Medical big data management method and system based on cloud edge collaboration

Abstract

The invention discloses a medical big data management method and system based on cloud edge collaboration, and relates to the technical field of cloud edge collaboration. The method comprises the following steps: constructing a cloud data computing center; the cloud data computing center performs feature extraction on cloud data to train a medical diagnosis model; the medical diagnosis model is combined with a federal cloud edge collaborative algorithm to carry out model training on an edge server; the cloud data computing center obtains a detection report or real-time vital signs uploaded by the patient; processing the acquired data, and inputting the processed data into a medical diagnosis model; and outputting a result by the medical diagnosis model, and performing vital sign abnormal value calculation, similarity value calculation and comprehensive similarity value calculation. The medical diagnosis model is trained through the cloud edge coordination algorithm, the vital sign abnormal value calculation, the vital sign similarity value calculation and the comprehensive similarity value calculation are performed according to the output result of the medical diagnosis model, accurate analysis of the patient is realized, and the medical rescue efficiency and quality are improved.

Description

Technical Field

[0001] The present invention belongs to the field of cloud-edge collaboration technology, and in particular, relates to a medical big data management method and system based on cloud-edge collaboration. Background Art

[0002] With the continuous development of medical business informatization, medical business systems are becoming increasingly diversified, various applications, equipment and sensors are gradually digitized, and the amount of medical and health-related data is growing exponentially. Patient information data, clinical data, drug data, electronic medical record data, etc. are scattered in various databases and different business systems. The fragmentation of medical data information is obvious, resulting in poor information flow. This makes it difficult for medical institutions to integrate and use this data, and it is difficult to quickly and efficiently obtain comprehensive and accurate medical information for diagnosis, treatment and decision-making.

[0003] The construction of various business systems lacks necessary industry standards, and medical information standards have not yet been universally implemented, resulting in serious heterogeneity in information systems developed and applied by different institutions. There are differences in data formats, encoding methods, data structures, etc. between different systems, making it difficult for them to communicate and be compatible with each other, and there are obstacles to data interconnection and interoperability. For example, the electronic medical record systems of different hospitals may use different formats and specifications, which requires complex conversion and adaptation when sharing and exchanging data, seriously affecting the circulation and utilization efficiency of medical data.

[0004] At the same time, in terms of real-time processing and analysis of medical data, traditional medical data management systems have gradually exposed problems such as high transmission delay, insufficient computing resources, and complex data storage and management when facing the scale of big data. Although traditional centralized cloud computing provides powerful computing power and large-scale data storage capabilities, it is limited by network conditions and bandwidth in terms of data transmission and real-time response. Especially when sharing data and conducting remote diagnosis between multi-center medical institutions, data transmission delays will seriously affect the efficiency of diagnosis and treatment and the timeliness of patient care. In addition, due to the limited processing power of terminals, the transmission and processing of large amounts of data also face information security issues.

[0005] Based on the above background, there is an urgent need for a new medical big data management method and system that can solve problems such as medical data fragmentation, heterogeneity, poor quality, and limitations of traditional architecture, and realize efficient management, secure sharing and value mining of medical data. Cloud-edge collaborative technology provides new ideas and approaches for solving these problems. Summary of the invention

[0006] The purpose of the present invention is to provide a medical big data management method based on cloud-edge collaboration, to train the medical diagnosis model through the cloud-edge collaboration algorithm, to calculate the vital signs abnormal values, vital signs similarity values ​​and comprehensive similarity values ​​according to the output results of the medical diagnosis model, thereby solving the problems of existing medical data fragmentation, heterogeneity and low utilization rate.

[0007] In order to solve the above technical problems, the present invention is achieved through the following technical solutions:

[0008] The present invention is a medical big data management method based on cloud-edge collaboration, comprising the following steps:

[0009] Step S1: Using data synchronization tools, based on the database logs or data update timestamps of major medical institutions, all business data of medical institutions are synchronized to the medical cloud center in real time or at a scheduled time to build a cloud data computing center;

[0010] Step S2: The cloud data computing center preprocesses the data uploaded to the cloud, and extracts features from the preprocessed data to train the medical diagnosis model;

[0011] Step S3: The medical diagnosis model is combined with the federated cloud-edge collaborative algorithm to perform model training on the edge server;

[0012] Step S4: The cloud data computing center obtains the test report or real-time vital signs uploaded by the patient, and retrieves the patient's historical disease data from the medical record database;

[0013] Step S5: Processing the acquired data and inputting it into the medical diagnosis model;

[0014] Step S6: The medical diagnosis model outputs the results, and performs vital sign abnormality value calculation, similarity value calculation and comprehensive similarity value calculation;

[0015] Step S7: Push the diagnosis result to the corresponding patient;

[0016] Among them, in step S1, the cloud data computing center establishes the fuzzy membership function of medical institution data mining under cloud-edge collaborative computing. The specific formula is:

[0017]

[0018] In the formula, e represents the number of sample categories of medical institution business data, x l represents the sample set of medical institution business data, m k represents the medical cloud center, t represents the weighted index;

[0019] According to the fuzzy membership function, the frequent feature distribution parameters of edge cloud computing are constructed. The specific calculation formula is as follows:

[0020]

[0021] In the formula, x i represents the initial distribution parameter of the medical institution business data, and n represents the length of the mining sample set;

[0022] According to the characteristic distribution parameters, the business data of medical institutions are matched to obtain the statistical data of medical institutions. The specific formula is as follows:

[0023]

[0024] According to the number of statistical coefficients, according to the code element sequence x(λ 0 +Δλ), construct the fuzzy recursive function M in the cloud-edge collaborative computing space v , the specific calculation formula is:

[0025]

[0026] Where M h Represents the fuzzy mining rule set of medical institution business data in cloud-edge collaborative computing space;

[0027] Then the output function of all business data of the medical institution is J(V), and its calculation formula is:

[0028]

[0029] In the formula, β represents the distribution feature dimension, V represents the distribution feature of medical institution business, ω represents the weighted coefficient of the semantic mapping of medical institution business data, (d) 2 Represents a set of similarity features.

[0030] As a preferred technical solution, in step S2, when the cloud data computing center pre-processes the data on the cloud, it parses, extracts, converts, cleans and reduces the data, uses multi-source heterogeneous data fusion technology to perform correlation analysis and normalization on the source data, saves it in Excel or txt document format, manually annotates it, and performs word segmentation processing through a word segmenter to build a medical abnormal data corpus.

[0031] As a preferred technical solution, in step S2, when extracting features from data, a semantic feature mapping model A between feature ontology of medical institution business data is constructed. X , the specific formula is:

[0032]

[0033] In the formula, G X(x, y) represents the directional function of the effective feature extraction direction of the medical institution business data, M (y = 1, 2, ...., M) and N (x = 1, 2, ...., N) represent the correlation dimension of the feature respectively;

[0034] Then the medical institution business data association feature vector s * The specific calculation formula is as follows:

[0035] s * ={x∈A X |F(x)=maxF(x)};

[0036] Where F(x) represents the optimal state parameter.

[0037] As a preferred technical solution, in step S3, the medical diagnosis model includes a self-attention mechanism and a multi-attention mechanism;

[0038] The specific workflow of the self-attention mechanism is as follows:

[0039] Step S31: Calculate a unified query vector Q for each input element;

[0040] Step S32: Calculate a unified key vector K for each input element;

[0041] Step S33: Calculate a unified value vector V for each input element;

[0042] Step S34: Calculate the similarity between each query vector and all key vectors, and normalize them using the softmax function to obtain weights;

[0043] Step S35: Use the weights obtained in step S33 to perform weighted summation on all value vectors to obtain the final output representation;

[0044] The calculation formula of the self-attention mechanism is as follows:

[0045]

[0046] Where Q is the query vector, K is the key vector, V is the value vector, and d k is the key vector dimension, and T represents the matrix transpose.

[0047] As a preferred technical solution, the specific workflow of the multi-attention mechanism is as follows:

[0048] Step D31: Perform three linear projections on the input sequence to obtain vectors Q, K, and V respectively;

[0049] Step D32: Perform a separate autonomous force calculation for each group (Q, K, V), and multi-head calculations are accelerated through parallel processing;

[0050] Step D33: concatenate the output results of each sub-attention;

[0051] Step D34: The concatenated result is again linearly projected to map it back to the same dimension as the original input;

[0052] The multi-attention mechanism concatenates the softmax output vectors obtained by n self-attentions, and the calculation formula is as follows:

[0053] addArrention(Q,K,V)=concat(Arrention 1 ,Arrention 2 ,...,Arrention n )×a;

[0054] Where a represents the d×d projection matrix.

[0055] As a preferred technical solution, in step S3, the cloud data computing center is used for training, management and data storage of the overall medical diagnosis model; the edge server obtains the business data of the local medical institution, and uses the federated learning algorithm to perform model training locally to obtain a local medical diagnosis model; each edge server independently trains the local medical diagnosis model;

[0056] The specific implementation process of the federated learning algorithm is as follows:

[0057] Step 1: The cloud data computing center obtains the parameter C(t) of the edge server node i. The specific calculation formula is as follows:

[0058]

[0059] Step 2: Each edge server obtains the parameter C(t-1) from the cloud data computing center at time t and calculates the parameter C of the local node i at time t i (t), the specific calculation formula is as follows:

[0060]

[0061] Where N represents the number of edge servers, D = {D 1 ,D 2 ,...,D n} represents the disease data set.

[0062] As a preferred technical solution, the edge server uploads the trained local medical diagnosis model to the cloud data computing center. The cloud data computing center receives the medical diagnosis model parameters from each edge server and uses a predetermined aggregation strategy to generate a global medical diagnosis model. The cloud data computing center optimizes and adjusts the global medical diagnosis model based on its own public data set and historical data, and then sends it to each edge server again for fine-tuning.

[0063] As a preferred technical solution, the calculation of vital sign abnormal values, vital sign similarity values ​​and comprehensive similarity values ​​is performed; when the vital sign abnormal values ​​are calculated, the vital sign indicators of each patient are standardized, and the formula is: Where, X j,min and X j,max Respectively represent X j The minimum and maximum values ​​of the index j; calculate the information entropy of In the formula, m is the number of samples; in calculating entropy weight So the abnormal values ​​of vital signs are:

[0064] When calculating the vital sign similarity value, assume that the vital sign abnormality value vector of the confirmed patient A is V A =(V A1 ,V A2 ,...,V An ), the abnormal value vector of vital signs of undiagnosed patient B is V B =(V B1 ,V B2 ,...,V Bn ), then the cosine similarity is The Mahalanobis distance is but In the formula, S is the covariance matrix and a is the weight;

[0065] When calculating the comprehensive similarity value, the weights of the abnormal vital sign value Sly and the historical disease similarity value Slyhis are determined by using the hierarchical analysis method. Sly and w Slyhis , construct the judgment matrix Calculate the maximum eigenvalue λ max And the eigenvector W, normalize the eigenvector to get the weight, and construct the fuzzy relationship matrix Where r ab If necessary, the membership degree from index a to evaluation level b is calculated, and the comprehensive similarity is calculated by fuzzy transformation B = W × R, B = (b 1 ,b 2 ,b 3 ), bb Indicates the degree of membership of the comprehensive similarity value to evaluation level b.

[0066] The present invention is a medical big data management system based on cloud-edge collaboration, including a cloud data computing center, an edge server and a terminal;

[0067] The cloud data computing center includes a cloud platform manager, a data storage server, a distributed cloud-edge collaborative system and a data processing system; the cloud platform management server is used for monitoring, managing and coordinating the entire terminal; the data storage server is used for medical data information, patient information, patient physical examination information and patient electronic medical records; the distributed cloud-edge collaborative system is used to complete collaborative operations with the edge server and send the trained medical diagnosis model to the edge server; the data processing system is used to train the medical diagnosis model;

[0068] The edge server is used to perform model training on the edge server in combination with the federated cloud-edge collaborative algorithm and provide feedback to the cloud data computing center;

[0069] The terminal is used to collect the patient's application information and receive the diagnosis information fed back by the cloud data computing center.

[0070] The present invention has the following beneficial effects:

[0071] (1) The present invention trains the medical diagnosis model through the cloud-edge collaboration algorithm, and calculates the abnormal values ​​of vital signs, the similarity values ​​of vital signs, and the comprehensive similarity values ​​according to the output results of the medical diagnosis model, so as to achieve accurate analysis of patients and improve the efficiency and quality of medical rescue;

[0072] (2) The present invention sends the global model optimized by the cloud data computing center to each edge server. After receiving the global model, the edge server performs fine-tuning training again in combination with local data to make the model more adaptable to the characteristics of local data. At the same time, it can also verify the performance of the global model locally, thereby improving the diagnostic efficiency and accuracy of the medical diagnosis model.

[0073] Of course, any product implementing the present invention does not necessarily need to achieve all of the advantages described above at the same time. BRIEF DESCRIPTION OF THE DRAWINGS

[0074] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for describing the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0075] Figure 1This is a flow chart of a medical big data management method based on cloud-edge collaboration of the present invention;

[0076] Figure 2 This is a structural schematic diagram of a medical big data management system based on cloud-edge collaboration of the present invention. DETAILED DESCRIPTION

[0077] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0078] In addition, the technical features involved in the various embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0079] In order to make the purpose, technical solution and advantages of this application clearer, the following Figure 1-2 The implementation methods of the present application are described in further detail.

[0080] Before introducing the embodiments of the present application, the cloud-edge collaboration technology is first described.

[0081] Cloud-edge collaboration refers to extending cloud computing capabilities to edge nodes close to terminal devices. By establishing edge nodes close to devices, the cloud and edge sides can work together organically. The cloud-edge collaboration platform manages edge nodes, provides cloud applications with the ability to extend to the edge side, links edge and cloud data, and realizes remote control, data processing, analytical decision-making, and intelligent applications of edge resources.

[0082] The advantages of cloud-edge collaboration are:

[0083] (1) Reduce costs: Reduce the demand and investment in enterprise IT infrastructure. Cloud computing services can save 30%-40% of costs.

[0084] (2) Improve response speed: Compared with traditional local infrastructure, it can process data better and respond to user needs faster, and can use data on-site for analysis, calculation, storage, and publishing.

[0085] (3) Save time: The use of key edge computing technologies, such as data collection, network access, and remote operation, can help companies complete tasks faster and more stably.

[0086] (4) Improved security: Sensitive data can be transferred from local storage to the cloud through device and environment security protocols. Data can also be stored on local servers and in cloud containers to ensure encrypted operations.

[0087] In order to make the purpose, technical solutions and advantages of this application more clear, the following Figure 1-2 It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0088] Embodiment 1

[0089] See also Figure 1 As shown, the present invention is a medical big data management method based on cloud-edge collaboration, comprising the following steps:

[0090] Step S1: Using data synchronization tools, based on the database logs or data update timestamps of major medical institutions, all business data of medical institutions are synchronized to the medical cloud center in real time or at a scheduled time to build a cloud data computing center;

[0091] Step S2: The cloud data computing center preprocesses the data uploaded to the cloud, and extracts features from the preprocessed data to train the medical diagnosis model;

[0092] Step S3: The medical diagnosis model is combined with the federated cloud-edge collaborative algorithm to perform model training on the edge server;

[0093] Step S4: The cloud data computing center obtains the test report or real-time vital signs uploaded by the patient, and retrieves the patient's historical disease data from the medical record database;

[0094] Step S5: Processing the acquired data and inputting it into the medical diagnosis model;

[0095] Step S6: The medical diagnosis model outputs the results, and performs vital sign abnormality value calculation, similarity value calculation and comprehensive similarity value calculation;

[0096] Step S7: Push the diagnosis result to the corresponding patient;

[0097] Because major hospitals have deficiencies in information construction and lack of construction of medical databases, it is not conducive to the unified and standardized storage of patient diagnostic information. The large scale and variety of electronic medical record data increase the difficulty of mining and utilizing medical data in the later stage. Therefore, after obtaining the electronic medical records of major hospitals, it is necessary to integrate and aggregate the electronic medical record data.

[0098] Among them, in step S1, the cloud data computing center establishes the fuzzy membership function of medical institution data mining under cloud-edge collaborative computing. The specific formula is:

[0099]

[0100] In the formula, e represents the number of sample categories of medical institution business data, x l represents the sample set of medical institution business data, m k represents the medical cloud center, t represents the weighted index;

[0101] Fuzzy membership function is a function that maps the state variables of objective things to fuzzy sets. Let μ kl is m k In x l The membership degree in the fuzzy set is μ kl That is m k The fuzzy membership function of . Its value range is usually [0,1], 0 means that the element does not belong to the fuzzy set at all, 1 means that the element completely belongs to the fuzzy set, and values ​​between 0 and 1 mean that the element partially belongs to the fuzzy set.

[0102] According to the fuzzy membership function, the frequent feature distribution parameter x of edge cloud computing is constructed. The specific calculation formula is as follows:

[0103]

[0104] In the formula, x i represents the initial distribution parameters of the medical institution's business data, and n represents the length of the mining sample set; by analyzing the frequent feature distribution parameters, it is possible to determine which features appear frequently and have an important impact on data classification, clustering and other tasks, thereby selecting the most representative features, reducing feature dimensions, and improving model training efficiency and accuracy; it is also possible to analyze the patient's symptoms, examination indicators and other frequent feature distribution parameters, which can assist doctors in disease diagnosis and condition prediction. In diabetes diagnosis, the distribution parameters of indicators such as blood sugar and glycosylated hemoglobin can help doctors determine whether a patient has diabetes and the severity of the disease;

[0105] According to the characteristic distribution parameters, the business data of medical institutions are matched to obtain the statistical data of medical institutions. The specific formula is as follows:

[0106]

[0107] In data mining and machine learning, based on the importance assessment of statistical coefficients, select features that have a greater impact on the target variable, remove redundant or irrelevant features, and improve model training efficiency and accuracy;

[0108] According to the number of statistical coefficients, according to the code element sequence x(λ 0+Δλ), construct the fuzzy recursive function M in the cloud-edge collaborative computing space v , the specific calculation formula is:

[0109]

[0110] Where M h Represents the fuzzy mining rule set of medical institution business data in the cloud-edge collaborative computing space; fuzzy recursive functions can be used to extract and analyze features, and by recursively processing feature information at different levels and combining fuzzy classification rules, the accuracy of recognition and classification can be improved to adapt to different pattern changes;

[0111] Then the output function of all business data of the medical institution is J(V), and its calculation formula is:

[0112]

[0113] In the formula, β represents the distribution feature dimension, V represents the distribution feature of medical institution business, ω represents the weighted coefficient of the semantic mapping of medical institution business data, (d) 2 Represents a set of similarity features.

[0114] In step S2, the cloud data computing center pre-processes the data on the cloud by parsing, extracting, converting, cleaning and reducing the data. It uses multi-source heterogeneous data fusion technology to perform correlation analysis and normalization on the source data, saves it in Excel or txt document format, and constructs a medical abnormality data corpus after manual annotation and word segmentation through a word segmenter.

[0115] In step S2, when extracting features from the data, a semantic feature mapping model A between the feature ontology of the medical institution business data is constructed. X , the specific formula is:

[0116]

[0117] In the formula, G X (x, y) represents the directional function of the effective feature extraction direction of the medical institution business data, M (y = 1, 2, ...., M) and N (x = 1, 2, ...., N) represent the correlation dimension of the feature respectively;

[0118] Then the medical institution business data association feature vector s * The specific calculation formula is as follows:

[0119] s * ={x∈A X |F(x)=maxF(x)};

[0120] Where F(x) represents the optimal state parameter.

[0121] In step S3, the medical diagnosis model includes a self-attention mechanism and a multi-attention mechanism;

[0122] The specific workflow of the self-attention mechanism is as follows:

[0123] Step S31: Calculate a unified query vector Q for each input element;

[0124] Step S32: Calculate a unified key vector K for each input element;

[0125] Step S33: Calculate a unified value vector V for each input element;

[0126] Step S34: Calculate the similarity between each query vector and all key vectors, and normalize them using the softmax function to obtain weights;

[0127] Step S35: Use the weights obtained in step S33 to perform weighted summation on all value vectors to obtain the final output representation;

[0128] The calculation formula of the self-attention mechanism is as follows:

[0129]

[0130] Where Q is the query vector, K is the key vector, V is the value vector, and d k is the key vector dimension, and T represents the matrix transpose.

[0131] The specific workflow of the multi-attention mechanism is as follows:

[0132] Step D31: Perform three linear projections on the input sequence to obtain vectors Q, K, and V respectively;

[0133] Step D32: Perform a separate autonomous force calculation for each group (Q, K, V), and multi-head calculations are accelerated through parallel processing;

[0134] Step D33: concatenate the output results of each sub-attention;

[0135] Step D34: The concatenated result is again linearly projected to map it back to the same dimension as the original input;

[0136] The multi-attention mechanism concatenates the softmax output vectors obtained by n self-attentions. The calculation formula is as follows:

[0137] addArrention(Q,K,V)=concat(Arrention 1 ,Arrention2 ,...,Arrention n )×a;

[0138] Where a represents the d×d projection matrix.

[0139] In step S3, the cloud data computing center is used for training, management and data storage of the overall medical diagnosis model; the edge server obtains the business data of the local medical institution and uses the federated learning algorithm to perform model training locally to obtain a local medical diagnosis model; each edge server independently trains the local medical diagnosis model;

[0140] The specific implementation process of the federated learning algorithm is as follows:

[0141] Step 1: The cloud data computing center obtains the parameter C(t) of the edge server node i. The specific calculation formula is as follows:

[0142]

[0143] Step 2: Each edge server obtains the parameter C(t-1) from the cloud data computing center at time t and calculates the parameter C of the local node i at time t i (t), the specific calculation formula is as follows:

[0144]

[0145] Where N represents the number of edge servers, D = {D 1 ,D 2 ,...,D n} represents the disease data set.

[0146] The edge server uploads the trained local medical diagnosis model to the cloud data computing center. The cloud data computing center receives the medical diagnosis model parameters from each edge server and uses a predetermined aggregation strategy to generate a global medical diagnosis model. The cloud data computing center optimizes and adjusts the global medical diagnosis model based on its own public data set and historical data, and then sends it to each edge server for fine-tuning.

[0147] Calculate the abnormal values ​​of vital signs, the similar values ​​of vital signs and the comprehensive similar values; when calculating the abnormal values ​​of vital signs, standardize the vital sign indicators of each patient, and the formula is: Where, X j,min and X j,max Respectively represent X j The minimum and maximum values ​​of the index j; calculate the information entropy of In the formula, m is the number of samples; in calculating entropy weight So the abnormal values ​​of vital signs are:

[0148] When calculating the similarity value of vital signs, let the abnormal value vector of vital signs of confirmed patient A be V A =(V A1 ,V A2 ,...,V An ), the abnormal value vector of vital signs of undiagnosed patient B is V B =(V B1 ,V B2 ,...,V Bn ), then the cosine similarity is The Mahalanobis distance is but In the formula, S is the covariance matrix and a is the weight;

[0149] When calculating the comprehensive similarity value, the weights of the abnormal value of vital signs Sly and the historical disease similarity value Slyhis are determined by using the hierarchical analysis method as w Sly and w Slyhis , construct the judgment matrix Calculate the maximum eigenvalue λ max And the eigenvector W, normalize the eigenvector to get the weight, and construct the fuzzy relationship matrix Where r ab If necessary, the membership degree from index a to evaluation level b is calculated, and the comprehensive similarity is calculated by fuzzy transformation B = W × R, B = (b 1 ,b 2 ,b 3 ), b b Indicates the degree of membership of the comprehensive similarity value to evaluation level b.

[0150] Embodiment 2

[0151] See also Figure 2 As shown, the present invention is a medical big data management system based on cloud-edge collaboration, which can be used to execute the method content of Example 1 of the present invention, including a cloud data computing center, an edge server and a terminal;

[0152] The cloud data computing center includes a cloud platform manager, a data storage server, a distributed cloud-edge collaborative system, and a data processing system; the cloud platform management server is used to monitor, manage, and coordinate the entire terminal; the data storage server is used to store medical data information, patient information, patient physical examination information, and the patient's electronic medical records; the distributed cloud-edge collaborative system is used to complete collaborative operations with the edge server and send the trained medical diagnosis model to the edge server; the data processing system is used to train the medical diagnosis model;

[0153] The edge server is used to combine the federated cloud-edge collaborative algorithm to train the model on the edge server and feed it back to the cloud data computing center;

[0154] The terminal is used to collect the patient's application information and receive diagnostic information fed back by the cloud data computing center.

[0155] It is worth noting that in the above system embodiment, the various units included are only divided according to functional logic, but are not limited to the above division, as long as the corresponding functions can be achieved; in addition, the specific names of the functional units are only for the convenience of distinguishing each other, and are not used to limit the scope of protection of the present invention.

[0156] In addition, those skilled in the art can understand that all or part of the steps in the above-mentioned embodiments can be completed by instructing related hardware through a program, and the corresponding program can be stored in a computer-readable storage medium.

[0157] The preferred embodiments of the present invention disclosed above are only used to help illustrate the present invention. The preferred embodiments do not describe all the details in detail, nor do they limit the invention to the specific implementation methods described. Obviously, many modifications and changes can be made according to the content of this specification. This specification selects and specifically describes these embodiments in order to better explain the principles and practical applications of the present invention, so that those skilled in the art can understand and use the present invention well. The present invention is limited only by the claims and their full scope and equivalents.

Claims

1. A medical big data management method based on cloud-edge collaboration, characterized in that: The steps include: Step S1: Using data synchronization tools, based on the database logs or data update timestamps of major medical institutions, all business data of medical institutions are synchronized to the medical cloud center in real time or at a scheduled time to build a cloud data computing center; Step S2: The cloud data computing center preprocesses the data uploaded to the cloud, and extracts features from the preprocessed data to train the medical diagnosis model; Step S3: The medical diagnosis model is combined with the federated cloud-edge collaborative algorithm to perform model training on the edge server; Step S4: The cloud data computing center obtains the test report or real-time vital signs uploaded by the patient, and retrieves the patient's historical disease data from the medical record database; Step S5: Processing the acquired data and inputting it into the medical diagnosis model; Step S6: The medical diagnosis model outputs the results, and performs vital sign abnormality value calculation, similarity value calculation and comprehensive similarity value calculation; Step S7: Push the diagnosis result to the corresponding patient; Among them, in step S1, the cloud data computing center establishes the fuzzy membership function of medical institution data mining under cloud-edge collaborative computing. The specific formula is: In the formula, e represents the number of sample categories of medical institution business data, x l represents the sample set of medical institution business data, m k represents the medical cloud center, t represents the weighted index; According to the fuzzy membership function, the frequent feature distribution parameter x of edge cloud computing is constructed. The specific calculation formula is as follows: In the formula, x i represents the initial distribution parameter of the medical institution business data, and n represents the length of the mining sample set; According to the characteristic distribution parameters, the business data of medical institutions are matched to obtain the statistical data of medical institutions. The specific formula is as follows: According to the number of statistical coefficients and the code element sequence x(λ0+Δλ), the fuzzy recursive function M in the cloud-edge collaborative computing space is constructed. v , the specific calculation formula is: Where M h Represents the fuzzy mining rule set of medical institution business data in cloud-edge collaborative computing space; Then the output function of all business data of the medical institution is J(V), and its calculation formula is: In the formula, β represents the distribution feature dimension, V represents the distribution feature of medical institution business, ω represents the weighted coefficient of the semantic mapping of medical institution business data, (d) 2 Represents a set of similarity features.

2. According to claim 1, a medical big data management method based on cloud-edge collaboration is characterized in that: In step S2, when the cloud data computing center pre-processes the data on the cloud, it parses, extracts, converts, cleans and reduces the data, uses multi-source heterogeneous data fusion technology to perform correlation analysis and normalization on the source data, saves it in Excel or txt document format, and constructs a medical abnormality data corpus after manual annotation and word segmentation through a word segmenter.

3. According to the medical big data management method based on cloud-edge collaboration according to claim 1, it is characterized in that: In step S2, when extracting features from the data, a semantic feature mapping model A between the feature ontology of the medical institution business data is constructed. X , the specific formula is: In the formula, G X (x, y) represents the directional function of the effective feature extraction direction of the medical institution business data, M (y = 1, 2, ...., M) and N (x = 1, 2, ...., N) represent the correlation dimension of the feature respectively; Then the medical institution business data association feature vector s * The specific calculation formula is as follows: s * ={x∈A X |F(X)=maxF(x)}; Where F(x) represents the optimal state parameter.

4. According to claim 1, a medical big data management method based on cloud-edge collaboration is characterized in that: In step S3, the medical diagnosis model includes a self-attention mechanism and a multi-attention mechanism; The specific workflow of the self-attention mechanism is as follows: Step S31: Calculate a unified query vector Q for each input element; Step S32: Calculate a unified key vector K for each input element; Step S33: Calculate a unified value vector V for each input element; Step S34: Calculate the similarity between each query vector and all key vectors, and normalize them using the softmax function to obtain weights; Step S35: Use the weights obtained in step S33 to perform weighted summation on all value vectors to obtain the final output representation; The calculation formula of the self-attention mechanism is as follows: Where Q is the query vector, K is the key vector, V is the value vector, and d k is the key vector dimension, and T represents the matrix transpose.

5. According to claim 4, a medical big data management method based on cloud-edge collaboration is characterized in that: The specific workflow of the multi-attention mechanism is as follows: Step D31: Perform three linear projections on the input sequence to obtain vectors Q, K, and V respectively; Step D32: Perform a separate autonomous force calculation for each group (Q, K, V), and multi-head calculations are accelerated through parallel processing; Step D33: concatenate the output results of each sub-attention; Step D34: The concatenated result is again linearly projected to map it back to the same dimension as the original input; The multi-attention mechanism concatenates the softmax output vectors obtained by n self-attentions, and the calculation formula is as follows: addRantion(Q,K,V)?concat(Ranntion1,Ranntion2,...,Ranntion n )×a Where a represents the d×d projection matrix.

6. The medical big data management method based on cloud-edge collaboration according to claim 1 is characterized in that: In step S3, the cloud data computing center is used for training, management and data storage of the overall medical diagnosis model; the edge server obtains the business data of the local medical institution, and uses the federated learning algorithm to perform model training locally to obtain a local medical diagnosis model; each edge server independently trains a local medical diagnosis model; The specific implementation process of the federated learning algorithm is as follows: Step 1: The cloud data computing center obtains the parameter C(t) of the edge server node i. The specific calculation formula is as follows: Step 2: Each edge server obtains the parameter C(t-1) from the cloud data computing center at time t and calculates the parameter C of the local node i at time t i (t), the specific calculation formula is as follows: C i (t)=C i (t-1)-▽L i (D,C); Where N represents the number of edge servers, D = {D1, D2, ..., D n } represents the disease data set.

7. The medical big data management method based on cloud-edge collaboration according to claim 6 is characterized in that: The edge server uploads the trained local medical diagnosis model to the cloud data computing center. The cloud data computing center receives the medical diagnosis model parameters from each edge server and generates a global medical diagnosis model using a predetermined aggregation strategy. The cloud data computing center optimizes and adjusts the global medical diagnosis model based on its own public data set and historical data, and then sends it to each edge server for fine-tuning.

8. The medical big data management method based on cloud-edge collaboration according to claim 1 is characterized in that: The calculation of abnormal vital sign values, similar vital sign values ​​and comprehensive similarity values ​​is performed; when the abnormal vital sign values ​​are calculated, the vital sign indicators of each patient are standardized, and the formula is: In the formula, X j,min and X j,max Respectively represent X j The minimum and maximum values ​​of the index j; calculate the information entropy of In the formula, m is the number of samples; in calculating entropy weight So the abnormal values ​​of vital signs are: When calculating the vital sign similarity value, assume that the vital sign abnormality value vector of the confirmed patient A is V A =(V A1 ,V A2 ,...,V An ), the abnormal value vector of vital signs of undiagnosed patient B is V B =(V B1 ,V B2 ,...,V Bn ), then the cosine similarity is The Mahalanobis distance is but In the formula, S is the covariance matrix and a is the weight; When calculating the comprehensive similarity value, the weights of the abnormal vital sign value Sly and the historical disease similarity value Slyhis are determined by using the hierarchical analysis method. Sly and w Slyhis , construct the judgment matrix Calculate the maximum eigenvalue λ max And the eigenvector W, normalize the eigenvector to get the weight, and construct the fuzzy relationship matrix Where r ab If necessary, the membership degree from indicator a to evaluation level b is calculated, and the comprehensive similarity is calculated by fuzzy transformation B = W × R, B = (b1, b2, b3), b b Indicates the degree of membership of the comprehensive similarity value to evaluation level b.

9. A medical big data management system based on cloud-edge collaboration, comprising a cloud data computing center, an edge server and a terminal, characterized in that: The cloud data computing center includes a cloud platform manager, a data storage server, a distributed cloud-edge collaborative system and a data processing system; the cloud platform management server is used for monitoring, managing and coordinating the entire terminal; the data storage server is used for medical data information, patient information, patient physical examination information and patient electronic medical records; the distributed cloud-edge collaborative system is used to complete collaborative operations with the edge server and send the trained medical diagnosis model to the edge server; the data processing system is used to train the medical diagnosis model; The edge server is used to perform model training on the edge server in combination with the federated cloud-edge collaborative algorithm and provide feedback to the cloud data computing center; The terminal is used to collect the patient's application information and receive the diagnosis information fed back by the cloud data computing center.

Images