Patient data management system based on cloud platform

Abstract

The invention provides a patient data management system based on a cloud platform, and relates to the field of intelligent management. And then performing semantic analysis and associated coding fusion on the basic information and the operation information of the patient by adopting a data analysis technology based on a natural language to obtain an antibacterial drug use reference information fusion semantic feature vector of the patient, and generating an antibacterial drug name suitable for the condition of the patient based on the vector obtained by fusion. In this way, the defects of manual medicine selection according to experience can be overcome to a certain extent, and then the accuracy and rationality of use of the antibacterial medicine are effectively improved.

Description

Technical Field

[0001] The present application relates to the field of intelligent management, and more specifically, to a patient data management system based on a cloud platform. Background Art

[0002] Antimicrobial drugs refer to drugs that have the function of inhibiting the activity of pathogens or disinfecting pathogens. Common clinical drugs include cephalosporins and imidazoles. Antimicrobial drugs can treat various diseases caused by bacterial infections and reduce mortality. Surgery is a common treatment method in general surgery and is a traumatic treatment method. During the treatment process, the surgical incision will be damaged, resulting in a large release of inflammatory factors, destroying the immune function and causing pathogen infection. General surgery antibiotics are mostly used in the perioperative period, which can effectively reduce the risk of incision infection in patients, help improve patient prognosis, and reduce the risk of postoperative complications. However, in clinical use, there is a certain irrationality in the management of antibiotics. Specifically, due to differences in medical experience among different doctors, there will inevitably be a certain degree of subjectivity in the selection of antibiotics, which may result in the prescription of antibiotics that are not suitable for the indications, which will hinder the patient's anti-infection treatment process, such as poor infection control and prolonged illness.

[0003] Therefore, an optimized patient data management solution is needed. Summary of the invention

[0004] In order to solve the above technical problems, this application is proposed. This application provides a patient data management system based on a cloud platform.

[0005] According to one aspect of the present application, a patient data management system based on a cloud platform is provided, comprising: The patient information collection module is used to obtain the patient's basic information and surgical information from the cloud platform. The basic information includes age, liver and kidney function, allergy history, and past medical history. The surgical information includes the type of surgery, surgery time, and surgical incision. A patient information encoding module, used for performing feature analysis and feature fusion on the basic information and surgical information of the patient to obtain a semantic feature vector of the patient's antibiotic use reference information; A feature association module, used for performing association coding on the semantic feature vector of the patient's antibiotic use reference information to obtain a semantic association feature vector of the patient's antibiotic use reference information; A feature fusion module, used for performing target domain data distribution matching based on principal component analysis on the semantic feature vector of the patient antibiotic use reference information and the semantic association feature vector of the patient antibiotic use reference information to obtain a fused semantic feature vector of the patient antibiotic use reference information; The antimicrobial drug recommendation result generation module is used to fuse the semantic feature vector based on the patient's antimicrobial drug use reference information to obtain the name of the antimicrobial drug suitable for the patient.

[0006] Furthermore, the patient information encoding module includes: A patient basic information semantic feature extraction unit, used for extracting semantic features from the patient's basic information to obtain a patient basic information semantic feature vector; A patient surgery information semantic feature extraction unit, used for extracting semantic features from the patient surgery information to obtain a patient surgery information semantic feature vector; The patient basic information and surgical information semantic feature fusion unit is used to fuse the patient basic information semantic feature vector and the patient surgical information semantic feature vector to obtain the patient antibiotic use reference information semantic feature vector.

[0007] Furthermore, the patient basic information semantic feature extraction unit is used to: segment the patient basic information and pass it through a patient basic information semantic encoder to obtain the patient basic information semantic feature vector.

[0008] Furthermore, the patient surgery information semantic feature extraction unit is used to: segment the patient surgery information and pass it through a patient surgery information semantic encoder to obtain the patient surgery information semantic feature vector.

[0009] Furthermore, the feature association module is used to: divide the semantic feature vector of the patient antibiotic use reference information into multiple sub-vectors and input them into a semantic information association encoder to obtain the semantic association feature vector of the patient antibiotic use reference information.

[0010] Furthermore, the patient basic information semantic encoder and the patient surgical information semantic encoder are bidirectional long short-term memory neural network models including a word embedding layer, and the semantic information association encoder is a semantic information association encoder based on a converter structure.

[0011] Furthermore, the feature fusion module includes: A semantic intrinsic decomposition unit for patient antibiotic use reference information, used for mapping the semantic feature vector of the patient antibiotic use reference information and the semantic association feature vector of the patient antibiotic use reference information to an intrinsic decomposition space to obtain the semantic intrinsic feature vector of the patient antibiotic use reference information and the semantic association intrinsic feature vector of the patient antibiotic use reference information; A semantic feature alignment unit for patient antibiotic use reference information, used for performing cross-domain feature alignment based on the target domain scale on the semantic intrinsic feature vector of the patient antibiotic use reference information and the semantic association intrinsic feature vector of the patient antibiotic use reference information to obtain a semantically modulated intrinsic feature vector of the patient antibiotic use reference information and a semantically associated modulated intrinsic feature vector of the patient antibiotic use reference information; A fine-grained interactive semantic information encoding unit, used to construct a fine-grained interactive semantic matrix of the patient antibiotic use reference information between the intrinsic feature vector after semantic modulation of the patient antibiotic use reference information and the intrinsic feature vector after semantic association modulation of the patient antibiotic use reference information; A patient antibiotic use reference information fusion semantic feature generation unit is used to fuse the patient antibiotic use reference information semantically modulated intrinsic feature vector and the patient antibiotic use reference information semantically associated modulated intrinsic feature vector based on the patient antibiotic use reference information fine-grained interactive semantic matrix to obtain the patient antibiotic use reference information fusion semantic feature vector.

[0012] Furthermore, the fine-grained interaction semantic information encoding unit is used to: The patient antibiotic use reference information semantically modulated intrinsic feature vector and the transposed vector of the patient antibiotic use reference information semantically modulated intrinsic feature vector are multiplied and then divided by the scale of the patient antibiotic use reference information semantically modulated intrinsic feature vector to obtain the patient antibiotic use reference information fine-grained interaction matrix; A convolution encoding operation based on a convolution kernel of 3×3 is performed on the fine-grained interaction matrix of the patient antibiotic use reference information to obtain a fine-grained interaction semantic matrix of the patient antibiotic use reference information.

[0013] Furthermore, the patient antimicrobial drug usage reference information fusion semantic feature generation unit is used to: Multiplying the fine-grained interactive semantic matrix of the patient antibiotic use reference information by the semantically modulated intrinsic feature vector of the patient antibiotic use reference information to obtain a semantic feature vector of the patient antibiotic use reference information after information field mapping; Multiplying the fine-grained interaction semantic matrix of the patient antibiotic use reference information by the semantic association modulated intrinsic feature vector of the patient antibiotic use reference information to obtain the semantic association feature vector of the patient antibiotic use reference information after information field mapping; The semantic feature vector of the patient's antibiotic use reference information after the information field mapping and the semantic association feature vector of the patient's antibiotic use reference information after the information field mapping are weighted fused to obtain the patient's antibiotic use reference information fused semantic feature vector.

[0014] Furthermore, the antibiotic recommendation result generation module is used to: fuse the patient's antibiotic usage reference information with a semantic feature vector through an antibiotic recommendation generator to obtain a generation result, and the generation result is used to represent the name of the antibiotic recommended to the patient.

[0015] Compared with the prior art, the patient data management system based on the cloud platform provided by the present application first obtains the basic information and surgical information of the patient from the cloud platform, and then uses the data analysis technology based on natural language to perform semantic analysis and associative coding fusion on the basic information and surgical information of the patient to obtain the patient's antibiotic use reference information fusion semantic feature vector, and generates the name of the antibiotic suitable for the patient's condition based on the fused vector. In this way, the shortcomings of manual drug selection based on experience can be overcome to a certain extent, thereby effectively improving the accuracy and rationality of the use of antibiotics. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] By describing the embodiments of the present application in more detail in conjunction with the accompanying drawings, the above and other purposes, features and advantages of the present application will become more apparent. The accompanying drawings are used to provide a further understanding of the embodiments of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation of the present application. In the accompanying drawings, the same reference numerals generally represent the same components or steps.

[0017] Figure 1 This is a system block diagram of a cloud platform-based patient data management system according to an embodiment of the present application.

[0018] Figure 2 This is a data flow diagram of a cloud platform-based patient data management system according to an embodiment of the present application.

[0019] Figure 3 It is a block diagram of a patient information encoding module in a cloud platform-based patient data management system according to an embodiment of the present application. DETAILED DESCRIPTION

[0020] Below, the exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the exemplary embodiments described here.

[0021] Antimicrobial drugs refer to a class of drugs that can inhibit or kill pathogens. Cephalosporins and imidazoles are more common in clinical practice. They play an important role in the treatment of a variety of diseases caused by bacterial infections and can significantly reduce the mortality rate of patients. Surgery, as a conventional treatment method in general surgery, is traumatic. The incision during the operation will cause damage, which will trigger the release of a large number of inflammatory factors, weaken the immune function, and increase the risk of pathogen infection. Therefore, in general surgery, antimicrobial drugs are often used in the perioperative period to effectively reduce the risk of infection of the patient's surgical incision, improve the patient's recovery, and reduce possible complications after surgery. However, in clinical practice, there are some unreasonable aspects in the management of antimicrobial drugs. Specifically, due to the different medical experience accumulated by each doctor, it is inevitable that there will be certain subjective judgments when choosing antimicrobial drugs, which may lead to the prescribed antimicrobial drugs not fully matching the patient's actual indications, thereby hindering the patient's anti-infection treatment process, such as unsatisfactory infection control, delayed illness, and other adverse consequences. Therefore, a patient data management solution based on a cloud platform is expected.

[0022] Figure 1 This is a system block diagram of a cloud platform-based patient data management system according to an embodiment of the present application. Figure 2 FIG. 1 is a data flow diagram of a patient data management system based on a cloud platform according to an embodiment of the present application. Figure 1 and Figure 2 As shown, in the patient data management system 100 based on the cloud platform, it includes: a patient information acquisition module 110, which is used to obtain the patient's basic information and surgical information from the cloud platform; a patient information encoding module 120, which is used to perform feature analysis and feature fusion on the patient's basic information and surgical information to obtain a patient's antibiotic use reference information semantic feature vector; a feature association module 130, which is used to perform association encoding on the patient's antibiotic use reference information semantic feature vector to obtain a patient's antibiotic use reference information semantic association feature vector; a feature fusion module 140, which is used to perform target domain data distribution matching based on principal component analysis on the patient's antibiotic use reference information semantic feature vector and the patient's antibiotic use reference information semantic association feature vector to obtain a patient's antibiotic use reference information fused semantic feature vector; an antibiotic recommendation result generation module 150, which is used to obtain the name of the antibiotic suitable for the patient based on the patient's antibiotic use reference information fused semantic feature vector.

[0023] In the embodiment of the present application, the patient information acquisition module 110 is used to obtain the basic information and surgical information of the patient from the cloud platform. The basic information includes age, liver and kidney function, allergy history, and past medical history. The surgical information includes the type of surgery, operation time, and surgical incision. It should be understood that age is an important consideration for the selection of antibiotics. The functions of various organs in children's bodies are not yet fully developed. For example, the liver drug enzyme system of newborns is immature and has a weak metabolic capacity for certain antibiotics; the elderly may have a variety of underlying diseases and organ function decline. For example, decreased renal function will affect the excretion of drugs. For children and the elderly, the safety and tolerability of the drugs need to be considered when choosing antibiotics. For example, aminoglycoside antibiotics are ototoxic and nephrotoxic, so they need to be used with caution for children and the elderly. Safer β-lactam drugs may be selected as substitutes, and the dosage should also be appropriately adjusted according to age. The liver is the main organ for drug metabolism, and the kidney is the main organ for drug excretion. Abnormal liver function may affect the metabolism of drugs, slowing down the metabolism of drugs in the body and leading to increased drug concentrations; renal insufficiency may affect the excretion of drugs and cause drug accumulation in the body. If the patient has liver insufficiency, it may be necessary to change the drug for antimicrobial drugs that are mainly metabolized by the liver (such as erythromycin). For patients with severe renal insufficiency, it may be necessary to avoid the use of certain nephrotoxic antimicrobial drugs. The allergy history records the patient's allergic reactions to various drugs. Antimicrobial drug allergies may cause serious adverse reactions, such as anaphylactic shock. If the patient is allergic to a certain antimicrobial drug, then the drug and its similar drugs must be avoided when recommending antimicrobial drugs. Past medical history includes various diseases that the patient has suffered in the past, such as diabetes, cardiovascular disease, and immune diseases. These diseases may affect the patient's immune function and overall physical condition. For example, diabetic patients have poor wound healing ability and a relatively high risk of infection. For such patients before and after surgery, it may be necessary to choose drugs with a broad antimicrobial spectrum and strong antimicrobial activity, and the effect of the drug on blood sugar should be considered. For patients with low immune function (such as those with immune diseases such as AIDS), more attention should be paid to preventing serious infections when choosing antibiotics. Preventive antibiotics may be used in advance, and drugs with little impact on their immune function should be selected. Surgery types are generally divided into clean surgery, clean-contaminated surgery, and contaminated surgery. The risk of infection in clean surgery is relatively low, and generally no preventive use of antibiotics is required. If they are needed, antibiotics against Staphylococcus aureus are usually selected; the risk of infection in clean-contaminated surgery is moderate, and preventive use of antibiotics is required. The choice of antibiotics should cover possible contaminating bacteria, such as gastrointestinal surgery may need to cover Gram-negative bacilli and anaerobic bacteria; the risk of infection in contaminated surgery is high, and the use of antibiotics is more stringent, and it is necessary to choose highly effective and broad-spectrum antibiotics for prevention and treatment.The duration of the patient's surgery directly affects the time the surgical incision is exposed to the external environment. The longer the time, the more opportunities for bacterial contamination and reproduction. For surgeries with longer operating times, long-acting formulations of antibiotics need to be selected. The surgical incision conditions include the size and depth of the incision, whether it is an open wound, etc. The larger and deeper the incision, the higher the risk of infection; open wounds are more likely to be contaminated by bacteria than closed wounds. For larger, deeper or open surgical incisions, when recommending antibiotics, consider choosing drugs that can penetrate into deep tissues and have a wider antibacterial spectrum. For example, for deep orthopedic implant surgical incisions, antibiotics that can reach effective concentrations in bone tissue may be selected, such as certain cephalosporins, while also considering covering common pathogens that may exist in the surgical site, such as Staphylococcus aureus and coagulase-negative Staphylococci.

[0024] The basic information and surgical information of patients come from a wide range of sources, including various department systems of the hospital (such as outpatient system, inpatient system, inspection and examination system, etc.). The cloud platform can serve as a data center, integrating patient data scattered in various systems for easy centralized acquisition. Obtaining information through the cloud platform can quickly collect the required data, reduce the time for manual collection and collation of data, and thus improve the efficiency of data acquisition. In addition, the data storage and management of the cloud platform usually have strict specifications and quality control measures to ensure the accuracy of the data. In addition, the cloud platform has accumulated a large amount of patient data, which can be used for subsequent big data analysis. By analyzing the basic information and surgical information of many patients and the corresponding antimicrobial drug use effects, the antimicrobial drug recommendation algorithm can be continuously optimized, thereby improving the scientificity and accuracy of the recommendation.

[0025] In the embodiment of the present application, the patient information encoding module 120 is used to perform feature analysis and feature fusion on the patient's basic information and surgical information to obtain a semantic feature vector of the patient's antibiotic use reference information. Specifically, Figure 3 FIG. 1 is a block diagram of a patient information encoding module in a patient data management system based on a cloud platform according to an embodiment of the present application. Figure 3 As shown, the patient information encoding module 120 includes: a patient basic information semantic feature extraction unit 121, which is used to extract semantic features from the patient's basic information to obtain a patient basic information semantic feature vector; a patient surgical information semantic feature extraction unit 122, which is used to extract semantic features from the patient's surgical information to obtain a patient surgical information semantic feature vector; a patient basic information and surgical information semantic feature fusion unit 123, which is used to fuse the patient basic information semantic feature vector and the patient surgical information semantic feature vector to obtain the patient antibiotic use reference information semantic feature vector.

[0026] In an embodiment of the present application, the patient basic information semantic feature extraction unit 121 is used to extract semantic features from the patient's basic information to obtain a patient basic information semantic feature vector. Specifically, in an embodiment of the present application, the patient basic information semantic feature extraction unit 121 is used to: segment the patient's basic information and pass it through the patient basic information semantic encoder to obtain the patient basic information semantic feature vector. It should be understood that the patient's basic information is stored in text form, and the computer cannot directly understand the semantics of such direct text data. In order to decompose complex text information into units that are easier for computers to process, it is first necessary to perform word segmentation on the patient's basic information. After word segmentation, the model can analyze each semantic unit more accurately in the future. For example, in the part of processing drug allergy history, after word segmentation, the specific allergic drug name or allergy-related descriptive vocabulary can be accurately identified to avoid treating the entire text as a vague whole and losing details. Moreover, considering that the textual expressions of different patients' basic information may be diverse, these texts can be processed in a unified manner after word segmentation, so that the model can better adapt to various expressions and improve the versatility and robustness of the model. Next, the patient's basic information data after word segmentation needs to be input into the patient's basic information semantic encoder for processing. Here, the patient's basic information semantic encoder is a bidirectional long short-term memory neural network model containing a word embedding layer. It should be explained that the word embedding layer is a structure that maps words in natural language to a low-dimensional real number vector space. Each word has a corresponding vector representation in this vector space. For example, the word "age" may be mapped to a specific vector [0.2, 0.3, -0.1, ...]. This vector representation can capture the semantic and grammatical information of the word. Words with similar semantics are relatively close in the vector space. When processing the patient's basic information, the word embedding layer can convert the words after word segmentation into a vector form that the computer can better understand. These vectors contain the semantic information of the words, so that the subsequent neural network model can use this semantic information for analysis. For example, for the two words "normal renal function" and "renal insufficiency", after the word embedding layer, their vector representations will be significantly different. This difference can be captured by the model, thereby distinguishing different renal function states. Long short-term memory neural network (LSTM) is a special recurrent neural network (RNN) that can effectively process sequence data. Based on LSTM, bidirectional LSTM considers both the forward and reverse information of sequence data. For a sequence, it can not only process the sequence from front to back, but also from back to front, which can better capture the long-distance dependencies between words.By processing the patient's basic information after word segmentation through the word embedding layer and the bidirectional long short-term memory neural network model, the complex semantic relationships in the original patient's basic information can be captured. This precise semantic capture and analysis helps to improve the accuracy and rationality of antimicrobial drug recommendations and make the recommendation results more in line with the patient's individual conditions.

[0027] In an embodiment of the present application, the patient surgical information semantic feature extraction unit 122 is used to extract semantic features from the patient's surgical information to obtain a semantic feature vector of the patient's surgical information. Specifically, in an embodiment of the present application, the patient surgical information semantic feature extraction unit 122 is used to: segment the patient's surgical information and pass it through a patient surgical information semantic encoder to obtain a semantic feature vector of the patient's surgical information. Similarly, considering that the patient's surgical information obtained from the cloud platform is also recorded in text form, in order to enable the computer to understand the semantic content in the surgical information, it is first necessary to segment the patient's surgical information. The surgical information obtained after the segmentation process can facilitate subsequent semantic encoding and avoid information ambiguity. Next, the segmented patient's surgical information is input into the patient's surgical information semantic encoder. Here, the patient's surgical information semantic encoder is a bidirectional long short-term memory neural network model containing a word embedding layer. The words in the surgical information obtained after word segmentation will be mapped to a low-dimensional real vector space after passing through the word embedding layer. This process can capture the semantic and grammatical information of the words; and through the bidirectional long short-term memory neural network model, the long-distance dependencies in the sequence of word embedding vectors obtained after mapping can be captured, and the deep semantic relationship between different words in the surgical information can be further explored. By capturing the semantic features of surgical information, the foundation can be laid for personalized recommendation of antimicrobial drugs.

[0028] In an embodiment of the present application, the patient basic information and surgical information semantic feature fusion unit 123 is used to fuse the patient basic information semantic feature vector and the patient surgical information semantic feature vector to obtain the patient antibiotic use reference information semantic feature vector. It should be understood that the patient's basic information and surgical information play an indispensable role in the reasonable recommendation of antibiotics. Basic information reflects the patient's own physiological condition; while surgical information reflects the infection risk associated with surgery. By combining these two aspects of information, the patient's demand for antibiotics can be fully considered. Therefore, in the technical solution of the present application, it is necessary to fuse the patient's basic information semantic feature vector and the patient's surgical information semantic feature vector. By fusing these two vectors, the selection range of antibiotics can be narrowed down based on considering multiple factors, and those drugs that are not suitable for the patient's basic condition or do not meet the risk control of surgical infection can be excluded. In this way, unreasonable recommendations caused by considering only a single factor can be avoided, and the safety and effectiveness of the use of antibiotics can be improved.

[0029] In an embodiment of the present application, the feature association module 130 is used to perform association coding on the semantic feature vector of the patient's antibiotic use reference information to obtain a semantic association feature vector of the patient's antibiotic use reference information. Specifically, in an embodiment of the present application, the feature association module 130 is used to: divide the semantic feature vector of the patient's antibiotic use reference information into multiple sub-vectors and input them into the semantic information association encoder to obtain the semantic association feature vector of the patient's antibiotic use reference information. Here, the semantic information association encoder is a semantic information association encoder based on a converter structure. It should be understood that the fused semantic feature vector of the patient's antibiotic use reference information contains a variety of complex semantic features of the patient's basic information and surgical information. Directly processing this vector will face problems such as high consumption of computing resources and difficulty in model training. In order to reduce the complexity of the data, the semantic feature vector of the patient's antibiotic use reference information can be segmented. After segmentation, multiple relatively simple sub-vectors can be obtained, each of which focuses on part of the information, and these local information can be deeply mined and associated with analysis, highlighting the local information association characteristics, so that the interaction relationship between different information can be more accurately discovered, and the patient's antibiotic needs can be better captured. Specifically, the multiple sub-vectors obtained after segmentation are input into the semantic information association encoder. Here, the semantic information association encoder is a semantic information association encoder based on the converter structure. The core of the converter structure is the self-attention mechanism, which can calculate the degree of association between each element in the input sequence (here, the sub-vector sequence after segmentation) and other elements, and then capture the complex association between the sub-vectors after segmentation. By accurately capturing the association information between different information, it can help recommend antimicrobial drugs more accurately.

[0030] In an embodiment of the present application, the feature fusion module 140 is used to perform target domain data distribution matching based on principal component analysis on the semantic feature vector of the patient's antibiotic use reference information and the semantic association feature vector of the patient's antibiotic use reference information to obtain a fused semantic feature vector of the patient's antibiotic use reference information. It should be understood that the semantic feature vector of the patient's antibiotic use reference information contains preliminary semantic features directly extracted and integrated from the patient's basic information and surgical information, reflecting the basic overview of the patient's individual condition and surgical situation. The semantic association feature vector of the patient's antibiotic use reference information focuses on mining the complex associations between these information. Both describe the relevant information of the patient's antibiotic use from different angles. The fusion of the semantic feature vector of the patient's antibiotic use reference information and the semantic association feature vector of the patient's antibiotic use reference information can combine the overall picture of the information with the internal connection to form a more comprehensive and in-depth information representation. That is, through fusion, a complete patient information portrait for accurate antibiotic recommendation is constructed. It not only retains the semantic features of the original information, but also incorporates the deep-level association logic between these information, so that the final fused semantic feature vector can serve as a comprehensive basis to fully reflect the various needs, risks and constraints of patients in the use of antibiotics. Based on this fusion vector, the model can make more accurate recommendations that are more in line with the actual situation of the patient. For example, when faced with a variety of antibiotics, the fusion vector can comprehensively consider the antibacterial spectrum of the drug, the patient's physical condition, the infection risk of the operation, and the relationship between various factors, so as to screen out the most suitable drug, rather than making decisions based on a single factor or a simple combination of factors.

[0031] In particular, considering that in the technical solution of the present application, the semantic feature vector of the patient's antibiotic use reference information is obtained by fusing the patient's basic information semantic feature vector and the patient's surgical information semantic feature vector, and the focus is on obtaining a comprehensive semantic representation; while the semantic association feature vector of the patient's antibiotic use reference information focuses on analyzing and encoding the internal structure of the original semantic feature vector, and the focus is on extracting the association information between sub-vectors, thereby forming a feature space different from the original semantic feature vector; when fusing the patient's antibiotic use reference information semantic feature vector and the patient's antibiotic use reference information semantic association feature vector, if a simple concatenation (such as directly connecting two vectors) or a simple weighted summation method is used, it is easy to ignore the more complex relationship between them. Because these simple fusion methods do not take into account the semantic differences and association methods of the features in the two feature spaces. Due to the ignoring of complex relationships, the generated patient antibiotic use reference information fusion semantic feature vector may not accurately reflect the patient's actual antibiotic demand, which will lead to the final recommended antibiotics not being very suitable for the patient. Therefore, in the technical solution of the present application, it is necessary to perform target domain data distribution matching based on principal component analysis on the semantic feature vector of the patient antibiotic use reference information and the semantic association feature vector of the patient antibiotic use reference information to obtain a fused semantic feature vector of the patient antibiotic use reference information.

[0032] Specifically, in the embodiment of the present application, the feature fusion module 140 includes: a semantic intrinsic decomposition unit of the patient antibiotic use reference information, which is used to map the semantic feature vector of the patient antibiotic use reference information and the semantic association feature vector of the patient antibiotic use reference information to the intrinsic decomposition space to obtain the semantic intrinsic feature vector of the patient antibiotic use reference information and the semantic association intrinsic feature vector of the patient antibiotic use reference information; and a semantic feature alignment unit of the patient antibiotic use reference information, which is used to perform cross-domain feature alignment based on the target domain scale on the semantic intrinsic feature vector of the patient antibiotic use reference information and the semantic association intrinsic feature vector of the patient antibiotic use reference information to obtain the semantically modulated intrinsic feature vector of the patient antibiotic use reference information. a fine-grained interactive semantic information encoding unit, used to construct a fine-grained interactive semantic matrix of patient antibiotic use reference information between the intrinsic feature vector after semantic modulation of the patient antibiotic use reference information and the intrinsic feature vector after semantic association modulation of the patient antibiotic use reference information; a patient antibiotic use reference information fused semantic feature generating unit, used to fuse the intrinsic feature vector after semantic modulation of the patient antibiotic use reference information and the intrinsic feature vector after semantic association modulation of the patient antibiotic use reference information based on the patient antibiotic use reference information fine-grained interactive semantic matrix to obtain the patient antibiotic use reference information fused semantic feature vector.

[0033] More specifically, in an embodiment of the present application, the fine-grained interaction semantic information encoding unit is used to: multiply the intrinsic feature vector after semantic modulation of the patient antibiotic use reference information and the transposed vector of the intrinsic feature vector after semantic association modulation of the patient antibiotic use reference information, and then divide the result by the square root of the scale of the intrinsic feature vector after semantic association modulation of the patient antibiotic use reference information to obtain a fine-grained interaction matrix of the patient antibiotic use reference information; perform a convolution encoding operation based on a convolution kernel of 3×3 on the fine-grained interaction matrix of the patient antibiotic use reference information to obtain a fine-grained interaction semantic matrix of the patient antibiotic use reference information.

[0034] More specifically, in an embodiment of the present application, the patient antibiotic use reference information fusion semantic feature generation unit is used to: multiply the patient antibiotic use reference information fine-grained interaction semantic matrix by the patient antibiotic use reference information semantically modulated eigenfeature vector to obtain the patient antibiotic use reference information semantic feature vector after information field mapping; multiply the patient antibiotic use reference information fine-grained interaction semantic matrix by the patient antibiotic use reference information semantic association modulated eigenfeature vector to obtain the patient antibiotic use reference information semantic association feature vector after information field mapping; perform weighted fusion on the patient antibiotic use reference information semantic feature vector after information field mapping and the patient antibiotic use reference information semantic association feature vector after information field mapping to obtain the patient antibiotic use reference information fusion semantic feature vector.

[0035] In an embodiment of the present application, specifically, the semantic feature vector of the patient's antibiotic use reference information and the semantic association feature vector of the patient's antibiotic use reference information are matched with the target domain data distribution based on principal component analysis to obtain a fused semantic feature vector of the patient's antibiotic use reference information, including: processing the semantic feature vector of the patient's antibiotic use reference information and the semantic association feature vector of the patient's antibiotic use reference information according to the following formula to obtain the fused semantic feature vector of the patient's antibiotic use reference information; wherein the formula is: ; ; ; ; ; ; ; ; in, The semantic feature vector representing the patient's antibiotic use reference information, The semantic association feature vector representing the patient's antibiotic use reference information, represents principal component analysis, A sequence of semantic intrinsic decomposition vectors representing patient antibiotic use reference information, represents the semantic diagonal matrix of patient antibiotic use reference information, and They represent the first and second positions of the semantic diagonal matrix of the patient's antibiotic use reference information. The eigenvalues ​​at the positions, represents the transpose of a vector, represents vector concatenation, , , The first, second and third sequences of the semantic eigendecomposition vectors of the patient's antibiotic use reference information are represented respectively. feature vectors, The semantic intrinsic feature vector representing the patient's antibiotic use reference information, A sequence of semantically associated intrinsic decomposition vectors representing patient antibiotic use reference information, represents the semantic diagonal matrix of patient antibiotic use reference information, and They represent the first and second positions of the semantic association diagonal matrix of the patient's antibiotic use reference information. The eigenvalues ​​at the positions, , , The first, second and third sequences of semantic association eigendecomposition vectors of the patient's antibiotic use reference information are represented respectively. feature vectors, The semantic association eigenvector representing the patient's antibiotic use reference information, The first semantic feature vector representing the patient's antibiotic use reference information eigenvalues, The first eigenvector representing the semantic association of the patient's antibiotic use reference information eigenvalues, represents the two-norm of a vector, represents the natural exponential function, Represents the intrinsic feature vector after semantic modulation of the patient's antibiotic use reference information, The intrinsic feature vector after semantic association modulation of the patient's antibiotic use reference information, represents matrix multiplication, represents the scale of the vector, Represents the fine-grained interaction matrix of patient antibiotic use reference information, Express Perform convolutional coding operation based on a convolution kernel of 3×3. A fine-grained interaction semantic matrix representing the patient's antibiotic use reference information, Represents the semantic feature vector of the patient's antimicrobial drug use reference information after information field mapping, Represents the semantic association feature vector of the patient's antimicrobial drug use reference information after information field mapping, and represents the weighted hyperparameter, Represents the patient's antimicrobial drug use reference information fused with semantic feature vectors.

[0036] That is, in the technical solution of the present application, the semantic feature vector of the patient's antibiotic use reference information and the semantic association feature vector of the patient's antibiotic use reference information are matched with the target domain data distribution based on principal component analysis. The process first maps the semantic feature vector of the patient's antibiotic use reference information and the semantic association feature vector of the patient's antibiotic use reference information to the intrinsic decomposition space to obtain the semantic intrinsic feature vector of the patient's antibiotic use reference information and the semantic association intrinsic feature vector of the patient's antibiotic use reference information. A person of ordinary skill in the art should know that the process of mapping to the intrinsic decomposition space is a combination of dimensionality reduction and feature extraction. In a specific example, the eigenvalue decomposition technology of principal component analysis (PCA) is used to find the intrinsic decomposition space, and the intrinsic decomposition space mapping is performed. The mapped semantic intrinsic feature vector of the patient's antibiotic use reference information and the semantic association intrinsic feature vector of the patient's antibiotic use reference information not only represent the projection of the original feature vector in the new coordinate system, but also reflect the main direction of change of the data, which helps to avoid the "dimensionality disaster", that is, the problem that the sample density drops sharply as the number of features increases.

[0037] Next, after mapping the feature vector to the intrinsic decomposition space, it is necessary to perform cross-domain feature alignment based on the target domain scale on the semantic intrinsic feature vector of the patient's antibiotic use reference information and the semantic association intrinsic feature vector of the patient's antibiotic use reference information to adjust the source domain feature vector to better match the data distribution of the target domain. This operation is intended to capture and compensate for the systematic deviations between the two domains to reduce the hidden disturbance of the class expression of the local feature distribution of the semantic intrinsic feature vector of the patient's antibiotic use reference information and the semantic association intrinsic feature vector of the patient's antibiotic use reference information to the overall class expression. After modulation, the semantically modulated intrinsic feature vector of the patient's antibiotic use reference information and the semantically associated modulated intrinsic feature vector of the patient's antibiotic use reference information can retain the source domain characteristics while being more in line with the actual needs of the target domain, thereby enhancing the model's understanding and application of the target domain data.

[0038] After completing the scale modulation of the feature vector, a fine-grained interactive semantic matrix of the patient's antibiotic use reference information is constructed between the intrinsic feature vector after semantic modulation of the patient's antibiotic use reference information and the intrinsic feature vector after semantic association modulation of the patient's antibiotic use reference information. The fine-grained interactive semantic matrix of the patient's antibiotic use reference information is used to describe a more detailed relationship between feature vectors. The introduction of the fine-grained interactive semantic matrix of the patient's antibiotic use reference information enables the model to go beyond simple point-to-point comparison and consider more complex patterns and contextual information, thereby enhancing the model's sensitivity to subtle differences and improving the quality of feature representation. Finally, based on the fine-grained interactive semantic matrix of the patient's antibiotic use reference information, the intrinsic feature vector after semantic modulation of the patient's antibiotic use reference information and the intrinsic feature vector after semantic association modulation of the patient's antibiotic use reference information are fused to obtain a fused semantic feature vector of the patient's antibiotic use reference information.

[0039] In an embodiment of the present application, the antimicrobial drug recommendation result generation module 150 is used to fuse the semantic feature vector based on the patient's antimicrobial drug use reference information to obtain the name of the antimicrobial drug suitable for the patient. Specifically, in an embodiment of the present application, the antimicrobial drug recommendation result generation module 150 is used to: pass the patient's antimicrobial drug use reference information fused semantic feature vector through an antimicrobial drug recommendation generator to obtain a generation result, and the generation result is used to represent the name of the antimicrobial drug recommended to the patient. It should be understood that although the patient's antimicrobial drug use reference information fused semantic feature vector has highly concentrated the key information related to the antimicrobial drug recommendation, it is still an abstract feature representation form that can be understood and processed by a computer, and cannot directly provide clinical medical staff with intuitive and clear antimicrobial drug recommendation content. In order to convert this abstract feature vector into a specific, practical antimicrobial drug name recommendation, thereby completing the key connection from data processing to actual decision support, it is necessary to input the patient's antimicrobial drug use reference information fused semantic feature vector into the antimicrobial drug recommendation generator. The antimicrobial drug recommendation generator is essentially a model built based on a machine learning algorithm. It can output the corresponding antimicrobial drug name as a recommendation result after receiving the patient's antimicrobial drug use reference information fused with semantic feature vectors obtained from the previous series of processing as input according to pre-trained patterns and rules. The generated antimicrobial drug name recommendation results provide a direct reference for clinical medical staff. When facing specific patients, especially in scenarios such as the perioperative period of general surgery where antimicrobial drugs need to be carefully selected to prevent and control infections, medical staff can accurately prescribe appropriate antimicrobial drugs based on this recommendation result, combined with their own clinical experience and the patient's real-time condition, to avoid irrational medication caused by blind medication or arbitrary selection of antimicrobial drugs based on experience. This is conducive to improving the accuracy and rationality of the use of antimicrobial drugs and can better meet the actual needs of patients.

[0040] In summary, the patient data management system 100 based on the cloud platform of the embodiment of the present application is explained, which first obtains the basic information and surgical information of the patient from the cloud platform, and then uses the data analysis technology based on natural language to perform semantic analysis and associative coding fusion on the basic information and surgical information of the patient to obtain the patient's antibiotic use reference information fusion semantic feature vector, and generates the name of the antibiotic suitable for the patient's condition based on the fused vector. In this way, the shortcomings of manual selection of drugs based on experience can be overcome to a certain extent, thereby effectively improving the accuracy and rationality of the use of antibiotics.

[0041] As described above, the patient data management system 100 based on the cloud platform according to the embodiment of the present application can be implemented in various terminal devices, such as a server for patient data management based on the cloud platform. In one example, the patient data management system 100 based on the cloud platform according to the embodiment of the present application can be integrated into the terminal device as a software module and / or a hardware module. For example, the patient data management system 100 based on the cloud platform can be a software module in the operating system of the terminal device, or can be an application developed for the terminal device; of course, the patient data management system 100 based on the cloud platform can also be one of the many hardware modules of the terminal device.

[0042] Alternatively, in another example, the cloud-based patient data management system 100 and the terminal device may also be separate devices, and the cloud-based patient data management system 100 may be connected to the terminal device via a wired and / or wireless network and transmit interactive information in accordance with an agreed data format.

Claims

1. A patient data management system based on a cloud platform, characterized in that: include: The patient information collection module is used to obtain the patient's basic information and surgical information from the cloud platform. The basic information includes age, liver and kidney function, allergy history, and past medical history. The surgical information includes the type of surgery, surgery time, and surgical incision. A patient information encoding module, used for performing feature analysis and feature fusion on the basic information and surgical information of the patient to obtain a semantic feature vector of the patient's antibiotic use reference information; A feature association module, used for performing association coding on the semantic feature vector of the patient's antibiotic use reference information to obtain a semantic association feature vector of the patient's antibiotic use reference information; A feature fusion module, used for performing target domain data distribution matching based on principal component analysis on the semantic feature vector of the patient antibiotic use reference information and the semantic association feature vector of the patient antibiotic use reference information to obtain a fused semantic feature vector of the patient antibiotic use reference information; The antimicrobial drug recommendation result generation module is used to fuse the semantic feature vector based on the patient's antimicrobial drug use reference information to obtain the name of the antimicrobial drug suitable for the patient.

2. The cloud-based patient data management system according to claim 1, characterized in that: The patient information encoding module comprises: A patient basic information semantic feature extraction unit, used for extracting semantic features from the patient's basic information to obtain a patient basic information semantic feature vector; A patient surgery information semantic feature extraction unit, used for extracting semantic features from the patient surgery information to obtain a patient surgery information semantic feature vector; The patient basic information and surgical information semantic feature fusion unit is used to fuse the patient basic information semantic feature vector and the patient surgical information semantic feature vector to obtain the patient antibiotic use reference information semantic feature vector.

3. The cloud-based patient data management system according to claim 2, characterized in that: The patient basic information semantic feature extraction unit is used to: segment the patient basic information and pass it through the patient basic information semantic encoder to obtain the patient basic information semantic feature vector.

4. The cloud-based patient data management system according to claim 3, characterized in that: The patient surgery information semantic feature extraction unit is used to: segment the patient surgery information and pass it through a patient surgery information semantic encoder to obtain the patient surgery information semantic feature vector.

5. The cloud-based patient data management system according to claim 4, characterized in that: The feature association module is used to: divide the semantic feature vector of the patient antibiotic use reference information into multiple sub-vectors and input them into a semantic information association encoder to obtain the semantic association feature vector of the patient antibiotic use reference information.

6. The cloud-based patient data management system according to claim 5, characterized in that: The patient basic information semantic encoder and the patient surgery information semantic encoder are bidirectional long short-term memory neural network models including a word embedding layer, and the semantic information association encoder is a semantic information association encoder based on a converter structure.

7. The cloud-based patient data management system according to claim 6, characterized in that: The feature fusion module comprises: A semantic intrinsic decomposition unit for patient antibiotic use reference information, used for mapping the semantic feature vector of the patient antibiotic use reference information and the semantic association feature vector of the patient antibiotic use reference information to an intrinsic decomposition space to obtain the semantic intrinsic feature vector of the patient antibiotic use reference information and the semantic association intrinsic feature vector of the patient antibiotic use reference information; A semantic feature alignment unit for patient antibiotic use reference information, used for performing cross-domain feature alignment based on the target domain scale on the semantic intrinsic feature vector of the patient antibiotic use reference information and the semantic association intrinsic feature vector of the patient antibiotic use reference information to obtain a semantically modulated intrinsic feature vector of the patient antibiotic use reference information and a semantically associated modulated intrinsic feature vector of the patient antibiotic use reference information; A fine-grained interactive semantic information encoding unit, used to construct a fine-grained interactive semantic matrix of the patient antibiotic use reference information between the intrinsic feature vector after semantic modulation of the patient antibiotic use reference information and the intrinsic feature vector after semantic association modulation of the patient antibiotic use reference information; A patient antibiotic use reference information fusion semantic feature generation unit is used to fuse the patient antibiotic use reference information semantically modulated intrinsic feature vector and the patient antibiotic use reference information semantically associated modulated intrinsic feature vector based on the patient antibiotic use reference information fine-grained interactive semantic matrix to obtain the patient antibiotic use reference information fusion semantic feature vector.

8. The cloud-based patient data management system according to claim 7, characterized in that: The fine-grained interactive semantic information encoding unit is used to: The patient antibiotic use reference information semantically modulated intrinsic feature vector and the transposed vector of the patient antibiotic use reference information semantically modulated intrinsic feature vector are multiplied and then divided by the scale of the patient antibiotic use reference information semantically modulated intrinsic feature vector to obtain the patient antibiotic use reference information fine-grained interaction matrix; A convolution encoding operation based on a convolution kernel of 3×3 is performed on the fine-grained interaction matrix of the patient antibiotic use reference information to obtain a fine-grained interaction semantic matrix of the patient antibiotic use reference information.

9. The cloud-based patient data management system according to claim 8, characterized in that: The patient's antimicrobial drugs use reference information fusion semantic feature generation unit, which is used to: Multiplying the fine-grained interactive semantic matrix of the patient antibiotic use reference information by the semantically modulated intrinsic feature vector of the patient antibiotic use reference information to obtain a semantic feature vector of the patient antibiotic use reference information after information field mapping; Multiplying the fine-grained interaction semantic matrix of the patient antibiotic use reference information by the semantic association modulated intrinsic feature vector of the patient antibiotic use reference information to obtain the semantic association feature vector of the patient antibiotic use reference information after information field mapping; The semantic feature vector of the patient antibiotic use reference information after the information field mapping and the semantic association feature vector of the patient antibiotic use reference information after the information field mapping are weighted fused to obtain the fused semantic feature vector of the patient antibiotic use reference information.

10. The cloud-based patient data management system according to claim 9, characterized in that: The antimicrobial drug recommendation result generation module is used to: fuse the patient's antimicrobial drug usage reference information with a semantic feature vector through an antimicrobial drug recommendation generator to obtain a generation result, and the generation result is used to represent the name of the antimicrobial drug recommended to the patient.

Images