CN114420292A - Medical health dynamic prediction method, system and equipment - Google Patents
Medical health dynamic prediction method, system and equipment Download PDFInfo
- Publication number
- CN114420292A CN114420292A CN202210045020.5A CN202210045020A CN114420292A CN 114420292 A CN114420292 A CN 114420292A CN 202210045020 A CN202210045020 A CN 202210045020A CN 114420292 A CN114420292 A CN 114420292A
- Authority
- CN
- China
- Prior art keywords
- data
- user
- medical
- medical health
- historical
- 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.)
- Pending
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
-
- 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
- G16H50/20—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for computer-aided diagnosis, e.g. based on medical expert systems
Landscapes
- Health & Medical Sciences (AREA)
- Engineering & Computer Science (AREA)
- Medical Informatics (AREA)
- Public Health (AREA)
- Biomedical Technology (AREA)
- Data Mining & Analysis (AREA)
- Databases & Information Systems (AREA)
- Pathology (AREA)
- Epidemiology (AREA)
- General Health & Medical Sciences (AREA)
- Primary Health Care (AREA)
- Measuring And Recording Apparatus For Diagnosis (AREA)
Abstract
The invention discloses a dynamic prediction method, a system and equipment for medical health, comprising the following steps: constructing a medical health data set, and processing to obtain a plurality of characteristic data sequences; constructing a medical health dynamic prediction model based on a gate control circulation unit and a knowledge attention mechanism, and inputting a plurality of characteristic data sequences into the medical health dynamic prediction model for training; acquiring a face image of a user, identifying the identity of the user, and extracting historical medical data of the user; and inputting the historical medical data of the user into a medical health dynamic prediction model, predicting the health condition of the user and outputting a prediction result. The invention can calibrate the feature importance of the feature data to output the interpretable factor, and further improves the accuracy and the interpretability of the prediction model by extracting the relation between medical knowledge and the user diagnosis and treatment feature vector.
Description
Technical Field
The invention belongs to the technical field, and particularly relates to a dynamic prediction method, a dynamic prediction system and dynamic prediction equipment for medical health.
Background
With the continuous development of scientific technology, big data and cloud computing are generally applied to various industries, great convenience is brought to the life of people, the healthy big data are new terms appearing with digital wave and information modernization in recent years, the purpose of the big data is to perform specialized processing and recycling on the healthy data, and the big data has positive significance on physical condition monitoring, disease prevention, health trend analysis and prediction.
The method for predicting the medical health data of the user by utilizing the machine learning technology is a current mainstream analysis method, scores the health state of the user based on the collected user data, and further judges whether the user is in an abnormal health state based on the score, for example, if the score is lower than a preset score threshold value, the user is judged to be in the abnormal health state; another example is: the relevance between the historical health data of the user is analyzed from the time dimension based on a deep learning technology, so that the health condition of the user is predicted. However, the above method for predicting the health condition of the user based on only the learning of the machine learning model has the problems of low confidence and interpretability.
Disclosure of Invention
The invention aims to provide a medical health dynamic prediction method, a medical health dynamic prediction system and medical health dynamic prediction equipment, which are used for solving at least one technical problem in the prior art.
In order to achieve the purpose, the invention adopts the following technical scheme:
in a first aspect, a medical health dynamic prediction method includes:
constructing a medical health data set, and processing the medical health data set to obtain a plurality of characteristic data sequences;
constructing a medical health dynamic prediction model based on a gate control circulation unit and a knowledge attention mechanism, and inputting a plurality of characteristic data sequences into the medical health dynamic prediction model for training;
acquiring a face image of a user, identifying the identity of the user according to the face image, and extracting historical medical data of the user according to an identification result;
and inputting the historical medical data of the user into the medical health dynamic prediction model, predicting the health condition of the user and outputting a prediction result.
In one possible design, constructing a medical health data set includes:
the method comprises the steps of obtaining baseline data, follow-up visit data and historical visit data of a plurality of users, and processing the baseline data, the follow-up visit data and the historical visit data to obtain a medical health data set.
In one possible design, the processing of the baseline data, the follow-up data, and the historical visit data results in a medical health data set comprising:
data aggregation and normalization are performed on the baseline data, the follow-up data, and the historical visit data to form structured and unstructured medical health data sets.
In one possible design, the processing of the medical health data set results in a number of characteristic data sequences, including:
arranging the historical follow-up data and the historical visit data of each user into a first medical record sequence according to time, and combining the first medical record sequence and the baseline data of the user into a second medical record sequence;
and constructing a characteristic matrix based on the time dimension according to the plurality of second medical record sequences, wherein the characteristic matrix comprises a plurality of characteristic data sequences.
In one possible design, a health dynamics prediction model based on a gated loop unit and a knowledge attention mechanism is constructed, including:
constructing a multi-channel gate control circulation unit, and respectively learning different characteristic data sequences by using the gate control circulation unit to obtain diagnosis and treatment characteristic hidden vectors corresponding to each user;
constructing a knowledge attention mechanism based on a Transformer, selectively aggregating the diagnosis and treatment characteristic vectors and medical knowledge, acquiring attention weights between the medical knowledge and baseline data and follow-up data of a user, and acquiring knowledge attention characteristic vectors;
and constructing a model output layer based on softmax, performing feature fusion on the diagnosis and treatment feature implicit vector and the knowledge attention feature vector, and inputting the fused feature vector to the model output layer for output.
In one possible design, the learning of different characteristic data sequences by using the gating cycle unit to obtain diagnosis and treatment characteristic hidden vectors corresponding to each user includes:
repeatedly receiving characteristic data sequences corresponding to historical diagnosis and treatment time steps of a user by using the gating cycle unit, and mapping each characteristic data sequence into an abstract first diagnosis and treatment characteristic hidden vector of hidden layer space coding according to historical medical health data coding;
and setting the influence of the historical hidden layer space coding result on the current diagnosis and treatment time step by utilizing an updating gate and a re-gating of the gating circulation unit to obtain a second diagnosis and treatment characteristic hidden vector.
In one possible design, acquiring a face image of a user, identifying the identity of the user according to the face image, and extracting historical medical data of the user according to an identification result, includes:
acquiring a face image of a user, and extracting facial features in the face image according to a preconfigured convolutional neural network;
according to the extracted facial features, the identity of the user is identified;
and extracting the historical medical data of the user from the medical database according to the identification result.
In a second aspect, the present invention provides a dynamic prediction system for medical health, comprising:
the data set construction module is used for constructing a medical health data set and processing the medical health data set to obtain a plurality of characteristic data sequences;
the prediction model training module is used for constructing a medical health dynamic prediction model based on a gating cycle unit and a knowledge attention mechanism, and inputting a plurality of characteristic data sequences into the medical health dynamic prediction model for training;
the data extraction module is used for acquiring a face image of a user, identifying the identity of the user according to the face image and extracting historical medical data of the user according to an identification result;
and the health condition prediction module is used for inputting the historical medical data of the user into the medical health dynamic prediction model, predicting the health condition of the user and outputting a prediction result.
In one possible design, in constructing the medical health data set, the data set construction module is specifically configured to:
the method comprises the steps of obtaining baseline data, follow-up visit data and historical visit data of a plurality of users, and processing the baseline data, the follow-up visit data and the historical visit data to obtain a medical health data set.
In one possible design, when the baseline data, the follow-up data, and the historical visit data are processed to obtain a medical health data set, the data set construction module is specifically configured to:
data aggregation and normalization are performed on the baseline data, the follow-up data, and the historical visit data to form structured and unstructured medical health data sets.
In one possible design, when the medical health data set is processed to obtain a plurality of characteristic data sequences, the data set construction module is specifically configured to:
arranging the historical follow-up data and the historical visit data of each user into a first medical record sequence according to time, and combining the first medical record sequence and the baseline data of the user into a second medical record sequence;
and constructing a characteristic matrix based on the time dimension according to the plurality of second medical record sequences, wherein the characteristic matrix comprises a plurality of characteristic data sequences.
In one possible design, when constructing a health dynamics prediction model based on a gated loop unit and a knowledge attention mechanism, the prediction model training module is specifically configured to:
constructing a multi-channel gate control circulation unit, and respectively learning different characteristic data sequences by using the gate control circulation unit to obtain diagnosis and treatment characteristic hidden vectors corresponding to each user;
constructing a knowledge attention mechanism based on a Transformer, selectively aggregating the diagnosis and treatment characteristic vectors and medical knowledge, acquiring attention weights between the medical knowledge and baseline data and follow-up data of a user, and acquiring knowledge attention characteristic vectors;
and constructing a model output layer based on softmax, performing feature fusion on the diagnosis and treatment feature implicit vector and the knowledge attention feature vector, and inputting the fused feature vector to the model output layer for output.
In a possible design, when the gating cycle unit is used to learn different feature data sequences respectively to obtain a diagnosis and treatment feature hidden vector corresponding to each user, the prediction model training module is specifically configured to:
repeatedly receiving characteristic data sequences corresponding to historical diagnosis and treatment time steps of a user by using the gating cycle unit, and mapping each characteristic data sequence into an abstract first diagnosis and treatment characteristic hidden vector of hidden layer space coding according to historical medical health data coding;
and setting the influence of the historical hidden layer space coding result on the current diagnosis and treatment time step by utilizing an updating gate and a re-gating of the gating circulation unit to obtain a second diagnosis and treatment characteristic hidden vector.
In one possible design, the data extraction module includes:
the facial feature extraction unit is used for acquiring a face image of a user and extracting facial features in the face image according to a preconfigured convolutional neural network;
the identity recognition unit is used for recognizing the identity of the user according to the extracted facial features;
and the data extraction unit is used for extracting the historical medical data of the user from the medical database according to the identification result.
In a third aspect, the present invention provides a computer device, comprising a memory, a processor and a transceiver, which are sequentially connected in communication, wherein the memory is used for storing a computer program, the transceiver is used for sending and receiving messages, and the processor is used for reading the computer program and executing the dynamic prediction method for medical health as described in any one of the possible designs of the first aspect.
In a fourth aspect, the present invention provides a computer-readable storage medium having stored thereon instructions for executing the method for dynamic prediction of medical health as set forth in any one of the possible designs of the first aspect when the instructions are run on a computer.
In a fifth aspect, the present invention provides a computer program product comprising instructions which, when run on a computer, cause the computer to perform a method of dynamic prediction of medical health as set forth in any one of the possible designs of the first aspect.
Has the advantages that:
the method comprises the steps of constructing a medical health data set, and processing the medical health data set to obtain a plurality of characteristic data sequences; then constructing a medical health dynamic prediction model based on a gating cycle unit and a knowledge attention mechanism, and inputting a plurality of characteristic data sequences into the medical health dynamic prediction model for training; then acquiring a face image of the user, identifying the identity of the user according to the face image, and extracting historical medical data of the user according to an identification result; and then inputting the historical medical data of the user into the medical health dynamic prediction model, predicting the health condition of the user and outputting a prediction result. According to the medical health characteristic data sequence of different users, the medical health characteristic data sequence of the user is processed, the characteristic data context information of the user is adaptively captured through a gate control circulation unit, and characteristic importance calibration is carried out on the characteristic data to output interpretable factors; and selectively aggregating the user diagnosis and treatment feature vectors and the medical knowledge by using a knowledge attention mechanism, and acquiring attention weights between the medical knowledge and baseline data and follow-up data of the user to extract the relation between the medical knowledge and the user diagnosis and treatment feature vectors, so that the accuracy and the interpretability of a prediction model are further improved, possible common diseases of the user can be managed and predicted by using the prediction model, and medication suggestions are provided for the user by combining the medical knowledge, so that the real-time monitoring and management of the health of the user are realized.
Drawings
Fig. 1 is a flowchart of a medical health dynamic prediction method in this embodiment.
Detailed Description
In order to make the objects, technical solutions and advantages of the embodiments of the present disclosure more clear, the technical solutions in the embodiments of the present disclosure will be clearly and completely described below with reference to the drawings in the embodiments of the present disclosure, and it is obvious that the described embodiments are some, but not all embodiments of the present disclosure. All other embodiments, which can be obtained by a person skilled in the art without inventive step based on the embodiments in the present description, belong to the protection scope of the present invention.
In order to solve the technical problems of low confidence coefficient and interpretability of a method for predicting the health condition of a user only based on the learning of a machine learning model in the prior art, the application provides a dynamic prediction method for medical health, the method processes a medical health characteristic data sequence of the user according to the conditions of different users, adaptively captures the characteristic data context information of the user through a gating cycle unit, and calibrates the characteristic importance of the characteristic data to output an interpretable factor; and selectively aggregating the user diagnosis and treatment feature vectors and the medical knowledge by using a knowledge attention mechanism, and acquiring attention weights between the medical knowledge and baseline data and follow-up data of the user to extract the relation between the medical knowledge and the user diagnosis and treatment feature vectors, so that the accuracy and the interpretability of a prediction model are further improved, possible common diseases of the user can be managed and predicted by using the prediction model, and medication suggestions are provided for the user by combining the medical knowledge, so that the real-time monitoring and management of the health of the user are realized. The present application will be specifically described below with reference to specific examples.
Examples
As shown in fig. 1, in a first aspect, the present embodiment provides a method for dynamically predicting medical health, including but not limited to steps S101 to S104:
s101, constructing a medical health data set, and processing the medical health data set to obtain a plurality of characteristic data sequences;
in a specific embodiment of step S101, constructing a medical health data set comprises:
the method comprises the steps of obtaining baseline data, follow-up visit data and historical visit data of a plurality of users, and processing the baseline data, the follow-up visit data and the historical visit data to obtain a medical health data set.
It should be noted that the baseline data includes: (1) and (4) a user's enrollment exclusion process. Roughly delineating the analyzed population by using an inclusion standard, and correcting the analyzed population by using an exclusion standard; (2) description and comparison of the user's baseline characteristics. Baseline characteristics often include social demographics, clinical characteristics, laboratory test indicators, and history of disease and medication. The follow-up visit data refers to data generated by the process that the patient who has been visited in the hospital is communicated or in other ways to know the disease condition change of the patient regularly and guide the patient to recover. The historical clinic data is the clinic data remained after the user establishes a file for clinic, and can be specifically acquired from the information systems of all hospitals.
In a specific embodiment of step S101, processing the baseline data and the follow-up visit data to obtain a medical health data set includes:
data aggregation and normalization are performed on the baseline data, the follow-up data, and the historical visit data to form structured and unstructured medical health data sets.
Specifically, the baseline data, the follow-up visit data and the historical visit data are sorted to obtain data to be processed with the same format, and the data to be processed is processed, wherein the data to be processed comprises data cleaning, data interpolation, data conversion, data standardization and/or data verification; and setting data specifications, such as a data name format, a data storage unit length and the like, so as to assemble and integrate the baseline data, the follow-up data and the historical visit data to form standardized data.
In a specific embodiment of step S101, processing the medical health data set to obtain a plurality of characteristic data sequences includes:
arranging the historical follow-up data and the historical visit data of each user into a first medical record sequence according to time, and combining the first medical record sequence and the baseline data of the user into a second medical record sequence;
and constructing a characteristic matrix based on the time dimension according to the plurality of second medical record sequences, wherein the characteristic matrix comprises a plurality of characteristic data sequences.
Based on the above disclosure, the historical follow-up data and the historical visit data of each user are arranged into a first medical record sequence according to time, the first medical record sequence and the baseline data of the user are combined into a second medical record sequence, a feature matrix based on a time dimension is constructed according to a plurality of second medical record sequences, and the feature matrix comprises a plurality of feature data sequences, so that the model can enhance the attention degree of the baseline data after the feature data sequences are input into the dynamic prediction model, so that the model is closer to a real medical scene, and the accuracy of model prediction is provided.
S102, constructing a medical health dynamic prediction model based on a gate control cycle unit and a knowledge attention mechanism, and inputting a plurality of characteristic data sequences into the medical health dynamic prediction model for training;
in a specific embodiment of step S102, constructing a health dynamics prediction model based on a gated loop unit and a knowledge attention mechanism includes:
s1021, constructing a multi-channel gate control circulation unit, and learning different characteristic data sequences by using the gate control circulation unit respectively to obtain diagnosis and treatment characteristic hidden vectors corresponding to each user;
specifically, the method for learning different characteristic data sequences by using the gating cycle unit to obtain diagnosis and treatment characteristic hidden vectors corresponding to each user includes:
repeatedly receiving characteristic data sequences corresponding to historical diagnosis and treatment time steps of a user by using the gating cycle unit, and mapping each characteristic data sequence into an abstract first diagnosis and treatment characteristic hidden vector of hidden layer space coding according to historical medical health data coding;
and setting the influence of the historical hidden layer space coding result on the current diagnosis and treatment time step by utilizing an updating gate and a re-gating of the gating circulation unit to obtain a second diagnosis and treatment characteristic hidden vector.
Based on the above disclosure, the embodiment utilizes the multi-channel coding framework of the gated cyclic unit to learn different feature data sequences respectively, so as to obtain diagnosis and treatment feature hidden vectors corresponding to each user, wherein the multi-channel coding framework can display and capture the relevance between features from the global view of the complete feature data sequence, and screen the most important features. Preferably, when the multi-channel coding part uses the cyclic neural network to code the dynamic data characteristic sequence, the multi-layer perceptron hidden layer neural network can be used to map the static information to obtain a denser intermediate representation which is input into the subsequent knowledge attention mechanism network.
S1022, a knowledge attention mechanism based on a Transformer is constructed, the diagnosis and treatment feature vectors and medical knowledge are selectively aggregated, attention weights among the medical knowledge, baseline data and follow-up data of a user are obtained, and the knowledge attention feature vectors are obtained; therefore, the knowledge attention mechanism is utilized to selectively aggregate the diagnosis and treatment feature vectors of the user with the medical knowledge, the attention weight between the medical knowledge and the baseline data and the follow-up data of the user is obtained, the relation between the medical knowledge and the diagnosis and treatment feature vectors of the user is extracted, and the accuracy and the interpretability of the prediction model are further improved.
And S1023, constructing a model output layer based on softmax, performing feature fusion on the diagnosis and treatment feature implicit vectors and the knowledge attention feature vectors, and inputting the fused feature vectors to the model output layer for output. Based on the diagnosis and treatment characteristic implicit vector and the knowledge attention characteristic vector, a prediction result with high accuracy can be output according to the influence of historical characteristic data on current time step output.
S103, acquiring a face image of a user, identifying the identity of the user according to the face image, and extracting historical medical data of the user according to an identification result;
in a specific implementation manner of step S103, acquiring a face image of a user, identifying a user identity according to the face image, and extracting historical medical data of the user according to an identification result, includes:
step S1031, acquiring a face image of a user, and extracting facial features in the face image according to a preconfigured convolutional neural network;
s1032, identifying the identity of the user according to the extracted facial features;
and S1033, extracting historical medical data of the user from the medical database according to the identification result.
Based on the above disclosure, based on the face image of the user, the facial features are extracted through the convolutional neural network, and historical medical data of the user is obtained based on the facial features, so that the user can automatically adopt the face scanning device to perform self-test on the health state without testing by a special testing mechanism, the operation is simple, and the user can conveniently monitor the self health condition in daily life.
And S104, inputting the historical medical data of the user into the medical health dynamic prediction model, predicting the health condition of the user and outputting a prediction result.
Based on the above disclosure, in the embodiment, a medical health data set is constructed, and the medical health data set is processed to obtain a plurality of characteristic data sequences; then constructing a medical health dynamic prediction model based on a gating cycle unit and a knowledge attention mechanism, and inputting a plurality of characteristic data sequences into the medical health dynamic prediction model for training; then acquiring a face image of the user, identifying the identity of the user according to the face image, and extracting historical medical data of the user according to an identification result; and then inputting the historical medical data of the user into the medical health dynamic prediction model, predicting the health condition of the user and outputting a prediction result. According to the medical health characteristic data sequence of different users, the medical health characteristic data sequence of the user is processed, the characteristic data context information of the user is adaptively captured through a gate control circulation unit, and characteristic importance calibration is carried out on the characteristic data to output interpretable factors; and selectively aggregating the user diagnosis and treatment feature vectors and the medical knowledge by using a knowledge attention mechanism, and acquiring attention weights between the medical knowledge and baseline data and follow-up data of the user to extract the relation between the medical knowledge and the user diagnosis and treatment feature vectors, so that the accuracy and the interpretability of a prediction model are further improved, possible common diseases of the user can be managed and predicted by using the prediction model, and medication suggestions are provided for the user by combining the medical knowledge, so that the real-time monitoring and management of the health of the user are realized.
In a second aspect, the present invention provides a dynamic prediction system for medical health, comprising:
the data set construction module is used for constructing a medical health data set and processing the medical health data set to obtain a plurality of characteristic data sequences;
the prediction model training module is used for constructing a medical health dynamic prediction model based on a gating cycle unit and a knowledge attention mechanism, and inputting a plurality of characteristic data sequences into the medical health dynamic prediction model for training;
the data extraction module is used for acquiring a face image of a user, identifying the identity of the user according to the face image and extracting historical medical data of the user according to an identification result;
and the health condition prediction module is used for inputting the historical medical data of the user into the medical health dynamic prediction model, predicting the health condition of the user and outputting a prediction result.
In one possible design, in constructing the medical health data set, the data set construction module is specifically configured to:
the method comprises the steps of obtaining baseline data, follow-up visit data and historical visit data of a plurality of users, and processing the baseline data, the follow-up visit data and the historical visit data to obtain a medical health data set.
In one possible design, when the medical health data set is obtained after processing the baseline data and the follow-up data, the data set construction module is specifically configured to:
data aggregation and normalization are performed on the baseline data and the follow-up data and historical visit data to form structured and unstructured medical health data sets.
In one possible design, when the medical health data set is processed to obtain a plurality of characteristic data sequences, the data set construction module is specifically configured to:
arranging the historical follow-up data of each user into a first medical record sequence according to time, and combining the first medical record sequence and the baseline data of the user into a second medical record sequence;
and constructing a characteristic matrix based on the time dimension according to the plurality of second medical record sequences, wherein the characteristic matrix comprises a plurality of characteristic data sequences.
In one possible design, when constructing a health dynamics prediction model based on a gated loop unit and a knowledge attention mechanism, the prediction model training module is specifically configured to:
constructing a multi-channel gate control circulation unit, and respectively learning different characteristic data sequences by using the gate control circulation unit to obtain diagnosis and treatment characteristic hidden vectors corresponding to each user;
constructing a knowledge attention mechanism based on a Transformer, selectively aggregating the diagnosis and treatment characteristic vectors and medical knowledge, acquiring attention weights between the medical knowledge and baseline data and follow-up data of a user, and acquiring knowledge attention characteristic vectors;
and constructing a model output layer based on softmax, performing feature fusion on the diagnosis and treatment feature implicit vector and the knowledge attention feature vector, and inputting the fused feature vector to the model output layer for output.
In a possible design, when the gating cycle unit is used to learn different feature data sequences respectively to obtain a diagnosis and treatment feature hidden vector corresponding to each user, the prediction model training module is specifically configured to:
repeatedly receiving characteristic data sequences corresponding to historical diagnosis and treatment time steps of a user by using the gating cycle unit, and mapping each characteristic data sequence into an abstract first diagnosis and treatment characteristic hidden vector of hidden layer space coding according to historical medical health data coding;
and setting the influence of the historical hidden layer space coding result on the current diagnosis and treatment time step by utilizing an updating gate and a re-gating of the gating circulation unit to obtain a second diagnosis and treatment characteristic hidden vector.
In one possible design, the data extraction module includes:
the facial feature extraction unit is used for acquiring a face image of a user and extracting facial features in the face image according to a preconfigured convolutional neural network;
the identity recognition unit is used for recognizing the identity of the user according to the extracted facial features;
and the data extraction unit is used for extracting the historical medical data of the user from the medical database according to the identification result.
In a third aspect, the present invention provides a computer device, comprising a memory, a processor and a transceiver, which are sequentially connected in communication, wherein the memory is used for storing a computer program, the transceiver is used for sending and receiving messages, and the processor is used for reading the computer program and executing the dynamic prediction method for medical health as described in any one of the possible designs of the first aspect.
In a fourth aspect, the present invention provides a computer-readable storage medium having stored thereon instructions for executing the method for dynamic prediction of medical health as set forth in any one of the possible designs of the first aspect when the instructions are run on a computer.
In a fifth aspect, the present invention provides a computer program product comprising instructions which, when run on a computer, cause the computer to perform a method of dynamic prediction of medical health as set forth in any one of the possible designs of the first aspect.
Finally, it should be noted that: the above description is only a preferred embodiment of the present invention, and is not intended to limit the scope of the present invention. Any modification, equivalent replacement, or improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention.
Claims (9)
1. A medical health dynamic prediction method, comprising:
constructing a medical health data set, and processing the medical health data set to obtain a plurality of characteristic data sequences;
constructing a medical health dynamic prediction model based on a gate control circulation unit and a knowledge attention mechanism, and inputting a plurality of characteristic data sequences into the medical health dynamic prediction model for training;
acquiring a face image of a user, identifying the identity of the user according to the face image, and extracting historical medical data of the user according to an identification result;
and inputting the historical medical data of the user into the medical health dynamic prediction model, predicting the health condition of the user and outputting a prediction result.
2. The dynamic medical health prediction method of claim 1, wherein constructing a medical health data set comprises:
the method comprises the steps of obtaining baseline data, follow-up visit data and historical visit data of a plurality of users, and processing the baseline data, the follow-up visit data and the historical visit data to obtain a medical health data set.
3. The dynamic prediction of medical health as claimed in claim 2, wherein the processing of the baseline data, follow-up data and historical visit data to obtain a medical health data set comprises:
data aggregation and normalization are performed on the baseline data, the follow-up data, and the historical visit data to form structured and unstructured medical health data sets.
4. The dynamic medical health prediction method according to claim 2 or 3, wherein the processing of the medical health data set to obtain a plurality of characteristic data sequences comprises:
arranging the historical follow-up data and the historical visit data of each user into a first medical record sequence according to time, and combining the first medical record sequence and the baseline data of the user into a second medical record sequence;
and constructing a characteristic matrix based on the time dimension according to the plurality of second medical record sequences, wherein the characteristic matrix comprises a plurality of characteristic data sequences.
5. The dynamic prediction method for medical health according to claim 4, wherein constructing a dynamic prediction model for health based on a gated loop unit and a knowledge attention mechanism comprises:
constructing a multi-channel gate control circulation unit, and respectively learning different characteristic data sequences by using the gate control circulation unit to obtain diagnosis and treatment characteristic hidden vectors corresponding to each user;
constructing a knowledge attention mechanism based on a Transformer, selectively aggregating the diagnosis and treatment characteristic vectors and medical knowledge, acquiring attention weights between the medical knowledge and baseline data and follow-up data of a user, and acquiring knowledge attention characteristic vectors;
and constructing a model output layer based on softmax, performing feature fusion on the diagnosis and treatment feature implicit vector and the knowledge attention feature vector, and inputting the fused feature vector to the model output layer for output.
6. The method according to claim 5, wherein the step of learning different characteristic data sequences by using the gated cycle unit to obtain diagnosis and treatment characteristic hidden vectors corresponding to each user comprises:
repeatedly receiving characteristic data sequences corresponding to historical diagnosis and treatment time steps of a user by using the gating cycle unit, and mapping each characteristic data sequence into an abstract first diagnosis and treatment characteristic hidden vector of hidden layer space coding according to historical medical health data coding;
and setting the influence of the historical hidden layer space coding result on the current diagnosis and treatment time step by utilizing an updating gate and a re-gating of the gating circulation unit to obtain a second diagnosis and treatment characteristic hidden vector.
7. The medical health dynamic prediction method of claim 1, wherein obtaining a face image of a user, recognizing the identity of the user according to the face image, and extracting historical medical data of the user according to the recognition result comprises:
acquiring a face image of a user, and extracting facial features in the face image according to a preconfigured convolutional neural network;
according to the extracted facial features, the identity of the user is identified;
and extracting the historical medical data of the user from the medical database according to the identification result.
8. A medical health dynamics prediction system, comprising:
the data set construction module is used for constructing a medical health data set and processing the medical health data set to obtain a plurality of characteristic data sequences;
the prediction model training module is used for constructing a medical health dynamic prediction model based on a gating cycle unit and a knowledge attention mechanism, and inputting a plurality of characteristic data sequences into the medical health dynamic prediction model for training;
the data extraction module is used for acquiring a face image of a user, identifying the identity of the user according to the face image and extracting historical medical data of the user according to an identification result;
and the health condition prediction module is used for inputting the historical medical data of the user into the medical health dynamic prediction model, predicting the health condition of the user and outputting a prediction result.
9. A computer device comprising a memory, a processor and a transceiver communicatively connected in sequence, wherein the memory is used for storing a computer program, the transceiver is used for transmitting and receiving messages, and the processor is used for reading the computer program and executing the dynamic prediction method for medical health according to any one of claims 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210045020.5A CN114420292A (en) | 2022-01-14 | 2022-01-14 | Medical health dynamic prediction method, system and equipment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210045020.5A CN114420292A (en) | 2022-01-14 | 2022-01-14 | Medical health dynamic prediction method, system and equipment |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114420292A true CN114420292A (en) | 2022-04-29 |
Family
ID=81273167
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210045020.5A Pending CN114420292A (en) | 2022-01-14 | 2022-01-14 | Medical health dynamic prediction method, system and equipment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114420292A (en) |
-
2022
- 2022-01-14 CN CN202210045020.5A patent/CN114420292A/en active Pending
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106934235B (en) | Patient's similarity measurement migratory system between a kind of disease areas based on transfer learning | |
| CN108899064A (en) | Electronic health record generation method, device, computer equipment and storage medium | |
| CN113241135A (en) | Disease risk prediction method and system based on multi-mode fusion | |
| CN110051324B (en) | Method and system for predicting death rate of acute respiratory distress syndrome | |
| CN111145910A (en) | Abnormal case identification method and device based on artificial intelligence and computer equipment | |
| WO2021151295A1 (en) | Method, apparatus, computer device, and medium for determining patient treatment plan | |
| CN111738302A (en) | System for classifying and diagnosing Alzheimer disease based on multi-modal data | |
| CN109543718B (en) | Method and device for modifying disease type description | |
| CN113657548A (en) | Medical insurance abnormity detection method and device, computer equipment and storage medium | |
| WO2020049404A2 (en) | Visualization of social determinants of health | |
| Skitsan et al. | Evaluation of the Informative Features of Cardiac Studies Diagnostic Data using the Kullback Method. | |
| CN114343585B (en) | Cognitive and behavioral disorder early warning method, device, equipment and storage medium | |
| CN117542474A (en) | Remote nursing monitoring system and method based on big data | |
| WO2023217737A1 (en) | Health data enrichment for improved medical diagnostics | |
| CN116884636A (en) | Infectious disease data analysis method, infectious disease data analysis device, computer equipment and storage medium | |
| CN114203312B (en) | Digital medical service analysis method and server combined with big data intelligent medical treatment | |
| CN114420292A (en) | Medical health dynamic prediction method, system and equipment | |
| CN115036034A (en) | Similar patient identification method and system based on patient characterization map | |
| CN113990514A (en) | Abnormality detection device for doctor diagnosis and treatment behavior, computer device and storage medium | |
| CN113724853A (en) | Intelligent medical system based on deep learning | |
| AU2021102832A4 (en) | System & method for automatic health prediction using fuzzy based machine learning | |
| Ojetunmibi et al. | Pneumonia disease detection and classification system using naive Bayesian technique | |
| CN117409978B (en) | Disease prediction model construction method, system, device and readable storage medium | |
| CN113223698B (en) | Emergency hierarchical processing method and device, electronic equipment and storage medium | |
| CN117235487B (en) | Feature extraction method and system for predicting hospitalization event of asthma patient |
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 |