CN112802573B - Medicine package recommendation method, device, computer system and readable storage medium - Google Patents
Medicine package recommendation method, device, computer system and readable storage medium Download PDFInfo
- Publication number
- CN112802573B CN112802573B CN202110116029.6A CN202110116029A CN112802573B CN 112802573 B CN112802573 B CN 112802573B CN 202110116029 A CN202110116029 A CN 202110116029A CN 112802573 B CN112802573 B CN 112802573B
- Authority
- CN
- China
- Prior art keywords
- drug
- characterization
- medicine
- package
- subgraph
- 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.)
- Active
Links
- 239000003814 drug Substances 0.000 title claims abstract description 350
- 229940079593 drug Drugs 0.000 title claims abstract description 185
- 238000000034 method Methods 0.000 title claims abstract description 73
- 238000012512 characterization method Methods 0.000 claims abstract description 171
- 239000013598 vector Substances 0.000 claims abstract description 119
- 206010013710 Drug interaction Diseases 0.000 claims abstract description 35
- 238000003062 neural network model Methods 0.000 claims abstract description 35
- 230000003993 interaction Effects 0.000 claims description 49
- 238000013528 artificial neural network Methods 0.000 claims description 36
- 230000006870 function Effects 0.000 claims description 28
- 238000012549 training Methods 0.000 claims description 16
- 230000000694 effects Effects 0.000 claims description 15
- 238000012545 processing Methods 0.000 claims description 15
- 238000010276 construction Methods 0.000 claims description 11
- 230000015654 memory Effects 0.000 claims description 11
- 230000005540 biological transmission Effects 0.000 claims description 10
- 230000002776 aggregation Effects 0.000 claims description 8
- 238000004220 aggregation Methods 0.000 claims description 8
- 230000008485 antagonism Effects 0.000 claims description 8
- 230000004927 fusion Effects 0.000 claims description 5
- 238000003556 assay Methods 0.000 claims description 4
- 239000011159 matrix material Substances 0.000 claims description 4
- 230000002195 synergetic effect Effects 0.000 claims description 4
- 230000009286 beneficial effect Effects 0.000 abstract description 3
- 238000010586 diagram Methods 0.000 description 11
- 238000004590 computer program Methods 0.000 description 10
- 238000004891 communication Methods 0.000 description 7
- 201000010099 disease Diseases 0.000 description 3
- 208000037265 diseases, disorders, signs and symptoms Diseases 0.000 description 3
- 238000012360 testing method Methods 0.000 description 3
- 238000004458 analytical method Methods 0.000 description 2
- 238000001914 filtration Methods 0.000 description 2
- 238000009472 formulation Methods 0.000 description 2
- 230000014509 gene expression Effects 0.000 description 2
- 239000000203 mixture Substances 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 230000003287 optical effect Effects 0.000 description 2
- 230000008569 process Effects 0.000 description 2
- 239000000758 substrate Substances 0.000 description 2
- 108010001267 Protein Subunits Proteins 0.000 description 1
- 230000002159 abnormal effect Effects 0.000 description 1
- 230000032683 aging Effects 0.000 description 1
- 230000003042 antagnostic effect Effects 0.000 description 1
- 238000013473 artificial intelligence Methods 0.000 description 1
- 239000003795 chemical substances by application Substances 0.000 description 1
- 238000007418 data mining Methods 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 238000003745 diagnosis Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 230000002452 interceptive effect Effects 0.000 description 1
- 239000004973 liquid crystal related substance Substances 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 230000001537 neural effect Effects 0.000 description 1
- 238000004806 packaging method and process Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 239000004065 semiconductor Substances 0.000 description 1
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
- G16H20/00—ICT specially adapted for therapies or health-improving plans, e.g. for handling prescriptions, for steering therapy or for monitoring patient compliance
- G16H20/10—ICT specially adapted for therapies or health-improving plans, e.g. for handling prescriptions, for steering therapy or for monitoring patient compliance relating to drugs or medications, e.g. for ensuring correct administration to patients
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- 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
- G06N3/047—Probabilistic or stochastic 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
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/049—Temporal neural networks, e.g. delay elements, oscillating neurons or pulsed inputs
-
- 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
- G16H20/00—ICT specially adapted for therapies or health-improving plans, e.g. for handling prescriptions, for steering therapy or for monitoring patient compliance
- G16H20/90—ICT specially adapted for therapies or health-improving plans, e.g. for handling prescriptions, for steering therapy or for monitoring patient compliance relating to alternative medicines, e.g. homeopathy or oriental medicines
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
- G16H50/00—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics
- G16H50/20—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for computer-aided diagnosis, e.g. based on medical expert systems
-
- 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
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
- Y02A90/00—Technologies having an indirect contribution to adaptation to climate change
- Y02A90/10—Information and communication technologies [ICT] supporting adaptation to climate change, e.g. for weather forecasting or climate simulation
Abstract
The present disclosure provides a medicine package recommendation method, device, computer system and computer readable storage medium, wherein the medicine package recommendation method includes: acquiring at least two drug characterizations and at least one drug interaction information in a drug package, and constructing a drug package subgraph; acquiring electronic medical record information and main complaint medical record information, and determining patient characterization according to a first multi-layer sensor and an LSTM model; obtaining mask vectors according to patient characterization, and updating the drug characterization in the drug package subgraph to obtain updated drug characterization; setting situation influence factors according to the updated drug characterization, and updating drug interaction information in the drug package subgraph; inputting the updated medicine package subgraph into a first neural network model to obtain subgraph characterization; splicing the sub-graph representation and the patient representation, inputting the sub-graph representation and the patient representation into a second multi-layer perceptron, and carrying out predictive scoring on the medicine package to obtain a medicine package recommendation result. The method and the device are beneficial to improving the accuracy of medicine recommendation.
Description
Technical Field
The present disclosure relates to the field of data mining, and in particular, to a medicine package recommendation method, device, computer system and computer readable storage medium based on interaction relation awareness.
Background
With the rapid growth and aging of population, the pressure of the medical system is increasing, and the intelligent diagnosis and treatment system based on the artificial intelligence technology shows great application potential, and the demand is particularly strong in the aspect of medicine recommendation. When a doctor diagnoses a patient, the doctor needs to comprehensively consider the basic information such as the sex, age, medical insurance and the like of the patient and the disease condition of the patient, and simultaneously gives a group of medicines, namely a medicine package, for the patient. The doctor needs to consider the interaction relation between medicines in the medicine package while ensuring that the medicine package has good treatment effect on the diseases of patients.
However, the prior art has two serious drawbacks. On one hand, the prior art method is generally based on traditional methods such as association rules, collaborative filtering and the like, and the modeling capability of the method for the illness state of a patient is insufficient, and the modeling for a plurality of medicines can not be performed simultaneously. On the other hand, there are wide interactions among medicines, such as synergism, antagonism, etc., which have great influence on the efficacy of medicines, and it is difficult for the prior art scheme to comprehensively analyze the interactions between the efficacy of medicines and medicines, which would certainly have influence on the accuracy of results. At present, although the modeling of the two layer factors and the medicine package recommendation problem are achieved by a certain result on the research problems of prior experience knowledge utilization, tabu detection and the like, the medicine package recommendation problem is converted into a constraint problem through the utilization of association rules, and the medicine package recommendation problem is recommended for different diseases by adopting a strategy for realizing definition, which is definitely inefficient.
Considering complex and changeable interaction between the patient's condition and the medicines in practical application, how to comprehensively model the patient's condition, the curative effect of the medicines and the interaction between the medicines so as to guide the use of clinical medicines has become an important task to be solved urgently by proposal analysis.
Disclosure of Invention
First, the technical problem to be solved
The present disclosure provides a medicine package recommendation method, device, computer system and computer readable storage medium, so as to solve the technical problems set forth above.
(II) technical scheme
According to one aspect of the present disclosure, there is provided a pharmaceutical package recommendation method including:
acquiring at least two drug characterizations and at least one drug interaction information in a drug package, and constructing a drug package subgraph; the node representation vector of the medicine package sub-graph is the medicine representation, and the edge representation vector of the medicine package sub-graph is the medicine interaction information;
acquiring electronic medical record information and main complaint medical record information, and determining patient characterization according to a first multi-layer sensor and an LSTM model;
obtaining a mask vector according to the patient representation, and updating the drug representation in the drug package subgraph to obtain an updated drug representation;
Setting situation influence factors according to the updated drug characterization, and updating the drug interaction information in the drug package subgraph to obtain an updated drug package subgraph;
inputting the updated medicine package subgraph into a first neural network model to obtain subgraph characterization;
and splicing the sub-graph representation and the patient representation, inputting the spliced sub-graph representation and the patient representation into a second multi-layer perceptron, estimating the matching degree of the sub-graph representation and the patient representation, and carrying out predictive scoring on the medicine package to obtain a medicine package recommendation result.
In some embodiments of the present disclosure, the obtaining at least two drug characterizations and at least one drug interaction information, constructing a drug steamed stuffed bun graph comprises:
acquiring a drug characterization corresponding to the drug ID according to the drug ID;
constructing the medicine package subgraph according to the medicine characterization and the medicine interaction information; the node representation vector of the medicine package sub-graph is the medicine representation, and the edge representation vector of the medicine package sub-graph is the medicine interaction information;
wherein the drug interaction information includes: unknown effects, antagonism, no effects, and synergy.
In some embodiments of the present disclosure, the obtaining electronic medical record information and complaint medical record information, determining patient characterization from the first multi-layer perceptron and the LSTM model includes:
acquiring electronic medical record information, and inputting the electronic medical record information into the first multi-layer sensor to obtain a first representation; wherein, the electronic medical record information includes: basic information and assay result information;
obtaining the medical record information of the main complaints, and inputting the LSTM model to obtain a second characterization;
and splicing the first characterization and the second characterization to obtain the patient characterization.
In some embodiments of the disclosure, the obtaining a mask vector according to the patient characterization, updating the drug characterization in the drug package sub-graph, and obtaining the updated drug characterization includes:
inputting the patient representation into a mask layer to obtain a mask vector;
and updating the drug characterization in the drug package subgraph according to the mask vector to obtain an updated drug characterization.
In some embodiments of the present disclosure, the setting a context impact factor according to the updated drug characterization, the updating the drug interaction information includes:
for edge e vu Acquiring contextual impact factors 
Wherein FNN is a feedforward neural network with a single hidden layer, a T Updating vectors for parameters with dimensions equal to the output dimensions of the feedforward neural network of the single hidden layer,
and />
Representing the updated medicine;
updating the medicine interaction information in the medicine package subgraph according to the situation influence factors to obtain updated medicine interaction information, namely updated edge weight
In some embodiments of the present disclosure, the training method of the first neural network model includes:
initializing a medicine subgraph, and initializing the corresponding characterization of the node characterization vector into an original medicine characterization; the medicine interaction information of two medicines is unknown or has no effect, the edge weight is initialized to the probability of the co-occurrence of the two medicines, and the edge weight e is initialized vu ＝p vu, wherein ,pvu Is the probability of co-occurrence of two drugs; the medicine interaction information of the two medicines is synergistic, and the edge weight e is initialized vu =1; and initializing the side weight e by taking the medicine interaction information of the two medicines as antagonism vu One or more of = -1;
updating the drug characterization in the drug package subgraph according to the mask vector to obtain an updated drug characterization 
The method comprises the following steps:
wherein FNN is a single hidden layer feedforward neural network, σ (MLP (u)) is a mask vector extracted by a mask layer, σ is a sigmoid function, and d is multiplication by element u Characterizing the original medicine;
setting situation influence factors according to the updated drug characterization, and updating the drug interaction information to obtain updated edge weights;
constructing the first neural network model as follows:
wherein ,
information vector transferred from node v to node u for layer i; w (W) 1 (l-1) Parameters to be learned are the model of the layer 1; />
Vector characterization at layer 1 for node v; />
Information vectors obtained by aggregation of the node u at the first layer; n (u) is a neighbor node of the node u; the GRU is a gating circulation unit; />
The updated edge weight; />
Information vector transferred from node v to node u in the first layer; />
A characterization vector of the node u at the first layer-1; />
Representing the vector of the node u at the first layer; FNN is a feedforward neural network with a single hidden layer; w (W) 0 (l-1) Parameters to be learned are the model of the layer 1; w and M are parameters to be learned;
inputting the updated medicine package subgraph into the first neural network model for information transmission, extracting node characterization of a layer of graph neural network, and obtaining subgraph characterization:
g＝∑ v∈V σ(FNN([d v ||h v ]))⊙(FNN([d v ||h v ]))；
V is the total node set in the updated medicine package subgraph, g is the subgraph representation; FNN ([ d) v ||h v ]) The method comprises the steps that fusion characterization is obtained by a node initial characterization and a characterization after information transmission through a feedforward neural network of a single hidden layer;
training the first neural network model using the following loss function:
wherein ,ui Characterization of patient corresponding to the ith patient, g i Representing a sub-graph corresponding to the ith medicine package; MLP is the second multilayer perceptron, MLP ([ u ] i ||g i ]) Scoring the model for patient i and drug package i;
regularizing the term for L2.
In some embodiments of the present disclosure, the training method of the first neural network model includes:
initializing a medicine subgraph, and initializing the corresponding characterization of the node characterization vector into an original medicine characterization; initializing the corresponding token of the edge token vector to e vu ＝FNN([d v ||d u ])；
wherein ,dv and du For drug characterization corresponding to different nodes, FNN is a feedforward neural network with a single hidden layer;
updating the medicine interaction information in the medicine package subgraph according to the mask vector to obtain updated medicine interaction information, namely an updated edge characterization vector 
Wherein, sigma is a sigmoid function, while, by multiplying element by element, FNN is a feedforward neural network of a single hidden layer;
constructing the first neural network model as follows:
wherein ,
information vector transferred from node v to node u for layer i; w (W) 1 (l-1) Parameters to be learned are the model of the layer 1; />
Representing vectors for the edges of layer 1; />
Information vectors obtained by aggregation of the node u at the first layer; n (u) is a neighbor node of the node u; />
Information vector transferred from node v to node u in the first layer; />
Representing the vector of the node u at the first layer-1; />
The characterization vector of the node u at the first layer is obtained; FNN is a feedforward neural network with a single hidden layer; w (W) 0 (l-1) Parameters to be learned are the model of the layer 1; w is a parameter to be learned;
inputting the updated medicine package subgraph into the first neural network model for information transmission, extracting node characterization of a layer of graph neural network, and obtaining subgraph characterization:
g＝∑ v∈V σ(FNN([d v ||h v ]))⊙(FNN([d v ||h v ]))；
v is the total node set in the updated medicine package subgraph, g is the subgraph representation; FNN ([ d) v ||h v ]) The method comprises the steps that fusion characterization is obtained by a node initial characterization and a characterization after information transmission through a feedforward neural network of a single hidden layer;
training the first neural network model using the following loss function:
Wherein MLP is the second multilayer perceptron, MLP ([ u ] i ||g i ]) Scoring the model for patient i and drug package i;
cross entropy loss function for performing edge classification tasks based on edge characterization;
regularizing the term for L2; e, e vu To characterize vectors for edges, R uv Drug interaction information corresponding to the edge representation vector; the matrix Q converts the edge token vector into a classified probability vector.
According to one aspect of the present disclosure, there is provided a medicine package recommending apparatus including:
the sub-graph construction module is used for acquiring at least two drug characterizations and at least one drug interaction information in the drug package and constructing a drug package sub-graph; the node representation vector of the medicine package sub-graph is the medicine representation, and the edge representation vector of the medicine package sub-graph is the medicine interaction information;
the acquisition module is used for acquiring the electronic medical record information and the complaint medical record information and determining patient characterization according to the first multi-layer perceptron and the LSTM model;
the first updating module is used for acquiring mask vectors according to the patient characterization, updating the drug characterization in the drug package subgraph and obtaining updated drug characterization;
the second updating module is used for setting situation influence factors according to the updated drug characterization, updating the drug interaction information in the drug package subgraph, and obtaining an updated drug package subgraph;
The processing module is used for inputting the updated medicine package subgraph into a first neural network model to obtain subgraph characterization;
and the recommending module is used for splicing the sub-graph representation with the patient representation, inputting the sub-graph representation into a second multi-layer perceptron, and carrying out predictive scoring on the medicine package to obtain a medicine package recommending result.
According to one aspect of the present disclosure, there is provided a computer system comprising:
one or more processors;
a memory for storing one or more programs;
wherein the one or more programs, when executed by the one or more processors, cause the one or more processors to implement the methods as described above.
According to one aspect of the present disclosure, there is provided a computer readable storage medium having stored thereon executable instructions which, when executed by a processor, cause the processor to implement a method as described above.
(III) beneficial effects
As can be seen from the above technical solutions, the method, apparatus, computer system and computer-readable storage medium for recommending pharmaceutical packages according to the present disclosure have at least one or a part of the following advantages:
(1) By introducing the interaction relation of the medicines and the influence of the patient conditions on the interaction effect of the medicines, the influence of the medicine interaction in the actual medical scene can be reflected more truly, and the accuracy of medicine recommendation is improved.
(2) The method and the device can perform complete interpretability analysis on various complex factors by using various methods including a context awareness mechanism, so that the effect of achieving two purposes is achieved.
Drawings
FIG. 1 schematically illustrates an exemplary system architecture to which the pharmaceutical package recommendation methods and apparatus of the present disclosure may be applied.
Fig. 2 schematically illustrates a schematic diagram of a pharmaceutical package recommendation method according to an embodiment of the present disclosure.
Fig. 3 is a schematic diagram of a method for recommending pharmaceutical packages according to an embodiment of the disclosure.
Fig. 4 schematically illustrates a schematic diagram of a pharmaceutical package recommendation apparatus according to an embodiment of the present disclosure.
Fig. 5 schematically illustrates a block diagram of a computer system suitable for implementing the above-described method according to an embodiment of the present disclosure.
Detailed Description
Hereinafter, embodiments of the present disclosure will be described with reference to the accompanying drawings. It should be understood that the description is only exemplary and is not intended to limit the scope of the present disclosure. In the following detailed description, for purposes of explanation, numerous specific details are set forth in order to provide a thorough understanding of the embodiments of the present disclosure. It may be evident, however, that one or more embodiments may be practiced without these specific details. In addition, in the following description, descriptions of well-known structures and techniques are omitted so as not to unnecessarily obscure the concepts of the present disclosure.
The terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the disclosure. The terms "comprises," "comprising," and/or the like, as used herein, specify the presence of stated features, steps, operations, and/or components, but do not preclude the presence or addition of one or more other features, steps, operations, or components.
All terms (including technical and scientific terms) used herein have the same meaning as commonly understood by one of ordinary skill in the art unless otherwise defined. It should be noted that the terms used herein should be construed to have meanings consistent with the context of the present specification and should not be construed in an idealized or overly formal manner.
Where expressions like at least one of "A, B and C, etc. are used, the expressions should generally be interpreted in accordance with the meaning as commonly understood by those skilled in the art (e.g.," a system having at least one of A, B and C "shall include, but not be limited to, a system having a alone, B alone, C alone, a and B together, a and C together, B and C together, and/or A, B, C together, etc.). Where a formulation similar to at least one of "A, B or C, etc." is used, in general such a formulation should be interpreted in accordance with the ordinary understanding of one skilled in the art (e.g. "a system with at least one of A, B or C" would include but not be limited to systems with a alone, B alone, C alone, a and B together, a and C together, B and C together, and/or A, B, C together, etc.).
The present disclosure provides a medicine package recommendation method, device, computer system and computer readable storage medium, wherein the medicine package recommendation method includes: acquiring at least two drug characterizations and at least one drug interaction information in a drug package, and constructing a drug package subgraph; acquiring electronic medical record information and main complaint medical record information, and determining patient characterization according to a first multi-layer sensor and an LSTM model; obtaining mask vectors according to patient characterization, and updating the drug characterization in the drug package subgraph to obtain updated drug characterization; setting situation influence factors according to the updated drug characterization, and updating drug interaction information in the drug package subgraph; inputting the updated medicine package subgraph into a first neural network model to obtain subgraph characterization; splicing the sub-graph representation and the patient representation, inputting the sub-graph representation and the patient representation into a second multi-layer perceptron, and carrying out predictive scoring on the medicine package to obtain a medicine package recommendation result. The method and the device are beneficial to improving the accuracy of medicine recommendation.
Fig. 1 schematically illustrates an exemplary system architecture 100 to which the pharmaceutical package recommendation methods and apparatus of the present disclosure may be applied. It should be noted that fig. 1 is only an example of a system architecture to which embodiments of the present disclosure may be applied to assist those skilled in the art in understanding the technical content of the present disclosure, but does not mean that embodiments of the present disclosure may not be used in other devices, systems, environments, or scenarios.
As shown in fig. 1, a system architecture 100 according to this embodiment may include terminal devices 101, 102, 103, a network 104, and a server 105. The network 104 is used as a medium to provide communication links between the terminal devices 101, 102, 103 and the server 105. The network 104 may include various connection types, such as wired and/or wireless communication links, and the like.
The user may interact with the server 105 via the network 104 using the terminal devices 101, 102, 103 to receive or send messages or the like. Various communication client applications may be installed on the terminal devices 101, 102, 103, such as shopping class applications, web browser applications, search class applications, instant messaging tools, mailbox clients and/or social platform software, to name a few.
The terminal devices 101, 102, 103 may be a variety of electronic devices having a display screen and supporting web browsing, including but not limited to smartphones, tablets, laptop and desktop computers, and the like.
The server 105 may be a server providing various services, such as a background management server (by way of example only) providing support for websites browsed by users using the terminal devices 101, 102, 103. The background management server may analyze and process the received data such as the user request, and feed back the processing result (e.g., the web page, information, or data obtained or generated according to the user request) to the terminal device.
It should be noted that, the medicine package recommendation method provided in the embodiments of the present disclosure may be generally performed by the server 105. Accordingly, the drug package recommendation device provided in the embodiments of the present disclosure may be generally provided in the server 105. The pharmaceutical package recommendation method provided by the embodiments of the present disclosure may also be performed by a server or a server cluster that is different from the server 105 and is capable of communicating with the terminal devices 101, 102, 103 and/or the server 105. Accordingly, the drug package recommendation apparatus provided by the embodiments of the present disclosure may also be provided in a server or a server cluster that is different from the server 105 and is capable of communicating with the terminal devices 101, 102, 103 and/or the server 105. Alternatively, the medicine package recommendation method provided by the embodiment of the present disclosure may be performed by the terminal device 101, 102, or 103, or may be performed by another terminal device different from the terminal device 101, 102, or 103. Accordingly, the medicine package recommending apparatus provided by the embodiment of the present disclosure may also be provided in the terminal device 101, 102, or 103, or in another terminal device different from the terminal device 101, 102, or 103.
For example, the medicine package information to be processed may be originally stored in any one of the terminal apparatuses 101, 102, or 103 (for example, but not limited to, the terminal apparatus 101), or stored on an external storage apparatus and may be imported into the terminal apparatus 101. Then, the terminal device 101 may locally perform the medicine package recommendation method provided by the embodiment of the present disclosure, or send the to-be-processed medicine package information to other terminal devices, servers, or server clusters, and perform the medicine package recommendation method provided by the embodiment of the present disclosure by the other terminal devices, servers, or server clusters that receive the to-be-processed medicine package information.
It should be understood that the number of terminal devices, networks and servers in fig. 1 is merely illustrative. There may be any number of terminal devices, networks, and servers, as desired for implementation.
Fig. 2 schematically illustrates a schematic diagram of a pharmaceutical package recommendation method according to an embodiment of the present disclosure.
As shown in fig. 2, the method includes operations S201 to S206.
In operation S201, at least two drug characterizations and at least one drug interaction information in a drug package are acquired, and a drug package sub-graph is constructed; the node representation vector of the medicine package sub-graph is medicine representation, and the edge representation vector of the medicine package sub-graph is medicine interaction information.
The following slave operations are further described in operation S201:
and acquiring the drug characterization corresponding to the drug ID according to the drug ID.
And constructing a medicine package subgraph according to the medicine characterization and medicine interaction information. Drug interaction information in particular embodiments of the present disclosure includes: unknown effects, antagonism, no effects and synergy. Constructing a drug interaction matrix R, wherein R ij Representing the type of interaction of drug i with drug j. R is R ij -2, indicating that the type of interaction of drug i with drug j is unknown. R is R ij -1, indicating that the type of interaction of drug i with drug j is antagonistic. R is R ij =0, indicating that the type of interaction of drug i with drug j is inactive. R is R ij =1, indicating that the type of interaction of drug i with drug j is synergistic.
In operation S202, electronic medical record information and complaint medical record information are acquired, and patient characterization is determined based on the first multi-layer perceptron and the LSTM model.
The operation S202 specifically includes the following operations:
and acquiring electronic medical record information, and inputting the electronic medical record information into a first multi-layer sensor to obtain a first representation. Acquiring electronic medical record information, and inputting the electronic medical record information into a first multi-layer sensor to obtain a first representation; wherein, the electronic medical record information includes: basic information and assay result information. Specifically, the electronic medical record information is converted into a 0/1 vector W by using a single thermal coding, and a first representation is obtained by using a first multi-layer sensor:
m w ＝MLP(W)；
wherein ,mw And for the first representation corresponding to the electronic medical record information, the MLP is a first multi-layer sensor for feature extraction. The multi-layer perceptron (MLP, multilayer Perceptron), which is a feedforward artificial neural network model known to those skilled in the art, maps multiple data sets of an input onto a single data set of an output by the multi-layer perceptron, and is not further developed herein.
In embodiments of the present disclosure, the electronic medical record information is structured patient base information and test result information. The structural patient basic information is segmented, for example, the age of the patient is divided into five segments of infants, teenagers, young, middle-aged and elderly people. The results obtained by the patient test on the test result information are compared with the standard range of normal people, and abnormal items in the results are extracted and classified into two types of higher and lower.
And acquiring the medical record information of the main complaints, and inputting the medical record information into an LSTM model to obtain a second characterization. Specifically, for each character in the complaint medical record information, setting a word embedding vector corresponding to the character, constructing a word vector sequence, and extracting features from the complaint medical record information by using the following char-LSTM model:
h q ＝LSTM(T)；
wherein T is a word vector sequence of the complaint case information.
Output h of final time step q And define the finalDeriving a splice of the patient characterization as two part characterization, i.e. u= [ m w ||h q ]Where u is the final patient characterization.
And splicing the first characterization and the second characterization to obtain the patient characterization.
In connection with the illustration of fig. 3, the patient characterization and the drug characterization are pre-trained in this embodiment from a one-to-one interactive relationship level. The method specifically comprises the following steps:
An embedded vector d corresponding to each medicine is set, and training is carried out by using a neural collaborative filtering method and a Bayesian ordering loss function. Specifically, the degree of matching of drug j to patient i is estimated as follows:
wherein ,ui Representing vectors for patient corresponding to patient i, d j For the drug characterization vector corresponding to drug j, MLP is the third multi-layer perceptron that predicts the matching of patient to drug. The following Bayesian ordering penalty function is then used:
wherein ,Pi The set of drugs used for patient i, N is the number of all patients, and l is the number of patients not present at P i All of which are in the same formula.
This loss function represents that the used drug should score higher than the unused drug and use L2 regularization to prevent the occurrence of the overfitting phenomenon, where Θ is the model parameter and λ is the regularization coefficient.
In operation S203, a mask vector is obtained according to the patient characterization, and the drug characterization in the drug package sub-graph is updated to obtain an updated drug characterization.
In operation S204, according to the updated drug characterization, a situation influencing factor is set, and drug interaction information in the drug package sub-graph is updated, so as to obtain an updated drug package sub-graph. In one embodiment of the present disclosure, specifically including:
For edge e vu Acquiring contextual impact factors
Wherein FNN is a feedforward neural network with a single hidden layer, a T Updating vectors for parameters with dimensions equal to the output dimensions of the feedforward neural network of the single hidden layer,
and />
Representing the updated medicine;
updating the medicine interaction information in the medicine package subgraph according to the situation influence factors to obtain updated medicine interaction information, namely updated edge weight
In operation S205, the updated drug package subgraph is input to the first neural network model to obtain the subgraph characterization.
The construction of the drug package subgraph and the first neural network model will be specifically described.
For a medicine package P containing several medicines, a medicine package sub-graph g= { V, E }, where V is a node set and E is an edge set, is defined in this embodiment. Each node V epsilon V has its corresponding drug representation d, each directed edge e vu E also has its corresponding drug interaction information, and the specific correspondence will be explained in more detail later. The topology of the medicine package sub-graph G, i.e. the presence or absence of edges, is defined as follows: for node v and node u, if the corresponding medicine has been marked, edge e vu Exists. If the two drugs are not labeled but co-occur at a frequency p vu Beyond a certain threshold, edge e vu Exists.
For the constructed medicine subgraph, a first neural network model structure to be used is defined as follows:
wherein ,
for the information vector transferred from node v to node u, -/->
and />
Is the node hidden vector of the layer 1, e vu N (u) is the neighbor node set of the node u, which is the feature vector corresponding to the edge, ++>
Information vector delivered at layer l for node u,/->
The MESSAGE is an information generating function, the AGGREGATION is an information AGGREGATION function, and the UPDATE is an information updating function.
The following two training methods of the first neural network model are provided as two exemplary embodiments, respectively.
In a first exemplary embodiment, a method of training a first neural network model includes:
initializing the medicine subgraph, and initializing the corresponding characterization of the node characterization vector to be the original medicine characterization; the medicine interaction information of two medicines is unknown or has no effect, the edge weight is initialized to the probability of the co-occurrence of the two medicines, and the edge weight e is initialized vu ＝p vu, wherein ,pvu Is the probability of co-occurrence of two drugs; the medicine interaction information of the two medicines is synergistic, and the edge weight e is initialized vu =1; and the medicine interaction information of the two medicines is antagonism, and initializing the side weight e vu One or more of = -1.
Updating the medicine characterization in the medicine package subgraph according to the mask vector to obtain updated medicine characterization
The method comprises the following steps:
wherein FNN is a single hidden layer feedforward neural network, σ (MLP (u)) is a mask vector extracted by a mask layer, σ is a sigmoid function, and d is multiplication by element u Is characterized by original medicines.
And setting situation influence factors according to the updated drug characterization, and updating drug interaction information to obtain updated edge weights. The specific method is consistent with the foregoing, and no further description is given here.
Constructing a first neural network model as follows:
wherein ,
information vector transferred from node v to node u for layer i; w (W) 1 (l-1) Parameters to be learned are the model of the layer 1; />
Vector characterization at layer 1 for node v; />
Information vectors obtained by aggregation of the node u at the first layer; n (u) is a neighbor node of the node u; the GRU is a gating circulation unit; />
The updated edge weight; />
Information vector transferred from node v to node u in the first layer; />
A characterization vector of the node u at the first layer-1; />
Representing the vector of the node u at the first layer; FNN is a feedforward neural network with a single hidden layer; w (W) 0 (l-1 ) Parameters to be learned are the model of the layer 1; w and M are parameters to be learned;
inputting the updated medicine package subgraph into a first neural network model for information propagation, extracting node characterization of a layer of graph neural network, and obtaining subgraph characterization:
g＝∑ v∈V σ(FNN([d v ||h v ]))⊙(FNN([d v ||h v ]))；
wherein V is all sections in the updated medicine package subgraphA point set, g is a subgraph representation; FNN ([ d) v ||h v ]) The method comprises the steps that fusion characterization is obtained by a node initial characterization and a characterization after information transmission through a feedforward neural network of a single hidden layer;
the first neural network model is trained using the following loss function:
wherein ,ui Characterization of patient corresponding to the ith patient, g i Representing a sub-graph corresponding to the ith medicine package; MLP is the second multilayer perceptron; MLP ([ u) i ||g i ]) Scoring the model for patient i and drug package i;
regularizing the term for L2.
In operation S206, the sub-graph representation and the patient representation are spliced and input to the second multi-layer perceptron, and the medicine package is predicted and scored to obtain a medicine package recommendation result. The second multi-layer sensor scoring the drug package is a value whose size reflects how well the patient is matched to the drug package.
In a second exemplary embodiment, a method of training a first neural network model includes:
Initializing the medicine subgraph, and initializing the corresponding characterization of the node characterization vector to be the original medicine characterization; initializing the corresponding token of the edge token vector to e vu ＝FNN([d v ||d u ])；
wherein ,dv and du For drug characterization corresponding to different nodes, FNN is a single hidden layer feedforward neural network.
Updating the medicine interaction information in the medicine package subgraph according to the mask vector to obtain updated medicine interaction information, namely updated edge characterization vector
Wherein the method comprises the steps ofSigma is a sigmoid function, while by element multiplication, FNN is a feed-forward neural network of a single hidden layer.
Constructing a first neural network model as follows:
wherein ,
information vector transferred from node v to node u for layer i; w (W) 1 (l-1) Parameters to be learned are the model of the layer 1; />
Representing vectors for the edges of layer 1; />
Information vectors obtained by aggregation of the node u at the first layer; n (u) is a neighbor node of the node u; />
Information vector transferred from node v to node u in the first layer; />
Representing the vector of the node u at the first layer-1; />
The characterization vector of the node u at the first layer is obtained; FNN is a feedforward neural network with a single hidden layer; w (W) 0 (l-1) To be used for the layer 1 modelLearning parameters; w is a parameter to be learned;
inputting the updated medicine package subgraph into a first neural network model for information propagation, extracting node characterization of a layer of graph neural network, and obtaining subgraph characterization:
g＝∑ v∈V σ(FNN([d v ||h v ]))⊙(FNN([d v ||h v ]))；
V is the total node set in the updated medicine package subgraph, g is the subgraph representation; FNN ([ d) v ||h v ]) The method comprises the steps that fusion characterization is obtained by a node initial characterization and a characterization after information transmission through a feedforward neural network of a single hidden layer;
the first neural network model is trained using the following loss function:
wherein MLP is the second multi-layer sensor, MLP ([ u ] i ||g i ]) Scoring the model for patient i and drug package i;
cross entropy loss function for performing edge classification tasks based on edge characterization; />
Regularizing the term for L2; e, e vu To characterize vectors for edges, R uv Drug interaction information corresponding to the edge representation vector; the matrix Q converts the edge token vector into a classified probability vector.
The edge attribute has different forms in the two model variants provided by the first exemplary embodiment, in which the edge attribute is a numerical value, and the second exemplary embodiment, in which the edge attribute is an edge token vector. In the DPR-AG provided in the second exemplary embodiment, the model loss function adds a classification task that classifies based on edge characterization, where the probability vector obtained is the edge classification task, and does not appear elsewhere in its entirety.
Fig. 4 schematically illustrates a schematic diagram of a pharmaceutical package recommendation apparatus according to an embodiment of the present disclosure.
As shown in fig. 4, the medicine package recommending apparatus 400 includes: the sub-graph construction module 410, the acquisition module 420, the first update module 430, the second update module 440, the processing module 450, and the recommendation module 460.
A sub-graph construction module 410, configured to acquire at least two drug characterizations and at least one drug interaction information in a drug package, and construct a drug package sub-graph; the node representation vector of the medicine package sub-graph is medicine representation, and the edge representation vector of the medicine package sub-graph is medicine interaction information. The sub-graph construction module 410 includes: a first acquisition unit and a first construction unit.
And the first acquisition unit is used for acquiring the drug characterization corresponding to the drug ID according to the drug ID.
And the first construction unit is used for constructing a medicine package sub-graph according to the medicine characterization and the medicine interaction information. Wherein the drug interaction information includes: unknown effects, antagonism, no effects, and synergy.
The obtaining module 420 is configured to obtain electronic medical record information and complaint medical record information, and determine a patient characterization according to the first multi-layer sensor and the LSTM model. The acquisition module 420 includes: the device comprises a second acquisition unit, a third acquisition unit and a first processing unit.
The second acquisition unit is used for acquiring electronic medical record information and inputting the electronic medical record information into the first multilayer perceptron to obtain a first representation; wherein, the electronic medical record information includes: basic information and assay result information.
And the third acquisition unit is used for acquiring the complaint medical record information and inputting the complaint medical record information into the LSTM model to obtain the second characterization.
And the first processing unit is used for splicing the first characterization and the second characterization to obtain the patient characterization.
The first updating module 430 is configured to obtain a mask vector according to the patient characterization, update the drug characterization in the drug package sub-graph, and obtain an updated drug characterization. The first update module 430 includes:
a second processing unit and a first updating unit.
And the second processing unit is used for inputting the patient representation into the mask layer to obtain mask vectors.
And the first updating unit is used for updating the drug characterization in the drug package subgraph according to the mask vector to obtain the updated drug characterization.
The second updating module 440 is configured to set a situation influencing factor according to the updated drug characterization, update drug interaction information in the drug package sub-graph, and obtain an updated drug package sub-graph. The second update module 440 includes: a fourth acquisition unit and a second updating unit.
A fourth acquisition unit for edge e vu Acquiring contextual impact factors
wherein ,aT Updating the vector for parameters having dimensions equal to the output dimensions of the multi-layer perceptron model,
and />
Is characterized by updated medicines.
A second updating unit, configured to update the drug interaction information in the drug package subgraph according to the situation influencing factor to obtain updated drug interaction information, i.e. updated edge weight
And the processing module 450 is used for inputting the updated medicine package subgraph into the first neural network model to obtain the subgraph representation.
And the recommendation module 460 is used for splicing the sub-graph representation and the patient representation, inputting the sub-graph representation and the patient representation into the second multi-layer perceptron, and carrying out predictive scoring on the medicine package to obtain a medicine package recommendation result.
Any number of modules, sub-modules, units, sub-units, or at least some of the functionality of any number of the sub-units according to embodiments of the present disclosure may be implemented in one module. Any one or more of the modules, sub-modules, units, sub-units according to embodiments of the present disclosure may be implemented as split into multiple modules. Any one or more of the modules, sub-modules, units, sub-units according to embodiments of the present disclosure may be implemented at least in part as a hardware circuit, such as a Field Programmable Gate Array (FPGA), a Programmable Logic Array (PLA), a system-on-chip, a system-on-substrate, a system-on-package, an Application Specific Integrated Circuit (ASIC), or in any other reasonable manner of hardware or firmware that integrates or encapsulates the circuit, or in any one of or a suitable combination of three of software, hardware, and firmware. Alternatively, one or more of the modules, sub-modules, units, sub-units according to embodiments of the present disclosure may be at least partially implemented as computer program modules, which when executed, may perform the corresponding functions.
For example, any of the sub-graph construction module 410, the acquisition module 420, the first update module 430, the second update module 440, the processing module 450, and the recommendation module 460 may be combined in one module/unit/sub-unit or any of the modules/units/sub-units may be split into a plurality of modules/units/sub-units. Alternatively, at least some of the functionality of one or more of these modules/units/sub-units may be combined with at least some of the functionality of other modules/units/sub-units and implemented in one module/unit/sub-unit. According to embodiments of the present disclosure, at least one of the sub-graph construction module 410, the acquisition module 420, the first update module 430, the second update module 440, the processing module 450, and the recommendation module 460 may be implemented at least in part as hardware circuitry, such as a Field Programmable Gate Array (FPGA), a Programmable Logic Array (PLA), a system-on-chip, a system-on-substrate, a system-on-package, an Application Specific Integrated Circuit (ASIC), or as hardware or firmware in any other reasonable manner of integrating or packaging the circuitry, or as any one of or a suitable combination of any of the three. Alternatively, at least one of the sub-graph construction module 410, the acquisition module 420, the first update module 430, the second update module 440, the processing module 450, and the recommendation module 460 may be at least partially implemented as a computer program module, which when executed may perform the corresponding functions.
It should be noted that, in the embodiment of the present disclosure, the training device portion of the intent recognition model corresponds to the training method portion of the intent recognition model in the embodiment of the present disclosure, and the description of the training device portion of the intent recognition model specifically refers to the training method portion of the intent recognition model, which is not described herein.
Accordingly, the intent recognition device portion in the embodiments of the present disclosure corresponds to the intent recognition method portion in the embodiments of the present disclosure, and the description of the intent recognition device portion specifically refers to the intent recognition method portion and is not described herein.
Fig. 5 schematically illustrates a block diagram of a computer system suitable for implementing the above-described method according to an embodiment of the present disclosure. The computer system illustrated in fig. 5 is merely an example, and should not be construed as limiting the functionality and scope of use of embodiments of the present disclosure.
As shown in fig. 5, a computer system 500 according to an embodiment of the present disclosure includes a processor 501, which can perform various appropriate actions and processes according to a program stored in a Read Only Memory (ROM) 502 or a program loaded from a storage section 508 into a Random Access Memory (RAM) 503. The processor 501 may include, for example, a general purpose microprocessor (e.g., a CPU), an instruction set processor and/or an associated chipset and/or a special purpose microprocessor (e.g., an Application Specific Integrated Circuit (ASIC)), or the like. The processor 501 may also include on-board memory for caching purposes. The processor 501 may comprise a single processing unit or a plurality of processing units for performing different actions of the method flows according to embodiments of the disclosure.
In the RAM 503, various programs and data required for the operation of the system 500 are stored. The processor 501, ROM 502, and RAM 503 are connected to each other by a bus 504. The processor 501 performs various operations of the method flow according to the embodiments of the present disclosure by executing programs in the ROM 502 and/or the RAM 503. Note that the program may be stored in one or more memories other than the ROM 502 and the RAM 503. The processor 501 may also perform various operations of the method flow according to embodiments of the present disclosure by executing programs stored in the one or more memories.
According to an embodiment of the present disclosure, the system 500 may further include an input/output (I/O) interface 505, the input/output (I/O) interface 505 also being connected to the bus 504. The system 500 may also include one or more of the following components connected to the I/O interface 505: an input section 506 including a keyboard, a mouse, and the like; an output portion 507 including a Cathode Ray Tube (CRT), a Liquid Crystal Display (LCD), and the like, and a speaker, and the like; a storage portion 508 including a hard disk and the like; and a communication section 509 including a network interface card such as a LAN card, a modem, or the like. The communication section 509 performs communication processing via a network such as the internet. The drive 510 is also connected to the I/O interface 505 as needed. A removable medium 511 such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, or the like is mounted on the drive 510 as needed so that a computer program read therefrom is mounted into the storage section 508 as needed.
According to embodiments of the present disclosure, the method flow according to embodiments of the present disclosure may be implemented as a computer software program. For example, embodiments of the present disclosure include a computer program product comprising a computer program embodied on a computer readable storage medium, the computer program comprising program code for performing the method shown in the flowcharts. In such an embodiment, the computer program may be downloaded and installed from a network via the communication portion 509, and/or installed from the removable media 511. The above-described functions defined in the system of the embodiments of the present disclosure are performed when the computer program is executed by the processor 501. The systems, devices, apparatus, modules, units, etc. described above may be implemented by computer program modules according to embodiments of the disclosure.
The present disclosure also provides a computer-readable storage medium that may be embodied in the apparatus/device/system described in the above embodiments; or may exist alone without being assembled into the apparatus/device/system. The computer-readable storage medium carries one or more programs which, when executed, implement methods in accordance with embodiments of the present disclosure.
According to embodiments of the present disclosure, the computer-readable storage medium may be a non-volatile computer-readable storage medium. Examples may include, but are not limited to: a portable computer diskette, a hard disk, a Random Access Memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing. In the context of this disclosure, a computer-readable storage medium may be any tangible medium that can contain, or store a program for use by or in connection with an instruction execution system, apparatus, or device.
For example, according to embodiments of the present disclosure, the computer-readable storage medium may include ROM 502 and/or RAM 503 and/or one or more memories other than ROM 502 and RAM 503 described above.
The flowcharts and block diagrams in the figures illustrate the architecture, functionality, and operation of possible implementations of systems, methods and computer program products according to various embodiments of the present disclosure. In this regard, each block in the flowchart or block diagrams may represent a module, segment, or portion of code, which comprises one or more executable instructions for implementing the specified logical function(s). It should also be noted that in some alternative implementations, the functions noted in the block may occur out of the order noted in the figures. For example, two blocks shown in succession may, in fact, be executed substantially concurrently, or the blocks may sometimes be executed in the reverse order, depending upon the functionality involved. It will also be noted that each block of the block diagrams or flowchart illustration, and combinations of blocks in the block diagrams or flowchart illustration, can be implemented by special purpose hardware-based systems which perform the specified functions or acts, or combinations of special purpose hardware and computer instructions. Those skilled in the art will appreciate that the features recited in the various embodiments of the disclosure and/or in the claims may be combined in various combinations and/or combinations, even if such combinations or combinations are not explicitly recited in the disclosure. In particular, the features recited in the various embodiments of the present disclosure and/or the claims may be variously combined and/or combined without departing from the spirit and teachings of the present disclosure. All such combinations and/or combinations fall within the scope of the present disclosure.
The embodiments of the present disclosure are described above. However, these examples are for illustrative purposes only and are not intended to limit the scope of the present disclosure. Although the embodiments are described above separately, this does not mean that the measures in the embodiments cannot be used advantageously in combination. The scope of the disclosure is defined by the appended claims and equivalents thereof. Various alternatives and modifications can be made by those skilled in the art without departing from the scope of the disclosure, and such alternatives and modifications are intended to fall within the scope of the disclosure.
Claims (9)
1. A pharmaceutical package recommendation method comprising:
acquiring at least two drug characterizations and at least one drug interaction information in a drug package, and constructing a drug package subgraph; the node representation vector of the medicine package sub-graph is the medicine representation, and the edge representation vector of the medicine package sub-graph is the medicine interaction information;
acquiring electronic medical record information and main complaint medical record information, and determining patient characterization according to a first multi-layer sensor and an LSTM model;
obtaining a mask vector according to the patient representation, and updating the drug representation in the drug package subgraph to obtain an updated drug representation;
Setting situation influence factors according to the updated drug characterization, and updating the drug interaction information in the drug package subgraph to obtain an updated drug package subgraph;
inputting the updated medicine package subgraph into a first neural network model to obtain subgraph characterization;
splicing the sub-graph representation and the patient representation, inputting the spliced sub-graph representation and the patient representation into a second multi-layer perceptron, estimating the matching degree of the sub-graph representation and the patient representation, and carrying out predictive scoring on the medicine package to obtain a medicine package recommendation result;
wherein, according to the updated drug characterization, setting a situation influencing factor, updating the drug interaction information includes:
for edge e vu Acquiring contextual impact factors
Wherein FNN is a feedforward neural network with a single hidden layer, a T Updating vectors for parameters with dimensions equal to the output dimensions of the feedforward neural network of the single hidden layer,
and />
Representing the updated medicine;
updating the medicine interaction information in the medicine package subgraph according to the situation influence factors to obtain updated medicine interaction information, namely updated edge weight 
2. The method of claim 1, wherein the obtaining at least two drug characterizations and at least one drug interaction information, constructing a drug steamed stuffed bun graph comprises:
acquiring a drug characterization corresponding to the drug ID according to the drug ID;
constructing the medicine package subgraph according to the medicine characterization and the medicine interaction information; the node representation vector of the medicine package sub-graph is the medicine representation, and the edge representation vector of the medicine package sub-graph is the medicine interaction information;
wherein the drug interaction information includes: unknown effects, antagonism, no effects, and synergy.
3. The medication package recommendation method of claim 1, wherein the obtaining electronic medical record information and complaint medical record information, determining patient characterization from a first multi-layer perceptron and LSTM model comprises:
acquiring electronic medical record information, and inputting the electronic medical record information into the first multi-layer sensor to obtain a first representation; wherein, the electronic medical record information includes: basic information and assay result information;
obtaining the medical record information of the main complaints, and inputting the LSTM model to obtain a second characterization;
and splicing the first characterization and the second characterization to obtain the patient characterization.
4. The drug package recommendation method of claim 1, wherein the obtaining a mask vector from the patient representation, updating the drug representation in the drug package sub-graph, obtaining an updated drug representation comprises:
inputting the patient representation into a mask layer to obtain a mask vector;
and updating the drug characterization in the drug package subgraph according to the mask vector to obtain an updated drug characterization.
5. The pharmaceutical package recommendation method of any one of claims 1 to 4, wherein the training method of the first neural network model comprises:
initializing a medicine subgraph, and representing the node representation vector correspondinglyInitializing original drug characterization; the medicine interaction information of two medicines is unknown or has no effect, the edge weight is initialized to the probability of the co-occurrence of the two medicines, and the edge weight e is initialized vu ＝p vu, wherein ,pvu Is the probability of co-occurrence of two drugs; the medicine interaction information of the two medicines is synergistic, and the edge weight e is initialized vu =1; and initializing the side weight e by taking the medicine interaction information of the two medicines as antagonism vu One or more of = -1;
updating the drug characterization in the drug package subgraph according to the mask vector to obtain an updated drug characterization
The method comprises the following steps:
wherein FNN is a feedforward neural network with a single hidden layer, sigma (MLP (u)) is a mask vector extracted by a mask layer, sigma is a sigmoid function, ☉ is multiplication element by element, and d u Characterizing the original medicine;
setting situation influence factors according to the updated drug characterization, and updating the drug interaction information to obtain updated edge weights;
constructing the first neural network model as follows:
wherein ,
information vector transferred from node v to node u for layer i; w (W) 1 (l-1) Parameters to be learned are the model of the layer 1; />
Vector characterization at layer 1 for node v; />
Information vectors obtained by aggregation of the node u at the first layer; n (u) is a neighbor node of the node u; the GRU is a gating circulation unit; />
The updated edge weight; />
Information vector transferred from node v to node u in the first layer; />
A characterization vector of the node u at the first layer-1; />
Representing the vector of the node u at the first layer; FNN is a feedforward neural network with a single hidden layer; w (W) 0 (l-1) Parameters to be learned are the model of the layer 1; w and M are parameters to be learned;
Inputting the updated medicine package subgraph into the first neural network model for information transmission, extracting node characterization of a layer of graph neural network, and obtaining subgraph characterization:
g＝∑ v∈V σ(FNN([d v ||h v ]))☉(FNN([d v ||h v ]))；
v is the total node set in the updated medicine package subgraph, g is the subgraph representation; FNN ([ d) v ||h v ]) Is thatThe node initial characterization and the characterization after information transmission are fused with each other through a feedforward neural network of a single hidden layer;
training the first neural network model using the following loss function:
wherein ,ui Characterization of patient corresponding to the ith patient, g i Representing a sub-graph corresponding to the ith medicine package; MLP is the second multilayer perceptron, MLP ([ u ] i ||g i ]) Scoring the model for patient i and drug package i;
regularizing the term for L2.
6. The pharmaceutical package recommendation method of any one of claims 1 to 4, wherein the training method of the first neural network model comprises:
initializing a medicine subgraph, and initializing the corresponding characterization of the node characterization vector into an original medicine characterization; initializing the corresponding token of the edge token vector to e vu ＝FNN([d v ||d u ])；
wherein ,dv and du For drug characterization corresponding to different nodes, FNN is a feedforward neural network with a single hidden layer;
Updating the medicine interaction information in the medicine package subgraph according to the mask vector to obtain updated medicine interaction information, namely an updated edge characterization vector
Wherein sigma is a sigmoid function, ☉ is multiplication element by element, and FNN is a feedforward neural network of a single hidden layer;
constructing the first neural network model as follows:
wherein ,
information vector transferred from node v to node u for layer i; w (W) 1 (l-1) Parameters to be learned are the model of the layer 1; />
Representing vectors for the edges of layer 1; />
Information vectors obtained by aggregation of the node u at the first layer; n (u) is a neighbor node of the node u; />
Information vector transferred from node v to node u in the first layer; />
Representing the vector of the node u at the first layer-1; />
The characterization vector of the node u at the first layer is obtained; FNN is a feedforward neural network with a single hidden layer; />
Parameters to be learned are the model of the layer 1; w is a parameter to be learned;
inputting the updated medicine package subgraph into the first neural network model for information transmission, extracting node characterization of a layer of graph neural network, and obtaining subgraph characterization:
g＝∑ v∈V σ(FNN([d v ||h v ]))⊙(FNN([d v ||h v ]))；
v is the total node set in the updated medicine package subgraph, g is the subgraph representation; FNN ([ d) v ||h v ]) The method comprises the steps that fusion characterization is obtained by a node initial characterization and a characterization after information transmission through a feedforward neural network of a single hidden layer;
training the first neural network model using the following loss function:
wherein MLP is the second multilayer perceptron, MLP ([ u ] i ||g i ]) Scoring the model for patient i and drug package i;
cross entropy loss function for performing edge classification tasks based on edge characterization; />
Regularizing the term for L2; e, e vu To characterize vectors for edges, R uv Drug interaction information corresponding to the edge representation vector; the matrix Q converts the edge token vector into a classified probability vector.
7. A pharmaceutical package recommendation device comprising:
the sub-graph construction module is used for acquiring at least two drug characterizations and at least one drug interaction information in the drug package and constructing a drug package sub-graph; the node representation vector of the medicine package sub-graph is the medicine representation, and the edge representation vector of the medicine package sub-graph is the medicine interaction information;
the acquisition module is used for acquiring the electronic medical record information and the complaint medical record information and determining patient characterization according to the first multi-layer perceptron and the LSTM model;
the first updating module is used for acquiring mask vectors according to the patient characterization, updating the drug characterization in the drug package subgraph and obtaining updated drug characterization;
The second updating module is used for setting situation influence factors according to the updated drug characterization, updating the drug interaction information in the drug package subgraph, and obtaining an updated drug package subgraph;
the processing module is used for inputting the updated medicine package subgraph into a first neural network model to obtain subgraph characterization;
the recommendation module is used for splicing the sub-graph representation with the patient representation, inputting the sub-graph representation into a second multi-layer perceptron, and carrying out predictive scoring on the medicine package to obtain a medicine package recommendation result;
wherein, according to the updated drug characterization, setting a situation influencing factor, updating the drug interaction information includes:
for edge e vu Acquiring contextual impact factors
Wherein FNN is a feedforward neural network with a single hidden layer, a T Updating vectors for parameters with dimensions equal to the output dimensions of the feedforward neural network of the single hidden layer,
and />
Representing the updated medicine;
updating the medicine interaction information in the medicine package subgraph according to the situation influence factors to obtain updated medicine interaction information, namely updated edge weight 
8. A computer system, comprising:
one or more processors;
a memory for storing one or more programs;
wherein the one or more programs, when executed by the one or more processors, cause the one or more processors to implement the method of any of claims 1 to 6.
9. A computer readable storage medium having stored thereon executable instructions which when executed by a processor cause the processor to implement the method of any of claims 1 to 6.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110116029.6A CN112802573B (en) | 2021-01-26 | 2021-01-26 | Medicine package recommendation method, device, computer system and readable storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110116029.6A CN112802573B (en) | 2021-01-26 | 2021-01-26 | Medicine package recommendation method, device, computer system and readable storage medium |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112802573A CN112802573A (en) | 2021-05-14 |
| CN112802573B true CN112802573B (en) | 2023-06-20 |
Family
ID=75812427
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110116029.6A Active CN112802573B (en) | 2021-01-26 | 2021-01-26 | Medicine package recommendation method, device, computer system and readable storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112802573B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113693610B (en) * | 2021-08-31 | 2022-10-14 | 平安科技（深圳）有限公司 | Method and device for processing few-lead electrocardiogram data, storage medium and computer equipment |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102789499A (en) * | 2012-07-16 | 2012-11-21 | 浙江大学 | Collaborative filtering method on basis of scene implicit relation among articles |
| CN109658208A (en) * | 2019-01-15 | 2019-04-19 | 京东方科技集团股份有限公司 | Recommended method, device, medium and the electronic equipment of drug |
| CN110837565A (en) * | 2019-11-14 | 2020-02-25 | 中山大学 | Model training method and device for realizing medicine recommendation and calculating equipment |
| CN111191020A (en) * | 2019-12-27 | 2020-05-22 | 江苏省人民医院（南京医科大学第一附属医院） | Prescription recommendation method and system based on machine learning and knowledge graph |
| CN111815609A (en) * | 2020-07-13 | 2020-10-23 | 北京小白世纪网络科技有限公司 | Pathological image classification method and system based on context awareness and multi-model fusion |
| CN111968715A (en) * | 2020-06-30 | 2020-11-20 | 厦门大学 | Drug recommendation modeling method based on medical record data and drug interaction risk |
Family Cites Families (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20180330808A1 (en) * | 2017-05-10 | 2018-11-15 | Petuum Inc. | Machine learning system for disease, patient, and drug co-embedding, and multi-drug recommendation |
| US20200243180A1 (en) * | 2019-01-25 | 2020-07-30 | International Business Machines Corporation | Optimized medication recommendation using a neural network |
-
2021
- 2021-01-26 CN CN202110116029.6A patent/CN112802573B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102789499A (en) * | 2012-07-16 | 2012-11-21 | 浙江大学 | Collaborative filtering method on basis of scene implicit relation among articles |
| CN109658208A (en) * | 2019-01-15 | 2019-04-19 | 京东方科技集团股份有限公司 | Recommended method, device, medium and the electronic equipment of drug |
| CN110837565A (en) * | 2019-11-14 | 2020-02-25 | 中山大学 | Model training method and device for realizing medicine recommendation and calculating equipment |
| CN111191020A (en) * | 2019-12-27 | 2020-05-22 | 江苏省人民医院（南京医科大学第一附属医院） | Prescription recommendation method and system based on machine learning and knowledge graph |
| CN111968715A (en) * | 2020-06-30 | 2020-11-20 | 厦门大学 | Drug recommendation modeling method based on medical record data and drug interaction risk |
| CN111815609A (en) * | 2020-07-13 | 2020-10-23 | 北京小白世纪网络科技有限公司 | Pathological image classification method and system based on context awareness and multi-model fusion |
Non-Patent Citations (4)
| Title |
|---|
| Safe Medicine Recommendation via Medical Knowledge Graph Embedding;Meng Wang et al;《arXiv》;全文 * |
| 基于BP神经网络的情境化信息推荐服务研究;周朴雄;张兵荣;赵龙文;;情报科学(03);全文 * |
| 基于多源情境协同感知的药品推荐;郑值;徐童;秦川;廖祥文;郑毅;刘同柱;童贵显;;计算机研究与发展(08);全文 * |
| 基于知识图谱的药品联用建模及风险评价;邵云婷;洪灵鸿;陈坚;陈龙彪;范晓亮;许中;赵敏;吴谨准;;中国数字医学(10);全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN112802573A (en) | 2021-05-14 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Raghupathi | Data mining in health care | |
| Zhang et al. | Privacy-preserving double-projection deep computation model with crowdsourcing on cloud for big data feature learning | |
| Alanazi | Identification and prediction of chronic diseases using machine learning approach | |
| Kim et al. | Neural-network based adaptive context prediction model for ambient intelligence | |
| US10978208B2 (en) | Patient risk stratification by combining knowledge-driven and data-driven insights | |
| US11276494B2 (en) | Predicting interactions between drugs and diseases | |
| Miller | Machine intelligence in cardiovascular medicine | |
| US11455543B2 (en) | Textual entailment | |
| US10902943B2 (en) | Predicting interactions between drugs and foods | |
| CN115858886B (en) | Data processing method, device, equipment and readable storage medium | |
| CN112802573B (en) | Medicine package recommendation method, device, computer system and readable storage medium | |
| Saadat et al. | Knowledge graph-based convolutional network coupled with sentiment analysis towards enhanced drug recommendation | |
| WO2022227171A1 (en) | Method and apparatus for extracting key information, electronic device, and medium | |
| Wen et al. | Integrated neural network model with pre-RBF kernels | |
| Aparicio et al. | Structure-based identification of sensor species for anticipating critical transitions | |
| Nelakurthi et al. | Finding cut from the same cloth: Cross network link recommendation via joint matrix factorization | |
| Nguyen et al. | Meta-learning and personalization layer in federated learning | |
| US20220367051A1 (en) | Methods and systems for estimating causal effects from knowledge graphs | |
| Parmar et al. | 5G-enabled deep learning-based framework for healthcare mining: State of the art and challenges | |
| Gaykar et al. | Faulty Node Detection in HDFS Using Machine Learning Techniques. | |
| US11823775B2 (en) | Hashing electronic records | |
| Lee et al. | Bayesian Nonparametric Collaborative Topic Poisson Factorization for Electronic Health Records-Based Phenotyping. | |
| US20210057097A1 (en) | Detection of product restrictions | |
| Unnikrishnan et al. | Efficient parameter tuning of neural foundation models for drug perspective prediction from unstructured socio-medical data | |
| Muthuraju et al. | Doctor patient assistance system using artificial intelligence |
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 |