Auxiliary evaluation method and device for hyperamylaemia after ERCP operation and storage medium

Abstract

The invention provides an ERCP postoperative hyperamylase auxiliary evaluation method and device and a storage medium, and the method comprises the steps: obtaining a clinical data sequence of P medical links in an ERCP postoperative nursing process, generating P analysis vertexes through the P medical links, generating an association path between the P analysis vertexes through the clinical data sequence, and obtaining a medical nursing graph structure; defining an analysis vertex associated with other medical links in the medical care diagram structure as a first analysis vertex; defining graph vertexes included in all association paths from the first analysis vertex to any analysis vertex as dominant control vertexes of the analysis vertexes, and defining a medical link of the analysis vertex with the highest occurrence frequency in the dominant control vertexes as a core medical link; performing postoperative risk assessment on the core medical link information, and determining a first risk assessment result in the ERCP postoperative nursing process. According to the method, the risk assessment speed of the hyperamylase after the ERCP operation can be increased.

Description

Technical Field

[0001] The present application relates to the field of data processing technology, and in particular to a method, device and storage medium for auxiliary evaluation of hyperamylasemia after ERCP. Background Art

[0002] Endoscopic retrograde cholangiopancreatography (ERCP) is an important method widely used in the diagnosis and treatment of pancreatic and biliary diseases. However, patients may face a variety of complications after ERCP, among which hyperamylasemia is a more common one. Hyperamylasemia may not only affect the patient's postoperative recovery process, but also cause more serious complications, such as acute pancreatitis, which brings potential risks to the patient's health and also poses challenges to medical care.

[0003] In the existing post-ERCP care process, the risk assessment of hyperamylasemia often involves a large amount of clinical data, which comes from many medical links, such as postoperative vital signs monitoring, drug management, dietary care, etc. Due to the large number of medical links and the large amount of data generated in each link, large-scale data needs to be analyzed when conducting risk assessment. Traditional analysis methods often require a lot of time and computing resources, and the analysis speed is slow, which makes it difficult to meet timely and efficient clinical needs.

[0004] In order to process a large amount of clinical data for risk assessment, some existing technologies may require the addition of additional data analysis branches or the use of high-performance hardware equipment. This undoubtedly increases medical costs and is a considerable burden for many medical institutions. For example, in order to increase the data processing speed, it may be necessary to purchase a dedicated large server or expensive data processing software license, which is unaffordable for medical institutions with limited resources, especially grassroots medical institutions. Summary of the invention

[0005] In view of this, the embodiment of the present application provides a method, device and storage medium for auxiliary assessment of hyperamylasemia after ERCP. The technical solution of the present application is implemented as follows:

[0006] On the one hand, the present application provides an auxiliary assessment method for hyperamylasemia after ERCP, the method comprising: obtaining a clinical data sequence generated by P medical links in the process of post-ERCP care, generating P analysis vertices with the P medical links, generating association paths between the P analysis vertices with the clinical data sequence, and obtaining a medical care graph structure; P≥1; defining the analysis vertex in the medical care graph structure that is associated with the remaining medical links as the first analysis vertex; obtaining the dominant control vertices corresponding to the P analysis vertices in the medical care graph structure, and defining the medical link corresponding to the analysis vertex with the highest appearance frequency among the dominant control vertices corresponding to the P analysis vertices as the core medical link in the P medical links; the dominant control vertex corresponding to any analysis vertex represents a graph vertex included in all association paths from the first analysis vertex to the analysis vertex; performing a post-operative risk assessment on the core medical link information contained in the core medical link, and determining the first risk assessment result in the process of post-ERCP care.

[0007] On the other hand, the present application provides an auxiliary assessment device, including a memory and a processor, wherein the memory stores a computer program that can be executed on the processor, and the processor implements the steps in the above method when executing the program.

[0008] On the other hand, the present application provides a computer-readable storage medium having a computer program stored thereon, and the computer program implements the steps in the above method when executed by a processor.

[0009] The technical effects of the present application include: the present application obtains a clinical data sequence generated by P medical links in the process of post-ERCP care, generates P analysis vertices with the P medical links, generates association paths between the P analysis vertices with the clinical data sequence generated by the P medical links, and obtains a medical care graph structure, P≥1, defines the analysis vertices in the medical care graph structure that are associated with the remaining medical links as the first analysis vertex, obtains the dominant control vertices corresponding to the P analysis vertices from the medical care graph structure, defines the medical link corresponding to the analysis vertex with the highest frequency among the dominant control vertices corresponding to the P analysis vertices as the core medical link among the P medical links, the dominant control vertex corresponding to any analysis vertex represents a graph vertex included in all association paths from the first analysis vertex to the analysis vertex, performs postoperative risk assessment on the core medical link information contained in the core medical link, and determines the first risk assessment result in the process of post-ERCP care. Based on this, the clinical indicator data generated between each medical link is analyzed to obtain the medical care graph structure constructed by each medical link. Because only the medical care graph structure is analyzed, the vertex identification method is simple, and there is no need to analyze large-scale data, which improves the analysis speed. At the same time, there is no need to add additional data analysis branches, and the hardware cost is low. In addition, after determining the core medical link, only the core medical link information in the core medical link is used for postoperative risk assessment, which reduces the data analysis scale of the risk assessment of the postoperative care process of ERCP and further improves the risk assessment speed.

[0010] It should be understood that the above general description and the following detailed description are merely exemplary and explanatory, and are not intended to limit the technical solutions of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] The drawings herein are incorporated into the specification and constitute a part of the specification. These drawings illustrate embodiments consistent with the present application and are used together with the specification to illustrate the technical solution of the present application.

[0012] Figure 1 A schematic diagram of the implementation flow of a method for auxiliary assessment of hyperamylasemia after ERCP provided in an embodiment of the present application.

[0013] Figure 2 A hardware entity schematic diagram of an auxiliary evaluation device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0014] The embodiment of the present application provides an auxiliary evaluation method for hyperamylasemia after ERCP, which can be executed by a processor of an auxiliary evaluation device. The auxiliary evaluation device can refer to a device with data processing capabilities such as a used device, a computer device, a server, a laptop computer, a tablet computer, a desktop computer, a mobile device, etc.

[0015] Figure 1 A schematic diagram of the implementation process of a method for assisting in the evaluation of hyperamylasemia after ERCP provided in an embodiment of the present application is shown in FIG. Figure 1 As shown, the method includes:

[0016] Step S100: Acquire clinical data sequences generated by P medical links during the ERCP postoperative care process, generate P analysis vertices with the P medical links, generate association paths between the P analysis vertices with the clinical data sequences, and obtain a medical care graph structure; P≥1.

[0017] In the embodiment of the present application, in the postoperative care process of ERCP (endoscopic retrograde pancreatocholangiopancreatography), there are multiple different medical links, such as postoperative vital signs monitoring link, drug administration link, diet management link, etc., where P is used to represent the number of medical links, and P is an integer greater than or equal to 1. Each medical link will generate a corresponding clinical data sequence during execution. The clinical data sequence contains rich information. Taking the postoperative vital signs monitoring link as an example, the clinical data sequence generated by this link may include temperature data, blood pressure data, heart rate data, etc. measured at different time points, and these data are arranged in the order of measurement to form a data sequence. For example, in the drug administration link, the clinical data sequence may include information such as the name, dosage, and administration time of the drug used. The diet management link may involve data such as the type of diet and intake of the patient in different time periods.

[0018] After the auxiliary evaluation device obtains these clinical data sequences, it then generates P analysis vertices based on P medical links. Analysis vertices are the basic elements for constructing the medical care graph structure, and each medical link corresponds to an analysis vertex. For example, if there are three medical links (P=3), namely vital sign monitoring, drug administration, and diet management, then three analysis vertices will be generated, representing these three medical links.

[0019] Then, the auxiliary evaluation device generates an association path between P analysis vertices with the clinical data sequence, thereby obtaining a medical care graph structure. The generation of the association path is based on the intrinsic connection between each medical link and the correlation between clinical data. Continuing with the above three medical links as an example, the data of the vital signs monitoring link may affect the decision-making of the drug administration link. For example, when the patient's body temperature is too high, the drug dosage may need to be adjusted. Then, an association path will be established between the analysis vertex representing vital signs monitoring and the analysis vertex representing drug administration. Similarly, the drug administration link may affect the patient's diet management. For example, some drugs may affect the patient's appetite or dietary taboos, which in turn establishes an association path between the analysis vertices corresponding to the drug administration link and the diet management link. In this way, the association paths between all P analysis vertices are constructed to form a complete medical care graph structure.

[0020] In actual operation, the auxiliary evaluation device can use a variety of implementation methods to achieve this process. For example, a database management system can be used to store and manage clinical data sequences of various medical links. The creation of analysis vertices can be represented by defining specific data structures, such as using structures (in C language) or classes (in object-oriented programming languages). Each instance of a structure or class represents an analysis vertex, which contains identification information related to the medical link.

[0021] For the generation of association paths, rule-based methods or data mining techniques can be used. The rule-based method refers to determining the association between various medical links based on pre-set medical knowledge and nursing specifications. For example, if medicine stipulates that a certain drug cannot be used when the patient's blood pressure is below a certain threshold, then the auxiliary evaluation device can establish a corresponding association path between the analysis vertices corresponding to the blood pressure monitoring link and the drug administration link according to this rule. Data mining technology can automatically discover potential association patterns from a large amount of clinical data. For example, an association rule mining algorithm (such as the Apriori algorithm) can be used to determine the association paths between them by analyzing the frequent item sets of clinical data in different medical links.

[0022] Step S200: defining the analysis vertex in the medical care graph structure that is associated with other medical links as the first analysis vertex.

[0023] In the medical care graph structure, each analysis vertex represents a medical link. The auxiliary evaluation device conducts a comprehensive investigation of these analysis vertices to determine the first analysis vertex. For example, suppose there are five analysis vertices in the medical care graph structure, representing the five medical links of postoperative vital signs monitoring, drug administration, diet management, wound care and psychological counseling. The auxiliary evaluation device will analyze each vertex to determine whether it is associated with the other four vertices. From the perspective of graph structure, association means that there is a relationship of information transmission or mutual influence in the logical relationship and actual operation process of medical care. Taking the association between vital signs monitoring and other links as an example, the results of vital signs monitoring will affect the dosage and type of drug administration (if the vital signs are abnormal, the drug may need to be adjusted), which is an association; at the same time, the condition of vital signs will also affect diet management (such as abnormal body temperature may change the temperature requirements of the diet), wound care (for example, high blood pressure may affect wound healing and thus adjust the nursing method) and psychological counseling (the patient's physical signs of discomfort may cause changes in the psychological state and thus affect the focus of psychological counseling).

[0024] For the auxiliary evaluation device to determine the first analysis vertex, a graph traversal algorithm such as depth-first search or breadth-first search can be used. The depth-first search algorithm starts from a vertex and explores the graph as deeply as possible along a path until it cannot continue or the target condition is reached, and then backtracks to the last incompletely explored vertex to continue exploring. The breadth-first search starts from the starting vertex and searches the vertices in the graph layer by layer.

[0025] Assuming that a depth-first search algorithm is used, the auxiliary evaluation device starts from any analysis vertex, such as the analysis vertex representing drug administration, and explores its connection relationship with other vertices. The auxiliary evaluation device will record the vertices that have been visited to avoid repeated visits. During the exploration process, if it is found that all other vertices can be reached from this vertex, then this vertex may be the first analysis vertex. However, the auxiliary evaluation device performs such a depth-first search operation on each vertex to ensure that the vertices related to the remaining medical links are accurately found.

[0026] The medical care graph structure can be represented by an adjacency matrix or an adjacency list. The adjacency matrix is ​​a two-dimensional matrix. If there is an edge between vertex i and vertex j (that is, there is an association), the element in the i-th row and j-th column of the matrix is ​​1, otherwise it is 0. The auxiliary evaluation device can find the first analysis vertex by checking each row of the adjacency matrix (corresponding to each analysis vertex) to see if all columns (corresponding to other analysis vertices) have 1 (indicating an association). The adjacency list establishes a linked list for each vertex, and the vertices in the linked list represent other vertices adjacent to (associated with) the vertex. The auxiliary evaluation device can traverse the adjacency list of each vertex and count the number of adjacent vertices. If the number is equal to the total number of vertices minus 1 (because it does not include itself), then the vertex may be the first analysis vertex.

[0027] In this process, when judging whether a vertex is associated with all other vertices, a simple logical judgment expression can be used. Let the vertex set be V. For vertex v∈V, the association relationship is represented by the function R(v,u) (if v and u are associated, R(v,u) is true, otherwise it is false). Judging whether v is the first analysis vertex can be expressed as: R(v,u)=true. Here stands for "for all", which means that for all vertices u except v, v and u are associated.

[0028] Step S300: Obtain the dominant control vertices corresponding to P analysis vertices in the medical care graph structure, and define the medical link corresponding to the analysis vertex with the highest appearance frequency among the dominant control vertices corresponding to the P analysis vertices as the core medical link among the P medical links; the dominant control vertex corresponding to any analysis vertex represents a graph vertex included in all associated paths from the first analysis vertex to the analysis vertex.

[0029] In the medical care graph structure, for any analysis vertex, its dominant control vertex represents the graph vertex included in all the association paths from the first analysis vertex to the analysis vertex. This means that the dominant control vertex has a special control or influence position in these association paths. For example, in a simple medical care graph structure, suppose there are analysis vertices A (representing the vital signs monitoring link), B (representing the drug administration link), and C (representing the diet management link), and there are multiple association paths from the first analysis vertex to the analysis vertex B. If a vertex (assuming it is vertex D) always exists in these paths, then vertex D is the dominant control vertex of analysis vertex B.

[0030] In order to obtain the dominant control vertex corresponding to each analysis vertex, the auxiliary evaluation device needs to perform a detailed analysis of the medical care graph structure. A feasible implementation method is to use a method that combines path traversal and statistics. The auxiliary evaluation device starts from the first analysis vertex and traverses all associated paths to the target analysis vertex for each target analysis vertex. For example, for the target analysis vertex B, the auxiliary evaluation device finds all possible paths from the first analysis vertex to B. In the process of traversing these paths, the graph vertices contained in the paths are recorded and counted.

[0031] Assuming that the structure of the medical care graph is represented by an adjacency matrix, for each associated path, the vertices on the path can be determined by analyzing the elements in the adjacency matrix. Let the adjacency matrix be M. If M[i][j]=1, it means that there is an edge (i.e., association) between vertex i and vertex j. When traversing the path from the first analysis vertex (set to the vertex with index k) to the target analysis vertex (set to the vertex with index m), the auxiliary evaluation device starts from vertex k and searches for the vertex corresponding to the element with a value of 1 in M[k][*]. Continue to repeat this operation for the found vertices until vertex m is reached, and all the vertices passed through are recorded in the process.

[0032] After completing the traversal of all associated paths, the auxiliary evaluation device performs statistical analysis on the recorded vertices. If a vertex appears in all associated paths, then this vertex is the dominant control vertex corresponding to the target analysis vertex. This process can be represented by a simple logical judgment. Suppose there are n associated paths from the first analysis vertex to the target analysis vertex, namely Path1, Path2,…, Pathn, and the vertex set is V. For vertex v∈V, the logic for judging whether v is the dominant control vertex of the target analysis vertex can be expressed as: v∈Pathi. Here stands for "for all", which means that for all the associated paths, vertex v is in them.

[0033] After obtaining the dominant control vertices corresponding to the P analysis vertices, the auxiliary evaluation device turns its attention to determining the core medical link. The core medical link is a special link that stands out from the P medical links. It is determined by analyzing the medical link corresponding to the analysis vertex with the highest frequency among the dominant control vertices. For example, suppose there are five medical links corresponding to five analysis vertices A, B, C, D, and E, and their corresponding dominant control vertices are D, D, A, D, and B. Among these dominant control vertices, vertex D appears most frequently, so the medical link corresponding to analysis vertex D is the core medical link.

[0034] Determining the core medical links is of great significance in the risk assessment of the entire post-ERCP care process. This is because the core medical links reflect the special connection with other links through the relationship of the dominant control vertex in the medical care graph structure. This connection may imply its key role in the entire post-operative care process, and it may have a special relevance to both the patient's recovery process and the potential impact on hyperamylasemia.

[0035] In actual operation, the auxiliary evaluation device can create a counter array to count the frequency of occurrence of each dominant control vertex. For example, create an array count[] with a size equal to the total number of graph vertices. After determining the dominant control vertex of each analysis vertex, add 1 to the corresponding count[] element. Finally, by traversing the count[] array to find the dominant control vertex corresponding to the maximum value, the core medical link is determined.

[0036] Step S400: Perform a postoperative risk assessment on the core medical link information included in the core medical link to determine a first risk assessment result in the ERCP postoperative care process.

[0037] The core medical link is determined in the previous step through an in-depth analysis of the medical care graph structure, and it is of special importance in the entire medical care process. The core medical link information contains various data related to the link, which reflects the various characteristics and conditions of this link in the postoperative care process. For example, in an ERCP postoperative care scenario, if the drug administration link is determined to be the core medical link, then its core medical link information may include the type of drugs used, dosage, administration time, drug interaction information, etc.

[0038] The process of risk assessment of these core medical information by the auxiliary evaluation device involves many considerations. First, for clinical indicator data, the auxiliary evaluation device will establish a data model to analyze its normal range and fluctuations. Taking drug dosage as an example, each drug has a recommended safe dosage range, which is determined based on a large amount of medical research and clinical practice. The auxiliary evaluation device can compare the actual dosage with this safety range. If the dosage exceeds the upper limit of the safety range or is lower than the lower limit, it may mean that there is a risk.

[0039] From the perspective of data processing implementation, the auxiliary evaluation device can use statistical analysis methods. Assuming that there is a series of drug dosage data, the mean, standard deviation and other statistics of these data can be calculated. Assume that the drug dosage data are x1, x2, …, x n , mean Standard Deviation If an actual administered dose x j With the mean The distance exceeds a certain multiple (for example, 3 times the standard deviation), that is, This dose might then be considered abnormal, indicating a risk.

[0040] In addition to dosage, drug interactions are also important risk factors. The auxiliary assessment device can build a drug interaction knowledge base to match the drug information in the core medical process with the information in the knowledge base. For example, if two drugs A and B are used at the same time in the core medical process, and the knowledge base indicates that there is an adverse interaction between A and B (such as increasing the incidence of a certain adverse reaction), then this is also considered a risk situation.

[0041] Consider the factor of administration time. Some drugs need to be administered within a specific time interval to achieve the best therapeutic effect and avoid adverse reactions. The auxiliary evaluation device can check whether the actual administration time meets the requirements according to the preset rules. For example, a certain drug needs to be administered every 8 hours. If the actual administration time interval deviates too much (such as a deviation of more than 1 hour), it may affect the efficacy of the drug or increase the risk of the patient.

[0042] Through detailed analysis of various aspects of these core medical link information, the auxiliary assessment device determines the first risk assessment result in the ERCP postoperative care process. This result reflects the risk level of the core medical link under the current conditions. For example, the risk results can be divided into low risk, medium risk and high risk levels. If the drug dosage is abnormal multiple times, there are multiple adverse drug interactions and the administration time is inaccurate during the drug administration link, it may be judged as a high risk level; if there are only a few minor dosage deviations or occasional time deviations, it may be judged as a low risk level.

[0043] This risk assessment of core medical information is an important part of the entire auxiliary assessment method for post-ERCP hyperamylasemia. It is based on accurate core medical information and uses scientific data processing and analysis to provide a key basis for subsequent medical decisions (such as whether to adjust the treatment plan, strengthen monitoring, etc.), which helps to improve the safety and effectiveness of post-ERCP care and reduce the risk of complications such as hyperamylasemia.

[0044] As an implementation mode, step S200 defines the analysis vertex in the medical care graph structure that is associated with other medical links as the first analysis vertex, including:

[0045] Step S210: In the medical care graph structure, P analysis vertices are used as trigger vertices, and all associated paths of each analysis vertex in the medical care graph structure are obtained one by one;

[0046] Step S220: If all the associated paths obtained by walking one by one have all the associated paths including P analysis vertices, then the analysis vertex corresponding to the path is defined as the first analysis vertex;

[0047] Step S230: If all the associated paths obtained by walking one by one do not contain all the associated paths including P analysis vertices, an auxiliary reference vertex is generated in the medical care graph structure, and an auxiliary associated path is generated between the auxiliary reference vertex and the medical care graph structure, and the auxiliary reference vertex is defined as the first analysis vertex.

[0048] For step S210, in the medical care graph structure, P analysis vertices are used as trigger vertices, and all associated paths of each analysis vertex in the medical care graph structure are obtained by walking one by one. The medical care graph structure here is constructed based on the previous steps, in which the analysis vertices correspond to different medical links, and the associated paths reflect the logical relationship and information interaction between the medical links. For example, suppose there are four medical links, namely vital signs monitoring, drug administration, diet management and psychological counseling, and the corresponding analysis vertices are A, B, C, and D. When the auxiliary evaluation device uses analysis vertex A as the trigger vertex, it will start walking according to the connection relationship in the medical care graph structure to find all paths to other vertices that can be reached from A.

[0049] The auxiliary evaluation device can use a depth-first search (DFS) or breadth-first search (BFS) algorithm to implement this walking process. Taking depth-first search as an example, it starts from the trigger vertex (such as A) and explores the graph structure as deeply as possible along a path until it cannot continue or reaches a predetermined end condition (such as reaching a vertex that has been visited or reaching the boundary of the graph), and then backtracks to the last unexplored vertex to continue exploring other paths. For breadth-first search, it starts from the trigger vertex and explores the vertices in the graph layer by layer, first exploring all vertices that are one step away from the trigger vertex, and then exploring all vertices that are two steps away, and so on. Assuming that the medical care graph structure is represented by an adjacency matrix, if the element M[i][j] in the adjacency matrix is ​​1, it means that there is a direct connection between the analysis vertex i and the analysis vertex j (that is, they are connected by an edge), which provides a basis for walking.

[0050] In this example, if we start from the analysis vertex A, through the depth-first search algorithm, we may find that there is one path to vertex B, two different paths to vertex C, and three different paths to vertex D. These paths constitute all the associated paths with A as the trigger vertex. The auxiliary evaluation device will repeat this process for each analysis vertex (i.e., A, B, C, D) to obtain all the associated paths for each analysis vertex.

[0051] Next is step S220. If all the associated paths obtained by walking one by one include all the associated paths of P analysis vertices, the corresponding analysis vertex is defined as the first analysis vertex. Continuing with the above example, for all the associated paths obtained by each analysis vertex through step S210, the auxiliary evaluation device will check whether there is such a situation: all the associated paths of a certain analysis vertex can cover all other analysis vertices (that is, P analysis vertices). Assume that among all the associated paths of analysis vertex A, there are some paths that can reach B, C and D, that is, the associated path of A includes all the other three analysis vertices, then analysis vertex A meets this condition.

[0052] From the perspective of logical judgment, let the analysis vertex set be V = {A, B, C, D}. For a certain analysis vertex v (such as A), the set of all its associated paths is Path(v). Whether v meets the conditions can be judged by the following logical expression: u∈p. Here means "for all", means "exists", which means that for all vertices u except v, there exists a path p in v's associated path set Path(v) such that vertex u is on path p. If this logical expression holds, then the analysis vertex v can be defined as the first analysis vertex.

[0053] However, in step S230, if all the association paths obtained by walking one by one do not have all the association paths including P analysis vertices, an auxiliary reference vertex is generated in the medical care graph structure, and an auxiliary association path between the auxiliary reference vertex and the medical care graph structure is generated, and the auxiliary reference vertex is defined as the first analysis vertex. Returning to the previous example, if after checking in steps S210 and S220, it is found that the association path of no analysis vertex can cover all other analysis vertices, then the auxiliary evaluation device will create an auxiliary reference vertex.

[0054] This auxiliary reference vertex is a virtual construction. Its main purpose is to meet the needs of defining the first analysis vertex and has no direct correspondence with the actual medical link. For example, the auxiliary evaluation device creates an auxiliary reference vertex named E, and then establishes auxiliary association paths for this vertex and other vertices in the medical care graph structure. The way to establish these auxiliary association paths can be based on certain rules. For example, if there is some kind of logical indirect connection between other vertices, an auxiliary association path can be constructed based on this connection. Assuming that there is an indirect influence relationship between analysis vertices A and B (for example, through comprehensive consideration of the patient's overall condition), although there is no direct association path in the original medical care graph structure, an auxiliary association path can be created from the auxiliary reference vertex E to A and B based on this indirect relationship.

[0055] For the creation of auxiliary reference vertices and auxiliary association paths, corresponding operations can be performed in the data structure representing the medical care graph structure. If the medical care graph structure is represented by an adjacency matrix, then a row and a column can be added to represent the relationship between the auxiliary reference vertex E and other vertices, and the corresponding matrix elements are set to 1 or 0 according to whether there is an auxiliary association path; if it is represented by an adjacency list, a new adjacency linked list is created for the auxiliary reference vertex E, and the vertices in the linked list represent other vertices that have auxiliary association paths with E.

[0056] In actual medical scenarios, this method of determining analysis vertices can better adapt to different medical care processes and relationship structures. For example, in a more complex ERCP postoperative care scenario, there may be more medical links, such as wound care, drainage tube management, complication monitoring, etc., and the corresponding number of analysis vertices increases and the relationship becomes more complex. Through the above steps, the auxiliary evaluation device can accurately handle such complex situations. Whether the various medical links are closely linked and there are many direct and indirect connections, or when some links are relatively loosely linked, the first analysis vertex can be determined according to predetermined rules, thereby promoting the smooth progress of the entire risk assessment process.

[0057] In addition, when dealing with large amounts of clinical data, this graph-based analysis method can effectively integrate data from different sources. For example, vital signs monitoring may generate a large amount of time series data, drug administration involves different attribute data of multiple drugs, and dietary management includes data such as diet type and intake. These data are interconnected in the medical care graph structure through analysis vertices and association paths. Determining the first analysis vertex can find a key entry point in this complex data relationship network, making subsequent data analysis and risk assessment more focused and efficient.

[0058] From the perspective of medical decision-making, the determination of the first analysis vertex helps medical staff better understand the key links and relationships in the entire ERCP postoperative care process. For example, if the first analysis vertex is finally determined to be the analysis vertex corresponding to vital signs monitoring, this may indicate that in the entire postoperative care process, the vital signs monitoring link occupies a special core position in the relationship with other links. Based on this result, medical staff can pay more attention to the analysis and utilization of vital signs monitoring data, as well as its mutual influence relationship with other medical links (such as drug adjustment, diet change, etc.), so as to make more reasonable medical decisions, improve the quality of ERCP postoperative care, and reduce the risk of complications such as hyperamylasemia.

[0059] At the same time, this method is also helpful to explore the internal mechanism of the post-ERCP care process in medical research. By analyzing the association paths between different medical links and determining the first analysis vertex, we can further study the impact pattern of each link on patient recovery. For example, is there a relationship between the length or number of association paths between certain medical links and the probability of patients with hyperamylasemia? Does the medical link represented by the first analysis vertex play a special regulatory role in this relationship? These research questions can be explored by analyzing a large number of clinical cases based on this method, providing a theoretical basis for improving the post-ERCP care plan.

[0060] As another implementation, step S200 defines the analysis vertex in the medical care graph structure that is associated with other medical links as the first analysis vertex, including:

[0061] Step S201: obtaining the rth analysis vertex in the medical care graph structure, taking the rth analysis vertex as the starting vertex, and tracing the entire process along the medical care graph structure to obtain the rth basic medical process network; wherein, 1≤r≤P, the starting vertex of the rth basic medical process network is the rth analysis vertex;

[0062] Step S202: If there is an analysis vertex among the P analysis vertices that does not belong to the rth basic medical process network, the sth analysis vertex is used as the starting vertex, and the entire process is traced along the medical care graph structure to obtain the sth basic medical process network; s=r+1;

[0063] Step S203: If the rth basic medical history network includes P analysis vertices, the rth analysis vertex is defined as the first analysis vertex.

[0064] For step S201, the rth analysis vertex is obtained in the medical care graph structure, and the rth analysis vertex is used as the starting vertex, and the entire process is traced along the medical care graph structure to obtain the rth basic medical process network. Wherein, 1≤r≤P, and the starting vertex of the rth basic medical process network is the rth analysis vertex.

[0065] For example, suppose there are five analysis vertices in a medical care graph structure for post-ERCP care, corresponding to five medical links: vital sign monitoring (vertex A), drug administration (vertex B), diet management (vertex C), wound care (vertex D) and psychological counseling (vertex E). When r = 1, the auxiliary evaluation device selects vertex A as the starting vertex. Full-process tracking means that the auxiliary evaluation device will explore all vertices that can be reached from vertex A according to the association relationship in the graph structure and the connection paths between them.

[0066] The auxiliary evaluation device can use a graph traversal algorithm to implement this process, such as a depth-first search (DFS) algorithm. The depth-first search algorithm starts from the starting vertex (here, the r-th analysis vertex) and explores the graph structure as deeply as possible along a path until it cannot continue (for example, it reaches a vertex that has been visited or a vertex with no outgoing edges), and then backtracks to the last incompletely explored vertex to continue exploring other paths. In this process, the auxiliary evaluation device records the visited vertices and the connection relationships between them. These recorded vertices and connection relationships constitute the r-th basic medical history network.

[0067] Assume that in the tracing process, starting from vertex A, first reaching vertex B, then from vertex B to vertex C, then from vertex C back to vertex A, then from vertex A to vertex D, and finally from vertex D to vertex E. In this way, vertices A, B, C, D, E and the connection paths between them constitute the first basic medical journey network when r=1.

[0068] Next is step S202. If there is an analysis vertex among the P analysis vertices that does not belong to the rth basic medical process network, the sth analysis vertex is used as the starting vertex and the entire process is traced along the medical care graph structure to obtain the sth basic medical process network; s=r+1.

[0069] Continuing with the above example, suppose that in the first basic medical process network (constructed with vertex A as the starting vertex), it is found that vertex E is not included in it. Then according to step S202, the auxiliary evaluation device will select s=r+1=2, that is, vertex B as the new starting vertex, and perform full-process tracking again.

[0070] Using the same depth-first search algorithm, starting from vertex B, the auxiliary assessment device begins to explore all vertices and connection paths that can be reached. Assuming that the path traced this time is from vertex B to vertex C, then to vertex D, and then to vertex E, and there is also a path from vertex E back to vertex B, then vertices B, C, D, E and the connection paths between them constitute the second basic medical history network when s=2.

[0071] In this process, the auxiliary evaluation device continuously checks whether there are analysis vertices that are not included in the constructed basic medical process network. If so, it continues to select the next analysis vertex as the starting vertex to construct a new basic medical process network. This method of gradually constructing the basic medical process network helps to fully cover all analysis vertices, thereby providing a basis for determining the first analysis vertex.

[0072] Finally, in step S203, if the rth basic medical history network includes P analysis vertices, the rth analysis vertex is defined as the first analysis vertex.

[0073] Going back to the previous example, if when constructing a basic medical process network with a certain analysis vertex as the starting vertex, for example, the first basic medical process network constructed with vertex A as the starting vertex, it is found that this network contains all five analysis vertices (A, B, C, D, E), then the auxiliary evaluation device will define vertex A as the first analysis vertex.

[0074] From the perspective of logical judgment, let the analysis vertex set be V = {A, B, C, D, E}, and for the rth basic medical process network, let the vertex set it contains be Vr. Whether the rth basic medical process network meets the conditions can be judged by the following logical expression: V = Vr. If this expression is true, then the rth analysis vertex can be defined as the first analysis vertex.

[0075] As an implementation mode, step S300, obtaining dominant control vertices corresponding to P analysis vertices in the medical care graph structure, includes:

[0076] Step S310: obtaining all Q associated paths a from the first analysis vertex to the ath analysis vertex in the medical care graph structure, and obtaining the associated graph vertices respectively included in all Q associated paths a; wherein 1≤a≤P, 1≤Q;

[0077] Step S320: defining, among the association graph vertices respectively included in the Q association paths a, the association graph vertices common to all the Q association paths a as the dominant control vertices corresponding to the a-th analysis vertex.

[0078] For step S310, all Q associated paths a from the first analysis vertex to the ath analysis vertex are obtained in the medical care graph structure, and the associated graph vertices respectively included in all Q associated paths a are obtained. 1≤a≤P, 1≤Q.

[0079] For example, suppose there are five analysis vertices (corresponding to five medical links) in the medical care graph structure, and the five analysis vertices are labeled A, B, C, D, and E, and vertex A has been determined to be the first analysis vertex. When a=3 (i.e., the third analysis vertex C is considered), the auxiliary evaluation device starts to find all associated paths from the first analysis vertex A to the analysis vertex C.

[0080] The auxiliary evaluation device can use a depth-first search (DFS) or breadth-first search (BFS) algorithm to find these associated paths. Taking breadth-first search as an example, starting from vertex A, the auxiliary evaluation device first explores the vertices directly connected to A, and then gradually expands outward until all possible paths to vertex C are found. Assume that through this search method, the auxiliary evaluation device finds 3 (i.e., Q=3) associated paths from vertex A to vertex C.

[0081] For the first association path, the association graph vertices that may be included are A, B, and C, which means starting from vertex A, passing through vertex B to reach vertex C; the second association path may be A, D, C, that is, starting from A, passing through D to reach C; the third association path may be A, E, C, starting from A, passing through E to reach C. The auxiliary evaluation device will record the association graph vertices included in each of the three association paths.

[0082] Next is step S320, defining the association graph vertices included in the Q association paths a as the dominant control vertices corresponding to the a-th analysis vertex.

[0083] Continuing with the above example, for the three association paths from vertex A to vertex C found above (including association graph vertices A, B, C; A, D, C; A, E, C respectively), the auxiliary evaluation device will check the association graph vertices that exist in all three paths. In this example, only vertex A appears in all three paths, so according to the definition of step S320, vertex A is the dominant control vertex corresponding to the analysis vertex C.

[0084] From the perspective of logical judgment, let the Q associated paths from the first analysis vertex to the ath analysis vertex be Path1, Path2, ..., PathQ, and the set of associated graph vertices contained in each path Pathi (i = 1, 2, ..., Q) be Vi. For an associated graph vertex v, whether it is the dominant control vertex corresponding to the ath analysis vertex can be determined by the following logical expression: v∈Vi. Here means "for all", this expression means that for all the association paths, the association graph vertex v is in it.

[0085] Taking ERCP postoperative care as an example, assume that analysis vertex A represents the vital signs monitoring link, and analysis vertex C represents the drug administration link. If vertex A is the dominant control vertex corresponding to vertex C, it means that in the various logical relationships and information transmission paths from the vital signs monitoring link to the drug administration link, the vital signs monitoring link has an influence that runs through the entire process. For example, changes in vital signs may be a key factor in determining the dose, type or time of drug administration. Regardless of the specific nursing process or individual patient situation, as long as it involves the associated path from vital signs monitoring to drug administration, the vital signs monitoring link plays a dominant control role in it.

[0086] As another implementation, step S300, obtaining dominant control vertices corresponding to P analysis vertices in the medical care graph structure, includes:

[0087] Step S301: defining a first analysis vertex as a dominant control vertex of the first analysis vertex;

[0088] Step S302: Obtain the first upstream associated vertex of the second analysis vertex, and define the key influencing vertex of the first upstream associated vertex of the second analysis vertex and the second analysis vertex as the dominant control vertices of the second analysis vertex; the second analysis vertex is an analysis vertex among P analysis vertices other than the first analysis vertex.

[0089] For step S301, the first analysis vertex is defined as the dominant control vertex of the first analysis vertex. In the medical care graph structure, the first analysis vertex has a special status. It is related to the remaining medical links. This association gives it an initial and basic status in its own dominant control relationship. For example, suppose there are four analysis vertices in a medical care graph structure for ERCP postoperative care, corresponding to the four medical links of vital signs monitoring, drug administration, diet management and psychological counseling. If the analysis vertex corresponding to vital signs monitoring is determined to be the first analysis vertex. From the concept of the dominant control vertex, for this first analysis vertex itself, due to its core relevance in the entire graph structure, the auxiliary evaluation device defines it as its own dominant control vertex. This is like in an organizational structure, a core department has a basic control role over its own operations under certain relationship definitions.

[0090] The auxiliary evaluation device can indicate that it is its own dominant control vertex by specially marking the first analysis vertex in the data structure or setting a specific attribute. For example, if the medical care graph structure is stored in the form of an adjacency list, for the vertex representing the first analysis vertex, an attribute field can be added to identify it as the dominant control vertex. This attribute value can be a specific identifier or a pointer to itself (under the concept of object-oriented programming). In the case of representing it in an adjacency matrix, this relationship can be represented by a specific matrix element value (such as a diagonal element). For example, setting the element to a special value indicates that this vertex is its own dominant control vertex.

[0091] Next is step S302, obtaining the first upstream associated vertex of the second analysis vertex, defining the key influencing vertex of the first upstream associated vertex of the second analysis vertex and the second analysis vertex as the dominant control vertex of the second analysis vertex. Here, the second analysis vertex is an analysis vertex other than the first analysis vertex among the P analysis vertices.

[0092] Taking the four analysis vertices mentioned earlier (vertices corresponding to vital signs monitoring, drug administration, diet management, and psychological counseling) as an example, if the vertex corresponding to vital signs monitoring is the first analysis vertex, then the vertex corresponding to drug administration is a second analysis vertex. For the second analysis vertex corresponding to drug administration, the auxiliary evaluation device begins to look for its first upstream associated vertex. The first upstream associated vertex refers to a vertex in the medical care graph structure that is directly associated with the second analysis vertex and is in an upstream position in a logical relationship. Assuming that in this example, the vital signs monitoring vertex is the first upstream associated vertex of the drug administration vertex, this means that the information or status generated by the vital signs monitoring link will directly affect the drug administration link.

[0093] The auxiliary evaluation device then determines the key influencing vertex of the first upstream associated vertex (vital sign monitoring vertex). The key influencing vertex is a vertex with a special influence in the association relationship from the first upstream associated vertex to the second analysis vertex. Assuming that in the association path from vital sign monitoring to drug administration, there is a vertex representing the evaluation result of the patient's basic physiological state (this vertex is part of the vital sign monitoring link), and the result of this vertex will greatly affect whether to administer drugs and what kind of drugs to administer, then this vertex representing the evaluation result of the patient's basic physiological state is the key influencing vertex of the vital sign monitoring vertex (the first upstream associated vertex).

[0094] Finally, the auxiliary evaluation device defines this key impact vertex and the second analysis vertex (drug administration vertex) together as the dominant control vertex of the second analysis vertex (drug administration vertex). This means that in the dominant control relationship of the medical link of drug administration, not only the upstream link directly related to it (vital sign monitoring) but also the part of this upstream link that has a key impact on it (the patient's basic physiological state assessment result) should be considered.

[0095] The auxiliary evaluation device can find the first upstream associated vertex by traversing the edge and vertex relationship in the medical care graph structure. For example, in the graph structure represented by the adjacency list, for the vertex representing the second analysis vertex, the adjacent vertices in its linked list are traversed, and which adjacent vertex is the first upstream associated vertex is determined according to the predefined medical link logical relationship. For determining the key influencing vertex of the first upstream associated vertex, it can be achieved by analyzing the clinical data and logical rules related to the vertex. If the clinical data is stored in a database, the auxiliary evaluation device can query the data table related to the first upstream associated vertex, and determine the key influencing vertex according to a specific field in the data (such as a field indicating the degree of influence) or based on a pre-set rule (such as a threshold value of a certain physiological indicator determines the key influence).

[0096] As an embodiment, the method further includes:

[0097] Step S303: obtaining a second upstream associated vertex of the first upstream associated vertex in the medical care graph structure, and obtaining a first all associated path including the first upstream associated vertex and the second upstream associated vertex in the medical care graph structure;

[0098] Step S304: Obtain the second all-association path in the first all-association path, and define the second upstream association vertex corresponding to the second all-association path as the key influencing vertex of the first upstream association vertex; the second all-association path is adjacent to the first upstream association vertex, and the first upstream association vertex does not have an edge with upstream association vertices other than the second upstream association vertex.

[0099] For step S303, a second upstream associated vertex of the first upstream associated vertex is obtained in the medical care graph structure, and a first all associated path including the first upstream associated vertex and the second upstream associated vertex is obtained in the medical care graph structure.

[0100] For example, in the medical care graph structure of ERCP postoperative care, it is assumed that a second analysis vertex (such as the vertex corresponding to the drug administration link) has been determined, and its first upstream associated vertex (such as the vertex corresponding to the vital signs monitoring link) has been found. Now the auxiliary evaluation device searches for the second upstream associated vertex of this first upstream associated vertex (vital signs monitoring vertex). Assume that in this graph structure, the vertex corresponding to the patient basic information collection link is the second upstream associated vertex of the vital signs monitoring vertex. This means that the patient basic information collection link logically precedes the vital signs monitoring link, and may have an impact on the vital signs monitoring link, and then indirectly affect the drug administration link.

[0101] Then the auxiliary evaluation device needs to obtain the first all-associated path including the first upstream associated vertex (vital sign monitoring vertex) and the second upstream associated vertex (patient basic information collection vertex). From the implementation point of view, the auxiliary evaluation device can use a graph traversal algorithm, such as a depth-first search (DFS) or a breadth-first search (BFS) algorithm to find these associated paths. Taking breadth-first search as an example, the auxiliary evaluation device starts from the second upstream associated vertex (patient basic information collection vertex) and searches for vertices connected to it layer by layer until the first upstream associated vertex (vital sign monitoring vertex) is found. In this process, all the paths passed are recorded, and these paths are the first all-associated paths. Assume that two such paths are found through breadth-first search: one is from the patient basic information collection vertex to a certain evaluation vertex in the middle, and then to the vital sign monitoring vertex; the other is from the patient basic information collection vertex directly to the vital sign monitoring vertex.

[0102] Next is step S304, in which the second all-association path is obtained from the first all-association path, and the second upstream association vertex corresponding to the second all-association path is defined as the key influencing vertex of the first upstream association vertex; the second all-association path here is that the first upstream association vertex is adjacent to the second upstream association vertex, and the first upstream association vertex does not have an edge with upstream association vertices other than the second upstream association vertex.

[0103] Continuing with the above example, among the two first all-associated paths found above, the auxiliary evaluation device must screen out the second all-associated paths that meet the conditions. For the first path (from the patient basic information collection vertex to a certain evaluation vertex in the middle, and then to the vital signs monitoring vertex), it does not meet the definition of the second all-associated path because there are other vertices in the middle (a certain evaluation vertex in the middle). For the second path (from the patient basic information collection vertex directly to the vital signs monitoring vertex), it satisfies the first upstream associated vertex (vital signs monitoring vertex) and the second upstream associated vertex (patient basic information collection vertex) are adjacent, and the vital signs monitoring vertex has no edges with upstream associated vertices other than the patient basic information collection vertex, so this path is the second all-associated path.

[0104] According to the definition in step S304, the auxiliary evaluation device defines the second upstream association vertex (patient basic information collection vertex) corresponding to the second all association paths as the key influence vertex of the first upstream association vertex (vital sign monitoring vertex). This means that in the logical relationship chain from patient basic information collection to vital sign monitoring to drug administration, the patient basic information collection vertex has a key influence on the vital sign monitoring vertex, which will indirectly affect the drug administration link.

[0105] Let the first set of all associated paths be PathAll, for path p∈PathAll, let the first upstream associated vertex be v1, and the second upstream associated vertex be v2. Whether path p is the second all associated path can be determined by the following logical expression: This expression means that path p contains the first upstream-related vertex v1 and the second upstream-related vertex v2, and for any vertex v3 other than v2, there is no edge from v3 to v1. If path p satisfies this expression, then it is the second-all-related path.

[0106] This method of determining the key influencing vertices is of great significance in the entire assisted evaluation method of post-ERCP hyperamylasemia. In actual medical scenarios, it helps to more accurately analyze the complex influence relationship between medical links. Taking ERCP post-operative care as an example, each medical link does not exist in isolation, but is affected by multiple upstream links, and there are also relationships between different upstream links. By determining the key influencing vertices in this way, this multi-level influence mechanism can be deeply explored.

[0107] As an embodiment, the method further includes:

[0108] Step S305: defining P analysis vertices as basic dominant control vertices of the second analysis vertex;

[0109] Based on this, step S302, obtaining the first upstream associated vertex of the second analysis vertex, defining the key influencing vertex of the first upstream associated vertex of the second analysis vertex and the second analysis vertex as the dominant control vertex of the second analysis vertex, includes:

[0110] Step S3021: when the dominant control vertex is obtained for the nth time, X first upstream associated vertices of the second analysis vertex are obtained in the medical care graph structure; wherein n and X are both natural numbers greater than 0;

[0111] Step S3022: obtaining the key influence vertices respectively corresponding to the X first upstream associated vertices of the second analysis vertex, and defining the overlapping parts of the key influence vertices respectively corresponding to the X first upstream associated vertices and the second analysis vertex as the nth iteration dominant control vertex of the second analysis vertex;

[0112] Step S3023: If the nth iteration dominant control vertex of the second analysis vertex is the same as the mth iteration dominant control vertex of the second analysis vertex, then the nth iteration dominant control vertex of the second analysis vertex is defined as the dominant control vertex of the second analysis vertex; if n is the value at startup, then the mth iteration dominant control vertex of the second analysis vertex is the basic dominant control vertex of the second analysis vertex, m=n-1.

[0113] For step S305, P analysis vertices are defined as the basic dominant control vertices of the second analysis vertex. In the medical care graph structure, each analysis vertex represents a medical link, and when the second analysis vertex (analysis vertex other than the first analysis vertex) considers the dominant control relationship, all P analysis vertices are first used as its basic dominant control vertices. This is a relatively broad initial definition, which lays the foundation for more precise determination of dominant control vertices in the future.

[0114] For example, in a medical care graph structure for ERCP postoperative care, there are five analysis vertices, corresponding to the five medical links of vital sign monitoring, drug administration, diet management, wound care and psychological counseling. If the analysis vertex corresponding to the drug administration link is considered as the second analysis vertex, then in this step, the auxiliary evaluation device will define these five analysis vertices (vertices corresponding to vital sign monitoring, drug administration, diet management, wound care and psychological counseling) as the basic dominant control vertices of the second analysis vertex of drug administration. This definition method is based on the integrity of the overall medical care graph structure. Considering that each medical link may have an impact on other links to some extent during the entire postoperative care process, all vertices are first included in the category of basic dominant control vertices.

[0115] The auxiliary evaluation device may create a data structure to store the basic dominant control vertex information of each second analysis vertex. For example, an array or a linked list may be used to store the identifiers or references of P analysis vertices in a corresponding data structure for each second analysis vertex.

[0116] Next is sub-step S3021 of step S302, in which, when the dominant control vertex is obtained for the nth time, X first upstream associated vertices of the second analysis vertex are obtained in the medical care graph structure; wherein n and X are both natural numbers greater than 0.

[0117] Continuing with the above example, assuming that the second dominant control vertex acquisition stage (n=2) has been entered, for the second analysis vertex of drug administration, the auxiliary evaluation device begins to look for its first upstream associated vertex in the medical care graph structure. Assume that in this graph structure, the auxiliary evaluation device finds that the analysis vertices corresponding to the two medical links of vital sign monitoring and diet management are the first upstream associated vertices of the drug administration vertex, where X=2. The first upstream associated vertex is the upstream vertex that directly affects the second analysis vertex in a logical relationship. They are located in the medical care process before the medical link corresponding to the second analysis vertex and are directly associated with it.

[0118] The auxiliary evaluation device can obtain the first upstream associated vertex by traversing the edges in the medical care graph structure. If the medical care graph structure is represented by an adjacency matrix, the auxiliary evaluation device can check the elements in the row representing the second analysis vertex and find the vertices corresponding to the columns with a value of 1 (indicating that there is an edge connected, that is, there is an association). These vertices are the first upstream associated vertices. If an adjacency list is used for representation, the auxiliary evaluation device can directly traverse the adjacency linked list of the vertex representing the second analysis vertex, and the vertex corresponding to the vertex in the linked list is the first upstream associated vertex.

[0119] Then comes sub-step S3022, obtaining the key influencing vertices corresponding to the X first upstream associated vertices of the second analysis vertex, and defining the overlapping parts of the key influencing vertices corresponding to the X first upstream associated vertices and the second analysis vertex as the nth iteration dominant control vertex of the second analysis vertex.

[0120] For example, for the second analysis vertex of drug administration mentioned above and its two first upstream associated vertices (vertices corresponding to vital signs monitoring and diet management). Assume that for the vital signs monitoring vertex, its key influence vertex is a specific vital signs indicator evaluation vertex; for the diet management vertex, its key influence vertex is the nutritional component evaluation vertex. If these two key influence vertices have overlapping parts (assuming that in some special cases, there is a specific association relationship between specific vital signs indicators and nutritional components, so that these two evaluation vertices have overlapping parts), then the auxiliary evaluation device defines this overlapping part and the drug administration vertex together as the second iterative dominant control vertex of the second analysis vertex of drug administration (because n=2).

[0121] From the perspective of logical judgment, let the second analysis vertex be v, the X first upstream associated vertices be u1, u2, ..., uX, and their corresponding key influence vertices be k1, k2, ..., kX. Whether a vertex v' belongs to the nth iteration dominant control vertex can be judged by the following logical expression: The meaning of this expression is that the vertex v' is either the second analysis vertex v itself, or there are two different first upstream associated vertices ui and uj, so that the vertex v' belongs to their corresponding key influence vertices ki and kj at the same time.

[0122] Finally, there is sub-step S3023. If the nth iteration dominant control vertex of the second analysis vertex is the same as the mth iteration dominant control vertex of the second analysis vertex, the nth iteration dominant control vertex of the second analysis vertex is defined as the dominant control vertex of the second analysis vertex; if n is the value at startup, the mth iteration dominant control vertex of the second analysis vertex is the basic dominant control vertex of the second analysis vertex, m=n-1.

[0123] Continuing with the example of the second analysis vertex of drug administration, if the iterative dominant control vertex obtained in the third iteration dominant control vertex acquisition (n=3) is the same as the iterative dominant control vertex obtained in the second iteration (m=2), then the auxiliary evaluation device defines the same iterative dominant control vertex as the dominant control vertex of drug administration of the second analysis vertex. This iterative approach helps to gradually converge to a more accurate definition of the dominant control vertex, by constantly comparing the results of different iteration stages, removing possible redundant or inaccurate parts, and finally determining the most logical dominant control vertex.

[0124] As an implementation manner, the number of the second analysis vertices is Y, where Y≥1; and the method further includes:

[0125] Step S306, taking the first analysis vertex as the starting vertex, tracing back the path of the medical care graph structure to obtain a graph vertex set corresponding to the P analysis vertices;

[0126] Based on this, step S302, obtaining the first upstream associated vertex of the second analysis vertex, defining the key influencing vertex of the first upstream associated vertex of the second analysis vertex and the second analysis vertex as the dominant control vertex of the second analysis vertex, includes:

[0127] Step S302A: obtaining the first upstream associated vertex of the g-th second analysis vertex in the graph vertex set, defining the key influencing vertex of the first upstream associated vertex of the g-th second analysis vertex and the g-th second analysis vertex as the dominant control vertex of the g-th second analysis vertex; wherein 1≤g≤Y;

[0128] Step S302B: if the g-th second analysis vertex is not the last analysis vertex in the graph vertex set, then obtain the h-th second analysis vertex in the graph vertex set, and obtain the dominant control vertex of the h-th second analysis vertex, where h=g+1;

[0129] Step S302C: If the g-th second analysis vertex is the last analysis vertex in the graph vertex set, jump to the step of defining the medical link corresponding to the analysis vertex with the highest frequency among the dominant control vertices corresponding to the P analysis vertices as the core medical link among the P medical links for execution.

[0130] For step S306, the path of the medical care graph structure is traced back with the first analysis vertex as the starting vertex to obtain a set of graph vertices corresponding to the P analysis vertices. In the medical care graph structure, the analysis vertices are connected to each other through associated paths, and the path tracing back is to explore the graph structure from the first analysis vertex in the direction opposite to the conventional path to find the graph vertices related to each analysis vertex.

[0131] For example, in a medical care graph structure for post-ERCP care, it is assumed that there are five analysis vertices corresponding to the five medical links of vital sign monitoring, drug administration, diet management, wound care and psychological counseling. If the analysis vertex corresponding to vital sign monitoring is determined as the first analysis vertex, the auxiliary evaluation device starts to trace the path backward from this vertex. For the analysis vertex of drug administration, the auxiliary evaluation device may find other graph vertices related to it through reverse tracing. For example, in the tracing process, it is found that the graph vertex corresponding to the previous medical order review link and the graph vertex corresponding to the patient's basic information evaluation link may be related to the drug administration link. In this way, for each analysis vertex (analysis vertices corresponding to vital sign monitoring, drug administration, diet management, wound care and psychological counseling), the auxiliary evaluation device will obtain a set of graph vertices related to it through this reverse tracing.

[0132] The auxiliary evaluation device can use a reverse depth-first search or a reverse breadth-first search algorithm to implement path reverse tracing. If a reverse depth-first search is used, the auxiliary evaluation device starts from the first analysis vertex, preferentially explores its direct reverse neighbor vertices, and then continues to explore the more distant reverse neighbor vertices along these vertices until a predetermined stop condition is reached (such as all possible reverse paths have been explored or the boundary of the graph has been reached). If a reverse breadth-first search is used, the auxiliary evaluation device starts from the first analysis vertex and explores the graph vertices layer by layer according to the reverse distance from the first analysis vertex, first exploring the reverse neighbor vertices with a distance of 1, then the distance of 2, and so on. In this process, the auxiliary evaluation device will record the set of graph vertices corresponding to each analysis vertex.

[0133] Next is sub-step S3021 of step S302A, which obtains the first upstream associated vertex of the g-th second analysis vertex in the graph vertex set, and defines the key influencing vertex of the first upstream associated vertex of the g-th second analysis vertex and the g-th second analysis vertex as the dominant controlling vertex of the g-th second analysis vertex; wherein, 1≤g≤Y, Y is the number of second analysis vertices, and Y≥1.

[0134] Assume that in the graph vertex set obtained above, for the second analysis vertex of drug administration (assuming g=1, that is, it is the first second analysis vertex considered), the auxiliary evaluation device searches for the first upstream associated vertex in its corresponding graph vertex set. For example, it is found that the graph vertex corresponding to the medical order review link is the first upstream associated vertex of the second analysis vertex of drug administration. Then the auxiliary evaluation device determines the key influencing vertex of this first upstream associated vertex (the graph vertex corresponding to the medical order review link). Assume that in the medical order review link, the qualification evaluation vertex of the reviewing doctor is the vertex that has a key influence on drug administration (because doctors with different qualifications may have differences in the degree of strictness when reviewing drugs, the degree of understanding of drug interactions, etc., which has a key impact on drug administration). Then the auxiliary evaluation device defines the qualification evaluation vertex of the reviewing doctor and the second analysis vertex of drug administration together as the dominant control vertex of the g-th second analysis vertex of drug administration.

[0135] From the perspective of logical judgment, let the g-th second analysis vertex be vg, and its graph vertex set be Vg. For vertex u∈Vg, if u satisfies the condition of being the first upstream associated vertex of vg, let its key influence vertex be k. To determine whether a vertex v' is the dominant control vertex of the g-th second analysis vertex, the following logical expression can be used: v'=vg||v'=k. The meaning of this expression is that vertex v' is either the g-th second analysis vertex vg itself, or the key influence vertex k of its first upstream associated vertex.

[0136] Next is sub-step S302B. If the g-th second analysis vertex is not the last analysis vertex in the graph vertex set, then obtain the h-th second analysis vertex in the graph vertex set, and obtain the dominant control vertex of the h-th second analysis vertex, h=g+1. Continuing with the above example, if drug administration is not the last second analysis vertex (assuming there are other second analysis vertices such as diet management), then when g=1 (drug administration), h=2, and the auxiliary evaluation device will then perform similar operations on the h-th second analysis vertex of diet management. That is, first find the first upstream associated vertex in the graph vertex set corresponding to diet management, then determine the key influencing vertex of this first upstream associated vertex, and finally define this key influencing vertex and the second analysis vertex of diet management as the dominant control vertex of the h-th second analysis vertex of diet management.

[0137] This way of processing each second analysis vertex in sequence helps to systematically determine the dominant control vertex of all second analysis vertices, thereby comprehensively analyzing the complex relationships between various analysis vertices in the medical care graph structure.

[0138] Finally, there is sub-step S302C. If the g-th second analysis vertex is the last analysis vertex in the graph vertex set, then jump to the step of defining the medical link corresponding to the analysis vertex with the highest frequency of appearance among the dominant control vertices corresponding to the P analysis vertices as the core medical link among the P medical links for execution. For example, if dietary management is the last second analysis vertex (assuming that the value of g corresponds to the second analysis vertex of dietary management and it is the last one), then after the auxiliary evaluation device completes the determination of the dominant control vertex of the second analysis vertex of dietary management, it will count the frequency of appearance of each dominant control vertex based on the dominant control vertices of all the P analysis vertices determined previously. Assuming that in this example, it is found through statistics that the dominant control vertex corresponding to the analysis vertex of drug administration has the highest frequency of appearance, then the auxiliary evaluation device will define the medical link of drug administration as the core medical link.

[0139] This method of determining the core medical links is of great significance in the entire auxiliary evaluation method of post-ERCP hyperamylasemia. In actual medical scenarios, the determination of core medical links helps medical staff focus on the links that have a greater impact on the entire postoperative care process. Taking ERCP post-operative care as an example, different medical links are interrelated and influence each other, but the core medical links may play a more critical role in the entire process. If drug administration is identified as a core medical link, medical staff can pay more attention to the impact of factors such as drug selection, dosage, and administration time on patient recovery, especially in the prevention of hyperamylasemia.

[0140] As an implementation mode, the first risk assessment result includes a high-risk medical link; step S400, performing a postoperative risk assessment on the core medical link information included in the core medical link, and determining the first risk assessment result in the ERCP postoperative care process, including:

[0141] Step S410: obtaining core medical link information in the core medical link; the core medical link information includes clinical indicator data corresponding to P medical links respectively;

[0142] Step S420: Generate graph vertex time series analysis charts corresponding to the P medical links respectively through the clinical indicator data corresponding to the P medical links respectively;

[0143] Step S430: defining the medical links with abnormalities in the time series analysis graph at the graph vertices as high-risk medical links.

[0144] For step S410, the core medical link information is obtained in the core medical link; the core medical link information includes clinical indicator data corresponding to P medical links. After the core medical link is determined, this link contains a lot of information related to medical care, among which clinical indicator data is the key part. For example, if the core medical link is the drug administration link, then its clinical indicator data may include the name of the drug used, the dosage, the administration time, the change data of the drug concentration in the body over time, etc.; if the core medical link is the vital signs monitoring link, the clinical indicator data may include the measurement values ​​of body temperature, blood pressure, heart rate, respiratory rate, etc. at different time points.

[0145] The auxiliary evaluation device obtains these core medical link information from the medical database or the relevant medical record system. In terms of implementation, data extraction can be achieved through database query statements. For example, if a relational database (such as MySQL) is used, an SQL query statement can be written to obtain relevant clinical indicator data from the corresponding data table according to the identifier of the core medical link (such as a specific code or name). Assuming that there is a table called "medical_data", which contains fields such as "medical_procedure" (medical link name), "clinical_indicator" (clinical indicator name) and "value" (indicator value), the query statement may be similar to "SELECT value FROMmedical_data WHERE medical_procedure = 'core medical link name' AND clinical_indicator = 'specific clinical indicator name'", so that the data of specific clinical indicators under a specific core medical link can be obtained.

[0146] Next is step S420, in which the time series analysis charts of the graph vertices corresponding to the P medical links are generated through the clinical indicator data corresponding to the P medical links. The clinical indicator data of each medical link will form a time series over time, and the auxiliary evaluation device uses these time series data to create the time series analysis charts of the graph vertices. Taking the drug administration link as an example, if the clinical indicator of drug concentration is concerned, the auxiliary evaluation device collects the drug concentration data at different time points, takes time as the horizontal axis and drug concentration as the vertical axis, and draws a curve of drug concentration changing with time. This is part of the time series analysis chart of the graph vertices corresponding to the drug administration link. For the vital signs monitoring link, if the clinical indicator of body temperature is considered, the curve is also drawn with time as the horizontal axis and body temperature as the vertical axis.

[0147] The auxiliary assessment device can use data visualization tools to create these charts. For example, the Matplotlib library in Python provides a wealth of functions to draw various types of charts. For the above drug concentration time series data, assuming that it is processed in Python, the code may be as follows:

[0148] import matplotlib.pyplot as plt

[0149] import numpy as np

[0150] Assume that this is the drug concentration time series data obtained from the database and stored in the drug_concentration array

[0151] Here, for example, some simulated data is simply generated:

[0152] time_points=np.arange(0,10,1)

[0153] drug_concentration=np.array([10,8,6,5,4,3.5,3,2.5,2,1.5])

[0154] plt.plot(time_points,drug_concentration)

[0155] plt.xlabel('Time')

[0156] plt.ylabel('Drug Concentration')

[0157] plt.title('Drug Concentration Time-Series Chart for DrugAdministration')

[0158] plt.show()

[0159] In this way, the auxiliary evaluation device creates graphs of time series analysis for the relevant clinical indicators of each medical link. These graphs can intuitively show the changing trend of clinical indicators over time, which helps to discover possible abnormal patterns or laws.

[0160] Finally, in step S430, the medical procedures with abnormal graph vertex time series analysis charts are defined as high-risk medical procedures. When analyzing these graph vertex time series analysis charts, the auxiliary evaluation device determines whether there are abnormalities through preset rules or algorithms. For example, for the time series chart of drug concentration, if the drug concentration suddenly increases or decreases by more than a certain threshold, it may be regarded as abnormal. Suppose the normal fluctuation range of drug concentration is [Cmin, Cmax]. If at a certain time point t, the drug concentration C(t) satisfies C(t) < Cmin or C(t) > Cmax, then this change in drug concentration is considered abnormal.

[0161] Taking body temperature in the vital sign monitoring procedure as another example, if the normal body temperature range is [36.0, 37.2] degrees Celsius and it is found that the body temperature exceeds this range multiple times in the graph vertex time series analysis chart, this indicates an abnormal situation. For some more complex clinical indicators, more complex algorithms may be required to determine abnormalities. For example, for an indicator such as heart rate variability (HRV) that reflects the function of the cardiac autonomic nervous system, the auxiliary evaluation device may need to use statistical methods, such as calculating statistical quantities such as the standard deviation and mean of HRV, and comparing them with the reference range of the normal population. If the standard deviation is too large or the mean deviates from the normal range, it may be determined as abnormal.

[0162] When the auxiliary evaluation device finds such an abnormal situation in a certain medical procedure in the graph vertex time series analysis chart, this medical procedure is defined as a high-risk medical procedure. The determination of such high-risk medical procedures is of great significance in the risk assessment during the postoperative care process of ERCP.

[0163] As an implementation manner, the method further includes:

[0164] Step S500A: Remove the high-risk medical procedures and execute the postoperative care process of ERCP again;

[0165] Or;

[0166] Step S500B, obtain the classification strategy passed by the postoperative care process of ERCP, obtain the first treatment category corresponding to the high-risk medical procedure under the classification strategy, obtain the treatment tasks corresponding to the first treatment category in the postoperative care process of ERCP, conduct a risk assessment when dealing with the treatment tasks for the first treatment category, and determine the second risk assessment result of the postoperative care process of ERCP.

[0167] For step S500A, the high-risk medical link is removed and the ERCP postoperative care process is performed again. When the auxiliary evaluation device determines the high-risk medical link through the previous steps, removing the link is an intervention strategy. For example, suppose that in the ERCP postoperative care process, the drug administration link is determined as a high-risk medical link. This may be because in the time series analysis chart of the vertex of the drug administration link, the fluctuation of the drug dosage exceeds the normal range or there is a potential adverse interaction between drugs.

[0168] After the auxiliary assessment device marks this drug administration link as high risk, it is theoretically removed in the intervention. From a practical perspective, it may mean adjusting the care plan, temporarily stopping the use of certain drugs or changing the method of administration, so that this link is no longer carried out in the original way that may be risky. Then the post-ERCP care process is performed again, which is equivalent to a revised nursing process attempt.

[0169] The auxiliary evaluation device can achieve this operation by modifying the algorithm or rules of the nursing process. If the entire post-ERCP care process is composed of a series of algorithm modules, each module corresponds to a medical link, then the auxiliary evaluation device can temporarily disable or modify the algorithm module corresponding to the high-risk medical link (such as the drug administration link). For example, in a nursing process management system based on a rule engine, the auxiliary evaluation device can update the rule set related to drug administration, set high-risk operations (such as high-dose use rules for specific drugs) to invalid or adjust them to safer alternative rules.

[0170] This method of removing high-risk medical links and re-executing the nursing process has certain significance in medical practice. In some cases, by removing factors that may cause risks, the patient's reaction can be observed to determine whether the link is indeed the key factor leading to adverse outcomes (such as hyperamylasemia). For example, if the patient's hyperamylasemia-related indicators are improved after removing high-risk factors in the drug administration link (such as stopping the use of a suspicious drug), then it can be more certain that the drug or the specific operation in the drug administration link is risky.

[0171] For step S500B, obtain the classification strategy adopted by the ERCP postoperative care process, obtain the first treatment category corresponding to the high-risk medical link under the classification strategy, obtain the treatment task corresponding to the first treatment category in the ERCP postoperative care process, perform risk assessment on the first treatment category when handling the treatment task, and determine the second risk assessment result of the ERCP postoperative care process.

[0172] For example, suppose that the ERCP postoperative care process adopts a classification strategy based on disease severity and individual patient characteristics. Under this classification strategy, high-risk medical links (such as the drug administration link mentioned above) may be classified into a specific first treatment category, such as "high-risk drug intervention category". This first treatment category corresponds to a series of treatment tasks in the entire ERCP postoperative care process. For example, for the "high-risk drug intervention category", the treatment tasks may include drug selection, dose adjustment, drug effect monitoring, etc.

[0173] The auxiliary evaluation device obtains the relevant information of the classification strategy, which can be obtained from the database or knowledge base storing the nursing process strategy. Then the first treatment category to which the high-risk medical link belongs is determined, also by querying the pre-set classification mapping relationship (such as a table storing the correspondence between medical links and treatment categories). Then the treatment task corresponding to the first treatment category is found.

[0174] For the risk assessment of this first treatment category when handling treatment tasks, the auxiliary evaluation device integrates many factors. For example, for the treatment task of drug selection, the auxiliary evaluation device must consider factors such as the safety, effectiveness, and interaction with other drugs of the selected drug. Suppose there is a drug safety assessment function Safety(drug), which returns a value indicating the degree of safety based on the properties of the drug (such as the chemical structure of the drug, known adverse reactions, etc.), and the value range can be [0,1], where 0 means extremely unsafe and 1 means very safe. For drug effectiveness evaluation, there can be a similar function Effectiveness(drug) that returns a value indicating the effectiveness of the drug.

[0175] When evaluating the risk of drug selection, a simple formula can be used: Risk_selection = 1-Safety(drug)*Effectiveness(drug). This formula indicates that the risk level is inversely proportional to the safety and effectiveness of the drug. If the drug is both safe and effective, the risk is low; conversely, if the drug is low in safety or low in effectiveness, the risk is high.

[0176] For the treatment task of dose adjustment, risk assessment may need to consider the patient's individual physiological parameters, such as weight, age, liver and kidney function, etc. Suppose there is a function Dose_adjustment_risk(weight, age, liver_function, kidney_function, dose) that calculates the risk value of dose adjustment based on these parameters and dose.

[0177] The auxiliary assessment device comprehensively obtains the second risk assessment result of the ERCP postoperative care process by evaluating the risk of each treatment task under the first treatment category. This second risk assessment result provides more detailed and targeted risk information.

[0178] In medical practice, this risk assessment method helps to more accurately understand the risk status of high-risk medical links in specific treatment categories and tasks. For example, if the risk of drug selection tasks in the "high-risk drug intervention category" is high, medical staff can consider changing the drug or further studying the applicability of the drug in this case. For situations where the risk of dose adjustment tasks is high, the dose can be adjusted more carefully or a more personalized dose adjustment plan can be adopted, which can help improve the quality of post-ERCP care and more effectively prevent the occurrence of complications such as hyperamylasemia.

[0179] As an embodiment, the method further includes:

[0180] Step S10: determining a grouping strategy in the patient care process, and dividing the P medical links into at least two data processing units and at least two algorithm processing units according to the grouping strategy; each data processing unit uses the same data set to execute the patient care process, and each algorithm processing unit uses the same prediction algorithm;

[0181] Step S20: determining the first data processing unit to which the current responsible unit belongs from at least two data processing units, and determining the first algorithm processing unit to which the current responsible unit belongs from at least two algorithm processing units;

[0182] Step S30: obtaining the proposed debugging algorithm associated with the first algorithm processing unit from the initial prediction algorithm corresponding to the patient care process, and optimizing the algorithm parameters of the proposed debugging algorithm using the data sample corresponding to the first data processing unit;

[0183] Step S40: If the algorithm parameter optimization process is terminated, restart the collection of clinical data generated by P medical links during the patient care process.

[0184] For step S10, determine the grouping strategy in the patient care process, and divide P medical links into at least two data processing units and at least two algorithm processing units according to the grouping strategy; each data processing unit uses the same data set to execute the patient care process, and each algorithm processing unit uses the same prediction algorithm.

[0185] During the post-ERCP care process, the patient's care involves multiple medical links. In order to process and analyze data more efficiently, the auxiliary evaluation device determines the grouping strategy. For example, one possible grouping strategy is to group based on the functional characteristics of the medical links. Assume that there are 8 medical links (P=8), namely vital signs monitoring, drug administration, diet management, wound care, psychological counseling, complication monitoring, rehabilitation training and equipment maintenance. The auxiliary evaluation device can divide these links into different data processing units and algorithm processing units according to their functions.

[0186] For example, vital sign monitoring and complication monitoring can be divided into one data processing unit, because they mainly involve data collection and monitoring operations and use the same data set (for example, all kinds of data about the patient's physical condition). Medication administration, diet management and rehabilitation training can be divided into another data processing unit, because they need to refer to some similar data types when performing intervention operations on patients, such as the patient's physical recovery status data.

[0187] For the algorithm processing unit, it is assumed that the prediction algorithm used is a risk prediction algorithm based on machine learning. The medical links corresponding to vital signs monitoring, wound care, and equipment maintenance can be classified into one algorithm processing unit, and the same machine learning-based algorithm is used in this unit to predict the possible risks in these links. For example, by analyzing vital signs data, it is predicted whether the patient may have physical changes related to hyperamylasemia, and the potential risks of equipment failure to patient care are predicted based on equipment maintenance data. The medical links corresponding to drug administration, diet management, psychological counseling, and rehabilitation training can be classified into another algorithm processing unit, and the same prediction algorithm is used to analyze the impact of factors in these links on patient recovery and risks, such as predicting the impact of drug administration on the risk of hyperamylasemia, and the impact of diet management on the overall recovery and risk of patients.

[0188] The auxiliary evaluation device can achieve this grouping through a predefined rule set. These rule sets can be stored in a database or configuration file. The auxiliary evaluation device matches the name, function description and other information of the medical link with the rules in the rule set to determine the data processing unit and algorithm processing unit to which each medical link belongs.

[0189] Next comes step S20, determining the first data processing unit to which the current responsible unit belongs from at least two data processing units, and determining the first algorithm processing unit to which the current responsible unit belongs from at least two algorithm processing units.

[0190] Continuing with the above example, during the entire nursing process, the auxiliary evaluation device needs to determine which data processing unit and algorithm processing unit are currently operating at a certain moment. Assuming that an operation closely related to the patient's physical condition data is currently being processed, the auxiliary evaluation device determines, based on the pre-set logic, that the first data processing unit to which the current responsible unit belongs is the unit that includes vital signs monitoring and complication monitoring. At the same time, if the current operation involves using a machine learning algorithm to predict the patient's risk of physical changes, the auxiliary evaluation device determines that the first algorithm processing unit to which the current responsible unit belongs is the unit that includes the medical links corresponding to vital signs monitoring, wound care, and equipment maintenance.

[0191] The auxiliary evaluation device can achieve this determination by maintaining an operation status record. This operation status record can record information such as the type of operation currently being performed and the data content involved. The auxiliary evaluation device matches this information with the definitions and functions of each data processing unit and algorithm processing unit to determine the unit to which it belongs.

[0192] Then comes step S30, in which the proposed debugging algorithm associated with the first algorithm processing unit is obtained from the initial prediction algorithm corresponding to the patient care process, and the algorithm parameters of the proposed debugging algorithm are optimized using the data samples corresponding to the first data processing unit.

[0193] For example, suppose that in an algorithm processing unit including medical links corresponding to vital sign monitoring, wound care, and equipment maintenance, the initial prediction algorithm is a risk prediction algorithm based on a decision tree. The auxiliary evaluation device obtains a proposed debugging algorithm associated with the algorithm processing unit from the initial prediction algorithm, which may be a sub-algorithm of the decision tree algorithm or an algorithm specifically improved based on the decision tree algorithm.

[0194] At the same time, the auxiliary evaluation device obtains patient physical condition data samples corresponding to the data processing unit including vital sign monitoring and complication monitoring. These data samples contain vital sign data (such as body temperature, blood pressure, heart rate, etc.) of many patients and records of whether complications occur or not.

[0195] For the optimization of algorithm parameters of the algorithm to be debugged, it is assumed that there are some key parameters in the algorithm to be debugged, such as the depth of the decision tree (set as d), the split vertex selection criterion (set as s), etc. The auxiliary evaluation device can be optimized in a variety of implementation methods, such as the cross-validation method. The data samples are divided into a training set and a validation set, and different parameter values ​​are used for training on the training set, and then the performance of the model is evaluated on the validation set. Assume that the mean square error (MSE) is used to measure the model performance, and for different combinations of d and s, calculate its MSE on the validation set. Let MSE(d,s) represent the mean square error when the depth of the decision tree is d and the split vertex selection criterion is s. The auxiliary evaluation device searches for different values ​​of d and s to find the parameter combination that minimizes MSE(d,s), thereby optimizing the algorithm parameters.

[0196] Finally, step S40 is performed. If the algorithm parameter optimization process is terminated, the clinical data generated by P medical links during the patient care process will be collected again.

[0197] During the optimization of algorithm parameters, the optimization process may be terminated due to various reasons. For example, the data sample is biased or incomplete, which makes it impossible to find a suitable parameter combination during the optimization process, so that the optimization process cannot converge or abnormal results appear. Assume that in the above-mentioned optimization process of the decision tree algorithm parameters, it is found that no matter how the decision tree depth and the split vertex selection criteria are adjusted, the mean square error is always large and there is no trend of improvement. The auxiliary evaluation device determines that the optimization process of the algorithm parameters is terminated.

[0198] At this point, the auxiliary assessment device restarts collecting clinical data generated by P medical links during the patient care process. This means returning to the source of the data to ensure the accuracy and completeness of the data. Technically speaking, the auxiliary assessment device can mark the previously collected data as invalid or inaccurate data, and then restart the data collection process, which may involve interaction with various medical devices and medical record systems to obtain the latest and most accurate clinical data, such as re-obtaining accurate vital signs data from vital signs monitoring devices, obtaining accurate drug use data from drug administration record systems, etc.

[0199] This operation flow from step S10 to step S40 has many important meanings in the entire ERCP postoperative hyperamylasemia auxiliary evaluation method. Dividing the medical links into different data processing units through grouping strategies helps to manage and analyze data more targetedly. The data of different units have similar characteristics, and more efficient data storage and processing methods can be used. For example, for data of vital signs monitoring and complication monitoring, time series data storage can be used, and special query and analysis tools can be developed for this type of data to improve data processing efficiency. In terms of algorithm application, dividing the medical links into algorithm processing units and optimizing algorithm parameters can improve the accuracy and applicability of the prediction algorithm. Different algorithm processing units use the same prediction algorithm for different medical link characteristics, and can better adapt to their respective situations by optimizing parameters. For example, in the algorithm processing unit including drug administration and other links, the optimized prediction algorithm can more accurately predict the effect of drugs on the risk of hyperamylasemia in patients, thereby providing a more reliable basis for adjusting drug treatment plans. In terms of risk assessment, this operation flow lays a more solid foundation for subsequent risk assessment. Accurate data and optimized algorithms can provide more reliable risk prediction results. For example, through the optimized prediction algorithm based on vital signs monitoring data, the risk of patients developing hyperamylasemia can be predicted more accurately, and preventive measures can be taken in advance, such as adjusting the care plan, increasing the monitoring frequency, etc. In terms of medical resource allocation, clear data processing units and algorithm processing units help to reasonably allocate medical resources. For example, if the algorithm of an algorithm processing unit (such as a unit that includes vital signs monitoring and other links) is optimized and found to have high requirements for computing resources, then more computing resources (such as more powerful servers, etc.) can be allocated to this unit in a targeted manner to ensure the efficient operation of the algorithm, thereby improving the quality and efficiency of the entire post-ERCP care process.

[0200] In summary, steps S10-S40 play an important role in the auxiliary evaluation of hyperamylasemia after ERCP in terms of data management, algorithm application, risk assessment, and medical resource allocation through reasonable grouping strategies, clear unit determination, algorithm parameter optimization, and re-collection of data when necessary, which helps to improve the scientificity, accuracy, and efficiency of the entire nursing process.

[0201] Figure 2 A hardware entity diagram of an auxiliary evaluation device provided in an embodiment of the present application is shown in FIG. Figure 2 As shown, the hardware entity of the auxiliary assessment device 1000 includes: a processor 1001 and a memory 1002, wherein the memory 1002 stores a computer program that can be run on the processor 1001, and the processor 1001 implements the steps of the method in any of the above embodiments when executing the program.

[0202] An embodiment of the present application provides a computer storage medium, which stores one or more programs. The one or more programs can be executed by one or more processors to implement the steps of the method for auxiliary assessment of hyperamylasemia after ERCP in any of the above embodiments.

[0203] It should be noted here that the description of the above storage medium and device embodiments is similar to the description of the above method embodiments, and has similar beneficial effects as the method embodiments. For the technical details not disclosed in the storage medium and device embodiments of the present application, please refer to the description of the method embodiments of the present application for understanding. The above processor can be at least one of a target application integrated circuit (Application Specific Integrated Circuit, ASIC), a digital signal processor (Digital Signal Processor, DSP), a digital signal processing device (Digital Signal Processing Device, DSPD), a programmable logic device (Programmable Logic Device, PLD), a field programmable gate array (Field Programmable Gate Array, FPGA), a central processing unit (Central Processing Unit, CPU), a controller, a microcontroller, and a microprocessor. It can be understood that the electronic device that realizes the above processor function can also be other, and the embodiments of the present application are not specifically limited.

[0204] The above is only an implementation method of the present application, but the protection scope of the present application is not limited thereto. Any technician familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application.

Claims

1. A method for assisting the evaluation of hyperamylasemia after ERCP, characterized in that: The method comprises: Acquire a clinical data sequence generated by P medical links in the ERCP postoperative care process, generate P analysis vertices with the P medical links, generate association paths between the P analysis vertices with the clinical data sequence, and obtain a medical care graph structure; P≥1; The analysis vertex in the medical care graph structure that is associated with other medical links is defined as a first analysis vertex; Obtaining the dominant control vertices corresponding to the P analysis vertices in the medical care graph structure, and defining the medical link corresponding to the analysis vertex with the highest appearance frequency among the dominant control vertices corresponding to the P analysis vertices as the core medical link among the P medical links; the dominant control vertex corresponding to any analysis vertex represents a graph vertex included in all associated paths from the first analysis vertex to the analysis vertex; Perform a postoperative risk assessment on the core medical link information contained in the core medical link to determine a first risk assessment result in the ERCP postoperative care process.

2. The method according to claim 1, characterized in that The analysis vertex in the medical care graph structure that is associated with other medical links is defined as the first analysis vertex, including: In the medical care graph structure, the P analysis vertices are respectively used as trigger vertices, and all associated paths of each analysis vertex in the medical care graph structure are obtained one by one; If all the associated paths obtained by walking one by one include all the associated paths including the P analysis vertices, the analysis vertex corresponding to the associated paths is defined as the first analysis vertex; If all the association paths obtained by walking one by one do not have all the association paths including the P analysis vertices, generating an auxiliary reference vertex in the medical care graph structure, generating an auxiliary association path between the auxiliary reference vertex and the medical care graph structure, and defining the auxiliary reference vertex as the first analysis vertex; or; The analysis vertex in the medical care graph structure that is associated with other medical links is defined as the first analysis vertex, including: Obtaining the rth analysis vertex in the medical care graph structure, taking the rth analysis vertex as the starting vertex, and tracing the entire process along the medical care graph structure to obtain the rth basic medical care process network; wherein, 1≤r≤P, the starting vertex of the rth basic medical care process network is the rth analysis vertex; If there is an analysis vertex among the P analysis vertices that does not belong to the r-th basic medical process network, the s-th analysis vertex is used as the starting vertex, and the entire process is traced along the medical care graph structure to obtain the s-th basic medical process network; s=r+1; If the rth basic medical history network includes the P analysis vertices, the rth analysis vertex is defined as the first analysis vertex.

3. The method according to claim 1, characterized in that The step of obtaining the dominant control vertices respectively corresponding to the P analysis vertices in the medical care graph structure includes: Obtaining Q all associated paths a from the first analysis vertex to the ath analysis vertex in the medical care graph structure, and obtaining the associated graph vertices respectively included in the Q all associated paths a; wherein 1≤a≤P, 1≤Q; Among the association graph vertices respectively included in the Q association paths a, the association graph vertices common to the Q association paths a are defined as the dominant control vertices corresponding to the a-th analysis vertex; or; The step of obtaining the dominant control vertices respectively corresponding to the P analysis vertices in the medical care graph structure includes: defining the first analysis vertex as a dominant control vertex of the first analysis vertex; Obtain the first upstream associated vertex of the second analysis vertex, and define the key influencing vertex of the first upstream associated vertex of the second analysis vertex and the second analysis vertex as the dominant control vertices of the second analysis vertex; the second analysis vertex is an analysis vertex among the P analysis vertices other than the first analysis vertex.

4. The method according to claim 3, characterized in that The method further comprises: Acquire a second upstream associated vertex of the first upstream associated vertex in the medical care graph structure, and acquire a first all associated path including the first upstream associated vertex and the second upstream associated vertex in the medical care graph structure; Obtain a second all-association path from the first all-association paths, and define a second upstream association vertex corresponding to the second all-association path as a key influencing vertex of the first upstream association vertex; the second all-association path is adjacent to the first upstream association vertex and the second upstream association vertex, and the first upstream association vertex does not have an edge with any upstream association vertex other than the second upstream association vertex.

5. The method according to claim 3, characterized in that The method further comprises: The P analysis vertices are defined as the basic dominant control vertices of the second analysis vertex; The acquiring the first upstream associated vertex of the second analysis vertex and defining the key influencing vertex of the first upstream associated vertex of the second analysis vertex and the second analysis vertex as the dominant controlling vertex of the second analysis vertex includes: When the dominant control vertex is obtained for the nth time, X first upstream associated vertices of the second analysis vertex are obtained in the medical care graph structure; wherein n and X are both natural numbers greater than 0; Obtaining the key influence vertices respectively corresponding to the X first upstream associated vertices of the second analysis vertex, and defining the overlapping parts of the key influence vertices respectively corresponding to the X first upstream associated vertices and the second analysis vertex as the nth iteration dominant control vertex of the second analysis vertex; If the nth iteration dominant control vertex of the second analysis vertex is the same as the mth iteration dominant control vertex of the second analysis vertex, the nth iteration dominant control vertex of the second analysis vertex is defined as the dominant control vertex of the second analysis vertex; if n is the value at startup, the mth iteration dominant control vertex of the second analysis vertex is the basic dominant control vertex of the second analysis vertex, m=n-1; The number of the second analysis vertices is Y, where Y≥1; and the method further includes: Taking the first analysis vertex as a starting vertex, tracing back the path of the medical care graph structure to obtain a graph vertex set corresponding to the P analysis vertices; The acquiring the first upstream associated vertex of the second analysis vertex and defining the key influencing vertex of the first upstream associated vertex of the second analysis vertex and the second analysis vertex as the dominant controlling vertex of the second analysis vertex includes: Obtaining the first upstream associated vertex of the g-th second analysis vertex in the graph vertex set, defining the key influencing vertex of the first upstream associated vertex of the g-th second analysis vertex and the g-th second analysis vertex as the dominant control vertices of the g-th second analysis vertex; wherein 1≤g≤Y; If the g-th second analysis vertex is not the last analysis vertex in the graph vertex set, then obtain the h-th second analysis vertex in the graph vertex set, and obtain the dominant control vertex of the h-th second analysis vertex, where h=g+1; If the g-th second analysis vertex is the last analysis vertex in the graph vertex set, then jump to the step of defining the medical link corresponding to the analysis vertex with the highest frequency among the dominant control vertices corresponding to the P analysis vertices as the core medical link among the P medical links for execution.

6. The method according to claim 1, characterized in that The first risk assessment result includes high-risk medical links; the postoperative risk assessment of the core medical link information included in the core medical links is performed to determine the first risk assessment result of the ERCP postoperative care process, including: Acquire core medical link information in the core medical link; the core medical link information includes clinical indicator data corresponding to the P medical links respectively; Generate graph vertex time series analysis charts corresponding to the P medical links respectively through the clinical indicator data corresponding to the P medical links respectively; The medical links with abnormalities in the time series analysis graph at the vertices of the graph are defined as high-risk medical links.

7. The method according to claim 6, characterized in that The method further comprises: The high-risk medical link is removed, and the ERCP postoperative care process is performed again; Alternatively, obtain the classification strategy adopted by the post-ERCP care process, obtain the first treatment category corresponding to the high-risk medical link under the classification strategy, obtain the treatment task corresponding to the first treatment category in the post-ERCP care process, perform a risk assessment on the first treatment category when handling the treatment task, and determine a second risk assessment result of the post-ERCP care process.

8. The method according to claim 1, characterized in that The method further comprises: Determine a grouping strategy in the patient care process, and divide the P medical links into at least two data processing units and at least two algorithm processing units according to the grouping strategy; each data processing unit uses the same data set to perform the patient care process, and each algorithm processing unit uses the same prediction algorithm; Determine the first data processing unit to which the current responsible unit belongs from the at least two data processing units, and determine the first algorithm processing unit to which the current responsible unit belongs from the at least two algorithm processing units; Obtaining a pseudo-debugging algorithm associated with the first algorithm processing unit from the initial prediction algorithm corresponding to the patient care process, and optimizing algorithm parameters of the pseudo-debugging algorithm using the data sample corresponding to the first data processing unit; If the algorithm parameter optimization process is terminated, the collection of clinical data generated by the P medical links during the patient care process will be restarted.

9. An auxiliary assessment device, comprising a memory and a processor, wherein the memory stores a computer program that can be run on the processor, characterized in that: When the processor executes the program, the steps in the method according to any one of claims 1 to 8 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps in the method according to any one of claims 1 to 8 are implemented.

Images