CN114822734A - Traditional Chinese medical record analysis method based on cyclic convolution neural network - Google Patents

Traditional Chinese medical record analysis method based on cyclic convolution neural network Download PDF

Info

Publication number
CN114822734A
CN114822734A CN202111012711.7A CN202111012711A CN114822734A CN 114822734 A CN114822734 A CN 114822734A CN 202111012711 A CN202111012711 A CN 202111012711A CN 114822734 A CN114822734 A CN 114822734A
Authority
CN
China
Prior art keywords
word
medical record
traditional chinese
neural network
text
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
CN202111012711.7A
Other languages
Chinese (zh)
Inventor
张明川
张玉
吴庆涛
王琳
朱军龙
冀治航
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.)
Henan University of Science and Technology
Original Assignee
Henan University of Science and Technology
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 Henan University of Science and Technology filed Critical Henan University of Science and Technology
Priority to CN202111012711.7A priority Critical patent/CN114822734A/en
Publication of CN114822734A publication Critical patent/CN114822734A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
    • G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
    • 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
    • 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
    • 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
    • G06N3/044—Recurrent networks, e.g. Hopfield networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—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
    • 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

Landscapes

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

Abstract

A traditional Chinese medical record analysis method based on a cyclic convolution neural network relates to the technical field of traditional Chinese medical record analysis, text classification is carried out by using CNN (convolutional neural network) so as to have the advantage of unbiased classification, characteristics are firstly extracted from each short text, and then the characteristics are obtained through a pooling layer, so that the semantic distinction of sentences or texts is realized. The invention has the beneficial effects that: the RCNN in the deep learning is applied to classification of the traditional Chinese medical record, and a corresponding classification model is trained and then tested, so that accuracy of class prediction is observed, and preliminary screening and decision support are provided for analysis of the traditional Chinese medical record.

Description

Traditional Chinese medical record analysis method based on cyclic convolution neural network
Technical Field
The invention belongs to the technical field of traditional Chinese medical record analysis, and particularly relates to a traditional Chinese medical record analysis method based on a cyclic convolution neural network.
Background
Currently, with the continuous improvement of medical level and the rapid development of information technology, the modernization of traditional Chinese medicine is an important way for the development of traditional Chinese medicine, and more novel technologies provide a plurality of methods and means for the modernization of traditional Chinese medicine.
According to market research, the traditional Chinese medical record analysis system in the market can only give a report after classifying the medical records of patients. The test sheet only writes information of disease names and some food therapy aspects, and does not give corresponding prescriptions. In addition, because the currently mainstream text analysis neural network uses a Long Short Term Memory artificial neural network (LSTM), the problem that a large amount of data cannot be processed in parallel exists in the traditional Chinese medical record analysis system in the market, and the time is greatly consumed.
Disclosure of Invention
The invention aims to solve the technical problem of providing a traditional Chinese medical record analysis method based on a cyclic convolution neural network, and solving the problems of the cyclic neural network in the prior art.
The technical scheme adopted by the invention for solving the technical problems is as follows: the traditional Chinese medical record analysis method based on the cyclic convolution neural network comprises the following steps:
collecting the existing medical records, carrying out noise processing on the collected medical records, eliminating noise data and reserving effective data;
step two, carrying out data preprocessing on the medical records subjected to noise processing;
step three, carrying out data training on the data preprocessed in the step two and detecting a training result to finally obtain a traditional Chinese medicine medical record analysis model;
and step four, inputting symptoms in the traditional Chinese medical record analysis model to carry out traditional Chinese medical record analysis to obtain a traditional Chinese medical record analysis result.
The specific process of the step of preprocessing the data of the medical record comprises the following steps:
(1) acquiring existing CSV data of a medical record text, labeling, performing jieba word segmentation, and removing stop words in a Chinese stop word bank;
(2) setting a frequency threshold, and if the frequency of a certain word appearing in the case text is less than the set frequency threshold, deleting the word;
(3) setting a characteristic selection threshold, selecting words with mutual information values higher than the characteristic selection threshold in the medical record text X, and regarding the words as characteristic values of the medical record text X to obtain characteristic vectors of the medical record text X;
(4) and (3) preprocessing the medical case text by using a word vector training model, wherein the preprocessing comprises word stopping, word segmentation and word vectorization, and the word vector training model is prepared for neural network training.
The specific method for training the data to obtain the medical record analysis model in the third step is as follows:
(1) definition of x l (t i ) As a sentence t in the case text i Left sentence of (1), x r (t i ) As a sentence t i Right sentence of (1), x l (t i ) And x r (t i ) Are all dense vectors with real valued elements, t i Left sentence x of l (t i ) Calculated from equation (1), the right sentence x r (t i ) Calculated from equation (2), where e (t) i-1 ) Is t i-1 Word embedding of, t i-1 Is a dense vector with | e | real-valued elements, the left-hand sentence of the first word in the case text uses the same sharing parameter x l (t 1 )，W (l) Is a matrix, W, that converts a hidden layer to the next hidden layer (yl) Is a matrix for combining the semantics of the current word with the left sentence of the next word, s is a non-linear activation function, and the right sentence of the last word in the case text shares a parameter x r (t n )；
x l (t i )＝s(W (l) x l (t i-1 )+W (yl) e(t i-1 )) (1)
x r (t i )＝s(W (r) x r (t i+1 )+W (yr) e(t i+1 )) (2)
Defining the case word m by equation (3) i It is the left context vector x l (t i ) Word embedding e (t) i ) And the right context vector x r (t i ) The connection of (2);
m i ＝[x l (t i )；e(t i )；x r (t i )] (3)
after obtaining the word m i After representation of (c), the result is sent to the next layer using a linear transformation and a relu activation function
Figure BDA0003239445580000021
Figure BDA0003239445580000022
Figure BDA0003239445580000023
Is a potential semantic vector, and determines the most useful part for representing the file by analyzing each semantic factor in the file;
representing the case by using convolutional neural network technology, and applying a maximum pool layer y after calculating to obtain representations of all words (3) ；
Figure BDA0003239445580000031
The max function is an element-level function, y (2) The jth element of (1)
Figure BDA0003239445580000032
Is that
Figure BDA0003239445580000033
(i∈[1,n]) The largest of the n elements;
the last layer of the model is the output layer y (4) Which is defined as
y (4) ＝W (4) y (3) +b (4) (6)
Finally, using the softmax function at y (4) Converting the output number into probabilities, wherein the sum of all the probabilities is 1, and taking an element corresponding to the maximum probability in the output result as a result;
Figure BDA0003239445580000034
in the model, a cross entropy loss function is used as the basis for the tuning parameter, as in equation (8), where p is the actual value and p is i Is the predicted value output of the network, resulting in a loss value L (p, p) i ) And parameters are adjusted according to the loss value, so that the model has better accuracy:
Figure BDA0003239445580000035
the invention has the beneficial effects that: by applying the recurrent convolutional neural network, the problem of the recurrent neural network on bias is effectively solved, the processing time is greatly shortened, and corresponding prescription recommendation can be given after the classification of the medical record text is finished, so that preliminary screening and decision support are provided for the traditional Chinese medicine diagnosis and analysis.
Drawings
FIG. 1 is a schematic diagram of a data training process of a medical record analysis model according to the present invention;
FIG. 2 is a schematic view illustrating a process of analyzing a medical record according to the medical record analysis model of the present invention;
FIG. 3 is a comparison chart of the test accuracy of the experimental training performed by the three methods in the embodiment of the present invention;
FIG. 4 is a graph showing the comparison of the test loss of the experimental training performed by the three methods in the example of the present invention;
FIG. 5 is a comparison graph of training accuracy for experimental training performed by three methods according to an embodiment of the present invention;
FIG. 6 is a comparison graph of training times for experimental training performed by three methods according to an embodiment of the present invention;
FIG. 7 is a graph showing a comparison of training losses for experimental training performed by three methods in an embodiment of the present invention.
Detailed Description
The invention provides a traditional Chinese medical record analysis method based on a cyclic convolution neural network, RCNN utilizes CNN to classify texts, has the advantage of unbiased classification, firstly extracts features on each short text, and then obtains the features through a pooling layer, thereby realizing semantic distinction of sentences or texts. The text classification algorithm based on the neural network in the big data analysis technology is improved. The RCNN in the deep learning is applied to classification of the traditional Chinese medical record, and a corresponding classification model is trained and then tested, so that accuracy of class prediction is observed, and preliminary screening and decision support are provided for analysis of the traditional Chinese medical record. The method specifically comprises the following steps:
1. data preprocessing stage
The process of data pre-processing is as follows:
(1) acquiring existing medical record text csv data, labeling, performing jieba word segmentation, and removing stop words in a Chinese stop word bank;
(2) setting a frequency threshold, and if the frequency of a certain word appearing in the case text is less than the set frequency threshold, deleting the word;
(3) setting a characteristic selection threshold, selecting words with mutual information values higher than the characteristic selection threshold in the medical case text X, and regarding the words as characteristic values of the text X to obtain characteristic vectors of the text X;
(4) and (3) preprocessing the medical record text by using a word vector training model, and preparing the medical record text for neural network training.
2. Data training phase
And putting the preprocessed data into a circular convolution neural network training model to train the data. The process of data training is as follows:
(1) definition of x l (t i ) As a sentence t in the case text i Left sentence of (1), x r (t i ) As a sentence t i Right sentence of (1), x l (t i ) Andx r (t i ) Are all dense vectors with real valued elements, t i Left sentence x of l (t i ) Calculated from equation (1), the right sentence x r (t i ) Calculated from equation (2), where e (t) i-1 ) Is t i-1 Word embedding of, t i-1 Is a dense vector with | e | real-valued elements, the left-hand sentence of the first word in the case text uses the same sharing parameter x l (t 1 )，W (l) Is a matrix, W, that converts a hidden layer to the next hidden layer (yl) Is a matrix for combining the semantics of the current word with the left sentence of the next word, s is a non-linear activation function, and the right sentence of the last word in the case text shares a parameter x r (t n )；
x l (t i )＝s(W (l) x l (t i-1 )+W (yl) e(t i-1 )) (1)
x r (t i )＝s(W (r) x r (t i+1 )+W (yr) e(t i+1 )) (2)
Defining the case word m by equation (3) i It is the left context vector x l (t i ) Word embedding e (t) i ) And the right context vector x r (t i ) The connection of (1);
m i ＝[x l (t i )；e(t i )；x r (t i )] (3)
after obtaining the word m i After representation of (c), the result is sent to the next layer using a linear transformation and a relu activation function
Figure BDA0003239445580000051
Figure BDA0003239445580000052
Figure BDA0003239445580000053
Is a potential semantic vector, and determines the most useful part for representing the file by analyzing each semantic factor in the file;
representing the case by using convolutional neural network technology, and applying a maximum pool layer y after calculating to obtain representations of all words (3) ；
Figure BDA0003239445580000054
The max function is an element-level function, y (2) The jth element of (1)
Figure BDA0003239445580000055
Is that
Figure BDA0003239445580000056
(i∈[1,n]) The largest of the n elements;
the last layer of the model is the output layer y (4) Which is defined as
y (4) ＝W (4) y (3) +b (4) (6)
Finally, the softmax function is used at y (4) Converting the output number into probabilities, wherein the sum of all the probabilities is 1, and taking an element corresponding to the maximum probability in the output result as a result;
Figure BDA0003239445580000057
in the model, a cross entropy loss function is used as the basis for the tuning parameter, as in equation (8), where p is the actual value and p is i Is the predicted value output of the network, and the differences of n groups of p sums are added to obtain the average value to obtain the loss value L (p, p) i ) And parameters are adjusted according to the loss value, so that the model has better accuracy:
Figure BDA0003239445580000061
3. construction of an analysis model of a medical record of traditional Chinese medicine
The trained model is applied to the analysis of the traditional Chinese medical record, and the specific searching process of the quantitative relation between the symptoms and the disease names is shown in the figure 2:
the first step is as follows: input (line offset, symptom data);
the second step is that: preprocessing input text data to stop operations such as words, word segmentation, word vectorization and the like;
the third step: carrying out cascade operation on the word vectors;
the fourth step: inputting the cascade word vectors into a trained neural network model for calculation, acquiring the feature vectors by using a pooling layer, calculating the corresponding probability of each feature word vector by using softmax, and calculating and outputting the corresponding disease name according to the finally obtained result.
The traditional Chinese medical record analysis method of the invention has better performance than the existing Bi-LSTM neural network which is widely used under the condition of taking accuracy (Precision Rate), Recall Rate (Recall Rate) and F value (F-measure) as judgment standards.
The evaluation standard accuracy (Precision Rate), Recall Rate (Recall Rate) and F value (F-measure) are calculated as follows:
Figure BDA0003239445580000062
Figure BDA0003239445580000063
Figure BDA0003239445580000064
wherein, X m Representing the text contained in the real category m, X n Indicating the number of texts, X, contained in the classified category n m,n Representing the number of documents belonging to the true category m in the classified category n, Precision (m, n) representing the accuracy, the ratio of the number of correctly classified information pieces to the number of all information pieces, Recall (m, n) representing the Recall rate, classificationThe ratio of the number of correct pieces of information to the number of correct pieces of information in the sample, and F represents the harmonic mean of accuracy and recall.
To verify the effectiveness of the method of the present invention, we used a medical record dataset for training and compared the results of the experiment. The data set comprises nine major categories of internal medicine, anorectal department, ophthalmology, otorhinolaryngology department, surgery, orthopedics and traumatology department, gynecology, dermatology and pediatrics of the traditional Chinese medicine, and adopts the standard disease diagnosis curative effect standard ZY/T001.1-94 of the Chinese medicine industry of the people's republic of China as the diagnosis standard. In addition, the data set is divided into a training set, a development set and a test set according to proportion, and the proportion of the training set, the development set and the test set is 64%, 16% and 20% respectively.
Before starting the training, we preprocessed the data set as follows. First, we use jieba participles to segment sentences, removing stop words and symbols in the text. Next, we pre-train word embedding, which is a distributed representation of words that represent text as numbers, so that a computer can read the content of the text and thus be suitable for input to a neural network. Conventional tokens, such as a hot token, will cause a dimensional disaster because a one-hot token occupies a large dimensional space, and will generally not be used because its invalid information is much more than valid information. We choose to use the Skip-gram model to pre-train word embedding. This model is a more common model for many NLP tasks.
The data set was previously divided into a training set and a test set. Three methods are respectively used for training, namely RNN and RNN Attention and the method (namely RCNN) of the invention, and the test results are compared, the accuracy is used as an evaluation index for comparison, and the Loss value and the time are also compared.
We set up the neural network according to the training data, in order to get a fair experimental result, we used bi-directional GRU in the cyclic neural network part of the three models, and 4 layers in the three models, learning rate is set to 0.001, to prevent the over-fitting phenomenon, we used drop-out and set its value to 0.5, cross entropy loss function selected for loss function, Adam optimizer used, hidden layer size is set to 256, input word vector length is 100, word embedded vector size and context vector size are both set to 50, finally, relu activation function used in our experiment. Training results as shown in fig. 3-7, the method (RCNN) of the present invention is advantageous in terms of test accuracy, test loss, training accuracy, training time, and training loss.

Claims (3)

1. The traditional Chinese medical record analysis method based on the cyclic convolution neural network is characterized by comprising the following steps of:
collecting the existing medical records, carrying out noise processing on the collected medical records, eliminating noise data and reserving effective data;
step two, carrying out data preprocessing on the medical records subjected to noise processing;
step three, carrying out data training on the data preprocessed in the step two and detecting a training result to finally obtain a traditional Chinese medicine medical record analysis model;
and step four, inputting symptoms in the traditional Chinese medical record analysis model to carry out traditional Chinese medical record analysis to obtain a traditional Chinese medical record analysis result.
2. The method for analyzing the medical records of traditional Chinese medicine based on the cyclic convolution neural network as claimed in claim 1, wherein the specific process of preprocessing the medical records data in the step is as follows:
(1) acquiring existing CSV data of a medical record text, labeling, performing jieba word segmentation, and removing stop words in a Chinese stop word bank;
(2) setting a frequency threshold, and if the frequency of a certain word appearing in the case text is less than the set frequency threshold, deleting the word;
(3) setting a characteristic selection threshold, selecting words with information values higher than the characteristic selection threshold in the medical record text X, and regarding the words as the characteristic values of the medical record text X to obtain characteristic vectors of the medical record text X;
(4) and (3) preprocessing the medical case text by using a word vector training model, wherein the preprocessing comprises word stopping, word segmentation and word vectorization, and the word vector training model is prepared for neural network training.
3. The method for analyzing the medical records of traditional Chinese medicine based on the cyclic convolution neural network of claim 1, wherein the specific method for training the data to obtain the medical record analysis model in the third step is as follows:
(1) definition of x l (t i ) As a medical record text sentence t i Left sentence of (1), x r (t i ) As a sentence t i Right sentence of (1), x l (t i ) And x r (t i ) Are all dense vectors with real valued elements, t i Left sentence x of l (t i ) Calculated from equation (1), the right sentence x r (t i ) Calculated from equation (2), where e (t) i-1 ) Is t i-1 Word embedding of, t i-1 Is a dense vector with | e | real-valued elements, the left sentence of the first word in the case text uses the same shared parameter x l (t 1 )，W (l) Is a matrix, W, that converts a hidden layer to the next hidden layer (yl) Is a matrix for combining the semantics of the current word with the left sentence of the next word, s is a non-linear activation function, and the right sentence of the last word in the case text shares a parameter x r (t n )；
x l (t i )＝s(W (l) x l (t i-1 )+W (yl) e(t i-1 )) (1)
x r (t i )＝s(W (r) x r (t i+1 )+W (yr) e(t i+1 )) (2)
Defining the case word m by equation (3) i It is the left context vector x l (t i ) Word embedding e (t) i ) And the right context vector x r (t i ) The connection of (1);
m i ＝[x l (t i )；e(t i )；x r (t i )] (3)
after obtaining the word m i After representation of (c), the result is sent to the next layer using a linear transformation and a relu activation function
Figure FDA0003239445570000021
Figure FDA0003239445570000022
Figure FDA0003239445570000023
Is a potential semantic vector, and determines the most useful part for representing the file by analyzing each semantic factor in the file;
representing the case by using convolutional neural network technology, and applying a maximum pool layer y after calculating to obtain representations of all words (3) ；
Figure FDA0003239445570000024
The max function is an element-level function, y (2) The jth element of (1)
Figure FDA0003239445570000025
Is that
Figure FDA0003239445570000026
(i∈[1，n]) The largest of the n elements;
the last layer of the model is the output layer y (4) Which is defined as
y (4) ＝W (4) y (3) +b (4) (6)
Finally, the softmax function is used at y (4) In the above, the output number is converted into probability, and the sum of all probabilities is 1, the output number is inputTaking the element corresponding to the maximum probability in the result as the result;
Figure FDA0003239445570000027
in the model, a cross entropy loss function is used as the basis for the tuning parameter, as in equation (8), where p is the actual value and p is i Is the predicted value output of the network, resulting in a loss value L (p, p) i ) The parameters are adjusted according to the loss value, so that the model has better accuracy,
Figure FDA0003239445570000031
CN202111012711.7A 2021-08-31 2021-08-31 Traditional Chinese medical record analysis method based on cyclic convolution neural network Pending CN114822734A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202111012711.7A CN114822734A (en) 2021-08-31 2021-08-31 Traditional Chinese medical record analysis method based on cyclic convolution neural network

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202111012711.7A CN114822734A (en) 2021-08-31 2021-08-31 Traditional Chinese medical record analysis method based on cyclic convolution neural network

Publications (1)

Publication Number Publication Date
CN114822734A true CN114822734A (en) 2022-07-29

Family

ID=82525874

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202111012711.7A Pending CN114822734A (en) 2021-08-31 2021-08-31 Traditional Chinese medical record analysis method based on cyclic convolution neural network

Country Status (1)

Country Link
CN (1) CN114822734A (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115620919A (en) * 2022-10-26 2023-01-17 河南科技大学 Coronary heart disease intelligent auxiliary dialectic device based on machine learning

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115620919A (en) * 2022-10-26 2023-01-17 河南科技大学 Coronary heart disease intelligent auxiliary dialectic device based on machine learning

Similar Documents

Publication Publication Date Title
CN109460473B (en) Electronic medical record multi-label classification method based on symptom extraction and feature representation
CN108831559B (en) Chinese electronic medical record text analysis method and system
CN111192680B (en) Intelligent auxiliary diagnosis method based on deep learning and collective classification
CN111414393B (en) Semantic similar case retrieval method and equipment based on medical knowledge graph
CN109935336B (en) Intelligent auxiliary diagnosis system for respiratory diseases of children
WO2016192612A1 (en) Method for analysing medical treatment data based on deep learning, and intelligent analyser thereof
Fang et al. Feature Selection Method Based on Class Discriminative Degree for Intelligent Medical Diagnosis.
CN109670177A (en) One kind realizing the semantic normalized control method of medicine and control device based on LSTM
CN108062978B (en) Method for predicting main adverse cardiovascular events of patients with acute coronary syndrome
CN110335653A (en) Non-standard case history analytic method based on openEHR case history format
CN109003677B (en) Structured analysis processing method for medical record data
CN109994201B (en) Diabetes and hypertension probability calculation system based on deep learning
CN110097096B (en) Text classification method based on TF-IDF matrix and capsule network
CN113610859B (en) Automatic thyroid nodule segmentation method based on ultrasonic image
CN111968741A (en) Diabetes complication high-risk early warning system based on deep learning and integrated learning
CN111191415A (en) Operation classification coding method based on original operation data
CN111858940A (en) Multi-head attention-based legal case similarity calculation method and system
CN112926332A (en) Entity relationship joint extraction method and device
Lu et al. Speech depression recognition based on attentional residual network
CN114428860A (en) Pre-hospital emergency case text recognition method and device, terminal and storage medium
Zou et al. Intelligent diagnosis method of bearing fault based on ICEEMDAN and Ghost-IRCNN
CN114822734A (en) Traditional Chinese medical record analysis method based on cyclic convolution neural network
US20220156489A1 (en) Machine learning techniques for identifying logical sections in unstructured data
CN113284627A (en) Medication recommendation method based on patient characterization learning
CN116403706A (en) Diabetes prediction method integrating knowledge expansion and convolutional neural network

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