Neurosurgery patient follow-up visit management system

Abstract

The invention relates to the technical field of neurosurgery patient follow-up visit management, and discloses a neurosurgery patient follow-up visit management system. The system comprises a patient information acquisition module, a follow-up plan generation module, a data cross validation module, a risk level judgment module and a follow-up structure output module. The system constructs an original information sequence template by extracting patient medical record text information elements; establishing an information directed path based on the template to form a follow-up visit plan path structure; extracting a path node sequence and screening cross data; matching risk labels according to the terminal node to obtain a risk affiliation label group; and finally generating a follow-up visit structure table. According to the system, structured processing, dynamic planning, data cross validation and accurate risk assessment of follow-up visit information of neurosurgical patients are realized, and scientificity and high efficiency of follow-up visit management are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of neurosurgery patient follow-up management, and in particular to a neurosurgery patient follow-up management system. Background Art

[0002] In neurosurgery clinical practice, long-term follow-up of patients after surgery is an important means of monitoring disease progression and evaluating treatment efficacy. Neurosurgical diseases are characterized by complex conditions, long recovery periods, and diverse complications. Patients may experience various risks at different stages after surgery, including abnormal intracranial pressure, neurological dysfunction, and infection. The onset and manifestation of these risks vary significantly from patient to patient, posing significant challenges to follow-up management. Currently, follow-up management of neurosurgery patients relies heavily on manual record-keeping and empirical judgment. Medical staff need to organize a large amount of information such as patients' medical records, test results, medication status, and develop follow-up plans based on their personal experience. This approach not only consumes a lot of manpower and material resources, but is also prone to lack of specificity in follow-up plans due to information omissions or misjudgment. For example, for patients undergoing surgery for gliomas, some medical staff may focus more on indicators of tumor recurrence while ignoring the potential risk of epileptic seizures. For patients undergoing surgery for cerebral hemorrhage, the timing of intervention for vasospasm may be delayed due to failure to track blood pressure changes in a timely manner. While some medical institutions have introduced electronic medical record systems to assist with follow-up, these systems primarily focus on information storage and simple querying, failing to conduct in-depth analysis and structured processing of patient information. Traditional systems typically employ fixed intervals for follow-up planning, such as one, three, or six months after surgery, failing to dynamically adjust follow-up schedules based on the patient's specific medical characteristics. This fixed model can result in excessively long follow-up intervals for high-risk patients, leading to missed opportunities for intervention, while placing unnecessary follow-up burdens on low-risk patients, resulting in a waste of medical resources. In addition, in the data verification process, the existing system lacks a cross-checking mechanism for multi-source data. Follow-up data for neurosurgery patients comes from a wide range of sources, including imaging examinations, laboratory tests, clinical symptom descriptions, etc., and data from different sources may be contradictory or duplicated. For example, a patient's imaging report shows that the tumor volume has shrunk, but the clinical symptoms have worsened. At this time, if there is a lack of effective data cross-validation, it is easy for medical staff to misjudge the condition. At the same time, due to the lack of a unified information element extraction and indexing method, different medical staff may have different interpretations of the same patient's medical records, which reduces the comparability of follow-up data and affects the reliability of follow-up results. Current risk assessment methods are often based on single indicators or subjective experience, making it difficult to comprehensively consider multiple clinical characteristics of patients. For example, risk assessments are based solely on the time since surgery, ignoring key factors such as age, underlying conditions, and surgical approach. This can lead to significant discrepancies between risk assessment results and the actual condition. This discrepancy directly impacts the accuracy of follow-up management and prevents the provision of personalized health guidance and interventions for patients. With the annual increase in the number of neurosurgery patients, the limitations of traditional follow-up management models have become increasingly prominent. How to achieve structured processing of follow-up information, dynamic plan formulation, data cross-validation and accurate risk assessment has become an urgent problem to be solved in current neurosurgery clinical practice. Summary of the Invention

[0003] The purpose of the present invention is to provide a neurosurgery patient follow-up management system to solve the problems raised in the above background technology.

[0004] To achieve the above objectives, the present invention provides a neurosurgery patient follow-up management system, the system comprising: The patient information collection module obtains the medical record text from the follow-up data of neurosurgery patients, extracts information element indexes through word segmentation, determines the initial position and frequency of key information, maps and generates the information element arrangement sequence, and constructs the patient's original information sequence template; A follow-up plan generation module, based on the patient's original information sequence template, sorts the information elements, establishes a directed information path from the beginning to the end according to the sorting result, collects the connection node numbers and adjacent node relationships of all edges in the path, and constructs a follow-up plan path structure; A data cross-validation module extracts a patient path node sequence according to the follow-up plan path structure, compares the intersection nodes and counts the frequency of the end nodes, and screens the cross-data to obtain a patient cross-data set; The risk level determination module collects risk labels of the endpoint nodes based on the endpoint nodes in the patient cross-data set, sorts them by frequency of occurrence, and matches the path endpoint labels to obtain a patient risk attribution label group; The follow-up structure output module counts the follow-up classification nodes to which each label belongs based on the patient risk attribution label group, divides the patient path into the corresponding nodes, establishes the node and patient path follow-up attribution relationship structure, and generates a neurosurgery follow-up structure table.

[0005] Preferably, the patient original information sequence template includes an information element index arrangement structure, a key information sequence fragment set, and a key information frequency weight model; the follow-up plan path structure includes an information element hierarchical mapping relationship, a directed path node chain, and a node connection relationship set; the patient cross data set includes a path intersection node set, an endpoint node occurrence frequency distribution, and an intersection node screening result; the patient risk attribution label group includes a risk label frequency ranking, a path endpoint risk label mapping, and an attribution grade label matching result; and the neurosurgery follow-up structure table includes a classification node identifier, a patient path grouping result, and a follow-up attribution relationship mapping result.

[0006] Preferably, the patient information collection module includes: The basic information extraction submodule obtains the medical records of neurosurgery patients' follow-up data, performs information element-level word segmentation on the medical records, extracts the information element set of each patient, records the index position of each information element in the medical record, compares the relationship between the first occurrence position of key information in the information element arrangement list and the number of information elements, classifies them by patient, and obtains the key information element index distribution results; The medical history record construction submodule extracts the information fragments corresponding to the key information in the original medical record text based on the key information element index distribution result, intercepts the information elements based on the index interval of the key information in the medical record, constructs a set of information fragments based on the position of each key information intercepted information element, and reorganizes the information fragments based on the patient to which the information element belongs to obtain a set of original key information fragments; The information template generation submodule counts the occurrence frequency values of all key information based on the set of key information original fragments, and performs position rearrangement processing on the information fragment set based on the original order of the key information in the medical record text. The original fragments of multiple key information in the same patient are sequentially spliced according to the first appearance position, and the sum of the patient's original information weights is calculated. The information weight results corresponding to each patient are combined for classification and integration to obtain the patient's original information sequence template.

[0007] Preferably, the follow-up plan generation module includes: The period sorting submodule compares and sorts all information element nodes according to their period label priority values based on the original patient information sequence template and the period label of each patient information element node, rearranges the positions of the patient information elements in order from the beginning to the end, establishes a rearranged sequence index table, and obtains the information element sorting index value; The plan construction submodule obtains the set of adjacent nodes in the information element rearrangement sequence according to the information element sorting index value, numbers each pair of adjacent nodes and records their connection direction, calculates the information directed path offset strength value, combines the connection relationship of all nodes in the path, integrates the structural edge information, and generates the information directed path follow-up data; The node structure extraction submodule collects the node numbers and adjacent node pair relationships in all connecting edges according to the directed path follow-up data of the information, constructs a node mapping table based on the adjacent structural relationship, stores the forward and backward association relationship types and connection directions between each information element, and obtains the follow-up plan path structure.

[0008] Preferably, the data cross-validation module includes: A path extraction submodule, based on the follow-up plan path structure, collects a chronological sequence of nodes in the complete follow-up paths of any two different patients, where each patient's path contains all node information from the initial follow-up to the current stage, extracts the node number information under each path in turn and establishes a patient node mapping set, labels the patient identifier and path length parameter of each path, and obtains the patient path node number value; The intersection comparison submodule calls the node number sequences of any two patient paths according to the patient path node number values, performs an intersection comparison operation on the node sets of the two paths, extracts all the terminal node numbers in the intersection, counts the number of times such nodes appear in different paths, calculates the deviation value of the intersection path terminal, compares it with the data intersection judgment benchmark value one by one, selects the path pair combinations with the deviation value less than or equal to the benchmark value, and establishes the path intersection number set that meets the conditions; The path screening submodule queries the original patient path identifier according to the set of path intersection numbers that meet the conditions and the path combination corresponding to the number, integrates the patient identifier and the path intersection node information, establishes a patient path relationship list, and generates a patient intersection data set.

[0009] Preferably, the risk level determination module includes: The risk label collection submodule collects the risk label set to which each endpoint node belongs based on the endpoint nodes in the patient cross-data set, performs index mapping between the patient pathway and its endpoint label, and generates a pathway endpoint risk label group; The label frequency statistics submodule performs a repeated count operation on all risk labels based on the pathway endpoint risk label group, records the number of times each risk label appears in the patient pathway set, and sorts them from high to low according to the number of occurrences to obtain a sorted risk label sequence; The category label determination submodule performs matching judgment on the label set corresponding to the terminal node in the patient path according to the sorted risk label sequence, selects the label item in each path that is at the front of the sorted sequence as the corresponding risk category of the path, integrates the attribution labels of all patient paths, and obtains the patient risk attribution label group.

[0010] Preferably, the follow-up structure output module includes: The node extraction submodule collects the follow-up classification nodes corresponding to each label according to the patient risk attribution label group, records the associated path number in each classification node and the number of corresponding patient path sets, determines the matching index between the risk label and the follow-up node, and obtains the label attribution node number value; The path classification submodule divides the corresponding patient path into each follow-up classification node based on the risk label based on the node number value of the label, establishes a bidirectional correspondence structure between the patient path number and the node number, extracts the path number list of each node, and obtains the node path attribution quantity value; The structure generation submodule integrates the follow-up classification nodes and the subordinate patient path numbers according to the node path attribution quantity value, outputs the classification node index, corresponding risk label and total number of paths, determines the attribution of nodes and patient paths, and generates a neurosurgery follow-up structure table.

[0011] Preferably, the basic information extraction submodule includes a basic information extraction unit for obtaining the patient's name, age, and operation date, an index recording unit for recording the chapter position of information elements in the medical record, a position comparison unit for comparing the first appearance position of key information with the total number of information elements, and an information classification unit for classifying by patient ID.

[0012] Preferably, the cycle sorting submodule includes a cycle label acquisition unit that combines outpatient follow-up, home follow-up, and reexamination follow-up, a priority comparison unit that sorts the cycles of 1 week, 1 month, and 3 months after surgery according to priority values, and an index establishment unit that establishes a rearranged sequence index table from the early postoperative period to long-term follow-up.

[0013] Preferably, the intersection comparison submodule includes a node sequence calling unit for calling the inpatient and outpatient pathways, an endpoint statistics unit for counting the number of occurrences of complication nodes and review result nodes in different pathways, a deviation screening unit for screening path pair combinations whose deviation values are less than a benchmark value, and a linked list establishment unit for establishing a correspondence between patient IDs and pathway numbers.

[0014] Compared with the prior art, the present invention has the following beneficial effects: This neurosurgery patient follow-up management system effectively addresses many issues in traditional follow-up management by establishing a systematic information processing and analysis process. In terms of information processing, the patient information collection module performs word segmentation, extraction, and element indexing on medical record text, generating standardized raw information sequence templates. This eliminates the omissions and confusion that existed in previous manual information compilation. This structured information extraction method enables unified interpretation of medical record data from different sources and formats, reducing follow-up deviations caused by differences in information interpretation and enabling medical staff to quickly and accurately grasp the core characteristics of a patient's condition. The follow-up plan generation module constructs a directed path based on the original information sequence, forming a follow-up plan path structure from the beginning to the end, breaking the traditional follow-up model of fixed time intervals. This path structure can dynamically adjust follow-up nodes based on the patient's specific condition characteristics, making the follow-up plan more closely aligned with the patient's actual recovery process. For example, for patients who experience neurological deficits after surgery, the system will add a neurological function assessment node to the path, while for patients who recover smoothly, the follow-up interval will be appropriately extended. This ensures that high-risk patients can receive timely monitoring, avoids excessive follow-up of low-risk patients, and improves the efficient utilization of medical resources. The data cross-validation module resolves inconsistencies and duplications between multi-source follow-up data by extracting pathway node sequences and filtering intersecting data. This module performs statistical analysis on intersection nodes and endpoint nodes in the patient pathway, identifying outliers and inconsistencies in the data and ensuring the reliability of the intersecting data sets included in the analysis. For example, when imaging data conflict with clinical symptom descriptions, the system prioritizes more representative data by comparing the frequency and relationships of intersecting nodes, providing an accurate basis for subsequent risk assessment. The risk level determination module matches risk labels based on endpoint nodes in the cross-data set, achieving an accurate assessment of the patient's risk level. Unlike the traditional single-indicator evaluation method, this module comprehensively considers multiple clinical characteristics of the patient and the endpoint information of the follow-up path. By sorting and matching the frequency of risk labels, it can identify multiple potential risk factors for the patient. For example, for patients after cerebral aneurysm surgery, the system will not only pay attention to the risk label of aneurysm recurrence, but also combine labels such as blood pressure control and signs of vascular spasm to form a comprehensive risk attribution label group, allowing medical staff to formulate targeted intervention measures. The follow-up structure output module divides patient pathways into corresponding follow-up classification nodes, establishing a clear follow-up attribution structure and providing an intuitive visualization tool for neurosurgery follow-up management. The generated follow-up structure table clearly displays the distribution of follow-up nodes for patients at different risk levels, helping medical staff quickly identify patient groups requiring special attention and optimize the arrangement and execution of follow-up work. Furthermore, this structured follow-up output method facilitates information sharing between different departments and medical staff, reducing communication costs and improving the collaborative efficiency of follow-up management. Through systematic processing of follow-up information, dynamic planning, precise risk assessment and structured result output, the system has comprehensively improved the scientificity and efficiency of follow-up management of neurosurgical patients, enabling follow-up work to more closely meet the individual needs of patients and better adapt to the complex and changeable characteristics of neurosurgical diseases. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 This is a working principle diagram of the neurosurgery patient follow-up management system of the present invention; Figure 2 It is a flow chart of the data structure relationship of each module of the system; Figure 3 This is the flowchart of the patient information collection module; Figure 4 A flowchart for the refinement of the basic information extraction submodule; Figure 5 Flowchart of the refined intersection comparison submodule. DETAILED DESCRIPTION

[0016] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0017] See also Figure 1-Figure 5 The present invention provides a neurosurgery patient follow-up management system, which includes: a patient information collection module, a follow-up plan generation module, a data cross-validation module, a risk level determination module, and a follow-up structure output module. The specific implementation method is as follows: The patient information collection module obtains the medical record text from the follow-up data of neurosurgery patients, performs word segmentation on the medical record text to extract the information element index, and maps the information elements to generate a permutation sequence by determining the first appearance position and frequency of key information, thereby constructing the patient's original information sequence template.

[0018] The follow-up plan generation module sorts the information elements based on the patient's original information sequence template, establishes a directed information path from the beginning to the end according to the sorting results, collects the connection node numbers and adjacent node relationships of all edges in the path, and constructs the follow-up plan path structure.

[0019] The data cross-validation module extracts the patient path node sequence based on the follow-up plan path structure, compares the intersection nodes and counts the frequency of the end nodes, and obtains the patient cross-data set after screening the cross-data.

[0020] The risk level determination module collects the risk labels corresponding to the endpoint nodes in the patient cross-data set, sorts them by frequency of occurrence, and matches the path endpoint labels to obtain the patient risk attribution label group.

[0021] The follow-up structure output module counts the follow-up classification nodes to which each label belongs according to the patient risk attribution label group, divides the patient pathway into the corresponding node, establishes the node and patient pathway follow-up attribution relationship structure, and generates the neurosurgery follow-up structure table.

[0022] Example 1:

[0023] In a neurosurgery patient follow-up management system, the patient original information sequence template includes an information element index arrangement structure, a key information sequence fragment set, and a key information frequency weight model; the follow-up plan path structure includes an information element hierarchical mapping relationship, a directed path node chain, and a node connection relationship set; the patient intersection data set includes a path intersection node set, an endpoint node occurrence frequency distribution, and an intersection node screening result; the patient risk attribution label group includes a risk label frequency ranking, a path endpoint risk label mapping, and an attribution grade label matching result; and the neurosurgery follow-up structure table includes a classification node identifier, a patient path grouping result, and a follow-up attribution relationship mapping result.

[0024] The basic information extraction submodule within the patient information collection module retrieves medical records from neurosurgery patient follow-up data and performs element-level word segmentation on the records. During the word segmentation process, the medical record text is segmented according to standard vocabulary in the medical terminology database, breaking down continuous text content into independent elements. For example, "The patient underwent craniotomy for glioma and developed a headache 3 days after surgery" is segmented into "glioma," "craniotomy," "3 days after surgery," and "headache," among other elements. After extracting each patient's set of information elements, the index position of each element in the medical record is recorded. The index position is represented by the paragraph number and character offset in the medical record text. For example, if an information element is located between characters 12 and 18 in paragraph 5, its index position is marked as "5-12-18." The relationship between the first occurrence position of key information in the information element list and the number of information elements is compared. Key information includes disease diagnosis, surgical method, and postoperative complications. The ratio of the first occurrence position to the total number of information elements is calculated to determine the proportion of key information in the overall information. When classifying by patient, all information elements and their related data are integrated according to the patient's unique identification to form a key information element index distribution result based on the patient.

[0025] The medical history record construction submodule extracts the information fragments corresponding to the key information in the original text of the medical record based on the index distribution results of the key information elements. For each key information, the start and end range in the original text is determined according to its index position, and the complete sentence or paragraph within the range is intercepted as the information fragment. For example, the text interval corresponding to the index position of the key information "postoperative infection" is "On the 5th day after surgery, the patient's body temperature rose to 38.9℃, and the blood routine showed an increase in white blood cell count, and was diagnosed with postoperative infection." This paragraph of text is intercepted as the corresponding information fragment. After intercepting the information elements based on the index interval of the key information in the medical record, an information fragment set is constructed according to the position of each key information intercepted information element, and information fragments with similar positions are classified into the same subset. The information elements are reorganized based on the patients to whom they belong, and all information fragments of the same patient are arranged in chronological order or logical relationship to obtain a set of original fragments of key information.

[0026] The information template generation submodule calculates the frequency of occurrence of all key information elements based on the original set of key information segments. During this statistical process, repeated occurrences of the same key information for the same patient are combined and counted, while the same key information for different patients is counted separately and aggregated. Based on the original order of the key information in the medical record, the information segments are rearranged, preserving the order of each information element in the original text to avoid disrupting the timeline or logical chain. Multiple original segments of key information for the same patient are concatenated according to their first occurrence, with separators added between each segment to distinguish information from different sources. The sum of the weights of the patient's original information is calculated. The weights are pre-set based on the importance of the key information; for example, a disease diagnosis is weighted higher than a symptom description, and surgical information is weighted higher than general examination results. The weights of all information elements for the same patient are summed to obtain the sum. The resulting information weights for each patient are then categorized and integrated. Patients with similar summed weights are grouped together, ultimately generating a patient original information sequence template.

[0027] In the basic information extraction submodule, the basic information extraction unit obtains basic data such as the patient's name, age, and surgery date. When extracting names, the name field is matched against the ID number or medical insurance card information to ensure consistency with the patient's actual identity. Age is automatically calculated based on the birth date and the current date, accurate to the nearest year. The surgery date is based on the start and end times in the surgical record and recorded in the "YYYY-MM-DDHH:MM:SS" format. The index recording unit records the location of information elements in the medical record, including sections such as chief complaint, present medical history, past medical history, physical examination, auxiliary examinations, diagnosis, treatment plan, and postoperative notes. Each information element is labeled with the specific section name and sequence number. The position comparison unit compares the first occurrence of key information with the total number of information elements, calculating the difference and ratio between the two to analyze the relative position of key information in the information sequence. The information classification unit performs classification based on the patient ID, a unique identifier assigned by the system. All information elements, index positions, weights, and other data related to the patient are associated with the corresponding patient ID, forming a complete patient information file.

[0028] Example 2:

[0029] The period sorting submodule within the follow-up plan generation module uses a template of the original patient information sequence and the period labels of each patient information element node to compare and sort all information element nodes according to their period label priority values. Period labels encompass various time node identifiers related to patient follow-up, which are directly linked to the follow-up tasks required at different stages. During the comparative sorting process, information element nodes are ranked according to the priority values of their period labels. Priority values are determined based on the urgency and importance of the follow-up tasks, with nodes with higher priority values positioned higher in the sorting. Patient information elements are rearranged from initial to final, with the initial position corresponding to information about the patient's postoperative period or the initial stage of treatment, and the final position corresponding to the endpoint of long-term follow-up. The rearrangement process maintains the inherent logical connections between information elements. A rearranged sequence index table is established, recording the post-rearrangement position number of each information element node and its corresponding original information element identifier. This index table allows for rapid location of the original information corresponding to each rearranged node, thereby obtaining the information element sorting index value.

[0030] The plan construction submodule retrieves the set of adjacent nodes in the rearranged sequence of information elements based on the sorting index value of the information elements. Each set of adjacent nodes consists of two consecutive nodes in the rearranged sequence. Each set contains the identifiers of the two nodes and their position information in the rearranged sequence. Each pair of adjacent nodes is numbered in a numerical sequence, starting from the starting node and incrementing in sequence. The connection direction is also recorded, represented by an arrow symbol (e.g., "A→B" for node A pointing to node B), to clarify the order of the nodes. The offset strength value of the directed information path is calculated. The offset strength value is determined based on the logical connection between adjacent nodes. The closer the connection, the smaller the offset strength value, and vice versa. The calculation refers to the correlation parameters of the information represented by the nodes. The structural edge information, including node number, connection direction, and offset strength value, is integrated based on the connection relationships of all nodes in the path. This information is then arranged in order according to the path's direction to generate the information directed path follow-up data.

[0031] The node structure extraction submodule collects the node numbers and adjacent node pair relationships in all connecting edges based on the information directed path follow-up data. The node numbers are directly extracted from the path follow-up data, and the adjacent node pair relationships are determined by the starting node and ending node of the connecting edge. For example, if the starting node corresponding to the connecting edge is 1 and the ending node is 2, the adjacent node pair relationship is expressed as "1-2". A node mapping table is constructed based on the adjacent structural relationship. The mapping table is in the form of a two-dimensional table, with rows and columns representing node numbers respectively. The cell content in the table records whether there is an adjacent relationship between the two corresponding nodes and the connection direction. The types of the preceding and following relationships and the connection direction between each information element are stored. The preceding and following relationship types include causal relationships, temporal relationships, and subordinate relationships. The connection direction is still marked with an arrow symbol. This information is classified and stored through a special storage unit to finally obtain the follow-up plan path structure.

[0032] In the cycle sorting submodule, the cycle label acquisition unit combines the cycle labels of outpatient follow-up, home follow-up, and reexamination follow-up. The outpatient follow-up label corresponds to the follow-up activities that patients undergo at the hospital outpatient clinic, including information such as the outpatient department and appointment time; the home follow-up label corresponds to the follow-up that patients receive at home, covering home care, remote monitoring, etc.; the reexamination follow-up label corresponds to the various reexamination items that patients undergo, including the type of reexamination examination, time interval, etc. The priority comparison unit sorts the cycles of 1 week, 1 month, and 3 months after surgery based on the priority value. The priority value of 1 week after surgery is the highest, followed by 1 month after surgery, and the lowest is 3 months after surgery. The corresponding cycle labels are arranged in this priority order. The index establishment unit establishes a rearranged sequence index table from the early postoperative period to long-term follow-up. The early postoperative period corresponds to the follow-up node within 1 week after surgery, and the long-term follow-up corresponds to the follow-up node six months or more after surgery. The index table records in detail the node identification, priority value, and position in the rearranged sequence of each follow-up stage. This index table can clearly present the time progression order of the follow-up plan.

[0033] The patient raw information sequence template includes an information element index arrangement structure, a key information sequence segment set, and a key information frequency weighting model. The information element index arrangement structure records the index position and arrangement order of information elements; the key information sequence segment set consists of text segments related to key information; and the key information frequency weighting model assigns corresponding weight parameters based on the frequency of occurrence of key information. The follow-up plan pathway structure includes information element hierarchical mapping relationships, directed path node chains, and a node connection relationship set. The information element hierarchical mapping relationships reflect the correspondence between information elements at different levels; the directed path node chain consists of sequentially arranged nodes, reflecting the direction of the pathway; and the node connection relationship set records the connection methods and associated attributes between all nodes. The patient intersection data set includes a path intersection node set, an endpoint node frequency distribution, and intersection node screening results. The path intersection node set is the set of nodes commonly included in different patient pathways; the endpoint node frequency distribution counts the number and proportion of occurrences of each endpoint node across all pathways; and the intersection node screening results are the intersection node information that meets the criteria after screening. The patient risk attribution label set includes risk label frequency rankings, pathway endpoint risk label mappings, and attribution level label matching results. The risk label frequency ranking is based on the frequency of risk labels, from high to low. Pathway endpoint risk label mapping establishes the correspondence between pathway endpoints and risk labels. The attribute level label matching result shows the matching between risk labels and attribute level labels. The neurosurgery follow-up structure table includes classification node identifiers, patient pathway grouping results, and follow-up attribute relationship mapping results. The classification node identifier uniquely marks the follow-up classification node; the patient pathway grouping result records the group to which each patient pathway belongs; and the follow-up attribute relationship mapping result reflects the attribute association between nodes and patient pathways.

[0034] Example 3:

[0035] The path extraction submodule in the data cross-validation module collects node sequences from any two patient pathways based on the follow-up plan pathway structure. The node sequence consists of nodes arranged in chronological order or logical relationship in the follow-up plan pathway structure, and each node contains corresponding information elements and numbers. The node number information under each path is extracted in turn and a patient node mapping set is established. The node mapping set is stored in the form of key-value pairs, where the key is the node number and the value is the information element content corresponding to the node. The patient identifier and path length parameter of each path are marked. The patient identifier is the patient's unique code in the system, and the path length parameter is the total number of nodes contained in the path. Through the above operations, the patient path node number value is obtained.

[0036] The intersection comparison submodule calls the node number sequence of any two patient paths based on the patient path node number value. An intersection comparison operation is performed on the node sets of the two paths. The intersection comparison is achieved by comparing the node numbers in the two sets one by one, and filtering out the node numbers that exist in both sets. All the terminal node numbers in the intersection are extracted. The terminal node is the last node in the path, and its number is at the end of the path sequence. The number of times this type of node appears in different paths is counted separately. When counting, the number is classified by path type. The same node is counted only once if it appears multiple times in the same path. The deviation value of the intersection path endpoint is calculated using the following formula:

[0037] in, Indicates the deviation value of the intersection path end point, represents the number of occurrences of the terminal node in the first patient pathway, represents the number of occurrences of the same endpoint node in the second patient pathway, Indicates taking and The maximum value in the denominator is added by 1 to avoid the situation where the denominator is 0. The calculated deviation value is compared with the data cross-judgment benchmark value one by one. The data cross-judgment benchmark value is a pre-set fixed value. The setting of the benchmark value needs to be combined with the clinical characteristics of neurosurgical diseases and the characteristics of follow-up data to ensure that it can effectively screen out pathway pairs with high consistency and provide a reliable basis for subsequent risk assessment. Specifically, the benchmark value can be determined based on the distribution of pathway deviations within the normal range in historical follow-up data. By analyzing the follow-up pathways of a large number of past patients, the distribution of deviation values of the end nodes between different pathway pairs is statistically analyzed, and the deviation value range of most normal pathway pair combinations is used as a reference, from which a reasonable value is selected as the benchmark value. For example, when analyzing the historical follow-up data of patients with common neurosurgical diseases such as glioma and cerebral hemorrhage, if it is found that the endpoint deviation values of the vast majority of pathway pair combinations without significant data contradictions are concentrated in a certain interval, then the upper limit of the interval can be set as the benchmark value. At the same time, the benchmark value also needs to be appropriately adjusted considering the characteristics of different types of neurosurgical diseases. For disease types with complex conditions and large differences in follow-up data, the benchmark value can be appropriately increased to include more valuable cross-data. For disease types with relatively stable conditions and high data consistency, the benchmark value can be appropriately lowered to strictly screen cross-data, thereby ensuring that the screened patient cross-data set can accurately reflect the patient's actual condition characteristics, provide reliable data support for the risk level determination module, and ultimately achieve accurate risk assessment and efficient follow-up management, which is consistent with the technical effect of this system to improve the scientificity and efficiency of follow-up management through structured processing and dynamic planning. Path pair combinations with deviation values less than or equal to the benchmark value are screened to establish a set of path intersection numbers that meet the conditions. The set contains the path pair numbers that meet the conditions and the corresponding intersection node numbers.

[0038] The path screening submodule retrieves the original patient path identifiers based on the set of path intersection numbers that meet the criteria and the corresponding path combinations. The original patient path identifiers correspond one-to-one with the path numbers, allowing traceability to the corresponding patient and complete path information. The patient identifiers and path intersection node information are integrated to create a patient path relationship linked list. Each node in the linked list contains the patient identifier, path number, intersection node number, and pointers to adjacent nodes. This integration and association generates a patient intersection data set.

[0039] In the intersection comparison submodule, the node sequence call unit calls the node sequences of the inpatient and outpatient pathways. The inpatient pathway node sequence includes nodes such as admission diagnosis, surgical record, postoperative care, and discharge assessment; the outpatient pathway node sequence includes nodes such as outpatient diagnosis, examination appointment, treatment plan adjustment, and follow-up record. The endpoint statistics unit counts the number of occurrences of complication nodes and review result nodes in different pathways. Complication nodes include nodes related to postoperative infection, bleeding, and neurological dysfunction; review result nodes include nodes related to imaging examinations, laboratory tests, and functional assessments. The statistics distinguish between different complication types and review items, and record their occurrences in the inpatient and outpatient pathways separately. The deviation screening unit selects pathway pairs with deviation values less than the baseline value. During the screening process, the deviation value of each pathway pair is evaluated, and only matching pairs are retained. The linked list creation unit establishes a correspondence between patient IDs and pathway numbers. The patient ID is the patient's unique identifier, and the pathway number is the unique code for each pathway. This correspondence is stored in a linked list format, facilitating quick query of all pathways belonging to a patient and the corresponding patient information.

[0040] The original patient information sequence template includes an information element index arrangement structure, a key information sequence segment set, and a key information frequency weight model. The information element index arrangement structure records the index position and arrangement order of the information elements, with the index position represented by the paragraph and character position in the medical record text; the key information sequence segment set consists of text segments related to the key information, and the segments are arranged in order of appearance; the key information frequency weight model assigns corresponding weights based on the frequency of occurrence of the key information, with the higher the frequency, the greater the weight. The follow-up plan path structure includes an information element hierarchical mapping relationship, a directed path node chain, and a node connection relationship set. The information element hierarchical mapping relationship reflects the correspondence between information elements at different levels, such as the mapping between major disease categories and specific disease types; the directed path node chain consists of nodes connected in a direction, reflecting the order of the follow-up process; the node connection relationship set records the connection method and association strength between nodes. The patient cross-data set includes a path intersection node set, the frequency distribution of the end node, and the cross-node screening results. The path intersection node set is the nodes commonly included in different paths; the endpoint node frequency distribution displays the frequency of occurrence of each endpoint node in all paths in a graph; the intersection node screening result is the intersection node information after deviation screening. The patient risk attribution label group includes the risk label frequency ranking, the path endpoint risk label mapping, and the attribution level label matching result. The risk label frequency ranking is arranged from high to low according to the number of times the label appears; the path endpoint risk label mapping establishes the correspondence between the path endpoint and the risk label; the attribution level label matching result is the matching status of the risk label and the preset level label. The neurosurgery follow-up structure table includes the classification node identifier, the patient pathway grouping result, and the follow-up attribution relationship mapping result. The classification node identifier is the unique code for each follow-up classification node; the patient pathway grouping result records the grouping category to which each patient pathway belongs; the follow-up attribution relationship mapping result displays the attribution association between the node and the patient pathway in the form of a matrix.

[0041] Example 4: The risk label collection submodule in the risk level determination module collects the risk label set belonging to each endpoint node based on the endpoint nodes in the patient cross-data set. Each endpoint node corresponds to a specific clinical outcome or status, such as the "epileptic seizure 3 months after surgery" node corresponds to the "epileptic risk" label, the "imaging review suggests tumor recurrence" node corresponds to the "tumor progression risk" label, and the "neurological function score decline" node corresponds to the "dysfunctional risk" label. An index mapping is performed between the patient path and its endpoint label. During the mapping process, a label index table is created for each patient path. The table records the path number, endpoint node number and corresponding risk label. The endpoint risk label of a certain path can be directly queried through this index table to generate a path endpoint risk label group.

[0042] The label frequency statistics submodule performs a repeated count operation on all risk labels based on the path endpoint risk label group. The statistical scope covers the endpoint labels of all patient pathways. The same risk label is counted once if it appears once in different pathways, regardless of the path source or patient characteristics. The number of times each risk label appears in the patient pathway set is recorded. For example, if the "epilepsy risk" label appears in 50 pathways, it is recorded as 50 times, and if the "tumor progression risk" label appears in 30 pathways, it is recorded as 30 times. Sorting is performed from high to low based on the number of occurrences, and an ordered label sequence is formed after sorting. For example, the sorting result may be "infection risk (60 times), epilepsy risk (50 times), dysfunction risk (40 times), tumor progression risk (30 times)", resulting in a sorted risk label sequence.

[0043] The category label determination submodule performs matching judgment on the label set corresponding to the terminal node in the patient path according to the sorted risk label sequence. The terminal node of each path may correspond to multiple risk labels. For example, the node "postoperative infection with neurological decline" may be associated with both the "infection risk" and "dysfunctional risk" labels. The label item in each path that is at the front of the sorted sequence is selected as the corresponding risk category of the path. For example, if the terminal labels of a path are "infection risk" and "dysfunctional risk", and the "infection risk" position is higher in the sorted sequence, then the path belongs to the "infection risk" category. Integrate the attribution labels of all patient paths, classify the paths with the same attribution label together, and obtain the patient risk attribution label group.

[0044] The basic information extraction submodule within the patient information collection module acquires medical records from neurosurgery patient follow-up data and performs element-level word segmentation on the records. This word segmentation breaks down the medical record text into independent medical terms or phrases. For example, "A 65-year-old male patient underwent hematoma evacuation for cerebral hemorrhage and developed right limb weakness two weeks after surgery" is broken down into information elements such as "male," "65 years old," "cerebral hemorrhage," "hematoma evacuation," "two weeks after surgery," and "right limb weakness." A set of information elements is extracted for each patient, containing all relevant information elements for that patient. The index position of each information element in the medical record is recorded, represented by chapter and paragraph numbers. For example, if "cerebral hemorrhage" is located in the second paragraph of the "Present Medical History" section, it is labeled "Present Medical History - 2." The relationship between the first occurrence position of key information in the information element arrangement list and the number of information elements is compared. For key information such as "cerebral hemorrhage" and "hematoma evacuation," the ratio of the first occurrence sequence number to the total number of elements is calculated. The information elements are categorized by patient, and all information elements and index data for the same patient are integrated to obtain the index distribution of key information elements.

[0045] The medical history record construction submodule extracts the information fragments corresponding to the key information in the original text of the medical record based on the index distribution results of the key information elements. For example, if the index position of the key information "hematoma removal" points to the first paragraph of the "operation record" chapter, the content related to the operation in this paragraph is extracted as the information fragment. The information elements are intercepted based on the index interval of the key information in the medical record. The interception range is centered on the key information and includes the content related to it before and after. According to the position of each key information intercepted information element, a set of information fragments is constructed, and the information fragments with adjacent positions are grouped together. The information elements are reorganized based on the patients to which they belong, and the information fragments of the same patient are arranged in chronological order to obtain a set of original fragments of key information.

[0046] The information template generation submodule counts the frequency of occurrence of all key information based on the set of original fragments of key information. For example, if "hematoma removal" appears 80 times in 100 medical records, the frequency value is 80. Based on the original order of the key information in the medical record text, the information fragment set is rearranged, and the chronological order of the information elements is maintained during the rearrangement. The original fragments of multiple key information in the same patient are sequentially spliced according to the position of their first appearance, and a complete patient information chain is formed after splicing. The sum of the weights of the patient's original information is calculated. The weights are set according to the importance of the information elements, such as the disease diagnosis weight is 5, the surgical information weight is 4, and the symptom description weight is 3. The weights of all information elements of the same patient are added together to obtain the total. The information weight results corresponding to each patient are combined for classification and integration. Patients with similar weight sums are grouped together to obtain the patient's original information sequence template.

[0047] The patient's original information sequence template includes an information element index arrangement structure, a key information sequence segment set, and a key information frequency weight model. The information element index arrangement structure records the index position and arrangement order of information elements; the key information sequence segment set consists of text segments corresponding to key information; and the key information frequency weight model contains the occurrence frequency of each key information and its corresponding weight. The follow-up plan path structure includes information element hierarchical mapping relationships, directed path node chains, and a node connection relationship set. The information element hierarchical mapping relationship reflects the correspondence between information elements at different levels; the directed path node chain consists of nodes connected by direction; and the node connection relationship set records the connection method between nodes. The patient intersection data set includes a path intersection node set, an endpoint node frequency distribution, and intersection node screening results. The path intersection node set contains nodes shared by different paths; the endpoint node frequency distribution records the number of occurrences of each endpoint node; and the intersection node screening results are the filtered intersection nodes. The patient risk attribution label set includes risk label frequency rankings, path endpoint risk label mappings, and attribution level label matching results. The risk label frequency ranking lists labels by number of occurrence; the pathway endpoint risk label mapping records the correspondence between pathway endpoints and labels; and the attribution level label matching results show the matching between labels and levels. The neurosurgery follow-up structure table contains classification node identifiers, patient pathway grouping results, and follow-up attribution relationship mapping results. The classification node identifier is the unique code of the node; the patient pathway grouping results show the path grouping; and the follow-up attribution relationship mapping results record the attribution relationship between nodes and pathways.

[0048] Example 5: The node extraction submodule in the follow-up structure output module collects the follow-up classification nodes corresponding to each label according to the patient risk attribution label group. The follow-up classification nodes are pre-set based on the type of risk label. Different risk labels correspond to different classification nodes, and each classification node has its own specific identification information. Record the number of associated path numbers and the number of corresponding patient path sets in each classification node. The path number corresponds to the patient path one-to-one, and the quantity statistics are based on the unique count of the path number. Determine the matching index between the risk label and the follow-up node. The matching index is established by associating the label with the node attributes, so that each label can accurately point to the corresponding node. Through the above operations, the label attribution node number value is obtained.

[0049] The path classification submodule is based on the node number value of the label attribution and divides the corresponding patient path into each follow-up classification node according to the risk label. During the division process, all patient paths with a certain risk label are centrally assigned to the node corresponding to the label based on the matching index between the label and the node. A bidirectional correspondence structure is established between the patient path number and the node number. In this structure, the corresponding node number can be queried through the path number, and all the included path numbers can also be queried through the node number. The list of path numbers belonging to each node is extracted, and the list is arranged in the order of the generation time of the path number. Through the above processing, the number value of the node path attribution is obtained.

[0050] The structure generation submodule integrates the follow-up classification nodes and their subordinate patient pathway numbers through structural mapping based on the node path attribution count. This integration takes the node number as the core, associates the corresponding risk label with all subordinate pathway numbers, and forms a hierarchical structure. The module outputs the classification node index, the corresponding risk label, and the total number of pathways. The classification node index uniquely identifies the node in the system, and the total number of pathways is the sum of the patient pathways contained within that node. The attribution of nodes and patient pathways is determined, clarifying which node each pathway belongs to and which pathways each node contains. Through this integration and determination, a neurosurgery follow-up structure table is generated.

[0051] The risk label collection submodule in the risk level determination module collects the risk label set associated with each endpoint node based on the endpoint nodes in the patient cross-data set. The endpoint node reflects the final status or result of the patient follow-up process. Each endpoint node is associated with a corresponding risk label based on the information content it contains. An endpoint node may be associated with multiple risk labels. An index mapping is performed between the patient pathway and its endpoint label. The mapping relationship is achieved by recording the association between the path number and the label, so that each path can correspond to all the risk labels of its endpoint node. Through the above collection and mapping, a path endpoint risk label group is generated.

[0052] The label frequency statistics submodule performs a repetition count operation on all risk labels based on the path endpoint risk label group. The statistics cover all patient pathway endpoint risk labels, and each risk label is counted as one occurrence if it appears once in different pathways. The number of times each risk label appears in the patient pathway set is recorded to the nearest digit. Labels are sorted from highest to lowest based on the number of occurrences, forming an ordered sequence. Labels positioned higher in the sequence have a higher number of occurrences. This statistical and sorting process yields a sorted risk label sequence.

[0053] The category label determination submodule performs a matching judgment on the label set corresponding to the terminal node in the patient path based on the sorted risk label sequence. During the judgment process, the terminal label set of each path is compared with the sorted sequence to find the label in the set that is at the top of the sequence. The label item at the top of the sorted sequence in each path is selected as the corresponding risk category of the path. Each path only determines one risk category. The attribution labels of all patient paths are integrated, and paths belonging to the same risk category are centrally managed. Through the above judgment and integration, the patient risk attribution label group is obtained.

[0054] The path extraction submodule within the data cross-validation module collects node sequences from any two patient pathways based on the follow-up plan pathway structure. The node sequence consists of all nodes in the pathway, arranged in order; each node contains a specific information element and a number. The node number information for each pathway is extracted sequentially, and a patient node mapping set is established. In this mapping set, the node number is associated with the corresponding information element. Each pathway is labeled with the patient ID and path length parameter. The patient ID is the patient's unique identity code, and the path length parameter is the total number of nodes in the path. Through this collection and labeling, the node number values of the patient pathway are obtained.

[0055] The intersection comparison submodule calls the node number sequence of any two patient paths based on the patient path node number value. Perform an intersection comparison operation on the node sets of the two paths. The comparison is achieved by checking the node numbers one by one, and the node numbers that exist in both sets are screened out. Extract all the terminal node numbers in the intersection. The terminal node is the last node in the path, and its number is at the end of the sequence. Count the number of times this type of node appears in different paths, and the number statistics are performed separately for each path. Calculate the deviation value of the intersection path terminal, and the deviation value is obtained by a preset calculation method. Compare it one by one with the data intersection judgment benchmark value, and the benchmark value is a fixed reference value. Screen the path pair combination with a deviation value less than or equal to the benchmark value. The combination contains the numbers of the two paths and the intersection node information. Establish a set of path intersection numbers that meet the conditions, and store all qualified path pairs and related node information in the set.

[0056] The path screening submodule retrieves the original patient path identifiers based on the set of path intersection numbers that meet the criteria and the corresponding path combinations. The original patient path identifiers correspond one-to-one with the paths, allowing complete path information to be retrieved through the identifiers. The patient identifiers and path intersection node information are integrated to create a patient path relationship linked list. In this linked list, each element is sequentially connected according to its association, forming a chain structure. Through this query and integration, a patient intersection data set is generated.

[0057] The patient raw information sequence template includes an information element index arrangement structure, a key information sequence segment set, and a key information frequency weight model. The information element index arrangement structure records the index position and arrangement order of information elements; the key information sequence segment set consists of text segments related to key information in the medical record; and the key information frequency weight model contains the occurrence frequency of key information and corresponding weight parameters. The follow-up plan path structure includes information element hierarchical mapping relationships, directed path node chains, and a node connection relationship set. The information element hierarchical mapping relationships reflect the correspondence between information elements at different levels; the directed path node chain consists of nodes connected in a certain direction; and the node connection relationship set records the connection mode and association type between nodes. The patient intersection data set includes a path intersection node set, an endpoint node frequency distribution, and intersection node filtering results. The path intersection node set represents common nodes in different paths; the endpoint node frequency distribution records the number of occurrences and distribution of each endpoint node; and the intersection node filtering results are the filtered intersection node information. The patient risk attribution label set includes risk label frequency rankings, path endpoint risk label mappings, and attribution level label matching results. The risk label frequency ranking lists risk labels by number of occurrence; the pathway endpoint risk label mapping records the correspondence between pathway endpoints and risk labels; and the attribution level label matching results show the matching between risk labels and attribution levels. The neurosurgery follow-up structure table contains the classification node identifier, patient pathway grouping results, and follow-up attribution relationship mapping results. The classification node identifier is the unique code for the follow-up classification node; the patient pathway grouping results record the grouping of each pathway; and the follow-up attribution relationship mapping results reflect the attribution association between nodes and pathways.

[0058] It should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprises," "comprising," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that includes a list of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, method, article, or apparatus.

[0059] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A neurosurgery patient follow-up management system, characterized in that: The system comprises: The patient information collection module obtains the medical record text from the follow-up data of neurosurgery patients, extracts information element indexes through word segmentation, determines the initial position and frequency of key information, maps and generates the information element arrangement sequence, and constructs the patient's original information sequence template; A follow-up plan generation module, based on the patient's original information sequence template, sorts the information elements, establishes a directed information path from the beginning to the end according to the sorting result, collects the connection node numbers and adjacent node relationships of all edges in the path, and constructs a follow-up plan path structure; A data cross-validation module extracts a patient path node sequence according to the follow-up plan path structure, compares the intersection nodes and counts the frequency of the end nodes, and screens the cross-data to obtain a patient cross-data set; The risk level determination module collects risk labels of the endpoint nodes based on the endpoint nodes in the patient cross-data set, sorts them by frequency of occurrence, and matches the path endpoint labels to obtain a patient risk attribution label group; The follow-up structure output module counts the follow-up classification nodes to which each label belongs based on the patient risk attribution label group, divides the patient path into the corresponding nodes, establishes the node and patient path follow-up attribution relationship structure, and generates a neurosurgery follow-up structure table.

2. The neurosurgery patient follow-up management system according to claim 1, characterized in that: The patient original information sequence template includes an information element index arrangement structure, a key information sequence fragment set, and a key information frequency weight model; the follow-up plan path structure includes an information element hierarchical mapping relationship, a directed path node chain, and a node connection relationship set; the patient cross-data set includes a path intersection node set, an endpoint node occurrence frequency distribution, and an intersection node screening result; the patient risk attribution label group includes a risk label frequency ranking, a path endpoint risk label mapping, and an attribution grade label matching result; and the neurosurgery follow-up structure table includes a classification node identifier, a patient path grouping result, and a follow-up attribution relationship mapping result.

3. The neurosurgery patient follow-up management system according to claim 1, characterized in that: The patient information collection module includes: The basic information extraction submodule obtains the medical records of neurosurgery patients' follow-up data, performs information element-level word segmentation on the medical records, extracts the information element set of each patient, records the index position of each information element in the medical record, compares the relationship between the first occurrence position of key information in the information element arrangement list and the number of information elements, classifies them by patient, and obtains the key information element index distribution results; The medical history record construction submodule extracts the information fragments corresponding to the key information in the original medical record text based on the key information element index distribution result, intercepts the information elements based on the index interval of the key information in the medical record, constructs a set of information fragments based on the position of each key information intercepted information element, and reorganizes the information fragments based on the patient to which the information element belongs to obtain a set of original key information fragments; The information template generation submodule counts the occurrence frequency values of all key information based on the set of key information original fragments, and performs position rearrangement processing on the information fragment set based on the original order of the key information in the medical record text. The original fragments of multiple key information in the same patient are sequentially spliced according to the first appearance position, and the sum of the patient's original information weights is calculated. The information weight results corresponding to each patient are combined for classification and integration to obtain the patient's original information sequence template.

4. The neurosurgery patient follow-up management system according to claim 1, characterized in that: The follow-up plan generation module includes: The period sorting submodule compares and sorts all information element nodes according to their period label priority values based on the original patient information sequence template and the period label of each patient information element node, rearranges the positions of the patient information elements in order from the beginning to the end, establishes a rearranged sequence index table, and obtains the information element sorting index value; The plan construction submodule obtains the set of adjacent nodes in the information element rearrangement sequence according to the information element sorting index value, numbers each pair of adjacent nodes and records their connection direction, calculates the information directed path offset strength value, combines the connection relationship of all nodes in the path, integrates the structural edge information, and generates the information directed path follow-up data; The node structure extraction submodule collects the node numbers and adjacent node pair relationships in all connecting edges according to the directed path follow-up data of the information, constructs a node mapping table based on the adjacent structural relationship, stores the forward and backward association relationship types and connection directions between each information element, and obtains the follow-up plan path structure.

5. The neurosurgery patient follow-up management system according to claim 1, characterized in that: The data cross-validation module includes: A path extraction submodule, based on the follow-up plan path structure, collects a chronological sequence of nodes in the complete follow-up paths of any two different patients, where each patient's path contains all node information from the initial follow-up to the current stage, extracts the node number information under each path in turn and establishes a patient node mapping set, labels the patient identifier and path length parameter of each path, and obtains the patient path node number value; The intersection comparison submodule calls the node number sequences of any two patient paths according to the patient path node number values, performs an intersection comparison operation on the node sets of the two paths, extracts all the terminal node numbers in the intersection, counts the number of times such nodes appear in different paths, calculates the deviation value of the intersection path terminal, compares it with the data intersection judgment benchmark value one by one, selects the path pair combinations with the deviation value less than or equal to the benchmark value, and establishes the path intersection number set that meets the conditions; The path screening submodule queries the original patient path identifier according to the set of path intersection numbers that meet the conditions and the path combination corresponding to the number, integrates the patient identifier and the path intersection node information, establishes a patient path relationship list, and generates a patient intersection data set.

6. The neurosurgery patient follow-up management system according to claim 1, characterized in that: The risk level determination module includes: The risk label collection submodule collects the risk label set to which each endpoint node belongs based on the endpoint nodes in the patient cross-data set, performs index mapping between the patient pathway and its endpoint label, and generates a pathway endpoint risk label group; The label frequency statistics submodule performs a repeated count operation on all risk labels based on the pathway endpoint risk label group, records the number of times each risk label appears in the patient pathway set, and sorts them from high to low according to the number of occurrences to obtain a sorted risk label sequence; The category label determination submodule performs matching judgment on the label set corresponding to the terminal node in the patient path according to the sorted risk label sequence, selects the label item in each path that is at the front of the sorted sequence as the corresponding risk category of the path, integrates the attribution labels of all patient paths, and obtains the patient risk attribution label group.

7. The neurosurgery patient follow-up management system according to claim 1, characterized in that: The follow-up structure output module includes: The node extraction submodule collects the follow-up classification nodes corresponding to each label according to the patient risk attribution label group, records the associated path number in each classification node and the number of corresponding patient path sets, determines the matching index between the risk label and the follow-up node, and obtains the label attribution node number value; The path classification submodule divides the corresponding patient path into each follow-up classification node based on the risk label based on the node number value of the label, establishes a bidirectional correspondence structure between the patient path number and the node number, extracts the path number list of each node, and obtains the node path attribution quantity value; The structure generation submodule integrates the follow-up classification nodes and the subordinate patient path numbers according to the node path attribution quantity value, outputs the classification node index, corresponding risk label and total number of paths, determines the attribution of nodes and patient paths, and generates a neurosurgery follow-up structure table.

8. The neurosurgery patient follow-up management system according to claim 3, characterized in that: The basic information extraction submodule includes a basic information extraction unit for obtaining the patient's name, age, and operation date, an index recording unit for recording the chapter position of information elements in the medical record, a position comparison unit for comparing the first appearance position of key information with the total number of information elements, and an information classification unit for classifying by patient ID.

9. The neurosurgery patient follow-up management system according to claim 4, characterized in that: The cycle sorting submodule includes a cycle label acquisition unit that combines outpatient follow-up, home follow-up, and reexamination follow-up, a priority comparison unit that sorts the cycles of 1 week, 1 month, and 3 months after surgery according to the priority value, and an index establishment unit that establishes a rearranged sequence index table from the early postoperative period to long-term follow-up.

10. The neurosurgery patient follow-up management system according to claim 5, characterized in that: The intersection comparison submodule includes a node sequence calling unit for calling the inpatient and outpatient pathways, an endpoint statistics unit for counting the number of occurrences of complication nodes and review result nodes in different pathways, a deviation screening unit for screening pathway pair combinations with deviation values less than a benchmark value, and a linked list establishment unit for establishing a correspondence between patient IDs and pathway numbers.

Images