WO2022041729A1 - Medication recommendation method, apparatus and device, and storage medium - Google Patents

Abstract

A medication recommendation method, apparatus and device, and a storage medium, which are applied to the field of smart medical care. The method comprises: inputting a diagnosis result into a drug recommendation model for drug matching, so as to obtain a candidate drug list; performing screening on the basis of the candidate drug list combined with inquiry data, so as to obtain recommended drugs; and finally selecting drugs, that meet conditions, on the basis of the actual condition of a patient, and storing the drugs in an inquiry form for subsequent use and consultation. The automatic drug recommendation is realized, and the phenomenon that a physician prescribes a useless drug due to the false memory of drug properties is also avoided. When issuing a prescription, the physician can quickly determine a corresponding treatment prescription according to recommended drugs in an inquiry form, thereby improving the diagnosis efficiency for physicians.

Description

Recommended methods, devices, equipment and storage media for medication

This application claims the priority of the Chinese patent application with the application number 202010896613.3 and the invention title "Method, Apparatus, Equipment and Storage Medium for Recommending Medication", which was filed with the China Patent Office on August 31, 2020, the entire contents of which are incorporated by reference in Applying.

technical field

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

Background technique

With the continuous development of artificial intelligence technology, this technology has received attention and use in various fields. For different application scenarios, various fields have developed corresponding intelligent systems based on artificial intelligence technology to help people improve work efficiency. The medical system in hospitals also No exception. The inventor realized that since the drug storage in the hospital is calculated in the quantity of tens of thousands, every time a doctor selects a drug that meets the patient's condition and diagnosis result, he needs to find a drug in a large-scale drug library that meets the patient's current situation. The drugs required for treatment and the drug names, descriptions, inventory and other information of the drugs. In this regard, in order to help doctors to query drugs from the drug library, the current method is to query the drug-related information in the drug management system based on the keywords of the drugs. The query function can only be queried based on the drug name or specific keywords. The query process depends on the doctor's experience and memory. In such a large-scale drug library, it is difficult for doctors to remember the drug name and efficacy of each drug. , dosage and other drug information, doctors use this method of drug selection based on drug name or keyword query, the efficiency of prescribing drugs is low, and the pertinence and accuracy of prescribing drugs are difficult to guarantee.

SUMMARY OF THE INVENTION

The main purpose of this application is to solve the technical problem of low drug prescribing efficiency due to the difficulty of doctors inquiring about drugs in the existing medical system.

In order to achieve the above purpose, a first aspect of the present application provides a method for recommending medication, including: obtaining historical prescriptions from multiple medical databases through a data crawler tool, and using machine learning algorithms to perform drug properties and medication rules on the historical prescriptions. Learning and constructing a drug recommendation model; obtaining the current patient's medical questionnaire, and extracting the diagnosis results and data from the medical questionnaire; inputting the diagnosis results into the drug recommendation model for drug matching, and generating candidates Drug list; sort and filter the drugs in the candidate drug list according to the consultation data to obtain the final drug recommendation result; according to the received drug use request, select the drug recommendation results that satisfy the patient's requirements Medicines are generated, and the prescriptions are generated and stored in the questionnaire.

A second aspect of the present application provides a medication recommendation device, comprising a memory, a processor, and computer-readable instructions stored on the memory and executable on the processor, and the processor executes the computer-readable instructions The following steps are implemented when instructing: obtaining historical prescriptions from multiple medical databases through a data crawler tool, and using machine learning algorithms to learn the medicinal properties and medication rules of the historical prescriptions to build a drug recommendation model; obtaining the current patient's consultation form , and extract the diagnosis result and inquiry data in the inquiry form; input the diagnosis result into the drug recommendation model for drug matching, and generate a candidate drug list; The drugs in the list are sorted and filtered to obtain the final drug recommendation result; according to the received drug use request, the drug that satisfies the patient is selected from the drug recommendation result, and the prescription is generated and stored in the consultation sheet.

A third aspect of the present application provides a computer-readable storage medium, where computer instructions are stored in the computer-readable storage medium, and when the computer instructions are executed on a computer, the computer is caused to perform the following steps: Obtain historical prescriptions from multiple medical databases, and use machine learning algorithms to learn the drug properties and medication rules of the historical prescriptions, and build a drug recommendation model; obtain the current patient's medical questionnaire, and extract the diagnosis in the medical questionnaire results and consultation data; input the diagnosis results into the drug recommendation model for drug matching to generate a candidate drug list; sort, filter and filter the drugs in the candidate drug list according to the consultation data to obtain the final According to the received drug use request, the drug that satisfies the patient is selected from the drug recommendation result, and the prescription is generated and stored in the consultation list.

A fourth aspect of the present application provides a medication recommendation device, including: a training module for obtaining historical prescriptions from multiple medical databases through a data crawler tool, and using machine learning algorithms to perform drug properties and medication rules on the historical prescriptions learning and constructing a drug recommendation model; an extraction module, used to obtain the current patient's medical questionnaire, and extract the diagnosis results and data in the medical questionnaire; a matching module, used to input the diagnosis results into the In the drug recommendation model, drug matching is performed to generate a candidate drug list; the screening module is used to sort, filter and filter the drugs in the candidate drug list according to the consultation data to obtain the final drug recommendation result; the recommendation module is used for According to the received drug use request, the drug that satisfies the patient is selected from the drug recommendation result, and the prescription is generated and stored in the medical questionnaire.

In the technical solution provided by this application, the model is used to implement drug recommendation to improve the efficiency of doctors in prescribing drugs. Specifically, the diagnostic data of the patient is obtained by the doctor, and the diagnostic data is input into the drug recommendation model, and the drug recommendation model is based on the diagnosis. The data recommends targeted medicines for doctors to choose and use. This method not only realizes the automatic recommendation of medicines, but also avoids the phenomenon of useless medicines caused by doctors' false memory of the medicinal properties of medicines.

Description of drawings

Fig. 1 is the schematic diagram of the first embodiment of the method for recommending medication in the embodiment of the application;

2 is a schematic diagram of the second embodiment of the method for recommending medication in the embodiment of the application;

3 is a schematic diagram of the third embodiment of the method for recommending medication in the embodiment of the application;

4 is a schematic diagram of the fourth embodiment of the method for recommending medication in the embodiment of the application;

FIG. 5 is a schematic diagram of an embodiment of the medication recommendation device in the embodiment of the present application;

6 is a schematic diagram of another embodiment of the medication recommendation device in the embodiment of the present application;

FIG. 7 is a schematic diagram of an embodiment of a medication recommendation device in an embodiment of the present application.

detailed description

In view of the defects in the prior art, the present application proposes an intelligent drug recommendation method with self-learning ability. The information such as the doctor's diagnosis result automatically recommends candidate drugs for the doctor that are suitable for the patient's current medication needs; this method helps the doctor reduce the time consumed by the drug query process by recommending the candidate drug for the doctor, and improves the efficiency of the doctor's prescription; this method recommends drugs. , extract the characteristic information of the consultation form from the patient's personal information, the patient's chief complaint, the consultation dialogue and the doctor's diagnosis result, and use the pre-trained intelligent drug recommendation model to recommend candidate drugs. The recommended drugs consider the patient's personal information and chief complaint content. and diagnosis results, which are more compatible with the patient’s current treatment scene, helping doctors to improve the pertinence of prescribing drugs; at the same time, because the drug push method considers the patient’s age, gender, pregnancy, allergies, contraindications, etc., when recommending candidate drugs, helping doctors Avoid prescribing medicines that conflict with the patient's situation, improving the safety of physicians prescribing medicines.

The terms "first", "second", "third", "fourth", etc. (if any) in the description and claims of this application and the above-mentioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It is to be understood that data so used may be interchanged under appropriate circumstances so that the embodiments described herein can be practiced in sequences other than those illustrated or described herein. Furthermore, the terms "comprising" or "having" and any variations thereof are intended to cover non-exclusive inclusion, for example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those expressly listed steps or units, but may include other steps or units not expressly listed or inherent to these processes, methods, products or devices.

For ease of understanding, the specific flow of the embodiment of the present application will be described below. Please refer to FIG. 1 . The first embodiment of the method for recommending medication in the embodiment of the present application includes:

101. Obtain historical prescriptions from multiple medical databases through data crawler tools, and use machine learning algorithms to learn the medicinal properties and medication rules of historical prescriptions, and build a drug recommendation model;

In this step, the doctor's historical consultation data can be obtained from multiple medical databases through a data crawler tool, and historical prescriptions can be extracted from the historical consultation data to construct a medicine push data set;

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

A machine learning algorithm is used to perform deep learning on the medicines in the training set and the medicine characteristic information corresponding to the medicines to construct a medicine recommendation model.

In practical applications, the drug recommendation model can be constructed by using the XGBOOST algorithm and the neural network deep learning algorithm. The specific implementation steps include:

First, the prescription is extracted through the filling template of the diagnosis sheet. For example, the prescription filling field specified in the template is used to identify the prescription filling field of the diagnosis sheet in the historical consultation data, and the information is read based on the corresponding relationship between the filling field and the content location of the prescription, and the generated historical prescriptions;

Then, the TextRank keyword extraction algorithm is used to extract the keywords in the prescription, such as the name of the drug and the usage and dosage in the drug; in practical applications, when extracting the name of the drug, the knowledge map of the drug can also be combined, based on the knowledge map of the drug. The recorded entity compares and identifies the entity name of the drug name in the historical prescription, so as to extract the name of the drug;

Further, based on the keyword extraction algorithm, the disease type corresponding to the historical prescription and the onset symptoms of the patient in the medical record are extracted to form a disease feature database;

Construct the corresponding relationship between the disease feature library and the drug;

Based on the corresponding relationship, a machine learning algorithm is used to learn the corresponding relationship to obtain a drug recommendation model.

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

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

Based on the training data set, use machine learning algorithms, such as neural network deep learning algorithms, to learn drugs and medication rules, and obtain a drug recommendation model;

Use the verification data set to verify the drug recommendation model. After the verification is passed, the final drug recommendation model is output; if it fails, the training data set and the verification data set are scrambled and reassigned for retraining.

102. Obtain the medical questionnaire of the current patient, and extract the diagnosis results and data of the medical consultation in the medical questionnaire;

In this step, the consultation sheet refers to a disease diagnosis receipt issued by a doctor when the patient seeks a doctor, and the consultation sheet includes at least patient information, consultation data and diagnosis results, and the patient information includes the gender of the patient , age, allergy history, contraindications and pregnancy (female patients) and other information, the consultation data is some symptom information, disease history that the doctor asks the user when the patient seeks a doctor, and the disease information that the patient voluntarily stated to the doctor, etc.

In practical applications, when obtaining a medical consultation form, it monitors the operation information and input information on each doctor's prescription interface, generates a corresponding medical consultation form based on the monitored information, and then analyzes the medical consultation form. The content in the specified extraction.

Of course, when extracting the diagnostic results and data from the medical questionnaire, in addition to the above-mentioned methods for extracting the diagnostic results and data from the medical questionnaire, it is also possible to monitor the doctor's prescription page. The extraction is specifically performed by monitoring information on specific items, for example, monitoring and identifying only the input box at the position of the diagnosis result item and monitoring and identifying the input box at the position of the doctor's question item.

In practical applications, the extraction of medical consultation data can be extracted in real time, specifically from the dialogue between the doctor and the patient, in addition to the data obtained from the medical consultation sheet.

103. Input the diagnosis result into the drug recommendation model for drug matching, and generate a list of candidate drugs;

In this embodiment, when performing drug matching, the drug recommendation model searches from the drug repository according to the input diagnosis result for all the drugs whose pharmacology or properties conform to the diagnosis result, and then calculates the queried drugs one by one. The matching degree between the drug and the diagnosis result, based on the matching degree, select the drug whose matching degree is greater than the preset percentage as the data structure of the drug recommendation model, and generate the candidate drug list.

In practical applications, while generating the candidate drug list, it also includes sorting the selected drugs according to the matching degree from large to small, so as to facilitate the subsequent selection and use of the drugs.

Further, after generating the candidate drug list, this step also includes extracting the medicinal properties of each drug in the candidate drug list, and judging the risk level of each medicinal property during use, and then marking according to the risk level for subsequent identification. Choice of use of medicines.

In practical applications, since there are multiple properties of medicines, each property may be used for the treatment of one or more diseases, and the determination of the risk level is specifically determined according to the current diagnosis results, that is to say According to the specific disease, the degree of influence of its chemical reaction on the human body is determined, so as to determine the corresponding risk level.

104. Sort, filter and filter the drugs in the candidate drug list according to the consultation data to obtain a final drug recommendation result;

In practical applications, this step can be achieved by extracting characteristic phrases in the consultation data for matching and filtering. Specifically, a model can be used to achieve screening and filtering. For example, a drug matching model is used. The drug matching model includes the mapping relationship of various drug features. The mapping relationship can be, but is not limited to, drug feature words extracted from information such as drug names, drug numbers, objects of use, usage, functions, dosages, and contraindications of various drugs, and the corresponding drug can be uniquely determined through the mapping relationship. The drug matching model can realize the feature matching between the feature phrase and the drug feature, which can perform drug matching according to the input feature phrase, and output the matched drug. Specifically, the drug matching model may be a naive Bayesian probability model obtained based on a Bayesian algorithm, which can count the probability of each drug according to the input feature phrase, and output the drug with the highest probability.

In another embodiment, the drug matching model may also be based on a drug matching neural network obtained by an artificial neural network algorithm, and the drug matching neural network may be a multi-layer structure. For example, if the feature word group is divided into three priority levels, high, medium and low, the drug matching neural network can be correspondingly set to a structure of three hidden layers to correspond to the priority division of the feature word group.

In the specific implementation, the drug matching models corresponding to the functional departments of each hospital may be different. In this case, the drug matching model corresponding to the functional departments of the hospital can be queried first, and then the feature phrases are input for feature matching to obtain the corresponding output results.

105. According to the received drug use request, from the drug recommendation result, select a drug that satisfies the patient, and generate a prescription and save it in the consultation list.

In this embodiment, the obtained drug recommendation result is displayed on the prescription position in the doctor's prescription interface. The doctor selects the displayed drug, generates a prescription based on the selected drug, and adds it to the consultation list , so as to facilitate subsequent drug packaging.

In practical applications, the prescription in this step can also be understood as a prescription drug list or an over-the-counter drug list, and the prescription drug list is a list of drugs issued by a doctor for a patient, a written document for the doctor to administer the current patient, and is used by pharmacists to allocate drugs. Based on this, the prescription recommendation can be used as a reference when a doctor prescribes a prescription. In particular, if the list of drugs in the prescription recommendation is appropriate, it can be used directly as a prescription. After the drug matching model performs feature matching according to the input feature phrases, the output matching results include various drugs, generate a drug list according to the drugs, and obtain prescription recommendations. After the doctor, the doctor selects one of them according to the actual situation, and finally the system automatically saves it in the consultation list for recording.

Through the embodiment of the above method, by obtaining the diagnosis data of the doctor on the patient, the diagnosis data is input into the drug recommendation model, and the drug recommendation model recommends targeted drugs for the doctor to select and use based on the diagnosis data. This method not only realizes The automatic recommendation of medicines also avoids the phenomenon of useless medicines caused by doctors' false memory of the medicinal properties of medicines;

Further, the recommended medicines can be used for reference by doctors when prescribing, so that the prescription generation process does not require doctors to comprehensively consider various factors to perform complicated operations, simplifies the generation process of medical prescriptions, and avoids the problem of repeated prescription revisions due to doctor errors. , improving the efficiency of prescription generation.

Please refer to Fig. 2, the second embodiment of the method for recommending medication in the embodiment of the present application includes:

201. Use a text recognition algorithm to extract the patient information, visiting department, chief complaint information, the content of the consultation dialogue between the doctor and the patient, and the doctor's diagnosis result recorded in the consultation sheet;

In this step, when using the text recognition algorithm for recognition, it can be realized by obtaining the template of the medical consultation form. The template is set with multiple field information, and the corresponding setting and filling content of different field information settings, such as the corresponding name field Fill in the name, the identity field corresponds to the ID number, and the diagnosis content field corresponds to the diagnosis data and consultation data.

Based on the field information in the template, use the text recognition algorithm to identify each field in the medical consultation form, and then extract the attribute value of the corresponding field in the medical consultation form according to the corresponding position relationship between the field and the filled content, and determine based on the attribute value. Patient information, visiting department, chief complaint information, the content of the consultation dialogue between the doctor and the patient, and the diagnosis result.

In practical applications, the content of the consultation dialogue and the diagnosis results between the doctor and the patient can be matched and identified by using the disease knowledge graph, specifically, multiple key features in the corresponding field information are extracted, and the key features and the disease The entities recorded in the knowledge graph are paired and identified, and then the corresponding symptoms are selected according to the identification results, and the content corresponding to the symptoms and the field information is continued to compare, so as to obtain the final result, such as the content of the consultation dialogue, first from which After extracting several key features, match all the symptoms of the disease with several key features from the knowledge graph, and then return the content of the consultation dialogue based on all symptoms for full-text matching, so as to obtain an effective and complete consultation dialogue content.

202. According to a preset priority classification condition, prioritize the patient information, the visiting department, the chief complaint information, the content of the consultation dialogue between the doctor and the patient, and the doctor's diagnosis result, to obtain a query condition list;

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

In practical applications, the conditional information in the medical questionnaire is extracted, specifically, pathological feature word extraction can be performed on the information recorded in the medical questionnaire based on the TextRank keyword extraction algorithm. For example, if the medical questionnaire includes the expression "cold leads to a cold, fever, headache, and nasal congestion", the verbs "cause", "cause", "appear" and other redundant data can be eliminated, and "catch cold" can be extracted. , "cold", "fever", "headache" and "nasal congestion" and other core pathological feature words, based on the case feature words to determine the diagnosis results in the inquiry form, and by extracting the pathological feature words from the inquiry, it can be effectively eliminated Redundant and useless data in order to ensure the processing efficiency and accuracy of subsequent prescription generation.

Further, the inquiry data in the extraction inquiry form is the content of the doctor's question or the patient's statement, such as "whether I am in this hospital for medical treatment or surgery", "whether there is an allergy history or whether there is any drug allergy", through the TextRank key The word extraction algorithm extracts the words "medical treatment in this hospital", "surgery", "allergic history" and "allergic medicine", and even includes keyword extraction processing such as patient information, visiting department, diagnosis doctor, etc., according to the preset Keywords Sort and combine these keywords to obtain characteristic phrases.

204. Input the diagnosis result into the drug recommendation model to perform drug matching, and generate a candidate drug list;

In this step, the specific drug recommendation model inquires from the drug repository according to the input diagnosis results all the drugs whose pharmacology or properties conform to the diagnosis results, and then calculates the queried drugs and the diagnosis results one by one. The matching degree between the two, and based on the matching degree, a drug with a matching degree greater than a preset percentage is selected as the data structure of the drug recommendation model, and a candidate drug list is generated.

205. Screening and filtering the candidate drug list according to the conditions in the characteristic phrase, eliminating the drugs whose drugs do not meet the patient information, and obtaining a new drug sequence;

In this step, the candidate drug list is firstly screened and filtered according to the priority according to the conditions in the characteristic phrase, specifically, the part of the drug that does not have a high matching degree for the diagnosis result is screened out, and then the screening is performed. The candidate drugs are further eliminated according to the patient information, and the drugs that do not satisfy the patient's use in pharmacology are eliminated, and the last drug is left to form a new drug sequence. At this time, the drug sequence takes the patient information as the highest priority, and the condition in the characteristic phrase Sort by priority.

206. Score and sort each drug in the new drug sequence according to the matching situation between the drug and the medical questionnaire to obtain a candidate drug sequence;

In practical applications, the matching situation refers to the degree of matching between information such as the object of use, usage, function, dosage, and contraindication of the drug and the specific information of the patient, wherein the patient information includes age, gender, pregnancy status, and allergy history. and contraindications, by comparing whether the matching degree of the object of use, usage, function, dosage and contraindication of the drug in the new drug sequence and the age, sex, pregnancy, allergy history and contraindication in the patient information meets the preset threshold, the The matching situation is input into the scoring model for the scoring of recommendation prediction, and the scoring score is obtained, and the new drug sequence is sorted based on the scoring score to obtain the candidate drug sequence.

207. Select the top N drugs from the candidate drug sequence to obtain a drug recommendation result;

In practical applications, when selecting drugs according to the ranking, it also includes checking whether the inventory quantity of each drug is sufficient. If it is not enough, the next drug in the ranking will be filled until the number of N drugs is completed. Of course, in When selecting N drugs, in addition to considering the inventory, it also includes judging whether the drug has special pharmacological side effects, such as the presence of high antibiotics or high pain-relieving ingredients, and further, it can also be based on preference. The rule is selected as the main, followed by Western medicine, and finally the medicines that meet all the above conditions are output.

208. According to the received drug use request, from the drug recommendation result, select a drug that satisfies the patient, and generate a prescription and save it in the consultation list.

In this embodiment, the selection of suitable medicines for patients can be specifically realized through a medicine matching model, and the specific medicine matching model is a medicine matching model obtained based on an artificial neural network algorithm. For example, if the feature word group is divided into three priority levels: high, medium and low, the drug matching neural network can be set to a three-layer hidden layer structure corresponding to the priority of the feature word group. correspond.

In the specific implementation, the drug matching model corresponding to each patient may be different. In this case, the drug matching model corresponding to the disease type of the patient can be queried first, and then the characteristic phrases of the extracted disease type can be input for feature matching to obtain the corresponding The output results, and the range of the model matching drugs is the range of the drug recommendation results provided above to be matched.

Through the implementation of the above method, by inputting the diagnosis results into the drug recommendation model for drug matching, a list of candidate drugs is obtained, and based on the candidate drug list combined with the consultation data, the recommended drugs are obtained, and finally the selection is based on the actual situation of the patient. Eligible medicines are stored in the consultation sheet for subsequent use and reference; this method not only realizes the automatic recommendation of medicines, but also avoids the phenomenon of useless medicines caused by doctors' false memory of the medicinal properties of medicines. When prescribing a prescription, the physician can quickly determine the corresponding treatment prescription according to the recommended drugs in the consultation sheet, thereby improving the physician's diagnostic efficiency.

Please refer to FIG. 3 , the third embodiment of the method for recommending medication in the embodiments of the present application includes:

301. Obtain the medical consultation form of the current patient, and extract the diagnosis results and medical consultation data in the medical consultation document;

302. Input the diagnosis result into the drug recommendation model to perform drug matching, and generate a candidate drug list;

303. Query the corresponding patient diagnosis history record from the diagnosis database of the hospital diagnosis system according to the patient information;

304. Determine the patient's allergy history and contraindication information according to the history of diagnosis;

In this embodiment, the screening of allergens is mainly to ensure the safety of prescribing the prescription and the safety of the patient's life, specifically through the query in historical diagnosis records.

Specifically, the process of inquiring the history of medical treatment in the diagnosis database is started by identifying whether there is a characteristic word for medical treatment in the hospital in the diagnosis data. Information or medical insurance card information to query the corresponding medical history records from the diagnosis database of the hospital system, extract the information of the doctor's remarks from the medical history records, and determine the corresponding precautions based on the information of the remarks, that is, Allergy history and contraindication information, or even the diagnosis results of historical medical treatment, further screening of candidate drug sequences based on the query results.

305. Screen and filter the candidate drug list according to the conditions in the characteristic phrase, and remove the drugs whose drugs do not meet the patient information to obtain a new drug sequence;

306. Perform secondary screening and filtering on the new drug sequence according to the allergy history and contraindication information to obtain a second drug sequence;

307. Use a preset scoring model to score the matching degree of each drug in the second drug sequence, and sort the scores from high to low to obtain a candidate drug sequence;

Specifically, the pathological feature words extracted from the medical history records and the file feature words extracted from the patient's personal file information involve many categories, and the influence weights of each category of feature words on the final prescription generation are different. For 18-year-old minor patients, the drugs for adults are not suitable. At this time, the age of the patient has a great influence on the prescription drugs, and the priority is high; for patients whose gender is male, the drugs for gynecological diseases are used. It is also not applicable. At this time, the gender of the patient has a high priority; for another example, when the patient's disease site is the stomach, the drugs targeting the brain or kidneys and other organs are also not suitable as the corresponding prescription drugs. In this embodiment, the pathological feature words and the archive feature words are prioritized to distinguish the weights of each category of feature words when generating prescriptions, so as to improve the efficiency and accuracy of drug matching.

In one embodiment, the pathological feature word may further include disease location, disease name and symptom expression, and the file feature word includes patient object, allergen and past medical history; the step of prioritizing the pathological feature word and the file feature word Including: classifying patient objects and allergens as high-priority feature words; classifying disease location and disease name as medium-priority feature words; classifying symptoms and past medical history as low-priority feature words.

308. Select the top N drugs from the candidate drug sequence to obtain a drug recommendation result;

309. According to the received drug use request, from the drug recommendation result, select a drug that satisfies the patient, and generate a prescription and save it in the consultation list.

To sum up, the method of this embodiment automatically learns the drug recommendation model from the doctor's historical medication data; extracts the feature information of the consultation sheet from the consultation information such as the patient's personal information, the main complaint, the content of the consultation dialogue, the doctor's diagnosis result, etc. The trained drug recommendation model recommends candidate drugs for doctors; the drug recommendation process considers the patient's gender, age, contraindication, pregnancy and other information to avoid conflicts between the recommended drugs and the patient's own situation; the drug recommendation method realizes the automatic recommendation of drugs, helping Doctors improve the efficiency of prescribing, and at the same time help doctors avoid medication errors caused by false memory of drug properties, and improve the accuracy and safety of medication.

In practical applications, in addition to recommending medicines to physicians through the above methods, recommendations can also be made by training a model, which can be more quickly identified. Please refer to FIG. 4 . A fourth embodiment of the recommended method includes:

401. Acquire historical consultation data of doctors from multiple medical databases through a data crawler tool, and extract historical prescriptions from the historical consultation data to construct a medicine push data set;

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

403. Use a machine learning algorithm to perform deep learning on the medicines in the training set and the medicine characteristic information corresponding to the medicines, to construct a medicine recommendation model;

In the present embodiment, the historical consultation data can be a doctor's handwritten diagnosis sheet, or it can be obtained by using this method. Specifically, it can be realized through the following steps:

Step 1, feature extraction: extract the features of the consultation from the information of the patient (including age, gender, pregnancy, allergy history, contraindications), the department of consultation, the content of the main complaint, the content of the consultation dialogue, and the doctor's diagnosis result. ;

Step 2, model training: the input of the model is the characteristics of the consultation, and the prediction target is the candidate drug for the target consultation sheet. The task of the model training module is to use the historical inquiries and prescription results as training data, and use machine learning algorithms to train and build a drug recommendation model that predicts candidate drugs;

Further, test whether the drug recommendation model meets the specified online conditions, and if the specified online conditions are met, the drug recommendation model obtained online;

Then, perform drug recommendation based on the online drug recommendation model: for the given consultation data, use the trained drug prediction model to predict the list of candidate drugs suitable for the consultation form; specifically, collect the patient information input into the consultation form through the collection (including age, gender, pregnancy status, allergy history, contraindications), visiting department, chief complaint information, the content of the consultation dialogue between doctors and patients, the doctor's diagnosis results and other information, use the method of step 1 to extract the characteristics of the consultation sheet; then The special diagnosis is input into the drug recommendation model for prediction, and the prediction result is obtained.

Further, candidate drug screening and sorting: according to the patient's age, gender, pregnancy, allergy history, contraindications and other conditions in the medical questionnaire, the candidate drugs recommended by the drug recommendation module are screened, and then based on the matching of the drug and the medical questionnaire Scoring and sorting of candidate drugs after screening;

Finally, according to the scoring results, it is pushed to the doctor to choose the use of drugs, and the doctor's use of the recommended results is recorded.

404. Obtain the medical consultation form of the current patient, and extract the diagnosis results and medical consultation data in the medical consultation document;

405. Input the diagnosis result into the drug recommendation model to perform drug matching, and generate a candidate drug list;

406. Sort and filter the drugs in the candidate drug list according to the consultation data to obtain a final drug recommendation result;

407. According to the received drug use request, select the drug that satisfies the patient from the drug recommendation result, and generate a prescription and save it in the consultation list;

408. Randomly select several medical questionnaires, and extract the usage of the drug recommendation results in each medical questionnaire;

409. If the usage does not meet the preset adoption rate, read the consultation sheets of all the drugs that are not used in the drug recommendation results, and retrain the drug recommendation model based on the consultation sheets to obtain iteratively optimized drug recommendations Model.

In this embodiment, the use of medicines may have different results and prescription ratios according to the staged medical treatment of patients. In order to adapt to the recommendation of medicines in real time, it also includes:

Randomly select a number of the medical questionnaires, and extract the usage of the drug recommendation results in each of the medical questionnaires;

If the usage does not meet the preset adoption rate, read the questionnaires of all the medicines that are not used in the drug recommendation results, and retrain the drug recommendation model based on the questionnaires to obtain iterative optimization The post-drug recommendation model.

That is to record the use of the drugs recommended by the doctor's drug recommendation model each time. If the doctor does not accept the recommended results, save the information of the medical questionnaire. When the number of saved non-adopted medical questionnaires reaches the given condition, re-execute Phase 1. Perform model training and iteratively optimize the drug recommendation model.

Through the implementation of the above scheme, in the process of prescribing the medicine, the doctor can automatically recommend candidate medicines suitable for the patient's current medication needs according to the patient's personal information, the content of the main complaint, the content of the doctor's patient's consultation dialogue, the doctor's diagnosis result and other information; Method By recommending candidate drugs for doctors, it helps doctors to reduce the time consumed by the drug query process and improve the efficiency of doctors' prescribing. When recommending drugs, the method extracts questions from patients' personal information, patient complaints, consultation dialogues and doctors' diagnosis results. The feature information of the medical bill, and the pre-trained intelligent drug recommendation model is used to recommend candidate drugs. The recommended drugs take into account the patient's personal information, main complaints and diagnosis results, which are more suitable for the patient's current medical treatment scene, helping doctors to improve their prescribing experience. Targeted; at the same time, because the drug push method considers the patient's age, gender, pregnancy, allergies, contraindications, etc. when recommending candidate drugs, it helps doctors avoid prescribing drugs that conflict with the patient's situation, and improves the safety of doctors prescribing drugs .

The method for recommending medication in the embodiment of the present application has been described above. The following describes the device for recommending medication in the embodiment of the present application. Please refer to FIG. 5 . The first embodiment of the device for recommending medication in the embodiment of the present application includes:

The training module 501 is used for obtaining historical prescriptions from multiple medical databases through a data crawler tool, and using a machine learning algorithm to learn the properties and medication rules of the historical prescriptions to build a drug recommendation model;

The extraction module 502 is used to obtain the medical questionnaire of the current patient, and extract the diagnosis result and the medical consultation data in the medical questionnaire;

A matching module 503, configured to input the diagnosis result into the drug recommendation model for drug matching, and generate a candidate drug list;

The screening module 504 is configured to sort, filter and filter the drugs in the candidate drug list according to the consultation data to obtain a final drug recommendation result;

The recommendation module 505 is configured to select a medicine that satisfies the patient from the medicine recommendation result according to the received medicine use request, and generate a prescription and save it in the consultation list.

In this embodiment, the drug recommendation device runs the above-mentioned drug recommendation method, the method performs drug matching by inputting the diagnosis result into the drug recommendation model, obtains a list of candidate drugs, and performs screening based on the candidate drug list combined with the consultation data, Obtain recommended drugs, and finally select qualified drugs based on the actual situation of the patient, and save them in the consultation sheet for subsequent use and review; this method not only realizes the automatic recommendation of drugs, but also avoids doctors When prescribing a prescription, the physician can quickly determine the corresponding treatment prescription according to the recommended drug in the consultation sheet, thereby improving the physician's diagnosis efficiency.

Please refer to FIG. 6 , the second embodiment of the medication recommendation device in the embodiment of the present application, the medication recommendation device specifically includes:

The training module 501 is used for obtaining historical prescriptions from multiple medical databases through a data crawler tool, and using a machine learning algorithm to learn the properties and medication rules of the historical prescriptions to build a drug recommendation model;

The extraction module 502 is used to obtain the medical questionnaire of the current patient, and extract the diagnosis result and the medical consultation data in the medical questionnaire;

A matching module 503, configured to input the diagnosis result into the drug recommendation model for drug matching, and generate a candidate drug list;

The screening module 504 is configured to sort, filter and filter the drugs in the candidate drug list according to the consultation data to obtain a final drug recommendation result;

The recommendation module 505 is configured to select a medicine that satisfies the patient from the medicine recommendation result according to the received medicine use request, and generate a prescription and save it in the consultation list.

Optionally, the extraction module 502 includes:

The identification unit 5021 is used for extracting the patient information, the visiting department, the chief complaint information, the content of the consultation dialogue between the doctor and the patient, and the doctor's diagnosis result recorded in the consultation sheet by using a character recognition algorithm;

The configuration unit 5022 is configured to prioritize the patient information, the visiting department, the chief complaint information, the content of the consultation dialogue between the doctor and the patient, and the doctor's diagnosis result according to the preset priority classification conditions to obtain a query condition list;

The combining unit 5023 is configured to combine the conditions in the query condition list except the diagnosis result according to the priority level to obtain the characteristic phrase.

Optionally, the screening module 504 includes:

A filtering unit 5041, configured to screen and filter the candidate drug list according to the conditions in the characteristic phrase, and remove the drugs whose drugs do not meet 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 the matching situation between the drug and the medical questionnaire, to obtain a candidate drug sequence;

The selecting unit 5043 is configured to select the top N drugs from the candidate drug sequence to obtain a drug recommendation result.

Wherein, the drug recommendation device further includes a query module 506, which is specifically used for:

According to the patient information, query the corresponding patient diagnosis history record from the diagnosis database of the hospital diagnosis system;

Based on the diagnostic history, the patient's allergy history and contraindication information are determined.

Optionally, the sorting unit 5042 is specifically used for:

Perform secondary screening and filtering on the new drug sequence according to the allergy history and contraindication information to obtain a second drug sequence;

The matching degree of each drug in the second drug sequence is scored using a preset scoring model, and the scores are sorted from high to low to obtain a candidate drug sequence.

In this embodiment, the training module 501 is specifically used for:

Obtain the doctor's historical consultation data from multiple medical databases through a data crawler tool, and extract historical prescriptions from the historical consultation data to construct a drug push data set;

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

A machine learning algorithm is used to perform deep learning on the medicines in the training set and the medicine characteristic information corresponding to the medicines to construct a medicine recommendation model.

In this embodiment, the medication recommendation device further includes an optimization module 507, which is specifically used for:

Randomly select a number of the medical questionnaires, and extract the usage of the drug recommendation results in each of the medical questionnaires;

If the usage does not meet the preset adoption rate, read the questionnaires of all the medicines that are not used in the drug recommendation results, and retrain the drug recommendation model based on the questionnaires to obtain iterative optimization The post-drug recommendation model.

5 and 6 above describe in detail the medication recommendation device in the embodiment of the present application from the perspective of a modular functional entity, and the medication recommendation device in the embodiment of the present application is described in detail below from the perspective of hardware processing, and the medication recommendation device can be plug-in. The form setting and the drug recommendation device realize that after the diagnosis is completed, the doctor can quickly provide the doctor with the recommendation of related drugs.

FIG. 7 is a schematic structural diagram of a medication recommendation device provided by an embodiment of the present application. The medication recommendation device 700 may vary greatly due to different configurations or performances, and may include one or more processors (central processing units, CPUs) ) 710 (eg, one or more processors) and memory 720, one or more storage media 730 (eg, one or more mass storage devices) that store applications 733 or data 732. Among them, the memory 720 and the storage medium 730 may be short-term storage or persistent storage. The program stored in the storage medium 730 may include one or more modules (not shown in the figure), and each module may include a series of instruction operations on the medication recommendation device 700 . Further, the processor 710 may be configured to communicate with the storage medium 730, and execute a series of instruction operations in the storage medium 730 on the medication recommendation device 700 to implement the steps of the above-mentioned medication recommendation method.

Medication recommendation device 700 may also include one or more power supplies 740, one or more wired or wireless network interfaces 750, one or more input and output interfaces 760, and/or, one or more operating systems 731, such as Windows Server, Mac OS X, Unix, Linux, FreeBSD and more. Those skilled in the art can understand that the structure of the medication recommendation device shown in FIG. 7 does not constitute a limitation on the medication recommendation device provided by the present application, and may include more or less components than those shown in the figure, or combine some components, or different component layout.

The blockchain referred to in this application is a new application mode of computer technologies such as distributed data storage, point-to-point transmission, consensus mechanism, and encryption algorithm. Blockchain, essentially a decentralized database, is a series of data blocks associated with cryptographic methods. Each data block contains a batch of network transaction information to verify its Validity of information (anti-counterfeiting) and generation of the next block. The blockchain can include the underlying platform of the blockchain, the platform product service layer, and the application service layer.

The present application also provides a medication recommendation device, comprising: a memory and at least one processor, wherein instructions are stored in the memory, and the memory and the at least one processor are interconnected through a line; the at least one processor calls the The instructions in the memory are used to cause the medication recommendation device to perform the steps in the above-mentioned medication recommendation method.

The present application also provides a computer-readable storage medium, and the computer-readable storage medium may be a non-volatile computer-readable storage medium or a volatile computer-readable storage medium. The computer-readable storage medium stores computer instructions, and when the computer instructions are executed on the computer, the computer performs the following steps:

Obtain historical prescriptions from multiple medical databases through data crawler tools, and use machine learning algorithms to learn the medicinal properties and medication rules of the historical prescriptions to build a drug recommendation model;

Obtain the medical questionnaire of the current patient, and extract the diagnosis result and the medical consultation data in the medical questionnaire;

Inputting the diagnosis result into the drug recommendation model for drug matching, and generating a candidate drug list;

Sorting, screening and filtering the drugs in the candidate drug list according to the consultation data to obtain a final drug recommendation result;

According to the received drug use request, the drug that satisfies the patient is selected from the drug recommendation result, and the prescription is generated and stored in the medical questionnaire.

Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working process of the system, device and unit described above may refer to the corresponding process in the foregoing method embodiments, which will not be repeated here.

The integrated unit, if implemented in the form of a software functional unit and sold or used as an independent product, may be stored in a computer-readable storage medium. Based on this understanding, the technical solutions of the present application can be embodied in the form of software products in essence, or the parts that contribute to the prior art, or all or part of the technical solutions, and the computer software products are stored in a storage medium , including several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), magnetic disk or optical disk and other media that can store program codes .

As mentioned above, the above embodiments are only used to illustrate the technical solutions of the present application, but not to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand: The technical solutions recorded in the embodiments are modified, or some technical features thereof are equivalently replaced; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions in the embodiments of the present application.

Claims (20)

1. A method of drug recommendation that includes:

   Obtain historical prescriptions from multiple medical databases through data crawler tools, and use machine learning algorithms to learn the medicinal properties and medication rules of the historical prescriptions to build a drug recommendation model;

   Obtain the medical questionnaire of the current patient, and extract the diagnosis result and the medical consultation data in the medical questionnaire;

   Inputting the diagnosis result into the drug recommendation model for drug matching, and generating a candidate drug list;

   Sorting, screening and filtering the drugs in the candidate drug list according to the consultation data to obtain a final drug recommendation result;

   According to the received drug use request, the drug that satisfies the patient is selected from the drug recommendation result, and the prescription is generated and stored in the medical questionnaire.
2. The method for recommending medication according to claim 1, wherein the extracting the diagnosis result and the inquiry data in the inquiry form comprises:

   Extract the patient information, visiting department, chief complaint information, the content of the consultation dialogue between the doctor and the patient, and the doctor's diagnosis result recorded in the consultation sheet by using a text recognition algorithm;

   According to the preset priority classification conditions, the patient information, the visiting department, the chief complaint information, the content of the consultation dialogue between the doctor and the patient, and the doctor's diagnosis result are prioritized to obtain a query condition list;

   The conditions in the query condition list except the diagnosis result are combined according to the priority level to obtain the characteristic phrase.
3. The drug recommendation method according to claim 2, wherein the sorting, screening and filtering of the drugs in the candidate drug list according to the consultation data to obtain a final drug recommendation result comprises:

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

   Scoring and sorting each drug in the new drug sequence according to the matching of the drug and the medical questionnaire to obtain a candidate drug sequence;

   The top N drugs are selected from the candidate drug sequence to obtain a drug recommendation result.
4. The method for recommending medication according to claim 3, wherein, in the pre-set priority classification condition, the patient information, the visiting department, the chief complaint information, the content of the consultation dialogue between the doctor and the patient, and the doctor's diagnosis result are divided into Prioritize and obtain the query condition list, including:

   According to the patient information, query the corresponding patient diagnosis history record from the diagnosis database of the hospital diagnosis system;

   Based on the diagnostic history, the patient's allergy history and contraindication information are determined.
5. The drug recommendation method according to claim 4, wherein the scoring and sorting of each drug in the new drug sequence according to the matching situation between the drug and the medical questionnaire, and obtaining the candidate drug sequence comprises:

   Perform secondary screening and filtering on the new drug sequence according to the allergy history and contraindication information to obtain a second drug sequence;

   The matching degree of each drug in the second drug sequence is scored using a preset scoring model, and the scores are sorted from high to low to obtain a candidate drug sequence.
6. The method for recommending medication according to any one of claims 1-5, wherein the historical prescriptions are obtained from a plurality of medical databases through a data crawler tool, and a machine learning algorithm is used to perform the drug properties and medication rules on the historical prescriptions. Learning to build a drug recommendation model includes:

   Obtain the doctor's historical consultation data from multiple medical databases through a data crawler tool, and extract historical prescriptions from the historical consultation data to construct a drug push data set;

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

   A machine learning algorithm is used to perform deep learning on the medicines in the training set and the medicine characteristic information corresponding to the medicines to construct a medicine recommendation model.
7. The drug recommendation method according to claim 6, wherein, according to the received drug use request, the drug that satisfies the patient is selected from the drug recommendation result, and the prescription is generated and stored in the medical questionnaire. ,Also includes:

   Randomly select a number of the medical questionnaires, and extract the usage of the drug recommendation results in each of the medical questionnaires;

   If the usage does not meet the preset adoption rate, read the questionnaires of all the medicines that are not used in the drug recommendation results, and retrain the drug recommendation model based on the questionnaires to obtain iterative optimization The post-drug recommendation model.
8. A medication recommendation device, comprising a memory, a processor, and computer-readable instructions stored on the memory and executable on the processor, and the processor implements the following steps when executing the computer-readable instructions:

   Obtain historical prescriptions from multiple medical databases through data crawler tools, and use machine learning algorithms to learn the medicinal properties and medication rules of the historical prescriptions to build a drug recommendation model;

   Obtain the medical questionnaire of the current patient, and extract the diagnosis result and the medical consultation data in the medical questionnaire;

   Inputting the diagnosis result into the drug recommendation model for drug matching, and generating a candidate drug list;

   Sorting, screening and filtering the drugs in the candidate drug list according to the consultation data to obtain a final drug recommendation result;

   According to the received drug use request, the drug that satisfies the patient is selected from the drug recommendation result, and the prescription is generated and stored in the medical questionnaire.
9. The device for recommending medication according to claim 8, wherein the processor further implements the following steps when executing the computer program:

   Extract the patient information, visiting department, chief complaint information, the content of the consultation dialogue between the doctor and the patient, and the doctor's diagnosis result recorded in the consultation sheet by using a text recognition algorithm;

   According to the preset priority classification conditions, the patient information, the visiting department, the chief complaint information, the content of the consultation dialogue between the doctor and the patient, and the doctor's diagnosis result are prioritized to obtain a query condition list;

   The conditions in the query condition list except the diagnosis result are combined according to the priority level to obtain the characteristic phrase.
10. The medication recommendation device according to claim 9, wherein the processor further implements the following steps when executing the computer program:

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

    Scoring and sorting each drug in the new drug sequence according to the matching of the drug and the medical questionnaire to obtain a candidate drug sequence;

    The top N drugs are selected from the candidate drug sequence to obtain a drug recommendation result.
11. The device for recommending medication according to claim 10, wherein the processor further implements the following steps when executing the computer program:

    According to the patient information, query the corresponding patient diagnosis history record from the diagnosis database of the hospital diagnosis system;

    Based on the diagnostic history, the patient's allergy history and contraindication information are determined.
12. The device for recommending medication according to claim 11, wherein the processor further implements the following steps when executing the computer program:

    Perform secondary screening and filtering on the new drug sequence according to the allergy history and contraindication information to obtain a second drug sequence;

    The matching degree of each drug in the second drug sequence is scored using a preset scoring model, and the scores are sorted from high to low to obtain a candidate drug sequence.
13. The medication recommendation device according to any one of claims 8-12, wherein the processor also implements the following steps when executing the computer program:

    Obtain the doctor's historical consultation data from multiple medical databases through a data crawler tool, and extract historical prescriptions from the historical consultation data to construct a drug push data set;

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

    A machine learning algorithm is used to perform deep learning on the medicines in the training set and the medicine characteristic information corresponding to the medicines to construct a medicine recommendation model.
14. The device for recommending medication according to claim 13, wherein the processor further implements the following steps when executing the computer program:

    Randomly select a number of the medical questionnaires, and extract the usage of the drug recommendation results in each of the medical questionnaires;

    If the usage does not meet the preset adoption rate, read the questionnaires of all the medicines that are not used in the drug recommendation results, and retrain the drug recommendation model based on the questionnaires to obtain iterative optimization The post-drug recommendation model.
15. A computer-readable storage medium, storing computer instructions in the computer-readable storage medium, when the computer instructions are executed on a computer, the computer is made to perform the following steps:

    Obtain historical prescriptions from multiple medical databases through data crawler tools, and use machine learning algorithms to learn the medicinal properties and medication rules of the historical prescriptions to build a drug recommendation model;

    Obtain the medical questionnaire of the current patient, and extract the diagnosis result and the medical consultation data in the medical questionnaire;

    Inputting the diagnosis result into the drug recommendation model for drug matching, and generating a candidate drug list;

    Sorting, screening and filtering the drugs in the candidate drug list according to the consultation data to obtain a final drug recommendation result;

    According to the received drug use request, the drug that satisfies the patient is selected from the drug recommendation result, and the prescription is generated and stored in the medical questionnaire.
16. The computer-readable storage medium of claim 15, when the computer instructions are executed on a computer, causing the computer to further perform the following steps:

    Extract the patient information, visiting department, chief complaint information, the content of the consultation dialogue between the doctor and the patient, and the doctor's diagnosis result recorded in the consultation sheet by using a text recognition algorithm;

    According to the preset priority classification conditions, the patient information, the visiting department, the chief complaint information, the content of the consultation dialogue between the doctor and the patient, and the doctor's diagnosis result are prioritized to obtain a query condition list;

    The conditions in the query condition list except the diagnosis result are combined according to the priority level to obtain the characteristic phrase.
17. The computer-readable storage medium of claim 16, when the computer instructions are executed on a computer, causing the computer to further perform the following steps:

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

    Scoring and sorting each drug in the new drug sequence according to the matching of the drug and the medical questionnaire to obtain a candidate drug sequence;

    The top N drugs are selected from the candidate drug sequence to obtain a drug recommendation result.
18. The computer-readable storage medium of claim 17, which, when executed on a computer, causes the computer to further perform the following steps:

    According to the patient information, query the corresponding patient diagnosis history record from the diagnosis database of the hospital diagnosis system;

    Based on the diagnostic history, the patient's allergy history and contraindication information are determined.
19. The computer-readable storage medium of claim 18, which, when executed on a computer, causes the computer to further perform the following steps:

    Perform secondary screening and filtering on the new drug sequence according to the allergy history and contraindication information to obtain a second drug sequence;

    The matching degree of each drug in the second drug sequence is scored using a preset scoring model, and the scores are sorted from high to low to obtain a candidate drug sequence.
20. A medication recommendation device, the medication recommendation device includes:

    The training module is used to obtain historical prescriptions from multiple medical databases through a data crawler tool, and use machine learning algorithms to learn the medicinal properties and medication rules of the historical prescriptions, and build a drug recommendation model;

    The extraction module is used to obtain the medical questionnaire of the current patient, and extract the diagnosis result and the medical consultation data in the medical questionnaire;

    a matching module, configured to input the diagnosis result into the drug recommendation model for drug matching, and generate a candidate drug list;

    A screening module, configured to sort, screen and filter the drugs in the candidate drug list according to the consultation data to obtain a final drug recommendation result;

    The recommendation module is used for selecting medicines that satisfy the patient from the medicine recommendation results according to the received medicine use request, and generating a prescription and saving it in the consultation list.

Images