Internet medical data authority management method based on multi-level user group

Abstract

The invention belongs to the technical field of medical data authority management, and discloses an internet medical data authority management method based on a multi-level user group. Comprising the following steps: acquiring medical data information in an internet medical platform and user information of a user; dividing the users into n hierarchical user groups based on the user information; for each hierarchical user group, clustering the users in the hierarchical user group by adopting a search algorithm to obtain N user sub-groups, and calculating an access permission evaluation value of each user sub-group to the medical data information; based on the access permission evaluation value of the user subgroup, an overall data permission management scheme of the Internet medical platform is constructed by using a directed graph algorithm, so that the efficiency and accuracy of permission management are improved, and the method has good universality and expansibility and has good effects in the aspects of maintaining medical data security, protecting patient privacy rights and interests and the like.

Description

Technical Field

[0001] The present invention relates to the technical field of medical data permission management. More specifically, the present invention relates to an Internet medical data permission management method based on a multi-level user group. Background Art

[0002] The patent with the application publication number CN111062051A discloses a permission management method for a medical data system, including: storing the permissions of medical data as records by adopting a Merkle directed acyclic graph. The records include content, content hash value, left previous record hash value, right previous record hash value, and record hash value. The content includes first record hash value, authorized person, authorized record hash value, permission, authorization date, and validity period. The record corresponding to the permission when the medical data is generated in the medical data system is used as the first record. The first record hash value is the record hash value of the first record. The authorized record hash value is the record hash value of the record in which the authorized person obtains the authorized right. The record hash value is the hash value generated by hashing the concatenation of the content hash value, left previous record hash value, and right previous record hash value of this record; it can quickly detect historical tampering of permissions.

[0003] However, in the existing Internet medical data permission management, there is still a lack of comprehensive analysis of user attributes and a refined classification management mechanism, making it difficult to accurately depict and identify the differentiated data access needs of different types of users. Secondly, even within the same user category, the existing methods cannot quantify and evaluate the internal access demand differences of the user group. For example, the different needs of different doctors for medical history data, inspection reports, etc. cannot be reflected. Moreover, the formulation of data permission policies relies too much on manual experience and lacks an automated and intelligent analysis and decision-making mechanism, making it difficult to balance various complex factors such as user attributes, data characteristics, and access needs, resulting in poor rationality and effectiveness of the policies. In addition, the existing medical data permission management policies are relatively single and static, unable to formulate differentiated and refined access permissions for different users and data, and prone to situations of over-opening or over-restriction, affecting data security and controllability. Finally, the adjustment and optimization of policies require a large amount of human costs, making it difficult to meet the development needs of medical informatization and difficult to fully protect patient privacy and data security.

[0004] In view of this, the present invention proposes an Internet medical data permission management method based on a multi-level user group to solve the above problems. Summary of the Invention

[0005] To overcome the above-mentioned defects of the prior art and to achieve the above object, the present invention provides the following technical solution: An Internet medical data permission management method based on a multi-level user group, including: S1, obtaining medical data information in an Internet medical platform and user information of users; S2. Divide users into n hierarchical user groups based on user information; S3. For each hierarchical user group, use a search algorithm to cluster the users within the hierarchical user group to obtain N user subgroups, and calculate the access permission evaluation value of each user subgroup for medical data information; S4. Based on the access permission evaluation value of the user subgroup, use a directed graph algorithm to construct an overall data permission management scheme for the Internet medical platform.

[0006] Further, the user information includes personal identity information, medical-related identity and permission level information; Medical data information includes patient medical history data, patient examination report data, patient treatment plan data, patient medication data, patient surgery data, doctor diagnosis data and medical insurance data.

[0007] Further, the division method of the hierarchical user group includes: preprocessing the user information to obtain standard user information; Transform the standard user information corresponding to each user into an ordered combination between vectors, that is, a combined vector; construct a grid according to the number of users and the number n of hierarchical user groups; Each node of the grid is assigned a node weight vector with the same dimension as the input data; traverse each combined vector of the standard user information as a sample, and calculate the similarity between the sample and each node weight vector; find the node with the highest similarity to the sample as the best matching unit; Update the node weight vectors of the best matching unit and its neighboring nodes to move them in the direction of the sample; when the learning rate parameter and the neighborhood function both decay to the minimum value, the grid converges, visualize the grid, obtain the area composed of adjacent nodes, calculate the similarity degree of the node weight vectors inside it, and ensure that the calculated similarity degree is greater than the preset similarity degree threshold, that is, complete the preliminary hierarchical division, mark the preliminary hierarchy according to the characteristics of the node weight vector, that is, the preliminary hierarchical user group; traverse each standard user information, calculate its distance from each preliminary hierarchical user group, and classify it into the nearest preliminary hierarchical user group, that is, complete the division of the hierarchical user group.

[0008] Further, the method for preprocessing the user information includes: For each non-quantized data field in the user information, count the frequency distribution of all its values, denoted as the value frequency; Sort the values in descending order according to the frequency of occurrence. For the values with a frequency of occurrence, perform a temporary sort in alphabetical order to obtain a preliminary sorting result; traverse all the values of the corresponding field, record their sorting results, and form an ordered list; in the ordered list, traverse each value from beginning to end in turn, determine and record its serial number position in the list; after sorting the serial numbers from smallest to largest, the alphabetical order arrangement result of the values of the corresponding field is obtained; record the sorted values one by one to form a sorted value list; based on the sorted value list, define the number of dimensions of the encoding and establish an encoding correspondence table; Traverse all user information, and according to the encoding correspondence table, convert the non-quantitative data into corresponding encoding vectors; that is, obtain the partially quantified user information; Merge the partially quantified user information with the original quantified data fields to form the standard user information.

[0009] Furthermore, the calculation formula for the similarity is: ; where is a sample, is the node weight vector, , and are adjustment coefficients, and ; is the attenuation parameter, is the non-linear similarity function between the sample and the node weight vector; is the sample and the node weight vector the similarity between; ; where is the bandwidth parameter, is the weight parameter, is the frequency parameter, is the sample scale parameter of, is scale parameter of.

[0010] Furthermore, the formula for updating the node weight vector of the best matching unit and its neighboring nodes is: ; where is the learning rate parameter, controlling the update step size; is the update adjustment parameter, is the neighborhood function, is the distance from the node to the best matching unit, is the current traversal iteration number; is the correction function; ; wherein, is a time-varying bandwidth parameter, is a time-varying adjustment parameter; is a time-varying center parameter; is a scale parameter; Time-varying bandwidth parameter ; wherein, is a preset initial bandwidth, is a bandwidth attenuation rate parameter, is a period adjustment parameter; is an angular frequency parameter, is a phase parameter; Correction function ; wherein, and are adjustment weight parameters, and ; is a preference function, is a distance function; is and the distance between; Preference function ; wherein, is a norm adjustment parameter; is the transpose of, is a positive semi-definite matrix, is the Lp norm of, is the norm order; ; wherein, , and are parameters for controlling the shape of the hyperbolic tangent function, is the hyperbolic tangent function.

[0011] Furthermore, the method for obtaining the user subgroup includes: Initially define the scale N3 of the clustering population and generate an initial clustering population. N3 cuckoos are defined in the clustering population, and each cuckoo corresponds to a candidate clustering scheme. Integer string encoding is performed on each cuckoo, that is, N3 candidate clustering schemes; Define the evolution function of the clustering population, calculate the value of the evolution function of each cuckoo, denoted as the evolution value, and sort each cuckoo in descending order according to the size of the evolution value, and select the top m4 cuckoos as the cuckoos with extremely good voices; Calculate the distance between each cuckoo and the cuckoos with extremely good voices. The formula for calculating the distance is: ; wherein, Integer string encoding for cuckoos The th element of is the integer string encoding for a cuckoo with excellent voice The th element of is The distance between and For the th cuckoo, find the cuckoo with the closest distance to it among the cuckoos with excellent voice as the guiding cuckoo for the th cuckoo; Guide the th cuckoo to move towards the position of the guiding cuckoo. The guiding formula is: ; where is the position of the th cuckoo after movement, is the position of the th cuckoo before movement, is the preset maximum number of iterations, is the current number of iterations, is the position of the guiding cuckoo for the th cuckoo ; is the weight coefficient of the population gravitational field; is the population gravitational field of the th cuckoo; ; where is the position of the th cuckoo before movement, is a positive real number parameter; the position after movement is to adjust the candidate clustering scheme to obtain a new cuckoo; Recalculate the evolutionary value of the new cuckoo, sort the new cuckoos in ascending order according to the size of the evolutionary value, and eliminate the first n4 new cuckoos; repeat until the maximum number of iterations is reached; calculate the evolutionary value of the new cuckoo obtained in the last iteration, and take the new clustering scheme corresponding to the new cuckoo with the largest evolutionary value as the optimal clustering scheme, that is, obtain N user subgroups.

[0012] Furthermore, the formula of the evolutionary function is: ; where and are clustering weight parameters; is the compactness function of the clustering ; is the clustering and the clustering The separation function between is the density penalty term for clustering ; ; where is the user index belonging to the cluster ; is the user 's low-dimensional user feature vector; is the centroid embedding vector of the cluster ; is the transpose of is the covariance matrix of the cluster ; is the smoothing bandwidth parameter; The separation function ; where and are the separation weight parameters; is the user 's low-dimensional user feature vector. The user belongs to the cluster ; is the deep neural network model used to output the similarity score between the user and the user ; is the parameter of the deep neural network model; is the entropy of the user relative to the cluster ; ; where and are positive constant parameters; is the total number of all users; is the dispersion term weight parameter; is the scale of the cluster ;

[0013] Furthermore, the calculation method of the access permission evaluation value includes: Define the access weight matrix of the user for different types of medical data information , is an m5×n5 matrix, where m5 is the number of users and n5 is the number of types of medical data information; the element of the access weight matrix represents the access weight vector of the th user for the th type of medical data information; For the th user subgroup, extract the access weight vectors corresponding to all users in the th type of medical data information and take the average to obtain the th user subgroup's average access demand for the th type of medical data information ; then obtain the average access demand of each user subgroup for all medical data information as the access permission evaluation value.

[0014] Furthermore, the construction method of the overall data permission management scheme includes: Represent all medical data information in the Internet medical platform as a weighted directed graph GH=(V, E), where V is the set of nodes, E is the set of all edges. If medical data information a needs to access medical data information v, then there is a directed edge e'=(a, v) in the weighted directed graph; each directed edge e' has a non - negative weight w(e'). Define the source point d as connected to all nodes without predecessors, and define the sink point f as connected to all nodes without successors; For each user subgroup C, construct a directed cost network GH'=(V', E') from the source point to the sink point, where V' is the union of the set of all medical data information and the set of relay points, and the set of relay points is the set of all source points and sink points; E' is the set of comprehensive weight edges; The definition of the set of comprehensive weight edges is: For each v belonging to V and there exists an edge LP, define the weight of the edge LP as the v - th element of the access permission evaluation value; For each v belonging to V and there exists an edge LU, define the weight of the edge LU as 0; the edge LP is the edge connecting v and the source point d, and the edge LU is the edge connecting v and the sink point f; Calculate the shortest path from d to f in the directed cost network GH', and this shortest path corresponds to the best access scheme for the user subgroup C to access the required medical data information; For each user subgroup C, obtain a best access scheme path(C), and merge all path(C) to obtain the overall data permission management scheme.

[0015] The technical effects and advantages of the Internet medical data permission management method based on multi - level user groups in the present invention: The present invention realizes the automated, intelligent, and refined management of a large amount of medical data on the Internet medical platform, thereby greatly improving the security and controllability of data access. Specifically, first, based on the multi-dimensional analysis of user information, users are divided into multiple hierarchical user groups, realizing refined user classification management. Then, within each hierarchical user group, clustering and evaluation models are used to further segment users, quantify the access requirements of each user subgroup for various types of medical data, and lay a foundation for formulating differential data access permission policies. Secondly, using graph theory algorithms, an overall data permission management scheme for the Internet medical platform is constructed. This scheme comprehensively considers various factors such as user attributes, data characteristics, and access requirements, and can formulate refined and reasonable access permission policies for different types of users and data, not only improving the efficiency and accuracy of permission management, but also having good versatility and scalability, and playing a good role in maintaining medical data security and protecting the privacy rights and interests of patients. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 FIG. is a schematic diagram of a method for managing Internet medical data permissions based on multi-level user groups according to the present invention; Figure 2 FIG. is a schematic diagram of a system for managing Internet medical data permissions based on multi-level user groups according to the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0017] The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0018] Embodiment 1 Please refer to Figure 1 As shown, a method for managing Internet medical data permissions based on multi-level user groups in this embodiment includes: S1. Obtain medical data information and user information of users in the Internet medical platform; S2. Based on the user information, divide the users into n hierarchical user groups; S3. For each hierarchical user group, use a search algorithm to cluster the users within the hierarchical user group to obtain N user subgroups, and calculate the access permission evaluation value of each user subgroup for the medical data information; S4. Based on the access permission evaluation values of the user subgroups, use a directed graph algorithm to construct an overall data permission management scheme for the Internet medical platform.

[0019] User information includes personal identity information (name, age, gender), medical-related identity (doctor level, nurse level, patient), and permission level information (administrator, ordinary user); Medical data information includes patient medical history data (number of past medical history entries, number of medical visits), patient examination report data (number of reports, types of examinations), patient treatment plan data (number of treatment types, treatment cycles), patient medication data (number of medication types, dosage statistics), patient surgery data (number of surgery types, surgery duration), doctor diagnosis data (number of diagnosis types, diagnosis time), and medical insurance data (compensation amount, number of claims).

[0020] It should be noted that multiple dimensions of user information can more comprehensively depict the attribute characteristics of users; while medical data information covers all aspects of patients' medical history, examinations, treatments, medications, surgeries, etc., and can also more completely describe the content and characteristics of medical data. Detailed user information helps to conduct fine-grained grouping of users, and obtaining medical data information helps to classify and manage the data.

[0021] The division methods of hierarchical user groups include: Preprocess the user information to obtain standard user information; specifically, for each non-quantitative data field in the user information (such as name, medical-related identity, permission level information, etc.), count the frequency distribution of all its values, denoted as the value frequency.

[0022] According to the magnitude of the value frequency, sort the values in descending order, and arrange the sorted values in alphabetical order to obtain a sorted value list. Specifically, for the values with value frequency, perform a temporary sort in alphabetical order to obtain a preliminary sorting result; traverse all the values of the corresponding field, record their sorting results, and form an ordered list; in the ordered list, traverse each value from beginning to end, determine and record its serial number position in the list; after sorting the serial numbers from small to large, the alphabetical order arrangement result of the values of the corresponding field is obtained; record the sorted values one by one to form a sorted value list; based on the sorted value list, define the number of dimensions of the encoding and establish an encoding correspondence table. Traverse all the user information, and according to the encoding correspondence table, convert the non-quantitative data into corresponding encoding vectors; that is, the partially quantified user information is obtained.

[0023] Merge the partially quantified user information with the original quantified data fields (such as age) to form standard user information; it should be noted that each user corresponds to a standard user information.

[0024] Transform the standard user information corresponding to each user into an ordered combination between vectors, that is, a combined vector.

[0025] Construct a grid according to the number of users and the number of hierarchical user groups \(n\). The number of rows and columns of the grid, such as \(5\times5\), \(8\times8\), etc., is adjusted appropriately according to the actual situation. Each node of the grid is assigned a node weight vector with the same dimension as the input data (standard user information).

[0026] Traverse each combined vector of standard user information as a sample, and calculate the similarity between the sample and each node weight vector; find the node with the highest similarity to the sample as the best matching unit.

[0027] The calculation formula for similarity is: ; where is the sample (combined vector), is the node weight vector, , and are adjustment coefficients to control the weights of the three parts, and ; is the attenuation parameter to control the attenuation rate of the function, is the non - linear similarity function between the sample and the node weight vector; is the sample and the node weight vector between the similarities.

[0028] ; where is the bandwidth parameter to control the smoothness of the function, is the weight parameter to control the weight of the periodic term, is the frequency parameter to control the frequency of the periodic term, is the sample scale parameter, is scale parameter, used to adaptively control the bandwidth of different dimensions.

[0029] Update the node weight vectors of the best matching unit and its neighboring nodes, making them move in the direction of the sample. The update formula is: ; where is the learning rate parameter to control the update step size; is the update adjustment parameter to make the update process of the node weight vector more stable and controllable, avoiding violent fluctuations; is the neighborhood function, is the distance from the node to the best matching unit, is the current traversal iteration number; is the correction function.

[0030] ; where is a time-varying bandwidth parameter that controls the decay of the neighborhood range. is a time-varying adjustment parameter that controls the weight of the new term; is a time-varying center parameter that controls the center position of the distance term; is a scale parameter that controls the scale of the distance term.

[0031] Time-varying bandwidth parameter ; where is a preset initial bandwidth, is a bandwidth decay rate parameter that controls the decay speed of the bandwidth, is a period adjustment parameter that controls the weight of the periodic term; is an angular frequency parameter, is a phase parameter that controls the initial phase of the periodicity.

[0032] The time-varying bandwidth parameter can not only reflect the basic non-linear decay trend, but also superimpose a periodic oscillation related to the number of iterations, making the change process of the bandwidth more dynamic and diverse, thus endowing the neighborhood range with a more complex spatio-temporal structure.

[0033] Correction function ; where and are adjustment weight parameters, and ; is a preference function, is a distance function; is and the distance between.

[0034] Preference function ; where is a norm adjustment parameter that controls the weight of the norm term; is the transpose of, is a positive semi-definite matrix used to perform a linear transformation on the difference between the sample and the weight. The positive semi-definite matrix can encode various prior knowledge, such as feature importance, expected clustering structure, etc.; for example, if certain features are more important, the elements of the corresponding positive semi-definite matrix are set to larger values to enhance the contribution of these features to the correction term; is the Lp norm of, is the norm order and can take different positive real values.

[0035] ; where , and are parameters that control the shape of the hyperbolic tangent function, is the hyperbolic tangent function. By adjusting the parameters , and , different S-shaped curves can be presented, and the influence on the distance shows a non-linear change. When the distance is small, this term is close to 1 and has little influence on the correction term; when the distance is large, this term may be greater than 1 or less than 1, and the influence on the correction term is amplified or reduced. The above parameters can be adjusted to make the function adapt to different data distributions and play a greater potential.

[0036] When the learning rate parameter and the neighborhood function both decay to the minimum value, the grid converges. Visualize the grid, obtain the region composed of adjacent nodes, calculate the similarity degree (average of the similarity degrees) of the node weight vectors inside it, and ensure that the calculated similarity degrees are all greater than the preset similarity degree threshold, that is, the preliminary hierarchical division is completed. Mark the preliminary hierarchy according to the characteristics of the node weight vectors, which is the preliminary hierarchical user group, such as the administrator group, the doctor group, etc.

[0037] It should be noted that in the process of dividing users into preliminary hierarchical user groups, each preliminary hierarchical user group is composed of adjacent nodes, and the weight vectors of these nodes have similar characteristics. Analyze the characteristics of the node weight vectors inside each preliminary hierarchical user group, so as to assign a label or identity to this preliminary hierarchical user group as the mark of the group.

[0038] For example, divide users into 5 preliminary hierarchies; the main elements of the node weight vectors in the first preliminary hierarchy are characteristics of high level and relatively high management authority, such as administrators, chief physicians, etc.; the main elements of the node weight vectors in the second preliminary hierarchy are characteristics of medium level and ordinary diagnosis and treatment authority, such as attending physicians, resident physicians, etc.; the main elements of the node weight vectors in the third preliminary hierarchy are nursing-related characteristics, such as nurses, nursing staff, etc.; the main elements of the node weight vectors in the fourth preliminary hierarchy are patient-related characteristics such as medical treatment behavior, medical history, etc.; the main elements of the node weight vectors in the fifth preliminary hierarchy are characteristics related to medical insurance, claims, etc.

[0039] Traverse each standard user information, calculate its distance from each preliminary hierarchical user group, and classify it into the nearest preliminary hierarchical user group (used to adjust outliers), that is, the hierarchical user group division is completed.

[0040] Furthermore, the method for obtaining the user subgroup includes: Initially define the scale N3 of the clustering population and generate an initial clustering population. Define N3 cuckoos in the clustering population, and each cuckoo corresponds to a candidate clustering scheme. Encode each cuckoo (candidate clustering scheme) with an integer string, that is, N3 candidate clustering schemes.

[0041] Define the evolution function of the clustering population ; Among them, and are clustering weight parameters, which are adjusted according to actual needs; is the compactness function of clustering, is the and separation function between clusters, is the density penalty term of clustering, which is used to prevent clustering from being too dense.

[0042] ; Among them, is the user index belonging to the cluster, is the low-dimensional user feature vector of the user. Specifically, for each user, their features are extracted to construct a feature vector. The features include the user's personal information (such as age, gender, etc.), behavioral information (such as medical records, medication records, etc.), and social information (such as friend relationships, etc.). Using an unsupervised or self-supervised method, the feature vector is mapped to a low-dimensional embedding space to obtain the low-dimensional user feature vector; is the centroid embedding vector of the cluster, which refers to the vector representation of the center point of the cluster in the low-dimensional embedding space; is the transpose of, is the covariance matrix of the cluster, which describes the shape and direction of the data distribution within the cluster; is the smoothing bandwidth parameter, which controls the smoothness of the function; the optimal parameter combination is obtained through fitting.

[0043] Separation function ; Among them, and are separation weight parameters, is the low-dimensional user feature vector of the user, and the user belongs to the cluster, is the deep neural network model used to output the similarity score between the user and the user, are the parameters of the deep neural network model. The input of the deep neural network model is the concatenation of two embedding vectors, and the output is a scalar score. Specifically, during the training process, the value of the separation function is used as the supervision signal to minimize the similarity within the cluster and maximize the separation between clusters, learning the optimal similarity calculation model to improve the discriminative ability of the separation; For the user relative to the cluster entropy.

[0044] Entropy ; ; where and are normal constant parameters, is the total number of all users, is the dispersion term weight parameter, is the cluster scale (number of users).

[0045] Calculate the value of the evolutionary function for each cuckoo (candidate clustering scheme), denoted as the evolutionary value. According to the magnitude of the evolutionary value, sort each cuckoo in descending order, and select the top m4 cuckoos (optimal clustering scheme) as the cuckoos with extremely good voices.

[0046] Calculate the distance between each cuckoo and the cuckoo with extremely good voice. The formula for the distance is: ; where is the integer string encoding of the cuckoo the th element (the th bit of the encoding), is the integer string encoding of the cuckoo with extremely good voice the th element, is and the distance between (the distance between the cuckoo and the cuckoo with extremely good voice).

[0047] For the th cuckoo, among the cuckoos with extremely good voices, find the one closest to it and use it as the guiding cuckoo for the th cuckoo.

[0048] Guide the th cuckoo to move towards the position of the guiding cuckoo. The guiding formula is: ; where is the position of the th cuckoo after moving, is the position of the th cuckoo before moving, is the preset maximum number of iterations, is the current number of iterations, is the th cuckoo's guiding cuckoo position, is the weight coefficient of the group gravitational field; is the group gravitational field of the $i$-th cuckoo; ; where is the position of the $i$-th cuckoo before moving, is a positive real parameter that controls the attenuation rate of gravity; the group gravitational field represents the resultant force of the gravity exerted on a cuckoo by all other cuckoos. The magnitude of the gravity exerted by each cuckoo on it is inversely proportional to the distance between them, and the direction of the gravity points to the position of that cuckoo; the position after moving is to adjust the candidate clustering scheme to obtain a new cuckoo (a new clustering scheme).

[0049] Recalculate the evolutionary value of the new cuckoos, sort the new cuckoos in ascending order according to the magnitude of the evolutionary value, and eliminate the first $n_4$ new cuckoos; repeat until the maximum number of iterations is reached; calculate the evolutionary value of the new cuckoos obtained in the last iteration, and take the new clustering scheme corresponding to the new cuckoo with the largest evolutionary value as the optimal clustering scheme, that is, obtain $N$ user subgroups.

[0050] Furthermore, the calculation method of the access permission evaluation value includes: Define the access weight matrix of users for different types of medical data information , is an $m_5\times n_5$ matrix, where $m_5$ is the number of users and $n_5$ is the number of types of medical data information; the access weight matrix 's element represents the -th user's access weight vector for the -th type of medical data information; For the -th user subgroup, extract the access weight vectors corresponding to all users in the -th type of medical data information and take the average to obtain the -th user subgroup's average access demand for the -th type of medical data information ; then obtain the average access demand of each user subgroup for all medical data information; as the access permission evaluation value.

[0051] It should be noted that the access weight matrix is given according to the actual situation, such as assigning reasonable weight values according to factors such as user identity and access history; other constraints are introduced when calculating the average, such as the accessibility of certain data to specific user groups being 0, etc.

[0052] Furthermore, the construction method of the overall data permission management scheme includes: All medical data information in the Internet medical platform is represented as a weighted directed graph GH=(V, E), where V is the set of nodes (the set of all medical data information), each node v represents a kind of medical data information, E is the set of all edges. If medical data information a needs to access medical data information v, then there is a directed edge e'=(a, v) in the weighted directed graph.

[0053] Each directed edge e' has a non - negative weight w(e'), which represents the cost required to access medical data information l (fitted based on data acquisition, processing difficulty, data value, and access frequency).

[0054] Define the source point d as connected to all nodes without predecessors (data that does not need to access other medical data information), and define the sink point f as connected to all nodes without successors (medical data information not accessed by other data); both the source point and the sink point are virtual nodes.

[0055] For each user subgroup C, construct a directed cost network GH'=(V', E') from the source point to the sink point, where V' is the union of the set of all medical data information and the set of relay points, and the set of relay points is the set of all source points and sink points; E' is the set of comprehensive weight edges.

[0056] The definition of the set of comprehensive weight edges is as follows: For each v belonging to V and there exists an edge LP, then define the weight of edge LP as the v - th element of the access permission evaluation value; it should be noted that for each user subgroup C, its access permission evaluation value ev(C) is a vector of length |V|, and the v - th element represents the average access demand of user subgroup C for data v.

[0057] For each v belonging to V and there exists an edge LU, then define the weight of edge LU as 0; edge LP is the edge connecting v and the source point d, and edge LU is the edge connecting v and the sink point f.

[0058] Calculate the shortest path from d to f in the directed cost network GH' (using the weights of the edges as the basis for calculating the shortest path), and this shortest path corresponds to the best access plan for user subgroup C to access the required medical data information.

[0059] For each user subgroup C, obtain a best access plan path(C), and merge all path(C) to get the overall data permission management plan.

[0060] The overall data permission management plan contains information on which user subgroups can access each type of medical data on the Internet medical platform, thus constructing the overall data permission management plan.

[0061] In this embodiment, the automated, intelligent, and refined management of a large amount of medical data on the Internet medical platform is realized, thus greatly improving the security and controllability of data access. Specifically, first, based on the multi-dimensional analysis of user information, users are divided into multiple hierarchical user groups, realizing refined user classification management; then, within each hierarchical user group, clustering and evaluation models are used to further segment users, quantify the access requirements of each user subgroup for various types of medical data, and lay a foundation for formulating differentiated data access permission policies. Secondly, using graph theory algorithms, an overall data permission management scheme for the Internet medical platform is constructed. This scheme comprehensively considers various factors such as user attributes, data characteristics, and access requirements, and can formulate refined and reasonable access permission policies for different types of users and data, not only improving the efficiency and accuracy of permission management, but also having good generality and scalability, and playing a good role in maintaining medical data security and protecting patients' privacy rights and interests.

[0062] Embodiment 2 Please refer to Figure 2 As shown, for the parts not described in detail in this embodiment, refer to the description content of Embodiment 1. A data permission management system for Internet medical data based on multi-level user groups is provided, including: A data collection module for obtaining medical data information and user information of users in the Internet medical platform; A preliminary grouping module that divides users into n hierarchical user groups based on user information; A clustering and evaluation module for clustering users within each hierarchical user group using a search algorithm to obtain N user subgroups and calculating the access permission evaluation values of each user subgroup for medical data information; A management scheme regulation module that constructs an overall data permission management scheme for the Internet medical platform based on the access permission evaluation values of user subgroups. Each module is connected by wired and / or wireless means to achieve efficient production and data transmission between modules.

[0063] Embodiment 3 This embodiment publicly provides an electronic device, including a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the computer program, it realizes the operation mode of the above-provided method for managing Internet medical data permissions based on multi-level user groups.

[0064] Since the electronic device introduced in this embodiment is the electronic device adopted in implementing a method for managing Internet medical data permissions based on multi-level user groups in the embodiments of the present application, based on a method for managing Internet medical data permissions based on multi-level user groups introduced in the embodiments of the present application, those skilled in the art can understand the specific implementation manners of the electronic device in this embodiment and its various variations. Therefore, the specific implementation of how this electronic device implements the method in the embodiments of the present application will not be described in detail here. As long as those skilled in the art implement the electronic device adopted in a method for managing Internet medical data permissions based on multi-level user groups in the embodiments of the present application, it falls within the scope of protection of the present application.

[0065] The above formulas are all dimensionless and take their numerical values for calculation. The formulas are obtained by collecting a large amount of data for software simulation to obtain a formula that is closest to the actual situation. The preset parameters and threshold selection in the formulas are set by those skilled in the art according to the actual situation.

[0066] The above description is only a preferred implementation manner of the present invention. The protection scope of the present invention is not limited to the above embodiments. Any technical solutions falling within the concept of the present invention belong to the protection scope of the present invention. It should be noted that for ordinary technical users in the technical field, several improvements and refinements made without departing from the principle of the present invention should also be regarded as within the protection scope of the present invention.

Claims

1. A method for managing Internet medical data rights based on multi-level user groups, characterized in that: include: S1. Obtain medical data information and user information of users in the Internet medical platform; S2. Based on user information, users are divided into n levels of user groups; S3. For each hierarchical user group, a search algorithm is used to cluster the users in the hierarchical user group to obtain N user subgroups, and an evaluation value of the access rights of each user subgroup to the medical data information is calculated; S4. Based on the access rights evaluation values ​​of user subgroups, a directed graph algorithm is used to build an overall data rights management solution for the Internet medical platform.

2. According to claim 1, a method for managing Internet medical data rights based on multi-level user groups is characterized in that: The user information includes personal identity information, medical-related identity and authority level information; Medical data information includes patient medical history data, patient examination report data, patient treatment plan data, patient medication data, patient surgery data, doctor diagnosis data and medical insurance data.

3. The Internet medical data rights management method based on multi-level user groups according to claim 2 is characterized in that: The division method of the hierarchical user groups includes: preprocessing the user information to obtain standard user information; Transform the standard user information corresponding to each user into a sequential combination of vectors, namely, a combination vector; construct a grid according to the number of users and the number of hierarchical user groups n; Each node of the grid is assigned a node weight vector with the same dimension as the input data; the combination vector of each standard user information is traversed as a sample, and the similarity between the sample and each node weight vector is calculated; the node with the highest similarity to the sample is found as the best matching unit; Update the node weight vectors of the best matching unit and its neighborhood nodes to move them in the direction of the sample; when the learning rate parameter and the neighborhood function decay to the minimum value, the grid converges, and the grid is visualized to obtain the area composed of adjacent nodes, and the similarity of the node weight vectors inside it is calculated, and it is ensured that the calculated similarity is greater than the preset similarity threshold, that is, the division of the preliminary level is completed, and the preliminary level is marked according to the characteristics of the node weight vector, that is, the preliminary level user group; traverse each standard user information, calculate its distance from each preliminary level user group, and classify it into the nearest preliminary level user group, that is, the division of the level user group is completed.

4. The Internet medical data rights management method based on multi-level user groups according to claim 3 is characterized in that: The method of preprocessing the user information includes: For each non-quantitative data field in the user information, the frequency distribution of all its values ​​is counted and recorded as the value frequency; According to the size of the value frequency, the values ​​are sorted in descending order. For the values ​​with value frequency, they are temporarily sorted in alphabetical order to obtain a preliminary sorting result; all the values ​​of the corresponding field are traversed, and their sorting results are recorded to form an ordered list; in the ordered list, each value is traversed from the beginning to the end, and its serial number position in the list is determined and recorded; after sorting the serial numbers from small to large, the alphabetical arrangement result of the values ​​of the corresponding field is obtained; the sorted values ​​are recorded one by one to form a sorted value list; based on the sorted value list, the number of encoding dimensions is defined, and a coding correspondence table is established; Traverse all user information and convert the non-quantized data into corresponding encoding vectors according to the encoding correspondence table; that is, obtain the quantized part of the user information; The quantified partial user information is merged with the original quantified data field to form standard user information.

5. The Internet medical data rights management method based on multi-level user groups according to claim 4 is characterized in that: The calculation formula of the similarity is: ;in, For the sample, is the node weight vector, , and is the adjustment coefficient, and ; is the attenuation parameter, is the nonlinear similarity function between samples and node weight vectors; For sample and the node weight vector The similarity between ;in, is the bandwidth parameter, is the weight parameter, is the frequency parameter, For sample The scale parameter, for The scale parameter of .

6. The Internet medical data rights management method based on multi-level user groups according to claim 5 is characterized in that: The formula for updating the node weight vector of the best matching unit and its neighboring nodes is: ;in, is the learning rate parameter, which controls the update step size; To update the tuning parameters, is the neighborhood function, is the distance from the node to the best matching unit, The number of current traversal iterations; is the correction function; ;in, is the time-varying bandwidth parameter, is the time-varying adjustment parameter; is the time-varying central parameter; is the scale parameter; Time-varying bandwidth parameters ;in, is the preset initial bandwidth, is the bandwidth attenuation parameter, is the period adjustment parameter; is the angular frequency parameter, is the phase parameter; Correction function ;in, and is the adjustment weight parameter, and ; is the preference function, is the distance function; for and The distance between Preference function ;in, is the norm adjustment parameter; for The transpose of is a positive semidefinite matrix, for The Lp norm of is the norm degree; ;in, , and is the parameter that controls the shape of the hyperbolic tangent function, is the hyperbolic tangent function.

7. The Internet medical data rights management method based on multi-level user groups according to claim 6 is characterized in that: The method of obtaining the user subgroup includes: Initially define the size of the cluster population N3, and generate the initial cluster population. Define N3 cuckoos in the cluster population. Each cuckoo corresponds to a candidate clustering scheme. Encode each cuckoo with an integer string, that is, N3 candidate clustering schemes. Define the evolution function of the cluster population , calculate the value of the evolution function of each cuckoo, record it as the evolution value, sort each cuckoo in descending order according to the size of the evolution value, and select the first m4 cuckoos as the cuckoos with excellent voices; Calculate the distance between each cuckoo and the cuckoo with the best voice. The distance calculation formula is: ;in, Encode the integer string for Cuckoo No. elements, Encode the integer string for the cuckoo with the best voice No. elements, for and The distance between For cuckoo, find the one closest to it among the cuckoos with the best voice, and make it the first cuckoo's guide cuckoo; guide the first The cuckoo moves to the position of the guiding cuckoo, and the guiding formula is: ;in, For the The position of the cuckoo after it moves, For the The position of the cuckoo before it moves, is the preset maximum number of iterations, is the current iteration number, For the Cuckoo's Guide Cuckoo location, is the weight coefficient of the group gravitational field; For the The collective gravitational field of cuckoos; ;in, For the The position of the cuckoo before it moves, is a positive real number parameter; the position after moving is to adjust the candidate clustering scheme to obtain a new cuckoo; Recalculate the evolutionary value of the new cuckoo, and sort the new cuckoos in ascending order according to the size of the evolutionary value, and eliminate the first n4 new cuckoos; repeat until the maximum number of iterations is reached; calculate the evolutionary value of the new cuckoo obtained in the last iteration, and take the new clustering scheme corresponding to the new cuckoo with the largest evolutionary value as the optimal clustering scheme, that is, obtain N user subgroups.

8. The Internet medical data rights management method based on multi-level user groups according to claim 7 is characterized in that: The evolution function The formula is: ;in, and is the clustering weight parameter; For clustering The compactness function of For clustering and clustering The separation function between For clustering The density penalty term of ; ;in, Belong to cluster The user index of For users The low-dimensional user feature vector, For clustering The centroid embedding vector of ; for The transpose of For clustering The covariance matrix of is the smoothing bandwidth parameter; Separation function ; in, and is the separation weight parameter, For users The low-dimensional user feature vector of user Is a cluster Users within For output user and users A deep neural network model for similarity scores between are the parameters of the deep neural network model, For users Compared with clustering Entropy of ； ;in, and is a positive constant parameter, is the total number of all users, is the weight parameter of the discreteness term, For clustering scale.

9. The Internet medical data rights management method based on multi-level user groups according to claim 8 is characterized in that: The calculation method of the access rights evaluation value includes: Define the user access weight matrix for different types of medical data information , is an m5×n5 matrix, where m5 is the number of users and n5 is the number of types of medical data information; the access weight matrix Elements Indicates User to Access weight vector of medical data-like information; For user subgroups, extract all users in the The access weight vectors corresponding to the medical data information of this type are averaged to obtain the User subgroups Average access requirements for medical data information ; Then the average access requirement of each user subgroup to all medical data information is obtained as the access permission evaluation value.

10. The Internet medical data rights management method based on multi-level user groups according to claim 9 is characterized in that: The overall data rights management solution is constructed in the following ways: All medical data information in the Internet medical platform is represented as a weighted directed graph GH=(V, E), where V is the set of nodes and E is the set of all edges. If medical data information a needs to access medical data information v, there is a directed edge e'=(a, v) in the weighted directed graph; each directed edge e' has a non-negative weight w(e'); Define the source point d to be connected to all nodes without predecessors, and define the sink point f to be connected to all nodes without successors; For each user subgroup C, a directed cost network GH'=(V', E') from source to sink is constructed, where V' is the union of the set of all medical data information and the set of relay points, and the set of relay points is the set of all source points and sink points; E' is the set of comprehensive weighted edges; The definition of the comprehensive weight edge set is: For each v that belongs to V and there is an edge LP, the weight of the edge LP is defined as the vth element of the access right evaluation value; For each v that belongs to V and there is an edge LU, the weight of the edge LU is defined as 0; the edge LP is the edge connecting v and the source point d, and the edge LU is the edge connecting v and the sink point f; Calculate the shortest path from d to f in the directed cost network GH', which corresponds to the best access plan for the user subgroup C to access the required medical data information; For each user subgroup C, an optimal access plan path(C) is obtained, and all paths(C) are combined to obtain the overall data permission management plan.

Images