A disease pre-examination triage auxiliary method and system

Abstract

The application discloses a disease pre-examination triage auxiliary method and system, relates to the technical field of disease pre-examination triage, collects patient treatment data to generate a data group, extracts features of the data group to obtain a disease feature vector, constructs a dynamic routing layer to convert the disease feature vector into a plastic weight matrix, constructs a processing cluster layer including a quick response cluster and a deep analysis cluster, inputs the plastic weight matrix into the processing cluster layer to obtain an analysis result, and outputs a treatment guide report by fusing analysis results of multiple clusters through a decision fusion layer.

Description

Technical Field

[0001] This invention relates to the field of disease pre-screening and triage technology, specifically to an auxiliary method and system for disease pre-screening and triage. Background Technology

[0002] Currently, the triage systems of medical institutions mainly rely on static triage models based on rule engines and single machine learning algorithms, which have significant technical limitations.

[0003] Traditional systems typically employ fixed threshold judgment mechanisms, which cannot adapt to the dynamic evolution of patients' conditions and real-time changes in medical resources. At the data processing level, existing technologies lack the ability to integrate multimodal medical data, and heterogeneous data such as vital signs, imaging features, biochemical indicators, and symptom descriptions are often analyzed in isolation, leading to the omission of key clinical clues. In terms of system architecture, centralized processing models face resource contention issues in high-concurrency scenarios. When emergency patients surge, system response time increases exponentially, severely impacting triage efficiency. Existing technologies have also failed to establish effective cross-module collaboration mechanisms, with preliminary screening and in-depth diagnosis processes being isolated from each other, making it impossible to achieve dynamic verification and knowledge iteration of diagnostic evidence.

[0004] Therefore, developing an efficient auxiliary method for disease pre-screening and triage is of great significance. Summary of the Invention

[0005] The purpose of this invention is to provide a disease pre-examination and triage auxiliary method and system to address the shortcomings of the prior art.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a disease pre-screening and triage auxiliary method, comprising:

[0007] Collect patient medical data to generate data clusters, and extract features from the data clusters to obtain disease feature vectors;

[0008] A dynamic routing layer is constructed to transform disease feature vectors into a plastic weight matrix;

[0009] The processing cluster layer is constructed, which includes a fast response cluster and a deep analysis cluster. The analysis results are obtained by inputting the plasticity weight matrix into the processing cluster layer.

[0010] The analysis results from multiple clusters are combined through a decision fusion layer to output a medical consultation guide report.

[0011] In a preferred embodiment, the step of collecting patient medical data to generate a data cluster, and extracting features from the data cluster to obtain a disease feature vector, is as follows:

[0012] The patient's medical data collected includes vital signs, symptom descriptions, imaging data, and biochemical indicators;

[0013] Preprocessing of medical data to extract disease feature vectors includes core pathological features, quantitative symptom features, imaging features, and environmental correlation features;

[0014] By combining knowledge graphs, entities associated with disease feature vectors and their relationships are mapped into a mapping graph retrieval table.

[0015] In a preferred embodiment, the step of constructing the dynamic routing layer to transform the disease feature vector into a plastic weight matrix is ​​as follows:

[0016] Quantifying the feature strength of disease feature vectors using neural networks The real-time load rate of the system is obtained based on the resource monitoring unit. ;

[0017] The plasticity weights for constructing the dynamic routing layer to generate disease feature vectors are:

[0018]

[0019] in Let be the plasticity weight of the i-th disease feature vector. As a clinical prior weight, For load sensitivity coefficient, Age decay constant, For time increments;

[0020] A plasticity weight matrix is ​​constructed based on the plasticity weights of disease feature vectors.

[0021] In a preferred embodiment, the steps of constructing the processing cluster layer, including a fast response cluster and a deep analytics cluster, are as follows:

[0022] Building a rapid response cluster involves multiple rapid processing nodes, and the rapid response cluster is used to handle acute and critical illnesses as well as ordinary cases;

[0023] Building a deep analytics cluster involves multiple deep analytics nodes, and the deep analytics cluster is used to process complex cases.

[0024] The fast processing nodes and deep analysis nodes are connected via a dynamic transmission channel.

[0025] In a preferred embodiment, the step of inputting the plasticity weight matrix into the processed cluster layer to obtain the analysis results is as follows:

[0026] Preset response thresholds, matching thresholds, and a database of important disease features;

[0027] The disease matching degree is obtained by matching disease feature vectors with an important disease feature database using a semantic matching algorithm;

[0028] The rapid response cluster generates preliminary diagnostic reports and diagnostic features based on disease feature vector input values ​​that are equal to or exceed the response threshold or matching threshold.

[0029] Inputting disease feature vectors below the response threshold into the deep analysis cluster yields etiology analysis reports and etiology analysis features;

[0030] Simultaneously, a cluster interaction mechanism is constructed to determine whether there is interaction between the fast processing nodes and the deep analysis nodes.

[0031] In a preferred embodiment, the step of constructing a cluster interaction mechanism and determining whether there is interaction between the fast processing node and the deep analysis node based on the cluster interaction mechanism is as follows:

[0032] The deep analysis node uses the mapping graph retrieval table to obtain the associated disease feature vectors and sends an interaction request to the fast processing node that stores the required disease feature vectors.

[0033] Upon receiving the interaction request, the fast processing node will transmit the stored disease feature vector to the deep analysis node that issued the interaction request via a dynamic transmission channel.

[0034] Deep analysis nodes combine disease feature vectors to update etiology analysis reports and generate complex etiology reports;

[0035] Simultaneously, the correlation degree between the disease feature vectors associated with the deep analysis nodes and the fast processing nodes is recorded and calculated, generating correlation indicators. ;

[0036] When the fast processing node and the deep analysis node receive the same diagnostic features and etiology analysis features respectively, the dynamic transmission channel between the fast processing node and the deep analysis node is strengthened:

[0037]

[0038] in, The weights are the enhanced weights for the dynamic transmission channel. The weights before dynamic transmission channel enhancement. Indicates the learning rate;

[0039] The bandwidth resources of the dynamic transmission channel are adjusted based on the weight. Higher weights have more bandwidth resources, while lower weights have less bandwidth resources.

[0040] In a preferred embodiment, the step of outputting a medical guide report by fusing the analysis results of multiple clusters through a decision fusion layer is as follows:

[0041] The decision fusion layer is constructed by including a fast track layer and a regular track layer;

[0042] The fast channel layer is used to receive preliminary diagnostic reports and diagnostic features from fast processing nodes, and uses a neural network model to comprehensively analyze the preliminary diagnostic reports to generate an emergency medical guidance report;

[0043] The standard channel layer is used to receive preliminary diagnostic reports and complex etiology reports, and to generate a standard medical visit guide report through comprehensive analysis.

[0044] The present invention also provides a disease pre-screening and triage auxiliary system, including

[0045] Data acquisition module: Collects patients' medical data to generate data clusters, and extracts features from the data clusters to obtain disease feature vectors;

[0046] Dynamic routing module: Constructs a dynamic routing layer to transform disease feature vectors into a plasticity weight matrix;

[0047] Processing Cluster Module: The processing cluster layer is constructed, including a fast response cluster and a deep analysis cluster. The analysis results are obtained after inputting the plasticity weight matrix into the processing cluster layer.

[0048] Fusion Decision Module: The decision fusion layer integrates the analysis results of multiple clusters and outputs a medical guide report.

[0049] The technical effects and advantages provided by the present invention in the above technical solution are as follows:

[0050] 1. This invention integrates multiple data sources and performs preprocessing to extract disease feature vectors, including core pathological features, symptom quantification features, imaging features, and environmental association features. Combined with a knowledge graph, the system can map these features to relevant entities and their relationships into a knowledge graph retrieval table, thereby achieving a comprehensive understanding and rapid retrieval of disease features. The system quantifies the feature strength of disease feature vectors through neural networks and dynamically adjusts the plasticity weights of these vectors based on the system's real-time load rate. This mechanism enables the system to intelligently allocate computing resources according to the urgency of the disease and the system's load. For acute, critical, and common cases, the system rapidly generates preliminary diagnostic reports and diagnostic features through a rapid response cluster, ensuring timely treatment for patients. For complex cases, the system assigns them to a deep analysis cluster for in-depth analysis, thus avoiding delays or misdiagnosis caused by uneven resource allocation. This intelligent triage and rapid response capability allows the system to quickly identify patients requiring urgent treatment in complex medical environments and provide them with timely medical guidance, thereby significantly improving the quality and efficiency of medical services.

[0051] 2. This invention constructs a processing cluster layer, including a rapid response cluster and a deep analysis cluster, which can work collaboratively to achieve comprehensive disease analysis. The rapid response cluster is mainly responsible for handling common diseases and acute and critical illnesses, while the deep analysis cluster focuses on handling complex cases. This division of labor can fully utilize system resources and improve processing efficiency. Deep analysis nodes can obtain the required disease feature vectors through a mapping graph retrieval table and send interactive requests to the rapid processing nodes that store these feature vectors; the rapid processing nodes then transmit relevant information to the deep analysis nodes through a dynamic transmission channel. This interaction mechanism enables the two clusters to share information and conduct collaborative analysis, thereby improving the accuracy and comprehensiveness of diagnosis. In addition, the system generates correlation indicators and strengthens the dynamic transmission channel between the rapid processing nodes and deep analysis nodes based on these indicators, allowing the system to adaptively adjust resource allocation and improve information transmission efficiency. The deep analysis cluster combines disease feature vectors to update the etiology analysis report, generating complex etiology reports and providing doctors with more comprehensive diagnostic information. Attached Figure Description

[0052] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0053] Figure 1 This is a flowchart of the method of the present invention.

[0054] Figure 2 This is a system block diagram of the present invention. Detailed Implementation

[0055] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0056] Example 1, please refer to Figure 1 As shown in this embodiment, a disease pre-screening and triage auxiliary method includes:

[0057] S1. Collect patient medical data to generate a data cluster, and extract features from the data cluster to obtain disease feature vectors;

[0058] S2. Construct a dynamic routing layer to convert disease feature vectors into a plastic weight matrix;

[0059] S3. Construct a processing cluster layer including a fast response cluster and a deep analysis cluster. Input the plasticity weight matrix into the processing cluster layer to obtain the analysis results.

[0060] S4. By fusing the analysis results of multiple clusters through the decision fusion layer, a medical visit guide report is output.

[0061] As described in steps S1-S4 above, currently, the pre-examination and triage system of medical institutions mainly relies on static triage models based on rule engines and single machine learning algorithms, which have significant technical limitations.

[0062] Traditional systems typically employ fixed threshold judgment mechanisms, which cannot adapt to the dynamic evolution of patients' conditions and real-time changes in medical resources. At the data processing level, existing technologies lack the ability to integrate multimodal medical data, and heterogeneous data such as vital signs, imaging features, biochemical indicators, and symptom descriptions are often analyzed in isolation, leading to the omission of key clinical clues. In terms of system architecture, centralized processing models face resource contention issues in high-concurrency scenarios. When emergency patients surge, system response time increases exponentially, severely impacting triage efficiency. Existing technologies have also failed to establish effective cross-module collaboration mechanisms, with preliminary screening and in-depth diagnosis processes being isolated from each other, making it impossible to achieve dynamic verification and knowledge iteration of diagnostic evidence.

[0063] This invention integrates and preprocesses multiple data sources to extract disease feature vectors, including core pathological features, quantitative symptom features, imaging features, and environmental association features. Combined with a knowledge graph, the system maps these features to relevant entities and their relationships into a knowledge graph retrieval table, enabling a comprehensive understanding and rapid retrieval of disease features. By quantifying the feature strength of the disease feature vectors through neural networks and dynamically adjusting their plasticity weights based on the system's real-time load, the system can intelligently allocate computing resources according to the urgency of the disease and the system's load. For acute, critical, and common cases, the system rapidly generates preliminary diagnostic reports and diagnostic features through a rapid response cluster, ensuring timely treatment for patients. For complex cases, the system assigns them to a deep analysis cluster for in-depth analysis, avoiding delays or misdiagnosis due to uneven resource allocation. This intelligent triage and rapid response capability allows the system to quickly identify patients requiring urgent treatment in complex medical environments and provide them with timely medical guidance, significantly improving the quality and efficiency of medical services.

[0064] This invention constructs a processing cluster layer, comprising a rapid response cluster and a deep analysis cluster, enabling collaborative work to achieve comprehensive disease analysis. The rapid response cluster primarily handles common diseases and acute and critical illnesses, while the deep analysis cluster focuses on complex cases. This division of labor fully utilizes system resources and improves processing efficiency. Deep analysis nodes can obtain the required disease feature vectors through a mapping graph lookup table and send interactive requests to the rapid processing nodes that store these feature vectors. The rapid processing nodes then transmit relevant information to the deep analysis nodes through a dynamic transmission channel. This interactive mechanism allows the two clusters to share information and conduct collaborative analysis, thereby improving the accuracy and comprehensiveness of diagnosis. Furthermore, the system generates correlation indicators and strengthens the dynamic transmission channel between the rapid processing nodes and deep analysis nodes based on these indicators, enabling the system to adaptively adjust resource allocation and improve information transmission efficiency. The deep analysis cluster updates the etiology analysis report based on the disease feature vectors, generating complex etiology reports and providing doctors with more comprehensive diagnostic information.

[0065] In one embodiment, step S1, which involves collecting patient body data to generate a data cluster and extracting features from the data cluster to obtain a disease feature vector, includes:

[0066] S11. Collect patient medical data including vital signs, symptom descriptions, imaging data, and biochemical indicators;

[0067] S12. Preprocess the medical data to extract disease feature vectors, including core pathological features, symptom quantification features, imaging features, and environmental association features;

[0068] S13. Combining the knowledge graph, the entities associated with the disease feature vector and their relationships are mapped into a mapping graph retrieval table;

[0069] As described in steps S11-S13 above, firstly, through a combination of automated and manual data entry, the patient's vital signs (including but not limited to body temperature, pulse, respiration, blood pressure, blood oxygen saturation), symptom descriptions (including but not limited to pain, dyspnea, cough, etc.), imaging data (including but not limited to X-rays, CT scans, MRI scans, etc.), and biochemical indicators (including but not limited to complete blood count, urinalysis, liver and kidney function tests, electrolytes, etc.) are collected and stored in a structured database. Next, the collected data undergoes preprocessing operations such as cleaning, transformation, and standardization to eliminate missing, outlier, and duplicate values, and continuous data is discretized. Then, using technologies such as natural language processing and computer vision, core pathological features (including but not limited to tumor type, differentiation degree, lymph node metastasis, and gene mutations) are extracted from the preprocessed data. Symptom quantification features include, but are not limited to, symptom keywords, severity, duration, and combination features; image features include, but are not limited to, shape, texture, grayscale, and deep learning features; and environmental association features include, but are not limited to, geographical location, air quality, meteorological data, lifestyle habits, and occupational information. These features together constitute a comprehensive description of the patient's condition, providing a basis for intelligent triage. Finally, to better utilize medical knowledge, disease feature vectors are linked to entities in the knowledge graph, and relationships between entities are extracted to construct a mapping graph retrieval table. This table stores entities, relationships, and their attributes for easy subsequent retrieval, providing a more comprehensive basis for disease diagnosis and triage. Through these steps, the system can extract valuable information from the raw data, providing a data foundation and knowledge support for subsequent intelligent triage.

[0070] In one embodiment, step S2, which involves constructing a dynamic routing layer to convert disease feature vectors into plasticity weight matrices, includes:

[0071] S21. Quantifying the feature strength of disease feature vectors using neural networks. The real-time load rate of the system is obtained based on the resource monitoring unit. ;

[0072] S22. The plasticity weights for constructing the dynamic routing layer to generate disease feature vectors are:

[0073]

[0074] S23, among which Let be the plasticity weight of the i-th disease feature vector. As a clinical prior weight, For load sensitivity coefficient, Age decay constant, For time increments;

[0075] S24. Construct a plasticity weight matrix based on the plasticity weights of disease feature vectors;

[0076] As described in steps S21-S24 above, the core function of the dynamic routing layer is to transform disease feature vectors into plasticity weight matrices so that the subsequent processing cluster can intelligently allocate resources and schedule tasks based on these weights. First, a neural network is used to quantify the feature strength of the disease feature vectors. This neural network is typically a multilayer perceptron (MLP) or convolutional neural network (CNN), whose input is the disease feature vector and output is the strength value of each feature. Training this neural network requires a large amount of clinical data to ensure it can accurately assess the importance of features. Second, the real-time system load rate is obtained based on the resource monitoring unit. The resource monitoring unit is responsible for monitoring the usage of system resources such as CPU utilization, memory usage, and network bandwidth, and summarizing this information into a comprehensive load rate index. This index reflects the system's busy level and is used to dynamically adjust the plasticity weights. Then, the plasticity weight of each disease feature vector is calculated according to a given formula, where... This is the clinical prior weight, representing the clinical importance of the feature, and is pre-set by medical experts. The Sigmoid function, representing feature strength, maps feature strength values ​​to the [0, 1] interval, indicating the activation level of the feature. Finally, the plasticity weights of all disease feature vectors are combined into a plasticity weight matrix. Each row of this matrix represents a disease feature vector, each column represents a feature, and the elements in the matrix are the plasticity weights of the corresponding features. This matrix is ​​passed to the processing cluster layer to guide resource allocation and task scheduling. In this way, the dynamic routing layer realizes the dynamic adjustment of resource allocation strategy based on the strength of disease features, system load, and time factors, while ensuring that disease feature vectors with high clinical prior weights are given higher weights under high system load, thus improving the efficiency and robustness of the system.

[0077] In one embodiment, step S3, which involves building a processing cluster layer and a fast response cluster and a deep analytics cluster, includes:

[0078] S31. Construct a rapid response cluster consisting of multiple rapid processing nodes. The rapid response cluster is used to handle acute and critical illnesses as well as ordinary cases.

[0079] S32. Construct a deep analysis cluster consisting of multiple deep analysis nodes. The deep analysis cluster is used to process complex cases.

[0080] S33. Fast processing nodes and deep analysis nodes are connected via a dynamic transmission channel;

[0081] As described in steps S31-S33 above, the core of building the processing cluster layer lies in designing an efficient fast response cluster and a deep analysis cluster, and establishing a dynamic connection between the two to achieve optimized processing of different types of cases. The fast response cluster design scheme is as follows: fast processing nodes typically use high-performance, low-latency servers equipped with large-capacity memory and high-speed solid-state drives to meet the needs of rapid data processing; the number of CPU cores can be adjusted according to the actual load, but it is usually necessary to ensure that each node has sufficient computing power to handle concurrent requests. The fast response cluster typically uses in-memory databases such as Redis, Memcached, or high-speed caches to store frequently accessed data, such as diagnostic rules for common diseases, basic patient information, etc., which can reduce access to the backend database and improve response speed. The deep analysis cluster design scheme is as follows: Deep analytics nodes typically utilize GPU servers, equipped with high-performance GPUs and ample memory to support complex computational tasks. While the number of CPU cores can be relatively small, each node must possess sufficient computing power to coordinate GPU computation. Deep analytics nodes are primarily responsible for executing complex tasks, such as image recognition, natural language processing, and gene sequence analysis using deep learning models. These tasks typically require significant computation time and resources. Deep analytics clusters often need access to large amounts of medical data, such as imaging data, genomic data, and pathology reports. This data can be stored in a distributed file system or object storage service to support high-throughput data access. Fast processing nodes and deep analytics nodes are connected via a dynamic transport channel. This design allows for the construction of a highly efficient processing cluster layer that intelligently allocates computing resources based on the type and complexity of cases, enabling rapid response and deep analysis. The dynamic transport channel ensures data exchange between the fast processing nodes and deep analytics nodes, allowing the two clusters to work collaboratively to complete complex medical tasks.

[0082] In one embodiment, step S3, which involves inputting the plasticity weight matrix into the processed cluster layer to obtain the analysis results, includes:

[0083] S34, preset response threshold, matching threshold, and important disease feature database;

[0084] S35. Use a semantic matching algorithm to match the disease feature vector with an important disease feature database to obtain the disease matching degree;

[0085] S36. Quickly generate preliminary diagnostic reports and diagnostic features from disease feature vector input values ​​that are equal to or exceed the response threshold or matching threshold;

[0086] S37. Input the disease feature vectors below the response threshold into the deep analysis cluster to obtain the etiology analysis report and etiology analysis features;

[0087] S38. Simultaneously construct a cluster interaction mechanism, and determine whether there is interaction between the fast processing node and the deep analysis node based on the cluster interaction mechanism.

[0088] As described in steps S34-S38 above, the plasticity weight matrix is ​​input into the processing cluster layer. This aims to intelligently allocate disease feature vectors to either the rapid response cluster or the deep analysis cluster for processing based on preset thresholds and an important disease feature library. A cluster interaction mechanism is also constructed to optimize diagnostic results. The key to the rapid response cluster lies in its rapid diagnostic capability. Disease feature vectors equal to or exceeding the response threshold or matching threshold indicate an urgent condition or obvious disease characteristics and are directly input into the rapid response cluster. Each rapid processing node is pre-loaded with diagnostic rules and templates for common diseases and critical illnesses. Through an efficient rule engine or a pre-trained lightweight model, a preliminary diagnostic report is quickly generated, and diagnostic features (such as disease type and severity) are extracted. The deep analysis cluster targets disease feature vectors below the response threshold; these vectors represent cases with more complex or unclear conditions that require further analysis. The deep analysis cluster utilizes high-performance computing resources and complex deep learning models to perform etiological analysis on disease feature vectors, such as 3D reconstruction of image data and lesion identification, and mutation analysis of gene sequences. The analysis results generate etiological analysis reports and etiological features, which may reveal potential disease risks or more precise diagnostic information. Through this hierarchical processing approach, the system can rationally allocate computing resources according to the urgency and complexity of the condition, improving diagnostic efficiency and accuracy. At the same time, the construction of the cluster interaction mechanism enables information sharing and collaborative analysis between rapid processing nodes and deep analysis nodes, further optimizing diagnostic results.

[0089] In one embodiment, step S38, which involves constructing a cluster interaction mechanism and determining whether fast processing nodes and deep analysis nodes interact based on the cluster interaction mechanism, includes:

[0090] S381. The deep analysis node uses the mapping graph retrieval table to obtain the associated disease feature vectors and sends an interaction request to the fast processing node that stores the required disease feature vectors.

[0091] S382. Upon receiving the interaction request, the fast processing node transmits the stored disease feature vector to the deep analysis node that issued the interaction request via a dynamic transmission channel.

[0092] S383: Deep analysis nodes combine disease feature vectors to update the etiology analysis report and generate complex etiology reports;

[0093] S384. Simultaneously, record and calculate the correlation degree between the disease feature vectors associated with the deep analysis node and the fast processing node, and generate correlation indicators. ;

[0094] S385. When the fast processing node and the deep analysis node receive the same diagnostic features and etiology analysis features respectively, the dynamic transmission channel between the fast processing node and the deep analysis node is strengthened:

[0095]

[0096] S386, among which, The weights are the enhanced weights for the dynamic transmission channel. The weights before dynamic transmission channel enhancement. Indicates the learning rate;

[0097] S387. The bandwidth resources of the dynamic transmission channel are adjusted based on the weight. The higher the weight, the more bandwidth resources are available, and the lower the weight, the less bandwidth resources are available.

[0098] As described in steps S381-S387 above, the cluster interaction mechanism aims to achieve information sharing and collaborative analysis between the rapid response cluster and the deep analysis cluster, thereby improving the diagnostic accuracy and efficiency of complex cases. The core of this mechanism lies in the deep analysis node actively initiating an interaction request and dynamically adjusting the bandwidth resources of the transmission channel according to the correlation and channel enhancement strategy. When performing etiological analysis, the deep analysis node may encounter insufficient information. At this time, the deep analysis node uses the mapping graph retrieval table to find other disease feature vectors associated with the currently analyzed disease feature vector. Through the mapping graph retrieval table, the fast processing node storing the required disease feature vector can be found, and an interaction request is sent to it. The interaction request includes information such as the ID of the required disease feature vector and the address of the deep analysis node. The fast processing node that receives the interaction request first verifies the legality of the request, for example, by checking whether the request comes from a trusted deep analysis node. Then, based on the disease feature vector ID in the request, the corresponding disease feature vector is retrieved from local storage. After retrieving the disease feature vector, the fast processing node transmits it to the deep analysis node that issued the interaction request via a dynamic transmission channel. To ensure the reliability of data transmission, TCP protocol can be used for data transmission, and data integrity verification can be performed using techniques such as checksums. After receiving the disease feature vector from the fast processing node, the deep analysis node merges it with its own disease feature vector, updates the etiology analysis model, and re-performs etiology analysis. The merged etiology analysis report contains more comprehensive information and can more accurately reflect the patient's condition. The deep analysis node records detailed information about this interaction, such as interaction time, requesting node, responding node, and amount of data transmitted, for subsequent correlation calculation and channel enhancement. To evaluate the effectiveness of the interaction, the system records and calculates the correlation between the disease feature vectors associated with the deep analysis node and the fast processing node, generating correlation indicators.Association metrics can be calculated using various methods, such as cosine similarity and Pearson correlation coefficient. Higher association indicates stronger data correlation between two nodes and higher interaction value. Association metrics can be used for subsequent channel enhancement decisions. Dynamic transmission channel enhancement is a key component of the cluster interaction mechanism. When the fast processing node and the deep analysis node receive the same diagnostic and etiological analysis features respectively, it indicates that data transmission between these two nodes is crucial for improving diagnostic accuracy. Therefore, it is necessary to enhance the dynamic transmission channel between them. The bandwidth resources of the dynamic transmission channel are adjusted according to the weight; channels with higher weights have higher bandwidth resources, while channels with lower weights have fewer bandwidth resources. By dynamically adjusting bandwidth resources, important channels can be prioritized for bandwidth, thereby improving the efficiency and reliability of data transmission. Through these steps, the cluster interaction mechanism achieves information sharing and collaborative analysis between the fast processing node and the deep analysis node, improving the diagnostic accuracy and efficiency of complex cases. The dynamic transmission channel enhancement strategy can dynamically adjust bandwidth resources according to the actual situation, ensuring that important channels are prioritized for bandwidth, thus improving the efficiency and reliability of data transmission.

[0099] In one embodiment, step S4, which involves fusing the analysis results of multiple clusters through a decision fusion layer to output a patient guidance report, includes:

[0100] S41. Construct a decision fusion layer including a fast channel layer and a regular channel layer;

[0101] S42, the fast channel layer is used to receive the preliminary diagnostic report and diagnostic features from the fast processing node, and uses a neural network model to comprehensively analyze the preliminary diagnostic report to generate an emergency medical guidance report;

[0102] S43, the routine channel layer is used to receive preliminary diagnostic reports and complex etiology reports, and to generate routine medical visit guide reports through comprehensive analysis.

[0103] As described in steps S41-S43 above, the construction of the decision fusion layer aims to integrate the analysis results of the rapid response cluster and the deep analysis cluster to generate the final medical guidance report, providing patients with clear medical guidance. The decision fusion layer includes a fast channel layer and a regular channel layer, which handle different types of reports respectively. The fast channel layer mainly processes the preliminary diagnostic reports and diagnostic features from the rapid processing nodes. This layer uses a neural network model, such as a multilayer perceptron (MLP) or a recurrent neural network (RNN), to perform semantic understanding and information extraction on the preliminary diagnostic reports. Combined with diagnostic features (such as disease type and severity), it comprehensively analyzes and generates an emergency medical guidance report. This report highlights the emergency situation and key information to guide patients to seek medical attention immediately or take emergency measures. The regular channel layer processes the preliminary diagnostic reports from the rapid processing nodes and the complex etiology reports from the deep analysis cluster. This layer also uses a neural network model, but the model structure may be more complex to accommodate more comprehensive information fusion. The neural network model comprehensively analyzes the preliminary diagnostic report and the complex etiology report to identify potential disease risks and causes, and generates a routine medical visit guidance report. This report provides more detailed medical advice, such as which department to visit and what tests are needed. Through the collaborative work of the fast track layer and the routine track layer, the decision fusion layer can generate corresponding medical visit guidance reports based on different conditions and analysis results, providing patients with timely and accurate medical guidance, thereby improving the efficiency and quality of medical services.

[0104] Example 2, please refer to Figure 2 As shown in this embodiment, a disease pre-screening and triage auxiliary system includes:

[0105] Data acquisition module: Collects patients' medical data to generate data clusters, and extracts features from the data clusters to obtain disease feature vectors;

[0106] Dynamic routing module: Constructs a dynamic routing layer to transform disease feature vectors into a plasticity weight matrix;

[0107] Processing Cluster Module: The processing cluster layer is constructed, including a fast response cluster and a deep analysis cluster. The analysis results are obtained after inputting the plasticity weight matrix into the processing cluster layer.

[0108] Fusion Decision Module: The decision fusion layer integrates the analysis results of multiple clusters and outputs a medical guide report.

[0109] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for assisting in disease pre-screening and triage, characterized in that: Collect patient medical data to generate data clusters, and extract features from the data clusters to obtain disease feature vectors; A dynamic routing layer is constructed to transform disease feature vectors into a plastic weight matrix; The processing cluster layer is constructed, which includes a fast response cluster and a deep analysis cluster. The analysis results are obtained by inputting the plasticity weight matrix into the processing cluster layer. The analysis results from multiple clusters are combined through a decision fusion layer to output a medical visit guide report; The steps for constructing the dynamic routing layer to transform disease feature vectors into plasticity weight matrices are as follows: The feature strength of disease feature vectors is quantified using neural networks, and the real-time system load rate is obtained based on the resource monitoring unit. A dynamic routing layer is constructed, which generates plasticity weights for disease feature vectors based on clinical prior weights, feature strengths of disease feature vectors, and real-time system load rates. The specific calculation formula is as follows: ; in, Let be the plasticity weight of the i-th disease feature vector. As a clinical prior weight, The feature strength of the disease feature vector. For load sensitivity coefficient, Age decay constant, For time increments, This refers to the system's real-time load rate. A plasticity weight matrix is ​​constructed based on the plasticity weights of disease feature vectors.

2. The disease pre-screening and triage auxiliary method according to claim 1, characterized in that: The steps for collecting patient medical data to generate a data cluster, and extracting disease feature vectors from the data cluster are as follows: The patient's medical data collected includes vital signs, symptom descriptions, imaging data, and biochemical indicators; Preprocessing of medical data to extract disease feature vectors includes core pathological features, quantitative symptom features, imaging features, and environmental correlation features; By combining knowledge graphs, entities associated with disease feature vectors and their relationships are mapped into a mapping graph retrieval table.

3. The disease pre-screening and triage auxiliary method according to claim 1, characterized in that: The steps for constructing the processing cluster layer, which includes a fast response cluster and a deep analysis cluster, are as follows: Building a rapid response cluster involves multiple rapid processing nodes, and the rapid response cluster is used to handle acute and critical illnesses as well as ordinary cases; Building a deep analytics cluster involves multiple deep analytics nodes, and the deep analytics cluster is used to process complex cases. The fast processing nodes and deep analysis nodes are connected via a dynamic transmission channel.

4. The disease pre-screening and triage auxiliary method according to claim 1, characterized in that: The steps for inputting the plasticity weight matrix into the cluster layer to obtain the analysis results are as follows: Preset response thresholds, matching thresholds, and a database of important disease features; The disease matching degree is obtained by matching disease feature vectors with an important disease feature database using a semantic matching algorithm; The rapid response cluster generates preliminary diagnostic reports and diagnostic features based on disease feature vector input values ​​that are equal to or exceed the response threshold or matching threshold. Inputting disease feature vectors below the response threshold into the deep analysis cluster yields etiology analysis reports and etiology analysis features; Simultaneously, a cluster interaction mechanism is constructed to determine whether there is interaction between the fast processing nodes and the deep analysis nodes.

5. The disease pre-screening and triage auxiliary method according to claim 4, characterized in that: The steps for constructing the cluster interaction mechanism and determining whether there is interaction between the fast processing nodes and the deep analysis nodes based on the cluster interaction mechanism are as follows: The deep analysis node uses the mapping graph retrieval table to obtain the associated disease feature vectors and sends an interaction request to the fast processing node that stores the required disease feature vectors. Upon receiving the interaction request, the fast processing node will transmit the stored disease feature vector to the deep analysis node that issued the interaction request via a dynamic transmission channel. Deep analysis nodes combine disease feature vectors to update etiology analysis reports and generate complex etiology reports; Simultaneously, the correlation degree between the disease feature vectors associated with the deep analysis nodes and the fast processing nodes is recorded and calculated, generating correlation indicators. ; When the fast processing node and the deep analysis node receive the same diagnostic features and etiology analysis features respectively, the dynamic transmission channel between the fast processing node and the deep analysis node is strengthened based on the weights, learning rate, and correlation indicators before the dynamic transmission channel strengthening. The specific calculation formula is as follows: ; in, The weights are the enhanced weights for the dynamic transmission channel. The weights before dynamic transmission channel enhancement. Indicates the learning rate; The bandwidth resources of the dynamic transmission channel are adjusted based on the weight. Higher weights have more bandwidth resources, while lower weights have less bandwidth resources.

6. The disease pre-screening and triage auxiliary method according to claim 5, characterized in that: The steps for outputting a patient guidance report by fusing the analysis results of multiple clusters through the decision fusion layer are as follows: The decision fusion layer is constructed by including a fast track layer and a regular track layer; The fast channel layer is used to receive preliminary diagnostic reports and diagnostic features from fast processing nodes, and uses a neural network model to comprehensively analyze the preliminary diagnostic reports to generate an emergency medical guidance report; The standard channel layer is used to receive preliminary diagnostic reports and complex etiology reports, and to generate a standard medical visit guide report through comprehensive analysis.

7. A disease pre-screening and triage assistance system, used to implement the disease pre-screening and triage assistance method according to any one of claims 1-6, characterized in that: Data acquisition module: Collects patients' medical data to generate data clusters, and extracts features from the data clusters to obtain disease feature vectors; Dynamic routing module: Constructs a dynamic routing layer to transform disease feature vectors into a plasticity weight matrix; Processing Cluster Module: The processing cluster layer is constructed, including a fast response cluster and a deep analysis cluster. The analysis results are obtained after inputting the plasticity weight matrix into the processing cluster layer. Fusion Decision Module: The decision fusion layer integrates the analysis results of multiple clusters and outputs a medical guide report.

Images