Intelligent medicine management system and method based on RFID

Abstract

The invention relates to the field of medicine management, and particularly discloses an intelligent medicine management system and method based on RFID, and the method comprises the steps: firstly obtaining the medicine information collected by the RFID and the medicine inventory information collected by a database, then carrying out the feature extraction and correlation analysis of the two information through a deep learning technology, and finally obtaining a classification result through a classifier. Therefore, the inventory change can be quickly responded, the state of the medicine can be ensured to be mastered at any time, the time and energy of manual checking can be greatly reduced, and the overall operation efficiency can be improved.

Description

Technical Field

[0001] The present application relates to the field of drug management, and more specifically, to an RFID-based intelligent drug management system and method. Background Art

[0002] Drugs are chemical substances or biological agents used to prevent, treat or alleviate diseases. In order to achieve scientific and standardized management of drug warehouses, it is necessary to store them reasonably to ensure the storage quality of drugs. At the same time, it is also very important to take inventory of drug inventory regularly. At present, drug inventory is usually carried out manually, and the contents of the inventory include the quantity of drugs, expiration date, and possible misplacement.

[0003] However, manual inventory carries the risk of errors and omissions, is inefficient, and may pose safety risks, making it difficult to effectively prevent erroneous, expired, or damaged drugs from entering the market.

[0004] Therefore, an intelligent drug management system and method based on RFID is desired. Summary of the invention

[0005] In order to solve the above technical problems, this application is proposed. The embodiment of this application provides an RFID-based intelligent drug management system and method, which first obtains drug information collected by RFID and drug inventory information collected by a database, then uses deep learning technology to perform feature extraction and association analysis on the two, and finally obtains classification results through a classifier to determine whether to issue a drug inventory warning, thereby quickly responding to inventory changes, ensuring that the drug status is mastered at any time, and greatly reducing the time and energy of manual inventory, thereby improving overall operational efficiency.

[0006] According to one aspect of the present application, there is provided an RFID-based intelligent drug management system, comprising:

[0007] A drug data acquisition module is used to acquire drug information collected by RFID and drug inventory information collected by the database;

[0008] A drug data extraction module, used to extract drug information phrase granularity feature vectors and drug inventory information text global feature vectors from the drug information collected by the RFID and the drug inventory information collected by the database;

[0009] The drug inventory warning judgment module is used to judge whether to issue a drug inventory warning based on the drug information phrase granularity feature vector and the drug inventory information text global feature vector.

[0010] According to another aspect of the present application, there is provided an RFID-based intelligent drug management method, comprising:

[0011] Obtain drug information collected by RFID and drug inventory information collected by database;

[0012] Extracting a drug information phrase granularity feature vector and a drug inventory information text global feature vector from the drug information collected by the RFID and the drug inventory information collected by the database;

[0013] Based on the drug information phrase granularity feature vector and the drug inventory information text global feature vector, it is determined whether to issue a drug inventory warning.

[0014] Compared with the prior art, the present application provides an RFID-based intelligent drug management system and method, which first obtains drug information collected by RFID and drug inventory information collected by a database, then uses deep learning technology to perform feature extraction and correlation analysis on the two, and finally obtains classification results through a classifier to determine whether to issue a drug inventory warning, thereby quickly responding to inventory changes, ensuring that the status of drugs is mastered at any time, and greatly reducing the time and effort of manual inventory, thereby improving overall operational efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] 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.

[0016] Figure 1 It is a block diagram of an RFID-based intelligent drug management system according to an embodiment of the present application.

[0017] Figure 2 It is a block diagram of a drug data extraction module in an RFID-based intelligent drug management system according to an embodiment of the present application.

[0018] Figure 3 It is a block diagram of a drug information feature extraction unit in an RFID-based intelligent drug management system according to an embodiment of the present application.

[0019] Figure 4 It is a block diagram of a drug inventory information feature extraction unit in an RFID-based intelligent drug management system according to an embodiment of the present application.

[0020] Figure 5 It is a block diagram of a drug inventory warning judgment module in an RFID-based intelligent drug management system according to an embodiment of the present application.

[0021] Figure 6 This is a flow chart of an RFID-based smart drug management method according to an embodiment of the present application. DETAILED DESCRIPTION

[0022] 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.

[0023] Figure 1 FIG. 1 is a block diagram of an RFID-based intelligent drug management system according to an embodiment of the present application. Figure 1 As shown, according to the embodiment of the present application, the RFID-based intelligent drug management system 100 includes: a drug data acquisition module 110, which is used to acquire drug information collected by RFID and drug inventory information collected by a database; a drug data extraction module 120, which is used to extract drug information phrase granularity feature vectors and drug inventory information text global feature vectors from the drug information collected by RFID and the drug inventory information collected by the database; a drug inventory warning judgment module 130, which is used to judge whether to issue a drug inventory warning based on the drug information phrase granularity feature vector and the drug inventory information text global feature vector.

[0024] In the above-mentioned RFID-based intelligent drug management system 100, the drug data acquisition module 110 is used to obtain drug information collected by RFID and drug inventory information collected by the database. It should be understood that as a chemical substance or biological agent used to prevent, treat or alleviate diseases, the warehouse management of drugs needs to be scientific and standardized to ensure the storage quality of drugs. To this end, reasonable storage measures are essential. In addition, regular inventory of drug inventory is also an indispensable link. At present, drug inventory mainly relies on manual operation, involving the quantity of drugs, expiration date, and possible misplacement. However, manual inventory has certain limitations, such as easy errors and omissions, low efficiency, and possible safety hazards. These factors make it difficult for manual inventory to effectively prevent wrong drugs, expired drugs or damaged drugs from entering the market.

[0025] In the technical solution of the present application, it is considered that the RFID smart medical cabinet system uses advanced scientific and technological means and equipment such as software technology, Internet technology, identity recognition technology, sensor technology, radio frequency identification (RFID), and artificial intelligence large model management technology to effectively plan, execute and control the management activities of the entry and exit, storage, distribution and information of various medical supplies. Therefore, by obtaining the drug information collected by RFID and the drug inventory information collected by the database, and combining deep learning technology, it is determined whether it is necessary to issue a drug inventory warning, so as to timely identify and respond to inventory changes, ensure real-time control of drug status, and significantly reduce the time and effort required for manual inventory, thereby improving overall operational efficiency. Among them, RFID technology can realize real-time and automated drug information collection. By adopting ultra-high frequency radio frequency identification (RFID), the advanced radio frequency signal transceiver mechanism enables the system to accurately capture the rich information contained in the RFID tag attached to the medical item in an instant. The system can quickly identify and record the unique identification of the drug, including but not limited to the unique identification of the item, detailed attributes, production batches, and key data such as the expiration date. Compared with the traditional manual recording method, RFID has higher accuracy and efficiency, and can update the storage location and status of the drug in real time, thereby reducing the risk of human errors and omissions. At the same time, RFID technology also supports dynamic monitoring of drug flow, ensuring the traceability of drugs throughout the entire management process. On the other hand, the database provides more comprehensive drug inventory information, including the total inventory of drugs, the number of drugs in transit, historical inventory data, and statistics on expired drugs. In this way, by combining RFID technology and database management systems, drug management can achieve rapid acquisition and accurate storage of information, improving the efficiency and safety of drug management. The integration of this dual source of information enhances the real-time and visualization of drug management.

[0026] In the above-mentioned RFID-based intelligent drug management system 100, the drug data extraction module 120 is used to extract drug information phrase granularity feature vectors and drug inventory information text global feature vectors from the drug information collected by RFID and the drug inventory information collected by the database. In this way, a machine learning model can be constructed to realize intelligent drug management, thereby making drug management more efficient and accurate, and providing strong support for improving patient drug safety.

[0027] Figure 2 FIG. 1 is a block diagram of a drug data extraction module in an RFID-based intelligent drug management system according to an embodiment of the present application. Figure 2As shown, in a specific embodiment of the present application, the drug data extraction module 120 includes: a drug information feature extraction unit 121, used to perform feature extraction on the drug information collected by RFID to obtain the drug information phrase granularity feature vector; a drug inventory information feature extraction unit 122, used to perform feature extraction on the drug inventory information collected by the database to obtain the drug inventory information text global feature vector.

[0028] It should be understood that RFID technology can collect basic information of drugs in real time, such as drug name, production batch number, expiration date, dosage form and inventory status. This information is an indispensable element in drug management. However, the original data is often unstructured or semi-structured, and it is not efficient to directly use this data for analysis or modeling. Therefore, by converting drug information into a vector form that can be used for machine learning and data analysis through feature extraction, data processing capabilities can be effectively improved. Specifically, first, the data collected by RFID needs to be preprocessed, the data needs to be cleaned, and redundant information and noise need to be removed. Next, the drug information is segmented and annotated through natural language processing (NLP) technology to extract meaningful words and phrases. For example, methods such as Bag of Words or TF-IDF (Term Frequency-Inverse Document Frequency) can be used to convert key information such as drug name, ingredients, and dosage into vector representation. These vectors can represent the characteristics of the drug and retain the correlation between information, providing a basis for subsequent machine learning models. In addition, by combining contextual information and specific properties of drugs, deep learning models such as Word2Vec or BERT can be used to map text information to a high-dimensional space and generate richer phrase-granular feature vectors. This approach not only takes into account the independent meaning of each word, but also captures the contextual relationship between words, thereby improving the expressive power of the vector. Ultimately, the resulting phrase-granular feature vector can provide strong support for the intelligent management of drugs, including drug classification, similarity retrieval, and usage trend analysis.

[0029] Furthermore, drug inventory information usually contains multiple dimensions, including total inventory, inventory turnover rate, expired drug ratio, and in-and-out frequency, which can reflect the dynamic changes of drug inventory. By performing feature extraction on drug inventory information in the database, these global features can be extracted and converted into vector representations, thereby better describing the overall situation of drug inventory.

[0030] Figure 3 FIG. 1 is a block diagram of a drug information feature extraction unit in an RFID-based intelligent drug management system according to an embodiment of the present application. Figure 3As shown, in a specific embodiment of the present application, the drug information feature extraction unit 121 includes: a drug information text encoding subunit 1211, which is used to pass the drug information collected by RFID through a converter-based drug information context encoder to obtain multiple drug information text word feature vectors; a drug information text one-dimensional arrangement subunit 1212, which is used to arrange the multiple drug information text word feature vectors into a one-dimensional drug information text word feature vector; a drug information text word multi-scale neighborhood feature extraction subunit 1213, which is used to pass the one-dimensional drug information text word feature vector through a drug information text word multi-scale neighborhood feature extraction module to obtain the drug information phrase granularity feature vector.

[0031] It should be understood that the drug information collected by RFID often exists in the form of unstructured text, and it is not efficient to directly use this data for analysis and modeling. Through the encoder based on the transformer, the drug information can be deeply understood in the context, thereby generating word vectors with high-dimensional features, which is crucial for further machine learning and natural language processing tasks. Among them, the transformer-based context encoder (such as BERT or GPT series models) uses the self-attention mechanism to capture the relationship between each word in the text. Specifically, first, the drug information collected by RFID is input into the encoder, and the model will perform word segmentation on the input text and generate an initial word vector representation. Next, the encoder dynamically calculates the relationship between each word and other words through multiple self-attention layers to generate a word feature vector with context information. In this process, the representation of each word not only depends on its own meaning, but also considers its context position and semantic relationship in the sentence. Finally, the transformer extracts features layer by layer by performing multiple iterative calculations on the input drug information text, and generates a word feature vector containing rich context information to capture the meaning of the drug name, the characteristics of the drug ingredients, and the semantic association of other related information to form a multi-dimensional representation. Specifically, the drug information collected by RFID is passed through a converter-based drug information context encoder to obtain multiple drug information text word feature vectors, including: performing word segmentation processing on the drug information collected by RFID to obtain a drug information word sequence; using the embedding layer of the converter-based drug information context encoder to respectively map each drug information word in the drug information word sequence into a word embedding vector to obtain a sequence of drug information word embedding vectors; using the converter-based Bert model of the converter-based drug information context encoder to perform global context semantic encoding on the sequence of drug information word embedding vectors to obtain multiple drug information text word feature vectors.

[0032] Furthermore, the text data of drug information is usually complex and contains a combination of multiple words. Using a one-dimensional vector representation can effectively compress this information into a more manageable form while maintaining the semantic characteristics of the original information. Specifically, all word feature vectors are added one by one and then divided by the number of words to obtain an average vector representing the entire drug information text. This method is simple and effective and can retain the overall semantic information. In this way, the key features of drug information can be retained, and the dimension of the data is simplified, which is convenient for model processing and understanding.

[0033] Furthermore, the one-dimensional drug information text word feature vector is processed by the drug information text word multi-scale neighborhood feature extraction module, which can capture more detailed semantic features in drug information, thereby providing more effective data support for intelligent drug management and decision-making. Among them, in the text analysis of drug information, drug information is usually represented in the form of words, and its semantic features may change in different contexts. Therefore, a single one-dimensional vector may not fully express the richness of drug information. Through the multi-scale neighborhood feature extraction module, information can be extracted at different scales to capture the local and global contextual relationships between words. Specifically, the module considers the contextual information of surrounding words, including the combination and mutual influence of phrases, so as to generate a more semantically deep phrase granular feature vector. In the implementation process, the one-dimensional drug information text word feature vector is first input into the multi-scale neighborhood feature extraction module. The module observes the text data by setting different window sizes (or "neighborhoods"), and the size of the window represents different feature scales. For example, a small window can focus on the subtle relationship between adjacent words, while a large window can capture the semantic association between words and distant texts. By performing feature extraction at multiple scales, the module can generate multiple different feature representations, which can be combined into a more comprehensive phrase-granular feature vector in subsequent processing. Next, these feature vectors can be fused by weighting or concatenation to retain the information extracted at each scale. Ultimately, the resulting phrase-granular feature vector contains fine-grained features of drug information, enabling machine learning models to more effectively understand and process drug-related data, providing a reliable basis for intelligent decision-making in drug management.

[0034] In a specific embodiment of the present application, the drug information text word multi-scale neighborhood feature extraction module includes: a first convolution layer, a second convolution layer parallel to the first convolution layer, and a cascade layer connected to the first convolution layer and the second convolution layer, wherein the first convolution layer uses a one-dimensional convolution kernel with a first scale, and the second convolution layer uses a one-dimensional convolution kernel with a second scale. More specifically, the one-dimensional drug information text word feature vector is one-dimensionally convolutionally encoded using the first convolution layer of the drug information text word multi-scale neighborhood feature extraction module to obtain a first scale feature vector; the one-dimensional drug information text word feature vector is one-dimensionally convolutionally encoded using the second convolution layer of the drug information text word multi-scale neighborhood feature extraction module to obtain a second scale feature vector; the first scale feature vector and the second scale feature vector are cascaded to obtain the drug information phrase granularity feature vector.

[0035] It should be understood that the design of the multi-scale neighborhood feature extraction module for drug information text words includes the first convolution layer, the parallel second convolution layer, and the cascade layer connected to these two layers. The diversity of its structure and function enables the module to effectively capture different semantic features of drug information text. The core concept of this module is to obtain the feature representation of drug information at different context scales through multi-scale convolution operations, thereby improving the accuracy and depth of information extraction. Among them, the first convolution layer uses a one-dimensional convolution kernel with the first scale, which means that it can focus on a smaller context range. Through this layer, the model can extract subtle relationships between adjacent words. The second convolution layer is parallel to the first convolution layer, and it uses a one-dimensional convolution kernel with the second scale. The setting of this layer is designed to capture a wider range of contextual relationships, focus on the larger distance between words, and can identify semantic associations at the phrase or sentence level. Finally, the cascade layer combines the features extracted by the two convolution layers to form a unified feature representation, so that the final generated feature vector contains both fine-grained semantic information and the semantic features of the overall context.

[0036] Figure 4 FIG. 1 is a block diagram of a drug inventory information feature extraction unit in an RFID-based intelligent drug management system according to an embodiment of the present application. Figure 4As shown, in a specific embodiment of the present application, the drug inventory information feature extraction unit 122 includes: a drug inventory text encoding subunit 1221, which is used to pass the drug inventory information collected from the database through a drug inventory information text semantic encoder including an embedding layer to obtain multiple drug inventory information text feature vectors; a drug inventory information two-dimensional arrangement subunit 1222, which is used to perform two-dimensional arrangement on the multiple drug inventory information text feature vectors to obtain a drug inventory information text feature matrix; a drug inventory information convolution encoding subunit 1223, which is used to pass the drug inventory information text feature matrix through a convolutional neural network as a drug inventory information text feature extractor to obtain the drug inventory information text global feature vector.

[0037] It should be understood that the drug inventory information collected from the database is processed by a drug inventory information text semantic encoder containing an embedding layer, aiming to convert the drug inventory data into a more expressive feature vector. Among them, considering that drug inventory information is usually irregular and unstructured, it is not efficient to use them directly for analysis. By introducing a text semantic encoder containing an embedding layer, these text information can be converted into dense feature vectors, thereby capturing richer semantic information. In the specific implementation process, the main function of the embedding layer is to convert discrete drug inventory text information into low-dimensional continuous vectors. First, the embedding layer maps each word or phrase to a high-dimensional space to generate corresponding word vectors. These word vectors are obtained through training and learning, and can reflect the semantic relationship and similarity between words. Then, these word vectors are input into the subsequent layers of the drug inventory information text semantic encoder for further feature extraction and integration. After that, the drug inventory information text semantic encoder uses various feature extraction mechanisms, such as convolution, recursion or self-attention mechanism, to process these embedded word vectors. In this process, the encoder can capture the contextual relationship between different words and generate a text feature vector containing global and local features of drug inventory information. These feature vectors can represent the overall status, circulation information and storage conditions of the drugs. Specifically, the drug inventory information collected from the database is segmented to obtain a drug inventory word sequence; the embedding layer of the drug inventory information text semantic encoder including the embedding layer is used to map each drug inventory word in the drug inventory word sequence to a word embedding vector to obtain a sequence of drug inventory word embedding vectors; the converter-based Bert model of the drug inventory information text semantic encoder including the embedding layer is used to perform global context semantic encoding on the sequence of drug inventory word embedding vectors to obtain multiple drug inventory information text feature vectors.

[0038] Furthermore, multiple drug inventory information text feature vectors are arranged in two dimensions to improve the organizational structure of the data. Among them, the generation of the feature matrix can effectively retain the multidimensional characteristics of drug inventory information, thereby providing solid data support for intelligent decision-making. Specifically, the two-dimensional arrangement method usually takes the feature vector of each drug as a row of the matrix, where each row represents the information of a drug. The advantage of this arrangement method is that it can bring together multiple samples of the same category, so that the model can process these data in batches during training and testing. At the same time, the number of columns of the feature matrix represents the feature dimensions contained in each drug, which can be feature dimensions generated by the embedding layer, feature dimensions extracted by the convolutional layer, etc. Here, the structured form of the feature matrix not only improves the efficiency of data processing, but also enhances the performance of the model in analyzing drug inventory information.

[0039] Furthermore, the drug inventory information text feature matrix is ​​processed through a convolutional neural network (CNN) as a feature extractor, which can effectively extract the deep semantic features of drug inventory information. It should be understood that the drug inventory information text feature matrix contains multiple aspects of drug information, but to convert this information into a more expressive global feature vector, it is necessary to use the powerful feature extraction capability of the convolutional neural network. Among them, the convolutional neural network can effectively capture the important patterns and structures in the feature matrix through local perception and weight sharing.

[0040] In a specific embodiment of the present application, the drug inventory information convolutional encoding subunit 1223 is used to: use each layer of the convolutional neural network as a drug inventory information text feature extractor to perform convolution processing, mean pooling processing based on local feature matrices and nonlinear activation processing on the input data in the forward transfer of the layer to output the drug inventory information text global feature vector by the last layer of the convolutional neural network as a drug inventory information text feature extractor, wherein the input of the convolutional neural network as a drug inventory information text feature extractor is the drug inventory information text feature matrix. It should be understood that the drug inventory information text feature matrix is ​​fed to the convolutional neural network as input. Each row of the feature matrix represents a multidimensional feature of a drug, and the design of the convolutional neural network allows the model to learn the complex relationship between these features. In the forward transfer process, the convolution layer performs convolution operations on the feature matrix by sliding multiple convolution kernels to generate multiple feature maps. These convolution operations can help identify local patterns between drug features, such as potential associations between features such as inventory quantity, expiration date and manufacturer. Next, after the convolution process, mean pooling processing is usually performed. This process reduces the feature dimension and extracts the most significant features by calculating the mean of the local feature matrix. This pooling method can effectively retain the importance of features while reducing the computational complexity and overfitting risk, thereby enhancing the generalization ability of the model. Through mean pooling, the convolutional encoding subunit further extracts comprehensive features about drug inventory information, allowing the model to better adapt to the performance of various drugs. In addition, a nonlinear activation function (such as ReLU or Sigmoid) is usually applied after each convolution layer. This step introduces nonlinear factors, allowing the model to learn more complex feature relationships, thereby improving the model's expressive power. The activation function converts linear outputs into nonlinear outputs, allowing the network to more effectively capture the nonlinear relationship between features. Finally, after multiple layers of convolution processing, mean pooling, and nonlinear activation, the last layer of the convolutional neural network outputs a global feature vector of drug inventory information text. This global feature vector integrates the key information in the input feature matrix, can effectively represent the overall status and characteristics of drug inventory information, and provides strong support for subsequent analysis and decision-making.

[0041] In the above-mentioned RFID-based intelligent drug management system 100, the drug inventory warning judgment module 130 is used to determine whether to issue a drug inventory warning based on the drug information phrase granularity feature vector and the drug inventory information text global feature vector. It should be understood that the drug information phrase granularity feature vector contains specific information about the drug, while the drug inventory information text global feature vector summarizes the inventory status, circulation situation and demand forecast as a whole. The combination of these two feature vectors involves the establishment of feature fusion and decision-making models. Through the comprehensive analysis of these two feature vectors, a comprehensive perspective is provided, which helps to timely discover inventory problems and take corresponding measures, thereby ensuring the stability and safety of drug supply, and thus improving the quality and efficiency of medical services.

[0042] Figure 5 FIG. 1 is a block diagram of a drug inventory early warning judgment module in an RFID-based intelligent drug management system according to an embodiment of the present application. Figure 5 As shown, in a specific embodiment of the present application, the drug inventory warning judgment module 130 includes: a drug data feature fusion unit 131, which is used to quantify the topological association relationship feature expression of the drug information phrase granularity feature vector and the drug inventory information text global feature vector based on the perspective co-projection space to obtain a drug inventory warning judgment classification feature vector; a drug inventory warning classification judgment unit 132, which is used to pass the drug inventory warning judgment classification feature vector through a classifier to obtain a classification result, and the classification result is used to determine whether to issue a drug inventory warning.

[0043] It should be understood that the fusion of the drug information phrase granularity feature vector and the drug inventory information text global feature vector to obtain the drug inventory warning judgment classification feature vector is to capture the drug inventory status and its potential risks more comprehensively and accurately. This fusion process can give full play to the advantages of the two feature vectors, and improve the effect and reliability of the model in judging inventory warnings by integrating information from different dimensions. Among them, the drug information phrase granularity feature vector contains detailed attributes about the drug, such as the drug category, dosage form, manufacturer, expiration date, and historical sales data. This information can provide an in-depth understanding of the drug itself, thereby helping to analyze the performance of a specific drug in the market. The drug inventory information text global feature vector reflects the current status of drug inventory as a whole, including inventory quantity, turnover speed, and demand forecast. The combination of the two can effectively reveal the supply and demand dynamics of drugs and help identify possible inventory shortages or surpluses.

[0044] In particular, in the technical solution of the present application, due to the different generation processes and information dimensions represented by the drug information phrase granularity feature vector and the drug inventory information text global feature vector, the two have significant feature heterogeneity. Among them, the drug information phrase granularity feature vector processes the data collected by RFID, focusing on the information details and text content of specific drugs, usually including fine-grained information such as drug name, dosage form, specification, and manufacturer, emphasizing the characteristics and uses of drugs, with high text feature richness, and can reflect the semantics and context of drugs. In contrast, the drug inventory information text global feature vector is obtained by processing drug inventory information through a convolutional neural network, and mainly focuses on the overall status and distribution of the inventory, including macro-level information on inventory quantity, drug circulation frequency, and expiration status, emphasizing the management and monitoring of inventory, so its features are more holistic and global. Due to this heterogeneity, the problem of feature manifold and feature space incompatibility may occur when performing feature fusion. If there is a large distance or distribution difference between the drug information phrase granularity feature vector and the drug inventory information text global feature vector in the feature space, it may cause the fused drug inventory warning judgment classification feature vector to be difficult to form a smooth manifold, thereby affecting the certainty of the feature expression. Specifically, the heterogeneity of the features may make it difficult for the classifier to effectively capture potential association patterns, resulting in instability and inaccuracy in the classification results. Therefore, in the technical solution of the present application, the drug information phrase granularity feature vector and the drug inventory information text global feature vector are quantified based on the topological association relationship feature expression of the perspective co-projection space to obtain the drug inventory warning judgment classification feature vector.

[0045] Among them, the drug information phrase granularity feature vector and the drug inventory information text global feature vector are quantified based on the topological association relationship feature expression of the perspective co-projection space to obtain the drug inventory early warning judgment classification feature vector, including: constructing a drug distance topological matrix between the drug information phrase granularity feature vector and the drug inventory information text global feature vector; constructing a drug feature value granularity association matrix between the drug information phrase granularity feature vector and the drug inventory information text global feature vector; performing topological association modulation on the drug feature value granularity association matrix based on the drug distance topological matrix to obtain a drug topological modulation association matrix; using the drug topological modulation association matrix as the perspective co-projection space, projecting the drug information phrase granularity feature vector and the drug inventory information text global feature vector to the perspective co-projection space to obtain the perspective modulated drug information phrase granularity feature vector and the perspective modulated drug inventory information text global feature vector; fusing the perspective modulated drug information phrase granularity feature vector and the perspective modulated drug inventory information text global feature vector to obtain the drug inventory early warning judgment classification feature vector.

[0046] The fusion step is specifically expressed as:

[0047] M t =D(V 1 ,V 2 )

[0048] D(V 1 ,V 2 ) is the drug distance topological matrix between the drug information phrase granularity feature vector and the drug inventory information text global feature vector, that is, d i,j =d(v 1i ,v 2j ), and V 1 and V 2 are column vectors.

[0049]

[0050]

[0051] V c =α·V 1 ′+β·V 2 '

[0052] Among them, V 1 represents the granularity feature vector of the drug information phrase, V 2 represents the global feature vector of the drug inventory information text, M t represents the drug distance topology matrix, T represents the transpose of the vector, Represents matrix multiplication, M v represents the drug feature value granularity association matrix, M c represents the drug topological modulation association matrix, V 1 ′ represents the granularity feature vector of drug information phrase modulated by viewing angle, V 2 ′ represents the global feature vector of the drug inventory information text modulated by perspective, α and β represent weighted hyperparameters, V c Represents the classification feature vector of drug inventory warning judgment.

[0053] In the technical solution of the present application, the drug information phrase granularity feature vector and the drug inventory information text global feature vector have relatively significant feature heterogeneity, which results in that when the feature fusion of the drug information phrase granularity feature vector and the drug inventory information text global feature vector is performed, the feature manifold and feature space mismatch caused by the heterogeneity will cause the drug inventory warning judgment classification feature vector to fail to have feature manifold smoothness, thereby affecting the certainty of its feature expression.

[0054] Based on this, in the technical solution of the present application, the granular feature vector of the drug information phrase and the global feature vector of the drug inventory information text are quantitatively expressed in terms of the topological association relationship based on the perspective co-projection space, which first constructs a drug distance topological matrix between the granular feature vector of the drug information phrase and the global feature vector of the drug inventory information text. This step utilizes the idea of ​​graph theory, takes the eigenvalues ​​of each position in the granular feature vector of the drug information phrase and the global feature vector of the drug inventory information text as nodes, and uses the distance metric function to determine the low-dimensional embedding expression of the edges between nodes, so as to construct a feature expression that can quantify the topological association relationship between the granular feature vector of the drug information phrase and the global feature vector of the drug inventory information text. This step is crucial for understanding the spatial distribution of data points and their mutual connections, and also provides basic structural information for subsequent steps.

[0055] Next, a drug feature value granularity association matrix is ​​constructed between the drug information phrase granularity feature vector and the drug inventory information text global feature vector. This step focuses on exploring the correlation between the internal attributes of the feature vector. In a specific example, the product between the drug information phrase granularity feature vector and the transposed vector of the drug inventory information text global feature vector is calculated to obtain the drug feature value granularity association matrix. Furthermore, the drug feature value granularity association matrix is ​​topologically modulated based on the drug distance topological matrix to obtain a drug topological modulation association matrix. That is, the feature value granularity association information of the drug information phrase granularity feature vector and the drug inventory information text global feature vector is driven to wander on the distance topological map to construct a low-dimensional modulated perspective co-projection space domain for mapping the drug information phrase granularity feature vector and the drug inventory information text global feature vector to a continuous high-dimensional regression space attribute.

[0056] Next, the drug topology modulation association matrix is ​​used as the perspective co-projection space, and the drug information phrase granularity feature vector and the drug inventory information text global feature vector are projected into the perspective co-projection space to obtain the perspective modulated drug information phrase granularity feature vector and the perspective modulated drug inventory information text global feature vector. That is, the key to this technical solution is to find an implicit third space with modulation capability, in which as much information of the original high-dimensional data as possible is retained, so that the converted new feature vector not only reduces the dimension, but also maintains the relative position relationship between the original features.

[0057] Finally, the perspective-modulated drug information phrase granularity feature vector and the perspective-modulated drug inventory information text global feature vector are fused to obtain the drug inventory early warning judgment classification feature vector. Fusion can be achieved in a variety of forms, ranging from simple arithmetic average to complex ensemble learning algorithms, depending on the application scenario and the goals pursued. The final drug inventory early warning judgment classification feature vector combines the main features of the two sets of input data and provides a more comprehensive data representation, which is not only conducive to improving the performance of machine learning models, but also promotes the ability of cross-modal or multi-source data analysis, thereby significantly enhancing the understanding and utilization efficiency of complex data structures.

[0058] Furthermore, the drug inventory early warning judgment classification feature vector contains multi-dimensional feature information. After fusion, these features can fully reflect the current inventory status, market demand and historical sales data of the drug. Using a classifier to process this feature vector can efficiently identify inventory problems and make corresponding early warning decisions. In the implementation process, the fused drug inventory early warning judgment classification feature vector is input into the classifier for training. In the training stage, the classifier uses the existing labeled data to learn the relationship between the feature vector and the early warning decision. Then, through training, the classifier can establish a mapping relationship, that is, predict the corresponding classification results based on the input drug inventory early warning judgment classification feature vector. These classification results are usually binary classification ("issue an early warning" or "no early warning"). Once the classifier is trained, the new drug inventory early warning judgment classification feature vector can be input into the classifier in actual applications to obtain the classification results in real time. Finally, when the classification results are obtained, these results can be further analyzed to determine whether an inventory early warning is needed. This process not only improves the efficiency of inventory management, but also reduces the risks caused by insufficient inventory, ensuring the stability and safety of drug supply.

[0059] In a specific embodiment of the present application, an RFID-based intelligent drug management system includes a coding end, a client and a cloud platform management end. Among them, the client, as the core service implementation unit of the system, provides a variety of services such as inventory management, item traceability, authority control and data analysis. The various algorithms and data processing capabilities required by the client are supported and completed by the cloud platform management end, thereby achieving efficient information management and resource optimization. The architecture design of the system ensures the real-time and accuracy of the data, which helps to improve the overall operational efficiency. Among them, the RFID-based intelligent drug management system installs RFID reading and writing equipment at a specific location of the storage cabinet to collect RFID tag information on items or medicines. In addition, during the process of accessing the cabinet, the device can effectively check the item information to ensure the accuracy of the read cargo information. After the pallet label information is successfully collected, it is sent to the cloud platform management system in real time through the network, which can control the opening and closing of the control lock and the cabinet door, and interact with the big database in real time to ensure the accuracy and integrity of the data. In this way, the management efficiency and security of the storage cabinet are improved.

[0060] Furthermore, the software platform developed by the project of this application can be integrated with the hospital's information system (such as HIS or EMR), and the data can be automatically transmitted to the hospital information system, real-time tracking of the entry and exit of drugs, and automatic updating of the cloud library. Authorized personnel can check the hospital's material status on their mobile phones or computers anytime and anywhere.

[0061] It is worth mentioning that in the technical solution of this application, compact compartments are designed for small medical devices, while spacious exclusive areas are planned for large equipment. By carefully selecting RFID tags, reasonably installing locations and improving reading stability, it is ensured that each medical item can be accurately identified and tracked. Among them, the design focus of the intelligent lock control system is to achieve safe and reliable authority management, and only authorized personnel can open the cabinet smoothly. At the same time, the research on temperature and humidity control systems aims to ensure that the internal environment of the medical cabinet meets the storage requirements of various medical supplies, especially for certain medicines, which require precise control within a specific temperature and humidity range.

[0062] Specifically, the technical solution of this application adopts an optimized antenna design to maintain good reading performance in a metal cabinet. Among them, it adopts 860 to 960MHz full-band high-frequency radio frequency identification technology, and transmits and receives radio frequency signals through a high-frequency reader. The maximum identification distance can exceed one meter, and the accuracy rate is above 99.9%, realizing accurate reading of medical item information.

[0063] In summary, the embodiment of the present application first obtains the drug information collected by RFID and the drug inventory information collected by the database, and then uses deep learning technology to perform feature extraction and correlation analysis on the two, and finally obtains the classification result through the classifier to determine whether to issue a drug inventory warning, so as to quickly respond to inventory changes, ensure that the status of drugs is mastered at any time, and greatly reduce the time and effort of manual inventory, thereby improving overall operational efficiency.

[0064] As described above, the RFID-based smart drug management system 100 according to the embodiment of the present application can be implemented in various terminal devices. In one example, the RFID-based smart drug management system 100 can be integrated into the terminal device as a software module and / or a hardware module. For example, the RFID-based smart drug management system 100 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 RFID-based smart drug management system 100 can also be one of the many hardware modules of the terminal device.

[0065] Alternatively, in another example, the RFID-based smart drug management system 100 and the terminal device may also be separate devices, and the RFID-based smart drug 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.

[0066] Figure 6 FIG. 1 is a flow chart of an RFID-based intelligent drug management method according to an embodiment of the present application. Figure 6 As shown, according to the RFID-based intelligent drug management method of the embodiment of the present application, it includes: S110, acquiring drug information collected by RFID and drug inventory information collected by a database; S120, extracting drug information phrase granularity feature vectors and drug inventory information text global feature vectors from the drug information collected by RFID and the drug inventory information collected by the database; S130, judging whether to issue a drug inventory warning based on the drug information phrase granularity feature vectors and the drug inventory information text global feature vectors.

[0067] Here, those skilled in the art can understand that the specific operations of each step in the above-mentioned RFID-based intelligent drug management method have been referred to above. Figures 1 to 5 The description of the RFID-based intelligent drug management system has been introduced in detail, and therefore, its repeated description will be omitted.

Claims

1. An intelligent drug management system based on RFID, characterized in that: include: A drug data acquisition module is used to acquire drug information collected by RFID and drug inventory information collected by the database; A drug data extraction module, used to extract drug information phrase granularity feature vectors and drug inventory information text global feature vectors from the drug information collected by the RFID and the drug inventory information collected by the database; The drug inventory warning judgment module is used to judge whether to issue a drug inventory warning based on the drug information phrase granularity feature vector and the drug inventory information text global feature vector.

2. The RFID-based intelligent drug management system according to claim 1 is characterized in that: The drug data extraction module comprises: A drug information feature extraction unit, used for extracting features from the drug information collected by the RFID to obtain a feature vector of the drug information phrase granularity; The drug inventory information feature extraction unit is used to extract features from the drug inventory information collected from the database to obtain a global feature vector of the drug inventory information text.

3. The RFID-based intelligent drug management system according to claim 2 is characterized in that: The drug information feature extraction unit comprises: A drug information text encoding subunit, used for passing the drug information collected by the RFID through a converter-based drug information context encoder to obtain a plurality of drug information text word feature vectors; a drug information text one-dimensional arrangement subunit, used to arrange the plurality of drug information text word feature vectors into a one-dimensional drug information text word feature vector; The drug information text word multi-scale neighborhood feature extraction subunit is used to pass the one-dimensional drug information text word feature vector through the drug information text word multi-scale neighborhood feature extraction module to obtain the drug information phrase granularity feature vector.

4. The RFID-based intelligent drug management system according to claim 3 is characterized in that: The drug information text word multi-scale neighborhood feature extraction module includes: a first convolution layer, a second convolution layer parallel to the first convolution layer, and a cascade layer connected to the first convolution layer and the second convolution layer, wherein the first convolution layer uses a one-dimensional convolution kernel with a first scale, and the second convolution layer uses a one-dimensional convolution kernel with a second scale.

5. The RFID-based intelligent drug management system according to claim 4 is characterized in that: The drug inventory information feature extraction unit includes: A drug inventory text encoding subunit, configured to pass the drug inventory information collected from the database through a drug inventory information text semantic encoder including an embedding layer to obtain a plurality of drug inventory information text feature vectors; A drug inventory information two-dimensional arrangement subunit, used for performing two-dimensional arrangement on the plurality of drug inventory information text feature vectors to obtain a drug inventory information text feature matrix; The drug inventory information convolutional encoding subunit is used to obtain the drug inventory information text global feature vector by passing the drug inventory information text feature matrix through a convolutional neural network serving as a drug inventory information text feature extractor.

6. The RFID-based intelligent drug management system according to claim 5, characterized in that: The drug inventory information convolutional encoding subunit is used to: use each layer of the convolutional neural network as the drug inventory information text feature extractor to perform convolution processing, mean pooling processing based on the local feature matrix and nonlinear activation processing on the input data in the forward transmission of the layer so that the last layer of the convolutional neural network as the drug inventory information text feature extractor outputs the drug inventory information text global feature vector, wherein the input of the convolutional neural network as the drug inventory information text feature extractor is the drug inventory information text feature matrix.

7. The RFID-based intelligent drug management system according to claim 6, characterized in that: The drug inventory early warning judgment module includes: A drug data feature fusion unit is used to quantify the topological association relationship feature expression based on the perspective co-projection space of the drug information phrase granularity feature vector and the drug inventory information text global feature vector to obtain a drug inventory early warning judgment classification feature vector; The drug inventory warning classification judgment unit is used to pass the drug inventory warning judgment classification feature vector through a classifier to obtain a classification result, and the classification result is used to determine whether to issue a drug inventory warning.

8. The RFID-based intelligent drug management system according to claim 7, characterized in that: The drug data feature fusion unit includes: Constructing a drug distance topology matrix between the drug information phrase granularity feature vector and the drug inventory information text global feature vector; Constructing a drug feature value granularity association matrix between the drug information phrase granularity feature vector and the drug inventory information text global feature vector; Performing topological association modulation on the drug characteristic value granularity association matrix based on the drug distance topological matrix to obtain a drug topological modulation association matrix; Taking the drug topological modulation association matrix as the perspective co-projection space, projecting the drug information phrase granularity feature vector and the drug inventory information text global feature vector into the perspective co-projection space to obtain the perspective-modulated drug information phrase granularity feature vector and the perspective-modulated drug inventory information text global feature vector; The perspective-modulated drug information phrase granularity feature vector and the perspective-modulated drug inventory information text global feature vector are fused to obtain the drug inventory early warning judgment classification feature vector.

9. An intelligent drug management method based on RFID, characterized in that: include: Obtain drug information collected by RFID and drug inventory information collected by database; Extracting a drug information phrase granularity feature vector and a drug inventory information text global feature vector from the drug information collected by the RFID and the drug inventory information collected by the database; Based on the drug information phrase granularity feature vector and the drug inventory information text global feature vector, it is determined whether to issue a drug inventory warning.

10. The RFID-based intelligent drug management method according to claim 9, characterized in that: Extracting a drug information phrase granularity feature vector and a drug inventory information text global feature vector from the drug information collected by the RFID and the drug inventory information collected by the database, including: Performing feature extraction on the drug information collected by the RFID to obtain a feature vector of the drug information phrase granularity; Feature extraction is performed on the drug inventory information collected from the database to obtain a global feature vector of the drug inventory information text.

Images