CN109697285A - Enhance the hierarchical B iLSTM Chinese electronic health record disease code mask method of semantic expressiveness - Google Patents

Enhance the hierarchical B iLSTM Chinese electronic health record disease code mask method of semantic expressiveness Download PDF

Info

Publication number
CN109697285A
CN109697285A CN201811523661.7A CN201811523661A CN109697285A CN 109697285 A CN109697285 A CN 109697285A CN 201811523661 A CN201811523661 A CN 201811523661A CN 109697285 A CN109697285 A CN 109697285A
Authority
CN
China
Prior art keywords
word
vector
feature
electronic health
health record
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
Application number
CN201811523661.7A
Other languages
Chinese (zh)
Other versions
CN109697285B (en
Inventor
王建新
余颖
李敏
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.)
Central South University
Original Assignee
Central South 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 Central South University filed Critical Central South University
Priority to CN201811523661.7A priority Critical patent/CN109697285B/en
Publication of CN109697285A publication Critical patent/CN109697285A/en
Application granted granted Critical
Publication of CN109697285B publication Critical patent/CN109697285B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F40/00—Handling natural language data
    • G06F40/20—Natural language analysis
    • G06F40/279—Recognition of textual entities
    • G06F40/289—Phrasal analysis, e.g. finite state techniques or chunking
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F40/00—Handling natural language data
    • G06F40/30—Semantic 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
    • G06N3/04—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
    • G06N3/08—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
    • 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

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Health & Medical Sciences (AREA)
  • Health & Medical Sciences (AREA)
  • Artificial Intelligence (AREA)
  • General Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • Computational Linguistics (AREA)
  • Computing Systems (AREA)
  • Software Systems (AREA)
  • Molecular Biology (AREA)
  • Data Mining & Analysis (AREA)
  • Biophysics (AREA)
  • Biomedical Technology (AREA)
  • Mathematical Physics (AREA)
  • Evolutionary Computation (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Audiology, Speech & Language Pathology (AREA)
  • Medical Informatics (AREA)
  • Primary Health Care (AREA)
  • Public Health (AREA)
  • Epidemiology (AREA)
  • Medical Treatment And Welfare Office Work (AREA)

Abstract

The invention discloses a kind of hierarchical B iLSTM Chinese electronic health record disease code mask methods for enhancing semantic expressiveness, after being pre-processed to the electronic health record text of input, in considering that Chinese word is constituted, individual Chinese character includes specific semantic, extracting character level feature vector using the BiLSTM for introducing concern mechanism indicates, obtains the semanteme and word-building characteristic of individual Chinese character；Character level term vector is indicated to splice with the other vector expression of word-level obtained using word2vec training, the word vectors for obtaining character feature enhancing indicate；Using the text sequence that Feature Words vector indicates as input, learn the contextual feature in entire electronic health record using BiLSTM again, and use concern mechanism, calculates the contribution degree of each Feature Words, the text vector for obtaining contextual feature weighting indicates, improves prediction effect.Method of the invention is suitable for the disease labeling task based on Chinese electronic health record text, and effectively increases classifying quality.

Description

Enhance the hierarchical B iLSTM Chinese electronic health record disease code mask method of semantic expressiveness
Technical field
The present invention relates to medical informatics field, especially a kind of hierarchical B iLSTM Chinese electronics disease for enhancing semantic expressiveness Go through disease code mask method.
Background technique
Electronic health care case history (Electronic Health Records, EHRs, abbreviation electronic health record) has become medicine and faces One of the significant data resource of bed research.Various information during it sees a doctor patient are stored with digitized data, Facilitate us using computer to analyze clinical data and handle.For a electronic health record, need to be described patient The unified Label specifications of disease condition, to be conducive to carry out patient information reasonable classification to help clinical decision.By generation The International Classification of Diseases of the publication of boundary's health organization and continuous updating encodes (International Classification of Diseases, ICD) be international disease code scheme, it often by the label as clinography, for identify symptom, Sign, disease, anomaly or operation etc..Currently, the ICD newly revised encodes the 10th edition hospital for being widely used in China In information system.
Marking ICD coding for electronic health record is an important and basic job using electronic health record.Electronic health record The missing of middle diagnosis name and ICD coding, is unfavorable for our analysis and research to clinical data.In general, the mark work of ICD coding Make the clinical diagnosis provided by the medical worker of each case of hospital room according to doctor to describe to carry out artificial cognition.H coding is not It requires nothing more than coder and grasps certain medical knowledge, coding rule and medical terminology, and is time-consuming and laborious.Therefore, meter is utilized Calculation machine can provide effective auxiliary to carry out autocoding for coding mark work, improve the annotating efficiency of ICD coding.
Most disease code automatic marking work is all based on clinical text data to carry out, such as the report of dept. of radiology at present Announcement, death certificate, discharge abstract etc..But most research work concentrates on English corpus, in Chinese clinical text On disease code prediction work it is less, and main method is that character string semanteme based on diagnosis name compares.It is semantic similar Property comparison quality requirement that diagnosis name is described it is higher, and autocoding can not be carried out in the case where diagnosis name missing Mark.There is presently no correlative study work to mark task for the disease code that neural network model is used for Chinese electronic health record.
There are two features for the processing of Chinese electronic health record text: first is that electronic health record text is longer, the context of long text Acquisition of information is more difficult；Second is that Chinese character is different from English, individual Chinese character also has semanteme, especially in medical terms, such as Orientation, physical feeling etc. are all a Chinese characters to describe, and therefore, the semantic expressiveness comprising character feature can preferably express word It is semantic.
Summary of the invention
The technical problem to be solved by the present invention is in view of the shortcomings of the prior art, provide a kind of layer for enhancing semantic expressiveness Secondary BiLSTM Chinese electronic health record disease code mask method completes automatic marking in a manner of end to end, improves prediction effect.
In order to solve the above technical problems, the technical scheme adopted by the invention is that:
A kind of hierarchical B iLSTM Chinese electronic health record disease code mask method enhancing semantic expressiveness, including following step It is rapid:
1) Chinese word segmentation tool is utilized, the customized clinical medicine of user is introduced and is segmented with dictionary, remove stop words, And Feature Words are filtered out according to word frequency；
2) Feature Words are carried out with character rank and the other vectorization of word-level respectively indicates, splices character level vector and word Grade vector, the character Enhanced feature vector for constructing word indicate；
3) contextual feature of entire text is obtained using spliced Feature Words, and uses concern mechanism, calculated each The contribution degree of Feature Words, the contextual feature weighing vector for obtaining entire text indicate.
In step 1), the Feature Words are chosen according to following rule:Wherein SfwIt indicates Feature set of words,Indicate word wiFrequency, NdIndicate electronic health record total sample number.
In step 2), indicated using the character level feature vector of the two-way LSTM training characteristics word of fusion concern mechanism, benefit The term vector representation method word2vec indicated with word-based distribution obtains the word-level vector representation of Feature Words.
The way of output of two-way shot and long term memory network training are as follows:WhereinIt indicates Forward direction LSTM is exported in the hidden layer of t-th of unit or t moment,It is exported to LSTM in the hidden layer of t-th of unit after being then.
The calculation of concern mechanism are as follows:
uij=tanh (Wchij+bc)；
hijFor i-th of word j-th of character BiLSTM training after hidden layer output, WcFor weight matrix, bcFor biasing Vector, ucFor the contextual feature vector of random initializtion character level, αijFor j-th be calculated using softmax function Character for i-th of word weight size,It is indicated for the context weighted feature vector of i-th of word.
In step 3), the method for calculating the contextual feature weighing vector of entire text includes: by spliced Feature Words The two-way shot and long term memory network of the text input second layer that vector indicates, study obtains the contextual feature of entire text, and adopts With concern mechanism, the weight of each Feature Words is calculated, obtains the Text eigenvector of contextual information weighting.
The calculation of concern mechanism are as follows:
ui=tanh (Whi+bw)；
V=∑iαihi；
hiIt is that the character of i-th of word of text sequence reinforces the output for the hidden layer that feature vector obtains after BiLSTM training, W For weight matrix, bwIt is corresponding to introduce simultaneously one other text of word-level of random initializtion in application concern mechanism for bias vector Shelves contextual feature vector uwTo complete the calculating of weight, αiFor the corresponding weight of each word, v is that the context of entire text adds Weighing feature vector indicates, which is inputted full articulamentum, the appearance that each disease code is calculated by sigmoid function is general Rate.
Compared with prior art, the advantageous effect of present invention is that: the present invention is directed to Chinese own characteristic, will be single The feature vector that the semantic feature of Chinese character incorporates word indicates, and combines concern mechanism, to contributive spy real in list entries Sign word is weighted, and improves the prediction effect of disease code；This method is suitable for the clinical text data of Chinese, utilizes nerve Network model automatically extracts text feature, and automatic marking is completed in a manner of end to end.
Detailed description of the invention
Flow chart Fig. 1 of the invention；
The hierarchical B iLSTM feature learning model of Fig. 2 fusion concern mechanism；
The calculating of Fig. 3 concern mechanism；(a) by hijBecome uij；(b) each u is calculated using contextual feature vectorijPower Weight；(c)hijWeighted sum be applied mechanism of paying close attention to feature vector indicate；
Fig. 4 is that the present invention implements experimental result picture.
Specific embodiment
One, the pretreatment of clinical text data
Using Chinese word segmentation tool " stammerer " and the customized medicine dictionary of user, the discharge abstract text of input is carried out After participle, stop words is removed, the word frequency of effective word is counted, selects Feature Words after sorting from large to small based on word frequency, by following rule Then choose:Wherein SfwIndicate feature set of words,Indicate word wiFrequency, NdIndicate electronics disease Go through sum.
Two, the term vector of Feature Words indicates
1) term vector based on character indicates
Firstly, initializing vector for each character indicates, the then BiLSTM of input fusion concern mechanism, trained Character level term vector to each Feature Words indicates, each neural unit state value c in BiLSTMtWith output valve htSpecific meter Calculation process is (t=1,2 ..., n, t indicate the neural unit of t-th of neural unit or t moment in network):
it=sigmoid (Wi[xt；ht-1]+bi) (1)
ft=sigmoid (Wf[xt；ht-1]+bf) (2)
gt=tanh (Wg[xt；ht-1]+bg) (3)
ot=sigmoid (Wo[xt；ht-1]+bo) (4)
ct=ft*ct-1+it*gt (5)
ht=ot*tanh(ct) (6)
Each neural unit includes an input gate i, an out gate o, a forgetting door f, a storage unit g, and one The unit c of an a preservation state and hidden state h, they are vector, Wi,Wf,Wg,WoFor weight matrix, bi,bf,bg,bo For bias vector, "；" indicate connection operation, " * " indicates element dot product, and sigmoid function is calculated asTanh function is calculated asThe way of output of BiLSTM For
2) application of attention mechanism
Concern mechanism calculation method are as follows:
uij=tanh (Wchij+bc) (7)
hijFor i-th of word j-th of character BiLSTM training after hidden layer output, WcFor weight matrix, bcFor biasing Vector, ucFor the contextual feature vector of random initializtion character level, αijThe jth being as calculated using softmax function A character for i-th of word weight size,The context weighted feature vector of as i-th word indicates.
3) the character level term vector that training obtains is spliced with the term vector generated using word2vec, obtains character The word feature vector that grade contextual feature is reinforced.
Three, contextual feature is extracted
The BiLSTM for the characteristic vector sequence input second layer fusion concern mechanism that character is reinforced, extracts text context Information characteristics, the calculating of the calculating of BiLSTM neural unit and contextual feature weighting, phase when being indicated with character level term vector Together, specific calculation formula is as follows:
ui=tanh (Whi+bw) (10)
V=∑iαihi (12)
hiIt is that the character of i-th of word of text sequence reinforces the output for the hidden layer that feature vector obtains after BiLSTM training, W For weight matrix, bwIt is corresponding to introduce simultaneously one other text of word-level of random initializtion in application concern mechanism for bias vector Shelves contextual feature vector uwTo complete the calculating of weight, αiFor the corresponding weight of each word, v is that the context of entire text adds Weighing feature vector indicates, which is inputted full articulamentum, the appearance that each disease code is calculated by sigmoid function is general Rate.
Four, experimental verification
1) experimentation
In order to verify the validity of this method, we test on true Chinese electronic health record clinical data Card.The data set includes 7732 discharge records, is related to 1177 ICD-10 disease code labels altogether, ICD-10 coding is by word Female and number composition point minute six codings, with beginning of letter, front three is encoded to level encoder, indicates disease classification.Discharge The average length of brief summary is 610 words, average corresponding 3.6 disease codes of each discharge abstract.
Experiment is completed on a server, which includes 256GB memory and NVIDIA GeForce Titan X Pascal CUDA GPU processor.Data set is divided into training set and test set according to the ratio of 9:1 by us, and is passed through ten times Upset data at random to be verified.Evaluation index has selected micro- average accuracy (P), recall rate (R) and the two synthesis Index F1 value, and the Hamming penalty values for reporting situation by mistake are evaluated from the angle of sample.F1 value is higher, Hamming penalty values more It is low to illustrate that model performance is better.
2) experimental result
Because correlative study work has had been pointed out deep learning method better than traditional machine learning method, we mainly and its He has carried out comparative experiments by common neural network model, and the results are shown in Table 1, and MA-BiLSTM indicates our model, D2V+ CNN is the method in correlative study work, and this method is obtained on disclosed English data set MIMIC III and preferably imitated at present Fruit.The experimental results showed that MA-BiLSTM is superior to other neural network models in every evaluation index, illustrate to combine concern machine The BiLSTM of system can effectively capture the contextual information feature of long text, and improve prediction effect.
1 contrast and experiment of table
Model Micro_P (CI:95%) Micro_R (CI:95%) Micro_F1 (CI:95%) HLoss (CI:95%)
CBOW 0.614(±6.43e-03) 0.522(±5.30e-03) 0.564(±4.52e-03) 0.00248(±3.14e-05)
CNN 0.647(±6.67e-03) 0.509(±6.51e-03) 0.569(±4.71e-03) 0.00237(±3.52e-05)
D2V+CNN 0.661(±9.57e-03) 0.514(±8.74e-03) 0.579(±7.14e-03) 0.00231(±3.70e-05)
MA-BiLSTM 0.704(±1.13e-02) 0.586(±5.84e-03) 0.639(±4.45e-03) 0.00204(±3.47e-05)
For the effect of the performance of analysis model modules, we devise ablation experiment and analyze, as a result such as 2 institute of table Show.From the experimental results, only term vector or character vector indicate that the feature of word in text, prediction result all have occurred down Drop, therefore, the term vector expression that character vector is reinforced bring better Text Representation really.Concern mechanism is in a model Important function is played, concern mechanism is eliminated, the performance decline of model is obvious.
It is predicted in ICD-10 full coding and level encoder, 7732 samples, corresponding level encoder is 488 It is a.Experimental result is as shown in Figure 4.Prediction result on level encoder has reached 80.5% in accuracy, can preferably assist The disease code of Record room medical worker marks work.
2 model of table melts experimental result

Claims (7)

1. a kind of hierarchical B iLSTM Chinese electronic health record disease code mask method for enhancing semantic expressiveness, which is characterized in that packet Include following steps:
1) Chinese word segmentation tool is utilized, the customized clinical medicine of user is introduced and is segmented with dictionary, removes stop words, and root Feature Words are filtered out according to word frequency；
2) Feature Words are carried out with character rank and the other vectorization of word-level respectively indicates, splice character level vector and word-level to Amount, the character Enhanced feature vector for constructing word indicate；
3) sequence is indicated using the term vector that spliced Feature Words obtain entire text, and use concern mechanism, calculate each The contribution degree of Feature Words, the contextual feature weighing vector for obtaining entire text indicate.
2. the hierarchical B iLSTM Chinese electronic health record disease code mark side of enhancing semantic expressiveness according to claim 1 Method, which is characterized in that in step 1), the Feature Words are chosen according to following rule:Wherein SfwIndicate feature set of words,Indicate word wiFrequency, NdIndicate electronic health record total sample number.
3. the hierarchical B iLSTM Chinese electronic health record disease code mark side of enhancing semantic expressiveness according to claim 1 Method, which is characterized in that in step 2), utilize the character level feature vector table of the BiLSTM training characteristics word of fusion concern mechanism Show, indicates shape using the word-level vector that the word-based distributed term vector representation method word2vec indicated obtains Feature Words Formula.
4. the hierarchical B iLSTM Chinese electronic health record disease code mark side of enhancing semantic expressiveness according to claim 3 Method, which is characterized in that the way of output of BiLSTM are as follows:WhereinExist before indicating to LSTM The output of the hidden layer of t-th of unit or t moment,It is exported to LSTM in the hidden layer of t-th of unit after being then.
5. the hierarchical B iLSTM Chinese electronic health record disease code mark side of enhancing semantic expressiveness according to claim 3 Method, which is characterized in that pay close attention to the calculation of mechanism are as follows:
uij=tanh (Wchij+bc)；
hijFor i-th of word j-th of character BiLSTM training after hidden layer output, WcFor weight matrix, bcFor bias vector, ucFor the contextual feature vector of random initializtion character level, αijFor j-th of character pair being calculated using softmax function In the weight size of i-th of word,It is indicated for the context weighted feature vector of i-th of word.
6. the hierarchical B iLSTM Chinese electronic health record disease code mark side of enhancing semantic expressiveness according to claim 1 Method, which is characterized in that in step 3), the method for calculating the contextual feature weighing vector of entire text includes: will be spliced The two-way shot and long term memory network of the text input second layer that Feature Words vector indicates, the context that study obtains entire text are special Sign, and concern mechanism is used, the weight of each Feature Words is calculated, the Text eigenvector of contextual information weighting is obtained.
7. the hierarchical B iLSTM Chinese electronic health record disease code mark side of enhancing semantic expressiveness according to claim 6 Method, which is characterized in that pay close attention to the calculation of mechanism are as follows:
ui=tanh (Whi+bw)；
V=∑iαihi；
hiIt is that the character of i-th of word of text sequence reinforces the output for the hidden layer that feature vector obtains after BiLSTM training, W is power Value matrix, bwIt is corresponding to introduce and on one other document of word-level of random initializtion in application concern mechanism for bias vector Following traits vector uwTo complete the calculating of weight, αiFor the corresponding weight of each word, v is that the context of entire text weights spy Levying vector indicates, which is inputted full articulamentum, the probability of occurrence of each disease code is calculated by sigmoid function.
CN201811523661.7A 2018-12-13 2018-12-13 Hierarchical BilSt Chinese electronic medical record disease coding and labeling method for enhancing semantic representation Active CN109697285B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201811523661.7A CN109697285B (en) 2018-12-13 2018-12-13 Hierarchical BilSt Chinese electronic medical record disease coding and labeling method for enhancing semantic representation

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201811523661.7A CN109697285B (en) 2018-12-13 2018-12-13 Hierarchical BilSt Chinese electronic medical record disease coding and labeling method for enhancing semantic representation

Publications (2)

Publication Number Publication Date
CN109697285A true CN109697285A (en) 2019-04-30
CN109697285B CN109697285B (en) 2022-06-21

Family

ID=66231615

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201811523661.7A Active CN109697285B (en) 2018-12-13 2018-12-13 Hierarchical BilSt Chinese electronic medical record disease coding and labeling method for enhancing semantic representation

Country Status (1)

Country Link
CN (1) CN109697285B (en)

Cited By (23)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110427610A (en) * 2019-06-25 2019-11-08 平安科技（深圳）有限公司 Text analyzing method, apparatus, computer installation and computer storage medium
CN110491499A (en) * 2019-07-10 2019-11-22 厦门大学 Clinical aid decision-making method and system towards mark electronic health record
CN110491465A (en) * 2019-08-20 2019-11-22 山东众阳健康科技集团有限公司 Classification of diseases coding method, system, equipment and medium based on deep learning
CN110633470A (en) * 2019-09-17 2019-12-31 北京小米智能科技有限公司 Named entity recognition method, device and storage medium
CN110781407A (en) * 2019-10-21 2020-02-11 腾讯科技（深圳）有限公司 User label generation method and device and computer readable storage medium
CN110837494A (en) * 2019-10-12 2020-02-25 云知声智能科技股份有限公司 Method and device for identifying unspecified diagnosis coding errors of medical record home page
CN110866401A (en) * 2019-11-18 2020-03-06 山东健康医疗大数据有限公司 Chinese electronic medical record named entity identification method and system based on attention mechanism
CN110867231A (en) * 2019-11-18 2020-03-06 中山大学 Disease prediction method, device, computer equipment and medium based on text classification
CN110895580A (en) * 2019-12-12 2020-03-20 山东众阳健康科技集团有限公司 ICD operation and operation code automatic matching method based on deep learning
CN111429204A (en) * 2020-03-10 2020-07-17 携程计算机技术（上海）有限公司 Hotel recommendation method, system, electronic equipment and storage medium
CN112052646A (en) * 2020-08-27 2020-12-08 安徽聚戎科技信息咨询有限公司 Text data labeling method
CN112185564A (en) * 2020-10-20 2021-01-05 福州数据技术研究院有限公司 Ophthalmic disease prediction method based on structured electronic medical record and storage device
CN112183104A (en) * 2020-08-26 2021-01-05 望海康信(北京)科技股份公司 Code recommendation method, system and corresponding equipment and storage medium
CN112259260A (en) * 2020-11-18 2021-01-22 中国科学院自动化研究所 Intelligent medical question and answer method, system and device based on intelligent wearable equipment
CN112380863A (en) * 2020-10-29 2021-02-19 国网天津市电力公司 Sequence labeling method based on multi-head self-attention mechanism
WO2021057133A1 (en) * 2019-09-24 2021-04-01 北京国双科技有限公司 Method for training document classification model, and related apparatus
CN112632911A (en) * 2021-01-04 2021-04-09 福州大学 Chinese character coding method based on character embedding
CN113012774A (en) * 2019-12-18 2021-06-22 医渡云（北京）技术有限公司 Automatic medical record encoding method and device, electronic equipment and storage medium
CN113593709A (en) * 2021-07-30 2021-11-02 江先汉 Disease coding method, system, readable storage medium and device
CN116884630A (en) * 2023-09-06 2023-10-13 深圳达实旗云健康科技有限公司 Method for improving disease automatic coding efficiency
CN116955628A (en) * 2023-08-08 2023-10-27 武汉市万睿数字运营有限公司 Complaint event classification method, complaint event classification device, computer equipment and storage medium
CN117438024A (en) * 2023-12-15 2024-01-23 吉林大学 Intelligent acquisition and analysis system and method for acute diagnosis patient sign data
CN116955628B (en) * 2023-08-08 2024-05-03 武汉市万睿数字运营有限公司 Complaint event classification method, complaint event classification device, computer equipment and storage medium

Citations (13)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20080288292A1 (en) * 2007-05-15 2008-11-20 Siemens Medical Solutions Usa, Inc. System and Method for Large Scale Code Classification for Medical Patient Records
WO2015084615A1 (en) * 2013-12-03 2015-06-11 3M Innovative Properties Company Constraint-based medical coding
US20160132648A1 (en) * 2014-11-06 2016-05-12 ezDI, LLC Data Processing System and Method for Computer-Assisted Coding of Natural Language Medical Text
CN106484674A (en) * 2016-09-20 2017-03-08 北京工业大学 A kind of Chinese electronic health record concept extraction method based on deep learning
CN106844308A (en) * 2017-01-20 2017-06-13 天津艾登科技有限公司 A kind of use semantics recognition carries out the method for automating disease code conversion
EP3273373A1 (en) * 2016-07-18 2018-01-24 Fresenius Medical Care Deutschland GmbH Drug dosing recommendation
CN107644014A (en) * 2017-09-25 2018-01-30 南京安链数据科技有限公司 A kind of name entity recognition method based on two-way LSTM and CRF
CN107731269A (en) * 2017-10-25 2018-02-23 山东众阳软件有限公司 Disease code method and system based on raw diagnostic data and patient file data
CN107977361A (en) * 2017-12-06 2018-05-01 哈尔滨工业大学深圳研究生院 The Chinese clinical treatment entity recognition method represented based on deep semantic information
CN108460013A (en) * 2018-01-30 2018-08-28 大连理工大学 A kind of sequence labelling model based on fine granularity vocabulary representation model
CN108536754A (en) * 2018-03-14 2018-09-14 四川大学 Electronic health record entity relation extraction method based on BLSTM and attention mechanism
CN108628824A (en) * 2018-04-08 2018-10-09 上海熙业信息科技有限公司 A kind of entity recognition method based on Chinese electronic health record
CN108628823A (en) * 2018-03-14 2018-10-09 中山大学 In conjunction with the name entity recognition method of attention mechanism and multitask coordinated training

Patent Citations (13)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20080288292A1 (en) * 2007-05-15 2008-11-20 Siemens Medical Solutions Usa, Inc. System and Method for Large Scale Code Classification for Medical Patient Records
WO2015084615A1 (en) * 2013-12-03 2015-06-11 3M Innovative Properties Company Constraint-based medical coding
US20160132648A1 (en) * 2014-11-06 2016-05-12 ezDI, LLC Data Processing System and Method for Computer-Assisted Coding of Natural Language Medical Text
EP3273373A1 (en) * 2016-07-18 2018-01-24 Fresenius Medical Care Deutschland GmbH Drug dosing recommendation
CN106484674A (en) * 2016-09-20 2017-03-08 北京工业大学 A kind of Chinese electronic health record concept extraction method based on deep learning
CN106844308A (en) * 2017-01-20 2017-06-13 天津艾登科技有限公司 A kind of use semantics recognition carries out the method for automating disease code conversion
CN107644014A (en) * 2017-09-25 2018-01-30 南京安链数据科技有限公司 A kind of name entity recognition method based on two-way LSTM and CRF
CN107731269A (en) * 2017-10-25 2018-02-23 山东众阳软件有限公司 Disease code method and system based on raw diagnostic data and patient file data
CN107977361A (en) * 2017-12-06 2018-05-01 哈尔滨工业大学深圳研究生院 The Chinese clinical treatment entity recognition method represented based on deep semantic information
CN108460013A (en) * 2018-01-30 2018-08-28 大连理工大学 A kind of sequence labelling model based on fine granularity vocabulary representation model
CN108536754A (en) * 2018-03-14 2018-09-14 四川大学 Electronic health record entity relation extraction method based on BLSTM and attention mechanism
CN108628823A (en) * 2018-03-14 2018-10-09 中山大学 In conjunction with the name entity recognition method of attention mechanism and multitask coordinated training
CN108628824A (en) * 2018-04-08 2018-10-09 上海熙业信息科技有限公司 A kind of entity recognition method based on Chinese electronic health record

Non-Patent Citations (5)

* Cited by examiner, † Cited by third party
Title
AITZIBER ATUTXA: "Machine Learning Approaches on Diagnostic Term Encoding With the ICD for Clinical Documentation", 《IEEE JOURNAL OF BIOMEDICAL AND HEALTH INFORMATICS》 *
HAORAN SHI ET AL: "Towards Automated ICD Coding Using Deep Learning", 《HTTPS://ARXIV.ORG/ABS/1711.04075V3》 *
MIN LI ET AL: "Automated ICD-9 Coding via A Deep Learning Approach", 《IEEE/ACM TRANSACTIONS ON COMPUTATIONAL BIOLOGY AND BIOINFORMATICS》 *
TAL BAUMEL ET AL: "Multi-Label Classification of Patient Notes: Case Study on ICD Code Assignment", 《HTTPS://ARXIV.ORG/ABS/1709.09587》 *
钟楠祎: "基于深度学习的数据特征的提取与预测研究", 《中国优秀硕士学位论文全文数据库信息科技辑》 *

Cited By (31)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110427610A (en) * 2019-06-25 2019-11-08 平安科技（深圳）有限公司 Text analyzing method, apparatus, computer installation and computer storage medium
CN110491499A (en) * 2019-07-10 2019-11-22 厦门大学 Clinical aid decision-making method and system towards mark electronic health record
CN110491465A (en) * 2019-08-20 2019-11-22 山东众阳健康科技集团有限公司 Classification of diseases coding method, system, equipment and medium based on deep learning
WO2021032219A3 (en) * 2019-08-20 2021-04-15 山东众阳健康科技集团有限公司 Method and system for disease classification coding based on deep learning, and device and medium
CN110633470A (en) * 2019-09-17 2019-12-31 北京小米智能科技有限公司 Named entity recognition method, device and storage medium
WO2021057133A1 (en) * 2019-09-24 2021-04-01 北京国双科技有限公司 Method for training document classification model, and related apparatus
CN110837494A (en) * 2019-10-12 2020-02-25 云知声智能科技股份有限公司 Method and device for identifying unspecified diagnosis coding errors of medical record home page
CN110837494B (en) * 2019-10-12 2022-03-25 云知声智能科技股份有限公司 Method and device for identifying unspecified diagnosis coding errors of medical record home page
CN110781407A (en) * 2019-10-21 2020-02-11 腾讯科技（深圳）有限公司 User label generation method and device and computer readable storage medium
CN110866401A (en) * 2019-11-18 2020-03-06 山东健康医疗大数据有限公司 Chinese electronic medical record named entity identification method and system based on attention mechanism
CN110867231A (en) * 2019-11-18 2020-03-06 中山大学 Disease prediction method, device, computer equipment and medium based on text classification
CN110895580A (en) * 2019-12-12 2020-03-20 山东众阳健康科技集团有限公司 ICD operation and operation code automatic matching method based on deep learning
CN113012774A (en) * 2019-12-18 2021-06-22 医渡云（北京）技术有限公司 Automatic medical record encoding method and device, electronic equipment and storage medium
CN111429204A (en) * 2020-03-10 2020-07-17 携程计算机技术（上海）有限公司 Hotel recommendation method, system, electronic equipment and storage medium
CN112183104A (en) * 2020-08-26 2021-01-05 望海康信(北京)科技股份公司 Code recommendation method, system and corresponding equipment and storage medium
CN112052646B (en) * 2020-08-27 2024-03-29 安徽聚戎科技信息咨询有限公司 Text data labeling method
CN112052646A (en) * 2020-08-27 2020-12-08 安徽聚戎科技信息咨询有限公司 Text data labeling method
CN112185564A (en) * 2020-10-20 2021-01-05 福州数据技术研究院有限公司 Ophthalmic disease prediction method based on structured electronic medical record and storage device
CN112185564B (en) * 2020-10-20 2022-09-06 福州数据技术研究院有限公司 Ophthalmic disease prediction method based on structured electronic medical record and storage device
CN112380863A (en) * 2020-10-29 2021-02-19 国网天津市电力公司 Sequence labeling method based on multi-head self-attention mechanism
CN112259260B (en) * 2020-11-18 2023-11-17 中国科学院自动化研究所 Intelligent medical question-answering method, system and device based on intelligent wearable equipment
CN112259260A (en) * 2020-11-18 2021-01-22 中国科学院自动化研究所 Intelligent medical question and answer method, system and device based on intelligent wearable equipment
CN112632911B (en) * 2021-01-04 2022-05-13 福州大学 Chinese character coding method based on character embedding
CN112632911A (en) * 2021-01-04 2021-04-09 福州大学 Chinese character coding method based on character embedding
CN113593709A (en) * 2021-07-30 2021-11-02 江先汉 Disease coding method, system, readable storage medium and device
CN113593709B (en) * 2021-07-30 2022-09-30 江先汉 Disease coding method, system, readable storage medium and device
CN116955628A (en) * 2023-08-08 2023-10-27 武汉市万睿数字运营有限公司 Complaint event classification method, complaint event classification device, computer equipment and storage medium
CN116955628B (en) * 2023-08-08 2024-05-03 武汉市万睿数字运营有限公司 Complaint event classification method, complaint event classification device, computer equipment and storage medium
CN116884630A (en) * 2023-09-06 2023-10-13 深圳达实旗云健康科技有限公司 Method for improving disease automatic coding efficiency
CN117438024A (en) * 2023-12-15 2024-01-23 吉林大学 Intelligent acquisition and analysis system and method for acute diagnosis patient sign data
CN117438024B (en) * 2023-12-15 2024-03-08 吉林大学 Intelligent acquisition and analysis system and method for acute diagnosis patient sign data

Also Published As

Publication number Publication date
CN109697285B (en) 2022-06-21

Similar Documents

Publication Publication Date Title
CN109697285A (en) Enhance the hierarchical B iLSTM Chinese electronic health record disease code mask method of semantic expressiveness
CN111192680B (en) Intelligent auxiliary diagnosis method based on deep learning and collective classification
CN111966917B (en) Event detection and summarization method based on pre-training language model
CN106844308B (en) Method for automatic disease code conversion using semantic recognition
CN106980609A (en) A kind of name entity recognition method of the condition random field of word-based vector representation
CN108182295A (en) A kind of Company Knowledge collection of illustrative plates attribute extraction method and system
CN109635280A (en) A kind of event extraction method based on mark
CN109753660B (en) LSTM-based winning bid web page named entity extraction method
CN108399163A (en) Bluebeard compound polymerize the text similarity measure with word combination semantic feature
CN109471895A (en) The extraction of electronic health record phenotype, phenotype name authority method and system
CN109508459B (en) Method for extracting theme and key information from news
CN109522557A (en) Training method, device and the readable storage medium storing program for executing of text Relation extraction model
CN110083710A (en) It is a kind of that generation method is defined based on Recognition with Recurrent Neural Network and the word of latent variable structure
CN110162779A (en) Appraisal procedure, device and the equipment of quality of case history
CN109003677B (en) Structured analysis processing method for medical record data
CN110321563A (en) Text emotion analysis method based on mixing monitor model
CN110134946A (en) A kind of machine reading understanding method for complex data
CN108563725A (en) A kind of Chinese symptom and sign composition recognition methods
CN110298036A (en) A kind of online medical text symptom identification method based on part of speech increment iterative
CN108345583A (en) Event recognition and sorting technique based on multi-lingual attention mechanism and device
CN111881256B (en) Text entity relation extraction method and device and computer readable storage medium equipment
CN111859938B (en) Electronic medical record entity relation extraction method based on position vector noise reduction and rich semantics
CN110046356A (en) Label is embedded in the application study in the classification of microblogging text mood multi-tag
CN109993227A (en) Method, system, device and the medium of automatic addition International Classification of Diseases coding
CN108920446A (en) A kind of processing method of Engineering document

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