obtain a second weighting vector, wherein the Softmax-like function value is a weighted sum of natural index function values raised by negative values of feature values of respective positions in the feature matrix divided by a weighted sum of natural index function values raised by negative values of feature values of respective positions in the first feature map;
weighting the first feature map in a time dimension and in a sample dimension with the first weighting vector and the second weighting vector, respectively, to obtain a second feature map; and
passing the second feature map through a classifier to obtain a classification result, wherein the classification result is used for representing a health trend estimation result;
wherein weighting the first feature map in a time dimension and in a sample dimension with the first weighting vector and the second weighting vector, respectively, to obtain a second feature map comprises: calculating the weight of each feature matrix of the first feature map in the time dimension by using the first weight vector; and calculating the weighting of each feature matrix of the first feature map matrix weighted by the first weighting vector on a sample dimension by the second weighting vector to obtain the second feature map.
2. The vital sign big data-based health trend estimation method according to claim 1, wherein training the convolutional neural network using data matrices of adjacent time periods as input matrices for training and real values, respectively, until parameters of the convolutional neural network converge comprises:
inputting the training input matrix into a convolutional neural network to obtain a training feature map; and
training the convolutional neural network based on a mean square error loss function value between the training feature map and the real value until parameters of the convolutional neural network converge.
3. The vital sign big data-based health trend estimation method according to claim 1, wherein calculating Softmax-like classification function values of respective feature matrices of the first feature map in a time dimension with respect to the first feature map as first weighting values to obtain a first weighting vector comprises:
calculating Softmax-like classification function values of the respective feature matrices of the first feature map in the time dimension relative to the first feature map as first weighting values to obtain a first weighting vector, with the following formula: p is a radical ofa,a∈t=∑xi∈Rs*cexp(-xi)/ ∑yi∈Rt*s*cexp (-yi), where xi represents the eigenvalue of each position in the eigen matrix, and yi represents the eigenvalue of each position in the first eigen map.
4. The vital sign big data-based health trend estimation method according to claim 3, wherein calculating an inverse of the Softmax-like function value of each feature matrix of the first feature map in a sample dimension relative to the first feature map as a second weighting value to obtain a second weighting vector comprises:
calculating the reciprocal of each feature matrix of the first feature map in sample dimension relative to the Softmax-like function value of the first feature map as a second weighting value to obtain a second weighting vector according to the following formula: p is a radical ofb,b∈s=α*1/[∑xi∈Rt*cexp(-xi)/ ∑yi∈Rt*s*cexp(-yi)]Where α is for pb,b∈sAdjustment ofTo [0,1]Xi represents a feature value of each position in the feature matrix, and yi represents a feature value of each position in the first feature map.
5. The vital sign big data-based health trend estimation method according to claim 4, wherein the second feature map is passed through a classifier to obtain a classification result, the classification result is used for representing a health trend estimation result, and the method comprises:
passing the second feature map through one or more fully-connected layers to encode the second feature map through the one or more fully-connected layers to obtain a classified feature vector; and
inputting the classification feature vector into a Softmax classification function to obtain the classification result.
6. The vital signs big data based health trend estimation method according to claim 5, wherein the classification result comprises: the health trend becomes better, stable and worse.
7. The vital sign big data-based health trend estimation method according to claim 1, wherein the convolutional neural network is a deep residual network.
8. A health trend prediction system based on vital sign big data, comprising:
a training module comprising:
the training data unit is used for obtaining a plurality of aspects of vital sign data of a plurality of time periods arranged along a time sequence and constructing the aspects of the vital sign data of each time period into a plurality of data matrixes;
the training unit is used for respectively taking the data matrixes of the adjacent time periods obtained by the data acquisition and construction unit as input matrixes for training and real values to train the convolutional neural network until the parameters of the convolutional neural network are converged; and
a prediction module comprising:
a detection data unit for obtaining a plurality of aspects of vital sign data arranged along a time series and constructing the plurality of aspects of vital sign data into an input matrix;
a first feature map generation unit, configured to input the input matrix obtained by the detection data unit into the convolutional neural network trained in the training stage to obtain a first feature map, where a scale of the first feature map is t × s × c, t is a time dimension, s is a sample dimension, and c is a number of channels of the convolutional neural network;
a first weighting vector generation unit configured to calculate, as a first weighting value, a Softmax-like classification function value of each feature matrix in a time dimension of the first feature map generated by the first feature map generation unit with respect to the first feature map, to obtain a first weighting vector, wherein the Softmax-like classification function value is a weighted sum of natural exponent function values raised to the negative values of the feature values at each position in the feature matrix divided by a weighted sum of natural exponent function values raised to the negative values of the feature values at each position in the first feature map;
a second weighting vector generation unit configured to calculate, as a second weighting value, a reciprocal of a Softmax-like function value of each feature matrix of the first feature map generated by the first feature map generation unit in a sample dimension with respect to the first feature map, to obtain a second weighting vector, where the Softmax-like function value is a weighted sum of natural exponent function values raised to a power of a negative value of the feature value at each position in the feature matrix divided by a weighted sum of natural exponent function values raised to a power of a negative value of the feature value at each position in the first feature map;
a second feature map generation unit configured to weight the first feature map from a time dimension and a sample dimension with the first weighting vector generated by the first weighting vector generation unit and the second weighting vector generated by the second weighting vector generation unit, respectively, to obtain a second feature map; and
and the classification result generating unit is used for enabling the second feature map generated by the second feature map generating unit to pass through a classifier to obtain a classification result, and the classification result is used for representing a health trend estimation result.
9. An electronic device, comprising:
a processor; and
memory in which computer program instructions are stored, which, when executed by the processor, cause the processor to carry out a vital signs big data based health trend estimation method according to any of claims 1-7.
CN202110633034.4A 2021-06-07 2021-06-07 Health trend estimation method and prediction system based on vital sign big data Active CN113393934B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202110633034.4A CN113393934B (en) 2021-06-07 2021-06-07 Health trend estimation method and prediction system based on vital sign big data

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202110633034.4A CN113393934B (en) 2021-06-07 2021-06-07 Health trend estimation method and prediction system based on vital sign big data

Publications (2)

Publication Number Publication Date
CN113393934A CN113393934A (en) 2021-09-14
CN113393934B true CN113393934B (en) 2022-07-12

Family

ID=77618477

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202110633034.4A Active CN113393934B (en) 2021-06-07 2021-06-07 Health trend estimation method and prediction system based on vital sign big data

Country Status (1)

Country Link
CN (1) CN113393934B (en)

Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107436993A (en) * 2017-05-05 2017-12-05 陈昕 Establish the method and server of ICU conditions of patients assessment models
CN109891517A (en) * 2016-10-25 2019-06-14 皇家飞利浦有限公司 The clinical diagnosis assistant of knowledge based figure
CN111028934A (en) * 2019-12-23 2020-04-17 科大讯飞股份有限公司 Diagnostic quality inspection method, diagnostic quality inspection device, electronic equipment and storage medium
CN112017771A (en) * 2020-08-31 2020-12-01 吾征智能技术(北京)有限公司 Method and system for constructing disease prediction model based on semen routine examination data
CN112204671A (en) * 2018-05-30 2021-01-08 国际商业机器公司 Personalized device recommendation for active health monitoring and management
CN112599246A (en) * 2021-03-03 2021-04-02 四川华迪信息技术有限公司 Vital sign data processing method, system, device and computer readable medium
CN112768090A (en) * 2021-01-05 2021-05-07 山东福来克思智能科技有限公司 Filtering system and method for chronic disease detection and risk assessment

Family Cites Families (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20200104641A1 (en) * 2018-09-29 2020-04-02 VII Philip Alvelda Machine learning using semantic concepts represented with temporal and spatial data

Patent Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109891517A (en) * 2016-10-25 2019-06-14 皇家飞利浦有限公司 The clinical diagnosis assistant of knowledge based figure
CN107436993A (en) * 2017-05-05 2017-12-05 陈昕 Establish the method and server of ICU conditions of patients assessment models
CN112204671A (en) * 2018-05-30 2021-01-08 国际商业机器公司 Personalized device recommendation for active health monitoring and management
CN111028934A (en) * 2019-12-23 2020-04-17 科大讯飞股份有限公司 Diagnostic quality inspection method, diagnostic quality inspection device, electronic equipment and storage medium
CN112017771A (en) * 2020-08-31 2020-12-01 吾征智能技术(北京)有限公司 Method and system for constructing disease prediction model based on semen routine examination data
CN112768090A (en) * 2021-01-05 2021-05-07 山东福来克思智能科技有限公司 Filtering system and method for chronic disease detection and risk assessment
CN112599246A (en) * 2021-03-03 2021-04-02 四川华迪信息技术有限公司 Vital sign data processing method, system, device and computer readable medium

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
An intelligent healthcare monitoring framework using wearable sensors and social networking data;Farman Ali et al.;《Future Generation Computer Systems》;20200724;全文 *
基于大数据框架的老年人心血管疾病预测系统研究与实现;周晨晨;《中国优秀硕士学位论文全文数据库-医药卫生科技专辑》;20210515;全文 *

Also Published As

Publication number Publication date
CN113393934A (en) 2021-09-14

Similar Documents

Publication Publication Date Title
CN112199717B (en) Privacy model training method and device based on small amount of public data
Park et al. Tackling overfitting in boosting for noisy healthcare data
Zhou et al. Optimizing autoencoders for learning deep representations from health data
CN114548246A (en) Control method and system of intelligent device and electronic device
KR102423483B1 (en) An artificial intelligence system that provides predictive information on depression based on EEG data
Yu et al. Semi-supervised learning and data augmentation in wearable-based momentary stress detection in the wild
US20230317257A1 (en) Method and system for predicting needs of patient for hospital resources
CN114566190A (en) Early warning method and system based on voice characteristics and electronic equipment thereof
Mishra et al. Cognitive inference device for activity supervision in the elderly
Zhang et al. Developing smart buildings to reduce indoor risks for safety and health of the elderly: A systematic and bibliometric analysis
Shah et al. Evaluating contrastive learning on wearable timeseries for downstream clinical outcomes
CN113393934B (en) Health trend estimation method and prediction system based on vital sign big data
CN116098595B (en) System and method for monitoring and preventing sudden cardiac death and sudden cerebral death
CN110598789A (en) Human fatigue state prediction method and system based on fuzzy perceptron
US20220375618A1 (en) Method and apparatus of calculating comprehensive disease index
Zaki et al. Smart medical chatbot with integrated contactless vital sign monitor
Chien et al. Continuous-time self-attention in neural differential equation
CN113892910A (en) Internet of things-based intelligent closestool old people monitoring method
CN113855008A (en) Wearable body fluid monitoring device and operation method thereof
Villmann Neural networks approaches in medicine-a review of actual developments.
CN117393153B (en) Shock real-time risk early warning and monitoring method and system based on medical internet of things time sequence data and deep learning algorithm
KR102461646B1 (en) Method of generating augmented data based on transfer learning for EEG data
Zhu et al. Comparison of parametric and non‐parametric Bayesian inference for fusing sensory estimates in physiological time‐series analysis
Sweeney-Fanelli et al. Automated Emotion Recognition Employing Wearable ECG Sensor and Deep-Learning
WO2024014875A1 (en) Method, program and device for improving accuracy of feature extraction for heart rate variability

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant
PE01 Entry into force of the registration of the contract for pledge of patent right

Denomination of invention: Health trend estimation method and prediction system based on vital signs big data

Effective date of registration: 20221123

Granted publication date: 20220712

Pledgee: Xiaoshan sub branch of Hangzhou United Rural Commercial Bank Co.,Ltd.

Pledgor: Yijin (Hangzhou) Health Technology Co.,Ltd.

Registration number: Y2022980023058

PE01 Entry into force of the registration of the contract for pledge of patent right