Medical equipment identification, connection and management method based on central control screen

Abstract

The invention belongs to the technical field of medical equipment management, and discloses a medical equipment identification, connection and management method based on a central control screen. Comprising the following steps: collecting equipment information and state information of medical equipment, and constructing a medical equipment information set; performing feature coding on the medical equipment information set to obtain a medical equipment feature vector set; based on the medical equipment feature vector set, performing split hierarchical clustering on the medical equipment to obtain n medical equipment category clusters; constructing a respective corresponding medical directed graph model for each medical equipment category cluster; applying an optimization algorithm to each medical directed graph model to obtain an optimal connection scheme of the medical equipment; according to the optimal connection scheme of the medical equipment, the connection and management of the medical equipment are realized through the central control screen, so that the equipment maintenance cost is reduced, and the continuity and reliability of medical services are ensured.

Description

Technical Field

[0001] This invention relates to the field of medical device management technology, and more specifically, to a method for identifying, connecting, and managing medical devices based on a central control screen. Background Technology

[0002] Patent application CN110232970A discloses a shared medical device and a shared medical device system and its usage method. It includes a medical device body, an identification mark set on the medical device body, and an information collection and transmission device. The information collection and transmission device is network-connected to the medical device body and is used to collect usage log data of the medical device body. The paper identification mark includes both paper and electronic identification marks. The paper identification mark is used for the borrowing and bed management of the shared medical device, while the electronic identification mark is used for the location management and security management of the medical device. The provided shared medical device and system allow users to view the location and status of the shared medical device via an app. Through the app or PC, users can perform functions such as device confirmation, reservation locking, bed management, device borrowing, usage management, disinfection management, device return, and record query. This system can be widely applied in the field of shared medical devices.

[0003] However, existing medical equipment management is inefficient and inaccurate, resulting in low resource utilization and frequent situations where equipment is idle or unavailable during emergencies. Secondly, the lack of intelligent connectivity between devices prevents them from working collaboratively, leading to low efficiency. Furthermore, existing systems have complex operating procedures and unfriendly user interfaces, making them prone to errors and reducing work efficiency. In addition, equipment allocation lacks flexibility, making it difficult to cope with emergencies or peak demand. In terms of equipment maintenance, reactive repair methods are often used, resulting in high failure rates and short lifespans, which not only increase maintenance costs but may also affect the continuity and reliability of medical services. Finally, the lack of comprehensive and accurate equipment usage data and analysis reports makes it difficult to conduct refined management and scientific decision-making, hindering the effective optimization of medical resource allocation.

[0004] In view of this, the present invention proposes a method for medical device identification, connection and management based on a central control screen to solve the above problems. Summary of the Invention

[0005] In order to overcome the above-mentioned defects of the prior art and to achieve the above objectives, the present invention provides the following technical solution: a method for medical device identification, connection and management based on a central control screen, comprising: S1, collecting device information and status information of medical devices, and constructing a medical device information set;

[0006] S2. Encode the medical equipment information set using features to obtain a set of medical equipment feature vectors;

[0007] S3. Based on the feature vector set of medical devices, perform hierarchical clustering of medical devices to obtain n medical device category clusters;

[0008] S4. Construct a corresponding directed graph model for each medical device category cluster;

[0009] S5. Apply an optimization algorithm to each medical directed graph model to obtain the optimal connection scheme for the medical devices; based on the optimal connection scheme for the medical devices, realize the connection and management of the medical devices through the central control screen.

[0010] Furthermore, the equipment information includes equipment type, equipment model, equipment interface type, equipment location information, and equipment service life;

[0011] The status information includes device operating status, device usage status, device connection status, and device data transmission status;

[0012] The methods for constructing the medical device information set include:

[0013] The collected equipment and status information of medical devices are cleaned and formatted to obtain standard equipment and status information.

[0014] Construct an empty preliminary medical equipment information set, iterate through the standard equipment information and standard status information, and add the standard equipment information and standard status information of each medical device as a dictionary object to the preliminary medical equipment information set; that is, complete the construction of the medical equipment information set.

[0015] Furthermore, the method for feature encoding the medical device information set includes:

[0016] The method of feature encoding the medical device information set includes:

[0017] For categorical features, one-hot encoding is used to convert them into M binary features; for numerical features, the original numerical values ​​are used directly; for text features, string similarity is used to convert them into numerical vector features, thus forming a set of medical device feature vectors.

[0018] The conversion methods for the numerical vector features include:

[0019] Extract strings with all distinct textual features from the medical device information set and construct a string set H;

[0020] Calculate the transformation distance between any string in the string set H and all preset metastrings, and construct an m1×M1 distance matrix D, where m1 is the number of strings in the string set H;

[0021] The element values ​​in the distance matrix D are normalized to obtain the normalized matrix D'. The distance similarity matrix HG = 1 - D' is constructed based on the normalized matrix D'.

[0022] For each text-type feature string, find its index i in the string set H. Then the feature vector of that string is the numerical vector feature corresponding to the text-type feature.

[0023] Furthermore, the calculation method for the conversion distance includes:

[0024] Construct a two-dimensional matrix DH, with the number of rows equal to the length of the first string plus 1, and the number of columns equal to the length of the second string plus 1. Initialize the values ​​of the first row and first column of DH; that is, assign the value 0 to DH[0,0] in the first row and first column; assign the value j to the remaining elements of the first row DH[j,0], indicating that it takes j insertion operations to convert an empty string into the first string; assign the value l to the remaining elements of the first column DH[0,l], indicating that it takes l deletion operations to convert the second string into an empty string.

[0025] Starting from DH[1,1], calculate the value of each element DH[j,l] in row-major order; the value of DH[i,j] is equal to the minimum of DH[j-1,l-1]+de, ​​DH[j-1,l]+1, and DH[j,l-1]+1; where de = 0 if the characters at this position are the same in the two strings, otherwise de = 1; DH[j-1,l]+1 represents the number of edits required to delete the j-th character of the first string; DH[j,l-1]+1 represents the number of edits required to insert the i-th character of the second string before the j-th character of the first string; DH[j-1,l-1]+de represents the number of edits required to replace the j-th character of the first string with the l-th character of the second string;

[0026] The two-dimensional matrix DH is continuously filled until DH[m3, n3] is calculated, where m3 and n3 are the lengths of the string and the metastring, respectively, and DH[m3, n3] is the conversion distance between the string and the metastring.

[0027] Furthermore, the preset method for the metastring includes:

[0028] Create a set of reference strings, cluster the reference strings in the set to obtain M1 clusters, and select the reference string with the highest average similarity to other reference strings in the cluster as the metastring.

[0029] The method of clustering reference strings within the reference string set includes:

[0030] For any two reference strings ri and rj in the set of reference strings, calculate the similarity between them. Where |ri∩rj| represents the number of elements in the intersection of ri and rj, and |ri∪rj| represents the number of elements in the union of ri and rj; calculate the similarity of all reference strings pairwise and construct an N×N similarity matrix S;

[0031] The similarity matrix S is propagated and updated to obtain the responsibility matrix R and availability matrix A of each reference string to other reference strings, and the updates are iteratively updated until convergence is achieved; the propagation responsibility matrix and propagation availability matrix are obtained; based on the propagation responsibility matrix and propagation availability matrix, M1 clusters are identified.

[0032] The methods for identifying M1 clusters include:

[0033] For each reference string ri, calculate re(ri) as the sum of R'(ri,rj) of all other reference strings rj; find the reference string ri* corresponding to the maximum value of re(ri) + ∑A'(ri,rj), and take ri* as the first cluster center; for each remaining reference string rj and all determined cluster centers ri*, calculate its maximum responsibility value with all current cluster centers; assign rj to the cluster of the cluster center with the maximum responsibility value; repeat until all reference strings have been assigned; that is, the identification of M1 clusters is completed.

[0034] Furthermore, the method for propagating and updating the similarity matrix S includes:

[0035] Initialize the responsibility matrix R and the availability matrix A as matrices containing all zeros, that is, the elements R(ri,rj) in the responsibility matrix R = 0; the elements A(ri,rj) in the availability matrix A = 0.

[0036] Update the elements in the responsibility matrix R according to their responsibilities; the formula for updating responsibilities is:

[0037] R(ri,rj) = r(ri,rj) - max(0,r(rj,ri)), where r(ri,rj) is the responsibility of reference string ri to reference string rj; r(rj,ri) is the responsibility of reference string rj to reference string ri.

[0038] Responsibility r(ri,rj) = Jd(ri,rj) - max(A(ri,rk) + A(rj,rk)); where A(ri,rk) represents the degree of advantage of choosing the reference string rk as a member of a cluster other than the cluster centered at ri; A(rj,rk) represents the degree of advantage of choosing the reference string rk as a member of a cluster other than the cluster centered at rj.

[0039] Update the availability of the elements in availability matrix A using the following formula:

[0040] A(ri,rj)=∑(max(0,r(rk,rj)));where r(rk,rj) is the responsibility of reference string rk to reference string rj; A(ri,rj) represents the availability value of selecting reference string rj as a member of other clusters besides ri;

[0041] Calculate the sum of responsibility and availability of cluster centers to reference string ri, chi(ri), and the sum of responsibility and availability of cluster centers to reference string rj, chi(rj);

[0042] chi(ri)=∑(R(ri,rj)+A(ri,rj));

[0043] chi(rj)=∑(R(rj,ri)+A(rj,ri)); where R(rj,ri) represents the responsibility value for assigning the reference string ri to the cluster centered at rj; A(rj,ri) is the availability value representing the selection of the reference string ri as a member of a cluster other than rj.

[0044] The responsibility matrix R and the availability matrix A are normalized based on chi(ri) and chi(rj), i.e., R′(ri,rj)=R(ri,rj)+chi(ri), A′(ri,rj)=A(rj,ri)+chi(rj); where R′(ri,rj) are the elements of the normalized responsibility matrix R, and A′(ri,rj) are the elements of the normalized availability matrix A.

[0045] After each iteration, the sum of all chi(ri) is calculated. If the change is less than the preset change threshold, the convergence has been achieved. Otherwise, the next iteration continues until the change is less than the preset change threshold, thus obtaining the propagation responsibility matrix corresponding to the responsibility matrix R and the propagation availability matrix corresponding to the availability matrix A.

[0046] Furthermore, the method of performing hierarchical clustering of medical devices includes:

[0047] The medical device feature vector set X is standardized to obtain a standardized data matrix Z; principal component analysis is performed on the standardized data matrix Z to obtain the principal component factors and their corresponding factor loading matrices.

[0048] Calculate the correlation coefficient matrix R of the standardized data Z, and perform eigenvalue decomposition on the correlation coefficient matrix R to obtain the eigenvalues ​​λ and the corresponding eigenvectors V. Based on the eigenvalues ​​λ, calculate the contribution rate of each principal component factor p. Where λ(p) is the eigenvalue corresponding to the principal component factor p;

[0049] If a preset threshold th is set, then the smallest u value that makes ∑(co(1:u))≥th is defined as the number of principal component factors; where co(1:u) represents the contribution rate of the first u principal component factors.

[0050] Extract the first u feature vectors and construct the factor loading matrix L; project the standardized data matrix Z onto a low-dimensional space composed of the first u principal component factors to obtain the dimensionality-reduced data representation Y = Z·L;

[0051] Perform core matrix decomposition on Y to obtain the core matrix SH and two projection matrices PH and QH, such that Y = PH·SH·QH'; where QH' is the transpose of QH;

[0052] Preliminary clustering is performed on the core matrix SH to obtain n5 initial clusters; the cluster mean vector y_mean of each initial cluster is calculated, and the cluster mean vector y_mean is projected back to the original space to obtain the cluster center point x_mean = y_mean·L′, where L′ is the transpose of the factor loading matrix L; the clusters are re-divided according to the distance of each point to the cluster center; this process is repeated until the cluster division is stable; the final n medical device category clusters are output.

[0053] Furthermore, the method for performing preliminary clustering on the core matrix SH includes:

[0054] Calculate the core similarity matrix WH of the core matrix SH; the elements in the core similarity matrix WH are WH(I,J);

[0055] Where SH(I,:) represents the row vector of the I-th data point in the core matrix SH; SH(I,:)′ is the transpose of SH(I,:); SH(J,:) represents the row vector of the J-th data point in the core matrix SH; norm(SH(I,:)) represents the magnitude of SH(I,:); norm(SH(J,:)) represents the magnitude of SH(J,:);

[0056] The co-occurrence similarity matrix CH is calculated based on the core similarity matrix WH. The element CH(z,v) in the co-occurrence similarity matrix CH is defined as ∑(WH(z,:)*WH(v,:)); where WH(z,:) is a row vector composed of the similarity values ​​between the z-th data point and all other data points; WH(v,:) is a row vector composed of the similarity values ​​between the v-th data point and all other data points; * represents element-wise multiplication.

[0057] By integrating WH and CH, we obtain the integrated similarity matrix SP = δ·WH + (1-δ)·CH; where δ is the weight parameter.

[0058] Based on SP, calculate the normalized Laplacian matrix LH and the first kp smallest non-zero eigenvectors VP of LH; perform k-means clustering on the row vectors of VP to obtain the initial cluster.

[0059] Furthermore, the construction method of the medical directed graph model includes:

[0060] For each medical device category cluster, construct an empty directed graph, traverse all medical devices within the cluster, and add each medical device as a node to the directed graph;

[0061] Obtain the interface types and number supported by each medical device; define constraints including for any two medical devices r and b; if the number of interfaces they support together is greater than 0, then connect; otherwise, do not connect.

[0062] Extract the location information of each medical device and represent it using hierarchical coding; for any two medical devices, calculate their weighted distance according to the different levels of the hierarchical coding.

[0063] Set a location distance threshold thre; if the weighted distance between locations is less than or equal to thre, then connect them.

[0064] If medical device r is in use and b is idle, then r is not allowed to connect to b. Furthermore, if r is already connected to another medical device, then it is not allowed to connect to b again. The combined constraints determine whether any two medical devices r and b are connected or not.

[0065] For a pair of interconnected medical devices (r, b), add a directed edge to the directed graph; assign a weight value to each directed edge;

[0066] The formula for assigning the weight values ​​is as follows:

[0067] E(r,b)=a1×T(r,b)+a2×Q(r,b)+a3×F(r,b)+a4×P(r,b); where E(r,b) is the weight value of the edge from medical device r to medical device b, a1 is the type weight coefficient, a2 is the usage state weight coefficient, a3 is the connection state weight coefficient, and a4 is the location distance weight coefficient;

[0068] T(r,b) is the type priority function between medical device r and medical device b;

[0069] Q(r,b) is the priority function for the usage status between medical device r and medical device b. If medical device r is idle and medical device b is in use, then Q(r,b) = 1; otherwise, Q(r,b) = 0.

[0070] F(r,b) is the connection state priority function between medical device r and medical device b; if medical device r is not connected and medical device b is connected, then F(r,b) = 1; otherwise, F(r,b) = 0.

[0071] P(r,b) is the positional distance priority function between medical device r and medical device b;

[0072] Where d0 is the distance scale parameter, and dist(r,b) is the actual distance between medical device r and medical device b;

[0073] Repeat this process until all medical devices within the medical device category cluster have been added to the directed graph, resulting in a medical directed graph model.

[0074] Furthermore, the optimal connection scheme is obtained through the following methods:

[0075] Initialize the parameters of the optimization algorithm, including the number of bats in the colony m6, the maximum number of iterations mer, the sonar pulse emission rate rop, the lower limit of sonar frequency fmin, and the upper limit of sonar frequency fmax;

[0076] Randomly generate n initial solutions, each of which is an n×n adjacency matrix; treat each solution as a bat;

[0077] Define the fitness function of the optimization algorithm;

[0078] Where toe is the sum of the weights of all directed edges in the connection scheme; prs is the device importance priority score; cts is the connectivity score; α1, β1 and γ1 are the corresponding weight coefficients;

[0079] Where w1(r,b) is the weight of the directed edge between medical device r and medical device b, and tyo(r,b) is the type priority function from medical device r to medical device b.

[0080] Where nu is the number of medical devices successfully connected under the connection scheme, and tod is the total number of all medical devices;

[0081] Update the position and speed of each bat:

[0082] v_h (t+1) =v_h (t) +(x_h (t) -x_best)·f_h;x_h (t+1) =x_h (t) +v_h (t+1) ;

[0083] Among them, v_h (t) x_h is the speed of the h-th bat in the t-th iteration. (t) v_h is the position of the h-th bat in the t-th iteration. (t+1) Let f_h be the velocity of the h-th bat in the (t+1)-th iteration; x_best is the position of the current optimal solution; f_h is the sonar frequency of the h-th bat.

[0084] A subset of bats is randomly selected based on the sonar pulse emission rate rop, and new connection schemes are generated through random perturbation.

[0085] Calculate the fitness function value for each solution, denoted as the fitness function value, compare the fitness function values ​​of all new solutions with those of the old solutions, and select the one with the largest fitness function value as x_best;

[0086] If the maximum number of iterations mer has been reached, then output the current x_best as the final optimal connection scheme.

[0087] The technical effects and advantages of the medical device identification, connection and management method based on a central control screen of the present invention are as follows:

[0088] This invention improves the efficiency and accuracy of medical equipment management, enabling healthcare professionals to better understand and utilize existing equipment resources. Secondly, it achieves intelligent connectivity between devices, significantly increasing equipment utilization efficiency and reducing idle time, thereby optimizing medical resource allocation. Thirdly, it simplifies operating procedures and provides an intuitive and user-friendly interface, allowing medical staff to manage and use various medical devices more conveniently, effectively reducing operational errors and improving work efficiency. Furthermore, optimal connectivity facilitates flexible equipment allocation, improving equipment applicability and flexibility. Simultaneously, it prevents potential malfunctions and extends equipment lifespan. This not only reduces equipment maintenance costs but also ensures the continuity and reliability of medical services. It provides strong support for refined medical management and scientific decision-making, contributing to improved medical quality. Attached Figure Description

[0089] Figure 1 This is a schematic diagram of a medical device identification, connection, and management method based on a central control screen according to the present invention;

[0090] Figure 2 This is a schematic diagram of a medical device identification, connection and management system based on a central control screen according to the present invention. Detailed Implementation

[0091] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0092] Example 1

[0093] Please see Figure 1 As shown in this embodiment, a method for identifying, connecting, and managing medical devices based on a central control screen includes:

[0094] S1. Collect equipment information and status information of medical devices, and construct a medical device information set;

[0095] S2. Encode the medical equipment information set using features to obtain a set of medical equipment feature vectors;

[0096] S3. Based on the feature vector set of medical devices, perform hierarchical clustering of medical devices to obtain n medical device category clusters;

[0097] S4. Construct a corresponding directed graph model for each medical device category cluster;

[0098] S5. Apply an optimization algorithm to each medical directed graph model to obtain the optimal connection scheme for the medical devices; based on the optimal connection scheme for the medical devices, realize the connection and management of the medical devices through the central control screen.

[0099] Furthermore, the equipment information includes equipment type (such as electrocardiograph, ventilator, monitor, etc.), equipment model, equipment interface type (such as USB, serial port, network interface, etc.), equipment location information (such as the ward, department, etc.) and equipment service life.

[0100] The status information includes device operating status (such as normal operation, fault, idle, etc.), device usage status (such as in use, idle, etc.), device connection status (such as connected to other devices, not connected), and device data transmission status (such as normal data transmission, packet loss, etc.).

[0101] Furthermore, the method for constructing the medical device information set includes:

[0102] The collected medical equipment information and status information are cleaned and formatted to obtain standard equipment information and standard status information, so as to ensure the integrity and consistency of the data.

[0103] Specifically, it iterates through the acquired device and status information, checks if each field has a value, and if a required field is empty, it is marked as invalid data and discarded or the default value is used.

[0104] Use regular expressions or other string processing functions to remove invalid characters, such as control characters and special symbols, from device and status information.

[0105] Based on the expected data type of each field, the retrieved data is converted to a different type, such as converting a string to an integer, floating-point number, or date / time type.

[0106] Standardize the data format of certain fields, such as converting date and time to the standard ISO format, and replacing special characters in device model numbers with standard formats.

[0107] If some fields contain multiple values, they need to be split into multiple separate fields. For example, the "Device Interface Type" field can be split into multiple fields such as USB interface, serial port interface, and network interface.

[0108] For the values ​​of certain fields, use standardized codes or abbreviations for replacement, such as replacing "normal operation" with "0" and "fault" with "1" in the device operating status.

[0109] Construct an empty preliminary medical device information set to store the processed standard device information and standard status information;

[0110] Iterate through the standard equipment information and standard status information, and add the standard equipment information and standard status information of each medical device as a dictionary (key-value pair) object to the initial medical device information set; that is, complete the construction of the medical device information set.

[0111] Furthermore, the method for feature encoding the medical device information set includes:

[0112] For categorical features (such as device type, interface type, etc.), one-hot encoding is used to convert them into M binary features; for numerical features (such as years of use, etc.), the original numerical values ​​are used directly; for textual features (such as device model, etc.), they are converted into numerical vector features, thus forming a set of medical device feature vectors.

[0113] Specifically, retrieve all the different text-based strings from the medical device information set and construct a string set H.

[0114] Calculate the transformation distance between any string in the string set H and all preset metastrings, and construct an m1×M1 distance matrix D, where m1 is the number of strings in the string set H;

[0115] The element values ​​in the distance matrix D are normalized to obtain the normalized matrix D'. The distance similarity matrix HG = 1 - D' is then constructed based on the normalized matrix D'.

[0116] For each text-type feature string, find its index i in the string set H. Then the feature vector of that string (the i-th row of the distance similarity matrix HG) is the numerical vector feature corresponding to the text-type feature.

[0117] The calculation method for the conversion distance includes:

[0118] Construct a two-dimensional matrix DH with the number of rows equal to the length of the first string plus 1 and the number of columns equal to the length of the second string plus 1. Initialize the values ​​of the first row and first column of DH; that is, assign the value 0 to DH[0,0] in the first row and first column; assign the value j to the remaining elements of the first row DH[j,0], indicating that it takes j insertion operations to convert an empty string into the first string; assign the value l to the remaining elements of the first column DH[0,l], indicating that it takes l deletion operations to convert the second string into an empty string.

[0119] Starting from DH[1,1], calculate the value of each element DH[j,l] in row-major order; the value of DH[i,j] is equal to the minimum of DH[j-1,l-1]+de, ​​DH[j-1,l]+1, and DH[j,l-1]+1; where de = 0 if the characters at this position are the same in the two strings, otherwise de = 1; DH[j-1,l]+1 represents the number of edits required to delete the j-th character of the first string; DH[j,l-1]+1 represents the number of edits required to insert the i-th character of the second string before the j-th character of the first string; DH[j-1,l-1]+de represents the number of edits required to replace the j-th character of the first string with the l-th character of the second string.

[0120] The two-dimensional matrix DH is continuously filled until DH[m3, n3] is calculated, where m3 and n3 are the lengths of the string and the metastring, respectively, and DH[m3, n3] is the conversion distance between the string and the metastring.

[0121] Create a reference string set. Specifically, extract a set of representative strings from domain knowledge as the reference string set; for example, for medical devices, extract common strings from fields such as device model, brand, and model abbreviation as the reference string set.

[0122] Extract all possible reference strings from the domain knowledge to construct an initial string set; perform deduplication and preprocessing on the initial string set, including removing empty strings, strings that are too short, and strings that contain only numbers or special characters.

[0123] From the filtered string set, a certain number of strings are randomly selected or extracted according to a certain strategy to form the final reference string set.

[0124] Cluster the reference strings in the reference string set to obtain M1 clusters, and select the reference string with the highest average similarity (calculated using a string kernel function) with other reference strings in the cluster as the meta string.

[0125] Methods for clustering reference strings within a reference string set include:

[0126] For any two reference strings ri and rj in the set of reference strings, calculate the similarity between them. Where |ri∩rj| represents the number of elements in the intersection of ri and rj, and |ri∪rj| represents the number of elements in the union of ri and rj; calculate the similarity of all reference strings pairwise and construct an N×N similarity matrix S.

[0127] The similarity matrix S is propagated and updated. The responsibility matrix R and availability matrix A of each reference string to other reference strings are calculated, and the updates are iteratively updated until convergence.

[0128] Specifically, the responsibility matrix R and the availability matrix A are initialized as all zero matrices, that is, the element R(ri,rj) in the responsibility matrix R is 0, which means that the responsibility value of assigning the reference string rj to the cluster centered at ri is 0; the element A(ri,rj) in the availability matrix A is 0, which means that the availability value of selecting the reference string rj as a member of a cluster other than the ri cluster is 0.

[0129] Update the elements in the responsibility matrix R according to their responsibilities; the formula for updating responsibilities is:

[0130] R(ri,rj) = r(ri,rj) - max(0,r(rj,ri)), where r(ri,rj) is the responsibility of reference string ri to reference string rj (interpreted as the similarity of adding rj to the cluster centered on ri minus the total evidence that ri and rj are assigned to other optimal clusters when not assigning ri and rj to the same cluster). The larger the value, the more advantageous it is to add rj to the cluster centered on ri; r(rj,ri) is the responsibility of reference string rj to reference string ri.

[0131] Responsibility r(ri,rj) = Jd(ri,rj) - max(A(ri,rk) + A(rj,rk)); where A(ri,rk) represents the advantage of choosing the reference string rk as a member of a cluster other than the cluster centered at ri (an element in the availability matrix A); A(rj,rk) represents the advantage of choosing the reference string rk as a member of a cluster other than the cluster centered at rj.

[0132] Update the availability of the elements in availability matrix A using the following formula:

[0133] A(ri,rj)=∑(max(0,r(rk,rj)));where r(rk,rj) is the responsibility of reference string rk to reference string rj;

[0134] A(ri,rj) represents the availability value of selecting reference string rj as a member of a cluster other than ri; it is only added to the summation if r(rk,rj) is greater than 0, ensuring that only favorable evidence is considered.

[0135] Calculate the sum of responsibility and availability of cluster centers to reference string ri, chi(ri), and the sum of responsibility and availability of cluster centers to reference string rj, chi(rj);

[0136] chi(ri)=∑(R(ri,rj)+A(ri,rj));

[0137] chi(rj)=∑(R(rj,ri)+A(rj,ri)); where R(rj,ri) is an element in the responsibility matrix R, representing the responsibility value for assigning the reference string ri to the cluster centered at rj; A(rj,ri) is an element in the availability matrix A, representing the availability value for selecting the reference string ri as a member of other clusters besides rj.

[0138] The responsibility matrix R and the availability matrix A are normalized based on chi(ri) and chi(rj), i.e., R′(ri,rj)=R(ri,rj)+chi(ri), A′(ri,rj)=A(rj,ri)+chi(rj); where R′(ri,rj) are the elements of the normalized responsibility matrix R, and A′(ri,rj) are the elements of the normalized availability matrix A.

[0139] After each iteration, the sum of all chi(ri) is calculated. If the change is less than the preset change threshold, the convergence has been achieved. Otherwise, the next iteration continues until the change is less than the preset change threshold. The propagation responsibility matrix corresponding to the responsibility matrix R and the propagation availability matrix corresponding to the availability matrix A are obtained.

[0140] Through continuous iteration, the values ​​of the responsibility matrix R and the availability matrix A will gradually converge; eventually, the value of R′(ri,rj) is larger, indicating that ri and rj should be assigned to the same cluster; the value of A′(ri,rj) is larger, indicating that rj should be assigned to the cluster centered on ri; based on A′(ri,rj) and R′(ri,rj), M1 clusters are identified.

[0141] The methods for identifying M1 clusters include:

[0142] For each reference string ri, calculate re(ri) as the sum of R'(ri,rj) of all other reference strings rj; find the reference string ri* corresponding to the maximum value of re(ri) + ∑A'(ri,rj), and take ri* as the first cluster center; for each remaining reference string rj and all determined cluster centers ri*, calculate its maximum responsibility value with all current cluster centers; assign rj to the cluster of the cluster center with the maximum responsibility value; repeat until all reference strings have been assigned; that is, the identification of M1 clusters is completed.

[0143] Furthermore, methods for hierarchical clustering of medical devices include:

[0144] The set of feature vectors X of medical devices is standardized to obtain a standardized data matrix Z; the standardization method can be Z-Score standardization or Min-Max standardization, etc.

[0145] Principal component analysis was performed on the standardized data matrix Z to obtain the principal component factors and their corresponding factor loading matrices.

[0146] Specifically, the correlation coefficient matrix R of the standardized data Z is calculated, and the correlation coefficient matrix R is decomposed into eigenvalues ​​to obtain eigenvalues ​​λ and their corresponding eigenvectors V. Based on the eigenvalues ​​λ, the contribution rate of each principal component factor p is calculated. Where λ(p) is the eigenvalue corresponding to the principal component factor p.

[0147] The correlation coefficient matrix R can be calculated in the following ways:

[0148] Calculate the covariance matrix Cov of the standardized data Z, and then calculate the correlation coefficient matrix based on the covariance matrix Cov. Here, DB is a diagonal matrix whose diagonal elements are the diagonal elements of Cov.

[0149] If a preset threshold th is set, then the smallest u value that makes ∑(co(1:u))≥th is defined as the number of principal component factors; usually th is set to 0.8-0.9; where co(1:u) represents the contribution rate of the first u principal component factors.

[0150] Extract the first u feature vectors and construct the factor loading matrix L; project the standardized data matrix Z onto a low-dimensional space composed of the first u principal component factors to obtain the dimensionality-reduced data representation Y = Z·L.

[0151] Perform core matrix decomposition on Y to obtain the core matrix SH and two projection matrices PH and QH, such that Y = PH·SH·QH', where QH' is the transpose of QH.

[0152] Preliminary clustering is performed on the core matrix SH to obtain n5 initial clusters; specifically, the core similarity matrix WH of the core matrix SH is calculated; the elements in the core similarity matrix WH are WH(I,J);

[0153] Wherein, SH(I,:) represents the row vector of the i-th data point (the data point within SH) in the core matrix SH, which is a 1×c row vector, where c is the number of columns in the core matrix SH; SH(I,:)' is the transpose of SH(I,:), which is a c×1 column vector; SH(J,:) represents the row vector of the j-th data point (the data point within SH) in the core matrix SH, which is also a 1×c row vector; norm(SH(I,:)) represents the 2-norm of SH(I,:), i.e., the magnitude of SH(I,:); norm(SH(J,:)) represents the 2-norm of SH(J,:), i.e., the magnitude of SH(J,:).

[0154] The co-occurrence similarity matrix CH is calculated based on the core similarity matrix WH. The element CH(z,v) in the co-occurrence similarity matrix CH is given by ∑(WH(z,:)*WH(v,:)). Here, WH(z,:) is the row vector composed of the similarity values ​​between the z-th data point (i.e., the row vector of the core matrix S) and all other data points; WH(v,:) is the row vector composed of the similarity values ​​between the v-th data point and all other data points; * represents element-wise multiplication, reflecting the sum of the similarity products of data point z and all other data points.

[0155] By integrating WH and CH, we obtain the integrated similarity matrix SP = δ·WH + (1-δ)·CH; where δ is the weight parameter.

[0156] Based on SP, calculate the normalized Laplacian matrix LH and the first kp smallest non-zero eigenvectors VP of LH; perform k-means clustering on the row vectors of VP to obtain the initial cluster.

[0157] Calculate the cluster mean vector y_mean for each initial cluster, and project the cluster mean vector y_mean back into the original space to obtain the cluster center point x_mean = y_mean·L′, where L′ is the transpose of the factor loading matrix L; re-divide the clusters according to the distance of each point to the cluster center; repeat until the cluster division is stable; output the final n medical device category clusters.

[0158] Furthermore, the construction method of the medical directed graph model includes:

[0159] For each medical device category cluster, construct an empty directed graph, traverse all medical devices within the cluster, and add each medical device as a node to the directed graph.

[0160] Obtain the interface types and quantities supported by each medical device from the device information; for example, {"USB": 2, "Ethernet": 1} indicates that the medical device has 2 USB ports and 1 Ethernet port;

[0161] The defined constraints include: for any two medical devices r and b, if the number of commonly supported interfaces is greater than 0, then a connection is established; otherwise, no connection is established.

[0162] Extract the location information of each medical device from the equipment information, such as the ward number and department; use hierarchical coding to represent it, such as "2A03" representing room 3 in area A of building 2;

[0163] For any two medical devices, calculate their location-weighted distance according to different levels of hierarchical coding.

[0164] Specifically, the location codes of medical devices r and b are split into multiple levels. For example, 2A03 can be split into three levels: 2 (floor level), A (area level), and 03 (room number level).

[0165] At each level, calculate the distance between r and b:

[0166] Floor level: If r and b are located on different floors, the distance can be set to the distance between floors, such as 100 meters; otherwise, the distance is 0.

[0167] Area level: If r and b are located on the same floor but in different areas, the distance can be set to the distance between areas, such as 50 meters; otherwise, the distance is 0.

[0168] Room number level: The absolute value of the difference between the numerical values ​​of room numbers r and b, multiplied by the standard distance of each room, such as 5 meters.

[0169] Assign a weight coefficient to each distance level to reflect the importance of that level of distance; for example: floor level has a weight of 0.6, area level has a weight of 0.3, and room number level has a weight of 0.1.

[0170] Multiply the distances at each level by their corresponding weight coefficients, and then sum them to obtain the position-weighted distance between r and b.

[0171] Set a location distance threshold thre, such as 50 meters. If the weighted distance between the two medical devices is less than or equal to thre, then the two devices are within the allowed connection range and a connection is established.

[0172] If medical device r is in use and b is idle, then r is not allowed to connect to b. Furthermore, if r is already connected to another medical device, then it is not allowed to connect to b again. The combined constraints determine whether any two medical devices r and b are connected or not.

[0173] For a pair of interconnected medical devices (r, b), add a directed edge to the directed graph to represent the connection from r to b; assign a weight value to each directed edge to represent the priority of the connection.

[0174] The formula for assigning the weight values ​​is as follows:

[0175] E(r,b) = a1×T(r,b) + a2×Q(r,b) + a3×F(r,b) + a4×P(r,b); where E(r,b) is the weight value of the edge from medical device r to medical device b, a1 is the type weight coefficient, a2 is the usage state weight coefficient, a3 is the connection state weight coefficient, and a4 is the location distance weight coefficient. The values ​​of the above weight coefficients are in the range of [0, 1]. These weights can be adjusted according to actual needs to reflect the importance of different factors.

[0176] T(r,b) is the type priority function between medical device r and medical device b. The type priority function is obtained by setting a priority mapping table for device types. For example, the monitor has the highest priority with a score of 5; the ventilator is next with a score of 4; the electrocardiograph has a score of 3; the infusion pump has a score of 2; and other devices have a score of 1. The priority mapping table reflects the subjective judgment of the importance of different device types.

[0177] Q(r,b) is the priority function for the usage status between medical device r and medical device b. If medical device r is idle and medical device b is in use, then Q(r,b) = 1; otherwise, Q(r,b) = 0. Idle devices have higher priority than devices in use.

[0178] F(r,b) is the connection priority function between medical device r and medical device b; if medical device r is not connected and medical device b is connected, then F(r,b) = 1; otherwise, F(r,b) = 0; the unconnected device has a higher priority than the connected device.

[0179] P(r,b) is the positional distance priority function between medical device r and medical device b;

[0180] Where d0 is a distance scale parameter, and dist(r,b) is the actual distance between medical device r and medical device b.

[0181] This process is repeated until all medical devices within the cluster are added to the directed graph and their connections and priority weights are determined. Finally, a directed weighted graph model corresponding to a medical device category cluster is obtained, which is the medical directed graph model.

[0182] Furthermore, the optimal connection scheme is obtained through the following methods:

[0183] Initialize the parameters of the optimization algorithm, including the number of bats in the colony m6, the maximum number of iterations mer, the sonar pulse emission rate rop, the lower limit of sonar frequency fmin, and the upper limit of sonar frequency fmax.

[0184] We randomly generate n initial solutions (individual bats), which are n different connection schemes. Each initial solution is an n×n adjacency matrix, representing the connection relationship between medical devices. Each solution corresponds to a bat.

[0185] Define the fitness function of the optimization algorithm;

[0186] Where toe is the sum of the weights of all directed edges under a given connectivity scheme; prs is the device importance priority score; cts is the connectivity score, which measures the completeness of the connectivity scheme; α1, β1 and γ1 are the corresponding weight coefficients.

[0187] Here, w1(r,b) is the weight of the directed edge between medical device r and medical device b, and tyo(r,b) is the type priority function from medical device r to medical device b, with a larger value indicating a higher priority.

[0188] Where nu is the number of medical devices successfully connected under this connection scheme, and tod is the total number of all medical devices.

[0189] Update the position (connection scheme) and velocity of each bat (i.e., each solution):

[0190] v_h (t+1) =v_h (t) +(x_h (t) -x_best)·f_h;x_h (t+1) =x_h (t) +v_h (t+1) ;

[0191] Among them, v_h (t) x_h is the velocity (rate of change of connection scheme) of the h-th bat in the t-th iteration. (t) v_h is the position (connection scheme) of the h-th bat in the t-th iteration. (t+1) Let f_h be the speed of the h-th bat in the (t+1)th iteration; x_best is the position of the current optimal solution; f_h is the sonar frequency of the h-th bat, which varies randomly between fmin and fmax.

[0192] A subset of bats is randomly selected based on the sonar pulse emission rate rop, and a new connection scheme (solution) is generated through random perturbation.

[0193] Calculate the fitness function value for each solution, denoted as the fitness function value, compare the fitness function values ​​of all new solutions with those of the old solutions, and select the one with the largest fitness function value as x_best;

[0194] If the maximum number of iterations mer has been reached, output the current x_best as the final optimal connection scheme; otherwise, continue iterating until the maximum number of iterations mer is reached.

[0195] Throughout the iterative optimization process, the position and velocity of each individual bat in the bat swarm (connection scheme) are continuously updated, and global and local searches are performed in the solution space to gradually approach the optimal solution.

[0196] The aforementioned method of connecting and managing medical devices via a central control screen specifically involves the central control screen displaying all connectable medical devices and their status information, visually displaying the current connection topology of the medical devices, and providing functions for adding, modifying, and deleting device connections.

[0197] Secondly, it displays the operating data and alarm information of connected devices in real time and communicates with medical devices to obtain device information and status information, and issues control commands to the devices when necessary.

[0198] The central control screen periodically scans the local area network or other known communication channels to automatically discover and identify new medical devices connected to the system, and adds the discovered devices to the device list.

[0199] Based on the previously calculated optimal connection scheme, the central control screen automatically or manually establishes a connection between medical devices; after the connection is established, the central control screen monitors the connection status and automatically attempts to reconnect if the connection is abnormal.

[0200] This embodiment improves the efficiency and accuracy of medical equipment management, enabling healthcare professionals to better understand and utilize existing equipment resources. Secondly, it achieves intelligent connectivity between devices, significantly increasing equipment utilization efficiency and reducing idle time, thereby optimizing medical resource allocation. Thirdly, it simplifies operating procedures and provides an intuitive and user-friendly interface, allowing medical staff to manage and use various medical devices more conveniently, effectively reducing operational errors and improving work efficiency. Furthermore, optimal connectivity facilitates flexible equipment allocation, improving equipment applicability and flexibility. Simultaneously, it prevents potential malfunctions and extends equipment lifespan. This not only reduces equipment maintenance costs but also ensures the continuity and reliability of medical services, providing strong support for refined medical management and scientific decision-making, and contributing to improved medical quality.

[0201] Example 2

[0202] Please see Figure 2 As shown, for parts not described in detail in this embodiment, please refer to the description in Embodiment 1. A medical device identification, connection, and management system based on a central control screen is provided, including:

[0203] The data acquisition module is used to collect equipment information and status information of medical devices and to build a medical device information set;

[0204] The feature extraction module is used to encode the medical device information set into features to obtain a set of medical device feature vectors;

[0205] The clustering module, based on the feature vector set of medical devices, performs hierarchical clustering of medical devices to obtain n medical device category clusters;

[0206] The model building module is used to construct a corresponding directed graph model for each medical device category cluster;

[0207] The connection management module applies optimization algorithms to each medical directed graph model to obtain the optimal connection scheme for the medical devices; based on the optimal connection scheme, the connection and management of the medical devices are realized through the central control screen; the various modules are connected to each other via wired and / or wireless means to realize data transmission between modules.

[0208] Example 3

[0209] This embodiment discloses an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the operation mode of the above-described method for identifying, connecting, and managing medical devices based on a central control screen.

[0210] Since the electronic device described in this embodiment is used to implement the medical device identification, connection, and management method based on a central control screen as described in this application embodiment, those skilled in the art can understand the specific implementation methods and various variations of the electronic device in this embodiment based on the medical device identification, connection, and management method based on a central control screen as described in this application embodiment. Therefore, how the electronic device implements the method in this application embodiment will not be described in detail here. Any electronic device used by those skilled in the art to implement the medical device identification, connection, and management method based on a central control screen as described in this application embodiment falls within the scope of protection of this application.

[0211] The above formulas are all dimensionless calculations. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters and thresholds in the formulas are set by those skilled in the art according to the actual situation.

[0212] The above description is merely a preferred embodiment of the present invention, and the scope of protection of the present invention is not limited to the above embodiments. All technical solutions falling within the scope of the present invention's concept are within the scope of protection of the present invention. It should be noted that for users of ordinary technical skills, any improvements and modifications made without departing from the principles of the present invention should also be considered within the scope of protection of the present invention.

Claims

1. A method for identifying, connecting, and managing medical devices based on a central control screen, characterized in that, include: S1. Collect equipment information and status information of medical devices, and construct a medical device information set; S2. Encode the medical equipment information set using features to obtain a set of medical equipment feature vectors; S3. Based on the feature vector set of medical devices, perform hierarchical clustering of medical devices to obtain n medical device category clusters; S4. Construct a corresponding directed graph model for each medical device category cluster; S5. Apply an optimization algorithm to each medical directed graph model to obtain the optimal connection scheme for the medical devices; based on the optimal connection scheme for the medical devices, realize the connection and management of the medical devices through the central control screen.

2. The method for identifying, connecting, and managing medical devices based on a central control screen according to claim 1, characterized in that, The equipment information includes equipment type, equipment model, equipment interface type, equipment location information, and equipment service life; The status information includes device operating status, device usage status, device connection status, and device data transmission status; The methods for constructing the medical device information set include: The collected equipment and status information of medical devices are cleaned and formatted to obtain standard equipment and status information. Construct an empty preliminary medical equipment information set, iterate through the standard equipment information and standard status information, and add the standard equipment information and standard status information of each medical device as a dictionary object to the preliminary medical equipment information set; that is, complete the construction of the medical equipment information set.

3. The method for identifying, connecting, and managing medical devices based on a central control screen according to claim 2, characterized in that, The method of feature encoding the medical device information set includes: The method of feature encoding the medical device information set includes: For categorical features, one-hot encoding is used to convert them into M binary features; for numerical features, the original numerical values ​​are used directly; for text features, string similarity is used to convert them into numerical vector features, thus forming a set of medical device feature vectors. The conversion methods for the numerical vector features include: Extract all distinct text-based strings from the medical device information set and construct a string set H; Calculate the transformation distance between any string in the string set H and all preset metastrings, and construct an m1×M1 distance matrix D, where m1 is the number of strings in the string set H; The element values ​​in the distance matrix D are normalized to obtain the normalized matrix D'. The distance similarity matrix HG = 1 - D' is constructed based on the normalized matrix D'. For each text-type feature string, find its index i in the string set H. Then the feature vector of that string is the numerical vector feature corresponding to the text-type feature.

4. The method for identifying, connecting, and managing medical devices based on a central control screen according to claim 3, characterized in that, The calculation method for the conversion distance includes: Construct a two-dimensional matrix DH, with the number of rows equal to the length of the first string plus 1, and the number of columns equal to the length of the second string plus 1. Initialize the values ​​of the first row and first column of DH; that is, assign the value 0 to DH[0,0] in the first row and first column; assign the value j to the remaining elements of the first row DH[j,0], indicating that it takes j insertion operations to convert an empty string into the first string; assign the value l to the remaining elements of the first column DH[0,l], indicating that it takes l deletion operations to convert the second string into an empty string. Starting from DH[1,1], calculate the value of each element DH[j,l] in row-major order; the value of DH[i,j] is equal to the minimum of DH[j-1,l-1]+de, ​​DH[j-1,l]+1, and DH[j,l-1]+1; where de = 0 if the characters at this position are the same in the two strings, otherwise de = 1; DH[j-1,l]+1 represents the number of edits required to delete the j-th character of the first string; DH[j,l-1]+1 represents the number of edits required to insert the i-th character of the second string before the j-th character of the first string; DH[j-1,l-1]+de represents the number of edits required to replace the j-th character of the first string with the l-th character of the second string; The two-dimensional matrix DH is continuously filled until DH[m3, n3] is calculated, where m3 and n3 are the lengths of the string and the metastring, respectively, and DH[m3, n3] is the conversion distance between the string and the metastring.

5. The method for identifying, connecting, and managing medical devices based on a central control screen according to claim 4, characterized in that, The preset methods for the metastring include: Create a set of reference strings, cluster the reference strings in the set to obtain M1 clusters, and select the reference string with the highest average similarity to other reference strings in the cluster as the metastring. The method of clustering reference strings within the reference string set includes: For any two reference strings ri and rj in the set of reference strings, calculate the similarity between them. Where |ri∩rj| represents the number of elements in the intersection of ri and rj, and |ri∪rj| represents the number of elements in the union of ri and rj; calculate the similarity of all reference strings pairwise and construct an N×N similarity matrix S; The similarity matrix S is propagated and updated to obtain the responsibility matrix R and availability matrix A of each reference string to other reference strings, and the updates are iteratively updated until convergence is achieved; the propagation responsibility matrix and propagation availability matrix are obtained; based on the propagation responsibility matrix and propagation availability matrix, M1 clusters are identified. The methods for identifying M1 clusters include: For each reference string ri, calculate re(ri) as the sum of R'(ri,rj) of all other reference strings rj; find the reference string ri* corresponding to the maximum value of re(ri) + ∑A′(ri,rj), and take ri* as the first cluster center; for each remaining reference string rj and all determined cluster centers ri*, calculate its maximum responsibility value with all current cluster centers; assign rj to the cluster of the cluster center with the maximum responsibility value; repeat until all reference strings have been assigned; that is, the identification of M1 clusters is completed.

6. The method for identifying, connecting, and managing medical devices based on a central control screen according to claim 5, characterized in that, The methods for propagating and updating the similarity matrix S include: Initialize the responsibility matrix R and the availability matrix A as matrices containing all zeros, that is, the elements R(ri,rj) in the responsibility matrix R = 0; the elements A(ri,rj) in the availability matrix A = 0. Update the elements in the responsibility matrix R according to their responsibilities; the formula for updating responsibilities is: R(ri,rj) = r(ri,rj) - max(0,r(rj,ri)), where r(ri,rj) is the responsibility of reference string ri to reference string rj; r(rj,ri) is the responsibility of reference string rj to reference string ri. Responsibility r(ri,rj) = Jd(ri,rj) - max(A(ri,rk) + A(rj,rk)); where A(ri,rk) represents the degree of advantage of choosing the reference string rk as a member of a cluster other than the cluster centered at ri; A(rj,rk) represents the degree of advantage of choosing the reference string rk as a member of a cluster other than the cluster centered at rj. Update the availability of the elements in availability matrix A using the following formula: A(ri,rj)=∑(max(0,r(rk,rj)));where r(rk,rj) is the responsibility of reference string rk to reference string rj; A(ri,rj) represents the availability value of selecting reference string rj as a member of other clusters besides ri; Calculate the sum of responsibility and availability of cluster centers to reference string ri, chi(ri), and the sum of responsibility and availability of cluster centers to reference string rj, chi(rj); chi(ri)=∑(R(ri,rj)+A(ri,rj)); chi(rj)=∑(R(rj,ri)+A(rj,ri)); where R(rj,ri) represents the responsibility value for assigning the reference string ri to the cluster centered at rj; A(rj,ri) is the availability value representing the selection of the reference string ri as a member of a cluster other than rj. The responsibility matrix R and the availability matrix A are normalized based on chi(ri) and chi(rj), i.e., R′(ri,rj)=R(ri,rj)+chi(ri), A′(ri,rj)=A(rj,ri)+chi(rj); where R′(ri,rj) are the elements of the normalized responsibility matrix R, and A′(ri,rj) are the elements of the normalized availability matrix A. After each iteration, the sum of all chi(ri) is calculated. If the change is less than the preset change threshold, the convergence has been achieved. Otherwise, the next iteration continues until the change is less than the preset change threshold, thus obtaining the propagation responsibility matrix corresponding to the responsibility matrix R and the propagation availability matrix corresponding to the availability matrix A.

7. A method for identifying, connecting, and managing medical devices based on a central control screen according to claim 6, characterized in that, The method of performing hierarchical clustering of medical devices includes: The medical device feature vector set X is standardized to obtain a standardized data matrix Z; principal component analysis is performed on the standardized data matrix Z to obtain the principal component factors and their corresponding factor loading matrices. Calculate the correlation coefficient matrix R of the standardized data Z, and perform eigenvalue decomposition on the correlation coefficient matrix R to obtain the eigenvalues ​​λ and the corresponding eigenvectors V. Based on the eigenvalues ​​λ, calculate the contribution rate of each principal component factor p. Where λ(p) is the eigenvalue corresponding to the principal component factor p; If a preset threshold th is set, then the smallest u value that makes ∑(co(1:u))≥th is defined as the number of principal component factors; where co(1:u) represents the contribution rate of the first u principal component factors. Extract the first u feature vectors and construct the factor loading matrix L; project the standardized data matrix Z onto a low-dimensional space composed of the first u principal component factors to obtain the dimensionality-reduced data representation Y = Z·L; Perform core matrix decomposition on Y to obtain the core matrix SH and two projection matrices PH and QH, such that Y = PH·SH·QH'; where QH' is the transpose of QH; Preliminary clustering is performed on the core matrix SH to obtain n5 initial clusters; the cluster mean vector y_mean of each initial cluster is calculated, and the cluster mean vector y_mean is projected back to the original space to obtain the cluster center point x_mean = y_mean·L′, where L′ is the transpose of the factor loading matrix L; the clusters are re-divided according to the distance of each point to the cluster center; this process is repeated until the cluster division is stable; the final n medical device category clusters are output.

8. The method for identifying, connecting, and managing medical devices based on a central control screen according to claim 7, characterized in that, The methods for performing preliminary clustering on the core matrix SH include: Calculate the core similarity matrix WH of the core matrix SH; the elements in the core similarity matrix WH are WH(I,J); Where SH(I,:) represents the row vector of the I-th data point in the core matrix SH; SH(I,:)' is the transpose of SH(I,:); SH(J,:) represents the row vector of the J-th data point in the core matrix SH; norm(SH(I,:)) represents the magnitude of SH(I,:); norm(SH(J,:)) represents the magnitude of SH(J,:); The co-occurrence similarity matrix CH is calculated based on the core similarity matrix WH. The element CH(z,v) in the co-occurrence similarity matrix CH is defined as ∑(WH(z,:)*WH(v,:)); where WH(z,:) is a row vector composed of the similarity values ​​between the z-th data point and all other data points; WH(v,:) is a row vector composed of the similarity values ​​between the v-th data point and all other data points; * represents element-wise multiplication. By integrating WH and CH, we obtain the integrated similarity matrix SP = δ·WH + (1-δ)·CH; where δ is the weight parameter. Based on SP, calculate the normalized Laplacian matrix LH and the first kp smallest non-zero eigenvectors VP of LH; perform k-means clustering on the row vectors of VP to obtain the initial cluster.

9. A method for identifying, connecting, and managing medical devices based on a central control screen according to claim 8, characterized in that, The construction methods of the medical directed graph model include: For each medical device category cluster, construct an empty directed graph, traverse all medical devices within the cluster, and add each medical device as a node to the directed graph; Obtain the interface types and number supported by each medical device; define constraints including for any two medical devices r and b; if the number of interfaces they support together is greater than 0, then connect; otherwise, do not connect. Extract the location information of each medical device and represent it using hierarchical coding; for any two medical devices, calculate their weighted distance according to the different levels of the hierarchical coding. Set a location distance threshold thre; if the weighted distance between locations is less than or equal to thre, then connect them. If medical device r is in use and b is idle, then r is not allowed to connect to b. Furthermore, if r is already connected to another medical device, then it is not allowed to connect to b again. The combined constraints determine whether any two medical devices r and b are connected or not. For a pair of interconnected medical devices (r, b), add a directed edge to the directed graph; assign a weight value to each directed edge; The formula for assigning the weight values ​​is as follows: E(r,b)=a1×T(r,b)+a2×Q(r,b)+a3×F(r,b)+a4×P(r,b); where E(r,b) is the weight value of the edge from medical device r to medical device b, a1 is the type weight coefficient, a2 is the usage state weight coefficient, a3 is the connection state weight coefficient, and a4 is the location distance weight coefficient; T(r,b) is the type priority function between medical device r and medical device b; Q(r,b) is the priority function for the usage status between medical device r and medical device b. If medical device r is idle and medical device b is in use, then Q(r,b) = 1; otherwise, Q(r,b) = 0. F(r,b) is the connection state priority function between medical device r and medical device b; if medical device r is not connected and medical device b is connected, then F(r,b) = 1; otherwise, F(r,b) = 0. P(r,b) is the positional distance priority function between medical device r and medical device b; Where d0 is the distance scale parameter, and dist(r,b) is the actual distance between medical device r and medical device b; Repeat this process until all medical devices within the medical device category cluster have been added to the directed graph, resulting in a medical directed graph model.

10. A method for identifying, connecting, and managing medical devices based on a central control screen according to claim 9, characterized in that, The methods for obtaining the optimal connection scheme include: Initialize the parameters of the optimization algorithm, including the number of bats in the colony m6, the maximum number of iterations mer, the sonar pulse emission rate rop, the lower limit of sonar frequency fmin, and the upper limit of sonar frequency fmax; Randomly generate n initial solutions, each of which is an n×n adjacency matrix; treat each solution as a bat; Define the fitness function of the optimization algorithm; Where toe is the sum of the weights of all directed edges in the connection scheme; prs is the device importance priority score; cts is the connectivity score; α1, β1 and γ1 are the corresponding weight coefficients; Where w1(r,b) is the weight of the directed edge between medical device r and medical device b, and tyo(r,b) is the type priority function from medical device r to medical device b. Where nu is the number of medical devices successfully connected under the connection scheme, and tod is the total number of all medical devices; Update the position and speed of each bat: v_h (t+1) =v_h (t) +(x_h (t) -x_best)·f_h;x_h (t+1) =x_h (t) +v_h (t+1) ; Among them, v_h (t) x_h is the speed of the h-th bat in the t-th iteration. (t) v_h is the position of the h-th bat in the t-th iteration. (t+1) Let f_h be the velocity of the h-th bat in the (t+1)-th iteration; x_best is the position of the current optimal solution; f_h is the sonar frequency of the h-th bat. A subset of bats is randomly selected based on the sonar pulse emission rate rop, and new connection schemes are generated through random perturbation. Calculate the fitness function value for each solution, denoted as the fitness function value, compare the fitness function values ​​of all new solutions with those of the old solutions, and select the one with the largest fitness function value as x_best; If the maximum number of iterations mer has been reached, then output the current x_best as the final optimal connection scheme.

Images