An intelligent management system for medication during the recovery period of neurological diseases and a method for generating digital prescriptions

Abstract

This application provides an intelligent management system for medication during the rehabilitation period of neurological diseases and a method for generating digital prescriptions, relating to the field of medical data processing technology. It obtains the target handwritten prescription by enhancing the contrast of the light-interference-affected area of ​​the prescription and determining the enhancement cost of the handwritten handwriting structure loss during the contrast enhancement process. Semantic tags are extracted for each drug name in the target handwritten prescription. Based on fuzzy reasoning, the semantic tags of different drug names are clinically matched to obtain the tag membership degree of different drug names, thereby determining the semantic tag loss generated during medical terminology segmentation during semantic tag extraction. Based on the enhancement cost and semantic tag loss, multi-objective optimization is performed on the medical semantic compensation features of the semantic tags to obtain the semantic compensation vector of the semantic tags. A digital prescription for neurological diseases is generated based on the semantic compensation vector. This application can realize the cost constraint of recognition loss during the transcription of handwritten prescriptions.

Description

Technical Field

[0001] This application relates to the field of medical data processing technology, and more specifically, to an intelligent management system for medication during the recovery period of neurological diseases and a method for generating digital prescriptions. Background Technology

[0002] With the continuous development of medical and health management technology, digital prescription generation has become an important means to improve the efficiency of drug management. Especially during the rehabilitation period of neurological diseases, patients' medication needs become more complex, and the automated recognition and management of handwritten prescriptions becomes particularly important. Traditional handwritten prescriptions often rely on doctors' manual writing, and the recording of drug information is cumbersome and prone to errors. Intelligent prescription generation systems based on digital technology can not only improve the safety and accuracy of drug use, but also alleviate the shortage of medical resources and reduce the occurrence of medical errors to a certain extent.

[0003] In existing technologies, handwritten prescription recognition technology mainly relies on optical character recognition (OCR) technology. However, its recognition accuracy is still limited by various factors. Illumination interference and blurred handwriting are common problems, which directly lead to the loss or misrecognition of prescription information. Existing technologies usually solve these problems through image preprocessing and enhancement algorithms, but these methods cannot completely avoid the loss of handwriting structure and may still lead to partial loss or misrecognition of prescription information. In addition, prescriptions usually contain a large number of medical terms and professional names, and current handwritten prescription recognition systems have difficulty in accurately recognizing specific medical terms, especially in the segmentation process of medical terminology scenarios. Existing technologies often cannot effectively distinguish and extract these terms, resulting in missegmentation or inaccurate processing of terms. Therefore, how to effectively reduce the loss of handwriting structure during the transcription of handwritten prescriptions has become an urgent problem to be solved. Summary of the Invention

[0004] This application provides an intelligent management system for medication during the rehabilitation period of neurological diseases and a method for generating digital prescriptions, which can realize the cost constraint of recognition loss during the transcription of handwritten prescriptions.

[0005] In a first aspect, this application provides a method for generating digital prescriptions for neurological diseases, used in an intelligent management system for medication during the recovery period of neurological diseases to generate digital prescriptions. The method includes the following steps:

[0006] Handwritten prescriptions for neurological diseases are collected using scanning equipment. The contrast of the light-interference-affected areas of the handwritten prescriptions is enhanced to obtain the target handwritten prescription with enhanced contrast. Then, the enhancement cost of the loss of handwriting structure during the contrast enhancement process is determined.

[0007] Semantic tags are extracted from the names of each drug in the target handwritten prescription. Based on a pre-defined neurological disease atlas, all semantic tags are embedded in a medical context to obtain a semantic vector for each semantic tag.

[0008] Based on fuzzy reasoning, the semantic tags of different drug names are clinically matched by combining all semantic vectors to obtain the tag membership degree of different drug names. Then, the semantic tag loss generated during the medical term scene segmentation in the semantic tag extraction process is determined by all tag membership degrees and the co-occurrence probability among various drugs in the neurological disease prescription database.

[0009] Based on the enhancement cost and the semantic label loss, multi-objective optimization is performed on the medical semantic compensation features of all semantic labels to obtain the semantic compensation vector of each semantic label. Based on all semantic compensation vectors, a digital prescription for neurological diseases is generated.

[0010] Preferably, the contrast enhancement of the light-interference-affected area of ​​the handwritten prescription to obtain a contrast-enhanced target handwritten prescription specifically includes:

[0011] The pixel distribution of the handwritten prescription is analyzed by grayscale histogram analysis to identify areas of light interference.

[0012] An adaptive gamma correction algorithm is used to enhance the contrast of the light-interference area in layers.

[0013] The enhanced handwritten prescription is binarized to obtain the target handwritten prescription with enhanced contrast.

[0014] Preferably, the enhancement cost for determining the loss of handwriting structure during contrast enhancement specifically includes:

[0015] Extract the edge feature map of the original handwritten prescription and mark the stroke intersections and turning points as topological nodes;

[0016] By comparing the structural differences of topological nodes in the edge feature maps before and after enhancement, the broken areas of the handwriting can be located.

[0017] The enhancement cost of handwriting structure loss during contrast enhancement is determined based on the pixel loss rate of the broken handwriting area.

[0018] Preferably, the semantic tags extracted from each drug name in the target handwritten prescription specifically include:

[0019] A pre-trained handwritten character recognition model is used to segment characters and words in the target handwritten prescription and extract text features of the drug name.

[0020] Semantic validation is performed on the text features of drug names to remove spelling errors and non-drug-related text features;

[0021] The removed text features are standardized and encoded to generate semantic tags for each drug name.

[0022] Preferably, based on a pre-defined neurological disease atlas, all semantic tags are subjected to medical context embedding processing to obtain the semantic vector of each semantic tag, which specifically includes:

[0023] For each semantic tag, the semantic tag is associated with the drug entity in the preset neurological disease atlas, thereby generating the atlas node tag of the semantic tag;

[0024] Generate initial embedding vectors for semantic tags based on the graph node tags;

[0025] By adjusting the vector weights of the initial embedding vector using the contextual information of the target handwritten prescription, the semantic vectors of the semantic tags are output, thus obtaining the semantic vector of each semantic tag.

[0026] Preferably, based on fuzzy reasoning, the semantic tags of different drug names are clinically matched using all semantic vectors to obtain the tag membership degree of different drug names, specifically including:

[0027] Determine the cosine similarity between each semantic vector, and then construct the drug compatibility relationship matrix;

[0028] Potential synergistic combinations in the relationship matrix were screened based on a clinical compatibility rule base.

[0029] The rationality of synergistic combinations is determined by fuzzy logic algorithms, thereby obtaining the label membership degree of different drug names.

[0030] Preferably, the semantic label loss generated during medical terminology scene segmentation in the semantic label extraction process is determined by using all label membership degrees and the co-occurrence probabilities among various drugs in the neurology disease prescription database. Specifically, this includes:

[0031] The co-occurrence probabilities of various drugs in the prescription database for neurological diseases were statistically analyzed, and a co-occurrence probability matrix of drug combinations was constructed.

[0032] Abnormal matching pairs with semantic bias are determined by all tag membership degrees and the co-occurrence probability matrix;

[0033] The semantic label loss during medical terminology scene segmentation in the semantic label extraction process is obtained by backtracking back to the degree of deviation of the abnormal matching pairs.

[0034] Preferably, the medical semantic compensation features of all semantic labels are optimized using a multi-objective approach based on the enhancement cost and the semantic label loss to obtain the semantic compensation vector for each semantic label, specifically including:

[0035] We construct a cost function by taking the semantic compensation features of all semantic labels as optimization variables and minimizing the enhancement cost and semantic label loss.

[0036] The enhancement cost and the semantic label loss are introduced into the cost function in the form of a weighted penalty term;

[0037] The cost function is iteratively solved using a preset optimization algorithm until it converges to the optimal solution, thereby obtaining the semantic compensation vector for each semantic tag.

[0038] Preferably, the scanning device is a flatbed scanner.

[0039] Secondly, this application provides an intelligent management system for medication during the recovery period of neurological diseases. This system includes a digital prescription generation unit, which comprises:

[0040] The acquisition module is used to acquire handwritten prescriptions for neurological diseases through a scanning device, enhance the contrast of the light interference area of ​​the handwritten prescription, obtain the target handwritten prescription with enhanced contrast, and then determine the enhancement cost of the loss of handwriting structure during the contrast enhancement process.

[0041] The processing module is used to extract the semantic tags of each drug name in the target handwritten prescription, and to perform medical context embedding processing on all semantic tags based on the preset neurological disease atlas to obtain the semantic vector of each semantic tag.

[0042] The processing module is also used to perform clinical matching of semantic tags for different drug names based on fuzzy reasoning and combining all semantic vectors to obtain the tag membership degree of different drug names, and then determine the semantic tag loss generated during the medical term scene segmentation in the semantic tag extraction process by using all tag membership degrees and the co-occurrence probability among various drugs in the neurological disease prescription database.

[0043] The execution module is used to perform multi-objective optimization on the medical semantic compensation features of all semantic labels based on the enhancement cost and the semantic label loss, thereby obtaining the semantic compensation vector of each semantic label, and generating a digital prescription for neurological diseases based on all semantic compensation vectors.

[0044] Thirdly, this application provides a computer device including a memory and a processor, the memory storing code, and the processor being configured to acquire the code and execute the above-described method for generating digital prescriptions for neurological diseases.

[0045] Fourthly, this application provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described method for generating digital prescriptions for neurological diseases.

[0046] The technical solutions provided by the embodiments disclosed in this application have the following beneficial effects:

[0047] In this embodiment, handwritten prescriptions for neurological diseases are collected using a scanning device. The contrast of the light-interference-affected areas of the handwritten prescriptions is enhanced to obtain a target handwritten prescription with enhanced contrast. The enhancement cost of the handwritten handwriting structure loss during the contrast enhancement process is then determined. Semantic tags for each drug name in the target handwritten prescription are extracted. Based on a preset neurological disease atlas, all semantic tags are processed using medical context embedding to obtain a semantic vector for each semantic tag. Based on fuzzy reasoning, the semantic tags for different drug names are clinically matched using all semantic vectors to obtain the tag membership degree for different drug names. Then, the semantic tag loss generated during medical terminology scene segmentation during semantic tag extraction is determined by using all tag membership degrees and the co-occurrence probability among various drugs in the neurological disease prescription database. Based on the enhancement cost and the semantic tag loss, multi-objective optimization is performed on the medical semantic compensation features of all semantic tags to obtain a semantic compensation vector for each semantic tag. A digital prescription for neurological diseases is generated based on all semantic compensation vectors.

[0048] Therefore, this application performs multi-objective optimization of the medical semantic compensation features of all semantic labels through the enhancement cost and the semantic label loss, thereby obtaining the semantic compensation vector of each semantic label. First, by enhancing the contrast of the light interference area of ​​the handwritten prescription and determining the enhancement cost of the handwritten handwriting structure loss during the contrast enhancement process, the impact of light interference on the accuracy of handwritten handwriting structure recognition can be reduced. At the same time, the calculation mechanism of the enhancement cost is introduced to quantify the handwriting loss during the enhancement process, effectively avoiding the problem of prescription information loss or misidentification caused by simply relying on the enhancement algorithm, thus ensuring that the original prescription content is recognized as completely and clearly as possible. Second, by extracting the semantic labels of drug names in the target handwritten prescription and performing medical context embedding processing on these labels based on the preset neurological disease atlas, accurate medical semantic information can be given to each drug name in the prescription, overcoming the problem of inaccurate understanding of medical terminology in traditional handwritten prescription recognition technology, thereby effectively reducing the risk of terminology misidentification. Then, by combining all semantic vectors based on fuzzy reasoning to perform clinical matching of semantic labels of different drug names, it is possible to... This method intelligently processes the semantic relationships and interactions of drug names to ensure their semantic accuracy in medical scenarios. By calculating tag membership and co-occurrence probabilities between drugs, it effectively reduces semantic tag loss during semantic tag extraction and medical terminology scene segmentation. This optimizes drug compatibility rules, improves the accuracy and reliability of prescription generation, and avoids missegmentation of drug names and terms, thus significantly enhancing the accuracy and intelligence of the digital prescription generation system. Finally, by combining enhancement costs and semantic tag losses to perform multi-objective optimization on the medical semantic compensation features of all semantic tags, the compensation vector of each semantic tag is precisely adjusted. This optimization process effectively reduces information loss or errors during prescription recognition, ensuring that each drug name and its semantic tag are accurately expressed in a specific medical context. Through multi-objective optimization, the scheme maintains semantic accuracy while further improving the overall quality and reliability of prescription generation, avoiding inaccurate prescriptions or prescriptions that do not meet clinical needs due to tag loss. In summary, this application's scheme can achieve cost constraints on recognition loss during the transcription of handwritten prescriptions. Attached Figure Description

[0049] Figure 1 This is an exemplary flowchart of a digital prescription generation method for neurological diseases according to some embodiments of this application;

[0050] Figure 2 This is a schematic diagram illustrating the application scenario of the intelligent management system for medication during the rehabilitation period of neurological diseases, based on some embodiments of this application.

[0051] Figure 3This is a flowchart illustrating the process of determining tag membership according to some embodiments of this application;

[0052] Figure 4 This is a schematic diagram of the structure of a digital prescription generation unit according to some embodiments of this application;

[0053] Figure 5 This is a schematic diagram of the structure of a computer device for implementing a digital prescription generation method for neurological diseases, according to some embodiments of this application. Detailed Implementation

[0054] To better understand the technical solution of this application, the technical solution of this application will be described in detail below with reference to the accompanying drawings and specific embodiments.

[0055] refer to Figure 1 The figure is an exemplary flowchart of a digital prescription generation method for neurological diseases according to some embodiments of this application. The digital prescription generation method 100 for neurological diseases mainly includes the following steps:

[0056] In step 101, a handwritten prescription for neurological diseases is acquired by scanning equipment, and the contrast of the light interference area of ​​the handwritten prescription is enhanced to obtain the target handwritten prescription with enhanced contrast, thereby determining the enhancement cost of the loss of handwriting structure during the contrast enhancement process.

[0057] In some embodiments, reference Figure 2 As shown in the figure, this figure is a schematic diagram of the application scenario of the intelligent management system for medication during the rehabilitation period of neurological diseases, as illustrated in some embodiments of this application. The data acquisition device is responsible for collecting handwritten prescriptions and sending the data to the server for processing through a communication network. The server uses intelligent algorithms to generate digital prescriptions and stores the results in a data storage device, thereby realizing automated management from the collection of handwritten prescriptions to the generation of digital prescriptions.

[0058] It should be noted that the scanning device in this application is a flatbed scanner; the use of a scanning device to collect handwritten prescriptions for neurological diseases in this application specifically refers to using a flatbed scanner to acquire images of paper prescriptions, ensuring high-resolution image acquisition is completed under natural or auxiliary lighting conditions.

[0059] In some embodiments, contrast enhancement of the light-interference-affected area of ​​the handwritten prescription to obtain a contrast-enhanced target handwritten prescription can be achieved through the following steps:

[0060] The pixel distribution of the handwritten prescription is analyzed by grayscale histogram analysis to identify areas of light interference.

[0061] An adaptive gamma correction algorithm is used to enhance the contrast of the light-interference area in layers.

[0062] The enhanced handwritten prescription is binarized to obtain the target handwritten prescription with enhanced contrast.

[0063] It should be noted that the layered contrast enhancement in this application refers to enhancing the contrast of local areas by using different enhancement parameters according to the characteristics of different brightness areas in the image.

[0064] In specific implementation, firstly, the pixel distribution of the handwritten prescription is analyzed through grayscale histogram analysis to identify areas of light interference. This can be achieved as follows: The acquired handwritten prescription image is converted to grayscale using a conventional weighted average method (e.g., Y=0.299R+0.587G+0.114B, where Y represents the pixel's brightness value, R represents the pixel's red component value, G represents the pixel's green component value, and B represents the pixel's blue component value) to convert the color image (Red-Green-Blue, RGB) into a single-channel grayscale image. Its grayscale histogram is then constructed, and the number of pixels corresponding to each grayscale value (0–255) is counted. The image is divided into several local regions (e.g., 64×64 pixel blocks), and the mean and standard deviation of the grayscale value for each region are calculated. If the grayscale mean of a region is lower than the average of the entire image, or its standard deviation is much smaller than the standard deviation of neighboring regions, then the region is determined to be a light interference region (e.g., a shadow area or a reflective area). Secondly, the adaptive gamma correction algorithm can be used to perform layered contrast enhancement on the light interference region. This can be achieved by: for the identified light interference region, using the adaptive gamma correction algorithm for layered enhancement, that is, dynamically adjusting the gamma value according to the average grayscale of the region; using a gamma value less than 1 to increase brightness in low-brightness areas, and using a gamma value greater than 1 to suppress overexposure in high-brightness areas, ensuring local contrast enhancement without structural distortion. Then, the enhanced handwritten prescription is binarized to obtain the target handwritten prescription with enhanced contrast. This can be achieved by: performing full binarization on the enhanced handwritten prescription image. Figure 2 Value-enhanced processing, using algorithms such as Otsu's to adaptively set thresholds, effectively separates the handwriting from the background, and outputs the target image with enhanced contrast, i.e., the target handwritten prescription.

[0065] In some embodiments, the enhancement cost of the loss of handwriting structure during contrast enhancement can be determined in the following manner:

[0066] Extract the edge feature map of the original handwritten prescription and mark the stroke intersections and turning points as topological nodes;

[0067] By comparing the structural differences of topological nodes in the edge feature maps before and after enhancement, the broken areas of the handwriting can be located.

[0068] The enhancement cost of handwriting structure loss during contrast enhancement is determined based on the pixel loss rate of the broken handwriting area.

[0069] It should be noted that the handwriting breakage area in this application refers to the missing pixels at the edge of the handwriting and the detachment of topological nodes caused by brightness adjustment during the contrast enhancement process, resulting in discontinuous areas such as stroke interruption and connection breakage in the enhanced image. This is used to reflect the degree of damage to the integrity of handwriting structure recognition caused by image enhancement. The enhancement cost in this application is a cost indicator that measures the degree of damage to the integrity of handwriting structure during the contrast enhancement process of handwritten prescription images.

[0070] In specific implementation, firstly, the edge feature map of the original handwritten prescription is extracted, and the stroke intersections and turning points are marked as topological nodes. This can be achieved in the following way: the Canny edge detection algorithm in the existing technology can be used to extract the edges of the original handwritten prescription image, identify the edge feature map of the handwriting, and then perform morphological analysis on the edge feature map. Combining pixel connectivity and curvature changes, the stroke intersections (multi-path pixel intersections) and turning points (points with significant curvature changes) in each character are marked as topological nodes to construct the structural skeleton of the handwriting. Then, the structural differences of the topological nodes in the edge feature maps before and after enhancement are compared to locate the broken areas of the handwriting. This can be achieved in the following way: the same edge extraction and topological node marking operations are performed on the enhanced image, and the images before and after enhancement are registered (based on a preset topological node structure template). The edge feature maps are aligned, and then compared node by node to identify node regions with large positional offsets, topological breaks, or disappearances. These identified regions are then used as handwriting breakage regions. Finally, the enhancement cost of handwriting structure loss during contrast enhancement is determined based on the pixel loss rate of the handwriting breakage regions. This can be achieved in the following way: For each handwriting breakage region, the change in the number of edge pixels within the node coverage area before and after enhancement is counted, and the change is used as the pixel loss rate of that handwriting breakage region, i.e., pixel loss rate = (original number of pixels - enhanced number of pixels) / original number of pixels. The pixel loss rates of all handwriting breakage regions are weighted, and the weight of each handwriting breakage region can be determined using the entropy weight method, which will not be elaborated here. Finally, the weighted value is used as the enhancement cost of handwriting structure loss during contrast enhancement.

[0071] It should also be noted that by comparing the topological changes in the handwriting structure before and after the contrast enhancement of the handwritten prescription image, this application can accurately assess the degree of damage to key handwriting information caused by image enhancement, thereby providing a precise quantitative basis for structural loss for subsequent semantic label compensation, avoiding character recognition errors or incomplete drug name extraction due to excessive enhancement, and ensuring the recognition accuracy and reliability of medication information in the digital prescription generation process.

[0072] In step 102, semantic tags for each drug name in the target handwritten prescription are extracted. Based on a preset neurological disease atlas, all semantic tags are processed by medical context embedding to obtain the semantic vector of each semantic tag.

[0073] In some embodiments, extracting semantic tags for each drug name in a target handwritten prescription can be achieved using the following steps:

[0074] A pre-trained handwritten character recognition model is used to segment characters and words in the target handwritten prescription and extract text features of the drug name.

[0075] Semantic validation is performed on the text features of drug names to remove spelling errors and non-drug-related text features;

[0076] The removed text features are standardized and encoded to generate semantic tags for each drug name.

[0077] It should be noted that the handwritten character recognition model in this application is a deep learning model used to automatically convert handwritten characters in images into text information. Its core technical principle is to combine convolutional neural networks and recurrent neural networks to extract spatial and temporal features from images, and to achieve character alignment and decoding by connecting a temporal classifier. Specifically, the convolutional neural network is responsible for extracting local handwriting feature maps from the image, the recurrent neural network captures the contextual dependencies between characters, and the temporal classifier layer achieves character sequence output without character segmentation annotation, thereby adapting to the irregular layout and continuous stroke features in natural handwriting and achieving high accuracy in recognizing continuous characters in handwritten prescriptions. The semantic label in this application refers to standardized annotation information used to identify the meaning of entities in text. In this solution, it is used to uniquely correspond to the drug name and express its medical semantic attributes.

[0078] In practical implementation, firstly, a pre-trained handwritten character recognition model is used to segment the target handwritten prescription into characters and phrases. Extracting the textual features of the drug name can be achieved as follows: the image of the target handwritten prescription is input into the pre-trained handwritten character recognition model, and the convolutional layers in the model extract local image features. A recurrent neural network captures the temporal relationships between characters, and finally, a temporal classifier decodes and outputs a continuous character sequence, achieving automatic character-to-phrase segmentation. Secondly, semantic verification is performed on the textual features of the drug name to remove spelling errors and non-drug-related textual features. This can be achieved as follows. In short: the identified phrases are matched with medical dictionaries (such as the National Pharmacopoeia), and word vector similarity calculations (such as using Word2Vec) are used to determine whether ambiguous spellings are likely to be legitimate drug names. By setting similarity thresholds and joint verification with context rules (such as frequency descriptions), spelling errors, non-drug entities, or irrelevant phrases are eliminated. Finally, the text features after elimination are standardized and encoded to generate semantic tags for each drug name. This can be achieved by querying a unified standard encoding system (such as pharmacopoeia encoding) for the retained text features, mapping each drug name to a unique, structured semantic tag.

[0079] In some embodiments, the semantic vector of each semantic tag can be obtained by embedding medical context into all semantic tags based on a preset neurological disease atlas in the following manner:

[0080] For each semantic tag, the semantic tag is associated with the drug entity in the preset neurological disease atlas, thereby generating the atlas node tag of the semantic tag;

[0081] Generate initial embedding vectors for semantic tags based on the graph node tags;

[0082] By adjusting the vector weights of the initial embedding vector using the contextual information of the target handwritten prescription, the semantic vectors of the semantic tags are output, thus obtaining the semantic vector of each semantic tag.

[0083] It should be noted that the neurological disease atlas in this application is a structured knowledge carrier used to express the relationships between drug entities related to neurological diseases; the atlas node labels in this application refer to the drug entity nodes matched with semantic labels in the neurological disease atlas, used to characterize the semantic connection relationship between semantic labels and medical entities in the atlas; the semantic vectors of semantic labels in this application are low-dimensional vector representations generated by fusing atlas structural information and prescription context features.

[0084] In specific implementation, for each semantic tag, firstly, the semantic tag is associated with drug entities in a preset neurological disease atlas. The generation of the atlas node label for the semantic tag can be achieved by using Euclidean distance to perform semantic similarity matching between the semantic tag and drug entities in the preset neurological disease atlas, and using the semantic similarity matching result as the atlas node label for the semantic tag. Then, the generation of the initial embedding vector for the semantic tag based on the atlas node label can be achieved by using a conventional graph neural network model to learn the embedding of the constructed atlas node label, and generating an initial embedding vector for each semantic tag containing its semantic relationships and drug entity structural information. Specifically, the preset neurological disease atlas is represented as a graph structure containing drug entities (nodes) and their semantic relationships (edges). Each semantic tag corresponds to a node label in the graph. A graph convolutional network model is then used to train the graph structure. The graph convolutional network model represents the contextual structure between nodes by aggregating the features of adjacent nodes. The translation embedding model models the triples in the manner of "entity + relationship ≈ target entity". The model continuously adjusts the embedding parameters during training by minimizing node similarity loss or triplet distance loss. This ensures that the generated initial embedding vector retains the basic semantics of the semantic label while embedding its structural position in the neurological disease atlas and upstream and downstream drug association information, providing a structural semantic foundation for subsequent semantic compensation and contextual fusion. Finally, the vector weights of the initial embedding vector are adjusted using the contextual information of the target handwritten prescription, and the semantic vector of the semantic label is output. The semantic vector of each semantic label can be obtained in the following way: First, the full text of the target handwritten prescription is semantically encoded, usually by using a bidirectional encoder to extract the contextual representation of each word or phrase. Then, the contextual representation is fused with the initial embedding vector output by the graph neural network. Common methods include attention mechanisms, which calculate the similarity score between the semantic label embedding and the context vector, assign dynamic weights to different dimensions in the initial embedding vector, and then use the initial embedding vector obtained by dynamically adjusting the weights as the semantic vector of the semantic label. In this way, the semantic vector of each semantic label can be obtained.

[0085] In step 103, based on fuzzy reasoning, the semantic tags of different drug names are clinically matched by combining all semantic vectors to obtain the tag membership degree of different drug names. Then, the semantic tag loss generated during the medical term scene segmentation in the semantic tag extraction process is determined by all tag membership degrees and the co-occurrence probability among various drugs in the neurological disease prescription database.

[0086] In some embodiments, reference Figure 3As shown in the figure, this is a flowchart illustrating the process of determining label membership in some embodiments of this application. In this embodiment, based on fuzzy reasoning and combining all semantic vectors, the semantic labels of different drug names are clinically matched to obtain the label membership of different drug names. This can be achieved through the following steps:

[0087] In step 1031, the cosine similarity between each semantic vector is determined, and then the drug compatibility relationship matrix is ​​constructed.

[0088] In step 1032, potential synergistic combinations in the relationship matrix are screened based on the clinical compatibility rule base;

[0089] In step 1033, the rationality of the synergistic combination is determined by a fuzzy logic algorithm, thereby obtaining the label membership degree of different drug names.

[0090] It should be noted that clinical compatibility in this application refers to the reasonable association and matching of different drugs according to clinical compatibility rules; the label membership degree in this application is an indicator that measures the degree of suitability of drug combinations in a specific clinical context.

[0091] In specific implementation, firstly, the cosine similarity between each semantic vector is determined, and then the drug compatibility relationship matrix is ​​constructed. This can be achieved as follows: the cosine similarity formula can be used to calculate the semantic vector of each pair of drug names to obtain the relationship value of each pair of drug names. Then, the symmetric matrix composed of all relationship values ​​is used as the drug compatibility relationship matrix. Next, the potential synergistic combinations in the relationship matrix are screened based on a clinical compatibility rule base. This can be achieved as follows: first, a structured clinical compatibility rule base is constructed, which contains information on drug compatibility relationships, co-occurrence of indications, pharmacological synergy, and complementary toxic side effects. During implementation, the system reads each drug pair in the compatibility relationship matrix one by one, and matches them with records in the rule base by name or code (such as anatomy-treatment code) to determine whether there are rule items with synergistic relationships, such as "Drug A + Drug B enhances the efficacy of treating disease X" or "Drug C + Drug D reduces the risk of adverse reactions." The matching process can be implemented using Boolean rule judgment and regular expression filtering. Finally, drug combinations that meet the synergistic effect conditions are screened out, and the screened drug combinations are used as... The potential synergistic combinations in the relation matrix; finally, the rationality of the synergistic combinations is determined by fuzzy logic algorithms, and the label membership degree of different drug names can be obtained in the following way: fuzzy membership functions can be set for the synergistic indicators defined in the clinical compatibility rule base, such as the degree of efficacy enhancement, the degree of toxicity and side effects complementarity, and the overlap rate of action targets. Triangular or Gaussian functions can usually be used to represent the fuzzy levels of "low," "medium," and "high." Then, the drug combinations selected from the compatibility relation matrix are used as input to extract their relevant indicator values ​​(as shown in the inference in the atlas). The co-occurrence intensity is fuzzified and input into the fuzzy inference system. Fuzzy inference is then performed based on a preset fuzzy rule base (e.g., "if the efficacy is enhanced, it is high; if the side effects are complementary, it is medium; then the rationality is high"). Common methods include the Mamdani model. The fuzzy inference results are then defuzzified to obtain the compatibility rationality value of each drug combination. All compatibility rationality values ​​are then normalized, and the normalized values ​​are used as the label membership degree of the corresponding drug combination. The label membership degree can reflect the compatibility adaptability of drugs in the current disease context.

[0092] It should be noted that this application constructs a drug compatibility matrix by introducing semantic vector similarity, enabling the potential association between drugs to be measured at the semantic level, thus overcoming the limitations of literal matching. On this basis, it combines a clinical compatibility rule base to screen drug combinations that may have synergistic effects in the relationship matrix, introduces medical knowledge to enhance the professionalism of drug association judgment, and then uses a fuzzy logic algorithm to calculate the rationality of each drug combination. This has a stronger adaptability in handling fuzziness and uncertainty in compatibility. The final tag membership degree can reflect the compatibility of drugs in a specific clinical context, ensuring that the generated prescription is more consistent and reasonable in semantic expression and clinical value, thereby improving the intelligence and accuracy of digital prescription construction.

[0093] In some embodiments, determining the semantic label loss generated during medical terminology scene segmentation in the semantic label extraction process by using all label membership degrees and the co-occurrence probability among various drugs in the neurology disease prescription database specifically includes:

[0094] The co-occurrence probabilities of various drugs in the prescription database for neurological diseases were statistically analyzed, and a co-occurrence probability matrix of drug combinations was constructed.

[0095] Abnormal matching pairs with semantic bias are determined by all tag membership degrees and the co-occurrence probability matrix;

[0096] The semantic label loss during medical terminology scene segmentation in the semantic label extraction process is obtained by backtracking back to the degree of deviation of the abnormal matching pairs.

[0097] It should be noted that the co-occurrence probability in this application refers to the probability that two drugs appear simultaneously in the same prescription in the neurology disease prescription database; the semantic label loss in this application refers to the degree to which the semantic label is inconsistent with its expression in the actual medical context due to semantic segmentation errors that occur during the medical terminology scene segmentation process.

[0098] In specific implementation, firstly, the co-occurrence probabilities of various drugs in the neurology disease prescription database are statistically analyzed. The co-occurrence probability matrix of drug combinations can be constructed as follows: by traversing the existing historical prescriptions in the neurology disease prescription database, the probability of any two drugs co-occurring in the same prescription is calculated using frequency statistics methods, further forming the co-occurrence probability matrix of drug combinations. This matrix reflects the rationality of clinical combination use of drugs. Then, abnormal pairings with semantic bias are identified based on all tag membership degrees and the co-occurrence probability matrix. This can be achieved as follows: by comparing the elements in the co-occurrence probability matrix with the tag membership degrees of the drug combinations, abnormal pairings with high co-occurrence probabilities but low membership degrees, or high membership degrees but low co-occurrence probabilities, are identified—that is, drug combinations with significant semantic bias. Finally, through the... The semantic label loss generated during medical terminology scene segmentation in the semantic label extraction process can be achieved by backtracking the deviation of abnormal matching pairs. This can be done in the following way: the co-occurrence probability of each abnormal matching pair and the numerical deviation between its corresponding label membership degree can be calculated. Then, combined with the context window information in the semantic vector generation process, the actual context of the corresponding drug name in the handwritten recognition text can be located. The semantic relationship and semantic boundary division at that time can be restored by the existing bidirectional encoder. By comparing the context division with the drug description in the standard medical terminology dictionary, the semantic offset caused by terminology segmentation errors, semantic ambiguity, or improper phrase classification can be identified. The semantic offset is mapped to the semantic label in the form of an error score, and the output error score is used as the semantic label loss generated during medical terminology scene segmentation in the corresponding semantic label extraction process.

[0099] It should be noted that this application's solution constructs a drug co-occurrence probability matrix from a neurological disease prescription database, introducing statistical patterns of drug co-occurrence in real prescription data. This addresses the lack of medical context support in traditional semantic segmentation models. Furthermore, it compares drug compatibility relationships using label membership degrees, identifying deviations between co-occurrence probabilities and semantic label expressions. This leads to the discovery of conflicts between semantic understanding and medical knowledge, enabling accurate identification of abnormal semantic pairs. Based on the degree of deviation in abnormal pairs, it traces back to the source of deviation in the semantic segmentation process, quantifying semantic label loss and establishing a closed-loop mechanism from medical knowledge feedback to semantic segmentation quality optimization. This improves the accuracy and clinical rationality of term extraction, achieving a traceable, explainable, and correctable semantic extraction process.

[0100] In step 104, the medical semantic compensation features of all semantic labels are optimized in multiple objectives based on the enhancement cost and the semantic label loss, thereby obtaining the semantic compensation vector of each semantic label, and generating a digital prescription for neurological diseases based on all semantic compensation vectors.

[0101] In some embodiments, the semantic compensation vector for each semantic label can be obtained by performing multi-objective optimization on the medical semantic compensation features of all semantic labels based on the enhancement cost and the semantic label loss, using the following steps:

[0102] We construct a cost function by taking the semantic compensation features of all semantic labels as optimization variables and minimizing the enhancement cost and semantic label loss.

[0103] The enhancement cost and the semantic label loss are introduced into the cost function in the form of a weighted penalty term;

[0104] The cost function is iteratively solved using a preset optimization algorithm until it converges to the optimal solution, thereby obtaining the semantic compensation vector for each semantic tag.

[0105] It should be noted that the semantic compensation features in this application refer to optimizable representation variables introduced to correct structural damage during image enhancement and recognition errors during semantic label extraction.

[0106] In specific implementation, firstly, the semantic compensation features of all semantic labels are used as optimization variables. A cost function is constructed with the goal of minimizing the enhancement cost and semantic label loss. The enhancement cost and semantic label loss are introduced into the cost function as weighted penalty terms. This can be achieved as follows: the semantic compensation features of all semantic labels are used as optimization variables. The goal of the cost function is to minimize the enhancement cost and semantic label loss. The enhancement cost reflects the structural loss caused by contrast enhancement of the handwritten prescription image, such as broken handwriting. The semantic label loss represents errors that occur during semantic extraction, such as incorrect term segmentation or semantic understanding bias. These two are combined in a single cost function, and weighted penalty terms reflect the damage to image structure (such as missing pixels or interrupted handwriting) and the impact on semantic accuracy (such as inaccurate label recognition), respectively. The weighting aims to adjust the relative importance of these two costs according to the actual situation. It should be further noted that the cost function in this application = Q * enhancement cost + W * semantic label loss, where Q and W represent weights. Q and W determine which type of error (image structure damage or semantic recognition bias) is assigned a higher weight in the optimization. The priority is adjusted by introducing the enhancement cost and semantic label loss into the cost function as a weighted penalty term. This can be used to reflect the constraints of image structure damage and semantic recognition errors on the compensation features. Then, the cost function is iteratively solved using a preset optimization algorithm until it converges to the optimal solution, thus obtaining the semantic compensation vector for each semantic label. This can be achieved by using an optimization algorithm (such as particle swarm optimization) to iteratively solve the cost function. For each iteration, the algorithm calculates the value of the cost function based on the current optimization variable (i.e., the semantic compensation feature). Specifically, in each iteration, the algorithm evaluates the impact of the current semantic compensation feature on two objectives: how it reduces the loss of image structure (enhancement cost) and how it reduces the error of semantic labels (semantic label loss). Then, the value of the semantic compensation feature is adjusted according to the optimization criterion (such as minimizing the cost function value) to gradually improve the compensation effect. During the iteration process, a preset convergence condition is used to determine when to stop the optimization process. For example, when the change in the cost function is less than a set threshold, the optimization can be considered to have converged and reached the optimal solution. Finally, after multiple iterations, the optimization algorithm will output the final semantic compensation vector for each semantic label.

[0107] It should be noted that this application's solution introduces two metrics, enhancement cost and semantic label loss, to perform multi-objective optimization of semantic compensation features, achieving the technical effect of improving the accuracy of image semantic expression and the integrity of visual structure. First, enhancement cost measures the structural damage caused during image contrast enhancement (such as broken handwriting or missing pixels), while semantic label loss quantifies the semantic bias generated in medical terminology recognition, achieving comprehensive constraints on both structural and semantic dimensions. Second, by constructing the two costs into a unified cost function and setting a weighted penalty term, the optimization process has flexible control capabilities, dynamically adjusting the trade-off between structural protection and semantic accuracy according to the actual needs of the task. Finally, the optimization algorithm iteratively solves the cost function, accurately outputting the compensation vector for each semantic label, thereby achieving the unification of image structure restoration and semantic information correction, providing a high-quality semantic representation foundation for subsequent prescription understanding.

[0108] It should also be noted that generating digital prescriptions for neurological diseases based on all semantic compensation vectors refers to accurately converting handwritten prescriptions into structured and standardized digital prescriptions based on all semantic compensation vectors.

[0109] On the other hand, in some embodiments, this application provides an intelligent management system for medication during the recovery period of neurological diseases, which includes a digital prescription generation unit. Figure 4 The figure is a schematic diagram of the structure of a digital prescription generation unit according to some embodiments of this application. The digital prescription generation unit 400 includes: a data acquisition module 401, a processing module 402, and an execution module 403, which are described below:

[0110] The acquisition module 401 in this application is mainly used to acquire handwritten prescriptions for neurological diseases through a scanning device, enhance the contrast of the light interference area of ​​the handwritten prescription, obtain the target handwritten prescription with enhanced contrast, and then determine the enhancement cost of the loss of handwritten handwriting structure during the contrast enhancement process.

[0111] Processing module 402, in this application, is used to extract the semantic tags of each drug name in the target handwritten prescription, and to perform medical context embedding processing on all semantic tags based on a preset neurological disease atlas to obtain the semantic vector of each semantic tag.

[0112] In this application, the processing module 402 is also used to perform clinical matching of semantic tags for different drug names based on fuzzy reasoning and combining all semantic vectors to obtain the tag membership degree of different drug names, and then determine the semantic tag loss generated during the medical term scene segmentation in the semantic tag extraction process by using all tag membership degrees and the co-occurrence probability among various drugs in the neurological disease prescription database.

[0113] The execution module 403 in this application is mainly used to perform multi-objective optimization on the medical semantic compensation features of all semantic labels based on the enhancement cost and the semantic label loss, thereby obtaining the semantic compensation vector of each semantic label, and generating a digital prescription for neurological diseases based on all semantic compensation vectors.

[0114] In addition, this application also provides a computer device, the computer device including a memory and a processor, the memory storing code, and the processor being configured to acquire the code and execute the above-described method for generating digital prescriptions for neurological diseases.

[0115] In some embodiments, reference Figure 5 The figure is a schematic diagram of the structure of a computer device for implementing a digital prescription generation method for neurological diseases according to some embodiments of this application. The digital prescription generation method for neurological diseases in the above embodiments can be implemented through... Figure 5 The computer device shown is used to implement this, and the computer device 500 includes at least one processor 501, a communication bus 502, a memory 503, and at least one communication interface 504.

[0116] Processor 501 can be a general-purpose central processing unit (CPU) or an application-specific integrated circuit (ASIC).

[0117] The communication bus 502 can be used to transmit information between the aforementioned components.

[0118] Memory 503 may be a read-only memory (ROM) or other type of static storage device capable of storing static information and instructions, random access memory (RAM) or other type of dynamic storage device capable of storing information and instructions, or electrically erasable programmable read-only memory (EEPROM), compact disc read-only memory (CD-ROM) or other optical disc storage, optical disc storage (including compressed optical discs, laser discs, optical discs, digital versatile optical discs, Blu-ray discs, etc.), magnetic disks or other magnetic storage devices, or any other medium capable of carrying or storing desired program code in the form of instructions or data structures and accessible by a computer, but not limited thereto. Memory 503 may exist independently and be connected to processor 501 via communication bus 502. Memory 503 may also be integrated with processor 501.

[0119] The memory 503 stores program code for executing the scheme of this application, and its execution is controlled by the processor 501. The processor 501 executes the program code stored in the memory 503. The program code may include one or more software modules. In the above embodiment, the digital prescription generation method for neurological diseases can be implemented by the processor 501 and one or more software modules in the program code in the memory 503.

[0120] Communication interface 504 uses any transceiver-like device to communicate with other devices or communication networks, such as Ethernet, radio access network (RAN), wireless local area networks (WLAN), etc.

[0121] In a specific implementation, as one example, a computer device may include multiple processors, each of which may be a single-core (single-CPU) processor or a multi-core (multi-CPU) processor. Here, a processor may refer to one or more devices, circuits, and / or processing cores used to process data (e.g., computer program instructions).

[0122] The aforementioned computer device can be a general-purpose computer device or a special-purpose computer device. In specific implementations, the computer device can be a desktop computer, a portable computer, a network server, a handheld digital assistant (PDA), a mobile phone, a tablet computer, a wireless terminal device, a communication device, or an embedded device. This application does not limit the type of computer device.

[0123] In addition, this application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described method for generating digital prescriptions for neurological diseases.

[0124] Although the preferred embodiments of the present application have been described, those skilled in the art may make additional changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present application.

[0125] Obviously, those skilled in the art can make various modifications and variations to this application without departing from the spirit and scope of this application. Therefore, if such modifications and variations fall within the scope of the claims of this application and their equivalents, this application also intends to include such modifications and variations.

Claims

1. A method for generating digital prescriptions for neurological diseases, used in an intelligent management system for medication during the recovery period of neurological diseases to generate digital prescriptions, characterized in that, The method includes the following steps: Handwritten prescriptions for neurological diseases are collected using scanning equipment. The contrast of the light-interference-affected areas of the handwritten prescriptions is enhanced to obtain the target handwritten prescription with enhanced contrast. Then, the enhancement cost of the loss of handwriting structure during the contrast enhancement process is determined. Semantic tags are extracted from the names of each drug in the target handwritten prescription. Based on a pre-defined neurological disease atlas, all semantic tags are embedded in a medical context to obtain a semantic vector for each semantic tag. Based on fuzzy reasoning, the semantic tags of different drug names are clinically matched by combining all semantic vectors to obtain the tag membership degree of different drug names. Then, the semantic tag loss generated during the medical term scene segmentation in the semantic tag extraction process is determined by all tag membership degrees and the co-occurrence probability among various drugs in the neurological disease prescription database. Based on the enhancement cost and the semantic label loss, multi-objective optimization is performed on the medical semantic compensation features of all semantic labels to obtain the semantic compensation vector of each semantic label. Based on all semantic compensation vectors, a digital prescription for neurological diseases is generated. Specifically, the enhancement cost for determining the loss of handwriting structure during contrast enhancement includes: Extract the edge feature map of the original handwritten prescription and mark the stroke intersections and turning points as topological nodes; By comparing the structural differences of topological nodes in the edge feature maps before and after enhancement, the broken areas of the handwriting can be located. The enhancement cost of handwriting structure loss during contrast enhancement is determined based on the pixel loss rate of the broken handwriting area. Specifically, the process of performing multi-objective optimization on the medical semantic compensation features of all semantic labels based on the enhancement cost and the semantic label loss to obtain the semantic compensation vector for each semantic label includes: We construct a cost function by taking the semantic compensation features of all semantic labels as optimization variables and minimizing the enhancement cost and semantic label loss. The enhancement cost and the semantic label loss are introduced into the cost function in the form of a weighted penalty term; The cost function is iteratively solved using a preset optimization algorithm until it converges to the optimal solution, thereby obtaining the semantic compensation vector for each semantic tag.

2. The method as described in claim 1, characterized in that, Contrast enhancement is applied to the light-interference-affected areas of the handwritten prescription to obtain a target handwritten prescription with enhanced contrast, specifically including: The pixel distribution of the handwritten prescription is analyzed by grayscale histogram analysis to identify areas of light interference. An adaptive gamma correction algorithm is used to enhance the contrast of the light-interference area in layers. The enhanced handwritten prescription is binarized to obtain the target handwritten prescription with enhanced contrast.

3. The method as described in claim 1, characterized in that, The semantic tags extracted from each drug name in the target handwritten prescription specifically include: A pre-trained handwritten character recognition model is used to segment characters and words in the target handwritten prescription and extract text features of the drug name. Semantic validation is performed on the text features of drug names to remove spelling errors and non-drug-related text features; The removed text features are standardized and encoded to generate semantic tags for each drug name.

4. The method as described in claim 1, characterized in that, Based on a pre-defined neurological disease atlas, all semantic tags are embedded using medical context, resulting in a semantic vector for each tag that specifically includes: For each semantic tag, the semantic tag is associated with the drug entity in the preset neurological disease atlas, thereby generating the atlas node tag of the semantic tag; Generate initial embedding vectors for semantic tags based on the graph node tags; By adjusting the vector weights of the initial embedding vector using the contextual information of the target handwritten prescription, the semantic vectors of the semantic tags are output, thus obtaining the semantic vector of each semantic tag.

5. The method as described in claim 1, characterized in that, Based on fuzzy reasoning, the semantic tags of different drug names are clinically matched by combining all semantic vectors to obtain the tag membership degree of different drug names, which specifically includes: Determine the cosine similarity between each semantic vector, and then construct the drug compatibility relationship matrix; Potential synergistic combinations in the relationship matrix were screened based on a clinical compatibility rule base. The rationality of synergistic combinations is determined by fuzzy logic algorithms, thereby obtaining the label membership degree of different drug names.

6. The method as described in claim 1, characterized in that, The semantic label loss generated during medical terminology scene segmentation in the semantic label extraction process is determined by using all label membership degrees and the co-occurrence probabilities among various drugs in the neurology disease prescription database. Specifically, this loss includes: The co-occurrence probabilities of various drugs in the prescription database for neurological diseases were statistically analyzed, and a co-occurrence probability matrix of drug combinations was constructed. Abnormal matching pairs with semantic bias are determined by all tag membership degrees and the co-occurrence probability matrix; The semantic label loss during medical terminology scene segmentation in the semantic label extraction process is obtained by backtracking back to the degree of deviation of the abnormal matching pairs.

7. The method as described in claim 1, characterized in that, The scanning device is a flatbed scanner.

8. A smart management system for medication during the recovery period of neurological diseases, the system comprising a digital prescription generation unit, which generates digital prescriptions using the method described in any one of claims 1 to 7, characterized in that, The digital prescription generation unit includes: The acquisition module is used to acquire handwritten prescriptions for neurological diseases through a scanning device, enhance the contrast of the light interference area of ​​the handwritten prescription, obtain the target handwritten prescription with enhanced contrast, and then determine the enhancement cost of the loss of handwriting structure during the contrast enhancement process. The processing module is used to extract the semantic tags of each drug name in the target handwritten prescription, and to perform medical context embedding processing on all semantic tags based on the preset neurological disease atlas to obtain the semantic vector of each semantic tag. The processing module is also used to perform clinical matching of semantic tags for different drug names based on fuzzy reasoning and combining all semantic vectors to obtain the tag membership degree of different drug names, and then determine the semantic tag loss generated during the medical term scene segmentation in the semantic tag extraction process by using all tag membership degrees and the co-occurrence probability among various drugs in the neurological disease prescription database. The execution module is used to perform multi-objective optimization on the medical semantic compensation features of all semantic labels based on the enhancement cost and the semantic label loss, thereby obtaining the semantic compensation vector of each semantic label, and generating a digital prescription for neurological diseases based on all semantic compensation vectors.

Images