A remote multidisciplinary collaborative decision-making method and system for emergency treatment of cerebral hemorrhage

Abstract

This invention discloses a remote multidisciplinary collaborative decision-making method and system for emergency cerebral hemorrhage. It synchronously integrates patient imaging data and clinical signs to construct a dynamic diagnostic benchmark. Based on this, it analyzes the differences in assessments across departments, identifies the locations of disagreements, extracts effective consultation features, and locates target feature items. Departmental treatment labels are defined based on these target feature items. Decision complexity weights are generated by tracking the patterns of departmental opinion changes, and a collaborative decision-making graph is established. The collaborative decision-making graph is combined with the dynamic diagnostic benchmark and complexity weights to complete the solution difference completion. Consistency verification generates a treatment conflict index and determines the dispute label table. Finally, arbitration rules are established based on historical arbitration patterns, weighted voting matching is completed, and a multidisciplinary collaborative treatment plan with multiple evolution path warnings is output. This provides quantifiable and executable multidisciplinary collaborative decision support for emergency remote consultations for cerebral hemorrhage.

Description

Technical Field

[0001] This invention relates to the field of emergency medical decision support technology, and in particular to a remote multidisciplinary collaborative decision-making method and system for emergency treatment of cerebral hemorrhage. Background Technology

[0002] Intracerebral hemorrhage is one of the most deadly and disabling critical illnesses in the emergency department, characterized by rapid disease progression and an extremely limited treatment window. Clinical treatment typically requires the simultaneous involvement of multiple specialties, including neurosurgery, neurology, critical care medicine, and radiology, to form a unified diagnostic and treatment judgment. However, in remote consultation scenarios, patient imaging data and continuous monitoring vital signs data are often stored in different systems, lacking a unified temporal integration mechanism. This results in differences in the information base obtained by participating departments, making it difficult to conduct collaborative judgment within the same data framework. Existing consultation processes lack quantitative tracking methods for changes in departmental opinions, failing to identify frequently adjusted departments and fundamental interdisciplinary conflicts at critical stages of the disease. When conflicts arise from multiple departments, there is no way to structurally assess the intensity of the conflict, nor is there an effective mediation path based on historical arbitration patterns. Ultimately, this leads to low efficiency in forming collaborative decision-making solutions and difficulty in quantifying the degree of consensus among departments.

[0003] Therefore, a method is urgently needed to solve at least one of the above problems. Summary of the Invention

[0004] This invention discloses a method and system for remote multidisciplinary collaborative decision-making in the emergency treatment of cerebral hemorrhage, aiming to solve problems such as the difficulty in integrating multi-source heterogeneous data, the inability to structurally track departmental disagreements, and the lack of a systematic arbitration mechanism for interdisciplinary diagnostic conflicts in remote multidisciplinary consultations. This invention constructs a unified temporal fusion benchmark, quantifies the patterns of departmental opinion changes, and establishes a conflict intensity assessment and history-driven arbitration framework. Ultimately, it generates a multidisciplinary collaborative treatment plan that includes a consensus-driven main opinion and early warnings of multiple evolution paths, providing structured and quantifiable decision support for remote consultations in the emergency treatment of cerebral hemorrhage.

[0005] The first aspect of this invention proposes a remote multidisciplinary collaborative decision-making method for emergency treatment of cerebral hemorrhage, comprising the following steps:

[0006] Acquire patient imaging data and clinical sign data, and synchronously fuse the patient imaging data and clinical sign data to construct a dynamic diagnostic benchmark;

[0007] The dynamic diagnostic benchmark is used to analyze the fluctuation of the disease and identify the pattern of opinion divergence. Based on the pattern of opinion divergence, the excessively consistent intervals are identified and the heterogeneous opinions are retained to form effective consultation features. The disease feature spectrum is analyzed to locate the target feature items of the effective consultation features.

[0008] Based on the target feature items, departmental diagnosis and treatment labels are defined. Departments with high change frequency in the departmental diagnosis and treatment labels are marked as core change departments to generate decision complexity weights. The departmental diagnosis and treatment labels and the decision complexity weights are used to perform partitioning and calibration on the dynamic diagnostic benchmark to establish a collaborative decision graph.

[0009] The collaborative decision-making graph and the dynamic diagnostic benchmark are combined with the decision complexity weight to complete the scheme difference and form a modified scheme set. The consistency check of the modified scheme set is performed to form a diagnosis-treatment conflict index. The dispute labeling table is determined based on the disciplinary opinion deviation distribution of the diagnosis-treatment conflict index.

[0010] Arbitration rules are established by mapping the disputed labeling table with the departmental diagnosis and treatment labels. Based on the arbitration rules, a weighted voting matching is performed on the disputed labeling table to generate decision validity. The low consensus interval of the decision validity is used to mark multiple evolution path warnings and incorporate them into the output of multidisciplinary collaborative diagnosis and treatment plans.

[0011] A second aspect of this invention proposes an emergency remote multidisciplinary collaborative decision-making system for cerebral hemorrhage, comprising:

[0012] A benchmark construction unit is used to acquire patient image data and clinical sign data, and synchronously fuse the patient image data and clinical sign data to construct a dynamic diagnostic benchmark.

[0013] The feature recognition unit is used to analyze the fluctuation of the disease through the dynamic diagnostic benchmark to identify the pattern of opinion divergence, identify the excessively consistent intervals based on the pattern of opinion divergence, reduce the weight of heterogeneous opinions to form effective consultation features, and perform disease feature spectrum analysis on the effective consultation features to locate target feature items.

[0014] The graph construction unit is used to delineate departmental diagnosis and treatment labels based on the target feature items, assign high-frequency change departments in the departmental diagnosis and treatment labels as core change departments to generate decision complexity weights, and perform partitioning and calibration of the dynamic diagnostic benchmark with the departmental diagnosis and treatment labels and the decision complexity weights to establish a collaborative decision graph.

[0015] The conflict detection unit is used to combine the collaborative decision-making graph with the dynamic diagnostic benchmark and the decision complexity weight to complete the scheme difference completion to form a modified scheme set, perform consistency verification on the modified scheme set to form a diagnosis and treatment conflict index, and determine the dispute labeling table based on the disciplinary opinion deviation distribution of the diagnosis and treatment conflict index.

[0016] The arbitration output unit is used to establish arbitration rules by mapping the dispute label table with the departmental diagnosis and treatment labels, perform weighted voting matching on the dispute label table based on the arbitration rules to generate decision validity, and use the low consensus interval of the decision validity to mark multiple evolution path warnings and incorporate them into the output of multidisciplinary collaborative diagnosis and treatment plans.

[0017] The beneficial effects of this invention are reflected in the following points: 1. By temporally aligning and fusing patient imaging data with clinical sign data, a dynamic diagnostic benchmark covering multiple zones is constructed. The consistency of assessment dimensions across departments is dynamically calculated, distinguishing between two types of sudden drop patterns: initial objections from a single department and simultaneous disagreements from multiple departments, and assigning weights accordingly. Excessively consistent intervals are identified while retaining heterogeneous opinions, and effective consultation features that balance the intensity of disagreements and heterogeneity are extracted, improving the accuracy of identifying key diagnostic and treatment elements. 2. Based on target feature items, diagnostic and treatment responsibilities are assigned to each responsible department. By tracking the shrinking pattern of opinion changes in each department and combining it with continuous risk grading, a core change department table is generated. After dual weighting of historical change frequency and diagnostic and treatment contribution, a decision complexity weight is formed. Departmental diagnostic and treatment labels and decision complexity weights are jointly mapped to the diagnostic benchmark partition structure, establishing a collaborative decision-making graph that combines departmental responsibility positioning and temporal complexity characterization capabilities. 3. The segment-by-segment scheme matching identifies missing coverage areas, and differentiated completion is implemented by combining the strength of associated features and complexity weights to form a set of corrected schemes and generate a diagnosis and treatment conflict index; based on the deviation distribution of departmental opinion amplitude, multi-dimensional synchronous deviations from departments are identified, and arbitration rules are established by extracting temporal arbitration patterns from historical consultation records. Weighted voting matching is completed through internal contradiction detection and temporary weight freezing, and multiple evolution path warnings are marked for low consensus intervals. Multidisciplinary collaborative diagnosis and treatment schemes containing the main opinion and candidate evolution paths are output to improve the structured nature and clinical feasibility of remote consultation decision-making. Attached Figure Description

[0018] Figure 1 This is a flowchart illustrating a remote multidisciplinary collaborative decision-making method for emergency treatment of cerebral hemorrhage according to the present invention.

[0019] Figure 2 This is a structural block diagram of a remote multidisciplinary collaborative decision-making system for emergency treatment of cerebral hemorrhage according to the present invention. Detailed Implementation

[0020] In the following description, specific details such as particular system architectures and techniques are set forth for illustrative purposes and not for limitation, in order to provide a thorough understanding of the embodiments of this application. However, those skilled in the art will understand that this application may also be implemented in other embodiments without these specific details. In other instances, detailed descriptions of well-known systems, apparatuses, circuits, and methods have been omitted so as not to obscure the description of this application with unnecessary detail.

[0021] It should be understood that, when used in this application specification and the appended claims, the term "comprising" indicates the presence of the described features, integrals, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or a collection thereof.

[0022] References to "one embodiment" or "some embodiments" as described in this specification mean that one or more embodiments of this application include a specific feature, structure, or characteristic described in connection with that embodiment. Therefore, the phrases "in one embodiment," "in some embodiments," "in other embodiments," "in still other embodiments," etc., appearing in different parts of this specification do not necessarily refer to the same embodiment, but rather mean "one or more, but not all, embodiments," unless otherwise specifically emphasized. The terms "comprising," "including," "having," and variations thereof mean "including but not limited to," unless otherwise specifically emphasized.

[0023] The technical solutions of the embodiments of this application will be described below.

[0024] like Figure 1 As shown, this embodiment of the invention provides a remote multidisciplinary collaborative decision-making method for emergency treatment of cerebral hemorrhage, including the following steps S110-S150:

[0025] Step S110: Acquire patient imaging data and clinical sign data, and synchronously fuse the patient imaging data and clinical sign data to construct a dynamic diagnostic benchmark.

[0026] Specifically, patient imaging data and clinical signs data are acquired. Both types of data are transmitted from the patient's hospital to the remote consultation platform via an encrypted channel for retrieval by experts from participating departments. The acquisition of patient imaging data covers CT plain scan sequences, CT angiography sequences, and MRI diffusion-weighted sequences, with scans covering the skull base to the skull top. Each acquisition includes 32 to 48 axial planes. Patient imaging data is stored in DICOM format, carrying scan parameters and acquisition timestamps with second-level precision. The image acquisition time window is set within 30 minutes of the patient's arrival at the emergency department. CT angiography sequences are used to reflect intracranial vascular morphology and the source of bleeding, while diffusion-weighted sequences are used to reflect the degree of diffusion restriction of the edema zone around the hematoma. The patient imaging data modal sequences are arranged in a time-ordered set according to the acquisition timestamp. Data from multiple modal sequences with an acquisition time difference of no more than 15 minutes are considered valid data from the same batch. Clinical vital signs data include two categories: dynamic monitoring indicators and static assessment indicators. Dynamic monitoring indicators cover continuous time-series acquisition of blood pressure, heart rate, and blood oxygen saturation, with a acquisition frequency of once every 5 minutes. Blood pressure records are presented in a dual-column time sequence of systolic and diastolic blood pressure, accompanied by nursing station identification. The Glasgow Coma Scale score is assessed every 30 minutes as a dynamic assessment indicator. Coagulation function indicators include prothrombin time, activated partial thromboplastin time, and international normalized ratio. Each indicator of clinical vital signs data carries an acquisition timestamp, which uses the same time reference as the acquisition timestamp of the patient's imaging data to ensure the temporal correspondence between the two types of data.

[0027] A dynamic diagnostic benchmark is constructed by synchronously fusing patient imaging data and clinical sign data. Synchronous fusion uses the image acquisition time point as the alignment benchmark, and various indicators of clinical sign data are mapped to the corresponding image time point using a nearest neighbor interpolation strategy, with the fusion time error controlled within 2.5 minutes. Taking the CT plain scan acquisition time as an example, the corresponding systolic blood pressure record value at that time is 157 mmHg. The hematoma volume measurement value at that section and the blood pressure value at the same moment are associated and bound through synchronous fusion, forming a multi-dimensional joint feature at a single time point. For each scan batch of patient imaging data, synchronous fusion associates the clinical sign data indicator vector corresponding to the acquisition time of that batch with it. The scan batch number and batch acquisition timestamp together constitute a two-dimensional index of the fusion record. The dynamic diagnostic benchmark organizes the fusion results with the time axis as the main index, and each time point is associated with the section coordinate index of the patient imaging data and the corresponding clinical sign data indicator vector. The dynamic diagnostic benchmark is structurally divided into three sections: baseline, acute phase, and observation phase. The baseline section corresponds to the time from the patient's arrival at the emergency department to the first imaging acquisition; the acute phase covers the time from the first imaging acquisition to the completion of the first interventional procedure; and the observation phase covers the time from the completion of the first interventional procedure to the 24-hour follow-up. Each section of the dynamic diagnostic benchmark sets indicator subsets according to the assessment dimensions of four participating departments: neurosurgery, neurology, critical care medicine, and radiology. The indicator subsets for each department's assessment dimensions are extracted from the imaging features and physical signs time series within the corresponding section. The indicator subsets for different departments within each section are stored independently and do not overlap. The boundary for indicator extraction between departments is determined by the scope of each department's consultation responsibilities. At the same time point, each department reads assessment information from its own responsible indicator subset, without cross-departmental indicator cross-references.

[0028] Step S120: Use dynamic diagnostic benchmarks to analyze disease fluctuations and identify patterns of opinion divergence. Based on the patterns of opinion divergence, identify excessively consistent intervals, reduce weights to retain heterogeneous opinions, and form effective consultation features. Then, perform disease feature spectrum analysis on the effective consultation features to locate target feature items.

[0029] In some embodiments, the step of analyzing and identifying the patterns of disagreement through the dynamic diagnostic benchmark includes: extracting the time series of disease indicators from the dynamic diagnostic benchmark according to the assessment dimensions of each department to generate a time series group of departmental indicators; performing inter-departmental consistency calculation on the time series group of departmental indicators to form a consistency assessment sequence; detecting consistency drop segments in the consistency assessment sequence to identify two types of drop segments: single-department-led objection triggering and multi-departmental synchronous disagreement triggering, to form a drop segment classification label table; and assigning weights to the two types of drop segments based on the drop segment classification label table to determine the location of the disagreement and form the pattern of disagreement.

[0030] The dynamic diagnostic benchmark extracts disease indicators time series according to the assessment dimensions of each department to generate departmental indicator time series groups. Each of the four departmental assessment dimensions of the dynamic diagnostic benchmark corresponds to an independent subset of indicators: the neurosurgery dimension includes hematoma volume, midline shift distance, and hematoma expansion rate; the neurology dimension includes neurological deficit score and cerebral perfusion pressure; the intensive care unit dimension includes mean arterial pressure, blood oxygen saturation, and intracranial pressure; and the radiology dimension includes CT density value and edema band width. Taking the acute phase of the dynamic diagnostic benchmark as an example, the neurosurgery dimension has hematoma volume measurements at each of the eight time nodes from T1 to T8, with the hematoma volume increasing from 28.5 mL to 34.2 mL. Time series extraction is performed node-by-node from the corresponding partition of the dynamic diagnostic benchmark, indexed by the acquisition timestamp. The departmental indicator time series groups are organized using department number as the primary index, indicator name as the secondary index, and time node as the tertiary index. Each time series record also carries a data quality marker. Time points with missing data in the dynamic diagnostic benchmark are filled in using linear interpolation. Time series of indicators with a missing rate exceeding 30% are labeled as low quality. The time series length of the departmental indicator time series group is consistent with the time span of each partition of the dynamic diagnostic benchmark. The time series length of the baseline segment and the time series length of the observation period are consistent with the time span of the corresponding partition of the dynamic diagnostic benchmark. The time series length of the acute phase is 8 to 12 nodes.

[0031] The consistency assessment sequence is formed by calculating the inter-departmental consistency of the departmental indicator time series groups. Inter-departmental consistency is measured by the consistency rate of indicator change direction among departments in the same time node. The change direction of each department is determined by the change direction of more than half of the indicators in its subset. The consistency rate is defined as the ratio of the number of department pairs with the same change direction to the total number of department pairs. Taking node T3 as an example, the trends of hematoma volume increase in neurosurgery, intracranial pressure increase in intensive care unit, and CT density value increase in radiology are consistent, while the neurological functional deficit score shows no synchronous change direction, differing from the other three departments. These three departments form three consistent department pairs, while the other three pairs are inconsistent. The inter-departmental consistency rate at this node is 0.50. The consistency calculation iterates through the directional consistency of all six department pairs among the four departments at each time node of the departmental indicator time series group, and takes the average of the directional consistency of the six department pairs as the overall consistency rate for that node. The consistency assessment sequence is calculated by arranging the nodes along the time axis to determine the overall consistency rate. The sequence length is consistent with the time series length of the departmental indicator time series group. For indicator time series marked as low quality within the departmental indicator time series group, a weighting factor of 0.4 is applied during consistency calculation, and the overall consistency rate of the corresponding node is replaced by the weighted mean instead of the arithmetic mean. The closer the consistency assessment sequence value is to 1, the more similar the assessment trends of each department are at that time node; the closer the value is to 0, the more dispersed the opinions among departments.

[0032] For consistency assessment sequences, two types of sudden drops in consistency are identified: those triggered by initial objections from a single department and those triggered by simultaneous divergence from multiple departments. A drop in the consistency assessment sequence is defined as a decrease in the overall consistency rate between adjacent time points exceeding 0.25. Drop detection calculates the difference between adjacent points along the time axis, and nodes with absolute differences exceeding a threshold are marked as candidate drops. Further backtracking of the directional consistency details of each department at the candidate drops determines the type of drop: if the directional change of a single department is inconsistent with the other three departments, and the other three departments are consistent with each other, it is determined to be a case of initial objection from a single department; if two or more departments simultaneously exhibit directional changes with inconsistent directions, it is determined to be a case of simultaneous divergence from multiple departments. Taking node T6 as an example, the overall consistency rate of the consistency assessment sequence drops sharply from 0.83 to 0.50. The detailed values ​​show that the direction of change in the radiology department's CT density value is different from that of the other three departments; this node is determined to be a case of initial objection from a single department. The sudden drop classification label table records the time coordinates, sudden drop type label, triggering department number, and consistency rate decrease for each sudden drop candidate point. For segments in the consistency assessment sequence where multiple consecutive nodes remain in a low consistency rate state, all nodes within that segment are grouped into the same sudden drop event record, with the segment's starting time coordinate as the event location identifier. The sudden drop classification label table comprehensively includes all sudden drop events identified in the consistency assessment sequence, distinguishing between single-department initial objection triggers and multi-department simultaneous disagreement triggers using type labels.

[0033] Based on the sudden drop classification marker table, the location of opinion disagreement is assigned weights to determine the form of opinion disagreement for two types of sudden drops. The weighting coefficient for single-department initiating objections in the sudden drop classification marker table is set to 0.6, while the weighting coefficient for multi-department simultaneous disagreement triggering events is set to 1.0. The latter is higher than the former, reflecting the more significant impact of fundamental cross-disciplinary opinion opposition on decision-making. At the zoning level, the weighting of sudden drops in the acute phase is multiplied by a zoning amplification factor of 1.2 based on the original coefficients, while the weighting of sudden drops in the baseline and observation phases is not amplified. The form of opinion disagreement is constructed using the event locations in the sudden drop classification marker table as the framework. Each event location corresponds to a disagreement node, which is associated with a time coordinate, disagreement type label, and weighted disagreement intensity. For sudden drops in the sudden drop classification marker table merged from consecutive segments, the position of the opinion disagreement form is marked by the time coordinate of the midpoint of the segment, and the average decrease in the consensus rate of each node within the segment is used as the pre-weighted intensity base. The weighted divergence intensity of the opinion divergence pattern is determined by the product of the decrease in the consensus rate and the weighting coefficient after partitioning. In the acute phase, the weighting coefficient is multiplied by an amplification factor of 1.2 based on the original value of 0.6 or 1.0 before being used in the calculation. Divergence nodes with a weighted divergence intensity exceeding 0.35 are marked as high-intensity divergence, and the rest are marked as low-intensity divergence.

[0034] Based on the identification of opinion divergence patterns, excessively consistent intervals are reduced in weight to retain heterogeneous opinions, forming effective consultation features. After arranging the weighted divergence intensity sequence along the time axis, segments with consecutive intensity values ​​below 0.15 are defined as excessively consistent intervals. Within these intervals, the assessment conclusions of various departments tend to be homogeneous, with low heterogeneity information. Consultation feature vectors within excessively consistent intervals are reduced in weight by a reduction factor of 0.3, resulting in feature weights less than one-third of those in normal intervals. Consultation feature vectors corresponding to high-intensity divergence nodes retain their original weights. For low-intensity divergence nodes with a weighted divergence intensity not lower than 0.15, the consultation feature vector weights are multiplied by a retention factor of 0.7. Excessively consistent intervals (intensity below 0.15) have already been processed with a reduction factor of 0.3 and are not subject to a repeated application of the 0.7 factor. Within the excessively consistent interval, if a single department's assessment indicator deviates from the mean by more than 1.5 standard deviations, this deviation record will be extracted and retained separately from the detailed data of opinion divergence patterns as a heterogeneous opinion candidate, avoiding the loss of valuable heterogeneous judgments due to overall weight reduction. Effective consultation features are selected from all consultation feature vectors of opinion divergence patterns, with weights exceeding a retention threshold defined as the 25th percentile of all feature vector weights. The sources of effective consultation features include three categories: original weight feature records of high-intensity divergence nodes in opinion divergence patterns, records with retained coefficients after processing of weights exceeding the retention threshold in low-intensity divergence nodes, and heterogeneous opinion candidates in the excessively consistent interval. Each effective consultation feature record is associated with a time coordinate, department identifier, indicator name, and normalized weight.

[0035] Target feature items are located using disease feature spectrum analysis on effective consultation features. The time axis of effective consultation features spans three partitions of the dynamic diagnostic benchmark, with the most concentrated records in the acute phase. Disease feature spectrum analysis counts the frequency of each indicator in effective consultation features within each partition. For example, the neurosurgical hematoma volume indicator appears 8 times in the acute phase, and the intensive care unit intracranial pressure indicator appears 7 times. Indicators with a frequency exceeding 1.5 times the average of all indicators are identified as high-frequency indicators, which constitute one of the frequency screening criteria for candidate features. A frequency distribution map is constructed with indicator name as the horizontal axis and frequency as the vertical axis. High-frequency indicators form obvious peak areas in the frequency distribution map. Based on this, disease feature spectrum analysis further calculates the weight concentration of each indicator, defined as the proportion of the sum of the weights of all records for that indicator to the total weight of all effective consultation features. High-frequency indicators with a weight concentration exceeding 0.08 are identified as candidate features, which are located in the high-frequency and high-weight dual-high region in the disease feature spectrum. The target feature is further filtered from the candidate feature. The filtering criteria are that it spans at least two partitions and the department identifier covers at least two departments. Appearance across partitions indicates that the indicator is continuously present throughout the disease progression, and coverage of multiple departments indicates that the indicator has received attention from multiple disciplines. The filtered target feature is associated with the indicator name, the combination of departments, the frequency of occurrence, and the concentration of weight, which serves as the basis for subsequent departmental diagnosis and treatment labeling.

[0036] Step S130: Delineate departmental diagnosis and treatment labels based on target feature items, assign weights to departments with high change frequency in the departmental diagnosis and treatment labels as core change departments to generate decision complexity weights, and perform zonal calibration on the dynamic diagnostic benchmark using departmental diagnosis and treatment labels and decision complexity weights to establish a collaborative decision map.

[0037] Specifically, departmental treatment labels were defined based on the target feature items. The indicator with the highest weight concentration among the target feature items was hematoma volume in neurosurgery, with a concentration value of 0.14. This indicator was recorded in both the acute and baseline phases, spanning two zones and covering the departmental identifiers of neurosurgery and radiology. Departmental treatment labels were assigned based on the indicator name of the target feature item, assigning each indicator to the responsible department, which was the department contributing the most to the weight concentration of that indicator. The intracranial pressure indicator covered the departments of critical care medicine and neurosurgery. Critical care medicine contributed 0.62 to the weight concentration of this indicator, and the departmental treatment labels assigned intracranial pressure to critical care medicine as the primary responsibility. A total of six indicators were identified as target feature items, and the departmental treatment labels assigned them to four participating departments: neurosurgery was responsible for hematoma volume and midline shift distance; critical care medicine was responsible for intracranial pressure and mean arterial pressure; neurology was responsible for neurological deficit scores; and radiology was responsible for CT density values. Each record in the departmental diagnosis and treatment annotation is associated with the name of the target feature indicator, the responsible department number, and the zone coverage area. The zone coverage area indicates which zones the indicator is effectively involved in the diagnosis and treatment evaluation. The number of indicators assigned to each department reflects the department's intervention density in this consultation. Neurosurgery and Critical Care Medicine are each responsible for two target feature indicators, and their intervention density is tied for the highest among the four departments.

[0038] In some embodiments, the step of weighting high-frequency change departments in the departmental diagnosis and treatment labels as core change departments to generate decision complexity weights includes: extracting opinion change records of each department from the departmental diagnosis and treatment labels in chronological order to generate a change frequency sequence; identifying high-frequency change departments in the change frequency sequence to form a core change department table; weighting the core change department table based on diagnosis and treatment contribution to generate a core weight distribution; and weighting and integrating the core weight distribution with the change frequency sequence to generate decision complexity weights.

[0039] The change frequency sequence is generated by extracting the records of opinion changes from each department's treatment labels in chronological order. The chronological dimension of the department's treatment labels covers the three partitions of the dynamic diagnostic benchmark. Each department may adjust its treatment opinions on its primary indicators at various time points, and these adjustments are recorded as change events in the department's treatment labels. In the department's treatment labels, the neurosurgery department adjusted its opinion on the timing of surgery related to hematoma volume three times consecutively at the T3 to T5 nodes in the acute phase, changing from conservative observation to active intervention and then to graded intervention. These three adjustments resulted in three change event records. Each change event record carries a change timestamp, the department number, and the opinion tags before and after the change, and is arranged in ascending order by the change timestamp to form a sequence of opinion change records. The change frequency sequence uses the department number and time node as a two-level index. At each index position, the cumulative number of changes for that department at that time node is counted. The cumulative number of changes is grouped and counted from the opinion change record sequence of the department's treatment labels by department and time node. In the frequency sequence of changes, the Intensive Care Unit (ICU) showed 3 changes at the T4 node during the acute phase, and the Neurosurgery Department showed 2 changes at the T3 node. Both of these frequency changes were higher than those of the Neurology and Radiology Departments during the same period. If there were no changes between adjacent opinion records with a time interval exceeding 60 minutes in the departmental treatment annotations, the corresponding position in the frequency sequence of changes was entered as 0, ensuring that the time coverage of the frequency sequence of changes was completely consistent with the temporal range of the departmental treatment annotations.

[0040] For example, the step of identifying high-frequency changing departments in the frequency sequence to form a core changing department table includes: extracting the frequency sequence by department segment by segment of the change interval to generate an interval time series group; identifying departments whose change intervals are continuously contracted to within a preset tolerance by the interval time series group to form a contraction zone marker; marking the contraction zone marker with a continuous risk classification to generate change risk annotations; and generating a core changing department table based on the change risk annotations.

[0041] The frequency variation sequence is segmented by department to extract variation intervals and generate interval time series groups. After arranging the variation events for each department in the frequency variation sequence by timestamp, the time distance between two adjacent variation events is defined as the variation interval, calculated in minutes. In the frequency variation sequence, the timestamps of the variation events in neurosurgery during the acute phase are T3:00, T3:45, T4:10, T4:32, T4:52, T5:12, and T5:37, respectively. The variation intervals between adjacent events are 45 minutes, 25 minutes, 22 minutes, 20 minutes, 20 minutes, and 25 minutes, respectively, showing an overall shrinking trend. The interval time series groups use the department number as the primary index and the variation event number as the secondary index, extracting the variation interval values ​​between adjacent events by difference. Time nodes with 0 variation occurrences in the frequency variation sequence do not generate variation interval records; these interval time series groups only contain adjacent pairs of actually occurring variation events. Department interval time series groups with 1 variation occurrence are empty sequences. In the interval time series, the radiology and neurology departments had few variable events, resulting in interval time series lengths of 0 and 1 respectively, which did not meet the minimum sequence length requirement for subsequent contraction zone identification. The sequence length of the interval time series is equal to the cumulative number of changes for the corresponding department in the frequency change sequence minus one. The interval time series length for neurosurgery is 6, and the interval time series length for intensive care unit is 4.

[0042] Departments whose variable intervals continuously contract to within a preset tolerance are identified by the interval time series group, forming a contraction zone marker. Continuous contraction in the interval time series group is defined as three or more consecutive monotonically decreasing variable interval values. The preset tolerance is set at 15 minutes, and the contraction terminates when the variable interval value falls within the tolerance. In the neurosurgery interval series, the variable interval sequence from the 2nd to the 4th position continuously decreases from 25 minutes to 20 minutes after 22 minutes, with all three values ​​showing a strict monotonically decreasing trend, meeting the continuous contraction condition. Furthermore, the 4th position value at 20 minutes is above the tolerance by 15 minutes, indicating the contraction zone has not yet bottomed out. In the intensive care unit interval series, the variable interval sequence from the 2nd to the 4th position decreases from 30 minutes to 12 minutes. The 4th position value at 12 minutes is below the tolerance by 15 minutes, confirming that the contraction zone bottomed out at this position and forming a contraction zone marker. The contraction zone marker records the department number that triggered contraction identification, the contraction zone initiation event number, the contraction zone termination event number, and the variable interval value at termination. Departments with interval time series lengths of less than 3 were excluded from contraction zone identification. The contraction zone markers only included records from departments that met the criteria. Radiology and neurology were excluded due to insufficient interval time series lengths. The number of contraction zone markers in the interval time series was positively correlated with the number of highly active departments in the frequency sequence. One contraction zone record each was identified in neurosurgery and intensive care unit.

[0043] Persistent risk grading and variable risk annotations are generated for contraction zone markers. The persistent risk of a contraction zone marker is measured by how close the variation interval at the termination of the contraction zone is to the preset tolerance. The smaller the variation interval at termination, the closer the variation rhythm is to the limit frequency, and the higher the persistent risk. The risk score R is calculated using the formula R=L / E, where L is the preset tolerance value and E is the variation interval at the termination of the contraction zone. In the contraction zone markers, the variation interval at the termination of the contraction zone in the intensive care unit is 12 minutes, which is 3 minutes different from the tolerance of 15 minutes, resulting in a risk score of 1.25. In the contraction zone markers, the contraction zone in the neurosurgery unit did not reach the bottom, and the termination interval was taken as the last value of the contraction zone, 20 minutes, resulting in a risk score of 0.75, lower than the risk score of the intensive care unit. Variable risk annotations are indexed by the department number of the contraction zone marker, with departments having a risk score greater than 1.0 marked as high persistent risk level, and departments with a risk score between 0.5 and 1.0 marked as medium persistent risk level. In the contraction zone markers, the Intensive Care Unit (ICU) risk score of 1.25 exceeds the high-risk threshold and is marked as high-persistent risk in the variable risk annotations; the Neurosurgery department risk score of 0.75 falls within the medium-risk range and is marked as medium-persistent risk in the variable risk annotations. The variable risk annotations are linked to the department number, risk level label, risk score value, and contraction zone time range. The contraction zone time range of departments with high-persistent risk in the risk annotations is prioritized for inclusion in the core variable department table selection criteria.

[0044] A core list of departments with variable risk is generated based on the variable risk annotations. Departments with a high persistent risk level in the variable risk annotations are directly included in the core list of departments with variable risk. Departments with a medium persistent risk level are also included if their risk score in the variable risk annotation exceeds 0.65. The Intensive Care Unit (ICU) has a high persistent risk level in the variable risk annotations and is directly entered into the core list of departments with variable risk. The Neurosurgery department has a risk score of 0.75, which is higher than the medium risk inclusion threshold of 0.65, and is also entered into the core list of departments with variable risk. The risk score value in the variable risk annotation is converted into a basic risk weight for that department in the core list of departments with variable risk. The basic risk weight for ICU is 1.25, and the basic risk weight for Neurosurgery is 0.75. The two records in the core list of departments with variable risk correspond to ICU and Neurosurgery, respectively. Departments with a risk score below 0.65 in the variable risk annotations are not entered into the core list of departments with variable risk. Neurology and Radiology are excluded because no valid variable risk annotation records were generated. The number of departments in the core change department table is determined by the risk classification results of the change risk annotation, and ultimately includes two department records: critical care medicine and neurosurgery.

[0045] The core weight distribution is generated by weighting the treatment contribution of departments in the core change department table. The basic risk weight of each department in the core change department table reflects the degree of activity in the change. The weighting of treatment contribution is achieved by adding the contribution of the department's primary responsibility indicators in the target feature item to the basic risk weight. The formula for calculating the comprehensive weight W is W=R+a×N, where R is the risk score (basic risk weight) of the department in the core change department table, N is the number of primary responsibility indicators for the department in the target feature item, and a is the single indicator contribution coefficient with a value of 0.2, representing the incremental contribution of each primary responsibility indicator to the comprehensive weight. In the core change department table, neurosurgery is primarily responsible for the hematoma volume and midline shift distance target feature items, with an additional contribution value of 0.4; critical care medicine is primarily responsible for intracranial pressure and mean arterial pressure, both with an additional contribution value of 0.4. After weighting, the comprehensive weight of neurosurgery in the core change department table is 1.15, and the comprehensive weight of critical care medicine is 1.65. The core weight distribution stores the weighted comprehensive weights using the department number as an index. Neurology and Radiology, which were not included in the core change department table, were entered into the core weight distribution with a baseline weight of 1.0. This baseline weight was not adjusted for weighting and reflects the routine decision-making influence of non-core change departments. The weight values ​​for the four departments in the core weight distribution are as follows: Intensive Care Unit 1.65, Neurosurgery 1.15, Neurology 1.0, and Radiology 1.0, with a total weight of 4.8. Subsequent proportion calculations are based on the proportion of each department's weight to the total weight of 4.8. The weighting results driven by the core change department table resulted in higher core weight distribution values ​​for Intensive Care Unit and Neurosurgery than for departments without weighting, placing them in a priority position in terms of decision-making influence within the core weight distribution.

[0046] The decision complexity weight is generated by weighted integration of the core weight distribution and the frequency sequence of changes. The core weight distribution provides the static weight benchmark for each department, while the frequency sequence of changes provides the dynamic activity level of each department at different time points. The weighted integration of the two forms the decision complexity weight that changes over time. For any time point t, the original complexity score C(t) is calculated as C(t) = sum(W(i) × f(i,t)), where W(i) is the weight of the i-th department in the core weight distribution, and f(i,t) is the number of changes of the i-th department at node t in the frequency sequence of changes. The summation is performed by traversing all K participating departments. The product of the weight of the ICU (1.65) in the core weight distribution and the number of changes of the ICU at node T4 in the acute phase (3 times) in the frequency sequence of changes is 4.95. This value is used as the contribution of the ICU to the decision complexity weight at node T4. The original complexity score for each node is obtained by multiplying the number of changes in each department at each time point in the frequency of change sequence by the corresponding department weight in the core weight distribution, and then summing the products over the four departments. The decision complexity weight D(t) is normalized for each node using the formula D(t) = C(t) / Cmax, where Cmax is the maximum value of the original complexity score across the entire time series, compressing the numerical range of the decision complexity weight to between 0 and 1. The weight differences in the core weight distribution make the impact of changes in high-weight departments more significant on the decision complexity weight; the concentrated changes in the intensive care unit during the acute phase lead to an overall higher decision complexity weight during this period. The decision complexity weight is stored as a one-dimensional sequence indexed by time nodes, with the sequence length consistent with the time coverage of the frequency of change sequence. Nodes in the frequency of change sequence with all changes being 0 correspond to a decision complexity weight of 0.

[0047] A collaborative decision-making graph is established by partitioning the dynamic diagnostic benchmark with departmental treatment labels and decision complexity weights. The three partitions of the dynamic diagnostic benchmark serve as spatial carriers in the partitioning process. Departmental treatment labels map the coverage of each department's primary responsibility indicator to the corresponding partition of the dynamic diagnostic benchmark, forming an index of the indicator's affiliation within the partition. In the departmental treatment labels, the hematoma volume indicator, primarily under the responsibility of neurosurgery, covers the baseline and acute phases. The CT density value indicator, primarily under the responsibility of radiology, covers all three partitions. Partitioning marks the coverage partitions of each indicator on the partition structure of the dynamic diagnostic benchmark. The time sequence of decision complexity weights is mapped to each partition node of the dynamic diagnostic benchmark. The mean of the decision complexity weight in the acute phase is 0.78, the mean in the baseline phase is 0.31, and the mean in the observation phase is 0.45. Partitioning appends these three means to the weight field of the corresponding partition of the dynamic diagnostic benchmark. After receiving the index of indicator attribution from departmental diagnosis and treatment annotations and the partition average of decision complexity weights, each partition of the dynamic diagnostic benchmark forms a partition structure with departmental identifiers and weight labels. The collaborative decision graph uses this partition structure as its framework to organize all the labeling results. Each partition node in the collaborative decision graph carries the list of indicators covering that partition from the departmental diagnosis and treatment annotations, the node-level weight sequence of decision complexity weights in that partition, and the original temporal index of the dynamic diagnostic benchmark. The indicator attribution driven by departmental diagnosis and treatment annotations and the partition weights driven by decision complexity weights form a two-dimensional labeling in the collaborative decision graph. This two-dimensional labeling enables the collaborative decision graph to simultaneously possess both departmental responsibility positioning and temporal complexity characterization query capabilities.

[0048] Step S140: Combine the collaborative decision-making graph with the dynamic diagnostic benchmark and the decision complexity weight to complete the scheme difference supplementation and form a modified scheme set. Perform consistency verification on the modified scheme set to form a diagnosis-treatment conflict index. Determine the dispute labeling table based on the disciplinary opinion deviation distribution of the diagnosis-treatment conflict index.

[0049] In some embodiments, the step of combining the collaborative decision graph with the dynamic diagnostic benchmark and the decision complexity weight to complete the scheme difference completion and form a corrected scheme set includes: performing segment-by-segment initial scheme matching on the collaborative decision graph and the dynamic diagnostic benchmark to form a preliminary scheme sequence; detecting the scheme coverage missing segments in the preliminary scheme sequence to form a scheme missing label table; extracting missing correlation features from the scheme missing label table and the dynamic diagnostic benchmark to generate associated missing labels; and performing differential completion based on the associated missing labels and the decision complexity weight to form a corrected scheme set.

[0050] The collaborative decision-making atlas and the dynamic diagnostic benchmark are matched segment by segment to form a preliminary protocol sequence. The partition structure of the collaborative decision-making atlas and the dynamic diagnostic benchmark share the same partition boundary definitions. The time boundaries of the baseline segment, acute phase segment, and observation phase correspond completely in both. Segment-by-segment matching is performed one by one on a partition-by-partition basis. The indicator list of the baseline segment of the collaborative decision-making atlas includes two entries: hematoma volume and CT density value. The indicator time sequence nodes of the baseline segment of the dynamic diagnostic benchmark include the above two indicators and prothrombin time. The initial protocol matching identifies prothrombin time in the baseline segment as an unmatched indicator that is recorded in the dynamic diagnostic benchmark but not covered by the collaborative decision-making atlas. The preliminary protocol sequence uses the partition number as the primary index and the indicator name as the secondary index. The matching status is marked at each index position. A successful match is marked as matched. If the collaborative decision-making atlas has a protocol but the dynamic diagnostic benchmark does not have a corresponding time sequence, it is marked as one-way coverage. If the dynamic diagnostic benchmark has a time sequence but is not covered by the collaborative decision-making atlas, it is marked as unmatched. In the acute phase of the dynamic diagnostic benchmark, 10 time-series nodes of indicators participated in matching, and in the acute phase of the collaborative decision-making graph, 14 scheme entries participated in matching. The preliminary scheme sequence formed 10 matched positions, 4 one-way covered positions, and 0 unmatched positions in the acute phase. The matching results of the preliminary scheme sequence for the three partitions were stored sequentially. The total length of the sequence is equal to the sum of the number of time-series nodes of all partition indicators of the dynamic diagnostic benchmark (including matched and unmatched positions) and the number of one-way covered entries of the collaborative decision-making graph (positions where the graph has a scheme but the benchmark does not have a corresponding time-series node). The two parts do not overlap.

[0051] The preliminary protocol sequence is used to detect missing segments and create a protocol missing marker table. Positions in the preliminary protocol sequence that show no matching are defined as protocol missing points. Missing point detection is performed position-by-position scanning along the partition order of the preliminary protocol sequence, merging consecutive adjacent missing points into missing segments. The prothrombin time position in the baseline segment of the preliminary protocol sequence is an independent missing point; there are no missing points in the acute phase. In the observation phase, two adjacent positions of cerebral perfusion pressure and neurological deficit score constitute a consecutive missing segment. The protocol missing marker table uses the missing segment number as an index, linking the partition identifier of the missing segment, a list of missing indicator names, the start and end positions of the missing segment, and the length of the missing segment. Missing segments of length 1 in the preliminary protocol sequence are marked as punctate missing segments in the protocol missing marker table; missing segments of length 2 or greater are marked as consecutive missing segments. Since consecutive missing segments involve multiple adjacent indicator positions, all missing indicators within the segment must be processed simultaneously during completion. The protocol missing marker table contains one missing record for the baseline segment (pointed missing) and one missing record for the observation period (continuous missing). The table contains two records covering the three missing indicator locations. One-way coverage locations in the preliminary protocol sequence are not included in the protocol missing marker table. One-way coverage indicates that the collaborative decision-making map already has protocol entries, but the dynamic diagnostic benchmark lacks a corresponding time sequence; this is considered map over-coverage rather than a missing sequence.

[0052] The missing data table and dynamic diagnostic benchmarks were compared using a missing data feature extraction method to generate associated missing data labels. For each missing indicator in the missing data table, associated features were extracted at the corresponding time-series node of the dynamic diagnostic benchmark partition. These associated features were defined as the correlation coefficient between the time-series of the missing indicator and the time-series of indicators covered by existing collaborative decision-making maps within the same partition. The correlation coefficient between the time-series of prothrombin time and hematoma volume in the same partition of the dynamic diagnostic benchmark was 0.71, and the correlation coefficient with CT density value was 0.58. The association strength for prothrombin time was the average of these two correlation coefficients, 0.645. The correlation coefficient between cerebral perfusion pressure in consecutive missing segments of the missing data table and intracranial pressure during the observation period of the dynamic diagnostic benchmark was 0.82, and the correlation coefficient between neurological deficit score and CT density value in the same partition was 0.55. The association strength for each missing indicator was determined by its highest correlation coefficient. The associated missing data labels were indexed by the missing segment number in the missing data table, and the association strength value and the name of the highest-correlated indicator were appended to the original missing data records. In the dynamic diagnostic benchmark, covered indicators with a correlation coefficient exceeding 0.6 with the missing indicator are marked as strongly correlated indicators. The presence of strongly correlated indicators indicates that the missing solution can be derived and completed based on existing solution entries. The correlation strength value of the missing markers will determine the initial confidence value of the completion solution during the differential completion stage. Missing positions with a correlation strength higher than 0.6 will be completed with high confidence, while those with a correlation strength lower than 0.6 will be completed with low confidence.

[0053] A modified scheme set is formed by implementing differentiated completion based on association missing markers and decision complexity weights. Completion schemes for high-confidence completion positions in association missing markers are generated by mapping from existing scheme entries of strongly correlated indicators. The mapping rule is to migrate the diagnosis and treatment opinion labels and department weights of strongly correlated indicators to the missing indicators according to the correlation coefficient ratio. Completion schemes for baseline prothrombin time in association missing markers are generated by mapping from hematoma volume scheme entries. The department identifier for the completed entry is retained as neurosurgery, and the diagnosis and treatment opinion label is migrated to "coagulation function monitoring - hematoma stability association," with an initial confidence value set to 0.645. The final weight P of the completed entry is calculated as P = Ds × p, where Ds is the mean of decision complexity weights within the partition s of the missing indicator, and p is the association strength of the association missing marker, i.e., the initial confidence value. The final completion weight for the baseline segment is 0.20. Completion schemes for low-confidence completion positions in association missing markers are written into the modified scheme set after adding low-confidence markers, referencing the opinion labels of adjacent entries in the same partition of the collaborative decision graph. The revised scheme set integrates all existing entries in the collaborative decision-making graph with new entries generated through differential completion. The weights of the original entries remain unchanged from their values ​​in the collaborative decision-making graph, while the weights of completed entries driven by missing association markers use the final completion weights calculated through differential completion. The total number of entries in the revised scheme set is the sum of the original number of entries in the collaborative decision-making graph and the number of entries completed by missing association markers. The revised scheme set uses a unique index of a combination of partition identifier and indicator name to ensure that the same indicator does not appear repeatedly within the same partition.

[0054] A consistency check is performed on the revised protocol set to generate a treatment conflict index. If different departments within the same partition of the revised protocol set have conflicting opinions on the same indicator, these opinions constitute a conflict pair within that partition. Conflict pairs are identified based on the semantic opposition of the treatment opinion labels; for example, "active intervention" and "conservative observation" constitute an opposition. In the acute phase of the revised protocol set, there are two records for the hematoma volume indicator: "active surgery" from neurosurgery and "conservative treatment" from neurology. The treatment opinion labels of these two records constitute a conflict pair. The intensity of this conflict pair is measured by the absolute value of the weight difference between the corresponding departments of the two records in the core weight distribution. A larger difference indicates a more significant advantage for the dominant opinion; the weight difference is 0.15. All conflict pairs within each partition of the revised protocol set are statistically analyzed along the indicator dimension. The number of conflict pairs for each indicator and the mean intensity of each conflict pair together constitute the conflict component of that indicator. The treatment conflict index is constructed using time partitions as rows and indicator dimensions as columns to form a conflict component matrix. A value of 0 in the conflict component matrix indicates no disagreement regarding that indicator in that partition; higher values ​​indicate stronger interdepartmental disagreements. When low-confidence completion items from the revised scheme set are included in conflict pair identification, the conflict intensity is multiplied by a reduction factor of 0.5. The low-confidence marker indicates that the reliability of the item is questionable; the reduction process prevents low-confidence completion items from excessively inflating the treatment conflict index. In the conflict component matrix of the treatment conflict index, the values ​​for the acute phase in the revised scheme set are generally higher than those for the baseline and observation phases.

[0055] In some embodiments, determining the dispute labeling table based on the disciplinary opinion deviation distribution of the treatment conflict index includes: statistically analyzing the opinion magnitude of each department to form a departmental opinion magnitude map; calculating the symmetry axis offset of the departmental opinion magnitude map to form a deviation degree sequence; identifying cases where the same department simultaneously deviates from the median line in multiple treatment dimensions from the deviation degree sequence to generate multidimensional synchronous deviation markers; and extracting the corresponding treatment conflict index from the departmental opinion magnitude map based on the multidimensional synchronous deviation markers to form a dispute labeling table.

[0056] The opinion magnitude of the treatment conflict index is statistically analyzed across departments to form a departmental opinion magnitude map. After slicing the conflict component matrix of the treatment conflict index by department, each department's slice matrix contains the opinion deviation values ​​for that department across all partitions and indicator dimensions. Departmental opinion magnitude is defined as the set of semantic distance scores between the department's opinion labels at each indicator position and the median label of the entire department's opinions. A larger opinion magnitude value indicates that the department's opinion deviates more from the overall consensus. The median label of the entire department's opinions is based on the median item of the opinion labels of participating departments at each indicator position, sorted by intervention intensity from strongest to weakest. The semantic distance score maps the difference in intervention intensity levels to a range of 0 to 1, with a difference of 0.33 for level one, 0.67 for level two, and 1.0 for level three (applicable to extended cases where the intervention intensity level exceeds level three). In the treatment conflict index, the opinion amplitude values ​​for neurosurgery in the hematoma volume and midline shift dimensions were 0.68 and 0.54, respectively, while the opinion amplitude value for critical care medicine in the intracranial pressure dimension was 0.72. The opinion amplitude values ​​for all four departments were statistically summarized at each position in the treatment conflict index. The departmental opinion amplitude graph was constructed using department number as the row index and indicator name as the column index. Each cell represented the weighted average of the opinion amplitude values ​​for the corresponding department across all regions for the corresponding indicator. The weighted average was calculated using the regional average of decision complexity as the weighting coefficient. The cell value corresponding to a position with a conflict component of 0 in the treatment conflict index was set to 0 in the departmental opinion amplitude graph. This position indicates no departmental disagreement, and zero-value cells do not generate subsequent deviation sequence records.

[0057] The departmental opinion amplitude chart is used to calculate the deviation sequence by applying the symmetry axis offset. The numerical distribution of each column in the departmental opinion amplitude chart is based on the median value of the entire departmental opinion amplitude as the symmetry axis benchmark, which reflects the center position of the opinion amplitude for each indicator dimension. The offset V is calculated using the formula V=AM, where A is the opinion amplitude value of that department for that indicator in the departmental opinion amplitude chart, M is the median value of the entire departmental opinion amplitude for that indicator column, and a positive V indicates a deviation above the median, while a negative V indicates a deviation below the median. The opinion amplitude values ​​for the four departments in the hematoma volume column of the departmental opinion amplitude chart are 0.68, 0.54, 0.21, and 0.09, with a median value of 0.375. The offset between the neurosurgery opinion amplitude of 0.68 and the median value of 0.375 is 0.305, and the offset between the radiology opinion amplitude of 0.09 and the median value of 0.375 is -0.285. The deviation sequence is indexed by a combination of department number and indicator name. After sorting the offsets in each column of the department opinion amplitude diagram by absolute value, the department with the largest absolute value ranks first in the deviation sequence value for that indicator dimension. Neurosurgery ranks first in the hematoma volume dimension. Positions with an absolute value exceeding 0.25 in the deviation sequence are marked as significant deviation points, and the corresponding department in the department opinion amplitude diagram is included in the subsequent multidimensional synchronous analysis for that indicator dimension. The deviation sequence covers all non-zero cells in the department opinion amplitude diagram, and the total sequence length equals the number of non-zero cells in the department opinion amplitude diagram. Zero-value cells in the department opinion amplitude diagram do not generate deviation sequence records.

[0058] Multidimensional synchronous deviation markers are generated to identify situations where the same department simultaneously deviates from the median across multiple diagnostic and treatment dimensions in the deviation degree sequence. Multidimensional synchronous deviation is defined as the absolute value of the offset in two or more indicator dimensions of the same department simultaneously exceeding the significant deviation threshold of 0.25, and the deviation directions are consistent. In the deviation degree sequence, the neurosurgery department has an offset of +0.305 in the hematoma volume dimension and +0.29 in the midline shift distance dimension; both dimensions are positively skewed and their absolute values ​​exceed the threshold, triggering multidimensional synchronous deviation identification. In the deviation degree sequence, the intensive care unit has an offset of +0.345 in the intracranial pressure dimension and +0.18 in the mean arterial pressure dimension; the absolute value of the mean arterial pressure dimension does not exceed the threshold of 0.25, and therefore does not constitute multidimensional synchronous deviation. The multidimensional synchronous deviation markers are indexed by the department number, linking to the list of indicator dimensions triggering synchronous deviation, the offset values ​​for each dimension, and the deviation direction label. In the multidimensional synchronous deviation markers, a single record for neurosurgery includes both hematoma volume and midline shift distance dimensions. After scanning the offsets of each department and dimension in the deviation degree sequence one by one, only one department, Neurosurgery, met the criteria for multidimensional synchronous deviation, and the multidimensional synchronous deviation marker ultimately contained one record. Departments with significant deviations in a single dimension in the deviation degree sequence did not generate multidimensional synchronous deviation marker records, and opinions from departments with single-dimensional deviations were not written into the multidimensional synchronous deviation markers. The subsequent dispute annotation table was generated solely by the multidimensional synchronous deviation markers, and significant single-dimensional deviations in intracranial pressure in the Intensive Care Unit were not included in the dispute annotation table.

[0059] Based on multidimensional synchronous deviation markers, a dispute annotation table is generated by extracting corresponding treatment conflict indices from the departmental opinion amplitude diagram. In the multidimensional synchronous deviation markers, the neurosurgery dimension triggering synchronous deviation is hematoma volume and midline shift distance. The corresponding neurosurgery cells for these two indicators are extracted from the departmental opinion amplitude diagram, with cell values ​​of 0.68 and 0.54, respectively. The conflict components of the treatment conflict index corresponding to the cell positions in the departmental opinion amplitude diagram hit by the multidimensional synchronous deviation markers are extracted from the conflict component matrix by partition and indicator name index. The conflict component of the hematoma volume indicator is 0.21 in the acute phase and 0.08 in the baseline phase. The dispute annotation table uses the combination of the department number and indicator dimension hit by the multidimensional synchronous deviation markers as the primary key, and associates the department number, indicator name, partition identifier, departmental opinion amplitude diagram cell value, treatment conflict index conflict component, and deviation direction label. The length of the list of indicator dimensions triggering synchronous deviation in the multidimensional synchronous deviation marker determines the number of records generated in the dispute annotation table. Neurosurgery triggers two dimensions: hematoma volume and midline shift distance. The effective partitions for the hematoma volume indicator are the baseline segment and the acute phase segment, and the effective partitions for the midline shift distance indicator are also the baseline segment and the acute phase segment. Two records are generated for each dimension, resulting in a total of four records in the dispute annotation table. Cells not hit by the multidimensional synchronous deviation marker in the departmental opinion amplitude diagram do not generate records in the dispute annotation table, ensuring that the dispute annotation table only includes high-confidence dispute locations verified by multidimensional synchronous deviation. The deviation direction label in the dispute annotation table is determined by the positive or negative sign of the corresponding position in the deviation degree sequence. All four records from neurosurgery have positive deviation direction labels, indicating that neurosurgery's opinion is generally biased towards the positive intervention side in the indicator dimensions involved in the dispute annotation table.

[0060] Step S150: Map the dispute labeling table with the departmental diagnosis and treatment labels to establish arbitration rules. Based on the arbitration rules, perform weighted voting matching on the dispute labeling table to generate decision validity. Mark multiple evolution paths and include them in the output of multidisciplinary collaborative diagnosis and treatment plans by labeling the low consensus interval of decision validity.

[0061] In some embodiments, the step of mapping the dispute labeling table to the departmental diagnosis and treatment labels to establish arbitration rules includes: performing co-occurrence frequency statistics on the dispute labeling table and the departmental diagnosis and treatment labels to form a co-occurrence frequency map; combining the co-occurrence frequency map with the dispute labeling table to extract high-co-occurrence departmental opinion groups and historical voting records to form a joint dispute pattern table; identifying opinion combinations in the joint dispute pattern table that have never been adopted in historical arbitration results to generate invalid co-occurrence markers; and based on the invalid co-occurrence markers, performing a merging strategy mapping to establish arbitration rules after removing invalid opinions from the joint dispute pattern table.

[0062] Co-occurrence frequency statistics were performed on the disputed labeling table and the departmental diagnosis and treatment labels to form a co-occurrence frequency map. Each of the four records in the disputed labeling table includes the deviating department number and the indicator name. The responsible department number for the same indicator name in the departmental diagnosis and treatment labels and the deviating department number in the disputed labeling table form a co-occurrence pair. The frequency of these co-occurrence pairs is cumulatively counted in historical consultation records. The combination of neurosurgery and hematoma volume in the disputed labeling table appeared 23 times in historical consultation records. The hematoma volume indicator, for which neurosurgery is primarily responsible in the departmental diagnosis and treatment labels, forms the highest frequency co-occurrence pair, with a frequency value of 23 in the co-occurrence frequency map. The co-occurrence frequency statistics are constructed using the deviating department number of each record in the disputed labeling table as the row index and the indicator name of each record in the departmental diagnosis and treatment labels as the column index. Each cell counts the number of times the corresponding co-occurrence pair appears in historical consultation records. When disputes occur in multiple indicator dimensions within the same department in the disputed labeling table, the number of non-zero cells in the corresponding row of the co-occurrence frequency map increases, reflecting a wider range of indicators involved in the disputes within that department. Departments with a higher number of primary responsibility indicators in their departmental diagnosis and treatment annotations show higher cumulative frequencies in the corresponding columns of the co-occurrence frequency map. For example, the neurosurgery department has two primary responsibility indicators, and the total frequency of the neurosurgery-related columns in the co-occurrence frequency map is 40. Cells in the co-occurrence frequency map whose frequency values ​​exceed 1.5 times the mean of the entire matrix are marked as high co-occurrence positions. The opinion groups of the departments corresponding to high co-occurrence positions are given priority in extraction in the subsequent joint dispute model table.

[0063] A joint dispute pattern table is formed by combining the co-occurrence frequency map with the dispute annotation table to extract high-co-occurrence departmental opinion groups and historical voting records. In the co-occurrence frequency map, the departmental opinion group corresponding to a high co-occurrence position is defined as an opinion pair between a department and an indicator whose participation frequency exceeds 1.5 times the mean. Two high-co-occurrence positions in the co-occurrence frequency map, (neurosurgery, hematoma volume) and (neurosurgery, midline shift distance), each correspond to one departmental opinion group. The deviation opinion tags corresponding to high-co-occurrence positions in the dispute annotation table represent the opinion content of the departmental opinion group. Historical voting records are retrieved from the consultation archive by combining the department number and indicator name, extracting the voting result sequence of that opinion group in historical arbitration. The joint dispute pattern table uses the departmental opinion group as a row index, linking the department number, indicator name, deviation opinion tags of the corresponding records in the dispute annotation table, the number of historical votes adopted, and the number of historical votes rejected. Departmental opinion groups corresponding to non-high-co-occurrence positions in the co-occurrence frequency map are not included in the joint dispute pattern table. Records with low deviation intensity in the dispute annotation table are also excluded if their corresponding co-occurrence positions do not reach the high co-occurrence threshold. The number of records in the joint dispute pattern table is determined by the number of high co-occurrence positions in the co-occurrence frequency spectrum. If the co-occurrence frequency spectrum identifies two high co-occurrence positions, the joint dispute pattern table contains two departmental opinion group records. The deviation direction labels from the dispute labeling table are retained as supplementary information in the joint dispute pattern table. These labels distinguish whether the opinion leans towards active intervention or conservative observation, serving as reference information for verifying the consistency of opinion direction during the implementation of arbitration rules.

[0064] For example, the step of generating invalid co-occurrence markers for identifying opinion combinations whose historical arbitration results have never been adopted in the joint dispute pattern table includes: extracting historical individual arbitration failure rates from the opinions of each department in the joint dispute pattern table to form a failure rate sequence; identifying department opinions whose failures are concentrated in the rapid progression stage of the disease from the failure rate sequence to generate a progression-stage failure marker; establishing high-risk phase arbitration files for the department opinions corresponding to the progression-stage failure markers to generate a phase arbitration strategy group; and filtering opinion combinations that have never been adopted based on the phase arbitration strategy group to generate invalid co-occurrence markers.

[0065] Historical individual arbitration failure rates were extracted from the joint dispute pattern table to form a failure rate sequence. Two records in the joint dispute pattern table correspond to two departmental opinion groups: (Neurosurgery, Hematoma Volume) and (Neurosurgery, Midline Shift Distance). The single voting results of each opinion group in historical arbitration are arranged chronologically to form a voting result sequence. The failure rate F is calculated using the formula F = Nr / Nt, where Nr is the number of historical rejections for that opinion group, and Nt is the total number of historical votes. In the joint dispute pattern table, the (Neurosurgery, Midline Shift Distance) opinion group had a total of 17 historical votes and 14 rejections, resulting in a historical individual arbitration failure rate of 0.82%. The (Neurosurgery, Hematoma Volume) opinion group had a failure rate of 0.30, showing a significant difference between the two. The failure rate sequence is indexed by the record number in the joint dispute pattern table. Each position carries the historical individual arbitration failure rate value for the corresponding departmental opinion group and the time phase label for each vote. The time phase labels are divided into two categories: stable disease period and rapid disease progression period. In the Joint Controversy Pattern Table, historical voting records of each opinion group are grouped by phase label to calculate failure rates. The difference between the failure rate during the progression phase and the failure rate during the stable phase reflects the additional unreliability of that opinion group during the disease progression phase. Each record in the failure rate sequence also includes a consecutive failure marker for the opinion group's five most recent votes. Opinion groups with more than four consecutive failures have a recent high failure marker added to the failure rate sequence. In the Joint Controversy Pattern Table (Neurosurgery, Midline Shift Distance), the recent high failure marker for that record is activated in the failure rate sequence because all five of the opinion groups' most recent votes were failures.

[0066] The failure rate sequence identifies departmental opinions where failures are concentrated in the rapid disease progression phase, generating a progression-phase failure marker. When the difference between the progression-phase failure rate and the stable-phase failure rate in each record of the failure rate sequence exceeds 0.3, the departmental opinions are identified as belonging to the opinion group where failures are concentrated in the rapid disease progression phase. In the failure rate sequence (Neurosurgery, Midline Shift Distance) opinion group, the progression-phase failure rate is 0.92 and the stable-phase failure rate is 0.60; the difference of 0.32 exceeds the threshold of 0.3, triggering the generation of a progression-phase failure marker. In the failure rate sequence (Neurosurgery, Hematoma Volume) opinion group, the progression-phase failure rate is 0.38 and the stable-phase failure rate is 0.25; the difference of 0.13 does not exceed the threshold, and the progression-phase failure marker is not triggered. The progression-phase failure marker uses a combination of department number and indicator name as an index, associating the progression-phase failure rate, stable-phase failure rate, difference value, and recent high-failure marker status. Opinion groups in the failure rate sequence that simultaneously meet both the conditions of progress-stage failure identification trigger and recent high-failure marker activation are marked with a double high-risk label in the progress-stage failure marker. The double high-risk label indicates that the arbitration unreliability of this opinion group is persistent during the progress stage. The source record number of the joint dispute pattern table in the failure rate sequence is retained as a correlation field in the progress-stage failure marker, which facilitates the lookup of the progress-stage failure marker and the original historical voting records in the joint dispute pattern table.

[0067] For opinions from departments corresponding to the failure markers in the progression phase, high-risk phase arbitration archives were established to generate phase arbitration strategy groups. Opinion groups (neurosurgery, midline shift distance) among the failure markers in the progression phase met the dual high-risk labeling criteria. The high-risk phase arbitration archives extracted all arbitration records for this opinion group during the rapid disease progression phase from historical consultation records, with the extraction scope limited to the progression phase label node corresponding to the failure marker. The archive extraction driven by the failure markers in the progression phase yielded 14 progression phase arbitration records, each carrying an arbitration timestamp, a combination of participating department numbers, the voting result, and the disease progression rate at that time. The phase arbitration strategy groups organized the high-risk phase arbitration archives around progression rate segmentation, with one phase strategy entry corresponding to each of the three progression rate intervals: below 2 mL / h, 2 to 4 mL / h, and above 4 mL / h. In the failure markers for the progression phase, 14 arbitration records were assigned to three intervals based on the progression rate. The interval exceeding 4 mL / h contained 7 records, and the arbitration failure rate for the opinion group within this interval (neurosurgery, midline shift distance) was 1.0, indicating that the opinion group was never adopted during the rapid progression phase. Each strategy entry in the phase arbitration strategy group was associated with the progression rate interval, the arbitration failure rate within the interval, the frequency distribution of participating department numbers within the interval, and the strategy recommendation label. The strategy recommendation label was automatically generated into two categories based on the failure rate value: high-risk avoidance or conditional adoption. The double high-risk labeling of the failure markers for the progression phase was transformed into a high-risk avoidance label for strategy entries in the phase arbitration strategy group exceeding the 4 mL / h interval. This label directly affected the generation and judgment of subsequent invalid co-occurrence markers.

[0068] Invalid co-occurrence markers are generated based on the selection of opinion combinations that have never been adopted within the temporal arbitration strategy group. In the temporal arbitration strategy group, strategy entries with a progression rate exceeding 4 mL / h have an arbitration failure rate of 1.0. Within this range (neurosurgery, midline shift distance), opinion groups have never been adopted in 7 arbitrations, meeting the generation criteria for invalid co-occurrence markers. In the temporal arbitration strategy group, strategy entries with a progression rate between 2 and 4 mL / h have an arbitration failure rate of 0.80. One out of five records was adopted, failing to meet the condition of never being adopted, and the invalid co-occurrence marker does not cover this range. Invalid co-occurrence markers are indexed by the combination of departmental opinion groups and progression rate ranges. The departmental opinion group and range combination corresponding to strategy entries with an arbitration failure rate of 1.0 in the temporal arbitration strategy group are written into the invalid co-occurrence marker. The invalid co-occurrence marker is associated with the department number, indicator name, applicable progression rate range, historical number of arbitration failures, and the source entry number of the temporal arbitration strategy group. The applicable progression rate range limits the validity conditions of this invalid marker. The phase-based arbitration strategy group-driven generation of invalid co-occurrence markers produced one record, corresponding to invalid co-occurrence scenarios in the (neurosurgery, midline shift distance) opinion group within a progression rate range exceeding 4 mL / h. The effective conditions of the invalid co-occurrence markers ensure that the arbitration rules only filter out the corresponding opinion group under specific phase conditions, rather than categorically excluding all phases. This refined application of invalid co-occurrence markers enhances the clinical adaptability of the arbitration rules.

[0069] Arbitration rules are established by mapping merging strategies after removing invalid opinions from the joint dispute pattern table based on invalid co-occurrence markers. When the opinion group (Neurosurgery, Midline Shift Distance) in the invalid co-occurrence markers meets the effective condition of a progression rate exceeding 4 mL / h, the corresponding record in the joint dispute pattern table is marked as removed under that condition. After removing records with invalid co-occurrence markers from the joint dispute pattern table, the remaining valid opinion groups are arranged in descending order of the department's historical adoption rate, forming a priority sequence of valid opinions. The opinion group (Neurosurgery, Hematoma Volume) has the highest adoption rate of 0.70. The effective condition of invalid co-occurrence markers is transformed into condition triggering logic in the arbitration rules. For each dispute marker record, the arbitration rules first determine whether the effective condition of the invalid co-occurrence marker is triggered in the current phase, and then decide whether to include it in the corresponding opinion group. The priority sequence of valid opinions in the joint dispute pattern table and the department's historical adoption rate together constitute the weight mapping basis of the arbitration rules. The arbitration rules use the adoption rate of each opinion group in the priority sequence of valid opinions as the basic weight, and the deviation intensity value of the dispute marker record as an adjustment factor to calculate the final voting weight. The arbitration rules are structured in three layers: a condition elimination layer driven by invalid co-occurrence markers, a valid opinion screening layer driven by the joint dispute pattern table, and a weight mapping layer driven by historical adoption rates. These three layers are executed sequentially to complete the arbitration rule calculation for each record in the dispute marker table. The number of records with invalid co-occurrence markers and the number of valid records in the joint dispute pattern table jointly determine the total number of rule entries in the arbitration rules. The current arbitration rules include two valid opinion weight mapping rules and one condition elimination rule.

[0070] In some embodiments, the step of generating decision validity by performing weighted voting matching on the dispute labeling table based on the arbitration rules includes: allocating departmental weight regions to the dispute labeling table based on the arbitration rules to form a departmental weight distribution table; identifying departments whose opinions contradict each other in terms of region and time sequence in the departmental weight distribution table to generate internal contradiction markers; temporarily freezing the weights of the departments corresponding to the internal contradiction markers to form a valid voting rights reorganization; and performing weighted voting matching on the dispute labeling table based on the valid voting rights reorganization to generate decision validity.

[0071] Based on the arbitration rules, a departmental weight distribution table is formed by allocating departmental weights to the dispute labeling table. The weight mapping layer of the arbitration rules assigns a basic weight to the participating departments for each valid record in the dispute labeling table. This basic weight is derived from the product of the historical adoption rate in the joint dispute pattern table and the primary responsibility amplification coefficient. The primary responsibility amplification coefficient is set to 1.2, applicable to the primary responsible department for the disputed indicator; for non-primary responsible departments, the historical adoption rate is directly used as the basic weight without multiplying by the amplification coefficient. In the dispute labeling table (Neurosurgery, Hematoma Volume), the participating departments for the acute phase records include Neurosurgery, Neurology, Intensive Care Unit, and Radiology. The arbitration rules allocate voting weights based on the product of each department's historical adoption rate for the hematoma volume indicator and the primary responsibility amplification coefficient (1.2 for the primary responsible department, no amplification for non-primary responsible departments): Neurosurgery 0.84, Intensive Care Unit 0.59, Neurology 0.42, and Radiology 0.25. The department weight distribution table is constructed using the disputed annotation table record number as the primary index and the participating department number as the secondary index. The number of rows in the department weight distribution table equals the number of valid records in the disputed annotation table, and the number of columns equals the number of participating departments. When the arbitration rules assign weights to records in different sections of the disputed annotation table, if the number of participating departments in the baseline segment records is less than that in the acute phase segment records, the number of non-zero columns in the corresponding row of the department weight distribution table will be reduced accordingly. Records in the disputed annotation table that are excluded by the arbitration rules are not generated rows in the department weight distribution table; the department weight distribution table only covers disputed annotation table records that are deemed valid by the arbitration rules. The columns of the department weight distribution table summarize the total weight of each department in all disputed records. The total weight reflects the cumulative influence of the department in the overall arbitration process, with neurosurgery having the highest total weight.

[0072] The department weight distribution table identifies departments with conflicting opinions across regional and temporal dimensions, generating internal conflict markers. If the opinion labels of the same department in different dispute annotation tables are opposite in direction, an internal conflict exists between these two records. For example, in the department weight distribution table, the neurosurgery department's opinion label for (hematoma volume, acute phase) is "active intervention," but its opinion label for (hematoma volume, baseline phase) is "conservative observation," indicating a reversal of opinion direction across the two intervals of the hematoma volume index. Internal conflict markers are indexed by department number, linking the dispute annotation table record number pair that triggered the conflict, the combination of conflicting opinion labels, and the index dimension where the conflict occurs. The presence of internal conflict markers indicates questionable consistency in the department's judgments across different treatment stages. Departments found to have internal conflicts in the department weight distribution table are downgraded in the arbitration reliability assessment, and their voting weight is temporarily reduced in the effective voting rights reorganization calculation. Departments without internal conflicts in the department weight distribution table are not recorded in the internal conflict marker. Departments without internal conflicts maintain their original weights assigned by the arbitration rules during the effective voting rights reorganization. The internal conflict marker detection covers all department columns in the department weight distribution table. Opinion tags across rows in each column are compared pairwise to form conflict pair sets. Departments with non-empty conflict pair sets are marked with internal conflict markers. In this case, the direction of the department opinion tags in each record of the dispute marker table remains consistent between the baseline and acute phases; no reversal of opinion direction was detected. Therefore, no internal conflict marker is generated, and the weights of each department remain unchanged during the effective voting rights reorganization.

[0073] A temporary weight freeze is applied to the departments corresponding to the internal conflict markers to form a valid voting rights reorganization. The department weight corresponding to each record in the internal conflict marker is multiplied by a temporary freeze coefficient of 0.4 in the valid voting rights reorganization. This freeze coefficient compresses the department's voting weight to 40% of its original value, reflecting a discount on the reliability of the opinions of the conflicting departments. If neurosurgery triggers a freeze in the internal conflict marker, its base weight of 0.84 in the department weight distribution table is reduced to 0.34 after the freeze coefficient is applied, and its voting weight is updated to 0.34 in the valid voting rights reorganization. Departments without records in the internal conflict marker directly inherit the base weight from the department weight distribution table in the valid voting rights reorganization without any adjustments. The weight values ​​of departments without internal conflicts in the valid voting rights reorganization are completely consistent with the department weight distribution table. The structure of the valid voting rights reorganization is the same as the department weight distribution table, but the weight values ​​may be adjusted through a freeze. The freeze scope of the internal conflict marker is precise to the department level. The weight of the same department is uniformly adjusted in all disputed record rows in the valid voting rights reorganization, without distinguishing between specific disputed records. The sum of the weight values ​​of the valid voting weight groups is renormalized within each row to ensure that the sum of the voting weights of the participating departments recorded in each dispute labeling table is equal to 1.0. The normalization benchmark is the sum of the adjusted weights of all participating departments in that row.

[0074] Based on the reorganization of effective voting rights, a weighted voting matching process is performed on the disputed label table to generate decision validity. After normalizing the voting weights of the participating departments corresponding to each disputed label record in the effective voting rights reorganization, the weighted voting matching uses the opinion labels of each department as voting options and the normalized weights as voting weights to calculate the weighted vote rate for each opinion label. In the acute phase records of the disputed label tables (Neurosurgery, Hematoma Volume), among the participating departments, Neurosurgery's opinion is active intervention, Neurology's opinion is conservative observation, Critical Care Medicine's opinion is active intervention, and Radiology's opinion is conditional intervention. After normalization of the effective voting rights reorganization, the weights of each department are 0.40, 0.20, 0.28, and 0.12, respectively, and the weighted vote rate for active intervention is 0.68. Decision validity is represented by the adoption probability of the opinion label with the highest weighted vote rate in the disputed label table records. The decision validity for the acute phase records (Neurosurgery, Hematoma Volume) is 0.68, exceeding the high validity threshold of 0.6, and is marked as high validity. In the effective voting rights restructuring, the opinion tags of departments with internal conflicts are frozen, but their weight is low. This freezing reduces the influence of conflicting departments on decision-making effectiveness. The decision-making effectiveness values ​​recorded in the dispute labeling table more accurately reflect the consensus level of departments without conflicts. Records with decision-making effectiveness values ​​below 0.4 in the decision-making effectiveness sequence are marked as low consensus positions. The dispute labeling table records corresponding to these low consensus positions are included in the key analysis scope when generating multi-evolution path warnings. After calculating the decision-making effectiveness of all valid records in the dispute labeling table, the mean of the decision-making effectiveness sequence reflects the overall consensus level of this multidisciplinary consultation. A mean below 0.55 triggers a global low consensus warning.

[0075] The low consensus interval of decision validity is used to mark multiple evolution path warnings and incorporate them into the output of multidisciplinary collaborative treatment plans. Low consensus positions in the decision validity sequence (below 0.4) correspond to the indicator dimensions with the greatest departmental disagreement in the dispute labeling table. The distribution of opinion labels at low consensus positions exhibits a multi-peaked characteristic, with no single dominant opinion clearly defined. Inapplicable records excluded by arbitration rules and low consensus positions in the decision validity sequence together define the coverage of the multi-evolution path warning. Disputed indicators in inapplicable records have the highest evolution uncertainty because effective arbitration cannot be completed. The multi-evolution path warning uses the record number in the dispute labeling table as an index, linking the corresponding decision validity value, the estimated number of evolution paths, and the combination of departmental opinion labels for each path. The estimated number of evolution paths equals the number of different values ​​of opinion labels in the participating departments at the low consensus position. When a global low consensus warning is triggered when the mean decision validity is below 0.55, the multidisciplinary collaborative treatment plan adds an uncertainty warning label to the global warning field. The warning label includes a list of low consensus indicators and a suggested review time window. The multidisciplinary collaborative treatment plan integrates all calculation results of the decision validity sequence. Opinion tags corresponding to high-validity records are written into the multidisciplinary collaborative treatment plan as principal recommendations, while low-validity records and multi-evolutionary path warnings are written into the multidisciplinary collaborative treatment plan as parallel candidate opinions. The multidisciplinary collaborative treatment plan organizes its output according to the three partitions of the dynamic diagnostic benchmark. Each partition contains high-validity principal recommendations, low-validity candidate opinions, and multi-evolutionary path warnings within that partition. The partition structure of the multidisciplinary collaborative treatment plan completely corresponds to the partition boundaries of the dynamic diagnostic benchmark.

[0076] To implement the above-described method embodiments, a remote multidisciplinary collaborative decision-making method for emergency treatment of cerebral hemorrhage is proposed to achieve the corresponding functional and technical effects. See also... Figure 2 , Figure 2 This application provides a structural block diagram of a remote multidisciplinary collaborative decision-making system 200 for emergency treatment of cerebral hemorrhage, comprising:

[0077] The benchmark construction unit 201 is used to acquire patient image data and clinical sign data, and synchronously fuse the patient image data and clinical sign data to construct a dynamic diagnostic benchmark.

[0078] Feature recognition unit 202 is used to analyze the fluctuation of the disease condition through the dynamic diagnostic benchmark to identify the pattern of opinion divergence, identify the excessively consistent intervals based on the pattern of opinion divergence, reduce the weight of heterogeneous opinions to form effective consultation features, and perform disease feature spectrum analysis on the effective consultation features to locate target feature items.

[0079] The graph construction unit 203 is used to delineate departmental diagnosis and treatment labels based on the target feature items, assign high-frequency change departments in the departmental diagnosis and treatment labels as core change departments to generate decision complexity weights, and perform partitioning and calibration of the dynamic diagnostic benchmark with the departmental diagnosis and treatment labels and the decision complexity weights to establish a collaborative decision graph.

[0080] The conflict detection unit 204 is used to combine the collaborative decision-making graph with the dynamic diagnostic benchmark and the decision complexity weight to complete the scheme difference completion to form a modified scheme set, perform consistency verification on the modified scheme set to form a diagnosis and treatment conflict index, and determine the dispute labeling table based on the disciplinary opinion deviation distribution of the diagnosis and treatment conflict index.

[0081] Arbitration output unit 205 is used to establish arbitration rules by mapping the dispute label table with the departmental diagnosis and treatment labels, perform weighted voting matching on the dispute label table based on the arbitration rules to generate decision validity, and use the low consensus interval of the decision validity to mark multiple evolution path warnings and incorporate them into the output of multidisciplinary collaborative diagnosis and treatment plan.

[0082] The aforementioned emergency remote multidisciplinary collaborative decision-making system 200 for cerebral hemorrhage can implement one of the above-described method embodiments for emergency remote multidisciplinary collaborative decision-making for cerebral hemorrhage. The options in the above method embodiments are also applicable to this embodiment and will not be detailed here. The remaining content of this application embodiment can be referred to the content of the above method embodiments, and will not be repeated in this embodiment.

[0083] The above embodiments are not an exhaustive list based on the present invention, and there may be many other embodiments not listed. Any substitutions and improvements made without departing from the concept of the present invention are within the protection scope of the present invention.

Claims

1. A remote multidisciplinary collaborative decision-making method for emergency treatment of cerebral hemorrhage, characterized in that, include: Acquire patient imaging data and clinical sign data, and synchronously fuse the patient imaging data and clinical sign data to construct a dynamic diagnostic benchmark; The dynamic diagnostic benchmark is used to analyze the fluctuation of the disease and identify the pattern of opinion divergence. Based on the pattern of opinion divergence, the excessively consistent intervals are identified and the heterogeneous opinions are retained to form effective consultation features. The disease feature spectrum is analyzed to locate the target feature items of the effective consultation features. Based on the target feature items, departmental diagnosis and treatment labels are defined. Departments with high change frequency in the departmental diagnosis and treatment labels are marked as core change departments to generate decision complexity weights. Based on the departmental diagnosis and treatment labels and the decision complexity weights, the dynamic diagnostic benchmark is partitioned and a collaborative decision graph is established. The collaborative decision-making graph and the dynamic diagnostic benchmark are combined with the decision complexity weight to complete the scheme difference and form a modified scheme set. The consistency check of the modified scheme set is performed to form a diagnosis-treatment conflict index. The dispute labeling table is determined based on the disciplinary opinion deviation distribution of the diagnosis-treatment conflict index. Arbitration rules are established by mapping the dispute labeling table with the departmental diagnosis and treatment labels. Based on the arbitration rules, weighted voting matching is performed on the dispute labeling table to generate decision validity. The low consensus interval of the decision validity is used to mark multiple evolution paths and output multidisciplinary collaborative diagnosis and treatment plans.

2. The method according to claim 1, characterized in that, The process of analyzing disease fluctuations and identifying patterns of disagreement using the dynamic diagnostic benchmark includes: The dynamic diagnostic benchmark is used to extract disease condition indicators in time series according to the assessment dimensions of each department to generate departmental indicator time series groups; The time series groups of the departmental indicators are used to calculate inter-departmental consistency and form a consistency evaluation sequence. The consistency assessment sequence detection identifies two types of sudden drops: those triggered by initial objections from a single department and those triggered by simultaneous disagreements from multiple departments, forming a drop classification label table. Based on the aforementioned sudden drop classification label table, the two types of sudden drops are weighted and labeled to determine the location where the disagreement arises, thus forming the disagreement pattern.

3. The method according to claim 1, characterized in that, The step of assigning high-frequency change departments as core change departments in the departmental diagnosis and treatment labels to generate decision complexity weights includes: The change frequency sequence is generated by extracting the records of opinion changes in each department according to the time sequence of the departmental diagnosis and treatment labels. The frequency sequence of changes is used to identify high-frequency changing departments to form a core changing department table. The core change department table is weighted by the contribution of diagnosis and treatment to generate a core weight distribution; The decision complexity weights are generated by weighting and integrating the core weight distribution with the frequency sequence of changes.

4. The method according to claim 1, characterized in that, The step of combining the collaborative decision-making graph with the dynamic diagnostic benchmark and the decision complexity weights to complete the scheme difference completion and form a corrected scheme set includes: The collaborative decision-making graph and the dynamic diagnostic benchmark are matched segment by segment to form a preliminary scheme sequence. The preliminary scheme sequence detection scheme covers the missing segments to form a scheme missing marker table; The missing label table of the scheme and the dynamic diagnostic benchmark are used to extract missing correlation features to generate associated missing labels; Based on the missing association markers and the decision complexity weights, differentiated completion is performed to form a set of correction schemes.

5. The method according to claim 1, characterized in that, The dispute labeling table based on the disciplinary opinion deviation distribution of the diagnosis and treatment conflict index includes: The opinion magnitude of each department was statistically analyzed based on the aforementioned medical conflict index to form a departmental opinion magnitude chart; The deviation degree sequence is generated by calculating the symmetry axis offset of the aforementioned departmental opinion amplitude diagram. The deviation degree sequence identifies situations where the same department simultaneously deviates from the median in multiple diagnostic and treatment dimensions, generating multidimensional synchronous deviation markers. Based on the multidimensional synchronous deviation markers, the corresponding diagnosis and treatment conflict index is extracted from the departmental opinion amplitude diagram to form a dispute annotation table.

6. The method according to claim 1, characterized in that, The process of mapping the dispute labeling table to the departmental diagnosis and treatment labels to establish arbitration rules includes: A co-occurrence frequency map was generated by statistically analyzing the co-occurrence frequency of the disputed labeling table and the departmental diagnosis and treatment labels. The co-occurrence frequency map is combined with the dispute annotation table to extract the opinion groups of highly co-occurring departments and historical voting records to form a joint dispute pattern table; The joint dispute pattern table identifies combinations of opinions whose historical arbitration results have never been adopted, generating invalid co-occurrence markers. Arbitration rules are established by mapping a merging strategy after removing invalid opinions from the joint dispute pattern table based on the invalid co-occurrence markers.

7. The method according to claim 1, characterized in that, The step of generating decision validity by weighted voting matching of the dispute labeling table based on the arbitration rules includes: Based on the arbitration rules, the dispute labeling table is assigned departmental weight regions to form a departmental weight distribution table; The department weight distribution table identifies departments where opinions on the regional and temporal dimensions contradict each other, generating internal contradiction markers. The weights of the departments corresponding to the aforementioned internal conflict markers are temporarily frozen to form a valid reorganization of voting rights; Based on the reorganization of valid voting rights, a weighted voting match is performed on the dispute labeling table to generate decision validity.

8. The method according to claim 3, characterized in that, The step of identifying high-frequency changing departments from the frequency sequence to form a core changing department table includes: The frequency variation sequence is processed by department to extract the variation intervals segment by segment to generate interval time series groups; The departments whose interval time sequence group continuously shrinks to within a preset tolerance are marked as shrinkage zones. The contraction zone markers are continuously labeled with risk classification to generate variable risk annotations; A core list of departments with changes is generated based on the aforementioned risk annotations.

9. The method according to claim 6, characterized in that, The process of generating invalid co-occurrence markers by identifying combinations of opinions whose historical arbitration results have never been adopted from the joint dispute pattern table includes: The failure rates of historical individual arbitrations were extracted from the opinions of each department in the joint dispute model table to form a failure rate sequence; The failure rate sequence identifies departmental opinions that focus on the rapid progression phase of the disease, generating a progression-phase failure marker. For the opinions of the departments corresponding to the failure markers in the progress phase, establish a high-risk phase arbitration file generation phase arbitration strategy group; Invalid co-occurrence markers are generated by selecting opinion combinations that are never adopted based on the aforementioned temporal arbitration strategy group.

10. A remote multidisciplinary collaborative decision-making system for emergency treatment of cerebral hemorrhage, characterized in that, include: A benchmark construction unit is used to acquire patient image data and clinical sign data, and synchronously fuse the patient image data and clinical sign data to construct a dynamic diagnostic benchmark. The feature recognition unit is used to analyze the fluctuation of the disease through the dynamic diagnostic benchmark to identify the pattern of opinion divergence, identify the excessively consistent intervals based on the pattern of opinion divergence, reduce the weight of heterogeneous opinions to form effective consultation features, and perform disease feature spectrum analysis on the effective consultation features to locate target feature items. The graph construction unit is used to delineate departmental diagnosis and treatment labels based on the target feature items, assign high-frequency change departments in the departmental diagnosis and treatment labels as core change departments to generate decision complexity weights, and perform partitioning and calibration on the dynamic diagnostic benchmark based on the departmental diagnosis and treatment labels and the decision complexity weights to establish a collaborative decision graph. The conflict detection unit is used to combine the collaborative decision-making graph with the dynamic diagnostic benchmark and the decision complexity weight to complete the scheme difference completion to form a modified scheme set, perform consistency verification on the modified scheme set to form a diagnosis and treatment conflict index, and determine the dispute labeling table based on the disciplinary opinion deviation distribution of the diagnosis and treatment conflict index. The arbitration output unit is used to establish arbitration rules by mapping the dispute label table with the departmental diagnosis and treatment labels, perform weighted voting matching on the dispute label table based on the arbitration rules to generate decision validity, and output multidisciplinary collaborative diagnosis and treatment plans by marking multiple evolution paths through the low consensus interval of the decision validity.

Images