CN110164519A - A kind of classification method for being used to handle electronic health record blended data based on many intelligence networks - Google Patents
A kind of classification method for being used to handle electronic health record blended data based on many intelligence networks Download PDFInfo
- Publication number
- CN110164519A CN110164519A CN201910372303.9A CN201910372303A CN110164519A CN 110164519 A CN110164519 A CN 110164519A CN 201910372303 A CN201910372303 A CN 201910372303A CN 110164519 A CN110164519 A CN 110164519A
- Authority
- CN
- China
- Prior art keywords
- data
- electronic health
- health record
- convolutional neural
- type data
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
- G06F16/21—Design, administration or maintenance of databases
- G06F16/215—Improving data quality; Data cleansing, e.g. de-duplication, removing invalid entries or correcting typographical errors
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/35—Clustering; Classification
- G06F16/355—Class or cluster creation or modification
-
- 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
- G16H50/70—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for mining of medical data, e.g. analysing previous cases of other patients
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Theoretical Computer Science (AREA)
- Databases & Information Systems (AREA)
- Public Health (AREA)
- Health & Medical Sciences (AREA)
- Medical Informatics (AREA)
- Epidemiology (AREA)
- General Health & Medical Sciences (AREA)
- Primary Health Care (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Pathology (AREA)
- Quality & Reliability (AREA)
- Biomedical Technology (AREA)
- Medical Treatment And Welfare Office Work (AREA)
- Measuring And Recording Apparatus For Diagnosis (AREA)
Abstract
The classification method for being used to handle electronic health record blended data based on many intelligence networks that the present invention relates to a kind of, can effectively using have processing numeric type data method, the classification effectiveness of the blended data in electronic health record is improved, doctor is helped to improve the quality and efficiency of diagnosis.Method includes: the original electron medical record data collection in the step 1. extraction original electron database of case history;Step 2, the judgement of character type data is carried out to the electronic health record data set after cleaning, and symbol data is carried out with the conversion of numeric type data;Step 3. extracts original electron medical record data and concentrates key feature field, and training classification diagnosis model;Step 4. extracts key feature field in electronic health record to be diagnosed, and is input in trained model, output category result and probability of illness.The present invention can effectively excavate valuable information in electronic health record and help doctor's quick diagnosis state of an illness, have important theory significance and applied value.
Description
Technical field
The present invention relates to a kind of many intelligence networks and electronic health record and big data fields.It is a kind of processing processing blended data
Analysis method, can mixed type data to magnanimity electronic health record carry out classification processing.
Background technique
Many intelligence networks (Crowd Network) are to probe into this swarm intelligence under thread environment on a large scale of many intelligence science
The large scale emulation and experiment porch that activity is built.In many intelligence networks, the heterogeneous interconnection such as people, machine, article, while in net
There is different types of data in network.Electronic health record (EMR, Electronic Medical Record) also named computerization
Medical record system or computer based patient record.It is with electronic equipment (computer, health card etc.) save, management,
Transmission and the digitized medical records reappeared, to replace hand-written paper case history.Its content includes all of paper case history
Information.US National Institute for Medical Research will be defined as: EMR is the electronic patient record based on a particular system, the system
The user data, the ability of warning, prompt and Clinical Decision Support Systems that access complete and accurate, the spy of electronic health record text are provided
Point be there is more field term phrase and abbreviation, and due to patient category diversification, but its data type mostly with
Based on data type and font data.
For numeric type data, all data can be converted into the vector in theorem in Euclid space, several with clearly space
What structure, the similarity degree or difference degree between data are measured by Euclidean distance or COS distance etc..Its correlation is ground
Study carefully and have been achieved for very significant effect, produce many effective algorithms, as SVM algorithm, convolutional neural networks algorithm,
KNN algorithm.
For character type data, compared to general text, the character of the data value finite state of the type, classification or
Person's numerical value.Such as patient blood type (A type, Type B, AB type are O-shaped), way of paying (medical insurance, at one's own expense etc.), type of credential (residential identity
Card, Hong Kong, Macao and Taiwan identity card, passport etc.) and whether there is or not passing medical histories (being not have) etc..This kind of data generally cannot directly into
Row numerical operation.The method of process symbol type data is also seldom at present.
With the arrival of many intelligence cybertimes and electronic health record epoch, more and more hospitals (doctor) are added to many intelligence nets
Network, while more hospitals are ready that the flower more sub- case history of more options electricity consumption replaces the hand written case histories of trivial operations.In many intelligence nets
In network, more and more doctors can carry out information exchange with other doctors in a network.Increase understanding of the doctor to patient, has
The quality of the promotion medical level of effect.So how electronic health record medical resource abundant and numeric type method are effectively utilized,
Effective information in electronic health record is excavated, help is provided for the diagnosis of doctor, is one of the hot spot solved instantly.
Summary of the invention
The classification for being used to handle electronic health record blended data based on many intelligence networks that the purpose of the present invention is to provide a kind of
Method.
How another object of the present invention improves in many intelligence networks effectively using having processing numeric type data method
The classification effectiveness of the blended data of electronic health record;
In order to achieve the above objectives, technical scheme is as follows:
A kind of classification method for being used to handle electronic health record blended data based on many intelligence networks, comprising:
Step 1. extracts the original electron medical record data collection in many intelligence networks in electronic health record database, carries out to data set
Data cleansing;
Electronic health record data set after step 2. pair cleaning carries out the judgement of character type data, and symbol data need to be into if it exists
Row character type data turn the conversion of numeric type data;
Step 3. extracts original electron medical record data and concentrates key feature field, generates training sample, and training sample is defeated
Enter into convolutional neural networks, is trained and generates subsidiary classification diagnostic model;
Step 4. extracts key feature field in electronic health record to be diagnosed, and generates test sample, test sample is inputted
Into trained convolutional neural networks, the probability of illness of output category result and each illness;
Further, described step 1 extracts the original electron medical record data collection in many intelligence networks in electronic health record database,
Data cleansing is carried out to data set:
Due to the electronic health record data set in the original electron database of case history, there may be unit difference, words for data set
Section redundancy, has the problems such as invalid value and missing values in data, for the consistency and correctness for guaranteeing data, so data need to be carried out
Cleaning;
Further, the electronic health record data set after described step 2 pair cleaning carries out the judgement of character type data, accords with if it exists
Number property data need to carry out the conversion that character type data turn numeric type data:
Firstly, for a large amount of character type numerical value present in electronic health record, character type data dictionary, data dictionary need to be constructed
(part) form is as shown in the table:
If in data set, for character present in dictionary, then needing to use one-hot coding (One-Hot) by character type data
It indicates in theorem in Euclid space, is translated into 0 and 1 and is the coding of composition, such as convert 00 for type of credential resident identification card,
Hong Kong, Macao and Taiwan resident identification card is converted into 01, and resident's residence booklet is 10, passport 11;
Secondly, calculating the frequency relation (similitude) between attribute and label by mutual information and conditional entropy method.This
Patent defines two different correlation calculations method I () and H ().Attribute ajProperties value ajkWith the frequency of label c
Relationship,WithCalculation method is as follows:
Wherein, p (yc) indicate attribute ajkProbability, p (yc) label yiValue be c probability, p (yc,ajk) and p (yc|
ajk) it is ajkWith ycJoint probability and conditional probability.
In the label and attribute value frequency relation for obtaining electronic health record, according to the law of large numbers, frequency of use carrys out approximate representation
Corresponding similarity relation:
Calculation method is as follows:
Wherein, IF () here is an indicator function, i.e. IF (true)=1, IF (false)=0.
By the available two attribute value a of formula (1), (2) and (3)jkWith label ycCorrelation matrix O-I (ajk)
Representation is as follows, realizes and converts numeric type data for character type data.Expression process is as follows:
O-H(ajk) expanded form and O-I (ajk) similar.It is expressed as follows:
This patent is using frequency relation between label and attribute as a result, approximate similarity degree between it, to realize
Character type data are converted into numeric type data.
Further, described step 3 extracts original electron medical record data and concentrates key feature field, generates training sample,
Training sample is input in convolutional neural networks, is trained and generates subsidiary classification diagnostic model, steps are as follows for specific step:
Feature extraction is carried out to the electronic health record data set after conversion, extracts key feature field of the doctor in diagnosis
(critical field includes patient main suit and passing medical history etc.);
Word segmentation processing is carried out to the key feature field at previous step extraction, generates training sample;
Generated training sample is input in convolutional neural networks model and is trained, constantly return adjusting parameter into
Row right value update generates subsidiary classification diagnostic model to reduce error.
Convolutional neural networks model used for previous step, selects 6 layers of convolutional neural networks model, comprising: input layer,
Convolutional layer, fused layer, pond layer, Softmax layers and output layer.Feature extraction is carried out using different size of convolution kernel, it will
The different characteristic of extraction is merged reaches next layer again.K is used in this patent setting1,k2Indicate convolution kernel size, value range
It is set as [1,5], b=0, model initialization parameterWherein k=k1Or k2.Convolution
After layer operation, wherein characteristic sequence F is plus biasing b, and is mapped with Relu function;
The generation of the phenomenon that prevent model from over-fitting or poor fitting occur uses 5 folding cross validation methods;
Specifically, this patent in the training process, by preset algorithms such as back-propagation algorithms (BP algorithm), updates
The parameter and classifier parameters of convolutional neural networks model.
Finally, judge whether the training error of the convolutional neural networks model after the training is greater than the error of true value,
If being less than, deconditioning;If more than then adjusting preset ratio, be trained again to model.This patent uses cross entropy letter
Number measures model predication value and true value y as loss function Li;
Further, described step 4 extracts key feature field in electronic health record to be diagnosed, and generates test sample, will
Test sample is input in trained convolutional neural networks, and the probability of illness of output category result and each illness includes:
Key feature field is extracted to electronic health record to be detected;
Word segmentation processing is carried out to the feature field at extraction, obtains the test set of electronic health record;
Generated test set is input to, in the convolutional neural networks model being trained to, and selects Softmax points
Class device is classified, and finally output obtains the probability of illness of each illness;
The present invention has the beneficial effect that:
The classification method for being used to handle electronic health record blended data based on many intelligence networks that the present invention relates to a kind of, using
There is processing numeric type data method, improves the classification effectiveness of the blended data in electronic health record.Firstly, utilizing one-hot coding
(One-Hot) algorithm indicates the symbol data in electronic health record in theorem in Euclid space, utilizes two kinds of sides of mutual information and conditional entropy
Method probes into the frequency relation (correlation) between characteristic attribute and label, successfully converts numeric type for character type data, and
The data classification to electronic health record is realized using convolutional neural networks, so that doctor is helped to improve the efficiency and quality diagnosed, with
Achieve the purpose that " auxiliary diagnosis ".
Detailed description of the invention
Fig. 1 is a kind of flow chart of classification method for handling electronic health record blended data based on many intelligence networks
Fig. 2 is electronic health record schematic diagram disclosed in a network
Specific embodiment
To keep the purpose of the present invention, technical solution and better effect explicit, in conjunction with attached drawing to of the invention further detailed
It describes in detail bright.It should be appreciated that specific implementation described herein is only used to explain the present invention, it is not intended to limit the present invention.
The reality for the classification method for handling electronic health record blended data based on many intelligence networks that the present invention provides a kind of
Flow chart is applied, as shown in Figure 1, electronic health record schematic diagram is Fig. 2, process includes:
Step 1. extracts the original electron medical record data collection in many intelligence networks in electronic health record database, carries out to data set
Data cleansing;
Electronic health record data set after step 2. pair cleaning carries out the judgement of character type data, and symbol data need to be into if it exists
Row character type data turn the conversion of numeric type data;
Step 3. extracts original electron medical record data and concentrates key feature field, generates training sample, and training sample is defeated
Enter into convolutional neural networks, is trained and generates subsidiary classification diagnostic model;
Step 4. extracts key feature field in electronic health record to be diagnosed, and generates test sample, test sample is inputted
Into trained convolutional neural networks, the probability of illness of output category result and each illness;
Step 1. extracts the original electron medical record data collection in many intelligence networks in electronic health record database, carries out to data set
Data cleansing:
Due to the electronic health record data set in the original electron database of case history, there may be unit difference, words for data set
Section redundancy, has the problems such as invalid value and missing values in data, for the consistency and correctness for guaranteeing data, so data need to be carried out
Cleaning;
This patent carries out data processing using data scrubbing software DataWrangler.
Electronic health record data set after step 2 pair cleaning carries out the judgement of character type data, and symbol data need to be into if it exists
Row character type data turn the conversion of numeric type data:
Firstly, for a large amount of character types numerical value present in electronic health record (such as :), data dictionary, data word need to be constructed
Allusion quotation (part) form is as shown in the table:
If in data set, for character present in dictionary, then needing to use one-hot coding (One-Hot) by character type data
It indicates in theorem in Euclid space, is translated into 0 and 1 and is the coding of composition, such as convert 00 for type of credential resident identification card,
Hong Kong, Macao and Taiwan resident identification card is converted into 01, and resident's residence booklet is 10, passport 11;
Secondly, calculating the frequency relation (similitude) between attribute and label by mutual information and conditional entropy method.This
Patent defines two different correlation calculations method I () and H ().Attribute ajProperties value ajkWith the frequency of label c
Relationship,WithCalculation method is as follows:
Wherein, p (yc) indicate attribute ajkProbability, p (yc) label yiValue be c probability, p (yc,ajk) and p (yc|
ajk) it is ajkWith ycJoint probability and conditional probability.
In the label and attribute value frequency relation for obtaining electronic health record, according to the law of large numbers, frequency of use carrys out approximate representation
Corresponding similarity relation:
Calculation method is as follows:
Wherein, IF () here is an indicator function, i.e. IF (true)=1, IF (false)=0.
By the available two attribute value a of formula (1), (2) and (3)jkWith label ycCorrelation matrix O-I (ajk)
Representation is as follows, realizes and converts numeric type data for character type data.Expression process is as follows:
O-H(ajk) expanded form and O-I (ajk) similar.It is expressed as follows:
This patent is using frequency relation between label and attribute as a result, approximate similarity degree between it, to realize
Character type data are converted into numeric type data.
Step 3 extracts original electron medical record data and concentrates key feature field, generates training sample, training sample is inputted
It into convolutional neural networks, is trained and generates subsidiary classification diagnostic model, steps are as follows for specific step:
Feature extraction is carried out to the electronic health record data set after conversion, extracts key feature field of the doctor in diagnosis
(critical field includes patient main suit and passing medical history etc.);
Word segmentation processing is carried out to the key feature field at previous step extraction, generates training sample;
Generated training sample is input in convolutional neural networks model and is trained, constantly return adjusting parameter into
Row right value update generates subsidiary classification diagnostic model to reduce error.
Convolutional neural networks model used for previous step, selects 6 layers of convolutional neural networks model, comprising: input layer,
Convolutional layer, fused layer, pond layer, Softmax layers and output layer.Feature extraction is carried out using different size of convolution kernel, it will
The different characteristic of extraction is merged reaches next layer again.K is used in this patent setting1,k2Indicate convolution kernel size, value range
It is set as [1,5], b=0, model initialization parameterWherein k=k1Or k2.Convolution
After layer operation, wherein characteristic sequence F is plus biasing b, and is mapped with Relu function, and calculation formula is as follows:
Relu (F)=max (0, F+b)
The generation of the phenomenon that prevent model from over-fitting or poor fitting occur, this patent use 5 folding cross validation methods;
Specifically, this patent in the training process, by preset algorithms such as back-propagation algorithms (BP algorithm), updates
The parameter and classifier parameters of convolutional neural networks model.
Finally, judge whether the training error of the convolutional neural networks model after the training is greater than the error of true value,
If being less than, deconditioning;If more than then adjusting preset ratio, be trained again to model.This patent uses cross entropy letter
Number measures model predication value and true value y as loss function Li, calculation formula is as follows:
Step 4 extracts key feature field in electronic health record to be diagnosed, and generates test sample, test sample is input to
In trained convolutional neural networks, the probability of illness of output category result and each illness includes:
Key feature field is extracted to electronic health record to be detected;
Word segmentation processing is carried out to the feature field at extraction, obtains the test set of electronic health record;
Generated test set is input to, in the convolutional neural networks model being trained to, and selects Softmax points
Class device is classified, and finally output obtains the probability of illness of each illness, and calculation formula is as follows;
Wherein, piIndicate the probability of i-th of illness of the electronic health record diagnosed, yiIndicate i-th of the element and electronics of y
The corresponding feature vector of i-th of illness, y in case historyjIndicate j-th of element of y, i.e., j-th of illness is corresponding in electronic health record
Feature vector.
With the above-mentioned ideal according to invention, example is enlightenment in real time, and through the above description, relevant staff is complete
Various changes and amendments can be carried out without departing from the scope of the technological thought of the present invention'.This invention it is technical
Range is not limited to the contents of the specification, it is necessary to which the technical scope thereof is determined according to the scope of the claim.
Claims (5)
1. it is a kind of based on many intelligence networks for handling the classification method of electronic health record blended data, which is characterized in that including with
Lower step: step 1. extracts the original electron medical record data collection in the original electron database of case history, and it is clear to carry out data to data set
It washes;
Electronic health record data set after step 2. pair cleaning carries out the judgement of character type data, and symbol data need to be accorded with if it exists
Number type data turn the conversion of numeric type data;
Step 3. extracts original electron medical record data and concentrates key feature field, generates training sample, training sample is input to
In convolutional neural networks, it is trained and generates subsidiary classification diagnostic model;
Step 4. extracts key feature field in electronic health record to be diagnosed, and generates test sample, test sample is input to instruction
In the convolutional neural networks perfected, the probability of illness of output category result and each illness.
2. a kind of classification method for being used to handle electronic health record blended data based on many intelligence networks according to claim 1,
It is characterized by:
Described step 1 extracts the original electron medical record data collection in the original electron database of case history, carries out data to data set
Cleaning;
There may be unit difference, field redundancies to have invalid value and missing values problem in data for data set, is guarantee data one
Cause property and correctness carry out data cleansing.
3. a kind of classification method for being used to handle electronic health record blended data based on many intelligence networks according to claim 1,
It is characterized by:
Electronic health record data set after described step 2 pair cleaning carries out the judgement of character type data, and symbol data need if it exists
The conversion that character type data turn numeric type data is carried out, specific features are as follows:
(1) firstly, for a large amount of character type numerical value present in electronic health record, character type data dictionary need to be constructed;
(2) it if in data set, for character present in dictionary, then needs to be indicated character type data European with one-hot coding
In space, being translated into 0 and 1 is the coding formed;
(3) secondly, calculating the frequency relation between attribute and label i.e. similitude by mutual information and conditional entropy method;Definition
Two different correlation calculations method I () and H ();Attribute ajProperties value ajkWith the frequency relation of label c,WithCalculation method is as follows:
Wherein, p (yc) indicate attribute ajkProbability, p (yc) label yiValue be c probability, p (yc,ajk) and p (yc|ajk) be
ajkWith ycJoint probability and conditional probability;
(4) label and attribute value frequency relation of electronic health record are obtained again, and according to the law of large numbers, frequency of use carrys out approximate representation phase
The similarity relation answered:
Wherein, IF () here is an indicator function, i.e. IF (true)=1, IF (false)=0;
Two attribute value a are obtained by formula (1), (2) and (3)jkWith label ycCorrelation matrix O-I (ajk) representation is such as
Under, it realizes and converts numeric type data for character type data;Expression process is as follows:
[0077]O-H(ajk) expanded form and O-I (ajk) similar;It is expressed as follows:
Using the frequency relation between label and attribute, approximate similarity degree between it, to realize that character type data convert
For numeric type data.
4. a kind of classification method for being used to handle electronic health record blended data based on many intelligence networks according to claim 1,
It is characterized by:
Retouched step 3 extracts original electron medical record data and concentrates key feature field, generates training sample, training sample is inputted
It into convolutional neural networks, is trained and generates subsidiary classification diagnostic model, steps are as follows for specific step:
(1) feature extraction is carried out to the electronic health record data set after conversion, extracts key feature field of the doctor in diagnosis, closes
Key field includes patient main suit and passing medical history;
(2) word segmentation processing is carried out to the key feature field at previous step extraction, generates training sample;
(3) generated training sample is input in convolutional neural networks model and is trained, constantly return adjusting parameter into
Row right value update generates subsidiary classification diagnostic model to reduce error;
(4) convolutional neural networks model used for previous step, selects 6 layers of convolutional neural networks model, comprising: input layer,
Convolutional layer, fused layer, pond layer, Softmax layers and output layer;Feature extraction is carried out using different size of convolution kernel, it will
The different characteristic of extraction is merged reaches next layer again;K is used in setting1,k2Indicate that convolution kernel size, value range are set as
[1,5], b=0, model initialization parameterWherein k=k1Or k2, convolution layer operation
Afterwards, wherein characteristic sequence F adds biasing b, and is mapped with Relu function;
(5) 5 folding cross validation methods are used;
(6) in the training process, by back-propagation algorithm, the parameter and classifier parameters of convolutional neural networks model are updated;
(7) finally, judging whether the training error of the convolutional neural networks model after the training is greater than the error of true value, if
It is less than, then deconditioning;If more than the error of true value, then preset ratio is adjusted, model is trained again;Using intersection
Entropy function measures model predication value and true value y as loss function Li。
5. a kind of classification method for being used to handle electronic health record blended data based on many intelligence networks according to claim 1,
It is characterized by:
Described step 4 extracts key feature field in electronic health record to be diagnosed, and generates test sample, test sample is inputted
Into trained convolutional neural networks, the probability of illness of output category result and each illness, feature includes:
(1) key feature field is extracted to electronic health record to be detected;
(2) word segmentation processing is carried out to the feature field at extraction, obtains the test set of electronic health record;
(3) generated test set is input to, in the convolutional neural networks model being trained to, and selects Softmax points
Class device is classified, and finally output obtains the probability of illness of each illness.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910372303.9A CN110164519B (en) | 2019-05-06 | 2019-05-06 | Classification method for processing electronic medical record mixed data based on crowd-sourcing network |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910372303.9A CN110164519B (en) | 2019-05-06 | 2019-05-06 | Classification method for processing electronic medical record mixed data based on crowd-sourcing network |
Publications (2)
Publication Number | Publication Date |
---|---|
CN110164519A true CN110164519A (en) | 2019-08-23 |
CN110164519B CN110164519B (en) | 2021-08-06 |
Family
ID=67633493
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910372303.9A Active CN110164519B (en) | 2019-05-06 | 2019-05-06 | Classification method for processing electronic medical record mixed data based on crowd-sourcing network |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN110164519B (en) |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111128375A (en) * | 2020-01-10 | 2020-05-08 | 电子科技大学 | Tibetan medicine diagnosis auxiliary device based on multi-label learning |
WO2021114635A1 (en) * | 2020-05-13 | 2021-06-17 | 平安科技(深圳)有限公司 | Patient grouping model constructing method, patient grouping method, and related device |
WO2021120934A1 (en) * | 2019-12-18 | 2021-06-24 | 浙江大学 | Convolutional neural network-based method for automatically grouping drgs |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106845529A (en) * | 2016-12-30 | 2017-06-13 | 北京柏惠维康科技有限公司 | Image feature recognition methods based on many visual field convolutional neural networks |
CN107833629A (en) * | 2017-10-25 | 2018-03-23 | 厦门大学 | Aided diagnosis method and system based on deep learning |
CN107958257A (en) * | 2017-10-11 | 2018-04-24 | 华南理工大学 | A kind of Chinese traditional medicinal materials recognition method based on deep neural network |
CN109378066A (en) * | 2018-12-20 | 2019-02-22 | 翼健(上海)信息科技有限公司 | A kind of control method and control device for realizing disease forecasting based on feature vector |
-
2019
- 2019-05-06 CN CN201910372303.9A patent/CN110164519B/en active Active
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106845529A (en) * | 2016-12-30 | 2017-06-13 | 北京柏惠维康科技有限公司 | Image feature recognition methods based on many visual field convolutional neural networks |
CN107958257A (en) * | 2017-10-11 | 2018-04-24 | 华南理工大学 | A kind of Chinese traditional medicinal materials recognition method based on deep neural network |
CN107833629A (en) * | 2017-10-25 | 2018-03-23 | 厦门大学 | Aided diagnosis method and system based on deep learning |
CN109378066A (en) * | 2018-12-20 | 2019-02-22 | 翼健(上海)信息科技有限公司 | A kind of control method and control device for realizing disease forecasting based on feature vector |
Non-Patent Citations (1)
Title |
---|
浦东旭: ""基于病历文本语义分析的智能肝病辅助诊疗系统研究"", 《中国优秀硕士学位论文全文数据库 医药卫生科技辑》 * |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2021120934A1 (en) * | 2019-12-18 | 2021-06-24 | 浙江大学 | Convolutional neural network-based method for automatically grouping drgs |
CN111128375A (en) * | 2020-01-10 | 2020-05-08 | 电子科技大学 | Tibetan medicine diagnosis auxiliary device based on multi-label learning |
CN111128375B (en) * | 2020-01-10 | 2021-11-02 | 电子科技大学 | Tibetan medicine diagnosis auxiliary device based on multi-label learning |
WO2021114635A1 (en) * | 2020-05-13 | 2021-06-17 | 平安科技(深圳)有限公司 | Patient grouping model constructing method, patient grouping method, and related device |
Also Published As
Publication number | Publication date |
---|---|
CN110164519B (en) | 2021-08-06 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Jaiswal et al. | Classification of the COVID-19 infected patients using DenseNet201 based deep transfer learning | |
Lindsey et al. | Deep neural network improves fracture detection by clinicians | |
Reshi et al. | An efficient CNN model for COVID‐19 disease detection based on x‐ray image classification | |
CN109599185B (en) | Disease data processing method and device, electronic equipment and computer readable medium | |
Liu et al. | Medical-vlbert: Medical visual language bert for covid-19 ct report generation with alternate learning | |
Hu et al. | Deep supervised learning using self-adaptive auxiliary loss for COVID-19 diagnosis from imbalanced CT images | |
Ahn et al. | Unsupervised deep transfer feature learning for medical image classification | |
Zhao et al. | TCM herbal prescription recommendation model based on multi-graph convolutional network | |
CN110164519A (en) | A kind of classification method for being used to handle electronic health record blended data based on many intelligence networks | |
Band et al. | Application of explainable artificial intelligence in medical health: A systematic review of interpretability methods | |
Alahmari et al. | A comprehensive review of deep learning-based methods for COVID-19 detection using chest X-ray images | |
Murphy et al. | Visual transformers and convolutional neural networks for disease classification on radiographs: a comparison of performance, sample efficiency, and hidden stratification | |
CN114781382A (en) | Medical named entity recognition system and method based on RWLSTM model fusion | |
Mahajan | Applications of pattern recognition algorithm in health and medicine | |
Lu | Computer‐Aided Diagnosis Research of a Lung Tumor Based on a Deep Convolutional Neural Network and Global Features | |
Singh et al. | Early diagnosis of COVID-19 patients using deep learning-based deep forest model | |
Wu et al. | BLCov: A novel collaborative–competitive broad learning system for COVID-19 detection from radiology images | |
Mall et al. | Credence-Net: a semi-supervised deep learning approach for medical images | |
Nneji et al. | Fine-tuned siamese network with modified enhanced super-resolution gan plus based on low-quality chest x-ray images for covid-19 identification | |
Liu et al. | Multi-branch fusion auxiliary learning for the detection of pneumonia from chest X-ray images | |
Pellegrini et al. | Xplainer: From x-ray observations to explainable zero-shot diagnosis | |
CN104933446B (en) | A method of it is verified for computer-aided diagnosis breast sonography characteristic validity | |
CN112216379A (en) | Disease diagnosis system based on intelligent joint learning | |
Jagan Mohan et al. | Gil-cnn: A novel multipath features for covid-19 detection using ct-scan images | |
Saha et al. | LM-DNN: pre-trained DNN with LSTM and cross fold validation for detecting viral pneumonia from chest CT |
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 | ||
TR01 | Transfer of patent right |
Effective date of registration: 20220920 Address after: Room 801, 85 Kefeng Road, Huangpu District, Guangzhou City, Guangdong Province Patentee after: Yami Technology (Guangzhou) Co.,Ltd. Address before: 100124 No. 100 Chaoyang District Ping Tian Park, Beijing Patentee before: Beijing University of Technology |
|
TR01 | Transfer of patent right |