CN117438104A - Intelligent medicine early warning method, electronic equipment and computer storage medium - Google Patents
Intelligent medicine early warning method, electronic equipment and computer storage medium Download PDFInfo
- Publication number
- CN117438104A CN117438104A CN202311765430.8A CN202311765430A CN117438104A CN 117438104 A CN117438104 A CN 117438104A CN 202311765430 A CN202311765430 A CN 202311765430A CN 117438104 A CN117438104 A CN 117438104A
- Authority
- CN
- China
- Prior art keywords
- information
- taken
- layer
- medicines
- medicine
- 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
- 239000003814 drug Substances 0.000 title claims abstract description 222
- 238000000034 method Methods 0.000 title claims abstract description 33
- 229940079593 drug Drugs 0.000 claims abstract description 126
- 230000007774 longterm Effects 0.000 claims abstract description 53
- 238000003062 neural network model Methods 0.000 claims abstract description 40
- 238000012549 training Methods 0.000 claims abstract description 15
- 239000013598 vector Substances 0.000 claims description 13
- 230000000694 effects Effects 0.000 claims description 9
- 230000006870 function Effects 0.000 claims description 9
- 238000004422 calculation algorithm Methods 0.000 claims description 8
- 238000000605 extraction Methods 0.000 claims description 8
- 238000013528 artificial neural network Methods 0.000 claims description 7
- 230000015654 memory Effects 0.000 claims description 7
- 230000000857 drug effect Effects 0.000 claims description 6
- 238000005457 optimization Methods 0.000 claims description 6
- 238000012546 transfer Methods 0.000 claims description 6
- 229940126680 traditional chinese medicines Drugs 0.000 claims description 5
- 229940126673 western medicines Drugs 0.000 claims description 5
- 230000004913 activation Effects 0.000 claims description 4
- 230000008569 process Effects 0.000 claims description 4
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 claims description 3
- 230000002457 bidirectional effect Effects 0.000 claims description 3
- 238000004590 computer program Methods 0.000 claims description 3
- 230000000007 visual effect Effects 0.000 claims description 3
- 238000013473 artificial intelligence Methods 0.000 abstract description 3
- 238000010276 construction Methods 0.000 abstract description 2
- 238000010801 machine learning Methods 0.000 abstract description 2
- 238000006243 chemical reaction Methods 0.000 description 4
- 238000002372 labelling Methods 0.000 description 4
- 230000001133 acceleration Effects 0.000 description 2
- 230000003213 activating effect Effects 0.000 description 2
- 238000004458 analytical method Methods 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 230000007787 long-term memory Effects 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 230000004075 alteration Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 230000001684 chronic effect Effects 0.000 description 1
- 238000004140 cleaning Methods 0.000 description 1
- 230000019771 cognition Effects 0.000 description 1
- 230000008034 disappearance Effects 0.000 description 1
- 201000010099 disease Diseases 0.000 description 1
- 208000037265 diseases, disorders, signs and symptoms Diseases 0.000 description 1
- 230000004927 fusion Effects 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000011176 pooling Methods 0.000 description 1
- 238000012216 screening Methods 0.000 description 1
- 230000011218 segmentation Effects 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
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
- G16H70/00—ICT specially adapted for the handling or processing of medical references
- G16H70/40—ICT specially adapted for the handling or processing of medical references relating to drugs, e.g. their side effects or intended usage
-
- 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
- G06N3/0442—Recurrent networks, e.g. Hopfield networks characterised by memory or gating, e.g. long short-term memory [LSTM] or gated recurrent units [GRU]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/048—Activation functions
-
- 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
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N5/00—Computing arrangements using knowledge-based models
- G06N5/02—Knowledge representation; Symbolic representation
- G06N5/022—Knowledge engineering; Knowledge acquisition
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Computing Systems (AREA)
- Data Mining & Analysis (AREA)
- Software Systems (AREA)
- General Health & Medical Sciences (AREA)
- Mathematical Physics (AREA)
- General Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Computational Linguistics (AREA)
- Molecular Biology (AREA)
- Life Sciences & Earth Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Public Health (AREA)
- Pharmacology & Pharmacy (AREA)
- Toxicology (AREA)
- Medicinal Chemistry (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Primary Health Care (AREA)
- Medical Informatics (AREA)
- Epidemiology (AREA)
- Chemical & Material Sciences (AREA)
- Medical Treatment And Welfare Office Work (AREA)
Abstract
The invention provides an intelligent medicine early warning method, electronic equipment and a computer storage medium, which aim to analyze the medicine information to be taken by identifying the medicine information to be taken currently uploaded by a user, and analyze whether medicines which cannot be taken simultaneously exist in the medicine to be taken under the condition that the user does not fix long-term medicine information; under the condition that the user has fixed long-term medication information, analyzing whether medicines which cannot be taken simultaneously exist in the medicines which are taken for a fixed long time and the medicines which are to be taken simultaneously, if the medicines which cannot be taken simultaneously exist, warning the doctor and the user simultaneously, and then communicating the doctor and the user; by constructing the first neural network model and the second neural network model and adopting a knowledge graph construction mode, the advantages of artificial intelligence and big data identification are fully utilized, machine learning training is performed in advance, a large number of different kinds of medicine information is collected and judged, and early warning is performed in advance.
Description
Technical Field
The invention belongs to the field of artificial intelligence, and particularly relates to acquisition and intelligent identification of medicine information.
Background
At present, people have the habit of purchasing medicines by themselves, and aiming at the taboo of medicines, most people can only acquire cognition through medicine specifications, but in the medicine specifications, only side effects of taking the medicines are often mentioned, and the medicines cannot be taken together with the medicines. For many chronic patients, the habit of taking medicine for a long time is that the medicine purchased at a pharmacy is not clear of the medicine taking contraindications with other medicines; some patients often forget to explain the medicine being taken by themselves with doctors when the patients are in a hospital consultation, and for the elderly patients, the names and contraindications of taking the medicine are difficult to remember. The medicine prescribed by the doctor in the hospital is prescribed for the disease being diagnosed, and the contraindication of the prescribed medicine cannot be made without knowing that the patient is taking other medicines. The medicines taken by the patients at present are generally relatively miscellaneous, and western medicines, traditional Chinese medicines, tibetan medicines and the like are included, so that patients or doctors in a certain department can not be aware of insufficient medicines which are forbidden or unsuitable to take together, and careful screening cannot be achieved.
Disclosure of Invention
In view of the above-mentioned shortcomings of the prior art, the present invention proposes an intelligent medicine method, an electronic device and a computer storage medium, which aim to analyze the current medicine information to be taken uploaded by a user by identifying the medicine information, and analyze whether medicines which cannot be taken simultaneously exist in the medicines to be taken under the condition that the user does not fix the long-term medicine information; under the condition that the fixed long-term medication information exists in the user, analyzing whether medicines which cannot be taken simultaneously exist in the medicines which are taken for a fixed long time and the medicines which are to be taken simultaneously, if the medicines which cannot be taken simultaneously exist, giving an early warning to doctors and the user at the same time, and communicating the doctors with the user, so that the patients cannot take medicines which are contraindicated mutually due to the fact that the medicine properties of various medicines are not known. Meanwhile, as the built knowledge graph is continuously updated in the background database, various medicine information such as traditional Chinese medicines, western medicines, tibetan medicines and the like and the contraindications of the medicine properties of the medicine information can be updated, the safe use of the medicine can be ensured in a large data layer, and misjudgment of a patient or a doctor in a certain department due to lack of knowledge of a certain medicine can be prevented.
The detailed technical scheme of the invention is as follows:
s1, acquiring to-be-taken medicine information uploaded by a user, and extracting characteristic information from the to-be-taken medicine information, wherein the characteristic information of the to-be-taken medicine information comprises names of to-be-taken medicines, the taking quantity of the to-be-taken medicines and components of the to-be-taken medicines;
s2, acquiring fixed long-term medication information of a user, and extracting characteristic information from the fixed long-term medication information, wherein the characteristic information of the fixed long-term medication information comprises names of the fixed long-term medication, the administration quantity of the fixed long-term medication and components of the fixed long-term medication;
s3, if the fixed long-term medication information of the user is not obtained, inputting the characteristic information of the to-be-taken medicine information into various characteristic models in a knowledge-graph library, identifying the to-be-taken medicine by the first neural network model, judging whether the input to-be-taken medicine has medicines which cannot be taken simultaneously, and if so, sending first early warning information to a user equipment end and a doctor equipment end; if the fixed long-term medication information of the user is obtained, the characteristic information of the to-be-taken medicine information and the characteristic information of the fixed long-term medication information are input into a second neural network model, the second neural network model identifies the to-be-taken medicine and the fixed long-term medication, whether the to-be-taken medicine and the fixed long-term medication have medicines which cannot be taken simultaneously or not is judged, and if the medicines which cannot be taken simultaneously exist, second early warning information is sent to a user equipment side and a doctor equipment side. Further, the medicines in the step S1 comprise traditional Chinese medicines, western medicines and Tibetan medicines.
Further, in the step S1, the information of the medicine to be taken is uploaded in a voice, text or image manner; the medicine information to be taken comprises a user's own description or a medicine instruction book;
further, in step S2, the fixed long-term medication information of the user allows the user to upload itself or connect to a hospital background system to obtain.
Further, in the step S3, the method for forming the first neural network model includes: acquiring the names, components and drug effects of the drugs, and taking contraindications of different drugs; the taking contraindications comprise medicines which cannot be taken simultaneously and specific side effects which can be generated after the medicines are taken simultaneously; constructing a knowledge graph base according to the acquired names, components and drug effects of the drugs and the taking contraindications of different drugs, extracting various characteristics in the knowledge graph base, and training a first neural network model according to the characteristics; the first neural network model selects an LSTM as a basic learner, and a basic unit of the LSTM comprises an output gate, an input gate, a forgetting gate and a cell state, wherein the cell state can store long-term history information, and effective circulation of the information in the training process of the first neural network model is ensured; the first neural network model comprises two LSTM layers, wherein each LSTM layer comprises a basic unit of the LSTM serving as a basic learner, and the basic unit comprises an output gate, an input gate, a forgetting gate and a cell state; the update formula of each component of the LSTM basic unit is as follows:
；
wherein:、/>、/>、/>candidate cell information, input gate, forget gate and output gate of the LSTM structure at the t moment respectively; />、/>、/>And->Respectively represent the output gate, the input gate, the forgetting gate and the input of the cell state to the current momentWeight of->、/>、/>And->Respectively show the output door, the input door, the forgetting door and the hidden layer of the cell stateWeight of->Representing a bias term; />Activating a function for Sigmoid; />And->Respectively representing the cell states of the LSTM structure at the current time t and the last time t-1, ++>Hidden information representing LSTM structure.
When the first neural network model is trained, various features in a knowledge graph base are input into a first LSTM layer, hidden states of all time steps of the first LSTM layer are output into a second LSTM layer, then the second LSTM layer only takes the hidden state of the last time step as output, all the outputs are spliced and then input into a full-connection layer for feature combination, and finally the output of the full-connection layer is input into a sigmoid layer for final classification output.
Further, in the step S3, the second neural network model includes five layers, specifically: an input layer, a word vector extraction layer, a Bi-LSTM layer, a CRF layer and an output layer;
the first layer is an input layer, preprocessed source Wen Benshu data is input, and the source text data is obtained by converting the to-be-taken medicine information and the fixed long-term medicine information; the second layer is a word vector extraction layer and is mainly used for representing words in an input word and sentence sequence as entity vectors with n dimensions; the third layer is a Bi-LSTM layer, the Bi-LSTM neural network is trained to serve as a classifier, an Adam algorithm is selected to serve as an optimization algorithm of the model, the tanh function is used as an activation function, the Bi-LSTM layer is designed to be overlapped with the bidirectional LSTM layer, parameters are initialized after the basic structure is completed, and the optimization model is further adjusted in subsequent training; the fourth layer is a CRF layer, and the CRF layer calculates the optimal joint probability among the labels through learning the label marks and the text transfer rule so as to optimize the label sequence; the fifth layer is an output layer for outputting medicine information and corresponding labels which cannot be taken simultaneously.
Further, in step S3, if the user equipment end and the doctor equipment end receive the early warning information, the doctor may manually update the early warning information at the doctor equipment end, and send the updated early warning information to the user equipment end, and the user may communicate with the doctor according to the updated early warning information sent by the doctor, where the updating includes adding, modifying, and deleting; the early warning information is displayed on the visual interfaces of the user equipment end and the doctor equipment end at the same time, and comprises names and shapes of medicines which cannot be taken at the same time, and specific side effects and other tabu instructions which can be generated after the medicines are taken at the same time. If the user equipment end and the doctor equipment end do not receive the early warning information, the reminding information is sent to the doctor equipment end and the user equipment end; the doctor can communicate the related medication with the user based on the reminding information; the reminding information comprises the information of the medicines to be taken and the fixed long-term medication information.
An electronic device comprising a processor, a memory, and one or more programs, wherein the one or more programs are stored in the memory and configured to perform the intelligent medicine warning method of steps S1-S3 by the one or more processors.
A computer storage medium for storing a computer program for executing the intelligent medicine early warning method of steps S1-S3.
Compared with the traditional method, the invention creatively proposes: judging whether medicines which cannot be taken simultaneously exist or not by judging the medicine information to be taken which is uploaded by a user, and carrying out early warning according to the medicine information; the fixed long-term medication information of the user can be obtained and judged together with the information of the medicine to be taken, so that the tabu reaction of the medicine which is fixed and taken for a long time by the user and the medicine which is newly taken is avoided.
By constructing the first neural network model and the second neural network model and adopting a knowledge graph construction mode, the advantages of artificial intelligence and big data identification are fully utilized, machine learning training is performed in advance, a large number of different kinds of medicine information is collected and judged, and under the condition that the tabu reactions of a plurality of medicines such as western medicines, traditional Chinese medicines, tibetan medicines and the like are complex, the tabu reactions among the medicines can be accurately pre-judged, and early warning is performed in advance.
According to the different data quantities to be judged, different neural network algorithms are adopted, and when the data quantity is large, a Bi-LSTM-CRF model is adopted for judging, and the Bi-LSTM-CRF model can furthest reserve upper and lower Wen Yuyi and solve the problem of labeling bias, so that the method is widely applied to entity labeling research. The text related to the medicine is different in time point of collection, information recorded by each module is different, the text body quantity is also different, and long-term dependence is needed to be processed through long-term memory for a longer text segment, so that the entity recognition effect is effectively improved by combining with surrounding contexts. The Bi-LSTM-CRF model effectively meets the above needs.
Drawings
FIG. 1 is a flow chart of the intelligent medicine early warning method of the invention.
Fig. 2 is a diagram of a first neural network model in the intelligent medicine early warning method of the present invention.
FIG. 3 is a diagram of a second neural network model in the intelligent drug early warning method of the present invention.
Detailed Description
The present application will now be described more fully hereinafter with reference to the accompanying drawings, in which embodiments of the invention are shown.
The specific method is shown in fig. 1, and comprises the following steps: s1, acquiring to-be-taken medicine information uploaded by a user, and extracting characteristic information from the to-be-taken medicine information, wherein the characteristic information of the to-be-taken medicine information comprises names of to-be-taken medicines, the taking quantity of the to-be-taken medicines and components of the to-be-taken medicines;
further, the medicine comprises traditional Chinese medicine, western medicine and Tibetan medicine.
Further, in the step S1, the information of the medicine to be taken is uploaded in a voice, text or image manner; the medicine information to be taken comprises a user's own description or a medicine instruction book;
s2, acquiring fixed long-term medication information of a user, and extracting characteristic information from the fixed long-term medication information, wherein the characteristic information of the fixed long-term medication information comprises names of the fixed long-term medication, the administration quantity of the fixed long-term medication and components of the fixed long-term medication;
further, in step S2, the fixed long-term medication information of the user allows the user to upload itself or connect to a hospital background system to obtain.
S3, if the fixed long-term medication information of the user is not obtained, inputting the characteristic information of the to-be-taken medicine information into a first neural network model, identifying the to-be-taken medicine by the first neural network model, judging whether the input to-be-taken medicine has medicines which cannot be taken simultaneously, and if so, sending first early warning information to a user equipment end and a doctor equipment end; if the fixed long-term medication information of the user is obtained, the characteristic information of the to-be-taken medicine information and the characteristic information of the fixed long-term medication information are input into a second neural network model, the second neural network model identifies the to-be-taken medicine and the fixed long-term medication, whether the to-be-taken medicine and the fixed long-term medication have medicines which cannot be taken simultaneously or not is judged, and if the medicines which cannot be taken simultaneously exist, second early warning information is sent to a user equipment side and a doctor equipment side. Further, in the step S3, the first neural network model is configured as shown in fig. 2, and the configuring step includes: acquiring the names, components and drug effects of the drugs, and taking contraindications of different drugs; the taking contraindications comprise medicines which cannot be taken simultaneously and specific side effects which can be generated after the medicines are taken simultaneously; constructing a knowledge graph base according to the acquired names, components and drug effects of the drugs and the taking contraindications of different drugs, extracting various characteristics in the knowledge graph base, and training a first neural network model according to the characteristics; the first neural network model selects an LSTM as a basic learner, and a basic unit of the LSTM comprises an output gate, an input gate, a forgetting gate and a cell state, wherein the cell state can store long-term history information, and effective circulation of the information in the training process of the first neural network model is ensured; the first neural network model comprises two LSTM layers, wherein each LSTM layer comprises a basic unit of the LSTM serving as a basic learner, and the basic unit comprises an output gate, an input gate, a forgetting gate and a cell state; the update formula of each component of the LSTM basic unit is as follows:
；
wherein:、/>、/>、/>candidate cell information, input gate, forget gate and output gate of the LSTM structure at the t moment respectively; />、/>、/>And->Respectively represent the output gate, the input gate, the forgetting gate and the input of the cell state to the current momentWeight of->、/>、/>And->Respectively show the output door, the input door, the forgetting door and the hidden layer of the cell stateWeight of->Representing a bias term; />Activating a function for Sigmoid; />And->Respectively representing the cell states of the LSTM structure at the current time t and the last time t-1, ++>Hidden information representing LSTM structure.
When the first neural network model is trained, various features in a knowledge graph base are input into a first LSTM layer, hidden states of all time steps of the first LSTM layer are output into a second LSTM layer, then the second LSTM layer only takes the hidden state of the last time step as output, all the outputs are spliced and then input into a full-connection layer for feature combination, and finally the output of the full-connection layer is input into a sigmoid layer for final classification output.
In particular, an expert database is provided, and the content in the knowledge graph database can be corrected and updated manually.
Further, in the step S3, the second neural network model is formed as shown in fig. 3, and the forming method includes: although the LSTM can acquire the characteristic information of a long distance, the information acquired by the LSTM is information before the output time, and reverse information is not utilized. In text semantic analysis, unidirectional LSTM can only use previous words for semantic analysis, while Bi-LSTM can use past and future information to make more complete and detailed decisions. Generally, bi-LSTM can simultaneously utilize forward and backward information, so that more information can be obtained when text information is processed, which is more beneficial to subsequent classification tasks.
Particularly, the Bi-LSTM-CRF model can not only reserve upper and lower Wen Yuyi to the greatest extent, but also solve the problem of labeling bias, and is widely applied to entity labeling research. The text related to the medicine is different in time point of collection, information recorded by each module is different, the text body quantity is also different, and long-term dependence is needed to be processed through long-term memory for a longer text segment, so that the entity recognition effect is effectively improved by combining with surrounding contexts. The Bi-LSTM model effectively meets the requirements, long-distance information utilization and gradient disappearance problem alleviation can be achieved through a memory unit and a threshold mechanism in the LSTM, and context characteristics can be captured through a forward LSTM network and a backward LSTM network, so that a more comprehensive semantic vector is generated. The complete context information is obtained through forward and backward LSTM calculation, and the entity labels and the probabilities corresponding to the characters are obtained by inputting the full-connection neural network. In addition, the CRF layer calculates the optimal joint probability among the labels through learning the label marks and the text transfer rules to optimize the label sequence, so that the problem of unreasonable label prediction sequence is effectively solved, and further, the entity label of the fusion structure information under the granularity of words is obtained. On the basis, the final entity is extracted by means of the entity boundary and is used as a domain dictionary, so that the accuracy of text word segmentation of the subsequent medicines is improved.
Specifically, the second neural network model comprises five layers, specifically: an input layer, a word vector extraction layer, a Bi-LSTM layer, a CRF layer and an output layer; the first layer is an input layer, preprocessed source Wen Benshu data is input, and the source text data is obtained by converting the to-be-taken medicine information and the fixed long-term medicine information; the second layer is a word vector extraction layer and is mainly used for representing words in an input word and sentence sequence as entity vectors with n dimensions; the third layer is a Bi-LSTM layer, and the Bi-LSTM neural network is trained to serve as a classifier. An Adam algorithm is selected as an optimization algorithm of the model, and a tanh function is an activation function. The detailed model building procedure is as follows. Step one: and taking the comment text which is not segmented after cleaning as a data set, and dividing the comment text into a training data set and a test data set according to a proportion. Inputting the training data set into a model, and constructing a data input layer of the neural network model; step two: and constructing a Bi-LSTM model basic structure based on TensorFlow, and establishing a convolution layer, an activation layer, a pooling layer, a full connection layer and the like. The word vector model pre-trained is selected for use herein as the word embedding layer. The Bi-LSTM layer is designed to be overlapped with the bidirectional LSTM layer, parameters are initialized after the basic structure is completed, and an optimization model is further adjusted in subsequent training. The Bi-LSTM neural network structure model is divided into 2 independent LSTM, the input sequence is respectively input into 2 LSTM neural networks in positive sequence and reverse sequence for feature extraction, word vectors formed after 2 output vectors (namely, forward hidden layer output and reverse hidden layer output) are spliced are used as hidden layer output at the current moment, and the final feature expression of the word is obtained; the fourth layer is a CRF layer, the CRF layer calculates the optimal joint probability among labels to optimize the label sequence through learning the label labels and the text transfer rules, the CRF algorithm can well capture the context semantic dependence of the sequence in an implicit state, and the conditional random field is built on the state transfer rules, so that the condition that the Bi-LSTM model can not learn the state transfer rules in the text can be well compensated. As the Bi-LSTM model has various GPU acceleration and multi-machine asynchronous SGD and other large-scale data training acceleration methods, the Bi-LSTM can rapidly finish feature extraction of an observation sequence when facing large-scale data. The fifth layer is an output layer for outputting medicine information and corresponding labels which cannot be taken simultaneously.
In particular, the cross entropy loss function is mainly used in the training process in this embodiment:
；
where i and j are the index of the sentence and the category index, respectively,is a predictive tag,>is regularized item, +.>Is an optimizable set of parameters for the model.
In particular, the source text data is obtained by converting the to-be-taken medicine information and the fixed long-term medication information, and the feature information can be extracted from the voice and the image uploaded by the user and converted into text data by a common multimedia conversion mode, which is not limited herein.
Further, in step S3, if the user equipment end and the doctor equipment end receive the early warning information in step S3, the doctor may manually update the early warning information at the doctor equipment end, and send the updated early warning information to the user equipment end, and the user may communicate with the doctor according to the updated early warning information sent by the doctor, where the updating includes adding, modifying, and deleting.
Furthermore, the early warning information is displayed on the visual interface of the user equipment end and the doctor equipment end at the same time, and comprises names and shapes of medicines which cannot be taken at the same time, and specific side effects and other tabu instructions which can be generated after the medicines are taken at the same time.
Further, in the step S3, if the user equipment end and the doctor equipment end do not receive the early warning information, the reminding information is sent to the doctor equipment end and the user equipment end; the doctor can communicate the related medication with the user based on the reminding information; the reminding information comprises the information of the medicines to be taken and the fixed long-term medication information.
An electronic device comprising a processor, a memory and one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the one or more processors as the intelligent medicine early-warning method of steps S1-S3 and a computer storage medium of intelligent medicine early-warning for storing a computer program for executing the above-described intelligent medicine early-warning method based on.
Various modifications and alterations of this invention may be made by those skilled in the art without departing from the scope of this invention.
Claims (10)
1. An intelligent medicine early warning method is characterized by comprising the following steps:
s1, acquiring to-be-taken medicine information uploaded by a user, and extracting characteristic information from the to-be-taken medicine information, wherein the characteristic information of the to-be-taken medicine information comprises names of to-be-taken medicines, the taking quantity of the to-be-taken medicines and components of the to-be-taken medicines; the medicine comprises traditional Chinese medicines, western medicines and Tibetan medicines;
s2, acquiring fixed long-term medication information of a user, and extracting characteristic information from the fixed long-term medication information, wherein the characteristic information of the fixed long-term medication information comprises names of the fixed long-term medication, the administration quantity of the fixed long-term medication and components of the fixed long-term medication;
s3, if the fixed long-term medication information of the user is not obtained, inputting the characteristic information of the to-be-taken medicine information into a first neural network model, identifying the to-be-taken medicine by the first neural network model, and judging whether the input to-be-taken medicine has medicines which cannot be taken simultaneously; if medicines which cannot be taken at the same time exist, first early warning information is sent to a user equipment end and a doctor equipment end; if the fixed long-term medication information of the user is obtained, the characteristic information of the to-be-taken medicine information and the characteristic information of the fixed long-term medication information are input into a second neural network model, the second neural network model identifies the to-be-taken medicine and the fixed long-term medication, whether the to-be-taken medicine and the fixed long-term medication have medicines which cannot be taken simultaneously or not is judged, and if the medicines which cannot be taken simultaneously exist, second early warning information is sent to a user equipment side and a doctor equipment side.
2. The method according to claim 1, wherein in step S1, the information of the medicine to be taken is uploaded in a voice, text or image manner; the medicine information to be taken comprises a user's own description or a medicine instruction book; in step S2, the fixed long-term medication information of the user allows the user to upload itself or connect to a hospital background system to obtain.
3. The method according to claim 1, wherein in the step S3, the method for constructing the first neural network model includes: acquiring the names, components and drug effects of the drugs, and taking contraindications of different drugs; the taking contraindications comprise medicines which cannot be taken simultaneously and specific side effects which can be generated after the medicines are taken simultaneously; constructing a knowledge graph base according to the acquired names, components and drug effects of the drugs and the taking contraindications of different drugs, extracting various characteristics in the knowledge graph base, and training a first neural network model according to the characteristics; the first neural network model selects an LSTM as a basic learner, and a basic unit of the LSTM comprises an output gate, an input gate, a forgetting gate and a cell state, wherein the cell state can store long-term history information, and effective circulation of the information in the training process of the first neural network model is ensured; the first neural network model comprises two LSTM layers, wherein each LSTM layer comprises a basic unit of the LSTM serving as a basic learner, and the basic unit comprises an output gate, an input gate, a forgetting gate and a cell state; when the first neural network model is trained, various features in a knowledge graph base are input into a first LSTM layer, hidden states of all time steps of the first LSTM layer are output into a second LSTM layer, then the second LSTM layer only takes the hidden state of the last time step as output, all the outputs are spliced and then input into a full-connection layer for feature combination, and finally the output of the full-connection layer is input into a sigmoid layer for final classification output.
4. A method according to claim 3, characterized in that an expert database is provided, the contents of which are manually corrected and updated.
5. The method according to claim 1, wherein in step S3, the second neural network model is composed of five layers, in particular: an input layer, a word vector extraction layer, a Bi-LSTM layer, a CRF layer and an output layer; the first layer is an input layer, preprocessed source Wen Benshu data is input, and the source text data is obtained by converting the to-be-taken medicine information and the fixed long-term medicine information; the second layer is a word vector extraction layer and is mainly used for representing words in an input word and sentence sequence as entity vectors with n dimensions; the third layer is a Bi-LSTM layer, the Bi-LSTM neural network is trained to serve as a classifier, an Adam algorithm is selected to serve as an optimization algorithm of the model, the tanh function is used as an activation function, the Bi-LSTM layer is designed to be overlapped with the bidirectional LSTM layer, parameters are initialized after the basic structure is completed, and the optimization model is further adjusted in subsequent training; the fourth layer is a CRF layer, and the CRF layer calculates the optimal joint probability among the labels through learning the label marks and the text transfer rule so as to optimize the label sequence; the fifth layer is an output layer for outputting medicine information and corresponding labels which cannot be taken simultaneously.
6. The method according to claim 1, wherein in the step S3, if the user equipment side and the doctor equipment side receive the early warning information, the doctor can manually update the early warning information at the doctor equipment side and send the updated early warning information to the user equipment side, and the user can communicate with the doctor according to the updated early warning information sent by the doctor, and the updating includes adding, modifying and deleting.
7. The method of claim 6, wherein the pre-warning information is displayed on a visual interface of the user equipment side and the doctor equipment side simultaneously, and comprises names and shapes of medicines which cannot be taken simultaneously, specific side effects and other tabu instructions which can be generated after the medicines are taken simultaneously.
8. The method according to claim 1, wherein in the step S3, if the user equipment side and the doctor equipment side do not receive the early warning information, the reminding information is sent to the doctor equipment side and the user equipment side; the doctor can communicate the related medication with the user based on the reminding information; the reminding information comprises the information of the medicines to be taken and the fixed long-term medication information.
9. An electronic device comprising a processor, a memory, and one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the processor for the intelligent medicine warning method of any of claims 1-8.
10. A computer storage medium for storing a computer program for performing the intelligent drug early warning method of one of claims 1-8.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311765430.8A CN117438104B (en) | 2023-12-21 | 2023-12-21 | Intelligent medicine early warning method, electronic equipment and computer storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311765430.8A CN117438104B (en) | 2023-12-21 | 2023-12-21 | Intelligent medicine early warning method, electronic equipment and computer storage medium |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN117438104A true CN117438104A (en) | 2024-01-23 |
| CN117438104B CN117438104B (en) | 2024-03-22 |
Family
ID=89558673
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311765430.8A Active CN117438104B (en) | 2023-12-21 | 2023-12-21 | Intelligent medicine early warning method, electronic equipment and computer storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117438104B (en) |
Citations (21)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106022004A (en) * | 2016-05-20 | 2016-10-12 | 北京千安哲信息技术有限公司 | Method and system for detecting mixed drug conflict |
| CN107330270A (en) * | 2017-06-29 | 2017-11-07 | 广东医睦科技有限公司 | Electronic medicine box system, medication reminding method and device |
| CN108363774A (en) * | 2018-02-09 | 2018-08-03 | 西北大学 | A kind of drug relationship sorting technique based on multilayer convolutional neural networks |
| CN109166610A (en) * | 2018-08-30 | 2019-01-08 | 成都宇亨智能设备科技有限公司 | A kind of intelligent medical order management system and method based on prescription information |
| CN110223751A (en) * | 2019-05-16 | 2019-09-10 | 平安科技（深圳）有限公司 | Prescription evaluation method, system and computer equipment based on medical knowledge map |
| CN110390021A (en) * | 2019-06-13 | 2019-10-29 | 平安科技（深圳）有限公司 | Drug knowledge mapping construction method, device, computer equipment and storage medium |
| CN110929511A (en) * | 2018-09-04 | 2020-03-27 | 清华大学 | Intelligent matching method for personalized traditional Chinese medicine diagnosis and treatment information and traditional Chinese medicine information based on semantic similarity |
| CN111145862A (en) * | 2019-12-30 | 2020-05-12 | 广州享药户联优选科技有限公司 | Medication safety reminding method and device |
| US20200184286A1 (en) * | 2018-12-11 | 2020-06-11 | Southern Taiwan University Of Science And Technology | Smart medication identifying system |
| CN111445976A (en) * | 2020-03-24 | 2020-07-24 | 屹嘉智创(厦门)科技有限公司 | Intelligent rational medication system and method |
| CN111710388A (en) * | 2020-06-18 | 2020-09-25 | 丁玉 | Artificial intelligence medicine prescription system, medicine prescription method and artificial intelligence pharmacy |
| CN112182252A (en) * | 2020-11-09 | 2021-01-05 | 浙江大学 | Intelligent medication question-answering method and device based on medicine knowledge graph |
| CN112820359A (en) * | 2021-02-24 | 2021-05-18 | 北京中医药大学东直门医院 | Liver injury prediction method, apparatus, device, medium, and program product |
| CN112951362A (en) * | 2021-02-23 | 2021-06-11 | 上海商汤智能科技有限公司 | Medicine recommendation method, device, equipment and storage medium |
| CN113707261A (en) * | 2021-08-30 | 2021-11-26 | 平安国际智慧城市科技股份有限公司 | Artificial intelligence-based medicine recommendation method and device and related equipment |
| US20210407642A1 (en) * | 2020-06-24 | 2021-12-30 | Beijing Baidu Netcom Science And Technology Co., Ltd. | Drug recommendation method and device, electronic apparatus, and storage medium |
| CN114724622A (en) * | 2022-03-18 | 2022-07-08 | 清华大学 | Medicine interaction prediction method and device based on medicine knowledge graph |
| WO2022213690A1 (en) * | 2021-04-10 | 2022-10-13 | 浙江大学 | Medication decision support method, apparatus and device based on graphic state machine, and medium |
| CN115223731A (en) * | 2022-07-15 | 2022-10-21 | 桂林理工大学 | Drug interaction prediction method based on DS evidence theory |
| CN115862892A (en) * | 2022-12-14 | 2023-03-28 | 深圳市宝安区妇幼保健院 | Medicine education system based on mobile terminal and application method thereof |
| CN116630062A (en) * | 2023-05-10 | 2023-08-22 | 重庆医药数据信息科技有限公司 | Medical insurance fraud detection method, system and storage medium |
-
2023
- 2023-12-21 CN CN202311765430.8A patent/CN117438104B/en active Active
Patent Citations (21)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106022004A (en) * | 2016-05-20 | 2016-10-12 | 北京千安哲信息技术有限公司 | Method and system for detecting mixed drug conflict |
| CN107330270A (en) * | 2017-06-29 | 2017-11-07 | 广东医睦科技有限公司 | Electronic medicine box system, medication reminding method and device |
| CN108363774A (en) * | 2018-02-09 | 2018-08-03 | 西北大学 | A kind of drug relationship sorting technique based on multilayer convolutional neural networks |
| CN109166610A (en) * | 2018-08-30 | 2019-01-08 | 成都宇亨智能设备科技有限公司 | A kind of intelligent medical order management system and method based on prescription information |
| CN110929511A (en) * | 2018-09-04 | 2020-03-27 | 清华大学 | Intelligent matching method for personalized traditional Chinese medicine diagnosis and treatment information and traditional Chinese medicine information based on semantic similarity |
| US20200184286A1 (en) * | 2018-12-11 | 2020-06-11 | Southern Taiwan University Of Science And Technology | Smart medication identifying system |
| CN110223751A (en) * | 2019-05-16 | 2019-09-10 | 平安科技（深圳）有限公司 | Prescription evaluation method, system and computer equipment based on medical knowledge map |
| CN110390021A (en) * | 2019-06-13 | 2019-10-29 | 平安科技（深圳）有限公司 | Drug knowledge mapping construction method, device, computer equipment and storage medium |
| CN111145862A (en) * | 2019-12-30 | 2020-05-12 | 广州享药户联优选科技有限公司 | Medication safety reminding method and device |
| CN111445976A (en) * | 2020-03-24 | 2020-07-24 | 屹嘉智创(厦门)科技有限公司 | Intelligent rational medication system and method |
| CN111710388A (en) * | 2020-06-18 | 2020-09-25 | 丁玉 | Artificial intelligence medicine prescription system, medicine prescription method and artificial intelligence pharmacy |
| US20210407642A1 (en) * | 2020-06-24 | 2021-12-30 | Beijing Baidu Netcom Science And Technology Co., Ltd. | Drug recommendation method and device, electronic apparatus, and storage medium |
| CN112182252A (en) * | 2020-11-09 | 2021-01-05 | 浙江大学 | Intelligent medication question-answering method and device based on medicine knowledge graph |
| CN112951362A (en) * | 2021-02-23 | 2021-06-11 | 上海商汤智能科技有限公司 | Medicine recommendation method, device, equipment and storage medium |
| CN112820359A (en) * | 2021-02-24 | 2021-05-18 | 北京中医药大学东直门医院 | Liver injury prediction method, apparatus, device, medium, and program product |
| WO2022213690A1 (en) * | 2021-04-10 | 2022-10-13 | 浙江大学 | Medication decision support method, apparatus and device based on graphic state machine, and medium |
| CN113707261A (en) * | 2021-08-30 | 2021-11-26 | 平安国际智慧城市科技股份有限公司 | Artificial intelligence-based medicine recommendation method and device and related equipment |
| CN114724622A (en) * | 2022-03-18 | 2022-07-08 | 清华大学 | Medicine interaction prediction method and device based on medicine knowledge graph |
| CN115223731A (en) * | 2022-07-15 | 2022-10-21 | 桂林理工大学 | Drug interaction prediction method based on DS evidence theory |
| CN115862892A (en) * | 2022-12-14 | 2023-03-28 | 深圳市宝安区妇幼保健院 | Medicine education system based on mobile terminal and application method thereof |
| CN116630062A (en) * | 2023-05-10 | 2023-08-22 | 重庆医药数据信息科技有限公司 | Medical insurance fraud detection method, system and storage medium |
Non-Patent Citations (2)
| Title |
|---|
| 姜众会;孟闫燕;杨巧宁;高铸烨;: "冠心病中西药联用交互作用分析方法的思考", 世界中医药, no. 12, pages 3179 - 3181 * |
| 李绪夫;: "基于深度学习的医药专利标签分类方法", 计算机时代, no. 05, 15 May 2020 (2020-05-15), pages 50 - 53 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN117438104B (en) | 2024-03-22 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110297908B (en) | Diagnosis and treatment scheme prediction method and device | |
| CN109471895B (en) | Electronic medical record phenotype extraction and phenotype name normalization method and system | |
| CN107977361B (en) | Chinese clinical medical entity identification method based on deep semantic information representation | |
| CN110210037B (en) | Syndrome-oriented medical field category detection method | |
| US20240203599A1 (en) | Method and system of for predicting disease risk based on multimodal fusion | |
| CN106682397B (en) | Knowledge-based electronic medical record quality control method | |
| CN111834014A (en) | Medical field named entity identification method and system | |
| CN110532398B (en) | Automatic family map construction method based on multi-task joint neural network model | |
| CN111222340A (en) | Breast electronic medical record entity recognition system based on multi-standard active learning | |
| CN111881292B (en) | Text classification method and device | |
| CN110277167A (en) | The Chronic Non-Communicable Diseases Risk Forecast System of knowledge based map | |
| WO2023029502A1 (en) | Method and apparatus for constructing user portrait on the basis of inquiry session, device, and medium | |
| CN113707339B (en) | Method and system for concept alignment and content inter-translation among multi-source heterogeneous databases | |
| CN111858940A (en) | Multi-head attention-based legal case similarity calculation method and system | |
| TWI723868B (en) | Method for applying a label made after sampling to neural network training model | |
| CN116682553A (en) | Diagnosis recommendation system integrating knowledge and patient representation | |
| CN117095802B (en) | Intelligent management system and method for accompanying personnel | |
| CN115391494B (en) | Intelligent traditional Chinese medicine syndrome identification method and device | |
| CN115293161A (en) | Reasonable medicine taking system and method based on natural language processing and medicine knowledge graph | |
| CN112732872A (en) | Biomedical text-oriented multi-label classification method based on subject attention mechanism | |
| CN117542467B (en) | Automatic construction method of disease-specific standard database based on patient data | |
| CN112216379A (en) | Disease diagnosis system based on intelligent joint learning | |
| CN114359656A (en) | Melanoma image identification method based on self-supervision contrast learning and storage device | |
| CN117438104B (en) | Intelligent medicine early warning method, electronic equipment and computer storage medium | |
| Ning et al. | Research on a vehicle-mounted intelligent TCM syndrome differentiation system based on deep belief 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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |