CN116092668A - Prediction method for heart failure patient readmission fused with multi-element heterogeneous data - Google Patents

Prediction method for heart failure patient readmission fused with multi-element heterogeneous data Download PDF

Info

Publication number
CN116092668A
CN116092668A CN202310039139.6A CN202310039139A CN116092668A CN 116092668 A CN116092668 A CN 116092668A CN 202310039139 A CN202310039139 A CN 202310039139A CN 116092668 A CN116092668 A CN 116092668A
Authority
CN
China
Prior art keywords
patient
heart failure
readmission
patients
information
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
Application number
CN202310039139.6A
Other languages
Chinese (zh)
Inventor
车超
张未秀
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.)
Dalian University
Original Assignee
Dalian University
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 Dalian University filed Critical Dalian University
Priority to CN202310039139.6A priority Critical patent/CN116092668A/en
Publication of CN116092668A publication Critical patent/CN116092668A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
    • A61B5/7271—Specific aspects of physiological measurement analysis
    • A61B5/7275—Determining trends in physiological measurement data; Predicting development of a medical condition based on physiological measurements, e.g. determining a risk factor
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
    • A61B5/7235—Details of waveform analysis
    • A61B5/7264—Classification of physiological signals or data, e.g. using neural networks, statistical classifiers, expert systems or fuzzy systems
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
    • A61B5/7235—Details of waveform analysis
    • A61B5/7264—Classification of physiological signals or data, e.g. using neural networks, statistical classifiers, expert systems or fuzzy systems
    • A61B5/7267—Classification of physiological signals or data, e.g. using neural networks, statistical classifiers, expert systems or fuzzy systems involving training the classification device
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
    • A61B5/7271—Specific aspects of physiological measurement analysis
    • 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
    • G06N3/088—Non-supervised learning, e.g. competitive learning
    • 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
    • G16H10/00—ICT specially adapted for the handling or processing of patient-related medical or healthcare data
    • G16H10/60—ICT specially adapted for the handling or processing of patient-related medical or healthcare data for patient-specific data, e.g. for electronic patient records
    • 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
    • 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

Landscapes

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

Abstract

The invention provides a prediction method for re-hospital admission of heart failure patients fused with multi-component heterogeneous data; preprocessing the structured information and clinical discharge records in the electronic health records, and extracting all discharge record data of the patient; static information, time sequence data and clinical records are respectively generated into multi-mode characteristics of a patient through a one-hot code, a Doc2Vec and a transformerler Encoder module, and then the multi-mode characteristics are transmitted to a maximum pooling layer for splicing; calculating similarity among patients based on cosine similarity, constructing a patient network diagram, inputting the hidden features of the node features and edges of the patients after linear transformation into a GIN layer, and aggregating space-time features of a neighborhood to generate final features of the patients; and sending the extracted characteristic information of the patient into a classification model for training, storing the model, and predicting the patient to be admitted again. The invention effectively extracts the characteristic information of the heterogeneous data of the patient, has higher accuracy rate when the prediction of the hospital is carried out within 30 days, and effectively reduces the morbidity and mortality of heart failure patients.

Description

Prediction method for heart failure patient readmission fused with multi-element heterogeneous data
Technical Field
The invention relates to the technical field of medical artificial intelligence processing, in particular to a heart failure patient readmission prediction method fusing multivariate heterogeneous data.
Background
Heart failure is a epidemic disease, the worldwide number of patients is about 3770 ten thousand, although the mortality rate of heart failure is obviously reduced in recent years, the heart failure is still the most common disease in readmission total cause analysis, and related researches show that the readmission rate is 25% within 30 days. And high readmission rates can lead to shortage of hospital resources and economic burden on patients. Therefore, accurate prediction is performed according to the actual condition of a patient so as to judge whether the patient needs to be admitted again, and the medical resource shortage condition can be greatly relieved.
In recent years, medical data mining has made great progress in the context of rapid development of deep learning, clinical decision assistance and personalized diagnosis and treatment have also become the most interesting research directions for researchers, and recently research methods are mainly divided into two categories, namely an EHR-based time series data prediction method and a graph-based neural network prediction method. Both of these methods suffer from certain drawbacks. For example, EHR-based temporal data prediction methods often use only a single temporal data, ignoring the impact of mining patient multimode heterogeneous information on the prediction results. Most of the prediction methods based on the graph neural network use a simple graph neural network method, only the logical relations among different data are considered, and enough attention is not paid to the spatial structure of the graph nodes.
Disclosure of Invention
The invention aims to provide a model for predicting 30-day readmission of heart failure patients by fusing heterogeneous medical records, which improves the accuracy of a prediction result and can better assist doctors to make clinical diagnosis decisions.
In order to achieve the above purpose, the technical scheme of the application is as follows: a prediction method for heart failure patients fusing multiple heterogeneous data includes:
step 1: preprocessing the structured information in the electronic health record and the clinical discharge record, and extracting the patient calendar admission visit record;
step 2: generating multimode characteristics of a patient by using static information, time sequence data and clinical discharge records through a one-hot code, a Doc2Vec module and a transformerler encoding module respectively, and transmitting the multimode characteristics to a maximum pooling layer for splicing;
step 3: acquiring similarity among patients based on cosine similarity, constructing a patient network graph, inputting the hidden features of the node features and edges of the patient after linear transformation into a graph isomorphic network GIN, and aggregating space-time features of a neighborhood to generate patient feature information;
step 4: the extracted characteristic information of the patient is sent into a classification model for training, the model is stored, and the patient is admitted again for prediction;
step 5: and loading the model, inputting patient hospitalization information to be predicted, predicting the patient readmission situation, and outputting a prediction result.
Further, the step 1 specifically includes:
step 1.1: screening physiological detection indexes, population basic information and discharge records of admission and visit in the electronic health records, and deleting patient data with incomplete key indexes;
step 1.2: screening the processed data for heart failure patients according to ICD-9 codes 398.91, 402.01, 402.11, 402.91, 425.X,428. Xx;
step 1.3: the screened patients admitted again within 30 days are marked as 1, otherwise, are marked as 0;
step 1.4: the marked data set is according to the training set: verification set: test set was 8:1: 1.
Further, step 2 uses three encoders to respectively perform embedded representation on the patient's history of admissions to the patient, and specifically includes:
the time sequence encoder takes the historical medical time sequence record of the patient as input to obtain a characteristic F i ＝{f 1 ,f 2 ,...,f z -wherein z is the total number of features; by transformerlenThe encoder module generates health features H for a patient temproal ＝{h 1 ,h 2 ,...,h t -a }; the module captures internal dependencies in each of the sequences of visits using a self-attention mechanism and learns external links between the sequences of visits through a plurality of different attention heads; the attention function of the transformerlencoder module is defined as follows:
Figure BDA0004050467060000031
wherein the method comprises the steps of
Figure BDA0004050467060000032
Representing a query matrix->
Figure BDA0004050467060000033
Representing a key matrix, < >>
Figure BDA0004050467060000034
Representing a value matrix, wherein D represents a dimension;
further, the recording encoder pre-processes the clinical text information to form a patient medical text block C= { C 1 ，c 2 ，...，c k K is the total number of blocks; then, the Doc2Vec module is used for carrying out primary unsupervised learning representation on the text block to generate a paragraph hiding state vector as H notes ＝{h 1 ，h 2 ，...，h m }，h i ∈R 1×s Where m is the total number of paragraphs, m=k, s is the matrix dimension; the embedded representation of the different text blocks is obtained by a transformerlencoder module in the same manner as the time sequence encoder.
Further, the demographic information is pre-processed by the demographic encoder and then encoded into a one-hot hidden state vector H static ＝{h 1 ，h 2 ，...，h s }。
Further, the multimode characteristics of the patient are represented by health characteristics H temproal Paragraph hidden state vector H notes And one-hot hidden status directionQuantity H static Three parts are formed, the three hidden features are respectively subjected to dimension reduction through a maximum pooling layer, and final representation information of a patient is obtained: these final representation information are combined together and represented as:
Z patient ＝Concat(Z temporal ,Z static ,Z notes ). (2)。
further, the step 3 specifically includes:
step 3.1: first, a patient network graph g= (V, E) is constructed, where v= { h 1 ，h 2 ，...，h m As the only set of admission nodes, E is the set of all sides; using adjacency matrix A.epsilon.R |m|×|m| To illustrate the construction process of the figure: if the cosine similarity between the features of the ith case and the jth case exceeds a threshold of 0.99, it indicates that there is a border between i and j, i.e., a ij =1, otherwise a ij ＝0；
Step 3.2: and (3) taking the edge set and the vertex set obtained in the step (3.1) as the input of the graph isomorphic network GIN, and extracting the patient health condition characterization by using the graph isomorphic network GIN.
Further, the step 3.2 specifically includes:
step 3.2.1: the graph isomorphic network GIN not only learns the neighborhood characteristics, but also captures the space structure information among nodes, and in the GIN, the aggregation function of the k-th layer for aggregating and updating the node characteristics is as follows:
Figure BDA0004050467060000041
wherein f is a multiple set at a node, and phi represents a single-shot function; the GIN aggregation mode adopts SUM mode, and the uniqueness of the function is ensured through the MLP layer; the final vector representation of the patient based on the GIN framework of mlp+sum is therefore:
Figure BDA0004050467060000051
where ε is a learnable parameter.
Further, the step 4 specifically includes:
inputting the final vector representation and the label obtained in the step 3.2.1 into a fully connected neural network, and training a classification model; optimizing the classification model by adopting a binary cross entropy function, and storing a model_best with the best effect:
Figure BDA0004050467060000052
further, the step 5 specifically includes:
and loading a model_best, inputting patient data in the verification data into the model, judging whether the patient is admitted again within 30 days, and outputting corresponding evaluation indexes.
By adopting the technical scheme, the invention can obtain the following technical effects: the invention adopts a deep learning model, utilizes the structured data and unstructured data of the patient, and automatically predicts the readmission situation of the heart failure patient through the model. The method effectively digs the characteristic information of the patient multi-element heterogeneous data, has higher accuracy when the patient is readmission predicted within 30 days, and can assist doctors to better carry out clinical diagnosis decision so as to effectively reduce the morbidity and mortality of heart failure patients.
Drawings
Fig. 1 is a flowchart of a method for predicting readmission of heart failure patients fused with multivariate heterogeneous data.
Detailed Description
The implementation of the invention is carried out on the premise of the technical scheme of the invention, and detailed implementation and specific operation processes are given, but the protection scope of the invention is not limited to the following examples.
The present invention is described in detail below with reference to examples so that those skilled in the art can practice the same with reference to the present specification.
Example 1
In the embodiment, a Windows system is used as a development environment, pycharm is used as a development platform, python is used as a development language, and the prediction of the heart failure patient is carried out within 30 days.
In this embodiment, a prediction method for re-hospital admission of heart failure patients with fusion of multiple heterogeneous data includes the following steps:
screening out the history information of the visit by giving one patient information, wherein the history information comprises 29 physiological check indexes, 6 static population information and a summary of patient discharge records, if a readmission record exists within 30 days, setting a data tag to be 1, otherwise setting the data tag to be 0; and taking the multi-component heterogeneous information of the patient as input, loading the stored model, and obtaining an evaluation index of a patient prediction result, wherein the evaluation index comprises an area under a False Positive Rate (FPR) -true rate (TPR) curve (AUROC), a Recall rate (Recall) -Precision rate (Precision) curve (AUPRC) and F1.TPR, FPR, recall, precision, F1 is defined as follows:
Figure BDA0004050467060000061
Figure BDA0004050467060000062
/>
Figure BDA0004050467060000063
Figure BDA0004050467060000064
where TP represents the number of patients correctly predicted to be readmitted, TN represents the number of patients correctly identified to be non-readmitted, FP represents the number of patients labeled to be readmitted but not correctly predicted in the experiment, and FN represents the number of non-readmitted patients not correctly identified in the experiment.
According to the above steps, the present invention compares the readmission prediction effect with the LR model, fusion-LSTM, fusion-CNN model, clinic bert model, deep-GNN model, and TransMT model. As can be seen from Table 1, the methods presented herein are significantly better in AUROC, AUPRC and F1 values than the other methods.
Table 1 comparison of different models against readmission prediction results
Figure BDA0004050467060000071
The foregoing description of specific embodiments of the invention has been presented to illustrate specific steps of implementation. It is not intended to limit the invention to the precise form disclosed, and obviously many modifications and variations are possible in light of the above teaching. The exemplary embodiments were chosen and described in order to explain the specific principles of the invention and its practical application to thereby enable one skilled in the art to make and utilize the invention in various exemplary embodiments and with various modifications as are suited to the particular use contemplated. It is intended that the scope of the invention be defined by the claims and their equivalents.

Claims (10)

1. The prediction method for the readmission of the heart failure patient fused with the multi-element heterogeneous data is characterized by comprising the following steps of:
step 1: preprocessing the structured information in the electronic health record and the clinical discharge record, and extracting the patient calendar admission visit record;
step 2: generating multimode characteristics of a patient by using static information, time sequence data and clinical discharge records through a one-hot code, a Doc2Vec module and a transformerler encoding module respectively, and transmitting the multimode characteristics to a maximum pooling layer for splicing;
step 3: acquiring similarity among patients based on cosine similarity, constructing a patient network graph, inputting the hidden features of the node features and edges of the patient after linear transformation into a graph isomorphic network GIN, and aggregating space-time features of a neighborhood to generate patient feature information;
step 4: the extracted characteristic information of the patient is sent into a classification model for training, the model is stored, and the patient is admitted again for prediction;
step 5: and loading the model, inputting patient hospitalization information to be predicted, predicting the patient readmission situation, and outputting a prediction result.
2. The method for predicting readmission of heart failure patients fused with multivariate heterogeneous data according to claim 1, wherein the step 1 specifically comprises:
step 1.1: screening physiological detection indexes, population basic information and discharge records of admission and visit in the electronic health records, and deleting patient data with incomplete key indexes;
step 1.2: screening the processed data for heart failure patients according to ICD-9 codes 398.91, 402.01, 402.11, 402.91, 425.X,428. Xx;
step 1.3: the screened patients admitted again within 30 days are marked as 1, otherwise, are marked as 0;
step 1.4: the marked data set is according to the training set: verification set: test set was 8:1: 1.
3. The method for predicting readmission of heart failure patients with fusion of multivariate heterogeneous data according to claim 1, wherein step 2 uses three encoders to respectively embed and represent the historic readmission records of the patients, and specifically comprises:
the time sequence encoder takes the historical medical time sequence record of the patient as input to obtain a characteristic F i ＝{f 1 ,f 2 ,...,f z -wherein z is the total number of features; generating health features H of a patient by a TransformaerEncoder Module temproal ＝{h 1 ,h 2 ,...,h t -a }; the module captures internal dependencies in each of the sequences of visits using a self-attention mechanism and learns external links between the sequences of visits through a plurality of different attention heads; the attention function of the transformerlencoder module is defined as follows:
Figure FDA0004050467050000021
wherein the method comprises the steps of
Figure FDA0004050467050000022
Representing a query matrix->
Figure FDA0004050467050000023
Representing a key matrix, < >>
Figure FDA0004050467050000024
Represents a matrix of values, D represents a dimension.
4. The method for predicting readmission of heart failure patient fused with multi-component heterogeneous data according to claim 3, wherein the recording encoder pre-processes the clinical text information to form a patient medical text block c= { C 1 ，c 2 ，...，c k K is the total number of blocks; then, the Doc2Vec module is used for carrying out primary unsupervised learning representation on the text block to generate a paragraph hiding state vector as H notes ＝{h 1 ，h 2 ，...，h m }，h i ∈R 1×s Where m is the total number of paragraphs, m=k, s is the matrix dimension; the embedded representation of the different text blocks is obtained by a transformerlencoder module in the same manner as the time sequence encoder.
5. The method for predicting readmission of heart failure patient with fused multivariate heterogeneous data as claimed in claim 3, wherein the demographic information is preprocessed by the demographic encoder and then encoded into one-hot hidden state vector H static ＝{h 1 ，h 2 ，...，h s }。
6. The method for predicting readmission of heart failure patient based on multi-component heterogeneous data as claimed in claim 3, wherein the multi-mode characteristics of the patient are represented by health characteristics H temproal Paragraph hidden state vector H notes And one-hot hidden status directionQuantity H static Three parts are formed, the three hidden features are respectively subjected to dimension reduction through a maximum pooling layer, and final representation information of a patient is obtained: these final representation information are combined together and represented as:
Z patient ＝Concat(Z temporal ,Z static ,Z notes ). (2)。
7. the method for predicting readmission of heart failure patients based on multi-component heterogeneous data according to claim 1, wherein the step 3 specifically comprises:
step 3.1: first, a patient network graph g= (V, E) is constructed, where v= { h 1 ，h 2 ，...，h m As the only set of admission nodes, E is the set of all sides; using adjacency matrix A.epsilon.R |m|×|m| To illustrate the construction process of the figure: if the cosine similarity between the features of the ith case and the jth case exceeds a threshold, it indicates that there is a border between i and j, i.e., a ij =1, otherwise a ij ＝0；
Step 3.2: and (3) taking the edge set and the vertex set obtained in the step (3.1) as the input of the graph isomorphic network GIN, and extracting the patient health condition characterization by using the graph isomorphic network GIN.
8. The method for predicting readmission of heart failure patients with fusion of multivariate heterogeneous data according to claim 7, wherein the step 3.2 specifically comprises:
step 3.2.1 in GIN, the aggregation function of the k-th layer to aggregate and update node characteristics is as follows:
Figure FDA0004050467050000031
wherein f is a multiple set at a node, and phi represents a single-shot function; the GIN aggregation mode adopts SUM mode, and the uniqueness of the function is ensured through the MLP layer; the final vector representation of the patient based on the GIN framework of mlp+sum is therefore:
Figure FDA0004050467050000041
where ε is a learnable parameter.
9. The method for predicting readmission of heart failure patients with fusion of multivariate heterogeneous data according to claim 8, wherein the step 4 specifically comprises:
inputting the final vector representation and the label obtained in the step 3.2.1 into a fully connected neural network, and training a classification model; optimizing the classification model by adopting a binary cross entropy function, and storing a model_best with the best effect:
Figure FDA0004050467050000042
10. the method for predicting readmission of heart failure patients with fusion of multivariate heterogeneous data according to claim 9, wherein the step 5 specifically comprises:
and loading a model_best, inputting patient data in the verification data into the model, judging whether the patient is admitted again within 30 days, and outputting corresponding evaluation indexes.
CN202310039139.6A 2023-01-13 2023-01-13 Prediction method for heart failure patient readmission fused with multi-element heterogeneous data Pending CN116092668A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202310039139.6A CN116092668A (en) 2023-01-13 2023-01-13 Prediction method for heart failure patient readmission fused with multi-element heterogeneous data

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202310039139.6A CN116092668A (en) 2023-01-13 2023-01-13 Prediction method for heart failure patient readmission fused with multi-element heterogeneous data

Publications (1)

Publication Number Publication Date
CN116092668A true CN116092668A (en) 2023-05-09

Family

ID=86213492

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202310039139.6A Pending CN116092668A (en) 2023-01-13 2023-01-13 Prediction method for heart failure patient readmission fused with multi-element heterogeneous data

Country Status (1)

Country Link
CN (1) CN116092668A (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116959715A (en) * 2023-09-18 2023-10-27 之江实验室 Disease prognosis prediction system based on time sequence evolution process explanation

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116959715A (en) * 2023-09-18 2023-10-27 之江实验室 Disease prognosis prediction system based on time sequence evolution process explanation
CN116959715B (en) * 2023-09-18 2024-01-09 之江实验室 Disease prognosis prediction system based on time sequence evolution process explanation

Similar Documents

Publication Publication Date Title
CN107977361B (en) Chinese clinical medical entity identification method based on deep semantic information representation
Sun et al. Disease prediction via graph neural networks
CN113241135B (en) Disease risk prediction method and system based on multi-modal fusion
CN109637669B (en) Deep learning-based treatment scheme generation method, device and storage medium
US11581067B2 (en) Method and apparatus for generating a chemical structure using a neural network
US8996428B2 (en) Predicting diagnosis of a patient
US20210287141A1 (en) Training Diverse and Robust Ensembles of Artificial Intelligence Computer Models
CN104572583B (en) Method and system for data densification
CN113396457A (en) System, method and apparatus for biophysical modeling and response prediction
US10936628B2 (en) Automatic processing of ambiguously labeled data
Ma et al. A general framework for diagnosis prediction via incorporating medical code descriptions
US11670421B2 (en) Method and system enabling digital biomarker data integration and analysis for clinical treatment impact
Gupta et al. A novel deep similarity learning approach to electronic health records data
CN116092668A (en) Prediction method for heart failure patient readmission fused with multi-element heterogeneous data
CN114300081B (en) Prediction device, system and storage medium based on electronic medical record multi-modal data
CN109887606B (en) Attention-based diagnosis and prediction method for bidirectional recurrent neural network
Wu et al. MEGACare: Knowledge-guided multi-view hypergraph predictive framework for healthcare
CN116403728B (en) Data processing device for medical treatment data and related equipment
Murad et al. AI powered asthma prediction towards treatment formulation: An android app approach
CN116884612A (en) Intelligent analysis method, device, equipment and storage medium for disease risk level
Sengupta et al. Analyzing historical diagnosis code data from NIH N3C and RECOVER Programs using deep learning to determine risk factors for Long Covid
Lu et al. Multi-modal Intermediate Fusion Model for diagnosis prediction
Theodorou et al. Synthesize extremely high-dimensional longitudinal electronic health records via hierarchical autoregressive language model
Landi et al. The evolution of mining electronic health records in the era of deep learning
CN117095811B (en) Prediction method, device and storage medium based on electronic medical case data

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