Information query method and device, computer equipment and storage medium

Abstract

The invention relates to an information query method and device, computer equipment and a storage medium. The method comprises the steps that a first entity generates a query request packet according to case information of a target patient; based on the query request packet, the first entity determines a plurality of candidate entities from a pre-stored searchable entity set through a constructed target entity prediction model and sends the query request packet; the candidate entity responds to the received query request packet, generates a response abstract packet according to the local case library and feeds back the response abstract packet to the first entity; the first entity responds to the received multiple response abstract packages, similarity verification is executed on the multiple response abstract packages in sequence, one or more reference cases are screened out, and the corresponding candidate entities serve as second entities; and the first entity and the second entity establish a data encryption transmission channel so as to obtain the full-amount case information of the referenceable case stored by the second entity. By adopting the method, the accuracy of case matching and the data transmission efficiency can be improved, and meanwhile, the patient privacy and the data security are guaranteed.

Description

Technical Field

[0001] The present application relates to the field of information query technology, and in particular to an information query method, apparatus, computer equipment, and storage medium. Background Art

[0002] With the advancement of medical technology and the promotion of regional medical collaboration, patients are increasingly seeking treatment and being referred between different medical institutions. In actual clinical work, hospitals are often limited by their own experience and insufficient case resources, and need to rely on historical case records, medical imaging, and other information from other hospitals.

[0003] However, existing case information query technologies still face numerous technical and privacy challenges in practical applications. For example, first, matching efficiency is low. Currently, most platforms rely on a "broadcast search" approach, requesting relevant case data from other hospitals nationwide. This is time-consuming, has a low response rate, and lacks precise location capabilities. Second, data redundancy is severe, especially for medical images such as CT and MRI, which are bulky. Direct transmission results in high bandwidth usage and significant system latency. Third, the risk of privacy leakage is high. Some platforms primarily transmit raw data in plaintext, lacking necessary encryption and authentication mechanisms, making patient privacy vulnerable. Therefore, a new information query method is urgently needed to address these issues. Summary of the Invention

[0004] Based on this, it is necessary to provide an information query method, device, computer equipment and storage medium to address the above technical problems, which can improve the accuracy of case matching and data transmission efficiency while protecting patient privacy and data security.

[0005] In one aspect, an information query method is provided, the method comprising: The first entity extracts disease feature data based on the case information of the target patient and generates a query request packet by combining it with the image feature data; The first entity determines, based on the query request packet and using a constructed target entity prediction model, a plurality of candidate entities from a pre-stored searchable entity set and sends the query request packet; In response to receiving the query request packet, the candidate entity generates a response summary packet according to the local case database and feeds it back to the first entity; In response to receiving the plurality of response summary packages, the first entity sequentially performs similarity verification on the plurality of response summary packages, selects one or more reference cases, and uses the corresponding candidate entities as the second entity; The first entity generates a key based on the interaction parameters between the query request packet and the response summary packet fed back by the second entity, and establishes a data encryption transmission channel with the second entity to obtain the full case information of the reference case stored by the second entity.

[0006] In one embodiment, the first entity extracts disease feature data based on the case information of the target patient and combines it with the image feature data to generate a query request packet, including: Parsing the case information of the target patient to obtain the disease characteristic data, wherein the disease characteristic data includes at least one of the following: International Classification of Diseases code, chief symptom, initial diagnosis record, and physiological parameters; generating a symptom query summary feature according to the symptom feature data; Acquiring imaging examination data of the target patient, and performing image pyramid feature extraction at a first scale on the imaging examination data to obtain image feature data at the first scale; An image query summary feature is generated according to the image feature data at the first scale, and a case summary feature is generated by combining the disease query summary feature and constructing the query request package.

[0007] In one embodiment, the first entity determines a plurality of candidate entities from a pre-stored searchable entity set based on the query request packet using a constructed target entity prediction model and sends the query request packet, including: Extracting the case summary features in the query request packet, and parsing them to obtain a structured disease summary vector and an image summary feature vector; fusing the structured disease summary vector with the image summary feature vector to construct a case joint input vector; Obtain historical case sharing distribution information of multiple searchable entities in the searchable entity set, and construct corresponding entity attribute vectors; For each of the plurality of searchable entities, feature fusion is performed on the case joint input vector and the corresponding entity attribute vector to generate case entity matching items and input them into the target entity prediction model, and correlation probability scores are calculated between each of the plurality of searchable entities and the case information of the target patient; The plurality of searchable entities are sorted based on the relevance probability scores, and a number of the searchable entities with the highest scores are selected as the candidate entities.

[0008] In one embodiment, for each of the plurality of searchable entities, the case joint input vector and the corresponding entity attribute vector are subjected to feature fusion, case entity matching items are generated and input into the target entity prediction model, and correlation probability scores between the plurality of searchable entities and the case information of the target patient are calculated, including: The case summary embedding network, graph neural network propagation layer and output scoring layer are sequentially set to form the target entity prediction model; Based on the case joint input vector, generating a case summary embedding vector through the case summary embedding network, wherein the case summary embedding network at least includes a multi-layer perceptron network for encoding structured disease features and an image feature encoding network for encoding image summary features; Fusing the case summary embedding vector and the corresponding entity attribute vector to obtain the case entity matching item; Taking the plurality of searchable entities as graph nodes, setting the corresponding case entity matching items as initialization features of the graph nodes, and setting the historical referral relationships and collaboration relationships between the searchable entities as edges, to construct an entity association graph; Based on the entity association graph, performing neighbor node feature aggregation on the graph nodes through the graph neural network propagation layer to generate a graph embedding vector of the case entity relationship; Based on the graph embedding vector of the case entity relationship, the output scoring layer calculates and outputs the correlation probability scores between the multiple searchable entities and the cases.

[0009] In one embodiment, in response to receiving the query request packet, the candidate entity generates a response summary packet based on the local case database and feeds it back to the first entity, including: The candidate entity parses the query request packet to obtain the symptom query summary feature and the image query summary feature; Retrieving historical case records in the local case database whose field matching rate with the symptom query summary feature is greater than or equal to a preset structured matching threshold as a candidate reference case set; constructing a query image vector based on the image query summary features; Obtaining medical imaging data corresponding to a plurality of cases in the candidate reference case set, performing image pyramid feature extraction at the first scale, and generating a first response image vector; performing a first-scale image similarity calculation on the query image vector and the first response image vector, and in response to the image feature similarity between the first response image vector and the query image vector being greater than or equal to a preset image matching threshold, taking the corresponding case as a response case; Performing image pyramid feature extraction at a second scale on the physical image data of the response case to obtain a second response image vector and generate a response image summary feature; Generate a structure field summary according to the structured data of the response case in the local case database; A unique identification code is generated according to the case identification of the response case and the entity identification of the candidate entity, and the response summary package is generated by combining the response image summary feature, the structure field summary and the image feature similarity.

[0010] In one embodiment, in response to receiving multiple response summary packages, the first entity sequentially performs similarity verification on the multiple response summary packages, screens out one or more reference cases, and uses the corresponding candidate entities as the second entity, including: The first entity parses the received multiple response summary packets to obtain multiple identification codes and corresponding response image summary features, the structure field summary, and the image feature similarity; Sort the plurality of identification codes based on the similarity of the image features to generate a verification sequence; performing image pyramid feature extraction at the second scale on the imaging examination data of the target patient to generate an image verification vector; Based on the verification sequence, performing multi-layer matching analysis on the plurality of response image summary features and the image verification vector in sequence to calculate a second-scale image verification similarity; Performing field alignment on the structure field summary and the disease characteristic data, and calculating the structure field matching degree according to a preset field weight rule; Based on a preset joint scoring function, the second scale image verification similarity and the structure field matching degree are combined to calculate the comprehensive similarity corresponding to the identification code; In response to the comprehensive similarity being greater than or equal to a preset available threshold, the corresponding response case is determined to be the reference case, the corresponding response image summary features and the structure field summary are recorded as reference case summary information, and the corresponding candidate entity is determined to be the second entity.

[0011] In one embodiment, the first entity generates a key based on interaction parameters between the query request packet and the response summary packet fed back by the second entity, and establishes a data encryption transmission channel with the second entity to obtain the full case information of the referenced case stored by the second entity, including: The first entity extracts the sending timestamp and the negotiation random factor of the query request packet, and constructs an interaction parameter set in combination with the identification code and the receiving timestamp of the response summary packet; Inputting the interaction parameter set into a preset elliptic curve key agreement algorithm to generate a one-time communication key; A temporary encrypted communication channel is established based on the one-time communication key, and mutual identity authentication and handshake confirmation operations are completed during the channel initialization process.

[0012] In another aspect, an information query device is provided, comprising: A query generation module, configured for the first entity to extract symptom feature data based on the case information of the target patient and combine it with the image feature data to generate a query request packet; A candidate entity determination module, configured for the first entity to determine a plurality of candidate entities from a pre-stored searchable entity set based on the query request packet and using a constructed target entity prediction model, and to send the query request packet; a response receiving module, configured for the candidate entity to generate a response summary packet based on a local case database and feed the response summary packet back to the first entity in response to receiving the query request packet; a similarity verification module configured for the first entity to, in response to receiving a plurality of the response summary packets, sequentially perform similarity verification on the plurality of the response summary packets, screen out one or more reference cases, and use the corresponding candidate entities as the second entity; A key negotiation communication module is used for the first entity to generate a key based on the interaction parameters between the query request packet and the response summary packet fed back by the second entity, and to establish a data encryption transmission channel with the second entity to obtain the full case information of the reference case stored by the second entity.

[0013] In another aspect, a computer device is provided, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the following steps are implemented: The first entity extracts disease feature data based on the case information of the target patient and generates a query request packet by combining it with the image feature data; The first entity determines, based on the query request packet and using a constructed target entity prediction model, a plurality of candidate entities from a pre-stored searchable entity set and sends the query request packet; In response to receiving the query request packet, the candidate entity generates a response summary packet according to the local case database and feeds it back to the first entity; In response to receiving the plurality of response summary packets, the first entity sequentially performs similarity verification on the plurality of response summary packets, selects one or more reference cases, and uses the corresponding candidate entities as the second entity; The first entity generates a key based on the interaction parameters between the query request packet and the response summary packet fed back by the second entity, and establishes a data encryption transmission channel with the second entity to obtain the full case information of the reference case stored by the second entity.

[0014] In another aspect, a computer-readable storage medium is provided, on which a computer program is stored, and when the computer program is executed by a processor, the following steps are implemented: The first entity extracts disease feature data based on the case information of the target patient and generates a query request packet by combining it with the image feature data; The first entity determines, based on the query request packet and using a constructed target entity prediction model, a plurality of candidate entities from a pre-stored searchable entity set and sends the query request packet; In response to receiving the query request packet, the candidate entity generates a response summary packet according to the local case database and feeds it back to the first entity; In response to receiving the plurality of response summary packets, the first entity sequentially performs similarity verification on the plurality of response summary packets, selects one or more reference cases, and uses the corresponding candidate entities as the second entity; The first entity generates a key based on the interaction parameters between the query request packet and the response summary packet fed back by the second entity, and establishes a data encryption transmission channel with the second entity to obtain the full case information of the reference case stored by the second entity.

[0015] The above-mentioned information query method, apparatus, computer device and storage medium, through which the first entity constructs a structured query request package based on the symptom characteristic data and image characteristic data of the target patient, replaces the direct transmission of the original case, not only effectively avoids the plaintext dissemination of sensitive medical data across institutions, but also realizes the compressed expression of the query information, thereby reducing the complexity of the matching calculation and system overhead while protecting the patient's privacy, and significantly improving the matching efficiency; at the same time, the first entity selects a number of candidate entities for precise request through the constructed target entity prediction model, avoiding the resource waste and low response rate problems caused by the "broadcast search" method in the prior art, and fundamentally solving the problem of low matching efficiency; further, by adopting the method of candidate entities returning response summary packages, an encrypted channel is established to obtain full case information only after confirming similarity, avoiding the redundant burden caused by the frequent transmission of large-volume images or full medical records in the early stage, and effectively solving the problem of serious data redundancy; in addition, by generating a one-time communication key based on the interaction parameters to establish a temporary encrypted channel, point-to-point encryption is achieved, ensuring identity authentication and data encryption during the patient information transmission process, significantly improving the information security level, and avoiding the privacy leakage risks that are prone to occur in the prior art. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the embodiments of the present application, the following is a brief introduction to the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0017] Figure 1 is an application environment diagram of an information query method in one embodiment; Figure 2 1 is a flow chart of an information query method according to an embodiment; Figure 3 Schematic diagram of the process of information query step in one embodiment; Figure 4 is a structural block diagram of an information query device in one embodiment; Figure 5 FIG. 1 is a diagram showing the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION

[0018] In order to make the purpose, technical solutions and advantages of this application more clear, the following further describes this application in detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain this application and are not intended to limit this application.

[0019] The information query method provided by this application can be applied to Figure 1In the application environment shown. Among them, the terminal 102 communicates with the server 104 through the network. The user of the first entity enters the case information of the target patient through the front-end system deployed on the terminal 102, and the terminal 102 sends the case information to the server 104 for processing. The server 104 constructs a query request package locally, and then performs intelligent screening and scoring on the preset searchable entity set based on the constructed target entity prediction model, selects several candidate entities, and pushes the query request package to the system to which these candidate entities are connected; after the candidate entity receives the query request, it executes the response summary generation logic through the local case library, and feeds the response summary package back to the server 104 for unified processing and similarity verification, and selects the second entity; thereafter, the server 104 generates a one-time communication key based on the interaction parameters in the query request package and the response summary package, and establishes an encrypted transmission channel with the second entity to obtain complete reference case information for clinical reference. Among them, the entity can be a single device or a cluster of devices, such as a hospital server; the terminal 102 can be but is not limited to various personal computers, laptops, smart phones, tablets and portable wearable devices, which are used by the entity user to initiate case query requests and receive feedback results; the server 104 can be implemented as an independent server or a server cluster composed of multiple servers, which is used to support the operation of the target entity prediction model, the screening of candidate entities, the collection of response summaries and similarity calculation and other functions.

[0020] In one embodiment, Figure 2 As shown, an information query method is provided, which is applied to Figure 1 Taking the server 104 in the example as an example, the following steps are included: Step 201: The first entity extracts disease feature data based on the case information of the target patient and generates a query request packet by combining it with the image feature data. Step 202: The first entity determines multiple candidate entities from a pre-stored searchable entity set based on the query request packet using the constructed target entity prediction model and sends the query request packet. Step 203: In response to receiving the query request packet, the candidate entity generates a response summary packet based on the local case database and feeds it back to the first entity; Step 204: In response to receiving multiple response summary packages, the first entity performs similarity verification on the multiple response summary packages in sequence, selects one or more reference cases, and uses the corresponding candidate entities as the second entity; In step 205, the first entity generates a key based on the interaction parameters between the query request packet and the response summary packet fed back by the second entity, and establishes a data encryption transmission channel with the second entity to obtain the full case information of the reference case stored by the second entity.

[0021] In the above-mentioned information query method, the first entity constructs a structured query request package based on the symptom characteristic data and image characteristic data of the target patient, replacing the direct transmission of the original case. This not only effectively avoids the plaintext dissemination of sensitive medical data across institutions, but also realizes the compressed expression of the query information, thereby reducing the complexity of the matching calculation and system overhead while protecting the patient's privacy, and significantly improving the matching efficiency. At the same time, the first entity selects several candidate entities for precise requests through the constructed target entity prediction model, avoiding the resource waste and low response rate problems caused by the "broadcast search" method in the existing technology, and fundamentally solving the problem of low matching efficiency. Furthermore, by adopting the method of candidate entities returning response summary packages, an encrypted channel is established to obtain full case information only after confirming similarity, avoiding the redundant burden caused by the frequent transmission of large-volume images or full medical records in the early stage, and effectively solving the problem of serious data redundancy. In addition, by generating a one-time communication key based on the interaction parameters to establish a temporary encryption channel, point-to-point encryption is achieved, ensuring identity authentication and data encryption during the patient information transmission process, significantly improving the information security level, and avoiding the privacy leakage risks that are prone to occur in the existing technology.

[0022] In one embodiment, the first entity extracts disease feature data based on the case information of the target patient and combines it with the image feature data to generate a query request packet, including: Parse the target patient's case information to obtain symptom characteristic data, which includes at least one of the following: International Classification of Diseases code, chief symptom, initial diagnosis record, and physiological parameters; Generate symptom query summary features based on symptom feature data; Acquire imaging examination data of a target patient, and perform image pyramid feature extraction at a first scale on the imaging examination data to obtain image feature data at the first scale; An image query summary feature is generated based on the image feature data at the first scale, and a case summary feature is generated by combining the disease query summary feature and constructing a query request package.

[0023] Specifically, in this embodiment, by performing structured analysis on the case information of the target patient, disease characteristics including International Classification of Diseases codes, chief symptoms, initial diagnosis records, physiological parameters, etc. are extracted, and combined with the patient's imaging examination data, multi-scale image features are extracted through an image pyramid method. On this basis, representative disease summary features and image summary features are generated to construct a standardized query request package. This not only significantly simplifies the data dimension and scale required for matching, improves the response speed of the subsequent query matching process, but also avoids the direct transmission of the original case data, effectively reduces the risk of privacy leakage, and achieves an optimal balance between privacy protection and matching efficiency.

[0024] In one embodiment, the first entity determines multiple candidate entities from a pre-stored searchable entity set based on a query request packet using a constructed target entity prediction model and sends the query request packet, including: Extract the case summary features in the query request package and parse them to obtain the structured disease summary vector and image summary feature vector respectively; The structured disease summary vector is fused with the image summary feature vector to construct a joint case input vector; Obtain the historical case sharing distribution information of multiple searchable entities in the searchable entity set and construct the corresponding entity attribute vector; For multiple searchable entities, the case joint input vector and the corresponding entity attribute vector are fused to generate case entity matching items and input them into the target entity prediction model. The correlation probability scores between the multiple searchable entities and the case information of the target patient are calculated respectively. Multiple searchable entities are sorted based on the relevance probability scores, and several searchable entities with the highest scores are selected as candidate entities.

[0025] Specifically, in this embodiment, by constructing a target entity prediction model, a structured matching mechanism is established based on the joint input vector of cases extracted from the query request package and the historical shared attribute information of the searchable entities, thereby achieving accurate screening of highly relevant candidate entities from the preset entity set. Compared with the traditional broadcast matching or manual selection mechanism, the accuracy and automation of target entity positioning are significantly improved, the frequency of participation of irrelevant entities is reduced, the response cycle is shortened, and the technical bottleneck of "low response rate and inaccurate positioning" in the existing technology is overcome, providing intelligent support for efficient case flow in multi-center collaborative scenarios.

[0026] In one embodiment, for multiple searchable entities, feature fusion is performed on the case joint input vector and the corresponding entity attribute vector to generate case entity matching items and input them into the target entity prediction model. The correlation probability scores between the multiple searchable entities and the case information of the target patient are calculated, including: The case summary embedding network, graph neural network propagation layer, and output scoring layer are sequentially set up to form the target entity prediction model; Based on the case joint input vector, a case summary embedding vector is generated by a case summary embedding network, where the case summary embedding network at least includes a multi-layer perceptron network for encoding structured disease features and an image feature encoding network for encoding image summary features; Fuse the case summary embedding vector and the corresponding entity attribute vector to obtain the case entity matching item; Multiple searchable entities are used as graph nodes, the corresponding case entity matching items are set as the initialization features of the graph nodes, and the historical referral relationships and collaboration relationships between the searchable entities are set as edges to construct an entity association graph; Based on the entity association graph, the graph neural network propagation layer performs neighbor node feature aggregation on the graph nodes to generate the graph embedding vector of the case entity relationship; Based on the graph embedding vector of case entity relationships, the output scoring layer calculates and outputs the correlation probability scores between multiple searchable entities and cases.

[0027] Specifically, in this embodiment, by introducing a multimodal summary embedding network and a graph neural network propagation mechanism into the target entity prediction model, utilizing the collaborative embedding capabilities of structured disease information and image summaries, and combining the real referral and collaboration relationships between entities, an entity graph network model is constructed. The features between nodes are propagated and aggregated through the graph neural network, so that a comprehensive judgment can be made based on structural features and graph structure information, and the correlation score between each entity and the target case is output, which enhances the model's ability to understand medical network relationships, improves the accuracy of candidate entity predictions, and ensures better entity recommendation results under limited medical resources.

[0028] In a specific embodiment, case entity matching items are generated and input into a target entity prediction model, and correlation probability scores between multiple searchable entities and case information of target patients are calculated, including: The input of the target entity prediction model is set to the case summary embedding vector, which is denoted as: ; Where X represents the case summary embedding vector, represents a structured disease summary vector, represents the image summary feature vector, MLP(·) represents the multi-layer perceptron network function, which is used to perform nonlinear feature encoding on the structured disease summary vector to achieve dimensionality compression and semantic enhancement, and output the embedding vector of the structured information. CNN(·) represents the image feature encoding network function, which is used to extract features and reduce the dimensionality of the image summary feature vector, retaining only the key image pattern information related to disease representation, and output the image embedding vector. The embedding vector of the structured information and the image embedding vector are concatenated to obtain the case summary embedding vector X.

[0029] The case summary embedding vector is fused with the corresponding entity attribute vector to obtain the case entity matching item, which is denoted as: ;in, represents the case entity matching item of the i-th searchable entity, H(i) represents the entity attribute vector of the i-th searchable entity, that is, the frequency statistics and disease distribution of related cases provided by the entity in history, ReLU(·) represents the activation function of the rectified linear unit, which is used to perform nonlinear transformation on the fused feature vector, setting all values ​​less than 0 in the feature vector to 0, and keeping the part greater than or equal to 0 unchanged to enhance feature sparsity and model expression ability. The output result is a non-negative feature vector with the same dimension as the input as the initial representation of the case entity matching item, which is subsequently used to construct node features in the graph neural network; Based on the entity association graph, the graph neural network propagation layer performs neighbor node feature aggregation on the graph nodes to generate the graph embedding vector of the case entity relationship. The process of neighbor node feature aggregation is based on the following formula: ;in, is the feature vector of the i-th graph node (i.e., the i-th searchable entity) in the k-th layer, initially , is the new feature of the i-th graph node in the k+1-th layer, which is the weighted aggregation result of the neighbor node information; is the feature vector of node j, node j is the neighbor of node i (i.e., it may be a cooperative entity in the region); N(i) is the set of neighbors of node i; W is the weight matrix, which is a learnable linear transformation matrix used to transform the features of neighbor nodes; is a weight coefficient that identifies the degree of influence of node j on node i, which can be weighted by the shared frequency / geographic relationship or learned through the attention mechanism; It is an activation function that performs nonlinear mapping on the feature vectors of neighboring nodes aggregated in the propagation layer of a graph neural network to enhance the model's expressiveness and ability to fit complex patterns. It is a nonlinear transformation function, such as ReLU, LeakyReLU, or ELU. The output result is still a vector of the same dimension as the input, but only undergoes nonlinear transformation processing to provide stable and trainable feature input for the subsequent scoring layer. The feature vector of the i-th image node under the final K-layer aggregation is the graph embedding vector, denoted as ; Based on the graph embedding vector of the case entity relationship, the following formula is used to calculate and output the correlation probability score between multiple searchable entities and cases: ; in, represents the probability score of the correlation between the i-th searchable entity and the case, that is, given a case entity match, the probability that the i-th searchable entity has similar cases; sigmoid(·) represents the activation function, which limits the output result to 0-1; Represents the weight vector corresponding to the graph embedding vector, which is used to embed the graph Perform weighted summation; it is a bias term used to adjust the output of the model and is obtained from the model training together with the weight vector.

[0030] In a specific embodiment, before generating case entity matching items and inputting them into the target entity prediction model, the method further includes: Obtain historical case request records and response entity sets, annotate them with true labels, and construct a training set; Based on a preset loss function, the target entity prediction model is trained on the training set. In response to the target entity prediction model converging to a preset credibility threshold on the preset loss function, it is determined that the training of the target entity prediction model is completed.

[0031] Specifically, in one embodiment, the preset loss function is: ; in, Represents the prediction deviation value, that is, the total difference between the overall prediction result of the model and the actual result in the prediction of the target entity of the current case; Indicates the total number of searchable entities, Indicates whether the i-th entity is a candidate entity, the value is 0 or 1; Based on the loss function, the target entity prediction model is trained on the training set. In response to the target entity prediction model converging on the preset loss function to a change in the prediction deviation value that is less than 0.1, the training of the target entity prediction model is determined to be completed.

[0032] Specifically, in this embodiment, the target entity prediction model is trained based on a training set annotated with real labels, and the model is optimized using a preset loss function. This can effectively determine whether the model has converged and completed the training, ensuring that the model has achieved a balance between prediction accuracy and convergence speed, improving the model's prediction performance, and ensuring its reliability and effectiveness in practical applications, especially in information queries between entities, which can improve the accuracy of case matching.

[0033] In one embodiment, in response to receiving the query request packet, the candidate entity generates a response summary packet based on the local case database and feeds it back to the first entity, including: The candidate entity parses the query request packet to obtain the symptom query summary feature and the image query summary feature; Retrieve historical case records in the local case database whose field matching rate with the symptom query summary feature is greater than or equal to a preset structured matching threshold as a candidate reference case set; Construct a query image vector based on the image query summary features; Obtain medical imaging data corresponding to multiple cases in the candidate reference case set, perform image pyramid feature extraction at a first scale, and generate a first response image vector; Performing a first-scale image similarity calculation on the query image vector and the first response image vector, and in response to the image feature similarity between the first response image vector and the query image vector being greater than or equal to a preset image matching threshold, taking the corresponding case as the response case; Performing image pyramid feature extraction at a second scale on the physical image data of the response case to obtain a second response image vector and generate a response image summary feature; Generate a structure field summary based on the structured data of the response case in the local case database; A unique identification code is generated according to the case ID of the response case and the entity ID of the candidate entity, and a response summary package is generated by combining the response image summary features, structure field summary and image feature similarity.

[0034] Specifically, in this embodiment, the original case data is replaced by the response summary package as the information carrier for the early interaction, the structured matching fields and image pyramid features in the local case library are extracted through the candidate entity, and the preliminary image similarity judgment is performed. Only the response cases whose similarity meets the threshold requirements are fed back as candidate results, which significantly reduces the amount of data transmitted over the network and the interference of invalid data. While improving the matching efficiency, it effectively reduces the bandwidth load and system delay caused by the transmission of the full amount of cases, and at the same time controls the scope of information exposure and improves the security of patient data use.

[0035] In one embodiment, in response to receiving multiple response summary packages, the first entity sequentially performs similarity verification on the multiple response summary packages, screens out one or more reference cases, and uses the corresponding candidate entities as the second entity, including: The first entity parses the received multiple response summary packets to obtain multiple identification codes and corresponding response image summary features, structure field summaries, and image feature similarities; Sort multiple identification codes based on image feature similarity to generate a verification sequence; Performing image pyramid feature extraction at a second scale on the imaging examination data of the target patient to generate an image verification vector; Based on the verification sequence, multi-layer matching analysis is performed on multiple response image summary features and image verification vectors in sequence to calculate the second-scale image verification similarity; Align the structure field summary with the disease characteristic data, and calculate the structure field matching degree according to the preset field weight rules; Based on the preset joint scoring function, the second scale image verification similarity and structure field matching are combined to calculate the comprehensive similarity corresponding to the identification code; In response to the comprehensive similarity being greater than or equal to a preset available threshold, the corresponding response case is determined to be a reference case, the corresponding response image summary features and structure field summary are recorded as reference case summary information, and the corresponding candidate entity is determined as the second entity.

[0036] Specifically, in this embodiment, after the first entity receives multiple response summary packages, a two-way comparison mechanism is implemented between the image summary features and the structure field summary, a multi-scale image similarity analysis is performed through the image verification vector, and the field alignment and weight matching calculation of the structure field are combined to comprehensively output the final similarity score. Compared with the traditional single image comparison or field matching method, this embodiment can take into account both image details and clinical semantic information, ensuring the high credibility of the screened reference cases.

[0037] In one embodiment, Figure 3 As shown, after taking the corresponding candidate entity as the second entity, the following is further included: Step 301: The first entity generates a confirmation information package based on the determined second entity and the information summary of the reference case and sends it to the second entity; Step 302: In response to the second entity receiving the confirmation information packet, a second query request packet is generated based on the determined information summary of the referenced case, and sent to a second searchable entity in the second searchable entity set associated with the second entity to obtain a secondary response summary packet; Step 303: In response to the second entity receiving one or more secondary response summary packages, similarity verification is performed to screen out one or more second reference cases and feed them back to the first entity.

[0038] It is worth noting that after completing the preliminary query and matching from the first entity to the second entity, the second entity system will automatically extract the core structural disease features and image summary features from the selected reference case as the corresponding information summary, and regenerate a second query request package, which contains the reference case features alone or is fused with the original request features of the target patient to form a mixed feature expression; When the second query request packet is generated, the second entity sends the second query request packet to its associated second queryable entity set, where the second queryable entity set includes historical collaboration entities, subordinate entities, associated community entities, special authority entities, etc.; After the second checkable entity receives the second query request packet, it performs the same local summary matching, response summary packet generation and encryption return steps as the first round, and feeds back the secondary response summary packet to the second entity, which then summarizes and securely forwards it to the first entity at one time.

[0039] Specifically, in this embodiment, by having the second hospital generate a second query request package based on its internal matched reference cases, it is possible to avoid the risk of mismatching due to incomplete information, atypical symptoms, or entry differences in the original target case characteristics. The generated query request package has stronger clinical representativeness and structural standardization characteristics, which is conducive to improving the ability of the next-level query hospital to distinguish case similarity; furthermore, through the control mechanism that the first hospital sends a confirmation information package to the second hospital before allowing it to initiate a secondary query request, it realizes the refined management of data exchange permissions between hospitals and avoids unauthorized information. Diffusion enhances the security compliance of the system; at the same time, the second searchable hospital set introduced constitutes a secondary association node in the hospital network, covering historical collaborative hospitals, subordinate institutions, community hospitals and special authority institutions, etc., which is conducive to further expanding the breadth of matchable data sources and improving the response coverage of rare diseases, marginal cases or referral problems; in addition, the second hospital uniformly summarizes and feeds back all secondary response summary packages, which can predict, re-screen and summarize the results of the second query process, thereby ensuring that the first hospital obtains high-value, highly relevant refined feedback, reducing the processing burden of the first hospital. Therefore, this multi-level query mechanism not only enhances the depth capability and intelligent evolution characteristics of the entire information query system, but also improves the comprehensive performance of the existing case information collaboration mechanism from multiple dimensions such as privacy protection, efficiency control and precise matching, with outstanding technical effects and practical value.

[0040] In one embodiment, the first entity generates a key based on the interaction parameters between the query request packet and the response summary packet fed back by the second entity, and establishes a data encryption transmission channel with the second entity to obtain the full case information of the referenced case stored by the second entity, including: The first entity extracts the sending timestamp and negotiation random factor of the query request packet, and constructs an interaction parameter set by combining the identification code and receiving timestamp of the response summary packet; Input the interaction parameter set into the preset elliptic curve key agreement algorithm to generate a one-time communication key; A temporary encrypted communication channel is established based on a one-time communication key, and mutual authentication and handshake confirmation operations are completed during the channel initialization process.

[0041] In a specific embodiment, the interaction parameter set is input into a preset elliptic curve key agreement algorithm to generate a one-time communication key, including: The first entity concatenates parameters such as the query request sending timestamp, the negotiated random factor, the identification code of the response digest packet, and the receiving timestamp to form a key derivation seed. The key derivation seed is input into a preset key derivation function to obtain a first private key. The corresponding first public key is calculated based on the first private key and the elliptic curve base point. The key derivation function is an HMAC-based key derivation function (HKDF), which is highly secure and supports multi-purpose key extensions. It is suitable for high-security negotiation scenarios involving dynamic session keys. The second entity generates a second private key and a second public key based on the same key derivation function as the first entity and also based on the parameters in the interaction parameter set; The first entity requests the second public key of the second entity, and inputs the second public key into the preset elliptic curve key agreement algorithm in combination with the first private key generated by the first entity to generate a one-time communication key; The second entity requests the first public key of the first entity, and inputs the preset elliptic curve key agreement algorithm into the second private key generated by itself to generate a one-time communication key consistent with the first entity; In a further embodiment, a temporary encrypted communication channel is established based on a one-time communication key, and mutual authentication and handshake confirmation operations are completed during the channel initialization process, including: The two parties input the shared communication key into a preset key derivation function, which further derives multiple subkeys used for data encryption, message authentication, and identity authentication, including encryption keys, authentication keys, and handshake keys; Based on the derived encryption key, the first entity and the second entity use a symmetric encryption algorithm (such as AES-GCM or the SM4 algorithm that complies with national encryption standards) to establish a temporary encrypted communication channel for subsequent encrypted transmission of sensitive data; During the channel initialization phase, the first and second entities perform identity authentication through a handshake interaction. Authentication includes: The two parties exchange digital certificates or preset identity credentials; Both parties use the derived handshake key to sign or perform HMAC operation on the communication digest, thereby completing the legitimacy confirmation of the identity of the communicating party; If the identity verification is successful, both parties send a "Finished" message to confirm that the key negotiation and authentication process is complete, and officially enter the encrypted communication phase. The temporary encrypted communication channel is automatically closed after the session ends or the preset timeout period is reached, and the communication keys and related intermediate materials used during the session are destroyed to achieve the confidentiality, integrity and forward security of the communication data.

[0042] Specifically, in this embodiment, by extracting interactive parameters such as timestamps, random factors, and identification codes from the query request and response summary during the data acquisition phase, and combining them with the elliptic curve key negotiation algorithm to generate a one-time communication key, a secure temporary encryption channel is established. On this basis, identity authentication and data transmission are completed, effectively preventing the original case information from being intercepted, tampered with, or forged during network transmission, solving the privacy leakage problem caused by the lack of encryption mechanism in the existing technology, and ensuring the security and compliance of medical data in cross-institutional sharing.

[0043] It should be understood that although Figure 2-Figure 3 The steps in the flowchart are shown in the order indicated by the arrows, but these steps are not necessarily executed in the order indicated by the arrows. Unless otherwise specified in this document, there is no strict order restriction for the execution of these steps, and these steps can be executed in other orders. In addition, Figure 2-Figure 3 At least part of the steps may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily executed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be executed in turn or alternately with other steps or at least part of the sub-steps or stages of other steps.

[0044] In one embodiment, Figure 4 As shown, an information query device is provided, comprising: a query generation module, a candidate entity determination module, a response receiving module, a similarity verification module and a key agreement communication module, wherein: A query generation module, configured for the first entity to extract symptom feature data based on the case information of the target patient and combine it with the image feature data to generate a query request packet; The candidate entity determination module is used for the first entity to determine multiple candidate entities from the pre-stored searchable entity set based on the query request packet and through the constructed target entity prediction model, and send the query request packet; A response receiving module, configured for the candidate entity to generate a response summary package based on the local case database and feed the response summary package back to the first entity in response to receiving the query request package; a similarity verification module configured for the first entity to, in response to receiving multiple response summary packets, sequentially perform similarity verification on the multiple response summary packets, screen out one or more reference cases, and use the corresponding candidate entities as the second entity; The key negotiation communication module is used for the first entity to generate a key based on the interaction parameters between the query request packet and the response summary packet fed back by the second entity, and to establish a data encryption transmission channel with the second entity to obtain the full case information of the reference case stored by the second entity.

[0045] For the specific definition of the information query device, please refer to the definition of the information query method above and will not be repeated here. The various modules in the above-mentioned information query device can be implemented in whole or in part through software, hardware, or a combination thereof. The above-mentioned modules can be embedded in or independent of the processor in the computer device in hardware form, or can be stored in the memory of the computer device in software form, so that the processor can call and execute the corresponding operations of the above-mentioned modules.

[0046] In one embodiment, a computer device is provided. The computer device may be a server, and its internal structure diagram may be as follows: Figure 5 As shown. The computer device includes a processor, memory, network interface, and database connected via a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for the operation of the operating system and computer program in the non-volatile storage medium. The database of the computer device is used to store information query data. The network interface of the computer device is used to communicate with an external terminal via a network connection. When the computer program is executed by the processor, it implements an information query method.

[0047] Those skilled in the art will understand that Figure 5 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than shown in the figure, or combine certain components, or have a different component arrangement.

[0048] In one embodiment, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the following steps are performed: The first entity extracts disease feature data based on the case information of the target patient and generates a query request packet by combining it with the image feature data; The first entity determines multiple candidate entities from a pre-stored searchable entity set based on the query request packet using the constructed target entity prediction model and sends the query request packet; In response to receiving the query request packet, the candidate entity generates a response summary packet according to the local case database and feeds it back to the first entity; In response to receiving multiple response summary packages, the first entity sequentially performs similarity verification on the multiple response summary packages, selects one or more reference cases, and uses the corresponding candidate entities as the second entity; The first entity generates a key based on the interaction parameters between the query request packet and the response summary packet fed back by the second entity, and establishes a data encryption transmission channel with the second entity to obtain the full case information of the reference case stored by the second entity.

[0049] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the following steps are implemented: The first entity extracts disease feature data based on the case information of the target patient and generates a query request packet by combining it with the image feature data; The first entity determines multiple candidate entities from a pre-stored searchable entity set based on the query request packet using the constructed target entity prediction model and sends the query request packet; In response to receiving the query request packet, the candidate entity generates a response summary packet according to the local case database and feeds it back to the first entity; In response to receiving multiple response summary packages, the first entity sequentially performs similarity verification on the multiple response summary packages, selects one or more reference cases, and uses the corresponding candidate entities as the second entity; The first entity generates a key based on the interaction parameters between the query request packet and the response summary packet fed back by the second entity, and establishes a data encryption transmission channel with the second entity to obtain the full case information of the reference case stored by the second entity.

[0050] Those skilled in the art will understand that all or part of the processes in the above-mentioned embodiments can be implemented by instructing the relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application may include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in many forms such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), Synchronous Link DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0051] The technical features of the above embodiments can be combined arbitrarily. In order to make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0052] The above embodiments merely represent several implementation methods of the present application. While the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the present invention. It should be noted that a person skilled in the art could make various modifications and improvements without departing from the spirit of the present application, all of which fall within the scope of protection of the present application. Therefore, the scope of protection of the present patent application shall be determined by the appended claims.

Claims

1. An information query method, characterized in that: include: The first entity extracts disease feature data based on the case information of the target patient and generates a query request packet by combining it with the image feature data; The first entity determines, based on the query request packet and using a constructed target entity prediction model, a plurality of candidate entities from a pre-stored searchable entity set and sends the query request packet; In response to receiving the query request packet, the candidate entity generates a response summary packet according to the local case database and feeds it back to the first entity; In response to receiving the plurality of response summary packets, the first entity sequentially performs similarity verification on the plurality of response summary packets, selects one or more reference cases, and uses the corresponding candidate entities as the second entity; The first entity generates a key based on the interaction parameters between the query request packet and the response summary packet fed back by the second entity, and establishes a data encryption transmission channel with the second entity to obtain the full case information of the reference case stored by the second entity.

2. The information query method according to claim 1, characterized in that: The first entity extracts disease feature data based on the case information of the target patient and combines it with the image feature data to generate a query request packet, including: Parsing the case information of the target patient to obtain the disease characteristic data, wherein the disease characteristic data includes at least one of the following: International Classification of Diseases code, chief symptom, initial diagnosis record, and physiological parameters; generating a symptom query summary feature according to the symptom feature data; Acquiring imaging examination data of the target patient, and performing image pyramid feature extraction at a first scale on the imaging examination data to obtain image feature data at the first scale; An image query summary feature is generated according to the image feature data at the first scale, and a case summary feature is generated by combining the disease query summary feature and constructing the query request package.

3. The information query method according to claim 2, characterized in that: The first entity determines a plurality of candidate entities from a pre-stored searchable entity set based on the query request packet using a constructed target entity prediction model and sends the query request packet, including: Extracting the case summary features in the query request packet, and parsing them to obtain a structured disease summary vector and an image summary feature vector; fusing the structured disease summary vector with the image summary feature vector to construct a case joint input vector; Obtaining historical case sharing distribution information of multiple searchable entities in the searchable entity set, and constructing corresponding entity attribute vectors; For each of the plurality of searchable entities, feature fusion is performed on the case joint input vector and the corresponding entity attribute vector to generate case entity matching items and input them into the target entity prediction model, and correlation probability scores are calculated between each of the plurality of searchable entities and the case information of the target patient; The plurality of searchable entities are sorted based on the relevance probability scores, and several searchable entities with the highest scores are selected as the candidate entities.

4. An information query method according to claim 3, characterized in that: For each of the plurality of searchable entities, the case joint input vector and the corresponding entity attribute vector are subjected to feature fusion to generate case entity matching items and input them into the target entity prediction model, and the correlation probability scores between the plurality of searchable entities and the case information of the target patient are calculated, including: The case summary embedding network, graph neural network propagation layer and output scoring layer are sequentially set to form the target entity prediction model; Based on the case joint input vector, generating a case summary embedding vector through the case summary embedding network, wherein the case summary embedding network at least includes a multi-layer perceptron network for encoding structured disease features and an image feature encoding network for encoding image summary features; Fusing the case summary embedding vector and the corresponding entity attribute vector to obtain the case entity matching item; Taking the plurality of searchable entities as graph nodes, setting the corresponding case entity matching items as initialization features of the graph nodes, and setting the historical referral relationships and collaboration relationships between the searchable entities as edges, to construct an entity association graph; Based on the entity association graph, performing neighbor node feature aggregation on the graph nodes through the graph neural network propagation layer to generate a graph embedding vector of the case entity relationship; Based on the graph embedding vector of the case entity relationship, the output scoring layer calculates and outputs the correlation probability scores between the multiple searchable entities and the cases.

5. The information query method according to claim 2, characterized in that: In response to receiving the query request packet, the candidate entity generates a response summary packet based on the local case database and feeds it back to the first entity, including: The candidate entity parses the query request packet to obtain the symptom query summary feature and the image query summary feature; Retrieving historical case records in the local case database whose field matching rate with the symptom query summary feature is greater than or equal to a preset structured matching threshold as a candidate reference case set; constructing a query image vector based on the image query summary features; Obtaining medical imaging data corresponding to a plurality of cases in the candidate reference case set, performing image pyramid feature extraction at the first scale, and generating a first response image vector; performing a first-scale image similarity calculation on the query image vector and the first response image vector, and in response to the image feature similarity between the first response image vector and the query image vector being greater than or equal to a preset image matching threshold, taking the corresponding case as a response case; Performing image pyramid feature extraction at a second scale on the physical image data of the response case to obtain a second response image vector and generate a response image summary feature; Generate a structure field summary according to the structured data of the response case in the local case database; A unique identification code is generated according to the case identification of the response case and the entity identification of the candidate entity, and the response summary package is generated by combining the response image summary feature, the structure field summary and the image feature similarity.

6. An information query method according to claim 5, characterized in that: In response to receiving the plurality of response summary packages, the first entity sequentially performs similarity verification on the plurality of response summary packages, selects one or more reference cases, and uses the corresponding candidate entities as the second entity, including: The first entity parses the received multiple response summary packets to obtain multiple identification codes and corresponding response image summary features, the structure field summary, and the image feature similarity; Sort the plurality of identification codes based on the similarity of the image features to generate a verification sequence; performing image pyramid feature extraction at the second scale on the imaging examination data of the target patient to generate an image verification vector; Based on the verification sequence, performing multi-layer matching analysis on the plurality of response image summary features and the image verification vector in sequence to calculate a second-scale image verification similarity; Performing field alignment on the structure field summary and the disease characteristic data, and calculating the structure field matching degree according to a preset field weight rule; Based on a preset joint scoring function, the second scale image verification similarity and the structure field matching degree are combined to calculate the comprehensive similarity corresponding to the identification code; In response to the comprehensive similarity being greater than or equal to a preset available threshold, the corresponding response case is determined to be the reference case, the corresponding response image summary features and the structure field summary are recorded as reference case summary information, and the corresponding candidate entity is determined to be the second entity.

7. An information query method according to claim 6, characterized in that: The first entity generates a key based on interaction parameters between the query request packet and the response summary packet fed back by the second entity, and establishes a data encryption transmission channel with the second entity to obtain the full case information of the referenced case stored by the second entity, including: The first entity extracts the sending timestamp and the negotiation random factor of the query request packet, and constructs an interaction parameter set in combination with the identification code and the receiving timestamp of the response summary packet; Inputting the interaction parameter set into a preset elliptic curve key agreement algorithm to generate a one-time communication key; A temporary encrypted communication channel is established based on the one-time communication key, and mutual identity authentication and handshake confirmation operations are completed during the channel initialization process.

8. An information query device, characterized in that: The device comprises: A query generation module, configured for the first entity to extract symptom feature data based on the case information of the target patient and combine it with the image feature data to generate a query request packet; A candidate entity determination module, configured for the first entity to determine a plurality of candidate entities from a pre-stored searchable entity set based on the query request packet and using a constructed target entity prediction model, and to send the query request packet; a response receiving module, configured for the candidate entity to generate a response summary packet based on a local case database and feed the response summary packet back to the first entity in response to receiving the query request packet; a similarity verification module configured for the first entity to, in response to receiving a plurality of the response summary packets, sequentially perform similarity verification on the plurality of the response summary packets, screen out one or more reference cases, and use the corresponding candidate entities as the second entity; A key negotiation communication module is used for the first entity to generate a key based on the interaction parameters between the query request packet and the response summary packet fed back by the second entity, and to establish a data encryption transmission channel with the second entity to obtain the full case information of the reference case stored by the second entity.

9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images