CN114203295B - Cerebral apoplexy risk prediction intervention method and system - Google Patents

Cerebral apoplexy risk prediction intervention method and system Download PDF

Info

Publication number
CN114203295B
CN114203295B CN202111395059.1A CN202111395059A CN114203295B CN 114203295 B CN114203295 B CN 114203295B CN 202111395059 A CN202111395059 A CN 202111395059A CN 114203295 B CN114203295 B CN 114203295B
Authority
CN
China
Prior art keywords
cnn
neural network
lstm
module
feature
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
CN202111395059.1A
Other languages
Chinese (zh)
Other versions
CN114203295A (en
Inventor
张静莎
李增勇
张腾宇
李慧媛
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.)
National Research Center for Rehabilitation Technical Aids
Original Assignee
National Research Center for Rehabilitation Technical Aids
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 National Research Center for Rehabilitation Technical Aids filed Critical National Research Center for Rehabilitation Technical Aids
Priority to CN202111395059.1A priority Critical patent/CN114203295B/en
Publication of CN114203295A publication Critical patent/CN114203295A/en
Application granted granted Critical
Publication of CN114203295B publication Critical patent/CN114203295B/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
    • G16H50/20—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
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
    • G06F18/2415—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on parametric or probabilistic models, e.g. based on likelihood ratio or false acceptance rate versus a false rejection rate
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/25—Fusion techniques
    • G06F18/253—Fusion techniques of extracted features
    • 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
    • G06N3/044—Recurrent networks, e.g. Hopfield networks
    • 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
    • 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
    • G06N3/049—Temporal neural networks, e.g. delay elements, oscillating neurons or pulsed inputs
    • 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
    • G06N3/08—Learning methods
    • 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
    • G16H20/00—ICT specially adapted for therapies or health-improving plans, e.g. for handling prescriptions, for steering therapy or for monitoring patient compliance
    • 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
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
    • Y02A90/00—Technologies having an indirect contribution to adaptation to climate change
    • Y02A90/10—Information and communication technologies [ICT] supporting adaptation to climate change, e.g. for weather forecasting or climate simulation

Abstract

The application relates to a method and a system for predicting and intervening cerebral apoplexy risks. The system comprises: the information collection module is used for collecting blood biochemical data, carotid artery color Doppler ultrasound data and daily continuous monitoring data of high-risk people suffering from cerebral apoplexy; the signal preprocessing and analyzing module is used for preprocessing the data collected by the information collecting module and extracting and analyzing the characteristics; the characteristic fusion module is used for carrying out characteristic fusion on the characteristics obtained through processing and analysis; the stroke risk prediction module is used for predicting the stroke risk according to the characteristics obtained by the characteristic fusion module; the intervention module is used for recommending an intervention scheme according to the risk level obtained by the stroke risk prediction module; and the human-computer interaction module is used for performing human-computer interaction according to the intervention scheme generated by the intervention module, wherein the feature fusion module performs feature fusion on the features obtained by the signal preprocessing analysis module by using a multi-source feature deep neural network fusion algorithm, and finally obtains fusion features DCNN-LSTM.

Description

Cerebral apoplexy risk prediction intervention method and system
Technical Field
The invention relates to the field of health management, in particular to a stroke risk prediction intervention method and system based on multi-source heterogeneous information fusion.
Background
The stroke is a serious chronic non-infectious disease seriously harming the health of China, is the first cause of death and disability of adults in China, and has the five characteristics of high morbidity, high disability rate, high mortality, high recurrence rate and high economic burden. With the development of social economy, the national life style is remarkably changed, particularly, the aging of population and the urbanization progress are accelerated, the epidemic trend of risk factors of cerebrovascular diseases is obvious, and the number of people suffering from cerebrovascular diseases is continuously increased.
The stroke has a long onset incubation period, early symptoms are slight, and the symptoms are difficult to discover, so that the medical field generally considers that the most effective measure for the diagnosis and treatment of the stroke is to intervene in high-risk groups of the stroke through effective risk assessment before the stroke occurs so as to achieve the purpose of preventing the stroke from occurring. At present, the wide cerebral apoplexy risk prediction clinically applied is mainly a risk assessment scale such as the Fuminghan stroke and the like for predicting the stroke incidence risk in the future 10 years, but the scale has obvious effect difference on the evaluation of different nationalities and regions, and the scale evaluation focuses on the prediction of the whole incidence and has limited cerebral apoplexy prevention and screening capability. With the rapid development of artificial intelligence technology and big data in recent years, researchers at home and abroad begin to apply artificial intelligence technology such as machine learning to early risk prediction of stroke, and carry out big data mining and analysis on information such as clinical biochemical data indexes and imaging to predict early risk of stroke, which can be roughly divided into two categories. The first type is that deep learning algorithm such as convolutional neural network is used for modeling and predicting data in biochemistry or image, and single data hardly represents the high-risk factors of stroke completely; the second type is that modeling prediction is carried out on multi-source data in the aspects of biochemistry, images and the like by using a certain algorithm in deep learning, and a single neural network cannot simultaneously learn the characteristics of the multi-source heterogeneous data, so that the prediction result is not ideal.
Therefore, in order to fully play the relevance and complementarity of multi-source heterogeneous data, the method and the system for predicting and intervening the stroke risk of multi-source heterogeneous information fusion are researched and designed, and have important significance for predicting and preventing the stroke risk.
Disclosure of Invention
The invention aims to provide a stroke risk prediction intervention system based on multi-source heterogeneous information fusion, which comprises an information collection module, an information preprocessing analysis module, a feature fusion module, a stroke risk prediction module, an intervention module and a man-machine interaction module, wherein:
the information collection module is used for collecting blood biochemical data, carotid artery color Doppler ultrasound data and daily continuous monitoring data of high-risk people suffering from cerebral apoplexy;
the signal preprocessing and analyzing module is used for preprocessing blood biochemical data, carotid artery color Doppler ultrasound data and daily continuous monitoring data of the information collecting module and extracting and analyzing features;
the characteristic fusion module is used for carrying out characteristic fusion on the characteristics obtained by the preprocessing and analysis of the signal preprocessing and analysis module;
the stroke risk prediction module is used for predicting the stroke risk according to the characteristics obtained by the characteristic fusion module;
the intervention module is used for recommending an intervention scheme according to the risk level obtained by the stroke risk prediction module;
and the human-computer interaction module is used for performing human-computer interaction according to the intervention scheme generated by the intervention module. The feature fusion module performs feature fusion on the features obtained by the signal preprocessing analysis module by using a deep neural network model through a multi-source feature deep neural network fusion algorithm to finally obtain fusion features DCNN-LSTM。
In the stroke risk prediction and intervention system, high-risk groups in the stroke in the information collection module are screened according to the basic information of clinical cases, and the groups have more than 3 high-risk factors in the stroke.
Preferably, the high risk factors in stroke include: age is high, hypertension, hyperglycemia, hyperlipemia, heart disease, family history of apoplexy, smoking, obesity, high salt and high oil eating habit, and less exercise 10 items; the blood biochemical data include: high density lipoproteins, low density lipoproteins, cholesterol, triglycerides, homocysteine, D dimers, glycated serum proteins, and the like; the daily continuous monitoring data comprises: heart rate, blood pressure, electrocardiogram, exercise, blood sugar, body fat ratio.
In the system for predicting and intervening stroke risk, the signal preprocessing analysis performed by the signal preprocessing analysis module comprises the preprocessing analysis of biochemical data, carotid artery color Doppler ultrasound data and daily continuous monitoring data, wherein,
the biochemical data preprocessing analysis comprises the following steps: normalization processing is carried out by utilizing dispersion normalization, and data is compressed to [0,1] through linear transformation of original data, wherein the formula is as follows:
Figure GDA0003587898990000031
wherein X' is the compressed value, X is the original value of a biochemical index, XmaxIs the maximum value of index X, XmiIs the minimum value of index X.
The preprocessing of carotid color Doppler ultrasound data comprises the following steps: and vectorizing the report result words of the carotid artery color Doppler ultrasound data by using word2vec word vector technology.
The preprocessing analysis of daily continuous monitoring data comprises the following steps: the heart rate variability index is calculated by utilizing the electrocardiogram data, the blood pressure variability index is calculated by utilizing the continuously monitored blood pressure data, and the average value of the heart rate, the blood pressure, the blood sugar, the exercise and the body fat ratio every 7 days is calculated.
In a stroke risk prediction intervention system, an information preprocessing analysis module analyzes and processes information obtained by the information preprocessing module, and the analysis comprises the following steps: performing feature extraction on the preprocessed biochemical data information through a convolutional neural network model to obtain a feature vector XCNN。
XCNN＝GCNN(Net1CNN，Tr1CNN，X)
[Net1CNN，Tr1CNN]＝Feedforward(W1CNN,B1CNN；M1CL,M1pool；Z)
Wherein, XCNNIs a feature vector, G, extracted from a convolutional neural networkCNNIs the output function of the convolutional neural network, Net1CNNIs a trained convolutional neural network, Tr1CNNIs a trained convolutional neural network parameter, W1CNNIs a convolution kernel matrix parameter, B1CNNIs a bias parameter, M1CLIs the convolution kernel size and number of layers, M1poolIs the maximum pooled nuclear size and number of layers, and X is the input biochemical data.
Extracting the characteristics of the preprocessed daily physiological monitoring signals through a long-time and short-time memory network, and outputting a characteristic vector YLSTM。
YLSTM＝GLSTM(NetLSTM，TrLSTM，Y)
[NetLSTM,TrLSTM]＝Feedforward(WLSTM1,WLSTM2,BLSTM1,BLSTM2；Y)
Wherein, YLSTMIs a characteristic vector, Net, extracted by a long-time and short-time memory networkLSTMIs a long and short term memory network after training, TrLSTMIs the network parameter memorized after training, and the feed forward is the Feedforward neural network function, WLSTM1、WLSTM2For long and short term memory of the weight parameters of the network input gate and the forgetting gate, BLST1、BLSTM2The method is characterized in that bias parameters of a network input gate and a forgetting gate are memorized at long time, and Y is an input daily physiological monitoring signal.
Performing feature extraction on the preprocessed carotid artery color Doppler ultrasound data through a convolutional neural network model to obtain a feature vector ZCNN。
ZCNN＝GCNN(Net2CNN，Tr2CNN，Z)
[Net2CNN，Tr2CNN]＝Feedforward(W2CNN,B2CNN；M2CL,M2pool；Z)
Wherein Z isCNNIs a feature vector, Net, extracted by a convolutional neural network2CNNIs a trained convolutional neural network, Tr2CNNIs a trained convolutional neural network parameter, W2CNNIs a convolution kernel matrix parameter, B2CNNIs a bias parameter, M2CLIs the convolution kernel size and number of layers, M2poolThe maximum pool nucleus size and the number of layers are adopted, and Z is input carotid artery color Doppler ultrasound data.
In the stroke risk prediction intervention system, a feature fusion module performs feature fusion on the features obtained by a signal analysis module by using a multi-source feature deep neural network fusion algorithm to finally obtain a fusion feature DCNN-LSTMThe method comprises the following steps:
s1: constructing a deep neural network structure, which consists of an input layer, a hidden layer and an output layer;
s2 initializing deep neural network parameters
Figure GDA0003587898990000041
Wherein, TrCNN-LSTM0Is a deep neural network initialization parameter, R is the number of input layer neurons, WCNN-LSTM1Is the weight of the input layer to the hidden layer, bCNN-LSTM1Is a bias of the hidden layer 1, WCNN-LSTM2Is the weight of hidden layer 1 to hidden layer 2, bCNN-LSTM2Is a bias of the hidden layer 2, WCNN-LSTM3Is the weight of hidden layer 2 to hidden layer 3, bCNN-LSTM3Is the bias of the hidden layer 3, hCNN_LSTM1、hCNN_LSTM2、hCNN_LSTM3The number of neurons in the hidden layer 1, the hidden layer 2, and the hidden layer 3; rand is a random number function between randomly generated (0, 1) and zeros is a zero function.
S3: mixing XCNNInputting the data into an initialized deep neural network, continuously updating network parameters and obtaining trained network parameters TrCNN-LSTM1And new feature XCNN,1
XCNN，1＝GCNN-LSTM1(NetCNN-LSTM1,TrCNN-LSTM1,XCNN)
GCNN-LSTM1Is a deep neural network NetCNN-LSTM1Is used to generate the output function of (1).
S4: will depth neural network TrCNN-LSTM1Value of as network layer NetCNN-LSTM2Initial value of (2), feature YLSTMInput to network layer NetCNN-LSTM2In the method, updating network parameters is carried out to complete the characteristic XCNNAnd YLSTMIs optimized alternately, then Z isCNNInput to network layer NetCNN-LSTM3In (3), updating network parameters is performed to complete feature XCNN、YLSTMAnd ZCNNFirst alternate optimization of (a).
Network layer NetCNN-LSTM1For feature XCNNIs measured for the reconstruction error Lθ1Comprises the following steps:
Figure GDA0003587898990000051
UcCNN1＝σ(W′CNN-LSTM1*σ(W′CNN-LSTM2*σ(W′CNN-LSTM3*XCNN，1+b′CNN-LSTM3))+b′CNN-LSTM2))+b′CNN-LSTM1)
Figure GDA0003587898990000052
Figure GDA0003587898990000053
where n is the total number of neurons, UcCNN1Is network layer NetCNN-LSTM1To XCNNThe reconstructed value of (1), W'CNN-LSTM1Is the weight, W ', passed from the hidden layer 1 to the output layer'CNN-LSTM2Is the weight, W ', that the hidden layer 2 transfers to the output layer'CNN-LSTM3Is the weight, b ', that the hidden layer 3 transfers to the output layer'CNN-LSTM1Is the bias of the output layer corresponding to the hidden layer 1,b′CNN-LSTM2is the offset of the output layer corresponding to the hidden layer 2, b'CNN-LSTM3Is the offset of the output layer corresponding to the hidden layer 2, bCNN-LSTM1Is the bias of the hidden layer 1, σ is the activation function, and α is the learning rate.
In the same way for feature YLSTMAnd feature ZCNNIs measured for the reconstruction error Lθ2And Lθ3Thereby obtaining XCNN、YLSTMAnd ZCNNError L of the first alternating optimization1:
L1＝Lθ1+Lθ2+Lθ3
S5: repeating step X according to steps S3 and S4CNN、YLSTMAnd ZCNNInputting the data into a deep neural network for alternate optimization until the error L is smaller than a set threshold value. At the moment, the final deep neural network parameters and fusion characteristics D are obtainedCNN-LSTM。
DCNN-LSTM＝GCNN-LSTM(NetCNN-LSTM,TrCNN-LSTM；XCNN,YLSTM,ZCNN)
Wherein, GCNN-LSTMIs a deep neural network output function, NetCNN-LSTMIs a deep neural network, Tr, after trainingCNN-LSTMAre the trained deep neural network parameters.
The risk prediction module carries out risk prediction according to the fusion characteristics obtained by the characteristic fusion module and comprises the following steps:
the obtained final fusion characteristics DCNN-LSTMInputting the risk prediction into a softmax classifier for risk prediction, and classifying the risk prediction into the following levels: the risk prediction method comprises the following steps of high risk level one, high risk level two and high risk level three:
P(result|DCNN-LSTM)＝softmax(Wp*DCNN-LSTM+bp)
Wp、bpis a parameter of the classifier softmax, P (result | D)CNN-LSTM) Is the predicted probability of a certain risk level, and the predicted risk level is obtained by decoding according to the predicted probability value.
The intervention module comprises a health information management library, an intelligent learning module and an intervention scheme, wherein the health management information library stores personal basic health information of a user and an intervention recommendation scheme library matched with the personal health information; the intelligent learning module is a pre-designed intelligent learning module, and recommends an individualized intervention scheme from an intervention recommendation scheme library according to the stroke risk prediction grade and the personal health basic condition of the user, and finally, a doctor adjusts and confirms the intervention scheme; the intervention program is an intervention program that is ultimately confirmed by a physician and recommended to the user.
Preferably, an intervention scheme is provided for a user for the first time and stored in the health information management base, and the intervention scheme is updated and adjusted periodically according to the stroke risk prediction result and the health management information updating condition.
And the human-computer interaction module is used for informing the doctor of the final intervention scheme to the user and carrying out communication interaction with the user in the modes of voice, text and the like.
Compared with the prior art, the invention has the beneficial effects that:
(1) and performing feature fusion on the biochemical data information, the carotid artery color Doppler ultrasound information and the basic physiological monitoring information by adopting a multi-source feature deep neural network fusion algorithm, and fully exerting relevance and complementarity of multi-source heterogeneous data.
(2) According to the screening of the clinical case basic information, people with more than 3 high-risk factors in stroke are screened, and then the stroke risk prediction is carried out on the people, so that the prediction result has pertinence and practical application value.
(3) By the method and the system, the risk level of the stroke high risk group can be obtained by collecting biochemical data information, carotid artery color Doppler ultrasound information and basic physiological monitoring information of the stroke high risk group, different intervention schemes are recommended according to the risk level of the stroke high risk group, the personalized self-adaptive intervention scheme is recommended through intelligent learning, man-machine interaction is carried out with the old people, personalized intervention is realized, and the stroke risk is reduced.
Drawings
In the drawings, like reference numerals refer to the same or similar parts or elements throughout the several views unless otherwise specified. The figures are not necessarily to scale. It is appreciated that these drawings depict only some embodiments in accordance with the disclosure and are therefore not to be considered limiting of its scope.
Fig. 1 is a schematic diagram of a stroke risk prediction intervention system based on multi-source heterogeneous information fusion according to the present invention;
fig. 2 is a flowchart of a stroke risk prediction intervention method based on multi-source heterogeneous information fusion according to the present invention;
FIG. 3 is a flow chart of feature fusion according to FIG. 2;
FIG. 4 is a flow chart of adaptive human-machine interaction according to the present invention.
Detailed Description
In the following, only certain exemplary embodiments are briefly described. As those skilled in the art will recognize, the described embodiments may be modified in various different ways, all without departing from the spirit or scope of the present invention. Accordingly, the drawings and description are to be regarded as illustrative in nature, and not as restrictive.
The following detailed description of the present invention will be made with reference to the accompanying drawings 1-4.
As shown in fig. 1, the invention provides a stroke risk prediction intervention system based on multi-source heterogeneous information fusion, which includes an information collection module, an information preprocessing analysis module, a feature fusion module, a stroke risk prediction module, an intervention module and a human-computer interaction module.
The information collection module is used for collecting blood biochemical data, carotid artery color Doppler ultrasound data and daily continuous monitoring data of high-risk groups suffering from cerebral apoplexy.
The signal preprocessing and analyzing module is used for preprocessing blood biochemical data, carotid artery color Doppler ultrasound data and daily continuous monitoring data of the information collecting module and extracting and analyzing features.
And the feature fusion module performs feature fusion on the features obtained by the signal preprocessing analysis module by using the deep neural network model.
And the stroke risk prediction module is used for inputting the features obtained according to the feature fusion module into the softmax module to predict the stroke risk.
The intervention module is used for recommending an intervention scheme according to the risk level obtained by the stroke risk prediction module.
And the human-computer interaction module is used for performing human-computer interaction according to the intervention scheme generated by the intervention module.
The method for predicting the risk of stroke by using the stroke risk prediction module based on multi-source heterogeneous information fusion according to the present invention is described below with reference to fig. 2 to 4.
The information collection module collects biochemical data, carotid artery color Doppler ultrasound data and daily continuous monitoring data of high-risk groups in the stroke, the groups are screened according to the basic information of clinical cases, and the groups have more than 3 high-risk factors in the stroke.
Specifically, high risk factors in stroke include: age is high, hypertension, hyperglycemia, hyperlipemia, heart disease, family history of apoplexy, smoking, obesity, high salt and high oil eating habit, and less exercise 10 items; the blood biochemical data include: high density lipoproteins, low density lipoproteins, cholesterol, triglycerides, homocysteine, D dimers, glycated serum proteins, and the like; the daily continuous monitoring data comprises: heart rate, blood pressure, electrocardiogram, exercise, blood sugar, body fat ratio.
The signal preprocessing analysis performed by the signal preprocessing analysis module comprises preprocessing analysis on biochemical data, carotid artery color Doppler ultrasound data and daily continuous monitoring data, wherein:
the biochemical data preprocessing comprises the following steps: normalization processing is carried out by utilizing dispersion normalization, and data is compressed to [0,1] through linear transformation of original data, wherein the formula is as follows:
Figure GDA0003587898990000081
where X' is a compressed value and X is an itemOriginal value of biochemical index, XmaxIs the maximum value of index X, XmiIs the minimum value of index X.
The preprocessing of carotid color Doppler ultrasound data comprises the following steps: and vectorizing the report result words of the carotid artery color Doppler ultrasound data by using word2vec word vector technology.
The preprocessing analysis of daily continuous monitoring data comprises the following steps: the heart rate variability index is calculated by utilizing the electrocardiogram data, the blood pressure variability index is calculated by utilizing the continuously monitored blood pressure data, and the average value of the heart rate, the blood pressure, the blood sugar, the exercise and the body fat ratio every 7 days is calculated.
The signal preprocessing and analyzing module is used for analyzing and processing the preprocessed information and comprises the following steps: performing feature extraction on the preprocessed biochemical data information through a convolutional neural network model to obtain a feature vector XCNN。
XCNN＝GCNN(Net1CNN,Tr1CNN,X)
[Net1CNN,Tr1CNN]＝Feedforward(W1CNN,B1CNN；M1CL,M1pool；X)
Wherein, XCNNIs a feature vector, Net, extracted by a convolutional neural network1CNNIs a trained convolutional neural network, Tr1CNNIs the parameter of the convolutional neural network after training, and feed forward is the function of the Feedforward neural network, W1CNNIs a convolution kernel matrix parameter, B1CNNIs a bias parameter, M1CLIs the size and number of convolution kernel layers set to 3 x 2 convolution kernel, the number of convolution layers is 2, M1poolThe maximum pooled nucleus size and number of layers were set to 2X 2 pooled nuclei, the number of pooled layers was 2, and X was the input biochemical data.
Extracting the characteristics of the preprocessed daily physiological monitoring signals through a long-time and short-time memory network, and outputting a characteristic vector YLSTM。
YLSTM＝GLSTM(NetLSTM，TrLSTM，Y)
[NetLSTM,TrLSTM]＝Feedforward(WLSTM1,WLSTM2,BLSTM1,BLSTM2；Y)
Wherein, YLSTMIs a characteristic vector, Net, extracted by a long-time and short-time memory networkLSTMIs a long and short term memory network after training, TrLSTMIs the network parameter memorized after training, and the feed forward is the Feedforward neural network function, WLSTM1、WLSTM2For long and short term memory of the weight parameters of the network input gate and the forgetting gate, BLST1、BLSTM2The method is characterized in that bias parameters of a network input gate and a forgetting gate are memorized at long time, and Y is an input daily physiological monitoring signal.
Performing feature extraction on the preprocessed carotid artery color Doppler ultrasound data through a convolutional neural network model to obtain a feature vector ZCNN。
ZCNN＝GCNN(Net2CNN，Tr2CNN，Z)
[Net2CNN，Tr2CNN]＝Feedforward(W2CNN,B2CNN；M2CL,M2pool；Z)
Wherein Z isCNNIs a feature vector, Net, extracted by a convolutional neural network2CNNIs a trained convolutional neural network, Tr2CNNIs the parameter of the trained convolutional neural network, feed forward is the function of the Feedforward neural network, W2CNNIs a convolution kernel matrix parameter, B2CNNIs a bias parameter, M2CLIs the size and number of convolution kernels, the convolution kernel is 3 x 3, the number of convolution layers is 3, M2poolThe maximum pool nucleus size and the number of layers are shown, the pool nucleus is 2 x 2, the pool layer is 2, and Z is input carotid artery color ultrasound data.
The feature fusion module performs feature fusion on the features obtained by the signal preprocessing analysis module by using a multi-source feature deep neural network fusion algorithm to finally obtain a fusion feature DCNN-LSTMAs shown in fig. 3:
s1: constructing a deep neural network structure, wherein the deep neural network is composed of an input layer, a hidden layer and an output layer;
s2 initializing deep neural network parameters
Figure GDA0003587898990000091
Wherein, TrCNN-LSTM0Is a deep neural network initialization parameter, R is the number of input layer neurons, WCNN-LSTM1Is the weight of the input layer to the hidden layer, bCNN-LSTM1Is a bias of the hidden layer 1, WCNN-LSTM2Is the weight of hidden layer 1 to hidden layer 2, bCNN-LSTM2Is a bias of the hidden layer 2, WCNN-LSTM3Is the weight of hidden layer 2 to hidden layer 3, bCNN-LSTM3Is the bias of the hidden layer 3, hCNN_LSTM1、hCNN_LSTM2、hCNN_LSTM3The number of neurons in the hidden layer 1, the hidden layer 2, and the hidden layer 3; rand is a random number function between randomly generated (0, 1) and zeros is a zero function.
S3: mixing XCNNInputting the data into an initialized deep neural network, continuously updating network parameters and obtaining trained network parameters TrCNN-LSTM1And new feature XCNN,1,
XCNN，1＝GCNN-LSTM1(NetCNN-LSTM1,TrCNN-LSTM1,XCNN)
GCNN-LSTM1Is a deep neural network NetCNN-LSTM1Is used to generate the output function of (1).
S4: will depth neural network TrCNN-LSTM1As the value of network layer NetCNN-LSTM2Initial value of (2), feature YLSTMInput to network layer NetCNN-LSTM2In the method, updating network parameters is carried out to complete the characteristic XCNNAnd YLSTMThen the deep neural network Tr is optimizedCNN-LSTM2As the value of network layer NetCNN-LSTM3Initial value of (1), feature ZCNNInput to network layer NetCNN-LSTM3In (3), updating network parameters is performed to complete feature XCNN、YLSTMAnd ZCNNFirst alternate optimization of (a).
Network layer NetCNN-LSTM1For feature XCNNIs measured for the reconstruction error Lθ1Comprises the following steps:
Figure GDA0003587898990000101
UcCNN1＝σ(W′CNN-LSTM1*σ(W′CNN-LSTM2*σ(W′CNN-LSTM3*XCNN，1+b′CNN-LSTM3))+b′CNN-LSTM2))+b′CNN-LSTM1)
Figure GDA0003587898990000102
Figure GDA0003587898990000103
where n is the total number of neurons, UcCNN1Is network layer NetCNN-LSTM1To XCNNThe reconstructed value of (1), W'CNN-LSTM1Is the weight, W ', that hidden layer 1 transfers to the output layer'CNN-LSTM2Is the weight, W ', that the hidden layer 2 transfers to the output layer'CNN-LSTM3Is the weight, b ', that the hidden layer 3 transfers to the output layer'CNN-LSTM1Is the offset of the output layer corresponding to the hidden layer 1, b'CNN-LSTM2Is the offset of the output layer corresponding to the hidden layer 2, b'CNN-LSTM3Is the offset of the output layer corresponding to the hidden layer 2, bCNN-LSTM1Is the bias of the hidden layer 1, σ is the activation function, and α is the learning rate.
In the same way for feature YLSTMAnd feature ZCNNIs reconstructed error Lθ2And Lθ3Thereby obtaining XCNN、YLSTMAnd ZCNNError L of the first alternating optimization1:
L1＝Lθ1+Lθ2+Lθ3
S5: fitting the deep neural network parameters TrCNN-LSTM3As the next initial value, repeat X according to steps S3 and S4CNN、YLSTMAnd ZCNNInputting the data into a deep neural network for repeated alternate optimization until the error L is less than a set threshold value. At this point the finalDeep neural network parameters and fusion characteristics DCNN-LSTM。
DCNN-LSTM＝GCNN-LSTM(NetCNN-LSTM,TrCNN-LSTM；XCNN,YLSTM,ZCNN)，
Wherein G isCNN-LSTMIs a deep neural network output function, NetCNN-LSTMIs a deep neural network, Tr, after trainingCNN-LSTMAre the trained deep neural network parameters.
The risk prediction module carries out risk prediction according to the fusion characteristics obtained by the characteristic fusion module, and the risk prediction comprises the following steps:
the obtained final fusion characteristics DCNN-LSTMInputting the risk prediction into a softmax classifier for risk prediction, and classifying the risk prediction into the following levels: the risk prediction method comprises the following steps of high risk level one, high risk level two and high risk level three:
P(result|DCNN-LSTM)＝softmax(Wp*DCNN-LSTM+bp)
Wp、bpis a parameter of the classifier softmax, P (result | D)CNN-LSTM) Is the predicted probability of a certain risk level, and the predicted risk level is obtained by decoding according to the predicted probability value.
The intervention module comprises a health information management library, an intelligent learning module and an intervention scheme, wherein the health management information library stores personal basic health information of a user and an intervention recommendation scheme library matched with the personal health information; the intelligent learning module is a pre-designed intelligent learning module, and recommends an individualized intervention scheme from an intervention recommendation scheme library according to the stroke risk prediction grade and the personal health basic condition of the user, and finally, a doctor adjusts and confirms the intervention scheme; the intervention program is an intervention program that is ultimately confirmed by a physician and recommended to the user.
Preferably, an intervention scheme is provided for a user for the first time and stored in the health information management base, and the intervention scheme is updated and adjusted periodically according to the stroke risk prediction result and the health management information updating condition.
And the human-computer interaction module is used for informing the doctor of the final intervention scheme to the user and carrying out communication interaction with the user in the modes of voice, text, video and the like.
Preferably, the human-computer interaction module timely informs the user of the change of the stroke risk prediction result and the intervention scheme recommended by the intervention module according to the change after adjustment and confirmation of the doctor, and simultaneously increases the adjusted intervention scheme to the intervention scheme library. For example, when the stroke risk prediction module predicts that the risk level of the user is changed from the high risk level to the high risk level in the last prediction result, the intervention module recommends a new intervention scheme of the user through intelligent learning of the risk level of the user and basic health information of the user and sends the new intervention scheme to a doctor, the doctor adjusts and confirms the intervention scheme by comprehensively analyzing health management information of the old, informs the intervention scheme to the user in a voice, text or video mode and the like, and simultaneously reminds the user to timely take medical examinations and state personal basic health conditions to the doctor, so that the user can take medical examinations and interventions early, and the probability of stroke occurrence is reduced.
The foregoing summary is provided for purposes of illustration only and is not intended to be limiting in any way. In addition to the illustrative aspects, embodiments, and features described above, further aspects, embodiments, and features of the present invention will be readily apparent by reference to the drawings and following detailed description.

Claims (6)

1. A stroke risk prediction intervention system, comprising:
the information collection module is used for collecting blood biochemical data, carotid artery color Doppler ultrasound data and daily continuous monitoring data of high-risk people suffering from cerebral apoplexy;
the signal preprocessing and analyzing module is used for preprocessing blood biochemical data, carotid artery color Doppler ultrasound data and daily continuous monitoring data collected by the information collecting module and extracting and analyzing characteristics;
the characteristic fusion module is used for carrying out characteristic fusion on the characteristics obtained by the preprocessing and analysis of the signal preprocessing and analysis module;
the stroke risk prediction module is used for predicting the stroke risk according to the characteristics obtained by the characteristic fusion module;
the intervention module is used for recommending an intervention scheme according to the risk level obtained by the stroke risk prediction module; and
a human-computer interaction module for performing human-computer interaction according to the intervention scheme generated by the intervention module,
the feature fusion module performs feature fusion on the features obtained by the signal preprocessing analysis module by using a deep neural network model through a multi-source feature deep neural network fusion algorithm to finally obtain fusion features DCNN-LSTM，
In the signal preprocessing analysis module, the preprocessed biochemical data information is subjected to feature extraction through a convolutional neural network model to obtain a feature vector XCNNPerforming feature extraction on the preprocessed daily physiological monitoring signals through a long-time and short-time memory network to obtain a feature vector YLSTMPerforming feature extraction on the preprocessed carotid artery color Doppler ultrasound data through a convolutional neural network model to obtain a feature vector ZCNN，
In the feature fusion Module, XCNNIs input into the initialized deep neural network, network parameters are updated, and trained network parameters Tr are obtainedCNN-LSTM1And new feature XCNN,1(ii) a Will depth neural network TrCNN-LSTM1Value of as network layer NetCNN-LSTM2Initial value of, feature YLSTMIs input to the network layer NetCNN-LSTM2In the method, updating network parameters is carried out to complete the characteristic XCNNAnd YLSTMAlternate optimization of (2); then the deep neural network TrCNN-LSTM2Value of as network layer NetCNN-LSTM3Initial value of (1), ZCNNIs input to the network layer NetCNN-LSTM3In (2), updating network parameters to complete feature XCNN、YLSTMAnd ZCNNThe first alternate optimization of (1); computing network layer NetCNN-LSTM1For feature XCNNReconstruction error of
Figure DEST_PATH_IMAGE001
Network layer NetCNN-LSTM2For feature YLSTMReconstruction error of
Figure 40219DEST_PATH_IMAGE002
And network layer NetCNN-LSTM3For feature ZCNNReconstruction error of
Figure DEST_PATH_IMAGE003
Thereby obtaining XCNN、YLSTMAnd ZCNNError L of the first alternating optimization1：
Figure 144310DEST_PATH_IMAGE004
(ii) a Fitting the deep neural network parameters TrCNN-LSTM3As the initial value of the next time, X is repeatedly setCNN、YLSTMAnd ZCNNInputting the data into a deep neural network for alternate optimization until the error L is smaller than a set threshold value, and obtaining final parameters of the deep neural network and fusion characteristics D at the momentCNN-LSTM,
Figure DEST_PATH_IMAGE005
，
Wherein G isCNN-LSTMIs a deep neural network output function, NetCNN-LSTMIs a deep neural network, Tr, after trainingCNN-LSTMIs the parameter of the deep neural network after training, and the final fusion characteristic D is obtainedCNN-LSTMAnd inputting the data into a stroke risk prediction module for risk prediction.
2. A prediction system using the stroke risk prediction intervention system of claim 1, comprising:
collecting blood biochemical data, carotid artery color Doppler ultrasound data and daily continuous monitoring data of high-risk people with stroke by using an information collection module;
preprocessing and characteristic extraction and analysis are carried out on the collected blood biochemical data, carotid artery color Doppler ultrasound data and daily continuous monitoring data by utilizing a signal preprocessing and analyzing module;
performing feature fusion on the extracted and analyzed features by using a feature fusion module; and
the stroke risk prediction module is used for predicting the stroke risk according to the characteristics obtained by the characteristic fusion,
performing feature fusion on the obtained features by using a deep neural network model through a multi-source feature deep neural network fusion algorithm to finally obtain fusion features DCNN-LSTM。
3. The prediction system of claim 2,
performing feature extraction on the preprocessed biochemical data information through a convolutional neural network model to obtain a feature vector XCNN，
Figure 596151DEST_PATH_IMAGE006
Wherein, XCNNIs a feature vector, Net, extracted by a convolutional neural network1CNNIs a trained convolutional neural network, Tr1CNNIs the parameter of the convolutional neural network after training, and feed forward is the function of the Feedforward neural network, W1CNNIs a convolution kernel matrix parameter, B1CNNIs a bias parameter, M1CLIs the convolution kernel size and number of layers, M1poolIs the maximum pool nucleus size and the layer number, X is the input biochemical data,
extracting the characteristics of the preprocessed daily physiological monitoring signals through a long-time and short-time memory network, and outputting a characteristic vector YLSTM，
Figure DEST_PATH_IMAGE007
Wherein, YLSTMIs a characteristic vector, Net, extracted by a long-time and short-time memory networkLSTMIs after trainingLong and short term memory network of, TrLSTMIs the network parameter memorized after training, and the feed forward is the Feedforward neural network function, WLSTM1、WLSTM2For long and short term memory of the weight parameters of the network input gate and the forgetting gate, BLST1、BLSTM2For memorizing the bias parameters of the input gate and the forgetting gate of the network at long time, Y is an input daily physiological monitoring signal,
performing feature extraction on the preprocessed carotid artery color Doppler ultrasound data through a convolutional neural network model to obtain a feature vector ZCNN，
Figure 679775DEST_PATH_IMAGE008
Wherein Z isCNNIs a feature vector, Net, extracted by a convolutional neural network2CNNIs a trained convolutional neural network, Tr2CNNIs the parameter of the convolutional neural network after training, and feed forward is the function of the Feedforward neural network, W2CNNIs a convolution kernel matrix parameter, B2CNNIs a bias parameter, M2CLIs the convolution kernel size and number of layers, M2poolThe maximum pool nucleus size and the number of layers, and Z is input carotid artery color Doppler ultrasound data.
4. The prediction system of claim 3, wherein X isCNNInputting the data into the initialized deep neural network, updating the network parameters and obtaining the trained network parameters TrCNN-LSTM1And new feature XCNN,1
Figure DEST_PATH_IMAGE009
GCNN-LSTM1Is a deep neural network NetCNN-LSTM1The output function of (a) is selected,
will depth neural network TrCNN-LSTM1As the value of network layer NetCNN-LSTM2Initial value of (2), feature YLSTMInput to network layer NetCNN-LSTM2In the method, updating network parameters is carried out to complete the characteristic XCNNAnd YLSTMThe alternating optimization of (a) and (b),
deep neural network TrCNN-LSTM2Value of as network layer NetCNN-LSTM3Initial value of (2), ZCNNInput to network layer NetCNN-LSTM3In (3), updating network parameters is performed to complete feature XCNN、YLSTMAnd ZCNNFirst alternate optimization of (a).
5. The prediction system of claim 4, wherein computing network layer NetCNN-LSTM1For feature XCNNReconstruction error of
Figure 431831DEST_PATH_IMAGE010
Calculating network layer NetCNN-LSTM2For feature YLSTMReconstruction error of (2)
Figure DEST_PATH_IMAGE011
And network layer NetCNN-LSTM3For feature ZCNNReconstruction error of
Figure 175665DEST_PATH_IMAGE012
Thereby obtaining XCNN、YLSTMAnd ZCNNError L of the first alternating optimization of1:
Figure DEST_PATH_IMAGE013
，
Fitting the deep neural network parameters TrCNN-LSTM3As the initial value of the next time, X is repeatedly setCNN、YLSTMAnd ZCNNInputting the data into a deep neural network for alternate optimization until the error L is smaller than a set threshold value, and obtaining final parameters of the deep neural network and fusion characteristics DCNN-LSTM，
Figure 131114DEST_PATH_IMAGE014
Wherein G isCNN-LSTMIs a deep neural network output function, NetCNN-LSTMIs a deep neural network, Tr, after trainingCNN-LSTMAre the trained deep neural network parameters.
6. Prediction system according to claim 4, characterized in that the final fused feature D to be obtainedCNN-LSTMInputting the risk prediction data into a stroke risk prediction module for risk prediction, wherein the risk prediction mode is as follows:
Figure DEST_PATH_IMAGE015
Wp、bpis a parameter of the classifier, P (result | D)CNN-LSTM) Is the predicted probability of a certain risk level, and the predicted risk level is obtained by decoding according to the predicted probability value.
CN202111395059.1A 2021-11-23 2021-11-23 Cerebral apoplexy risk prediction intervention method and system Active CN114203295B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202111395059.1A CN114203295B (en) 2021-11-23 2021-11-23 Cerebral apoplexy risk prediction intervention method and system

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202111395059.1A CN114203295B (en) 2021-11-23 2021-11-23 Cerebral apoplexy risk prediction intervention method and system

Publications (2)

Publication Number Publication Date
CN114203295A CN114203295A (en) 2022-03-18
CN114203295B true CN114203295B (en) 2022-05-20

Family

ID=80648444

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202111395059.1A Active CN114203295B (en) 2021-11-23 2021-11-23 Cerebral apoplexy risk prediction intervention method and system

Country Status (1)

Country Link
CN (1) CN114203295B (en)

Families Citing this family (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115019919B (en) * 2022-06-16 2023-04-07 江苏省人民医院（南京医科大学第一附属医院） Deep learning-based method for establishing function prediction model after early recovery of stroke
CN116269249B (en) * 2023-02-14 2023-11-17 国家康复辅具研究中心 Cerebral apoplexy risk prediction method and system
CN116705306B (en) * 2023-08-03 2023-10-31 首都医科大学附属北京天坛医院 Method for monitoring cerebral apoplexy, device for monitoring cerebral apoplexy and storage medium

Citations (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107220506A (en) * 2017-06-05 2017-09-29 东华大学 Breast cancer risk assessment analysis system based on depth convolutional neural networks
CN109117864A (en) * 2018-07-13 2019-01-01 华南理工大学 Coronary heart disease risk prediction technique, model and system based on heterogeneous characteristic fusion
CN111161882A (en) * 2019-12-04 2020-05-15 深圳先进技术研究院 Breast cancer life prediction method based on deep neural network
CN111242288A (en) * 2020-01-16 2020-06-05 浙江工业大学 Multi-scale parallel deep neural network model construction method for lesion image segmentation
CN111430029A (en) * 2020-03-24 2020-07-17 浙江达美生物技术有限公司 Multi-dimensional stroke prevention screening method based on artificial intelligence
CN111968746A (en) * 2020-07-21 2020-11-20 华中科技大学协和深圳医院 Cerebral apoplexy risk prediction method and device based on hybrid deep transfer learning
CN112216391A (en) * 2020-10-22 2021-01-12 深圳市第二人民医院(深圳市转化医学研究院) Method and device for assessing cerebral stroke morbidity risk based on carotid atherosclerosis condition
CN113241135A (en) * 2021-04-30 2021-08-10 山东大学 Disease risk prediction method and system based on multi-mode fusion
CN113362959A (en) * 2021-06-03 2021-09-07 重庆南鹏人工智能科技研究院有限公司 Sudden respiratory infectious disease risk prediction model for regional epidemic prevention and control

Family Cites Families (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20120163693A1 (en) * 2010-04-20 2012-06-28 Suri Jasjit S Non-Invasive Imaging-Based Prostate Cancer Prediction
US20110257545A1 (en) * 2010-04-20 2011-10-20 Suri Jasjit S Imaging based symptomatic classification and cardiovascular stroke risk score estimation

Patent Citations (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107220506A (en) * 2017-06-05 2017-09-29 东华大学 Breast cancer risk assessment analysis system based on depth convolutional neural networks
CN109117864A (en) * 2018-07-13 2019-01-01 华南理工大学 Coronary heart disease risk prediction technique, model and system based on heterogeneous characteristic fusion
CN111161882A (en) * 2019-12-04 2020-05-15 深圳先进技术研究院 Breast cancer life prediction method based on deep neural network
CN111242288A (en) * 2020-01-16 2020-06-05 浙江工业大学 Multi-scale parallel deep neural network model construction method for lesion image segmentation
CN111430029A (en) * 2020-03-24 2020-07-17 浙江达美生物技术有限公司 Multi-dimensional stroke prevention screening method based on artificial intelligence
CN111968746A (en) * 2020-07-21 2020-11-20 华中科技大学协和深圳医院 Cerebral apoplexy risk prediction method and device based on hybrid deep transfer learning
CN112216391A (en) * 2020-10-22 2021-01-12 深圳市第二人民医院(深圳市转化医学研究院) Method and device for assessing cerebral stroke morbidity risk based on carotid atherosclerosis condition
CN113241135A (en) * 2021-04-30 2021-08-10 山东大学 Disease risk prediction method and system based on multi-mode fusion
CN113362959A (en) * 2021-06-03 2021-09-07 重庆南鹏人工智能科技研究院有限公司 Sudden respiratory infectious disease risk prediction model for regional epidemic prevention and control

Non-Patent Citations (5)

* Cited by examiner, † Cited by third party
Title
"A remote quantitative Fugl-Meyer assessment framework for stroke patients based on wearable sensor networks";Yu L 等;《Computer Methods and Programs in Biomedicine》;20160531;第128卷;第100-110页 *
"卒中后运动神经反馈康复训练研究进展与前景";王萌亚 等;《中国生物医学工程学报》;20191220;第38卷(第6期);第742-752页 *
"基于多源异构大数据机器学习的风险预测方法研究";杨帆;《中国优秀硕士学位论文全文数据库（信息科技辑）》;20190415;第I140-74页 *
"融合脑功能和运动评估的脑卒中康复训练处方推荐模型构建";张腾宇 等;《中国生物医学工程学报》;20210820;第40卷(第4期);第394-400页 *
"轴承故障的多源异构数据特征级融合诊断方法";徐济宣;《机械设计与制造》;20210915(第9期);第150-154页 *

Also Published As

Publication number Publication date
CN114203295A (en) 2022-03-18

Similar Documents

Publication Publication Date Title
CN114203295B (en) Cerebral apoplexy risk prediction intervention method and system
Patil et al. A comparative analysis on the evaluation of classification algorithms in the prediction of diabetes
CN111105860B (en) Accurate motion big data intelligent prediction, analysis and optimization system for chronic disease rehabilitation
Chakradar et al. A non-invasive approach to identify insulin resistance with triglycerides and HDL-c ratio using machine learning
Zohora et al. Forecasting the risk of type ii diabetes using reinforcement learning
Li et al. Identifying informative risk factors and predicting bone disease progression via deep belief networks
JP2006518062A (en) Prediction algorithm training and testing database optimization system and method
Afsaneh et al. Recent applications of machine learning and deep learning models in the prediction, diagnosis, and management of diabetes: a comprehensive review
CN111445978B (en) Motion scheme reactivity prediction model, modeling method and electronic equipment
CN111951975A (en) Sepsis early warning method based on deep learning model GPT-2
CN112932501A (en) Method for automatically identifying insomnia based on one-dimensional convolutional neural network
Barhate et al. Analysis of classifiers for prediction of type ii diabetes mellitus
Junaid et al. Explainable machine learning models based on multimodal time-series data for the early detection of Parkinson’s disease
Fakih et al. An efficient prediction of diabetes using artificial neural networks
Singh et al. Collaboration of features optimization techniques for the effective diagnosis of glaucoma in retinal fundus images
CN114300126A (en) Cancer prediction system based on early cancer screening questionnaire and feed-forward neural network
CN112712895B (en) Data analysis method of multi-modal big data aiming at type 2 diabetes complications
Saravanakumar et al. An effective convolutional neural network-based stacked long short-term memory approach for automated Alzheimer’s disease prediction
CN111916179A (en) Method for carrying out 'customized' diet nourishing model based on artificial intelligence self-adaption individual physical sign
CN115394448B (en) Modeling method, model and equipment of coronary heart disease motion reactivity prediction model
CN116898451A (en) Method for realizing atrial fibrillation prediction by using neural network with multi-scale attention mechanism
Chaturvedi et al. An Innovative Approach of Early Diabetes Prediction using Combined Approach of DC based Bidirectional GRU and CNN
Venkatrao et al. HDLNET: A Hybrid Deep Learning Network Model With Intelligent IOT for Detection and Classification of Chronic Kidney Disease
CN115547502A (en) Hemodialysis patient risk prediction device based on time sequence data
Kalaivani et al. Heart disease diagnosis using optimized features of hybridized ALCSOGA algorithm and LSTM classifier

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