A prescription review method, device, system and storage medium

Abstract

The application discloses a prescription auditing method, device, system and storage medium, the method comprises the following steps: obtaining a to-be-audited prescription and a document associated with the to-be-audited prescription; scoring the to-be-audited prescription based on a drug knowledge graph and the document to obtain a first score of the to-be-audited prescription; obtaining clinical medication knowledge associated with the to-be-audited prescription based on the document, and scoring the to-be-audited prescription based on the clinical medication knowledge to obtain a second score of the to-be-audited prescription; and obtaining a final score of the to-be-audited prescription based on the first score and the second score. The application has the effect of improving the accuracy and rationality of prescription auditing.

Description

Technical Field

[0001] This application relates to the field of smart healthcare technology, and more specifically to a prescription review method, apparatus, system, and storage medium. Background Technology

[0002] Prescription review is a crucial function in healthcare settings. In reality, after a doctor prescribes medication based on a patient's condition, a pharmacist conducts a secondary review. If the prescription is found to be irregular or inappropriate, the pharmacist will refuse to dispense the medication. Compared to tertiary hospitals, primary care hospitals often face issues such as irregular prescriptions, antibiotic overuse, and inappropriate dosages for special populations (the elderly, children, pregnant women, etc.), posing significant challenges to pharmacists' prescription review work.

[0003] Most of the review methods in related technologies are based on constructing reasoning rules from drug knowledge graphs to review prescriptions, or on obtaining a set of recommended drugs from a drug knowledge graph and then scoring the prescription based on the similarity between the recommended drugs and the drugs in the prescription to be reviewed, thereby realizing prescription review.

[0004] Current prescription review methods, while capable of reviewing prescriptions, neglect the mining of vast amounts of clinical medication experience and its role in prescription review. Their review dimensions are too narrow to cover the necessary knowledge for prescription review, and they struggle to filter out key disease characteristics relevant to the final task from a large pool of features. Consequently, their evaluation effectiveness fails to meet the needs of real-world scenarios and easily leads to a waste of human resources. Therefore, a prescription review method is needed to address at least one of the aforementioned problems. Summary of the Invention

[0005] This application is made to address the aforementioned problems. According to one aspect of this application, a prescription review method is provided, the method comprising the following steps:

[0006] Obtain the prescription to be reviewed and the documents associated with the prescription; score the prescription based on the drug knowledge graph and the documents to obtain a first score for the prescription; obtain clinical drug knowledge associated with the prescription based on the documents, and score the prescription based on the clinical drug knowledge to obtain a second score for the prescription; obtain the final score for the prescription based on the first score and the second score.

[0007] In one embodiment of this application, the step of obtaining clinical medication knowledge associated with the prescription to be reviewed based on the document includes: constructing a patient profile based on the document; mining historical prescription data from the hospital based on the patient profile to obtain prescriptions from similar patients; and obtaining clinical medication knowledge associated with the prescription to be reviewed based on the prescriptions from similar patients.

[0008] In one embodiment of this application, the step of mining historical prescription data of the hospital based on the patient profile to obtain similar patient prescriptions includes: mining historical prescription data of the hospital based on the patient profile to obtain relevant patient condition profiles; calculating the similarity between the patient profile and the relevant patient condition profiles, and taking the prescriptions corresponding to the relevant patient condition profiles with similarity higher than a threshold as similar patient prescriptions.

[0009] In one embodiment of this application, obtaining clinical medication knowledge associated with the prescription to be reviewed based on the similar patient prescriptions includes: calculating the prescription text for each prescription among the similar patient prescriptions; and obtaining a knowledge representation of the clinical medication knowledge associated with the prescription to be reviewed based on the prescription text of each prescription.

[0010] In one embodiment of this application, calculating the prescription text of each prescription includes: calculating the initial prescription text of different drugs in each prescription; concatenating the initial prescription texts of different drugs in each prescription to obtain the prescription text of each prescription; and obtaining a knowledge representation of clinical medication knowledge associated with the prescription to be reviewed based on the prescription text of each prescription includes: inputting the prescription text of each prescription into a text encoding model to obtain a knowledge representation of clinical medication knowledge associated with the prescription to be reviewed.

[0011] In one embodiment of this application, a second score is obtained by scoring the prescription to be reviewed based on the clinical medication knowledge, including: calculating the prescription representation of the prescription to be reviewed and the knowledge representation of the clinical medication knowledge; calculating the similarity between the prescription representation and the knowledge representation, and obtaining the second score of the prescription to be reviewed based on the similarity.

[0012] In one embodiment of this application, calculating the prescription representation of the prescription to be reviewed includes: calculating the initial prescription text of different drugs in the prescription to be reviewed; concatenating the initial prescription text of different drugs in the prescription to be reviewed to obtain the prescription text of the prescription to be reviewed; and inputting the prescription text of the prescription to be reviewed into a text encoding model to obtain the prescription representation of the prescription to be reviewed.

[0013] In one embodiment of this application, the initial prescription texts for the different drugs are obtained by concatenating the drug information of each drug to obtain the initial prescription text.

[0014] In one embodiment of this application, the step of scoring the prescription to be reviewed based on the drug knowledge graph and the document to obtain a first score for the prescription to be reviewed includes: obtaining a drug knowledge representation of the prescription to be reviewed based on the drug knowledge graph; identifying the document to obtain key information about the illness, and obtaining a text representation of the illness based on the key information about the illness; calculating the similarity between the drug knowledge representation and the text representation of the illness, and obtaining a first score for the prescription to be reviewed based on the similarity.

[0015] In one embodiment of this application, obtaining the final score of the prescription to be reviewed based on the first score and the second score includes: weighted summing of the first score and the second score to obtain the final score of the prescription to be reviewed.

[0016] In one embodiment of this application, before performing a weighted summation of the first score and the second score, the method further includes: calculating the attention weights of the first score and the second score respectively for use in the weighted summation; wherein the attention weight of the first score is calculated based on the similarity between the drug knowledge representation and the disease text representation; and the attention weight of the second score is calculated based on the similarity between the patient profile and the relevant patient disease profile.

[0017] According to another aspect of this application, this application provides a prescription review device, the device comprising: an information acquisition module for acquiring a prescription to be reviewed and documents associated with the prescription to be reviewed; a first scoring module for scoring the prescription to be reviewed based on the drug knowledge graph and the documents to obtain a first score for the prescription to be reviewed; a second scoring module for acquiring clinical drug knowledge associated with the prescription to be reviewed based on the documents, and scoring the prescription to be reviewed based on the clinical drug knowledge to obtain a second score for the prescription to be reviewed; and a comprehensive scoring module for obtaining a final score for the prescription to be reviewed based on the first score and the second score.

[0018] According to another aspect of this application, this application provides a prescription review system, the system including a memory and a processor, the memory storing a computer program executed by the processor, the computer program, when executed by the processor, causing the processor to perform the above-described prescription review method.

[0019] According to another aspect of this application, this application provides a storage medium storing a computer program that, when executed, performs the prescription review method as described above.

[0020] According to the prescription review method of this application, the first dimension of review and scoring is performed based on a drug knowledge graph, and the second dimension of review and scoring is performed based on clinical drug use knowledge. Finally, the final score is obtained by combining the first and second scores to complete the prescription review. This method considers different review dimensions, ensuring the coverage of drug knowledge and the accuracy of prescription review, thereby improving the accuracy and rationality of prescription review. Attached Figure Description

[0021] The above and other objects, features, and advantages of this application will become more apparent from the more detailed description of the embodiments of this application in conjunction with the accompanying drawings. The drawings are provided to further illustrate the embodiments of this application and form part of the specification. They are used together with the embodiments of this application to explain this application and do not constitute a limitation thereof. In the drawings, the same reference numerals generally represent the same components or steps.

[0022] Figure 1 This is a flowchart of the prescription review method in the embodiments of this application.

[0023] Figure 2 This is an example diagram of a drug knowledge graph.

[0024] Figure 3 This is a flowchart of the first scoring process in an embodiment of this application.

[0025] Figure 4 This is a schematic diagram of the first scoring process in an embodiment of this application.

[0026] Figure 5 This is a flowchart of the second scoring process in an embodiment of this application.

[0027] Figure 6 This is a schematic diagram of the second scoring process in an embodiment of this application.

[0028] Figure 7 This is a schematic diagram of the prescription review device in the embodiments of this application.

[0029] Figure 8 This is a schematic diagram of the prescription review system in an embodiment of this application. Detailed Implementation

[0030] To make the objectives, technical solutions, and advantages of this application more apparent, exemplary embodiments according to this application will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are merely a part of the embodiments of this application, and not all of the embodiments of this application. It should be understood that this application is not limited to the exemplary embodiments described herein. Based on the embodiments of this application described herein, all other embodiments obtained by those skilled in the art without inventive effort should fall within the protection scope of this application.

[0031] Please refer to Figure 1 This application provides a prescription review method, which includes the following steps:

[0032] S11. Obtain the prescription to be reviewed and the documents associated with the prescription to be reviewed.

[0033] S12. Based on the drug knowledge graph and documents, score the prescriptions to be reviewed to obtain the first score of the prescriptions to be reviewed.

[0034] S13. Obtain clinical medication knowledge related to the prescription to be reviewed based on the documents, and score the prescription to be reviewed based on the clinical medication knowledge to obtain the second score of the prescription to be reviewed.

[0035] S14. Based on the first score and the second score, the final score of the prescription to be reviewed is obtained.

[0036] According to the prescription review method of this application, the first dimension of review and scoring is performed based on a drug knowledge graph, and the second dimension of review and scoring is performed based on clinical drug use knowledge. Finally, the final score is obtained by combining the first and second scores to complete the prescription review. This method considers different review dimensions, ensuring the coverage of drug knowledge and the accuracy of prescription review, thereby improving the accuracy and rationality of prescription review.

[0037] The method is illustrated below by way of example.

[0038] The initial scoring process is based on knowledge graph reasoning. Its inputs are medication knowledge and documents, and the output is the first score. For example... Figure 2 The diagram shown is an example of a single drug knowledge graph. For drug information such as instructions for use, pharmacopoeias, and related literature, the process begins with structured parsing to obtain a structured drug knowledge graph for each drug. This graph is then reviewed and supplemented to arrive at the final drug knowledge graph. For example, the structured parsing of drug information is achieved through relationships such as Named Entity Recognition (NER) models based on span annotation. Figure 2 The example provided illustrates a drug knowledge graph for omeprazole. This graph includes the drug's dosage regimen, basic information, and indications. Basic information includes the generic name, code, specifications, and manufacturer. The dosage regimen includes the timing, duration, frequency, and dosage of administration. Indications include the symptoms targeted by the drug, such as gastric ulcers and duodenal ulcers. Searching through a large number of knowledge graphs can yield the drug knowledge graph for the desired medication.

[0039] like Figure 3 As shown, Figure 3 The flowchart for the initial scoring of prescriptions awaiting review includes the following steps:

[0040] S31. Obtain the drug knowledge representation of the prescription to be reviewed based on the drug knowledge graph.

[0041] S32. Recognize the document to obtain key information about the patient's condition, and obtain a document representation based on the key information about the patient's condition.

[0042] S33. Calculate the similarity between the drug knowledge representation and the document representation, and obtain the first score of the prescription to be reviewed based on the similarity.

[0043] like Figure 4 As shown, Figure 4 This illustrates the process of initially scoring prescriptions awaiting review. Drug knowledge representation is obtained through a drug knowledge graph. Key information is extracted from the patient's medical records, and this key information is encoded to obtain a textual representation of the patient's condition. Then, similarity is calculated between the textual representation of the condition and the drug knowledge representation to obtain the first score.

[0044] For example, for each drug in the prescription to be reviewed, the relevant drug graph is retrieved from the drug knowledge graph, and the entities in the drug knowledge graph are concatenated to obtain the drug knowledge graph knowledge text x1. For example, the entities in the drug knowledge graph refer to knowledge information such as medication regimen, basic information, and indications. x1 is input into the text encoding model for encoding to obtain the drug knowledge representation y1 of the prescription to be reviewed in the drug knowledge graph. For example, the text encoding model is a bidirectional encoder representation from transformer (BERT) model, and the encoding process for x1 is shown in the following formula:

[0045] y1 = Encoder(x1)(1)

[0046] Here, Encoder(·) represents any encoding function.

[0047] Next, the prescription to be reviewed and related documents are obtained. For example, these documents can be medical records, including medical history, test reports, etc. The documents are then identified using entity recognition and relation recognition models to extract key information related to the illness. This key information is then concatenated to obtain the key information text x2 of the document. Finally, x2 is encoded using a text encoding model to obtain the illness text representation y2. For example, the encoding process is shown in the following formula:

[0048] y2 = Encoder(x2)(2)

[0049] Here, Encoder(·) represents any encoding function.

[0050] After obtaining the drug knowledge representation y1 and the disease text representation y2, the similarity between y1 and y2 is calculated to obtain the similarity result s'1 for a single drug. Then, the similarity result s'1 for each drug in the prescription to be reviewed is calculated to obtain the first score. The calculation result s'1 based on y1 and y2 is shown in the following formula:

[0051] s'1=α1·S(y1,y2)(3)

[0052] Where S represents any metric function, and α1 represents trainable parameters.

[0053] Based on the above process, drug knowledge representation was obtained through drug knowledge graph, and document representation was obtained through patient medical records and other documents. Then, the first score was calculated by combining the drug knowledge representation and the document representation, thus completing the first scoring and review process.

[0054] like Figure 5 The diagram shows a flowchart for the second scoring based on clinical medication knowledge. The process of the second scoring based on clinical medication knowledge includes:

[0055] S51. Obtain clinical medication knowledge related to the prescription to be reviewed based on the documents.

[0056] S52. Calculate the prescription representation of the prescription to be reviewed.

[0057] S53. Calculate the second score based on the similarity between the knowledge representation of clinical medication and the prescription representation of the prescription to be reviewed.

[0058] like Figure 6 As shown, Figure 6The process of the second scoring is illustrated. First, clinical medication knowledge needs to be acquired. This process includes: constructing a patient profile based on patient medical records and other documents; encoding the patient profile; and mining historical prescription data from the hospital based on the patient profile to obtain prescriptions from similar patients. Based on these similar patient prescriptions, a knowledge representation of clinical medication knowledge associated with the prescription to be reviewed is obtained. Next, the prescription to be reviewed is encoded to obtain a prescription representation. Finally, the second score is calculated based on the prescription representation and the knowledge representation of clinical medication knowledge.

[0059] For example, the process of obtaining prescriptions from similar patients is as follows:

[0060] First, the hospital's historical prescription data is mined based on patient profiles to obtain relevant patient condition profiles; the similarity between the patient profiles and relevant patient condition profiles is calculated, and prescriptions corresponding to relevant patient condition profiles with similarity scores higher than a threshold are considered as similar patient prescriptions.

[0061] After constructing patient profiles, patient profile features por are built based on the profile labels. A key information extraction model is used to mine historical prescription data from hospitals to obtain relevant patient condition profiles. Based on the features of these relevant patient condition profiles, related profile features are constructed. Similarity can be calculated between these related profile features and the patient profile features por. Prescriptions corresponding to patient condition profiles with similarity scores exceeding a threshold are selected and considered similar patient prescriptions. For example, the average of the profile features of similar patient prescriptions is calculated as the average patient profile feature por'.

[0062] For example, the process of obtaining clinical medication knowledge associated with a prescription to be reviewed based on similar patient prescriptions includes:

[0063] First, calculate the prescription text for each prescription in the similar patient prescriptions. For each drug in the prescription, concatenate the drug information to obtain the initial prescription text. For example, the drug information includes the drug name, frequency, route of administration, and duration of treatment. Concatenate the drug name a1, frequency a2, route of administration a3, and duration of treatment a4 of each individual drug in the similar patient prescriptions to obtain the initial prescription text b1. Concatenate the initial prescription text b1 of each different drug in each prescription to obtain the prescription text x3 for each prescription.

[0064] The process of calculating the initial prescription text b1 and the prescription text x3 for each prescription is shown in the following formula:

[0065] b1=a1||a2||a3||a4(4)

[0066] x3=b1||…||b n (5)

[0067] Where || represents the text concatenation method, and b1, b n This represents the initial prescription text for different drugs within the same prescription, where n represents the number of drugs in the same prescription.

[0068] Next, the prescription text x3 for each prescription is input into the text encoding model to obtain the knowledge representation z1 of clinical medication knowledge, as shown in the following formula:

[0069] z1 = Encoder(x3)(6)

[0070] Here, Encoder(·) represents any encoding function.

[0071] To calculate the second score, the prescription representation to be reviewed also needs to be calculated. This second score is calculated using the prescription representation z2 and the knowledge representation z1 of clinical medication knowledge. For example, the process of calculating the prescription representation z2 is as follows:

[0072] First, the drug information of individual drugs in the prescription to be reviewed is concatenated to obtain the initial prescription text b1. ’ The initial prescription texts of different drugs in the prescription to be reviewed are concatenated to obtain the prescription text x4 of the prescription to be reviewed. The prescription text of the prescription to be reviewed is then input into the text encoding model to obtain the prescription representation z2 of the prescription to be reviewed, as shown in the following formula:

[0073] z2 = Encoder(x4)(7)

[0074] Here, Encoder(·) represents any encoding function.

[0075] After obtaining the knowledge representation z1 of clinical drug use knowledge and the prescription representation z2 of the prescription to be reviewed, a second score is obtained through similarity calculation. For example, the similarity between z1 and z2 is calculated, and the second score is obtained based on this similarity. The formula is as follows:

[0076] s2=α2·S(z1,z2)(8)

[0077] Where S represents any metric function, α2 represents trainable parameters, and the product of the similarity between z1 and z2 is used as the second score after calculating the similarity between z1 and z2.

[0078] For example, after obtaining the first score and the second score, the first score and the second score are weighted and summed to obtain the final score of the prescription to be reviewed. Before performing the weighted summation, the attention weights of the first score and the second score need to be calculated for use in the weighted summation.

[0079] For example, the attention weights for the first score and the second score are calculated as follows:

[0080] The attention weight for the first score is calculated based on the similarity between the drug knowledge representation and the disease text representation. The attention weight att1 for the first score is calculated using the similarity between the drug knowledge representation y1 and the disease text representation y2 of the prescription to be reviewed in the drug knowledge graph. First, the similarity between y1 and y2 is calculated, and the result is substituted into the attention calculation function to obtain the attention weight for the first score, as shown in the following formula:

[0081] att1 = ATT(S(y1,y2))(9)

[0082] Where S represents any metric function, ATT represents any attention calculation function, and att1 represents the attention weight corresponding to the first score.

[0083] The attention weight for the second score is calculated based on the similarity between the patient profile and related patient condition profiles. The patient profile feature `por` and the average patient profile feature `por'` were obtained previously. The attention weight `att2` for the second score is calculated using `por` and `por'`. First, the similarity between `por` and `por'` is calculated. The similarity calculation structure is then substituted into the attention calculation function to obtain the attention weight for the second score, as shown in the following formula:

[0084] att2 = ATT(S(por,por'))(10)

[0085] Where S represents any metric function, ATT represents any attention calculation function, and att2 represents the attention weight corresponding to the second score.

[0086] After obtaining the attention weights for the first score, the second score, and the first and second scores respectively, the final score is calculated by weighted summation, as shown in the following formula:

[0087] s=pool(att1·s1,att2·s2)(11)

[0088] Where pool(·) represents any pooling function, and s represents the final score of the prescription to be reviewed.

[0089] The final score was obtained based on the first score, the second score, and their respective attention weights, thus completing the review of the prescription to be reviewed.

[0090] Based on the above description, the prescription review method provided in this application first scores the prescription based on a drug knowledge graph and written documents to obtain a first score, and then scores it based on clinical drug knowledge to obtain a second score. Next, the attention weights of the first and second scores are calculated, and finally, the final score is obtained based on the first and second scores and their respective weights. This method covers different knowledge dimensions, addresses the problem of insufficient drug knowledge during prescription review, and the final score obtained through different dimensions and weights can reflect the rationality of the prescription to the greatest extent possible, significantly improving the effectiveness of prescription review.

[0091] This application also provides a prescription review device, such as... Figure 7 As shown, the prescription review device 700 includes: an information acquisition module 701, used to acquire the prescription to be reviewed and the documents associated with the prescription to be reviewed.

[0092] The first scoring module 702 is used to score the prescription to be reviewed based on the drug knowledge graph and documents, and obtain the first score of the prescription to be reviewed.

[0093] The second scoring module 703 is used to obtain clinical medication knowledge related to the prescription to be reviewed based on the documents, and to score the prescription to be reviewed based on the clinical medication knowledge to obtain the second score of the prescription to be reviewed.

[0094] The comprehensive scoring module 704 is used to obtain the final score of the prescription to be reviewed based on the first score and the second score.

[0095] This application provides a prescription review device that performs review and scoring based on a drug knowledge graph for the first dimension and on clinical medication knowledge for the second dimension. The final score is obtained by combining the first and second scores to complete the prescription review. This method considers different review dimensions, ensuring comprehensive coverage of drug knowledge and high accuracy of prescription review, reflecting the rationality of the prescription to the greatest extent possible, and significantly improving the effectiveness of prescription review.

[0096] This application also provides a prescription review system, such as... Figure 8 As shown, the prescription review system 800 includes a memory 802 and a processor 801. The memory 802 stores a computer program that runs on the processor. When the computer program is run by the processor, it causes the processor 801 to execute the prescription review method described above.

[0097] This application also provides a storage medium storing a computer program, which executes the above-described prescription review method when running.

[0098] Although exemplary embodiments have been described herein with reference to the accompanying drawings, it should be understood that the above exemplary embodiments are merely illustrative and are not intended to limit the scope of this application. Various changes and modifications can be made therein by those skilled in the art without departing from the scope and spirit of this application. All such changes and modifications are intended to be included within the scope of this application as claimed in the appended claims.

[0099] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0100] In the several embodiments provided in this application, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another device, or some features may be ignored or not executed.

[0101] Numerous specific details are set forth in the specification provided herein. However, it will be understood that embodiments of this application may be practiced without these specific details. In some instances, well-known methods, structures, and techniques have not been shown in detail so as not to obscure the understanding of this specification.

[0102] Similarly, it should be understood that, in order to streamline this application and aid in understanding one or more of the various inventive aspects, features of this application may sometimes be grouped together in a single embodiment, figure, or description thereof in the description of exemplary embodiments of this application. However, this approach should not be construed as reflecting an intention that the claimed application requires more features than are expressly recited in each claim. Rather, as reflected in the corresponding claims, its inventive point lies in solving the corresponding technical problem with features fewer than all features of a single disclosed embodiment. Therefore, the claims following the detailed description are hereby expressly incorporated into that detailed description, wherein each claim itself is a separate embodiment of this application.

[0103] Those skilled in the art will understand that, apart from the mutual exclusion of features, all features disclosed in this specification (including the accompanying claims, abstract, and drawings) and all processes or units of any method or apparatus so disclosed can be combined in any combination. Unless otherwise expressly stated, each feature disclosed in this specification (including the accompanying claims, abstract, and drawings) may be replaced by an alternative feature that serves the same, equivalent, or similar purpose.

[0104] Furthermore, those skilled in the art will understand that although some embodiments described herein include certain features included in other embodiments but not others, combinations of features from different embodiments are intended to be within the scope of this application and form different embodiments. For example, in the claims, any one of the claimed embodiments can be used in any combination.

[0105] The various component embodiments of this application can be implemented in hardware, or as software modules running on one or more processors, or a combination thereof. Those skilled in the art will understand that microprocessors or digital signal processors (DSPs) can be used in practice to implement some or all of the functions of some modules according to the embodiments of this application. This application can also be implemented as an apparatus program (e.g., a computer program and computer program product) for performing part or all of the methods described herein. Such an implementation of this application can be stored on a computer-readable medium, or can be in the form of one or more signals. Such signals can be downloaded from an Internet website, provided on a carrier signal, or provided in any other form.

[0106] It should be noted that the above embodiments are illustrative of this application and not restrictive, and that those skilled in the art can devise alternative embodiments without departing from the scope of the appended claims. In the claims, any reference signs placed between parentheses should not be construed as limiting the claims. The word "comprising" does not exclude the presence of elements or steps not listed in the claims. The word "a" or "an" preceding an element does not exclude the presence of a plurality of such elements. This application can be implemented by means of hardware comprising several different elements and by means of a suitably programmed computer. In the unit claims enumerating several means, several of these means may be embodied by the same item of hardware. The use of the words first, second, and third, etc., does not indicate any order. These words can be interpreted as names.

[0107] The above description is merely a specific embodiment or illustration of the embodiments of this application. The scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. The scope of protection of this application shall be determined by the scope of the claims.

Claims

1. A prescription review method, characterized in that, The method includes the following steps: Obtain the prescription to be reviewed and the documents associated with the prescription; The drug knowledge representation of the prescription to be reviewed is obtained based on the drug knowledge graph; the document is identified to obtain key information about the illness, and a textual representation of the illness is obtained based on the key information about the illness; the similarity between the drug knowledge representation and the textual representation of the illness is calculated, and a first score of the prescription to be reviewed is obtained based on the similarity. A patient profile is constructed based on the document; historical prescription data from the hospital is mined based on the patient profile to obtain relevant patient condition profiles; the similarity between the patient profile and the relevant patient condition profiles is calculated, and prescriptions corresponding to relevant patient condition profiles with similarity higher than a threshold are taken as similar patient prescriptions; clinical medication knowledge associated with the prescription to be reviewed is obtained based on the similar patient prescriptions, and the prescription to be reviewed is scored based on the clinical medication knowledge to obtain a second score for the prescription to be reviewed; The attention weights of the first score and the second score are calculated, wherein the attention weight of the first score is calculated based on the similarity between the drug knowledge representation and the disease text representation; and the attention weight of the second score is calculated based on the similarity between the patient profile and the relevant patient disease profile. The first score and the second score are weighted and summed according to the attention weight to obtain the final score of the prescription to be reviewed.

2. The method according to claim 1, characterized in that, The process of obtaining clinical medication knowledge associated with the prescription to be reviewed based on the similar patient prescriptions includes: For each prescription among the similar patient prescriptions, calculate the prescription text for each prescription; Based on the prescription text of each prescription, a knowledge representation of clinical medication knowledge associated with the prescription to be reviewed is obtained.

3. The method according to claim 2, characterized in that, The calculation of the prescription text for each prescription includes: Calculate the initial prescription text of each drug in each prescription; concatenate the initial prescription texts of each drug in each prescription to obtain the prescription text of each prescription; The step of obtaining a knowledge representation of clinical medication knowledge associated with the prescription to be reviewed based on the prescription text of each prescription includes: inputting the prescription text of each prescription into a text encoding model to obtain a knowledge representation of clinical medication knowledge associated with the prescription to be reviewed.

4. The method according to any one of claims 1-3, characterized in that, Based on the aforementioned clinical medication knowledge, the prescription to be reviewed is scored to obtain a second score for the prescription to be reviewed, including: Calculate the prescription representation of the prescription to be reviewed and the knowledge representation of the clinical medication knowledge; Calculate the similarity between the prescription representation and the knowledge representation, and obtain a second score for the prescription to be reviewed based on the similarity.

5. The method according to claim 4, characterized in that, The calculation of the prescription representation of the prescription to be reviewed includes: Calculate the initial prescription text for different drugs in the prescription to be reviewed; The initial prescription texts of different drugs in the prescription to be reviewed are concatenated to obtain the prescription text of the prescription to be reviewed; The prescription text of the prescription to be reviewed is input into the text encoding model to obtain the prescription representation of the prescription to be reviewed.

6. The method according to claim 3 or 5, characterized in that, The initial prescription texts for the different drugs were obtained in the following manner: For each drug, the drug information is concatenated to obtain the initial prescription text.

7. A prescription verification device, characterized in that, The device includes: The information acquisition module is used to acquire the prescription to be reviewed and the documents associated with the prescription to be reviewed; The first scoring module is used to obtain the drug knowledge representation of the prescription to be reviewed based on the drug knowledge graph; to identify the document and obtain key information about the illness, and to obtain a text representation of the illness based on the key information about the illness; to calculate the similarity between the drug knowledge representation and the text representation of the illness, and to obtain a first score for the prescription to be reviewed based on the similarity. The second scoring module is used to construct a patient profile based on the document; mine historical prescription data from the hospital based on the patient profile to obtain relevant patient condition profiles; calculate the similarity between the patient profile and the relevant patient condition profiles, and take the prescriptions corresponding to the relevant patient condition profiles with similarity higher than a threshold as similar patient prescriptions; obtain clinical medication knowledge associated with the prescription to be reviewed based on the similar patient prescriptions, and score the prescription to be reviewed based on the clinical medication knowledge to obtain a second score for the prescription to be reviewed. The weight calculation module is used to calculate the attention weights of the first score and the second score respectively, wherein the attention weight of the first score is calculated based on the similarity between the drug knowledge representation and the disease text representation, and the attention weight of the second score is calculated based on the similarity between the patient profile and the relevant patient disease profile. The comprehensive scoring module is used to sum the first score and the second score according to the attention weight to obtain the final score of the prescription to be reviewed.

8. A prescription review system, characterized in that, The system includes a memory and a processor, the memory storing a computer program executed by the processor, the computer program, when executed by the processor, causing the processor to perform the prescription review method as described in any one of claims 1-6.

9. A storage medium, characterized in that, The storage medium stores a computer program that, when executed, performs the prescription review method as described in any one of claims 1-6.

Images