CN112037880B

CN112037880B - Medication recommendation method, device, equipment and storage medium

Info

Publication number: CN112037880B
Application number: CN202010896613.3A
Authority: CN
Inventors: 侯永帅; 王垂新; 赵建双
Original assignee: Kangjian Information Technology Shenzhen Co Ltd
Current assignee: Kangjian Information Technology Shenzhen Co Ltd
Priority date: 2020-08-31
Filing date: 2020-08-31
Publication date: 2022-07-15
Anticipated expiration: 2040-08-31
Also published as: WO2022041729A1; CN112037880A

Abstract

The invention relates to the field of data processing, and discloses a medication recommendation method, device, equipment and storage medium, which are applied to the field of intelligent medical treatment. Inputting a diagnosis result into a medicine recommendation model to perform medicine matching to obtain a candidate medicine list, performing screening based on the candidate medicine list and inquiry data to obtain a recommended medicine, and finally selecting a medicine meeting the conditions based on the actual condition of a patient and storing the medicine in an inquiry list for subsequent use and inquiry; by the method, automatic recommendation of the medicines is realized, the phenomenon that doctors cannot use the medicines due to wrong memory of medicine properties of the medicines is avoided, and when prescriptions are provided, the doctors can quickly determine the corresponding treatment prescriptions according to the recommended medicines in the inquiry list, so that the diagnosis efficiency of the doctors is improved. In addition, the invention also relates to a block chain technology, and the diagnosis result, the candidate medicine list and the medicine recommendation result can be stored in the block chain.

Description

Medication recommendation method, device, equipment and storage medium

Technical Field

The application relates to the field of data processing, in particular to a medication recommendation method, device, equipment and storage medium.

Background

With the continuous development of the artificial intelligence technology, the technology is concerned about and used in various fields, and for different application scenes in various fields, a corresponding intelligent system is developed based on the artificial intelligence technology to help people to improve the working efficiency, and the technology is not exceptional in medical systems of hospitals. Since the drug storage in hospitals is calculated in quantities of tens of thousands, each time a doctor selects a drug that satisfies a patient's condition and a diagnosis result, the medicines meeting the current treatment requirements of patients and the information of the medicine names, descriptions, stock conditions and the like of the medicines need to be found out in a large-scale medicine library, in order to help doctors to search drugs from drug libraries, the currently used method is to search drug related information in a drug management system according to the keywords of the drugs, the inquiry function can only inquire according to the name of the medicine or specific keywords, the inquiry process completely depends on the experience and memory of doctors, in such a large-scale medicine library, doctors are difficult to remember medicine information such as medicine names, medicine effects and dosage of each medicine, and the doctors select medicines by the method of inquiring according to the medicine names or key words, so that the medicine prescription efficiency is low, and the medicine prescription pertinence and accuracy are difficult to guarantee.

Disclosure of Invention

The invention mainly aims to solve the technical problem of low drug administration efficiency caused by difficulty in drug inquiry of doctors in the conventional medical system.

The invention provides a medication recommendation method in a first aspect, which comprises the following steps:

acquiring historical prescriptions from a plurality of medical databases through a data crawler tool, and learning the drug properties and drug administration rules of the historical prescriptions by adopting a machine learning algorithm to construct a drug recommendation model;

acquiring an inquiry sheet of a current patient, and extracting a diagnosis result and inquiry data in the inquiry sheet;

inputting the diagnosis result into the drug recommendation model for drug matching to generate a candidate drug list;

sorting, screening and filtering the medicines in the candidate medicine list according to the inquiry data to obtain a final medicine recommendation result;

and selecting the medicines meeting the requirements of the patient from the medicine recommendation results according to the received medicine use request, generating a prescription and storing the prescription in the inquiry list.

Optionally, in a first implementation manner of the first aspect of the present invention, the extracting the diagnosis result and the inquiry data in the inquiry sheet includes:

extracting the patient information, the visiting department, the chief complaint information, the inquiry dialogue contents among the doctors and the patients and the diagnosis result of the doctor recorded in the inquiry list by using a character recognition algorithm;

according to preset priority classification conditions, carrying out priority classification on the patient information, the clinic of seeing a doctor, the chief complaint information, the inquiry dialogue contents among the patients of the doctor and the diagnosis results of the doctor to obtain a query condition list;

and combining the conditions except the diagnosis result in the query condition list according to the priority level to obtain the feature phrase.

Optionally, in a second implementation manner of the first aspect of the present invention, the sorting, screening, and filtering the drugs in the candidate drug list according to the inquiry data to obtain a final drug recommendation result includes:

screening and filtering the candidate drug list according to the conditions in the feature phrase, and removing drugs which do not meet the patient information to obtain a new drug sequence;

according to the matching condition of the medicines and the inquiry list, scoring and sequencing each medicine in the new medicine sequence to obtain a candidate medicine sequence;

and selecting N medicines ranked at the top from the candidate medicine sequence to obtain a medicine recommendation result.

Optionally, in a third implementation manner of the first aspect of the present invention, before the prioritizing the patient information, the visit department, the chief complaint information, the content of the inquiry dialogue among the doctors and the patients, and the diagnosis result of the doctor according to the preset prioritization condition to obtain the query condition list, the method further includes:

inquiring a corresponding patient diagnosis historical record from a diagnosis database of a hospital diagnosis system according to the patient information;

determining allergy history and contraindication information for the patient based on the diagnostic history.

Optionally, in a fourth implementation manner of the first aspect of the present invention, the scoring and sorting each drug in the new drug sequence according to a matching condition between the drug and an inquiry sheet to obtain a candidate drug sequence includes:

performing secondary screening and filtering on the new medicine sequence according to the allergy history and the taboo information to obtain a second medicine sequence;

and scoring the matching degree of each medicine in the second medicine sequence by using a preset scoring model, and sequencing the medicines from high to low according to scores to obtain a candidate medicine sequence.

Optionally, in a fifth implementation manner of the first aspect of the present invention, the obtaining, by a data crawler tool, historical prescriptions from a plurality of medical databases, and learning, by using a machine learning algorithm, drug properties and drug administration rules of the historical prescriptions, and constructing the drug recommendation model includes:

acquiring historical inquiry data of doctors from a plurality of medical databases through a data crawler tool, extracting historical prescriptions from the historical inquiry data, and constructing a medicine pushing data set;

extracting the medicines in the medicine pushing data set and medicine characteristic information corresponding to the medicines to generate a training set;

and performing deep learning on the medicines in the training set and the medicine characteristic information corresponding to the medicines by using a machine learning algorithm to construct a medicine recommendation model.

Optionally, in a sixth implementation manner of the first aspect of the present invention, after the selecting, according to the received drug usage request, a drug that satisfies a patient from the drug recommendation results, and generating a prescription to store in the questionnaire, the method further includes:

randomly drawing a plurality of inquiry lists, and extracting the use condition of the medicine recommendation result in each inquiry list;

and if the service condition does not meet the preset acceptance rate, reading the inquiry sheets of all the medicines in the non-adopted medicine recommendation result, and retraining the medicine recommendation model based on the inquiry sheets to obtain the medicine recommendation model after iterative optimization.

A second aspect of the present invention provides a medication recommendation device, including:

the training module is used for acquiring historical prescriptions from a plurality of medical databases through a data crawler tool, and learning the drug properties and drug administration rules of the historical prescriptions by adopting a machine learning algorithm to construct a drug recommendation model;

the extraction module is used for acquiring an inquiry sheet of a current patient and extracting a diagnosis result and inquiry data in the inquiry sheet;

the matching module is used for inputting the diagnosis result into the drug recommendation model to perform drug matching so as to generate a candidate drug list;

the screening module is used for sorting, screening and filtering the medicines in the candidate medicine list according to the inquiry data to obtain a final medicine recommendation result;

and the recommending module is used for selecting the medicines meeting the requirements of the patient from the medicine recommending results according to the received medicine using request, generating a prescription and storing the prescription into the inquiry sheet.

Optionally, in a first implementation manner of the second aspect of the present invention, the extracting module includes:

the recognition unit is used for extracting the patient information, the visiting department, the chief complaint information, the inquiry dialogue contents among the doctors and the patients and the diagnosis result of the doctor recorded in the inquiry list by using a character recognition algorithm;

the configuration unit is used for carrying out priority division on the patient information, the visiting department, the chief complaint information, the inquiry dialogue contents among the doctors and the patients and the diagnosis results of the doctors according to preset priority division conditions to obtain a query condition list;

and the combination unit is used for combining the conditions except the diagnosis result in the query condition list according to priority levels to obtain the characteristic phrases.

Optionally, in a second implementation manner of the second aspect of the present invention, the screening module includes:

the filtering unit is used for screening and filtering the candidate medicine list according to the conditions in the feature phrases, and eliminating medicines which do not meet the patient information to obtain a new medicine sequence;

the sorting unit is used for scoring and sorting each medicine in the new medicine sequence according to the matching condition of the medicines and the inquiry list to obtain a candidate medicine sequence;

and the selection unit is used for selecting N medicines ranked at the top from the candidate medicine sequence to obtain a medicine recommendation result.

Optionally, in a third implementation manner of the second aspect of the present invention, the medication recommendation device further includes a query module, which is specifically configured to:

Optionally, in a fourth implementation manner of the second aspect of the present invention, the sorting unit is specifically configured to:

carrying out secondary screening and filtering on the new medicine sequence according to the allergy history and the taboo information to obtain a second medicine sequence;

Optionally, in a fifth implementation manner of the second aspect of the present invention, the training module is specifically configured to:

Optionally, in a sixth implementation manner of the second aspect of the present invention, the medication recommendation device further includes an optimization module, which is specifically configured to:

randomly extracting a plurality of inquiry sheets, and extracting the use condition of the medicine recommendation result in each inquiry sheet;

and if the service condition does not meet the preset acceptance rate, reading all inquiry lists of the medicines in the non-adopted medicine recommendation result, and retraining the medicine recommendation model based on the inquiry lists to obtain the iteratively optimized medicine recommendation model.

A third aspect of the present invention provides a medication recommendation apparatus, including: a memory having instructions stored therein and at least one processor, the memory and the at least one processor interconnected by a line;

the at least one processor invokes the instructions in the memory to cause the medication recommendation device to perform the medication recommendation method described above.

A fourth aspect of the present invention provides a computer-readable storage medium having stored therein a computer program which, when run on a computer, causes the computer to execute the medication recommendation method described above.

According to the technical scheme provided by the invention, medicine recommendation is realized through the model so as to improve the medicine prescription efficiency of a doctor, and specifically, diagnosis data of the doctor on a patient is acquired and input into the medicine recommendation model, and the medicine recommendation model recommends a targeted medicine to the doctor for selective use based on the diagnosis data, so that the automatic recommendation of the medicine is realized, and the phenomenon that the doctor does not use the medicine due to the wrong memory of the medicine property of the medicine is avoided

Drawings

FIG. 1 is a schematic diagram of a first embodiment of a medication recommendation method in an embodiment of the invention;

FIG. 2 is a diagram of a second embodiment of a medication recommendation method in accordance with an embodiment of the present invention;

FIG. 3 is a diagram of a third embodiment of a medication recommendation method in an embodiment of the invention;

FIG. 4 is a diagram of a fourth embodiment of a medication recommendation method in an embodiment of the invention;

FIG. 5 is a schematic diagram of one embodiment of a medication recommendation device in accordance with embodiments of the present invention;

FIG. 6 is a schematic diagram of another embodiment of a medication recommendation device in accordance with an embodiment of the present invention;

fig. 7 is a schematic diagram of an embodiment of a medication recommending apparatus according to an embodiment of the present invention.

Detailed Description

Aiming at the defects in the prior art, the application provides an intelligent medication recommendation method with self-learning capability, and in the medication prescription process of a doctor, a candidate medicine suitable for the current medication requirement of a patient can be automatically recommended to the doctor according to the personal information, the chief complaint content, the inquiry dialogue content of the patient and the doctor, the diagnosis result of the doctor and other information of the patient; the method helps the doctor to reduce the time consumed in the drug inquiry process and improve the drug prescription efficiency of the doctor by recommending candidate drugs for the doctor; when recommending medicines, the method extracts characteristic information of an inquiry list from personal information of a patient, chief complaints of the patient, inquiry dialogue and diagnosis results of a doctor, recommends candidate medicines by using an intelligent medicine pushing model trained in advance, takes the personal information, the chief complaint content and the diagnosis results of the patient into consideration for the recommended medicines, is more matched with the current diagnosis scene of the patient, and helps the doctor improve the pertinence of prescription; meanwhile, the condition of age, sex, inoculation, allergy, contraindication and the like of the patient is considered when the candidate medicine is recommended by the medicine pushing method, so that the medicine is prevented from being prescribed by a doctor, and the safety of the prescription of the doctor is improved.

The terms "first," "second," "third," "fourth," and the like in the description and in the claims, as well as in the drawings, if any, are used for distinguishing between similar elements and not necessarily for describing a particular sequential or chronological order. It will be appreciated that the data so used may be interchanged under appropriate circumstances such that the embodiments described herein may be implemented in other sequences than those illustrated or described herein. Moreover, the terms "comprises," "comprising," or "having," and any variations thereof, are intended to cover a non-exclusive inclusion, such that a process, method, system, article, or apparatus that comprises a list of steps or elements is not necessarily limited to those steps or elements expressly listed, but may include other steps or elements not expressly listed or inherent to such process, method, article, or apparatus.

For understanding, the following describes a specific process of an embodiment of the present invention, and with reference to fig. 1, a first embodiment of a medication recommendation method according to an embodiment of the present invention includes:

101. acquiring historical prescriptions from a plurality of medical databases through a data crawler tool, and learning the drug properties and drug administration rules of the historical prescriptions by adopting a machine learning algorithm to construct a drug recommendation model;

in the step, historical inquiry data of doctors can be obtained from a plurality of medical databases through a data crawler tool, and a historical prescription is extracted from the historical inquiry data to construct a medicine pushing data set;

extracting medicines in the medicine pushing data set and medicine characteristic information corresponding to the medicines to generate a training set;

In practical application, the medicine recommendation model can be specifically constructed by adopting an XGB OST algorithm and a neural network deep learning algorithm, and the specific implementation steps comprise:

extracting prescriptions through a filling template of a diagnosis order, for example, identifying the prescription filling fields of the diagnosis order in historical inquiry data through prescription filling fields specified in the template, reading information based on the corresponding relationship between the prescription filling fields and content positions, and generating a historical prescription;

then, extracting keywords in the prescription, such as the name of the medicine and the usage and the dosage in the medicine, by using a TextRank keyword extraction algorithm; in practical application, when the name of the medicine is extracted, the name of the medicine can be extracted by comparing and identifying the entity name of the medicine in the historical prescription based on the entity recorded in the knowledge map by combining the knowledge map of the medicine;

further, extracting disease species corresponding to the historical prescription and disease symptoms of the patient of the medical record based on a keyword extraction algorithm to form a disease characteristic library;

constructing a corresponding relation between a disease characteristic library and medicines;

and based on the corresponding relation, learning the corresponding relation by using a machine learning algorithm to obtain a medicine recommendation model.

Further, when learning the corresponding relationship, the method specifically includes:

dividing the corresponding relation into a training data set and a verification data set;

based on the training data set, learning the medicine and the medication rule by using a machine learning algorithm, such as a neural network deep learning algorithm, to obtain a medicine recommendation model;

verifying the drug recommendation model by using a verification data set, and outputting a final drug recommendation model after the verification is passed; if not, the training data set and the verification data set are disturbed, and are redistributed for training again.

102. Acquiring an inquiry sheet of a current patient, and extracting a diagnosis result and inquiry data in the inquiry sheet;

in this step, the questionnaire refers to a disease diagnosis receipt prescribed by a doctor at the time of the patient's visit, and the questionnaire includes at least patient information, questionnaire data, and diagnosis results, and the patient information includes information such as sex, age, allergy history, contraindications, and pregnancy (female patient) of the patient, and the questionnaire data is some symptom information, disease history, and disease information that the patient actively states to the doctor at the time of the patient's visit, which are inquired by the doctor.

In practical application, when the inquiry sheet is obtained, the operation information and the input information on the prescription interface of each doctor are monitored, the corresponding inquiry sheet is generated based on the monitored information, and then the content in the inquiry sheet is specified and extracted.

Of course, when extracting the diagnosis result and the inquiry data in the inquiry list, in addition to the manner of extracting the diagnosis result and the inquiry data from the inquiry list provided above, the method may also be implemented by monitoring and extracting the prescription page of the doctor, specifically by monitoring and extracting information on a specific item, for example, only monitoring and identifying the input frame on the position of the diagnosis result item and only monitoring and identifying the input frame on the position of the inquiry item of the doctor.

In practical applications, the inquiry data may be extracted in real time, in particular from a dialog between a doctor and a patient, in addition to being obtained from an inquiry list.

103. Inputting the diagnosis result into a medicine recommendation model for medicine matching to generate a candidate medicine list;

in this embodiment, when matching medicines, the medicine recommendation model queries all medicines, of which pharmacology or drug property meets the use requirement of the diagnosis result, from a medicine repository according to the input diagnosis result, then calculates the matching degree between the queried medicines and the diagnosis result one by one, and selects the medicines, of which the matching degree is greater than a preset percentage, as the data structure of the medicine recommendation model based on the matching degree to generate a candidate medicine list.

In practical application, the method also comprises the step of sorting the selected medicines from large to small according to the matching degree while generating the candidate medicine list so as to facilitate the subsequent selection and use of the medicines.

Further, after the candidate medicine list is generated, the steps further include extracting the drug property of each medicine in the candidate medicine list, judging the risk level of each drug property in the using process, and labeling according to the risk level so as to select the use of the medicine in the following process.

In practical applications, since a plurality of drug properties exist in a drug, each drug property may correspond to treatment of one or more diseases, and the determination of the risk level is specifically determined according to a current diagnosis result, that is, according to a specific disease, the degree of influence of a chemical reaction on a human body is determined, so as to determine a corresponding risk level.

104. Sorting, screening and filtering the medicines in the candidate medicine list according to the inquiry data to obtain a final medicine recommendation result;

in practical application, the step may be implemented by extracting feature phrases in the inquiry data to match and filter, and specifically, the step may be implemented by using a model, for example: the method comprises the steps of using a medicine matching model, wherein the medicine matching model comprises a mapping relation of various medicine characteristics, the mapping relation can be composed of medicine characteristic words extracted from information such as medicine names, medicine numbers, using objects, usage, functions, usage amounts, contraindications and the like of various medicines, and the corresponding medicines can be uniquely determined through the mapping relation. The characteristic matching of the characteristic phrases and the characteristics of the medicines can be realized through the medicine matching model, medicine matching can be carried out according to the input characteristic phrases, and matched medicines are output. Specifically, the drug matching model may be a naive bayesian probability model obtained based on a bayesian algorithm, which may count probabilities of drugs according to input feature phrases and output the drug with the highest probability.

In another embodiment, the drug matching model may also be a drug matching neural network obtained based on an artificial neural network algorithm, and the drug matching neural network may be a multilayer architecture, for example, the neural network layer structure may be divided according to the priority division of the feature phrases, for example, if the feature phrases are divided into three priority levels, namely high, medium, and low, the drug matching neural network may be correspondingly set as a structure of three hidden layers to correspond to the priority division of the feature phrases.

In specific implementation, the medicine matching models corresponding to the hospital departments may be different, and at this time, the medicine matching models corresponding to the hospital departments may be queried first, and then the feature phrases may be input for feature matching, so as to obtain corresponding output results.

105. According to the received medicine use request, selecting the medicine meeting the patient from the medicine recommendation result, generating a prescription and storing the prescription in an inquiry list.

In this embodiment, the obtained medicine recommendation result is displayed at the position of the prescription in the prescription interface of the doctor, and the doctor selects the displayed medicine, generates the prescription based on the selected medicine, and adds the prescription to the questionnaire, so as to package the subsequent medicines.

In practical applications, the prescription in this step may be a prescription or an over-the-counter prescription, and the prescription is a list of medicines that the doctor has made for the patient, a written document for the doctor to take medicine for the current patient is a basis for the pharmacy staff to prepare the medicines, and the prescription recommendation may be a reference for the doctor to make a prescription, and particularly, if the prescription recommendation is a proper list of medicines, the prescription may be directly used as the prescription. After the medicine matching model carries out feature matching according to the input feature phrases, the output matching result comprises various medicines, a medicine list is generated according to the medicines, prescription recommendation is obtained, a plurality of recommended prescriptions may exist, after the medicine is recommended to a doctor, the doctor selects one of the recommended prescriptions according to the actual situation, and finally the system automatically stores the recommended prescriptions in an inquiry list for recording.

According to the method, the diagnosis data of the doctor on the patient are acquired, the diagnosis data are input into the medicine recommendation model, and the medicine recommendation model recommends the targeted medicine to the doctor for selective use based on the diagnosis data, so that the automatic recommendation of the medicine is realized, and the phenomenon of useless medicine caused by mistaken memory of the medicine property of the doctor is avoided;

furthermore, the recommended medicine can be referred to when a doctor makes a prescription, so that the prescription generation process does not need the doctor to comprehensively consider various factors for carrying out complicated operation, the generation process of the medical prescription is simplified, the problem of repeated prescription modification caused by doctor errors is avoided, and the generation efficiency of the prescription is improved.

Referring to fig. 2, a second embodiment of the medication recommending method according to the embodiment of the present invention includes:

201. extracting the patient information, the visiting department, the chief complaint information, the inquiry dialogue contents among the doctors and the patients and the diagnosis result of the doctor recorded in the inquiry list by using a character recognition algorithm;

in the step, when the text recognition algorithm is used for recognition, the recognition can be specifically realized by obtaining a template of the inquiry sheet, a plurality of field information are arranged in the template, different field information is provided with corresponding filling contents, for example, a name field corresponds to a filling name, an identity field corresponds to a filling identity card number, and a diagnosis content field corresponds to filling diagnosis data and inquiry data.

And identifying each field in the inquiry bill by using a character recognition algorithm based on the field information in the template, extracting the attribute value of the corresponding field in the inquiry bill according to the corresponding position relation between the field and the filling content, and determining patient information, a doctor seeing a doctor, chief complaint information, inquiry dialogue content between doctors and patients and a diagnosis result based on the attribute value.

In practical application, the inquiry dialogue content and the diagnosis result between a doctor and a patient can be specifically matched and identified by using a disease knowledge graph, specifically, a plurality of key features in corresponding field information are extracted, the key features and entities recorded in a disease knowledge graph are matched and identified based on the key features, then, according to the identified result, a corresponding disease is selected, and the corresponding content of the disease and the field information is continuously compared to obtain a final result, for example, inquiry dialogue content.

202. Prioritizing the patient information, the visiting department, the chief complaint information, the inquiry dialogue contents among the doctors and the patients and the diagnosis results of the doctors according to preset prioritization conditions to obtain a query condition list;

203. combining conditions except the diagnosis result in the query condition list according to the priority level to obtain a characteristic phrase;

in practical application, the condition information in the inquiry list is extracted, and the pathological feature words of the information recorded in the inquiry list can be specifically extracted based on a TextRank keyword extraction algorithm. For example, if the questionnaire includes expressions of "catching a cold and causing a cold, causing fever, causing headache, and having nasal obstruction", redundant data such as verbs "causing", and "having" may be removed, core pathological feature words such as "catching a cold", "fever", "headache", and "having nasal obstruction" may be extracted, a diagnosis result in the questionnaire may be determined based on the case feature words, and by extracting the pathological feature words from the questionnaire, redundant and useless data may be effectively removed, so as to ensure processing efficiency and accuracy of subsequent prescription generation.

Further, the inquiry data in the extracted inquiry list, that is, the content of the inquiry of the doctor or the statement of the patient, such as "whether to seek medical attention or operation in the hospital", "whether there is an allergy history or whether there is a drug allergy", is extracted by the TextRank keyword extraction algorithm, and even further includes keyword extraction processing of patient information, a visiting department, a diagnostician, and the like, and the keywords are sorted and combined according to preset keywords to obtain a feature phrase.

204. Inputting the diagnosis result into the medicine recommendation model for medicine matching to generate a candidate medicine list;

in this step, the specific drug recommendation model queries, from a drug repository, all the drugs whose pharmacology or drug property meets the use of the diagnosis result according to the input diagnosis result, then calculates the matching degree between the queried drugs and the diagnosis result one by one, and selects the drugs whose matching degree is greater than a preset percentage as a data structure of the drug recommendation model based on the matching degree to generate a candidate drug list.

205. Screening and filtering the candidate drug list according to conditions in the characteristic word group, and removing drugs which do not meet the patient information to obtain a new drug sequence;

in the step, the candidate medicine list is firstly screened and filtered according to the conditions in the feature phrase and the priority, specifically, a part of medicines which are not very high in matching degree with the diagnosis result is screened, then the screened candidate medicines are further removed according to the patient information, the medicines which are not satisfied with the pharmacology are removed, the last medicines are left to form a new medicine sequence, the medicine sequence is obtained by sequencing with the patient information as the highest priority and the conditions in the feature phrase as the second best priority.

206. According to the matching condition of the medicines and the inquiry list, scoring and sequencing each medicine in the new medicine sequence to obtain a candidate medicine sequence;

in practical application, the matching condition refers to the matching degree of information such as usage object, usage, function, usage amount and contraindication of the medicine and specific information of a patient, wherein the patient information includes age, sex, inoculation condition, allergy history and contraindication, the matching condition of the matching condition is input into a scoring model to perform recommended and predicted scoring by comparing whether the usage object, usage, function, usage amount and contraindication of the medicine in a new medicine sequence and the age, sex, inoculation condition, allergy history and contraindication in the patient information meet preset thresholds, so as to obtain scoring scores, and the new medicine sequence is ranked based on the scoring scores to obtain candidate medicine sequences.

207. Selecting N medicines ranked in the front from the candidate medicine sequence to obtain a medicine recommendation result;

in practical application, when selecting the medicines according to the sorting, the method further comprises the steps of detecting whether the stock quantity of each medicine is enough, if not, completing the next medicine in the sorting until the quantity of N medicines is completed, and certainly, when selecting N medicines, besides considering the stock quantity, judging whether the medicines have special pharmacological side effects, such as high antibiotic or high pain-relieving component, further, selecting the medicines according to the rule that Chinese patent medicines are selected as main medicines and western medicines are selected in a priority mode, and finally outputting the medicines meeting all the conditions.

208. According to the received medicine use request, selecting the medicine meeting the patient from the medicine recommendation result, generating a prescription and storing the prescription in an inquiry list.

In this embodiment, selecting a drug suitable for a patient may be specifically realized by a drug matching model, where the specific drug matching model is a drug matching model obtained based on an artificial neural network algorithm, and the drug matching model may be a multilayer architecture, for example, if the feature phrases are divided into three priority levels, i.e., high, medium, and low, the drug matching neural network may be correspondingly set as a structure with three hidden layers to correspond to the priority division of the feature phrases.

In the concrete implementation, the medicine matching models corresponding to the patients may be different, at this time, after the medicine matching models corresponding to the disease categories of the patients are inquired, the extracted feature phrases of the disease categories are input for feature matching, and a corresponding output result is obtained, and the range of the model matching medicines is the range of the provided medicine recommendation result for matching.

Through the implementation of the method, the diagnosis result is input into a medicine recommendation model to carry out medicine matching to obtain a candidate medicine list, screening is carried out on the basis of the candidate medicine list and inquiry data to obtain recommended medicines, and finally, medicines meeting the conditions are selected on the basis of the actual condition of a patient and stored in an inquiry list for subsequent use and inquiry; the method not only realizes the automatic recommendation of the medicine, but also avoids the phenomenon that the doctor causes useless medicine due to the wrong memory of the medicine property, and when the prescription is made, the doctor can quickly determine the corresponding treatment prescription according to the recommended medicine in the inquiry list, thereby improving the diagnosis efficiency of the doctor.

Referring to fig. 3, a third embodiment of the medication recommending method according to the embodiment of the present invention includes:

301. acquiring an inquiry sheet of a current patient, and extracting a diagnosis result and inquiry data in the inquiry sheet;

302. inputting the diagnosis result into the medicine recommendation model for medicine matching to generate a candidate medicine list;

303. inquiring a corresponding patient diagnosis historical record from a diagnosis database of a hospital diagnosis system according to the patient information;

304. determining the allergy history and contraindication information of the patient according to the diagnosis history record;

in this embodiment, the screening of the allergens mainly ensures the safety of the prescription and the life safety of the patient, specifically by querying in the historical diagnosis record.

Specifically, the process of inquiring the medical history in the diagnosis database is started by identifying whether the characteristic words for hospitalizing in the hospital exist in the diagnosis data, if the characteristic words for hospitalizing in the hospital exist in the diagnosis data, the corresponding medical history record is inquired from the diagnosis database of the hospital system according to the identity card information of the patient or the medical insurance card information, the remark item information of the doctor is extracted from the medical history record, the corresponding cautionary items, namely the allergy history and contraindication information, and even the diagnosis result of the historical hospitalization can be determined based on the remark item information, and the candidate medicine sequence is further screened based on the inquiry result.

305. Screening and filtering the candidate drug list according to conditions in the characteristic phrases, and eliminating drugs which do not meet the information of the patient to obtain a new drug sequence;

306. carrying out secondary screening and filtering on the new medicine sequence according to the allergy history and the taboo information to obtain a second medicine sequence;

307. scoring the matching degree of each medicine in the second medicine sequence by using a preset scoring model, and sequencing the medicines from high to low according to scores to obtain a candidate medicine sequence;

specifically, the categories related to extracting pathological feature words from the medical history record and profile feature words extracted from the personal profile information of the patient are numerous, and the influence weight of the feature words of each category on the final prescription is different, for example, for an underage patient with the age of less than 18 years old, the medicine for adults is not suitable, and at this time, the influence of the age of the patient on the prescription medicine is large, and the priority is high; for a patient with male sex, the drug for gynecological diseases is also not applicable, and the sex priority of the patient is high; furthermore, if the diseased part of the patient is the stomach, the drugs for the brain or kidney are not suitable for the corresponding prescription. In the embodiment, the pathological feature words and the archive feature words are subjected to priority division, so that the occupied weight of each category of feature words during prescription generation can be distinguished, and the efficiency and the accuracy of medicine matching are improved.

In one embodiment, the pathological feature words may further include disease location, disease name and symptom expression, and the archival feature words include patient subjects, allergens and past medical history; the method for carrying out priority division on the pathological characteristic words and the archival characteristic words comprises the following steps: dividing the patient object and the allergen into high-priority feature words; dividing the disease part and the disease name into feature words with medium priority; the symptom expression and the past medical history are divided into low-priority characteristic words.

308. Selecting N medicines ranked at the front from the candidate medicine sequence to obtain a medicine recommendation result;

309. according to the received medicine use request, selecting the medicine meeting the patient from the medicine recommendation result, generating a prescription and storing the prescription in an inquiry list.

In conclusion, the method of the embodiment automatically learns the medicine recommendation model from the historical medication data of the doctor; extracting inquiry list characteristic information from inquiry information such as patient personal information, chief complaints, inquiry dialogue contents, diagnosis results of doctors and the like, and recommending candidate medicines for the doctors by using a trained medicine recommendation model; in the drug recommendation process, information of sex, age, contraindication, pregnancy and the like of a patient is considered, so that conflict between the recommended drug and the condition of the patient is avoided; the medicine recommending method realizes automatic medicine recommending, helps doctors to improve prescription efficiency, and also helps doctors to avoid the condition of medicine taking errors caused by wrong memory of medicine properties, and improves the accuracy and safety of medicine taking.

In practical applications, besides the recommendation of a medicine to a physician is realized in the above manner, the recommendation can be performed in a manner of training a model, and the model can be recognized more quickly, please refer to fig. 4, in which a fourth embodiment of the medication recommendation method in the embodiment of the present invention includes:

401. acquiring historical inquiry data of doctors from a plurality of medical databases through a data crawler tool, extracting historical prescriptions from the historical inquiry data, and constructing a medicine pushing data set;

402. extracting the medicines in the medicine pushing data set and medicine characteristic information corresponding to the medicines to generate a training set;

403. deep learning is carried out on the medicines in the training set and medicine characteristic information corresponding to the medicines by using a machine learning algorithm, and a medicine recommendation model is constructed;

in this embodiment, the historical inquiry data may be a handwritten diagnosis form of a doctor, or may be an inquiry form obtained by using the method, which may specifically be implemented by the following steps:

step 1, feature extraction: extracting the characteristics of the inquiry from the information of the patients (including age, sex, pregnancy, allergy history and taboo), the departments of visiting the doctor, the content of chief complaints, the content of inquiry dialogue, the diagnosis result of doctors and the like;

step 2, model training: the inputs to the model are the characteristics of the interrogation and the predicted targets are drug candidates for the target interrogation list. The task of the model training module is to use a historical inquiry sheet and a prescription result thereof as training data and adopt a machine learning algorithm to train and construct a medicine recommendation model for predicting candidate medicines;

further, whether the drug recommendation model reaches a specified online condition is tested, and if the drug recommendation model reaches the specified online condition, the drug recommendation model is obtained online;

then, recommending the medicine based on an online medicine recommendation model: for given inquiry data, predicting a candidate drug list suitable for the inquiry list by using a trained drug prediction model; specifically, by collecting the information (including age, sex, pregnancy, allergy history and taboo), the department of visit, chief complaint information, the contents of inquiry dialogue between doctors and patients, the diagnosis result of the doctor and the like of the patient who inputs the inquiry list, the method of step 1 is used for extracting the characteristics of the inquiry list; and inputting the special diagnosis into a drug recommendation model for prediction to obtain a prediction result.

Further, candidate drugs are screened and ranked: screening candidate medicines recommended by a medicine recommendation module according to conditions such as age, sex, inoculation condition, allergy history and contraindication of patients in the inquiry list, and then grading and sequencing the screened candidate medicines according to matching conditions of the medicines and the inquiry list;

and finally, pushing the scoring result to a doctor for use selection of the medicine, and recording the use condition of the doctor on the recommended result.

404. Acquiring an inquiry sheet of a current patient, and extracting a diagnosis result and inquiry data in the inquiry sheet;

405. inputting the diagnosis result into the medicine recommendation model for medicine matching to generate a candidate medicine list;

406. sorting, screening and filtering the medicines in the candidate medicine list according to the inquiry data to obtain a final medicine recommendation result;

407. selecting a medicine meeting the requirements of a patient from medicine recommendation results according to a received medicine use request, generating a prescription and storing the prescription in an inquiry list;

408. randomly extracting a plurality of inquiry sheets, and extracting the use condition of the recommended result of the medicine in each inquiry sheet;

409. and if the use condition does not meet the preset acceptance rate, reading the inquiry lists of all the medicines in the non-adopted medicine recommendation result, and retraining the medicine recommendation model based on the inquiry lists to obtain the medicine recommendation model after iterative optimization.

In this embodiment, the use of the medicine may have different true results and prescription ratios according to the stage-based medical treatment of the patient, and for this purpose, in order to adapt to the recommendation of the medicine in real time, the method further includes:

The method comprises the steps of recording the use condition of the medicine recommended by a medicine recommendation model of a doctor every time, storing the information of the questionnaire if the doctor does not adopt the recommendation result, and when the number of the stored unadopted questionnaires reaches a given condition, re-executing the first-stage model training and iteratively optimizing the medicine recommendation model.

Through the implementation of the scheme, in the process of dispensing the medicine by a doctor, the candidate medicine suitable for the current medicine taking demand of the patient can be automatically recommended to the doctor according to the personal information, the chief complaint content, the inquiry dialogue content of the doctor and the patient, the diagnosis result of the doctor and other information of the patient; the method helps the doctor to reduce the time consumed in the drug inquiry process and improve the drug prescription efficiency of the doctor by recommending the candidate drug for the doctor; when recommending medicines, the method extracts characteristic information of an inquiry list from personal information of a patient, chief complaints of the patient, inquiry dialogue and diagnosis results of a doctor, recommends candidate medicines by using an intelligent medicine pushing model trained in advance, takes the personal information, the chief complaint content and the diagnosis results of the patient into consideration for the recommended medicines, is more matched with the current diagnosis scene of the patient, and helps the doctor improve the pertinence of prescription; meanwhile, the condition of age, sex, inoculation, allergy, contraindication and the like of the patient is considered when the candidate medicine is recommended by the medicine pushing method, so that the medicine is prevented from being prescribed by a doctor, and the safety of the prescription of the doctor is improved.

With reference to fig. 5, the medication recommending method in the embodiment of the present invention is described above, and a medication recommending apparatus in the embodiment of the present invention is described below, where a first embodiment of the medication recommending apparatus in the embodiment of the present invention includes:

the training module 501 is used for acquiring historical prescriptions from a plurality of medical databases through a data crawler tool, and learning drug properties and drug administration rules of the historical prescriptions by adopting a machine learning algorithm to construct a drug recommendation model;

an extracting module 502, configured to obtain an inquiry list of a current patient, and extract a diagnosis result and inquiry data in the inquiry list;

the matching module 503 is configured to input the diagnosis result into the drug recommendation model to perform drug matching, so as to generate a candidate drug list;

a screening module 504, configured to perform sorting, screening, and filtering on the drugs in the candidate drug list according to the inquiry data to obtain a final drug recommendation result;

and the recommending module 505 is configured to select a medicine meeting the patient from the medicine recommending result according to the received medicine use request, generate a prescription, and store the prescription in the inquiry list.

In this embodiment, the medication recommendation device operates the medication recommendation method, and the method includes inputting a diagnosis result into a drug recommendation model to perform drug matching, obtaining a candidate drug list, performing screening based on the candidate drug list in combination with inquiry data to obtain recommended drugs, and finally selecting drugs meeting conditions based on actual conditions of patients, and storing the drugs in an inquiry list for subsequent use and review; by the method, automatic recommendation of the medicines is realized, the phenomenon that doctors cannot use the medicines due to wrong memory of medicine properties of the medicines is avoided, and when prescriptions are provided, the doctors can quickly determine the corresponding treatment prescriptions according to the recommended medicines in the inquiry list, so that the diagnosis efficiency of the doctors is improved.

Referring to fig. 6, a medication recommending apparatus according to a second embodiment of the present invention specifically includes:

the training module 501 is used for acquiring historical prescriptions from a plurality of medical databases through a data crawler tool, and performing learning of drug properties and drug administration rules on the historical prescriptions by adopting a machine learning algorithm to construct a drug recommendation model;

Optionally, the extracting module 502 includes:

the recognition unit 5021 is used for extracting the patient information, the clinic of the doctor, the chief complaint information, the inquiry dialogue content among the doctors and the patients and the diagnosis result of the doctor recorded in the inquiry list by using a character recognition algorithm;

a configuration unit 5022, configured to prioritize the patient information, the office of visit, the chief complaint information, the content of the inquiry dialogue among the patients of the doctors and the diagnosis results of the doctors according to preset prioritization conditions, so as to obtain a query condition list;

a combining unit 5023, configured to combine the conditions in the query condition list except for the diagnosis result according to the priority level, so as to obtain the feature phrase.

Optionally, the screening module 504 includes:

a filtering unit 5041, configured to filter and screen the candidate drug list according to the condition in the feature phrase, and remove a drug that does not satisfy the patient information to obtain a new drug sequence;

a sorting unit 5042, configured to score and sort each drug in the new drug sequence according to a matching condition between the drug and the questionnaire, so as to obtain a candidate drug sequence;

a selecting unit 5043, configured to select top-ranked N drugs from the candidate drug sequence, so as to obtain a drug recommendation result.

The medication recommendation device further includes a query module 506, which is specifically configured to:

Optionally, the sorting unit 5042 is specifically configured to:

In this embodiment, the training module 501 is specifically configured to:

and deeply learning the medicines in the training set and medicine characteristic information corresponding to the medicines by using a machine learning algorithm, and constructing a medicine recommendation model.

In this embodiment, the medication recommending apparatus further includes an optimizing module 507, which is specifically configured to:

The medication recommending device in the embodiment of the present invention is described in detail in the above fig. 5 and fig. 6 from the perspective of a modular functional entity, and the medication recommending device in the embodiment of the present invention is described in detail in the following from the perspective of hardware processing, and the medication recommending device can be set in a plug-in form and implemented by the medication recommending device to provide a doctor with a recommendation of a relevant drug quickly after a diagnosis is completed.

Fig. 7 is a schematic structural diagram of a medication recommendation device 600 according to an embodiment of the present invention, which may have a relatively large difference due to different configurations or performances, and may include one or more processors (CPUs) 610 (e.g., one or more processors) and a memory 620, and one or more storage media 630 (e.g., one or more mass storage devices) for storing applications 633 or data 632. Memory 620 and storage medium 630 may be, among other things, transient or persistent storage. The program stored on the storage medium 630 may include one or more modules (not shown), each of which may include a sequence of instructions operating on the medication recommendation device 600. Still further, the processor 610 may be configured to communicate with the storage medium 630 and execute a series of instruction operations in the storage medium 630 on the medication recommendation device 600 to implement the steps of the medication recommendation method described above.

The medication recommendation device 600 may also include one or more power supplies 640, one or more wired or wireless network interfaces 650, one or more input-output interfaces 660, and/or one or more operating systems 631, such as Windows Server, Mac OS X, Unix, Linux, FreeBSD, and so forth. It will be appreciated by those skilled in the art that the medication recommendation device configuration illustrated in FIG. 7 is not intended to be limiting of the medication recommendation devices provided herein, and may include more or fewer components than shown, or some components may be combined, or a different arrangement of components.

The block chain is a novel application mode of computer technologies such as distributed data storage, point-to-point transmission, a consensus mechanism, an encryption algorithm and the like. A block chain (Blockchain), which is essentially a decentralized database, is a string of data blocks associated by using a cryptographic method, and each data block contains information of a batch of network transactions, which is used for verifying the validity (anti-counterfeiting) of the information and generating a next block. The blockchain may include a blockchain underlying platform, a platform product service layer, an application service layer, and the like.

The present invention also provides a computer-readable storage medium, which may be a non-volatile computer-readable storage medium, or a volatile computer-readable storage medium, where instructions are stored in the computer-readable storage medium, and when the instructions are executed on a computer, the instructions cause the computer to perform the steps of the medication recommendation method provided in each of the above embodiments.

It can be clearly understood by those skilled in the art that, for convenience and simplicity of description, the specific working processes of the above-described systems, apparatuses and units may refer to the corresponding processes in the foregoing method embodiments, and are not described herein again.

The integrated unit, if implemented in the form of a software functional unit and sold or used as a stand-alone product, may be stored in a computer readable storage medium. Based on such understanding, the technical solution of the present invention, which is substantially or partly contributed by the prior art, or all or part of the technical solution may be embodied in a software product, which is stored in a storage medium and includes instructions for causing a computer device (which may be a personal computer, a server, or a network device) to perform all or part of the steps of the method according to the embodiments of the present invention. And the aforementioned storage medium includes: various media capable of storing program codes, such as a usb disk, a removable hard disk, a read-only memory (ROM), a Random Access Memory (RAM), a magnetic disk, or an optical disk.

The above-mentioned embodiments are only used for illustrating the technical solutions of the present invention, and not for limiting the same; although the present invention has been described in detail with reference to the foregoing embodiments, it should be understood by those of ordinary skill in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some technical features may be equivalently replaced; and such modifications or substitutions do not depart from the spirit and scope of the corresponding technical solutions of the embodiments of the present invention.

Claims

1. A medication recommendation method is characterized by comprising the following steps:

acquiring historical prescriptions from a plurality of medical databases through a data crawler tool, and learning the drug properties and drug administration rules of the historical prescriptions by adopting a machine learning algorithm to construct a drug recommendation model; wherein, obtain historical prescriptions from a plurality of medical databases through the data crawler instrument to adopt machine learning algorithm to right historical prescriptions carries out the study of property of a medicine and rule of using medicine, and it includes to construct the medicine recommendation model: acquiring historical inquiry data of doctors from a plurality of medical databases through a data crawler tool, extracting historical prescriptions from the historical inquiry data, and extracting disease species corresponding to the historical prescriptions and disease symptoms of patients of the disease species based on a keyword extraction algorithm to form a disease feature library; constructing a corresponding relation between a disease characteristic library and a medicine; based on the corresponding relation, learning the corresponding relation by using a machine learning algorithm to obtain a medicine recommendation model;

inputting the diagnosis result into the medicine recommendation model for medicine matching to generate a candidate medicine list;

according to the received medicine use request, selecting a medicine meeting the requirements of the patient from the medicine recommendation result, generating a prescription and storing the prescription in the inquiry list;

2. The medication recommendation method according to claim 1, wherein said extracting the diagnosis result and the inquiry data in the inquiry sheet comprises:

extracting the patient information, the visiting department, the chief complaint information, the inquiry dialogue contents among the doctors and the patients and the diagnosis result of the doctors recorded in the inquiry list by using a character recognition algorithm;

prioritizing the patient information, the visiting department, the chief complaint information, the inquiry dialogue contents among the doctors and the patients and the diagnosis results of the doctors according to preset prioritization conditions to obtain a query condition list;

and combining the conditions except the diagnosis result in the query condition list according to priority levels to obtain a characteristic phrase.

3. The medication recommendation method according to claim 2, wherein the sorting, screening and filtering the drugs in the candidate drug list according to the inquiry data to obtain a final drug recommendation result comprises:

screening and filtering the candidate drug list according to the conditions in the feature phrase, and eliminating drugs which do not meet the patient information to obtain a new drug sequence;

according to the matching condition of the medicines and the inquiry bill, scoring and sequencing each medicine in the new medicine sequence to obtain a candidate medicine sequence;

4. The medication recommendation method according to claim 3, wherein before the prioritizing the patient information, the visit departments, the chief complaint information, the contents of the inquiry dialogue between the doctors and the patients and the diagnosis results of the doctors according to the preset prioritization conditions to obtain the query condition list, the method further comprises:

5. The medication recommendation method according to claim 4, wherein the step of ranking and sorting each drug in the new drug sequence according to the matching condition between the drug and the questionnaire to obtain the candidate drug sequence comprises:

6. A medication recommendation device, comprising:

the training module is used for acquiring historical prescriptions from a plurality of medical databases through a data crawler tool, and learning drug properties and drug administration rules of the historical prescriptions by adopting a machine learning algorithm to construct a drug recommendation model; wherein, acquire historical prescription from a plurality of medical databases through data crawler instrument to adopt machine learning algorithm to right the historical prescription carries out the study of property of a medicine and rule of using medicine, and it includes to construct the medicine recommendation model: acquiring historical inquiry data of doctors from a plurality of medical databases through a data crawler tool, extracting historical prescriptions from the historical inquiry data, and extracting disease species corresponding to the historical prescriptions and disease symptoms of patients of the disease species based on a keyword extraction algorithm to form a disease feature library; constructing a corresponding relation between a disease characteristic library and a medicine; learning the corresponding relation by utilizing a machine learning algorithm based on the corresponding relation to obtain a medicine recommendation model;

the extraction module is used for acquiring an inquiry list of a current patient and extracting a diagnosis result and inquiry data in the inquiry list;

the recommending module is used for selecting medicines meeting the requirements of the patient from the medicine recommending result according to the received medicine using request, generating a prescription and storing the prescription into the inquiry list;

the optimization module is used for randomly drawing a plurality of inquiry lists and extracting the use condition of the medicine recommendation result in each inquiry list; and if the service condition does not meet the preset acceptance rate, reading all inquiry lists of the medicines in the non-adopted medicine recommendation result, and retraining the medicine recommendation model based on the inquiry lists to obtain the iteratively optimized medicine recommendation model.

7. A medication recommendation device, characterized in that the medication recommendation device comprises: a memory having instructions stored therein and at least one processor, the memory and the at least one processor interconnected by a line;

the at least one processor invokes the instructions in the memory to cause the medication recommendation control device to perform the medication recommendation method of any one of claims 1-5.

8. A computer-readable storage medium, on which a computer program is stored, which, when being executed by a processor, carries out a medication recommendation method according to any one of claims 1-5.