A smart management method and system applicable to drug clinical projects

Abstract

This invention discloses an intelligent management method and system applicable to drug clinical trials, relating to the field of drug clinical trial management technology. It includes collecting raw project activity records containing information such as trial site number and subject identification from the drug clinical trial execution system to form a collection. This collection undergoes multi-dimensional structured processing to generate a unified project view comprising subject time-series activity logs, trial site operational status snapshots, clinical document association networks, and key event sequences. Compliance modeling, effectiveness assessment, compliance checks, and logical verification are performed on the four types of views respectively. After generating corresponding specialized results, these are input into an intelligent analysis model for fusion and deduction, ultimately outputting a comprehensive project report including project risk warnings, process deviation location, and resource gap alerts. This invention achieves multi-dimensional integration and specialized fusion analysis of drug clinical trial data, optimizing the data analysis and application methods of clinical trial management.

Description

Technical Field

[0001] This invention belongs to the field of drug clinical project management technology, specifically an intelligent management method and system applicable to drug clinical projects. Background Technology

[0002] Currently, in the field of drug clinical trial management, raw data related to trial sites, subjects, examinations, test values, timestamps, and clinical documents are mostly collected manually from the project execution system. This data is recorded and organized in a single dimension, without multi-dimensional structured integration. Clinical project data is presented in a fragmented and independent manner. Existing technologies can only perform basic analysis on single types of data and cannot create a unified data view covering subject activities, trial site operations, clinical documents, and key events, resulting in significant limitations in data application dimensions.

[0003] Current drug clinical trial management methods cannot provide differentiated and customized processing based on subject activity timelines, trial site status, clinical document networks, and key event sequences, making it difficult to generate corresponding specialized analysis results. The analysis results of various data types are independent and lack fusion and extrapolation through intelligent analysis models, failing to simultaneously output project risk warnings, process deviation locations, and resource gap alerts. It is necessary to perform multi-dimensional structured processing on the original project activity records to form a unified project view with four fixed structures. Simultaneously, it is necessary to perform specialized processing on each of the four unified project views and fuse and extrapolate the results to overcome the shortcomings of existing technologies. Summary of the Invention

[0004] This invention aims to solve at least one of the technical problems existing in the prior art;

[0005] Therefore, this invention proposes an intelligent management method suitable for drug clinical projects, comprising:

[0006] The original project activity records, including trial site number, subject identification, examination type, test value, event timestamp, and clinical document identifier, are collected from the drug clinical project execution system to form a set of original activity records;

[0007] Multi-dimensional structured processing is performed on the original activity record set to generate a unified project view that includes subject time-series activity flow, trial site operation status snapshots, clinical document association networks, and key event sequences;

[0008] Compliance modeling is performed on the subject's time-series activity flow to generate individual subject compliance trajectories; efficacy assessment is performed on the snapshot of the trial site's operational status to generate trial site efficacy data; compliance checks are performed on the clinical document association network to generate a document compliance status map; and logical verification is performed on the key event sequence to generate an event logic consistency report.

[0009] The individual subject compliance trajectory, the test site effectiveness data, the document compliance status map, and the event logic consistency report are input into the intelligent analysis model for fusion and deduction, generating a comprehensive project report that includes project risk warnings, process deviation location, and resource gap prompts.

[0010] Furthermore, compliance modeling is performed on the temporal activity flow of the subjects to generate individual subject compliance trajectories, including:

[0011] From the subject's time-series activity log, extract the subject's actual activity records according to the preset experimental protocol milestone nodes;

[0012] The extracted records of the subjects' actual activities were compared with the predefined standard timeline of the trial protocol to calculate the compliance deviation measure for each activity.

[0013] By connecting the compliance deviation measures of each activity in chronological order and introducing a time decay function to model the continuous impact of historical deviations, the individual subject compliance trajectory is formed, which reflects the dynamic change process of subject compliance.

[0014] Furthermore, a performance evaluation is performed on the snapshot of the test site's operational status to generate test site performance data, including:

[0015] The data sequences of indicators such as group entry / exit rate, data entry timeliness rate, question clarification cycle, and scheme deviation rate are analyzed from the snapshot of the operational status of the test site.

[0016] Benchmark reference lines are established for the indicators of group entry rate, data entry timeliness rate, question clarification cycle and plan deviation rate, respectively. The benchmark reference lines are determined based on historical project data or industry standard data.

[0017] Compare the current indicator data series with the corresponding benchmark reference line to calculate the performance deviation of each indicator at different time points;

[0018] The performance deviation of all indicators is integrated and coded in terms of time and indicator dimensions to form the performance data of the test point, which characterizes the overall and individual sub-item performance of the test point.

[0019] Furthermore, a compliance check is performed on the clinical document association network to generate a document compliance status map, including:

[0020] The clinical document association network is analyzed to identify source documents, derived documents, and the citation, verification, and supplementary relationship chains between documents;

[0021] The citation, verification, and supplementary relationship chains between each document are matched with the pre-defined clinical document management standards to check the completeness of the chain, the correctness of the logical order, and the completeness of the necessary metadata.

[0022] Mark the compliance status of each link in the citation, verification and supplementation relationship between documents, and weight the compliance importance of the document itself according to the centrality of the document node in the network;

[0023] All document nodes and their compliance status relationships are visualized and encoded to form a document compliance status graph presented in the form of a network graph, where the node size represents the importance of compliance and the edge attributes represent the specific compliance status.

[0024] Furthermore, logical verification is performed on the key event sequence to generate an event logical consistency report, including:

[0025] Identify event pairs or groups of events with causal, temporal, or conditional constraints from the key event sequence;

[0026] Based on the trial protocol and clinical logic rules, define the expected constraints for each identified event pair or event group;

[0027] By comparing the actual occurrence time and state of events in the key event sequence with the expected constraint relationship, logical conflicts of events that violate the expected constraint relationship are detected.

[0028] All detected logical conflicts in events are attributed, categorized, and classified according to severity, resulting in a structured report on the logical consistency of the events.

[0029] Furthermore, the individual subject compliance trajectory, the test site efficacy data, the document compliance status map, and the event logic consistency report are input into an intelligent analysis model for fusion and deduction, generating a comprehensive project report that includes project risk warnings, process deviation location, and resource gap alerts, including:

[0030] A knowledge graph reasoning engine is established in the intelligent analysis model. The nodes of the knowledge graph reasoning engine are composed of the individual subject compliance trajectory, test point performance data, document compliance status graph, and key entities and states in the event logic consistency report.

[0031] The propagation and influence rules between entity states are defined in the knowledge graph reasoning engine to simulate how the deterioration of the individual subject's compliance trajectory is transmitted to the test point performance data, and how the defects in the document compliance state graph cause conflicts in the event logic consistency report.

[0032] The knowledge graph reasoning engine is started to perform multiple rounds of state propagation and deduction. When the deduced entity state exceeds the preset risk threshold, the corresponding entity and its state are marked as risk points.

[0033] Aggregate all marked risk points, and generate project risk warnings, process deviation locations, and resource gap prompts based on the type and correlation of the risk points, and integrate them into the comprehensive project report;

[0034] The knowledge graph reasoning engine defines rules for the propagation and influence between entity states, including:

[0035] Define a negative influence rule between subject compliance status and trial site enrollment rate: when the subject compliance trajectory shows that multiple subjects at a specific trial site have compliance consistently below a threshold, the expected enrollment rate of the trial site will be reduced.

[0036] Define the triggering rules between document compliance status and event logic conflict. That is, when the verification link of a key source document is missing in the document compliance status graph, a logical conflict about data traceability in the event logic consistency report will be triggered.

[0037] Define a weighted contribution rule between the deviation rate of the scheme in the performance data of the test sites and the overall risk level of the project. That is, the deviation rates of different test sites are accumulated according to their weights, which affects the assessment of the overall risk level of the project.

[0038] Furthermore, it also includes:

[0039] Based on the comprehensive project report, a targeted management intervention plan is generated through a decision support engine. The management intervention plan includes a visit plan adjustment plan for specific trial sites, a follow-up enhancement plan for specific subjects, a list of rectification requirements for designated clinical documents, and a review and verification process for key events.

[0040] The management intervention plan is converted into a queue of operation instructions that can be recognized and executed by the drug clinical project execution system, and the queue of operation instructions is distributed to the corresponding drug clinical project execution system terminal.

[0041] Furthermore, the generation of targeted management intervention plans based on the comprehensive project report through a decision support engine includes:

[0042] The comprehensive project report is analyzed to extract specific project risk warning items, process deviation location coordinates, and resource gap prompts.

[0043] Each project risk warning item, process deviation location coordinate, and resource gap prompt detail is matched with a pre-set intervention measure knowledge base to identify one or more candidate intervention actions for each warning, location, or prompt.

[0044] Based on the performance data and real-time resource status of the test sites, the execution cost, expected effect, and feasibility of the candidate intervention actions are evaluated and ranked.

[0045] The candidate intervention actions with the highest evaluation ranking are selected and combined according to different intervention subjects to form a visit plan adjustment plan for specific trial sites, a follow-up enhancement plan for specific subjects, a list of rectification requirements for specified clinical documents, and a review and verification process for key events, which together constitute the management intervention plan.

[0046] The evaluation and ranking of candidate intervention actions based on the effectiveness data and real-time resource status at the test sites, considering their execution cost, expected effects, and feasibility, includes:

[0047] Obtain the current workload and response capability indicators of the target test point from the performance data of the test point;

[0048] Obtain information on the number of available inspectors, budget margin, and time window from the real-time status of resources;

[0049] Calculate the increment of each candidate intervention action on the current workload, the degree of improvement on the response capability index, the number of monitors required, the budget consumed, and the time window occupied;

[0050] The calculated increment, lift, number of inspectors, budget consumption, and time window occupancy are compared with the preset constraints to screen out feasible candidate intervention actions that meet all constraints.

[0051] Among the feasible candidate intervention actions, a comprehensive score is calculated based on a preset utility function, and the evaluation and ranking are carried out according to the comprehensive score from high to low.

[0052] Furthermore, the management intervention plan is converted into a queue of operational instructions that can be recognized and executed by the drug clinical program execution system, including:

[0053] The management intervention plan includes adjustments to the visitation plan for specific trial sites, enhanced follow-up plans for specific subjects, a list of rectification requirements for designated clinical documents, and a review and verification process for key events, which are broken down into atomic-level operational tasks.

[0054] Each atomic-level operation task is mapped to a corresponding standard application programming interface call command or database operation statement in the drug clinical project execution system;

[0055] Based on the logical dependencies and time requirements between atomic-level operation tasks, all standard application programming interface call commands and database operation statements are scheduled and orchestrated to generate the operation instruction queue with a clear execution order.

[0056] Furthermore, the present invention also includes an intelligent management system suitable for drug clinical projects, the system including a memory, a processor, and a computer program stored in the memory and running on the processor, wherein the processor, when executing the computer program, implements the steps of the intelligent management method suitable for drug clinical projects described above.

[0057] Compared with the prior art, the beneficial effects of the present invention are:

[0058] Multi-dimensional structured processing is performed on the original project activity record set containing trial site number, subject identification, examination type, test value, event timestamp, and clinical document identifier. This process generates a unified project view that includes subject time-series activity flow, trial site operation status snapshot, clinical document association network, and key event sequence. This transforms fragmented original project activity data into a multi-dimensional, collaborative, and structured presentation, constructing a complete data system covering subjects, trial sites, clinical documents, and key events. This eliminates the information fragmentation problem caused by single-dimensional data processing and achieves standardized integration and presentation of full-dimensional clinical project data.

[0059] The system performs compliance modeling on the subject's time-series activities to generate individual subject compliance trajectories; performs performance evaluation on snapshots of the trial site's operational status to generate trial site performance data; performs compliance checks on the clinical document association network to generate document compliance status maps; and performs logical verification on key event sequences to generate event logic consistency reports. These four types of specialized results are input into an intelligent analysis model to complete fusion and deduction, forming a comprehensive project report that includes project risk warnings, process deviation location, and resource gap alerts. This adapts to the analysis and processing needs of data from different dimensions, enabling the collaborative integration of multi-dimensional specialized analysis results and simultaneously outputting multiple types of project management reference information, breaking through the application boundaries of single-dimensional analysis results. Attached Figure Description

[0060] Figure 1 This is a flowchart illustrating the steps of an intelligent management method applicable to drug clinical projects as described in this invention.

[0061] Figure 2 A flowchart for generating performance data at test sites;

[0062] Figure 3 A line graph showing the trajectory of participant compliance;

[0063] Figure 4 Grouped bar chart for resource allocation in drug clinical trial management intervention programs;

[0064] Figure 5 This is a trend chart showing the relationship between task priority and deadline pressure. Detailed Implementation

[0065] The technical solution of the present invention will be clearly and completely described below with reference to the embodiments. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0066] See Figure 1 The overall implementation scheme of an intelligent management method suitable for drug clinical projects is as follows:

[0067] This method collects raw project activity records from the drug clinical trial execution system. These records include trial site number, subject identification, examination type, test values, event timestamps, and clinical document identifiers, forming a raw activity record set. Multi-dimensional structured processing is performed on this raw activity record set, integrating and associating the scattered activity records to generate a unified project view containing subject time-series activity logs, trial site operational status snapshots, clinical document association networks, and key event sequences. Compliance modeling is performed on the subject time-series activity logs, generating individual subject compliance trajectories. Performance evaluation is performed on the trial site operational status snapshots, generating trial site performance data. Compliance checks are performed on the clinical document association networks, generating document compliance status maps. Logical verification is performed on the key event sequences, generating event logical consistency reports. The individual subject compliance trajectories, trial site performance data, document compliance status maps, and event logical consistency reports are input into an intelligent analysis model for fusion and deduction. The intelligent analysis model analyzes and infers from these multi-source inputs, generating a comprehensive project report including project risk warnings, process deviation location, and resource gap alerts.

[0068] In one embodiment of the invention, compliance modeling is performed on the subject's time-series activity flow to generate an individual subject compliance trajectory. Actual activity records of the subjects are extracted from the subject's time-series activity flow according to preset trial protocol milestone nodes. The extracted actual activity records are compared with a predefined trial protocol standard timeline, and this comparison process calculates a compliance deviation metric for each activity. The compliance deviation metrics for each activity are connected in chronological order, and a time decay function is introduced to model the continuous impact of historical deviations. This modeling process forms an individual subject compliance trajectory that reflects the dynamic changes in subject compliance.

[0069] A performance evaluation is performed on snapshots of the test site's operational status, generating test site performance data. (See also...) Figure 2 The data sequences of enrollment rate, data entry timeliness rate, clarification period for questions, and deviation rate were extracted from snapshots of the experimental site's operational status. Benchmark reference lines were established for each of these indicators, based on historical project data or industry standard data. The current indicator data sequence was compared with the corresponding benchmark reference line, and this comparison process calculated the performance deviation of each indicator at different time points. The performance deviations of all indicators were integrated and encoded along the time and indicator dimensions. This integration and encoding process formed experimental site performance data characterizing the overall operational performance of the experimental site and its individual components.

[0070] In practice, compliance modeling is performed on the subject's time-series activity log to generate individual subject compliance trajectories. From the subject's time-series activity log, actual activity records are extracted according to pre-defined trial protocol milestones. The original project activity records (follow-up, medication, sampling, etc.) recorded in the drug clinical trial execution system are structured and aggregated into a time-ordered subject time-series activity log for each subject. The process of extracting actual activity records from the subject's time-series activity log is based on milestones defined in the trial protocol, such as screening visits, first-time medication visits, and week 8 assessment visits. The system automatically identifies activity entries in the log that match these milestones and filters out activity records associated with pre-defined milestones. The extracted actual activity records are compared with a pre-defined trial protocol standard timeline. The drug clinical trial protocol standard timeline pre-defines the standard execution time for each milestone activity; for example, the first-time medication visit should be completed within 3 days of successful screening, and the time window for the week 8 assessment visit is week 8 plus or minus 7 days. The adherence deviation metric for each activity is calculated by comparing the timestamps of events recorded in the participants' actual activity logs with the corresponding standard time points in the standard timeline of the trial protocol. This deviation metric can be an absolute time deviation measured in days. In practice, the adherence deviation metrics for each activity are connected chronologically, and a time decay function is introduced to model the ongoing impact of historical deviations. This modeling process creates an individual participant adherence trajectory that reflects the dynamic changes in participant adherence. It can be understood that the time series of adherence deviation metrics reflects the participants' adherence performance at different time points, and the time decay function is used to simulate the gradual weakening of the impact of early deviations on the current overall adherence assessment; for example, an exponential decay function can be used. Calculate historical deviation over time Residual effects:

[0071]

[0072] in: It is the first The absolute value of compliance deviation measure for each activity This refers to the time when the event took place. It is the current time. It is the decay coefficient. The individual subject compliance trajectory is a quantitative time series of each subject's compliance status, used to characterize the trend of compliance changes as the project progresses.

[0073] In practice, performance evaluation is performed on snapshots of the pilot site's operational status to generate pilot site performance data. Data sequences of indicators such as enrollment rate, data entry timeliness rate, clarification period, and protocol deviation rate are extracted from these snapshots. The pilot site operational status snapshots in the unified project view contain the raw values ​​of various operational indicators for each pilot site at a specific time slice. The analysis process identifies and extracts the enrollment rate indicator data sequence from the snapshots. The enrollment rate is the number of subjects successfully screened and enrolled per unit time; the data entry timeliness rate is the percentage of activities that complete data entry within a specified time window; the clarification period is the average time from issuing a data-related question to receiving a satisfactory answer; and the protocol deviation rate is the ratio of the number of recorded protocol deviation events to the total number of subject visits. These indicators are extracted in time series form. Benchmark reference lines are established for the enrollment rate, data entry timeliness rate, clarification period, and protocol deviation rate indicators, determined based on historical project data or industry standard data. In some embodiments, the baseline reference line is set based on statistical values ​​of historical performance data from similar drug clinical trials. For example, the baseline for enrollment rate is set as 80% of the average enrollment rate of historical trials during the same period. In some embodiments, for entirely new trials, the baseline reference line can be determined based on recommended values ​​in industry benchmark survey reports or regulatory guidelines. The current indicator data sequence is compared with the corresponding baseline reference line to calculate the efficacy deviation of each indicator at different time points. Optionally, the efficacy deviation is calculated as the difference or ratio between the current indicator value and the baseline reference line value. The efficacy deviation can be a scalar value used to quantify the direction and magnitude of the deviation of the current performance relative to the baseline. The efficacy deviations of all indicators are integrated and encoded in both time and indicator dimensions to form trial point efficacy data characterizing the overall and individual sub-item performance of the trial point. The integration and encoding process can be represented by a multi-dimensional vector, where each element in the vector corresponds to the efficacy deviation of a specific indicator at a specific time point, or by extracting features from the time series data to form a fixed-dimensional feature vector. It is understandable that the performance data of the pilot site is a structured data object that comprehensively reflects the performance status of the pilot site in multiple dimensions such as enrollment, data management, question handling, and protocol adherence, as well as its changes over time.

[0074] In one embodiment of the present invention, a compliance check is performed on the clinical document association network to generate a document compliance status graph. The clinical document association network is parsed, identifying source documents, derived documents, and the citation, verification, and supplementary relationship chains between documents. Each citation, verification, and supplementary relationship chain between documents is matched against preset clinical document management specifications. This matching process checks the completeness of the chain, the correctness of the logical order, and the completeness of necessary metadata. Each citation, verification, and supplementary relationship chain between documents is marked with a compliance status, and the compliance importance of the document itself is weighted according to the centrality of the document node in the network. All document nodes and their compliance status relationships are visualized and encoded to form a document compliance status graph presented in the form of a network graph, where node size represents compliance importance and edge attributes represent specific compliance statuses.

[0075] Logical validation is performed on the critical event sequence, generating an event logical consistency report. Event pairs or groups with causal, temporal, or conditional constraints are identified from the critical event sequence. Based on the trial protocol and clinical logic rules, expected constraints are defined for each identified event pair or group. The actual timing and state of events in the critical event sequence are compared with the expected constraints; this comparison process detects logical conflicts that violate the expected constraints. All detected logical conflicts are attributed, categorized, and graded for severity, resulting in a structured event logical consistency report.

[0076] In practice, compliance checks are performed on the clinical document association network to generate a document compliance status graph. The clinical document association network is then parsed, identifying source documents, derived documents, and the citation, verification, and supplementary relationship chains between documents. This network is constructed from the clinical document identifiers in the original project activity records and the logical relationships between documents. Nodes in the network represent independent clinical documents, and edges represent citation, verification, or supplementary relationships between documents. The parsing operation traverses the network topology, distinguishing between the originally generated source document nodes (e.g., scanned copies of informed consent forms signed by subjects) and the derived document nodes generated by processing the source documents (e.g., data entry forms generated from original laboratory reports). The edges connecting these nodes and their attributes are extracted, with attributes indicating the specific type of citation, verification, or supplementary relationship. Each citation, verification, and supplementary relationship chain between documents is matched against pre-defined clinical document management specifications, checking the completeness of the links, the correctness of the logical order, and the completeness of necessary metadata. In some embodiments, the pre-defined clinical document management specifications explicitly stipulate that specific types of derived documents must reference the unique identifier of their source documents, and that citation relationships must be explicitly recorded in the document metadata. The matching process checks whether each relationship chain conforms to the specifications. For example, it checks whether the verification relationship chain from the scanned copy of the test report to the test result entry form in the electronic data acquisition system is complete. The completeness of the verification relationship chain means that the verification operation from the source document to the derived document is clearly recorded. The correctness of the logical order means that the generation time of the source document must be earlier than the generation time of the derived document that performed the verification operation. The completeness of necessary metadata includes whether key metadata such as document version number, signature, and date are complete. Each link in the reference, verification, and supplement relationship chain between documents is marked with a compliance status, and the compliance importance of the document itself is weighted according to the centrality of the document node in the network. Each relationship chain is marked as compliant, missing, incorrect, or pending based on the matching results. The centrality of a document node in the network can be calculated based on the number of times the document is referenced by other documents. The higher the centrality of a document in the network, the higher the weight of its compliance status. All document nodes and their compliance status relationships are visualized and encoded to form a document compliance status graph presented in the form of a network graph, where the node size represents the compliance importance and the edge attribute represents the specific compliance status. The visualization coding process maps network structure, node attributes, and edge attributes to graphical elements. The visual size of a node is proportional to its weighted value for document compliance importance, and the color or line type of the edge is used to distinguish different specific compliance states. In essence, the document compliance status map provides a structured global view of the compliance status of the clinical document system, allowing users to intuitively identify key document nodes and relationship links in the network that exhibit abnormal compliance status.

[0077] In practice, logical verification is performed on the critical event sequence to generate an event logic consistency report. Event pairs or groups with causal, temporal, or conditional constraints are identified from the critical event sequence. The critical event sequence is a list of events that have a significant impact on the project, extracted from a unified project view and ordered chronologically, such as subject signing informed consent, randomization, occurrence of a serious adverse event, and withdrawal from the study. The identification process is based on a predefined event logic pattern library, which defines common constraint types between events. For example, in a causal relationship, the "occurrence of a serious adverse event" event must be followed by the "investigator-reported serious adverse event" event; in a temporal relationship, the "first administration" event must occur after the "completion of screening eligibility" event; and in a conditional constraint relationship, the "performance of surgery" event can only occur after the "obtaining surgical consent" event. Based on the trial protocol and clinical logic rules, expected constraints are defined for each identified event pair or group. Trial protocols and clinical logic rules define the expected constraints. For example, the trial protocol explicitly states that subjects must receive their first dose within 72 hours of passing all screening assessments. Clinical logic rules require that any laboratory anomaly determination during a visit must be based on the test report of the sample collected during that visit. By comparing the actual occurrence time and state of events in the critical event sequence with the expected constraints, event logic conflicts that violate the expected constraints are detected. The comparison process involves comparing event timestamps and matching states. If the timestamp of event A is later than that of event B, but the expected constraints require event A to occur before event B, an event logic conflict violating the time constraint is detected. If the state of event C is "occurred," but the state of the prerequisite event D on which the expected constraints depend is "not occurred," an event logic conflict violating the condition constraint is detected. All detected event logic conflicts are attributed, categorized, and classified by severity, resulting in a structured event logic consistency report. Attribution classification is based on the root cause of the conflict, such as classification as "time sequence error", "missing premise" or "state contradiction", and severity classification is based on the potential impact of the conflict on data integrity or subject safety, such as classification into three levels: "high", "medium" and "low".

[0078] In some embodiments, the severity level classification employs a quantitative scoring method to calculate the conflict severity score. formula:

[0079]

[0080] in: Indicates the conflict type weight. This represents a coefficient indicating the scope of impact of the events involved in the conflict. and These are preset weighting coefficients. Optionally, the event logic consistency report is presented in tabular form, with columns including at least the conflict event identifier, conflict type, description of the violated constraint, attribution category, severity level, and occurrence timestamp. In essence, the event logic consistency report systematically lists the event logic inconsistencies detected in the project, providing clear guidance for subsequent problem tracing and process correction.

[0081] In one embodiment of the present invention, individual subject compliance trajectories, trial site efficacy data, document compliance status graphs, and event logic consistency reports are input into an intelligent analysis model for fusion and deduction, generating a comprehensive project report that includes project risk warnings, process deviation location, and resource gap alerts. The fusion and deduction of the intelligent analysis model is implemented based on a knowledge graph reasoning engine. The nodes of the knowledge graph reasoning engine consist of key entities and states from the individual subject compliance trajectories, trial site efficacy data, document compliance status graphs, and event logic consistency reports. The nodes of the knowledge graph reasoning engine may include entity nodes such as "Subject-001," "Trial Site-A," "Source Document-Lab Report 123," and "Key Event-Serious Adverse Event Report," as well as state attribute nodes such as "Compliance-Low," "Enrollment Rate-Decreased," "Compliance Status-Abnormal," and "Logical Conflict-Existence." These nodes are connected by describing the relationships between entities or the hierarchical relationship between states and entities, forming a multi-dimensional interconnected knowledge network. In the knowledge graph reasoning engine, rules for the propagation and influence between entity states are defined. These rules simulate how the deterioration of individual subject compliance trajectories is transmitted to the performance data at the test site, and how defects in the document compliance state graph cause conflicts in the event logic consistency report. The defined rules form the basis for the logical reasoning performed by the knowledge graph reasoning engine. These rules are encoded in the form of "IF-THEN" or production rules, describing the potential causal relationships and influence paths between different dimensions of information. The knowledge graph reasoning engine is launched to perform multiple rounds of state propagation and deduction. During these rounds, information is transferred and states are updated within the knowledge graph network according to the defined propagation and influence rules. When the deduced entity state exceeds a preset risk threshold, the corresponding entity and its state are marked as a risk point. The risk threshold is a pre-set numerical boundary for different types of state attributes. For example, the quantitative value of the "compliance-low" state is continuously below the threshold of 0.6 for more than 3 time periods, or the performance deviation of the "solution deviation rate" indicator continuously exceeds the threshold + 0.3. All marked risk points are aggregated, and based on their type and relationships, project risk warnings, process deviation locations, and resource gap alerts are generated separately, integrated into a comprehensive project report. The project risk warning describes the identified potential risks and their possible impact; the process deviation location identifies the specific process steps or operations that lead to the risk; and the resource gap alert indicates the type or quantity of resources needed to address the risk or correct the deviation that are currently lacking. These three parts together constitute a structured comprehensive project report.

[0082] A negative influence rule is defined between subject compliance status and trial site enrollment rate. This rule stipulates that when the subject compliance trajectory shows that multiple subject compliance levels at a specific trial site are consistently below a threshold, the expected enrollment rate of the trial site will be reduced. In specific implementation, if the knowledge graph inference engine detects that more than 30% of the subject nodes under "Trial Site-A" have the "Compliance - Low" state activated, and this state persists for more than a preset time window, the inference engine, based on the negative influence rule, infers that the "Expected Enrollment Rate - Reduced" state of the "Trial Site-A" node should be activated, and propagates the state change information to the associated "Trial Site Performance Data" node. A trigger rule is also defined between document compliance status and event logic conflict. This trigger rule stipulates that when the document compliance status graph shows a missing verification link for a key source document, a logic conflict regarding data traceability in the event logic consistency report will be triggered. In some embodiments, the knowledge graph inference engine identifies the "Verification Link - Missing" status of the key source document "Informed Consent Form - Subject 005" node from the input document compliance status graph. Based on triggering rules, the inference engine creates or activates a "Data Source Logic Conflict" node in the knowledge graph and connects this node to the "Event Logic Consistency Report" entity and the related "Subject - 005" entity, marking the existence of the conflict. A weighted contribution rule is defined between the protocol deviation rate in the experimental site performance data and the overall project risk level. This rule stipulates that the protocol deviation rates of different experimental sites are summed according to their weights, affecting the assessment of the overall project risk level. The knowledge graph inference engine parses the performance deviation value of the "Protocol Deviation Rate" indicator for each experimental site from the input experimental site performance data. Each experimental site is assigned a weight coefficient, which can be set based on the planned number of subjects to be enrolled or historical performance. Overall Project Risk Level The calculation can be based on the formula:

[0083]

[0084] in: This is the total number of test sites in the project. It is the first The weighting coefficients of each test point It is the first The performance deviation of the deviation rate index for each test site. Optionally, refer to Table 1 for the test site weight coefficients and the performance deviation of the deviation rate index.

[0085] Table 1: Experimental Point Weights and Deviation Rate / Effectivity Table of Experimental Point Weights

[0086]

[0087] It is understandable that, through the propagation and influence rules defined and executed in the knowledge graph inference engine, isolated risk signals from different dimensions—such as individual subject compliance trajectories, experimental site efficacy data, document compliance status graphs, and event logic consistency reports—are correlated, simulating the process of risk transmission and amplification among various project elements. In some embodiments, the inference process of the knowledge graph inference engine is multi-round iterative. The state changes triggered by the first round of rules may serve as input, triggering other rules in subsequent rounds, thereby revealing deeper, cross-dimensional composite risks. Optionally, the output format of the comprehensive project report includes structured text, dashboard charts, or interactive knowledge graph views, from which users can view a detailed list of risk points, a risk propagation path diagram, and specific warnings, locations, and prompts.

[0088] See Figure 3 This is a line graph showing the trajectory of participant compliance, accurately depicting the changes in compliance scores for three participants over the 12-month trial period. It visually reflects the relationship between the dynamic trend of compliance and the risk threshold. Participant-010 initially had the highest compliance score (approximately 0.98), showing a slow downward trend, reaching approximately 0.68 at 12 months, and never falling below the compliance threshold (0.6). Participant-001 initially had a score of approximately 0.98, which declined continuously in the early stages, bottoming out at 8 months (approximately 0.58), and then gradually recovered, reaching 0.80 at 12 months, showing a clear upward trend. Participant-005 initially had a score of approximately 0.95, showing the largest decline, reaching its lowest point at 9 months (approximately 0.53), and then slowly recovered, reaching approximately 0.67 at 12 months. The 0.6 threshold, marked by the red dashed line, is the core reference line; scores below this line are considered within the compliance abnormality risk range.

[0089] In one embodiment of the present invention, a targeted management intervention plan is generated based on a comprehensive project report using a decision support engine. The management intervention plan includes adjustments to the visit plan for specific trial sites, enhanced follow-up plans for specific subjects, a list of rectification requirements for designated clinical documents, and a review and verification process for key events. The process of generating the management intervention plan by the decision support engine includes the following steps: The decision support engine parses the comprehensive project report, extracting specific project risk warning items, process deviation location coordinates, and resource gap prompt details. The comprehensive project report is a structured data file. The decision support engine obtains specific warning, location, and prompt information such as "Trial Site-C Enrollment Rate Warning," "Subject-012 Follow-up Compliance Deviation," "Missing Laboratory Report Verification Link," and "Serious Adverse Event and Medication Record Timing Conflict" by parsing specific data fields or tags in the report. Each project risk warning item, process deviation location coordinate, and resource gap prompt detail is pattern matched with a pre-set intervention knowledge base to identify one or more candidate intervention actions for each warning, location, or prompt. The intervention knowledge base is a rule base or case base that stores the mapping relationship between "problem patterns" and "intervention actions." For example, when pattern matching identifies the entry "trial site enrollment rate warning," the intervention knowledge base may return multiple candidate intervention actions such as "increase on-site recruitment support," "optimize subject screening process," and "conduct investigator training." Based on trial site effectiveness data and real-time resource status, the implementation cost, expected effects, and feasibility of candidate intervention actions are evaluated and ranked. The evaluation and ranking process requires quantitative or qualitative analysis of the implementation conditions and potential benefits of each candidate intervention action in the current project environment. The top-ranked candidate intervention actions are selected and, based on different intervention subjects, combined into visit plan adjustment schemes for specific trial sites, follow-up enhancement schemes for specific subjects, rectification requirement lists for designated clinical documents, and review and verification processes for key events, collectively constituting the management intervention plan. For example, in response to the warning for "Trial Site-C", the two high-ranking candidate intervention actions of "optimizing the subject screening process" and "conducting researcher training" may be combined to form a detailed adjustment plan for the visitation plan for "Trial Site-C", which clearly defines the adjusted screening visitation process and the new training plan.

[0090] In practical implementation, the execution cost, expected effects, and feasibility of candidate intervention actions are evaluated and ranked based on the effectiveness data of pilot sites and the real-time status of resources. The current workload and response capability indicators of the target pilot sites are obtained from the pilot site effectiveness data. Target pilot sites refer to the pilot sites where the management intervention plan will be implemented. Current workload indicators can be "data entry backlog" or "on-duty participant visit rate" parsed from the pilot site effectiveness data. Response capability indicators can be "average time to resolve questions" or "deviation of plan from review cycle." The number of available monitors, budget surplus, and time window information are obtained from the real-time resource status. Real-time resource status information is obtained from the project management system or resource management database. The number of available monitors refers to the number of monitors currently available to execute new tasks. Budget surplus refers to the remaining funds available for emergency management. Time window refers to the number of days remaining until the next key project milestone. The incremental impact of each candidate intervention action on the current workload, the improvement in response capability indicators, the number of monitors required, the budget consumed, and the time window occupied are calculated. These parameters for each candidate intervention can be estimated through historical data analysis, expert experience, or a standard work time library. The calculated increment, boost, number of monitors, budget consumption, and time window occupancy are compared with preset constraints to filter out feasible candidate interventions that meet all constraints. Preset constraints include the maximum acceptable workload increment, minimum required responsiveness boost, maximum number of monitors available, budget ceiling, and latest completion time. Among the feasible candidate interventions, a comprehensive score is calculated based on a preset utility function, and they are evaluated and ranked from highest to lowest comprehensive score. The utility function is used to quantify the comprehensive value of each feasible candidate intervention. An example of a utility function is as follows:

[0091]

[0092] in: This represents the overall score of the candidate intervention action. This represents the normalized evaluation value that indicates the degree to which the action improves the responsiveness index. The required number of inspectors, The normalized value representing the consumed budget. The normalized value representing the time window occupied. , , , These are preset weighting coefficients used to balance the relative importance of effects, human resource costs, financial costs, and time costs. It can be understood that, through utility function calculations, the intervention action with the highest overall benefit or lowest overall cost can be prioritized while satisfying constraints. In some embodiments, referring to Table 2, the evaluation and ranking process can generate the analysis results shown in Table 2.

[0093] Table 2: Ranking Table of Candidate Intervention Actions

[0094]

[0095] It is understandable that the decision support engine transforms high-level, comprehensive project reports into specific, actionable, and optimized management intervention plans by parsing reports, matching them to knowledge bases, evaluating resources, and calculating utility. In some embodiments, the management intervention plan is output as a structured task list, clearly defining the action description, responsible party, required resources, timeline, and success criteria for each task. The process of generating management intervention plans by the decision support engine is dynamic; as the project progresses and resource status changes, the decision support engine can periodically re-evaluate and update the management intervention plans.

[0096] See Figure 4 This is a bar chart showing the resource allocation groupings for intervention programs in drug clinical project management. It visually presents the differences among the four intervention programs in terms of task quantity, priority, and estimated completion time, providing data support for project resource scheduling and priority decisions. Document rectification tasks have the most (15 items), the highest priority (level 5), and the longest estimated completion time (20 days), making it the core focus of the current project and requiring priority resource allocation to ensure progress. Visit plan adjustments / event reviews both have a priority of level 4, with a moderate workload, and need to be coordinated with the document rectification plan. Follow-up reinforcement tasks have a moderate workload (12 items) and a low priority (level 3), and can be scheduled as secondary priority tasks during gaps in the core program's implementation. The time-risk document rectification plan takes the longest, requiring close monitoring of schedule deviation risks to avoid impacting overall project milestones; event reviews take the shortest time and can be quickly closed.

[0097] In one embodiment of the invention, the management intervention plan is converted into a queue of operation instructions that can be recognized and executed by the drug clinical trial execution system. This conversion involves parsing the visit plan adjustment scheme for specific trial sites, the follow-up enhancement scheme for specific subjects, the rectification requirement list for designated clinical documents, and the review and verification process for critical events within the management intervention plan, decomposing them into atomic-level operation tasks. Each atomic-level operation task is mapped to a corresponding standard application programming interface (API) call command or database operation statement in the drug clinical trial execution system. Based on the logical dependencies and time requirements between the atomic-level operation tasks, all standard API call commands and database operation statements are scheduled and orchestrated to generate an operation instruction queue with a clear execution order.

[0098] In practice, the management intervention plan is converted into a queue of operational instructions that can be recognized and executed by the drug clinical project execution system. This conversion process includes parsing the specific instructions within the management intervention plan, mapping them to system-recognizable commands, and arranging the execution sequence. The management intervention plan includes adjustments to visit plans for specific trial sites, enhanced follow-up plans for specific subjects, a list of rectification requirements for designated clinical documents, and verification procedures for critical events, breaking these down into atomic-level operational tasks. The management intervention plan is a structured document containing various types of intervention instructions. The management intervention plan generated by the decision support engine may exist in JSON, XML, or other structured data formats. The parsing process reads this document and identifies different sections or tag blocks such as "Visit Plan Adjustment Plan," "Follow-up Enhancement Plan," "Rectification Requirements List," and "Verification Procedure." The visitation plan adjustment plan for a specific trial site may include a description of "increasing the screening visit rate of trial site A from twice a week to four times a week". The follow-up enhancement plan for a specific subject may include an instruction to "arrange an additional telephone follow-up for subject ID-1005". The rectification requirements list for a specified clinical document may include a requirement that "the missing signature document Consent_Form_003 must be signed within 3 days". The review and verification process for critical events may stipulate that "all medication records related to event SAE-20250321 shall be checked a second time".

[0099] Each atomic-level operation task is mapped to a corresponding standard application programming interface (API) call command or database operation statement in the drug clinical trial execution system. The system predefines a set of standard APIs for interaction with external systems or modules. These standard APIs are well-defined, open functional call interfaces covering core business processes from subject management, visit planning, document management to data quality control. The mapping process locates and calls the corresponding standard API based on the type and required parameters of the atomic-level operation task, generating specific call commands.

[0100] Based on the logical dependencies and time requirements between atomic-level operation tasks, all standard application programming interface (API) calls and database operation statements are scheduled and orchestrated to generate an operation instruction queue with a clear execution order. Logical dependencies mean that certain atomic-level operation tasks can only be executed after the successful completion of other atomic-level operation tasks. For example, the atomic-level operation task "calling the API to update the screening and visitation plan for test point A" logically depends on the completion of two preceding atomic-level operation tasks: "calling the API to query the existing screening and visitation plan for test point A" and "calculating the new visitation time point based on the query results." Time requirements can originate from urgency indicators or business rules in management intervention plans. For example, the requirement that "missing signature file Consent_Form_003 must be signed within 3 days" translates into setting a "latest completion time" constraint on related atomic-level operation tasks. The scheduling and orchestration process analyzes the dependency network and time constraints between all atomic-level operation tasks. The dependency network can be represented as a directed acyclic graph, where nodes represent atomic-level operation tasks, and directed edges represent dependencies between tasks. The goal of the scheduling and orchestration process is to compute a linear sequence of task executions, i.e., an operation instruction queue, that satisfies all dependencies and time constraints for this directed acyclic graph.

[0101] See Figure 5This is a chart showing the trend of task priority and deadline pressure. Through a dynamic comparison of two curves, it intuitively reflects the priority fluctuations and deadline pressure accumulation of project tasks over 20 time points, providing a basis for task scheduling and resource allocation decisions. The task priority score generally exhibits high-frequency fluctuations, oscillating within the range of 0.62 to 0.99, with multiple peaks and troughs. Priority reaches its peak at time points 2 and 12 (approximately 0.98 to 0.99), representing the core critical phase of the project; priority drops to its lowest point at time points 5 and 15 (approximately 0.66 to 0.67), temporarily reducing task importance; later (time points 15 to 20), priority gradually recovers to 0.72 to 0.81, again increasing task importance. Deadline pressure shows a stable linear upward trend, continuously increasing from an initial value of 0.3 to a final value of 0.9 without significant fluctuations, reflecting the continuous accumulation of pressure as the deadline approaches.

[0102] The above embodiments are only used to illustrate the technical methods of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical methods of the present invention without departing from the spirit and scope of the technical methods of the present invention.

Claims

1. An intelligent management method applicable to drug clinical projects, characterized in that, The method includes: The original project activity records, including trial site number, subject identification, examination type, test value, event timestamp, and clinical document identifier, are collected from the drug clinical project execution system to form a set of original activity records; Multi-dimensional structured processing is performed on the original activity record set to generate a unified project view that includes subject time-series activity flow, trial site operation status snapshots, clinical document association networks, and key event sequences; Compliance modeling is performed on the subject's time-series activity flow to generate individual subject compliance trajectories; efficacy assessment is performed on the snapshot of the trial site's operational status to generate trial site efficacy data; compliance checks are performed on the clinical document association network to generate a document compliance status map; and logical verification is performed on the key event sequence to generate an event logic consistency report. The individual subject compliance trajectory, the test site effectiveness data, the document compliance status map, and the event logic consistency report are input into the intelligent analysis model for fusion and inference, generating a comprehensive project report that includes project risk warning, process deviation location, and resource gap prompts; Compliance modeling is performed on the temporal activity flow of the subjects to generate individual subject compliance trajectories, including: From the subject's time-series activity log, extract the subject's actual activity records according to the preset experimental protocol milestone nodes; The extracted records of the subjects' actual activities were compared with the predefined standard timeline of the trial protocol to calculate the compliance deviation measure for each activity. By connecting the compliance deviation measures of each activity in chronological order and introducing a time decay function to model the continuous impact of historical deviations, the individual subject compliance trajectory is formed, which reflects the dynamic change process of subject compliance.

2. The intelligent management method for drug clinical projects according to claim 1, characterized in that, A performance evaluation is performed on the snapshot of the operational status of the test site to generate test site performance data, including: The data sequences of indicators such as group entry / exit rate, data entry timeliness rate, question clarification cycle, and scheme deviation rate are analyzed from the snapshot of the operational status of the test site. Benchmark reference lines are established for the indicators of group entry rate, data entry timeliness rate, question clarification cycle and plan deviation rate, respectively. The benchmark reference lines are determined based on historical project data or industry standard data. Compare the current indicator data series with the corresponding benchmark reference line to calculate the performance deviation of each indicator at different time points; The performance deviation of all indicators is integrated and coded in terms of time and indicator dimensions to form the performance data of the test point, which characterizes the overall and individual sub-item performance of the test point.

3. The intelligent management method for drug clinical projects according to claim 2, characterized in that, Perform compliance checks on the aforementioned clinical document association network to generate a document compliance status map, including: The clinical document association network is analyzed to identify source documents, derived documents, and the citation, verification, and supplementary relationship chains between documents; The citation, verification, and supplementary relationship chains between each document are matched with the pre-defined clinical document management standards to check the completeness of the chain, the correctness of the logical order, and the completeness of the necessary metadata. Mark the compliance status of each link in the citation, verification and supplementation relationship between documents, and weight the compliance importance of the document itself according to the centrality of the document node in the network; All document nodes and their compliance status relationships are visualized and encoded to form a document compliance status graph presented in the form of a network graph, where the node size represents the importance of compliance and the edge attributes represent the specific compliance status.

4. The intelligent management method for drug clinical projects according to claim 3, characterized in that, Perform logical verification on the key event sequence and generate an event logical consistency report, including: Identify event pairs or groups of events with causal, temporal, or conditional constraints from the key event sequence; Based on the trial protocol and clinical logic rules, define the expected constraints for each identified event pair or event group; By comparing the actual occurrence time and state of events in the key event sequence with the expected constraint relationship, logical conflicts of events that violate the expected constraint relationship are detected. All detected logical conflicts in events are attributed, categorized, and classified according to severity, resulting in a structured report on the logical consistency of the events.

5. The intelligent management method for drug clinical projects according to claim 4, characterized in that, The individual participant compliance trajectory, the test site efficacy data, the document compliance status map, and the event logic consistency report are input into an intelligent analysis model for fusion and deduction, generating a comprehensive project report that includes project risk warnings, process deviation location, and resource gap alerts, including: A knowledge graph reasoning engine is established in the intelligent analysis model. The nodes of the knowledge graph reasoning engine are composed of the individual subject compliance trajectory, test point performance data, document compliance status graph, and key entities and states in the event logic consistency report. The propagation and influence rules between entity states are defined in the knowledge graph reasoning engine to simulate how the deterioration of the individual subject's compliance trajectory is transmitted to the test point performance data, and how the defects in the document compliance state graph cause conflicts in the event logic consistency report. The knowledge graph reasoning engine is started to perform multiple rounds of state propagation and deduction. When the deduced entity state exceeds the preset risk threshold, the corresponding entity and its state are marked as risk points. Aggregate all marked risk points, and generate project risk warnings, process deviation locations, and resource gap prompts based on the type and correlation of the risk points, and integrate them into the comprehensive project report; The knowledge graph reasoning engine defines rules for the propagation and influence between entity states, including: Define a negative influence rule between subject compliance status and trial site enrollment rate: when the subject compliance trajectory shows that multiple subjects at a specific trial site have compliance consistently below a threshold, the expected enrollment rate of the trial site will be reduced. Define the triggering rules between document compliance status and event logic conflict. That is, when the verification link of a key source document is missing in the document compliance status graph, a logical conflict about data traceability in the event logic consistency report will be triggered. Define a weighted contribution rule between the deviation rate of the scheme in the performance data of the test sites and the overall risk level of the project. That is, the deviation rates of different test sites are accumulated according to their weights, which affects the assessment of the overall risk level of the project.

6. The intelligent management method for drug clinical projects according to claim 5, characterized in that, Also includes: Based on the comprehensive project report, a targeted management intervention plan is generated through a decision support engine. The management intervention plan includes a visit plan adjustment plan for specific trial sites, a follow-up enhancement plan for specific subjects, a list of rectification requirements for designated clinical documents, and a review and verification process for key events. The management intervention plan is converted into a queue of operation instructions that can be recognized and executed by the drug clinical project execution system, and the queue of operation instructions is distributed to the corresponding drug clinical project execution system terminal.

7. The intelligent management method for drug clinical projects according to claim 6, characterized in that, Based on the comprehensive project report, a targeted management intervention plan is generated through a decision support engine, including: The comprehensive project report is analyzed to extract specific project risk warning items, process deviation location coordinates, and resource gap prompts. Each project risk warning item, process deviation location coordinate, and resource gap prompt detail is matched with a pre-set intervention measure knowledge base to identify one or more candidate intervention actions for each warning, location, or prompt. Based on the performance data and real-time resource status of the test sites, the execution cost, expected effect, and feasibility of the candidate intervention actions are evaluated and ranked. The candidate intervention actions with the highest evaluation ranking are selected and combined according to different intervention subjects to form a visit plan adjustment plan for specific trial sites, a follow-up enhancement plan for specific subjects, a list of rectification requirements for specified clinical documents, and a review and verification process for key events, which together constitute the management intervention plan. The evaluation and ranking of candidate intervention actions based on the effectiveness data and real-time resource status at the test sites, considering their execution cost, expected effects, and feasibility, includes: Obtain the current workload and response capability indicators of the target test point from the performance data of the test point; Obtain information on the number of available inspectors, budget margin, and time window from the real-time status of resources; Calculate the increment of each candidate intervention action on the current workload, the degree of improvement on the response capability index, the number of monitors required, the budget consumed, and the time window occupied; The calculated increment, lift, number of inspectors, budget consumption, and time window occupancy are compared with the preset constraints to screen out feasible candidate intervention actions that meet all constraints. Among the feasible candidate intervention actions, a comprehensive score is calculated based on a preset utility function, and the evaluation and ranking are carried out according to the comprehensive score from high to low.

8. The intelligent management method for drug clinical projects according to claim 7, characterized in that, The management intervention plan is converted into a queue of operational instructions that can be recognized and executed by the drug clinical program execution system, including: The management intervention plan includes adjustments to the visitation plan for specific trial sites, enhanced follow-up plans for specific subjects, a list of rectification requirements for designated clinical documents, and a review and verification process for key events, which are broken down into atomic-level operational tasks. Each atomic-level operation task is mapped to a corresponding standard application programming interface call command or database operation statement in the drug clinical project execution system; Based on the logical dependencies and time requirements between atomic-level operation tasks, all standard application programming interface call commands and database operation statements are scheduled and orchestrated to generate the operation instruction queue with a clear execution order.

9. An intelligent management system suitable for drug clinical projects, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the intelligent management method for drug clinical projects as described in any one of claims 1 to 8.

Images